April 2001 by Nature Magazine

toc_r

Page 1 of 2

CONTENTS

April 2001 Vol 2 No 4

227 | In this issue doi:10.1038/35067046

Highlights PDF

[1954K]

229 | CELL PROLIFERATION Adhesion isn't everything doi:10.1038/35067042

230 | IN BRIEF REPRODUCTION | PROTEIN METHYLATION | CELL DIVISION doi:10.1038/35067050

230 | DNA REPAIR Breaking the cycle doi:10.1038/35067052

231 | TECHNIQUE The visible cell project ... doi:10.1038/35067039

231 | WEB WATCH ... and the virtual cell project doi:10.1038/35067055

237 | THE CAP-TO-TAIL GUIDE TO MRNA TURNOVER Carol J. Wilusz, Michael Wormington & Stuart W. Peltz doi:10.1038/35067025 [693K]

247 | MRNA LOCALIZATION: MESSAGE ON THE MOVE Ralf-Peter Jansen doi:10.1038/35067016 [1050K]

257 | TIE RECEPTORS: NEW MODULATORS OF ANGIOGENIC AND LYMPHANGIOGENIC RESPONSES Nina Jones, Kristiina Iljin, Daniel J. Dumont & Kari Alitalo doi:10.1038/35067005 [3105K]

268 | A REAL-TIME VIEW OF LIFE WITHIN 100 NM OF THE PLASMA MEMBRANE J. A. Steyer & W. Almers doi:10.1038/35067069 [1037K]

232 | WEB WATCH Mad scientists doi:10.1038/35067057

232 | PRIONS A big jump across the species barrier doi:10.1038/35067059

232 | MICROBIOLOGY Flagellin shifts gear doi:10.1038/35067036

233 | CELL DIVISION Counting on EGFR doi:10.1038/35067061

233 | IN BRIEF MOLECULAR MOTORS | TECHNIQUE | MEMBRANE FUSION | TRANSLOCATION

276 | CELL SIGNALLING AT THE SHOOT MERISTEM Steven E. Clark doi:10.1038/35067079 [1017K]

285 | MULTIFUNCTIONAL STRANDS IN TIGHT JUNCTIONS Shoichiro Tsukita, Mikio Furuse & Masahiko Itoh doi:10.1038/35067088 [1740K]

294 | CBL: MANY ADAPTATIONS TO REGULATE PROTEIN TYROSINE KINASES Christine B. F. Thien & Wallace Y. Langdon doi:10.1038/35067100 [2274K]

doi:10.1038/35067064

234 | LIPID Shifting the fat? doi:10.1038/35067066

307 | OPINION SIGNAL PROCESSING AND TRANSDUCTION IN PLANT

http://www.nature.com.ezproxy.library.wisc.edu/nrm/journal/v2/n4/toc_r.html

12/3/2006

toc_r

Page 2 of 2

CELLS: THE END OF THE BEGINNING? Simon Gilroy & Anthony Trewavas

234 | APOPTOSIS Competing for XIAP doi:10.1038/35067002

235 | CELL DIVISION Battle at the p15 promoter

doi:10.1038/35067109 [873K]

315 | NatureView

doi:10.1038/35067034

doi:10.1038/35067115

235 | PROTEIN DEGRADATION Water, water everywhere ... doi:10.1038/35067013

NATURE REVIEWS | MOLECULAR CELL BIOLOGY

http://www.nature.com.ezproxy.library.wisc.edu/nrm/journal/v2/n4/toc_r.html

12/3/2006

HIGHLIGHTS HIGHLIGHTS ADVISORS JOAN S. BRUGGE HARVARD MEDICAL SCHOOL, BOSTON, MA, USA PASCALE COSSART INSTITUT PASTEUR, PARIS, FRANCE GIDEON DREYFUSS UNIVERSITY OF PENNSYLVANIA, PHILADELPHIA, PA, USA PAMELA GANNON CELL AND MOLECULAR BIOLOGY ONLINE JEAN GRUENBERG UNIVERSITY OF GENEVA, SWITZERLAND ULRICH HARTL MAX-PLANCK-INSTITUTE, MARTINSRIED, GERMANY NOBUTAKA HIROKAWA UNIVERSITY OF TOKYO, JAPAN STEPHEN P. JACKSON WELLCOME/CRC INSTITUTE, CAMBRIDGE, UK ROBERT JENSEN JOHNS HOPKINS UNIVERSITY, BALTIMORE, MD, USA VICKI LUNDBLAD BAYLOR COLLEGE OF MEDICINE, HOUSTON, TX, USA TONY PAWSON SAMUEL LUNENFELD RESEARCH INSTITUTE, TORONTO, CANADA NORBERT PERRIMON HARVARD MEDICAL SCHOOL, BOSTON, MA, USA THOMAS D. POLLARD THE SALK INSTITUTE, LA JOLLA, CA, USA JOHN C. REED THE BURNHAM INSTITUTE, LA JOLLA, CA, USA KAREN VOUSDEN NATIONAL CANCER INSTITUTE, FREDERICK, MD, USA JOHN WALKER MRC DUNN HUMAN NUTRITION UNIT, CAMBRIDGE, UK

C E L L P R O L I F E R AT I O N

Adhesion isn’t everything It has long been known that there is a correlation between the dysfunction of adherens junctions and tumour invasion and metastasis. But the general belief was that loss of cell–cell adhesion was only one of several factors necessary to cause a tumorigenic phenotype. Vasioukhin and colleagues now report in Cell that the ablation of a single adherens junction protein — α-catenin — in skin can cause a phenotype very similar to that observed in a pre-cancerous skin condition. As α-catenin is required during early development, Vasioukhin and co-workers used a conditional knockout approach to ablate α-catenin only after embryonic day 14, in skin epithelial stem cells and their progeny. The newborn knockout mice, in addition to missing whiskers and having underdeveloped limbs, lacked large segments of epidermis, so the investigators decided to study embryos to limit trauma. Epithelial cell differentiation was normal in these animals, but cell polarity was affected and there were marked defects in the control of cell proliferation. For example keratin 6 — a typical marker of hyperproliferating epidermis — was abundant (red in the picture) throughout the epidermis and in disorganized masses of keratinocytes in the dermis. In tissue culture, knockout keratinocytes grew much faster than wild-type cells, even in the absence of calcium, when adherens junctions cannot form, indicating that hyper-

Courtesy of V. Vasioukhin and E. Fuchs, Howard Hughes Medical Institute of the University of Chicago, USA.

proliferation is a direct consequence of α-catenin ablation, and is not a secondary effect of the loss of cell–cell adhesion. This was further confirmed by the fact that conditional ablation of the desmosomal protein desmoplakin in skin did not lead to a similar hyperproliferation of keratinocytes. To understand why α-catenin knockout cells proliferate so fast, the authors looked at several known signal-transduction pathways. They found that the levels of phosphorylated extracellular signal-regulated kinase (ERK, a MAPK) and Ras–GTP were increased, and that this signalling pathway is triggered by insulin or insulin-like growth factor 1. The corresponding tyrosine kinase receptors were not more active than in wild-type cells, but the signal seemed to be propagated more efficiently. How is all this connected to α-

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

catenin? In normal adherens junctions, α-catenin binds to β-catenin, which in turn binds to E-cadherin. You could imagine that β-catenin might be the culprit, but this is not the case, as neither the amounts nor the localization of this protein are altered in α-catenin knockout cells. What happens is that the phosphorylated insulin receptor substrate 1 (IRS-1) — a protein that relays signals from the insulin receptor — binds to the vacant E-cadherin–βcatenin complex. Finding the missing link between the formation of this complex and the increased activation of the MAPK pathway will be the next challenge. Raluca Gagescu References and links ORIGINAL RESEARCH PAPER Vasioukhin, V. et al. Hyperproliferation and defects in epithelial polarity upon conditional ablation of α-catenin in skin. Cell 104, 605–617 (2001)

VOLUME 2 | APRIL 2001 | 2 2 9

HIGHLIGHTS

IN BRIEF

D N A R E PA I R

Breaking the cycle

RE PRODUCTION

A sperm cytoskeletal protein that signals oocyte meiotic maturation and ovulation. Miller, M. A. et al. Science 291, 2144–2147 (2001)

In many animals, oocytes arrest during meiotic prophase. They are triggered to resume meiosis by sperm, which also cause the gonadal sheath cells (needed for ovulation) to contract. Miller and colleagues show here that the major sperm cytoskeletal protein (MSP) is a signal for both of these events. They go on to show that MSP has a function in sperm locomotion, analogous to that of actin, and conclude that MSP has acquired both extracellular signalling and intracellular cytoskeletal functions during evolution. P R OT E I N M E T H Y L AT I O N

Arginine methylation of STAT1 modulates IFNα/βinduced transcription. Mowen, K. A. et al. Cell 104, 731–741 (2001)

Previous analysis of STAT transcription factors has focused on their regulation by phosphorylation, shadowing the importance of another modification — arginine methylation. Here, Mowen and colleagues show that STAT1 is methylated on a conserved arginine residue, a modification mediated by the arginine methyltransferase PRMT1 and required for interferon α/βinduced transcription. Moreover, they find that the methyltransferase inhibitor MTA inhibits this methylation, leading to increased association of STAT with its inhibitor PIAS1 and so provide a model for the lack of interferon responsiveness observed in certain cancers. CELL DIVISION

Requirement of a centrosomal activity for cell cycle progression through G1 into S phase Hinchcliffe, E. H. et al. Science 291, 1547–1550 (2001)

A broken leg is a minor inconvenience — it may keep you in bed for a few days, but eventually you’ll need to get up and carry on as normal. So too, it seems, for yeast with broken DNA. A double-stranded DNA break (DSB) brings the cell cycle to a halt, allowing time for the break to be repaired. After 8–12 hours, however, the cells may escape this arrest — even if the DSB persists — in a process known as adaptation. How can cells do this? James Haber, Marco Foiani and colleagues have investigated this problem and, reporting in Molecular Cell, they show that adaptation is accompanied by a loss of the Rad53 checkpoint kinase activity. The DNA-damage checkpoint in Saccharomyces cerevisiae involves a cascade of protein kinases, including Rad53, Chk1, Mec1 and Cdc5 (which is phosphorylated by Rad53 and is known to be involved in adaptation). One way for cells to adapt would be to turn off this kinase cascade, even in the presence of a DSB, so the authors asked what happens to Rad53 and Chk1 in cells that adapt versus those that remain permanently arrested. They chose Rad53 and Chk1 because both are phosphorylated and activated by the upstream kinase Mec1 in response to DNA damage.

Centrosome-dependent exit of cytokinesis in animal cells Piel, M. et al. Science 291, 1550–1553 (2001)

The centrosome — the main microtubule organizing centre in animal cells — has long been thought to change its microtubule organizing properties in response to cell-cycle progression. These two papers now show that the centrosome, in turn, is required at two distinct stages of the cell cycle — the G1 to S transition and the completion of cytokinesis. Piel and colleagues show that, in animal cells, the centrosome is repositioned after anaphase, and is required for microtubule release from the midbody and the completion of cell division. Hinchcliffe et al. removed the centrosome from primate somatic cells, and found that they arrested before S phase, indicating that a centrosomal-associated factor mediates entry into S phase. They also find that, once in mitosis, the centrosome is not required for the G2 to M transition. Together these papers indicate cells may have checkpoints that monitor centrosome duplication or, say the authors,“core centrosomal structures could bind cell cycle regulatory molecules in a way that activates their function”.

230

The authors used a site-specific endonuclease to introduce one or two DSBs that could not be repaired. Although cells adapt to one DSB, they remain arrested with two. Interestingly, however, it has been previously shown that yeast cells do not decide to adapt by measuring the number of DSBs per se. Rather, they seem to monitor the amount of single-stranded DNA that is generated as exonucleases nibble back at the broken ends. So deletion of the DNAend-binding protein yKu70 prevents adaptation with one DSB, as exonucleases are more free to attack the exposed DNA ends. Haber, Foiani and colleagues found that, in cells that adapt, there is a concomitant loss of Rad53 kinase activity. Conversely, in those cells that fail to adapt — because they contain two DSBs or lack functional Cdc5 or yKu70 — the levels of Rad53 activity remain high. When cells adapt, phosphorylation of Chk1 also decreases, indicating that both the Rad53 and Chk1 pathways are involved in adaptation. Moreover, maintenance of the arrested state depends on continued activity of the Mec1 kinase. This means, say the authors, that adaptation is due to inactivation of the kinase cascade itself, rather than, for example, modification of a downstream receptor such that it is no longer sensitive to the cascade. But why do yeast cells adapt at all? The authors speculate that it may allow the DSB to be repaired in a subsequent cell cycle, and there is evidence that some repair mechanisms may be more efficient during S phase than at other times in the cell cycle. For the moment, though, a next step will be to find out how the kinase cascade is inactivated — by proteolytic destruction of phosphorylated kinases, perhaps, or by dephosphorylation of Rad53 and Chk1? Alison Mitchell References and links ORIGINAL RESEARCH PAPER Pellicioli, A. et al.

Regulation of Saccharomyces Rad53 checkpoint kinase during adaptation from DNA damageinduced G2/M arrest. Mol. Cell 7, 293–300 (2001)

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

HIGHLIGHTS

WEB WATCH

Organelles in the Golgi region. C1 (cis-Golgi cisterna), light blue; C5, dark blue; C7, red; ER, yellow; microtubules and mitochondria, green; dense core vesicles, blue; clathrin-negative vesicles, white; clathrin-positive vesicles and compartments, red; clathrin-negative compartments, purple; free ribosomes, orange. Reproduced with permission from National Academy of Sciences © (2001). TECHNIQUE

The visible cell project . . . Pancreatic β-cells (HIT-T15) were put to death through high-pressure freezing and freezesubstitution, cut into ribbons of serial 400-nm-thick sections and studied by electron tomography. Marsh and colleagues collected 480 images of three serial sections by tilting the grid that holds the specimen by 1.5° over a range of 120° about two orthogonal axes. They then combined all these images to produce a single, high resolution three-dimensional reconstruction of a 3.1 × 3.2 × 1.2 µm3 area around the Golgi apparatus of a cell. You would think this is a lot of work for a single image, but just look at the result! The freezing procedure immobilized all cellular activity within milliseconds, so the final image probably reflects the situation found in a live cell. In addition to being incredibly pretty, this snapshot of the cell also gives us some insights into the organization of organelles around the Golgi. There were not many surprises about the structure of the Golgi itself, which is here made up of seven cisternae (C1 to C7, going from cis to trans Golgi). One interesting observation is that the endoplasmic reticulum (ER) seems to traverse the Golgi stack through aligned openings in the cisternae. There are close contacts between the ER and the C5, C6 and C7 cisternae, which might point to direct exchange between these organelles instead of vesicular transport. Following up on these observations might shed some

light on the relationship between the ER and the Golgi, which has been controversial over the past years. Microtubules seem to follow closely the membranes of the C1 cisterna of the Golgi and of endosomal compartments. However, the ER seems to be anchored along microtubules at a few points only. Here again, this observation could provide a lead for further investigation into the interaction of organelles with microtubules. The technique is not only qualitative but also provides a means to quantify organelles in situ in three dimensions, and to measure accurately their physical associations with other organelles. For example, at first sight, you might get the impression that the Golgi region is very crowded, but Marsh and colleagues calculated that only about 34% of the volume is taken up by organelles, the rest being made up of cytoplasmic matrix. Although many of the findings in this study are not truly novel or revolutionary, being able to see organelles in three dimensions in their natural cellular context should have high impact on how we imagine life in the cell. Raluca Gagescu References and links ORIGINAL RESEARCH PAPER Marsh, B. J. et al. Organellar relationships in

the Golgi region of the pancreatic beta cell line, HIT-T15, visualized by high resolution electron tomography. Proc. Natl Acad. Sci. USA 98, 2399–2406 (2001)

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

… and the virtual cell project The Virtual Cell is described by its creators as “a general computational framework for modelling cell biological processes”. Access to this software is available over the internet through a JAVAbased interface, and although there is no charge for academic use of the Virtual Cell, users are required to register. So what is the Virtual Cell? It has been developed at the National Resource for Cell Analysis and Modeling (NRCAM), a US-based resource centre supported by the National Center for Research Resources. Those behind it claim that the technology links biochemical and electrophysiological data with experimentally derived microscopic images showing the subcellular localizations of the molecules involved. In this way, they say, “physiological results can be simulated within the empirically derived geometries, thus facilitating the direct comparison of model predictions with experiment”. There are two interfaces to the Virtual Cell — one biologically orientated, the other mathematical. The biological interface has been designed to allow users to define cellular geometry, create models and simulations, and to analyse the results of such simulations. For those unfamiliar with the technology or software there’s a ‘User’s Guide’, backed up by a tutorial designed to work in conjunction with it, and a ‘User Discussion’ page for troubleshooting. There are also examples of what the site can do — the applications shown include use of the Virtual Cell to study diffusion processes in mitochondrial cristae, and a simulation of a calcium wave in neuroblastoma cells. Movies are provided with some of the examples, although more explanation would enhance their value.

Alison Mitchell

VOLUME 2 | APRIL 2001 | 2 3 1

HIGHLIGHTS

WEB WATCH Mad scientists With the recent furore over foot and mouth disease in Europe, bovine spongiform encephalopathy (BSE) has, for the moment, stepped out of the public eye. Tucked away on the internet, however, is the Official Mad Cow Disease Home Page, which claims to host over 7,431 articles on Creutzfeldt–Jakob disease (CJD), prions, BSE, scrapie and other transmissible spongiform encephalopathies. What this site lacks in flashy graphics and design, it makes up for in the sheer volume of information — the news archives date back to March 1996, and the prion science archives to October of that year. New information is added at least once per week, from sources including international news agencies and newspapers as well as scientific journals. We especially like the ‘Best Links’ page, which directs readers to sites such as databases, genomes, journals, meetings and contacts. There is also a comprehensive — albeit slightly daunting for the nonspecialist — list of links to online tools. Government and public-interest sites are listed as well, although some links are of dubious quality. In addition to these links, the site itself hosts a number of useful resources. These include an image gallery of three-dimensional prion structures, a graphics index, a tutorial on genome annotation and a curated database of all available prion and prion-like sequences. One criticism, though, is that the site is not very easy to navigate, and a site map would be useful to highlight the wealth of resources available. Finally, the tantalisingly titled section on ‘mad scientists’ warrants a mention. Disappointingly, however, this is simply an eclectic bag of news snippets on a handful of researchers.

PRIONS

A big jump across the species barrier Yeast Sup35 self-propagates in a similar way to mammalian prions, transmitting the prion element [PSI+], which confers heritable suppression of nonsense mutations. There is a robust species barrier between the yeasts Saccharomyces cerevisiae and Candida albicans, and prions from one species cannot ‘infect’ prions of the other species. To examine the relationship between the primary structure of prions and their species specificity, Chien and Weissman constructed a chimeric (CHIM)

M I C R O B I O LO G Y

Flagellin shifts gear The swimming of bacteria such as Salmonella typhimurium and Escherichia coli is driven by the rotation of flagella, no more than 0.25 µm in diameter but as much as 60 times that in length. The bacteria alternate between ‘running’ in a straight line and chaotic ‘tumbling’, while the rotary motor at the base of a flagellum changes from anticlockwise rotation to clockwise and back again. Reporting in Nature, Keiichi Namba

Alison Mitchell

232

protein using the first 39 amino acids of the S. cerevisiae prion domain (SC) and amino acids 40–140 of the C. albicans prion domain (CA). To their surprise, they found that their chimera was able to jump the species barrier, as transient overexpression of CHIM could induce the formation of [PSI+] in yeast expressing SC, CA or CHIM. Conversely, SC, CA or CHIM could induce a [PSI+]like state in CHIM-expressing yeast. Remarkably, inducing the conversion of CHIM with either

and colleagues at the Protonic Nanomachine Project in Kyoto, Japan, have begun to work out the subtle atomic-level changes that couple these two events. The bacterial flagellum consists over almost all of its length of a single protein, flagellin. Thousands of flagellin molecules form a hollow tube composed of 11 simple polymer threads, known as protofilaments. Electron microscopy of flagella has revealed that the protofilaments can exist in two forms, the L and R forms. The L form is slightly longer, with its flagellin subunits being 0.8 Å more elongated. A flagellum made up entirely of protofilaments of either type is straight and, although good for structural studies, it makes a poor propeller. Mixing the two forms in a single flagellum, however, means that the difference in lengths sets up tensions that can be resolved only by supercoiling the flagel-

SC or CA gave rise to distinct prion strains — CHIM[SC] and CHIM[CA] — with different strain phenotypes in vivo and different fibre conformations in vitro. Second-generation fibres showed marked differences in seeding specificity: CHIM[SC] could seed only CHIM and SC but not CA, whereas CHIM[CA] could seed only CHIM and CA but not SC. So it seems that the diversity of strain phenotypes might rely on the ability of a prion protein to adopt several self-propagating

lum into a corkscrew shape. When bacteria are swimming in a straight line, the flagella usually have nine L-type and two R-type protofilaments, producing a left-handed corkscrew. These flagella can bundle together to form a coordinated propulsion unit. When the flagellar motors reverse direction, a number of L-type protofilaments change to Rtype protofilaments through a cooperative change in each flagellin molecule, right along the flagellum’s length. Right-handed supercoiled flagella are produced, breaking up the flagellar bundles and leaving the individual flagella to push in different directions — this produces the tumbling motion. A very small change in flagellin’s structure is thus at the heart of the bacterium’s change in behaviour. Proteins that form polymers pose a particular problem to structural studies; rather than forming wellordered crystals, they tend to produce poorly ordered aggregates of their polymers. Consequently, Namba and colleagues studied a version of flagellin from Salmonella lacking 52 amino acids from its amino terminus and 44 amino acids from its carboxyl terminus. On solving its structure at 2.0-Å resolution, the authors found that the flagellin proteins were arranged as if single protofilaments were running throughout the crystals. This high-resolution crystal structure

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

HIGHLIGHTS

IN BRIEF

conformations. The species barrier might result from the fact that prions that differ at the amino-acid level tend to adopt distinct, non-interacting conformations. But the barrier breaks down if a prion protein is promiscuous and can adopt both conformations.

M O L E C U L A R M OTO R S

Cargo of kinesin identified as JIP scaffolding proteins and associated signaling molecules. Verhey, K. J. et al. J. Cell Biol. 152, 959–970 (2001)

Raluca Gagescu References and links

CELL DIVISION

ORIGINAL RESEARCH PAPER Chien, P. &

Weissman, J. S. Conformational diversity in a yeast prion dictates its seeding specificity. Nature 410, 223–227 (2001) FURTHER READING Aguzzi, A. et al. Prions: health scare and biological challenge. Nature Rev. Mol. Cell Biol. 2, 118–126 (2001) | Serio, T. R. & Lindquist, S. L. [PSI+]: an epigenetic modulator of translation termination efficiency. Annu. Rev. Cell Biol. 15, 661–703 (1999) ELS LINK Prions AUTHOR LAB PAGE Weissman laboratory

could be easily fitted into lower-resolution electron microscope data to produce a model for the straight flagella formed by R-type protofilaments alone (as in the figure). To understand how the R-type protofilament converts into the L type, the Japanese group used computer modelling to stretch their structure. Three flagellin subunits were taken, arranged as in a protofilament. The end subunits were kept rigid while being pulled apart in 0.1-Å steps, and the central subunit was allowed to relax to its lowest energy state. At first there were no major changes in structure, but then, over a 0.2-Å stretch, the central flagellin underwent a subtle conformational change. A β-hairpin shifted to allow the 0.8-Å expansion required to lengthen the protofilament. The bacterial flagellum has been one of the most intensely studied structures in biology, and what has piqued the curiosity of biochemists, biophysicists and engineers alike is its changes between forward swimming and reverse tumbling. At long last we are beginning to see exactly how subtle this switch in gear really is. Christopher Surridge Senior Editor, Nature References and links ORIGINAL RESEARCH PAPER Samatey, F. A. et al. Structure of the bacterial flagellar protofilament and implications for a switch for supercoiling. Nature 410, 331–337 (2001)

Counting on EGFR During development, it is just as crucial to have the right number of cells as it is to specify their fates correctly. Extracellular cues have long been thought to regulate cell number, but what are these cues? A paper in Cell now reveals that, in the Drosophila melanogaster eye, activation of the epidermal growth factor receptor (EGFR) triggers division and that this same signal regulates cell survival. Baker and Yu focused on a group of cells that form late in the eye disc. By forming mutant clones within a wild-type tissue, they found that spitz, the ligand for EGFR, is important for division. Looking at different markers for cells that have entered the cell cycle, they then showed that the cells were arrested in G2. This led them to wonder whether EGFR is required for G2/M progression. As suspected, EGFR mutant clones did not enter mitosis. Using EGFR overexpression, they then confirmed that autonomous EGFR is sufficient to trigger division. Where does the trigger for EGFR activation come from? The authors showed that neighbouring cells provide this signal — a mechanism that might ensure coordination between the number of cell types in the tissue. Once the cells are formed, the authors showed, EGFR mediates their survival — a role previously suspected and which is distinct from its function in G2/M progression. So it seems that EGFR has a dual function — ensuring that enough cells form and that, once formed, they do not die. Alison Schuldt References and links ORIGINAL RESEARCH PAPER Baker, N. E. &

Yu, S.-Y. The EGF receptor defines domains of cell cycle progression and survival to regulate cell number in the developing Drosophila eye. Cell 104, 699–708 (2001)

We understand in some detail how kinesin moves, but we know little about its cargo. Verhey et al. show that kinesin binds to three scaffolding proteins in the JNK signalling pathway, JIP-1, JIP-2 and JIP-3. Several signalling molecules are present in the complex, indicating that kinesin might transport pre-assembled signalling complexes. Could signalling molecules, in turn, regulate kinesin? TECHNIQUE

Delivery of proteins into living cells by reversible membrane permeabilization with streptolysin-O. Walev, I. et al. Proc. Natl Acad. Sci. USA 98, 3185–3190 (2001)

It is not easy to introduce reagents into cells without permanently damaging them. Permeabilization with streptolysin-O — a standard cell-biology technique — might be a solution to this problem. Walev and colleagues report that it is possible to introduce whole proteins into cells permeabilized with low concentrations of streptolysin-O. The cells reseal and remain viable for days in culture, allowing the effects of the introduced proteins (or reagents) to be studied in living cells. MEMBRANE FUSION

SNARE complex oligomerization by synaphin/ complexin is essential for synaptic vesicle exocytosis. Tokumaru, H, et al. Cell 104, 421–432 (2001)

Membrane fusion involves the assembly of SNARE transcomplexes to which VAMPs and syntaxins each contribute one αhelix, and SNAP-25 contributes two. Tokumaru et al. propose that synaphin facilitates the assembly of SNARE trans-complexes and the oligomerization of several SNARE complexes into a ‘rosette’. One SNAP-25 molecule might take part in two neighbouring SNARE complexes, contributing one helix to each. In this way, several SNARE complexes could be serially connected, and this might increase the speed of fusion at nerve termini. T R A N S LO C AT I O N

The three modules of ADP/ATP carrier cooperate in receptor recruitment and translocation into mitochondria. Wiedemann, N. et al. EMBO J. 20, 951–960 (2001)

The translocation of ADP/ATP carriers of the inner mitochondrial membrane is puzzling as, instead of one aminoterminal pre-sequence, they contain targeting information in all three of their modules. Which signal is, then, the most important? Here, the authors report that signals in all three modules cooperate at each stage of the translocation process, including binding to the receptor Tom70, translocation (through loop formation), and dimerization in the inner membrane.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 3 3

HIGHLIGHTS

LIPIDS

Shifting the fat? Three groups report in The Journal of Cell Biology that, under some conditions, caveolins — cholesterol-binding proteins characteristic of caveolae — localize to intracellular lipid droplets. This localization raises more questions than it answers, but the new findings will certainly open avenues for investigation. Pol and colleagues were following up their previous studies, in which they discovered that a truncation mutant of caveolin-3 (CavDGV) inhibits signalling through H-Ras but not through K-Ras. They now find that CavDGV localizes to lipid droplets, and that it redistributes cholesterol from the plasma membrane (where it is important for maintaining functional lipid raft signalling domains) to late endosomes. Ostermeyer and colleagues investigated how caveolin-1 travels from the endoplasmic reticulum (ER) to the plasma membrane. They

added an ER-retention signal to the carboxyl terminus of the molecule (Cav–KKSL), and obtained a clear-cut result: Cav–KKSL did not reach the plasma membrane, so its transport is through the biosynthetic pathway. But to their surprise, Cav–KKSL was not retained in the ER; it accumulated on lipid droplets. Last, Fujimoto and colleagues studied the intracellular route of overexpressed caveolin2β, and found that it too localized to lipid droplets, whereas its close relative caveolin-2α did not. Many of the results obtained in the three studies coincide, even if the authors interpret their results in different ways. Ostermeyer and colleagues propose that caveolins accumulate in lipid droplets only if they are slow to exit the ER. This is the case for mutant proteins (which might not fold fast enough), for overexpressed proteins (which might saturate the export machinery), and for cells treated with brefeldin A, a fungal metabolite that collapses the Golgi into the ER. Pol and colleagues, on the other hand, argue that, as a small proportion of endogeneous caveolins can reach lipid droplets, and CavDGV acts as a

CavDGV on lipid droplets. Courtesy of A. Pol and R. Parton.

dominant-negative mutant, these structures might represent an intermediate station on the normal intracellular route of caveolins. The new findings are intriguing and bring lipid droplets under the spotlight. But many questions remain. Are caveolins targeted to lipid droplets under physiological conditions? How are they targeted to this compartment? And above all, what is the function of this localization? Raluca Gagescu References and links ORIGINAL RESEARCH PAPERS Pol, A. et al. A caveolin

dominant negative mutant associates with lipid bodies and induces intracellular cholesterol imbalance. J. Cell Biol. 152, 1057–1070 (2001) | Ostermeyer, A. G. et al. Accumulation of caveolin in the endoplasmic reticulum redirects the protein to lipid storage droplets. J. Cell Biol. 152, 1071–1078 (2001) | Fujimoto, T. et al. Caveolin-2 is targeted to lipid droplets, a new ‘membrane domain’ in the cell. J. Cell Biol. 152, 1079–1085 (2001)

A P O P TO S I S

Competing for XIAP

Caspases are powerful destructive agents which, once activated, lead to certain death. Or do they? According to a report in Nature by Emad Alnemri and colleagues, a cell can be snatched from the jaws of death through the recruitment, by activated caspase-9, of its own inhibitor. Death is, as Tennyson famously said, the end of life. It’s no surprise, then, that caspase pathways must be tightly regulated. The so-called intrinsic death pathway, which responds to most death stimuli, is switched on through

234

recruitment of the weakly active procaspase-9 by an adaptor molecule called Apaf-1. Procaspase-9 is then processed by autoproteolytic cleavage to yield active caspase-9. Alnemri and colleagues knew that caspase-9 can be inhibited by the X-linked inhibitor of apoptosis protein (XIAP). To find out how this inhibition works, they reconstituted, in vitro, caspase-9–Apaf-1 complexes containing either fully processed caspase-9 or the procaspase form, and then studied their interactions with XIAP. Although both complexes were catalytically active, only the fully processed caspase-9 could be inhibited by XIAP. This, it turns out, is because XIAP cannot associate with the unprocessed procaspase-9. Does this mean that processing is required for inhibition by XIAP? To test this idea, the authors identified a conserved motif of four amino acids, which becomes exposed at the amino terminus of the caspase-9 small subunit after proteolytic processing. This motif also has considerable homology to a sequence that binds IAPs in the Drosophila death proteins Grim, Reaper and Hid. Systematic mutation of the residues in this motif confirmed that they are needed for binding of XIAP to caspase-9.

Inhibition of caspase-9 can be overcome by a protein called Smac/DIABLO, which also interacts with XIAP and also contains the conserved four residues at its amino terminus. Alnemri and colleagues wondered whether the binding of Smac/DIABLO and caspase-9 to XIAP is mutually exclusive, and found that, indeed, the caspase-9 IAP-binding motif abolishes the binding of Smac/DIABLO to XIAP (and vice versa). The inference, say the authors, is that “Smac competes with caspase-9 for binding the same pocket on the surface of XIAP, which could explain the ability of Smac to promote the catalytic activity of caspase-9 in the presence of XIAP”. They also point out that, unlike for other caspases, proteolytic processing of caspase-9 is required for its inhibition rather than its activation. And, as Donald Nicholson discusses in the accompanying News and Views article, these results could have exciting implications for cancer therapy, as synthetic mimics of XIAP-binding peptides might sensitize cancer cells to apoptotic stimuli. Alison Mitchell References and links ORIGINAL RESEARCH PAPER Srinivasula, S. M. et al. A

conserved XIAP-interaction motif in caspase-9 and Smac/DIABLO regulates caspase activity and apoptosis. Nature 410, 112–116 (2001) FURTHER READING Nicholson, D. W. Caspases bait inhibitors. Nature 410, 33–34 (2001) | Chai, J. et al. Structural basis of caspase-7 inhibition by XIAP. Cell 104, 769–780 (2001) | Huang, Y. et al. Structural basis of caspase inhibition by XIAP: differential roles of the linker versus the BIR domain. Cell 104, 781–790 (2001) | Riedl, S. J. et al. Structural basis for the inhibition of caspase-3 by XIAP. Cell 104, 791–800 (2001)

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

CELL DIVISION

Battle at the p15 promoter The company we keep often affects our behaviour, and the same is true of the oncoprotein Myc: in the company of Max, Myc can activate the transcription of genes that generally stimulate proliferation; but Myc can also behave as a trancriptional repressor. The proteins that bring out this side of Myc’s personality have remained elusive, but two papers in the April issue of Nature Cell Biology identify them. Miz-1 is a zinc-finger-containing protein that induces cell-cycle arrest and activates the transcription of several genes that are repressed by Myc. Staller and colleagues reasoned that Miz-1 might activate transcription of a cyclin-dependent kinase inhibitor, and used PCR and western blots to identify p15Ink4b as a target of Miz-1. Deletion mapping identified a Miz-1binding region in the promoter of Cdkn2b, the gene that encodes p15Ink4b. Co-expression of Myc blocked transactivation of Cdkn2b by Miz-1, and co-immunoprecipitations identified a complex of Miz-1, Myc and Max. But how does Myc block the ability of Miz-1 to activate transcription of Cdkn2b? The authors wondered whether Myc might prevent Miz-1 from recruiting a co-activator and their speculations proved correct: inactivation of the co-activator protein p300 inhibited Miz-1’s ability to activate Cdkn2b transcription, and Miz-1 co-immunoprecipitated with p300. Mapping of the p300-binding site on Miz-1 showed that Myc and p300 both bind to overlapping sites on Miz-1, explaining how Myc blocks reruitment of p300 to the Cdkn2b promoter. But does Myc use this mechanism to make cells proliferate? To find out, the authors made chimeras of Myc and Mad-1 that have Myc’s transcriptional activator activity, but can’t bind Miz-1. Unlike wild-type Myc, the chimeras couldn’t block accumulation of p15Ink4b, and couldn’t transform p53–/– cells. Myc’s oncogenic abilities threfore depend on its ability to block Cdkn2b transcription.

Transforming growth factor-β (TGF-β) can also induce expression of p15Ink4b by activating Smad transcription factors, but why this response is blocked by overexpression of c-Myc has remained a mystery. Joan Seoane and colleagues now explain why. Treatment of keratinocytes with TGF-β downregulated Myc and decreased levels of the Myc–Miz-1 complex. To see whether this upregulates transcription of Cdkn2b, they created a dominantnegative form of Miz-1, Miz-dZF, that binds Myc but not the Cdkn2b promoter. Overexpression of MizdZF increased expression from a Cdkn2b reporter contruct, but not as well as treatment with TGF-β. So TGF-β must activate transcription of Cdkn2b, as well as relieving repression of its transcription by Myc. Reasoning that Smads were the most likely candidates for this bipartite effect of TGF-β, the authors located a Smad-binding region in the Cdkn2b promoter, and immunoprecipitated a Smad-containing complex from the region. Forced expression of Myc didn’t prevent formation of this complex, but it did block its ability to transactivate Cdkn2b, so Myc’s ability to repress transcription dominates the ability of Smads to activate it. The authors identified a ternary complex of Myc, Miz-1 and Smad4, so Myc and Smads don’t battle it out by binding to the same site on Miz. The most likely explanation is that Myc’s ability to represss transcription — at least of Cdkn2b — is due to its ability to block the interaction between Miz-1 and the co-activator p300. Whether this is a general mechanism by which Myc represses transcription is an exciting possibility. Cath Brooksbank Editor, Nature Reviews Cancer References and links ORIGINAL RESEARCH PAPERS Staller, P. et al. Repression of p15Ink4b expression by Myc via association with Miz-1. Nature Cell Biol. 3, 392–399 (2001) | Seoane, J. et al. Concerted TGFβ inputs via Myc, Miz and Smad proteins control the Cdk inhibitor p15Ink4b. Nature Cell Biol. 3, 400–408 (2001)

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

P R OT E I N D E G R A D AT I O N

Water, water everywhere … Cells in water-transporting tissues, such as the kidney and the airways of the lung, achieve a remarkable feat — they survive large fluctuations in osmolarity without either exploding or shrivelling up. To do this, the cell recruits the help of water channel proteins — the aquaporins (AQPs) — which allow extremely high water permeability at the plasma membrane. When cells are under hypertonic stress, the expression of AQP1 is induced. Little is known about the role that AQP1 protein stability or post-translational modifications might have during this process. Reporting in Proceedings of the National Academy of Sciences, Leitch and colleagues now describe how expression of AQP1 is upregulated under hypertonic stress, not simply by inducing its expression, but also by preventing its degradation. Many proteins in the cell are targeted for degradation by a post-translational modification — multi-ubiquitylation — which recruits the proteasome, sending proteins to their death. This pathway, although known to regulate numerous mammalian cytosolic and membrane receptor proteins, has been implicated for only a few membrane transport proteins. The first hint that the ubiquitylation–proteasome pathway might be important for degrading AQP1 came when the authors looked at the effects of proteasome inhibitors on AQP1 levels. Inhibiting the proteasome, they found, increased AQP1 expression. This led them to wonder whether AQP1 is ubiquitylated. To address this, they precipitated AQP1 with antibodies and then looked for the presence of ubiquitin — a test that confirmed their suspicions. So if ubiquitylation regulates the levels of AQP1 in the cell, what happens when cells are exposed to hypertonic stress? Leitch and co-workers found that ubiquitylation of AQP1 actually decreased during hypertonic stress. This suggests that one way the cell might react to stress is to hang on to any AQP1 that it already has, by preventing its degradation. To test this idea, the authors used metabolic labelling to follow the fate of AQP1 in different conditions. And, consistent with this model, they found that the half-life of AQP1 increases markedly under conditions of hypertonic stress compared with normal conditions. The conclusion, say the authors, is that this mechanism “functions to facilitate protein induction at a time when the general pressure on the cell is to reduce protein synthesis”. But how general a mechanism is this for the induction of proteins that are required during stress? Understanding how AQP1 and the other members of the aquaporin family are regulated is imperative, particularly in light of pathophysiological conditions in which their expression is altered. Alison Schuldt References and links ORIGINAL RESEARCH PAPER Leitch, V. et al. Altered ubiquitination and stability of

aquaporin-1 in hypertonic stress. Proc. Natl Acad. Sci. USA 98, 2894–2898 (2001) FURTHER READING Weissman, A. M. Themes and variation on ubiquitylation. Nature Rev.

Mol. Cell Biol. 2, 169–178 (2001)

VOLUME 2 | APRIL 2001 | 2 3 5

REVIEWS THE CAP-TO-TAIL GUIDE TO mRNA TURNOVER Carol J. Wilusz*, Michael Wormington* ‡ and Stuart W. Peltz* ‡ The levels of cellular messenger RNA transcripts can be regulated by controlling the rate at which the mRNA decays. Because decay rates affect the expression of specific genes, they provide a cell with flexibility in effecting rapid change. Moreover, many clinically relevant mRNAs — including several encoding cytokines, growth factors and proto-oncogenes — are regulated by differential RNA stability. But what are the sequence elements and factors that control the half-lives of mRNAs?

*Department of Molecular Genetics and Microbiology, Robert Wood Johnson Medical School–UMDNJ, Piscataway, New Jersey 08854, USA. ‡PTC Therapeutics, 100 Corporate Court, South Plainfield, New Jersey 07080, USA. Correspondence to S.W.P. e-mail: speltz@ptcbio.com

The flow of genetic information in all cells progresses from DNA to messenger RNA to protein. Even for genes with the simplest form of regulation, many events must occur precisely to generate the protein product accurately and efficiently. As a consequence, cells have evolved a ‘gene-expression factory’ that encompasses the routing of a nascent transcript through multimeric mRNA–protein complexes that mediate its splicing, polyadenylation, nuclear export, translation and ultimate degradation. Regulation of mRNA decay rates is an important control point in determining the abundance of cellular transcripts. Decay rates of individual mRNAs differ extensively — whereas some mRNAs decay with halflives that are 100-fold shorter than cellular generation times, others have half-lives spanning several cell cycles1. Moreover, although the decay rates of most transcripts such as those encoding housekeeping genes are invariant, the half-lives of numerous mRNAs change markedly in response to environmental cues. These differences in mRNA decay rates have notable effects on the expression of specific genes, and provide the cell with flexibility in effecting rapid change in transcript abundance. For example, upon alterations in the transcription rate, unstable mRNAs will attain new steady-state levels in the least amount of time. The biosynthesis of a translationally competent eukaryotic mRNA requires many events to be initiated by RNA polymerase II (REF. 2). A 5′ 7-methylguanosine cap structure is incorporated co-transcriptionally, and a 3′ poly(A) tail is added by the poly(A) polymerase

immediately after transcription (FIG. 1). These two modifications are crucial determinants of efficient processing, nuclear export and translation, and are also important in maintaining transcript stability. Many clinically relevant mRNAs are regulated by differential RNA stability3, and the aberrant control of mRNA stability has been implicated in disease states, including cancer, chronic inflammatory responses and coronary disease. Turnover of mRNA can also be used to block the expression of certain mutant genes. Many genetic diseases, such as cystic fibrosis and Duchenne muscular dystrophy, can be caused by mutations that generate premature stop codons. Transcripts encoded by such nonsense-containing alleles are targeted for rapid degradation by a mechanism known as nonsense-mediated decay. This surveillance process prevents the generation of truncated and potentially deleterious proteins. The goal of this review is to summarize the mechanisms of mRNA decay, highlighting a subset of sequence elements and factors that determine the halflives of mRNAs. Our aim is to highlight key events in mRNA decay, and interested readers should also consult other review articles on this subject4–6. Deadenylation-dependent mRNA decay

A principal mRNA-degradation pathway used in both yeast and higher eukaryotes is initiated by removal of the 3′ poly(A) tail (FIG. 1). Studies in yeast have been particularly fruitful in dissecting the mechanism of this deadenylation-dependent decay pathway. Two distinct deadenylating activities have been identified

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 3 7

REVIEWS

7-me GpppG

AAAAAAAAAAAAAAAAAAAA

5′ cap

3′ poly(A) tail Deadenylation

7-me GpppG

AAAAAAAAAAAAA

Decapping enzyme

Decapping

Deadenylase

e pp pG 7-m G

exo 5′

3′ exo

Exonucleolytic decay

Figure 1 | Deadenylation-dependent mRNA decay. When mRNA processing is complete, the mRNA bears a 5′ cap structure and 3′ poly(A) tail that protect the message from exonucleolytic decay. The first step in the decay of most wild-type mRNAs is shortening of the poly(A) tail by a deadenylase (blue). Once poly(A) shortening is complete, the 5′ 7methylguanosine cap is rapidly removed and the rest of the mRNA is attacked by 5′ and 3′ exonucleases (green and pink, respectively).

in yeast. The poly(A)-binding protein (Pab)1-dependent Pan2/Pan3 (poly(A) nuclease) complex is proposed to trim nascent poly(A) tails in the nucleus before export7, but might also participate in cytoplasmic deadenylation8,9. A second deadenylating activity is postulated to consist of two previously defined transcriptional regulators (carbon catabolite repression 4 (Ccr4)/ Ccr4-associated factor (Caf1))9,10. However, although Ccr4/Caf1 are required for optimal poly(A) nuclease activity both in vivo and in vitro, neither protein has been shown directly to be a component of the deadenylating activity. Notably, both the Pan2/Pan3 and Ccr4/Caf1 complexes must be disrupted to have an appreciable effect on deadenylation rates in vivo, indicating a possible redundancy in function 9. Nonetheless, this report raises the intriguing possibility that transcriptional and post-transcriptional gene regulation might be coupled either by the involvement of dual-function proteins such as CCR4/CAF1, or by coupling these processes through protein–protein interactions such as those previously characterized between RNA polymerase II and various splicing/polyadenylation factors2.

Table 1 | Factors required for mRNA decapping in yeast Factor

Function and phenotypes

Homologies and motifs

Decay pathway

Dcp1

Decapping enzyme11,12 Catalyses removal of a 7-methyl-GDP moiety from the 5′ cap11 Deletion suppresses lethality of a pab1∆ strain15

Novel protein

Required for all mRNA decay11,12

Dcp2

Required for production of active Dcp1 (REF. 13) Interacts with Dcp1 directly13

MutT motif13

Required for all mRNA decay13

Vps16

Vacuolar protein-sorting factor90

Coiled-coil domains

Required for all mRNA decay14

Pat1/ Mrt1/ Spb10

Protein associated with topoisomerase II (REF. 91) A pat1∆ strain has defects in translation initiation89 Deletion can also suppress the lethality of a pab1∆ strain15,89

Proline and glutamine rich

Required only for deadenylationdependent decay15

Lsm proteins

Sm-like proteins form seven-member ring like complexes27 Nuclear function (Lsm2–8) is in pre-mRNA splicing27 Cytoplasmic function (Lsm1–7) required for mRNA decapping17,18 lsm1∆ strain suppresses the lethality of a pab1∆ strain42 Lsm proteins interact with Dcp1 and Pat1 (REF. 17)

Similar to Sm proteins Homologues in higher eukaryotes and archaea92

Required only for deadenylationdependent decay17

Edc1 and Edc2

Enhancers of decapping93 Suppressors of conditional alleles of dcp1 and dcp2 respectively93 Might facilitate assembly or activation of the decapping complex93 Co-immunoprecipitate with Dcp1 and Dcp2 (REF. 93)

Homologous to each other93

Required for deadenylationdependent decay93 not tested Nonsense-mediated decay (NMD)

Upf1

RNA-dependent ATPase/helicase94 Component of the surveillance complex required for NMD95 Interacts with Upf2, Upf3, Hrp1 and the release factor eRF3 (REF. 70) Plays a role in nonsense suppression94

Class I helicase94 Human RENT1/UPF177,96 C. elegans smg-2 (REF. 74)

Required for NMD only98

Upf2/Nmd2

Component of the surveillance complex required for NMD97,98 Plays a role in nonsense suppression70 Interacts with Upf1, Upf3 and eRF3 (REF. 70)

Contains 4G homology domains80,99 Required for NMD only98 Acidic carboxyl terminus98 Human UPF2 (REFS 76,80)

Upf3

Small shuttling protein100 Component of the surveillance complex required for NMD97,101 Plays a role in nonsense suppression70 Interacts with Upf, Upf2 and eRF3 (REF. 70)

Nuclear export and import signals100 Required for NMD only97,101 Human UPF3A, UPF3B76,81

Hrp1/Nab4

Essential hnRNP-like RNA-binding protein102 Involved in poly(A) site selection72 and NMD71 Perhaps a marker protein that targets a transcript for NMD71

Two RNA recognition motifs Homologous to hnRNP A1 102

238

Required for NMD only71

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

CAP ANALOGUES

Dinucleotides such as 7-meGpppG or GpppG that resemble the 5′ cap structure of messenger RNAs. They can be used to analyse the specificity of cap-binding proteins or to compete them away from the 5′ cap structure.

Poly(A) removal triggers rapid cleavage of the 5′ cap by a decapping enzyme (Dcp1)11,12. Biochemical and genetic analyses have identified several factors involved in decapping, including Dcp2 (REF. 13), Vps16 (REF. 14), Pat1 (REF. 15,16) and the Lsm proteins17,18 (TABLE 1). Although mutations in each of the genes encoding these factors prevent decapping, their roles in either catalysing or regulating this reaction remain to be determined. After decapping, the body of the transcript is degraded by the 5′ exonuclease Xrn1 (REF. 19). There are 3′ exonuclease activities in yeast, but they seem to have only minor functions in normal mRNA decay20. Deadenylation-dependent mRNA decay is important for regulating transcript stability in mammalian cells21. A poly(A)-specific deadenylating nuclease — initially termed DAN, but subsequently designated as PARN (poly(A) ribonuclease) — has been purified and biochemically characterized from both mammalian cells and Xenopus laevis oocytes22–24. The subsequent steps of mRNA decay in multicellular organisms are not

Decapping enzyme 4E

RN PA

PABP

AAAAAAAAAAAAAAAAAAAA

Deadenylation

Decapping

Figure 2 | The deadenylase as an inhibitor of translation initiation and decapping. During translation, the mRNA is thought to be circularized by its interaction with the translationinitiation factors eIF4E (4E), eIF4G (4G) and the poly(A)-binding protein (PABP). The eIF4E protein binds to the 5′ cap structure and this interaction is promoted by its binding to eIF4G. eIF4G is a large protein that can interact with both eIF4E and PABP simultaneously. This conformation protects the 5′ and 3′ ends of the mRNA from attack by the deadenylase and decapping enzymes. We postulate that the deadenylase can somehow invade this closed loop and interact with the cap while simultaneously removing the poly(A) tail. The interaction of poly(A) ribonuclease (PARN) with the cap perpetuates the closed loop and thereby blocks both translation initiation and decapping. When poly(A) shortening is complete, PARN dissociates, allowing the decapping enzyme to hydrolyse the 5′ cap of the message.

as well defined as in yeast, owing to the transient nature of intermediate decay products. However, decapped mRNA decay intermediates have been isolated from mouse liver cells25. Moreover, a decapping activity that can be inhibited by the poly(A) tail has recently been characterized in HeLa cell extracts26, indicating that decapping might be an important step in the deadenylation-dependent turnover of mammalian mRNAs. Although mammalian homologues of the yeast Lsm proteins27 and the Dcp2 protein13 have been identified, their functions remain to be determined. It is also not clear whether the body of deadenylated and decapped RNA is degraded by 5′ or 3′ exonucleases. Although a mammalian homologue of yeast Xrn1 exoribonuclease has been identified28, only 3′ exonuclease activities have been observed in vitro29–31. The poly(A) tail inhibits mRNA decay through its interaction with the poly(A)-binding protein (PABP)32 (FIG. 2). As well as binding poly(A), PABP interacts with a specific region of the translation-initiation factor eIF4G, which in turn forms a ternary complex with the cap-binding protein eIF4E33 (FIG. 2). This complex circularizes the mRNA in vitro34, can promote translation, and might simultaneously stabilize mRNAs by preventing access of deadenylating and decapping enzymes to their targets. Indeed, evidence from recent analyses indicates that accessibility of the 5′ cap to both deadenylating and decapping activities might be a major determinant of mRNA stability26,35,36. But the role of PABP in deadenylation remains to be established. PABP inhibits deadenylation in mammalian cell-free assays32,37 and when overexpressed in Xenopus oocytes39, consistent with its function as a negative regulator of deadenylation. By contrast, yeast strains lacking the PAB1 gene actually show slower deadenylation, suggesting that Pab1 might contribute to optimal poly(A) shortening38. Biochemical studies have shown that the deadenylating nuclease PARN binds directly to the 5′ cap structure on a substrate RNA, and that this interaction stimulates deadenylase activity both in vitro and in vivo36. In addition, both CAP ANALOGUES and eIF4E inhibit PARN in vitro by competing for binding to the 5′ cap35,36. Intriguingly, cap analogues stimulate decapping activity by preventing the binding of both PARN and eIF4E26,35. So one can imagine a sequential process in which an initial event must destabilize the poly(A)–PABP and cap–eIF4E complexes, allowing PARN to bind simultaneously to both the 5′ cap and the 3′ poly(A) tail. We suggest that the interaction of PARN with the cap and poly(A) tail can inhibit decapping, even if the ternary eIF4E, eIF4G and PABP complex is disrupted. Once deadenylation is complete, PARN dissociates and the decapping enzyme can now recognize the cap. These observations indicate that the translational efficiency of an mRNA might be directly coupled to mRNA decay, and that both processes are determined by interactions assembled on the 5′ cap and 3′ poly(A) tail. Translation and mRNA turnover

As noted in the previous section, translation and mRNA decay are intimately linked. Additional support for such

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 3 9

REVIEWS

3′ UNTRANSLATED REGION

Non-coding region that lies 3′ to the protein-coding part of a messenger RNA. Often contains sequences involved in RNA regulation.

linkage comes from the observation that, whereas inhibition of translation initiation destabilizes mRNAs40, inhibiting translation elongation with drugs such as cycloheximide promotes mRNA stabilization41. Perhaps the interaction of the initiation factors with the mRNA is reduced during translation elongation or termination, thereby providing access to decay components. Several studies have investigated the roles of translation factors in mRNA decay in yeast. These found that mutations in the PAB1 gene lead to premature decapping of mRNAs, presumably because Pab1 is required for the poly(A) tail–cap interaction that prevents decapping38. Consistent with this idea, mutations in PAT1, DCP1 or LSM1, any of which reduce decapping, can suppress the lethality of a pab1 deletion15,42. The eIF4E

Box 1 | A+U-rich elements and their associated factors A+U-rich elements (AREs) are defined by their ability to promote rapid deadenylationdependent mRNA decay. However, their sequence requirements are only loosely conserved. AREs were grouped into three classes experimentally by Xu and colleagues103. Bakheet and co-workers45 recently compiled a database of AREcontaining mRNAs, and divided them into five groups. This database is based on reiterations of the AUUUA pentamer and therefore all of these elements fall within class II of the experimental classification. We have attempted to reconcile the two classifications in the Table below. Group

Motif

Examples

WAUUUAW and a U-rich region

c-fos, c-myc

IIA

AUUUAUUUAUUUAUUUAUUUA

GM-CSF, TNF-α

IIB

AUUUAUUUAUUUAUUUA

Interferon-α

IIC

WAUUUAUUUAUUUAW

cox-2, IL-2, VEGF

IID

WWAUUUAUUUAWW

FGF2

IIE

WWWWAUUUAWWWW

u-PA receptor

III

U-rich, non-AUUUA

c-jun

Given the large sequence variation of the ARE, it is not surprising that an abundance of ARE-binding factors have been identified. Some of these factors and their associated functions are summarized below. • AUF1 / hnRNP D Four isoforms generated by alternative splicing104 Binds as a tetramer in vitro105 Binding to ARE correlates with instability51 Has been linked to mRNA stabilization during heat shock53 • HuR/ HuA Ubiquitous ELAV-like protein106 Binding to ARE correlates with stabilization47,47,107 Shuttling protein108 RNA binding modulated by SETα, SETβ, pp32 and APRIL109 Binding also modulated by heat shock110 • HuD, Hel-N1(HuB), HuC Neuronal-specific ARE-binding ELAV-like proteins Hel-N1 influences translation of certain mRNAs111 • Tristetraprolin CCCH zinc-finger protein56 Destabilizes ARE-containing mRNAs49 • TIA-1, TIAR Modulate translation of TNF-α mRNA48,112

240

and eIF4G proteins are also involved in inhibiting mRNA decay, as mutations in these factors accelerate deadenylation and decapping40. The interaction of recombinant eIF4E with the cap can inhibit decapping in vitro, apparently by blocking access of Dcp1 to the cap43. It is unlikely, however, that inhibition of decapping by the poly(A) tail is mediated entirely by the interaction of Pab1 with the eIF4F complex. A temperature-sensitive allele of eIF4E (cdc33-42) that reduces its interaction with the cap does not lead to premature decapping, only to more rapid turnover40. Also, removal of eIF4E from the cap by competition with a cap analogue does not abolish the inhibition of decapping by the poly(A) tail in mammalian extracts26. Dcp1 and eIF4G have been shown to interact (biochemically and genetically), raising the possibility that eIF4G promotes decapping by recruitment of Dcp1 (REF. 44). But it is plausible that eIF4G also inhibits decapping as reduced levels of eIF4G promote more rapid decapping in vivo40. Regulating mRNA decay by cis-acting elements

The turnover of mRNAs is a highly controlled process. Several sequence elements can regulate the rate of turnover of a transcript, either by promoting (destabilizer elements) or by inhibiting (stabilizer elements) decay. One of the best studied and most prevalent is the A+U-rich element (ARE), found in the 3′ UNTRANSLATED REGIONS (3′ UTRs) of some mRNAs encoding cytokines, proto-oncogenes and growth factors3,45 (BOX 1). There are several classes of ARE with slightly different sequence determinants that are characterized by their abilities to promote the rapid deadenylation and subsequent decay of a transcript3. A subset of AREs within certain mRNAs can also mediate mRNA stabilization in the presence of certain stimuli, illustrating a dual role for these regulatory elements. A plethora of ARE-binding proteins have been identified, characterized and cloned, including AUF1/hnRNPD46, HuR47, TIA-1 (REF. 48) and tristetraprolin49 (BOX 1). Binding of these factors to transcripts bearing an ARE can have either negative or positive effects on such diverse processes as stability, translation and subcellular localization of the mRNA. The ARE may therefore be considered a crucial regulator of mRNA function at several steps during the lifetime of a transcript. For example, binding of HuR stabilizes some ARE-containing mRNAs in vivo 50 and in vitro 37. By contrast, the binding of certain isoforms of AUF1 correlates with transcript instability under normal conditions51,52 but can enhance mRNA stability under stress responses such as heat shock53. Tristetraprolin also promotes the decay of ARE-containing tumour necrosis factor (TNF)-α and granulocyte macrophage colony-stimulating factor (GM-CSF) transcripts49,54. Knockout strains of mice harbouring a deletion of the tristetraprolin gene show increased stability of both TNF-α and GM-CSF mRNAs, and show many inflammatory symptoms as a consequence of overexpressing TNF-α54–56. Although AREs and their associated RNA-binding factors have been intensively investigated, it is still unclear how binding of these proteins alters mRNA www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

5′ UNTRANSLATED REGION

ARE

Non-coding region that lies in front of (5′ to) the proteincoding part of a messenger RNA.

4E 4G

PABP

AAAAAAAAAAAAAAAAAAAA

Binding of destabilizing complex (AUF1)

RN PA

Binding of stabilizing complex (HuR)

ARE

ARE AR

P -B

ARE-BP

AAAAAAAAAAAAAAAAAAAA

Rapid deadenylationdependent decay

ARE

A AAAAA

AAAAA

Figure 3 | Model for how the A+U-rich element mediates stability and instability. Interaction of the A+U-rich element (ARE) with a destabilizing factor, such as AUF1, might promote rapid deadenylation by reducing the affinity of the poly(A) binding protein (PABP) for the poly(A) tail. Conversely, stabilizing factors, such as HuR, might enhance binding of the PABP to the poly(A) tail, thus blocking deadenylation.

decay rates (FIG. 3). One possibility is that ARE–protein complexes alter interactions between PABP and poly(A), or between eIF4E and the 5′ cap, thereby providing access to PARN. Alternatively, the ARE-binding proteins might interact directly with the deadenylase itself, modulating its activity. The ARE can also accelerate decapping in the absence of deadenylation in vitro 26, suggesting that it might stimulate deadenylationdependent turnover at many steps in the decay pathway. So perhaps the deadenylation and decapping activities are recruited as a multifunctional complex to ARE-containing substrate mRNAs. Sequence elements can also stabilize mRNAs by inhibiting specific degradation pathways. For example, the α-globin transcript contains a cytosine-rich element in its 3′ UTR that nucleates the formation of the stabilizing α-complex, which contains the α-complex proteins α-CP-1 and α-CP-257. This complex protects the α-globin mRNA from both deadenylation-dependent decay and endoribonucleolytic cleavage58,59 (BOX 2). The α-stability complex interacts cooperatively with PABP such that both show a higher affinity for mRNA59. The enhanced stability of the poly(A)–PABP complex probably inhibits deadenylation. Cis-acting elements that modulate transcript stability can also be found in the 5′ UNTRANSLATED REGION (5′ UTR) and coding region of mRNAs. In some cases, these elements act in concert with 3′ UTR elements to regulate mRNA decay. For example, the c-fos mRNA contains both an ARE in the 3′ UTR and a destabilizing sequence known as the major protein-coding-region

determinant (mCRD) within its coding region60. A complex of five proteins (PABP, PABP-interacting protein (PAIP), AUF1, NS1-associated protein (NSAP1) and Unr (Upstream of N-ras)) assembles on the mCRD and promotes rapid deadenylation and decay of the mRNA61. The mCRD needs to be at least 450 nucleotides proximal to the poly(A) tail and requires continuing translation for its destabilizing function. One model proposes that transit of ribosomes through the mCRD element disrupts the complex and triggers the decay pathway61. In this case, the mCRD complex might protect untranslated c-fos mRNA from the rapid deadenylation-dependent decay that is otherwise promoted by the ARE in its 3′ UTR. What triggers changes in mRNA stability?

Several signalling pathways have been implicated in regulating the decay of specific mRNAs. The interleukin-2 (IL-2) mRNA, for example, has stability determinants in both its 5′ and 3′ UTRs. These sequences act in concert to regulate its stability in response to various extracellular stimuli62, such as phorbol esters, lipopolysaccharide or calcium ionophores. In unstimulated T cells, IL-2 mRNA is inherently unstable. This instability requires the presence of an ARE in the 3′ UTR. After Tcell activation, the mRNA is stabilized through the c-Jun amino-terminal kinase (JNK) signalling pathway62 — a JNK-responsive element (JRE) in the 5′ UTR of IL-2 interacts with two factors, nucleolin and YB-1 (REF. 63). Although binding of both proteins is required for stabilization of IL-2 mRNA, they are probably not direct

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 4 1

REVIEWS

DOWNSTREAM SEQUENCE ELEMENT

Box 2 | Endoribonucleolytic decay

A degenerate sequence element found in the coding region of most mRNAs, which can promote nonsense-mediated decay when located downstream of a premature stop codon.

There are a few messenger Protein-binding site RNAs that degrade by a AAAAAAAAAAAAAAA minor pathway known as Endonuclease endoribonucleolytic decay. recognition site Endoribonucleases Recognition by Recognition by stabilizer protein endonuclease recognize specific sequence elements within the transcript and cleave the AAAAAAAAAAAAAAA AAAAAAAAAAAAAAA mRNA internally. The Stable mRNA cleavage event generates free 3′ and 5′ ends that are easily accessible to exonucleases and the products of the AAAAAAAAAAAAAAA cleavage reaction are Exonucleolytic decay therefore rapidly degraded. Interestingly, several mRNAs that are degraded by this pathway also interact with proteins that block access of the endoribonuclease to its cleavage site. For example, an endonuclease from Xenopus laevis hepatocytes, PMR1, can cleave the vitellogenin mRNA but its action is prevented by binding of the vigilin protein to a site that overlaps the PMR1 cleavage site113. Similarly, the α-globin mRNA is cleaved at a site in its 3′ UTR by an erythroid-enriched endonuclease. In this case, cleavage is inhibited by binding of the α-CP complex of proteins to an overlapping sequence58.

HETEROLOGOUS NUCLEAR RIBONUCLEOPROTEIN

RNA-binding protein with a nuclear function. Many hnRNPs can shuttle between the nucleus and the cytoplasm, indicating that they might function in nuclear export of RNA.

targets of the JNK pathway as they bind to the JRE in both unstimulated and activated T cells. As both nucleolin and YB-1 have RNA-unfolding activity, they have been proposed to remodel the RNA to allow a JNK-activated stabilization factor to bind63. By contrast, the JNK pathway does not mediate stabilization of IL-6 or IL-8 mRNAs, both of which nonetheless contain AREs in their 3′ UTRs64. Cytokineinduced stabilization of these transcripts apparently requires the p38 mitogen-activated protein kinase (MAPK) pathway, as constitutive activation of kinases within this pathway (for example, MEKK1, MKK6, MK2) mimics the stabilizing effect of cytokine treatment64. GM-CSF and c-fos mRNAs (which contain AREs) are also stabilized by these activated kinases. Another ARE-containing mRNA, cox-2, can be destabilized by the glucocorticoid dexamethasone, which inhibits the p38 MAPK pathway65. These results implicate a role for the p38 MAPK pathway in the general regulation of ARE-dependent mRNA stability. However, increases in intracellular calcium levels (elicited by many stimuli in diverse cell types, including nonimmune cells) stabilize ARE-containing transcripts — presumably by altering the activities of either decay components or stability factors. Thus, ARE-mediated mRNA stabilization is not specific to either immune cells or interleukin mRNAs66. Nonsense-mediated decay

Perhaps the strongest evidence for a link between translation and turnover is the nonsense-mediated decay (NMD) pathway. This pathway ensures that mRNAs bearing premature stop codons are eliminated as templates for translation. NMD substrates include mutant mRNAs as well as several wild-type transcripts containing upstream open reading frames67 that mimic premature termination codons. So how are premature and normal stop codons distinguished?

242

The pathway has been studied extensively in yeast, in which a premature stop codon is recognized during translation termination, resulting in rapid, deadenylation-independent decapping68 (FIG. 4). The same decapping enzyme hydrolyses the cap structure as for the deadenylation-dependent decay pathway, but at least some of the regulatory factors differ. For instance, lossof-function mutations in Lsm proteins and Pat1 do not affect NMD (TABLE 1). Four factors — Upf1, Upf2, Upf3 and Hrp1 — are essential for NMD in yeast. All interact with the translation-release factor RF3, and they are also involved in suppressing nonsense codons69,70 (TABLE 1). According to current hypotheses, a ‘surveillance’ complex (consisting of the Upf proteins) associates with the release factors at translation termination. After hydrolysis of the peptidyl–tRNA bond, the complex scans the downstream part of the transcript for a signal that the termination was premature. In yeast, this signal consists of a loosely conserved DOWNSTREAM SEQUENCE ELEMENT (DSE). Hrp1, a shuttling HETEROLOGOUS NUCLEAR RIBONUCLEOPROTEIN (hnRNP)-like factor, interacts specifically with both the DSE and Upf1 (REF. 71). A temperature-sensitive allele of Hrp1 stabilized nonsense-containing transcripts and prevented Upf1 binding. It is possible that Hrp1 acts as a ‘molecular tag’ to target nonsense-containing transcripts for rapid NMD. When the DSE is located within normal coding sequences, the Hrp1–DSE interaction would be disrupted by ribosome transit. But if translation is terminated prematurely, the DSE is now positioned in an untranslated region and Hrp1 binding persists. Hrp1 has also been implicated in poly(A)-site selection of pre-mRNAs in the nucleus72. So Hrp1 might bind during mRNA processing and be exported to the cytoplasm while bound to the transcript. Although nonsense-mediated decay clearly occurs in higher organisms, the pathway is less well characterized www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

MS2 COAT PROTEIN

A specific mRNA-binding protein from bacteriophage MS2 that recognizes a stem–loop structure. MS2 coat protein is often used to tether other proteins to RNAs. POLYSOME

Polyribosome; two or more ribosomes attached to different points on the same strand of mRNA.

than in yeast. For example, we do not know the mechanism of decay (rapid deadenylation-independent decapping?) or even its subcellular localization (nuclear or cytoplasmic?). Several genes (smg-1 to smg-7 (REF.73)) in Caenorhabditis elegans are required for NMD, and, when mutated, they allow normally recessive nonsense-containing alleles of the unc-54 gene to act dominantly by stabilizing the nonsense-containing mRNA and thereby allowing the synthesis of cytotoxic, truncated unc-54 protein. Only three smg genes have been cloned and characterized; smg-2 encodes a phosphoprotein with homology to Upf1 (REF. 74), smg-7 encodes a novel protein with no obvious yeast homologue75 and smg-4 is homologous to yeast Upf3 (REF. 76). The human homologue of Upf1 (UPF1, encoded by the RENT1/HUPF1 gene) has similar enzymatic characteristics77 and, like SMG-2, it is a phosphoprotein78. Expression of a dominant-negative UPF1 in vivo impairs the rapid decay of nonsense-containing mRNAs79, indicating its functional conservation. Human homologues of yeast Upf2 and Upf3 have also recently been identified76,80, and these are likely to be functionally equivalent to their yeast homologues. Lykke-Andersen and colleagues81 recently showed that

when UPF2 or UPF3 are fused to MS2 COAT PROTEIN and expressed together in vivo with a β-globin mRNA bearing MS2-binding sites in its 3′ UTR, they can promote rapid decay similar to that caused by premature stop codons. Although UPF1 is cytoplasmic and associated with POLYSOMES, UPF2 localizes to the perinuclear membrane and UPF3 shuttles between the nucleus and cytoplasm. These observations indicate that UPF3 might associate with nascent transcripts in the nucleus, remain bound during nuclear export, and perhaps interact with UPF2 at the nuclear membrane. One hypothesis81 is that the first round of translation occurs while the mRNA is still being translocated through the nuclear pore. This might explain why it is not clear whether NMD occurs in the nucleus or the cytoplasm. There is some evidence for DSE-like sequences (also called fail-safe sequences) in mammalian cells. However, an intron found 3′ of the termination codon has been shown to function like yeast DSEs82. To be recognized as premature, a termination codon must lie upstream of the last intron, and this has led to the hypothesis that a marker protein is deposited at the splice junction during splicing. This marker protein could have the same role as Hrp1 in yeast, and interact

STOP

AAA

STOP

Surveillance

Processing 2 1 Nuclear pore

PABP

STOP

AAAAAAAAAA

Stop codon recognition 2

Export

STOP

Marker recognition

STOP

AAAAAAAAAA STOP

Translation Ribosome commences

Premature termination codon

AAAAAAAAAA

Splicing-dependent marker (UPF3?) Release of cap and decapping

Nuclear cap-binding complex Poly(A)-binding protein 1

UPF1

UPF2

1 Nuclear membrane

AAAAAAAAAA

Release factors elF4G Decapping enzyme

Figure 4 | Nonsense-mediated mRNA decay. The figure shows one possible model for how premature stop codons are recognized and trigger rapid mRNA decay in mammalian cells. During splicing, specific marker proteins are deposited on the mRNA close to the exon–exon junctions. The mRNA is then exported to the cytoplasm where translation ensues, even before translocation through the nuclear pore is complete and perhaps while the message is still associated with the nuclear capbinding complex. The passage of the ribosome can disrupt the binding of the marker proteins such that a wild-type mRNA will no longer be marked after one round of translation. At translation termination, a ‘surveillance’ complex assembles that contains UPF1, and this complex scans downstream of the stop codon for marker proteins. If the surveillance complex encounters a mark then the mRNA is remodelled, leading to disruption of the cap-binding complex and rapid decapping of the mRNA.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 4 3

REVIEWS with the surveillance complex to trigger NMD. Consistent with this theory, UPF3 was found associated with spliced mRNA in vivo, but not with unspliced transcripts81. As UPF3 is a shuttling protein, it might act as a splicing-dependent marker. Several other proteins have also been identified that interact with mRNA in a splicing-dependent manner — Y14 (REFS 83,84), ALY /REF84,85, SRm160 (REFS 84,86), RNPS1 (REF. 84) and DEK84,87. At least two of these, Y14 and ALY, are also exported to the cytoplasm. It is tempting to speculate that these proteins are involved in marking the exon–exon boundaries so that the surveillance machinery can determine whether a stop codon is bona fide. However, the association of these proteins could simply be a prerequisite for mRNA export. One question in both yeast and mammalian systems is how the recognition of a premature stop codon triggers NMD. The idea that NMD occurs only during the first round of translation might provide a clue. Last year, Fortes and colleagues88 showed that the yeast nuclearcap-binding complex (CBC) can interact with eIF4G and promote translation initiation. The CBC is exported with the mRNA and is exchanged in the cytoplasm for the eIF4F complex that normally mediates translation initiation. Perhaps the nuclear CBC mediates the first round of translation and this is how it is distinguished from later rounds. Another link to translation initiation comes from the UPF2 protein, which contains two regions with weak homology to the translation-initiation factor eIF4G within its eIF4A- and eIF3-binding domain80. These domains seem to be functionally significant and are conserved in Saccharomyces cerevisiae and Schizosaccharomyces pombe. Mendell and colleagues80 have proposed that, on recognition of a premature stop codon, the UPF2 protein displaces eIF4G from the translation-initiation complex. This could lead to disruption of the cap-binding complex and allow the decapping enzyme access to the 5′ cap.

unclear. They might be required for assembly of the decapping complex, or perhaps for communication between the deadenylase and the decapping enzyme. In vitro assays might also be invaluable in dissecting the mechanisms of regulation, and perhaps isolation of in vivo decay intermediates will be informative. The interplay between translation and mRNA decay requires more detailed investigation. As translation-initiation factors are clearly involved in regulating mRNA stability, mRNA decay factors probably affect translation. For example, the interaction of the mammalian PARN with the 5′ cap might be expected to compete with eIF4E binding and inhibit translation initiation. This ‘invasion of the closed loop’ might indeed be the initial event that triggers deadenylation and subsequent decapping. Support for this model comes from analyses of mutant alleles of the yeast decapping factor Pat1, which show translation-initiation defects89. Another crucial question is how mRNA decapping is triggered in response to a premature stop codon. Although the evidence is strong for marker proteins deposited during mRNA splicing, we do not yet understand how these factors are recognized or how this leads to rapid decapping. All of these issues apply to general mechanisms of mRNA decay, but there are also questions about how decay of specific mRNAs is regulated. How do cis-acting elements within the mRNA sequence promote or inhibit mRNA decay? We need to identify more of the factors binding to these elements and to study their interactions with the degradation machinery and PABP. Work is already in progress to develop drugs that interfere with both NMD and deadenylation-dependent decay in the hope of alleviating disease.An understanding of mRNA decay should allow us to modulate mRNA stability and thereby regulate gene expression in the future.

Links Future directions

Although much progress has been made towards identifying the factors involved in mRNA decay and their mechanisms of action, there are still many questions. We have yet to purify and reconstitute the full complement of enzymes and regulatory cofactors that are required to degrade an mRNA. The development of an in vitro decapping assay in extracts from mammalian cells should facilitate the purification and cloning of the relevant decapping enzyme. The role of auxiliary decapping factors — such as the Lsm proteins and Pat1 — is

2. 3.

4. 5.

244

Herrick, D., Parker, R. & Jacobson, A. Identification and comparison of stable and unstable mRNAs in Saccharomyces cerevisiae. Mol. Cell. Biol. 10, 2269–2284 (1990). Proudfoot, N. Connecting transcription to messenger RNA processing. Trends Biochem. Sci. 25, 290–293 (2000). Chen, C. Y. & Shyu, A. B. AU-rich elements: characterization and importance in mRNA degradation. Trends Biochem. Sci. 20, 465–470 (1995). Hilleren, P. & Parker, R. Mechanisms of mRNA surveillance in eukaryotes. Annu. Rev. Genet. 33, 229–260 (1999). He, W. & Parker, R. Functions of Lsm proteins in mRNA

DATABASE LINKS cystic fibrosis | Duchenne muscular dystrophy | Pab1 | Pan2 | Pan3 | Ccr4 | Caf1 | Dcp1 | Dcp2 | Vps16 | Pat1 | PARN | PABP | eIF4G | eIF4E | Lsm1 | AUF1 | HuR | TIA-1 | tristetraprolin | GM-CSF | NSAP1 | Unr | IL-2 | MEKK1 | MKK6 | Upf1 | Upf2 | Upf3 | Hrp1 | unc-54 | smg-2 | smg-7 | smg-4 | UPF1 | UPF3 | Y14 | ALY | SRm160 | RNPS1 | DEK | FURTHER INFORMATION Peltz home page | 3′ UTR database | mRNA decay resource page ENCYCLOPEDIA OF LIFE SCIENCES mRNA stability | mRNA turnover

degradation and splicing. Curr. Opin. Cell Biol. 12, 346–350 (2000). Conne, B., Stutz, A. & Vassalli, J. D. The 3′ untranslated region of messenger RNA: a molecular ‘hotspot’ for pathology? Nature Med. 6, 637–641 (2000). Brown, C. E. & Sachs, A. B. Poly(A) tail length control in Saccharomyces cerevisiae occurs by message-specific deadenylation. Mol. Cell. Biol. 18, 6548–6559 (1998). Boeck, R. et al. The yeast Pan2 protein is required for poly(A)-binding protein-stimulated poly(A)-nuclease activity. J. Biol. Chem. 271, 432–438 (1996). Tucker, M. et al. The transcription factor associated Ccr4

and Caf1 proteins are components of the major cytoplasmic mRNA deadenylase in Saccharomyces cerevisiae. Cell 104, 377–386 (2001). 10. Chang, M. et al. A complex containing RNA polymerase II, Paf1p, Cdc73p, Hpr1p, and Ccr4p plays a role in protein kinase C signaling. Mol. Cell. Biol. 19, 1056–1067 (1999). 11. LaGrandeur,T. E. & Parker, R. Isolation and characterization of Dcp1p, the yeast mRNA decapping enzyme. EMBO J. 17, 1487–1496 (1998). 12. Beelman, C. A. et al. An essential component of the decapping enzyme required for normal rates of mRNA turnover. Nature 382, 642–646 (1996).

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS 13. Dunckley, T. & Parker, R. The DCP2 protein is required for mRNA decapping in Saccharomyces cerevisiae and contains a functional MutT motif. EMBO J. 18, 5411–5422 (1999). 14. Zhang, S., Williams, C. J., Hagan, K. & Peltz, S. W. Mutations in VPS16 and MRT1 stabilize mRNAs by activating an inhibitor of the decapping enzyme. Mol. Cell. Biol. 19, 7568–7576 (1999). 15. Hatfield, L., Beelman, C. A., Stevens, A. & Parker, R. Mutations in trans-acting factors affecting mRNA decapping in Saccharomyces cerevisiae. Mol. Cell. Biol. 16, 5830–5838 (1996). 16. Bonnerot, C., Boeck, R. & Lapeyre, B. The two proteins Pat1p (Mrt1p) and Spb8p interact in vivo, are required for mRNA decay, and are functionally linked to Pab1p. Mol. Cell. Biol. 20, 5939–5946 (2000). 17. Tharun, S. et al. Yeast Sm-like proteins function in mRNA decapping and decay. Nature 404, 515–518 (2000). 18. Bouveret, E., Rigaut, G., Shevchenko, A., Wilm, M. & Seraphin, B. A Sm-like protein complex that participates in mRNA degradation. EMBO J. 19, 1661–1671 (2000). 19. Hsu, C. L. & Stevens, A. Yeast cells lacking 5′→3′ exoribonuclease 1 contain mRNA species that are poly(A) deficient and partially lack the 5′ cap structure. Mol. Cell. Biol. 13, 4826–4835 (1993). 20. Jacobs, J. S., Anderson, A. R. & Parker, R. P. The 3′ to 5′ degradation of yeast mRNAs is a general mechanism for mRNA turnover that requires the SKI2 DEVH box protein and 3′ to 5′ exonucleases of the exosome complex. EMBO J. 17, 1497–1506 (1998). 21. Shyu, A. B., Belasco, J. G. & Greenberg, M. E. Two distinct destabilizing elements in the c-fos message trigger deadenylation as a first step in rapid mRNA decay. Genes Dev. 5, 221–231 (1991). 22. Korner, C. G. et al. The deadenylating nuclease (DAN) is involved in poly(A) tail removal during the meiotic maturation of Xenopus oocytes. EMBO J. 17, 5427–5437 (1998). 23. Korner, C. G. & Wahle, E. Poly(A) tail shortening by a mammalian poly(A)-specific 3′-exoribonuclease. J. Biol. Chem. 272, 10448–10456 (1997). 24. Martinez, J. et al. A 54-kDa fragment of the poly(A)specific ribonuclease is an oligomeric, processive, and cap-interacting poly(A)-specific 3′ exonuclease. J. Biol. Chem. 275, 24222–24230 (2000). 25. Couttet, P., Fromont-Racine, M., Steel, D., Pictet, R. & Grange, T. Messenger RNA deadenylation precedes decapping in mammalian cells. Proc. Natl Acad. Sci. USA 94, 5628–5633 (1997). 26. Gao, M., Wilusz, C. J., Peltz, S. W. & Wilusz, J. A novel mRNA decapping activity in HeLa cytoplasmic extracts is regulated by AU-rich elements. EMBO J. 20, 1134–1143 (2001). The first biochemical evidence for an mRNAdecapping activity in higher eukaryotes. Also the first demonstration that A+U-rich elements can promote decapping as well as deadenylation. 27. Achsel, T. et al. A doughnut-shaped heteromer of human Sm-like proteins binds to the 3′-end of U6 snRNA, thereby facilitating U4/U6 duplex formation in vitro. EMBO J. 18, 5789–5802 (1999). 28. Bashkirov, V. I., Scherthan, H., Solinger, J. A., Buerstedde, J. M. & Heyer, W. D. A mouse cytoplasmic exoribonuclease (mXRN1p) with preference for G4 tetraplex substrates. J. Cell Biol. 136, 761–773 (1997). 29. Caruccio, N. & Ross, J. Purification of a human polyribosome-associated 3′ to 5′ exoribonuclease. J. Biol. Chem. 269, 31814–31821 (1994). 30. Kwan, C. N. A cytoplasmic exoribonuclease from HeLa cells. Biochim. Biophys. Acta 479, 322–331 (1977). 31. Brouwer, R. et al. Three novel components of the human exosome. J. Biol. Chem. 276, 6177–6184 (2001). 32. Bernstein, P., Peltz, S. W. & Ross, J. The poly(A)-poly(A)binding protein complex is a major determinant of mRNA stability in vitro. Mol. Cell. Biol. 9, 659–670 (1989). 33. Gingras, A. C., Raught, B. & Sonenberg, N. eIF4 initiation factors: effectors of mRNA recruitment to ribosomes and regulators of translation. Annu. Rev. Biochem. 68, 913–963 (1999). 34. Wells, S. E., Hillner, P. E., Vale, R. D. & Sachs, A. B. Circularization of mRNA by eukaryotic translation initiation factors. Mol. Cell 2, 135–140 (1998). 35. Gao, M., Fritz, D. T., Ford, L. P. & Wilusz, J. Interaction between a poly(A)-specific ribonuclease and the 5′ cap influences mRNA deadenylation rates in vitro. Mol. Cell 5, 479–488 (2000). Shows that the mammalian deadenylase is a cap-binding protein and that its activity is stimulated by interaction with the cap in vitro (also see reference 36).

36. Dehlin, E., Wormington, M., Korner, C. G. & Wahle, E. Cap-dependent deadenylation of mRNA. EMBO J. 19, 1079–1086 (2000). Shows that mammalian and Xenopus PARN is a capbinding protein and that activity is stimulated by interaction with the cap in vivo and in vitro (also see reference 35). 37. Ford, L. P., Watson, J., Keene, J. D. & Wilusz, J. ELAV proteins stabilize deadenylated intermediates in a novel in vitro mRNA deadenylation/degradation system. Genes Dev. 13, 188–201 (1999). 38. Caponigro, G. & Parker, R. Multiple functions for the poly(A)-binding protein in mRNA decapping and deadenylation in yeast. Genes Dev. 9, 2421–2432 (1995). 39. Wormington, M., Searfoss, A. M. & Hurney, C. A. Overexpression of poly(A) binding protein prevents maturation-specific deadenylation and translational inactivation in Xenopus oocytes. EMBO J. 15, 900–909 (1996). 40. Schwartz, D. C. & Parker, R. Mutations in translation initiation factors lead to increased rates of deadenylation and decapping of mRNAs in Saccharomyces cerevisiae. Mol. Cell. Biol. 19, 5247–5256 (1999). 41. Beelman, C. A. & Parker, R. Differential effects of translational inhibition in cis and in trans on the decay of the unstable yeast MFA2 mRNA. J. Biol. Chem. 269, 9687–9692 (1994) 42. Boeck, R., Lapeyre, B., Brown, C. E. & Sachs, A. B. Capped mRNA degradation intermediates accumulate in the yeast spb8-2 mutant. Mol. Cell. Biol. 18, 5062–5072 (1998). 43. Schwartz, D. C. & Parker, R. mRNA decapping in yeast requires dissociation of the cap binding protein, eukaryotic translation initiation factor 4E. Mol. Cell. Biol. 20, 7933–7942 (2000). 44. Vilela, C., Velasco, C., Ptushkina, M. & McCarthy, J. E. The eukaryotic mRNA decapping protein dcp1 interacts physically and functionally with the eIF4F translation initiation complex. EMBO J. 19, 4372–4382 (2000). 45. Bakheet, T., Frevel, M., Williams, B. R., Greer, W. & Khabar, K. S. ARED: human AU-rich element-containing mRNA database reveals an unexpectedly diverse functional repertoire of encoded proteins. Nucleic Acids Res. 29, 246–254 (2001). A comprehensive database of A+U-rich elements found in 3′ UTRs of human mRNAs. The database includes both predicted and known AREs and should be an invaluable resource to researchers. 46. Zhang, W. et al. Purification, characterization, and cDNA cloning of an AU-rich element RNA-binding protein, AUF1. Mol. Cell. Biol. 13, 7652–7665 (1993). 47. Myer, V. E., Fan, X. C. & Steitz, J. A. Identification of HuR as a protein implicated in AUUUA-mediated mRNA decay. EMBO J. 16, 2130–2139 (1997). 48. Piecyk, M. et al. TIA-1 is a translational silencer that selectively regulates the expression of TNF-α. EMBO J. 19, 4154–4163 (2000). 49. Lai, W. S. et al. Evidence that tristetraprolin binds to AUrich elements and promotes the deadenylation and destabilization of tumor necrosis factor-α mRNA. Mol. Cell. Biol. 19, 4311–4323 (1999). 50. Peng, S. S., Chen, C. Y., Xu, N. & Shyu, A. B. RNA stabilization by the AU-rich element binding protein, HuR, an ELAV protein. EMBO J. 17, 3461–3470 (1998). 51. DeMaria, C. T. & Brewer, G. AUF1 binding affinity to A+Urich elements correlates with rapid mRNA degradation. J. Biol. Chem. 271, 12179–12184 (1996). 52. Loflin, P., Chen, C. Y. & Shyu, A. B. Unraveling a cytoplasmic role for hnRNP D in the in vivo mRNA destabilization directed by the AU-rich element. Genes Dev. 13, 1884–1897 (1999). 53. Laroia, G., Cuesta, R., Brewer, G. & Schneider, R. J. Control of mRNA decay by heat-shockubiquitin–proteasome pathway. Science 284, 499–502 (1999). 54. Carballo, E., Lai, W. S. & Blackshear, P. J. Evidence that tristetraprolin is a physiological regulator of granulocytemacrophage colony-stimulating factor messenger RNA deadenylation and stability. Blood 95, 1891–1899 (2000). 55. Taylor, G. A. et al. A pathogenetic role for TNF-α in the syndrome of cachexia, arthritis, and autoimmunity resulting from tristetraprolin (TTP) deficiency. Immunity 4, 445–454 (1996). 56. Carballo, E., Lai, W. S. & Blackshear, P. J. Feedback inhibition of macrophage tumor necrosis factor-α production by tristetraprolin. Science 281, 1001–1005 (1998). 57. Wang, X., Kiledjian, M., Weiss, I. M. & Liebhaber, S. A. Detection and characterization of a 3′ untranslated region ribonucleoprotein complex associated with human α-

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

58.

59.

60.

61.

62.

63.

64.

65.

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.

76.

77.

globin mRNA stability. Mol. Cell. Biol. 15, 1769–1777 (1995). Wang, Z. & Kiledjian, M. The poly(A)-binding protein and an mRNA stability protein jointly regulate an endoribonuclease activity. Mol. Cell. Biol. 20, 6334–6341 (2000). Wang, Z., Day, N., Trifillis, P. & Kiledjian, M. An mRNA stability complex functions with poly(A)-binding protein to stabilize mRNA in vitro. Mol. Cell. Biol. 19, 4552–4560 (1999). Veyrune, J. L., Carillo, S., Vie, A. & Blanchard, J. M. c-fos mRNA instability determinants present within both the coding and the 3′ non coding region link the degradation of this mRNA to its translation. Oncogene 11, 2127–2134 (1995). Grosset, C. et al. A mechanism for translationally coupled mRNA turnover: interaction between the poly(A) tail and a c-fos RNA coding determinant via a protein complex. Cell 103, 29–40 (2000). Identification of a multi-protein complex that is required for translation-dependent destabilization of mRNAs containing the c-fos coding region determinant. PABP, PAIP1, Unr, AUF1 and NSAP1 were involved and overexpression of these proteins inhibited deadenylation of mCRD-containing mRNAs. Chen, C. Y., Gatto-Konczak, F., Wu, Z. & Karin, M. Stabilization of interleukin-2 mRNA by the c-Jun NH2terminal kinase pathway. Science 280, 1945–1949 (1998). Chen, C. Y. et al. Nucleolin and YB-1 are required for JNKmediated interleukin-2 mRNA stabilization during T-cell activation. Genes Dev. 14, 1236–1248 (2000). Winzen, R. et al. The p38 MAP kinase pathway signals for cytokine-induced mRNA stabilization via MAP kinaseactivated protein kinase 2 and an AU-rich region-targeted mechanism. EMBO J. 18, 4969–4980 (1999). A study of the kinase pathways that are required for ARE-mediated stabilization. The authors use dominant kinase mutants to distinguish between signalling pathways required for mRNA stabilization in vivo. Lasa, M., Brook, M., Saklatvala, J. & Clark, A. R. Dexamethasone destabilizes cyclooxygenase 2 mRNA by inhibiting mitogen-activated protein kinase p38. Mol. Cell. Biol. 21, 771–780 (2001). Klein, N., Curatola, A. M. & Schneider, R. J. Calciuminduced stabilization of AU-rich short-lived mRNAs is a common default response. Gene Expr. 7, 357–365 (1999). Ruiz-Echevarria, M. J. & Peltz, S. W. The RNA binding protein Pub1 modulates the stability of transcripts containing upstream open reading frames. Cell 101, 741–751 (2000). Muhlrad, D. & Parker, R. Premature translational termination triggers mRNA decapping. Nature 370, 578–581 (1994). Weng, Y., Czaplinski, K. & Peltz, S. W. Identification and characterization of mutations in the UPF1 gene that affect nonsense suppression and the formation of the Upf protein complex but not mRNA turnover. Mol. Cell. Biol. 16, 5491–5506 (1996). Wang, W., Czaplinski, K., Rao, Y. & Peltz, S. W. The role of Upf proteins in modulating the translation read-through of nonsense-containing transcripts. EMBO J. 20, 880–890 (2001). Gonzalez, C. I., Ruiz-Echevarria, M. J., Vasudevan, S., Henry, M. F. & Peltz, S. W. The yeast hnRNP-like protein Hrp1/Nab4 marks a transcript for nonsense-mediated mRNA decay. Mol. Cell 5, 489–499 (2000). Minvielle-Sebastia, L. et al. Control of cleavage site selection during mRNA 3′ end formation by a yeast hnRNP. EMBO J. 17, 7454–7468 (1998). Pulak, R. & Anderson, P. mRNA surveillance by the Caenorhabditis elegans smg genes. Genes Dev. 7, 1885–1897 (1993). Page, M. F., Carr, B., Anders, K. R., Grimson, A. & Anderson, P. SMG-2 is a phosphorylated protein required for mRNA surveillance in Caenorhabditis elegans and related to Upf1p of yeast. Mol. Cell. Biol. 19, 5943–5951 (1999). Cali, B. M., Kuchma, S. L., Latham, J. & Anderson, P. smg-7 is required for mRNA surveillance in Caenorhabditis elegans. Genetics 151, 605–616 (1999). Serin, G., Gersappe, A., Black, J. D., Aronoff, R. & Maquat, L. E. Identification and characterization of human orthologues to Saccharomyces cerevisiae upf2 protein and upf3 protein (Caenorhabditis elegans SMG-4) Mol. Cell. Biol. 21, 209–223 (2001). Bhattacharya, A. et al. Characterization of the biochemical properties of the human Upf1 gene product that is involved in nonsense-mediated mRNA decay. RNA 6, 1226–1235 (2000).

VOLUME 2 | APRIL 2001 | 2 4 5

REVIEWS 78. Pal, M., Ishigaki, Y., Nagy, E. & Maquat, L. E. Evidence that phosphorylation of human Upf1 protein varies with intracellular location and is mediated by a wortmanninsensitive and rapamycin-sensitive PI 3-kinase-related kinase signalling pathway. RNA 7, 5–15 (2001). 79. Sun, X., Perlick, H. A., Dietz, H. C. & Maquat, L. E. A mutated human homologue to yeast Upf1 protein has a dominant-negative effect on the decay of nonsensecontaining mRNAs in mammalian cells. Proc. Natl Acad. Sci. USA 95, 10009–10014 (1998). 80. Mendell, J. T., Medghalchi, S. M., Lake, R. G., Noensie, E. N. & Dietz, H. C. Novel upf2p orthologues suggest a functional link between translation initiation and nonsense surveillance complexes. Mol. Cell. Biol. 20, 8944–8957 (2000). Discovery of human and S. pombe homologues of Upf2 allowed the authors to identify regions of the protein with homology to eIF4G. They were able to show interaction of UPF2 with both human eIF4A and human SUI1 by yeast two-hybrid assays. 81. Lykke-Andersen, J., Shu, M. D. & Steitz, J. A. Human Upf proteins target an mRNA for nonsense-mediated decay when bound downstream of a termination codon. Cell 103, 1121–1131 (2000). An elegant demonstration of the role of human Upf proteins in NMD. The authors show that when tethered downstream of a stop codon, Upf proteins can promote NMD of a wild-type transcript and also show that UPF3 is a shuttling protein that associates specifically with spliced mRNAs. 82. Zhang, J., Sun, X., Qian, Y. & Maquat, L. E. Intron function in the nonsense-mediated decay of β-globin mRNA: indications that pre-mRNA splicing in the nucleus can influence mRNA translation in the cytoplasm. RNA 4, 801–815 (1998). 83. Kataoka, N. et al. Pre-mRNA splicing imprints mRNA in the nucleus with a novel RNA-binding protein that persists in the cytoplasm. Mol. Cell 6, 673–682 (2000). 84. Le Hir, H., Izaurralde, E., Maquat, L. E. & Moore, M. J. The spliceosome deposits multiple proteins 20–24 nucleotides upstream of mRNA exon–exon junctions. EMBO J. 19, 6860–6869 (2000). Identification of several factors that are deposited on mRNAs by the splicing machinery at a consistent distance from the exon–exon junction. Such proteins are candidates for NMD ‘markers’. 85. Zhou, Z. et al. The protein Aly links pre-messenger-RNA splicing to nuclear export in metazoans. Nature 407, 401–405 (2000). 86. Le Hir, H., Moore, M. J. & Maquat, L. E. Pre-mRNA splicing alters mRNP composition: evidence for stable association of proteins at exon–exon junctions. Genes Dev. 14, 1098–1108 (2000). 87. McGarvey, T. et al. The acute myeloid leukemia-associated

246

88.

89.

90.

91.

92.

93.

94.

95.

96.

97.

98.

99.

100.

protein, DEK, forms a splicing-dependent interaction with exon-product complexes. J. Cell Biol. 150, 309–320 (2000). Fortes, P. et al. The yeast nuclear cap binding complex can interact with translation factor eIF4G and mediate translation initiation. Mol. Cell 6, 191–196 (2000). Wyers, F., Minet, M., Dufour, M. E., Vo, L. T. & Lacroute, F. Deletion of the PAT1 gene affects translation initiation and suppresses a PAB1 gene deletion in yeast. Mol. Cell. Biol. 20, 3538–3549 (2000). Horazdovsky, B. F. & Emr, S. D. The VPS16 gene product associates with a sedimentable protein complex and is essential for vacuolar protein sorting in yeast. J. Biol. Chem. 268, 4953–4962 (1993). Wang, X., Watt, P. M., Louis, E. J., Borts, R. H. & Hickson, I. D. Pat1: a topoisomerase II-associated protein required for faithful chromosome transmission in Saccharomyces cerevisiae. Nucleic Acids Res. 24, 4791–4797 (1996). Salgado-Garrido, J., Bragado-Nilsson, E., Kandels-Lewis, S. & Seraphin, B. Sm and Sm-like proteins assemble in two related complexes of deep evolutionary origin. EMBO J. 18, 3451–3462 (1999). Dunckley, T., Tucker, M. & Parker, R. Two related proteins, Edc1p and Edc2p, stimulate mRNA decapping in Saccharomyces cerevisiae. Genetics 157, 27–37 (2001). Czaplinski, K., Weng, Y., Hagan, K. W. & Peltz, S. W. Purification and characterization of the Upf1 protein: a factor involved in translation and mRNA degradation. RNA 1, 610–623 (1995). Leeds, P., Peltz, S. W., Jacobson, A. & Culbertson, M. R. The product of the yeast UPF1 gene is required for rapid turnover of mRNAs containing a premature translational termination codon. Genes Dev. 5, 2303–2314 (1991). Applequist, S. E., Selg, M., Raman, C. & Jack, H. M. Cloning and characterization of HUPF1, a human homolog of the Saccharomyces cerevisiae nonsense mRNAreducing UPF1 protein. Nucleic Acids Res. 25, 814–821 (1997). He, F., Brown, A. H. & Jacobson, A. Upf1p, Nmd2p, and Upf3p are interacting components of the yeast nonsensemediated mRNA decay pathway. Mol. Cell. Biol. 17, 1580–1594 (1997). Cui, Y., Hagan, K. W., Zhang, S. & Peltz, S. W. Identification and characterization of genes that are required for the accelerated degradation of mRNAs containing a premature translational termination codon. Genes Dev. 9, 423–436 (1995). Ponting, C. P. Novel eIF4G domain homologues linking mRNA translation with nonsense-mediated mRNA decay. Trends Biochem. Sci. 25, 423–426 (2000). Shirley, R. L., Lelivelt, M. J., Schenkman, L. R., Dahlseid, J. N. & Culbertson, M. R. A factor required for nonsensemediated mRNA decay in yeast is exported from the nucleus to the cytoplasm by a nuclear export signal

sequence. J. Cell Sci. 111, 3129–3143 (1998). 101. Ruiz-Echevarria, M. J., Yasenchak, J. M., Han, X., Dinman, J. D. & Peltz, S. W. The upf3 protein is a component of the surveillance complex that monitors both translation and mRNA turnover and affects viral propagation. Proc. Natl Acad. Sci. USA 95, 8721–8726 (1998). 102. Henry, M., Borland, C. Z., Bossie, M. & Silver, P. A. Potential RNA binding proteins in Saccharomyces cerevisiae identified as suppressors of temperaturesensitive mutations in NPL3. Genetics 142, 103–115 (1996). 103. Xu, N., Chen, C. Y. & Shyu, A. B. Modulation of the fate of cytoplasmic mRNA by AU-rich elements: key sequence features controlling mRNA deadenylation and decay. Mol. Cell. Biol. 17, 4611–4621 (1997). 104. Wagner, B. J., DeMaria, C. T., Sun, Y., Wilson, G. M. & Brewer, G. Structure and genomic organization of the human AUF1 gene: alternative pre-mRNA splicing generates four protein isoforms. Genomics 48, 195–202 (1998). 105. Wilson, G. M., Sun, Y., Lu, H. & Brewer, G. Assembly of AUF1 oligomers on U-rich RNA targets by sequential dimer association. J. Biol. Chem. 274, 33374–33381 (1999). 106. Ma, W. J., Cheng, S., Campbell, C., Wright, A. & Furneaux, H. Cloning and characterization of HuR, a ubiquitously expressed Elav-like protein. J. Biol. Chem. 271, 8144–8151 (1996). 107. Fan, X. C. & Steitz, J. A. Overexpression of HuR, a nuclear–cytoplasmic shuttling protein, increases the in vivo stability of ARE-containing mRNAs. EMBO J. 17, 3448–3460 (1998). 108. Fan, X. C. & Steitz, J. A. HNS, a nuclear–cytoplasmic shuttling sequence in HuR. Proc. Natl Acad. Sci. USA 95, 15293–15298 (1998). 109. Brennan, C. M., Gallouzi, I. E. & Steitz, J. A. Protein ligands to HuR modulate its interaction with target mRNAs in vivo. J. Cell Biol. 151, 1–14 (2000). 110. Gallouzi, I. E. et al. HuR binding to cytoplasmic mRNA is perturbed by heat shock. Proc. Natl Acad. Sci. USA 97, 3073–3078 (2000). 111. Antic, D., Lu, N. & Keene, J. D. ELAV tumor antigen, HelN1, increases translation of neurofilament M mRNA and induces formation of neurites in human teratocarcinoma cells. Genes Dev. 13, 449–461 (1999). 112. Gueydan, C. et al. Identification of TIAR as a protein binding to the translational regulatory AU-rich element of tumor necrosis factor-α mRNA. J. Biol. Chem. 274, 2322–2326 (1999). 113. Cunningham, K. S., Dodson, R. E., Nagel, M. A., Shapiro, D. J. & Schoenberg, D. R. Vigilin binding selectively inhibits cleavage of the vitellogenin mRNA 3′-untranslated region by the mRNA endonuclease polysomal ribonuclease 1. Proc. Natl Acad. Sci. USA 97, 12498–12502 (2000).

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

mRNA LOCALIZATION: MESSAGE ON THE MOVE Ralf-Peter Jansen Cytoplasmic messenger RNA localization is a key post-transcriptional mechanism of establishing spatially restricted protein synthesis. The characterization of cis-acting signals within localized mRNAs, and the identification of trans-acting factors that recognize these signals, has opened avenues towards identifying the machinery and mechanisms involved in mRNA transport and localization.

Zentrum für Molekulare Biologie der Universität Heidelberg (ZMBH), Im Neuenheimer Feld 282, D-69120 Heidelberg, Germany. e-mail: r.jansen@mail.zmbh.uniheidelberg.de

A key feature of eukaryotic cells is their organization into distinct compartments, each with a distinct set of proteins. Many student’s textbooks give the impression that sorting of proteins is mainly a post-translational event. Most proteins that are imported or integrated into organelles are, indeed, targeted to their destination on the basis of signals in the peptide sequence. But the past 15 years have shown that the sorting of several (mainly cytoplasmic) proteins involves an additional mechanism — messenger RNA localization. mRNA localization is a universal mechanism, with more than 90 localized mRNAs known so far. Most of these are found in oocytes and early embryos1; more than 20 are known in Drosophila melanogaster oocytes or embryos, and over 25 have been found in Xenopus laevis oocytes. The number of localized mRNAs detected in somatic cells is also increasing, with the dendritic compartment of neurons having been shown to accumulate a wealth of localized transcripts2. As well as transporting mRNA during oogenesis and targeting intracellular transcripts in somatic cells, plants seem to transport mRNAs over long distances through their vascular system3. And mRNA localization has been detected in single-celled organisms such as yeast4–6 and protozoa7. Progress has recently been made towards a detailed understanding of mRNA transport mechanisms. Some proteins that recognize localized mRNAs are now thought to assemble with the mRNA inside the nucleus, thereby ‘tagging’ the mRNA for subsequent recognition by the cytoplasmic transport apparatus. We know that cytoplasmic mRNA transport occurs in the form of large ribonucleoprotein (RNP) complexes, which might

contain a large number of mRNAs and proteins. And, finally, with the purification of large RNP complexes we should be able to identify their individual components. Biological function of mRNA localization

Why do cells localize mRNAs? Sometimes mRNA localization might be preferable to protein localization. Because one mRNA molecule can serve as a template for multiple rounds of translation, localizing mRNA rather than protein to the site where the protein is needed is obviously more energy efficient. In addition, there is a special requirement for high local levels of proteins that act as determinants to induce specific cell fates8. The synthesis of these cell-fate determinants must be restricted to defined cytoplasmic positions, as their mislocalization would be disastrous for the cell9,10. Examples of such cell-fate determinants include the products of the ASH1 gene in yeast, and the bicoid (bcd), nanos (nos) and oskar (osk) genes in Drosophila. mRNA localization might also be involved in the cotranslational assembly of supermolecular structures11. Some cytoskeletal proteins assemble, at least in part, during translation as nascent peptides12,13, and co-translational assembly would be facilitated by co-localization of the corresponding mRNAs. Co-translational assembly might explain why several transcripts co-localize in Naegleria, a single-celled protozoan that can switch between amoeboid and motile states. Upon the transition from the amoeboid to the flagellar states, localization of mRNAs encoding proteins of the microtubule cytoskeleton precedes assembly of the basal body, the microtubule-organizing centre7.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 4 7

REVIEWS

α-Stau

Syto 14

Studies on mRNAs that encode proteins from the nucleus, endoplasmic reticulum (ER) or mitochondria indicate that mRNA localization is not only required for cytoplasmic sorting of proteins, but that it also might facilitate efficient import into the nucleus or organelles14–16. In fibroblasts, some transcripts that encode nuclear proteins (including c-myc17 and metallothionein14) are enriched in a perinuclear region (FIG. 1), and this proximity could be important for efficient import into the nucleus14. There are also examples of mRNA targeting to organelles. For example, the yeast Atm1 protein is an ABC 18 TRANSPORTER of the inner mitochondrial membrane . Surprisingly, ATM1 mRNA, which is transcribed in the nucleus, accumulates in the proximity of mitochondria, indicating a role for mRNA localization in mitochondrial-targeting mechanisms — possibly by facilitating cotranslational import. Because the accumulation of ATM1 mRNA around mitochondria is independent of translation, localization by co-translational targeting through the nascent peptide chain has been excluded, indicating that the mRNA itself might be transported to the proximity of mitochondria. Another example is mRNA targeting to specific subdomains of the ER, which has been detected in rice seeds15. Prolamines, a specific class of rice storage protein, are located in distinct ER subdomains (‘prolamine bodies’), around which prolamine mRNAs have been found to be enriched. Finally, asymmetric distribution of mRNAs might be involved in sorting integral membrane proteins to specific subdomains of the plasma membrane. For example, Takizawa and colleagues6 have shown localization of the mRNA encoding a putative ion channel, Ist2. IST2 mRNA is localized to the growing bud of yeast cells. The mRNA is translated in the bud and the protein is eventually inserted into the bud’s plasma membrane. Diffusion into the plasma membrane of the mother cell is blocked by the function of septins, a set of proteins implicated in compartmentalization of the yeast plasma membrane. Although the reason for this asymmetric distribution of Ist2 is not known, it shows that mRNA localization, in conjunction with compartmentalization mechanisms, can result in the generation or maintenance of membrane subdomains. mRNA localization: movement and more

Figure 1 | Examples of localized mRNAs and zip-code-binding proteins. a | β-actin mRNA localized to the leading process in chicken fibroblasts. β-actin mRNA is shown in red, actin is shown in green and nuclei are shown in blue. (Image courtesy of R. H. Singer.) b | gurken mRNA in Drosophila stage 10a oocyte, localized to the anterior–dorsal corner of the oocyte. gurken mRNA is shown in green and Gurken protein is shown in red. (Image courtesy of R. S. Cohen.) c | oskar mRNA localized at the posterior end of a Drosophila oocyte. (Image courtesy of A. Ephrussi.) d | c-myc mRNA staining in a fibroblast cell, showing perinuclear accumulation of the mRNA. (Image courtesy of J. Hesketh.) e | Vg1 mRNA localization to the vegetal pole of a Xenopus stage III–IV oocyte. (Image courtesy of B. Schnapp.) f | ASH1 mRNA localized to the tip of the daughter cell of a mitotic cell pair. g | Co-localization of mammalian Staufen (mStau; red) and mRNA granules stained with SYTO-14 (arrowheads) in dendrites of rat hippocampal neurons. Mitochondria stained with SYTO-14 (arrows) do not co-localize with mStau. (Image courtesy of M. Kiebler.) h | Localization of MAP2 mRNA in the cell body and dendrites neurons. Small insert shows mRNA granules in a dendrite. (Image courtesy of S. Kindler.)

248

How is a non-random distribution of mRNAs achieved? There are three main mechanisms for the asymmetric enrichment of certain mRNAs at specific sites. The first — and most extensively studied — is active, directed transport of mRNAs (BOX 1), which requires a functional cytoskeleton as well as motor proteins that move along these cytoskeletal filaments. Our current view of this process is based on a model that was proposed by James Wilhelm and Ron Vale in 1993 (REF. 19), according to which mRNA localization is a three-step process. The first step is cytoplasmic assembly of the localized mRNA with mRNA-binding proteins that recognize the targeting signals. After mRNA recognition, the resulting RNP complex binds the cognate motor www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

Box 1 | The cytoskeleton and motor proteins Cytoplasmic transport of most localized mRNAs requires a functional cytoskeleton and motor proteins. mRNA transport in large cells — such as oocytes or neurons — is thought to rely on a different kind of track system (microtubules) to that used for transport in smaller cell types such as fibroblasts and yeast (actin microfilaments)70,87. This view has, however, recently been challenged61. In Drosophila blastoderm-stage embryos, transcripts of the pair-rule transcription factor Fushi-tarazu are localized to the apical side of the nucleus facing the outside of the embryo. Although these transcripts need to travel only a few micrometres (a similar distance to ASH1 mRNA during cytoplasmic localization in yeast74), their localization is independent of a functional actin cytoskeleton. But localization is disrupted by microtubule-depolymerizing drugs, indicating a role for microtubules in short-range transport too61. Different kinds of track require different types of train. Active transport along actin filaments requires myosin family motors, whereas directional microtubule-based transport depends on dynein- or kinesin-type motors. All three types of motor have been implicated in mRNA localization70, although yeast Myo4 is the only mRNAlocalizing myosin identified so far4. Brendza and colleagues88 have reported a microtubule-dependent motor that is essential for transport of oskar (osk) mRNA and Staufen protein to the posterior end of the Drosophila oocyte. This motor is the plusend-directed conventional kinesin (kinesin I), consistent with the idea that osk mRNA localizes to the plus end of microtubules in oocytes89. By contrast, dynein motors expected to move minus-end-directed mRNAs such as bicoid (bcd) have not yet been identified. However, the finding that Swallow, a putative bcd mRNA-binding protein, associates with a dynein subunit indicates that we might not have long to wait79. What about the specificity of motors involved in mRNA transport? Do mRNAspecific motors exist, or does the specificity rely on adaptors that connect ribonucleoprotein (RNP) complexes to motors that serve other cellular functions as well? There is, as yet, no clear answer. But yeast Myo4 seems to be an RNP-specific motor4–6; it is not an essential protein in yeast, and there is no indication that it participates in other transport processes such as vesicular transport90. By contrast, kinesin I definitely has functions other than mRNA transport; for example, in fast axonal transport of vesicles91.

ABC TRANSPORTER

Member of a membranespanning transporter protein family containing an ATPbinding cassette (ABC). 3′ UNTRANSLATED REGION

Non-coding region that lies 3′ to the protein-coding part of a messenger RNA; often contains sequences involved in mRNA regulation.

protein, and active transport occurs. Finally, once the RNP or mRNA has been delivered to its destination, it is anchored to block further diffusion of the localized transcript (FIG. 2). Although many new data are consistent with this model, other studies indicate that it must be modified. These include results that point to an essential role for nuclear proteins and proteins that shuttle between the nucleus and the cytoplasm, indicating that nuclear events might contribute to the cytoplasmic localization of mRNAs20–22. A second mechanism for asymmetric mRNA distribution is local stabilization of transcripts23. An example of this is the posterior localization of heat-shock protein (Hsp) 83 mRNA in early Drosophila embryos24. In the Drosophila oocyte, Hsp83 mRNAs are evenly distributed throughout the cytoplasm. After fertilization they are then degraded, with the exception of a fraction at the posterior cytoplasm. This results in a concentrated pool of Hsp83 transcripts at the posterior pole of the embryo. In the Drosophila oocyte, degradation mechanisms sometimes collaborate with active-transport processes to establish posterior localization of mRNAs. Enrichment of nos mRNA at the posterior pole does not depend exclusively on mRNA transport; it also requires removal of the unlocalized bulk of the mRNA25. The third mechanism is facilitated diffusion of mRNA in combination with local entrapment26. Here,

the crucial localization step is anchoring of the mRNA at the target site, about which little is known. Early studies in Xenopus oocytes and chicken fibroblasts revealed a function for the cortical actin cytoskeleton in anchoring27,28. Consistent with this idea, Drosophila oocytes mutant for cytoplasmic tropomyosin II, a protein of the actin cytoskeleton, fail to trap osk mRNA (which encodes a protein that determines posterior structures) at the posterior pole29. But, overall, the nature of the mRNA-immobilizing factors is obscure. One of the few players we know is Staufen (Stau), a double-stranded (ds)RNAbinding protein that is essential for anchoring bcd mRNA (which encodes an anterior cell-fate-determining transcription factor) in the anterior cytoplasm of Drosophila eggs30 (FIG. 3). Stau is probably a component of the bcd mRNA anchoring machinery because it is essential for keeping the mRNA at the anterior; it co-localizes with bcd mRNA at the anterior of the embryo; and it seems to bind dsRNA motifs within the bcd 3′ UNTRANSLATED REGION (3′ UTR)31. Individual mRNAs seem to use different mechanisms to dock to their target sites. In yeast, the tight cortical association of ASH1 mRNA at the tip of the bud (FIG. 1) requires the function of Bni1 and Bud6, two proteins of the cortical actin cytoskeleton. These two proteins have not yet been shown to anchor ASH1 mRNA directly32, but they at least act indirectly, as mutations in BNI1 affect many functions of the actin cytoskeleton33–35. It seems that local synthesis of an encoded protein can be involved in mRNA anchoring. For example, as well as requiring Bni1 and Bud6, capture of the ASH1 transcript at the bud tip requires translation of ASH1 mRNA or its polysomal association36. A similar dependence has been observed in Drosophila oocytes, where local synthesis of the Oskar protein is required to keep the osk mRNA at the posterior pole37. But it is not just proteins that can function as anchors — RNAs can too. Short untranslated RNAs in Xenopus called Xlsirt RNAs are essential for the association of Vg1 mRNA (FIG. 1) with the cell cortex at the vegetal pole of Xenopus oocytes, but not for the transport of Vg1 (REF. 38). Xlsirts contain stretches that are complementary to Vg1 mRNA38 and might mediate Vg1 mRNA association with the cortex. Zip codes and postmen

To reach its destination every letter needs an address, an essential part of which is the zip code (or postal code). The term ‘zip code’ as a definition for mRNA localization signals was proposed by Robert H. Singer in 1993 (REF. 39), and will be used in this review to specify sequences that mediate localization. Zip codes are essential for the localization of a transcript, and their removal or mutation severely impairs targeting and anchoring. Conversely, if a zip code is fused to an otherwise non-localized mRNA (generally a heterologous mRNA such as the lacZ mRNA from Escherichia coli), the zip code can target this mRNA to a specified destination. In most cases zip codes, like other regions involved in eukaryotic mRNA regulation, are

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 4 9

REVIEWS

b Cytoplasmic

a Nuclear 'core complex'

c Association with motor

RNP maturation

assembly

d Anchoring

and transport

Shuttling zip-codespecific hnRNP Cortical cytoskeleton

Cytoplasmic RNPs mRNA Zip code 3′ 5′

Mature transport complex

Cortical anchor

3′ 5′ 3‘ 5‘ General hnRNPs

Export factor

Adaptor protein Anchoring through generated polypeptide?

Motor

Figure 2 | Model for mRNA transport. a | Core complex assembly. Heterologous nuclear ribonucleoproteins (hnRNPs) bind to the transcript (blue) inside the nucleus. Apart from general hnRNPs (yellow circles), there are specific hnRNPs (red ovals) that recognize the mRNA’s localization signal (zip code, red). In the following step, localized (and non-localized) mRNAs assemble with proteins that are involved in mRNA export (orange oval) and the mRNA–RNP complex is exported to the cytoplasm. b | Cytoplasmic maturation. General hnRNPs and export factors shuttle back to the nucleus, whereas specific hnRNPs stay associated with the mRNA (lower part). Alternatively, they could be released from the transcript and replaced by cytoplasmic zipcode-specific RNPs (green ovals, upper part). c | Transport. The mature RNP complex associates with a motor protein (green triangle), most likely using adaptor proteins, and is transported to the target site. Alternatively, the RNP associates with membranous structures (for example, rough endoplasmic reticulum) and is transported by piggyback. d | Anchoring. Finally, the RNP is released from the motor and is tethered to the target site by specific proteins (‘cortical anchor’) or through a translationdependent process. Translation is allowed only at the target site. The model does not take into account the existence of additional mRNA-binding proteins that associate with the transported mRNA; for example, those that repress translation during transport. (Modified from REF. 19.)

STEM LOOP

Messenger RNA secondary structure containing a singlestranded loop region between the base-paired helical stem. NURSE CELL

Auxiliary cell that supplies the oocyte with synthesized messenger RNAs and proteins during insect oogenesis.

250

found in the 3′ UTR, and their function is mediated by mRNA-binding proteins1. What defines a zip code? Their apparent lack of a similarity — at the levels of either primary or secondary structure — means that it is hard to define a common theme (FIG. 4). Zip codes can be short segments with a defined nucleotide sequence40, or repeated short signals, such as in the case of Vg1 or β-actin mRNA41–43. They can also be secondary or tertiary structures such as STEM LOOPS, in which the primary sequence is less important than the structure of the mRNA36,44,45. Finally, localized mRNAs can contain more than one zip code. Multiple targeting signals might have overlapping function, or they might act in sequential targeting steps. For example, various maternal transcripts in Drosophila and Xenopus oocytes are localized through sequential events46–48, and the stepwise localization is reflected in the modular nature of the localization signals in their 3′ UTRs. A well-studied example is bcd mRNA (FIG. 4b). All the signals required for the proper localization of bcd

mRNA to the anterior of the oocyte reside within a 625-nucleotide segment of the 3′ UTR49,50. Deletion analysis defined a 50-nucleotide localization element called BLE1 (bcd localization element 1), which is necessary and sufficient (provided that it is present in two copies) for transport from NURSE CELLS into the oocyte, and for an initial accumulation of bcd at the anterior margin of the oocyte51. Together with BLE1, two stem loop structures — stems IV and V of the 3′ UTR — can drive early mRNA localization (accumulation of the transcript in the oocyte) as well as later stages of mRNA localization (accumulation of the transcript at the oocyte’s anterior end). But they are insufficient to ‘anchor’ the mRNA once it has been localized50. Anchoring requires the additional presence of stem III (REF. 52), completing the set of localization signals within the bcd 3′ UTR. Xenopus Xcat2 mRNA is another example of an oocyte mRNA with a modular localization signal in its 3′ UTR47,48,53. It encodes a protein with similarity to the www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

hnRNP A2 Squid S VgRBP60

ZBP-1 / Vera (VgRBP1)

(Dm) Staufen (Mm) Staufen

VILIP

RRM

dsRBD

Figure 3 | Modular structure of zip-code-binding proteins. Zip-code-binding proteins (or putative binding proteins) with known mRNA-binding motifs are shown. Three classes can be distinguished: hnRNP-like proteins with RRM (mRNA recognition motif) domains (green boxes), ZBP-1-like proteins with RRMs and KH (hnRNP K-like) domains103 (blue boxes), and double-stranded (ds) mRNA-binding proteins with dsRBD (dsRNA-binding domain) motifs (red boxes). Orange boxes represent a split dsRBD found in proteins homologous to Staufen. Proteins such as TB-RBP98 or She263,64, that do not have a recognizable mRNA-binding motif, are not included. (Dm, Drosophila melanogaster; Mm, Mus musculus.)

GERM PLASM

A special cytoplasmic region in (dividing) eggs that contains primary germ-cell-determining factors. MYELIN

Proteins produced by Schwann cells or oligodendrocytes that cause adjacent plasma membranes to stack tightly together. OLIGODENDROCYTE

Glial cell type in the central nervous system with myelincontaining processes that wrap around axons and help to facilitate conduction of electrical signals. 5′ UNTRANSLATED REGION

Non-coding region that lies 5′ to the protein-coding part of a messenger RNA.

Drosophila posterior determinant Nanos53. Xcat2 mRNA is initially targeted to the mitochondrial cloud of the oocyte (an aggregate of mitochondria and electron-dense granules that give rise to the GERM PLASM) by an ‘early’ localization element (also called the mitochondrial cloud localization element)47. Within the cloud, Xcat2 localizes to germinal granules with the help of a second distinct signal48. A similar modularity has been shown for mRNAs in other cell types. One example is the MYELIN basic protein (MBP) mRNA, which is localized in OLIGODENDROCYTES. In these cells, MBP mRNA is transported along microtubules into peripheral processes (the oligodendrocyte’s ‘myelin compartment’) that generate myelin sheets to surround neuronal axons. Its 3′ UTR contains two signals: the RNA-targeting sequence (RTS), which contains 21 nucleotides, and the RNA-localization region (RLR), which is composed of 340 nucleotides. Whereas the RTS is required for localizing MBP mRNA to the cell’s extensions, the RLR is involved in subsequent localization, targeting MBP mRNA to the myelin compartment54. The general view that elements required for mRNA localization are located in the 3′ UTR has, however, been challenged. There are examples in Drosophila55–57 and yeast16,45 of mRNAs with zip codes in the 5′ UNTRANS-

LATED REGION (5′ UTR), or even in the coding region. For example, the mRNA encoding Gurken (Grk), a Drosophila transforming growth factor (TGF)-α-like protein, is localized to the anterior–dorsal corner of the mature oocyte56,57 (FIG. 1). Its zip codes are dispersed throughout the transcript. Whereas a region in the 5′ UTR is necessary for the initial localization steps, the 3′UTR signal seems to be involved in later stages of mRNA localization57. As well as the 3′ UTR, final accumulation of grk mRNA requires a 230-nucleotide element within the coding region, which is thought to protect or anchor the mRNA57. Yeast ATM1 and ASH1 mRNAs also carry localization signals in the coding region16,45. One of the two ATM1 mRNA-targeting signals is found in the 5′-most part of the coding sequence, and it is enough to target a green fluorescent protein (GFP) reporter construct to mitochondria16. ASH1 contains four zip-code elements: three are embedded in the open reading frame, and a fourth element extends from the 3′ end of the coding region into the 3′ UTR36,45 (FIG. 4c). Each ASH1 localization element can transport a reporter mRNA into the daughter cell but, unlike the full-length transcript, the individual elements fail to restrict the reporter to the tip of the daughter cell. This indicates that the localization elements might be redundant for the actual transport, but that they must collaborate during the subsequent step — mRNA anchoring.

Deciphering the zip code

How does a cell interpret the information in a localization zip code? For this, it requires specific mRNA-binding proteins. Although more than 25 zip codes have been characterized1, the cognate zip-code-binding proteins are known for fewer than half 30,58–64. The Drosophila Stau protein is involved in localization of three transcripts (bcd, osk and prospero (pros)) at three different stages of embryogenesis65. Stau is required for anchoring bcd mRNA at the anterior of the embryo; for microtubule-dependent transport of osk mRNA to the posterior of the oocyte; and for actindependent localization of pros mRNA (which encodes a homeodomain transcription factor) in neuroblasts. Stau binds to dsRNA in vitro66, and contains five modules of the dsRBD (dsRNA-binding domain) type, three of which bind dsRNA on their own67 (FIG. 3). However, on its own Stau binds mRNA without apparent specificity, indicating that the specificity might be provided by auxiliary factors52. There are homologues of Drosophila Stau in Caenorhabditis elegans and mammals (TABLE 1), indicating a possible conserved function for this protein in mRNA localization in diverse cell types65, and also a common mechanism for deciphering zip codes during mRNA localization. For instance, Drosophila Stau is found in large RNP particles52 and, similarly, the rat homologue is detected in large, neuronal mRNA-containing granules (also named ‘mRNA granules’) that move into and within dendrites68. Two other proteins that recognize dsRNA and are involved in mRNA localization have also been identified — VILIP, a trkB

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 5 1

REVIEWS

a Vg1 mRNA

mRNAs are localized through different pathways — microtubule- versus actin-dependent transport, respectively27,28.

3′ UTR Vg1LE 3′

How big are transport complexes? UAUUUCUAC

UAUUUCUAC

UUCUAC

UUCAC

b bicoid mRNA (3′ UTR)

V IVb III IVa

Early localization (BLE1) Early and late localization RNA anchoring

I+II 5′

3′

c ASH1 mRNA 5′ UTR

3′ UTR

Coding region

E2a

E2b E3

5′ 5′ 3′ 3′

Figure 4 | Types of mRNA zip codes. a | The Vg1 3’ UTR contains a 340-nucleotide zip code that is sufficient for mRNA localization, termed the Vg1LE, or Vg1 localization element. The functional units in this type of element are short repetitive sequences (either UAUUUCUAC or UUCAC)59,104. b | The bicoid localization element (BLE) in the bicoid 3′ UTR is an example of a zip code with a modular architecture. bicoid mRNA undergoes several sequential transport steps, each involving different, partially overlapping, regions in the highly structured 3′ UTR. (Light blue, early localization; dark blue, early and late localization; purple, mRNA anchoring.) c | ASH1 mRNA is an example of a zip code element that lies in the coding region (E1, E2a, E2b) and in the 3’ UTR (E3). The E3 element also represents an example of a structure-based zip code, because the displayed stem loop structure but not the primary sequence is important for the function of the zip code36,45.

PERIKARYON

Cell body of neurons or glial cells, containing the nucleus and most organelles.

252

(tyrosine kinase B) mRNA-binding protein in dendrites69, and She2, which binds to stem-loop-containing zip codes of yeast ASH1 mRNA63,64. VILIP contains one Stau-like dsRBD module, but She2 contains no known dsRNA-binding or other mRNA-interacting domain. Similarly to Stau, other zip-code-binding proteins involved in mRNA localization also show cross-species conservation. Despite their conservation, they bind different zip codes in different cell types and transport them along different tracks. For example, ZBP-1 (actin zip-code-binding protein 1) binds to the β-actin localization element in chicken fibroblasts22. The Xenopus Vg1RBP (Vg1-mRNA-binding protein) is a homologue of ZBP-1 that recognizes the Vg1 mRNA zip code59,60. Vg1RBP (also called Vera) and ZBP-1 are 78% identical and share the same modular architecture59,60 (FIG. 3). This finding was surprising as the Vg1 and β-actin zip codes differ from one another and the

After their zip codes have been recognized by the corresponding mRNA-binding proteins, mRNAs are carried to their target sites in form of large containers rather than small packages. Large, localized mRNA-containing particles — ‘mRNA granules’ — have been observed in several cell types70. Such studies have relied mainly on visual analyses using mRNA-binding dyes68,71, in vivo detection of proteins that bind to transported mRNAs52,68,72 (see online movie of mStaufen), injection of fluorescently labelled mRNAs52,73 or, more recently, live imaging of RNPs using GFP fused to mRNA-binding proteins (‘green mRNA’; see online movie of Ash1)32,74. One model cell system used extensively to study the formation and transport of mRNA granules in vivo is the oligodendrocyte75. In these cells, MBP mRNA is transported into peripheral processes. An early step in the mRNA translocation is the formation of MBP mRNA granules in the cytoplasm of the PERIKARYON. But whereas mRNA granules are generated in oligodendrocytes irrespective of the mRNA species, only those particles with mRNAs that contain a proper localization signal are transported into the processes of the oligodendrocyte73. The localized particles have been calculated to be around 0.7 µm in diameter — many times the size of a ribosome76. This means that they could accommodate an enormous number of protein and mRNA molecules. Co-localization studies indicate that the MBP RNPs contain not only MBP mRNA and mRNA-binding proteins, but also components of the translation machinery, such as elongation factor-α, transfer mRNA synthetase and even ribosomes76. So in oligodendrocytes (and also in neurons77), a substantial fraction of the protein-synthesis machinery accompanies a localized mRNA to its target site. Large mRNA-containing particles that move in a directional fashion have also been observed in live yeast cells and in Drosophila. A fusion protein between GFP and Exuperantia (Exu), a protein involved in localizing bcd mRNA, was detected in large particles that move from nurse cells into the oocyte, and also within the oocyte72. These particles presumably also contain bcd mRNA. Yeast ASH1 mRNA was detected directly in moving particles using GFP-based reporter systems that recognize mRNA32,74. Co-localization studies showed that the large RNPs represent supermolecular assemblies that contain not only the green reporter mRNA, but also endogenous ASH1 transcripts and the proteins involved in ASH1 mRNA transport74. However, there is a caveat with these results. Size determination in the previous cases relied on optical methods and needs to be supported by biochemical measurements. Moreover, overexpression or microinjection of the corresponding mRNAs could have created artefacts that might not be physiologically meaningful. www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

Table 1 | Cross-species comparison of proteins involved in cytoplasmic mRNA localization Function/feature

Yeast

Drosophila melanogaster

Vertebrates

Zip-code-binding hnRNP protein (located in nucleus and cytoplasm)

Not yet identified

Squid (grk)20

hnRNP A2 (MBP)21 VgRBP60* (Vg1)62 Vera/VgRBP* (Vg1)59,60 ZBP-1* (ß-actin)22

Cytoplasmic zip-codebinding RNP

She2 (ASH1, IST2?)6,63,64

Staufen (osk, pros)65 mStaufen* (?)68,79,80 Swallow* (bcd)82 VILIP* (trk)69 Ypsilon-Schachtel* (osk)83 TB-RBP (CaMKII?)98,99

Motor protein for RNP

Myo4 (ASH1, IST2?)4,6

Kinesin I (osk)88 Dynein* (bcd)82

RNP motor adaptor

She3 (ASH1, IST2?)6,63

Dynein light chain (bcd)82 Not yet identified

mRNA/RNP anchor

Bni1*, Bud6* (ASH1)

Staufen (bcd)30 Oskar (osk)37

Kinesin* (MBP, CaMKII)75,99

Xlsirt mRNAs* (Vg1)38

Target mRNAs, where known, are indicated in parentheses. Proteins shown on the same line are not necessarily homologues, but fulfil similar functions. *Function or feature that has not yet been experimentally proven/reported. (bcd, bicoid; CaMKII, calcium/calmodulin-dependent kinase II; grk, gurken; MBP, myelin basic protein; osk, oskar; RNP, ribonucleoprotein; trk, tyrosine kinase.)

Why so big?

TYPE V MYOSIN

Subclass of the myosin protein family of actin-dependent motor proteins, required for transport of vesicles or messenger RNA cargo.

The most surprising feature of mRNA granules is their size. In some cases, they are bigger than macromolecular structures such as ribosomes. So what are these large, localized RNPs composed of? There are several possibilities. The most simple one assumes that they are, in part, composed of smaller complexes such as ribosomes or polysomes. RNPs of this kind have been detected in oligodendrocytes and neurons76,77. Alternatively, large RNPs could be created from hundreds of copies of localized mRNAs and their corresponding mRNA-binding proteins, which are somehow packed or glued together. An attractive model for how such a ‘glue’ might be generated is based on the assembly of particles containing bcd/Stau in Drosophila embryos78. The formation of large mRNA granules containing injected bcd 3′ UTR and endogenous Stau protein not only required dsRNA structures recognized by Stau, but also intermolecular mRNA–mRNA interactions. Both types of interaction are thought to be essential for chaining mRNAs and mRNA-binding proteins together to create the large mRNA granules that are observed78. Another possible explanation for the large size relies on the assumption that localized RNPs assemble around an as-yet-unknown scaffold. Such a scaffold could be composed of membrane structures, with transported mRNAs moving ‘piggyback’ on transported organelles. There is experimental support for such a speculation. During the development of Xenopus oocytes, Vg1 mRNA moves to the vegetal pole in a microtubule-dependent manner. Vera (Vg1-binding, ER-associated protein), a protein that binds to the Vg1 zip code, is involved in these processes. As its name suggests, Vera is associated with a specific compartment of the oocyte’s rough ER43. It has been proposed that Vera links Vg1 mRNA to the ER, which in turn is transported by kinesin-like motors towards the vegetal pole59. Similarly to Vera, mammalian Stau can co-localize with rough ER79,80. Association with the ER has also been observed for Drosophila grk mRNA56. During oogenesis,

a fraction of the ER that is associated with the nucleus moves from the posterior end of the oocyte to the anterior–dorsal corner. So there are examples that indicate a functional relationship between vesicular structures, such as the ER, and localized mRNAs. However, both Vg1 and grk encode TGF-like proteins that have to be secreted and imported into the ER. A question for the future is whether localized mRNAs are associated with the ER only if those mRNAs encode secreted proteins, or whether it is a general phenomenon that could be the basis for the observed large mRNA granules. Biochemical characterization

Another way to understand why the mRNA-localization granules are so big is to get a ‘biochemical grip’ by purifying them and identifying their components. By combining genetics with biochemical analyses in vivo and in vitro, steps are being taken to understand the protein components of the yeast ASH1 RNP4,5,63,64,81. Five SHE genes are essential for the transport of cytoplasmic ASH1 mRNA4 in yeast, and the functions of three of the corresponding She proteins within the transported RNP have been determined63,64. She1/Myo4 is a TYPE V MYOSIN and the molecular motor of the ASH1 RNP complex (BOX 1). She2 is a new type of dsRNA-binding protein that binds to the localization signals within ASH1 mRNA. She3 seems to be the adaptor that bridges the myosin heavy chain and She2 (REF. 63). The identification of a second mRNA, IST2, localization of which also depends on She2, She3 and Myo4 (REF. 6), indicates that the She proteins might associate with other localized mRNAs and are probably crucial trans-acting factors in the yeast mRNA localization machinery. Whereas She2 and She3 link ASH1 mRNA to a myosin motor protein in yeast, the Swallow (Swa) protein is a putative link between bcd mRNA and cytoplasmic dynein, the motor implicated in localization of bcd 82. Swa, which has two domains with weak similarity to the mRNA-recognition motif (RRM), binds directly to the

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 5 3

REVIEWS

Box 2 | A role for mRNA localization in synaptic plasticity? The dendrites on a typical neuron contain thousands of postsynaptic sites that can respond to and store many pieces of information. Long-lasting forms of activitydependent synaptic modifications (‘memory storage’) are believed to require the local synthesis of proteins at postsynaptic sites92,93. The first indications of local translation came from the discovery of synapse-associated polyribosomes94. Local translation of proteins also implies that mRNAs are transported from the site at which they are synthesized (the nucleus in the cell body) to dendritic sites that are rich in synapses. Localization of mRNA to dendrites has been observed for several mRNAs95,96, but in most cases there is no direct link between the activation of specific synapses and directed mRNA transport to them. Localization of mRNA-containing granules or specific mRNAs to dendrites has been observed2,93,95, and candidates for zip-code-binding proteins involved in neuronal mRNA transport have also been identified95. Among these mRNA-binding proteins is mammalian Staufen. A mammalian Staufen–GFP (green fluorescent protein) fusion protein was detected in two forms of mRNA-containing granules, the smaller ones being able to move within dendrites68. Other proteins implicated in mRNA transport include testis-brain mRNA-binding protein (TB-RBP) and zip-code-binding protein 1 (REFS 97,98). TB-RBP can associate with several mRNAs, among them α-CaMKII (αcalcium/calmodulin-dependent kinase II) mRNA98. Suppressing TB-RBP expression in cultured neurons with antisense oligonucleotides disrupts dendritic α-CaMKII mRNA localization99. A goal in the future will be to functionally link dendritically localized mRNAs to these candidate proteins so that they can be assigned molecular functions. Even if we observe mRNA localization in dendrites, what might be the function of mRNA transport in synaptic modification? Transport of specific mRNAs into dendrites could be a constitutive process, and local translational control could be the key regulatory mechanism for activity-induced protein synthesis. Alternatively, mRNAs could be directly localized to ‘tagged’ postsynaptic sites that have been marked by their activation100,101. In support of this idea, localization of Arc mRNA (encoding a synaptic-activity-regulated cytoskeletal protein) to specific dendritic domains is induced by high-frequency activation of synapses within these domains102. Like Arc mRNA, newly synthesized Arc protein accumulates in dendritic areas that have been activated. The inducible mRNA localization is independent of protein synthesis in the activated dendritic area, indicating a possible mRNA-based targeting mechanism.

dynein light chain82 (a subunit of the dynein motor complex), making it an ideal tool for a future isolation of the bcd-localization complex. In another approach, with the goal of isolating native localized RNP complexes from Drosophila, affinity purification has led to the isolation of Exu-containing RNP complexes. Like Swa, Exu is essential for the correct localization of bcd mRNA (although it might not bind directly to the mRNA), and it has been detected in putative RNP particles moving from nurse cells to the oocyte72. Exu is part of a large mRNA-containing complex, and it purifies together with seven other proteins83. Unexpectedly, the Exu complex is enriched not with bcd mRNA, but with the posteriorly directed osk mRNA. A closer examination of exu mutants indicated a (weak) defect in osk localization, leading to the proposal that different populations of Exu could be involved in localization of diverse RNPs; for example, bcd or osk RNPs83. Cytoplasmic mRNA localization and the nucleus NUCLEAR-EXPORT SIGNAL

Amino-acid sequence required for active transport of certain proteins from the nucleus to the cytoplasm.

254

mRNA localization is a cytoplasmic process, and it was originally assumed that trans-acting factors essential for localization reside in the cytoplasm19. But this assumption has now been challenged. Searches for proteins that recognize zip codes in oligodendrocytes and Xenopus

oocytes, or for genes essential for mRNA localization in Drosophila, have picked up heterologous nuclear ribonucleoproteins (hnRNPs)20–22,61,62. Originally described as mRNA-binding proteins with a nuclear function, it has now become clear that many hnRNPs can shuttle between the nucleus and the cytoplasm, indicating that they might function in the nuclear export of mRNA84,85. But hnRNPs reach out even further. They accompany localized mRNAs not only during nuclear export, but also during subsequent cytoplasmic localization86. A complex of six proteins that bind the zip code of MBP mRNA has been identified. The main protein in this complex is a ‘classical’ hnRNP protein, hnRNP A2 (FIG. 3), previously implicated in nuclear functions such as transcript packaging and splicing21. In oligodendrocytes, hnRNP A2 is preferentially located in the nucleus and cell body, but it can also be seen in particle-like structures in the cell extensions, indicating that it might accompany MBP mRNA to the cell extensions21. This surprising result was just the tip of an iceberg, and a growing body of experimental evidence now indicates distinct cytoplasmic roles for hnRNPs in mRNA localization. The Drosophila hnRNP A homologue, Squid (Sqd) is also required for mRNA localization20. During oogenesis, Sqd is essential for localization of grk mRNA, and the different Sqd isoforms produced by alternative splicing show different localization patterns. Sqd A is mainly cytoplasmic, whereas Sqd B and Sqd S are enriched in the nuclei of nurse cells or the oocyte, respectively. The nuclear isoform Sqd S, but not the cytoplasmic isoform Sqd A, is required for localization of grk mRNA20. By contrast, Sqd A is necessary for efficient control of grk translation20. These results might indicate a function of Sqd S in cytoplasmic mRNA transport. The ZBP-1 protein and its Xenopus homologue Vera/Vg1-RBP22,59,60 both contain NUCLEAR-EXPORT 85 SIGNALS , indicating that they might also shuttle between the nucleus and the cytoplasm, and that they might bind their targets inside the nucleus. A second Vg1 mRNAbinding protein, VgRBP60 (FIG. 3) is a Xenopus homologue of hnRNP I (REF. 62). Although mammalian hnRNP I has been implicated in nuclear mRNA biogenesis, VgRBP60 clearly co-localizes with Vg1 mRNA during cytoplasmic transport62. All of these examples indicate that initial assembly of the mRNA-localization machinery on the zip code might begin in the nucleus, and that the cytoplasmic machinery might recognize a localized mRNA only when it is bound to its partner proteins from the nucleus. Perspective

The mRNA-localization field is moving towards new frontiers. Attention is shifting from an ‘early’ phase, spanning the past ten years, characterized by the identification of localized mRNAs and analysis of the signals that mediate localization, towards a new phase to characterize the machinery that drives mRNA localization. The recently started approaches to purify mRNA complexes, in combination with molecular genetics, should lead to the identification of a ‘near complete’ set of www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS RNPs in the next few years. What will follow soon thereafter is the reconstitution and study of localized RNPs in vitro. New in vivo detection methods (for example, the current ‘green mRNA’ approach) will allow us to study localization in living cells, leading finally to the ultimate challenge of trying to understand, at the molecular level, how cells regulate mRNA localization in the context of other cellular events. In addition, we have just opened the door to the field of dendritic mRNA transport and its function in neural plasticity (BOX 2). Key questions include: what are the (ultrastructural) destination sites of transported mRNAs; what are the retrograde signals that are sent back to the cell body to induce the transport

2. 3.

10. 11. 12.

13.

14.

15.

16.

17.

18.

Bashirullah, A., Cooperstock, R. L. & Lipshitz, H. D. RNA localization in development. Annu. Rev. Biochem. 67, 335–394 (1998). Kuhl, D. & Skehel, P. Dendritic localization of mRNAs. Curr. Opin. Neurobiol. 8, 600–606 (1998). Xoconostle-Càzares, B. et al. Plant paralog to viral movement protein that potentiates transport of mRNA into the phloem. Science 283, 94–98 (1999). First evidence for an endogenous protein in pumpkin that mediates long-distance translocation of mRNAs through the plant’s vascular system. Long, R. M. et al. Mating type switching in yeast controlled by asymmetric localization of ASH1 mRNA. Science 277, 383–387 (1997). Takizawa, P. A., Sil, A., Swedlow, J. R., Herskowitz, I. & Vale, R. D. Actin-dependent localization of an RNA encoding a cell-fate determinant in yeast. Nature 389, 90–93 (1997). Takizawa, P. A., DeRisi, J. L., Wilhelm, J. E. & Vale, R. D. Plasma membrane compartmentalization in yeast by messenger RNA transport and a septin diffusion barrier. Science 290, 341–344 (2000). Han, J. W., Park, J. H., Kim, M. & Lee, J. mRNAs for microtubule proteins are specifically colocalized during the sequential formation of basal body, flagella, and cytoskeletal microtubules in the differentiation of Naegleria gruberi. J. Cell Biol. 137, 871–879 (1997). Driever, W. & Nusslein-Volhard, C. The bicoid protein determines position in the Drosophila embryo in a concentration-dependent manner. Cell 54, 95–104 (1988). Ephrussi, A., Dickinson, L. K. & Lehmann, R. oskar organizes the germ plasm and directs localization of the posterior determinant nanos. Cell 66, 37–50 (1991). Gavis, E. R. & Lehmann, R. Localization of nanos RNA controls embryonic polarity. Cell 71, 301–313 (1992). Singer, R. H. The cytoskeleton and mRNA sorting. Curr. Opin. Cell Biol. 4, 15–19 (1992). Isaacs, W. B., Coo, R. K., Van Atta, J. C., Redmond, C. M. & Fulton, A. B. Assembly of vimentin in cultured cells varies with cell type. J. Biol. Chem. 264, 17953–17960 (1989). L’Ecuyer, T. J., Tompach, P. C., Morris, E. & Fulton, A. B. Assembly of tropomyosin isoforms into the cytoskeleton of avian muscle cells. Pediatr. Res. 43, 813–822 (1998). Levadoux, M., Mahon, C., Beattie, J. H., Wallace, H. M. & Hesketh, J. E. Nuclear import of metallothionein requires its mRNA to be associated with the perinuclear cytoskeleton. J. Biol. Chem. 274, 34961–34966 (2000). Choi, S.-B. et al. Messenger RNA targeting of rice seed storage proteins to specific ER subdomains. Nature 407, 765–767 (2000). The first example of a role of mRNA localization to distinct compartments of the endoplasmic reticulum, thereby facilitating spatial separation of the encoded proteins. Corral-Debrinski, M., Blugeon, C. & Jacq, C. In yeast, the 3´ untranslated region or the presequence of ATM1 is required for the exclusive localization of its mRNA to the vicinity of mitochondria. Mol. Cell. Biol. 20, 7881–7892 (2000). Hesketh, J., Campbell, G., Piechaczyk, M. & Blanchard, J.-M. Targeting of c-myc and β-globin coding sequences to the cytoskeletal-bound polysomes by c-myc 3′ untranslated region. Biochem. J. 298, 143–148 (1994). Leighton, J. & Schatz, G. An ABC transporter in the mitochondrial inner membrane is required for normal growth of yeast. EMBO J. 14, 188–195 (1995).

of mRNAs to the dendritic compartment; and what are the molecular events that ‘tag’ a synapse as a destination site for localized mRNAs?

19. Wilhelm, J. E. & Vale, R. D. RNA on the move: the mRNA localization pathway. J. Cell Biol. 123, 269–274 (1993). 20. Norvell, A., Kelley, R. L., Wehr, K. & Schüpbach, T. Specific isoforms of Squid, a Drosophila hnRNP, perform distinct roles in Gurken localization during oogenesis. Genes Dev. 13, 864–876 (1999). 21. Hoek, K. S., Kidd, G. J., Carson, J. H. & Smith, R. hnRNP A2 selectively binds the cytoplasmic transport sequence of myelin basic protein mRNA. Biochemistry 37, 7021–7029 (1998). References 20 and 21 show for the first time a specific function of mainly nuclear-located hnRNP proteins in recognition and transport of localized mRNAs in oligodendrocytes or Drosophila oocytes. 22. Ross, A. F., Oleynikov, Y., Kislauskis, E. H., Taneja, K. L. & Singer, R. H. Characterization of a β-actin mRNA zipcodebinding protein. Mol. Cell. Biol. 17, 2158–2165 (1997). 23. Ding, D., Parhurst, S. M., Halsell, S. R. & Lipshitz, H. D. Dynamic Hsp83 RNA localization during Drosophila oogenesis and embryogenesis. Mol. Cell. Biol. 13, 3773–3781 (1993). 24. Bashirullah, A. et al. Joint action of two RNA degradation pathways controls the timing of maternal transcript elimination at the midblastula transition in Drosophila melanogaster. EMBO J. 18, 2610–2620 (1999). 25. Bergsten, S. E. & Gavis, E. R. Role for mRNA localization in translational activation but not spatial restriction of nanos RNA. Development 126, 659–669 (1999). 26. Glotzer, J. B., Saffrich, R., Glotzer, M. & Ephrussi, A. Cytoplasmic flows localize injected oskar RNA in Drosophila oocytes. Curr. Biol. 7, 326–337 (1997). 27. Sundell, C. L. & Singer, R. H. Requirement of microfilaments in sorting of actin messenger RNA. Science 252, 1275–1277 (1991). 28. Yisraeli, J. K., Sokol, S. & Melton, D. A. A two-step model for the localization of maternal mRNA in Xenopus oocytes: involvement of microtubules and microfilaments in the translocation and anchoring of Vg1 mRNA. Development 108, 289–298 (1990). 29. Erdelyi, M., Michon, A.-M., Guichet, A., Bogucka Glotzer, J. & Ephrussi, A. Requirement for Drosophila cytoplasmic tropomyosin in oskar mRNA localization. Nature 377, 524–527 (1995). 30. St Johnston, D., Beuchle, D. & Nüsslein-Volhard, C. staufen, a gene required to localize maternal RNAs in the Drosophila egg. Cell 66, 51–63 (1991). 31. Ramos, A. et al. RNA recognition by a Staufen doublestranded RNA-binding domain. EMBO J. 19, 997–1009 (2000). 32. Beach, D. L., Salmon, E. D. & Bloom, K. Localization and anchoring of mRNA in budding yeast. Curr. Biol. 9, 569–578 (1999). 33. Evangelista, M. et al. Bni1p, a yeast formin linking Cdc42p and the actin cytoskeleton during polarized morphogenesis. Science 276, 118–122 (1997). 34. Fujiwara, T. et al. Rho1p–Bni1p–Spa2p interactions: implication in localization of Bni1p at the bud site and regulation of the actin cytoskeleton in Saccharomyces cerevisiae. Mol. Biol. Cell 9, 1221–1233 (1998). 35. Fujiwara, T., Tanaka, K., Inoue, E., Kikyo, M. & Takai, Y. Bni1p regulates microtubule-dependent nuclear migration through the actin cytoskeleton in Saccharomyces cerevisiae. Mol. Cell. Biol. 19, 8016–8027 (1999). 36. Gonzalez, I., Buonomo, S. B. C., Nasmyth, K. & von Ahsen, U. ASH1 mRNA localization in yeast involves multiple secondary structural elements and Ash1 protein

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

translation. Curr. Biol. 9, 337–340 (1999). 37. Rongo, C., Gavis, E. R. & Lehmann, R. Localization of oskar RNA regulates oskar translation and requires Oskar protein. Development 121, 2737–2746 (1995). 38. Kloc, M. &. Etkin, L. D. Delocalization of Vg1 mRNA from the vegetal cortex in Xenopus oocytes after destruction of Xlsirt RNA. Science 265, 1101–1103 (1994). 39. Singer, R. H. RNA zipcodes for cytoplasmic adresses. Curr. Biol. 3, 719–721 (1993). 40. Chan, A. P., Kloc, M. & Etkin, L. D. fatvg encodes a new localized RNA that uses a 25-nucleotide element (FVLE1) to localize to the vegetal cortex of Xenopus oocytes. Development 126, 4943–4953 (1999). 41. Mowry, K. L. &. Melton, D. A. Vegetal messenger RNA localization directed by a 340-nt RNA sequence element in Xenopus oocytes. Science 255, 991–993 (1992). 42. Kislauskis, E. H., Zhu, X. & Singer, R. H. Sequences responsible for intracellular localization of β-actin messenger RNA also affect cell phenotype. J. Cell Biol. 127, 441–451 (1994). 43. Deshler, J. O., Highett, M. I. & Schnapp, B. J. Localization of Xenopus Vg1 mRNA by Vera protein and the endoplasmic reticulum. Science 276, 1128–1131 (1997). First evidence for an association of a zip-codebinding protein (Vera) with the endoplasmic reticulum (ER) of Xenopus oocytes. This work suggests co-transport of the localized Vg1 mRNA with the ER. 44. Serano, T. L. & Cohen, R. S. A small predicted stem-loop structure mediates oocyte localization of Drosophila K10 mRNA. Development 121, 3809–3818 (1995). 45. Chartrand, P., Meng, X.-H., Singer, R. H. & Long, R. M. Structural elements required for the localization of ASH1 mRNA and of a green fluorescent protein reporter particle in vivo. Curr. Biol. 9, 333–336 (1999). 46. Lasko, P. RNA sorting in Drosophila. FASEB J. 13, 421–433 (1999). 47. Zhou, Y. & King, M. L. Localization of Xcat-2 RNA, a putative germ plasm component, to the mitochondrial cloud in Xenopus stage I oocytes. Development 122, 2947–2953 (1996). 48. Kloc, M., Bilinski, S., Pui-Yee Chan, A. & Etkin, L. D. The targeting of Xcat2 mRNA to the germinal granules depends on a cis-acting germinal granule localization element within the 3′ UTR. Dev. Biol. 217, 221–229 (2000). 49. Macdonald, P. M. & Struhl, G. cis-acting sequences responsible for anterior localization of bicoid mRNA in Drosophila embryos. Nature 336, 595–598 (1988). 50. Macdonald, P. M. & Kerr, K. Redundant RNA recognition events in bicoid mRNA localization. RNA 3, 1413–1420 (1997). 51. Macdonald, P. M. RNA regulatory element BLE1 directs the early steps of bicoid mRNA localization. Development 118, 1233–1243 (1993). 52. Ferrandon, D., Elphick, L., Nüsslein-Volhard, C. & St Johnson, D. Staufen protein associates with the 3′ UTR of bicoid mRNA to form particles that move in a microtubuledependent manner. Cell 79, 1221–1232 (1994). 53. Mosquera, L., Forristall, C., Zhou,Y. & King M. L. A mRNA localized to the vegetal cortex of Xenopus oocytes encodes a protein with a nanos-like zinc finger domain. Development 117, 377–386 (1993). 54. Ainger, K. et al. Transport and localization elements in myelin basic protein mRNA. J. Cell Biol. 138, 1077–1087 (1997). 55. Capri, M., Santoni, M. J., Thomas-Delaage, M. & Ait-

VOLUME 2 | APRIL 2001 | 2 5 5

REVIEWS

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

66.

67.

68.

69.

70.

71.

256

Ahmed, O. Implication of a 5′ coding sequence in targeting maternal mRNA to the Drosophila oocyte. Mech. Dev. 68, 91–100 (1997). Saunders, C. & Cohen, R. S. The role of oocyte transcription, the 5′ UTR, and translation repression and derepression in Drosophila gurken mRNA and protein localization. Mol. Cell 3, 43–45 (1999). Thio, G. L., Ray, R. P., Barcelo, G. & Schupbach, T. Localization of gurken RNA in Drosophila oogenesis requires elements in the 5′ and 3′ regions of the transcript. Dev. Biol. 15, 435–446 (2000). Macdonald, P. M., Leask, A. & Kerr, K. exl protein specifically binds BLE1, a bicoid mRNA localization element, and is required for one phase of its activity. Proc. Natl Acad. Sci. USA 92, 10787–10791 (1995). Deshler, J. O., Highett, M. I., Abramson, T. & Schnapp, B. J. A highly conserved RNA-binding protein for cytoplasmic mRNA localization in vertebrates. Curr. Biol. 8, 489–496 (1998). Havin, L. et al. RNA-binding protein conserved in both microtubule- and microfilament-based RNA localization. Genes Dev. 12, 1593–1598 (1998). Reference 59, 60 and 22 led to the surprising finding that the zip-code-binding proteins (ZBPs) involved in chicken β-actin mRNA or Xenopus Vg1 mRNA localization are homologous, although the localization mechanisms seemed to be different. This led to the idea that ZBPs might be mRNAbinding modules that can associate with different transport machineries. Lall, S., Francis-Lang, H., Flament, A., Norvell, A.,Schüpbach, T. & Ish-Horowicz, D. Squid hnRNP protein promotes apical cytoplasmic transport and localization of Drosophila pair-rule transcripts. Cell 98, 171–180 (1999). Cote, C. A. et al. A Xenopus protein related to hnRNP I has a role in cytoplasmic RNA localization. Mol. Cell 4, 431–437 (1999). Böhl, F., Kruse, C., Frank, A., Ferring, D. & Jansen, R.-P. She2p, a novel RNA-binding protein tethers ASH1 mRNA to the Myo4p myosin motor via She3p. EMBO J. 19, 5514–5524 (2000). Long, R. M., Gu, W., Lorimer, E., Singer, R. H. & Chartrand, P. She2p is a novel RNA-binding protein that recruits the Myo4p/She3p complex to ASH1 mRNA. EMBO J. 19, 6592–6601 (2000). The two previous papers provide the most detailed picture yet of protein functions in a complex involved in mRNA localization, the yeast ASH1 RNP. Roegiers, F. & Jan, Y. N. Staufen: a common component of mRNA transport in oocytes and neurons? Trends Cell Biol. 10, 220–224 (2000). St Johnston, D., Brown, N. H., Gall, J. G. & Jantsch, M. A conserved double-stranded RNA-binding domain. Proc. Natl Acad. Sci. USA 89, 10979–10983 (1992). Micklem, D. R., Adams, J., Grünert, S. & St Johnston, D. Distinct roles of two conserved Staufen domains in oskar mRNA localization and translation. EMBO J. 19, 1366–1377 (2000). Köhrmann, M. et al. Microtubule-dependent recruitment of Staufen-green fluorescent protein into large RNAcontaining granules and subsequent dendritic transport in living hippocampal neurons. Mol. Biol. Cell 10, 2945–2953 (1999). Mathisen, P. M., Johnson, J. M., Kawczak, J. A. & Tuohy, V. K. Visinin-like protein (VILIP) is a neuron-specific calcium-dependent double-stranded RNA-binding protein. J. Biol. Chem. 274, 31571–31576 (1999). Bassell, G. J., Oleynikov, Y. & Singer, R. H. The travels of mRNAs through all cells large and small. FASEB J. 13, 447–454 (1999). Knowles, R. B. et al. Translocation of RNA granules in living

neurons. J. Neurosci. 16, 7812–7820 (1996). 72. Wang, S. & Hazelrigg, T. Implications for bcd mRNA localization from spatial distribution of exu protein in Drosophila oogenesis. Nature 369, 400–403 (1994). 73. Ainger, K. et al. Transport and localization of exogenous myelin basic protein mRNA microinjected into oligodendrocytes. J. Cell Biol. 123, 431–441 (1993). 74. Bertrand, E. et al. Localization of ASH1 mRNA particles in living yeast. Mol. Cell 2, 437–445 (1998). The first description of a new approach to visualize RNA localization in living cells. A system using a ‘tagged RNA’ and a protein hybrid of an RNA-binding module and GFP is applied to follow yeast ASH1 mRNA during transport. A similar system is described in reference 32. 75. Carson, J. H., Kwon, S. & Barbarese, E. RNA trafficking in myelinating cells. Curr. Opin. Neurobiol. 8, 607–612 (1998). 76. Barbarese, E. et al. Protein translation components are colocalized in granules in oligodendrocytes. J. Cell Sci. 108, 2781–2790 (1995). 77. Bassell, G. J. et al. Sorting of β-actin mRNA and protein to neurites and growth cones in culture. J. Neurosci. 18, 251–265 (1998). 78. Ferrandon, D., Koch, I., Westhof, E. & Nüsslein-Volhard, C. RNA–RNA interaction is required for the formation of specific bicoid mRNA 3´UTR–STAUFEN ribonucleoprotein particles. EMBO J. 16, 1751–1758 (1997). 79. Marión, R. M., Fortes, P., Beloso, A., Dotti, C. & Ortín, J. A human sequence homologue of staufen is an RNAbinding protein that is associated with polysomes and localizes to the rough endoplasmic reticulum. Mol. Cell Biol. 19, 2212–2219 (1999). 80. Wickham, L., Duchaine, T., Luo, M., Nabi, I. R. & desGroseillers, L. Mammalian staufen is a doublestranded-RNA- and tubulin-binding protein which localizes to the rough endoplasmic reticulum. Mol. Cell. Biol. 19, 2220–2230 (1999). 81. Takizawa, P. A. & Vale, R. D. The myosin motor, Myo4p, binds Ash1 mRNA via the adapter protein, She3p. Proc. Natl Acad. Sci. USA 97, 5273–5278 (2000). 82. Schnorrer, F., Bohmann, K. & Nuesslein-Volhard, C. The molecular motor dynein is involved in targeting Swallow and bicoid RNA to the anterior pole of Drosophila oocytes. Nature Cell Biol. 2, 185–190 (2000). Using a two-hybrid interaction approach, the authors provide evidence for the interaction of Drosophila Swallow, a putative RNP required for bicoid mRNA localization with the motor protein dynein. 83. Wilhelm, J. E. et al. Isolation of a ribonucleoprotein complex involved in mRNA localization in Drosophila oocytes. J. Cell Biol. 148, 427–439 (2000). The first example of a biochemical approach to purify localized RNP complexes from Drosophila. 84. Mattaj, I. W. & Englmeier, L. Nucleocytoplasmic transport: the soluble phase. Annu. Rev. Biochem. 67, 265–306 (1998). 85. Nakielny, S. & Dreyfuss, G. Transport of proteins and RNAs in and out of the nucleus. Cell 99, 677–690 (1999). 86. Schnapp, B. J. RNA localization: a glimpse of the machinery. Curr. Biol. 9, R725–R727 (1999). 87. Jansen, R.-P. RNA–cytoskeletal associations. FASEB J. 13, 455–466 (1999). 88. Brendza, R. P., Serbus, L. R., Duffy, J. B. & Saxton, W. M. A function for kinesin I in the posterior transport of oskar mRNA and Staufen protein. Science 289, 2120–2122 (2000). The identification of the long-sought microtubule motor required for localization of oskar mRNA and Staufen protein to the posterior

end of Drosophila oocytes. 89. Pokrywka, N. J. & Stephenson, E. C. Microtubules mediate the localization of bicoid RNA during Drosophila oogenesis. Development 113, 55–66 (1991). 90. Haarer, B. K., Petzold, A., Lillie, S. H. & Brown, S. S. Identification of MYO4, a second class V myosin gene in yeast. J. Cell Sci. 107, 1055–1064 (1994). 91. Hurd, D. D. & Saxton, W. M. Kinesin mutations cause motor neuron disease phenotypes by disrupting fast axonal transport in Drosophila. Genetics 144, 1075–1085 (1996). 92. Schuman, E. M. Synapse specificity and long-term information storage. Neuron 18, 339–342 (1997). 93. Schuman, E. M. mRNA trafficking and local protein synthesis at the synapse. Neuron 23, 645–648 (1999). 94. Steward, O. & Ribak, C. E. Polyribosomes associated with synaptic specializations on axon initial segments: localization of protein-synthetic machinery at inhibitory synapses. J. Neurosci. 6, 3079–3085 (1986). 95. Kiebler, M. A. & DesGroseillers, L. Molecular insights into mRNA transport and local translation in the mammalian nervous system. Neuron 25, 19–28 (2000). A recent review that summarizes the current view of putative functions and mechanisms of mRNA localization in neurons. 96. Crino, P. B. & Eberwine, J. Molecular characterization of the dendritic growth cone: regulated mRNA transport and local protein synthesis. Neuron 17, 1173–1187 (1996). 97. Dhanrajan, T. M., Oleynikov, Y., Martinez, M. M., Singer, R. H. & Bassell, G. J. Dynamics of ZIP-code binding protein 1 localization into dendrites, spines and filipodia of cultured neurons. Soc. Neurosci. Abstr. 187, 3 (1999). 98. Wu, X. A. & Hecht, N. B. Testis brain RNA-binding protein (TB-RBP) colocalizes with microtubules and immunoprecipitates with mRNAs encoding myelin basic protein, α-calmodulin kinase II and protamines 1 and 2. Biol. Reprod. 62, 720–725 (2000). 99. Severt, W. L. et al. The suppression of testis-brain RNA binding protein and kinesin heavy chain disrupts mRNA sorting in dendrites. J. Cell Sci. 112, 3691–3702 (1999). 100. Knowles, R. B. & Kosick, K. S. Neurotrophin-3 signals redistribute RNA in neurons. Proc. Natl Acad. Sci. USA 94, 14804–14808 (1997). 101. Muslimov, I. A., Banker, G., Brosius, J. & Tiedge, H. Activity-dependent regulation of dendritic BC1 RNA in hippocampal neurons in culture. J. Cell Biol. 141, 1601–1611 (1998). 102. Steward, O., Wallace, C. S., Lyford, G. L. & Worley, P. F. Synaptic activation causes the mRNA for the IEG Arc to localize selectively near activated postsynaptic sites on dendrites. Neuron 21, 741–751 (1998). Evidence for a regulated expression and localization of mRNAs in neurons. ARC mRNA is localized to postsynaptic sites upon stimulation of the corresponding dendritic area. 103. Adinolfi, S. et al. Novel RNA-binding motif: the KH module. Bipolymers (Peptide Sci.) 51, 153–164 (1999). 104. Gautreau, D., Cote, C. A. & Mowry, K. L. Two copies of a subelement from the Vg1 RNA localization sequence are sufficient to direct vegetal localization in Xenopus oocytes. Development 124, 5013–5020 (1997).

Acknowledgements I apologize to all colleagues whose work has not been properly discussed owing to limited space. I am grateful to J. Carson, R. Cohen, A. Ephrussi, J. Hesketh, M. Kiebler, S. Kindler, R. Long, B. Schnapp and R. Singer for providing figures and movies. I would like to thank especially A. Ephrussi, M. Kiebler, A. Jaedicke and three further colleagues for encouraging comments and suggestions, and members of my lab for their contribution to numerous discussions on the subject.

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

TIE RECEPTORS: NEW MODULATORS OF ANGIOGENIC AND LYMPHANGIOGENIC RESPONSES Nina Jones*, Kristiina Iljin‡, Daniel J. Dumont* and Kari Alitalo‡ Angiogenesis is required for normal embryonic vascular development and aberrant angiogenesis contributes to several diseases, including cancer, diabetes and tissue ischaemia. What are the molecular mechanisms that regulate this important process? The Tie family of receptors and their ligands, the angiopoietins, are beginning to provide insight into how vessels make decisions such as whether to grow or regress — processes that are important not only during development but throughout an organism’s life.

MESODERM

Third germ layer in the embryo, formed during the process of gastrulation.

*Department of Medical Biophysics, University of Toronto, Sunnybrook and Women’s College Health Sciences Centre, 2075 Bayview Avenue, S-227, Toronto, Ontario, Canada M4N 3M5. ‡ Molecular/Cancer Biology Laboratory, Ludwig Institute for Cancer Research, Biomedicum Institute, University of Helsinki, P.O. Box 21 (Haartmaninkatu 3), SF-00014 Helsinki, Finland. Correspondence to K.A. e-mail: Kari.Alitalo@Helsinki.FI

During embryogenesis, the cardiovascular system is the first organ system to develop and it is established through several complex processes that involve coordinated interactions between distinct cell lineages. Embryonic endothelial cells provide a framework around which the heart — as well as the arteries, veins and capillaries — are subsequently organized to supply essential oxygen and nutrients to developing tissues and organs. Expansion of the embryonic vasculature through the regulated proliferation, migration and differentiation of endothelial cells is divided into two discrete processes known as vasculogenesis and angiogenesis. During vasculogenesis, MESODERM-derived endothelial cell precursors known as angioblasts differentiate and assemble into discrete blood vessels in situ to form a primitive tubular network that consists of relatively uniformly sized endothelial channels. This PRIMARY CAPILLARY PLEXUS is subsequently remodelled through the process of angiogenesis whereby new capillaries arise from pre-existing vessels to give rise to a more complex vascular network with a hierarchy of both large and small vessels. Stabilization of the vasculature occurs when PERIENDOTHELIAL SUPPORT CELLS, such as smooth muscle cells and PERICYTES, are progressively recruited to the vessel wall, and the surrounding extracellular matrix (ECM) is reconstituted. In the absence of support-cell coverage, unprotected vessels undergo regression. Blood-vessel growth and regression are also important

in the adult for the continuous remodelling of the female reproductive system and for tissue repair, and perturbations in this delicate balance can contribute to pathological processes such as tumour growth (BOX 1). Coincident with maturation of the blood vascular system, the lymph vascular system develops in a parallel fashion from embryonic veins through a process referred to as LYMPHANGIOGENESIS. The lymph vascular system transports tissue fluid, extravasated plasma proteins and cells back into the blood circulation (FIG. 1)1. Excess fluid and macromolecules from the STROMA compartment initially drain into lymphatic capillaries, which differ from blood capillaries in that they consist of an extremely permeable, thin endothelial cell layer that is often devoid of surrounding support cells and basement membrane. From these small lymphatic vessels, the fluid is transferred to progressively larger collecting lymphatic vessels, consisting of endothelial, muscular and adventitial layers, and ultimately into the venous circulation through the thoracic duct. Movement of lymph in the lymphatic system is dependent on the intrinsic contractility of the smooth muscle layer of the larger lymphatic vessels, and lumenal valves prevent back-flow. In addition to the lymphatic vessels, the lymphatic system also contains several lymphoid organs (spleen, lymph nodes, tonsils and thymus) that are essential in immune responses. Throughout the past decade, studies focused on understanding the molecular

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 5 7

REVIEWS

Thoracic duct

Blood vessel

Lymphatic vessel

Endothelial cell

Pericyte

Anchoring filament

Erythrocyte

Endothelial cell

Leucocyte Basal lamina

(Basal lamina) Lymph node

Artery

Vein Bileaflet valve

Lymphatic vessel Capillaries Leucocyte

Cell

PRIMARY CAPILLARY PLEXUS

Simplified, honeycomb-like network of rudimentary blood vessels that is especially evident in the yolk sac vasculature. PERIENDOTHELIAL SUPPORT CELLS

Mesenchymal cells that surround the endothelial component of mature blood vessels. They include pericytes and vascular smooth muscle cells. PERICYTES

Support cells of capillaries (referred to as smooth muscle cells in larger vessels). LYMPHANGIOGENESIS

Development and growth of lymphatic vessels. STROMA

Connective tissue made up of cells, such as fibroblasts, and matrix, such as collagen.

258

Figure 1 | Blood and lymph vascular systems. a | The cardiovascular system consists of the heart as well as blood vessels (arteries, veins and capillaries) and lymphatic vessels. Arteries deliver oxygenated blood (red) to the capillaries where bidirectional exchange occurs between blood and tissues. Veins collect deoxygenated blood (blue) from the microvascular bed and carry it back to the heart. Lymphatic vessels (yellow) collect extravasated tissue fluid, filter it through lymph nodes and return it to the circulation through the thoracic and lymphatic ducts and the lymphaticovenous anastomoses (not shown). The lymphatic vascular system is not continuous like the blood vascular system. b | Shows extravasation of tissue fluid from blood to the lymphatic vessels. Within the bloodstream, fluid flows rapidly as a plasma suspension of erythrocytes whereas outside the bloodstream, it flows slowly as a tissue fluid–lymph suspension of immune cells through lymphatics. Blood vessels have a continuous basal lamina (black) with tight interendothelial junctions and they are supported by pericytes and smooth muscle cells. By contrast, lymphatic endothelial cells have a discontinuous basal lamina and have gaps between the lymphatic endothelial cells that can open to the adjacent connective tissue. In oedematous tissue, the endothelial cells are pulled by the anchoring filaments. Bileaflet valves prevent backflow of lymphatic fluid. Fluid flow is represented as large arrows.

mechanisms that govern vessel growth have revealed that endothelial-cell-specific polypeptide growth factors and receptors are fundamental for the development of both the blood and the lymph vascular systems. Endothelial cell growth factors and receptors

Assembly of a functional vascular system requires coordinated signalling between various growth factors and receptors. Vascular endothelial growth factor (VEGF, also known as vascular permeability factor) was initially identified in the late 1970s on the basis of its ability to induce transient vascular leakage and by its function as the first selective angiogenic factor for endothelial cells. Despite intense investigation into the function of VEGF in vascular development, it was not until the early 1990s that the endothelial cell-specific receptors that mediate the effects of VEGF were cloned. So far, five VEGF-type ligands (VEGFs A–E) have been identified that bind and activate three subsets of VEGF

receptors (VEGFRs): VEGFR-1 (Flt-1), VEGFR-2 (KDR/Flk-1) and VEGFR-3 (Flt-4) (FIG. 2). Neuropilin1 has also been identified as an isoform-specific receptor for VEGF, although this receptor is unique among VEGFRs in that it does not seem to have tyrosine kinase activity and it is expressed abundantly by both endothelial and certain non-endothelial cells, including neurons and tumour cells2. Recently, VEGFs have also been implicated in regulating growth control of the lymphatic endothelium: VEGF-C and VEGF-D can stimulate lymphangiogenesis3,4 and lymphatic metastasis5–7, and their receptor, VEGFR-3, has been linked to human hereditary LYMPHOEDEMA8,9. Concurrent with the identification of the VEGFRs, two additional receptors known as the Tie receptors were also isolated from endothelial cells (BOX 2), although these receptors were classed as a different subfamily on the basis of differences in their domain structure (FIG. 2). Gene targeting experiments in mice, aimed at underwww.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

Box 1 | Angiogenesis in the adult Endothelial cell proliferation and vascular endothelial growth factor (VEGF)-induced angiogenesis are required soon after birth for the growth of several organs and bones82,83 and in adults for hair growth (M. Detmar, personal communication). In the adult female reproductive cycle, development and endocrine function of the ovarian corpus luteum are dependent on VEGF and Ang2-regulated growth of capillary vessels11,84. Angiogenesis is also important during placental and endometrial growth, and is crucial for the pathogenesis of endometriosis, which results from endometrial tissue implanting in the peritoneal cavity. Neovascularization also has a central function in tissue repair to promote wound healing following the release of cytokines at the wound site. Neovascularization is needed to promote tissue remodelling through angiogenesis, and the formation of arteries and collateral blood vessels when VEGF is upregulated during ischaemia. In addition to its function in normal physiological processes, angiogenesis has also been associated with several diseases including tumour growth and metastasis and, in several tumours, increased vascularization has been directly correlated with poor prognosis. Although hypoxia is the primary inducer of VEGF secretion, other growth factors produced by tumour cells, as well as mutations in tumour suppressor genes, might contribute. Neovascularization of the eye, resulting in blindness, can occur in diabetes owing to upregulation of VEGF in retinal cells85. This can also occur in premature infants that have been exposed to high oxygen levels. In rheumatoid arthritis, new capillary blood vessels associated with the inflammatory condition invade the joint and destroy cartilage and in psoriasis — an inflammatory skin disease — hypervascular lesions are characterized by elongation and widening of dermal vessels.

standing the physiological function of these two subfamilies of receptors, revealed that they are indispensable for normal embryonic blood-vessel development (see below). Although these studies provided great insight into the genetic mechanisms that control the development of the vascular system, knowledge of the molecular pathways controlling these effects was still poorly understood. In recent years, a second family of growth factors that are also largely specific for the vascular endothelium has begun to emerge and these unique ligands are

Ligands ? Receptors Tie1

Ang1 Ang2 Ang3/4 Tie2

VEGF PIGF VEGF-B

VEGF VEGF-C VEGF-D VEGF-C VEGF-E VEGF-D

sVEGFR-1 VEGFR-1 VEGFR-2 VEGFR-3

LYMPHOEDEMA

Disease caused by obstruction or defects of the lymphatic vessels, resulting in accumulation of interstitial tissue fluid.

Downstream events

Survival, sprouting, Migration, growth, stabilization, permeability... differentiation, survival, permeability...

ENDOCARDIUM

Ig-like

Fibronectin III

Endothelial lining of the cardiac lumen.

EGF-like

Intracellular

known as the angiopoietins (FIG. 3). The angiopoietins isolated so far seem to bind exclusively to the Tie2 (also known as Tek) receptor tyrosine kinase (RTK), and the ligand for the closely related Tie1 receptor remains to be identified. Interestingly, these ligands seem to have opposing actions in endothelial cells as angiopoietin-1 (Ang1) and angiopoietin-4 (Ang4) function as agonistic or activating ligands for Tie2, whereas angiopoietin-2 (Ang2) and angiopoietin-3 (Ang3) behave as contextdependent competitive antagonists10–12. Mouse Ang3 and human ANG4 are actually interspecies orthologues that are more divergent than the mouse and human counterparts of Ang1 and Ang2, and this structural divergence is associated with marked differences in their function and expression12. The distinct but overlapping expression patterns of the angiopoietins might indicate that each has a unique function that is required for the spatial and temporal regulation of vascular development. Tie receptors in normal vessel development

The Tie receptors seem to be required for the angiogenic remodelling and vessel stabilization that occur subsequent to the initial vasculogenic actions of VEGFR-1 and VEGFR-2 (REFS 13,14). Embryos lacking Tie2 signalling pathways die between embryonic day 9.5 (E9.5) and E12.5 as a consequence of insufficient expansion and maintenance of the primary capillary plexus11,15–18. One of the most prominent defects in these mutant embryos is the incomplete development of the heart region, characterized by detachment of the ENDOCARDIUM from the underlying myocardial wall and the absence of myocardial projections known as trabeculae. Widespread vascular haemorrhage can also be observed and the remaining vasculature appears simplified with few support cells in regions of deficient

Figure 2 | Structure of endothelial-cell receptor tyrosine kinases and growth factors involved in vasculogenesis, angiogenesis and lymphangiogenesis. These structurally divergent receptors can be subgrouped into the Tie and the vascular endothelial growth factor (VEGF) receptor families. The specificity of ligand (marked in blue) binding to the receptors is indicated by arrows. The Tie receptor family has two members, Tie1 and Tie2 (also called Tek). The extracellular region of the Tie receptors consists of one complete and one incomplete immunoglobulin (Ig)-like domain that are separated by three tandem epidermal growth factor (EGF)-like cysteine repeats and are followed by three fibronectin type III homology domains. The VEGF receptor family consists of three transmembrane receptors, VEGFR-1, VEGFR-2 and VEGFR-3. A soluble form of VEGFR-1 (sVEGFR-1) has also been characterized. The extracellular regions of VEGFR-1 and VEGFR-2 contain seven immunoglobulin domains that are stabilized by disulphide links (SS) between paired cysteine residues and in VEGFR-3, the fifth domain is proteolytically processed into two disulphide-linked polypeptides. In the intracellular region of both VEGF and Tie receptors (red), the tyrosine kinase domains are interrupted by a small stretch of amino acids commonly referred to as a kinase insert (dark green). These receptors can participate in various biological processes, some of which are indicated in the figure. (PIGF, placenta growth factor.)

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 5 9

REVIEWS

Box 2 | Tie receptors in haematopoiesis In addition to their expression in endothelial cells, Tie1 and Tie2 have also been detected in certain haematopoietic cell types43,86, indicating a possible additional function for these receptors in the development of the haematopoietic system. Despite this observation, embryos lacking Tie1 have no major defects in embryonic haematopoiesis and haematopoietic cells lacking Tie1 contribute normally to haematopoietic cell populations in chimeric animals19,21. Consistent with the idea that Tie1 might not be required for haematopoiesis, transplantation experiments have shown that all haematopoietic lineages can be established in the absence of Tie1 (REF. 87). In contrast to the findings with Tie1, embryos lacking either Tie2 alone or both Tie1 and Tie2 are pale and anaemic20. Moreover, cell-culture experiments have shown that Tie2-null embryos cannot give rise to definitive haematopoietic cells88. These observations have indicated that Tie2 might be required during early haematopoiesis. Interestingly, however, haematopoietic cells derived from Tie1/Tie2 double-mutant embryos are not selected against in chimeric animals, indicating that this block in haematopoietic cell differentiation might not be cell autonomous20. Instead, these findings imply that Tie2 probably establishes the proper vascular microenvironment in which haematopoiesis occurs. Endothelial cells are important in the development of haematopoietic stem cells as Ang1 can induce adhesion of Tie2-expressing haematopoietic cells to fibronectin expressed on the surface of endothelial cells and thereby enhance proliferation of haematopoietic progenitor cells88. Conversely, Ang1-expressing bone marrow haematopoietic stem cells have recently been shown to recruit endothelial cells through Tie2-mediated migration, which indicates a possible novel mechanism of angiogenesis89. Despite the emerging interplay between haematopoiesis and the growth of blood vessels, rescue from the vascular defects observed in mutant embryos is needed to fully characterize the haematopoietic importance of both Tie1 and Tie2.

ANG1 ANG1-1.3 kB ANG1-0.9 kB ANG1-0.7 kB

ANG1*

ANG2

100%

69%

63%

ANG2B

70%

65%

ANG3

44%

57%

ANG4

42%

55%

ANG2443

ANG1 coiled-coil

ANG2 coiled-coil

ANG1 fibrinogen-like

ANG2 fibrinogen-like

Figure 3 | The angiopoietins. The opposing actions of these structurally similar ligands have been ascribed to the receptor-binding fibrinogen-like domains and the coiled-coil domains mediate distinct multimerization patterns of the ligands in vitro12,101. Human angiopoietin-1 (ANG1) is the prototypic angiopoietin and has three different splice isoforms: ANG1-1.3 kB, ANG1-0.9 kB and ANG1-0.7kB102. To facilitate purification of ANG1, a chimeric angiopoietin denoted ANG1* was engineered in which the first 73 amino acids of ANG2 (red) were fused to the region of ANG1 beginning at amino-acid residue 77 (yellow) with the inclusion of a cysteine-to-serine mutation at amino-acid residue 265 (asterisk)57. ANG1* is thought to behave in a similar fashion to native ANG1. ANG2 also has a splice variant, which results from alternative splicing of 52 amino acids from the coiled-coil domain (ANG2443)103. The structures of an avian isoform of ANG2, which is characterized by a partially truncated amino-terminal coiled-coil domain as a result of alternative splicing of a 5â&#x20AC;˛ intron (aANG2B)104, mouse Ang3 (mAng3) and its human orthologue ANG4 (REF. 12) are also shown. The numbers within each box represent the per cent identity to ANG1. (Figure concept courtesy of N. Ward).

260

vessel branching as well as a decreased number of endothelial cells. Embryos that lack Tie1 also show compromised vascular maintenance owing to impaired endothelial cell integrity and they die later in development â&#x20AC;&#x201D; between E13.5 and birth16,19. Cardiac morphogenesis proceeds normally in Tie1-deficient embryos, indicating that Tie1 might be required later than Tie2 by particular endothelial-cell types. A stringent requirement for Tie2 in heart development has been elucidated from mosaic analysis of Tie2-deficient embryos mixed with normal embryos in which cells lacking Tie2 were excluded from the endocardium of E10.5 chimeras20. Similar chimeric studies have shown a requirement for both Tie1 and Tie2 throughout embryonic vessel development and in the quiescent adult vasculature20,21. Close ultrastructural analysis of the vessels in embryos lacking Tie2 or Ang1 has indicated that endothelial cells fail to associate properly with underlying support cells17,18. By contrast, mice that overexpress Ang1 in the skin have larger and more highly branched vessels that are resistant to leakage induced by inflammatory stimuli, even in the presence of excess permeability factors such as VEGF22,23. These findings have led to the hypothesis that Tie2 signalling might facilitate recruitment of, and tight association with, adjacent periendothelial cells by stimulating the release or activation of chemoattractant growth factors from endothelial cells. Interestingly, however, mosaic analysis of double-mutant Tie1- and Tie2-deficient embryos mixed with normal embryos has shown that the presence of double-mutant endothelial cells does not affect the morphology of early embryonic blood vessels, although a continuous selection against these cells is observed in later-stage embryos and adult chimeras20. The inability of neighbouring wild-type cells to rescue the deficient cells indicates a possible cell-autonomous defect in these double-mutant cells, and it raises the possibility that the reduced recruitment of periendothelial cells in embryos lacking Tie2 sigwww.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

ÂŠ 2001 Macmillan Magazines Ltd

REVIEWS a

SRC

Cytoplasmic tyrosine kinase first identified as a transforming oncogene in an avian retrovirus. This kinase is the prototypic kinase from which srchomology regions were first described. ANOIKIS

Induction of programmed cell death by detachment of cells from the extracellular matrix.

Figure 4 | Venous malformation associated with a TIE2 mutation. a | Hematoxylin–eosin stained section of a venous malformation in the skin of an 80-year-old female with a familial inheritance of a heterozygous activating (R849W) mutation in TIE2 (REF. 35). These lesions have enlarged vascular channels (arrows) lined with a thin layer of flattened endothelial cells but with a relative lack of smooth muscle support cells (stained red using antibodies that recognize smooth muscle α-actin). For comparison, b | shows a section of uninvolved skin and subcutaneous tissue. This has a regular and uniform layer of smooth muscle around the veins (arrows) and artery (thick arrow). After the initial report on the association of venous malformations with TIE2 mutations35, similar mutations have been published by other groups36. Interestingly, the histology of the lesions has some resemblance with that described for the transgenic mice expressing Ang1 under a basal epidermal keratin promoter23. (Figure courtesy of Laurence Boon and Miikka Vikkula at the Laboratory of Human Molecular Genetics, Christian de Duve Institute of Cellular Pathology and Catholic University of Louvain, Brussels, Belgium).

nalling pathways might be secondary to the impaired survival of the endothelium. In support of this premise, conditional loss of Tek expression in vivo is concurrent with the onset of endothelial cell apoptosis and vascular haemorrhage24. Taken together, it seems likely that Tie2 signalling provides more than a secreted attraction signal for perivascular cells, and persistent expression and

phosphorylation of Tie2 in quiescent adult endothelium is consistent with a role for Tie2 in transducing a continuing survival stimulus to endothelial cells25. Angiopoietins and vascular stability

The discovery of a unique family of natural agonists and competitive antagonists to coordinate the stabilizing

Box 3 | Endothelial cell adhesion molecules Ang2

Endothelial cells use cell adhesion molecules such as cadherins, selectins and integrins to attach themselves to one another and to the inner lining of blood vessels. Tie2 VEGFR-2 Endothelial cell membrane Cadherins are localized to structures known as adherens p110 p110 junctions, and vascular endothelial (VE)-cadherin p85 p85 P P PI(3)K PI(3)K mediates the calcium-dependent interactions between neighbouring endothelial cells. These junctions are thought to provide a mechanical barrier to interendothelial leakage, whereas tight and gap junctions control permeability to plasma proteins. The short cytoplasmic tail Bcl-xL Bad Akt of VE-cadherin interacts with α-, β- and γ-catenin Ca2+ (plakoglobin), and this complex is co-localized with β-catenin γ-catenin vinculin and α-actinin to link the VE-cadherin–catenin α-catenin n VE-cadherin complex to the actin cytoskeleton. VE-cadherin can also in culi n i t Vin ac signal with VEGFR-2 to mediate phosphatidylinositol-3p110 α p85 OH kinase (PI(3)K)/Akt-dependent endothelial cell Actin FAK Paxillin survival90. In addition to cell–cell contacts, endothelial cells Cytochrome c CAS adhere to the extracellular matrix (ECM) or basal lamina αvβ3 through interactions with cell-surface heterodimeric integrins. The ECM is a dense latticework of collagen and MMP2 ECM elastin found within a complex mixture of proteoglycans and glycoproteins, whereas the basal lamina is a meshwork of type IV collagen and many other glycoproteins. The engagement of integrins with the ECM or basal lamina causes the activation of focal adhesion kinase (FAK). Phosphorylated FAK recruits SRC which, in turn, phosphorylates FAK on additional sites, allowing the recruitment of signalling molecules such as Sos, PI(3)K, CAS and paxillin to focal adhesions. The assembly of this signalling complex contributes to the role of FAK in cell survival and cell migration. In addition to adhesion-dependent survival, endothelial cells can also respond to receptor-tyrosine-kinase-mediated survival signals transmitted from VEGFR-2 and Tie2 to Akt, and both VEGF and Ang1 allow endothelial cells to escape ANOIKIS upon detachment from the ECM. Lumen of vessel

VEGF

Ang1

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 6 1

REVIEWS

Box 4 | Hypoxia and the angiogenic response Angiogenesis is induced as a result of oxygen deprivation or hypoxia resulting from an imbalance between oxygen supply and the metabolic demands of tissues. Hypoxia regulates the pathophysiology of many diseases such as ischaemic cardiovascular disorders and tumour growth through induction of vascular endothelial growth factor (VEGF). Tumour hypoxia and tumour-specific genetic alterations involving oncogenes and tumour suppressor genes increase VEGF levels in neoplastic cells through stimulation of transcription by hypoxia-inducible factors 1 and 2 (HIF-1/HIF-2)91 and, in some cases, through increased messenger RNA stability92. Under hypoxia, HIF-1α is stabilized and accumulated in the nucleus where it heterodimerizes with HIF-1β and interacts with several co-activators of transcription, resulting in transactivation of numerous target genes that are implicated in angiogenesis. HIF-2α can also form heterodimers with HIF-1β to bind to the regulatory sequences of certain hypoxiaresponsive genes47,91. HIF-1α-, HIF-1β- and HIF-2α-deficient mice die in utero as a result of vascular defects93–95, and HIF-1α-deficient tumours show a reduced level of vascularization with slower growth94,96,97. Constitutive expression of HIF-1α is observed following loss of the von Hippel–Lindau tumour suppressor gene98,99, which is associated with haemangioblastomas and highly vascularized renal cell carcinomas in patients. Interestingly, HIF-1α has also been found in a molecular complex with the p53 tumour suppressor protein and the ubiquitin ligase Mdm2, and loss of p53 decreases Mdm2-mediated degradation of HIF-1α96. This observed HIF-1α stabilization might, in turn, have a role in hypoxia-induced cell death, selecting for loss of p53 in tumour cells100. Other oncogenes and tumour suppressors that increase the phosphatidylinositol-3-OH kinase/Akt or mitogen-activated protein kinase pathways can also enhance HIF-1α stability and transactivation. By contrast, too much oxygen (hyperoxia) leads to excessive trimming of the vascular tree and this process has been implicated in retinopathy in premature infants, in whom vessels regress owing to endothelial cell apoptosis as a consequence of reduced VEGF levels26.

FIBRIN MATRICES

Cell-culture substratum used in in vitro assays of angiogenesis to examine endothelial cell differentiation. Based on the premise that endothelial cells invade blood clots during wound repair and that fibrin is the major component of a blood clot. OEDEMA

Accumulation of tissue fluid leading to swelling. STATS

Family of cytoplasmic transcription factors (signal transducers and activators of transcription) that dimerize upon phosphorylation and translocate to the nucleus to activate transcription of target genes. GLIOBLASTOMA

Malignant brain tumour, predominantly located in the cerebral hemispheres. KAPOSI´S SARCOMA

Angiogenic tumour composed of endothelial and spindle cells (elongated fibroblast-like shaped cells that usually express endothelial markers).

262

functions of the Tie2 receptor implies that the mechanisms surrounding vessel remodelling are precisely regulated in vivo. Angiogenic remodelling of the mature vasculature requires a progressive disengagement of endothelial cells from surrounding support cells and this destabilization can lead to vessel sprouting or vessel regression. During vessel sprouting, this detachment renders endothelial cells accessible to further angiogenic stimuli, such as VEGF, allowing the cells to overcome growth inhibition and invade the vessel wall through the release of proteases such as those from the matrix metalloproteinase (MMP) family (BOX 3). Migration through the ECM and assembly of endothelial cells into tubes is facilitated by contacts that are provided by cell adhesion molecules expressed on the surface of endothelial cells. Alternatively, during vessel regression, angiogenic inducers are absent and detachment of endothelial cells from the underlying matrix precedes apoptosis26,27. Endothelial cell apoptosis associated with vascular regression is controlled by numerous molecular pathways and some of these have been shown to be stimulated by the Tie receptors (see below). The distinct expression pattern of Ang2 at sites of active vascular remodelling11 and in highly vascularized tumours28–31 has implicated Ang2 in blockade of the constitutively stabilizing function of Ang1 to facilitate angiogenesis. Ang2 is selectively upregulated in tumour vessels before the expression of VEGF in adjacent tumour cells28–30 and Ang2 can synergize with VEGF to enhance neovascularization32, indicating that Ang2 might be an agonist in particular microenvironments. In support of this, Ang2 can promote tubule formation of endothelial cells when

they are cultured on FIBRIN MATRICES and Tie2 isolated from these Ang2-stimulated endothelial cells is in fact activated (as determined by tyrosine phosphorylation)33. Moreover, high concentrations of Ang2 can induce endothelial cell survival through a similar mechanism as that defined for Ang134. Surprisingly, recent data also indicate that Ang-2 might have a role in lymphatic development. In Ang-2 knockout mice, the peritoneal cavity is filled with chylous ascites, the mice develop severe peripheral OEDEMA, and the hindlimb and abdominal lymphatics appear nonfunctional and perhaps disconnected (N. Gale, C. Suri, M. Witte and G. Yancopoulos, personal communication). These new results should be considered in designing gene therapy for various diseases in which endothelial cells and vessel permeability are involved. Pathological angiogenesis

Coordinated expression of Tie1, Tie2 and the angiopoietins maintains vascular plasticity, and perturbations in this regulation can contribute to abnormal vascular growth. Gain-of-function mutations at the TIE2 locus have been identified in some families with inherited venous malformations35,36 (FIG. 4). Although the precise consequences of these mutations in signal-transduction pathways that are mediated by Tie2 have not been defined, it has recently been shown that one of these mutant forms of Tie2 can uniquely activate STAT1 (REF. 37). Furthermore, elevated expression of both Tie1 and Tie2 has been observed in the endothelium of the neovasculature in numerous solid tumours29,31,38–41 as well as in healing skin wounds25,42. Certain human leukaemia cell lines also express both Tie1 and Tie2 as well as Ang1 (REF. 43), and expression of Ang2 is markedly upregulated in highly angiogenic tumours including GLIOBLASTOMA and KAPOSI’S SARCOMA29,31. Ang2 expression in endothelial cells can be selectively induced by oxygen starvation (hypoxia, BOX 4) and VEGF44,45, which correlates with its expression in regions undergoing intense neovascularization. Similarly, Tie1 protein levels are also increased in response to hypoxia and after VEGF treatment46, and Tie2 is a target of hypoxic activation by the endothelial hypoxia-inducible factor-2 (HIF-2, also known as EPAS1/HRF)47,48. Specific expression of the Tie1 and Tie2 genes in endothelial cells is dependent on genomic regulatory regions including ETS-transcription-factorbinding sites49,50 (FIG. 5). The expression of these growth factors and receptors in tumour vasculature indicates that, in addition to the VEGFR pathways, Tie1 and Tie2 pathways might also be involved in tumour angiogenesis. In support of this idea, interruption of Tie2 signalling using soluble, dominant-negative receptors can inhibit angiogenic growth in tumour-bearing mice51–53. Inhibition of Tie2 and possibly Tie1 signalling can therefore provide potential treatments for tumours that do not respond well to anti-VEGF therapy54. Signalling through Tie receptors

The inability of the Tie receptors to functionally compensate for one another in vivo has shown that these receptors mediate overlapping but independent signalwww.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

b Tie1-EGFP

AflII

ApaI

Tie1

Ets

PEA

Oct

PEA AP2

0.1 kb

ATG NFS

bZIP CP2γ

1 kb

0.1 kb

NcoI XbaI

HindIII

Tie2

ATG

Figure 5 | Regulatory elements of Tie1 and Tie2 genes. a | Regulatory regions of Tie1 and Tie2 genes and elements associated with endothelial cell-specific expression are shown. The AflII–ApaI mouse Tie1 promoter fragment contains all the DNA elements necessary to direct expression of heterologous genes in endothelial cells50,105. In Tie2, the endothelial regulatory elements are distributed in both the proximal promoter and in the large first intron (in the NcoI–XbaI fragment)49,106. The consensus binding sites for factors that are important for the promoter/enhancer activities are indicated in the figure. The regulatory regions share several DNA motifs including putative binding sites for the Ets transcription factors (Ets and PEA) as well as an octamer (Oct) factor binding site. In addition, a putative transcription-factor-binding site for AP2 is found in the Tie1 promoter and sites for NFS (nuclear factor S), bZIP (basic leucine zipper protein) and CP2γ (CCAAT-binding protein) are found in the Tie2 intronic enhancer. White boxes indicate the non-coding portions of the first exons and black boxes indicate the coding sequences; the arrows with ATG indicate the sites of the translational initiation codons. b | A photomicrograph of Tie1 promoterdriven expression of enhanced green fluorescent protein in blood vessels of the limb of a transgenic mouse embryo at embryonic day 18 (K.I. and K.A., unpublished observations).

transduction pathways in the maintenance and survival of the vasculature. Identification of the individual cascades of molecular protein–protein interactions initiated by these RTKs is therefore of paramount importance in understanding how blood vessels develop. Recent isolation of the angiopoietins has triggered an explosion of intense investigation into the molecular mechanisms that underlie Tie2-mediated signalling. It was initially anticipated that Tie2 might provide a mitogenic signal to endothelial cells as the activated receptor associates with the Grb2 adaptor protein55,56. However, subsequent studies have shown that Ang1 does not activate mitogen-activated protein kinase (MAPK) nor does it stimulate cellular proliferation10,57–59. Instead, the angiopoietins seem to regulate vascular expansion and survival.

ETS

Group of winged helix–loop–helix transcription factors that contain an ETS DNA-binding domain. CHORIOALLANTOIC MEMBRANE

An extremely vascular envelope created by fusion of the allantoic membrane with the chorion. Functions in the exchange of wastes and metabolites in the embryo.

Endothelial cell survival. Phosphatidylinositol-3-OH kinase (PI(3)K) has been identified as a critical mediator of extracellular survival signals through its regulation of the serine–threonine kinase Akt (also known as protein kinase B). Activation of Akt in turn stimulates the phosphorylation and subsequent inhibition of proapoptotic proteins such as Bad and caspase-9 and, in endothelial cells, Akt influences the production of nitric oxide through phosphorylation of nitric oxide synthase to inactivate caspases. Ang1 and the modified Ang1* chimeric ligand (FIG. 3) can facilitate PI(3)K-dependent endothelial cell survival through stimulation of Akt56,60–64 (FIG. 6). Furthermore, expression of either Akt or the catalytic subunit of PI(3)K can induce angiogenesis in the CHORIOALLANTOIC MEMBRANE of the chick

embryo65. Ang1*-mediated protection from apoptosis has been associated with an upregulation of the inhibitor of apoptosis protein (IAP) survivin66, although the role of survivin in the control of apoptosis remains to be fully determined as the nematode survivin-like protein BIR-1 seems to function during cytokinesis67. Activation of PI(3)K and Akt requires a multisubstrate docking site on Tie2 that is conserved in Tie1 (REFS 56,60), and Tie1 might also signal through this pathway to promote cell survival (C. D. Kontos and K. G. Peters, personal communication). The role of these receptors in promoting endothelial cell survival complements observations in mice lacking Tie2 and/or Tie1, in which mutant endothelial cells are progressively lost15,20,21,24. Moreover, it raises the question as to whether the defects in vessel morphology that are noted in the absence of Ang1 signalling might actually be due to a loss of contact between the endothelial and smooth muscle cell layers as a direct consequence of endothelial cell apoptosis17,18,28. VEGF also functions as a strong survival factor for endothelial cells through VEGFR-2mediated activation of PI(3)K and Akt68, and addition of VEGF can augment the anti-apoptotic effect of Ang1 in endothelial cells59. So endothelial cells seem to have evolved to respond to two distinct survival cues, which might reflect the expression levels of Ang1 and VEGF in particular vascular beds. Endothelial cell migration. During sprouting angiogenesis, endothelial cells must alter their intracellular architecture to facilitate migration into the surrounding

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 6 3

REVIEWS Angiopoietin-1 VE-PTP

Endothelial cell membrane Ras

Tie 2

FAK Grb7

Grb2 P Ras

MAPK cascade

Cell proliferation

RasGAP

Pak

P p110

Dok-R

Akt

p85

P P

Caspase-9

P Nck

Bad

eNOS

Survivin

Shp-2 P

Grb14 Cell survival Cell migration

Figure 6 | Tie2 signal transduction, showing the known binding partners for Tie2. The p85 regulatory subunit of phosphatidylinositol-3-OH kinase (PI(3)K) associates with the phosphorylated Tie2 receptor, presumably through tyrosine residue 1100, resulting in stimulation of PI(3)K. Subsequent activation of the serine/threonine kinase Akt correlates with an increase in cell survival, although the exact mechanism by which this occurs remains to be determined. Alternatively, PI(3)K activity is coincident with focal adhesion kinase (FAK) activation and cell migration. Grb7 and the protein tyrosine phosphatase Shp2 might also use this pathway to potentiate cell migration and they seem to be recruited to the receptor through tyrosine residues 1100 and 1111, respectively55,56. The effects of Dok-R (docking protein) recruitment to Tie2 and its downregulation of mitogen-activated protein kinase (MAPK) and cell proliferation are inferred from our studies on the epidermal growth factor receptor107, and we have shown that Dok-R can enhance p21-activated (Pak)-dependent cell migration in response to angiopoietin-1 through its association with Nck. The significance of Grb2 and Grb14 binding to Tie2, as well as binding of the Ras GTPase-activating protein RasGAP to Dok-R, are now under investigation. The binding sites for Grb14 and Dok-R are from our unpublished results (N.J. and D.J.D.). Solid arrows indicate pathways that are known to exist downstream of Tie2 and dashed arrows indicate links that have been demonstrated using other receptor signalling models. The vascular endothelial protein tyrosine phosphatase (VE-PTP) is illustrated at the cell membrane and HCPTPA (human cellular protein tyrosine phosphatase A) has been omitted from this schematic. (NOS, nitric oxide synthase.)

NCK

SH2/SH3-domain-containing adaptor molecule that has been implicated in cell migration through transduction of signals to the actin cytoskeleton. SHP2

SH2-domain-containing cytoplasmic tyrosine phosphatase.

264

basement membrane after the secretion of matrixdegrading proteinases. Numerous studies have shown that activation of Tie2 by Ang1 results in the stimulation of endothelial cell migration. This role for Ang1 is consistent with the findings that Tie2- and Ang1-null mice have an angiogenic or migratory defect that is manifested as a lack of vessel sprouting and remodelling throughout the embryo15–18. Ang1 and Ang1* have been shown to stimulate endothelial cell migration, sprouting and tubule formation in vitro33,57–59,63,64, and Ang1* can synergize with VEGF during sprouting angiogenesis in vivo32. Ang1-induced endothelial cell motility also depends in part on PI(3)K activity56,61, inhibition of which reduces tyrosine phosphorylation of the cytoskeletal regulatory protein focal adhesion kinase (FAK) as well as the secretion of MMP-2 (REF. 69) (FIG. 6). Tie2 can also recruit additional signalling molecules that participate in cellular pathways that affect the shape and migratory properties of cells. For instance, the Tie2associated docking protein Dok-R (also known as p56Dok2 and FRIP) can potentiate NCK-dependent cell migration in response to Ang1 through stimulation of p21-activated kinase (Pak)70 (Z. Master, N.J. and D.J.D., unpublished observations) (FIG. 6). Furthermore, Tie2 has also been shown to interact with the adaptor protein Grb7 and the tyrosine phosphatase SHP2 (REFS 55,56), both of which can promote cell migration through associations with activated FAK (FIG. 6).

Tie2 regulation by phosphatases. Several tyrosine phosphatases, including Shp2, are expressed in endothelial cells, and differential recruitment of these signalling molecules to Tie2 might participate in the dynamic regulation of the receptor. Recruitment of Shp2 to Tie2 seems to modulate receptor activity as disruption of this interaction results in enhanced receptor phosphorylation and hyperactivation of Akt (C. D. Kontos, personal communication). In addition to Shp2, an endothelial cell-specific receptor-type phosphatase known as vascular endothelial protein tyrosine phosphatase (VE-PTP; the mouse orthologue of human protein tyrosine phosphatase-β, HPTPβ) and a third phosphatase known as human cellular protein tyrosine phosphatase A (HCPTPA) have also been shown to associate with Tie2 (REFS 71,72) (FIG. 6). Intriguingly, the extracellular domain of VE-PTP contains a series of motifs that have been shown to participate in receptor–ligand interactions72. It will be fascinating to examine whether the angiopoietins can simultaneously bind to both Tie2 and VE-PTP on the surface of endothelial cells, promoting the formation of Tie2/VE-PTP heterodimers. In this event, VE-PTP would prevent reciprocal phosphorylation of Tie2 and the heterodimers would be functionally inactive. Accordingly, if this ligand were preferentially Ang2 rather than Ang1, this would provide an attractive molecular model for the antagonistic properties of www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS Ang2 (FIG. 6). Structural insight into the regulation of Tie2 has recently revealed that the Tie2 kinase has numerous unique potential mechanisms of self-inhibition73. One particular phosphorylated tyrosine residue in the first kinase domain of Tie2 seems to provide negative regulation, perhaps through inhibiting receptor dimerization or through recruitment of a phosphatase such as Shp2, which could downregulate receptor activity. Interestingly, this residue is altered to serine in some families with venous malformations and this mutation leads to a marked increase in kinase activity36. Future structures of the activated form of Tie2 as well as mutated forms of Tie2 will aid in understanding the molecular regulation mechanisms of this important RTK. Tie1 signalling mechanisms. Identification of Tie1 receptor signal-transduction pathways is hampered by the fact that, despite much work, no ligand for this receptor has been found. However, recent findings indicate that Tie1 might also participate in ligand-independent signalling pathways. Tie1 is proteolytically cleaved in vivo in response to activation of protein kinase C, VEGF and inflammatory cytokines, resulting in the release of the extracellular domain74–76. The remaining fragment, composed of the transmembrane and intracellular domains of Tie1, persists as a cell-associated fragment for several hours and release of this fragment results in the association of Tie1 with several phosphoproteins, including Shp2 (REFS 75,77). It has been proposed that release of the soluble extracellular domain of Tie1 might downregulate ligand-induced signalling through Tie1 in endothelial cells while activating an alternative ligand-independent pathway involving the endodomain. Both this cleaved fragment and fulllength Tie1 exist as pre-formed complexes with Tie2 in endothelial cells, although Tie1 does not seem to be tyrosine phosphorylated under basal or stimulatory conditions, nor does it become trans-phosphorylated upon Tie2 activation78. Heterodimerization of Tie1 with Tie2 could therefore modulate Tie2 signalling by restricting the recruitment of signalling intermediates to Tie2 and preventing Tie2 activation. Alternatively, this heterodimerization might be required for the binding of an as-yet-unidentified ligand for Tie1. Further insight into the signalling pathways mediated by Tie1 awaits the identification of the activation mechanism of this enigmatic receptor.

signalling pathways in vivo. The creation of mice with mutant receptors that block binding to key downstream binding partners might be one way in which researchers can more accurately define the molecular functions of the Tie receptors in the context of a complex cellular environment. The absolute requirement for Tie1 and Tie2 during angiogenic remodelling and vessel maintenance has provided researchers with an enhanced understanding of the important biological decisions that contribute to the development of new blood vessels and the regression of existing vessels. Because altered angiogenesis is manifested by several important diseases, the ability to augment particular signal-transduction pathways will have implications in the therapeutic control of vascular growth. Interference with Tie2 signalling pathways through dominant-negative receptor approaches has already proved to be effective at inhibiting angiogenic growth in mice with tumours51–53. Ang1 also promotes formation of leakage-resistant vessels and, in combination with VEGF therapy, Ang1 might overcome the oedema caused by VEGF administration in patients with ischaemia22,23,79. Continued characterization of the specific signalling molecules recruited by these receptors to control vascular expansion and stabilization can undoubtedly lead to the development of more effective and refined targets for angiogenic therapies. But despite these advances, numerous basic questions are left unanswered and hypotheses await resolution. For instance, it remains to be established whether a highaffinity ligand exists for Tie1. Although it has recently been shown that Tie1 can heterodimerize with Tie2, it is curious that Tie1 can influence signalling pathways in the absence of receptor phosphorylation. Another challenge is to attain a greater understanding of the mechanisms of action of Ang2 as this one ligand seems to both stimulate and inhibit angiogenesis in different contexts. Ang2 might also have important functions in the development of the lymphatic vascular system. Finally, the emerging connection between the haematopoietic and angiogenic systems raises the issue of whether the signal-transduction pathways now being examined in endothelial cells also exist in Tiereceptor-expressing haematopoietic cells. Although the first reports on the Tie receptors were published almost a decade ago80,81, the relationship between the unique Tie receptors and their ligands, the angiopoietins, will undoubtedly continue to intrigue vascular biologists for years to come.

Loose ends

In recent years, enormous advances have been made towards elucidating the molecular and cellular events involved in transmitting signals from the Tie family of RTKs. It should be noted, however, that the models describing the signal-transduction mechanisms of the Tie receptors are based on in vitro studies. Because the vascular system is a complex organ system and angiogenesis involves interactions between endothelial cells and numerous other cell types, it will now be important to re-evaluate the findings outlined here using innovative approaches to study

Links DATABASE LINKS VEGF | VEGFR-1 | VEGFR-2 | VEGFR-3 | neuropilin-1 | VEGF-C | VEGF-D | angiopoietins | Tie2 | Tie1 | Ang1 | Ang4 | Ang2 | Ang3 | HIF-2 | Grb2 | MAPK | Akt | Bad | caspase-9 | nitric oxide synthase | survivin | BIR-1 | FAK | MMP-2 | Dok-R | Pak | Grb7 | Shp2 | protein kinase C | ANG2443 | aANG2B | mAng3 | cadherins | selectins | integrins | VE-cadherin | α-catenin | β-catenin | γ-catenin | vinculin | α-actinin | Sos | CAS | paxillin | HIF1α | HIF-1β | HIF-2α | p53 | Mdm2

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 6 5

REVIEWS 1.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

266

Witte, M. H., Way, D. L., Witte, C. L. & Bernas, M. Lymphangiogenesis: mechanisms, significance and clinical implications. EXS 79, 65–112 (1997). Soker, S., Takashima, S., Miao, H. Q., Neufeld, G. & Klagsbrun, M. Neuropilin-1 is expressed by endothelial and tumor cells as an isoform-specific receptor for vascular endothelial growth factor. Cell 92, 735–745 (1998). Jeltsch, M. et al. Hyperplasia of lymphatic vessels in VEGF-C transgenic mice. Science 276, 1423–1425 (1997). Karkkainen, M. J., Jussila, L., Ferrell, R. E., Finegold, D. N. & Alitalo, K. Molecular regulation of lymphangiogenesis and targets for tissue oedema. Trends Mol. Med. 7, 18–22 (2001). Mandriota, S. J. et al. Vascular endothelial growth factorC-mediated lymphangiogenesis promotes tumour metastasis. EMBO J. 20, 672–682 (2001). Stacker, S. A. et al. Vascular endothelial growth factor-D promotes the metastatic spread of cancer via the lymphatics. Nature Med. 7, 136–191 (2001). Skobe, M. et al. Induction of tumor lymphangiogenesis by VEGF-C promotes breast cancer metastasis. Nature Med. 7, 192–198 (2001) Karkkainen, M. J. et al. Missense mutations interfere with VEGFR-3 signalling in primary lymphoedema. Nature Genet. 25, 153–159 (2000). Makinen, T. et al. Inhibition of lymphangiogenesis with resulting lymphoedema in transgenic mice expressing soluble VEGF receptor-3. Nature Med. 7, 199–205 (2001) Davis, S. et al. Isolation of angiopoietin-1, a ligand for the TIE2 receptor, by secretion-trap expression cloning. Cell 87, 1161–1169 (1996). Description of the isolation of the first ligand for the Tie2 receptor using a new expression cloning technique. Maisonpierre, P. C. et al. Angiopoietin-2, a natural antagonist for Tie2 that disrupts in vivo angiogenesis. Science 277, 55–60 (1997). Shows that Ang2 is an antagonist of Tie2, providing the first evidence that related growth factors might have opposing actions towards receptor tyrosine kinases in vertebrates. Valenzuela, D. M. et al. Angiopoietins 3 and 4: diverging gene counterparts in mice and humans. Proc. Natl Acad. Sci. USA 96, 1904–1909 (1999). Shalaby, F. et al. Failure of blood-island formation and vasculogenesis in Flk-1-deficient mice. Nature 376, 62–66 (1995). Fong, G. H., Rossant, J., Gertsenstein, M. & Breitman, M. L. Role of the Flt-1 receptor tyrosine kinase in regulating the assembly of vascular endothelium. Nature 376, 66–70 (1995). Dumont, D. J. et al. Dominant-negative and targeted null mutations in the endothelial receptor tyrosine kinase, tek, reveal a critical role in vasculogenesis of the embryo. Genes Dev. 8, 1897–1909 (1994). Description of the importance of Tie2 in regulating the survival and integrity of the endothelium during development. Sato, T. N. et al. Distinct roles of the receptor tyrosine kinases Tie-1 and Tie-2 in blood vessel formation. Nature 376, 70–74 (1995). Describes the importance of Tie1 and Tie2 in vascular integrity and remodelling during embryogenesis, respectively. Suri, C. et al. Requisite role of angiopoietin-1, a ligand for the TIE2 receptor, during embryonic angiogenesis. Cell 87, 1171–1180 (1996). Shows that Ang1-deficient mice have a similar phenotype to Tie2-deficient mice and implies that Ang1 is required for endothelial–periendothelial paracrine communication during vascular development. Patan, S. TIE1 and TIE2 receptor tyrosine kinases inversely regulate embryonic angiogenesis by the mechanism of intussusceptive microvascular growth. Microvasc. Res. 56, 1–21 (1998). Puri, M., Rossant, J., Alitalo, K., Bernstein, A. & Partanen, J. The receptor tyrosine kinase TIE is required for integrity and survival of vascular endothelial cells. EMBO J. 14, 5884–5891 (1995). Describes the essential role of Tie1 in controlling the integrity of the embryonic microvascular endothelium. Puri, M. C., Partanen, J., Rossant, J. & Bernstein, A. Interaction of the TEK and TIE receptor tyrosine kinases during cardiovascular development. Development 126, 4569–4580 (1999).

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

42.

Shows that the functions mediated by the Tie1 and Tie2 receptors are not redundant through analysis of chimeric embryos. An interesting endocardial developmental phenotype is also revealed. Partanen, J. et al. Cell autonomous functions of the receptor tyrosine kinase TIE in a late phase of angiogenic capillary growth and endothelial cell survival during murine development. Development 122, 3013–3021 (1996). Suri, C. et al. Increased vascularization in mice overexpressing angiopoietin-1. Science 282, 468–471 (1998). Thurston, G. et al. Leakage-resistant blood vessels in mice transgenically overexpressing angiopoietin-1. Science 286, 2511–2514 (1999). Shows that Ang1 can block vascular permeability induced by VEGF. Jones, N. et al. Rescue of the early vascular defects in Tek/Tie2 null mice reveals an essential survival function. EMBO Rep. (in the press). Wong, A. L. et al. Tie2 expression and phosphorylation in angiogenic and quiescent adult tissues. Circ. Res. 81, 567–574 (1997). Alon, T. et al. Vascular endothelial growth factor acts as a survival factor for newly formed retinal vessels and has implications for retinopathy of prematurity. Nature Med. 1, 1024–1028 (1995). Benjamin, L. E., Hemo, I. & Keshet, E. A plasticity window for blood vessel remodelling is defined by pericyte coverage of the preformed endothelial network and is regulated by PDGF-B and VEGF. Development 125, 1591–1598 (1998). Holash, J. et al. Vessel cooption, regression, and growth in tumors mediated by angiopoietins and VEGF. Science 284, 1994–1998 (1999). Shows that Ang2 is associated with tumour angiogenesis in a model where the tumour co-opts existing host vessels. Stratmann, A., Risau, W. & Plate, K. H. Cell type-specific expression of angiopoietin-1 and angiopoietin-2 suggests a role in glioblastoma angiogenesis. Am. J. Pathol. 153, 1459–1466 (1998). Zagzag, D. et al. In situ expression of angiopoietins in astrocytomas identifies angiopoietin-2 as an early marker of tumor angiogenesis. Exp. Neurol. 159, 391–400 (1999). Brown, L. F., Dezube, B. J., Tognazzi, K., Dvorak, H. F. & Yancopoulos, G. D. Expression of Tie1, Tie2, and angiopoietins 1, 2, and 4 in Kaposi’s sarcoma and cutaneous angiosarcoma. Am. J. Pathol. 156, 2179–2183 (2000). Asahara, T. et al. Tie2 receptor ligands, angiopoietin-1 and angiopoietin-2, modulate VEGF-induced postnatal neovascularization. Circ. Res. 83, 233–240 (1998). Teichert-Kuliszewska, K. et al. Biological actions of angiopoietin-1 and–2 in a fibrin matrix model of angiogenesis. Cardiovasc. Res. 49, 659–670 (2001). Kim, I. et al. Angiopoietin-2 at high concentration can enhance endothelial cell survival through the phosphatidylinositol 3′-kinase/Akt signal transduction pathway. Oncogene 19, 4549–4552 (2000). Vikkula, M. et al. Vascular dysmorphogenesis caused by an activating mutation in the receptor tyrosine kinase TIE2. Cell 87, 1181–1190 (1996). Finds an activating mutation in the Tie2 kinase domain associated with venous malformations in two unrelated families. Calvert, J. T. et al. Allelic and locus heterogeneity in inherited venous malformations. Hum. Mol. Genet. 8, 1279–1289 (1999). Korpelainen, E. I., Karkkainen, M., Gunji, Y., Vikkula, M. & Alitalo, K. Endothelial receptor tyrosine kinases activate the STAT signaling pathway: mutant Tie-2 causing venous malformations signals a distinct STAT activation response. Oncogene 18, 1–8 (1999). Kaipainen, A. et al. Enhanced expression of the tie receptor tyrosine kinase messenger RNA in the vascular endothelium of metastatic melanomas. Cancer Res. 54, 6571–6577 (1994). Hatva, E. et al. Expression of endothelial cell-specific receptor tyrosine kinases and growth factors in human brain tumors. Am. J. Pathol. 146, 368–378 (1995). Salven, P. et al. Endothelial Tie growth factor receptor provides antigenic marker for assessment of breast cancer angiogenesis. Br. J. Cancer 74, 69–72 (1996). Peters, K. G. et al. Expression of Tie2/Tek in breast tumour vasculature provides a new marker for evaluation of tumour angiogenesis. Br. J. Cancer 77, 51–56 (1998). Korhonen, J. et al. Enhanced expression of the tie receptor tyrosine kinase in endothelial cells during neovascularization. Blood 80, 2548–2555 (1992).

43. Kukk, E. et al. Analysis of Tie receptor tyrosine kinase in haemopoietic progenitor and leukaemia cells. Br. J. Haematol. 98, 195–203 (1997). 44. Oh, H. et al. Hypoxia and vascular endothelial growth factor selectively up-regulate angiopoietin-2 in bovine microvascular endothelial cells. J. Biol. Chem. 274, 15732–15739 (1999). 45. Mandriota, S. J. & Pepper, M. S. Regulation of angiopoietin-2 mRNA levels in bovine microvascular endothelial cells by cytokines and hypoxia. Circ. Res. 83, 852–859 (1998). 46. McCarthy, M. J., Crowther, M., Bell, P. R. & Brindle, N. P. The endothelial receptor tyrosine kinase tie-1 is upregulated by hypoxia and vascular endothelial growth factor. FEBS Lett. 423, 334–338 (1998). 47. Tian, H., McKnight, S. L. & Russell, D. W. Endothelial PAS domain protein 1 (EPAS1), a transcription factor selectively expressed in endothelial cells. Genes Dev. 11, 72–82 (1997). 48. Willam, C. et al. Tie2 receptor expression is stimulated by hypoxia and proinflammatory cytokines in human endothelial cells. Circ. Res. 87, 370–377 (2000). 49. Schlaeger, T. M. et al. Uniform vascular-endothelial-cellspecific gene expression in both embryonic and adult transgenic mice. Proc. Natl Acad. Sci. USA 94, 3058–3063 (1997). 50. Iljin, K. et al. Role of ets factors in the activity and endothelial cell specificity of the mouse Tie gene promoter. FASEB J. 13, 377–386 (1999). 51. Lin, P. et al. Antiangiogenic gene therapy targeting the endothelium-specific receptor tyrosine kinase Tie2. Proc. Natl Acad. Sci. USA 95, 8829–8834 (1998). 52. Lin, P. et al. Inhibition of tumor angiogenesis using a soluble receptor establishes a role for Tie2 in pathologic vascular growth. J. Clin. Invest. 100, 2072–2078 (1997). References 51 and 52 indicate that Tie2 might be useful in tumour therapy. 53. Siemeister, G. et al. Two independent mechanisms essential for tumor angiogenesis: inhibition of human melanoma xenograft growth by interfering with either the vascular endothelial growth factor receptor pathway or the Tie-2 pathway. Cancer Res. 59, 3185–3191 (1999). 54. Millauer, B., Shawver, L. K., Plate, K. H., Risau, W. & Ullrich, A. Glioblastoma growth inhibited in vivo by a dominant-negative Flk-1 mutant. Nature 367, 576–579 (1994). 55. Huang, L., Turck, C. W., Rao, P. & Peters, K. G. GRB2 and SH-PTP2: potentially important endothelial signaling molecules downstream of the TEK/TIE2 receptor tyrosine kinase. Oncogene 11, 2097–2103 (1995). 56. Jones, N. et al. Identification of Tek/Tie2 binding partners. Binding to a multifunctional docking site mediates cell survival and migration. J. Biol. Chem. 274, 30896–30905 (1999). 57. Koblizek, T. I., Weiss, C., Yancopoulos, G. D., Deutsch, U. & Risau, W. Angiopoietin-1 induces sprouting angiogenesis in vitro. Curr. Biol. 8, 529–532 (1998). 58. Witzenbichler, B., Maisonpierre, P. C., Jones, P., Yancopoulos, G. D. & Isner, J. M. Chemotactic properties of angiopoietin-1 and -2, ligands for the endothelialspecific receptor tyrosine kinase Tie2. J. Biol. Chem. 273, 18514–18521 (1998). 59. Kwak, H. J., So, J. N., Lee, S. J., Kim, I. & Koh, G. Y. Angiopoietin-1 is an apoptosis survival factor for endothelial cells. FEBS Lett. 448, 249–253 (1999). 60. Kontos, C. D. et al. Tyrosine 1101 of Tie2 is the major site of association of p85 and is required for activation of phosphatidylinositol 3-kinase and Akt. Mol. Cell Biol. 18, 4131–4140 (1998). 61. Fujikawa, K., de Aos Scherpenseel, I., Jain, S. K., Presman, E. & Varticovski, L. Role of PI 3-kinase in angiopoietin-1-mediated migration and attachmentdependent survival of endothelial cells. Exp. Cell Res. 253, 663–672 (1999). 62. Kim, I. et al. Angiopoietin-1 regulates endothelial cell survival through the phosphatidylinositol 3′-kinase/Akt signal transduction pathway. Circ. Res. 86, 24–29 (2000). 63. Hayes, A. J. et al. Angiopoietin-1 and its receptor tie-2 participate in the regulation of capillary-like tubule formation and survival of endothelial cells. Microvasc. Res. 58, 224–237 (1999). 64. Papapetropoulos, A. et al. Direct actions of angiopoietin-1 on human endothelium: evidence for network stabilization, cell survival, and interaction with other angiogenic growth factors. Lab. Invest. 79, 213–223 (1999). 65. Jiang, B. H., Zheng, J. Z., Aoki, M. & Vogt, P. K. Phosphatidylinositol 3-kinase signaling mediates angiogenesis and expression of vascular endothelial growth factor in endothelial cells. Proc. Natl Acad. Sci.

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS USA 97, 1749–1753 (2000). 66. Papapetropoulos, A. et al. Angiopoietin-1 inhibits endothelial cell apoptosis via the Akt/survivin pathway. J. Biol. Chem. 275, 9102–9105 (2000). 67. Speliotes, E. K., Uren, A., Vaux, D. & Horvitz, H. R. The survivin-like C. elegans BIR-1 protein acts with the Auroralike kinase AIR-2 to affect chromosomes and the spindle midzone. Mol. Cell 6, 211–223 (2000). 68. Gerber, H. P. et al. Vascular endothelial growth factor regulates endothelial cell survival through the phosphatidylinositol 3′-kinase/Akt signal transduction pathway. Requirement for Flk-1/KDR activation. J. Biol. Chem. 273, 30336–30343 (1998). 69. Kim, I.et al. Angiopoietin-1 induces endothelial cell sprouting through the activation of focal adhesion kinase and plasmin secretion. Circ. Res. 86, 952–959 (2000). 70. Jones, N. & Dumont, D. J. The Tek/Tie2 receptor signals through a novel Dok-related docking protein, Dok-R. Oncogene 17, 1097–1108 (1998). 71. Huang, L. et al. HCPTPA, a protein tyrosine phosphatase that regulates vascular endothelial growth factor receptormediated signal transduction and biological activity. J. Biol. Chem. 274, 38183–38188 (1999). 72. Fachinger, G., Deutsch, U. & Risau, W. Functional interaction of vascular endothelial-protein-tyrosine phosphatase with the angiopoietin receptor tie-2. Oncogene 18, 5948–5953 (1999). 73. Shewchuk, L. M. et al. Structure of the Tie2 RTK domain. Self-inhibition by the nucleotide binding loop, activation loop, and C-terminal tail. Struct. Fold Des. 8, 1105–1113 (2001). 74. Yabkowitz, R. et al. Regulation of tie receptor expression on human endothelial cells by protein kinase C-mediated release of soluble tie. Blood 90, 706–715 (1997). 75. McCarthy, M. J. et al. Potential roles of metalloprotease mediated ectodomain cleavage in signaling by the endothelial receptor tyrosine kinase Tie-1. Lab. Invest. 79, 889–895 (1999). 76. Yabkowitz, R. et al. Inflammatory cytokines and vascular endothelial growth factor stimulate the release of soluble tie receptor from human endothelial cells via metalloprotease activation. Blood 93, 1969–1979 (1999). 77. Marron, M. B., Hughes, D. P., McCarthy, M. J., Beaumont, E. R. & Brindle, N. P. Tie-1 receptor tyrosine kinase endodomain interaction with SHP2: potential signalling mechanisms and roles in angiogenesis. Adv. Exp. Med. Biol. 476, 35–46 (2000). 78. Marron, M. B., Hughes, D. P., Edge, M. D., Forder, C. L. & Brindle, N. P. Evidence for heterotypic interaction between the receptor tyrosine kinases tie-1 and tie–2. J. Biol. Chem. 275, 39741–39746 (2000). 79. Thurston, G. et al. Angiopoietin-1 protects the adult vasculature against plasma leakage. Nature Med. 6, 460–463 (2000). 80. Partanen, J. et al. A novel endothelial cell surface receptor tyrosine kinase with extracellular epidermal

81.

82. 83.

84.

85.

86.

87.

88.

89. 90.

91.

92.

93.

94.

95.

96.

growth factor homology domains. Mol. Cell. Biol. 12, 1698–1707 (1992). Dumont, D. J., Gradwohl, G. J., Fong, G. H., Auerbach, R. & Breitman, M. L. The endothelial-specific receptor tyrosine kinase, tek, is a member of a new subfamily of receptors. Oncogene 8, 1293–1301 (1993). Gerber, H. P. et al. VEGF is required for growth and survival in neonatal mice. Development 126, 1149–1159 (1999). Gerber, H. P. et al. VEGF couples hypertrophic cartilage remodeling, ossification and angiogenesis during endochondral bone formation. Nature Med. 5, 623–628 (1999). Ferrara, N. et al. Vascular endothelial growth factor is essential for corpus luteum angiogenesis. Nature Med. 4, 336–340 (1998). Pe’er, J. et al. Upregulated expression of vascular endothelial growth factor in proliferative diabetic retinopathy. Br. J. Ophthalmol. 80, 241–245 (1996). Iwama, A. et al. Molecular cloning and characterization of mouse TIE and TEK receptor tyrosine kinase genes and their expression in hematopoietic stem cells. Biochem. Biophys. Res. Commun. 195, 301–309 (1993). Rodewald, H. R. & Sato, T. N. Tie1, a receptor tyrosine kinase essential for vascular endothelial cell integrity, is not critical for the development of hematopoietic cells. Oncogene 12, 397–404 (1996). Takakura, N. et al. Critical role of the TIE2 endothelial cell receptor in the development of definitive hematopoiesis. Immunity 9, 677–686 (1998). Associates Tie2 with the regulation of haematopoiesis. Takakura, N. et al. A role for hematopoietic stem cells in promoting angiogenesis. Cell 102, 199–209 (2001). Carmeliet, P. et al. Targeted deficiency or cytosolic truncation of the VE-cadherin gene in mice impairs VEGFmediated endothelial survival and angiogenesis. Cell 98, 147–157 (1999). Wiesener, M. S. et al. Induction of endothelial PAS domain protein-1 by hypoxia: characterization and comparison with hypoxia-inducible factor-1α. Blood 92, 2260–2268 (1998). Claffey, K. P. et al. Identification of a human VPF/VEGF 3′ untranslated region mediating hypoxia-induced mRNA stability. Mol. Biol. Cell 9, 469–481 (1998). Maltepe, E., Schmidt, J. V., Baunoch, D., Bradfield, C. A. & Simon, M. C. Abnormal angiogenesis and responses to glucose and oxygen deprivation in mice lacking the protein ARNT. Nature 386, 403–407 (1997). Ryan, H. E., Lo, J. & Johnson, R. S. HIF-1α is required for solid tumor formation and embryonic vascularization. EMBO J. 17, 3005–3015 (1998). Peng, J., Zhang, L., Drysdale, L. & Fong, G. H. The transcription factor EPAS-1/hypoxia-inducible factor 2α plays an important role in vascular remodeling. Proc. Natl Acad. Sci. USA 97, 8386–8391 (2000). Ravi, R. et al. Regulation of tumor angiogenesis by p53-

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

97.

98.

99.

100.

101.

102.

103.

104.

105.

106.

107.

induced degradation of hypoxia-inducible factor 1α. Genes Dev. 14, 34–44 (2000). Carmeliet, P. et al. Role of HIF-1α in hypoxia-mediated apoptosis, cell proliferation and tumour angiogenesis. Nature 394, 485–490 (1998). Maxwell, P. H. et al. The tumour suppressor protein VHL targets hypoxia-inducible factors for oxygen-dependent proteolysis. Nature 399, 271–275 (1999). Lisztwan, J., Imbert, G., Wirbelauer, C., Gstaiger, M. & Krek, W. The von Hippel-Lindau tumor suppressor protein is a component of an E3 ubiquitin-protein ligase activity. Genes Dev. 13, 1822–1833 (1999). Graeber, T. G. et al. Hypoxia-mediated selection of cells with diminished apoptotic potential in solid tumours. Nature 379, 88–91 (1996). Procopio, W. N., Pelavin, P. I., Lee, W. M. & Yeilding, N. M. Angiopoietin-1 and-2 coiled coil domains mediate distinct homo-oligomerization patterns, but fibrinogen-like domains mediate ligand activity. J. Biol. Chem. 274, 30196–30201 (1999). Huang, Y. Q., Li, J. J. & Karpatkin, S. Identification of a family of alternatively spliced mRNA species of angiopoietin-1. Blood 95, 1993–1999 (2000). Kim, I. et al. Characterization and expression of a novel alternatively spliced human angiopoietin-2. J. Biol. Chem. 275, 18550–18556 (2000). Mezquita, J., Mezquita, B., Pau, M. & Mezquita, C. Characterization of a novel form of angiopoietin-2 (Ang-2B) and expression of VEGF and angiopoietin-2 during chicken testicular development and regression. Biochem. Biophys. Res. Commun. 260, 492–498 (1999). Korhonen, J. et al. Endothelial-specific gene expression directed by the tie gene promoter in vivo. Blood 86, 1828–1835 (1995). Schlaeger, T. M., Qin, Y., Fujiwara, Y., Magram, J. & Sato, T. N. Vascular endothelial cell lineage-specific promoter in transgenic mice. Development 121, 1089–1098 (1995). Jones, N. & Dumont, D. J. Recruitment of Dok-R to the EGF receptor through its PTB domain is required for attenuation of Erk MAP kinase activation. Curr. Biol. 9, 1057–1060 (1999).

Acknowledgements We thank G. Yancopoulos, C. Kontos, M. Detmar and Z. Master for discussions of unpublished data. The authors also acknowledge the support of the Canadian Institutes of Health Research (the former Medical Research Council (MRC) of Canada), the National Cancer Institute of Canada, the Juvenile Diabetes Foundation International and Cancer Care Ontario to D.J.D. D.J.D. is a scientist of the MRC of Canada. N.J. was supported by an MRC studentship award. K.I. and K.A. are supported by the Finnish Academy, the Sigrid Juselieus Foundation, the University of Helsinki Hospital, the State Technology Development Center as well as by EU Biomed programme. We apologize for the failure to cite many important contributions to this field owing to space limitations.

VOLUME 2 | APRIL 2001 | 2 6 7

REVIEWS

A REAL-TIME VIEW OF LIFE WITHIN 100 NM OF THE PLASMA MEMBRANE J. A. Steyer* and W. Almers‡ The plasma membrane is a two-dimensional compartment that relays most biological signals sent or received by a cell. Signalling involves membrane receptors and their associated enzyme cascades as well as organelles such as exocytic and endocytic vesicles. Advances in light microscope design, new organelle-specific vital stains and fluorescent proteins have renewed the interest in evanescent field fluorescence microscopy, a method uniquely suited to image the plasma membrane with its associated organelles and macromolecules in living cells. The method shows even the smallest vesicles made by cells, and can image the dynamics of single protein molecules. CAVEOLAE

Flask-shaped, cholesterol-rich invaginations of the plasma membrane, thought to be involved in cell signalling. LIPID RAFTS

Micro-aggregates of cholesterol and sphingomyelin thought to occur in the plasma membrane. GREEN FLUORESCENT PROTEIN

Isolated from the jellyfish Aequorea victoria. Can be genetically conjugated with proteins to make them fluorescent. The most widely used mutant, EGFP, has an emission maximum at 510 nm.

*Department of Molecular and Cell Biology, University of California, Berkeley, California 94720, USA. ‡ Vollum Institute, Oregon Health Sciences University, Portland, Oregon 97210, USA. Correspondence to W.A. e-mail: almersw@ohsu.edu

268

The plasma membrane is a busy place. Exocytic vesicles insert receptors into the plasma membrane and release ligands into the extracellular space. Endocytic vesicles carry receptors with bound ligand to internal processing stations. CAVEOLAE are plasma-membrane-associated vesicles with a presumed role in cell signalling1. LIPID RAFTS are thought to populate the plasma membrane as small floating islands2 in which select membrane proteins meet in private to exchange signals. Finally, there is the universe of membrane receptors. Many are probably embedded in large molecular complexes that continually recruit and release downstream effector molecules. Most or all these structures are highly dynamic, do their jobs in milliseconds to minutes and sometimes disperse soon thereafter. Only rarely do organelles of the same type act in synchrony. To observe single events mediated by single organelles and signalling complexes, we require in vivo methods that image single organelles, detect molecules in small numbers and report their function at high resolution in time and space. Fluorescence microscopy is a natural choice in so far as some organelles may be stained specifically with dyes, and more and more proteins have been conjugated with fluorescent proteins such as GREEN FLUORES3 CENT PROTEIN (GFP) without impairing their function . However, most plasma membrane events involve inter-

actions with cytosolic proteins that have been recruited to the plasma membrane transiently and in small numbers. And organelles of a given type often inhabit the entire cell. Because even CONFOCAL MICROSCOPES look into cells to a depth of nearly half a micron when focused on the plasma membrane, these and more conventional fluorescence microscopes show strong ‘background’ fluorescence from the cytosol that obscures the weaker fluorescence from small structures or molecular assemblies near the plasma membrane. Evanescent field (EF) fluorescence microscopy overcomes this problem because it provides depth discrimination of near-molecular dimensions. This review focuses on recent applications of this method to membrane dynamics and signal transduction. We ignore a large number of interesting papers with a more biophysical orientation4, including those on cell–substrate contacts5 and on single molecule imaging6. Evanescent fields

An EF can form when a beam of light travelling in a medium of high REFRACTIVE INDEX, such as glass, encounters one of lower refractive index such as the adjoining water or an adherent cell. When the angle of incidence α is small, light is refracted and propagates through the interface. But when α exceeds a certain ‘critical angle’, www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

Box 1 | Evanescent fields Glass

Water

Exocytic vesicle

Intensity

α 0

200

400

Distance (nm)

Cell

Out

When a parallel beam of light in a medium of high refractive index (n1) strikes an interface with a medium of lower refractive index (n2) it suffers total internal reflection if the angle of incidence, α, exceeds the so-called critical angle. Total internal reflection generates an evanescent field in the medium of lower refractive index (a). The intensity of the evanescent field medium declines exponentially with distance from the interface (b), falling 37% within the so-called ‘penetration depth’ d: d=

(1)

4 π √NA i–n 2 2

in which λ is the wavelength of light and NAi = n1 sin α is the numerical aperture of incidence. The larger the difference between NAi and n2, the smaller is d. In one example33, light propagates through a coverslip of high refractive index glass (n1 = 1.8) at an angle of α = 66° onto adhering cells (typically n2 = 1.37). EQN 1 predicts d = 43 nm (b). EQN1 holds strictly in a homogenous medium such as water, and holds approximately in cells if local variations of n2 are small compared to (NAi–n2). Otherwise light-scattering structures can significantly increase the apparent penetration depth of an evanescent field43 and even cause it to become propagated19,44. CONFOCAL MICROSCOPE

A fluorescence microscope achieving improved depth discrimination by blocking fluorescence that originates outside the plane of focus by use of a ‘confocal’ pinhole. In most confocal microscopes, a laser beam focused to a small spot provides the excitation light and scans the image point by point. REFRACTIVE INDEX

In a transparent medium, the refractive index is defined as the speed of light in a vacuum (or air) divided by the speed of light in the medium. It determines the change in direction undergone by a beam of light when it strikes an interface between two media of different refractive indices. TOTAL INTERNAL REFLECTION

Highly efficient reflection occurring at the interface with a transparent medium of lower refractive index when light strikes the interface at a glancing angle.

light instead undergoes TOTAL INTERNAL REFLECTION (BOX 1). Classical electrodynamics does not allow an electromagnetic wave to vanish discontinuously at an interface, therefore total internal reflection sets up a thin layer of light in the water or cell, called the evanescent field (EF). An EF selectively illuminates fluorescent molecules near the interface and leaves more remote structures in the dark. In the example of BOX 1, the EF reaches from the plasma membrane into the cytosol for little more than 100 nm, a distance comparable to the thickness of ultrathin sections cut for electron microscopy. EF microscopes are extremely sensitive to movement of fluorescent objects vertical to the glass, as structures brighten when they approach the glass and dim when they retreat. Brightness is proportional to the illumination intensity within the evanescent field, but also to the availability of the emitted light for collection by a microscope objective (BOX 2). In the example of BOX 1, a 20-nm movement would produce a 37% change in illumination and, in the ‘through-the-lens’ configuration (see below), a 40% change in detected fluorescence (BOX 2b). In depth discrimination, EF fluorescence microscopy is up to tenfold better than confocal microscopy, the only other fluorescence microscopic technique developed for this purpose. Confocal and EF fluorescence microscopy are compared in TABLE 1.

There are two common configurations for EF microscopes — prism and through-the-lens (BOX 3). Prism microscopes can illuminate a larger field of view. This is an advantage when simultaneously imaging several cells or a single cell that spreads over large areas, such as a neuron with its axon and dendrites. Prism-type EF microscopes also have the least background light7. However, the sample is sandwiched in a narrow space between the prism and the objective lens and access to it is restricted. Moreover, the brightness of an object does not necessarily vary monotonically with its distance from the interface (BOX 2c). This is a disadvantage when interpreting brightness changes in terms of movement in and out of the evanescent field. The prism method is easy to implement, but we know of no commercial supplier. Through-the-lens microscopes allow free access to the specimen, collect fluorescent light more efficiently from objects near the interface (BOX 2) and have higher image quality and spatial resolution. Good through-thelens microscopes require unusual objectives that have become available only recently. So most early work with EF fluorescence has used the prism method (BOX 4). Attachments and objectives for through-the-lens microscopes are available from Olympus Co. Imaging secretory granules

The molecular mechanism of exocytosis has been intensely studied biochemically8. Although ELECTROPHYSIOLOGY has been the method of choice in functional studies 9,10, this method directly assays only exocytosis, the last step in a long sequence of events. Precursor steps such as the docking of vesicles at the plasma membrane and their preparation for exocytosis must be inferred indirectly by kinetic modelling. Hence it was desirable to image SECRETORY VESICLES and granules before exocytosis. It is not difficult to make secretory vesicles fluorescent. Vesicle-resident proteins can be conjugated with GFP (FIG. 1) 11,12 or one may use the fact that most vesicles are acidic inside and therefore accumulate fluorescent weak bases such as acridine orange or quinacrine. In MAST CELLS of mutant mice, secretory granules are so large (1–3 µm diameter) that they could be observed with simple EPIFLUORESCENCE MICROSCOPY while they discharged quinacrine by exocytosis13. The large granules of sea urchin eggs could also be observed by epifluorescence microscopy14. But fluorescence imaging of the ten times smaller and more densely packed secretory granules of endocrine cells required EF fluorescence microscopy 15. After CHROMAFFIN CELLS had accumulated acridine orange in their granules, stimulation caused the granules to release their fluorescence as a short-lived fluorescent cloud, and then to dim or vanish. Once the plasma membrane was stripped of docked granules, fresh granules arrived from the cytosol. Some of them approached by directed motion as if moving along filamentous tracks, and then lost mobility because they had either docked at the plasma membrane or become ensnared in the dense network of actin filaments that invests the plasma membrane of most cells.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 6 9

REVIEWS

Box 2 | Light collection near a glass–water interface

SECRETORY VESICLES

Used to sequester molecules within a cell and then deliver (secrete) them to the extracellular space by exocytosis. MAST CELL

A type of leukocyte with large secretory granules containing histamine and various protein mediators. Granules are especially large in the beige mouse mutant.

Glass

60 40

Water

c Image brightness

Three variants are used to study secretion. First, a glass micropipette penetrating a neuron records electric currents through plasma membrane ion channels opened by secreted neurotransmitter. Second, a carbon fibre placed near the surface of a secretory cell records a current as it oxidizes secreted catecholamines. Third, the increase in cell surface area following exocytosis is measured by monitoring the electrical capacitance of the plasma membrane. All three methods are sensitive enough to detect the exocytosis of single vesicles.

b 80

Image brightness

ELECTROPHYSIOLOGY

Light emitted (%)

1.0

Water 0.5

1.0

Water

0.5

Glass

Glass 0

100

200

300

Distance from glass (nm)

400

0.0

100

200

Distance from glass (nm)

0.0

100

200

300

400

Distance from glass (nm)

Apart from radiating light in all directions, a fluorescent molecule also generates an electromagnetic ‘near field’ that, like an evanescent field, does not propagate but declines rapidly with distance. Where the near field extends into the glass, it radiates into the glass as propagated light, entering at an angle larger than the critical angle45. In so far as the total fluorescence emitted by the molecule does not change, the energy thus captured from the near field diminishes the fluorescent light propagating elsewhere. Hence objects close to the interface emit most of their fluorescence into the glass at the expense of the aqueous phase. Objects at greater distance emit more fluorescence into the water, as propagated light is reflected from the glass back into the water. These effects occur regardless of whether fluorescence is excited by an EF or by propagated light. They are illustrated in panel a, which shows the light emitted into water (blue) and into glass (red). The blue curve was calculated for dipoles of random orientation as in References 46 and 47 but for λ = 520 nm and high refractive index glass (n1 = 1.8) contacting water (n2 = 1.33). The red curve is one minus the blue curve. A 1.65 NA objective is expected to collect most of the light emitted into the glass. The red and blue curves in panel b show how bright a fluorescent object appears at various distances when it is illuminated with an evanescent field as in Box 1 (re-drawn as the black curve). Brightness depends both on the intensity of illumination (black) and on what portion of the emitted fluorescence is available for collection (red and blue in panel a). In glass, both factors combine in dimming the object as it retreats from the interface. This is shown in the red curve, which multiplies the black curve by the red curve in a. In water, the two factors oppose each other, and the blue curve is the black curve multiplied by the blue curve in a. All three curves were scaled to coincide at the interface. We assumed that ideal objectives collect all light emitted into glass on one side and water on the other. Light emitted into glass falls more steeply with distance than the illumination intensity, and light emitted into water less steeply. Panel c is similar to panel b but with a more deeply penetrating evanescent field (d = 150 nm, black). Note that light detected through water first rises and then falls as the object moves into water.

EPIFLUORESCENCE

Most common fluorescence microscopy arrangement wherein excitation light is applied through the objective that is also used for viewing the fluorescent specimen. The method excites and collects fluorescence throughout the cell and has poor depth discrimination. CHROMAFFIN CELLS

Cells of the medulla of the adrenal gland. They store and secrete adrenaline, noradrenaline and protein hormones. They are termed ‘chromaffin’ because they can be stained by chromium salts. INS-1 CELLS

Cell line derived from pancreatic β-cells that secrete insulin. PC-12 CELLS

A cell line derived from a tumour of CHROMAFFIN CELLS. Used as a substitute for chromaffin cells in secretion studies. Compared with chromaffin cells, they have smaller and fewer secretory granules, but also contain vesicles similar to synaptic vesicles.

270

Occasionally granules left the plasma membrane after remaining there for tens of seconds. The replenishment of the cell surface with fresh granules took some 10–20 min, a surprisingly long time considering that a docked and release-ready granule can undergo exocytosis in tens of milliseconds16 or less17 after a rise in internal Ca2+ concentration. Related observations were made by others18–20. In INS-1 CELLS, the interior of secretory granules was labelled with acridine orange and their membrane was labelled with GFP-conjugated phogrin, a membrane protein whose function is unclear20. Both orange and green fluorescence were observed simultaneously. Interestingly, the green image of the empty granule membrane remained visible for several seconds after the release of acridine orange. Evidently the granule membrane does not flatten into the plasma membrane immediately after exocytosis. Sometimes the empty granule detached from the membrane and moved away, in direct support of the idea that empty granules can be retrieved intact by the cell21,22. Single secretory granules in live endocrine cells have also been observed by confocal microscopy12,23. Secretory granules were observed after deconvolution of epifluorescence images in live PC-12 (REF. 24) and pituitary cells25. Upon exocytosis, pituitary granule matrices stained brightly with FM1-43 and were retrieved intact in large endocytic vesicles25.

Constitutive exocytosis

This process had previously been studied almost entirely by biochemical methods and electron microscopy. Neither method easily provides information on how often and how fast exocytic events happen. Two groups expressed a fluorescently labelled membrane protein in exocytic organelles (TRANSPORT CONTAINERS) of epithelial cells and observed the cells by EF fluorescence26,27 and by epifluorescence27. To shorten the penetration depth, both groups plated cells on special high-refractive index glass and used special optics to achieve a large angle of incidence. Video imaging of single transport containers provided a wealth of new results. Transport containers undergoing exocytosis were both spherical and tubeshaped. Exocytosis was apparent as fluorescent membrane protein escaped from transport containers and spread into the plasma membrane by lateral diffusion. Transport containers moved around in the cytosol, arrived in the evanescent field, stopped at or close to the plasma membrane and then presumably docked. There they remained for tens of seconds or minutes until they either returned into the cytosol or underwent exocytosis. The large number of apparently docked vesicles was surprising, as it is normally a hallmark of REGULATED EXOCYTOSIS. As a second surprise, tubular transport containers released only a small portion of their membrane proteins as they fused transiently with the plasma memwww.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

Table 1 | Comparison of two microscopic techniques Confocal

Evanescent field

Floods entire cell with excitation light, but rejects most fluorescence in confocal pinhole.

Illuminates only the plane of interest. No out-of-focus fluorescence. All fluorescence usable for imaging.

Extensive bleaching and photodamage throughout the cell. Tens of useful pictures per cell.

Less bleaching and photodamage per collected photon. Hundreds of useful pictures per cell.

Limited vertical resolution blurs a point into a vertical ellipse about 500-nm long.

Illumination declines exponentially over distances to 40–50 nm when high refractive index glass is used.

Pixels imaged sequentially in laser scanning microscopes. Imaging relatively slow.

All pixels imaged simultaneously. Rate of imaging limited only by speed of camera and photon collection.

Can image entire cell.

Images only cell surface.

FM1-43

A water-soluble lipid that becomes fluorescent when it reversibly enters a lipid bilayer or the protein cores of some dense core granules. TRANSPORT CONTAINERS

Vesicles or tubules derived from the trans-Golgi network that can either undergo exocytosis or give rise to vesicles capable of exocytosis. REGULATED EXOCYTOSIS

In regulated exocytosis, vesicles accumulate beneath the plasma membrane and wait for a signal, such as an increase in cytosolic [Ca2+]i. By contrast, constitutive exocytosis is thought to occur as soon as vesicles arrive at the plasma membrane. FUSION PORE

Small opening that allows flux of cargo between two membrane-bounded compartments. Fusion pores form at an early stage of membrane fusion and widen when they lead to full fusion. DENSE CORE GRANULES

Large (100–1,000 nm diameter) secretory vesicles that concentrate and then secrete proteins. Because of their high protein content they stain heavily and hence appear to have a ‘dense core’ under the electron microscope.

brane and then closed their FUSION PORE. The time sequence in FIG. 2 shows an example, imaged both by epifluorescence to emphasize structures >100 nm from the plasma membrane, and by EF fluorescence for structures <100 nm. Both methods imaged the same fluorophore but the epifluorescence signal was coded red and the EF signal green. An arrow marks a transport container at some distance (red). At 7 s, part of it had turned green and hence had entered the evanescent field. Its red extension indicates that it was a tube extending beyond the evanescent field. Exocytosis appears as the spread of green fluorescence into the plasma membrane (8–12 s). After the green cloud faded (12 s), a green and red spot remained at the fusion site (41 s), indicating that the transport container remained

Box 3 | Two types of evanescent field microscope In prism-type microscopes (a), a prism directs light into a a coverglass bearing cells. Fluorescence excited by the evanescent field (EF) is collected opposite the reflecting interface, either with an objective dipping into an open Objective lens chamber as shown, or with an oil immersion objective if the chamber is covered by a coverslip. With a hemicylinder prism, light can easily be applied at various angles for varying the penetration depth of the Water EF19,44. Prism-type microscopes have completely separate paths for excitation and emission light and for Glass this reason have the lowest background light. With a slide wide laser beam, it is easy to generate an EF over large In Out areas, thereby illuminating a wide field of view. However, Prism prism-type microscopes must look through a cell to see its bottom surface and this tends to degrade the image quality. They also tend to view cells through objectives of b longer working distances and lower resolution. Water Through-the-lens microscopes (b) generate the EF with the objective lens that is used for viewing the cell. Glass How is this done? From basic optics, light diverging from slide a point source can be made parallel by placing the point Objective lens source in the focal plane of a converging lens. The same Focal plane is true for an objective lens and the focal plane at the back of the objective (dashed). The further off-axis the point source, the larger the angle at which the parallel Out beam leaves the objective. The largest angle α at which an objective can emit (and receive) light is expressed in terms of its numerical aperture (NA) as NA = n sin α, in which n is the refractive index for which the objective Dichroic mirror In has been designed. Total internal reflection requires that the NA be higher than the refractive index of the specimen, n2 = 1.37 for a typical cell. Until recently, the highest available NA was 1.4, barely enough to achieve total internal reflection at the interface between the glass and the cell43. However, a special objective with NA = 1.65 is now available (Olympus APO 100x O HR)48. Through-the-lens set-ups can be used like normal inverted epifluorescence microscopes and allow electrodes, pipettes or even an atomic force microscope on the stage without compromising optical quality. Because both the light collection efficiency and spatial resolution of an objective increase with its numerical aperture, through-thelens set-ups automatically excel on both counts. Because they view cells through glass, through-the-lens set-ups have two advantages over prism set-ups. First, they collect more light from near-by objects (BOX 2). Second, light collected diminishes monotonically with distance, so brightness changes can be converted, at least approximately, into movements. This cannot be guaranteed for light emitted into water. However, the 1.65 NA objective requires special and costly coverslips of high-refractive index glass as well as special immersion oil that passes blue light poorly. But objectives using normal coverglass and immersion oil are becoming available from Olympus and Zeiss Co. Their numerical aperture is only 1.45, but this exceeds the refractive index of cells (1.37) sufficiently to be useful for this application. Through-the-lens set-ups have been discussed in recent articles38,49.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 7 1

REVIEWS

Box 4 | Applications of the method • First study reporting the imaging of cells by evanescent field microscopy50. • First through-the-lens evanescent field microscope51. • Imaging clusters of a fluorescently labelled plasma membrane protein52. • Adhesion of cells to substrates52,53; see REF. 5 for review. • Imaging cells with through-the-lens evanescent field fluorescence54.

SYNAPTIC VESICLES

Small-diameter (20–65 nm) secretory vesicles that store and secrete neurotransmitters. Capable of exocytosis within fractions of a millisecond after a stimulus. They do not enclose protein and hence lack a dense core.

• Theory of evanescent field microscopy45,55,56. • Measuring the distance between substrates, plasma membranes and organelles by variable-angle evanescent field illumination19,44,47,57. • Fluorimetric tracking of a cell’s volume58. • Imaging Ca2+ concentration changes near the plasma membrane59,60.

ACTIVE ZONE

Structurally well-defined zone in presynaptic nerve terminals constituting a preferred site for the exocytosis of synaptic vesicles.

there and retained some or most of its cargo. The direct observation of an exocytic organelle fusing and then disconnecting adds to the extensive electrophysiological evidence on endocrine21,22 and mast cells28,29 in which DENSE CORE GRANULES undergo incomplete exocytosis as they open and then close transient fusion pores.

RETINAL BIPOLAR NEURONS

The predominant neurons in the inner nuclear layer of the retina. At the end of a short axon, they carry an unusually large synaptic terminal that can be directly studied by capacitance measurements.

Synaptic vesicles

By virtue of their small size, SYNAPTIC VESICLES can release all their neurotransmitter in fractions of a millisecond. Indeed, synaptic vesicles are the smallest membranous organelles made in any cell (diameter 30–50 nm in brain). Whereas transport containers fuse continuously and at apparently random locations over the entire cell surface26, synaptic vesicles dock preferentially at so-called ACTIVE ZONES, and fuse there within a millisecond after an electric stimulus. Synaptic vesicles have been intensely studied using biochemistry and electrophysiology but, as with secretory granules, no method existed to record signals from single synaptic vesicles before exocytosis. It therefore seemed desirable to image synaptic vesicles. After exocytosis, the membrane of synaptic vesicles is retrieved by endocytosis30,31, but may be loaded during its brief stay at the plasma membrane with a few hundred molecules of the fluorescent lipid FM1-43 (REF. 32). EF fluorescence was used to image FM1-43 stained giant synaptic terminals of RETINAL BIPOLAR NEU-

SYNAPTIC RIBBON

Proteinaceous structure at the active zones of some sensory neurons. Thought to transport and/or capture synaptic vesicles in preparation for exocytosis. VOLTAGE CLAMP

Electrophysiological amplifier that controls the plasma membrane voltage by electronic feedback, and reports the current that must pass across the plasma membrane to maintain the desired voltage. Used, for example, to open and close voltage-gated Ca2+ channels. SNARES

Proteins required for membrane fusion in exocytosis and other membrane traffic events. Vesicle SNAREs on the vesicle membrane bind to target SNAREs on the target plasma membrane. When such transSNARE complexes are formed, they pull the two membranes close together and presumably cause them to fuse. DIFFERENTIAL INTERFERENCE CONTRAST MICROSCOPY

Forms images of high contrast and resolution in unstained cells by using birefringent prisms and polarized light.

272

Time (ms)

–1,000

100

where so-called SYNAPTIC RIBBONS tether vesicles near active zones34. Single vesicles were observed in terminals plated on high refractive index glass33. Many vesicles made brief visits to the plasma membrane, rapidly bouncing into and out of the evanescent field. Others were bright and immobile as if docked at the plasma membrane. A stimulus delivered with a VOLTAGE CLAMP caused the docked vesicles to undergo rapid exocytosis, visible as the release of FM1-43 from the vesicle and its spread into the plasma membrane by lateral diffusion (FIG. 3). The stimulus also caused new vesicles to dock and replace those lost through exocytosis. They were ready to fuse after being docked for 0.2–0.3 s. Although vesicles fused and docked mostly at discrete active zones of submicron diameter, they occasionally did so elsewhere on the plasma membrane. As in chromaffin and epithelial cells, capture of vesicles at or near the plasma membrane was reversible. Interestingly, vesicles at active zones often appeared motionless but dim until a stimulus caused them to brighten as they moved to the plasma membrane and then fused. Evidently, active zones contain a cytosolic structure, probably the synaptic ribbon, that holds vesicles in reserve a short distance from the plasma membrane. Given the penetration depth of the evanescent field, it was calculated that vesicles were held about 20 nm away from the plasma membrane. This is close enough for v-SNARE proteins on the vesicle to reach out and touch t-SNARE proteins on the plasma membrane, and thereby initiate the formation of the SNARE complexes needed for fusion35. Whether or not the 20-nm movement actually reflects formation of the SNARE complex, the finding illustrates that EF fluorescence can image molecular-sized motions in living cells. RONS

Organelle movement

Most organelles travel extensively within cells. Classical work with DIFFERENTIAL INTERFERENCE CONTRAST (DIC) MICROSCOPY has shown how microtubules transport organelles over long distances and how they segregate chromosomes. DIC is less well suited, however, to explore submicron movement of densely packed organelles in the very periphery of a cell. Such movement must occur if secretory granules are to occupy their docking site beneath the plasma membrane, and if endocytic vesicles at the plasma membrane are to reach their processing stations in the cytosol. In either direc133

200

900

Figure 1 | Chromaffin cell expressing GFP-conjugated pro-neuropeptide Y (p-NPY) in its granules. p-NPY is normally contained in the secretory granules of chromaffin cells. Pictures taken at various times relative to a voltage jump from –70 mV to 0 mV that opened Ca2+ channels and stimulated exocytosis of a granule in the centre of the cell. Note the rapid spread of fluorescence, followed by disappearance of the granule. Scale bar, 2 µm. (Courtesy of I. Kleppe. Olympus APO 100x O HR 1.65 NA objective.)

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS a Golgi

Time (s)

1 0

Figure 2 | Constitutive exocytosis. Epithelial cell expressing vesicular stomatitis virus glycoprotein tagged with yellow fluorescent protein and targeted to the plasma membrane. a | Schematic of the structures imaged in b. A tubular transport container approaches the plasma membrane, fuses and then disconnects. b | Pictures with evanescent field (EF) fluorescence showing structures <100 nm from the plasma membrane are overlaid on epifluorescence pictures that also show structures further away. Zone imaged with EF fluorescence in green, that with epifluorescence in red. Superimposed pairs of EF (green) and epifluorescence images (red) taken at indicated times. Structures illuminated both by the EF and by epifluorescence appear yellow. One transport container is marked with an arrow in the first image. Images of this transport container are hypothesized to be vertical projections of the structures drawn in a. Lines connect images with the structures they are hypothesized to represent. Images taken using the prism method27. Scale bar, 2 µm.

PHALLOIDIN

Family of toxins present in the highly poisonous agaric fungus, Amanita phalloides. Phalloidin binds specifically to actin filaments and prevents their depolymerization. LATRUNCULIN

Toxin present in the sponge Latruncula magnifica, which binds to monomeric actin and depolymerizes actin filaments.

In permeabilized PC-12 cells37, granule motion stopped when ATP was replaced by a non-hydrolysable analogue, or when the turnover of cortical actin was blocked by PHALLOIDIN. These results indicate that the movement of granules might require energy and might be hindered by actin filaments. Surprisingly, granule motion diminished also when most cortical actin was disassembled by LATRUNCULIN. Actin seems to both help and hinder granule motion, as previously shown in pancreatic ACINAR 39 CELLS . These studies are a modest beginning, but they promise that future work will provide important insights. A surprising role for actin in endocytosis was found when MACROPINOCYTOSIS was observed in cultured mast cells expressing GFP-β-actin40. While separating from the plasma membrane, macropinocytic vesicles ignited a burst of actin polymerization that drove them into the cell interior much as actin ‘comet tails’ drive microorganisms such as Listeria monocytogenes through infected cells. Filamentous actin was long thought to function during endocytosis, but this was the first demonstration of actin-driven movement initiated by endocytosis.

tion, organelles must penetrate the so-called actin cortex beneath the plasma membrane, a dense meshwork of actin filaments that is a few hundred nanometres thick. To the extent that actin filaments constantly assemble and disassemble, the meshwork is dynamic and permeable to organelles. Control mechanisms regulating the assembly and disassembly would also regulate the permeability of the actin cortex. To study them, a simple method to track organelle movement in the actin cortex would be useful. EF fluorescence studies in PC-12 cells showed that granules beneath the plasma membrane come in three types36,37. A few move in a directed fashion over micron distances, others dither about without apparent direction and many do not measurably move at all. In resting chromaffin cells, all but a few per cent of the granules near the surface dither ~70 nm around a resting position as if tethered there or imprisoned in a cage43, most likely the actin network. Movement over longer distances is extremely slow. By comparison, some giant synaptic terminals contain large reservoirs of mobile synaptic vesicles33. a

Time (ms)

–342

–17

+17

+50

+204

ACINAR CELLS

Acinar cells in the mammalian pancreas are responsible for the secretion of digestive enzymes. Like mast cells, they have large secretory granules. MACROPINOCYTOSIS

Actin-dependent process by which cells engulf large volumes of fluids.

Figure 3 | Synaptic vesicles in a goldfish retinal bipolar nerve terminal. a | Footprint of a nerve terminal adhering to a coverslip. Synaptic vesicles stained with FM1-43. Average of 500 images. Bright spots show places frequently occupied by vesicles. Scale bar, 3 µm. b | Vesicle undergoing exocytosis; times are relative to a voltage jump from –60 mV to 0 mV that opened Ca2+ channels. Average from five spatially and temporally aligned vesicles. Scale bar, 1 µm. (Images taken using an Olympus APO 100x O HR NA 1.65 objective31.)

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 7 3

REVIEWS

DSRED

(dsRed). A red fluorescent protein (emission maximum at 583 nm) that, like GFP, can be genetically conjugated to proteins to make them fluorescent. First isolated from tropical corals of the Discosoma genus. CFP

Variant of GFP engineered to fluoresce in cyan. Its emission maximum is at 474 nm. YFP

Variant of GFP engineered to fluoresce in yellow. Its emission maximum is at 525 nm.

1. 2. 3. 4.

274

Signal transduction

An important use of EF fluorescence will be to image signal-transduction events. Although the potential of the method for this purpose remains to be fully exploited, it can be illustrated by two examples. In one study41, red fluorescent epidermal growth factor (Cy3–EGF) molecules were applied to cells. In an interesting but probably difficult variant of EF fluorescence, the authors sought to image the plasma membrane not where it adhered to the coverslip but instead at the opposite side where the cell faced the free solution. For generating the evanescent field, they took advantage of the minute refractive index differences between the cell and the surrounding medium. Dim spots of similar brightness appeared abruptly as epidermal growth factor (EGF) molecules were captured by the plasma membrane and then vanished some time later. Their persistence time varied inversely with the illumination intensity, as expected for the photobleaching of single molecules. Measurement of the fluorescence intensity of spots indicated two populations, one twice as bright as the other, as if the brighter spots originated from two EGF molecules bound to an EGF receptor dimer. These observations are a fine example of how single ligand molecules can be watched as they bind to the plasma membrane of an intact cell. Cells were next incubated with mixtures of red Cy3–EGF and deep-red Cy5–EGF, and then viewed with a camera system that separated the two fluorescence wavelengths. In spots emitting both wavelengths, fluorescence in the two channels fluctuated in opposite directions, indicating intermittent fluorescence energy transfer from a Cy3–EGF to a Cy5–EGF. Evidently pairs of chromophores can be imaged on the surface of living cells as they engage in fluorescence energy transfer with each other. The study shows that biochemistry can be done in living cells at the level of single molecules. Similar studies might directly reveal molecular insights that are not always possible with conventional biochemical methods. In the second study42, fibroblasts expressed a GFPconjugated pleckstrin homology domain of the Akt protein kinase (GFP–Akt-PH). Akt binds to 3-phosphorylated phosphoinositides (3-PPI), relatively rare lipids used by membranes for signalling and protein recruitment. Platelet-derived growth factor (PDGF) is known to increase turnover of 3-PPIs in fibroblasts and to initiate a directed migration to the source of the hormone. PDGF turned the surface of fibroblasts expressing GFP–Akt-PH bright green in minutes, indicating

Kurzchalia, T. V. & Parton, R. G. Membrane microdomains and caveolae. Curr. Opin. Cell Biol. 11, 424–431 (1999). Simons, K. & Toomre, D. Lipid rafts and signal transduction. Nature Rev. Mol. Cell Biol. 1, 31–39 (2000). Tsien, R. Y. The green fluorescent protein. Annu. Rev. Biochem. 67, 509–544 (1998). Thompson, N. L. & Lagerholm, C. B. Total internal reflection fluorescence: applications in cellular biophysics. Curr. Opin. Biotechnol. 8, 58–64 (1997). Burmeister, J. S., Olivier, L. A., Reichert, W. M. & Truskey, G. A. Application of total internal reflection fluorescence microscopy to study cell adhesion to biomaterials. Biomaterials 19, 307–325 (1998).

the capture of GFP–Akt-PH by freshly generated 3-PPI at the plasma membrane. The construct was used to image the local plasma membrane concentration of 3PPI. In a concentration gradient of PDGF, fluorescence (and hence 3-PPI local concentration) in any given cell was found highest on the side that faced the higher PDGF concentration. Future perspectives

The most recent wave of experiments with EF fluorescence has emphasized secretory vesicles but this reflects more the interests of the investigators than the potential of the method. The number of articles using the technique on living cells has all but doubled during each of the past three years. As the method becomes more available, it will be applied to other problems in cell biology. Perhaps the most promising development is the simultaneous EF imaging of two chromophores, each on a different molecule20,41. Pairs of genetically targetable chromophores, such as GFP and DSRED or CFP and YFP, will be particularly useful as they offer the opportunity to investigate protein–protein interactions on a nanometre scale. This will be particularly useful for studying reaction cascades at the plasma membrane with a spatially defined start or end point, such as exoor endocytosis, the budding of a caveola or the assembly/disassembly of a lipid raft. Labelling the contents of an exocytic vesicle with a red chromophore, for instance, defines both the time and the place of the exocytic event. Any protein of interest may be labelled green so that it can be determined precisely how long before exocytosis it is recruited, and how soon afterwards it is dismissed. Because many events may be observed and averaged after temporal and spatial alignment, the recruitment and dismissal of arbitrarily few molecules can be detected. Next, another protein can be labelled and followed in the same way. In time, we may learn for each known member of a reaction cascade when and how long they are present at the reaction site. Such information can be obtained at the level of single events and at sub-second time resolution, and will help significantly in determining the function of each participating protein. Links FURTHER INFORMATION Almers home page ENCYCLOPEDIA OF LIFE SCIENCES Light microscopy | Fluorescence microscopy | Synaptic vesicle traffic

Forkey, J. N., Quinlan, M. E. & Goldman, Y. E. Protein structural dynamics by single-molecule fluorescence polarization. Prog. Biophys. Mol. Biol. 74, 1–35 (2000). 7. Ambrose, W. P., Goodwin, P. M. & Nolan, J. P. Singlemolecule detection with total internal reflection excitation: comparing signal-to-background and total signals in different geometries. Cytometry 36, 224–231 (1999). 8. Jahn, R. & Südhof, T. C. Membrane fusion and exocytosis. Annu. Rev. Biochem. 68, 863–911 (1999). 9. Henkel, A. W. & Almers, W. Fast steps in exocytosis and endocytosis studied by capacitance measurements in endocrine cells. Curr. Opin. Neurobiol. 6, 350–357 (1996). 10. Neher, E. Vesicle pools and Ca2+ microdomains: new tools

11.

12. 13.

14.

for understanding their roles in neurotransmitter release. Neuron 20, 389–399 (1998). Lang, T. et al. Ca2+-triggered peptide secretion in single cells imaged with green fluorescent protein and evanescent-wave microscopy. Neuron 18, 857–863 (1997); erratum in 19, 463 (1997). Burke, N. V. et al. Neuronal peptide release is limited by secretory granule mobility. Neuron 19, 1095–1102 (1997). Breckenridge, L. J. & Almers, W. Final steps in exocytosis observed in a cell with giant secretory granules. Proc. Natl Acad. Sci. USA 84, 1945–1949 (1987). Whalley, T., Terasaki, M., Cho, M. S. & Vogel, S. S. Direct membrane retrieval into large vesicles after exocytosis in

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS sea urchin eggs. J. Cell Biol. 131, 1183–1192 (1995). 15. Steyer, J. A., Horstmann, H. & Almers, W. Transport, docking and exocytosis of single secretory granules in live chromaffin cells. Nature 388, 474–478 (1997). 16. Thomas, P., Wong, J. G., Lee, A. K. & Almers, W. A low affinity Ca2+ receptor controls the final steps in peptide secretion from pituitary melanotrophs. Neuron 11, 93–104 (1993). 17. Heinemann, C., Chow, R. H., Neher, E. & Zucker, R. S. Kinetics of the secretory response in bovine chromaffin cells following flash photolysis of caged Ca2+. Biophys J. 67, 2546–2557 (1994). 18. Oheim, M., Loerke, D., Stuhmer, W. & Chow, R. H. The last few milliseconds in the life of a secretory granule. Docking, dynamics and fusion visualized by total internal reflection fluorescence microscopy (TIRFM). Eur. Biophys. J. 27, 83–98 (1998). 19. Oheim, M. & Stuhmer, W. Tracking chromaffin granules on their way through the actin cortex. Eur. Biophys. J. 29, 67–89 (2000). 20. Tsuboi, T., Zhao, C., Terakawa, S. & Rutter, G. A. Simultaneous evanescent wave imaging of insulin vesicle membrane and cargo during a single exocytotic event. Curr. Biol. 10, 1307–1310 (2000). Describes simultaneous evanescent field fluorescence imaging with two colours. 21. Albillos, A. et al. The exocytotic event in chromaffin cells revealed by patch amperometry. Nature 389, 509–512 (1997). 22. Ales, E. et al. High calcium concentrations shift the mode of exocytosis to the kiss-and-run mechanism. Nature Cell Biol. 1, 40–44 (1999). 23. Wacker, I. et al. Microtubule-dependent transport of secretory vesicles visualized in real time with a GFP-tagged secretory protein. J. Cell Sci. 110, 1453–1463 (1997). 24. Lochner, J. E. et al. Real-time imaging of the axonal transport of granules containing a tissue plasminogen activator/green fluorescent protein hybrid. Mol. Biol. Cell 9, 2463–2476 (1998). 25. Angleson, J. K., Cochilla, A. J., Kilic, G., Nussinovitch, I. & Betz, W. J. Regulation of dense core release from neuroendocrine cells revealed by imaging single exocytic events. Nature Neurosci. 2, 440–446. (1999). 26. Schmoranzer, J., Goulian, M., Axelrod, D. & Simon, S. M. Imaging constitutive exocytosis with total internal reflection fluorescence microscopy. J. Cell Biol. 149, 23–32 (2000). 27. Toomre, D., Steyer, J. A., Keller, P., Almers, W. & Simons, K. Fusion of constitutive membrane traffic with the cell surface observed by evanescent wave microscopy. J. Cell Biol. 149, 33–40 (2000). References 26 and 27 image for the first time the exocytosis of constitutive secretory vesicles. 28. Fernandez, J. M., Neher, E. & Gomperts, B. D. Capacitance measurements reveal stepwise fusion events in degranulating mast cells. Nature 312, 453–455 (1984). 29. Spruce, A. E., Breckenridge, L. J., Lee, A. K. & Almers, W. Properties of the fusion pore that forms during exocytosis of a mast cell secretory vesicle. Neuron 4, 643–654 (1990). 30. von Gersdorff, H. & Matthews, G. Dynamics of synaptic

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

42.

43.

44.

45.

vesicle fusion and membrane retrieval in synaptic terminals. Nature 367, 735–739 (1994). Ryan, T. A., Reuter, H. & Smith, S. J. Optical detection of a quantal presynaptic membrane turnover. Nature 388, 478–482 (1997). Betz, W. J. & Bewick, G. S. Optical analysis of synaptic vesicle recycling at the frog neuromuscular junction. Science 255, 200–203 (1992). Zenisek, D., Steyer, J. A. & Almers, W. Transport, capture and exocytosis of single synaptic vesicles at active zones. Nature 406, 849–854 (2000). Describes results from imaging synaptic vesicles in a mature presynaptic terminal. Raviola, E. & Gilula, N. B. Intramembrane organization of specialized contacts in the outer plexiform layer of the retina. A freeze-fracture study in monkeys and rabbits. J. Cell Biol. 65, 192–222 (1975). Sutton, R. B., Fasshauer, D., Jahn, R. & Brunger, A. T. Crystal structure of a SNARE complex involved in synaptic exocytosis at 2.4 Å resolution. Nature 395, 347–353 (1998). Han, W., Ng, Y. K., Axelrod, D. & Levitan, E. S. Neuropeptide release by efficient recruitment of diffusing cytoplasmic secretory vesicles. Proc. Natl Acad. Sci. USA 96, 14577–14582 (1999). Lang, T. et al. Role of actin cortex in the subplasmalemmal transport of secretory granules in PC-12 cells. Biophys. J. 78, 2863–2877 (2000). Steyer, J. A. & Almers, W. in Imaging Neurons: A Laboratory Manual (eds Yuste, R., Lanni, F. & Konnerth, A.) 54.1–54.8 (Cold Spring Harbor Laboratory Press, Cold Spring Harbor, 2000). Muallem, S., Kwiatkowska, K., Xu, X. & Yin, H. L. Actin filament disassembly is a sufficient final trigger for exocytosis in nonexcitable cells. J. Cell Biol. 128, 589–598 (1995). Merrifield, C. J. et al. Endocytic vesicles move at the tips of actin tails in cultured mast cells. Nature Cell Biol. 1, 72–74 (1999). Sako, Y., Minoghchi, S. & Yanagida, T. Single-molecule imaging of EGFR signalling on the surface of living cells. Nature Cell Biol. 2, 168–172 (2000). Reports that two single molecules, each in a different colour, can be watched on a live cell as they engage in fluorescence resonance energy transfer. Haugh, J. M., Codazzi, F., Teruel, M. & Meyer, T. Spatial sensing in fibroblasts mediated by 3’ phosphoinositides. J. Cell Biol. 151, 1269–1280 (2000). Steyer, J. A. & Almers, W. Tracking single secretory granules in live chromaffin cells by evanescent-field fluorescence microscopy. Biophys. J. 76, 2262–2271 (1999). Rohrbach, A. Observing secretory granules with a multiangle evanescent wave microscope. Biophys. J. 78, 2641–2654 (2000). Axelrod, D., Hellen, E. H. & Fulbright, R. in Topics in Fluorescence Spectroscopy Vol. 3 (ed. Lakowicz, J. R.) 289–343 (Plenum, New York, 1992). An excellent and readable review covering most

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

46.

47.

48.

49.

50. 51.

52.

53.

54.

55. 56.

57.

58.

59.

60.

aspects of evanescent field fluorescence microscopy. Hellen, E. H. & Axelrod, D. Fluroescence emission at dieletric and metal-film interfaces. J. Opt. Soc. Am. 4, 337–350 (1987). Olveczky, B. P., Periasamy, N. & Verkman, A. S. Mapping fluorophore distributions in three dimensions by quantitative multiple angle-total internal reflection fluorescence microscopy. Biophys. J. 73, 2836–2847 (1997). Terakawa, S., Sakurai, T. & Abe, K. Development of an objective lens with a high numerical aperture for light microscopy. Bioimages 5, 24 (1997). Axelrod, D. Selective imaging of surface fluorescence with very high aperture microscope objectives. J. Biomed. Opt. 6, 6–13 (2001). Ambrose, E. J. The movements of fibrocytes. Exp. Cell Res. 8, 54–73 (1961). McCutchen, C. W. Optical systems for observing surface topography by frustrated total internal reflection and by interference. The Review of Scientific Instruments 35, 1340–1345 (1964). Axelrod, D. Cell–substrate contacts illuminated by total internal reflection fluorescence. J. Cell Biol. 89, 141–145 (1981). Lanni, F., Waggoner, A. S. & Taylor, D. L. Structural organization of interphase 3T3 fibroblasts studied by total internal reflection fluorescence microscopy. J. Cell Biol. 100, 1091–1102 (1985). Stout, A. L. & Axelrod, D. Evanescent field exitation of fluorescence by epi-illumination microscopy. Appl. Opt. 28, 5237–5242 (1989). Bryngdahl, O. in Progress in Optics (ed. Wolf, E.) 169–221 (North–Holland, Amsterdam, 1973). Gingell, D., Heavens, O. S. & Mellor, J. S. General electromagnetic theory of total internal reflection fluorescence: the quantitative basis for mapping cell–substratum topography. J. Cell Sci. 87, 677–693 (1987). Burmeister, J. S., Truskey, G. A. & Reichert, W. M. Quantitative analysis of variable-angle total internal reflection fluorescence microscopy (VA-TIRFM) of cell/substrate contacts. J. Microsc. 173, 39–51. (1994). Farinas, J., Simanek, V. & Verkman, A. S. Cell volume measured by total internal reflection microfluorimetry: application to water and solute transport in cells transfected with water channel homologs. Biophys. J. 68, 1613–1620 (1995). Omann, G. M. & Axelrod, D. Membrane-proximal calcium transients in stimulated neutrophils detected by total internal reflection fluorescence. Biophys. J. 71, 2885–2891 (1996). Cleemann, L., DiMassa, G. & Morad, M. Ca2+ sparks within 200 nm of the sarcolemma of rat ventricular cells: evidence from total internal reflection fluorescence microscopy. Adv. Exp. Med. Biol. 430, 57–65 (1997).

Acknowledgements We acknowledge support from the Max Planck Society (J.A.S.) and a grant from the NIH (W.A.).

VOLUME 2 | APRIL 2001 | 2 7 5

O N L I N E O N LY • Evanescent field (EF) fluorescence microscopy (also called total internal reflection fluorescence microscopy) is uniquely suited to image the plasma membrane with its associated organelles and macromolecules in living cells • Total internal reflection of a light beam generates an evanescent field — a thin layer of light that typically penetrates about 40–200 nm from a coverslip into an adhering cell. • EF fluorescence microscopy combines the specificity of confocal microscopy to detect fluorescent molecules with a depth discrimination of near-molecular dimensions. However, imaging is confined to the cell surface. • A recent wave of applications was aimed at membrane transport events in living cells. EF fluorescence microscopy offers new insights for its abilities to resolve even the smallest vesicles made by cells and to detect molecular-sized motions of fluorescent objects vertical to the glass. • In endocrine cells and synaptic nerve terminals, transport, exocytosis and replenishment of single secretory vesicles could be investigated. No method existed to record signals from single synaptic vesicles before the exocytic event. • EF fluorescence imaging of transport containers undergoing constitutive exocytosis provided a wealth of new results, including evidence for incomplete exocytosis. Other studies enhanced our understanding about the role of actin filaments for the movement of small organelles near the plasma membrane. • An important use of EF fluorescence microscopy will be to image signal-transduction events. Recent studies using EF fluorescence microscopy have shown that biochemistry can be done in living cells at the level of single molecules.

After studying physics in Mainz, Germany, Juergen A. Steyer moved into the field of cellular biophysics using two-photon microscopy during his yearlong stay at the AT&T Bell Laboratories in 1993/94. As a graduate student at the Max-Planck-Institute for Medical Research in Heidelberg, Germany, he explored, in the lab of Wolfhard Almers, the prospects of using evanescent field fluorescence microscopy to optically study transport and exocytosis of single vesicles in neuroendocrine cells. After receiving his Ph.D. in physics from the University of Heidelberg in 1997, he refined the method to image, in collaboration with David Zenisek, single synaptic vesicles and their fusion with the plasma membrane. Thereafter he continued to work on neurobiological applications of evanescent field fluorescence microscopy as a postdoctoral fellow at the Vollum Institute in Portland, Oregon and most recently, at the University of California at Berkeley. Wolfhard Almers received his Ph.D. from the University of Rochester, New York. He investigated excitation–contraction coupling in skeletal muscle, gating and permeability of ion channels and the mechanism of Ca2+ selectivity in Ca2+ channels, first at Cambridge, England and then at the University of Washington, Seattle, Washington. Since 1985, he focused on the mechanism of exocytosis in single cells and at the level of single secretory vesicles. In 1992, he became a Director at the Max Planck Institute for Medical Research in Heidelberg, Germany. He continued his interest in exoand endocytosis of single vesicles. Since 1999, he has been a Senior Scientist at the Vollum Institute, Oregon Health Sciences University, Portland, Oregon. ELS links Light microscopy http://www.els.net/elsonline/fr_article.jsp?id=A0002 634 Fluorescence microscopy http://www.els.net/elsonline/fr_article.jsp?id=A0002 637 Synaptic vesicle traffic http://www.els.net/elsonline/fr_loadarticle.jsp?available=1&ref=A0000215&orig=searching&page_number=1&page=search&Sitemap=exocytosis&searchtype =freetext&searchlevel=4

Wolhard Almers home page http://www.ohsu.edu/vollum/faculty/almers.htm

276

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

CELL SIGNALLING AT THE SHOOT MERISTEM Steven E. Clark The regulation of cell differentiation at meristems is crucial to developmental patterning in plants. Rapid progress has been made in identifying the genes that regulate differentiation and the receptor-mediated signalling events that have a key role in this process. In particular, we are now learning how the CLAVATA receptor kinase signalling pathway promotes stem cell differentiation in balance with the initiation of stem cells by the transcription factor WUSCHEL. MERISTEMS

Locations on a plant where stem cells are maintained and organogenesis occurs. Root meristems, shoot meristems and flower meristems fit this description. ORGAN PRIMORDIA

An organ (for example, a leaf, flower or petal) at an early stage of development, immediately after its initiation. ANGIOSPERMS

Most extant plants are angiosperms, or flowering plants. Non-angiosperms include gymnosperms (for example, pine and cycads), ferns and mosses.

Department of Biology, University of Michigan, Ann Arbor, Michigan 481091048, USA. e-mail: clarks@umich.edu

276

Plant development is fundamentally different from developmental patterning in most animals in that very little of the plant body plan is established during embryogenesis. Embryogenesis in higher plants establishes a very simple structure that contains two stemcell populations — the shoot meristem and the root meristem. Post-embryonic developmental patterning at these MERISTEMS is responsible for the morphology of the adult plant (FIG. 1). The shoot meristem is ultimately responsible for all of the ‘above-ground’ organs formed during the plant’s lifespan. In Arabidopsis thaliana, the shoot meristem initiates the leaves, flowers, vasculature and other tissues of the stem. The shoot meristem is able to form organs continuously by carefully balancing two activities. The first is the maintenance of undifferentiated stem cells at the very centre of the shoot meristem. The second is the direction of appropriately positioned progeny cells towards differentiation, so that they are competent to form ORGAN PRIMORDIA. The breakdown of either of these processes would be a morphological disaster for the plant, so the balance must be maintained even when variations in light, temperature or nutrient supply drive differences in growth and organ formation rates. In ANGIOSPERMS, the cells of the shoot meristem are found in three clonally distinct populations of cells called cell layers1,2 (FIG. 2). The outermost cell layer in Arabidopsis, referred to as the L1 layer, is the epidermal cell layer. Within the meristem, cells of this layer divide in a strictly ANTICLINAL fashion. As a result, the L1 cell layer in the meristem is one cell thick and remains so during organogenesis. Cells in the first subepidermal

layer, the L2, also divide in a largely anticlinal fashion, forming a completely separate population of cells from the other cell layers. Within developing organs, the L2 divides anticlinally and PERICLINALLY, but it still remains largely separate from the underlying L3 layer. The L3 layer is different, in that whereas the apical edge of the cell layer — the boundary between the L2 and L3 layers — is clearly defined, L3 cells frequently divide in various orientations. The net flow of cells is from the centre and the apex to the flanks and the basal regions of the shoot meristem (FIG. 2). This pattern of cell division indicates that cell signalling is required. First, a small number of stem cells give rise to all the differentiated cell types of the adult plant, ruling out any important role for cell lineage patterns in regulating cell fate. Second, the organs initiated on the flanks of the meristem are composed of cells from all three clonally distinct layers3,4, so these layers must communicate to execute organ formation in a coordinated fashion. Key events in the differentiation of cells at the shoot meristem include the commitment to differentiation, initiation of organ primordia and the establishment of polarities within each organ primordium. As a stem cell divides, leaving one daughter in the centre of the meristem and one daughter towards the flanks of the meristem, positional information must distinguish between these cells, such that the central daughter retains stem cell identity, and the peripheral daughter differentiates. Cells on the flanks of the meristem form either organ primordia or internodes. The PHYLLOTAXY of the individual plant species determines which peripheral cells of the meristem form organs1,2. In Arabidopsis, the sites of www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

Flower Stamen Petal Petal

Carpel

Stamen Sepal

Silique (fruit) Stem

Pith Cauline (stem) leaves Internode

Vascular bundle

Rosette leaves

Roots

Figure 1 | An adult Arabidopsis plant. A diagram of an adult Arabidopsis plant, showing organs such as roots, rosette leaves, cauline leaves and flowers. Internodes are the regions of differentiated tissue between successive organs. Shoot meristems are found at the tip of each active shoot, as well as in the axils of each leaf. Root meristems are found at the tip of the primary root and each lateral root.

ANTICLINAL

During anticlinal cell divisions, the new cell wall forms perpendicular to the layer of cells. This maintains cells in a single layer. PERICLINAL

In periclinal cell divisions, the new cell wall forms parallel to the cell layer, effectively thickening that cell layer. PHYLLOTAXY

The pattern of organ initiation by a shoot or flower meristem.

organ initiation form a spiral around the centre of the meristem (FIG. 3a). Once organs are initiated, the proximal/distal, medial/lateral and adaxial/abaxial (dorsal/ventral) asymmetries must be established within each organ primordium (FIG. 3b,c). This review focuses on our emerging understanding of the mechanism of these differentiation events, with particular emphasis on the signalling pathways that regulate meristem development and organ formation.

code for components of a signal-transduction pathway (TABLE 1). Plants mutant for any of the CLV loci progressively accumulate undifferentiated stem cells as development proceeds. Plants that are homozygous for strong loss-of-function alleles of CLV1 and CLV3 accumulate over 1,000-fold more undifferentiated cells than wildtype plants. Genetic analysis has revealed that these genes function in the same pathway6,7. CLV1 encodes a receptor kinase, with an extracellular domain composed of tandem leucine-rich repeats (LRRs)8. These LRRs are very similar in structure to several animal receptors, including thyroid-stimulating, luteinizing, and gonadotropin hormone receptors, although CLV1 contains more repeats (21 repeats) than the corresponding animal receptors (7–11 repeats)9. Early experiments established that the CLV1 kinase domain, when expressed in Escherichia coli, trans-phosphorylates multiple serine residues10,11. Although this is certainly consistent with the hypothesis that CLV1 acts as a receptor kinase, it far from establishes that CLV1 acts as a receptor, nor does it tell us anything about how CLV1 might interact with intracellular proteins. When purified from Arabidopsis, CLV1 is found in two protein complexes, one of ~185 kDa and a second of ~450 kDa (REF. 12). An attractive interpretation is that the 185-kDa complex contains inactive CLV1, and that the 450-kDa complex is composed of activated CLV1 that is associated with downstream signalling proteins. Genetic studies are consistent with this hypothesis. The clv1-1 mutant contains a missense mutation in the kinase-domain-coding region. This allele shows a partial loss-of-function phenotype, and the kinase domain has less than 50% of the autophosphorylation activity of wild type when expressed in E. coli. Similarly, the clv1-10 allele contains two missense mutations in the kinasedomain-coding region, shows a null or near-null phenotype, and has no autophosphorylation activity when expressed in E. coli. Extracts from clv1-1 plants have 50% less of the 450-kDa complex (and a corresponding increase in the accumulation of the 185-kDa complex), whereas clv1-10 extracts have no detectable 450-kDa complex. So, formation of the 450-kDa complex depends on the kinase activity of CLV1, and the 450kDa complex is very likely the active form of CLV1. CLV2 is similar to CLV1 in terms of the structure of the extracellular domain, although it is not very similar

Screens in Arabidopsis for mutants that lack stem cells or that accumulate ectopic stem cells have uncovered regulators of organogenesis. Many of the corresponding genes have been cloned, and a signal-transduction pathway that regulates stem-cell behaviour is beginning to emerge. The CLAVATA signalling pathway. The differentiation of stem cells in the shoot meristem is regulated by the CLAVATA genes (CLV1, CLV2, CLV3)5–7, which seem to

Stem cells

Leaf primordium

Stem-cell maintenance and differentiation

Leaf primordium Epidermis

L1 L2 L3

Subepidermis

Figure 2 | Cells layers and cell divisions. A longitudinal section through a shoot meristem, revealing the organization of the meristem into cell layers (L1, L2, L3). The location of the stem cells in each layer is indicated. The flow of cells as a result of cell growth and cell division is indicated with arrows. On the flanks of the meristem, cells form organ primordia, which become apparent (leaf primoridia) after rapid cell growth and division.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 7 7

REVIEWS

278

at the level of primary sequence13. However, CLV2 lacks a cytoplasmic domain, so how can it participate in CLV1 signalling? Genetic experiments established that CLV2 functions in the same pathway as CLV1 to regulate meristem development, and that CLV2 is required for the accumulation of CLV1 (REF. 13). This raises the possibility that CLV1 and CLV2 directly bind to each other. Consistent with this, the masses of CLV1 (105 kDa) and CLV2 (80 kDa) add up to 185 kDa, indicating that the inactive CLV1 complex might be a heterodimer of CLV1 and CLV2 (FIG. 4). CLV3 also acts in the same genetic pathway as CLV1, and encodes a small, probably secreted protein14, which is likely to act as the ligand for CLV1 (FIG. 4). In the absence of CLV3, CLV1 is detected only in the inactive, 185-kDa complex12, indicating that CLV3 is required for the formation of the active 450-kDa complex. Additional evidence that CLV3 acts upstream of CLV1 came from overexpression studies. Ectopic expression of CLV3 gives rise to the opposite phenotype to clv1 or clv3 mutants — a failure to maintain stem cells — and this phenotype depends on the presence of both CLV1 and CLV2 (REF. 15). In vivo biochemical and cell-culture experiments have shown that CLV3 is indeed the ligand for CLV1 (REF. 16). First, CLV3 and CLV1 immunoprecipitate together in vivo with the 450-kDa CLV1 complex, indicating that CLV3 binds only to active CLV1. Second, when intact yeast cells expressing CLV1 and CLV2 are incubated with plant extracts, CLV3 binds CLV1. Now that this ligand has been identified, it should be easier to manipulate CLV1 signalling in vivo and in cell culture.

CLV-interacting proteins. The CLV1 450-kDa complex presumably has several associated proteins, a subset of which are probably bound to phosphoserine residues in the kinase domain. The first protein identified that binds CLV1 was the kinase-associated protein phosphatase (KAPP)10,11. KAPP had originally been identified because it bound the kinase domain of an Arabidopsis LRR-containing receptor-like kinase, RLK5/HAE, in a phosphorylation-dependent manner17. KAPP contains three domains: a type I signal anchor, a kinase-interaction domain and a functional type 2C protein phosphatase domain. The kinase-interaction domain has been shown to contain a forkheadassociated (FHA) domain, which is a phosphothreonine/phosphoserine-binding domain18. KAPP binds directly to CLV1 in vitro10,11 and is a component of the 450-kDa CLV1 complex in vivo12. The presence of a phosphatase domain supports the hypothesis that KAPP negatively regulates CLV1 (FIG. 4). This was shown through two complimentary approaches. First, KAPP overexpression in wild-type plants recreates a weak clv– phenotype10. Second, when KAPP expression was suppressed in clv1-1 plants, the mutant phenotype was suppressed, depending on the level of KAPP suppression11. The second known component of the 450-kDa CLV1 complex is a Rho/Rac-GTPase-related protein. Although plants lack a protein that is clearly orthologous to the monomeric GTP-binding protein Ras, which carries out

Figure 3 | The Arabidopsis thaliana shoot meristem. A scanning electron micrograph of an Arabidopsis shoot meristem initiating leaf primordia is shown. The region containing stem cells is indicated. Organs (1–7) and incipient organ primordium (8) are numbered from oldest to youngest. Note the spiral arrangement of their initiation. b | Indicates the distal/proximal and the adaxial/abaxial axes. c | Indicates the lateral and medial axes.

4 1 7

Stem cells

8 5

Distal

Adaxial

Abaxial

Proximal

Lateral

Medial

many of the receptor-mediated signalling events for animal receptor tyrosine kinases (RTKs), they do have a family of small GTPases of the Ras superfamily, called ROP, which are similar to animal Rho/Rac proteins19,20. Using an antibody that cross-reacts with all of the Arabidopsis ROP isoforms tested, an appropriately sized cross-reacting protein was identified in the CLV1 450kDa complex12. So plants might have evolved a unique role for Rho proteins as a relay for RTK signalling. Identification of other interacting proteins, which should be possible using genetic screens combined with biochemical analysis, will allow us to dissect further the mechanisms of CLV1 signalling. For example, screens for suppressors of clv1 and clv3 alleles led to the identification of POLTERGEIST (POL), a gene that seems to function downstream as a negative regulator of the CLV pathway21. Characterization of the POL gene product and interacting proteins will certainly clarify the molecular events that relay the signal originated by CLV3. www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

EPISTATIC

Mutations that are epistatic mask the phenotype of other mutations.

The WUSCHEL/CLV3 feedback loop. The WUSCHEL (WUS) gene encodes a putative homeodomain-containing transcription factor. WUS is involved in initiating stem cells at the shoot meristem during embryogenesis, as wus mutants lack stem cells at shoot apices22,23. However, WUS is not expressed in stem cells, but in the cells underneath23. So, WUS either signals stem-cell fate to the overlying cells, or the loss of stem cells in wus mutants is an indirect consequence of the breakdown in meristem structure. The CLV proteins regulate the expression of WUS, at least indirectly. Genetic interactions revealed that wus mutations were largely EPISTATIC to clv mutations22, indicating that WUS might act downstream of CLV and that CLV negatively regulates WUS. A look at the expression of CLV1, CLV3 and WUS reveals that WUS is expressed in a basal and central subdomain of CLV1expressing cells (FIG. 5), whereas CLV3, encoding the ligand for CLV1, is expressed in cells adjacent to CLV1expressing cells that do not express WUS14,23. So CLV3 might be secreted from the apical cells and diffuse to the most apical and lateral CLV-expressing cells to repress

WUS expression. The hypothesis has been well supported in studies showing that WUS expression broadens both apically and laterally in clv mutants, indicating that the CLV pathway represses WUS in vivo, at least indirectly15,24 (FIG. 5). How could such a precise distinction between one cell and the next be achieved? CLV3 is expressed only two cell-lengths away from WUS-expressing cells. This suggests that there would need to be a mechanism to limit the range of diffusion of CLV3 to only those immediately adjacent cells. Over 75% of CLV3 is bound to CLV1, consistent with the idea that CLV1 titrates CLV3 from the soluble intercellular phase16, an effect known as ligand sequestration25. So, WUS-expressing cells might not detect a significant amount of CLV3 because much of it is sequestered by the overlying CLV1-expressing cells. Is WUS repression the only function for CLV1, and is WUS expression sufficient to establish stem cells? Both of these questions were addressed in experiments that tested the effects of misexpressing WUS within the meristem and organ primordia. WUS expression under

Table 1 | Genes involved in signalling at the meristem Gene

Type of protein

Function

Plant homologues

Animal homologues

CLV1

Receptor kinase

Promotes differentiation

Large gene family (>150 in Arabidopsis)

Kinase and LRR domains separately

CLV2

Receptor-like protein

Promotes differentiation, other functions

Large gene family (>40 in Arabidopsis)

LRR domains found in animal receptors

CLV3

Secreted ligand

Promotes differentiation

Putative secreted proteins in maize

WUS

Homeodomain transcription factor

Maintain and establish stem cells

Gene family

Yes

STM

Homeodomain transcription factor

Meristem identity, organ separation

Gene family

Yes

KAPP

FHA/phosphatase

Negatively regulates CLV1

Only for individual domains

FHA and protein phosphatase 2C domains separately

ROP

Rho/Rac-GTPase

Possible component of active CLV1 complex

Gene family (>11 in Arabidopsis)

Yes

PAN

B-zip transcription factor

Regulates organ number in flower

Large gene family (~80 in Arabidopsis)

Yes

CUC1

Unknown

Promotes organ separation

–

CUC2

NAC transcription factor

Promotes organ separation

Large gene family (>100 in Arabidopsis)

PHAN

MYB transcription factor

Establishes polarities in leaf primordia

Large gene family (~200 in Arabidopsis)

Yes

RS2

AS1 MYB transcription factor

Promotes differentiation of leaf primordia

PHAN homology (~7 in Arabidopsis)

Yes

AGO1

Novel

Organ polarity, other functions

Gene family

Yes

ZLL

Novel

Organ polarity, other functions

AGO homology

Yes

CRC

YABBY transcription factor

Polarity of carpels

Gene family (~6 in Arabidopsis)

Zinc finger and HMG domains separately

FIL

YABBY transcription factor

Polarity of organs, flower development

CRC homology

Zinc finger and HMG domains separately

REV

HD-zip class III

Polarity of organs, lateral meristems, vascular development

Gene family (~5 in Arabidopsis)

HD-zip and START domains separately

AGO, ARGONAUTE; CLV, CLAVATA; CRC, CRABS CLAW; FHA, forkhead-associated; FIL, FILAMENTOUS FLOWER; HD-zip, homeodomain plus leucine zipper; HMG, high mobility group; KAPP, kinase-associated protein phosphatase; LRR, Leucine-rich repeats; PAN, PERIANTHIA; PHAN, PHANTASTICA; REV, REVOLUTA; RS2, ROUGH SHEATH 2; START, StAR-related lipid transfer; WUS, WUSCHEL; ZLL, ZWILLE.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 7 9

REVIEWS from wus mutants could simply reflect the lack of stem cells in wus mutant plants. Distinguishing between these alternatives will require more subtle and inducible regulation and detection of gene expression.

X CLV3

S-S S-S

Continued differentiation

CLV2 P

KAPP

CLV1 P

P P

WUS

ROP

MAPKs? ? POL

Figure 4 | The CLAVATA1 signalling pathway. The CLV3 multimer binds to the extracellular domain of the putative CLV1/CLV2 heterodimer. Ligand binding drives CLV1 phosphorylation, which leads to the binding of the downstream effector molecules, kinase-associated protein phosphatase (KAPP) and ROP. KAPP is a negative regulator of CLV1, whereas the function of ROP is unknown. One possibility is that ROP acts through a mitogen-activated protein kinase cascade (MAPK) cascade to regulate WUSCHEL (WUS) expression. POLTERGEIST (POL) is another negative regulator of CLV1 signalling that functions downstream of CLV1 and in close association with WUS. P, phosphate.

the control of the CLV1 regulatory elements recreates the WUS expression seen in clv1 mutants, and mimicks the clv– phenotype24. So the accumulation of stem cells in clv1 mutants seems to result largely from WUS misexpression. ANT is expressed in nascent organ primordia, and WUS expressed under the control of ANT regulatory elements in incipient organs prevents their differentiation. Plants with WUS expression driven by the ANT promoter often form only a large mass of stem cells at their apex. These elegant experiments have revealed that WUS is sufficient to establish stem-cell fate in adjacent cells. How this occurs remains a mystery. Several observations point to the existence of a feedback loop between WUS and CLV3 that might aid in maintaining a stable population of stem cells: first, WUS and CLV3 expression is broader in clv mutants than in wild-type plants; second, CLV3 expression is downregulated in wus mutants; and last, CLV3 expression is broader in plants that overexpress WUS15,24. An attractive possibility is that CLV3 negatively regulates WUS expression, and WUS activates CLV3 expression. Thus, a downregulation of WUS would lead to a downregulation of CLV3, which in turn would lead to an upregulation of WUS. Such a system would move to an equilibrium point at which the expression of CLV3 and WUS would be stable. Imbalances in expression of one gene or the other would tend to return to equilibrium. One could even imagine that, in other species with larger meristems, the equilibrium point is shifted by altering the parameters of WUS and CLV3 interaction. However, the evidence for this hypothesis could also be interpreted as the indirect consequences of changes in cell identity. For example, the loss of CLV3 expression

280

The loss of stem-cell identity by cells on the flanks of the meristem is just the first step in the long road towards differentiation. We must also keep in mind that this is not a one-way street: some cells regain stem-cell status if they are incorporated into lateral shoot and flower meristems. Although all the cells on the flanks of the meristem are competent to form organ primordia, only some do; others form the internodes between organs. So how do the cells that make up organs acquire the correct proximal/distal, lateral/medial and adaxial/abaxial asymmetries? Organ initiation. Organs in higher plants are initiated in distinct patterns1,2. The Arabidopsis phyllotaxy is spiral, with a defined angle of 137° between each subsequent organ (FIG. 3a). Other species initiate organs in rings or in alternate or opposite patterns. The pattern of organ formation can vary between shoot and flower meristems (BOX 1), and between shoot meristems at different stages of development. The mechanisms that regulate the phyllotaxy of organogenesis has fascinated plant biologists for centuries, but they remain a mystery. Two well-discussed hypotheses invoke inhibitory signals from recently initiated organ primordia, and the role of biophysical stresses in marking the site of buckling and hence organ formation, respectively. Theorectical models of each can explain the various patterns of organ formation observed in nature1,2,26. But, so far, little experimental evidence is available to support any of the hypotheses. One recent study showed that nearly all cells on the flanks of the meristem might be competent to form organs. Exogenous application of an inhibitor that blocks transport of the plant hormone auxin to tomato a

CLV3

CLV1

WUS

Figure 5 | Stem-cell regulators. a | The approximate domains of mRNA expression for CLAVATA1 (CLV1), CLV3 and WUSCHEL (WUS) are shown. Note that WUS is expressed in the cells immediately underlying the stem cells. The function of CLV1 and CLV3 is to repress WUS, as implied by the expansion of WUS expression and stem cells in clv mutant plants. b | Expression of WUS in clv mutants. Arrows indicate that WUS expression expands both apically and laterally.

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

Box 1 | Flowers and shoots — functional and evolutionary relationships There are only two sites within a plant where distinct lateral organs can be initiated: the shoot meristem and the flower meristem. At each location, a population of centrally located stem cells give rise to more differentiated progeny on the flanks of the meristem that are then incorporated into organ primordia. The flower meristem is probably a modified shoot meristem, as indicated by similarities in structure and function, the fact that shoots evolved long before flowers, and the fact that flower organs seem to be modified leaves. More recent molecular genetic work on the development of Arabidopsis and other species have confirmed this. LEAFY and APETALA 1 in Arabidopsis are two genes that are involved in specifying flower meristems. Mutation of both genes converts the flower meristems into shoot meristems57–59. Conversely, mutations in the TERMINAL FLOWER (TFL) gene convert shoot meristems into flower meristems60,61. Given that the shoot and flower meristems are functionally very similar, genes that regulate a fundamental aspect of meristem function (for example, stem cell maintenance, differentiation and organ formation) might be expected to have similar functions in shoot and flower meristem development. Indeed, for the genes that are the focus of this review (TABLE 1), mutations result in very similar phenotypes within the shoot and flower meristems. For example, in clv1 mutants, both shoot and flower meristems accumulate stem cells5. However, many differences are found between shoot and flower meristems. The positioning of organs often varies between the two structures. For example, the Arabidopsis shoot meristem initiates organs in a spiral pattern, whereas the flower meristem initiates organs in a ring pattern. Whereas the Arabidopsis shoot meristem always maintains stem cells, flower meristems inevitably terminate in differentiated organs. Finally, the identity of the organs that are initiated differ widely between shoot and flower meristems.

b ab

apices, or use of an auxin mutant in Arabidopsis, resulted in shoot meristems that lack organs on the flanks without affecting differentiation27. An organ formed wherever a drop of auxin was placed on the meristem flanks, indicating that all the cells on the flanks were at least competent to form organ primordia. No mutants that specifically alter shoot meristem phyllotaxy have been identified so far in Arabidopsis. All the mutants that alter the pattern of organ initiation also notably alter the entire structure of the meristem, implying that the two processes might be inextricably linked. The only exception to this is the PERIANTHIA (PAN) gene that functions within the flower meristem. pan mutants alter the number of organs initiated in each ring of organs within the flower meristem, without altering the fundamental structure of the flower meristem28. Understanding how PAN functions might provide the first clear insight into how phyllotaxy is established.

Developing leaf (transverse section)

Leaf primordium (longitudinal section) ab ad

Apical meristem

MOSAIC ORGANS

Organs that have cell types characterisitic of two or more organs. FUSED ORGANS

Organs fused together at early stages of development, usually at the lateral edges.

Figure 6 | Establishment of organ polarities within organ primordia. a | Longitudinal section through the vegetative apex of an Arabidopsis thaliana shoot meristem. The expression of the developmental receptor FILAMENTOUS FLOWER (FIL) is revealed in longitudinal and transverse sections through developing leaf primordia. FIL is expressed specifically within the abaxial region of leaf primordia. b | Sketch of a indicating the polarity axes. (Ab, abaxial; ad, adaxial.) a taken with permission from REF. 41 (1999) © Company of Biologists Ltd.

Organ separation. Three genes involved in separating organs at the earliest stages in organ initiation have been identified. Plants mutant for STM, CUC1 or CUC2 result in the fusion of the earliest organs initiated — the cotyledons29–31. In addition, these genes also seem to have a role in separation of all the organs that are initiated by the shoot meristem. STM, although it is expressed in a central region of the meristem and is required for meristem maintenance, is also expressed between the early forming organ primordia31,32. Indeed, weak alleles of stm that form transient shoot meristems show a great deal of MOSAIC and FUSED ORGANS33,34. cuc1 cuc2 plants also give rise to postembryonic organ fusion30. CUC2 encodes a putative plant-specific transcription factor that is expressed between organs. The expression of these genes at the boundaries of organs could prevent these cells from being incorporated into the organ primordia, or could simply inhibit their growth.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 8 1

REVIEWS

Box 2 | Advantages of plants as a model system for studying signalling Studies of animals and yeast signal transduction have dominated the signalling field since its inception. However, the rapid development of modern molecular genetic approaches in plants, especially Arabidopsis thaliana, allows us to make some comparisons of the relative strengths of studying signal transduction in plants and animals. A curious consequence of the sequencing of the Arabidopsis genome was the discovery of nearly 200 receptor-like kinases that are similar to CLV1. This large diversification of receptors in a relatively simple organism might mean that plant receptors are much more specialized than in animals. Indeed, the available evidence on developmental receptors in plants indicates that many might regulate a single developmental process: CLV1 promotes differentiation; RLK5/HAESA promotes ABSCISSION62; ERECTA regulates organ length63; S-receptor kinase mediates pollen/pistil recognition64; and PRK1 regulates gamete development65. This has many important consequences for experiments. First, plants with null mutations in these receptors are generally viable. This allows simple genetic screening for interacting factors. Second, the process that is regulated by the receptor is usually a process that continues for much of the lifespan of the plant. For example, all growing plants initiate organs and use CLV1 signalling. These features combine to provide plants with their greatest advantage as an experimental system for studying signalling — the ability to characterize receptor function through in vivo biochemistry. Compare this with animals in which key receptors have a role in a plethora of developmental processes and act during very specific developmental stages. This means that examining the in vivo status of protein–protein interaction in a specific organ at a specific stage of development is technically difficult to say the least. Animal developmental biologists have, of course, developed powerful genetic and cell-culture systems in which to study signalling. However, the ability to study signalling in vivo in plants might well lead to the characterization of new features that are common to both plants and animals.

ABSCISSION

The process by which dead parts of a plant break off naturally (for example, leaves). MYB TRANSCRIPTION FACTOR

A type of transcription factor first identified in animals. The MYB gene family is greatly expanded in plants, and the proteins that these encode have been shown to control many developmental processes. ZYGOMORPHIC FLOWERS

Flowers that are asymmetric, and in which the development of specific organs varies depending on the polarities of the flowers. Snapdragon flowers are zygomorphic, whereas roses are not.

282

Organ polarity. The asymmetric growth at the earliest stages of organ development indicates that organs might rapidly acquire apical/basal, lateral/medial and adaxial/abaxial polarities. The field has progressed rapidly with the isolation of several key genes. One of the first polarity specification genes was PHANTASTICA (PHAN) from Antirrhinum majus (snapdragon). This predicted MYB TRANSCRIPTION FACTOR is proposed to establish adaxial/abaxial polarity in developing leaves35. An interesting hypothesis resulting from PHAN analysis was that the outgrowth of leaf blades might be promoted by the juxtaposition of abaxial and adaxial domains at the edges of young leaf primordia. This hypothesis was based on the frequent absence or ectopic formation of blade outgrowths in the leaves of phan mutant plants, thought to result from an absence or incomplete establishment of abaxial/adaxial polarity. Mutations in a homologue of PHAN from maize, ROUGH SHEATH 2 (RS2), also led to leaf defects, although these have been interpreted as a breakdown in proximal/distal polarity36. This difference might arise from the different architecture of leaves in Antirrhinum and maize. The identification of an Arabidopsis orthologue of PHAN, ASYMMETRIC LEAVES 1 (AS1), provided evidence linking organ differentiation to meristem function37,38. The as1 mutation disrupted leaf morphology, which had also been shown for mutations in PHAN and RS2. Interestingly, as1 also suppressed the lack of stem cells in stm mutants37. STM, and its homologue in maize KNOTTED 1 (KN1), had been shown to be necessary for stem-cell maintenance at the shoot meristem29,39. How STM carried out this activity was unclear, because stm mutations showed additive interactions with wus and clv mutations, indicating that the genes might function in parallel pathways22,33,34. The observations that as1 suppresses the stm meristem defect and that STM is expressed normally in the as1 mutant38, indicates that STM carries out stem-cell

maintenance by inhibiting the expression of AS1 in the meristem. Indeed, in stm mutant embryos, AS1 is expressed in the position that the shoot meristem would normally occupy37. So stem-cell maintenance requires two activities: WUS to promote stem-cell identity, and STM to inhibit differentiation. Later specification of adaxial/abaxial polarity seems to involve several genes in Arabidopsis. Several members of the YABBY family of putative transcription factors, such as CRABS CLAW (CRC) and FILAMENTOUS FLOWER (FIL) are also expressed in the abaxial portion of several organ types, including leaves and carpels40–42 (FIG. 6). As predicted by their expression patterns, mutant analysis of several YABBY genes indicates that these genes establish abaxial fate. Conversely, REVOLUTA (REV) is expressed in the adaxial region of each primordia43. The ZWILLE/PINHEAD (ZLL)44,45 and ARGONAUTE 1 (AGO1) genes seem to specify adaxial fate as well, and ZLL is expressed in the adaxial portion of leaf primordia. AGO1 has recently been implicated in the process of post-transcriptional gene silencing or RNA interference46,47. The PHABULOSA (PHB) gene, which has not yet been cloned, seems to promote adaxial fate48. Adaxial/abaxial polarity within the flower primordia is necessary for the development of ZYGOMORPHIC FLOWERS. Many flowers develop strikingly asymmetric floral organs, especially petals, that are essential for complex interactions with pollinators. In Antirrhinum, this asymmetry is established by the position of the flower primordia relative to the meristem, and requires several putative transcription factors49. Perspectives

Our understanding of differentiation and organ formation has progressed rapidly over the past few years. Several of the key regulatory elements that determine the fate of stem cells within the shoot meristem have been identified. And much more is now known about the function of receptor kinases in www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS plant signalling15,16,50–53. In most cases, the investigators benefit from the ability to carry out in vivo biochemical experiments in a genetic system. The combination of in vivo biochemistry and genetics is a great advantage for studying signal transduction in plants (BOX 2). But, as always, the results have created as many questions as answers, many of which concern the initial steps towards differentiation. How is CLV1 signalling relayed within the cell? How does WUS signal to the overlying cells to establish stem-cell fate? Which genes are expressed within stem cells to maintain their identity? Curiously, screens for single mutants that specifically affect meristem development have failed to identify many of these factors. Attention must therefore turn to mutations with pleiotropic effects54–56, and to designing genetically sensitized screens for mutations that do not lead to a phenotype on their own21. Only such a comprehensive approach will complete the cast of characters that regulate cell differentiation. Later differentiation events, touched on briefly in this review, are also being vigorously addressed by many labs. Exciting progress has been made on factors that

1. 2.

10.

11.

12.

13.

Steeves, T. A. & Sussex, I. M. Patterns in Plant Development (Cambridge Univ. Press, New York, 1989). Lyndon, R. F. The Shoot Apical Meristem: Its Growth and Development 1–277 (Cambridge Univ. Press, Cambridge, 1998). Satina, S., Blakeslee, A. F. & Avery, A. G. Demonstration of the three germ layers in the shoot apex of Datura by means of induced polyploidy in periclinal chimeras. Am. J. Bot. 27, 895–905 (1940). Jenik, P. D. & Irish, V. F. Regulation of cell proliferation patterns by homeotic genes during Arabidopsis floral development. Development 127, 1267–1276 (2000). Clark, S. E., Running, M. P. & Meyerowitz, E. M. CLAVATA1, a regulator of meristem and flower development in Arabidopsis. Development 119, 397–418 (1993). Clark, S. E., Running, M. P. & Meyerowitz, E. M. CLAVATA3 is a specific regulator of shoot and floral meristem development affecting the same processes as CLAVATA1. Development 121, 2057–2067 (1995). Kayes, J. M. & Clark, S. E. CLAVATA2, a regulator of meristem and organ development in Arabidopsis. Development 125, 3843–3851 (1998). Clark, S. E., Williams, R. W. & Meyerowitz, E. M. The CLAVATA1 gene encodes a putative receptor kinase that controls shoot and floral meristem size in Arabidopsis. Cell 89, 575–585 (1997). Grossman, M., Weintraub, B. D. & Szkudlinski, M. W. Novel insights into the molecular mechanisms of human thyrotropin action: structural, physiological, and therapeutic implications for the glycoprotein hormone family. Endocr. Rev. 18, 476–501 (1997). Williams, R. W., Wilson, J. M. & Meyerowitz, E. M. A possible role for kinase-associated protein phosphatase in the Arabidopsis CLAVATA1 signaling pathway. Proc. Natl Acad. Sci. USA 94, 10467–10472 (1997). Stone, J. M., Trotochaud, A. E., Walker, J. C. & Clark, S. E. Control of meristem development by CLAVATA1 receptor kinase and kinase-associated protein phosphatase interactions. Plant Physiol. 117, 1217–1235 (1998). Trotochaud, A. E., Hao, T., Guang, W., Yang, Z. & Clark, S. E. The CLAVATA1 receptor-like kinase requires CLAVATA3 for its assembly into a signaling complex that includes KAPP and a Rho-related protein. Plant Cell 11, 393–405 (1999). Characterization of CLV1 receptor protein complexes in wild-type and mutant plants, revealing inactive and active CLV1 complexes, the association of a putative Rho/Rac-GTPase, and the requirement for CLV3 in receptor activation. Jeong, S., Trotochaud, A. E. & Clark, S. E. The Arabidopsis CLAVATA2 gene encodes a receptor-like protein required for the stability of the CLAVATA1 receptorlike kinase. Plant Cell 11, 1925–1933 (1999).

regulate polarity. However, the interaction between the various factors or the type of genes that these potential transcription factors target has not been sorted out. Nor has an understanding been developed of the signals that establish the earliest polarity events. Although at an early stage of investigation, it is clear that many of the genes already identified will provide excellent starting points for understanding such processes.

Links DATABASE LINKS CLV1 | CLV2 | CLV3 | thyroidstimulating hormone receptor | luteinizing hormone receptor | gonadotropin hormone receptor | RLK5 | Rho/Rac | POL | WUS | auxin | PAN | STM | AS1 | CRC | FIL | REV | ZLL | AGO1 | PHB | LEAFY | APETALAI 1 | TFL | ERECTA FURTHER INFORMATION Clark lab home page ENCYCLOPEDIA OF LIFE SCIENCES Arabidopsis thaliana as an experimental organism | Arabidopsis: Flower development and patterning | Positional information in plant development

14. Fletcher, J. C., Brand, U., Running, M. P., Simon, R. & Meyerowitz, E. M. Signaling of cell fate decisions by CLAVATA3 in Arabidopsis shoot meristems. Science 283, 1911–1914 (1999). 15. Brand, U., Fletcher, J. C., Hobe, M., Meyerowitz, E. M. & Simon, R. Dependence of stem cell fate in Arabidopsis on a feedback loop regulated by CLV3 activity. Science 289, 617–619 (2000). Overexpression of CLV3 reveals a regulatory feedback loop between CLV3 and WUS, providing a possible mechanism for the regulation of a stable stem-cell population. 16. Trotochaud, A. E., Jeong, S. & Clark, S. E. CLAVATA3, a mulitmeric ligand for the CLAVATA1 receptor-kinase. Science 289, 613–617 (2000). 17. Stone, J. M., Collinge, M. A., Smith, R. D., Horn, M. A. & Walker, J. C. Interaction of a protein phosphatase with an Arabidopsis serine/threonine receptor kinase. Science 266, 793–795 (1994). 18. Li, J., Smith, G. P. & Walker, J. C. Kinase interaction domain of kinase-associated protein phosphatase, a phosphoprotein-binding domain. Proc. Natl Acad. Sci USA 96, 7821–7826 (1999). 19. Li, H., Wu, G., Ware, D., Davis, K. R. & Yang, Z. Arabidopsis Rop GTPases: Differential gene expression in pollen and polar localization in fission yeast. Plant Physiol. 118, 407–417 (1998). 20. Winge, P., Brembu, T., Kristensen, R. & Bones, A. M. Genetic structure and evolution of RAC-GTPases in Arabdopsis thaliana. Genetics 156, 1959–1971 (2000). 21. Yu, L. P., Simon, E. J., Trotochaud, A. E. & Clark, S. E. POLTERGEIST functions to regulate meristem development downstream of the CLAVATA loci. Development 127, 1661–1670 (2000). 22. Laux, T., Mayer, K. F. X., Berger, J. & Jurgens, G. The WUSCHEL gene is required for shoot and floral meristem integrity in Arabidopsis. Development 122, 87–96 (1996). 23. Mayer, K. L. et al. Role of WUSCHEL in regulating stem cell fate in the Arabidopsis shoot meristem. Cell 95, 805–815 (1998). 24. Schoof, H. et al. The stem cell population of Arabidopsis shoot meristems is maintained by a regualtory loop between the CLAVATA and WUSCHEL genes. Cell 100, 635–644 (2000). A superlative demonstration that the CLV loci regulate WUS gene expression, and that WUS is sufficient to establish stem-cell identity. 25. Chen, Y. & Struhl, G. Dual roles for patched in sequestering and transducing hedgehog. Cell 87, 553–563 (1996). 26. Green, P. B. Expression of pattern in plants: combining molecular and calculus-based biophysical paradigms. Am. J. Bot. 86, 1059–1076 (1999).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

27. Reinhardt, D., Mandel, T. & Kuhlemeier, C. Auxin regulates the initiation and radial position of plant lateral organs. Plant Cell 12, 507–518 (2000). A creative look at the role of auxin in organ formation, revealing a uniform compentency for organ formation among cells on the periphery of the mersitem. 28. Running, M. P. & Meyerowitz, E. M. Mutations in the PERIANTHIA gene of Arabidopsis specifically alter floral organ number and initiation pattern. Development 122, 1261–1269 (1996). The Arabidopsis PAN gene is the only gene so far that specifically affects the phyllotaxy of organ initiation. This paper is a careful phenotypic and genetic analysis of pan mutants. 29. Barton, M. K. & Poethig, R. S. Formation of the shoot apical meristem in Arabidopsis thaliana: an analysis of development in the wild type and shoot meristemless mutant. Development 119, 823–831 (1993). 30. Aida, M., Ishida, T., Fukaki, H., Fujisawa, H. & Tasaka, M. Genes involved in organ separation in Arabidopsis — an analysis of the cup-shaped cotyledon mutant. Plant Cell 9, 841–857 (1997). 31. Long, J. A. & Barton, M. K. The development of apical embryonic pattern in Arabidopsis. Development 125, 3027–3035 (1998). A carefully done analysis of the sequential establishment of cell fates and meristem identity in the apical domain of the Arabidopsis embryo. 32. Long, J. & Barton, M. K. Initiation of axillary and floral meristems in Arabidopsis. Dev. Biol. 218, 341–353 (2000). 33. Clark, S. E., Jacobsen, S. E., Levin, J. & Meyerowitz, E. M. The CLAVATA and SHOOT MERISTEMLESS loci competitively regulate meristem activity in Arabidopsis. Development 122, 1567–1575 (1996). 34. Endrizzi, K., Moussain, B., Haecker, A., Levin, J. Z. & Laux, T. The SHOOT MERISTEMLESS gene is required for maintenance of undifferentiated cells in Arabidopsis shoot and floral meristems and acts at a different regulatory level than the meristem genes WUSCHEL and ZWILLE. Plant J. 10, 967–979 (1996). 35. Waites, R., Selvadurai, H. R., Oliver, I. R. & Hudson, A. The PHANTASTICA gene encodes a MYB transcription factor involved in growth and dorsoventrality of lateral organs in Antirrhinum. Cell 93, 779–789 (1998). In this groundbreaking paper, the PHAN gene was shown to regulate organ polarity in snapdragon. PHAN orthlogues were later isolated from many species, including maize and Arabidopsis, and shown to retain conserved developmental functions. 36. Tsiantis, M., Schneeberger, R., Golz, J. F., Freeling, M. & Langdale, J. A. The maize rough sheath2 gene and leaf development programs in monocot and dicot plants.

VOLUME 2 | APRIL 2001 | 2 8 3

REVIEWS Science 284, 154–156 (1999). 37. Byrne, M. E. et al. Asymmetric leaves1 mediates leaf patterning and stem cell function in Arabidopsis. Nature 408, 967–971 (2000). An important analysis showing both the role of AS1 in leaf development, and the link between STM function and differentiation. 38. Ori, N., Eshed, Y., Chuck, G., Bowman, J. L. & Hake, S. Mechanisms that control knox gene expression in the Arabidopsis shoot. Development 127, 5523–5532 (2000). 39. Kerstetter, R. A., Laudencia-Chingcuanco, D., Smith, L. G. & Hake, S. Loss-of-function mutations in the maize homeobox gene, knotted1, are defective in shoot meristem maintenance. Development 124, 3045–3054 (1997). 40. Bowman, J. L. & Smyth, D. R. CRABS CLAW, a gene that regualtes carpel and nectary development in Arabidopsis, encodes a novel protein with zinc fingers and helix-loophelix domains. Development 126, 2387–2396 (1999). 41. Siegfried, K. R. et al. Members of the YABBY gene family specify abaxial cell fate in Arabidopsis. Development 126, 4117–4128 (1999). An excellent look at several key genes involved in specifying organ polarity in Arabidopsis. 42. Villanueva, J. M. et al. INNER NO OUTER regulates abaxial–adaxial patterning in Arabidopsis ovules. Genes Dev. 13, 3160–3169 (1999). 43. Ostuga, D., DeGuzman, B., Prigge, M., Drews, G. N. & Clark, S. E. REVOLUTA regulates meristem initiation at lateral positions. Plant J. 25, 223–236 (2001). 44. Moussian, B., Schoof, H., Haecker, A., Jürgens, G. & Laux, T. Role of the ZWILLE gene in the regulation of central shoot meristem cell fate during Arabidopsis embryogenesis. EMBO J. 17, 1799–1809 (1998). 45. Lynn, K. et al. The PINHEAD/ZWILLE gene acts pleiotropically in Arabidopsis development and has overlapping functions with the ARGONAUTE1 gene. Development 126, 469–481 (1999).

284

46. Fagard, M., Bouter, S., Morel, J. B., Bellini, C. & Vaucheret, H. AGO1, QDE-2 and RDE-1 are related proteins required for post-transcriptional gene silencing in plants, quelling in fungi, and RNA interference in animals. Proc. Natl Acad. Sci. USA 97, 11650–11654 (2000). 47. Bernstein, E., Caudy, A. A., Hammond, S. M. & Hannon, G. J. Role for bidentate ribonulcease in the initiation step of RNA interference. Nature 409, 363–366 (2001). 48. McConnell, J. R. & Barton, M. K. Leaf polarity and meristem formation in Arabidopsis. Development 125, 2935–2942 (1998). 49. Luo, D. et al. Control of organ asymmetry in flowers of Antirrhinum. Cell 99, 367–376 (1999). 50. He, Z. et al. Perception of brassinosteroids by the extracellular domain of the receptor kinase BRI1. Science 288, 2360–2363 (2000). 51. Schopfer, C. R., Nasrallah, M. E. & Nasrallah, J. B. The male determinant of self-incompatibility in Brassica. Science 286, 1697–1700 (1999). 52. Takasaki, T. et al. The S receptor kinase determines selfincompatibility in Brassica stigma. Nature 403, 913–916 (2000). 53. Giranton, J. L., Dumas, C., Cock, J. M. & Gaude, T. The integral membrane S-locus receptor kinase of Brassica has serine/threonine kinase activity in a membranous environment and spontaneously forms oligomers in Planta. Proc. Natl Acad. Sci. USA 0, 3759–3764 (2000). 54. Leyser, H. M. O. & Furner, I. J. Characterization of three shoot apical meristem mutants of Arabidopsis thaliana. Development 116, 397–403 (1992). 55. Callos, J. D. et al. The FOREVER YOUNG gene encodes an oxidoreductase required for proper development of the Arabidopsis vegetative shoot apex. Plant J. 6, 835–847 (1994). 56. Laufs, P., Dockx, J., Kronenberger, J. & Traas, J. MGOUN1 and MGOUN2-2 genes required for primordium initiation at the shoot apical and floral meristems in Arabidopsis thaliana. Development 125,

1253–1260 (1998). 57. Weigel, D., Alvarez, J., Smyth, D. R., Yanofsky, M. F. & Meyerowitz, E. M. LEAFY controls floral meristem identity in Arabidopsis. Cell 69, 843–859 (1992). 58. Bowman, J. L., Alvarez, J., Weigel, D., Meyerowitz, E. M. & Smyth, D. R. Control of flower development in Arabidopsis thaliana by APETALA1 and interacting genes. Development 119, 721–743 (1993). 59. Shannon, S. & Meeks-Wagner, D. R. Genetic interactions that regulate inflorescence development in Arabidopsis. Plant Cell 5, 639–655 (1993). 60. Shannon, S. & Meeks-Wagner, D. R. A mutation in the Arabidopsis TFL1 gene affects inflorescence meristem development. Plant Cell 3, 877–892 (1991). 61. Alvarez, J., Guli, C. L., Yu, X.-H. & Smyth, D. R. terminal flower: a gene affecting inflorescence development in Arabidopsis thaliana. Plant J. 2, 103–116 (1992). 62. Jinn, T.-S., Stone, J. M. & Walker, J. C. HAESA, an Arabidopsis leucine-rich repeat receptor kinase, controls floral organ abscission. Genes Dev. 14, 108–117 (2000). 63. Torii, K. U. et al. The Arabidopsis ERECTA gene encodes a putative receptor protein kinase with extracellular leucinerich repeats. Plant Cell 8, 735–746 (1996). 64. Stein, J. C., Howlett, B., Boyes, D. C., Nasrallah, M. E. & Nasrallah, J. B. Molecular cloning of a putative receptor protein kinase gene encoded at the self-incompatibility locus of Brassica oleracea. Proc. Natl Acad. Sci. USA 88, 8816–8820 (1991). 65. Mu, J.-H., Lee, H.-S. & Kao, T.-H. Characterization of a pollen-expressed receptor-like kinase gene of Petunia inflata and the activity of its encoded kinase. Plant Cell 6, 709–721 (1994).

Acknowledgements Work in the author’s lab on meristem development is supported by grants from the NSF Developmental Mechanisms programme and the DOE Energy Biosciences programme. The author thanks John Bowman for providing the photograph used in FIG. 6.

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

MULTIFUNCTIONAL STRANDS IN TIGHT JUNCTIONS Shoichiro Tsukita, Mikio Furuse and Masahiko Itoh Tight junctions are one mode of cell–cell adhesion in epithelial and endothelial cellular sheets. They act as a primary barrier to the diffusion of solutes through the intercellular space, create a boundary between the apical and the basolateral plasma membrane domains, and recruit various cytoskeletal as well as signalling molecules at their cytoplasmic surface. New insights into the molecular architecture of tight junctions allow us to now discuss the structure and functions of this unique cell–cell adhesion apparatus in molecular terms.

EPITHELIAL CELLS

Closely packed cells, arranged in one or more layers, that cover the outer surfaces of the body or line any internal cavities or tubes (except the blood vessels, heart and serous cavities). ENDOTHELIAL CELLS

Thin, flattened cells of mesoblastic origin that are arranged in a single layer lining the blood vessels and some body cavities, for example those of the heart. MESOTHELIAL CELLS

Flat cells derived from mesoderm that are arranged in a single layer, found lining some body cavities.

Department of Cell Biology, Kyoto University Faculty of Medicine, Yoshida-Konoe, Sakyo-ku, Kyoto 606-8501, Japan. Correspondence to S.T. e-mail: htsukita@mfour.med.kyoto -u.ac.jp

The existence of separate fluid compartments with different molecular compositions is particularly important for the development and maintenance of multicellular organisms. These compartments are delineated by various cellular sheets, which function as barriers to maintain the distinct internal environment of each compartment. For example, renal tubules, blood vessels and the peritoneal cavity are lined with EPITHELIAL, ENDOTHELIAL and MESOTHELIAL cellular sheets, respectively. Within these sheets, individual cells are mechanically linked with each other to maintain the structural integrity of the sheet, and the intercellular space between adjacent cells is sealed to prevent the diffusion of solutes through the intercellular space. The junctional complex of simple epithelial cells is located at the most-apical part of the lateral membrane and consists of three components: tight junctions, ADHERENS JUNCTIONS and DESMOSOMES1 (FIG. 1). On ultrathin section electron micrographs, tight junctions appear as a series of apparent fusions (‘kissing points’), involving the outer leaflets of the plasma membranes of adjacent cells (FIG. 1b and FIG. 2b). At the kissing points, the intercellular space is completely obliterated, whereas in adherens junctions and desmosomes, the apposing membranes are 15–20 nm apart (FIG. 1b). In simple epithelial cellular sheets, adherens junctions and desmosomes mechanically link adjacent cells, whereas tight junctions are responsible for intercellular sealing2,3. But many physiological situations require that various materials are selectively transport-

ed across cellular sheets, and this occurs either by TRANSCELLULAR TRANSPORT through the cell or by paracellular flux through tight junctions4 (BOX 1). So tight junctions are not simply impermeable barriers: they show ion as well as size selectivity, and vary in tightness depending on the cell type3,5. In addition to the ‘barrier function’, tight junctions are thought to function as a ‘fence’2,3. Plasma membranes are functionally divided into apical and basolateral domains that face the luminal and serosal compartments, respectively. Apical and basolateral membrane domains differ in the compositions of integral membrane proteins and lipids. However, because integral membrane proteins and lipids can diffuse laterally within the plane of the plasma membrane, some diffusion barrier is required at the border between apical and basolateral membrane domains. Tight junctions look like a fence in the most apical part of the lateral plasma membrane and they probably are the morphological counterpart of a localized diffusion barrier. In recent years, information on the molecular components of tight junctions, and in particular their cell- adhesion molecules, has accumulated. Here, we present an overview of our current understanding of the structure and functions of tight junctions in molecular terms. The composition of tight-junction strands

The morphology of tight junctions has been intensively analysed by FREEZE-FRACTURE REPLICA ELECTRON MICROSCOPY.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 8 5

REVIEWS

Microvilli

Tight junction Adherens junction

Gap junction

DS Desmosome

Hemidesmosome

Basal lamina

Figure 1 | Junctional complex and tight junctions. a | Schematic drawing of intestinal epithelial cells. The junctional complex, which is located at the most apical region of lateral membranes, is circled. b | Electron micrograph of the junctional complex in mouse intestinal epithelial cells. The tight junction is circled. (Mv, microvilli; TJ, tight junction; AJ, adherens junction; DS, desmosome.) Scale bar, 200 nm.

ADHERENS JUNCTIONS

Cell–cell adhesive junctions that are linked to cytoskeletal filaments of the microfilament type. DESMOSOMES

These patch-like intercellular junctions are found in vertebrate tissue, and are particularly abundant in tissues undergoing mechanical stress. The central plaque contains adhesion molecules, and represents an anchorage point for cytoskeletal filaments of the intermediate filament type.

On freeze-fracture replica electron micrographs, tight junctions appear as a set of continuous, ANASTOMOSING intramembranous particle strands or fibrils (tight-junction strands) on the P FACE with complementary vacant grooves on the E FACE6 (FIG. 2a). The number of tightjunction strands as well as the frequency of their ramification vary notably depending on cell type, producing marked variation in the morphology of tight-junction strand networks. These observations led to our understanding of the three-dimensional structure of tight junctions (FIG. 2c). Each tight-junction strand within the plasma membrane associates laterally with another tight-junction strand in the apposing membrane of adjacent cells to form ‘paired’ tight-junction strands, where the intercellular space is obliterated. Two models have been proposed to explain the

Box 1 | Two distinct pathways across cellular sheets Tight junction Paracellular There are two pathways pathway Apical through which materials membrane cross epithelial and endothelial cellular sheets: the transcellular and paracellular pathways. The transepithelial electric resistance of epithelia from different tissues varies by Transcellular Basolateral two orders of magnitude, pathway membrane which is thought largely to reflect variations in the permeability of tight junctions91. Tight junctions show charge and size selectivity in their permeability. Ion transport across tight junctions is cation selective92. Anion-selective transport was found in a few cases, such as the rabbit colon and frog skin92. Non-charged materials such as water and sucrose also move across tight junctions4,5. Under various physiological conditions, material transport across tight junctions occurs in a regulated fashion, and its regulation in certain states might be coupled to transcellular transport91,93. For example, during absorption of glucose from the intestine, a large fraction of glucose might be transported across tight junctions, particularly when luminal glucose concentrations are elavated, saturating the Na+–glucose co-transporter in the apical plasma membrane91. Activation of the Na+–glucose co-transporter is thought to alter the structure and function of tight junctions, but the molecular mechanism underlying this coupling remains unclear.

286

chemical nature of tight-junction strands (FIG. 3). In the ‘protein model’, tight-junction strands represent units of integral membrane proteins that are polymerized linearly within lipid bilayers whereas, in the ‘lipid model’, lipids organized in inverted cylindrical micelles are proposed to constitute tight-junction strands7. Recent identification of tight-junction-specific integral membrane proteins strongly supports the ‘protein’ model, although we cannot exclude the possibility that specific lipids might also be important for the formation of tightjunction strands. Occludin (~60 kDa) was identified as the first integral membrane protein localized at tight junctions in chicken8, and then also in mammals9. Occludin has four transmembrane domains, a long carboxy-terminal cytoplasmic domain and a short amino-terminal cytoplasmic domain (FIG. 4a). No occludin-related genes have been identified yet, but two isoforms of occludin are generated by alternative splicing10. In IMMUNO-REPLICA ELECTRON MICROSCOPY, antibodies against occludin exclusively labelled tight-junction strands11, indicating that occludin is probably incorporated directly into tight-junction strands. Furthermore, as the intensity of immunostaining with antibodies against occludin in various tissues correlates well with the number of tight-junction strands, the density of occludin molecules in tight-junction strands seems almost constant11. But tight-junction strands can also be formed without occludin, as in some cell types such as endothelial cells in non-neuronal tissue and in SERTOLI CELLS in some species, occludin was not detected in tightjunction strands12,13. More importantly, VISCERAL ENDODERM cells originating from occludin-deficient embryonic stem cells have well-developed networks of tight-junction strands14. At present, the physiological functions of occludin are not well understood. Its possible functions will be discussed in detail below. More recently, two other transmembrane proteins — claudin-1 and claudin-2 — were identified as integral components of tight-junction strands15. These proteins also have four transmembrane domains, but do not show any sequence similarity to occludin (FIG. 4b). So far, 24 members of the claudin family have been identified in mouse and human, mainly through database searches16,17 (TABLE 1). There is accumulating evidence that claudins constitute the backbone of tight-junction strands. Immuno-replica electron microscopy revealed that claudins are exclusively localized on tight-junction strands16,18,19. Exogenously expressed claudins conferred cell-aggregation activity on L FIBROBLASTS with their concomitant concentration at cell–cell contact planes20 and led to the formation of a large network of structures that look like tight-junction strands21 (FIG. 4d). Occludin itself cannot reconstitute such well-organized strands, but when it was introduced into claudin-expressing L transfectants, it was incorporated into reconstituted claudin-based strands21. The expression pattern of claudins varies considerably among tissues15,16 (TABLE 1). Some claudins are expressed in specific cell types; for example claudinwww.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

c Extracellular space

Mv Ap Cytoplasmic surface

Kissing point

Tight-junction strand

Figure 2 | Structure of tight junctions. a | Freeze-fracture replica electron microscopic image of intestinal epithelial cells. Tight junctions appear as a set of continuous, anastomosing intramembranous particle strands or fibrils (arrowheads) on the P face with complementary vacant grooves on the E face (arrows). (Mv, microvilli; Ap, apical membrane; Bl, basolateral membrane.) Scale bar, 200 nm. b | Ultrathin sectional view of tight junctions. At kissing points of tight junctions (arrowheads), the intercellular space is obliterated. Scale bar, 50 nm. c | Schematic of three-dimensional structure of tight junctions. Each tight-junction strand within a plasma membrane associates laterally with another tight-junction strand in the apposed membrane of an adjacent cell to form a paired tight-junction strand, obliterating the intercellular space (kissing point).

TRANSCELLULAR TRANSPORT

Transport of macromolecules across a cell, including transport through channels, pumps and transporters, as well as transcytosis (endocytosis of a macromolecule at one side of a monolayer and exocytosis at the other side). FREEZE-FRACTURE REPLICA ELECTRON MICROSCOPY

An electron-microscopic method that uses metal replicas to visualize the interior of cell membranes. This technique provides a convenient way to visualize the distribution of large integral membrane proteins as intramembranous particles in the plane of a membrane. ANASTOMOSIS

Cross-connection between adjacent channels, tubes, fibres or other parts of a network. P AND E FACE

When membranes are freezefractured, fracture planes run between the cytoplasmic and extracytoplasmic leaflets of plasma membranes, giving the Por E-face images of membranes. The P (protoplasmic) face is the inner leaflet viewed from the outside, whereas the E (extracytoplasmic) face is the outer leaflet viewed from the inside. IMMUNO-REPLICA ELECTRON MICROSCOPY

A form of electron microscopy, combining freeze-fracture replica electron microscopy and immune labelling of proteins.

5/TMVCF is expressed primarily in endothelial cells of blood vessels19, and claudin-11/OSP is expressed primarily in OLIGODENDROCYTES and Sertoli cells18. Most cell types, however, express more than two claudin species in various combinations to constitute tightjunction strands: within individual single strands, distinct species of claudin are polymerized together to form ‘heteropolymers’, and between adjacent strands within ‘paired’ strands, claudins adhere with each other in a homotypic as well as a heterotypic manner22,23. The last transmembrane component of tight junctions is JAM (junctional adhesion molecule; ~40 kDa)24. There are three JAM-related proteins25,26, which belong to the immunoglobulin superfamily: they have a single transmembrane domain and their extracellular portion is thought to be folded into two immunoglobulin-like domains (FIG. 4c). Preliminary freeze-fracture replica electron microscopy revealed that exogenously expressed JAM does not reconstitute tight-junction strands in L transfectants and that it associates laterally with the claudin-based backbone of tight-junction strands in epithelial cells. JAM is involved in cell–cell adhesion/junctional assembly of epithelial/endothelial cells24,25,27,28 as well as in the EXTRAVASATION of MONOCYTES through endothelial cells24, but our knowledge on its function is still fragmentary.

epithelial cells — MDCK I and MDCK II — show marked disparity in their TER. Stevenson et al. reported that MDCK I cells have a 30–60-fold higher TER than MDCK II cells, but the number of tight-junction strands in these strains is very similar33. These observations indicate that individually paired tight-junction strands vary not only in number, but also in quality. The number of strands. The number of tight-junction strands is an important factor in determining the barrier properties of tight junctions, but the molecular mechanism underlying regulation of the strand number remains unknown. When MDCK I cells, which express claudin-1 and claudin-4, were specifically depleted of claudin-4, a marked decrease was observed in the number of tight-junction strands and in their barrier function34. Mice lacking claudin-11/OSP, which is expressed specifically in oligodendrocytes and Sertoli cells in wild-type mice, were recently generated, and in these mice tight-junction strands were absent in MYELIN 35 SHEATHS as well as in Sertoli cells . Furthermore, when claudins were overexpressed in L fibroblasts, a large network of tight-junction strands was formed21. In addition, overexpression of occludin in MDCK cells was also shown to cause a slight increase in the number of tight-junction strands36. These findings indicate that the number of tight-junction strands might be determined by the total amount of expressed claudins (and occludin) in individual cells. However, the regulation of the number of tight-junction strands is probably more complicated. In epithelial cells that already express claudins, overexpression of claudins did not lead to an increase in the number of tight-junction strands37, indicating that an upper limit might exist. Interestingly, when a claudin-1 mutant that cannot bind the underlying cytoskeleton was overexpressed in MDCK cells, aberrant tight-junction strands were formed37. This finding indicates the possible involvement of the underlying cytoskeleton in the regulation of the tight-junction strand number, but how the upper limit is set remains a mystery. In addition to the strand number, the complexity of the network pattern might also be an important factor for determining the barrier properties of tight junctions30.There were marked differences in the network

A ziplock with diversified permeability

Tight junctions vary in tightness in a tissue-dependent manner2,3. Their tightness can be directly measured as TRANSEPITHELIAL ELECTRIC RESISTANCE (TER). The number of tight-junction strands was found to correlate well with the TER values of tight junctions in various tissues29,30. For example, in the kidney, epithelial cells of the proximal and distal tubules have 1–2 and 4–7 tight-junction strands, respectively, and the epithelial cells of the distal tubules have a much higher TER than those of the proximal tubules. However, exceptions to this correlation have also been reported31,32. For example, the two existing strains of Madin–Darby canine kidney (MDCK)

Protein model

Lipid model

Figure 3 | Protein versus lipid models. In the protein model, tight-junction strands represent units of integral membrane proteins polymerized linearly within lipid bilayers, whereas in the lipid model inverted cylindrical lipid micelles constitute tight-junction strands.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 8 7

REVIEWS

N JAM

Occludin

Claudin-1

P E

N C 211

299

521

Figure 4 | Integral membrane proteins localized at tight junctions. a | Occludin has four transmembrane domains with two extracellular loops. The first loop is characterized by a high content (~60%) of glycine and tyrosine residues. b | Claudin-1 also has four transmembrane domains, but shows no sequence similarity to occludin. Note that the cytoplasmic tail of claudin-1 is shorter than that of occludin. c | Junctional adhesion molecule (JAM) has a single transmembrane domain, and its extracellular portion bears two immunoglobulin-like loops that are formed by disulphide bonds. d | Freeze-fracture replica electron microscopy of L cell transfectants expressing exogenous claudin-1. At cell–cell contact planes, a huge network of tight-junction strand-like structures was reconstituted; strands (arrowheads) on the P face and complementary grooves (arrows) on the E face. Scale bar, 100 nm.

SERTOLI CELLS

A supporting cell of the mammalian testis that surrounds and nourishes developing sperm cells. VISCERAL ENDODERM

Cells that delineate the yolk sac cavity together with parietal endoderm cells in the egg cylinder stage of the mammalian embryo. L FIBROBLASTS

A mouse fibroblast line derived from connective tissue that does not show adhesion activity.

patterns of tight junctions reconstituted by expression of different types of claudin in fibroblasts. For example, claudin-1-induced tight-junction strands form a large network with frequent ramifications21, whereas claudin11/OSP-induced strands scarcely branched, running parallel to each other18. This observation is likely to be relevant in vivo, as claudin-11/OSP-based tight-junction strands in myelin sheaths and Sertoli cells are mostly parallel with little branching18,35. It is tempting to speculate that the complexity of the tight-junction strand network is determined by the combination and the mixing ratio of the different claudins. Extracellular aqueous pores. The extracellular portion of tight-junction strands probably functions as a ziplock to create a primary barrier against paracellular diffusion (FIG. 5). By comparing the TER and the mor-

phology of tight-junction strands in various epithelia, Claude et al. found that as tight-junction strands increase in number, the TER value increases logarithmically30. To explain this relationship, the existence of aqueous pores, taking both open and closed states, was postulated within the paired tight-junction strands5,30,38 (BOX 2). However, some exceptions to the relationship between the number of tight-junction strands and the TER value have been found31,32. The difference in tightness of individual tight-junction strands could be explained by the heterogeneity of aqueous pores in terms of their probability of being open or closed33. What is the chemical nature of these aqueous pores? Recent studies on hereditary hypomagnesemia have provided a clue to this question39. Most Mg2+ is resorbed from the urine through the paracellular pathway in the thick ascending LIMB OF HENLE, but in patients with heredity hypomagnesemia, this reabsorption is reduced. Positional cloning identified claudin-16/paracellin-1 as the gene responsible for this disease. Consistently, claudin-16/paracellin-1 is exclusively expressed in the thick ascending limb of Henle. This finding indicated that claudin-16/paracellin-1 might form aqueous pores that function as Mg2+ paracellular channels. The difference between MDCK I and MDCK II cells might be due to the expression of different claudins40: MDCK I express primarily claudin-1 and claudin-4, whereas MDCK II cells also express large amounts of claudin-2 in addition to claudin-1 and claudin-4. When claudin-2 was introduced into MDCK I cells, the TER value of these MDCK I transfectants fell to the level of MDCK II cells without any changes in the number of tight-junction strands. By contrast, exogenously expressed claudin-3 did not affect the TER value of MDCK I cells. Therefore, it is likely that claudin-2 constitutes aqueous pores with high conductance within paired tight-junction strands of MDCK II cells. These findings led to the speculation that claudins not only form the backbone of tight-junction strands but also form extracellular aqueous pores, and that the

OLIGODENDROCYTES

A type of glial cell that forms and supports the myelin sheath around axons in the central nervous system of vertebrates. EXTRAVASATE

Let or force out something from a vessel that naturally contains it. MONOCYTES

Large leukocyte of the mononuclear phagocyte system found in bone marrow and the bloodstream. Monocytes are derived from pluripotent stem cells and become macrophages when they enter the tissues. TRANSEPITHELIAL ELECTRIC RESISTANCE

Electric resistance across epithelial sheets, measured across the apical–basolateral axis of the cell.

288

Table 1 | Claudin gene family Claudins* Alternative names

Expression‡ Heart Brain Lung Liver Kidney Testis

Other

Claudin-1

–

Claudin-2

–

+ –

Claudin-3

RVP1 (REF. 94)

–

Claudin-4

CPE-R95

–

Claudin-5

TMVCF96

Endothelial cells Embyronic tissues

Claudin-6

–

Claudin-7

–

Claudin-8

–

Claudin-10 OSPL

Claudin-11 OSP97

–

Oligodendrocytes, Sertoli cells

Claudin-14 –

–

Hair cells in the Corti organ

Claudin-16 Paracellin-1 (REF. 39)

–

Thick ascending limb of Henle

*Claudin-9, -12, -13, -15 and 17–24 have not been characterized well. ‡Detected by northern blotting15,16,18,39,88. ND, not determined.

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

MYELIN SHEATH

The sheath that surrounds the axons of vertebrate nerves to prevent the leakage of electric current. It is formed by Schwann cells in peripheral nerves and by oligodendrocytes in the central nervous system. These cells wrap up to 100 concentric layers of their plasma membrane in a tight spiral around the axons.

Plasma membranes of adjacent cells

Ziplock Signalling

LIMB OF HENLE

U-shaped part of a nephron lying in the renal medulla. It comprises a thin descending tubule and an ascending tubule formed of both a thin and a thick segment. It has a role in the selective reabsorption of fluid and solutes. HYPERPLASIA

The increase in the size of a tissue or organ, resulting from an increase in the total number of cells present. The part that is affected retains its normal form.

PDZ-domain proteins

Fence

Magnetic bar Barrier

Figure 5 | Multiple functions of tight-junction strands. The extracellular and cytoplasmic portions of tight-junction strands function as a ziplock with diversified permeability (barrier function) and a magnetic bar attracting various PDZcontaining proteins (signalling function), respectively. Furthermore, within plasma membranes, the strand functions as a ‘fence’ to establish apical and basolateral membrane domains (fence function).

GAP JUNCTION

A junction between two cells consisting of pores that allow passage of molecules (up to 9 kDa).

combination and the mixing ratios of different claudins determine the tightness of individual tight-junction strands23. But it is also possible that tight-junction strands are simply repeatedly broken and annealed, and that this contributes to the tightness of individual strands. So far, no information is available about the

Box 2 | Aqueous pores within tight-junction strands a

Aqueous pores

b or

Tight-junction strands d

stability of strands, and the molecular details should be examined in future studies. Occludin has also been shown to be involved in the barrier function of tight junctions, but at present, how occludin is involved remains unclear. Transfection of carboxy-terminally truncated occludin into MDCK cells removed endogenous occludin from tight-junction strands. In these transfectants, the TER did not decrease41. These findings contradict those of another study, in which endogenous occludin was removed from tight junctions by addition of synthetic peptides corresponding to the second extracellular loop of occludin to the culture medium. This resulted in a marked decrease in the TER, accompanied by an increase in paracellular flux42. Interestingly, overexpression of full-length as well as carboxy-terminally truncated occludin in MDCK cells did not affect the TER, but, paradoxically, increased the paracellular flux of mannitol36,37,41. Although the precise relationship between occludin and the TER is still a mystery, these observations indicate that occludin might contribute to the electrical barrier function of tight junctions and possibly to the formation of aqueous pores within tight-junction strands through which flux of noncharged solutes occurs. These conclusions are compatible with those obtained from studies of occludin-deficient mice. These mice were born seemingly normal, but as they grew up, they began to show complex phenotypes, including marked growth retardation, chronic inflammation and HYPERPLASIA of the gastric epithelium, and mineral deposition in the brain43. Tight junctions in most organs of occludin-deficient mice that have been studied so far, such as intestinal epithelial cells, seem normal in terms of their morphology and TER. Although the paracellular flux of non-charged solutes has not yet been examined in these mice, it is possible that the defects found in these mice could be caused by the disappearance of occludin-dependent aqueous pores from tight-junction strands. Of course, it is also possible that occludin is important for other functions of tight junctions apart from regulating paracellular flux. These could be related to the fence function of tight junctions or to signalling events that take place at tight junctions, which could explain the pleiotropic defects observed in occludindeficient mice. A magnetic bar for PDZ-containing proteins

Paracellular pathway

Claude et al. found that transepithelial resistance (Rt) increases logarithmically with the number of parallel tight-junction strands (n) that must be crossed by a permeating ion in the transepithelial direction29,30. To explain this relationship, they postulated the existence of aqueous pores within tight-junction strands that fluctuate between open and closed states30 (a). If each parallel strand has one channel, and if the open probability (PO) of all channels is the same, Rt would depend on PO–n (b). However, along individual tight-junction strands, there would be a large number of channels, and under this condition Rt would increase linearly with n (c). To maintain the exponential relationship, Cereijido and colleagues suggested compartmentalization, corresponding to the freeze-fracture replica images of tight-junction strand networks38 (d).

The thickness of tight-junction strands6 (~10 nm; see FIG. 2a) is similar to the diameter of the GAP JUNCTIONAL channel (connexon), which consists of six connexin molecules that also have four transmembrane domains. Therefore, it is likely that, instead of being aligned in a single line to constitute tight-junction strands, claudins are packed more densely in the strands. It is then expected that the cytoplasmic surface of individual tight-junction strands appears as a toothbrush consisting of densely packed, numerous short carboxy-terminal cytoplasmic tails of claudins. In addition to these claudin tails, relatively long carboxy-terminal tails of occludin are probably intermingled.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 8 9

REVIEWS

PDZ1

ZO-1

PDZ2

PDZ3 SH3

GUK 1745

Claudins PDZ1

ZO-2

ZONAB Occludin PDZ3 SH3 GUK

PDZ2

Actin filament

ZO-2

1167 Claudins ZO-1 PDZ1 PDZ2 PDZ3 SH3

ZO-3

Claudins

PAR-3

ZO-1 PDZ1

Actin filament GUK

PDZ2 PDZ3 904 Atypical PKC

CRIB PDZ

PAR-6

345 Atypical Cdc42 PKC

MAGI-1

MAGI-2

PDZ0 GUK

PDZ1

PDZ2

PDZ3 PDZ4 PDZ5 1171

PDZ0 GUK

PDZ1

PDZ2

PDZ1

PTEN PDZ2

PDZ3

PDZ4

PDZ5 1455

PDZ0 GUK

PDZ3 PDZ4

PDZ5

MAGI-3

1150 PTEN

Figure 6 | PDZ-containing proteins localized at tight junctions. PDZ domains are represented by closed boxes, and the regions responsible for intermolecular association are indicated by arrowheads. Numbers show amino-acid positions. (GUK, guanylyl-kinase-like domain; MAGI, membrane-associated guanylyl kinase inverted; PKC, protein kinase C; PAR, partitioning defective; SH3, src homology region 3.)

PDZ DOMAIN

Protein–protein interaction domain first described in the proteins PSD-95, DLG and ZO-1. POSTSYNAPTIC DENSITY

Higher-order protein structure present in postsynaptic membranes that functions to concentrate neurotransmitter receptors.

Many cytosolic proteins have been reported to associate with the cytoplasmic surface of tight junctions. As the first component of tight junctions, a peripheral membrane protein with a molecular mass of 220 kDa was identified through monoclonal antibody production, and was named ZO-1 (zonula occludens-1)44. When ZO-1 was immunoprecipitated from cell lysates of MDCK cells, two proteins with molecular masses of 160 kDa and 130 kDa were co-precipitated45,46. As these proteins were also localized at tight junctions, they were

Table 2 | Proteins lacking PDZ domains recruited to tight junctions Proteins

Binding partners Possible functions in tight junctions

References

Cingulin

JAM, ZO-1, ZO-2, Cross-linking tight junctions and ZO-3 the actomyosin cytoskeleton

98, 99

7H6 antigen

Symplekin

100 101

Heterotrimeric G proteins

Regulation of tight-junction formation

102

Atypical PKC (PKCζ/λ)

PAR-3, PAR-6

Serine/threonine kinase Regulation of epithelial polarization

ZONAB

ZO-1

Transcription factor Regulation of ErbB-2 transcription

huASH1

Transcription factor

Rab3b, Rab-13

Polarized vesicle transport

Sec6/Sec8 homologues

Vesicle targetting

PTEN

MAGI-2, MAGI-3

Lipid phosphatase Tumour suppressor

103

104 69, 70 71 60, 61

JAM, junctional adhesion molecule; MAGI, membrane-associated guanylyl kinase inverted; PAR, partitioning defective; PKC, protein kinase C; ZO, zonula occludens; ZONAB, ZO-1-associated nucleic-acid-binding protein.

290

designated as ZO-2 and ZO-3, respectively. ZO-1, ZO-2 and ZO-3 have sequence similarity with each other47–50: they contain three PDZ DOMAINS (PDZ1, PDZ2 and PDZ3), one SH3 domain, and one guanylyl kinase-like (GUK) domain (FIG. 6). The PDZ domain was initially reported to bind specifically to the carboxy-terminal Glu–Ser/Thr– Asp–Val motif, but it is now known that it recognizes more diverse four amino-acid sequences, most often ending in Val. Interestingly, most claudin tails have a Val at their carboxyl termini15,16; the only known exception is claudin-12. This indicates that these carboxyl termini might directly bind to PDZ domains. If so, the cytoplasmic surface of tight-junction strands might function as a magnetic bar that strongly attracts and recruits many PDZ-containing proteins (FIG. 5). Indeed, the PDZ1 domains of ZO-1, ZO-2 and ZO-3 were recently shown to bind directly to the carboxyl termini of claudins51. No notable differences were detected in the affinity of different claudins for PDZ1 domains. ZO-1, ZO-2 and ZO-3 also directly bind to the carboxy-terminal tail of occludin, but their GUK domains, not their PDZ domains, are involved in this interaction 50,52–54. As ZO-1, ZO-2 and ZO-3 are localized to tight junctions in occludin-deficient mice, this interaction might not be essential for their recruitment to tight junctions43,51. Furthermore, JAM, which is concentrated around tightjunction strands, also ends in Val, and was recently shown to bind directly to ZO-155,56 and other PDZ-containing proteins57. So, it is possible that JAM is also involved in the recruitment of various PDZ-containing proteins to tight junctions. In addition to ZO-1, ZO-2 and ZO-3, several PDZcontaining proteins are recruited to the cytoplasmic surface of tight junctions, but it remains unknown whether these proteins directly bind to the carboxyl termini of claudins (FIG. 6). Examples include MAGI-1/-2/-3 (MAGUK inverted -1/-2/-3)58–61, and mammalian homologues of C. elegans PAR (partitioning defective) gene products, which function in asymmetric cell division62–65 (FIG. 6). The list of PDZ-containing proteins localized at tight junctions will probably continue to increase. These proteins might function as adaptors at the cytoplasmic surface of tight-junction strands to recruit other proteins including cytoskeletal and signalling molecules (FIG. 6, TABLE 2). Through the recruitment of various types of protein to tight junctions using PDZ-containing proteins, a huge macromolecular complex is probably formed at the cytoplasmic surface of tight-junction strands (FIG. 5). What are the physiological functions of this complex? First, as actin filaments bind to the carboxy-terminal portions of ZO-1 and ZO-2 (REFS 53,54,66), this complex probably crosslinks tight-junction strands to actomyosin cytoskeletons, and this interaction might have a role in the regulation of tight-junction functions. Interestingly, similar accumulation of PDZ-containing proteins occurs in the POSTSYNAPTIC DENSITY in neurons, where PDZ-containing proteins are directly involved in synaptic signal transduction and its regulation67. In cell–matrix adhesion, a huge macromolecular complex www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS is formed at INTEGRIN-based adhesion sites, which has crucial roles in extracellular-matrix-dependent signalling, but in this case, complex formation is not based on PDZ-containing proteins68. However, such macromolecular complexes are not well developed at CADHERIN-based cell–cell adhesion sites (adherens junctions). Therefore, it is tempting to speculate that the huge macromolecular complex formed at the cytoplasmic surface of tight junctions is central in the intercellular adhesion signalling of epithelial and endothelial cells, and is involved in the regulation of their proliferation, differentiation and polarization. In this context, it is interesting to point out that tight junctions recruit a tumour suppressor gene product (PTEN)60,61, cellpolarity-related gene products (PAR-3, PAR-6, cdc42)63–65 and vesicular-transport-related proteins (Rab3b, Rab13, Sec6/Sec8 products)69–71. A fence within the plasma membrane

INTEGRINS

A large family of heterodimeric transmembrane proteins that act as receptors for cell-adhesion molecules. CADHERINS

Calcium-dependent adhesion molecules that mediate homophilic adhesions. There are several subfamilies of cadherin. GLYCOSPHINGOLIPIDS

Any compound containing residues of a sphingoid and at least one monosaccharide. SPHINGOMYELIN

Any of a class of phospholipids in which the amino group of sphingosine is in amide linkage with a fatty acid, and the terminal hydroxyl group of sphingosine is esterified to phosphorylcholine.

Tight-junction strands are heteropolymers of integral membrane proteins, occludin and claudins, which are embedded within the plasma membrane, and encircle the top of individual epithelial/endothelial cells to delineate the border between the apical and basolateral membrane domains. Therefore, it is likely that tightjunction strands act as a ‘fence’, limiting the lateral diffusion of lipids and proteins between the apical and basolateral membrane domains (FIG. 5). The apical membrane of epithelial cells is enriched in GLYCOSPHINGOLIPIDS and SPHINGOMYELIN72,73. Interestingly, this membrane has a striking asymmetric organization of lipids across the lipid bilayer, and glycosphingolipids, as well as sphingomyelin, are concentrated in its outer leaflet74,75. Such polarized localization of lipids indicates the possible existence of a diffusion barrier, especially for the lipids in the outer leaflet. This was confirmed experimentally. When fluorescently labelled lipids were inserted into the outer leaflet of the apical membrane of cultured epithelial cells, they remained on the apical surface. By contrast, fluorescently labelled lipids inserted into the inner leaflet of the apical membrane quickly redistributed to the basolateral surface76,77. It is reasonable to speculate that tight junctions restrict the lateral diffusion of not only lipids but also integral membrane proteins. As the intercellular space is completely obliterated at tight junctions, integral membrane proteins with extended extracellular portions could not cross tight junctions. However, it is also clear that, in addition to tight junctions, there are other mechanisms behind the asymmetric distribution of certain integral membrane proteins within plasma membranes78: the cytoskeletal proteins underlying the plasma membranes can restrict the lateral diffusion of proteins within membrane domains, and the targeted delivery of exocytic vesicles is also important. Early studies showed that the disruption of intercellular junctions (for instance, by incubation with low-calcium medium) resulted in the intermixture of membrane proteins from the apical and basolateral domains79,80. In another study, the correct polarization of a basolateral protein depended on cell–cell junctions, whereas that of

an apical protein did not81. However, in these experiments, not only intercellular junctions but also cytoskeletal organization were affected. So, it is difficult to conclude that tight junctions act as a diffusion barrier for membrane proteins. In a more recent study, expression of constitutively active RhoA and Rac1 small GTPases in MDCK cells resulted in the disorganization of tight-junction-strand networks as well as the disruption of the junctional fence for lipids but not for integral membrane proteins82. So, the importance of tight junctions in the asymmetric distribution of integral proteins remains controversial. When carboxy-terminally truncated occludin was overexpressed in cultured MDCK cells, fluorescently labelled sphingomyelin added to the apical membrane domain was redistributed to the basolateral surface41. The polarized distribution of integral membrane proteins did not seem to be affected. These findings suggest a possible involvement of occludin in the diffusion barrier, especially for lipids. The relationship between claudins and the diffusion barrier in epithelial cells has not yet been examined. Future directions

Tight junctions have attracted a great deal of interest, but their study has been hampered by lack of information about tight-junction-specific integral membrane proteins. The recent identification of the main components of tight-junction strands has facilitated the molecular assessment of the morphological and physiological observations of tight junctions that have accumulated over years. On the basis of the accumulated information on occludin and claudins, we have discussed here several functions of tight-junction strands at the molecular level; the barrier, signalling and fence functions (FIG. 5). The challenge of answering many other questions lies ahead. The identification of occludin and claudins raises many basic questions about the structure of the tightjunction strand itself. How are occludin and heterogeneous claudins arranged in individual tight-junction strands? To what extent are tight-junction strands dynamically polymerized and depolymerized? How is this regulated? Are some lipids required for the polymerization of occludin and claudins? How can tight-junction strands restrict the lateral diffusion of lipids only in the outer leaflet of the membrane? As the polymerization of integral membrane proteins in a linear fashion is unique, the elucidation of the basic physico-chemical properties of tight-junction strands might constitute one of the big challenges for years to come. Several intriguing questions stand out at the cellular level. One of the most pressing questions concerns the molecular mechanism that underlies the polarized formation of tight junctions at the most apical region of the lateral membranes in epithelial cells. What is the relationship between tight junctions, adherens junctions and desmosomes during epithelial polarization? How are occludin, claudins, cadherins and their associated molecules integrated into the polarized junctional

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 9 1

REVIEWS

HETEROTRIMERIC G PROTEINS

Components of receptormediated activation or inhibition of adenylyl cyclase and other second messenger systems. MITOGEN-ACTIVATED PROTEIN KINASE CASCADE

Signalling cascade that relays signals from the plasma membrane to the nucleus. MAPKs, which represent the first step in the pathway, are activated by a wide range of proliferation- or differentiation-inducing signals.

3. 4. 5. 6.

7. 8.

10.

11.

12.

13.

14.

292

complex during epithelial polarization? It has been hypothesized that ZO-1, ZO-2 and ZO-3 might recruit tight-junction proteins such as claudins and occludin to their final destination at the interface between the apical and basolateral membrane domains, but compelling evidence is lacking so far. Another outstanding issue concerns the regulation of the tight-junction barrier. As indicated above, tight junctions vary in tightness in a cell-type-dependent manner. This tightness is also dynamically and finely regulated in individual cells, depending on various physiological and pathological requirements2,3 (BOX 1). The information of the molecular mechanism underlying these regulations is still fragmentary, but several signalling pathways such as serine/threonine phosphorylation, tyrosine phosphorylation, HETEROTRIMERIC G PROTEINS and small G proteins are thought to be involved in their regulation83. The transcription of occludin was reported to be downregulated by tumour necrosis factor-α and interferon-γ 84 and/or by activation of the MITOGEN-ACTI85,86 VATED PROTEIN KINASE CASCADE , but there is no information available about the transcriptional regulation of claudins by these or other signalling pathways. The cytoplasmic tail of occludin was shown to be heavily phosphorylated on serine and threonine residues87, whereas the phosphorylation of claudins has not yet been examined. Finally, another important challenge for future studies of tight junctions is to examine their possible involvement in various diseases. As mentioned above, mutations in claudin-16/paracellin-1 were shown to

Farquhar, M. G. & Palade, G. E. Junctional complexes in various epithelia. J. Cell Biol. 17, 375–412 (1963). This is the first electron microscopic description of the junctional complex consisting of tight junctions, adherens junctions and desmosomes. Schneeberger, E. E. & Lynch, R. D. Structure, function, and regulation of cellular tight junctions. Am. J. Physiol. 262, L647–L661 (1992). Gumbiner, B. Breaking through the tight junction barrier. J. Cell Biol. 123, 1631–1633 (1993). Spring, K. Routes and mechanism of fluid transport by epithelia. Annu. Rev. Physiol. 60, 105–119 (1998). Reuss, L. in Tight Junctions (ed. Cereijido, M.) 49–66 (CRC, London, 1992). Staehelin, L. A. Further observations on the fine structure of freeze-cleaved tight junctions. J. Cell Sci. 13, 763–786 (1973). Kachar, B. & Reese, T. S. Evidence for the lipidic nature of tight junction strands. Nature 296, 464–466 (1982). Furuse, M. et al. Occludin: a novel integral membrane protein localizing at tight junctions. J. Cell Biol. 123, 1777–1788 (1993). This paper reports identification of occludin as a first component of tight-junction strands. Ando-Akatsuka,Y. et al. Interspecies diversity of the occludin sequence: cDNA cloning of human, mouse, dog, and rat-kangaroo homologues. J. Cell Biol. 133, 43–47 (1996). Muresan, Z., Paul, D. L. & Goodenough, D. A. Occludin1B, a variant of the tight junction protein. Mol. Biol. Cell 11, 627–634 (2000). Saitou, M. et al. Mammalian occludin in epithelial cells: its expression and subcellular distribution. Eur. J. Cell Biol. 73, 222–231 (1997). Hirase, T. et al. Occludin as a possible determinant of tight junction permeability in endothelial cells. J. Cell Sci. 110, 1603–1613 (1997). Moroi, S. et al. Occludin is concentrated at tight junctions of mouse/rat but not human/guinea pig Sertoli cells in testes. Am. J. Physiol. 274, C1708–C1717 (1998). Saitou, M. et al. Occludin-deficient embryonic stem cells can differentiate into polarized epithelial cells bearing tight

cause hereditary hypomagnesemia39. Furthermore, recent positional cloning identified claudin-14 as the gene responsible for hereditary deafness88. This claudin species is expressed in hair cells in the cochlea of the inner ear. As tight junctions in these cells have crucial roles in the establishment of two compositionally distinct compartments in the inner ear, mutations in the claudin-14 gene would cause deafness. In addition to hereditary diseases, claudins seem to have something to do with various pathological conditions including inflammation89. Furthermore, the involvement of occludin86 as well as claudins90 in tumorigenesis has been suggested in recent years. We are only just beginning to understand the functions of tight junctions in molecular terms. Our picture of the molecular architecture of tight junctions remains incomplete, and other important constituents need to be identified. Further development of the molecular biology of tight junctions will lead to a better understanding of their functions, not only in normal physiology, but also in disease.

junctions. J. Cell Biol. 141, 397–408 (1998). 15. Furuse, M., Fujita, K., Hiiragi, T., Fujimoto, K. & Tsukita, S. Claudin-1 and-2: novel integral membrane proteins localizing at tight junctions with no sequence similarity to occludin. J. Cell Biol. 141, 1539–1550 (1998). Identification and cloning of the genes encoding claudin-1 and claudin-2. 16. Morita, K., Furuse, M., Fujimoto, K. & Tsukita, S. Claudin multigene family encoding four-transmembrane domain protein components of tight junction strands. Proc. Natl Acad. Sci. USA 96, 511–516 (1999). 17. Tsukita, S. & Furuse, M. Occludin and claudins in tight junction strands: leading or supporting players? Trend. Cell Biol. 9, 268–273 (1999). 18. Morita, K., Sasaki, H., Fujimoto, K., Furuse, M. & Tsukita, S. Claudin-11/OSP-based tight junctions in myelinated sheaths of oligodendrocytes and Sertoli cells in testis. J. Cell Biol. 145, 579–588 (1999). 19. Morita, K., Sasaki, H., Furuse, M. & Tsukita, S. Endothelial claudin: Claudin-5/TMVCF constitutes tight junction strands in endothelial cells. J. Cell Biol. 147, 185–194 (1999). 20. Kubota, K. et al. Ca2+-independent cell adhesion activity of claudins, integral membrane proteins of tight junctions. Curr. Biol. 9, 1035–1038 (1999). 21. Furuse, M., Sasaki, H., Fujimoto, K. & Tsukita, S. A single gene product, claudin-1 or -2, reconstitutes tight junction strands and recruits occludin in fibroblasts. J. Cell Biol. 143, 391–401 (1998). First report that tight-junction strands can be reconstituted within plasma membranes of cultured fibroblasts by a single gene product, claudin-1 or claudin-2. 22. Furuse, M., Sasaki, H. & Tsukita, S. Manner of interaction of heterogeneous claudin species within and between tight junction strands. J. Cell Biol. 147, 891–903 (1999). 23. Tsukita, S. & Furuse, M. Pores in the wall: claudins constitute tight junction strands containing aqueous pores. J. Cell Biol. 149, 13–16 (2000). 24. Martin-Padura, I. et al. Junctional adhesion molecule, a novel member of the immunoglobulin superfamily that distributes at intercellular junctions and modulates

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

monocyte transmigration. J. Cell Biol. 142, 117–127 (1998). Identification and cloning of genes that encode JAM. Palmeri, D., van Zante, A., Huang, C. C., Hemmerich, S. & Rosen, S. D. Vascular endothelial junction-associated molecule, a novel member of the immunoglobulin superfamily, is localized to intercellular boundaries of endothelial cells. J. Biol. Chem. 275, 19139–19145 (2000). Aurrand-Lons, M. A., Duncn, L., Du Pasquire, L. & Imhof, B. A. Cloning of JAM-2 and JAM-3; an emerging junctional adhesion molecular family? Curr. Top. Microbiol. Immunol. 251, 91–98 (2000). Bazzoni, G. et al. Homophilic interaction of junctional adhesion molecule. J. Biol. Chem. 275, 30970–30976 (2000). Liu, Y. et al. Human junction adhesion molecule regulates tight junction resealing in epithelia. J. Cell Sci. 113, 2363–2374 (2000). Claude, P. & Goodenough, D. A. Fracture faces of zonulae occludentes from ‘tight’ and ‘leaky’ epithelia. J. Cell Biol. 58, 390–400 (1973). Claude, P. Morphological factors influencing transepithelial permeability: a model for the resistance of the zonula occludens. J. Membr. Biol. 10, 219–232 (1978). Martinez-Palomo, A. & Erlij, D. Structure of tight junctions in epithelia with different permeability. Proc. Natl Acad. Sci. USA 72, 4487–4491 (1975). Mollgard, K., Malinowski, D. N. & Saunders, N. R. Lack of correlation between tight junction morphology and permeability properties in developing choroid plexus. Nature 264, 293–294 (1976). Stevenson, B. R., Anderson, J. M., Goodenough, D. A. & Mooseker, M. S. Tight junction structure and ZO-1 content are identical in two strains of Madin–Darby canine kidney cells which differ in transepithelial resistance. J. Cell Biol. 107, 2401–2408 (1988). Comprehensive examination of the tightness and structure of tight-junction strands in MDCK I and II cells. Sonoda, N. et al. Clostridium perfringens enterotoxin fragment removes specific claudins from tight junction strands: evidence for direct involvement of claudins in tight

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS junction barrier. J. Cell Biol. 147, 195–204 (1999). 35. Gow, A. et al. CNS myelin and Sertoli cell tight junction strands are absent in OSP/claudin-11 null mice. Cell 99, 649–659 (1999). Generation of mice lacking one claudin gene, confirming the importance of claudin and tight junctions at the level of the whole organism. 36. McCarthy, K. M. et al. Occludin is a functional component of the tight junction. J. Cell Sci. 109, 2287–2298 (1996). 37. McCarthy, K. M. et al. Inducible expression of claudin1–myc but not occludin–VSV-G results in aberrant tight junction strand formation in MDCK cells. J. Cell Sci. 113, 3387–3398 (2000). 38. Cereijido, M., Gonzales–Mariscal, L. & Contreras, G. Tight junction: barrier between higher organisms and environment. News in Physiol. Sci. 4, 72 (1989). 39. Simon, D. B. et al. Paracellin-1, a renal tight junction protein required for paracellular Mg2+ resorption. Science 285, 103–106 (1999). Report that mutations in claudin molecules cause a human hereditary disease, and that claudins might be involved in the formation of aqueous pores within tight-junction strands. 40. Furuse, M., Furuse, K., Sasaki, H. & Tsukita, S. Conversion of zonula occludentes from tight to leaky strand type by introducing claudin–2 into MDCK I cells. J. Cell Biol. (in the press). 41. Balda, M. S. et al. Functional dissociation of paracellular permeability and transepithelial electrical resistance and disruption of the apical–basolateral intramembrane diffusion barrier by expression of a mutant tight junction membrane protein. J. Cell Biol. 134, 1031–1049 (1996). 42. Wong, V. & Gumbiner, B. M. A synthetic peptide corresponding to the extracellular domain of occludin perturbs the tight junction permeability barrier. J. Cell Biol. 136, 399–409 (1997). 43. Saitou, M. et al. Complex phenotype of mice lacking occludin, a component of tight junction strands. Mol. Biol. Cell 11, 4131–4142 (2000). 44. Stevenson, B. R., Siliciano, J. D., Mooseker, M. S. & Goodenough, D. A. Identification of ZO-1: a high molecular weight polypeptide associated with the tight junction (zonula occludens) in a variety of epithelia. J. Cell Biol. 103, 755–766 (1986). Identification and characterization of ZO-1 as a first component of tight junctions. 45. Gumbiner, B., Lowenkopf, T. & Apatira, D. Identification of a 160-kDa polypeptide that binds to the tight junction protein ZO-1. Proc. Natl Acad. Sci. USA 88, 3460–3464 (1991). 46. Balda, M. S., González-Mariscal, L., Matter, K., Cereijido, M. & Anderson, J. M. Assembly of the tight junction: the role of diacylglycerol. J. Cell Biol. 123, 293–302 (1993). 47. Itoh, M. et al. The 220-kD protein colocalizing with cadherins in non-epithelial cells is identical to ZO-1, a tight junction-associated protein in epithelial cells: cDNA cloning and immunoelectron microscopy. J. Cell Biol. 121, 491–502 (1993). 48. Willott, E. et al. The tight junction protein ZO-1 is homologous to the Drosophila discs-large tumor suppresser protein of septate junctions. Proc. Natl Acad. Sci. USA 90, 7834–7838 (1993). 49. Jesaitis, L. A. & Goodenough, D. A. Molecular characterization and tissue distribution of ZO-2, a tight junction protein homologous to ZO-1 and the Drosophila discs-large tumor suppresser protein. J. Cell Biol. 124, 949–961 (1994). 50. Haskins, J., Gu, L., Wittchen, E. S., Hibbard, J. & Stevenson, B. R. ZO-3, a novel member of the MAGUK protein family found at the tight junction, interacts with ZO1 and occludin. J. Cell Biol. 141, 199–208 (1998). 51. Itoh, M. et al. Direct binding of three tight junctionassociated MAGUKs, ZO-1, ZO-2, and ZO-3, with the COOH termini of claudins. J. Cell Biol. 147, 1351–1367 (1999). 52. Furuse, M. et al. Direct association of occludin with ZO-1 and its possible involvement in the localization of occludin at tight junctions. J. Cell Biol. 127, 1617–1626 (1994). 53. Itoh, M., Morita, K. & Tsukita, S. Characterization of ZO-2 as a MAGUK family member associated with tight and adherens junctions with a binding affinity to occludin and α-catenin. J. Biol. Chem. 274, 5981–5986 (1999). 54. Fanning, A. S., Jameson, B. J., Jesaitis, L. A. & Anderson, J. M. The tight junction protein ZO-1 establishes a link between the transmembrane protein occludin and the actin cytoskeleton. J. Biol. Chem. 273, 29745–29753 (1998). 55. Bazzoni, G. et al. Interaction of junctional adhesion molecule with the tight junction components ZO-1,

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

66.

67. 68. 69.

70.

71.

72.

73.

74.

75.

76.

77.

78.

cingulin, and occludin. J. Biol. Chem. 275, 20520–20526 (2000). Ebnet, K., Schulz, C. U., Meyer Zu Brickwedde, M. K., Pendle, G. G. & Vestweber, D. Junctional adhesion molecule interacts with the PDZ domain-containing proteins AF-6 and ZO-1. J. Biol. Chem. 275, 27979–27988 (2000). Martinez-Estrada, O. M. et al. Association of junctional adhesion molecule with calcium/calmodulin-dependent serine protein kinase (CASK/LIN-2) in human epithelial Caco-2 cells. J. Biol. Chem. (in the press). Dobrosotskaya, I., Guy, R. K. & James, G. L. MAGI-1, a membrane-associated guanylate kinase with a unique arrangement of protein-protein interaction domains. J. Biol. Chem. 272, 31589–31597 (1997). Ide, N. et al. Localization of membrane-associated guanylate kinase (MAGI)-1/BAI-associated protein (BAP) 1 at tight junctions of epithelial cells. Oncogene 18, 7810–7815 (1999). Wu, X. et al. Evidence for regulation of the PTEN tumor suppressor by a membrane-localized multi-PDZ domain containing scaffold protein MAGI-2. Proc. Natl Acad. Sci. USA 97, 4233–4238 (2000). Wu, X. et al. Interaction of the tumor suppressor PTEN/MMAC with a PDZ domain of MAGI3, a novel membrane-associated guanylate kinase. J. Biol. Chem. 275, 21477–21485 (2000). Izumi, Y. et al. An atypical PKC directly associates and colocalizes at the epithelial tight junction with ASIP, a mammalian homologue of Caenorhabditis elegans polarity protein PAR-3. J. Cell Biol. 143, 95–106 (1998). Joberty, G., Petersen, C., Gao, L. & Macara, I. G. The cellpolarity protein Par6 links Par3 and atypical protein kinase C to Cdc42. Nature Cell Biol. 2, 531–539 (2000). Lin, D. et al. A mammalian PAR-3–PAR-6 complex implicated in Cdc42/Rac1 and aPKC signalling and cell polarity. Nature Cell Biol. 2, 540–547 (2000). Qiu, R. G., Abo, A. & Steven Martin, G. A human homolog of the C. elegans polarity determinant par-6 links rac and cdc42 to PKCζ signaling and cell transformation. Curr. Biol. 10, 697–707 (2000). Itoh, M., Nagafuchi, A., Moroi, S. & Tsukita, S. Involvement of ZO-1 in cadherin-based cell adhesion through its direct binding to α catenin and actin filaments. J. Cell Biol. 138, 181–192 (1997). Hata, Y., Nakanishi, H. & Takai, Y. Synaptic PDZ domaincontaining proteins. Neurosci. Res. 32, 1–7 (1998). Giancotti, F. G. & Ruoslahti, E. Integrin signaling. Science 285, 1028–1032 (1999). Weber, E. et al. Expression and polarized targeting of a rab3 isoform in epithelial cells. J. Cell Biol. 125, 583–594 (1994). Zahraoui, A. et al. A small rab GTPase is distributed in cytoplasmic vesicles in non polarized cells but colocalizes with the tight junction marker ZO-1 in polarized epithelial cells. J. Cell Biol. 124, 101–115 (1994). Grindstaff, K. K. et al. Sec6/8 complex is recruited to cell–cell contacts and specifies transport vesicle delivery to the basal–lateral membrane in epithelial cells. Cell 93, 731–740 (1998). Indicates that the tight junction region might function as a specific site for the polarized delivery of exocytic vesicles during epithelial cell polarization. Forstner, G. G. & Wherrett, J. R. Plasma membrane and mucosal glycosphingolipids in the rat intestine. Biochim. Biophys. Acta 306, 446–459 (1973). Chapelle, S. & Gilles-Baillien, M. Phospholipids and cholesterol in brush border and basolateral membranes from rat intestinal mucosa. Biochim. Biophys. Acta 753, 269–271 (1983). Barsukov, L. I., Bergelson, L. D., Spiess, M., Hauser, H. & Semenza, G. Phospholipid topology and flip-flop in intestinal brush-border membrane. Biochim. Biophys. Acta 862, 87–99 (1986). Rothman, J. E., Tsai, D. K., Dawidowicz, E. A. & Lenard, J. Transbilayer phospholipid asymmetry and its maintenance in the membrane of influenza virus. Biochemistry 15, 2361–2370 (1976). Dragsten, P. R., Blumenthal, R. & Handler, J. S. Membrane asymmetry in epithelia: is the tight junction a barrier to diffusion in the plasma membrane? Nature 294, 718–722 (1981). van Meer, G. & Simon, K. The function of tight junctions in maintaining differences in lipid composition between the apical and basolateral cell surface domains of MDCK cells. EMBO J. 5, 1455–1464 (1986). Nelson, W. J. Regulation of cell surface polarity from bacteria to mammals. Science 258, 948–954 (1992).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

79. Pisam, M. & Ripoche, P. Redistribution of surface macromolecules in dissociated epithelial cells. J. Cell Biol. 71, 909–920 (1976). 80. Ziomek, C. A., Shulman, S. & Edidin, M. Redistribution of membrane proteins in isolated mouse intestinal epithelial cell. J. Cell Biol. 86, 849–857 (1980). 81. Vega-Salas, D. E., Salas, P. J. I., Gundersen, D. & Rodriguez-Boulan, E. Formation of the apical pole of epithelial (Madin–Darby canine kidney) cells: polarity of an apical protein is independent of tight junctions while segregation of a basolateral marker requires cell–cell interaction. J. Cell Biol. 104, 905–916 (1987). 82. Jou, T.-S., Schneeberger, E. E. & Nelson, W. J. Structural and functional regulation of tight junctions by RhoA and Rac1 small GTPases. J. Cell Biol. 142, 101–115 (1998). 83. Tsukita, S., Furuse, M. & Itoh, M. Structural and signaling molecules come together at tight junctions. Curr. Opin. Cell Biol. 11, 628–633 (1999). 84. Mankertz, J. et al. Expression from the human occludin promoter is affected by tumor necrosis factor α and interferon γ. J. Cell Sci. 113, 2085–2090 (2000). 85. Chen, Yh., Lu, Q., Schneeberger, E. E. & Goodenough, D. A. Restoration of tight junction structure and barrier function by down-regulation of the mitogen-activated protein kinase pathway in ras-transformed Madin-Darby canine kidney cells. Mol. Biol. Cell 11, 849–862 (2000). 86. Li, D. & Mrsny, R. J. Oncogenic Raf-1 disrupts epithelial tight junctions via downregulation of occludin. J. Cell Biol. 148, 791–800 (2000). 87. Sakakibara, A., Furuse, M., Saitou, M., Ando-Akatsuka, Y. & Tsukita, S. Possible involvement of phosphorylation of occludin in tight junction formation. J. Cell Biol. 137, 1393–1401 (1997). 88. Wilcox, E. R. et al. Mutations in the gene encoding tight junction claudin-14 cause autosomal recessive deafness DFNB29. Cell 104, 165–172 (2001). 89. Kinugasa, T., Sakaguchi, T. & Reinecker, H. C. Claudins regulate the intestinal barrier in response to immune mediators. Gastroenterology 118, 1001–1011 (2000). 90. Hough, C. D. et al. Large-scale serial analysis of gene expression reveals genes differentially expressed in ovarian cancer. Cancer Res. 60, 6281–6287 (2000). 91. Madara, J. L. Tight junction dynamics: Is paracellular transport regulated? Cell 53, 497–498 (1988). 92. Powell, D. W. Barrier function of epithelia. Am. J. Physiol. 241, G275–G288 (1981). 93. Madara, J. L. Regulation of the movement of solutes across tight junctions. Annu. Rev. Physiol. 60, 143–159 (1998). 94. Briehl, M. M. & Miesfeld, R. L. Isolation and characterization of transcripts induced by androgen withdrawal and apoptotic cell death in the rat ventral prostate. Mol. Endocrinol. 5, 1381–1388 (1991). 95. Katahira, J., Inoue, N., Horiguchi, Y., Matsuda, M. & Sugimoto, N. Molecular cloning and functional characterization of the receptor for Clostridium perfringens enterotoxin. J. Cell Biol. 136, 1239–1247 (1997). 96. Sirotkin, H. et al. Identification, characterization, and precise mapping of a human gene encoding a novel membrane-spanning protein from the 22q11 region deleted in velo-cardio-facial syndrome. Genomics 42, 245–251 (1997). 97. Bronstein, J. M., Popper, P., Micevych, P. E. & Farber, D. B. Isolation and characterization of a novel oligodendrocytespecific protein. Neurology 47, 772–778 (1996). 98. Citi, S., Sabanay, H., Jakes, R., Geiger, B. & KendrickJones, J. Cingulin, a new peripheral component of tight junctions. Nature 333, 272–276 (1988). 99. Cordenonsi, M. et al. Cingulin contains globular and coiled-coil domains and interacts with ZO-1, ZO-2, ZO-3, and myosin. J. Cell Biol. 147, 1569–1582 (1999). 100. Zhong, Y. et al. Monoclonal antibody 7H6 reacts with a novel tight junction-associated protein distinct from ZO-1, cingulin, and ZO-2. J. Cell Biol. 120, 477–483 (1993). 101. Keon, B. H., Schäfer, S., Kuhn, C., Grund, C. & Franke, W. W. Symplekin, a novel type of tight junction plaque protein. J. Cell Biol. 134, 1003–1018 (1996). 102. Saha, C., Nigam, S. K. & Denker, B. M. Involvement of Gαi2 in the maintenance and biogenesis of epithelial cell tight junctions. J. Biol. Chem. 273, 21629–21633 (1998). 103. Balda, M. S. & Matter, K. The tight junction protein ZO-1 and an interacting transcription factor regulate ErbB-2 expression. EMBO J. 19, 2024–2033 (2000). 104. Nakamura, T. et al. huASH1 protein, a putative transcription factor encoded by a human homologue of the Drosophila ash1 gene, localizes to both nuclei and cell–cell tight junctions. Proc. Natl Acad. Sci. USA 97, 7284–7289 (2000).

VOLUME 2 | APRIL 2001 | 2 9 3

REVIEWS

CBL: MANY ADAPTATIONS TO REGULATE PROTEIN TYROSINE KINASES Christine B. F. Thien and Wallace Y. Langdon Responses to extracellular stimuli are often transduced from cell-surface receptors to protein tyrosine kinases which, when activated, initiate the formation of protein complexes that transmit signals throughout the cell. A prominent component of these complexes is the product of the proto-oncogene c-Cbl, which specifically targets activated protein tyrosine kinases and regulates their signalling. How, then, does this multidomain protein shape the responses generated by these signalling complexes?

UBIQUITIN-CONJUGATING ENZYME (E2)

An enzyme that accepts ubiquitin from a ubiquitin-activating enzyme and transfers it to a ubiquitin ligase which, in turn, transfers it to a substrate protein. MULTI-UBIQUITYLATION

The sequential addition of the small protein ubiquitin to a target protein, to form a chain of isopeptide-linked ubiquitin molecules. This is a signal for proteolytic cleavage in the proteasome. BONE RESORPTION

The digestion of mineralized bone tissue by specialized cells called osteoclasts.

Department of Pathology, University of Western Australia, 35 Stirling Highway, Crawley, Western Australia 6009, Australia. Correspondence to W.Y.L. e-mails: wlangdon@cyllene. uwa.edu.au canitbe@cyllene.uwa.edu.au

294

The c-Cbl proto-oncogene and other members of the Cbl family (FIG. 1) have created much interest in recent years, not only because some altered forms of c-Cbl promote tumorigenesis but also because of three findings that relate to their structure and function. First, a Cbl protein was identified in the nematode Caenorhabditis elegans that can negatively regulate signalling from a receptor tyrosine kinase (RTK); second, all Cbl proteins have a unique domain that recognizes phosphorylated tyrosine residues that are present on activated tyrosine kinases; and third, Cbl proteins recruit UBIQUITIN-CONJUGATING ENZYMES (E2s or ubiquitin-carrier enzymes (UBCs)) to, and directs MULTI-UBIQUITYLATION and degradation of, activated RTKs. Cbl proteins therefore function by specifically targeting activated RTKs and mediating their downregulation, thus providing a means by which signalling processes can be negatively regulated. However, there is evidence that Cbl is also involved in positive signalling events through its function as a multidomain adaptor protein that is required for functions associated with BONE RESORPTION, glucose uptake in response to insulin, and cell spreading in response to integrin engagement. In this review, we address these diverse aspects of Cbl function, as well as describing Cbl proteins as E2-dependent ubiquitin protein ligases and the ability of some altered forms to promote tumorigenesis.

Discovery of Cbl in a mouse retrovirus

The c-Cbl proto-oncogene was first identified through studies of haematopoietic tumours of mice infected with the ECOTROPIC Cas-Br-M retrovirus. These studies led to the isolation of a recombinant retrovirus that induced pre- and pro-B lymphomas and the transformation of rodent fibroblasts. The causative retrovirus was termed Cas NS-1 and its oncogene named v-Cbl, for Casitas Blineage lymphoma1. Subsequent cloning of the mouse c-Cbl gene revealed that v-Cbl encoded only the first 355 amino acids of the 913 that are present in the complete protein, and that overexpression of wild-type c-Cbl did not promote tumorigenesis2. The 120-kDa c-Cbl protein is ubiquitously expressed and localized to the cytoplasm, with highest expression levels in haematopoietic cells and testis. Domains of Cbl proteins

Since the identification of c-Cbl, two additional mammalian homologues (Cbl-b and Cbl-3) have been found, as well as their invertebrate orthologues in Drosophila melanogaster (D-Cbl, which has long and short isoforms) and C. elegans (SLI-1)3â&#x20AC;&#x201C;8 (FIG. 1a). These proteins share highly conserved amino-terminal regions that encompass the sequences found in v-Cbl (FIG. 1b), plus a small linker domain and a zinc-binding C3HC4 RING FINGER. The v-Cbl region can independently bind activated RTKs, such as the epidermal growth factor (EGF) and www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

ÂŠ 2001 Macmillan Magazines Ltd

REVIEWS a c-Cbl

TKB domain N–

RING

SH2

Proline-rich

UBA/LZ –C

906 Cbl-b N –

SH2

–C

982 Cbl-3 N –

SH2

–C

474 D-Cbl N – (short)

D-Cbl N – (long)

SLI-1

SH2

–C

448 EF

SH2

–C

878

N–

SH2

–C 582

TKB domain

70Z-Cbl N– (∆366–382)

889

SH2

L Y371

v-Cbl

N–

UBA/LZ –C

Y368 ∆Y371-Cbl N –

Proline-rich

17aa

L ∆Y368-Cbl N –

RING

SH2

–C 906

SH2 – C 357

Figure 1 | The Cbl protein family. a | The main domains common to Cbl proteins in mammals (c-Cbl, Cbl-b and Cbl-3), Drosophila melanogaster (D-Cbl, long and short) and Caenorhabditis elegans (SLI-1) are shown. All Cbl proteins have a high degree of sequence homology between their tyrosine kinase-binding (TKB), linker and RING finger domains, and most have extensive proline-rich regions (indicated by red bars or stripes) in their carboxy-terminal halves. The TKB domain is composed of three interacting domains comprising a four-helix bundle (4H), a Ca2+binding EF hand (EF), and a variant Src homology region 2 (SH2) domain15 that is connected to the RING finger by a short linker domain (L). The ubiquitin-associated (UBA)/LZ domain at the carboxyl terminus of c-Cbl, Cbl-b and D-Cbl (long) refers to a region with homology to ubiquitin-associated domains and leucine zippers. b | c-Cbl can be converted to an oncogenic protein either by introducing mutations within the α-helix of the linker domain that separates the SH2 and RING finger domains (702-Cbl, ∆Y368-Cbl and ∆Y371-Cbl) or by a large truncation that removes all sequences except for the TKB domain (v-Cbl). All oncogenic Cbl proteins found so far require a functional TKB domain to target activated tyrosine kinases. The linker-domain mutants transform cells more rapidly and with greater efficiency than v-Cbl.

ECOTROPIC

A classification for viruses that only replicate in cells of the species from which they were derived, for example, mouse ecotropic retroviruses replicate in mouse but not foreign cells. C3HC4 RING FINGER

A protein domain containing two interleaved zinccoordinating sites. In the first site, the zinc is coordinated by three cysteines and a histidine (C3H), whereas in the second it is coordinated by four cysteines (C4).

platelet-derived growth factor (PDGF) receptors, as well as the non-receptor tyrosine kinases ζ-chain-associated protein kinase 70 (ZAP-70) and Syk9–14 (FIG. 2). Binding is mediated by three domains comprising a four-helix bundle, a calcium-binding EF hand domain and a variant Src homology 2 (SH2) domain. Together, these domains recognize specific phosphotyrosine residues on activated tyrosine kinases15. As all three domains are required to form this unique phosphotyrosine-binding module, the entire region is commonly referred to as a tyrosine-kinase-binding (TKB) domain. The TKB domain’s crystal structure was resolved as a complex with a phosphopeptide representing its binding site in ZAP-70, to which it binds in a manner similarly to other

SH2–phosphopeptide complexes, with the specificitydetermining interactions lying carboxy-terminal to the phosphotyrosine15. Determining the purpose of TKB domain-targeting of activated tyrosine kinases is an important key to understanding Cbl function. Two other regions are highly conserved among all Cbl proteins; the RING finger and a short linker sequence that connects the TKB and RING finger domains. The RING finger domain has recently been shown to recruit UBCs16–20, and this observation was an important component in the discovery that Cbl proteins function as UBIQUITIN PROTEIN LIGASES (E3s). The only other protein known to bind the Cbl RING finger is Sprouty, a known antagonist of signalling from the fibroblast growth factor (FGF) and EGF receptors21–23. The sequences that lie carboxy-terminal to the RING finger contain a small acidic-rich domain in c-Cbl, Cblb and D-Cbl that is absent in Cbl-3 and SLI-1 and has no known function. Following this domain, the sequence homology among Cbl proteins is less extensive; however, with the exception of D-Cbl (short), they all have proline-rich regions that are involved in SH3DOMAIN interactions. The carboxy-terminal half of c-Cbl has an abundance of proline residues, which contribute 15 potential SH3-domain-binding motifs. The SH3domain-containing proteins that interact with c-Cbl are shown in FIG. 2. Cbl-b also has an extensive proline-rich region and shares three SH3-domain-binding motifs with c-Cbl. Within this proline-rich region of c-Cbl is a small serine-rich domain that contains two 14-3-3 proteinbinding motifs24. 14-3-3 proteins are ubiquitously expressed proteins that bind serine or threonine-phosphorylated target proteins implicated in the activation of protein kinases and the control of the cell cycle and apoptosis. The interaction with 14-3-3 requires both cCbl motifs and is dependent on serine phosphorylation by protein kinase C, which occurs after T-cell-receptor (TCR) stimulation25. The second motif is not present in Cbl-b and is the probable reason for the inability of Cblb to interact with 14-3-3 proteins. The role of c-Cbl’s association with 14-3-3 proteins remains a mystery; however, serine phosphorylation of c-Cbl does inhibit its tyrosine phosphorylation and might therefore regulate interactions with signalling intermediates such as SH2-domain-containing proteins26,27. c-Cbl is a prominent substrate of tyrosine kinases and is phosphorylated after the stimulation of a diverse array of cell-surface receptors (BOX 1). c-Cbl has 22 tyrosines and those at 700, 731 and 774 are the principal phosphorylation sites28. These sites are efficiently phosphorylated by Syk and the Src-family kinases Fyn, Yes and Lyn, but not by Lck and ZAP-70 (REFS 14,28,29). Phosphorylation of Y700 and Y774 provide docking sites for the SH2 domain of CRK — an adaptor protein that forms prominent associations with c-Cbl30–34. Y700 might also be recognized by the SH2 domain of the Abl tyrosine kinase35. Phosphorylation of Y700 also mediates an interaction with the SH2 domain of Vav1, a guanine nucleotide exchange factor for the Rho family GTPases Rac and Cdc42 (REF. 36). Like c-Cbl, Cbl-b is a

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 9 5

REVIEWS

p85

Sprouty Ub-conjugating enzymes (E2s)

PI(3)K

SH2

14-3-3

CMS SH3 P

TKB domain N–

RING

Proline-rich

SH2 L

Y 774

P 323 Y Syk 292 Y ZAP-70

UBA/LZ –C 906

SH3

731

S Y Y 700

Grb2, Nck

1045 Y EGFR

SH2 SH3

CrkL

Fyn, Src P Y PDGFR

SH3

SH2

CAP CIN85/SETA

Vav

Figure 2 | c-Cbl interacts with many signalling proteins. Some of these interactions occur after the stimulation of cell-surface receptors (BOX 1), which transduce signals into cells by activating tyrosine kinases. This results in recruitment of c-Cbl and many other signalling proteins to activated receptors. The tyrosine kinase-binding (TKB) domain can target phosphorylated tyrosines on tyrosine kinases, which allows their activity to be downregulated by c-Cbl-associated proteins such as ubiquitin (Ub)-conjugating enzymes recruited by the RING finger. c-Cbl also becomes phosphorylated on tyrosine (Y) and serine (S) residues, which promotes associations with Src homology region 2 (SH2)-domain-containing proteins and 143-3 proteins, respectively. Some associations are constitutive, such as those with the numerous SH3-domain-containing proteins that bind one or more of the 15 SH3-binding motifs within the extensive proline-rich regions. (CAP, Cbl-associated protein; CIN c-Cblinteracting protein; CMS, Cas ligand with multiple SH3 domains; EGFR, epidermal growth factor receptor; PDGFR, platelet-derived growth factor receptor; PI(3)K, phosphatidylinositol-3OH kinase; SETA, SH3-encoding, expressed in tumorigenic astrocytes; UBA, ubiquitinassociated; ZAP-70, ζ-chain-associated protein kinase 70.)

UBIQUITIN PROTEIN LIGASE (E3)

An enzyme that couples the small protein ubiquitin to lysine residues on a target protein, marking that protein for destruction by the proteasome. SH3 DOMAINS

(Src-homology region 3). Protein sequence of about 50 amino acids that recognizes and binds sequences rich in proline. CRK

Adaptor protein first described as the product of an avian oncogene, v-Crk, that contains an amino-terminal SH2 domain and two SH3 domains that function as binding sites for a diverse set of signalling proteins. VULVAL INDUCTION

A readily accessible genetic system in Caenorhabditis elegans for studying the induction and regulation of epidermal growth factor receptor signalling pathways in vivo.

296

substrate of tyrosine kinases37 and has tyrosines at 709 and 655 with flanking sequences that are homologous to Y700 and Y774, respectively. It is therefore likely that Cbl-b is a substrate for the same tyrosine kinases and binds identical SH2 domains as c-Cbl. The third major tyrosine phosphorylation site, Y731, is unique to c-Cbl and does not share flanking homology with Y700 or Y774. Tyrosine 731 has a downstream sequence (EAM) that provides a docking site for the SH2 domains of the p85 regulatory subunit of phosphatidylinositol-3-OH kinase (PI(3)K)38–41. Tyrosines with flanking sequences similar to Y700, Y731 or Y774 do not occur in Cbl-3, D-Cbl or Sli-1, indicating that c-Cbl, and Cbl-b might have additional functions not associated with Cbl-3 or Cbl proteins in less complex organisms. The carboxyl termini of c-Cbl, Cbl-b and D-Cbl (long) encompass a conserved domain known as a ubiquitin-associated (UBA) domain. The UBA domain was identified by sequence homology to regions found in subsets of UBCs, E3 ligases and ubiquitin hydrolases42. The discovery of a UBA domain is of interest because c-Cbl has been shown to undergo ubiquitylation43 and its overexpression directs RTK multi-ubiquitylation16,44–46. Whether this region has a direct role in Cbl’s function as an E3 ligase has not been investigated. However, truncated Cbl mutants with only the TKB and RING finger domains still efficiently promote RTK multi-ubiquitylation18,47. The carboxy-terminal region

of c-Cbl was also predicted to form a leucine zipper as it shows sequence similarity to domains found in DNAbinding proteins that dimerize through leucine sidechains2. c-Cbl can form homodimers through this domain and its deletion decreases c-Cbl tyrosine phosphorylation and its association with the EGF receptor48. SLI-1: a negative regulator of RTKs

Evidence that Cbl proteins have functions regulating tyrosine kinase signalling initially came from genetic studies in C. elegans 8,49. Loss-of-function mutations in sli-1 restored signalling for VULVAL INDUCTION and survival through a weakly active EGF receptor homologue (LET-23), whereas introducing additional copies of sli-1 suppressed vulval induction. These experiments indicated that SLI-1 is a negative regulator of LET-23 that functions upstream of Ras (known as LET-60 in C. elegans) at the level of LET-23 and the Grb2 adaptor protein SEM-5 (FIG. 3a,b). Studies in Drosophila also suggest that Cbl can negatively regulate RTK signalling: D-Cbl suppresses the development of R7 PHOTORECEPTOR CELLS in

Box 1 | Upstream receptors Antigen/immunoglobulin receptors • T-cell receptor • B-cell receptor • CD38 • CD19 • CD16 • CD5 • CD2 • CD26 • FcγRI, II, III • FcεRI Growth factor receptors • Epidermal growth factor receptor • Platelet-derived growth factor receptor • Nerve growth factor receptor • Fibroblast growth factor receptor • Hepatocyte growth factor receptor • TrkB receptor Cytokine receptors • Interleukin-2 receptor (IL-2R) • IL-3R • IL-4R • Erythropoietin receptor • Thrombopoietin receptor • Colony stimulating factor-1 receptor • Granulocyte-macrophage colony-stimulating factor receptor • Steel factor receptor (c-Kit) • Interferon-α receptor • FLT-3 receptor Hormone receptors • Insulin receptor • Prolactin receptor Integrins • α5β1 • αVβ3 • β1

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

b Mutant LET-23 (sy97)

LIN-3 (EGF)

Anchor cell

Kinase

Vulval precursor cell

LET-341 (Sos)

LET-23 (EGFR)

Y Y Y Y Y

LET-60 (Ras)

SEM5 (Grb2)

LIN-45 (Raf)

Kinase

SLI-1 (wt)

SLI-1–receptor association Weak signal

SLI-1 (Cbl)

Mutant LET-23 (sy97)

Y Y Y Y Y

SLI-1 (G315E)

No SLI-1–receptor association Weak signal enhanced

SUR1/MPK1 (MAPK) Vulval differentiation

No vulval development and reduced viability

Signalling for vulval development and survival is restored

Figure 3 | The genetic pathway of vulval induction in Caenorhabditis elegans is negatively regulated by sli-1. a | The vulval differentiation pathway is initiated by an anchor cell producing localized ligand (LIN-3, an epidermal growth factor (EGF) homologue), which activates LET-23 (EGF receptor, EGFR), resulting in downstream signalling through LET-60 (Ras) and LIN-45 (Raf). Loss-of-function mutations at the sli-1 locus suppress severe reduction-of-function mutations in let-23 and sem-5, whereas mutations in lin-3, let-60 or lin-45 are not rescued. These experiments identified SLI-1(Cbl) proteins as negative regulators of receptor tyrosine kinase (RTK) signalling that function in a ligand-dependent manner upstream of Ras at the level of LET-23 and SEM-5 (Grb2). Diagram adapted from REF. 49. b | The G315E loss-of-function mutation in sli-1 rescues vulval induction in C. elegans expressing a severe reduction-of-function mutation in let-23 (sy97). More than 80% of worms that are homozygous for the let-23 (sy97) mutation die, and those that escape show little vulval differentiation and are infertile. This mutation produces a carboxy-terminally truncated receptor that has lost three potential SEM-5-binding sites and is therefore markedly perturbed in its ability to promote survival, vulval development and egg-laying. It is proposed that, although wild-type (wt) SLI-1 can no longer be recruited by SEM-5, it can still associate with the mutant receptor through its tyrosine-kinase-binding (TKB) domain and so suppresses weak signals that are below the threshold for inducing biological responses. However, introducing loss-of-function mutations, such as G315E (in the variant SH2 domain), into the sli-1 locus restores signalling to a level at which vulval differentiation and survival can now occur8. Structural studies and analyses of the equivalent mutation in c-Cbl (G306E) indicate that this would disrupt TKB domain binding to phosphotyrosine, thus it is likely that SLI-1(G315E) would be unable to target LET-23. Consequently, signalling for vulval differentiation and survival reverts to wild-type levels. (MAPK, mitogen-activated protein kinase).

flies on a sensitized genetic background5 and functions as a negative regulator of a dose-sensitive EGF receptor pathway involved in dorsoventral patterning during oogenesis50. Do Cbl proteins in mammals also negatively regulate tyrosine kinases and, if so, what are the biochemical mechanisms involved? RTK downregulation by multi-ubiquitylation

R7 PHOTORECEPTOR CELLS

A readily accessible genetic system in Drosophila melanogaster for studying the induction and regulation of epidermal growth factor receptor signalling pathways in vivo.

The demonstration of a prominent and inducible association between Cbl proteins and the activated EGF receptor in mammalian cells clearly indicated that genetic studies in C. elegans had correctly predicted the site of Cbl action4,9,51–53. In addition, introducing a corresponding loss-of-function mutation from SLI-1 into c-Cbl (a glycine to glutamic acid substitution at residue 306) blocks interaction of the TKB domain with the EGF receptor11,12 (FIG. 3b). This mutation occurs near the phosphotyrosine-binding pocket of c-Cbl’s variant SH2 domain and is predicted to prevent its interaction with phosphotyrosine15. Furthermore, the identification of specific phosphotyrosine residues within the EGF receptor (pY1045)18, Syk (pY323)13,14 and ZAP-70 (pY292)54 that are recognized by Cbl TKB domains has supported the proposal that Cbl is involved in their regulation. The recognition of pY292 in ZAP-70 is particu-

larly pertinent because this residue has been characterized as a negative regulatory site55,56. An important breakthrough in identifying a regulatory mechanism that could explain the function of SLI1 came from observing the profound effect of c-Cbl overexpression on the promotion of EGF, PDGF and colony-stimulating factor 1 (CSF-1) receptor multiubiquitylation (BOX 2) and downregulation44–46. Receptor multi-ubiquitylation was found to be directly mediated by c-Cbl and dependent on the integrity of both the TKB and RING finger domains16,47. Furthermore, CSF-1 fails to stimulate CSF-1 receptor multi-ubiquitylation in c-Cbl–/– macrophages46. So, c-Cbl can promote the multi-ubiquitylation of activated RTKs, which targets them for degradation and prevents their recycling from early endosomes to the cell surface45,47 (FIG. 4). Precisely where Cbl acts on RTKs is still unclear as some studies indicate that receptor multi-ubiquitylation might occur in endosomal compartments57,58, whereas others indicate that this occurs at the plasma membrane before internalization46,59. Whether Cbl also regulates receptor internalization remains uncertain: its overexpression does not enhance EGF receptor internalization45,60 but the internalization of the CSF-1 receptor in c-Cbl–/– macrophages is delayed, although not blocked46.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 9 7

REVIEWS The development of in vitro reconstituted ubiquitylation assays allowed researchers to show that the c-Cbl RING finger has intrinsic E3 ligase activity and can independently recruit UBCs and direct ubiquitin transfer to substrates17–19. Furthermore, the structure of c-Cbl bound to the E2 enzyme UBCH7 has been solved, identifying multiple contacts between c-Cbl’s RING finger and linker domain and UBCH7 (REF. 20). These studies resulted in the definition of Cbl proteins as RING-type E3 ubiquitin protein ligases that direct RTK polyubiquitylation and downregulation. However, Cbl’s E3 activity is not solely restricted to RTKs or to TKB-interacting proteins as it was recently shown that Cbl-b overexpression can promote multi-ubiquitylation of the p85 subunit of PI(3)K. The Cbl-b–p85 interaction was not TKB dependent but mediated by proline-rich sequences in Cbl-b and the p85 SH3 domain61. Although c-Cbl overexpression clearly promotes EGF and PDGF receptor multi-ubiquitylation, there is some conflict over the downstream effects of c-Cbl on growth-factor responses. In some studies in NIH-3T3 fibroblasts, c-Cbl overexpression has no effect on proliferation11,62,63 whereas in another study, when cells were examined at low passage after transfection, c-Cbl is suppressive64. Micro-injection of NIH 3T3 fibroblasts with a c-Cbl expression plasmid inhibits PDGF and EGFinduced DNA synthesis, which supports a negative regulatory role65. At present, the effect of c-Cbl overexpression on the growth of fibroblasts does not seem as clear-cut as the profound effect it has in promoting RTK multi-ubiquitylation. c-Cbl’s negative regulation of cell growth is, however, more apparent in cells transformed by the ErbB2 oncogene, in which c-Cbl-directed multi-ubiquitylation and downregulation of this RTK suppress tumour formation58. Regulation of non-receptor tyrosine kinases

The discovery that c-Cbl is an E3 enzyme raised a key question: do all Cbl proteins primarily function as negative regulators by targeting active pools of tyrosine kinases for degradation? This is plausible, but whether degradation of tyrosine kinases is the universal mechanism for negative regulation by Cbl is not fully resolved. This uncertainty is due, in part, to conflicting results involving the non-receptor tyrosine kinases Syk and ZAP-70: in some cell types, c-Cbl overexpression suppresses the activity of these kinases without reducing their levels66,67 whereas in others c-Cbl overexpression enhances their degradation13,67,68. Conversely, in c-Cbldeficient B cells and thymocytes there is no increase in Syk or ZAP-70 protein, respectively, compared with c-Cbl-expressing cells69–72, but the activity of ZAP-70 is markedly enhanced in c-Cbl-deficient thymocytes. It remains to be determined whether c-Cbl can direct the multi-ubiquitylation of Syk or ZAP-70, so the mechanism of c-Cbl’s negative regulation of these kinases remains unsolved. c-Cbl is a prominent binding partner and substrate of Src family kinases, so does c-Cbl also negatively regulate their activity? Antisense experiments show that cCbl has a positive role downstream of Src in a signalling

298

Box 2 | Multi-ubiquitylation Multi-ubiquitylation is the process of conjugating several ubiquitin molecules to a substrate protein and often serves to tag proteins for degradation by the proteasomal/lysosomal pathway. Initially the ubiquitinactivating (E1) enzyme forms a thiol ester bond with a ubiquitin molecule before transferring it to ubiquitinconjugating (E2 or ubiquitin-carrier enzyme) enzymes. In most cases, the final step involves ubiquitin ligases (E3) that complex with specific protein substrates and facilitate the transfer of ubiquitin from the E2 to the target protein. Monoubiquitylation is thought to be sufficient as an internalization signal108,109 whereas addition of four or more ubiquitin moieties is required for recognition by the 26S proteasome110. Ubiquitylation can target proteins for degradation by either proteasomes or lysosomes111,112. For a more detailed description of ubiquitylation, see the March issue of Nature Reviews Molecular Cell Biology.

pathway necessary for bone resorption in osteoclasts73, whereas deletion of c-Src or c-Cbl leads to a decrease in osteoclast migration74 and c-Cbl is also required for Srcfamily kinases to mediate spreading and migration of macrophages75. Evidence from these cells therefore indicates that c-Cbl is a positive effector of Src. However, Sanjay et al. also show that c-Cbl can inhibit c-Src kinase activity in a TKB-dependent manner by binding pY416 of activated Src74. Interestingly, both 70Z-Cbl and v-Cbl, which lack an intact RING finger, were equally effective in inhibiting Src kinase activity, indicating that this is not dependent on c-Cbl’s E3 activity. By contrast, biochemical studies have shown that c-Cbl overexpression reduces the active pool of Fyn by enhancing its degradation, whereas increased Fyn levels were found in c-Cbl deficient T cells and fibroblasts76. Whether this is due to c-Cbl directed multi-ubiquitylation of Fyn is not known but its degradation was not observed in cells expressing oncogenic 70Z-Cbl, which lacks E3 activity76. These studies illustrate the complexity of c-Cbl’s regulation of Src signalling and show that c-Cbl can have both positive roles, by functioning as an adaptor to contribute to signal transduction, as well as negative roles, by downregulating Src kinase activity or enhancing its degradation. Regulation of T-cell signalling

Since the discovery of c-Cbl as a prominent tyrosine kinase substrate in TCR-stimulated Jurkat T cells77, its role in T-cell signalling has been extensively studied. Overexpression of c-Cbl in Jurkat cells suppresses the activation of two TCR-induced transcription factors: AP-1 (Fos–Jun) and nuclear factor of activated T cells (NF-AT), which is linked to suppressed tyrosine phosphorylation of ZAP-70 (REFS 67, 78–81) and requires a functional c-Cbl TKB domain67,82. However, just like in the thymocyte studies mentioned above, ZAP-70 protein levels are not reduced by c-Cbl, indicating that it has other tricks up its sleeve to negatively regulate the activity of ZAP-70. Intriguingly, similar experiments involving Cbl-b overexpression in Jurkat cells have shown that interaction of its TKB domain with ZAP-70 results in www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

c b

Clathrin-coated pit

a T K P B

P T K B

Clathrin

RF E2 P P P Y

Recycling vesicle

RF E2 P P P Y P

T K P B

RF E2 P P P Y P

T K P B

Endosome (low pH)

Coated vesicle

RF E2 P P P Y P

T K P B

Uncoated vesicle

RF E2 P P P Y P

h g

T K P B Lysosome

RF E2 P P P Y P

T K P B

RF E2 P P P Y P

Proteasome

Figure 4 | Proposed model for the role of c-Cbl in directing multi-ubiquitylation and downregulation of the epidermal growth factor receptor. a | Ligand binding induces tyrosine phosphorylation (P) of the epidermal growth factor (EGFR), and initiates movement of the activated EGFR to clathrin-coated membrane areas. b | Cbl is recruited to the tyrosine-phosphorylated receptor (by adaptor proteins, not shown) and the tyrosine-kinase-binding (TKB) domain binds to phosphorylated Y1045 on the EGFR18. At this time, c-Cbl also becomes tyrosine phosphorylated. c | The E3 (ubiquitin ligase) function of Cbl catalyses transfer of a ubiquitin molecule (red circle) from the RING-finger-bound E2 (ubiquitin-conjugating enzyme) to the EGFR. d | Continued addition of ubiquitin moieties leads to multi-ubiquitylation of the EGFR. e | Multi-ubiquitylated receptor–Cbl complex is internalized by clathrin-coated pits. f | Internalized vesicles are rapidly uncoated and fuse into early endosomes (g). In endosomes, ligand dissociation normally occurs and the receptor is recycled back to the cell surface (h). In cells overexpressing c-Cbl, however, the EGFR is prevented from recycling and is directed instead towards proteasomal/lysosomal degradation (i). Cbl has been reported to be transiently ubiquitylated in macrophages, and has been detected in endosomal compartments, but does not itself become degraded, while the polyubiquitinated CSF-1 receptor is lysosomally degraded43,46. Adapted from REF. 45.

ANERGY

A state in which T cells cannot respond to antigen.

the constitutive activation of NF-AT, which is further enhanced following TCR stimulation80. Importantly, in the same experiment, c-Cbl overexpression did not promote the constitutive activation of NF-AT but rather produced a marked suppression following TCR stimulation. So, under identical experimental conditions, the targeting of ZAP-70 by these two highly related proteins results in opposite effects on NF-AT activation. An explanation is not obvious. However, another study also noted opposite effects of c-Cbl and Cbl-b53. In these experiments, Cbl-b inhibited EGF-induced proliferation whereas the overexpression of c-Cbl had no effect, even though both seem equally effective in directing EGF receptor multi-ubiquitylation and downregulation18. These experiments are difficult to reconcile as both proteins would be expected to have functions conserved with SLI-1. Indeed, the fact that c-Cbl and Cbl-b deficient mice are viable and fertile but the loss of both is embryonic lethal suggests that they have important overlapping functions (H. Gu, M. Murphy and D.

Bowtell, personal communication). c-Cbl also might contribute to the generation of ANERGIC T cells in which Fyn is constitutively activated and mediates the tyrosine phosphorylation of c-Cbl. As Y700 and Y774 in c-Cbl are docking sites for CrkL, which forms a complex with the guanine nucleotide exchange factor C3G to catalyse the exchange of GTP for Rap1, it has been proposed that the predominance of Rap1–GTP over Ras–GTP in these cells is mediated by c-Cbl, and this results in the blockade of interleukin-2 transcription leading to anergy83. Cbl-deficient mice

Although c-Cbl- and Cbl-b-deficient mice are generally healthy and do not show developmental abnormalities, they do show phenotypic alterations, the most marked being associated with thymocyte and peripheral T-cell activation, respectively. Thymocytes in c-Cbl–/– mice show a marked activation of ZAP-70 in response to TCR stimulation, in contrast to wild-type thymocytes, which

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 2 9 9

CD4

a Thymocyte

TCR– CD3

ZAP-70

Lck/Fyn

Peripheral T cell

Cbl-b

PLCγ1

Vav c-Cbl

[Ca2+]i

CD28

REVIEWS

PKC

Slp76

[Ca2+]i

Rac–GDP

Cdc42–GDP

Rac–GTP

Cdc42–GTP

JNK

WASP

Lat Ras–GDP Ras–GTP

Grb2 SOS

MAPK

Actin-dependent structures: • Actin cytoskeleton reorganization • Lipid raft aggregation • Receptor clustering

Activation of transcription factors e.g. NF-AT, NF-κB, AP-1 NF-AT

• Expression of cytokines e.g. IL-2 • Cell proliferation

AP-1

• Expression of cytokines e.g. IL-2 • Cell proliferation

Figure 5 | Cbl proteins negatively regulate T-cell-receptor-coupled signalling pathways. a | In thymocytes, engagement of the T-cell receptor (TCR)–CD3 complex and coaggregation with the CD4 receptor results in activation of the Src family tyrosine kinases Lck and Fyn. These phosphorylate the CD3 receptor and the ζ-chain-associated protein kinase 70 (ZAP-70). This activates ZAP-70 kinase activity, which in turn phosphorylates Slp76 and Lat. TCR triggering also promotes Ca2+ mobilization and activation of protein kinase C (PKC), which combine to activate transcription factors that are important for cytokine gene expression and cellular proliferation. Although c-Cbl deficiency markedly enhances the phosphorylation and activation of ZAP-70, and its phosphorylation of Lat and Slp-76, this surprisingly does not lead to elevated interleukin-2 (IL-2) production or enhanced proliferation. (AP-1, activating protein 1; MAPK, mitogen-activated protein kinase; NF-AT, nuclear factor of activated T cells; NF-κB, nuclear factor of κ-light polypeptide gene enhancer in B cells; PLC, phospholipase C.) b | In peripheral T cells, engagement of the TCR/CD3 complex and the co-stimulatory receptor CD28 augments Ras signalling through the TCR (not shown) and promotes optimal phosphorylation of Vav. Vav is required to activate the c-Jun N-terminal kinase (JNK) and promote receptor clustering and lipid raft aggregations through Rac/Cdc42-dependent pathways. Studies of Cbl-b–/– T cells indicate that Cbl-b is a potent negative regulator of Vav phosphorylation and of signalling pathways leading to Rac and Cdc42 activation, but not Vavdependent calcium mobilization. Whether Cbl-b directly targets Vav or an upstream tyrosine kinase is not known. (WASP, Wiskott–Aldrich syndrome protein.)

T-CELL CO-RECEPTOR

Receptor on T cells that binds accessory molecules on antigenpresenting cells and, when engaged along with the T-cell receptor (TCR), provides either a co-stimulatory (for example, CD28) or an inhibitory (for example, CTLA-4) signal with respect to T-cell activation.

300

require co-stimulation of both the TCR and the coreceptor CD4 (REFS 70–72). The crucial role of CD4 stimulation is to activate Lck, which phosphorylates tyrosine 493 on ZAP-70 to trigger kinase activity84,85. Remarkably, however, in c-Cbl–/– thymocytes ZAP-70 activation can be attained in the absence of Lck activation70. The fact that this occurs without hyperactivation of signalling molecules upstream of ZAP-70 suggests that c-Cbl directly regulates ZAP-70, not its activators (FIG. 5a). The activation of ZAP-70 does, however, require a signal, in the form of TCR crosslinking, so the effect is similar to that seen in C. elegans, in which the loss of SLI-1 restores a weak signal from an attenuated LET-23 receptor8. The Cbl-b-deficient mouse shows a similar phenomenon, but it is in peripheral T cells as opposed to thymocytes. In this case, T-cell proliferation and interleukin 2 (IL-2) production are uncoupled from a requirement for activation of the CO-RECEPTOR CD28 (REFS 86,87). Consequently, Cbl-b-deficient mice are highly susceptible to autoimmune disease. So a lowered threshold for TCR signalling is a common theme in

both mutant mice. Intriguingly, however, Cbl-b-deficient mice show no perturbation in thymocyte signalling, and conversely the peripheral T cells of c-Cbldeficient mice are essentially normal and show no propensity to develop autoimmune disease. Another remarkable distinction between the two mutant mice is that the uncoupling of a CD28 requirement for T-cell activation in Cbl-b-deficient mice does not involve enhanced ZAP-70 activation, but instead results in a significant enhancement in the activation of Vav1. Cbl-b deficiency restores TCR-induced Cdc42 activity and cell proliferation, but not IL-2 production or Ca2+ mobilization, in Vav1–/– mice88. These findings support an earlier study identifying Cbl-b as an inhibitor of Vav-mediated c-Jun amino-terminal kinase activation89. Interestingly Fang et al. showed that PI(3)K, an upstream activator of Vav, could be multi-ubiquitylated by Cbl-b, indicating a possible mechanism for Cbl-b-mediated negative regulation of Vav61. Cbl knockout mice have therefore raised interesting questions about the abilities of c-Cbl and Cbl-b to negatively regulate different signalling molecules in distinct T-cell populations, and suggest that Cbl-b’s mode of action may be focused more at regulating exchange factors for Rho-family GTPases rather than tyrosine kinases (FIG. 5a,b). Given the conservation between c-Cbl and Cbl-b in their TKB and RING finger domains, these differences are surprising and could reflect a divergence of Cbl functions that are unique to haematopoietic cells of higher organisms. With this in mind, it is interesting to note that although Cbl-b has been shown to be tyrosine phosphorylated after EGF or TCR stimulation in cell lines37,53, it is only negligibly phosphorylated after antiCD3 stimulation of primary T cells86. This is in contrast to c-Cbl, which is a prominent phosphotyrosine substrate70,86. In addition, unlike c-Cbl, Cbl-b predominantly uses its proline-rich sequences to constitutively associate with, and regulate, Vav and p85 (REFS 61,89). So the contribution of tyrosine phosphorylation to the function of Cbl proteins is likely to differ and could explain the phenotypic differences seen in T cells and thymocytes of the knockout mice. However, embryonic death in the c-Cbl/Cbl-b double-mutant mouse indicates there is redundancy of key functions, which are likely to be conserved with SLI-1. Additional phenotypic and biochemical alterations of Cbl mutant mice are described in REFS 70–72,86,87 and summarized in TABLE 1. One notable phenotype of the c-Cbl mutant mouse is the striking increase in mammary duct density and branching, which could reflect enhanced signalling by the EGF/ErbB receptor family70. Whether the Cbl-b mutant mouse has a similar phenotype has not been reported. Oncogenic forms of Cbl

c-Cbl was originally identified through the isolation of the oncogenic v-Cbl protein consisting only of the TKB domain2. From information that has since emerged, it is widely accepted that v-Cbl functions as a dominantnegative protein by competing with wild-type Cbl proteins for binding sites on activated tyrosine kinases, thus www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

Table 1 | Contrasting phenotypes of c-Cbl- and Cbl-b-deficient mice c-Cbl–/–

Cbl-b–/–

Thymocytes Elevated cell-surface expression of TCR, CD3, CD4, CD5 and CD69

Expression levels of TCR, CD3, CD4, CD5 and CD69 normal

Enhanced positive selection of CD4+ thymocytes in TCR transgenic mice

Normal selection of CD4+ thymocytes in TCR transgenic mice

Activation of ZAP-70 enhanced and uncoupled from a requirement for CD4 co-stimulation and Lck activation

Activation of ZAP-70 is normal and requires CD4 costimulation and Lck activation

Peripheral T cells Anti-CD3 proliferative responses are slightly suppressed

Anti-CD3 proliferative responses are enhanced

IL-2 production is dependent on CD28 co-stimulation

IL-2 production is enhanced and uncoupled from requirement for CD28 co-stimulation

Normal activation of Vav

Enhanced activation of Vav independent of CD28 costimulation

Not determined

Receptor clustering and raft aggregation independent of Vav expression and CD28 co-stimulation

General phenotypes No evidence of autoimmune disease

Enhanced susceptibility to autoimmune disease

Enlarged spleens and lymph nodes owing to erythroid and Enlarged spleens and lymph nodes owing to infiltration of lymphoid hyperplasia activated T and B cells (mice >6 months, diseased) No apparent infiltration of organs

Infiltration of several organs with activated T and B cells

Increased ductal density and branching in mammary fat pads Not determined IL-2, interleukin-2; TCR, T-cell receptor; ZAP-70, ζ-chain-associated protein kinase 70.

preventing Cbl from negatively regulating these target kinases. This model is supported by the inability of vCbl to transform when its TKB domain is mutated (G306E) and can no longer bind activated tyrosine kinases11,12, and from studies in Drosophila where the Drosophila v-Cbl phenotype of rough eyes is markedly enhanced by a single copy of Drosophila Cbl 7. Even large carboxy-terminal truncations that disrupt the RING finger are insufficient to convert c-Cbl to an oncogenic protein90. Rather, deletion of the linker domain separating the TKB and RING finger domains seems to be the critical mutation. These findings indicate that the linker sequence is an important regulator of TKB domain activity although how it does so remains unanswered. However, part of the linker forms an α-helix that interacts with the TKB domain (FIG. 6a,b), and it is probably the absence of these interactions in v-Cbl that converts Cbl to an oncogenic protein20. The experiments investigating carboxy-terminal truncations also drew attention to a mutant form of c-Cbl isolated from the 70Z/3 mouse pre-B cell lymphoma90. This mutant protein, called 70Z-Cbl, has a 17amino-acid deletion between 366 and 382 that removes most residues in the linker domain that encompass the α-helix, plus the first cysteine of the RING finger2,20 (FIG. 1b). Expression of a Cbl protein with this 17-amino-acid deletion, or deletion of either of two tyrosine residues within the α-helix of the linker (Y368, Y371; FIG. 1b) induces more rapid and acute transformation of NIH 3T3 cells than does v-Cbl90.Furthermore, in contrast to v-Cbl, carboxy-terminal sequences contribute to transformation as ∆Y368- and ∆Y371-Cbl proteins truncated to residue 480 no longer transform (C.T. and W.L.,

unpublished observations). Although the mechanism of 70Z-Cbl transformation is not known, it has been proposed that it is due to the disruption of the RING finger, and therefore an inability to multi-ubiquitylate and downregulate RTKs. This, however, seems unlikely because deletions or substitutions in the RING finger are insufficient to convert Cbl to an oncogenic protein although these mutations abolish Cbl’s E3 activity60. By contrast, deletions in the linker α-helix are highly oncogenic60,90. The characteristics of the α-helix that seem to be responsible for this effect are the multiple contacts that it makes with both the TKB domain and the UBC recruited by the RING finger20 (FIG. 6a,b). The linker helix is likely to be a critical determinant of c-Cbl’s E3 function by precisely regulating the position and orientation of the TKB-bound substrate relative to the RING finger-associated E2 enzyme20. Thus mutations that alter linker–TKB interactions could conceivably prevent substrate multi-ubiquitylation, even in the presence of an intact RING finger. Indeed, although Y368 and Y371 form hydrogen bonds with residues in the variant SH2 domain and EF hand, respectively, and therefore do not contact the UBC, their deletion also abolishes Cbl’s ability to promote RTK multi-ubiquitylation. Oncogenic mutations are therefore predicted to be those that disrupt the structure of the α-helix and therefore abolish interactions with both the TKB domain and the UBC. This suggests that Cblinduced transformation requires both the loss of an ability to function as an E3 ligase and the disruption of the linker α-helix that permits its interaction with the TKB domain. An intriguing observation that emerged from these studies is the unique role of Y371 in directing EGF recep-

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 3 0 1

REVIEWS

Cbl linker sequence

ZAP-70 peptide

UBCH7 Loop 1

C86 L2

Loop 2

Cbl TKB domain Zn

Cbl RING domain

b Cbl TKB–linker–UBCH7 interfaces UBCH7

Figure 6 | Structure of the complex between c-Cbl and UBCH7. a | Complex of the c-Cbl tyrosine-kinase-binding (TKB) (purple), linker (pink) and RING finger (red) domains interacting with UBCH7 (green) and a tyrosine phosphorylated ζ-chain-associated protein kinase 70 (ZAP70) peptide (orange). The UBCH7 active-site cysteine is in yellow and the two zinc ions are shown as blue spheres. The structure of the TKB domain consists of a four-helix bundle, two EF hand domains and a variant Src homology region 2 (SH2) domain15,20. The linker domain forms an ordered loop and an α-helix that forms multiple contacts with the TKB domain and UBCH7 (see below). The RING finger provides a shallow groove formed by the α-helix and the two Zn-chelating loops (loops 1 and 2) and this groove accommodates the most important contacts with UBCH7. The RING-finger domain is also anchored onto the TKB domain through contacts with the four-helix bundle20. b | Close-up view of the interfaces between the linker helix and TKB domain and between the linker helix and the UBCH7 H1 helix20. The linker, the TKB domain and UBCH7 are coloured in orange, green and purple, respectively. The side chains of the residues involved in the interfaces are yellow from the linker helix, green from the TKB domain and cyan from UBCH7. The linker-helix–TKB domain interactions are centred on conserved Y368 and Y371 residues from the linker. These residues are in a buried environment and make several van der Waals contacts with hydrophobic TKB domain residues and also hydrogen bond with N259 from the SH2 domain and T227 from the EF hand, respectively. Disrupting the structure of the linker helix by deleting either Y368 or Y371 converts Cbl to an oncogenic protein and abolishes its E3 ligase activity. It is likely that mutations to other residues that disrupt the structure of the highly conserved linker helix would similarly abolish its interactions with the TKB domain and UBC and convert Cbl to an oncogenic protein60. Courtesy of Zheng and colleagues20.

302

tor multi-ubiquitylation18,60. Phenylalanine substitution abolishes c-Cbl-directed multi-ubiquitylation of the EGF receptor, a remarkable finding as this residue makes no contact with the UBC (FIG. 6b) and its substitution is unlikely to disrupt the linker α-helix20. By contrast, Y368F-Cbl directs EGF receptor multi-ubiquitylation as efficiently as wild-type c-Cbl60. Whether it is phosphorylation of Y371 that is essential for E3 activity remains to be determined, although from the structure it is not readily accessible to phosphorylation20.Additionally, unlike ∆Y371, the phenylalanine substitution of Y371 is not oncogenic, which is further evidence that loss of E3 activity is insufficient for transformation90. A characteristic of cells expressing oncogenic Cbl proteins is the constitutive activation of signalling proteins that might be effectors of transformation. In fibroblasts, this is seen as the constitutive phosphorylation of the EGF and PDGF receptors, Akt and Cbl itself11,60,91. Activation of these proteins is not solely due to an inability to direct RTK multi-ubiquitylation because non-oncogenic RING finger mutants do not mediate an equivalent effect60. The effect of 70Z-Cbl on EGF receptor and Akt activation is similar to that seen in cells transformed by activated Src, suggesting that some oncogenic Cbl proteins function by positively regulating RTKs, that is, as activated oncogenes with gainof-function mutations. It seems, however, that these effects are independent of mitogenic signalling because cells transformed by oncogenic Cbl have no proliferative advantage and remain dependent on serum63. However, as Cbl-transformed cells show anchorage independent growth, it is possible that Cbl might positively regulate signals from RTKs and integrin receptors that affect adhesion complexes and the organization of actin63,75,90,92,93. Positive signalling by 70Z-Cbl is not restricted to enhanced RTK activation because Jurkat T cells transfected with 70Z-Cbl also show constitutive activation of NF-AT, AP-1 and phospholipase Cγ-1 (REFS 40,67,81,94). ZAP-70 has been identified as an essential target; however, because its kinase activity and protein levels seem unaltered, the mechanism of enhanced signalling by 70Z-Cbl requires further investigation. Whether oncogenic forms of Cbl proteins exist in or contribute to human cancers has not been reported in the literature but it remains an issue that warrants investigation. Regulation of tyrosine kinase signalling

It is widely accepted that c-Cbl functions as a negative regulator of tyrosine kinase signalling. Nonetheless, there is evidence that c-Cbl can also have a positive role. For example, c-Cbl enhances proliferation and survival through PI(3)K-dependent pathways after cytokine stimulation95,96, and enhances MAPK activation in response to stimulation of the Met receptor (the RTK activated by scatter factor)97. Interaction with the SH3 domain of Cblassociated protein (CAP) preferentially recruits tyrosine phosphorylated c-Cbl to LIPID RAFTS after insulin stimulation of adipocytes,and this has been implicated in directing a second signal necessary for glucose transport in www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

LIPID RAFTS

Dynamic assemblies of cholesterol and sphingolipids in the plasma membrane.

response to insulin98,99. However, in adipocytes c-Cbl seems to be a substrate of Fyn rather than the insulin receptor, indicating its possible role in glucose transport is downstream of a Src family member98. A similar conclusion was drawn in studies characterizing c-Cbl’s role in bone resorption and osteoclast migration where c-Cbl was identified as a key downstream substrate and effector of Src73,75. Involvement of c-Cbl in positive signalling events is best shown in its regulation of the cytoskeletal responses.Although the mechanisms of these effects of cCbl have yet to be determined, it is likely that in such cases c-Cbl is being used as a complex multi-adaptor protein rather than as an E3 ligase. Regulation of the actin cytoskeleton

Non-transforming Cbl proteins are involved in the regulation of cell morphology and functional organization of the actin cytoskeleton75,93,100–102. Integrin-dependent adhesion induces phosphorylation of c-Cbl on tyrosine residues and its translocation to the membrane, where it is associated with enhanced PI(3)K activity75,100. c-Cbl localizes with actin at lamellipodia and leading edges of migrating cells93,103 and also interacts with proteins such as Crk, CAP and CMS/CD2AP/CIN85, which co-localize to actin structures and modulate cytoskeletal responses93,103–106. c-Cbl depletion inhibited Srcdependent osteoclast migration74, and blocked Srcand PI(3)K-dependent spreading of primary macrophages75. Furthermore, expression of a dominant-negative c-Cbl mutant lacking carboxy-terminal proline-rich sequences markedly perturbed cell morphology, reduced cell spreading and inhibited Racdependent actin lamellae formation in cycling NIH 3T3 fibroblasts93. Conversely, overexpression of c-Cbl facilitated cell adhesion and spreading of v-Abl-transformed fibroblasts, and this was ablated by treatment with PI(3)K inhibitors or by mutation of major tyrosine phosphorylation sites that normally recruit p85 and CrkL92. Notably, in these contexts, Cbl seems to have a positive regulatory function and act downstream of Src, linking PI(3)K and CrkL to the integrin signalling pathway. Cbl-b also seems to be a potent regulator of actin cytoskeletal processes as Cbl-b-deficient T cells show enhanced anti-CD3-induced T-cell receptor clustering, membrane raft aggregation and filopodia formation in the absence of CD28 co-stimulation88.

Langdon, W. Y., Hartley, J. W., Klinken, S. P., Ruscetti, S. K. & Morse, H. C., III. v-cbl, an oncogene from a dualrecombinant murine retrovirus that induces early B-lineage lymphomas. Proc. Natl Acad. Sci. USA 86, 1168–1172 (1989). Blake, T. J., Shapiro, M., Morse, H. C., III & Langdon, W. Y. The sequences of the human and mouse c-cbl protooncogenes show v-cbl was generated by a large truncation encompassing a proline-rich domain and a leucine zipperlike motif. Oncogene 6, 653–657 (1991). Keane, M. M., Rivero-Lezcano, O. M., Mitchell, J. A., Robbins, K. C. & Lipkowitz, S. Cloning and characterization of cbl-b: a SH3-binding protein with homology to the c-cbl proto-oncogene. Oncogene 10, 2367–2377 (1995). Keane, M. M. et al. cbl-3: a new mammalian cbl family protein. Oncogene 18, 3365–3375 (1999).

Future directions

A clearer understanding of Cbl function is emerging, but some key questions remain unanswered. One of the highest priorities is to understand whether the E3 ligase activity of Cbl is solely for the purpose of RTK degradation, or whether Cbl-directed polyubiquitylation alters the function and fate of RTKs in ways not yet appreciated. It will also be important to determine whether Cbl’s E3 activity is involved in the regulation of Syk, ZAP-70 and Src family kinases, as well as how ubiquitylation might affect other Cbl binding partners such as PI(3)K. We know remarkably little about the bewildering number of downstream effects and protein interactions of Cbl that do not seem to be associated with RTK multi-ubiquitylation. Indeed, recent findings that loss of E3 activity is insufficient to promote transformation60, and the fact that SLI-1’s RING finger is partially dispensable for its negative regulation of LET-23 (REF. 107), shows that E3 activity is not the sole inhibitory function of SLI-1/Cbl and highlights a major gap in our understanding. Clues about how best to approach this aspect of Cbl function may be revealed by more in-depth analyses of the TKB domain, as it is highly unlikely that such an extensive structure exists solely to recognize and bind phosphotyrosine. Furthermore, the limitations associated with interpreting results using cell lines that overexpress Cbl suggest a need for a greater reliance on genetically altered mice, Drosophila and C. elegans. An important factor that will aid these studies is the progress recently made in resolving the structure of the TKB, linker and RING finger domains, which now allows precise modifications using knowledge that was not available when the c-Cbl- and Cbl-b-deficient mice were first generated. The next few years promise to be exciting times for studying Cbl proteins and their roles in regulating tyrosine kinase signalling pathways. Links DATABASE LINKS c-Cbl | Cbl family | Cbl-b | Cbl-3 | D-Cbl | SLI-1 | EGFR | PDGFR | ZAP-70 | Syk | EF hand | SH2 domain | RING finger | FGFR | 14-3-3 protein | Fyn | Yes | Lyn | Abl | Vav1 | Cdc42 | UBA domain | LET-23 | LET-60 | Grb2 | SEM-5 | colony-stimulating factor 1 receptor | UBCH7 | ErbB2 | Fos | Jun | NF-AT | CrkL | C3G | Rap1 | CD28 | Akt | Met | scatter factor | CAP ENCYCLOPEDIA OF LIFE SCIENCES Antigen recognition by lymphocytes | Ubiquitin pathway

Meisner, H. et al. Interactions of Drosophilia Cbl with epidermal growth factor receptors and role of Cbl in R7 photoreceptor cell development. Mol. Cell. Biol. 17, 2217–2225 (1997). Hime, G. R., Dhungat, M. P., Ng, A. & Bowtell, D. D. L. D-Cbl, the Drosophila homologue of the c-Cbl protooncogene, interacts with the Drosophila EGF receptor in vivo, despite lacking C-terminal adaptor binding sites. Oncogene 14, 2709–2719 (1997). Robertson, H., Hime, G. H., Lada, H. & Bowtell, D. D. L. A Drosophila analogue of v-Cbl is a dominant-negative oncoprotein in vivo. Oncogene 19, 3299–3308 (2000). Yoon, C. H., Lee, J., Jongeward, G. D. & Sternberg, P. W. Similarity of sli-1, a regulator of vulval development in C. elegans, to the mammalian proto-oncogene c-cbl. Science 269, 1102–1105 (1995).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

A seminal publication that identified Cbl as a negative regulator of receptor tyrosine kinases. 9. Galisteo, M. L., Dikic, I., Batzer, A. G., Langdon, W. Y. & Schlessinger, J. Tyrosine phosphorylation of the c-cbl proto-oncogene product and association with epidermal growth factor (EGF) receptor upon EGF stimulation. J. Biol. Chem. 270, 20241–20245 (1995). First demonstration that the amino-terminal half of cCbl can bind directly to tyrosine-phosphorylated epidermal growth factor receptor. This finding was further characterized in reference 10. 10. Lupher, M. L., Jr, Reedquist, K. A., Miyake, S., Langdon, W. Y. & Band, H. A novel PTB domain in the N-terminal transforming region of Cbl interacts directly and selectively with ZAP-70 in T cells. J. Biol. Chem. 271, 24063–24068 (1996).

VOLUME 2 | APRIL 2001 | 3 0 3

REVIEWS 11. Bonita, D. P., Miyake, S., Lupher Jr, M. L., Langdon, W. Y. & Band, H. Phosphotyrosine binding domain-dependent upregulation of the platelet-derived growth factor receptor α signaling cascade by transforming mutants of Cbl: implications for Cbl’s function and oncogenicity. Mol. Cell. Biol. 17, 4597–4610 (1997). 12. Thien, C. B. F. & Langdon, W. Y. EGF receptor binding and transformation by v-cbl is ablated by the introduction of a loss-of-function mutation from the Caenorhabditis elegans sli-1 gene. Oncogene 14, 2239–2249 (1997). 13. Lupher, M. L., Jr et al. Cbl-mediated negative regulation of the Syk tyrosine kinase. J. Biol. Chem. 273, 35273–35281 (1998). 14. Deckert, M., Elly, C., Altman, A. & Liu, Y. C. Coordinated regulation of the tyrosine phosphorylation of Cbl by Fyn and Syk tyrosine kinases. J. Biol. Chem. 273, 8867–8874 (1998). 15. Meng, W., Sawasdikosol, S., Burakoff, S. J. & Eck, M. J. Structure of the amino-terminal domain of Cbl complexed to its binding site on ZAP-70 kinase. Nature 398, 84–90 (1999). The structure of Cbl’s novel TKB domain revealed three interacting domains comprising a four-helix bundle, a Ca2+-binding EF hand and a variant SH2 domain. 16. Waterman, H., Levkowitz, G., Alroy, I. & Yarden, Y. The RING finger of c-Cbl mediates desensitization of the epidermal growth factor. J. Biol. Chem. 274, 22151–22154 (1999). 17. Joazeiro, C. A. P. et al. The tyrosine kinase negative regulator c-Cbl as a RING-type, E2-dependent ubiquitinprotein ligase. Science 286, 309–312 (1999). This paper, along with references 18 and 19, was the first to define Cbl as an E3 ubiquitin protein ligase. 18. Levkowitz, G. et al. Ubiquitin ligase activity and tyrosine phosphorylation underlie suppression of growth factor signaling by c-Cbl/Sli-1. Mol. Cell 4, 1–20 (1999). 19. Yokouchi, M. et al. Ligand-induced ubiquitination of the epidermal growth factor receptor involves the interaction of the c-Cbl RING finger and UbcH7. J. Biol. Chem. 274, 31707–31712 (1999). 20. Zheng, N., Wang, P., Jeffrey, P. D. & Pavletich, N. P. Structure of a c-Cbl–UbcH7 complex: RING domain function in ubiquitin-protein ligases. Cell 102, 533–539 (2000). The structure revealed that UBCH7 has several interactions with c-Cbl’s RING finger domain and a highly conserved α-helix at the carboxyl terminus of the linker domain. This paper helped to explain the why deletions in the linker α-helix convert Cbl to an oncogenic protein. 21. Wong, E. S. M., Lim, J., Low, B. C., Chen, Q. & Guy, G. R. Evidence for direct interaction between Sprouty and Cbl. J. Biol. Chem. 276, 5866–5875 (2001). 22. Hacohen, N., Kramer, S., Sutherland, D., Hiromi, Y. & Kasnow, M. A. sprouty encodes a novel antagonist of FGF signaling that patterns apical branching of the Drosophila airways. Cell 92, 253–263 (1998). 23. Casci, T., Vinòs, J. & Freeman, M. Sprouty, an intracellular inhibitor of Ras signaling. Cell 96, 655–665 (1999). 24. Liu, Y.-C., Elly, C., Yoshida, H., Bonnefoy-Berard, N. & Altman, A. Activation-modulated association of 14-3-3 proteins with Cbl in T cells. J. Biol. Chem. 271, 14591–14595 (1996). 25. Liu, Y. C. et al. Serine phosphorylation of Cbl induced by phorbol ester enhances its association with 14-3-3 proteins in T cells via a novel serine-rich 14-3-3-binding motif. J. Biol. Chem. 272, 9979–9985 (1997). 26. Liu, Y. et al. Protein kinase C activation inhibits tyrosine phosphorylation of Cbl and its recruitment of Src homology 2 domain-containing proteins. J. Immunol. 162, 7095–7101 (1999). 27. Fernàndez, B., Czech, M. P. & Meisner, H. Role of protein kinase C in signal attenuation following T cell receptor engagement. J. Biol. Chem. 274, 20244–20250 (1999). 28. Feshchenko, E. A., Langdon, W. Y. & Tsygankov, A. Y. Fyn, Yes and Syk phosphorylation sites in c-Cbl map to the same tyrosine residues that become phosphorylated in activated T cells. J. Biol. Chem. 273, 8323–8331 (1998). 29. Tezuka, T. et al. Physical and functional association of the cbl protooncogene product with an Src-family protein tyrosine kinase, p53/56lyn, in the B cell antigen receptormediated signaling. J. Exp. Med. 183, 675–680 (1996). 30. de Jong, R., ten Hoeve, J., Heisterkamp, N. & Groffen, J. Crkl is complexed with tyrosine-phosphorylated cbl in Phpositive leukemia. J. Biol. Chem. 270, 21468–21471 (1995). 31. Buday, L., Khwaja, A., Sipeki, S., Farago, A. & Downward, J. Interactions of Cbl with two adaptor proteins, Grb2 and

304

32.

33.

34.

35.

36.

37.

38. 39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

53. 54.

Crk, upon T cell activation. J. Biol. Chem. 271, 6159–6163 (1996). Sattler, M. et al. The proto-oncogene product p120CBL and the adaptor proteins CRKL and c-CRK link c-ABL, p190BCR/ABL and p210BCR/ABL to the phosphatidylinositol-3′ kinase pathway. Oncogene 12, 839–846 (1996). Ribon, V., Hubbell, S., Herrera, R. & Saltiel, A. R. The product of the cbl oncogene forms stable complexes in vivo with endogenous crk in a tyrosine phosphorylationdependent manner. Mol. Cell. Biol. 16, 45–52 (1996). Andoniou, C. E., Thien, C. B. F. & Langdon, W. Y. The two major sites of cbl tyrosine phosphorylation in abltransformed cells select the crkL SH2 domain. Oncogene 12, 1981–1989 (1996). Jain, S. K., Langdon, W. Y. & Varticovski, L. Tyrosine phosphorylation of p120 cbl in BCR/abl transformed hematopoietic cells mediates enhanced association with phosphatidylinositol 3-kinase. Oncogene 14, 2217–2228 (1997). Marengére, L. E. M. et al. Proto-oncoprotein Vav interacts with c-Cbl in activated thymocytes and peripheral T cells. J. Immunol. 159, 70–76 (1997). Elly, C. et al. Tyrosine phosphorylation and complex formation of Cbl-b upon T cell receptor stimulation. Oncogene 18, 1147–1156 (1999). Songyang, Z. et al. SH2 domains recognize specific phosphopeptide sequences. Cell 72, 767–778 (1993). Hartley, D. & Corvera, S. Formation of c-Cblphosphatidylinositol 3-kinase complexes on lymphocyte membranes by a p56lck-independent mechanism. J. Biol. Chem. 271, 21939–21943 (1996). Liu, Y.-C., Elly, C., Langdon, W. Y. & Altman, A. Rasdependent, Ca2+-stimulated activation of nuclear factor of activated T cells by a constitutively active Cbl mutant in T cells. J. Biol. Chem. 272, 168–173 (1997). Hunter, S., Burton, E. A., Wu, S. C. & Anderson, S. M. Fyn associates with Cbl and phosphorylates tyrosine 731 in Cbl, a binding site for phosphatidylinositol 3-kinase. J. Biol. Chem. 274, 2097–2106 (1999). Hofmann, K. & Bucher, P. UBA domain: a sequence motif present in multiple enzyme classes of ubiquitin pathway. Trends. Biochem. Sci. 21, 172–173 (1996). Wang, Y., Yeung, Y. G., Langdon, W. Y. & Stanley, E. R. ccbl is transiently tyrosine-phosphorylated, ubiquitinated, and membrane-targeted following CSF-1 stimulation of macrophages. J. Biol. Chem. 271, 17–20 (1996). First paper to indicate a connection between c-Cbl and protein multi-ubiquitylation. Miyake, S., Lupher, M. L., Jr, Druker, B. & Band, H. The tyrosine kinase regulator Cbl enhances the ubiquitination and degradation of the platelet-derived growth factor receptor α. Proc. Natl Acad. Sci. USA 95, 7927–7932 (1998). Levkowitz, G. et al. c-Cbl/Sli-1 regulates endocytic sorting and ubiquitination of the epidermal growth factor receptor. Genes Dev. 12, 3663–3674 (1998). References 44 and 45 were the first to show that cCbl can function as a negative regulator of receptor tyrosine kinase signalling by directing the multiubiquitylation of activated receptors. Lee, P. S. W. et al. The Cbl protooncoprotein stimulates CSF-1 receptor multiubiquitination and endocytosis, and attenuates macrophage proliferation. EMBO J. 18, 3616–3628 (1999). Lill, N. L. et al. The evolutionarily conserved N-terminal region of Cbl is sufficient to enhance down-regulation of the epidermal growth factor receptor. J. Biol. Chem. 275, 367–377 (2000). Bartkiewicz, M., Houghton, A. & Baron, R. Leucine zippermediated homodimerization of the adaptor protein c-Cbl. J. Biol. Chem. 274, 30887–30895 (1999). Jongeward, G. D., Clandinin, T. R. & Sternberg, P. W. sli-1, a negative regulator of let-23-mediated signaling in C. elegans. Genetics 139, 1553–1566 (1995). Pai, L.-M., Barcelo, G. & Schüpbach, T. D-Cbl, a negative regulator of the Egfr pathway, is required for dorsoventral patterning in Drosophila oogenesis. Cell 103, 51–61 (2000). Tanaka, S., Neff, L., Baron, R. & Levy, J. Tyrosine phosphorylation and translocation of the c-cbl protein after activation of tyrosine kinase signaling pathways. J. Biol. Chem. 270, 14347–14351 (1995). Bowtell, D. D. L. & Langdon, W. Y. The protein product of the c-cbl oncogene rapidly complexes with the EGF receptor and is tyrosine phosphorylated following EGF stimulation. Oncogene 11, 1561–1567 (1995). Ettenberg, S. A. et al. cbl-b inhibits epidermal growth factor receptor signaling. Oncogene 18, 1855–1866 (1999). Lupher, M. L., Songyang, Z., Shoelson, S. E., Cantley, L. C. & Band, H. The Cbl phosphotyrosine-binding domain selects a D(N/D)XpY motif and binds to the TyrP292

55.

56.

57.

58. 59.

60.

61.

62.

63.

64.

65.

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.

76.

77.

negative regulatory phosphorylation site of ZAP-70. J. Biol. Chem. 272, 33140–33144 (1997). Kong, G. et al. Distinct tyrosine phosphorylation sites in ZAP-70 mediate activation and negative regulation of antigen receptor function. Mol. Cell. Biol. 16, 5026–5035 (1996). Zhao, Q. & Weiss, A. Enhancement of lymphocyte responsiveness by a gain-of-function mutation of ZAP-70. Mol. Cell. Biol. 16, 6765–6774 (1996). Galcheva-Gargova, Z., Theroux, S. J. & Davis, R. J. The epidermal growth factor receptor is covalently linked to ubiquitin. Oncogene 11, 2649–2655 (1995). Levkowitz, G. et al. c-Cbl is a suppressor of the Neu oncogene. J. Biol. Chem. 275, 35532–35539 (2000). Stang, E., Johannessen, L. E., Knardal, S. L. & Madshus, I. H. Polyubiquitination of the epidermal growth factor receptor occurs at the plasma membrane upon ligandinduced activation. J. Biol. Chem. 275, 13940–13947 (2000). Thien, C. B. F., Walker, F. & Langdon, W. Y. Ring finger mutations that abolish c-Cbl-directed polyubiquitination and downregulation of the EGF receptor are insufficient for cell transformation. Mol. Cell 7, 355–365 (2001). Fang, D. et al. Cbl-b, a RING-type E3 ubiquitin ligase, targets phosphatidylinositol 3-kinase for ubiquitination in T cells. J. Biol. Chem. 276, 4872–4878 (2001). Ueno, H. et al. Antisense repression of proto-oncogene cCbl enhances activation of the JAK-STAT pathway but not the Ras pathway in epidermal growth factor receptor signaling. J. Biol. Chem. 272, 8739–8743 (1997). Ojaniemi, M., Langdon, W. Y. & Vuori, K. Oncogenic forms of Cbl abrogate the anchorage requirement but not the growth factor requirement for proliferation. Oncogene 16, 3159–3167 (1998). Miyake, S., Mullane-Robinson, K. P., Lill, N. L., Douillard, P. & Band, H. Cbl-mediated negative regulation of plateletderived growth factor receptor-dependent cell proliferation. J. Biol. Chem. 274, 16619–16628 (1999). Broome, M. A., Galisteo, M. L., Schlessinger, J. & Courtneidge, S. A. The proto-oncogene c-Cbl is a negative regulator of DNA synthesis initiated by both receptor and cytoplasmic tyrosine kinases. Oncogene 18, 2908–2912 (1999). Ota, Y. & Samelson, L. E. The product of the protooncogene c-Cbl — a negative regulator of the Syk tyrosine kinase. Science 276, 418–420 (1997). First paper to show a negative regulatory role for Cbl towards a protein tyrosine kinase in mammalian cells. Rao, N. et al. The linker phosphorylation site Tyr292 mediates the negative regulatory effect of Cbl on ZAP-70 in T cells. J. Immunol. 164, 4616–4626 (2000). Ota, S. et al. The RING finger domain of Cbl is essential for negative regulation of the Syk tyrosine kinase. J. Biol. Chem. 275, 414–422 (2000). Yasuda, T. et al. Cbl suppresses B cell receptor-mediated phospholipase C (PLC)-γ2 activation by regulating B cell linker protein-PLCγ2 binding. J. Exp. Med. 191, 641–650 (2000). Murphy, M. A. et al. Tissue hyperplasia and enhanced T cell signalling via ZAP-70 in c-Cbl deficient mice. Mol. Cell. Biol. 18, 4872–4882 (1998). Naramura, M., Kole, H. K., Hu, R.-J. & Gu, H. Altered thymic positive selection and intracellular signals in Cbldeficient mice. Proc. Natl Acad. Sci. USA 95, 15547–15552 (1998). References 70 and 71 provided evidence that c-Cbl functions as a negative regulator of ZAP-70 and that ZAP-70 in c-Cbl–/– thymocytes can be activated in the absence of CD4 co-receptor stimulation. Thien, C. B. F., Bowtell, D. D. L. & Langdon, W. Y. Perturbed regulation of ZAP-70 and sustained tyrosine phosphorylation of LAT and SLP-76 in c-Cbl-deficient thymocytes. J. Immunol. 162, 7133–7139 (1999). Tanaka, S. et al. c-Cbl is downstream of c-Src in a signalling pathway necessary for bone resorption. Nature 383, 528–531 (1996). Sanjay, A. et al. Cbl associates with Pyk2 and Src to regulate Src kinase activity, αvβ3 integrin-mediated signaling, cell adhesion, and osteoclast motility. J. Cell. Biol. 152, 181–195 (2001). Meng, F. & Lowell, C. A. A β1 integrin signaling pathway involving Src-family kinases, Cbl and PI-3 kinase is required for macrophage spreading and migration. EMBO J. 17, 4391–4403 (1998). Andoniou, C. E. et al. The Cbl proto-oncogene product negatively regulates the Src-family tyrosine kinase Fyn by enhancing its degradation. Mol. Cell. Biol. 20, 851–867 (2000). Donovan, J. A., Wange, R. L., Langdon, W. Y. & Samelson,

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

REVIEWS

78.

79.

80.

81.

82.

83.

84.

85.

86.

87. 88.

89.

L. E. The protein product of the c-cbl protooncogene is the 120-kDa tyrosine-phosphorylated protein in Jurkat cells activated via the T cell antigen receptor. J. Biol. Chem. 269, 22921–22924 (1994). Rellahan, B. L., Graham, L. J., Stocia, B., DeBell, K. E. & Bonvini, E. Cbl-mediated regulation of T cell receptorinduced AP-1 activation. J. Biol. Chem. 272, 30806–30811 (1997). Zhang, Z., Elly, C., Altman, A. & Liu, Y.-C. Dual regulation of T cell receptor-mediated signaling by oncogenic Cbl mutant 70Z. J. Biol. Chem. 274, 4883–4889 (1999). Zhang, Z., Elly, C., Qiu, L., Altman, A. & Liu, Y.-C. A direct interaction between the adaptor protein Cbl-b and the kinase Zap-70 induces a positive signal in T cells. Curr. Biol. 9, 203–206 (1999). van Leeuwen, J. E. M., Paik, P. K. & Samelson, L. E. Activation of nuclear factor of activated T cells (NFAT) and activating protein 1 (AP-1) by oncogenic 70Z Cbl requires an intact phosphotyrosine binding domain but not Crk(L) or p85 phosphatidylinositol 3-kinase association. J. Biol. Chem. 274, 5153–5162 (1999). van Leeuwen, J. E. M., Paik, P. K. & Samelson, L. E. The oncogenic 70Z Cbl mutation blocks the phosphotyrosine binding domain-dependent negative regulation of ZAP-70 by c-Cbl in Jurkat T cells. Mol. Cell. Biol. 19, 6652–6664 (1999). Boussiotis, V. A., Freeman, G. J., Berezovskaya, A., Barber, D. L. & Nadler, L. M. Maintenance of human T cell anergy: blocking of IL-2 gene transcription by activated Rap1. Science 278, 124–128 (1997). Chan, A. C. et al. Activation of ZAP-70 kinase activity by phosphorylation of tyrosine 493 is required for lymphocyte antigen receptor function. EMBO J. 14, 2499–2508 (1995). Wange, R. L. et al. Activating and inhibitory mutations in adjacent tyrosines in the kinase domain of ZAP-70. J. Biol. Chem. 270, 18730–18733 (1995). Bachmaier, K. et al. Negative regulation of lymphocyte activation and autoimmunity by the molecular adaptor Cbl-b. Nature 403, 211–216 (2000). Chiang, Y. J. et al. Cbl-b regulates the CD28 dependence of T-cell activation. Nature 403, 216–220 (2000). Krawczyk, C. et al. Cbl-b is a negative regulator of receptor clustering and raft aggregation in T cells. Immunity 13, 463–473 (2000). References 86–88 provided evidence that Cbl-b selectively suppresses TCR-mediated Vav activation and that Cbl-b-deficient mice are highly susceptible to autoimmune disease. Bustelo, X. R., Crespo, P., López-Barahona, M., Gutkind, J.

90.

91.

92.

93.

94.

95.

96.

97.

98.

99.

S. & Barbacid, M. Cbl-b, a member of the Sli-1/c-Cbl protein family, inhibits Vav-mediated c-Jun N-terminal kinase activation. Oncogene 15, 2511–2520 (1997). Andoniou, C. E., Thien, C. B. F. & Langdon, W. Y. Tumour induction by activated abl involves tyrosine phosphorylation of the product of the cbl oncogene. EMBO J. 13, 4515–4523 (1994). Identified an acutely transforming c-Cbl protein in the 70Z/3 mouse pre-B cell lymphoma. This form of c-Cbl has 17-amino-acid deletion spanning the α-helix of the linker domain (that separates the SH2 and RING finger domains) and the first two residues of the RING finger. Thien, C. B. F. & Langdon, W. Y. Tyrosine kinase activity of the EGF receptor is enhanced by the expression of oncogenic 70Z-Cbl. Oncogene 15, 2909–2919 (1997). Feshchenko, E. A., Shore, S. K. & Tsygankov, A. Y. Tyrosine phosphorylation of c-Cbl facilitates adhesion and spreading while suppressing anchorage-independent growth of v-Abltransformed NIH3T3 fibroblasts. Oncogene 18, 3703–3715 (1999). Scaife, R. M. & Langdon, W. Y. c-Cbl localizes to actin lamellae and regulates lamellipodia formation and cell morphology. J. Cell Sci. 113, 215–226 (2000). Graham, L. J. et al. Differential effects of Cbl and 70Z/3 Cbl on T cell receptor-induced phospholipase Cγ-1 activity. FEBS Lett. 470, 273–280 (2000). Ueno, H. et al. c-Cbl is tyrosine-phosphorylated by interleukin-4 and enhances mitogenic and survival signals of interleukin-4 receptor by linking with the phosphatidylinositol 3’-kinase pathway. Blood 91, 46–53 (1998). Grishin, A. et al. Involvement of Shc and Cbl-PI 3-kinase in Lyn-dependent proliferative signaling pathways for G-CSF. Oncogene 19, 97–105 (2000). Garcia-Guzman, M., Larsen, E. & Vuori, K. The protooncogene c-Cbl is a positive regulator of Met-induced MAP kinase activation: a role for the adaptor protein Crk. Oncogene 19, 4058–4065 (2000). Mastick, C. C. & Saltiel, A. R. Insulin-stimulated tyrosine phosphorylation of caveolin is specific for the differentiated adipocyte phenotype in 3T3-L1 cells. J. Biol. Chem. 272, 20706–20714 (1997). Baumann, C. A. et al. CAP defines a second signalling pathway required for insulin-stimulated glucose transport. Nature 407, 202–207 (2000). Identifies c-Cbl as a key adaptor molecule in a complex with CAP and flotillin. The complex is localized to lipid raft subdomains and mediates a

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

second signal required for glucose transport. 100. Ojaniemi, M., Martin, S. S., Dolfi, F., Olefsky, J. M. & Vuori, K. The proto-oncogene product p120cbl links c-Src and phosphatidylinositol 3′-kinase to the integrin signaling pathway. J. Biol. Chem. 272, 3780–3787 (1997). 101. Sattler, M. et al. Differential signaling after β1 integrin ligation is mediated through binding of CrkL to p120CBL and p110HEF1. J. Biol. Chem. 272, 14320–14326 (1997). 102. Zell, T. et al. Regulation of β1-integrin-mediated cell adhesion by the Cbl adaptor protein. Curr. Biol. 8, 814–822 (1998). 103. Kirsch, K. et al. The adaptor type protein CMS/CD2AP binds to the proto-oncogenic protein c-Cbl through a tyrosine phosphorylation-regulated Src homology 3 domain interaction. J. Biol. Chem. 276, 4957–4963 (2001). 104. Uemura, N. & Griffin, J. D. The adapter protein Crkl links Cbl to C3G after integrin ligation and enhances cell migration. J. Biol. Chem. 274, 37525–37532 (1999). 105. Ribon, V., Herrera, R., Kay, B. K. & Saltiel, A. R. A role for CAP, a novel, multifunctional Src homology 3 domaincontaining protein in formation of actin stress fibers and focal adhesion. J. Biol. Chem. 273, 4073–4080 (1998). 106. Take, H. et al. Cloning and characterization of a novel adaptor protein, CIN85, that interacts with c-Cbl. Biochem. Biophys. Res. Comm. 268, 321–328 (2000). 107. Yoon, C. H., Chang, C., Hopper, N. A., Lesa, G. M. & Sternberg, P. W. Requirements of multiple domains of SLI1, a Caenorhabditis elegans homologue of c-Cbl, and an inhibitory tyrosine in LET–23 in regulating vulval differentiation. Mol. Biol. Cell 11, 4019–4031 (2000). 108. Terrell, J., Shih, S., Dunn, R. & Hicke, L. A function for monoubiquitination in the internalization of a G proteincoupled receptor. Mol. Cell 1, 193–202 (1998). 109. Nakatsu, F. et al. A di-leucine signal in the ubiquitin moiety. J. Biol. Chem. 275, 26213–26219 (2000). 110. Thrower, J. S., Hoffman, L., Rechsteiner, M. & Pickart, C. M. Recognition of the polyubiquitin proteolytic signal. EMBO J. 19, 94–102 (2000). 111. Bonifacino, J. S. & Weissman, A. M. Ubiquitin and the control of protein fate in the secretory and endocytic pathways. Annu. Rev. Cell Dev. Biol. 14, 19–57 (1998). 112. Ciechanover, A. The ubiquitin-proteasome pathway: on protein death and cell life. EMBO J. 17, 7151–7160 (1998).

Acknowledgements We thank N. Zheng and N. Pavletich for providing the structural figures, and H. Gu, D. Bowtell and M. Murphy for providing unpublished results on the Cbl-knockout mice.

VOLUME 2 | APRIL 2001 | 3 0 5

PERSPECTIVES OPINION

Signal processing and transduction in plant cells: the end of the beginning? Simon Gilroy and Anthony Trewavas Plants have a very different lifestyle to animals, and one might expect that unique molecules and processes would underpin plant-cell signal transduction. But, with a few notable exceptions, the list is remarkably familiar and could have been constructed from animal studies. Wherein, then, does lifestyle specificity emerge?

Plants and animals have solved the problems of being multicellular in different ways. Eukaryotic photosynthesis evolved some 2,000 million years ago in the oceans. The ubiquity of light over the surface of the globe is thought to have been responsible for a major evolutionary decision by the primordial plant eukaryotic cell; to remain sessile and, as a consequence, to tolerate inevitable predation1,2. When plants invaded the land, they found the supply and distribution of water, minerals and light much more variable than in the oceans. Among the primary advances made on land were an elaboration of tip branching and the evolution of a differentiated modular structure. The module elements — leaves, buds (dormant meristems), flowers, abscission zones and branch roots — are reiterated many times during development, as are their signal-transduction capacities. Such modularity ensures that predation and environmental damage are minimized because some modules usually survive to regenerate the individual. In general, tissue and cell functional specialization is minimized in plants to limit fatal predatory damage. However, some distribution of functions among different cells is

required; for example, the specialized cellular structure of the plant vascular system precludes its direct role in photosynthesis. But most plant cells can sense nearly all the signals to which the individual plant responds. Owing to their very different physical environments, however, there are some differences between signals perceived by the root compared with the shoot. How, then, can a plant cell process these myriad signals through to an appropriate response? Here we emphasize the structural and spatial characteristics of plant signal transduction, and conclude that organization emerges from the interrelationships of specific components. Exploitation of growth resources

The growing shoot can accurately perceive gradients of light, and reflected light from leaves is used to detect the position of neighbours3. A three-dimensional image is constructed by the shoot, and growth (and leaf angle) is redirected if necessary to optimize light capture. Each shoot cell acts like an indi-

vidual ommatidium of the insect eye. Below ground, recent observations have shown4 that plants prefer patchily distributed minerals in the soil. Remarkably, the plant can sense the volume of the patch, maximize growth when an optimal volume is sensed and perceive the steepness of the gradient across the patch boundary. How these soil variables are perceived is not understood4. But single plant cells can sense very slight gradients of many environmental factors (BOX 1). Perception of these important plant resources takes place within the context of an environment that changes from minute to minute. At least 17 environmental variables are sensed (FIG. 1), and each can modify the response to the others5. A complex array of external information is therefore either summed or integrated, whereas between other groups of variables, synergistic interactions are common6–8.

“What is required of plantcell signal-transduction studies … is to account for the capacity for ‘intelligent’ decision-making; computation of the right choice between close alternatives.”

Box 1 | Single cells can sense fine gradients The classic example of fine sensing is the zygote of the marine alga Fucus. This single cell can respond to remarkably slight gradients in temperature, osmotic pressure, light, pH, minerals (K+, Ca2+), solution flow, electrical fields, other chemicals, gases and probably gravity, and direct the orientation of growth accordingly many hours later69. These gradients usually have a narrow time window in which they are sensed by the single cell, although in a population of zygotes this window is stochastically distributed around a mean value of several hours. By contrast, the sperm entry site can be remembered for at least a day and be used to specify the direction of growth if no other cues have been detected70. A similar remarkable sensitivity to signal gradients is shown by single-celled euglenoids, which can sense their own cytoplasmic weight and modify swimming activity71.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | APRIL 2001 | 3 0 7

PERSPECTIVES

Exogenous signals

Endogenous signals

Light

Growth regulators

• Quality • Quantity • Duration • Direction

• Cytokinin • Ethylene • Gibberellin • Auxin • Abscisic acid • Brassinosteroids

Mechanical Variable • Wind • Herbivores Constant • Substrate • Support structures

Mechanical • Growth-related tissue compression and tension

Defence signals

Atmospheric humidity Other plant proximity Temperature Soil nutrients Soil water Pathogens Gravity CO2

• Jasmonic acid • Salicylic acid

Developmental regulators • Mobile RNAs

Metabolites • Sugars • Glutamate

contiguous wall. Because the cytoplasm contains a turgor pressure of about eight atmospheres, compression and tension gradients are common. Nodal points of tension/compression can be expected to elicit changes in development12. Such mechanical signals can act as specific morphogens13 and can contribute to the polarized nature of growth and development. Proteins that alter the mechanical characteristics of the cell wall, such as the recently described expansins14,15, might therefore act as plant-specific developmental regulators. Mechanical sensing by higher plants is extremely sensitive, and only slight movement or touch is necessary to induce immediate responses in cytoplasmic calcium16. The developmental production of successive modules can also programme changes in signal sensing. During cereal root development, for instance, the lateral roots grow horizontally at first, only later assuming a characteristic vertical direction17. Successive roots become progressively more vertical with respect to gravity, leading to a network of roots that efficiently mines the local soil around the stem.

Regulation of growth and development

Figure 1 | A wide range of disparate external and internal signals is monitored by plants and used to compute appropriate developmental responses. The molecular elements of the plant sensory apparatus and signal-transduction systems can integrate these signals and reach a finely balanced decision as to how to grow and develop to most successfully survive and exploit the environment. As plant responses are generally irreversible growth responses, these signalling systems must compute each developmental decision with extreme care.

Decisions about exploitation of basic nutrient resources can be made by plants before any nutritional benefit is derived. Dodder, a parasitic plant, can sense the level of circulating nutrients when it first touches a putative host9,10. Within one hour, it ‘decides’ whether it is worth initiating a developmental programme, which involves shoot-coiling around the host and the formation of haustoria several days later. Rejection of the putative host is frequent. Once haustoria penetrate the host vascular system, nutrients are gained and used for growth. Remarkably, the number of coils of the parasite around the host stem reflects with some accuracy the nutrients in the host and the likely subsequent return in growth resources. What is required of plantcell signal-transduction studies, then, is to account for the capacity for ‘intelligent’ decision-making; computation of the right choice between close alternatives. To exploit patchily distributed environmental resources, dormant meristems can be activated, and individual growing meristems on a single plant often show striking degrees of independence in growth and signal response.

308

A network of growing and branching meristems is constructed, and this efficiently mines local light, minerals and water. The overall organization of signal sensing–response and interactions between the growing regions is thus democratically arranged, with no overarching, controlling tissue like a nervous system. Competition between growing points is common. However, the resultant architecture of the plant is invariably highly functional, indicating that these sensing and control systems must also be highly coordinated. Internal signals: more complexity

A plethora of internal signals circulate around a vascular system in which flow rate can vary from minute to minute. These signals include growth regulators, ions, wall fragments, sugars, water and amino acids, all of which can modify development11. The degree of fine vascular branching can be limited and many cells can be 10–20 cells away from direct contact. So these active agents often arrive at individual growing cells in a polarized manner and might be perceived as gradients across them. Plant cells are permanently joined through a

Phenotypic plasticity

Because plants usually have little choice over their immediate growth environment, an ability to modify development to cope with an environment of enormous variability is believed to increase fitness. Phenotypic plasticity — that is, the capability of a single genotype to generate many phenotypes — is a pronounced and unusual characteristic of plant development1. It is also a crucial feature of plant-cell signal transduction. Specific phenotypic adaptations in morphology, physiology, anatomy, development, reproductive timing, breeding systems and offspring developmental patterns have all often been observed18. Enormous variability in module numbers is common. One view is that a direct coupling of signal transduction to gene expression regulates plasticity. However, the mechanism might not be straightforward and epigenetic processes or even cell individuality, as we indicate later, might be crucial to the response. Some aspects of development and morphology are strongly resistant to environmental variation. Numerous complex feedback controls must therefore be operative, but detection of these is clearly in its infancy19,20. Furthermore, redundancy in control elements will help strengthen reliability in the face of environmental disruption. Redundancy was an early control feature introduced into computer design to ensure reliable performance. Polyploidy seems to have had an important www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

PERSPECTIVES

-5

2.5

The importance of individuality

–100

7.5

12.5

Cytoplasmic calcium (nM)

–600

–200

1000 -

750 500 -

250 -

0Day: 10

Night: 5

–40

d 80

Amylase

% cells

60 8 40

6 4

20 2 0

0 11

9 8 –log [GA]

Root numbers per segment

α-amylase secreted from tissue (units mL–1)

The term individuality is used to describe situations in which morphologically similar cells, tissues or plants show non-similar or unique responses to signals. Commonly, individuality can be identified in situations in which development is ‘all-or-none’21. The cell or tissue does, or does not, respond to the inducing signal; flowering, abscission, germination, bud break and root formation are good examples. In these cases, an increasing strength of stimulus (light photoperiods, growth regulator concentrations) leads to a response from more of the population. The dose–response curve therefore represents the different sensitivities of the individuals of the population to the stimulus22. Cells of the stomatal complex23, aleurone24,25 and pericycle26,27 have all been observed as heterogeneous populations in a single tissue. When the concentration of a modifying stimulus such as auxin, gibberellin or abscisin is increased, progressively more cells respond (FIG. 2). Each cell therefore has its own sensing threshold and, when this is exceeded, a response is initiated. However, there is also variation in the lag period before individual cells respond, and in the duration of the response23,26. Individuality in regulation of the lac operon in bacterial populations was observed many years ago28. Partial expression of the lac operon represented simply the numbers of bacteria that had made the transition. Explanation of such individuality probably lies in certain stochastic processes during plant-cell development. The cytoplasm of a mature plant cell is little more than a few picolitres in volume and contains about 20% protein. Depending on the cell type and the signalling pathway, the numbers of molecules in each cell concerned with signal transduction and the control of gene expression are estimated to range from single figures to under a thousand29. Predictions of cellular properties are usually based on the assumption that the cytoplasm is a homogeneous, relatively dilute solution, containing statistically large numbers of molecules; cellular kinetics are assumed to rely on concentration and equilibrium constants to determine interactions. In neither case is this true for the cell. How accurately, for example, can a cell control the amount or behaviour of regulatory proteins or transcription factors that number

>1000

ABA-induced increase in cytosolic calcium above resting (nM)

% secreting cells

role in genome evolution in angiosperms. The common presence of several copies of genes (and gene products) — and thus potential redundancy in the plant genome — might be a reaction to the complexity of the environment as plants perceive it.

2 4 Auxin-time (days)

Figure 2 | Plant cell signalling systems show characteristics of cell-to-cell individuality, threshold phenomena and environmental entrainment. a | Cytosolic Ca2+ changes in stomatal guard cells of Arabidopsis thaliana treated with abscisic acid (ABA). Ca2+ levels were monitored by confocal ratio imaging of the Ca2+-sensitive ratiometric dye Indo-1 dextran microinjected into the cells. Note the spatial and temporal variability in the Ca2+ increase induced by ABA within a single cell. Calcium levels are pseudocolour-coded according to the inset scale. Numbers reflect time after addition of 20-mM ABA. Scale bar represents 5 µm (S.G., unpublished data). b | The precise kinetics of plant signalling systems can be environmentally determined. The maximal ABA-induced Ca2+ increase in guard cells of Commelina communis depends on the previous growth conditions of the plant. The lower the growth temperature, the less the guard cells seem to use a Ca2+-dependent ABA signalling system. (Data redrawn from REF 83.) c | Dose–response curve of α-amylase secretion from barley aleurone induced in response to gibberellin (GA) shows threshold phenomena for plant-cell responses. The aleurone is a secretory tissue that responds to the growth regulator gibberellin by producing hydrolases as part of the reserve mobilization system that supports cereal grain germination. Note the dose response of the whole tissue reflects the recruitment of more cells with higher threshold for activation to the secreting population as the gibberellin levels are increased24,25. d | Root segments (Haplopappus ravenii) were exposed to auxin (5 × 10–7 M) for the numbers of days shown (auxin-time) and then placed in auxin-free media for the remainder of the 6day total incubation period when lateral roots were counted. As roots are formed by division from root pericycle cells, these data (redrawn from REF. 27) indicate substantive time variation in sensitivity to auxin. The first derivative of these data produces a bell-shaped curve indicating stochastic variation in time sensitivity to auxin among a supposedly uniform tissue27.

fewer than ten to a hundred? Can we comprehend regulation when only a few dozen molecules are involved, and in cases where stochastic or chaotic events could be crucial in determining the outcome? How will environmental variation during cell or tissue specification modify the partition or synthesis of such small numbers of proteins at crucial cell cycles? Even when dealing with several thousand protein molecules, are cellular processes sufficiently accurate to ensure an identical number of copies between different cells? And

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

there are at least 10,000 cellular proteins. It is at crucial stages of cell specification that individuality emerges. The adult plant originates with the first division of the zygote. Many tissues in plants originate from just one or a few cells. And particular cells within tissues such as leaf guard cells certainly originate from single cells. Variations in the small number of crucial transcription factors at any of these stages will ensure the individuality that is subsequently observed. Such epigenetic ‘noise’ could be consid-

VOLUME 2 | APRIL 2001 | 3 0 9

PERSPECTIVES

Box 2 | Biological advantages to individuality Individuality allows phenotypically plastic responses to the environment. Plants can adjust the numbers of branch roots to best fit the prevailing circumstances72. Variation in individual aleurone cells allows amylase production in the germinating cereal seedling to be adjusted according to variable germination circumstances24–26. Trees can optimize the number of leaves to a reduced water supply simply by abscising the excess73. In the case of leaf guard cells, subpopulations more sensitive than others to light, abscisic acid, water deficit, humidity or carbon dioxide, for example, allow the leaf to optimize water relations26 by using different combinations of the sub-populations of guard cells. Individuality in signal transduction also allows plants to deal more effectively with herbivores and disease. The same herbivore signal causes different induced responses in different plants of the same species and in different tissues of the same individual74. Every leaf can assume a different phenotype owing to the expression of resistance genes. And even though each step might be transcriptionally regulated, the net effect of the induced response might seem random, without detailed knowledge of the position, age, history and chemical environment of the affected tissue. This moving target, n-phenotype strategy74 is a crucial example of plant individuality and plasticity.

ered irrelevant, a biological nuisance, but plants probably engineer such variation because it allows a graded response from the population of plants, tissues or cells and thus increases fitness (BOX 2). Clearly, individuality forms a basis for phenotypic plasticity in terms of numbers of roots, flowers or leaves. But during tissue specification, the crucial transcription factor numbers might instead ensure the production of a small population of mother cells with different potentialities. According to environmental conditions at the time, one or other mother cell could be cloned to produce the tissue most relevant to prevailing circumstances.

“Although there are many sites within the cell where signal integration and processing can occur … the unusual properties of the plant-cell plasma membrane make it a prime candidate for the location of the ‘cellular computer’.” Individuality in calcium signalling

Changes in cytosolic Ca2+ are recognized as ubiquitous regulators of cell function and provide some of the clearest indications of the individual behaviour of plant signalling molecules when viewed at the single-cell level. Calcium responses induced by the same signal are rarely identical between any two plant cells of the same type22,30,31. Such individuality probably results from the low numbers of channels and receptors involved in Ca2+ entry.

310

This cell-specific behaviour suggests that the cell orchestrates the essential Ca2+ kinetics of the signal necessary to elicit the appropriate response, rather than the reverse32. However, a further unusual property of plant signalling systems is that signals usually induce the synthesis of the proteins that are involved in mediating the response. One obvious example is that increases in the levels of Ca2+ induce the synthesis of calmodulin, but many others (phospholipases, calmodulin-like domain protein kinases, mitogen-activated protein kinases, resting levels of second messengers and so on) have been recorded33–35. One explanation is the construction of new signal-transduction equipment in each cell, designed to take account of the new circumstances after the first set of signals. As such, these changes in expression would represent a cellular ‘memory’ of the environmental history of the cell, perhaps providing a molecular explanation for how a plant can incorporate its growth history into its future developmental decisions. In addition, a form of cellular learning34 takes place because increased information flux through cytosolic Ca2+ should result. An alternative is that more cells might be slowly recruited into a signalling mode, so the answer lies in explanations from individuality. Whatever the functional basis for this signal-induced synthesis of signalling proteins, it seems an unusual and widespread feature of plant signalling systems worthy of further investigation. Principles of perception

Occupied receptors are usually considered the start of any signal-transduction network, and our knowledge of plant receptors has advanced considerably over the past 5–10 years. Small families of receptors for red and blue light, ethylene and brassinosteroids have been isolated32,36–38 (FIG. 3). Sugars, which can

act as internal plant morphogens, might be sensed by hexokinase39. Receptor-like kinases are prominent in the Arabidopsis thaliana genome40 and, with their putative ligands, they are thought to mediate processes such as incompatible pollen–stigma interactions41 and the maintenance of meristem structure19. Phytochrome, the red/far-red-light sensor, has some characteristics of a two-component-like phosphorylation system similar to those in bacteria, although it acts as a serine/threonine rather than histidine kinase5,25. The blue-light sensors cryptochrome and NPH (non-phototropic), which use flavins or pterins as chromophores, might couple into redox systems42,43. Ethylene is one of five main growth regulators, and its receptors (such as Eth-1) have been characterized as histidine kinases similar to bacterial two-component signalling systems44. A candidate cytokinin receptor has also recently been identified as a histidine-kinase-like protein45. A close relationship between auxin transport and perception has been predicted, and this may be clarified now that candidates for auxin receptors and auxin-transport proteins have emerged46,47. Despite such successes, however, we still lack specific candidate receptors for the growth regulators gibberellin and abscisin. But structure–activity relationships indicate that all of these growth substances might be sensed through proteinaceous receptors. Nearly all the receptors shown in FIG. 3 are located in the plasma membrane. The exceptions — for example, phytochrome and cryptochrome — must be fixed to some specific spatial cellular domain because cells can sense gradients of light. Sites near the plasma membrane, fixed perhaps to an attached cytoskeleton, have been suggested32. The plasma membrane as a computer

Perception of signals is, however, more complex than the limited families of receptors indicated above might suggest. For example, in the case of light, not only can red and blue light be easily distinguished, but plant cells can assess the total quantity of light received, the direction from which the light comes, the intensity during exposure, the time (minutes to many hours) that light was available, and the temporal order in which red or blue light has been perceived48,49,50. It has been speculated that an unknown group of PAS/kinase proteins revealed by the Arabidopsis genome initiative51 could be a new class of photoreceptors. However, it is likely that many of these complex light-perception events are done through interactions between the small number of receptors already identified. www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

PERSPECTIVES cell where signal integration and processing can occur, such as organelles, cytosol, cytoskeleton and endomembranes, the unusual properties of the plant-cell plasma membrane make it a prime candidate for the location of the ‘cellular computer’. Probably 500–1,000 proteins and enzymes at very high density are embedded or attached to the plasma membrane11,52. These include receptors (FIG. 3), protein kinases53,54, ion channels55, microfilament anchorage and signal-transduction proteins involved in second-messen-

The timing, direction and quantity characteristics are almost certainly shared in the perception of growth regulators, nitrate, water, gravity, temperature and mechanical signals. Whatever the receptors for these latter signals, transduction mechanisms have to account for a complexity of perception not easily explained by single classes of receptor. Furthermore, as indicated earlier, it is necessary to account for an ability to integrate many signals and to compute ‘intelligent’ decisions. Although there are many sites within the

Pathogenic Brassinosteroid bacteria elicitor

Ethylene

LRR

Ethylene

S-S S-S

(CKI1)

GAF His kinase

Cu Cu

GAF

GAF His kinase

(ETR1, ERS1)

Cu Cu

His kinase

GAF

His kinase

(CKI1)

S-S S-S

Cu Cu

His kinase Receiver

His kinase

(BRI1)

Receiver

Ser/Thr kinase

(XA21)

Receiver

Ser/Thr kinase

(GCR1, putative G-protein coupled receptor)

Receiver

His kinase

Cu Cu

Receiver

LRR

Cytokinin?

(ETR2, ERS2, EIN4)

Sugars (Hexokinase)

Abscisic acid

Gibberellic acid

CLAVATA3 Blue light 96 aa peptide LRR

LRR

(CLAVATA1)

Red/far red light Tetrapyrolle chromophore Phytochrome homodimer (PHYA–PHYE)

LOV2 LOV2

Plasma membrane associated by unknown mechanism

Ser/Thr kinase

FMN chromophore

Ser/Thr kinase

(NPH1) PAS

PAS

Ser/Thr kinase

PAS

Ser/Thr kinase

Auxin Glycosylation (ABP1)

Box A

Box B

KDEL

Blue light Pterin FAD chromophore chromophore

(CRY1, CRY2)

Structurally similar to photolyases

Tetrapyrolle chromophore

Figure 3 | The domain structures of several known plant receptor proteins, putative receptors or components of putative perception complexes, and their respective ligands. His kinase, histidine protein kinase domain; KDEL, endoplasmic-reticulum retention sequence; LOV, light/oxidation/voltage sensor-like protein domain; LRR, leucine-rich-repeat motif; FMN, flavin mononucleotide; FAD, flavin adenine dinucleotide; PAS, carboxy-terminal structural PAS repeat domain; Receiver, domain homologous to bacterial two-component signalling system receiver proteins; Ser/Thr kinase, serine/threonine protein kinase domain; ? unknown protein. For more detailed discussion of the structure and function of these receptors and putative receptors see: phytochromes (PHYA–E)38; cryptochromes (CRY1,2)42; auxin (ABP1)47; phototropin (NPH1)43; hexokinase39; LRR protein kinases as receptors for brassinosteroids (BRI1) and pathogens (XA21)84; ethylene (ETR1)44; cytokinin (CRE1, CKI1, GCR1)45,85; CLAVATA186 (also see the review by Steven E. Clark on page 276 of this issue); and references therein.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

ger production32. Scaffold proteins on the cytoplasmic face of the membrane similar to ankyrin are probably also present because ankyrin-binding regions in some plant proteins have been identified. The domain on the cytoplasmic surface is likely to be hydrophobic, with limited numbers of water molecules encouraging protein–protein interactions. Early electron microscope studies indicated that complexes of proteins were present and that the whole membrane was enormously differentiated56. Limited identification of these complexes has been made, although some are probably transient and formed only after signalling has commenced (BOX 3). Current views on plasma membrane behaviour suggest a fluid mosaic structure52. But these models are often derived from motile animal cells and there are reasons (such as polarized cell growth and tissue morphology, sensing the direction and gradients of incoming signals, lack of cell mobility) that indicate that many functions in plant cells might be fixed rather than mobile. Even if large protein complexes are effectively free to move, diffusion will be extremely slow. The wall provides an obvious anchor for proteins, particularly for those that straddle the membrane52. The plasma membrane is under turgor pressure and is compressed against the wall. Movement of proteins will be hindered by wall constituents and thus membrane fluidity will be reduced by the pressure. Changes in turgor (for example, from hypo-osmotic shock) or bending of the cell will concomitantly alter the conformation of structurally attached proteins by stretching or otherwise deforming the bilayer. Such treatments result in immediate transients in cytosolic Ca2+ (REFS 57,58). The implication is that channel proteins are either directly or indirectly anchored to the wall as the Ca2+ involved enters from outside the cell. Changes in wall–membrane protein interactions could provide the rapid channel gating observed under these conditions. The large numbers of protein kinases and phosphatases found in cells present serious problems for fluid mosaic models. At least 1,000 protein kinases are present in the Arabidopsis genome51 and the density at membrane surfaces is probably very high. For any signal to navigate, with fidelity, through the forest of protein kinases and phosphatases, requires severe spatial constraints on plasma membrane protein kinases to ensure specific modification of protein substrates32. Some kinases might be permanently tethered to scaffolds constructed around the plasma membrane and the associated cytoskeleton, but others might transiently

VOLUME 2 | APRIL 2001 | 3 1 1

PERSPECTIVES

Box 3 | Stable and transient protein complexes There are over 1,000 protein kinases in plant cells and, for a signal to navigate correctly its way through this morass of transduction proteins, spatial location is essential. Stable connections of protein kinases to the plasma membrane involve farnesylation, myristoylation or prenylation; less permanent ones involve autophosphorylation, phosphorylation of scaffold proteins, or localized high concentrations of activating second messengers. Simple, calcium-induced transduction complexes — such as prenylated calmodulin targeted to the plasma membrane67, or a membrane-associated calmodulin-like domain protein kinase (CDPK) regulating a membrane transporter54 — are probably dwarfed by semi-permanent structures such as those described in animal cells by caveolae and rafts75,76. These structures (transducons11) are nucleated around particular membrane lipids or even scaffold-like proteins, and contain many plant-cell transduction proteins that are involved in phosphoinositide production and phospholipid modification77,78, Ca2+-signalling proteins, calmodulin, kinases, water channels, nitric oxide synthase, anchorage proteins and some enzymes. The constituents of these transducons are dynamic, moving in and out of the complex after signalling. Although researchers have barely begun to define these structures in plants, the Cop9 signalosome (an eight-subunit complex regulating de-etiolation, and controlled by phosphorylation79) is a clear case for a stable plant-transduction complex. Three kinds of less stable signalling complex — but nonetheless associated with the plasma membrane — have also been reported. And many more can be expected. Complexes can form around pleckstrin homology (PH) domains80 in plant cells, and about ten genes in the Arabidopsis thaliana genome contain a PH-like domain, including protein kinases81. The PH domain binds to phospholipids, and aggregation is usually initiated by phosphorylation or autophosphorylation resulting from receptor occupation. The aggregate can ensure substrate activation or phosphorylation leading to the initiation of, for example, mitogen-activated protein kinase (MAPK) cascades35. A second set of less stable complexes has been reported to form with ClAVATA, Rop GTPase, other regulatory proteins and MAPKs40. Finally, the 14-3-3 proteins are represented in plant cells by a family of about ten genes. Usually such proteins cross-link others after phosphorylation, and CDPK has been reported to activate 14-3-3 proteins, which are probably involved in controlling ATPase activity within the plasma membrane82.

membrane allow summation, integration and computation of electrical properties; just as they do in nerve cells5,59,60. Like many other aspects of plant life, action potentials in plants are slow compared with those in animals. But even slower again, and lasting minutes, are the very pronounced transient falls in membrane

connect with their substrates only after specific binding sites have been exposed. Electrical properties

Although few plants use action potentials for communication, in those that do, the enzymatic and electrical properties of the plasma

GA Sly1 (transcription factor)

Root growth

Abi1 (Protein phosphatase)

ABA IAA

Ein2 (Putative metal transporter)

Jasmonic acid responses

Ein2 (Putative metal transporter)

Abi1

Ctr1 (Raf-like kinase)

Abi3 (Transcription factor)

Ctr1 (Raf-like kinase)

Etr1 Etr1 (receptor) (receptor)

Seed germination

Ein3 (Transcription factor)

C2H4 Transcriptional responses

C2H4

Oxidative stress responses

Cytokinin

Figure 4 | Plants show extensive cross-talk and interactions between signalling systems. Recent genetic analysis of the physiological responses of mutants of Arabidopsis thaliana has uncovered possible molecular elements of a complex interacting network of control allowing growth regulators, such as auxin (IAA), cytokinin, ethylene (C2H4), abscisic acid (ABA) and gibberellin (GA), to interact in the regulation of root growth, stress and defence responses (such as oxidative stress and jasmonic-acid responses), and seed germination87–89. Despite its complexity, this is a simplified view of the true regulatory interactions that occur in the plant cell.

312

potential that might be regarded as a kind of pseudo action potential. Many plant-cell signals induce these changes in membrane potential; auxin, gravity, abscisic acid, blue light and red light are excellent examples5,50. These electrical changes are unlikely to be used for cell–cell communication. But the electrical changes will be as profound on the properties of the plasma membrane as those of the genuine action potential itself. Furthermore, a change in membrane potential should allow signal integration as observed for action potentials. Much electrophysiological information indicates the presence of voltage-gated channels in the plant plasma membrane61. As the membrane potential changes, the channel proteins undergo conformational changes that promote opening (or closing) and subsequent altered ion flux. However, there is no reason to think that channels are the only proteins to undergo electrically dependent structural change. On the cytoplasmic face of the plasma membrane, higher rates of ion flux will radically alter the ionic strength, particularly near the mouth of channels; the availability of water will be changed, modifying protein–protein interaction and protein complex status; electrical changes will modify the three-dimensional conformation of many proteins, exposing groups for phosphorylation/phosphatase action and alterations in surface charge might even alter membrane lipid mobilities. We propose that one important result will be to modify receptor phosphorylation and diversify receptor behaviour. Phosphorylation alters both conformation and function and can generate proteins with different activities according to sites and numbers of phosphorylated amino acids. A simple feedback loop is closed in which perception modifies subsequent perception. The structure of many protein complexes and channels might be altered for extended periods by phosphorylation. Mechanisms for timing of the signal exposure, estimates of the quantity of signal arriving and for long-term modifications of plasma membrane function can therefore easily be constructed. In cases that there are no obvious receptor proteins (such as for water or nitrate), transporting proteins themselves (in this case, the water channels or nitrate transporters) might provide the necessary basis for perception11,62. The significance of a change in membrane potential itself to signalling is supported by older observations, which showed that many organic chemicals, thiol (SH)-group reagents and respiratory inhibitors can break seed and bud dormancy, or induce root formation for www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2

PERSPECTIVES example. The common property linking these chemicals is a demonstrated change in membrane potential2,63,64. Signal computation with attitude

If the plasma membrane is highly structured and activation of channels is localized, then a means to coordinate the behaviour of the membrane becomes necessary. Changes in Ca2+ fluxes across this membrane provide one example of how such coordination could be effected. The basis of Ca2+ signalling is the separation of different concentrations across a membrane, energized by Ca2+ ATPases. Upon signalling, channels open in the requisite membrane, allowing Ca2+ to move down its electrochemical potential gradient. Cytosolic Ca2+ has three properties that make it ideal for plasma membrane coordination. Calcium is not a mobile ion in the cytoplasm65; Ca2+ signals move as waves that are thought to be constrained to the cytoplasmic surface of the membrane33, and these waves are usually initiated at specific cellular sites57.

“The same panoply of building blocks can be used in transduction between both plants and animals but, by changing their relationships and interactions, different properties will emerge.” When plant cells are subjected to several signals, they seem able to access different sources of cytosolic Ca2+, producing a different response to each signal57. Single-cell imaging of Ca2+ changes in response to different signals confirms these observations30. A highly structured arrangement of channels, Ca2+ stores and wave direction is implied. These waves might have complex fractal-like forms. Only certain discrete spatial regions of the plasma membrane may be activated by Ca2+ elevation57 66 (FIG. 2). By varying the combinations of plasma membrane regions that are activated, considerable potential for the computation of signals emerges. Several important Ca2+-binding proteins, such as calmodulin66,67 and CDPK53,54, are attached to the plasma membrane. Evidence that structural rearrangements of the plasma membrane result from signalling

can be deduced from an experimental separation of signals from the associated Ca2+ transients and physiological effects. Plant cells given a hyperosmotic shock58 or exposed to red light68 will normally express some transient increase in cytosolic Ca2+. However, if the signal is imposed in the absence of extracellular Ca2+, no Ca2+ transient is observed. The physiological response and the Ca2+ transient are delayed until Ca2+ is added back to the cells, when both progress normally. Some ‘excited’ state is induced by the initial signal; this lasts 20 minutes with hyperosmotic shock and up to 4 hours with red light. Future directions

Fundamentally, life is organization. The cell is a product of the special properties that emerge from the complex interactions and spatial structures between the many thousands of molecules and enzymes of which it is composed. The same panoply of building blocks can be used in transduction between both plants and animals but, by changing their relationships and interactions, different properties will emerge. The plasma membrane (perhaps more so in a plant cell than others) acts as a relatively permanent structure on which many kinds of transduction structure can be made. This might represent part of the answer to the question posed in the Preface. Emphases on spatial relationships and cross-talk8 between signalling pathways seems to be crucial. The completion of the sequencing of various plant genomes will provide us with a phenomenally rich array of candidate regulators of plant-cell function. The direction now must be to define which molecules interact with each other and where these interactions occur in vivo. Emerging technologies, such as green fluorescent protein fusions, live cell imaging of fluorescence resonance energy transfer and fluorescence lifetime imaging31, are beginning to approach these questions in the only setting where these interactions can really be determined — the intact, functioning cell. By this means, we will slowly unravel the network of connections (FIG. 4) that provides unity to cellular and plant activities and that is undoubtedly present. The particular properties of the living cell are shared in some way or another through every constituent molecule, forming a highly integrated regulatory network. Equally, the environmental context, whether from outside the plant or from within, contributes to shaping how information is processed by each cell. In trying to understand signal transduction, we are doing no more than trying to understand life itself.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

Simon Gilroy is at the Biology Department, Penn State University, University Park, Pennsylvania 16802. USA. Anthony Trewavas is at the Institute of Cell and Molecular Biology, Kings Buildings, University of Edinburgh, Edinburgh, EH9 3JH, Scotland. Correspondence to A.T. e-mail: trewavas@ed.ac.uk

Links FURTHER INFORMATION Trewavas lab home

page | Gilroy lab home page ENCYCLOPEDIA OF LIFE SCIENCES Plant

growth factors and receptors | Plant plasma membrane

9. 10.

11.

12.

13.

14.

15. 16.

17.

18.

19.

20.

21.

Trewavas, A. J. in Plasticity in Plants (eds Jennings, D. & Trewavas, A. J.) Symp. Soc. Expt. Biol. Med. 40, 31–77 (1986). Trewavas, A. J., Sexton, R. & Kelly, P. Polarity calcium and abscission: molecular bases for developmental plasticity in plants. J. Embryol. Exp. Morphol. 83, S179–S195 (1984). Smith, H. & Whitlam, G. C. The shade avoidance syndrome: multiple responses mediated by multiple phytochromes. Plant Cell Environ. 20, 840–844 (1997). Wijensinghe, D. K. & Hutchings, M. J. The effects of environmental heterogeneity on the performance of Glechoma: the interactions between patch scale and patch contrast. J. Ecol. 87, 860–872 (1999). Trewavas, A. J. in Biochemistry and Molecular Biology of Plants (eds Buchanan, R. B., Gruissem, W. & Jones, R. L.) 930–988 (Am. Soc. of Plant Physiologists, Rockville Maryland, 2000). Jenkins, G. I. Signal transduction networks and the integration of responses to environmental stimuli. Adv. Bot. Res. 29, 54–74 (1999). Novoplansky, A. Developmental responses of Portulaca seedlings to conflicting spectral signals. Oecologia 88, 138–140 (1991). Novoplansky, A., Cohen, D. & Sachs, T. Ecological implications of correlative inhibition between plant shoots. Oecologia 82, 490–493 (1990). Kelly, C. E. Resource choice in Cuscuta europea. Proc. Natl Acad. Sci. USA 89, 12194–12197 (1992). Kelly, C. K. Plant foraging: a marginal value model and coiling response in Cuscuta subinclusa. Ecology 71, 1916–1925 (1990). Gilroy, S. & Trewavas, A. J. in The Plant Plasma Membrane (eds Larsson, C. & Moller, I. M.) 203–233 (Springer, Berlin, 1990). Trewavas, A. J. & Knight, M. R. Calcium, mechanical signalling and plant form. Plant Mol. Biol. 26, 1329–1341 (1994). Green, P. B. Expression of pattern in plants: combining molecular and calculus-based biophysical paradigms. Am. J. Bot. 86, 1059–1076 (1999). Fleming, A. J., Caderas, D., Wehrli, E., McQueen-Mason, S. & Kuhlemeier, C. Analysis of expansin-induced morphogenesis on the apical meristem of tomato. Planta 208, 166–174 (1999). Cosgrove, D. J. New genes and new biological roles for expansins. Curr. Opin. Plant Biol. 3, 73–78 (2000). Knight, M. R., Knight, H. & Watkins, N. J. Calcium and the generation of plant form. Phil. Trans. R. Soc. Lond. B 350, 83–86 (1995). Firn, R. & Digby, J. Solving the puzzle of gravitropism — has a lost piece been found? Planta 203, S159–S163 (1997). Sultan, S. E. Phenotypic plasticity for plant development, function and life history. Trends Plant Sci. 5, 537–542 (2000). Brand, U., Fletcher, J. C., Hobe, M., Meyerowitz, E. M. & Simon, R. Dependence of stem cell fate in Arabidopsis on a feedback loop regulated by CLV3 activity. Science 289, 617–619 (2000). School, H. et al. The stem cell population of Arabidopsis shoot meristem is maintained by a regulatory loop between the clavata and wuschel genes. Cell 100, 635–644 (2000). Trewavas, A. J. in Plant Responses to Environmental Stresses (ed. Lerner, H. R.) 27–42 (Marcel Dekker, New

VOLUME 2 | APRIL 2001 | 3 1 3

PERSPECTIVES York, 1999). 22. Trewavas, A. J. How do plant growth substances work? (II). Plant Cell Environ. 14, 1–12 (1991). 23. Mott, K. A. & Buckley, T. N. Patchy stomatal conductance emergent collective behaviour. Trends Plant Sci. 5, 258–262 (2000). 24. Hillmer, S., Gilroy, S. & Jones, R. L. Visualising enzyme secretion from individual barley protoplasts. Plant Physiol. 102, 279–286 (1993). 25. Ritchie, S., McCubbin, A., Ambrose, G., Kao, T.-H. & Gilroy, S. The sensitivity of barley aleurone tissue to gibberellin is heterogeneous and may be spatially determined. Plant Physiol. 120, 361–370 (1999). 26. Bradford, K. J. & Trewavas, A. J. Sensitivity thresholds and variable time scales in plant hormone action. Plant Physiol. 105, 1029–1036 (1994). 27. Blakeley, L. M., Rodaway, S. J., Hollen, L. B. & Croker, S. G. Control and kinetics of branch root formation in cultured root segments of Haplopappus ravenii. Plant Physiol. 50, 35–42 (1972). 28. Novick, A. & Weiner, M. Enzyme induction as an all or none phenomenon. Proc. Natl Acad. Sci. USA 43, 553–566 (1957). 29. McAdams, H. H. & Arkin, A. Genetic regulation at the nanomolar scale. Trends Genet. 15, 65–69 (1999). 30. Gilroy, S., Fricker, M., Read, N. & Trewavas, A. J. Role of calcium in signal transduction of Commelina guard cells. Plant Cell 3, 333–344 (1991). 31. Blancaflor, E. B. & Gilroy, S. Plant cell biology in the new millennium: new tools and new insights. Am. J. Bot. 87, 1547–1560 (2000). 32. Trewavas, A. J. & Malho, R. Perception and transduction: the origin of the phenotype. Plant Cell 9, 1181–1195 (1997). 33. Trewavas, A. J. Le calcium c’est la vie. Calcium makes waves. Plant Physiol. 120, 1–6 (1999). 34. Trewavas, A. J. How plants learn. Proc. Natl Acad. Sci. USA 96, 4216–4218 (1999). 35. Meskienne, I. & Hirt, H. MAP kinase pathways: molecular plug and play chips for the cell. Plant Mol. Biol. 42, 791–806 (2000). 36. McCarty, D. R. & Chory, J. Conservation and innovation in plant signalling pathways. Cell 103, 201–209 (2000). 37. He, Z. et al. Perception of brassinosteroids by the extracellular domain of the receptor kinase BRI1. Science 288, 2360–2361 (2000). 38. Genick, U. K. & Chory, J. Red light sensing in plants. Curr. Biol. 10, R651–R654 (2000). 39. Smeekens, S. Sugar-induced signal transduction in plants. Annu. Rev. Plant Physiol. Plant Mol. Biol. 51, 49–81 (2000). 40. Torii, K. U. Receptor kinase activation and signal transduction in plants: an emerging picture. Curr. Opin. Plant Biol. 3, 361–367 (2000). 41. Schopfer, C., Nasrallah, M. & Nasrallah, J. The male determinant of self incompatability in Brassica. Science 286, 1697–1700 (1999). 42. Cashmore, A. R., Jarillo, J. A., Wu, Y.-J. & Liu, D. Cryptochromes: blue light receptors for plants and animals. Science 284, 760–765 (1999). 43. Christie, J. M., Salomon, I., Nozue, K., Wada, M. & Briggs, W. R. LOV (light, oxygen, or voltage) domains of the blue-light photoreceptor phototropin (nph1): binding sites for the chromophore flavin mononucleotide. Proc. Natl Acad. Sci. USA 96, 8779–8783 (2000). 44. Stepanova, A. N. & Ecker, J. R. Ethylene signalling from mutants to molecules. Curr. Opin. Plant Biol. 3, 353–360 (2000). 45. Inoue, T. et al. Identification of CRE1 as a cytokinin receptor from Arabidopsis. Nature 409, 1060–1063 (2001).

46. Galwieler, L. et al. Regulation of polar auxin transport by AtPIN1 in Arabidopsis vascular tissue. Science 282, 2226–2230 (1998). 47. Jones, A. M. et al. ABP1: auxin-dependent cell expansion mediated by overexpressed auxin-binding protein 1. Science 282, 1114–1117 (1999). 48. Salisbury, F. B. & Ross, C. W. Plant Physiology 4th edn (Wadsworth, California, 1992). 49. Baum, G., Long, J., Jenkins, G. & Trewavas, A. J. Stimulation of the blue light receptor NPH1 causes a transient increase in cytosolic calcium. Proc. Natl Acad. Sci. USA 98, 13554–13559 (1999). 50. Taiz, L. & Zeiger, E. Plant Physiology 2nd edn (Sinauer Associates, Massachusetts, 1998). 51. Kaul, B. et al. Analysis of the genome sequence of the flowering plant Arabidopsis thaliana. Nature 408, 796–815 (2000). 52. Staehelin, L. A. & Newcomb, E. H. in Biochemistry and Molecular Biology of Plants (eds Buchannan, R. B., Gruissem, W. & Jones, R. L.) 2–50 (Am. Soc. of Plant Physiologists, Rockville, Maryland, 2000). 53. Harmon, A., Gribskov, M. & Harper, J. F. CDPK’s, a kinase for every signal. Trends Plant Sci. 5, 154–159 (2000). 54. Satterlee, J. S. & Sussman, M. R. Unusual membraneassociated protein kinases in higher plants. J. Memb. Biol. 164, 205–213 (1998). 55. Hirsch, R. E., Lewis, B. D., Spalding, E. P. & Sussman, M. R. A role for the AKT1 potassium channel in plant nutrition. Science 280, 918–921 (1998). 56. Volkmann, D. The plasma membrane of growing root hairs is composed of zones of local differentiation. Planta 162, 392–403 (1984). 57. Malho, R., Moutinho, A., Van der Luit, A. & Trewavas, A. J. Spatial characteristics of calcium signalling: the calcium wave as a basic unit in plant cell signaling. Phil. Trans. R. Soc. Lond. B 353, 1463–1473 (1998). 58. Takahashi, K., Isobe, M., Knight, M. R., Trewavas, A. J. & Muto, S. Hypoosmotic shock induces increases in cytosolic Ca2+ in tobacco suspension culture cells. Plant Physiol. 113, 587–594 (1997). 59. Alberts, B. et al. Molecular Biology of the Cell (Garland, New York, 1989). 60. Bose, J. C. Plant Response as Means of Physiological Investigation (Longmans and Co., London, 1906). 61. Sanders, D. & Bethke, P. in Biochemistry and Molecular Biology of Plants (eds Buchanan, B. B., Gruissem, W. & Jones, R. L.) 110–160 (Am. Soc. of Plant Physiologists, Rockville, Maryland, 2000). 62. Zhang, H. & Forde, B. G. An Arabidopsis MADS box gene that controls nitrate induced changes in root architecture. Science 279, 407–409 (1998). 63. Trewavas, A. J. Growth substances in context: a decade of sensitivity. Biochem. Soc. Trans 20, 102–108 (1991). 64. Trewavas, A. J. in The Molecular Biology of Plant Development (eds Smith, H. & Grierson, D.) 8–28 (Blackwell Scientific, Oxford, 1982). 65. Hodgkin, A. L. & Keynes, R. D. Movements of labelled calcium in giant squid axons. J. Physiol. 138, 253–281 (1957). 66. Zielinski, R. E. Calmodulin and calmodulin binding proteins in plants. Annu. Rev. Plant Physiol. Plant Mol. Biol. 49, 697–725 (1998). 67. Rodriguez-Concepcion, M., Yalovsky, S., Zik, M., Fromm, H. & Gruissem, W. The prenylation status of a novel plant calmodulin directs plasma membrane or nuclear localisation of the protein. EMBO J. 18, 1996–2007 (1999). 68. Viner, N., Whitlam, G. & Smith, H. Calcium and phytochrome control of leaf unrolling in dark grown barley seedings. Planta 175, 209–213 (1988).

69. Jaffe, L. F. Localization in the developing Fucus egg and the general role of localizing currents. Adv. Morphogen. 7, 295–328 (1968). 70. Kropf, D. L. Induction of polarity in fucoid zygotes. Plant Cell 9, 1011–1020 (1997). 71. Hader, D.-P. & Hemmersbacj, R. Graviperception and graviorientation in flagellates. Planta 203, S7–S10 (1997). 72. Zhang, H., Jennings, A., Barlow, P. W. & Forde, B. G. Dual pathways for regulation of root branching by nitrate. Proc. Natl Acad. Sci. USA 96, 6529–6534 (1999). 73. Addicott, F. T. Abscission (Univ. California Press, London, 1982). 74. Karban, R. & Baldwin, I. T. Induced Responses to Herbivory (Univ. Chicago Press, Chicago, 1997). 75. Brown, D. A. & London, E. Functions of lipid rafts in biological membranes. Annu. Rev. Cell Dev. Biol. 14, 111–136 (1998). 76. Anderson, R. G. W. The caveolae membrane system. Annu. Rev. Biochem. 67, 199–225 (1998). 77. Munnik, T. et al. Hyperosmotic stress stimulates phospholipase D activity and elevates the levels of phosphatidic acid and diacylglycerol pyrophosphate. Plant J. 22, 147–154 (2000). 78. Stevenson, J. M., Perera, I. Y., Heilmann, I., Persson, S. & Boss, W. F. Inositol signalling and plant growth. Trends Plant Sci. 5, 252–258 (2000). 79. Karniol, B. & Chamovitz, D. A. The COP9 signalosome: from the light signalling to general developmental regulation and back. Curr. Opin. Plant Biol. 3, 387–393 (2000). 80. Blomberg, N., Baraldi, E., Nilges, M. & Saraste, M. The PH superfold: a structural scaffold for multiple functions. Trends Biochem. Sci. 24, 441–445 (1999). 81. Deak, M., Casamayor, A., Currie, R. A., Downes, P. & Alessi, D. R. Characterisation of a plant phosphoinositide-dependent protein kinase-1 homologue which contains a pleckstrin homology domain. FEBS Lett. 451, 220–226 (1999). 82. Roberts, M. R. Regulatory 14-3-3-protein/protein interactions in plant cells. Curr. Opin. Plant Biol. 3, 400–405 (2000). 83. Allan, A. C., Fricker, M. D., Ward, J. L., Beale, M. H. & Trewavas, A. J. Two transduction pathways mediate rapid effects of abscisic acid in Commelina guard cells. Plant Cell 6, 1319–1328 (1994). 84. Zuhua, H. et al. Perception of brassinosteroids by the extracellular domain of the receptor kinase BRI1. Science 288, 2360–2363 (2000). 85. Estelle, M. Cytokinin action: two receptors better than one? Curr. Biol. 8, R539–R541 (1998). 86. Trotochaud, A. E., Jeong, S. & Clark, S. E. CLAVATA3, a multimeric ligand for the CLAVATA1 receptor-kinase. Science 289, 613–617 (2000). 87. Alonso, J. M., Hirayama, T., Roman, G., Nourizadeh, S. & Ecker, J. R. EIN2, a bifunctional transducer of ethylene and stress responses in Arabidopsis. Science 284, 2148–2152 (1999). 88. Beaudoin, N., Serizet, C., Gosti, F. & Giraudat, J. Interactions between abscisic acid and ethylene signaling cascades. Plant Cell 12, 1103–1116 (2000). 89. Ghassemian, M. et al. Regulation of abscisic acid signalling by the ethylene response pathway in Arabidopsis. Plant Cell 12, 1117–1126 (2000).

Acknowledgements The authors would like to thank Sian Ritchie for critical reading of the manuscript at a late stage in its construction. We also gratefully acknowledge the support of the USDA, NSF and NASA (S.G.) and BBSRC and Human Frontier Science Programme Organisation (A.T.).

Take a look online Online, all Nature Reviews articles are enhanced with hyperlinks, including the following: • References are linked to PubMed abstracts. • Gene and protein names, inherited diseases and protein domains are linked to public-domain databases such as LocusLink, Flybase, SGD, OMIM and InterPro. • Links to other related online resources are provided. • Web watch articles are linked to the resources discussed in the text.

314

www.nature.com/reviews/molcellbio

| APRIL 2001 | VOLUME 2