Nature Reviews - Molecular Cell Biology - by Nature Magazine

toc_r

Page 1 of 2

CONTENTS

November 2000 Vol 1 No 2

81 | In this issue doi:10.1038/35040108

Highlights PDF

[1847K]

83 | CELL DIVISION Foreman in the histone factory doi:10.1038/35040000

84 | WEB WATCH A forum for all doi:10.1038/35040018

84 | CYTOSKELETON Subunits 1: Polymers 0 doi:10.1038/35040020

84 | DNA REPAIR Recognition is crystal clear doi:10.1038/35040003

85 | RNA PROCESSING Control freak doi:10.1038/35040110

85 | IN BRIEF TRANSCRIPTION | CELL POLARITY | CELL SIGNALLING doi:10.1038/35040023

86 | WEB WATCH From signal to sequence doi:10.1038/35040025

86 | CELL MOTILITY The sting in WASP's tail doi:10.1038/35040027

87 | CHROMOSOME BIOLOGY Sisters find a slick way to stick doi:10.1038/35040030

87 | TRANSCRIPTION A molecular Swiss-army knife doi:10.1038/35040032

88 | DEVELOPMENT AND CANCER Tampering with the cell cycle's brakes doi:10.1038/35040034

91 | CADHERINS IN EMBRYONIC AND NEURAL MORPHOGENESIS Ulrich Tepass, Kevin Truong, Dorothea Godt, Mitsuhiko Ikura & Mark Peifer doi:10.1038/35040042 [1620K]

101 | THE EXPANDING POLYMERASE UNIVERSE Myron F. Goodman & Brigette Tippin doi:10.1038/35040051 [714K]

110 | SECRETS OF ACTIN-BASED MOTILITY REVEALED BY A BACTERIAL PATHOGEN Lisa A. Cameron, Paula A. Giardini, Frederick S. Soo & Julie A. Theriot doi:10.1038/35040061 [678K]

120 | APOPTOSIS IN NEURODEGENERATIVE DISORDERS Mark P. Mattson doi:10.1038/35040009 [1647K]

130 | GRABBING THE CAT BY THE TAIL: MANIPULATING MOLECULES ONE BY ONE Carlos Bustamante, Jed C. Macosko & Gijs J. L. Wuite doi:10.1038/35040072 [1148K]

137 | NUCLEAR COMPARTMENTALIZATION AND GENE ACTIVITY Claire Francastel, Dirk Sch端beler, David I. K. Martin & Mark Groudine doi:10.1038/35040083 [851K]

145 | TIMELINE THE METEORIC RISE OF REGULATED INTRACELLULAR

http://www.nature.com.ezproxy.library.wisc.edu/nrm/journal/v1/n2/toc_r.html

12/3/2006

toc_r

Page 2 of 2

88 | APOPTOSIS Viral pirates hijack Bcl-2 doi:10.1038/35040037

doi:10.1038/35040090 [523K]

89 | MEMBRANE TRAFFIC FYVE fingers grab endosomes

149 | TIMELINE BIOLOGICAL MACHINES: FROM MILLS TO MOLECULES Marco Piccolino

doi:10.1038/35040006

89 | IN BRIEF APOPTOSIS | TRANSLOCATION | STEM CELLS | TECHNOLOGY doi:10.1038/35040040

PROTEOLYSIS R. John Mayer

doi:10.1038/35040097 [815K]

153 | OPINION SLOW AXONAL TRANSPORT: STOP AND GO TRAFFIC IN THE AXON Anthony Brown doi:10.1038/35040102 [1117K]

157 | NatureView doi:10.1038/35040112

NATURE REVIEWS | MOLECULAR CELL BIOLOGY

http://www.nature.com.ezproxy.library.wisc.edu/nrm/journal/v1/n2/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

CELL DIVISION

Foreman in the histone factory The cell, like any factory, must often step up supply to meet demand. For example, during S phase of the cell cycle, the supply of histones has to be increased to decorate the newly synthesized DNA. Two papers in Genes and Development explain how the kinase CDK2 and its regulatory partner cyclin E coordinate the synthesis of histones and DNA through a CDK2 substrate called NPAT. Humans have two clusters of histone genes, on chomosomes 1 and 6, but the transcription factors driving histone expression vary. Zhao and colleagues reasoned that there must be a ‘master regulator’ of histone expression and set out to find it. Having previously identified NPAT in a screen for cyclin E–CDK2 substrates, they used immunofluorescence to study its cellular localization. This revealed two tiny dots of NPAT in non-S-phase cells, but four in S phase. This localization overlapped with that of coilin, a component of a nuclear organelle called the Cajal body or coiled body (see picture) — a finding corroborated by Ma and colleagues. Cajal bodies often associate with histone gene clusters, so this finding provided an intriguing link between cyclin E–CDK2 and histone genes. Furthermore, fluorescence in situ hybridization showed that NPAT’s association with the histone gene cluster on chromosome 1 was cell cycle dependent, explaining why the number of NPAT dots increases during S phase. Next, Zhao et al. found a large

increase in gene expression driven by the histone H4 promoter when the NPAT gene was cotransfected into the cells. NPAT also enhanced expression from the H2B and H3 promoters. For the H4 promoter, the authors narrowed down the NPAT-responsive region to a sequence that binds a putative transcription factor called H4TF-2. Mutations in this sequence that abolish H4TF-2 binding blocked the effect of NPAT, whereas cotransfection of cyclin E- and CDK2expressing plasmids enhanced NPATmediated transcriptional activation of histone genes. Ma and colleagues determined the CDK2 phosphorylation sites on NPAT, then used phospho-NPATspecific antibodies to show that phospho-NPAT colocalizes with both cyclin E and coilin in Cajal bodies, and that the combination of phospho-NPAT and cyclin E–CDK2 is present in Cajal bodies only during S phase. Furthermore, mutation of NPAT’s CDK2 phosphorylation sites to alanine reduced NPAT’s ability to activate transcription from the histone 2B promoter. So cyclin E–CDK2 gives the orders, and NPAT ensures that they’re carried out. Appreciating NPAT’s skills will be our next lesson in this tour of the histone factory: how does NPAT manage its staff — presumably the histone-gene-specific transcription factors — and does it have other teams with responsibilities beyond histone production? Cath Brooksbank

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

One of Cajal’s drawings of his ‘accessory bodies’, from a paper published in 1910. This organelle was then forgotten for 60 years until it was rediscovered and named the coiled body by W. Bernard. Image courtesy of Joseph Gall, Carnegie Institution, Baltimore, Maryland, USA.

References and links ORIGINAL RESEARCH PAPERS Zhao, J. et al.

NPAT links cyclin E–Cdk2 to the regulation of replication-dependent histone gene transcription. Genes Dev. 14, 2283–2297 (2000) | Ma, T. et al. Cell cycle-regulated phosphorylation of p220NPAT by cyclinE/Cdk2 in Cajal bodies promotes histone gene transcription. Genes Dev. 14, 2298–2313 (2000) FURTHER READING Ewan, M. E. Where the cell cycle and histones meet. Genes Dev. 14, 2265–2270 (2000)

VOLUME 1 | NOVEMBER 2000 | 8 3

HIGHLIGHTS

WEB WATCH A forum for all Sorting the wheat from the chaff when it comes to web sites is always tricky, especially in fields of public interest. But for anybody seeking to learn more about Alzheimer’s disease, the Alzheimer Research Forum is a good place to start. Although the site has been around for some years (indeed, the archive goes back to February 1996), it is regularly updated and contains an impressive variety of sections and links. First, though, the visitor must identify themself — layperson, physician or researcher? After logging in the layperson is directed to an ‘Alzheimer general information directory’ containing basic information about the latest research, and links to Alzheimer associations and support groups around the world. As a physician or researcher, however, you are directed to a different home page. Here you’ll find the expected lists of relevant papers, some of which have links through to PubMed, and ‘Abstracts in Advance’ from The Journal of Alzheimer’s Disease. There’s also a directory listing “genes that have been studied in relation to their role in Alzheimer’s disease” — again with useful links to public databases. Other features that catch the eye are the ‘Forum Interviews’ with wellknown researchers such as Dennis Selkoe and Bruce Yankner, the various mutations directories (APP, presenilins and tau), and the ‘Virtual Conferences’, where you can listen to recordings of the speakers. There are a few glitches in this otherwise excellent site. For example, many of the newest ‘Papers of the Week’ do not yet contain PubMed links, and the ‘Milestone Papers’ section needs updating (for instance, there is no mention of the recent γ-secretase studies). But overall this is an easily navigable, useful site.

Alison Mitchell

C Y TO S K E L E TO N

Subunits 1: Polymers 0 The field of slow axonal transport is divided between those who believe that cytoskeletal subunits are transported along the axon (the subunit model) and those who believe that entire filaments move (the polymer model). What both camps have in common is ignorance of the precise transport mechanism, whatever the cargo might be. Reporting in Cell, Hirokawa and colleagues now provide evidence that slow axonal movement requires kinesin motors and microtubule tracks. The visualization of slow axonal transport is not trivial (see Anthony Brown’s article on page 153 of this issue), and finding an appropriate model system is half the job. Hirokawa and colleagues used the squid giant axon in their studies because it has two advantages: it is translucent and it is big. They injected fluorescent tubulin into the axon and measured the speed at which it

moved away from the cell body. The values obtained with this experimental set-up were similar to those reported for mammalian axons. In a series of pharmacological studies, the speed of movement was considerably reduced when microtubules were depolymerized, but remained constant in the absence of polymerized actin microfilaments. Similarly, transport was slowed down when kinesin’s motor activity was inhibited, but myosin seemed to be dispensable for this process. Hirokawa and colleagues also observed that the diffusion rate of the transported tubulin is lower than the diffusion rate of creatine kinase (another cargo for slow axonal transport), but higher than the diffusion rate of taxol-stabilized microtubules. The authors interpret this finding as an indication that tubulin is transported in a complex that is large but different from a fully polymerized

D N A R E PA I R

The axonal cytoskeleton. Courtesy of N. Hirokawa, University of Tokyo, Japan.

microtubule. This would tilt the balance in favour of the subunit model again, at least for the transport of tubulin. Figuring out the polymerization state of tubulin transported in this complex will hopefully clarify this issue. Raluca Gagescu

C-l

Recognition is crystal clear

C-l'

N-l'

N-l

DNA polymerases are not infallible — they make mistakes while replicating DNA. Sometimes these errors are deliberate (see, for example, the review by Goodman and Tippin on page 101 of this issue). But often they are not, and that’s when mismatch repair kicks in to protect against mutation. This system identifies DNA bases that have been incorrectly paired up, and allows the correct base to be reinserted. But how does it recognize the mismatches? Two groups, reporting in the 12 October issue of Nature, have studied the bacterial mismatch-repair protein MutS to address this question. In the first paper, Obmolova et al. present the crystal structures of a Thermus aquaticus MutS homodimer, both alone and as a complex with DNA containing a single unpaired thymidine. The

authors show Domain I (B) that the DNA adopts an unusual kinked shape owing to interactions with two domains from each MutS monomer (domains I and IV). Domain IV (A) However, these N-IV interactions are asymmetric, N-IV' Domain IV (B) with domain I from monomer A in the diagram donating the structure of Escherichia coli MutS phenylalanine residue (yellow ring) binding to a G•T mismatch. that interacts specifically with the Mismatch binding is known to unpaired base. induce the uptake of ATP, and both This message — that the MutS groups show that the ATPase homodimer is actually a domains also differ between the heterodimer at the structural level two MutS monomers. This — also emerges from the second asymmetry between the monomers paper by Lamers and colleagues. in DNA and ATP binding could These authors report the crystal explain the specificity of MutS for

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

HIGHLIGHTS

RNA PROCE SSI NG

Control freak

References and links ORIGINAL RESEARCH PAPER Terada, S.,

Kinjo, M. & Hirokawa, N. Oligomeric tubulin in large transporting complex is transported via kinesin in squid giant axons. Cell 103, 141–155 (2000) FURTHER READING Hirokawa, N. The mechanisms of fast and slow transport in neurons: identification and characterization of the new kinesin superfamily motors. Curr. Opin. Neurobiol. 7, 605–614 (1997)

mispaired DNA bases; presumably the observed conformational changes occur only in response to binding a mismatch. Why are these papers so significant? The human cousins of bacterial MutS — MSH2 and MSH6 — are mutated in hereditary nonpolyposis colorectal cancer (HNPCC). Not only can these mutations be mapped to the new MutS structures but, according to Obmolova et al.,“the crystal structures … provide a molecular framework for understanding HNPCC mutations”. Alison Mitchell References and links ORIGINAL RESEARCH PAPERS Obmolova,

G. et al. Crystal structures of mismatch repair protein MutS and its complex with a substrate DNA. Nature 407, 703–710 (2000) | Lamers, M. H. et al. The crystal structure of DNA mismatch repair protein MutS binding to a G•T mismatch. Nature 407, 711–717 (2000) FURTHER READING Kolodner, R. D. Guarding against mutation. Nature 407, 687–689 (2000)

The yeast Saccharomyces cerevisiae likes to be in control, especially when it comes to gene expression. We already knew that transcription, splicing, messenger RNA export, mRNA stability, mRNA translation and post-translational modification are regulated processes. Tollervey and colleagues now report in Cell that degradation of unspliced premRNAs occurs in the nucleus in a regulated manner, providing yet another regulatory mechanism for gene expression. Using several mutants defective in mRNA processing, Tollervey and coworkers found that unspliced premRNAs are degraded in the nucleus in a 3′ to 5′ direction by a large protein complex containing several exoribonucleases (the exosome), or in a 5′ to 3′ direction by the exonuclease Rat1p. This is, in fact, similar to what is known to happen in the cytosol, where mRNAs are degraded either 3′ to 5′ by the exosome or 5′ to 3′ by the cytosolic exonuclease Xrn1p. But contrary to what happens in the cytosol, exosome-mediated degradation is predominant in the nucleus. Nuclear degradation of premRNA seems to compete with splicing. It is increased in the presence of glucose — yeast’s favourite food — indicating that it is a physiological regulatory pathway for gene expression. The 3′ to 5′ Rat1p-dependent pathway is probably inhibited by the cap structure of the pre-mRNA, providing another level of control. Degradation of inaccurately spliced pre-mRNAs has also been observed in mammalian cells, and there again, the activity seems to be nuclear. An obvious experiment will be to test whether homologues of the genes identified in yeast have the same function in mammalian cells. Raluca Gagescu References and links ORIGINAL RESEARCH PAPER

Bousquet–Antonelli C., Presutti, C. & Tollervey, D. Identification of a regulated pathway for nuclear pre-mRNA turnover. Cell 102, 765–775 (2000) FURTHER READING Mitchell, P. & Tollervey, D. Musing on the structural organization of the exosome complex. Nature Struct. Biol. 7, 843–846 (2000)

IN BRIEF TRANSCRIPTION

Dynamic association of capping enzymes with transcribing RNA polymerase II. Schroeder, C. et al. Genes Dev. 14, 2435–2440 (2000)

Different phosphorylated forms of RNA polymerase II and associated mRNA processing factors during transcription. Komarnitsky, P., Cho, E.-J. & Buratowski, S. Genes Dev. 14, 2452–2460 (2000)

RNA polymerase II is a molecular platform to which many messenger RNA-processing factors bind during transcription. Are such factors associated simultaneously with RNA pol II, or do they interact in a transient and sequential manner? Using mRNA-capping enzymes, both papers indicate that RNAprocessing enzymes associate dynamically with differently modified forms of the polymerase at different stages of the transcriptional cycle. CELL POLARITY

Plasma membrane compartmentalization in yeast by messenger RNA transport and a septin diffusion barrier. Takizawa, P. A. et al. Science 290, 341–344 (2000)

Saccharomyces cerevisiae restricts the cellular distribution of the transcription factor Ash1p by transporting its messenger RNA into the forming bud. Takizawa et al. now present a list of other mRNAs that also localize to the bud, and the transport mechanism for at least one of them — which encodes a transmembrane protein — is the same as for ASH1 mRNA. The protein is then retained in the bud by a diffusion barrier involving septins that is functionally similar to tight junctions in epithelial cells.

CELL SIGNALLING

RasGRP is essential for mouse thymocyte differentiation and TCR signalling. Dower, N. A. et al. Nature Immunol. 1, 317–321 (2000)

Control of pre-T cell proliferation and differentiation by the GTPase Rac-1. Gomez, M. et al. Nature Immunol. 1, 317–321 (2000)

These papers describe the actions of two small GTPases, Ras and Rac, in T-cell responses. The first sheds light on how RasGRP, a Ras activator that’s directly sensitive to diacylglycerol, mediates responses that had previously been assumed to be a result of the archetypal diacylglycerol-sensitive enzyme — protein kinase C. The second uses Rac mutants that can control actin dynamics, but not other downstream targets of Rac such as mitogenactivated protein kinases, to show that changes in actin dynamics are sufficient to drive some stages of T-cell differentiation.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 8 5

HIGHLIGHTS

WEB WATCH From signal to sequence No scientist is an island and, as the Human Genome Project testifies, we can achieve remarkable things when we work as a team. Biocarta has adopted a community approach to mapping cellular pathways: it provides the tools; the scientific community provides — and constantly updates — the data. The Biocarta website contains several types of information, but its pathway diagrams, which cover all areas of cellular regulation from cell division to apoptosis, immunology to neuroscience, are the main attraction. Clicking on a pathway category gives you a list of pathways: for example, clicking on cell-cycle regulation links to a menu containing the ATM pathway, regulation by cyclins, the p53 pathway and more. Clicking on any single pathway provides a stylized, information-packed diagram that evolves as users provide new information. In the future, pathways will also have ‘gurus’ responsible for assessing submitted information on their pathway. If you don’t agree with what you see, you can send comments to a discussion group and help the pathway to evolve. Each component of the pathway is a gateway to a wealth of information: clicking on any protein takes you to a table containing links to just about any public-domain database you can think of. Whether you want sequences, structures, information on genetic diseases, or just some relevant abstracts, you can link to them from here. If your favourite pathway isn’t in the list, you can submit it. Biocarta even provides a template with which to draw your pathway diagram. If the template doesn’t have the right components, you can send diagrams in any format, “even a paper napkin”. So don’t sit back and watch this site evolve: join in!

C E L L M OT I L I T Y

The sting in WASP’s tail A C How are the many stimuli that tell cells to move C Inactive WASP translated into changes in actin (Autoinhibited) polymerization? Evidence points to WH1 Polyproline N members of the Wiskott–Aldrich CDC42-binding Acidic syndrome protein (WASP) family as phospholipid the interpreters but, for scientists, the language of cell movement has proved difficult to learn. Two papers in The Journal of Cell Biology provide some clues as to how a lipid and a protein Prenyl group GTP-bound GDP-bound collaborate to activate two WASP-family CDC42 CDC42 members. The details seem protein specific Pi but the general message is the same — activation of WASPs Active A C WH1 Polyproline C N CDC42-binding WASP involves stopping them from biting their own tails. Actin nucleation is stimulated by the Arp2/3 complex, which is activated by WASPs. A Actin polymerization small GTPase, Cdc42, and a phospholipid, phosphatidylinositol-4,5-bisphosphate (PtdIns(4,5)P2), activate WASPs, but how they do so is controversial: recombinant WASPs often have some constitutive activity, and inactive, GDP-bound Cdc42 sometimes seems capable of activating them. Henry Higgs and Tom Pollard sought to end this controversy — at least for the haematopoieticHow Cdc42 and acidic phospholipids might activate WASP. cell-specific member of the family, WASP — by Inset shows an actin-filament halo (red) surrounding a vesicle (green). Photo courtesy of Henry Higgs and Tom Pollard, Salk purifying native WASP from bovine thymus. They find Institute, La Jolla, California, USA. that purified WASP alone has no effect on actin polymerization rates, but that micelles containing activity , including the C motif and an acidic region (C PtdIns(4,5)P2 activate polymerization through WASP. In and A in the figure), of the carboxyl terminus in trans. the presence of Cdc42, PtdIns(4,5)P2 micelles, or vesicles Full-length N-WASP, however, can’t inhibit N-WASP’s containing either PtdIns(4,5)P2 or another acidic carboxyl terminus, presumably because the full-length phospholipid, phosphatidylserine, produce halos of protein is folded into its autoinhibited conformation. polymerized actin surrounding the phospholipid (see Higgs and Pollard find that inhibition of WASP’s picture). This effect requires PtdIns(4,5)P2 or carboxyl terminus is relieved by GTP–Cdc42 but not by phosphatidylserine, and Cdc42 must be both GTPPtdIns(4,5)P2, whereas Rohatgi and co-workers find that bound and prenylated, indicating that it needs to be their intermolecular inhibitory complex is regulated in membrane associated to do its job. exactly the same way as wild-type N-WASP. Rohatgi and colleagues, working with a recombinant We’re still straining to understand what WASPs, Cdc42 form of the widely expressed N-WASP, have a different and PtdIns(4,5)P2 are saying to each other. Perhaps story: they find that Cdc42, but not PtdIns(4,5)P2, can WASP and N-WASP respond to their two activators partly activate N-WASP, and that the two molecules slightly differently, or maybe the discrepancies are due to synergize to activate N-WASP fully. They identify a basic variations between purified and recombinant proteins. region, close to the Cdc42-binding domain, that seems to Is WASP’s PtdIns(4,5)P2-binding domain equivalent to bind PtdIns(4,5)P2. In actin nucleation assays, a mutant N-WASP’s? And can N-WASP be activated by N-WASP lacking this domain remains sensitive to Cdc42 phosphatidylserine? Further studies should clarify how but is insensitive to the additive effects of Cdc42 and Cdc42 and acidic phospholipids unleash WASP’s sting. PtdIns(4,5)P2. These researchers previously showed that Cath Brooksbank PtdIns(4,5)P2 stimulates actin polymerization in Xenopus egg extracts. They now show that this effect References and links depends on N-WASP but, curiously, the deletion mutant ORIGINAL RESEARCH PAPERS Higgs, H. N. & Pollard, T. D. Activation by can also translate a PtdIns(4,5)P2 signal into limited Cdc42 and PIP2 of Wiskott–Aldrich syndrome protein (WASP) stimulates actin nucleation by Arp2/3 complex. J. Cell Biol. 150, 1311–1320 (2000) | Rohatgi, actin polymerization, albeit more slowly. R. et al. Mechanism of N-WASP activation by CDC42 and Phosphatidylinositol WASP’s carboxyl terminus is constitutively active, 4,5-bisphosphate. J. Cell Biol. 150, 1299–1309 (2000) suggesting an autoinhibitory mechanism. Both groups FURTHER READING Cameron, L. A. et al. Secrets of actin-based motility show that a separate Cdc42-binding domain can curb the revealed by a bacterial pathogen. Nature Rev. Mol. Cell Biol. 1, 110–119 (2000)

Cath Brooksbank

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

C H R O M O S O M E B I O LO G Y

Sisters find a slick way to stick In some ways, meiotic division is like country dancing. During meiosis I, homologous chromosome pairs line up opposite one another and then, at anaphase I, whirl apart to opposite poles of the cell. But in the next reel, meiosis II, the sister chromatids themselves separate. So what is the molecular ‘caller’ that tells the partners when to stick together and when to separate? Dean Dawson and colleagues, reporting in October’s Current Biology, have identified a key player in the process. Clues to what’s going on in meiosis can be gleaned from events during mitosis. Sister chromatids are initially bound along their length by a physical tie — the so-called cohesin complex. Consisting of four subunits, Smc1p, Smc3p, Scc1p and Scc3p, cohesin establishes this connection and maintains it until anaphase. At this point, sister-chromatid cohesion is lost — first along the arms and then at the centromeres — by proteolytic cleavage of the Scc1p subunit. This cohesion must be maintained during meiosis I and, in addition, the kinetochores (multiprotein complexes at the centromeres) must be regulated such that both sisters are pulled to one pole rather than being dragged away from each other. Dawson and colleagues realized that yeast with a mutation in the SLK19 (synthetic lethal kar3) gene show a defect in meiosis consistent with a failure in the control of sister-chromatid behaviour: rather than forming tetrads containing four haploid spores, these mutants form dyads with two diploid spores. The authors found that although slk19 mutants enter meiosis I efficiently, the sister chromatids then separate to opposite poles of the cell. Moreover, most of the mutants do not go through a second meiotic division — that’s why they form dyads. This could be due to a breakdown of sister-chromatid cohesion, to impaired kinetochore function, or to both. Dawson and co-workers next used indirect immunofluorescence to monitor the localization of Slk19p during meiosis. The picture shows synaptonemal complexes (red; these run along paired meiotic chromosomes) and Slk19p–GFP (green). The authors found that Slk19p localizes to centromeric regions during prophase I, and that it remains there until anaphase I. After this point, however, Slk19p loses its association with the centromeres, and localizes instead along the spindles, as it is left behind when the centromeres migrate polewards at meiosis I. These localization patterns indicate that

Slk19p might keep the sister chromatids together during the first meiotic division. But how? The authors speculated that it could affect Rec8p, which is the yeast meiosis-specific homologue of Scc1p. Rec8p must be localized at the centromeres to ensure sister-chromatid cohesion through meiosis I, so perhaps Slk19p prevents the degradation of Rec8p? To test this, Dawson and colleagues looked at the effect of Slk19p on localization of Rec8p. They found that, during the early stages of meiosis I, the localization of Rec8p is indistinguishable between slk19 and wild-type strains. In anaphase I, however, Rec8p is maintained at the centromeres in 80% of the wild-type cells, but in most of the slk19 cells there is almost no Rec8p staining. How can a protein found in mitotic cells specifically regulate the behaviour of sister chromatids during meiosis? The answer may be that Slk19p acts downstream of another meiosis-specific factor, and the authors’ top candidate is a putative meiotic regulatory protein, Spo13p. Both spo13 and slk19 mutants have defects in the behaviour of sister chromatids during meiosis. But the defect in spo13slk19 double mutants is no more serious than that in the spo13 mutant alone, indicating that the two proteins may be involved in a common pathway. The next step in studying the molecular dance of meiosis is to work out how Slk19p and Spo13p might prevent the partners from separating too soon. Alison Mitchell References and links ORIGINAL RESEARCH PAPER Kamieniecki, R. J., Shanks, R.

M. Q. & Dawson, D. S. Slk19p is necessary to prevent separation of sister chromatids. Curr. Biol. 10, 1182–1190 (2000) FURTHER READING Nasmyth, K., Peters, J.-M. & Uhlmann, F. Splitting the chromosome: cutting the ties that bind sister chromatids. Science 288, 1379–1384 (2000)

TRANSCRIPTION

A molecular Swiss-army knife Like a bottle of wine, heterochromatin is no use unless you can open it. TAF250 is one protein that helps to uncork the DNA by loosening histone H1’s grip on it. A paper by Pham and Sauer in Science suggests that TAF250 is more a Swiss-army knife than a corkscrew, though: it was already known to phosphorylate and acetylate histone H1, but now it adds ubiquitylation to its list of talents list. Polyubiquitylation — adding a chain of ubiquitin molecules to proteins — is a signal for destruction, but monoubiquitylation is less well understood. Knowing that histone H1 can be monoubiquitylated, Pham and Sauer set out to discover what was doing it, and found TAF250. Ubiquitylation requires a minimum of two enzyme activities, an activator and a conjugator, which are usually on separate proteins. In vitro assays revealed that TAF250 has both. But what about in vivo? Having identified the region of TAF250 that’s necessary for histone monoubiquitylation, they created mutants that lack the activity and are heterozygous for dorsal — a maternally expressed gene important for early Drosophila development. In these mutants, expression of the dorsalresponse genes twist and snail was significantly reduced, resulting in a twisted phenotype. Levels of monoubiquitylated histone were also reduced. More substrates of TAF250’s ubiquitylating activity, as well as factors that ubiquitylate the other histones, may await discovery. Cath Brooksbank References and links ORIGINAL RESEARCH PAPER Pham, A.-D. & Sauer, F.

Ubiquitin-activating/conjugating activity of TAF250, a mediator of activation of gene expression in Drosophila. Science 289, 2357–2360 (2000) FURTHER READING Mizzen, C. A. & Allis, C. D. New insights into an old modification. Science 289, 2290–2291 (2000)

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 8 7

HIGHLIGHTS

D E V E LO P M E N T A N D C A N C E R

Tampering with the cell cycle’s brakes To stop the cell cycle, tumour suppressors such as the retinoblastoma protein (Rb) apply the brakes. A Nature paper by Anna Lasorella and colleagues describes how Id2 — a dominant-negative inhibitor of helix–loop–helix DNA-binding proteins — gets the wheels turning again. Retinoblastoma protein is essential for mammalian development: knockouts die during embryogenesis. Lasorella et al. report that knocking out Id2 rescues Rb–/– embryos. Defective myogenesis kills Id2–/–Rb–/– mice shortly after birth, but they show none of the hallmarks of Rb–/– mice — too much proliferation and apoptosis in the haematopoietic and nervous systems. Id2 therefore perpetrates the Rb–/–phenotype; but how? Immunoprecipitates revealed that active, hypophosphorylated Rb binds Id2 and, to tip the balance against tumour suppression, Id2 has to be in molar excess of Rb. An intact Rb pathway is needed to prevent tumorigenesis, so Lasorella and colleagues

asked whether Id2 is overexpressed in neuroblastoma cell lines, in which N-myc amplification typically bypasses Rb. Remarkably, Id2 expression correlated with N-myc amplification. What’s more, Myc’s effect on Id2 expression (which also extends to c-Myc) is direct, owing to two high-affinity Myc-binding sites in the Id2 promoter. Deletion of these sites abolished Id2 expression in response to Myc. By producing a surfeit of Id, then, Myc can override Rb’s attempts to halt the cell cycle. But what targets of Id are responsible for the Rb–/– phenotype? And what about the other Id family members, Id1 and Id3? These questions, and others, await the next cycle tour. Cath Brooksbank References and links ORIGINAL RESEARCH PAPER Lasorella, A. et al. Id2 is a retinoblastoma protein target and mediates signalling by Myc oncoproteins. Nature 407, 592–598 (2000) FURTHER READING Yokota, Y. et al. Development of peripheral lymphoid organs and natural killer cells depends on the helix-loop-helix inhibitor Id2.Nature 397, 702--706 (1999)

A P O P TO S I S

Viral pirates hijack Bcl-2

Viruses are modern-day buccaneers. They ride the cellular seas, hijacking proteins and using them to promote their own survival. But how? Ojala et al., reporting in the November issue of Nature Cell Biology, describe how the Kaposi’s sarcoma herpesvirus (aka human herpesvirus 8; HHV8) interferes with apoptotic signalling pathways in its host.

The virus encodes a pirated cyclin (vcyclin), which forms a complex with a cellular cyclin-dependent kinase, CDK6. This complex can induce apoptosis, and Ojala et al. now show that it probably does so by phosphorylating — and inactivating — the cellular anti-apoptotic molecule Bcl-2. The authors find that Bcl-2 and CDK6 associate in cell extracts, and they show that the targets for phosphorylation are two serine residues in an unstructured ‘loop’ region of cellular Bcl-2. But herein lies a paradox. Why should the virus inactivate a molecule that protects the host cell from apoptotic death? One might think it in the best interests of the virus to keep its host alive. It turns out, however, that cellular Bcl-2 has other functions that the virus finds less than savoury; for instance, it has been reported to impair cell-cycle progression when overexpressed. Moreover, host cell death is advantageous to the virus in that it allows the spread of viral particles.

There is a problem, though — the risk that, when the virus blocks cellular Bcl-2, the host cell will die before the viral replication cycle is complete. Here, in an ultimate act of viral skulduggery, HHV8 produces its own Bcl-2 homologue. This virus-encoded protein lacks the crucial unstructured loop, so it cannot be phosphorylated or inactivated by v-cyclin–CDK6. Infection of host cells with HHV8 has previously been shown to be linked to apoptosis, and the lesions associated with Kaposi’s sarcoma contain some apoptotic cells. The next step, then, will be to work out how the acts of viral piracy uncovered by Ojala et al. link HHV8 to tumour formation and to the development of Kaposi’s sarcoma. Alison Mitchell References and links ORIGINAL RESEARCH PAPER Ojala, P. M. et al. The

apoptotic v-cyclin-CDK6 complex phosphorylates and inacticates Bcl-2. Nature Cell Biol. 2, 819–825 (2000) FURTHER READING Hardwick, J. M. Cyclin’ on the viral path to destruction. Nature Cell Biol. 2, E203–E204 (2000) | Desagher, S. & Martinou, J. C. Mitochondria as the central control point of apoptosis. Trends Cell Biol. 10, 369–377 (2000).

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

HIGHLIGHTS

IN BRIEF

MEMBRANE TRAFFIC

FYVE fingers grab endosomes Over the years, many investigators have set out to look for a protein regulator of their favourite protein and instead found a phosphoinositide. Most labs have therefore developed at least a peripheral interest in this family of lipids. So they will be happy to read in the EMBO Journal that the laboratories of Harald Stenmark and Rob Parton have devised a new probe to study the cellular localization of phosphatidylinositol-3-phosphate (PtdIns(3)P). The products of phosphatidylinositol-3-OH kinases function in processes as diverse as signal transduction, cytoskeletal organization and apoptosis. PtdIns(3)P is particularly interesting for membrane traffic aficionados, as it regulates transport along the endocytic pathway in all species where this has been studied. One of its activities is to recruit proteins that contain PtdIns(3)P-binding FYVE finger domains to membranes. But to which membranes? Gillooly et al. reasoned that if PtdIns(3)P binds FYVE domains, then FYVE domains should bind PtdIns(3)P. They built a probe (2XFYVE) consisting of two FYVE domains from Hrs, a protein that acts in the endocytic pathway. A series of control experiments showed that 2XFYVE binds selectively to PtdIns(3)P, both in vivo and in vitro. The probe effectively competes with endogenous proteins for PtdIns(3)P binding when transfected into cells, and can be used for immunofluorescence as well as for immunoelectron microscopy studies. The next step was to use 2XFYVE to localize PtdIns(3)P in the cell. The FYVE finger protein EEA1 is known to

localize exclusively to early endosomes, where it is involved in membrane fusion. So it was clear from the start that there must be substantial amounts of PtdIns(3)P in the membrane of early endosomes. This was confirmed in this study — 2XFYVE, shown in red in the picture, colocalized extensively with EEA1, shown in green. More surprisingly, immunoelectron microscopy using 2XFYVE revealed that PtdIns(3)P is also present on internal membranes of multivesicular late endosomes. This observation led the authors to speculate that the PtdIns(3)P-containing intralumenal vesicles arise from invagination of the endosomal membrane. This could sequester PtdIns(3)P away from the surface, stopping it from recruiting cytoplasmic proteins such as EEA1 to late endosomes. The origin of the convoluted morphology of multivesicular endosomes is still mysterious, and 2XFYVE might prove a useful tool to study this process. This probe is not the first on the market. The pleckstrin homology (PH) domain of phospholipase Cδ1 and the PH domains of ARNO and Bruton’s tyrosine kinase have been used to detect PtdIns(4,5)P2 and PtdIns(3,4,5)P3, respectively. But these probes cannot be used for electron microscopy, and neither of them shows the exquisite specificity for its target that 2XFYVE seems to have. Raluca Gagescu References and links ORIGINAL RESEARCH PAPER Gillooly, D. J. et

al. Localization of phosphatidylinositol 3phosphate in yeast and mammalian cells. EMBO J. 17, 4577–4588 (2000) FURTHER READING Corvera, S., D’arrigo, A. & Stenmark, H. Phosphoinositides in membrane

traffic. Curr. Opin. Cell Biol. 11, 460–465 (1999)

A P O P TO S I S

P53AIP1, a potent mediator of p53-dependent apoptosis, and its regulation by Ser-46-phosphorylated p53. Oda, K. et al. Cell 102, 849–862 (2000)

The star of this story — p53AIP1 — is a newly cloned gene, the protein product of which localizes to the mitochondria. It causes apoptotic cell death by dissipating the mitochondrial transmembrane potential, and it could help to mediate p53dependent apoptosis. After severe DNA damage, phosphorylation of a specific residue (serine 46) on p53 leads to apoptosis. The authors show that substitution of Ser 46 not only inhibits p53dependent apoptosis, but also blocks expression of p53AIP1. T R A N S LO C AT I O N

Two intermembrane space TIM complexes interact with different domains of Tim23p during its import into mitochondria. Davis, A. J. et al. J. Cell Biol. 150, 1271–1282 (2000)

How are mitochondrial proteins targeted for either insertion into the mitochondrial inner membrane or translocation into the matrix? Davis et al. show that the inner membrane protein Tim23p interacts with both known intermembrane space TIM complexes before reaching the Tim22p inner membrane translocon. But only its interaction with one of these complexes — Tim9p–Tim10p — is essential for correct targeting, leaving the mystery of what the Tim8p–Tim13p complex does intact. STEM CELLS

Effects of eight growth factors on the differentiation of cells derived from human embryonic stem cells. Schuldiner, M. et al. Proc. Natl Acad. Sci. USA 97, 11307–11312 (2000)

Until now there has been no systematic attempt to determine how growth factors affect the lineage choice of embryonic stem cells. This broad study correlates different growth-factor treatments with cell morphology and expression of markers for 11 tissues, derived from all three germ layers. It shows that what you put in biases, but doesn’t absolutely determine, what you get out, and has obvious implications for stem-cell therapy. T E C H N O LO G Y

Harnessing the ubiquitination machinery to target the degradation of specific cellular proteins. Zhou, P. et al. Mol. Cell 6, 751–756 (2000)

This paper uses a neat trick to functionally ‘knock out’ stable proteins — by targeting them for proteasomal degradation. By engineering specific protein–protein interaction domains into one component of the SCF complex, a multimeric ubiquitinconjugating machine, Zhou and colleagues can send proteins to their death in both yeast and mammalian cells, and can measure the phenotypic consequences. NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 8 9

REVIEWS CADHERINS IN EMBRYONIC AND NEURAL MORPHOGENESIS Ulrich Tepass*, Kevin Truong‡, Dorothea Godt*, Mitsuhiko Ikura‡ and Mark Peifer§ Cadherins not only maintain the structural integrity of cells and tissues but also control a wide array of cellular behaviours. They are instrumental for cell and tissue polarization, and they regulate cell movements such as cell sorting, cell migration and cell rearrangements. Cadherins may also contribute to neurite outgrowth and pathfinding, and to synaptic specificity and modulation in the central nervous system. GROWTH CONE

Exploratory tip of an extending neuronal process such as an axon. IMMUNOGLOBULIN-TYPE ADHESION MOLECULES

Family of adhesion molecules characterized by the presence of immunoglobulin-like domains, which are also found in antibody molecules.

Department of Zoology, University of Toronto, 25 Harbord Street,Toronto, Ontario M5S 3G5, Canada. ‡ Division of Molecular and Structural Biology, Ontario Cancer Institute and Department of Medical Biophysics, University of Toronto, Toronto M5G 2M9, Ontario, Canada. § Department of Biology, University of North Carolina, Chapel Hill, North Carolina 27599, USA. Correspondence to U.T. e-mail: utepass@zoo.utoronto.ca

In 1963, two publications appeared that focused attention on the adhesive mechanisms that contribute to embryonic morphogenesis and govern structural differentiation of the nervous system. Malcolm Steinberg laid the groundwork for the ‘differential adhesion hypothesis’, suggesting that the segregation or ‘sorting out’ of different embryonic cell types into separate tissues involves qualitative or quantitative differences in cell adhesion1. Roger Sperry proposed the ‘chemoaffinity hypothesis’, which holds that specific synaptic contacts form according to differences in the adhesive properties of individual GROWTH CONES and synapses2. Both hypotheses were the result of decades of experimental work, but were formulated before any adhesion molecules had been identified. It is now apparent that a large number of adhesion molecules exist that can be grouped into several superfamilies. The cadherin and IMMUNOGLOBULIN-TYPE ADHESION MOLECULES are the main groups of cell–cell adhesion receptors, whereas the integrins are the predominant contributors to cell–substrate adhesion3,4. Adhesive mechanisms that contribute to embryonic or neural morphogenesis share many similarities, revealing that Steinberg’s and Sperry’s hypotheses are essentially similar proposals applied to different cell populations. Morphogenesis involves two interrelated themes: structure and movement. For example, the different adhesive properties of two mixed cell populations induce cell movement, leading to the sorting out of the two groups of cells. After the sorting process is completed, adhesive differences maintain the segregation and relative position of the two cell groups, therefore pre-

serving a specific tissue architecture5–8. Neuronal morphogenesis follows a similar pattern, in which the neuronal growth cone has to move through a complex environment using differential adhesive cues. On reaching the target, synaptic contacts are formed and maintained by specific adhesive interactions9. The first cadherins to draw scientists’ attention were vertebrate classic cadherins, which were independently identified for their ability to mediate calcium-dependent adhesion among cultured cells and for their role in the epithelialization of the early mouse embryo10. So far, the sequences of over 300 vertebrate cadherins have been reported, and the virtually complete sets of cadherins encoded by the genomes of Caenorhabditis elegans and Drosophila melanogaster are now known. Structural diversity of the cadherin superfamily

The recent explosion in genomic sequencing of various animals has shed new light on the diversity of the cadherin superfamily. In humans, more than 80 members of the cadherin superfamily have been sequenced. Current annotation of the C. elegans and the Drosophila genomes reveals 14 and 16 cadherin genes, respectively. Cadherins are defined by the presence of the cadherin domain (CD), a roughly 110 amino-acid peptide that mediates calcium-dependent homophilic interactions between cadherin molecules (FIG. 1). The CD is typically organized in tandem repeats. Calcium ions associate with the linker region that connects two CDs, and require interaction with amino acids from both CDs (FIG. 1; see below). Here we present a classification of cadherins into

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 9 1

REVIEWS

CHORDATES

Phylum that comprises animals with a notochord and includes all vertebrates. DESMOSOME

A patch-like adhesive intercellular junction found in vertebrate tissues that is linked to intermediate filaments. METAZOANS

Refers to the kingdom Animalia (animals) that comprises roughly 35 phyla of multicellular organisms.

subfamilies on the basis of the domain layout of individual cadherins, which includes the number and sequence of CDs, and the presence of other conserved domains and sequence motifs (FIG. 2, TABLE 1). This analysis reveals that four cadherin subfamilies are conserved between C. elegans, Drosophila and humans: classic cadherins, Fat-like cadherins, seven-transmembrane cadherins and a new subfamily of cadherins that is related to Drosophila Cad102F. Classic cadherins break up in four subgroups, as listed in TABLE 1. Fat-like cadherins contain a subgroup of highly related molecules that we call Fatoid cadherins. These include all known vertebrate Fat-like cadherins, Drosophila Cad76E and C. elegans Cdh-4. Cadherins containing protein kinase domains are found in vertebrates (RET cadherins) and in Drosophila. Desmosomal cadherins are presumably derived from type I classic cadherins within the CHORDATE lineage, as neither desmosomal cadherins nor DESMOSOMES are found in Drosophila or C. elegans. Finally, protocadherins also seem to be limited to chordates. The grouping of cadherins into seven subfamilies, which is largely on the basis of the overall protein domain architecture, is corroborated by sequence comparison of CDs only (see online supplementary materials). Note that only about half of the cadherins found in C. elegans and Drosophila are part of identified subfamilies. Cadherins are not found in yeast, and only a single, poorly conserved CD has been reported in a secreted protein from Dictyostelium11, indicating that transmembrane proteins of the cadherin superfamily might have evolved to meet the need for the complex cell interacβA

R N

ECHINODERMS

ARTHROPODS

Largest animal phylum composed of invertebrates that have a segmented body, segmented appendages and an external skeleton. This includes insects, spiders and crustaceans. NEMATODES

Animal phylum of unsegmented roundworms. GENE RADIATION

Process that leads to the formation of gene families in which gene amplification through gene duplication events is followed by the diversification of gene structure and function.

G58 W G58 R E βC

βG

Animal phylum of marine invertebrates including sea urchins and starfish.

βB N

K25 I Q V L N K P18 αA

E31 K D

Cell–cell or cell–matrix adhesive junctions that are linked to microfilaments.

αA

I4 P5

ADHERENS JUNCTIONS

tions that are required for the multicellular organization of METAZOANS. The function of classic cadherins during the formation and maintenance of epithelial tissues and cell–cell ADHERENS JUNCTIONS — two other metazoan inventions — is of particular importance. Within the classic cadherin subfamily, an interesting shift in protein organization has taken place during evolution — chordate classic cadherins lack non-chordate classic cadherin domains (NCCDs), laminin G (LG) and epidermal growth factor (EGF) domains and consistently contain five CDs, in contrast to the domain structure of classic cadherins in three other phyla: ECHINODERMS, ARTHROPODS and NEMATODES (FIG. 2, TABLE 1). This finding indicates that, during the early evolution of chordates, the structure of classic cadherins was modified and that a single progenitor might have given rise to the numerous classic cadherins found in vertebrates today, a conclusion supported by the phylogenetic analysis of chordate cadherins12. The cadherin families of C. elegans and Drosophila are small and very similar in size, and these cadherin genes are scattered throughout the genome without any obvious clustering, indicating that gene duplication events might not have occurred recently. So most of these genes were presumably established early during metazoan evolution. In particular, the progenitors of the four conserved subfamilies of cadherins are the result of a GENE RADIATION event that occurred before nematodes, arthropods and chordates diverged. In contrast, classic cadherins and the closely related desmosomal cadherins, as well as the more divergent protocadherins, were all amplified within the chordate lineage, resulting

L I

V T63

F V G49

βE βD

V34 F Y S I T G Q41

D67 αB

R βF

D103 CalciumN E11 binding Q pocket D100

A70 E

S82 V A H S Y L I Y K A72 E86 A V E D P M E I V I T V98

Putative homophilic binding surface

βA-A'

βA' P6 I

S C9

αB αB

βB

αA

βC

βD

βE

αB

βF

βG

C Linker

E_CAD N_CAD

1 11 21 31 41 51 61 71 81 91 101 DWVIPPISC––PENEK–GEFPKNLVQIKSNRDKET–––KVFYSITGQGADKPPVGVFIIERETGWLKVTQ–PLDREAIAKYILYSHAVSSNGEAVEDPMEIVITVTDQNDNRPEF DWVIPPINL––PENSR–GPFPQELVRIRSDRDKNL–––SLRYSVTGPGADQPPTGIFIINPISGQLSVTK–PLDRELIARFHLRAHAVDINGNQVENPIDIVINVIDMNDNRPEF

E_CAD N_CAD

111 121 131 141 151 161 171 181 191 201 211 ––TQEVFEGSVAEGAVPGTSVMKVSATDADDDVNTYNAAIAYTIVSQDPELPHKNMFTVNRDTGVISVLTSGLDRESYPTYTLVVQAADLQG–––EGLSTTAKAVITVKDINDNAPVF ––LHQVWNGSVPEGSKPGTYVMTVTAIDADDPN–ALNGMLRYRILSQAPSTPSPNMFTINNETGDIITVAAGLDREKVQQYTLIIQATDMEGNPTYGLSNTATAVITVTDVNDNPPEF

Figure 1 | Structure of the cadherin domain. The structure of the first cadherin domain (CD) of mouse E-cadherin is shown in a | and b |. It was solved by NMR spectroscopy18 and, subsequently, the crystal structure of the first CD of N-cadherin revealed a similar topology19. The CD consists of a seven-strand β-sheet with the amino and carboxyl termini located at opposite ends of the molecule. The segment connecting strands B and C adopts an apparently helical structure consisting of a succession of β-turn and β-like hydrogen bonds. This unique quasi-β-helix structure is characteristic of the CD. a | Schematic of topology of the aminoterminal CD of mouse E-cadherin. βA, βA ′, βB, βE, and βD (green) and βC, βF, and βG (yellow) form β-sheets. The α-helices are shown in magenta. The putative homophilic binding surface including the amino acids HAV (red), and the Ca2+-binding pocket with the amino acids that interact with Ca2+ (blue), are indicated by the dotted lines. b | Ribbon structure of the CD. The colour coding is the same as in (a). c | Alignment of the first two N-terminal CDs of mouse E- and N-cadherin. The colour coding is the same as in (a). The CDs are connected by a roughly 10 amino-acid linker region. Note that the amino acids that form a single Ca2+ binding pocket (indicated in blue) are found in the first and second CDs and include animo acids in the linker region.

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS in numerous genes with pronounced clustering13,14. In humans, all six desmosomal cadherin genes are found in chromosomal region 18q12.1; three genes that encode more than 50 protocadherins are found in

region 5q31–32, and many classic cadherins are organized in gene clusters13–16. A similar gene amplification is seen in other unrelated gene clusters, such as in the C. elegans collagen genes. It has been proposed that the

Table 1 | Cadherin subfamilies Cadherin subfamily

Type

Cadherin

Species

No. of CDs

Examples include: E-cadherin (CDH1) P-cadherin (CDH3) N-cadherin (CDH2) R-cadherin (CDH4)

Hs Hs Hs Hs

5 5 5 5

Vertebrate Type II classic cadherins • No HAV motif in first CD

Examples include: VE-cadherin (CDH5) Cadherin-7 (CDH7) Cadherin-8 (CDH8) Br-cadherin (CDH12)

Hs Hs Hs Hs

5 5 5 5

Ascidian classic cadherins

CI-cadherin BS-cadherin

Ci Bs

5 5

Non-chordate classic cadherins • Variable number of CDs • LG and EGF domains • NCCD domain, only found in these cadherins

LvG-cadherin DE-cadherin DN-cadherin Hmr-1 (splice products 1a/1b)

Lv Dm Dm Ce

17 8 18 3/19*

Fatoid cadherins • More than 30 CDs, closely related in sequence • Flamingo box in some members • LG and EGF domains • Conserved region in the cytoplasmic domain (between vertebrates and fly)

Examples include: hFat1 hFat2 DCad76E (CG7749) Cdh4 (F25F2.2)

Hs Hs Dm Ce

34 34 34 32

Other Fat-like cadherins • Variable number of CDs • Flamingo box in Fat • LG and EGF domains in Fat and Cdh-3

Fat Dachsous Cdh-3 (ZK112.7) Cdh-1 (R10F2.2)

Dm Dm Ce Ce

34 27 19 25

(III) Seven-pass transmembrane cadherins

• Seven-pass transmembrane domain similar to G-protein linked receptors (Secretins) • Flamingo box • LG and EGF domains

hFlamingo1 hFlamingo2 Flamingo/Starry night Cdh-6 (F15B9.7)

Hs Hs Dm Ce

9 9 9 9

(IV) DCad102F-like cadherins

• Sequence conservation throughout much of the protein • LG domain • Glu/Ser-rich cytoplasmic domain

KIAA0911 KIAA0726 DCad102F (CG11059) Cdh-11 (B0034.3)

Hs Hs Dm Ce

2 2 2 2

(V) Protein kinase cadherins

• Tyrosine kinase domain (RET-Cadherins) • Putative Ser/Thr kinase

RET DRet (CG1061+CG14396) DCad96Ca (CG10244)

Hs Dm Dm

2 1‡ 1

(VI) Desmosomal cadherins • Only found in vertebrates • Localize at desmosomes • Interact with plakoglobin, desmoplakin and plakophillins

Desmocollins • Conserved cytoplasmic domain

Desmocollin-1 Desmocollin-2 Desmocollin-3 Desmoglein-1 Desmoglein-2 Desmoglein-3

Hs Hs Hs Hs Hs Hs

5 5 5 5 5 5

Protocadherins (Pcdh) α, β and γ • 52 protocadherins encoded by 3 genes • all Pcdh-α/CNR proteins have a constant C-terminal cytoplasmic domain that interacts with Fyn tyrosine kinase

Examples include: Pcdh-α3 Pcdh-β1 Pcdh-γA9

Hs Hs Hs

6 6 6

Other protocadherins

Examples include: Pcdh-1 Pcdh-8 (Arcadlin)

Hs Hs

6 6

(I) Classic cadherins Vertebrate Type I classic cadherins • Highly conserved cytoplasmic • HAV motif in first CD domain that binds to catenins • Often found at adherens junctions

(II) Fat-like cadherins • Very large extracellular domain with up to 34 CDs • Heterogeneous subfamily

(VII) Protocadherins • Only found in vertebrates

Desmogleins • Conserved cytoplasmic domain

*J. Petite, personal communication; ‡ R. Cagan, personal communication (Species: Hs, Homo sapiens; Dm, Drosophila melanogaster; Ce, Caenorhabditis elegans, Ci, Ciona intestinalis; Bs, Botryllus schlosseri; Lv, Lytechinus variegatus. Domains: CD, cadherin domain; CNR, cadherin-related neuronal receptor; EGF, epidermal growth factor; LG, laminin G; NCCD, non-chordate classic cadherin domain.) Uncharacterized cadherins of Drosophila are named according to their cytological map position (for example, DCad 102F).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 9 3

REVIEWS

I. Classic cadherins E-cadherin

Catenins

DE-cadherin

Catenins

II. Fat-like cadherins DCad76E

III. Seven-pass transmembrane cadherins Flamingo

IV. DCad102F-like cadherins

V. Protein kinase cadherins RET

Plakophillins Plakoglobin Desmoplakin

VI. Desmosomal cadherins Desmocollin

Fyn

VII. Protocadherins Pcdhα/CNR

Plasma membrane

Cadherin domain

G LG domain

Flamingo box

EGF domain/C-rich

NCCD domain

Tyrosine-kinase domain

Figure 2 | Structural diversity of the cadherin superfamily. Representatives of each of the seven subfamilies of cadherins are shown. Subfamilies I to VII are conserved between nematodes (Caernorhabditis elegans), arthropods (Drosophila melanogaster) and chordates (humans). Members of subfamily V are found in chordates and Drosophila, whereas cadherins of subfamilies VI and VII are at present only known in vertebrates. Binding partners for the cytoplasmic tail of cadherins have been characterized for classic cadherins (subfamily I), for desmosomal cadherins (subfamily VI), and for Pcdhα/CNR protocadherins (subfamily VII). These interacting factors are listed at the right. It was recently shown that DE-cadherin is proteolytically cleaved within the NCCD domain during maturation (arrowhead)95. (NCCD, nonchordate classic cadherin domain; EGF, epidermal growth factor; LG, Laminin G)

formation of collagen gene clusters was caused by the evolution of a complex extracellular matrix, the nematode cuticle17. The amplification of cadherins in vertebrates might be explained by the more complex tissue interactions found in humans and other vertebrates compared with invertebrates, particularly the large increase in size and complexity in the central nervous system. Structural basis of cell adhesion

Although it is broadly accepted that the predominant role of cadherins is to mediate adhesive interactions between cells, the mechanism of adhesive contact formation is still a matter of intense research. Structural studies have focused on vertebrate classic cadherins. These molecules are believed to form two types of dimers. Cadherins associate laterally within the same plasma membrane to form parallel cis dimers, and cadherins protruding from adjacent plasma membranes associate in an anti-parallel fashion to form trans dimers. The structure of the first CD of E-cadherin18 and of N-cadherin19 revealed that the cadherin fold consists of a seven-strand β-sheet with its amino and

carboxyl termini located at opposite ends of the molecule (FIG. 1b). The crystal structures of peptides containing the first and second CDs (CD1 and CD2) of E-cadherin20,21, or of N-cadherin22, indicate that calcium is central in cis-dimer formation. Each dimer associates with six calcium ions through residues that are located in the linker region between CD1 and CD2 (FIG. 3a). Single amino-acid substitutions in the calcium binding sites can disrupt cell aggregation in vivo23. Calcium binding makes CDs arrange in a rigid structure 20,21,24,25 that is resistant to proteolysis26. Crystallographic analysis on the first and second CDs of E-cadherin and N-cadherin have provided clues as to how cadherin molecules induce lateral clustering essential for the formation of a stable adhesive interface between adjacent cells. Although different mechanisms underlying cis dimerization have been observed in the crystal structures of different cadherin molecules19,20,22, an emerging theme is that two cadherin molecules form a cis dimer that functions as a building block for lateral clustering. These and other studies indicate that cis dimerization or more extensive lateral clustering is a prerequisite for stable cell adhesion27–29. Although cis dimers might primarily form as homodimers, the formation of functional cis heterodimers between N- and R-cadherin has been reported30. Adhesion between opposing cell membranes requires the formation of trans dimers (FIG. 3b). The mechanism of trans-dimer formation is, at present, controversial. Several studies indicate that trans dimers form by interactions between the amino-terminal CDs of opposing cadherin molecules19,21,28,31. These data are corroborated by early findings that located the homophilic binding specificity of classic cadherins within the amino-terminal CD32,33. On the basis of the crystal structure of the first CD from N-cadherin, a zipper model for trans dimerization was proposed that involved only the tip of the amino-terminal CD19. This model provided an early foundation for understanding the mechanics of the cell adhesion interface. However, the subsequent crystal structures of CD1 and CD2 from N-cadherin22 and E-cadherin20,21 did not show the adhesion interface seen in the first CD of N-cadherin. In addition, a recent biophysical study indicates a different type of trans-dimer association, in which the five CDs show variable degrees of lateral overlap, including the complete anti-parallel overlap of all five CDs (FIG. 3b)34. In the presence of calcium, the extracellular part of vertebrate classic cadherins forms a rod-like structure of about 20 nm in length with each individual CD spanning about 4.5 nm (REFS 19,20,25). Full lateral overlap of trans dimers would imply a distance between adjacent plasma membranes of 20–25 nm, a value consistent with the distance between plasma membranes at adherens junctions that is found in ultrastructural studies. Adhesive contacts and adherens junctions

Two types of adhesive contacts are mediated by classic cadherins: diffuse adhesive contacts all along a cell–cell contact surface, and more discrete contacts by ultrastructurally defined adherens junctions, such as the www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS

ZONULA ADHERENS

A cell–cell adherens junction that forms a circumferential belt around the apical pole of epithelial cells. PDZ DOMAINS

Protein–protein interaction domain, first found in PSD-95, DLG and ZO-1. SH3 DOMAINS

Src homology region 3 domains. Protein sequences of about 50 amino acids that recognize and bind sequences rich in proline. FILOPODIUM

Finger-like exploratory cell extension found in crawling cells and growth cones. LAMELLIPODIUM

Thin sheet-like cell extension found at the leading edge of crawling cells or growth cones

ZONULA ADHERENS. Diffuse adhesive contacts probably involve the oligomerization of cadherin trans dimers, as individual trans dimers provide little adhesive strength25,29. Adherens junctions could simply represent very large arrays of trans dimers. However, the situation seems to be more elaborate, as cadherins might not be the principal components of adherens junctions, at least in some cases. Indeed, adherens junctions can form in the absence of Hmr-1 cadherin, the only classic cadherin in C. elegans, or in the absence of its associated catenins35,36. In mouse and Drosophila embryos, where E-cadherin or DE-cadherin, respectively, are essential for adherens junction assembly and epithelial integrity, markedly reduced levels of these cadherins can still support the formation of normally sized adherens junctions37–39. These observations are inconsistent with a model in which adherens junctions simply represent a large array of cadherins and associated cytoplasmic proteins. Instead, cadherin trans dimers probably form small clusters separated by other proteins, and the density of these clusters in the adherens junctions may vary considerably without affecting the size of the adherens junction. A novel protein complex has recently been characterized that is concentrated at adherens junctions40–42. This complex is composed of Nectin, a transmembrane protein of the immunoglobulin superfamily that interacts with the PDZ-DOMAIN protein Afadin, which in turn can bind to Ponsin, a protein containing three SH3 DOMAINS (FIG. 4a). This complex can interact with the actin cytoskeleton. Nectin and cadherin complexes interact with each other and are recruited together to adherens junctions43. Initial functional studies indicate

that Afadin is important for junctional organization and epithelial integrity44. So adherens junctions contain two complexes that interact with each other and with the actin cytoskeleton. Cell and tissue polarity

Epithelial cells provide a clear example of cell polarity, with various molecules, including proteins, sorting to distinct apical and basolateral membrane domains. The crucial role of classic cadherins and their associated catenins in epithelial differentiation has been well documented, and these protein complexes seem to be broadly important for forming and maintaining epithelial tissues45,46. Conversely, downregulation of classic cadherins, such as E-cadherin or DE-cadherin, is often associated with a loss of epithelial morphology during normal development and in many carcinomas. The zinc-finger transcription factor Snail is important for repressing the expression of DE-cadherin and E-cadherin in non-epithelial cells47–49. Epithelial cells usually form a continuous tissue structure. However, at certain times during normal development, or in experimental cell-culture models, epithelial cells have free edges that approach each other to establish new lateral contacts36,50,51. The initial contact between cells is made by FILOPODIA or LAMELLIPODIA, and such contacts are stabilized by classic cadherins. When these contacts broaden, cadherins concentrate in discrete puncta. The adhesive interactions are further stabilized through linkage of cadherins to the cytoskeleton and, eventually, by the formation of mature adherens junctions. Cadherin-mediated adhesion leads to

N Plasma membrane

Ecad1

20–25 nm

Ca2+

Ecad2

Cis dimer

Trans dimer

Figure 3 | Ca2+-mediated cis- and trans-dimer formation of vertebrate classic cadherins. a | Dimer interface between two N-terminal repeats of E-cadherin domains 1 (Ecad1) and 2 (Ecad2). Each cadherin molecule binds three calcium ions that are important in the rigidification and cis-dimer association of cadherins20,21. b | A cis dimer consists of two cadherin molecules within the same plasma membrane that are associated laterally. The pairs of cadherin molecules from opposing cells that associate with one another are referred to as trans dimers. Different models for trans-dimer formation have been proposed that suggest different extents of lateral overlap between the extracellular regions. Red dots indicate the location of Ca2+ ions between adjacent CDs.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 9 5

REVIEWS recruitment of specific cytoskeletal factors, such as the actin-associated factor Mena51 and other transmembrane proteins, to cell–cell contact sites46. Cadherins seem to be directly involved in maintaining cell polarity by directing the localization of the sec6/8 complex, which specifies vesicle targeting to the lateral membrane52. This recruitment, and the continuous polarized delivery of specific molecular components to the lateral membrane, establishes and maintains the lateral membrane domain of epithelial cells and contributes to epithelial apical–basal polarity46,52. Interestingly, in fully polarized epithelial cells, the sec6/8 complex is not found along the entire lateral membrane but is concentrated in close association with apical adherens junctions, indicating a potentially direct molecular link between the cadherin–catenin complex and the vesicle targeting machinery (FIG. 4a)52.

FOLLICLE CELLS

In this review, the term follicle cells refers to cells that surround the developing insect egg and secrete the egg membranes, the chorion and vitelline envelope. GASTRULATION

Series of morphogenetic movements observed during the early development of most animals that leads to the formation of a multilayered embryo with an outer cell layer (ectoderm), an inner cell layer (endoderm), and an intermediate cell layer (mesoderm).

a Actin cytoskeleton Ponsin Afadin

Afadin Nectin Ponsin

Catenins

Ponsin Afadin

Afadin Ponsin

Catenins Sec6/8 complex

Sec6/8 complex Catenins

Classic cadherin

Actin cytoskeleton

b Actin cytoskeleton

Actin cytoskeleton ?

? Catenins

Classic cadherin

Fyn

Synaptic vesicle docking

Active zone (Sec6/8 complex)

Catenins ? Actin cytoskeleton

Catenins

Fyn

Catenins

? Actin cytoskeleton

Protocadherin

Cell movement

Figure 4 | Comparison between cadherin-mediated adhesive interactions at epithelial and synaptic adherens junctions. a | Schematic of the zonula adherens, a circumferential adherens junction found in epithelial cells. This junction contains the classic cadherin–catenin complex and the recently identified nectin/afadin/ponsin complex40–42. Both complexes interact with the actin cytoskeleton and with each other10,43. The zonula adherens is closely associated with a vesicle docking site that contains the Sec6/8 complex52. b | At the interneuronal synapse, we also find a close association between adherens junctions and vesicle docking zones. The sec6/8 complex was found to associate with the postsynaptic membrane only during synaptogenesis96. The classic cadherin–catenin complex is a principal component of synaptic adherens junctions similar to the zonula adherens. Protocadherins that localize to synapses include Arcadlin84 and the Pcdhα/CNR protocadherins75. Whether protocadherins contribute to synaptic adhesion remains to be established. The Pcdhα/CNR protocadherins interact with the cytoplasmic protein kinase Fyn, and seem to reside within the active zone, indicating that they might have a primary role in signalling rather than adhesion13,75.

In addition to their role in apical–basal polarity, cadherin superfamily members were recently implicated in a second form of cell polarity called planar epithelial polarity (FIG. 5). This property is found in many epithelia. One example is the fly wing epithelium, where each cell polarizes its actin cytoskeleton along the proximal–distal axis, such that a bundle of actin filaments polymerizes and projects from the surface at the distal-most vertex of each cell, ultimately forming a wing hair (FIG. 5a). Genes involved in establishing the planar polarity of the wing epithelium include components of the Wnt/Frizzled signalling pathway, Frizzled and Dishevelled53, and three different cadherins, Fat, Dachsous54 and Flamingo/Starry Night55,56. Although the mechanism by which cadherins affect planar polarity is unknown, it was found that Flamingo/Starry Night adopts an asymmetric distribution during polarity establishment, becoming enriched at the proximal and distal cell surfaces (FIG. 5b)55. Planar polarity also influences polymerization of the microtubule cytoskeleton, manifesting itself through orientation of the mitotic spindle and thereby the axis of cell division along the body axes. Studies in both C. elegans and Drosophila implicate the Wnt/Frizzled pathway in this process53. Many planar polarity genes have not been examined in this context, but it is at least clear that Flamingo/Starry Night is essential for spindle orientation57. Two other examples that highlight important roles of cadherins in generating asymmetric tissue organization are the contribution of DE-cadherin to the formation of the anterior–posterior axis in Drosophila, and the function of N-cadherin in setting up left–right asymmetry in the chick. A cell sorting process that is driven by different levels of DE-cadherin directs the oocyte to the posterior pole of the egg chamber during Drosophila oogenesis8,58. This highly reproducible positioning event allows the oocyte to interact with a specific group of FOLLICLE CELLS, thereby initiating a cascade of cell interactions that are crucial for the formation of the embryonic anterior–posterior axis. Disruption of the function of N-cadherin during chick GASTRULATION leads to a random orientation of the heart along the left–right axis59. Asymmetric N-cadherin expression and cell movements that prefigure the position of the heart and other organs along the left–right axis are seen during gastrulation. How N-cadherin contributes to these asymmetric cell movements remains a mystery.

Many of the changes in cell shape or movement observed during development occur while cells are in direct contact and require, therefore, dynamic changes in adhesive interactions. These changes may play a permissive role, as the release of adhesion is important for the relative movement of cells that are in contact. However, adhesive interactions also directly promote movement, as traction must be generated between cells for cell rearrangements to occur in solid tissues. To determine whether changes in cadherin activity play a permissive or a more active role can be difficult, as illustrated by the analysis of C-cadherin function during www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS

NEUROEPITHELIUM

Epithelial layer of cells that gives rise to the nervous system. NEURULATION

Morphogenetic process during which the progenitors of the nervous system segregate from the ectoderm.

convergent extension movements in Xenopus gastrulation. In this process, cells move towards the dorsal midline of the embryo, thereby rearranging by cell intercalation, leading to an extension of the embryo along the anterior–posterior axis (FIG. 6a)60. Adhesion mediated by the classic cadherin C-cadherin must be reduced to permit these movements to occur61. However, the disruption of C-cadherin activity causes defects not only during gastrulation movements, but also in tissue structure62, raising the possibility that the disruptions in cell movement might be a secondary consequence of a compromised cell architecture. Similar difficulties have emerged from the analysis of DE-cadherin in embryonic morphogenesis where its role in epithelial maintenance might mask a function in promoting cell rearrangements37,38. Paraxial protocadherin (PAPC) seems to be directly involved in convergent extension in Xenopus and zebrafish embryos, where it is expressed in the mesoderm during gastrulation63,64. PAPC, which can promote homotypic cell adhesion, is required for convergent extension of the mesoderm. Notably, overexpression of PAPC can promote convergent extension under certain experimental conditions63. These findings argue that PAPC acts as an adhesion receptor that directly promotes cell movement, perhaps providing traction needed for cell motility. Alternative and non-exclusive possibilities are that PAPC is primarily a signalling receptor, as was suggested for other protocadherins13. PAPC activity might also generate the tissue polarization that is observed during convergent extension (FIG. 6a)60,63, functioning similarly to the activity of other cadherins in planar epithelial polarization, outlined above. Intriguingly, cell polarization during convergent extension resembles planar polarity in that it also requires Wnt/Frizzled signalling65–67. The requirement for DE-cadherin in cell migration during Drosophila oogenesis is a convincing example for a direct role of classic cadherins in cell migration on a cellular substrate. DE-cadherin is involved in the migration of a small group of somatic cells, the ‘border’ cells, on the surface of the much larger germline cells68 (FIG. 6b). It is required in both the somatic and germline cells for this movement to occur. DE-cadherin is not required for the formation of the border cell cluster and, more importantly, is not required for maintaining integrity of the border cell cluster during migration. In the case of integrin-based cell migration, it was shown that intermediate levels of adhesion to the substrate promotes maximal migration speed, with both positive and negative deviations slowing or halting motility69. Similarly, reduction in the level of DE-cadherin reduces the speed of border cell migration68, indicating that DEcadherin might not have just a permissive role, but might be the key adhesion molecule that provides traction for border cells to travel over germline cells. Organization of the nervous system

Various cadherins are expressed in the nervous system in complex patterns. The first example was N-cadherin, which is broadly expressed in the NEUROEPITHELIUM,

beginning at NEURULATION. This expression pattern led to the speculation that N-cadherin might be critical for the segregation of neural and epidermal tissues during neural tube formation. However, an essential role in neurulation was disproved by the knockout of mouse N-cadherin70. After neurulation, but before neuronal differentiation, many classic cadherins, including Ncadherin, are expressed in the developing central nervous system (CNS) in a region-specific manner that often coincides with morphological boundaries71. The most direct evidence that these cadherins contribute to the subdivision of the neuroepithelum has come from analysis of the Xenopus type II classic cadherin F-cadherin72. The expression of F-cadherin confines neuroepithelial cells to the sulcus limitans, a region separating the dorsal and ventral halves of the caudal neural tube. One apparent consequence of F-cadherin expression in a

Planar polarity

Apical Zonula adherens Diffuse adhesive contact

Basal

Distal

Figure 5 | Cadherins in apicobasal and planar epithelial polarity. a | The example depicted here is the wing epithelium of Drosophila melanogaster, shown in cross section. Classic cadherins mediate lateral cell contacts (yellow) between epithelial cells that can take the form of either a diffuse adhesive contact, or an adherens junction, such as the zonula adherens, that can be seen in electron micrographs as an electron-dense specialization of the plasma membrane. Epithelial sheets are obviously different across their apical–basal axis, but many epithelial cells can also discern directions in the plane of the epithelium with respect to the organ or body axis of which they are a part. They use this information to polarize their actin and microtubule cytoskeletons along this axis. The most obvious indication of planar polarity in the wing epithelium is the hair that is formed by each cell. Hairs emerge from a distal region of the apical cell surface and all point distally. b | The array of cells in a Drosophila wing epithelium, viewed from above. A wild-type array is shown on the left, illustrating the uniform planar polarity (wing hair orientation) and the distribution of Flamingo/Starry Night (green), which accumulates at the proximal and distal surfaces of every cell. The right array shows a group of cells that contain Flamingo/Starry Night mutant cells (absence of green), in which the orientation of wing hairs and therefore the axis of planar polarity has shifted55,56.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 9 7

REVIEWS

Figure 6 | Cell movements that involve cadherins. a | Convergent extension is a cell rearrangement that involves the transient polarization of cells, which then move, converging towards the centre of the tissue. Cell intercalation leads to an extension of the tissue perpendicular to the axis of convergence. This movement is seen, for example, during frog gastrulation, where it may involve C-cadherin and paraxial protocadherin61,63. b | The border cell cluster (green) is a small group of Drosophila follicle cells that migrate on the surface of much larger germline cells (green/yellow). This movement is driven by DE-cadherin, which is required in both border cells and germline cells68. c | N-cadherin and DNcadherin mediate the movement of neuronal growth cones (red) on cellular substrates such as axon bundles (green)76–78.

NEURITE

Process extended by a nerve cell that can give rise to an axon or a dendrite. FASCICULATION

Bundling of axonal processes of neurons.

the sulcus limitans is that these cells remain a coherent group and do not participate in the extensive cell rearrangments that take place during neurulation. Protocadherins also contribute to CNS regionalization by controlling the migration of neurons that will organize into different cortical layers during brain morphogenesis. Although cadherins, including protocadherins, are generally viewed as homophilic adhesion molecules, recent work indicates that the Pcdhα/CNR protocadherins might also function as receptors or coreceptors for extracellular ligands in the brain13,73. This work began with studies of the secreted molecule Reelin, identified because mutant mice have a marked behavioural disorder. Two protein families have been shown to function as reelin receptors, perhaps as a heteromeric complex: members of the LDL-receptor-related protein family, which couple to the cytoplasmic adaptor protein mDab1; and members of the Pcdhα/CNR family, which bind the non-receptor tyrosine kinase Fyn73,74. As Pcdhα/CNR protocadherins show considerable molecular diversity and differential expression patterns within local brain areas, it is possible that Reelin receptor complexes that contain different Pcdhα/CNR protocadherins are instructive in positioning and differentiating neuronal sub-populations within the cortex73,75. Classic cadherins are also important during the outgrowth of NEURITES and during axonal patterning and FASCICULATION (FIG. 6c). Early studies that indicated that N-cadherin can function as a substrate for neurite extension in cultured cells were reinforced by the finding that N-cadherin is required for the normal outgrowth and guidance of retinal axons76,77. In Drosophila, DN-cadherin is the only classic cadherin

expressed in the developing embryonic CNS78. Null mutations in DN-cadherin and the Drosophila catenin gene armadillo affect axon outgrowth, although in a mild fashion, with many axons finding their targets appropriately78,79. In this respect, cadherins resemble various other axon guidance cues that direct axon outgrowth in a combinatorial fashion, with individual cues having subtle functions9. In addition, it was recently found that the patterning of dendrites that extend from Drosophila sensory neurons requires Flamingo/Starry Night80. Both classic and protocadherins localize to synapses, indicating that they may contribute to the generation of adhesive specificity needed to build complex neural networks75,81–84. The synapse is an adhesive contact between two neurons, with the transmitter release zone framed by adherens junctions that are ultrastructurally similar to epithelial adherens junctions, and that share with them the cadherin–catenin complex as a principal molecular component (FIG. 4b)82,83. In synapses, as in epithelial cells, adherens junctions are closely associated with vesiclerelease zones (FIG. 4). It is believed that classic cadherins are important during synaptic adhesion, whereas the adhesive role of protocadherins at synapses remains to be clarified. Recent intriguing evidence indicates that synaptic activity can change the distribution and adhesive state of N-cadherin85. Cadherins, in turn, can influence synaptic activity84,86,87. These findings indicate an intimate relationship between synaptic adhesion and activity, raising the possibility that cadherins are important regulators of synaptic plasticity and activity modulation13,89. Some classic cadherins, such as cadherin-6, are expressed in groups of neurons that form neural circuits, indicating that cadherins might functionally integrate such neural circuits71,90. Protocadherins are also expressed in divergent and restricted patterns in the CNS, indicating that they might have a function in integrating neural circuits91,92. Moreover, the protocadherin genes Pcdhα, Pcdhβ and Pcdhγ can give rise to over 50 protocadherins18 and the Pcdhα/CNR proteins have a differential expression pattern within individual brain areas75. These findings raise the possibility that a ‘cadherin code’ exists, which could identify individual neurons and their synaptic contacts75,89, although neurexins and immunoglobulin-type adhesion receptors have also been proposed to contribute to synaptic specificity93,94. The structures of the Pcdhα, Pcdhβ and Pcdhγ genes show provocative similarities to the gene organization of immunoglobulins or T-cell receptors, which has led to the proposal that gene rearrangement might function in determining protocadherin expression patterns13,15,16, a speculation that remains to be proved. The future

The analysis of cadherins emphasizes the similarities between embryonic and neural morphogenesis. Cadherins have emerged as the predominant group of cell–cell adhesion molecules involved in embryonic morphogenesis, determining cell and tissue architecture, and controlling dynamic changes in cell shape and position. The role of individual cadherins in several specific morphogenetic processes has been determined, which www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

1. 2.

4. 5.

10. 11.

12.

13.

14.

15.

16.

17.

18.

19. 20.

Steinberg, M. S. Reconstruction of tissue by dissociated cells. Science 141, 401–408. (1963). Sperry, R. W. Chemoaffinity in the orderly growth of nerve fiber patterns and connections. Proc. Natl Acad. Sci. USA 50, 703–710 (1963). Gumbiner, B. M. Cell adhesion: The molecular basis of tissue architecture and morphogenesis. Cell 84, 345–579 (1996). Hynes, R. O. Cell adhesion: Old and new questions. Trends Cell Biol. 9, M33–M37 (1999). Townes, P. L. & Holtfreter, J. Directed movements and selective adhesion of embryonic amphibian cells. J. Exp. Zool. 128, 53–120 (1955). Nose, A., Nagafuchi, A, & Takeichi, M. Expressed recombinant cadherins mediate cell sorting in model systems. Cell 54, 993–1001 (1988). This paper shows for the first time cadherinmediated cell sorting in cell-culture cells. Steinberg, M. S. & Takeichi, M. Experimental specification of cell sorting, tissue spreading, and specific spatial patterning by quantitative differences in cadherin expression. Proc. Natl Acad. Sci. USA 91, 206–209 (1994). Godt, D. & Tepass, U. Drosophila oocyte localization is mediated by differential cadherin-based adhesion. Nature 395, 387–391 (1998). This paper documents for the first time an in vivo cell sorting process that is driven by differential expression of a cadherin. Tessier-Lavigne, M. & Goodman, C. S. The molecular biology of axon guidance. Science 274, 1123–1133 (1996). Takeichi, M. Cadherin cell adhesion receptors as a morphogenetic regulator. Science 251, 1451–1455 (1991). Wong, E. F. S., Brar, S. K., Sesaki, H., Yang, C. & Siu, C. H. Molecular cloning and characterization of DdCAD-1, a Ca2+-dependent cell–cell adhesion molecule, in Dictyostelium discoideum. J. Biol. Chem. 271, 16399–16408 (1996). Gallin, W. J. Evolution of the ‘classical’ cadherin family of cell adhesion molecules in vertebrates. Mol. Biol. Evol. 15, 1099–1107 (1998). Yagi, T. & Takeichi, M. Cadherin superfamily genes: Functions, genomic organization, and neurologic diversity. Genes Dev. 14, 1169–1180 (2000). Nollet, F., Kools, P. & van Roy, F. Phylogenetic analysis of the cadherin superfamily allows identification of six major subfamilies besides several solitary members. J. Mol. Biol. 299, 551–72 (2000). Wu, Q. & Maniatis, T. A. striking organization of a large family of human neural cadherin-like cell adhesion genes. Cell 97, 779–790 (1999). Sugino, H. et al. Genomic organization of the family of CNR cadherin genes in mice and humans. Genomics 63, 75–87 (2000). Hutter, H. et al. Conservation and novelty in the evolution of cell adhesion and extracellular matrix genes. Science 287, 989–994 (2000). Overduin, M. et al. Solution structure of the epithelial cadherin domain responsible for selective cell adhesion. Science 267, 386–389 (1995). Shapiro, L. et al. Structural basis of cell–cell adhesion by cadherins. Nature 374, 327–337 (1995). Nagar, B., Overduin, M., Ikura, M. & Rini, J. M. Structural basis of calcium-induced E–cadherin rigidification and dimerization. Nature 380, 360–364 (1996).

21.

22.

23.

24.

25.

26.

27.

28.

29.

30. 31.

32.

33.

34.

35.

36.

37.

tant during neural morphogenesis, although the functional significance of cadherins in neural development remains less well understood. One important challenge will be to determine the exact mechanism underlying the specificity and stability of cadherin-mediated cell–cell adhesion, and to explore the variation in adhesive mechanisms and cellular responses of different types of cadherins. A second important challenge will be to substantiate the conjecture that cadherins provide an adhesive code that controls synaptic specificity. Elucidating whether and how different families of adhesion receptors cooperate in this process will represent an enormous advance in our understanding of complex neural network formation.

This is one of several papers (see also refs 18–26,31) that describe the three-dimensional structure of the cadherin domain, and analyse the role of calcium ions in the formation of cadherin dimers. Pertz, O. et al. A new crystal structure, Ca2+ dependence and mutational analysis reveal molecular details of E–cadherin homoassociation. EMBO J. 18, 1738–1747 (1999). Tamura, K., Shan, W. S., Hendrickson, W. A., Colman, D. R. & Shapiro, L. Structure–function analysis of cell adhesion by neural (N-) cadherin. Neuron 20, 1153–1163 (1998). Ozawa, M., Engel, J. & Kemler, R. Single amino acid substitutions in one Ca2+ binding site of uvomorulin abolish the adhesive function. Cell 63,1033–1038 (1990). Pokutta, S., Herrenknecht, K., Kemler, R. & Engel, J. Conformational changes of the recombinant extracellular domain of E–cadherin upon calcium binding. Eur. J. Biochem. 223, 1019–1026 (1994). Baumgartner, W. et al. Cadherin interaction probed by atomic force microscopy. Proc. Natl Acad. Sci. USA 97, 4005–4010 (2000). Hyafil, F., Babinet, C. & Jacob, F. Cell–cell interactions in early embryogenesis: a molecular approach to the role of calcium. Cell 26, 447–454 (1981) Brieher, W. M., Yap, A. S., & Gumbiner, B. M. Lateral dimerization is required for the homophilic binding activity of C-cadherin. J. Cell Biol. 135, 487–496 (1996). Alattia, J. R. et al. Lateral self-assembly of E-cadherin directed by cooperative calcium binding. FEBS Lett. 417, 405–408 (1997). Yap, A. S., Brieher, W. M., Pruschy, M. & Gumbiner, B. M. Lateral clustering of the adhesive ectodomain: a fundamental determinant of cadherin function. Curr. Biol. 7, 308–315 (1997). Shan, W. S. et al. Functional cis-heterodimers of N- and Rcadherins. J. Cell Biol. 148, 579–590 (2000). Tomschy, A., Fauser, C., Landwehr, R. & Engel, J. Homophilic adhesion of E-cadherin occurs by a cooperative two-step interaction of N–terminal domains. EMBO J. 15, 3507–3514 (1996). Nose, A., Tsuji, K. & Takeichi, M. Localization of specificity determining sites in cadherin cell adhesion molecules. Cell 61,147–155 (1990). Blaschuk, O. W., Sullivan, R., David, S. & Pouliot, Y. Identification of a cadherin cell adhesion recognition sequence. Dev. Biol. 139, 227–229 (1990). Sivasankar, S., Brieher, W., Lavrik, N., Gumbiner, B. & Leckband, D. Direct molecular force measurements of multiple adhesive interactions between cadherin ectodomains. Proc. Natl Acad. Sci. USA 96, 11820–11824 (1999). A recent paper that suggests a new model of cadherin trans-dimer formation. Costa, M. et al. A putative catenin–cadherin system mediates morphogenesis of the Caenorhabditis elegans embryo. J. Cell Biol. 141, 297–308 (1998). Raich, W. B., Agbunag, C. & Hardin, J. Rapid epithelialsheet sealing in the Caenorhabditis elegans embryo requires cadherin-dependent filopodial priming. Curr. Biol. 9, 1139–1146 (1999). This paper analyses the dynamics of cadherindependent contact formation between epithelial cells during C. elegans embryogenesis. Uemura, T. et al. Zygotic Drosophila E-cadherin expression is required for processes of dynamic epithelial cell

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

38.

39.

40.

41.

42.

43.

44.

45. 46.

47.

48.

49.

50.

51.

52.

53.

54.

rearrangement in the Drosophila embryo. Genes Dev. 10, 659–671 (1996). Tepass, U. et al. shotgun encodes Drosophila E–cadherin and is preferentially required during cell rearrangement in the neuroectoderm and other morphogenetically active epithelia. Genes Dev. 10, 672–685 (1996). Ohsugi, M., Larue, L., Schwarz, H. & Kemler, R. Celljunctional and cytoskeletal organization in mouse blastocysts lacking E-cadherin. Dev. Biol. 185, 261–271 (1997). Mandai, K. et al. Afadin: A novel actin filament-binding protein with one PDZ domain localized at cadherin-based cell-to-cell adherens junction. J. Cell Biol. 139, 517–528 (1997). Mandai, K. et al. Ponsin/SH3P12: An l-afadin- and vinculin-binding protein localized at cell-cell and cell-matrix adherens junctions. J. Cell Biol. 144, 1001–1017 (1999). Takahashi, K. et al. Nectin/PRR: An immunoglobulin-like cell adhesion molecule recruited to cadherin-based adherens junctions through interaction with Afadin, a PDZ domain-containing protein. J. Cell Biol. 145, 539–549 (1999). One of a series of papers (refs 40–44) that describe a new protein complex that localizes to adherens junctions, and that might interact with the cadherin–catenin complex functionally. Tachibana, K. et al. Two cell adhesion molecules, nectin and cadherin, interact through their cytoplasmic domainassociated proteins. J. Cell Biol. 150, 1161–1175 (2000). Ikeda, W. et al. Afadin: A key molecule essential for structural organization of cell–cell junctions of polarized epithelia during embryogenesis. J. Cell Biol. 146, 1117–1132 (1999). Tepass, U. Genetic analysis of cadherin function in animal morphogenesis. Curr. Opin. Cell Biol. 11, 540–548 (1999). Yeaman, C., Grindstaff, K. K. & Nelson, W. J. New perspectives on mechanisms involved in generating epithelial cell. Phys. Rev. 79, 73–98 (1999). Oda, H., Tsukita, S. & Takeichi, M. Dynamic behavior of the cadherin-based cell–cell adhesion system during Drosophila gastrulation. Dev. Biol. 203, 435–450 (1998). Cano, A. et al. The transcription factor snail controls epithelial–mesenchymal transitions by repressing Ecadherin expression. Nature Cell Biol. 2, 76–83 (2000). Batlle, E. et al. The transcription factor snail is a repressor of E–cadherin gene expression in epithelial tumour cells. Nature Cell Biol. 2, 84–89 (2000). Adams, C. L. et al. Mechanisms of epithelial cell–cell adhesion and cell compaction revealed by high-resolution tracking of E-cadherin–green fluorescent protein. J. Cell Biol. 142, 1105–1119 (1998). Vasioukhin, V., Bauer, C., Yin, M. & Fuchs, E. Directed actin polymerization is the driving force for epithelial cell–cell adhesion. Cell 100, 209–219 (2000). 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). This paper shows a close association between the sec6/8 complex, which is involved in lateral vesicle targeting, and cadherin-based adherens junctions. Peifer, M. & Polakis, P. Wnt signaling in oncogenesis and embryogenesis — a look outside the nucleus. Science 287, 1606–1609 (2000). Adler, P. N., Charlton, J. & Liu, J. Mutations in the cadherin superfamily member gene dachsous cause a tissue

VOLUME 1 | NOVEMBER 2000 | 9 9

REVIEWS

55.

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

66. 67.

68.

100

polarity phenotype by altering Frizzled signaling. Development 125, 959–968 (1998). Usui, T. et al. Flamingo, a seven-pass transmembrane cadherin, regulates planar cell polarity under the control of Frizzled. Cell 98, 585–595 (1999). Chae, J. et al. The Drosophila tissue polarity gene starry night encodes a member of the protocadherin family. Development 126, 5421–5429 (1999). References 55 and 56 show a role for the Flamingo/Starry Night cadherin in planar epithelial polarity. Reference 55 also shows that the subcellular distribution of Flamingo/Starry Night depends on the direction of Wnt/Frizzled signalling. Lu, B., Usui, T., Uemura, T., Jan, L. & Jan, Y. N. Flamingo controls the planar polarity of sensory bristles and asymmetric division of sensory organ precursors in Drosophila. Curr. Biol. 9, 1247–1250 (1999). Gonzalez-Reyes, A. & St Johnston, D. The Drosophila AP axis is polarised by the cadherin-mediated positioning of the oocyte. Development 125, 3635–3644 (1998). Garcia-Castro, M. I., Vielmetter, E. & Bronner-Fraser, M. N-Cadherin, a cell adhesion molecule involved in establishment of embryonic left–right asymmetry. Science 288, 1047–1051 (2000). Shih, J. & Keller, R. Cell motility driving mediolateral intercalation in explants of Xenopus laevis. Development 116, 901–914 (1992). Zhong, Y., Brieher, W. M. & Gumbiner, B. M. Analysis of Ccadherin regulation during tissue morphogenesis with an activating antibody. J. Cell Biol. 144, 351–359 (1999). Lee, C. H. & Gumbiner, B. M. Disruption of gastrulation movements in Xenopus by a dominant–negative mutant for C-cadherin. Dev. Biol. 171, 363–373 (1995). Kim, S. H., Yamamoto, A., Bouwmeester, T., Agius, E. & Robertis, E. M. The role of paraxial protocadherin in selective adhesion and cell movements of the mesoderm during Xenopus gastrulation. Development 125, 4681–4690 (1998). Experiments in this paper indicate an important role for a protocadherin in the convergent extension movements during frog gastrulation. Yamamoto, A. et al. Zebrafish paraxial protocadherin is a downstream target of spadetail involved in morphogenesis of gastrula mesoderm. Development 125, 3389–3397 (1998). Heisenberg, C. P. et al. Silberblick/Wnt11 mediates convergent extension movements during zebrafish gastrulation. Nature 405, 76–81 (2000). Wallingford, J. B. et al. Dishevelled controls cell polarity during Xenopus gastrulation. Nature 405, 81–85 (2000). Tada, M. & Smith, J. C. Xwnt11 is a target of Xenopus brachyury: regulation of gastrulation movements via dishevelled, but not through the canonical wnt pathway. Development 127, 2227–2238 (2000). Niewiadomska, P., Godt, D. & Tepass, U. DE–cadherin is required for intercellular motility during Drosophila

69.

70. 71. 72.

73.

74. 75.

76.

77.

78.

79.

80.

81.

82.

83.

oogenesis. J. Cell Biol. 144, 533–547 (1999). This paper documents a cadherin-dependent cell migration process. Palecek, S. P., Loftus, J. C., Ginsberg, M. H., Lauffenburger, D. A. & Horwitz, A. F. Integrin-ligand binding properties govern cell migration speed through cell–substratum adhesiveness. Nature 385, 537–540 (1997). Radice, G. L. et al. Developmental defects in mouse embryos lacking N-cadherin. Dev. Biol. 181, 64–78 (1997). Redies, C. Cadherins in the central nervous system. Prog. Neurobiol. 61, 611–648 (2000). Espeseth, A., Marnellos, G. & Kintner, C. The role of F–cadherin in localizing cells during neural tube formation in Xenopus embryos. Development 125, 301–312 (1998). Senzaki, K., Ogawa, M. & Yagi, T. Proteins of the CNR family are multiple receptors for Reelin. Cell 99, 635–647 (1999). Protocadherins of the CNR family are identified as receptors of the extracellular matrix protein Reelin, an interaction that might contribute to the migration and differentiation of neurons within the brain cortex. Gilmore, E. C. & Herrup, K. Cortical development: receiving reelin. Curr. Biol. 10, R162–R166 (2000). Kohmura, N. et al. Diversity revealed by a novel family of cadherins expressed in neurons at a synaptic complex. Neuron 20, 1137–1151 (1998). This paper identifies a closely related group of protocadherins that are differentially expressed in the brain and localize to synapses. Riehl, R. et al. Cadherin function is required for axon outgrowth in retinal ganglion cells in vivo. Neuron 17, 837–848 (1996). Inoue, A. & Sanes, J. R. Lamina-specific connectivity in the brain: Regulation by N-cadherin, neurotrophins, and glycoconjugates. Science 276, 1428–1431 (1997). Iwai, Y. et al. Axon patterning requires DN-cadherin, a novel neuronal adhesion receptor, in the Drosophila embryonic CNS. Neuron 19, 77–89 (1997). Loureiro, M. et al. Anomalous origin of the left pulmonary artery (Sling): A case report and review of the literature. Rev. Port. Cardiol. 17, 811–815 (1998). Gao, F. B., Brenman, J. E., Jan, L. Y. & Jan, Y. N. Genes regulating dendritic outgrowth, branching, and routing in Drosophila. Genes Dev. 13, 2549–2561 (1999). Yamagata, M., Herman, J. P. & Sanes, J. R. Laminaspecific expression of adhesion molecules in developing chick optic tectum. J. Neurosci. 15, 4556–4571 (1995). Fannon, A. M. & Colman, D. R. A model for central synaptic junctional complex formation based on the differential adhesive specificities of the cadherins. Neuron 17, 423–434 (1996). Uchida, N., Honjo, Y., Johnson, K. R., Wheelock, M. J. & Takeichi, M. The catenin/cadherin adhesion system is localized in synaptic junctions bordering transmitter release

84.

85. 86.

87.

88. 89.

90.

91.

92.

93. 94.

95.

96.

zones. J. Cell Biol. 135, 767–779 (1996). References 82 and 83 show that classic cadherins are components of synaptic adherens junctions. Yamagata, K. et al. Arcadlin is a neural activity-regulated cadherin involved in long term potentiation. J. Biol. Chem. 274, 19473–19479 (1999). Tanaka, H. et al. Molecular modification of N–cadherin in response to synaptic activity. Neuron 25, 93–107 (2000). Tang, L., Hung, C. P. & Schuman, E. M. A role for the cadherin family of cell adhesion molecules in hippocampal long-term potentiation. Neuron 20, 1165–1175 (1998). Manabe, T. et al. Loss of Cadherin–11 adhesion receptor enhances plastic changes in hippocampal synapses and modifies behavioral responses. Mol. Cell. Neurosci. 15, 534–546 (2000). Uemura, T. The cadherin superfamily at the synapse: More members, more missions. Cell 93, 1095–1098 (1998). Shapiro, L. & Colman, D. R. The diversity of cadherins and implications for a synaptic adhesive code in the CNS. Neuron 23, 427–430 (1999). Suzuki, S. C., Inoue, T., Kimura, Y., Tanaka, T. & Takeichi, M. Neuronal circuits are subdivided by differential expression of type-II classic cadherins in postnatal mouse brains. Mol. Cell. Neurosci. 9, 433–447 (1997). Obata, S. et al. A common protocadherin tail: Multiple protocadherins share the same sequence in their cytoplasmic domains and are expressed in different regions of brain. Cell. Adhes. Commun. 6, 323–333 (1998). Hirano, S., Yan, Q. & Suzuki, S. T. Expression of a novel protocadherin, OL–protocadherin, in a subset of functional systems of the developing mouse brain. J. Neurosci. 19, 995–1005 (1999). Missler, M. & Südhof, T. C. Neurexins: Three genes and 1001 products. Trends Genet. 14, 20–26 (1998). Schmücker, D. et al. Drosophila Dscam is an axon guidance receptor exhibiting extraordinary molecular diversity. Cell 101, 671–684 (2000). Oda, H., & Tsukita, S. Nonchordate classic cadherins have a structurally and functionally unique domain that is absent from chordate classic cadherins. Dev. Biol. 216, 406–422 (1999). Hazuka, C. D. et al. The sec6/8 complex is located at neurite outgrowth and axonal synapse-assembly domains. J. Neurosci. 19, 1324–1334 (1999).

Acknowledgements We would like to thank Y. Takai, J. Petite, R. Cagan and T. Uemura for communicating unpublished results. The work on cadherins in the authors’ laboratories is funded by grants from the National Cancer Institute of Canada with funds from the Canadian Cancer Society (to U.T. and M.I.), the Canadian Institute for Health Research (to U.T. and D.G.), University of Toronto Connaught Committee (to D.G.), the National Institutes of Health (to M.P.), the Human Frontier Science Program (to M.P.) and the US Army Breast Cancer Research Program (to M.P.).

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS

THE EXPANDING POLYMERASE UNIVERSE Myron F. Goodman and Brigette Tippin Over the past year, the number of known prokaryotic and eukaryotic DNA polymerases has exploded. Many of these newly discovered enzymes copy aberrant bases in the DNA template over which ‘respectable’ polymerases fear to tread. The next step is to unravel their functions, which are thought to range from error-prone copying of DNA lesions, somatic hypermutation and avoidance of skin cancer, to restarting stalled replication forks and repairing doublestranded DNA breaks. DCMP TRANSFERASE

The DNA template-directed reaction catalysed by the yeast Rev1 protein, where C is favoured for incorporation opposite an abasic template site, and to a much lesser extent opposite a normal G site. PROOFREADING

Excision of a misincorporated nucleotide at a growing 3′primer end by a 3′ exonuclease associated with the polymerase. T–T 6–4 PHOTOPRODUCT

A form of damage occurring when DNA is exposed to UV radiation, in which a covalent bond is formed between the 6 and 4 pyrimidine ring positions, coupling adjacent thymines on the same DNA strand.

University of Southern California, Department of Biological Sciences and Chemistry, Stauffer Hall of Science 172, Los Angeles, California 90089-1340, USA. e-mail: mgoodman@rcf.usc.edu

In the beginning there was just one DNA polymerase — Escherichia coli DNA polymerase I (pol I), discovered by Arthur Kornberg and colleagues1,2 in 1956. Thirteen years later, Paula de Lucia and John Cairns, at Stony Brook, New York, isolated an E. coli mutant, polA (its designation being a play on de Lucia’s first name, as proposed to Cairns by Julian Gross) that seemed to have less than 1% of the normal pol I activity3. From this strain, a new DNA-polymerizing enzyme, pol II, was isolated4. The polA strain was much more sensitive to ultraviolet (UV) radiation than wild-type cells, suggesting that pol I might be involved in DNA repair in addition to chromosomal replication. Shortly after, using this same polA strain, Thomas Kornberg and Malcolm Gefter5, and Friedrich Bonhoeffer, Heinz Schaller and colleagues6, independently discovered DNA polymerase III (pol III). Isolation of a conditionally lethal temperature-sensitive pol III mutant7 showed that this enzyme is required for replicating the E. coli chromosome6. In contrast, pol II remained an enigma until last year, when it was shown8 to be pivotal in restarting replication in UV-irradiated cells. Last year also saw the identification of a new class of DNA polymerases — the UmuC/DinB/Rev1p/Rad30 superfamily (TABLE 1) — on the basis of five conserved sequence motifs present in all of these proteins (FIG. 1). The yeast Rev1 protein had been shown to contain DNAtemplate-dependent DCMP TRANSFERASE activity nearly three years earlier9, but it was not until 1999 that the other family members were isolated and shown to be capable of replicating DNA using all four bases. Biological functions

have been established for some members, including the E. coli UmuD′2C complex (now known as pol V), the yeast Rev1 protein and human DNA polymerase eta (pol η/Rad30). However, the functions of the remaining members of the UmuC/DinB/Rev1p/Rad30 polymerase superfamily are less certain. A feature common to many of these polymerases is their tendency to copy undamaged DNA with remarkably poor fidelity, whether or not they are involved in translesion synthesis. As its name suggests, translesion synthesis is the unimpaired copying of aberrant bases (see below) at which other cellular polymerases stall. With undamaged DNA, these low-fidelity polymerases incorporate an incorrect nucleotide once every 100–1,000 bases on average10–12 (TABLE 1). For comparison, normal polymerases that do not PROOFREAD misincorporate nucleotides in the range of once every 104–106 bases13. Examples of low-fidelity polymerases include E. coli pol V, which preferentially misincorporates G opposite a 3′ T of a T–T 6–4 PHOTOPRODUCT; E. coli DNA polymerase IV (pol IV/DinB), which adds a nucleotide onto the end of a misaligned primer; Rev1p, which incorporates C opposite a non-coding ABASIC LESION; and human DNA polymerase iota (pol ι/Rad30B), which favours misincorporation of G opposite T on undamaged DNA. All of these events lead to mutation. There is also the remarkable case of pol η, which copies pyrimidine T–T DIMERS accurately, resulting in mutation avoidance at this type of DNA damage (FIG. 2). The number of DNA polymerases has now grown from 3 to 5 in E. coli, and from 5 to at least 14 and count-

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 0 1

REVIEWS ing in eukaryotes (TABLE 1). Indeed, in a ‘back to the future’ moment during a recent conversation with Bob Lehman, Arthur Kornberg remarked, “In 1955, who would have imagined that there could be five DNA polymerases in E. coli?”. So what were the events that led to the discovery of these polymerases, and what do we now know of their biochemical functions and cellular properties? And why are there so many of them in eukaryotic cells? Whereas prokaryotic cells have just one choice — replicate damaged DNA or die — eukaryotic cells can, in principle, use programmed cell death (apoptosis) as an ‘escape hatch’ to avoid a potential catastrophe.

ABASIC LESION

A common form of DNA damage in which a base is lost from a strand of DNA, spontaneously or by the action of DNA repair enzymes such as apurinic endonucleases or uracil glycosylase, while leaving the phosphodiester bond intact. T–T DIMER

A form of damage occurring when DNA is exposed to UV radiation, in which two covalent bonds are formed between both the 5 and 6 positions of the pyrimidine ring on adjacent thymines located on the same DNA strand.

A growing family

Genetic studies in Saccharomyces cerevisiae and E. coli have been instrumental in defining groups of proteins required for mutagenesis. For example, yeast lacking the REV3 (REF. 14), REV7 (REFS 15,16) or REV1 (REF. 17) genes show significantly decreased spontaneous and UVinduced mutation rates. In E. coli, SOS mutagenesis (BOX 1) requires the umuC and umuD genes18. In 1968, Dean Rupp and Paul Howard-Flanders19 observed discontinuities (daughter-strand gaps) in DNA synthesized in an excision-defective strain of E. coli after UV irradiation. Because these strains cannot carry out NUCLEOTIDE-EXCISION REPAIR20, Rupp and HowardFlanders suggested that a single pyrimidine dimer is enough to kill the cell, presumably by blocking DNA replication. But although it may be advantageous to copy a variety of template lesions as an alternative to cell death, there is no such thing as a free lunch. The associated cost of survival is an increased number of mutations, targeted at the lesion sites. In E. coli, this is referred to as UV-induced SOS error-prone repair (BOX 1). Because both E. coli and yeast were known to have three DNA polymerases, there was no reason to suspect

NUCLEOTIDE-EXCISION REPAIR

The main pathway for removal of UV-damaged bases. REPLICATION FORK

Site in double-stranded DNA at which the template strands are separated, allowing a newly formed copy of the DNA to be synthesized, with the fork moving in the direction of leading strand synthesis. DISTRIBUTIVE POLYMERASE

A polymerase that dissociates from the primer–template DNA after incorporating one (or at most a few) nucleotides.

a E. coli pol IV/DinB

H2N

b H. sapiens pol κ/ HDINB1

H2N

c H. sapiens pol η

H2N

III

IV V COOH COOH COOH

d S. cerevisae Rev1p H2N BRCT e E. coli UmuC f H. sapiens pol ι

H2N H2N

COOH COOH COOH

Figure 1 | Representative members of the UmuC/DinB/Rev1p/Rad30 superfamily. Five highly conserved domains (indicated by roman numerals I–V) are believed to contain the nucleotide binding and catalytic residues. The subgroups within the family can be easily distinguished by the presence or absence of unique domains. a, b | The DinB subgroup contains a further three small domains near the carboxyl terminus of the protein (red boxes), whereas zinc finger motifs are uniquely found in b | pol κ/HDINB1 (C2HC type, yellow diamonds) and c | pol η (C2H2 type, dark grey diamond) that may be involved in DNA binding and selective targeting. d | Rev1p is the longest member of the family and contains two regions that are only conserved within the Rev1p subgroup (light grey boxes) as well as a BRCT (BRCA-1 carboxyterminal) domain believed to mediate protein–protein interactions for cell cycle checkpoints and DNA repair. e | UmuC and f | pol ι are both characterized by unique carboxy-terminal ends in which no known functional domains have been identified. These unique regions could possibly mediate protein interactions that stimulate and target UmuC or pol ι to their cellular destinations.

102

that there is a special class of polymerases to copy damaged DNA templates. Instead, proteins such as UmuC and UmuD′ (the mutagenically active form of UmuD; BOX 1) were thought to reduce the fidelity of E. coli pol III, enabling a blocked REPLICATION FORK to carry out errorprone translesion synthesis21. We now know that this is not the case — indeed, the cellular function of several new, errant DNA polymerases is translesion DNA synthesis22–25 (TABLE 1). Further studies using these ‘sloppier copier’ DNA polymerases are now revealing a rich biochemical tapestry. For example, one member of this family — the E. coli UmuD′2C polymerase (pol V)26–28 — does not act alone, but requires three further proteins to catalyse translesion synthesis12,28. Moreover, yeast Rev1p requires DNA polymerase zeta (pol ζ) to copy past abasic sites9 and T–T 6–4 photoproducts29. Other family members probably also use accessory proteins. The E. coli pol V mutasome

A good place to start any tour of the new DNA polymerases is with the E. coli pol V mutasome. The DNAdamage-inducible SOS response in E. coli was discovered more than 25 years ago (BOX 1). Many of the 30 or more SOS-regulated genes are involved in repairing DNA damage30,31, but two genes, umuC and umuD, are instead required for SOS-induced mutagenesis18,32–34. Although SOS mutation rates are typically 100-fold higher than spontaneous rates31, increased mutagenesis cannot occur unless UmuD is first converted (by cleavage) to the mutagenically active UmuD′ protein in a reaction that depends on another SOS-induced protein, RecA (REF. 35). The UmuC and UmuD′ proteins then interact to form a tight complex36,37, UmuD′2C (pol V)27,28, with intrinsic DNA polymerase activity. Working alone on an undamaged primed DNA template, pol V is a poor DISTRIBUTIVE POLYMERASE26,27. However, pol V cannot copy damaged DNA by itself — it requires RecA, single-stranded DNA binding protein (SSB) and β/γ complex12,27 (where β is the PROCESSIVITY CLAMP and γ the CLAMP-LOADER component of the replicative pol III holoenzyme). This multiprotein system, consisting of pol V, RecA, SSB and β/γ, is called the pol V “mutasome” (FIG. 3), a term coined by Harrison Echols38. The specific activity of pol V is amplified by an extraordinary 15,000-fold in the presence of RecA-coated template, allowing it to copy past damaged DNA bases12. Although the SOS system typically introduces mutations at sites of DNA damage, there is also an increase in untargeted mutations in the absence of damage39. All three common forms of DNA damage (FIG. 2) are copied efficiently by the pol V mutasome, but synthesis by either the pol III holoenzyme or pol IV (DinB) is blocked12. The specificity of incorporation by the pol V mutasome opposite the three forms of lesion mimics the in vivo mutational data12. For example, the 3′ T of a T–T 6–4 photoproduct is a T→C mutational ‘hotspot’ caused by the misincorporation of G opposite T (FIG. 2b) — precisely the reaction favoured by the pol V mutasome12. In contrast, pols III and IV preferentially incorporate A, which agrees with the ‘A-rule’40, but not with the in vivo data. What is the mechanism of translesion synthesis by www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS

Table 1 | The expanding polymerase universe DNA polymerase

Error rate

Properties

Possible function

E. coli pol V (umuDC) 12,27,28,99

10–2 –10–3

Translesion synthesis Low-fidelity synthesis

SOS lesion-targeted and untargeted mutagenesis

E. coli pol IV (dinB) 12,46 H. sapiens pol κ/θ (HDINBI) 49,50,100

10–3 –10–4

Mismatch extension

Untargeted and lesion-targeted mutagenesis Rescues stalled replication forks

H. sapiens pol η (XPV) 11,61,62

10–2 –10–3

Error-free synthesis of T–T UV photodimers

Prevents sunlight-induced skin cancer xeroderma pigmentosum

101–10–4

Low-fidelity synthesis

Somatic hypermutation

Incorporation of C opposite abasic sites

UV mutagenesis

Mismatch extension at lesions

UV mutagenesis

UmuC/DinB/Rev1p/Rad30 superfamily

S. cerevisiae pol η (RAD30) 10 H. sapiens pol ι (HRAD30B) 59,79 56,57

H. sapiens Rev1 (HREV1) S. cerevisiae Rev1 (REV1) 9,29 Family B

S. cerevisiae pol ζ (REV3/REV7) 58,59 H. sapiens pol ζ (HREV3/HREV7) 55,66,67,101

10–4 –10–5

Family X H. sapiens pol λ/pol β2 (POLL/POLβ2) 82,83 H.sapiens pol µ (POLM)

Meiosis-associated DNA repair

Somatic hypermutation

S. cerevisiae pol κ (TRF4) 84*

Sister-chromatid cohesion

Family A H. sapiens pol θ (POLQ) 85* D. melanogaster MUS308 86

Repair of DNA crosslinks

* The HDINB1 polymerase has also been designated pol κ/pol θ.

PROCESSIVITY CLAMP

A doughnut-shaped protein complex that threads the DNA through its hole while tethering the polymerase to DNA, typically increasing the processivity of the polymerase (the number of nucleotides incorporated into DNA per polymerase–template binding event). CLAMP LOADER

A protein complex that binds and then assembles the processivity clamp onto the DNA at a 3′-OH primer end, in a reaction requiring ATP.

pol V? The key to arriving at quantitative, kinetic-based conclusions for the effects of RecA, SSB and β/γ on pol Vcatalysed lesion bypass is to have the pol V mutasome bound in a confined region just before the lesion41. Two interactions occur: the first is between pol V and RecA; the second is between pol V and SSB (FIG. 3). Assembly of a RecA filament requires ATP binding (it proceeds 5′→3′ along a single-stranded DNA template in the presence of ATP or a poorly hydrolysable analogue, ATPγS). But disassembly of the filament in the same direction requires ATP hydrolysis42. So in the presence of ATPγS, RecA is bound stably to DNA as a helical filament. The other polymerases (pols II, III or IV) cannot copy DNA in the form of a RecA filament, even if the template is undamaged. Remarkably, however, pol V, along with SSB and β/γ, copies damaged and undamaged stabilized filaments with high processivity12,41, perhaps providing the key to unlock the lesion-copying mechanism. The RecA filament is 100 Å in diameter, whereas the β-clamp has an inner diameter of only 35 Å. But processive synthesis takes place on the filament. The obvious explanation is that pol V, acting in conjunction with SSB, strips RecA off the DNA in a 3′→5′ direction — a 100 Å RecA molecule cannot be threaded through the eye of a 35 Å β-dimer ‘needle’41. The stripping process is akin to the action of a locomotive ‘cowcatcher’ (a pointed device attached to the front of trains to push obstacles off the track). In this case, the RecA ‘cow’ is pushed off the DNA template ahead of the advancing pol V–SSB ‘locomotive’41. We have recently proposed that bidirectional disassembly of the RecA

filament, driven in the 3′→5′ direction by pol V–SSB and in the opposite direction by ATP hydrolysis, confines SOS mutations to the sites of DNA damage41, although SOS untargeted mutations do occur, albeit at a much lower frequency. After disassembly of the RecA filament and dissociation of pol V, the pol III holoenzyme presumably resumes replication on undamaged DNA downstream from the lesion43. The DinB subfamily

Escherichia coli pol IV (DinB) is believed to copy undamaged DNA at stalled replication forks44, which arise in vivo from mismatched or misaligned primer ends that are not proofread. A function for pol IV in alleviating stalling of the pol III holoenzyme is potentially significant, given the estimate that E. coli replication forks probably stall at least once during each replication cycle45. Overexpression of pol IV results in increased frameshift mutagenesis44, consistent with the ability of the enzyme to extend misaligned primer termini46 (FIG. 2). Whereas DinB homologues are among the most conserved members of the UmuC/DinB/Rev1p/Rad30 superfamily, almost nothing is known about what they do in other organisms. A second function for pol IV has been found, however, in adaptive mutation, a process in which non-proliferating microbial populations accumulate mutations when placed under non-lethal selective pressure47. In E. coli, pol IV is responsible for roughly half the lacZ adaptive frameshift mutations occurring on a plasmid in a wild-type background, and essentially all of the increased frameshifts in the absence of pol II48. So muta-

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 0 3

REVIEWS tor polymerases provide flexibility in dealing with environmental stress, particularly in prokaryotic organisms. By investigating competition for survival using E. coli strains containing combinations of single, double and triple pol II, pol IV and pol V mutants, it should be possible to determine the contribution of each polymerase to the relative fitness of the organism. Little is known about the in vivo function of human DINB (DNA polymerase kappa, pol κ). Purified pol κ, with a carboxy-terminal truncation (polκ∆C) that deletes two zinc clusters (FIG. 1) found only in the higher eukaryote homologues49, retains its polymerase activity. This implies that the carboxy-terminal region is dispensable for binding and catalysis, but that it mediates protein

GENETIC COMPLEMENTATION GROUPS

A distinct group of genes coding for separate polypeptides (proteins) required in the same biological pathway.

pol V G A T T

Error-prone TLS

O HN 3 4 5

pol η

C X

O O HN HN 3 4 5 O2 N 1 6 O N

Figure 1

d 5'

G T

T–T (6-4) photoproduct

Rev1p

Error-prone synthesis

4 3 5 2 61

TT Error-prone dCMP incorporation

A A

Error-free TLS

2 1 6

T–T cis-syn photodimer

pol ι 5'

O-CH2 O

e Misaligned primer–template extension

pol IV A A A C T T T G

Abasic site

pol κ (HDINB1)

Error-prone TLS by primer–template misalignment

T G A C A C T G

Figure 2 | Biochemical properties of new polymerases. a | Error-prone translesion synthesis (TLS) by E. coli pol V results in misincorporation of G opposite the 3′ T of a T–T 6–4 photoproduct. b | DNA polymerase η incorporates two A bases as it replicates across a T–T cis–syn photodimer, thereby avoiding mutation. c | The DNA-dependent dCMP transferase activity of Rev1p incorporates C opposite an abasic site. d | Misincorporation of G opposite T on an undamaged template is carried out by pol ι, in preference to correct incorporation of A, resulting in a high incidence of A to G transitions. e | Misaligned primer–template ends are extended efficiently by E. coli pol IV, leading to frameshift errors. Extension of a mismatched primer end (not shown) would lead to a base-substitution mutation. f | Bypass of an abasic site by pol κ (HDINB1) results in a –1 frameshift mutation when the lesion is followed by T.

pol V

SSB RecA*

Figure 3 | The pol V mutasome. The pol V mutasome consists of pol V (UmuD′2C), activated RecA (RecA*), β sliding clamp, γ clamp loading complex and the single-stranded DNA binding protein (SSB). Pol V associates at a 3′-primer end (vacated by the pol III core), while establishing direct contact with SSB, and the 3′ tip of a RecA filament.

104

interactions that target or regulate the enzyme50. In vitro, polκ∆C can bypass an abasic site by preferential insertion of A opposite the lesion, creating a –1 frameshift mutation by a template loop-out mechanism. This occurs when the abasic site is followed by a T in the template; if the abasic site is followed by a template A, only a simple base substitution is observed (FIG. 2). Polκ∆C can also synthesize past an N-2-acetylaminofluorine (AAF)-modified G in an error-prone manner by preferential incorporation of T, without generating a frameshift mutation50. Rev1p C transferase

When confronted with a missing DNA template base — that is, an abasic site (FIG. 2) — most DNA polymerases favour an ‘A-rule’ default mechanism in which A is strongly preferred (about tenfold) for incorporation opposite the non-templating lesion40,51–53. However, when Christopher Lawrence and colleagues54 used plasmid DNA containing a site-directed abasic moiety to infect yeast cells, they observed preferential incorporation of C, not A (FIG. 2). This unexpected effect depends on the Rev1 protein, which is required for UV mutagenesis17. Mutations in REV1, REV3 or REV7 eliminate more than 95% of base-substitution mutations in yeast55,56. The human homologue of yeast Rev1p has since been found to be required for UV-induced mutagenesis56, and it also behaves as a template-dependent dCMP transferase57. There is very little bypass of an abasic site in vivo in the absence of Rev1p, and what little bypass does occur obeys the A-rule29. Rev1p also shows weak incorporation of C opposite G, at about a tenfold lower rate compared with incorporation opposite an abasic site. Because it incorporates only C, Rev1p is perhaps better characterized as a template-based dCMP transferase rather than a bona fide DNA polymerase9. But Rev1p does not act alone in catalysing translesion replication — for this it requires pol ζ (the Rev3 and Rev7 proteins)58 (FIG. 4). Pol ζ has the remarkable property of adding correct nucleotides onto mismatched 3′-primer ends with exceptionally high efficiencies, only 10–100fold less than observed for correct primer extension59. So it is likely that pol ζ takes over from Rev1p, which incorporates C opposite an abasic lesion but cannot go further58. Lawrence and co-workers have also reported29 that Rev1p is needed to copy past pyrimidine 6–4 photoproducts but, in contrast to bypass of abasic sites, C is not incorporated. Rev1p therefore seems to have two distinct functions in copying DNA damage. One requires its C transferase activity (FIG. 4a), whereas the other facilitates translesion synthesis by another polymerase, most probably pol ζ (FIG. 4b). However, a direct interaction between Rev1p and pol ζ has not been reported. DNA polymerase η

DNA pol η, a human homologue of the yeast Rad30 protein60, was identified as the product of the XPV gene61,62 last year. Xeroderma pigmentosum (XP) is characterized by mutations in eight GENETIC COMPLEMENTATION GROUPS, seven of which code for enzymes involved in nucleotide-excision repair31. The eighth is the XPV gene. Although XPV cells can carry out nucleotide-excision www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS

V(D)J RECOMBINATION

The site-specific recombination of immunoglobulin coding regions from multiple copies in the germ line to just one variable (V), one diversity (D) and one joining (J) region in the process of forming a functional immunoglobulin gene in B cells.

repair, they are deficient in copying UV-damaged DNA63. People who carry defects in XP genes show increased susceptibility to sunlight-induced skin cancer. The prevalent form of UV damage to DNA is the T–T cis–syn photodimer (FIG. 2). T–T photodimers block replication by various polymerases in vitro, but they do not significantly impede human or yeast pol η, both of which copy these photodimers by correctly incorporating two A bases opposite each T site64,65 (FIG. 2). This is consistent with a cellular role for pol η in the error-free replication of UV photodimers: error-free synthesis

Box 1 | SOS mutagenesis SOS off Escherichia coli keeps an LexA (free) arsenal of regulated pathways that help it to survive when under stress31. LexA One of these is the ‘SOS’ regulon, which is thought to (bound) x x be induced in response to umuD umuC regions of single-stranded LexA DNA — presumably a (free) hallmark of large-scale DNA damage. Normally, the LexA repressor binds to the LexA (bound) operators of more than 30 xumuD x umuC SOS genes and keeps them repressed. But in the RecA* presence of single-stranded DNA, the RecA protein LexA forms a close-packed (inactivated) ‘activated’ RecA filament, RecA*, which acts as a coprotease to cleave any LexA SOS on UmuD protein released from low-affinity operators. Further cleavage of LexA frees up the more weakly bound operators, umuD umuC RecA* and the SOS genes are relieved from repression (see figure). The SOS proteins are UmuC mainly involved in UmuD' protein (mutationally active) nucleotide-excision and recombination-repair pathways to remove the DNA damage. However, the two ‘UV mutagenesis’ UmuD'2C/ pol V (umu) genes, umuC and umuD, are instead required for replication past unrepaired lesions in the DNA template. They leave behind mutations targeted to sites of DNA damage. To be active, UmuD must be post-translationally cleaved to UmuD′ on the RecA* filament31,35,97,98 (see figure). UmuC and UmuD′ then form a tight complex, UmuD′2C, which has an intrinsic, low-fidelity DNA polymerase activity43. A replication fork blocked by DNA damage is dealt with by two SOS-induced DNA polymerases — pol II and pol V (UmuD′2C). About two minutes after SOS induction, pol II reinitiates replication downstream from the lesion, leaving a gapped structure that is resolved by homologous recombination43. Replication restart is an error-free repair process. Pol V appears 30–45 minutes later. It binds at the 3′-OH adjacent to the lesion, then copies past the lesion, often inserting the wrong base opposite it. This process also requires RecA*, single-stranded DNA binding protein (SSB), and the β/γ processivity proteins (FIG. 3).

when copying T–T photodimers does not translate into error-free synthesis on undamaged DNA. Indeed, error rates for pol η on natural DNA templates can be as high as about 5% for T→C mutations (T•dGMP mispairs)11, with most base-substitution errors in a range of around 0.5–1% (REFS 10,11). In comparison, most non-proofreading cellular polymerases (which do not have an associated 3′→5′ exonuclease activity for editing out misinserted nucleotides13) have error rates of about 10–3–10–5. So relaxed active-site specificity, enabling pol η to copy ‘blocking’ T–T dimers accurately, is probably responsible for its low fidelity on undamaged DNA. In XPV cells, error-prone replication of T–T photodimers by some other polymerase could potentially cause an increase in mutations. One candidate is a human homologue of the yeast pol ζ. The yeast polymerase is composed of a complex of the Rev3 and Rev7 proteins, and, as already discussed, is required in an errorprone translesion-synthesis pathway58. Indeed, human REV3 (REF. 66) and REV7 (Ref. 67) homologues have been identified, but even if human pol ζ is not responsible for error-prone replication in XPV cells, there are plenty of other candidates to choose from (TABLE 1). Candidates for somatic hypermutation

The kind of mutagenesis discussed so far is not the only process in which errors can be introduced into DNA. Take somatic hypermutation, for example, which is one of the processes responsible for generating the roughly one billion antibody variants in humans68. An initial repertoire of antibodies results from non-random V(D)J RECOMBINATION. After exposure to an antigen, activation of B cells expressing the correct antibody starts a second phase of diversity, termed ‘affinity maturation’, caused by somatic hypermutation in rapidly dividing GERMINAL 69 CENTRE cells . These mutations occur exclusively in the variable region of the immunoglobulin gene; they begin proximal to the promoter and diminish about 1–2 kilobases downstream70. The base-substitution error rate of around 3×10–4 per base pair per generation is about 106-fold above spontaneous background levels68. In other words, somatic hypermutation is exquisitely targeted, and is unaccompanied by a global alteration in the fidelity of B-cell replication. Two cis-acting transcriptional enhancers located downstream of the variable region in light and heavy chains regulate somatic hypermutation71 (FIG. 5). The intronic enhancer (Ei) and flanking matrix attachment region (MAR) of the κ light chain are both essential, eliminating somatic hypermutation completely when deleted72. Another κ light chain 3′ enhancer (E3′) affects mutations to a lesser extent73. A promoter sequence upstream of the immunoglobulin gene variable region is also essential, but any promoter can be used, and any DNA inserted into the variable region can act as the mutational target. Before the discovery of error-prone polymerases, several models for somatic hypermutation were proposed. One suggests that hypermutation rates could arise from repetitive application of transcription-coupled repair, by which stalled replication forks in the

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 0 5

REVIEWS

GERMINAL CENTRE

A highly organized structure that develops around follicles in peripheral lymphoid organs, such as the spleen and lymph nodes, in which B cells undergo rapid proliferation and selection on formation of antigen–antibody complexes during the immune response.

variable region recruit transcription-coupled repair proteins74. Repeated replication of the ‘repair’ region could increase the chance of mutation. But transcription-coupled repair shows a large bias for the actively transcribed strand, a feature that is uncharacteristic of somatic hypermutation75. Moreover, a huge amount of repetitive replication would be necessary to reach mutation rates of about 3×10–4 per base pair. In another model, reverse transcriptase is thought to synthesize a complementary DNA copy from an elongating messenger RNA that could replace the gene by homologous recombination76. However, this would be a convoluted way to obtain chromosomal mutations. Two other models invoke a collision between stalled transcription forks and moving replication forks, causing either a reduction in fidelity of a normal polymerase77, or a signal from a stalled fork for an errorprone polymerase to take over78. The idea that somatic hypermutation is caused by error-prone DNA polymerases has been given fresh impetus by the discovery of the errant polymerases (although the models invoking transcription-coupled repair, reverse transcripts, and collisions between transcription and replication machinery remain alive, albeit tenuously). At a meeting of The Royal Society on ‘Hypermutation in antibody genes’ (5–6 July 2000), which devoted one of its four sessions to the new polymerases, two favoured candidate polymerases emerged — pol ι (Rad30B) and DNA polymerase mu (pol µ) (TABLE 1). Interest in human pol ι stems from its preference for incorporating G opposite a template T, making a G•T wobble base pair with a 3:1 preference over a Watson–Crick A•T pair79 (FIG. 2). A T•T mispair is also easily formed, about 70% as efficiently as A•T (REF. 79). In a

Rev1p 5'

TRANSITION

A point mutation in which a purine base (A or G) is substituted for a different purine base, and a pyrimidine base (C or T) is substituted for a different pyrimidine base, for example, an A•T→G•C transition. TRANSVERSION

A point mutation in which a purine base is substituted for a pyrimidine base and vice versa, for example, an A•T→C•G transversion. TERMINAL TRANSFERASE

An enzyme found primarily in the thymus gland that incorporates nucleotides randomly onto the 3′ end of single-stranded DNA (a nontemplated reaction), in contrast to a polymerase, which incorporates nucleotides onto a 3′-primer-end in a doublestranded, template-directed reaction.

106

C 5' pol ζ

Rev1p

CA XT

pol ζ 5'

Rev1p

AA TT

another experiment59, incorporation of both G and T opposite a template T is favoured by about tenfold and fivefold, respectively, relative to incorporation of A opposite T. Most of the other mispairs occur in the range of 10–2–10–3. The pol ι misincorporation preferences seem consistent with immunoglobulin mutational spectra in which TRANSITIONS are favoured 2:1 over TRANSVERSIONS, and A mutates more often than T (REF. 80). Human pol µ is most closely related (41% aminoacid identity) to terminal deoxynucleotidyltransferase (TdT)81, a template-independent DNA-synthesizing enzyme. Some similiarity (23% identity) is also observed between pol µ and polymerase beta (pol β). Pol µ has weak intrinsic TERMINAL TRANSFERASE activity, and can also act as a DNA-dependent polymerase that shows poor base selection when manganese replaces magnesium as a cofactor in replication reactions in vitro81. This polymerase is expressed preferentially in peripheral lymphoid tissues and, based on analysis of the expressed sequence tag database, could be overrepresented in human B-cell germinal centres, which are critical for maturation of the immune response81. Should experiments using knockout mouse strains reveal a requirement for one (or perhaps several) of the errant polymerases in somatic hypermutation, it will be just the beginning of the story. Biochemical reconstitution of somatic hypermutation in vitro is likely to be a challenge. Any model for somatic hypermutation will have to account for the localization, polarity, magnitude and specificity of the point mutations — a tall order. Although it is premature to speculate on specific mechanisms of somatic hypermutation in vitro, at least two models can be envisaged. In the first, interactions between transcription factors, somatic hypermutationspecific enhancer elements and co-activator proteins result in formation of a DNA secondary structure that recruits a mutator polymerase (FIG. 6a). In the second model, the errant polymerase is recruited to the site of a DNA nick, short gap, or perhaps even to a double-strand break (FIG. 6b). In both models, DNA synthesis by the mutator enzyme across a short gap — analogous to baseexcision repair by pol β — generates mutations targeted to the variable-region gene. Moreover, nucleotide misincorporation and mismatch extension might require the action of separate polymerases. For example, pol ι often makes misincorporation errors, but seems to have difficulty in extending a mismatched primer end59. On the other hand, pol ζ synthesizes DNA with essentially nor-

Figure 4 | Rev1p/pol ζ lesion bypass. a | Rev1p dCMP transferase activity incorporates C opposite a noninstructional abasic site in a DNA template in the absence of pol ζ, but it cannot extend the primer beyond the mismatch. Pol ζ can then take over for Rev1p and efficiently extend the mismatched primer terminus. However, it is not known whether Rev1p stays associated with the DNA or directly interacts with pol ζ during this process. b | Pol ζ can efficiently incorporate two A bases opposite a T–T 6–4 photoproduct in vitro, resulting in error-free bypass of the template lesion, but it can only do so in the presence of Rev1p, although the C transferase activity of Rev1p is not involved.

MAR E i

Cκ

E3'

Figure 5 | Genetic elements required for somatic hypermutation in the kappa light-chain immunoglobulin gene. Both a promoter (P) and leader (L) sequence are required, but may be replaced with non-immunoglobulin counterparts from other genes. The intronic enhancer element (Ei) and associated nuclear matrix attachment region (MAR), as well as the 3′ enhancer (E3′), must be present for hypermutation in the variable region (VJ). This region is flanked by the upstream promoter and downstream MAR/Ei sequences. The constant domain of the kappa light chain (Cκ) is not a target for somatic hypermutation.

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS that cells lacking TRF function do not completely replicate the genome. In vitro, Trf4 has a DNA polymerase activity with an elevated Km (the substrate concentration that allows the reaction to proceed at half its maximum rate) for nucleotides, designated pol κ (this name has been applied to two other DNA polymerases; TABLE 1). One suggested function for Trf4 DNA polymerase is to facilitate progression of the replication fork through chromatid cohesion sites that might inhibit other DNA polymerases, such as pol δ or pol ε, causing the replication fork to collapse84.

a Ei

MAR

TAF L

pol pol

b pol P

VJ pol

MAR

Why so many polymerases? Figure 6 | Models for somatic hypermutation by an error-prone polymerase. a | Enhancer/primer-mediated targeting of somatic hypermutation. Strong interactions between enhancer-binding proteins (shown in white) and transcription-associated factors (TAFs) at the promoter (P) mediate the formation of a unique open DNA complex. Error-prone polymerases (pol), possibly pol ι or pol µ, may preferentially substitute for the normal replicative polymerases in this region of the DNA, resulting in mutations. Mutation is shown on both strands as there is no evidence for a strand bias. b | Nick- or gap-dependent targeting of somatic hypermutation. The variable region of the immunoglobulin genes may contain small nicks or gaps as a by-product of V(D)J recombination or very active transcription. The DNA breaks may be substrates for an error-prone polymerase to bind and generate mutations. (MAR, matrix attachment region; Ei, enhancer element.)

mal fidelity (10–4–10–5) but seems very efficient at mismatch extension (10–1–10–2). So the sequential action of both polymerases may be necessary for translesion synthesis in eukaryotic cells59. Other new polymerases

Pol λ/β2. Pol µ is not the only new polymerase related to pol β. Another DNA polymerase (pol λ) was identified in mouse82, and later in humans (POL β2)83, that shares a 32% amino-acid identity with pol β and contains the conserved family X (TABLE 1) residues critical for DNA and nucleotide binding, as well as catalysis82. Pol λ/β2 is expressed to significant levels only in the testes and ovaries, indicating that this enzyme may be involved in meiotic cell division82, but this remains to be shown. Purified pol λ/β2 has polymerase activity, although nothing is known about its fidelity or preferred DNA substrates82,83. A BRCA1-containing carboxy-terminal (BRCT) domain located in the aminoterminal region can also be deleted from human pol λ/β2 without significant reduction in polymerase activity83. The in vivo function of BRCT domains in both pol λ/β2 and Rev1p pathways is an area that awaits further investigation. Pol κ (Trf4). In S. cerevisiae, another β-like polymerase has been identified as the product of the TRF4 gene. With its close homologue Trf5, theTrf4 protein is involved in maintaining sister-chromatid cohesion during S-phase replication. Fluorescence in situ hybridization (FISH) in trf4 mutant cells84 has revealed a marked increase in nuclei that fail to maintain cohesion of sister chromatids near centromeres and on chromosome arms. Moreover, trf4ts/trf5 double mutant cells show delays in the G1 to S-phase transition and contain levels of DNA between those found in G1 and G2. These results imply

There are well established biological roles for pol V (UmuD′2C), pol η (Rad30) and Rev1p. Although less certain, pol IV (DinB) is probably required to rescue stalled replication complexes. What about the other new polymerases? The DNA polymerase theta (pol θ/POLQ)85 and MUS308 (REF. 86) proteins, both family A polymerases (TABLE 1), may be involved in repairing DNA crosslinks, but functions for pols ι (Rad 30B), µ, λ and κ (HDINB1) remain speculative (TABLE 1). It is reasonable to conclude that pol V and pol η usually copy damaged DNA templates, with pol V required for errorprone and pol η for error-free repair. Because pol V causes SOS untargeted mutations in the absence of DNA damage in E. coli 31, it is probably also important in natural selection and evolution87. In eukaryotes, the new polymerases might fill in short gaps emanating from non-homologous end-joining and from homologous recombination. Indeed, a marked increase in sister-chromatid exchange in transformed XPV cells led to the discovery88, earlier this year, of a relationship between the S-phase checkpoint of the cell cycle and an X-ray-induced recombination pathway for repairing double-strand breaks88. Even E. coli seems to have a DNA-damage checkpoint in which UmuC and uncleaved UmuD coordinate progression through the cell cycle, signalling when it is safe to switch from stationary phase to exponential growth89,90. Recently, two groups have independently demonstrated the embryonic lethality of disrupting the mouse homologue of the REV3 gene91,92, the presumed catalytic subunit of mouse pol ζ. These studies emphasize the potential importance of specialized DNA polymerases in development and raise even more interesting questions regarding the extent of lesions that may occur during rapid cell proliferation, or rather if pol ζ might be closely linked to the mitotic checkpoints in the absence of DNA damage91. Macromolecular traffic control

The fact that DNA polymerases have become a ‘growth industry’ in the cell raises concerns about traffic control. To copy damaged DNA, rescue blocked replication forks, catalyse somatic hypermutation or fill in gaps during homologous and non-homologous recombination, the enzymes have to show up where and when they are needed and then depart when finished — and not a moment later. The basic idea, and it is not a new one, is that DNA repair proteins might always be

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 0 7

REVIEWS

PRIMOSOME

A complex of proteins whose role is to initiate DNA synthesis by the de novo synthesis of an oligonucleotide RNA primer on a DNA template strand; a primosome is typically used to initiate synthesis at a replication origin or to re-initiate synthesis downstream of a stalled replication fork.

present at the replication fork, perhaps bound to the replication complex. A multiprotein complex, composed of two interconnected polymerase holoenzymes for coordinated leading- and lagging-strand synthesis, a lagging-strand PRIMOSOME, DNA helicase and SSB protein, would have the added baggage of other specialized polymerases, to be used sparingly when called for. This picture is easier to imagine if the replication complex is stationary, with the DNA moving through the ‘replication factory’93. In contrast, however, the textbook version of events is that polymerase bound to accessory proteins traverses a DNA track. Nevertheless, an in situ assay using a Bacillus subtilis polymerase (PolC) tagged with green fluorescent protein identified the enzyme at discrete intracellular loci94, indicating that the DNA may be moving through an anchored DNA polymerase. Interactions between proteins of the replication complex and a superfamily polymerase have indeed been found in E. coli. Last year, Graham Walker and co-workers95 reported differential binding between components of pol V (UmuD/UmuD′) and the α-, β- and ε-subunits of pol III. Further evidence27 comes from the stabilization of a thermolabile pol III α-subunit, at non-permissive temperature, in the presence of pol V. These data hint that there could be a coordinated exchange between high- and low-fidelity polymerases, acting as partners in a macromolecular complex at sites of DNA damage. However, the challenge, as Arthur Kornberg has often cautioned, is the need “to capture it alive” — that is, to reassemble an intact ‘replisome–mutasome’

Kornberg, A., Lehman, I. R., Bessman, M. J. & Simms, E. S. Enzymatic synthesis of deoxyribonucleic acid. Biochim. Biophys. Acta 21, 197–198 (1956). 2. Lehman, I. R., Bessman, M. J., Simms, E. S. & Kornberg, A. Enzymatic synthesis of deoxyribonucleic acid I. Preparation of substrates and partial purification of an enzyme from Escherichia coli. J. Biol. Chem. 233, 163–170 (1958). 3. de Lucia, P. & Cairns, J. Isolation of an E. coli strain with a mutation affecting DNA polymerase. Nature 224, 1164–1166 (1969). 4. Knippers, R. DNA polymerase II. Nature 228, 1050–1053 (1970). 5. Kornberg, T. & Gefter, M. L. Purification and DNA synthesis in cell-free extracts: properties of DNA polymerase II. Proc. Natl Acad. Sci. USA 68, 761–764 (1971). 6. Nusslein, V., Otto, B., Bonhoeffer, F. & Schaller, H. Function of DNA polymerase 3 in DNA replication. Nature New Biol. 234, 285–286 (1971). 7. Gefter, M. L., Hirota, Y., Kornberg, T., Wechsler, J. A. & Barnoux, C. Analysis of DNA polymerases II and III in mutants of Escherichia coli thermosensitive for DNA synthesis. Proc. Natl Acad. Sci. USA 68, 3150–3153 (1971). 8. Rangarajan, S., Woodgate, R. & Goodman, M. F. A phenotype for enigmatic DNA polymerase II: a pivotal role for pol II in replication restart in UV-irradiated Escherichia coli. Proc. Natl Acad. Sci. USA 96, 9224–9229 (1999). 9. Nelson, J. R., Lawrence, C. W. & Hinkle, D. C. Deoxycytidyl transferase activity of yeast REV1 protein. Nature 382, 729–731 (1996). Shows that yeast Rev1 protein has templatedependent dCMP transferase activity, and is the first member of the UmuC/DinB /Rev1p/Rad30 superfamily found to be involved in DNA synthesis. 10. Washington, M. T., Johnson, R. E., Prakash, S. & Prakash, L. Fidelity and processivity of Saccharomyces cerevisiae DNA polymerase η. J. Biol. Chem. 274, 36835–36838 (1999). Yeast Rad30 protein is shown to have low-fidelity DNA polymerase activity and is designated as pol η. 11. Matsuda, T., Bebenek, K., Masutani, C., Hanaoka, F. &

108

macromolecular complex from individually purified protein components. The technology is available — in situ immunofluorescence, immunoprecipitation and multiple-hybrid screening can all be used to identify interactions between the new polymerases and other proteins. Even if such a fishing expedition bears fruit, to mix a metaphor, it still will not be easy to reconstitute a replisome–mutasome macromolecular complex in vitro. But there are successful precedents. For example, prokaryotic and eukaryotic replication, repair and recombination complexes have been built from the ground up with purified polymerases, primases, processivity factors, DNA-binding proteins, mismatch-binding proteins, recombinases, helicases and ligases. Indeed, this tried and tested approach is the first of Arthur Kornberg’s96 Ten commandments: Lessons from the enzymology of DNA replication — “rely on enzymology to clarify biological questions”. Update — added in proof

M. Goldsmith et al.102 have shown that the MucB protein, a plasmid-encoded homologue of the E. coli. UmuC protein, is a DNA polymerase capable of translesion synthesis past an abasic site in the presence of MucA′ (the UmuD′ homologue), RecA and SSB protein. Links DATABASE LINKS REV3 | REV7 | REV1 | umuC

Kunkel, T. A. Low fidelity DNA synthesis by human DNA polymerase η. Nature 404, 1011–1013 (2000). 12. Tang, M. et al. Roles of E. coli DNA polymerases IV and V in lesion-targeted and untargeted mutagenesis. Nature 404, 1014–1018 (2000). The E. coli DNA pol V mutasome is shown to be responsible for SOS-induced mutagenesis through error-prone translesion synthesis. 13. Echols, H. & Goodman, M. F. Fidelity mechanisms in DNA replication. Annu. Rev. Biochem. 60, 477–511 (1991). 14. Lawrence, C. W. & Christensen, R. B. Ultraviolet-induced reversion of CYC1 alleles in radiation-sensitive strains of yeast. III. REV3 mutant strains. Genetics 92, 397–408 (1979). 15. Lawrence, C. W., Das, G. & Christensen, R. B. REV7, a new gene concerned with UV mutagenesis in yeast. Mol. Gen. Genet. 200, 80–85 (1985). 16. Lawrence, C. W., Nisson, P. E. & Christensen, R. B. UV and chemical mutagenesis in REV7 mutants of yeast. Mol. Gen. Genet. 200, 86–91 (1985). 17. Lawrence, C. W. & Christensen, R. Ultraviolet-induced reversion of cyc1 alleles in radiation sensitive strains of yeast. I. rev1 mutant strains. J. Mol. Biol. 122, 1–22 (1978). 18. Sommer, S., Knezevic, J., Bailone, A. & Devoret, R. Induction of only one SOS operon, umuDC, is required for SOS mutagenesis in E. coli. Mol. Gen. Genet. 239, 137–144 (1993). 19. Rupp, W. D. & Howard-Flanders, P. Discontinuities in the DNA synthesized in an excision-defective strain of Escherichia coli following ultraviolet radiation. J. Mol. Biol. 31, 291–304 (1968). 20. Sancar, A. & Sancar, G. B. DNA repair enzymes. Annu. Rev. Biochem. 57, 29–67 (1988). 21. Bridges, B. A. & Woodgate, R. The two-step model of bacterial UV mutagenesis. Mutat. Res. 150, 133–139 (1985). 22. Walker, G. Skiing the black diamond slope: Progress on the biochemistry of translesion DNA synthesis. Proc. Natl Acad. Sci. USA 95, 10348–10350 (1998). 23. Woodgate, R. A plethora of lesion-replicating DNA polymerases. Genes Dev. 13, 2191–2195 (1999). 24. Goodman, M. F. & Tippin, B. Sloppier copier DNA

polymerases involved in genome repair. Curr. Opin. Genet. Dev. 10, 162–168 (2000). 25. Friedberg, E., Feaver, W. J. & Gerlach, V. L. The many faces of DNA polymerases: Strategies for mutagenesis and for mutational avoidance. Proc. Natl Acad. Sci. USA 97, 5681–5683 (2000). 26. Tang, M. et al. Biochemical basis of SOS mutagenesis in Escherichia coli: Reconstitution of in vitro lesion bypass dependent on the UmuD′2C mutagenic complex and RecA protein. Proc. Natl Acad. Sci. USA 95, 9755–9760 (1998). The first evidence for the intrinsic polymerase activity in UmuD′2C (now pol V). 27. Tang, M. et al. UmuD′2C is an error-prone DNA polymerase, Escherichia coli pol V. Proc. Natl Acad. Sci. USA 96, 8919–8924 (1999). Purified UmuD′2C from a temperature-sensitive DNA polymerase III mutant possesses intrinsic polymerase activity and can do translesion synthesis, demonstrating that UmuD′2C is a bona fide DNA polymerase. 28. Reuven, N. B., Arad, G., Maor-Shoshani, A. & Livneh, Z. The mutagenic protein UmuC is a DNA polymerase activated by UmuD′, RecA, and SSB and is specialized for translesion replication. J. Biol. Chem. 274, 31763–31766 (1999). The E. coli UmuC protein was shown to have DNA polymerase activity. 29. Nelson, J. R., Gibbs, P. E. M., Nowicka, A. M., Hinkle, D. C. & Lawrence, C. W. Evidence for a second function for Saccharomyces cerevisiae Rev1p. Mol. Microbiol. 37, 549–553 (2000). 30. Walker, G. C. Inducible DNA repair systems. Annu. Rev. Biochem. 54, 425–457 (1985). 31. Friedberg, E. C., Walker, G. C. & Siede, W. in DNA Repair and Mutagenesis Vol. 1, 407–522 (ASM Press, Washington DC, 1995). 32. Kato, T. & Shinoura, Y. Isolation and characterization of mutants of Escherichia coli deficient in induction of mutagenesis by ultraviolet light. Mol. Gen. Genet. 156, 121–131 (1977). 33. Steinborn, G. Uvm mutants of Escherichia coli K12 deficient in UV mutagenesis. I. Isolation of uvm mutants

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS

and their phenotypical characterization in DNA repair and mutagenesis. Mol. Gen. Genet. 165, 87–93 (1978). 34. Bagg, A., Kenyon, C. J. & Walker, G. C. Inducibility of a gene product required for UV and chemical mutagenesis in Escherichia coli. Proc. Natl Acad. Sci. USA 78, 5749–5753 (1981). 35. Burckhardt, S. E., Woodgate, R., Scheuremann, R. H. & Echols, H. UmuD mutagenesis protein of Escherichia coli: Overproduction, purification, and cleavage by RecA. Proc. Natl Acad. Sci. USA 85, 1811–1815 (1988). 36. Woodgate, R., Rajagopalan, M., Lu, C. & Echols, H. UmuC mutagenesis protein of Escherichia coli: Purification and interaction with UmuD and UmuD′. Proc. Natl Acad. Sci. USA 86, 7301–7305 (1989). 37. Bruck, I., Woodgate, R., McEntee, K. & Goodman, M. F. Purification of a soluble UmuD’C complex from Escherichia coli: Cooperative binding of UmuD’C to single-stranded DNA. J. Biol. Chem. 271, 10767–10774 (1996). 38. Echols, H. & Goodman, M. F. Mutation induced by DNA damage: A many protein affair. Mutat. Res. 236, 301–311 (1990). 39. Witkin, E. M. Ultraviolet mutagenesis and inducible DNA repair in Escherichia coli. Bacteriol. Rev. 40, 869–907 (1976). 40. Strauss, B. S. The ‘A rule’ of mutagen specificity: a consequence of DNA polymerase bypass of noninstructional lesions? Bioessays 13, 79–84 (1991). 41. Pham, P., O’Donnell, M., Woodgate, R. & Goodman, M. F. A ‘cowcatcher’ model for SOS–lesion targeted mutations in E. coli involving pol V, RecA, SSB and β sliding clamp. (submitted). 42. Kuzminov, A. Recombinational repair of DNA damage in Escherichia coli and bacteriophage λ. Microbiol. Mol. Biol. Rev. 63, 751–813 (1999). 43. Goodman, M. F. Coping with replication ‘train wrecks’ in Escherichia coli using pol V, pol II, and RecA proteins. Trends Biochem. Sci. 25, 189–195 (2000). 44. Kim, S.-R. et al. Multiple pathways for SOS-induced mutagenesis in Escherichia coli: An overexpression of dinB/dinP results in strongly enhancing mutagenesis in the absence of any exogenous treatment to damage DNA. Proc. Natl Acad. Sci. USA 94, 13792–13797 (1997). 45. Cox, M. M. et al. The importance of repairing stalled replication forks. Nature 404, 37–41 (2000). 46. Wagner, J. et al. The dinB gene encodes a novel Escherichia coli DNA polymerase (DNA pol IV). Mol. Cell 4, 281–286 (1999). E. coli DinB/pol IV protein is shown to have distributive polymerase activity that can also extend misaligned 3′-primer ends. 47. Foster, P. L. Adaptive mutation: The uses of adversity. Annu. Rev. Microbiol. 47, 467–504 (1993). 48. Foster, P. L. Adaptive mutation in Escherichia coli. Cold Spring Harb. Symp. Quant. Biol. (in the press). 49. Gerlach, V. L. et al. Human and mouse homologs of Escherichia coli DinB (DNA polymerase IV), members of the UmuC/DinB superfamily. Proc. Natl Acad. Sci. USA 96, 11922–11927 (1999). 50. Ohashi, E. et al. Error-prone bypass of certain DNA lesions by the human DNA polymerase κ. Genes Dev. 14, 1589–1594 (2000). 51. Boiteux, S. & Laval, J. Coding properties of poly(deoxycycytidylic acid) templates containing uracil or apyrimidinic sites: In vitro modulation of mutagenesis by DNA repair enzymes. Biochemistry 21, 6746–6751 (1982). 52. Schaaper, R. M., Kunkel, T. A. & Loeb, L. A. Infidelity of DNA synthesis associated with bypass of apurinic sites. Proc. Natl Acad. Sci. USA 80, 487–491 (1983). 53. Sagher, D. & Strauss, B. S. Insertion of nucleotides opposite apurinic/apyrimidinic sites in deoxyribonucleic acid during in vitro synthesis: uniqueness of adenine nucleotides. Biochemistry 22, 4518–4526 (1983). 54. Lawrence, C. W., Borden, A., Banerjee, S. K. & LeClerc, J. E. Mutation frequency and spectrum resulting from a single abasic site in a single-stranded vector. Nucleic Acids Res. 18, 2153–2157 (1990). 55. Gibbs, P. E., McGregor, W. G., Maher, V. M., Nisson, P. & Lawrence, C. W. A human homolog of the Saccharomyces cerevisiae REV3 gene, which encodes the catalytic subunit of DNA polymerase ζ. Proc. Natl Acad. Sci. USA 95, 6876–6880 (1998). 56. Gibbs, P. E. M. et al. The function of the human homolog of Saccharomyces cerevisiae REV1 is required for mutagenesis induced by UV light. Proc. Natl Acad. Sci. USA 97, 4186–4191 (2000). 57. Lin, W. et al. The human REV1 gene codes for a DNA template-dependent dCMP transferase. Nucleic Acids Res. 27, 4468–4475 (1999). 58. Nelson, J. R., Lawrence, C. W. & Hinkle, D. C. Thymine–thymine dimer bypass by yeast DNA polymerase ζ. Science 272, 1646–1649 (1996).

59. Johnson, R. E., Washington, M. T., Haracska, L., Prakash, S. & Prakash, L. Mutagenic bypass of DNA lesions by sequential action of eukaryotic DNA polymerases ι and ζ. Nature 406, 1015–1019 (2000). 60. Johnson, R. E., Prakash, S. & Prakash, L. Requirement of DNA polymerase activity of yeast Rad30 protein for its biological function. J. Biol. Chem. 274, 15975–15977 (1999). 61. Masutani, C. et al. The XPV (xeroderma pigmentosum variant) gene encodes human DNA polymerase η. Nature 399, 700–704 (1999). Purification of human xeroderma pigmentosum variant (XPV) and demonstration that it is the human homologue of yeast DNA pol η. 62. Johnson, R. E., Kondratick, C. M., Prakash, S. & Prakash, L. hRAD30 mutations in the variant form of Xeroderma pigmentosum. Science 285, 263–265 (1999). 63. Cordeiro-Stone, M., Zaritskaya, L. S., Price, L. K. & Kaufmann, W. K. Replication fork bypass of a pyrimidine dimer blocking leading strand DNA synthesis. J. Biol. Chem. 272, 13945–13954 (1997). 64. Masutani, C. et al. Xeroderma pigmentosum variant (XP-V) correcting protein from HeLa cells has a thymine dimer bypass DNA polymerase activity. EMBO J. 18, 3491–3501 (1999). 65. Johnson, R. E., Prakash, S. & Praskah, L. Efficient bypass of a thymine–thymine dimer by yeast DNA polymerase, Pol η. Science 283, 1001–1004 (1999). 66. Lin, W., Wu, X. & Wang, Z. A full-length cDNA of hREV3 is predicted to encode DNA polymerase ζ for damageinduced mutagenesis in humans. Mutat. Res. 433, 89–98 (1999). 67. Murakumo, Y. et al. A human REV7 homolog that interacts with the polymerase ζ catalytic subunit hREV3 and the spindle assembly checkpoint protein hMAD2. J. Biol. Chem. 275, 4391–4397 (2000). 68. Berek, C. & Milstein, C. The dynamic nature of the antibody repertoire. Immunol. Rev. 105, 5–26 (1988). 69. Zhang, J., MacLennan, I. C., Liu, Y. J. & Lane, P. J. Is rapid proliferation in B centroblasts linked to somatic mutation and memory B cell clones? Immunol. Lett. 18, 2393–2400 (1988). 70. Kim, S., Davis, M., Sinn, E., Patten, P. & Hood, L. Antibody diversity: somatic hypermutation of rearranged VH genes. Cell 27, 573–581 (1981). 71. Neuberger, M. S. et al. Monitoring and interpreting the intrinsic features of somatic hypermutation. Immunol. Rev. 162, 107–116 (1998). 72. Betz, A. G. et al. Elements regulating somatic hypermutation of an immunoglobulin kappa gene: critical role for the intron enhancer/matrix attachment region. Cell 77, 239–248 (1994). 73. Winter, D. B. & Gearhart, P. J. Another piece in the hypermutation puzzle. Curr. Biol. 5, 1345–1346 (1995). 74. Storb, U. The molecular basis of somatic hypermutation of immunoglobulin genes. Curr. Opin. Immunol. 8, 206–214 (1996). 75. Milstein, C., Neuberger, M. S. & Staden, R. Both DNA strands of antibody genes are hypermutation targets. Proc. Natl Acad. Sci. USA 95, 8791–8794 (1998). 76. Steel, E. J., Rothenfluh, H. S. & Blanden, R. V. Mechanism of antigen-driven somatic hypermutation of rearranged immunoglobulin V(D)J genes in the mouse. Immunol. Cell. Biol. 75, 82–95 (1997). 77. Winter, D. B., Sattar, N. & Gearhart, P. J. in Current topics in Microbiology and Immunology (eds Kelsoe, G. & Flajnits, M. F.) 1–10 (Springer, Berlin, 1998). 78. Diaz, M. & Flajnik, M. F. Evolution of somatic hypermutation and gene conversion in adaptive immunity. Immunol. Rev. 162, 13–24 (1998). An excellent summary of the key features of somatic hypermutation spectra. 79. Tissier, A., McDonald, J. P., Frank, E. G. & Woodgate, R. pol ι, a remarkably error-prone human DNA polymerase. Genes Dev. 14, 1642–1650 (2000). Demonstration that human Rad30B/pol ι prefers to make the wobble G–T base pair over the correct A–T base pair, and overall has extremely poor fidelity. 80. Foster, S. J., Dorner, T. & Lipsky, P. E. Somatic hypermutation of VκJκ rearrangements: targeting of RGYW motifs on both DNA strands and preferential selection of mutated codons within RGYW motifs. Eur. J. Immunol. 29, 4011–4021 (1999). 81. Dominguez, O. et al. DNA polymerase mu (Pol µ), homologous to TdT, could act as a DNA mutator in eukaryotic cells. EMBO J. 19, 1731–1742 (2000). 82. Garcia-Diaz, M. et al. DNA polymerase lambda (Pol lambda), a novel eukaryotic DNA polymerase with a potential role in meiosis. J. Mol. Biol. 301, 851–867 (2000). 83. Nagasawa, K.-I. et al. Identification and characterization of

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

human DNA polymerase β2, a DNA polymerase β–related enzyme. J. Biol. Chem. 275, 31233–31238 (2000). 84. Wang, Z., Castano, I. B., de Las Penas, A., Adams, C. & Christman, M. F. Pol κ: A DNA polymerase required for sister chromatid cohesion. Science 289, 774–779 (2000). Discovery that the yeast Trf4 protein is a polymerase essential for sister-chromatid cohesion, indicating that possible polymerase exchange mechanisms may be needed to replicate chromosomes fully. 85. Sharief, F. S., Vojta, P. J., Ropp, P. A. & Copeland, W. C. Cloning and chromosomal mapping of the human DNA polymerase θ (POLQ), the eighth human DNA polymerase. Genomics 59, 90–96 (1999). 86. Oshige, M., Aoyage, N., Harris, P. V., Burtis, K. C. & Sakaguchi, K. A new DNA polymerase species from Drosophila melanogaster: a probable mus308 gene product. Mutat. Res. 433, 183–192 (1999). 87. Radman, M. Enzymes of evolutionary change. Nature 401, 866–869 (1999). 88. Limoli, C. L., Giedzinski, E., Morgan, W. F. & Cleaver, J. E. Polymerase η deficiency in the Xeroderma pigmentosum variant uncovers an overlap between the S phase checkpoint and double-strand break repair. Proc. Natl Acad. Sci. USA 97, 7939–7946 (2000). 89. Opperman, T., Murli, S., Smith, B. T. & Walker, G. C. A model for umuDC-dependent prokaryotic DNA damage checkpoint. Proc. Natl Acad. Sci. USA 96, 9218–9223 (1999). 90. Murli, S., Opperman, T., Smith, B. T. & Walker, G. C. A role for the umuDC gene products of Escherichia coli in increasing resistance to DNA damage in stationary phase by inhibiting the transition to exponential growth. J. Bacteriol. 182, 1127–1135 (2000). 91. Bemark, M., Khamlichi, A. A., Davies, S. L. & Neuberger, M. S. Disruption of mouse polymerase ζ (Rev3) leads to embryonic lethality and impairs blastocyst development in vitro. Curr. Biol. 10, 1213–1216 (2000). 92. Wittschieben, J. et al. Disruption of the developmentally regulated Rev3I gene causes embryonic lethality. Curr. Biol. 10, 1217–1220 (2000). 93. Shapiro, L. & Losick, R. Dynamic spatial regulation in the bacterial cell. Cell 100, 89–98 (2000). 94. Lemon, K. P. & Grossman, A. D. Localization of bacterial DNA polymerase: evidence for a factory model of replication. Science 282, 1516–1519 (1998). 95. Sutton, M. D., Opperman, T. & Walker, G. C. The Escherichia coli SOS mutagenesis proteins UmuD and UmuD′ interact physically with the replicative DNA polymerase. Proc. Natl Acad. Sci. USA 96, 12373–12378 (1999). 96. Kornberg, A. Ten commandments: Lessons from the enzymology of DNA replication. J. Bacteriol. 182, 3613–3618 (2000). 97. Perry, K. L., Elledge, S. J., Mitchell, B. B., Marsh, L. & Walker, G. C. umuDC and mucAB operons whose products are required for UV light- and chemical-induced mutagenesis: UmuD, MucA, and LexA proteins share homology. Proc. Natl Acad. Sci. USA 82, 4331–4335 (1985). 98. Shinagawa, H., Iwasaki, T., Kato, T. & Nakata, A. RecA protein-dependent cleavage of UmuD protein and SOS mutagenesis. Proc. Natl Acad. Sci. USA 85, 1806–1810 (1988). 99. Maor-Shoshani, A., Reuven, N. B., Tomer, G. & Livneh, Z. Highly mutagenic replication by DNA polymerase V (UmuC) provides a mechanistic basis for SOS untargeted mutagenesis. Proc. Natl Acad. Sci. USA 97, 565–570 (2000). 100. Johnson, R. E., Prakash, S. & Prakash, L. The human DINB1 gene encodes the DNA polymerase pol θ. Proc. Natl Acad. Sci. USA 97, 3838–3843 (2000). 101. Morelli, C., Mungall, A. J., Negrini, M., Barbanti–Brodano, G. & Croce, C. M. Alternative splicing, genomic structure, and fine chromosome localization of REV3L. Cytogenet. Cell Genet. 83, 18–20 (1998). 102. Goldsmith, M., Sarov-Blat, L. & Livneh, Z. Plasmidencoded MucB protein is a DNA polymerase (pol RI) specialized for lesion bypass in the presence of MucA′, RecA, and SSB. Proc. Natl Acad. Sci. USA 97, 11227–11231 (2000).

Acknowledgements This work was supported by grants from the National Institutes of Health. I want to express heartfelt gratitude to my collaborators, Roger Woodgate, Mike O’Donnell, John-Stephen Taylor, Kevin McEntee and especially to Hatch Echols. I also want to express my sincere appreciation to the students in my laboratory, Phuong Pham, Mengjia Tang, Xuan Shen, Irina Bruck and Jeffrey Bertram.

VOLUME 1 | NOVEMBER 2000 | 1 0 9

REVIEWS

SECRETS OF ACTIN-BASED MOTILITY REVEALED BY A BACTERIAL PATHOGEN Lisa A. Cameron, Paula A. Giardini, Frederick S. Soo and Julie A. Theriot Actin-based cell motility is a complex process involving a dynamic, self-organizing cellular system. Experimental problems initially limited our understanding of this type of motility, but the use of a model system derived from a bacterial pathogen has led to a breakthrough. Now, all the molecular components necessary for dynamic actin self-organization and motility have been identified, setting the stage for future mechanistic studies.

AMOEBOID MOTILITY

A distinctive form of cell crawling typified by Amoeba proteus, which involves extension of pseudopodia and cytoplasmic streaming. FIBROBLAST

Common cell type found in connective tissue in many parts of the body, which secretes an extracellular matrix rich in collagen and other macromolecules and connects cell layers.

Department of Biochemistry, Stanford University School of Medicine, 279 Campus Drive West, Stanford, California 94305-5307, USA. e-mail: theriot@cmgm.stanford.edu Correspondence to: J.A.T.

110

The ability to move in a directed, purposeful manner is one of the properties we most closely associate with living cells. Many forms of cell motility, such as the intracellular movement of organelles by molecular motors, rely on discrete and stable protein machines. By coupling energy release to a protein conformational change, myosin and kinesin carry their cargo a single step along their substrate, and larger-scale movement is simply the linear addition of many such discrete steps. These types of motility have been extensively studied in purified or semi-purified systems, and a great deal is known about the molecular and biophysical requirements for movement1. In contrast, AMOEBOID MOTILITY is not driven by discrete machines acting additively. Instead, it is a complex process involving an interconnecting network of nonequilibrium, dynamic, whole-cell events. Although detailed descriptive studies of amoeboid motility have graced the cell biological literature for over fifty years, it could not be easily investigated at the molecular level using classic biochemical or genetic techniques. However, the past ten years have seen remarkable advancements in our understanding of the molecular basis of amoeboid motility. The breakthrough came, oddly, from a bacterial pathogen called Listeria monocytogenes. Like many cell biologists, this pathogen chose actin-based motility as its field of study, but it has had the advantage of millions of years of evolutionary experimentation. This review tells the story of how the

secrets of amoeboid motility known to this tiny bacterium have been revealed. Actin dynamics in locomoting cells

Crawling cells, such as epithelial cells, FIBROBLASTS or neurons, have at their front a broad, flat region, usually less

Cell body

Lamellipodium

Figure 1 | A rapidly moving cell; a keratocyte from the skin of a fish. This is a phase-contrast micrograph, a single frame from a video sequence. The lamellipodium and cell body are labelled. This cell is moving in the direction of the large arrow. Movie online

www.nature.com/reviews/molcellbio

| OCTOBER 2000 | VOLUME 1

ÂŠ 2000 Macmillan Magazines Ltd

REVIEWS

KERATOCYTE

A small, motile cell type found in the epidermis of fish and amphibians. LEADING EDGE

The thin margin of a lamellipodium spanning the area of the cell from the plasma membrane to about 1 µm back into the lamellipodium. GROWTH CONE

Motile tip of the axon or dendrite of a growing nerve cell, which spreads out into a large cone-shaped appendage.

than one micrometre thick, filled with a dense meshwork of actin filaments. Referred to as a lamellipodium, this region of the cell contains all of the machinery necessary for amoeboid motility. Rapidly moving cells, such as the fish epidermal KERATOCYTE, consist basically of a large lamellipodium that carries the cell body on its dorsal surface (FIG. 1) (Movie 1). Small lamellipodial fragments sliced off from a cell body can crawl on their own, essentially forming a tiny nucleus-free cell2,3. The crawling process can be broken down into three subprocesses: the assembly of actin into a coherent meshwork at the LEADING EDGE of the lamellipodium, the coupling of this meshwork to the external substrate, and the controlled depolymerization of the meshwork for recycling and reuse of the actin monomer. Understanding how each of these subprocesses is regulated and how they interconnect and work together is critical to the study of how cells crawl. Several experiments established the principle that actin assembly (BOX 1) occurs primarily at the front of lamellipodia. In fibroblasts and neuronal GROWTH CONES, a spot photobleached in the actin meshwork of the lamellipodium translocates backwards slowly and moves rearward relative to the leading edge in a coherent fashion4,5. When a stationary neuronal growth cone is allowed to recover after the actin meshwork is completely depolymerized by treatment with the toxin cytochalasin, the actin network reforms exclusively at the leading edge and then moves rearward6. Across cell types, the rate of rearward flux of the actin meshwork is negatively correlated with the speed of forward protrusion. In rapidly moving fish keratocytes, photoactivated spots of fluorescent actin that have incorporated into the lamellipodium remain stationary with respect to the substrate as the leading edge moves

Box 1 | Actin filament dynamics Actin is one of the most abundant proteins in eukaryotic cells, and is a primary determinant of cell shape and cytoplasmic structure. It exists in two forms, G-actin (for globular), the soluble 43 kDa protein subunit, and F-actin (for filamentous), a helical polymer of arbitrary length where individual subunits self-associate in a head-to-tail fashion. About half the actin in a typical cell (up to 50 µM) is in the form of G-actin, and the other half is in the form of F-actin. Actin is an ATPase, and ATP hydrolysis affects the kinetics of polymerization. The rate-limiting step in the formation of F-actin from a solution of pure G-actin is the formation of a stable ‘nucleus’. When two molecules of G-actin collide in solution, they will form a dimer, but the dimer comes apart rapidly and no filament can grow. When three or four molecules collide simultaneously, they form a more stable trimer or tetramer, which can be rapidly elongated by further collisions of individual subunits with either end of the growing filament. In cells, spontaneous nucleation is rare. Cells regulate the location of new F-actin formation by regulating nucleation. Within the cell, the dynamic behaviour of F-actin and G-actin is modified and regulated by a group of over 100 actin-binding proteins. These include proteins that bind to G-actin and prevent it from polymerizing, proteins that bind to F-actin and prevent it from depolymerizing, accessory proteins that affect the rate of nucleotide hydrolysis, proteins that sever long filaments into smaller bits, proteins that bypass the slow steps of nucleation, myosin motors that carry cargo along filaments … in short, proteins to speed up or slow down every dynamic behaviour of this remarkable polymer. In addition, F-actin crosslinking proteins can assemble multiple filaments into larger-scale structures, including bundles where all the filaments align in parallel and meshworks where the filaments cross orthogonally.

Comet tail Bacterium

Figure 2 | Movement of Listeria monocytogenes in an infected host cell. This is a phase-contrast micrograph, a single frame from a video sequence. The kidney epithelial cell was infected about five hours before the acquisition of this video sequence. All of the bacteria in this cell are clonal descendants of a single individual. A bacterium and its associated comet tail are labelled. Bacteria are moving in the direction of the blue arrows. Movie online

forward7. In slowly moving fibroblasts, the actin meshwork moves rearward with respect to the substrate, even as the leading edge continues to move forward8. To use the treadmilling activity of actin assembly in the lamellipodium for efficient forward extension of its leading edge, the cell must anchor the actin meshwork through the plasma membrane to the underlying substrate. A keratocyte has an efficient ‘clutch’ mechanism that allows rapid forward protrusion and little or no rearward flux, whereas a fibroblast has a ‘slippery clutch’ that results in significant rearward flux and slow forward protrusion. Within a given cell type, there is no correlation between the rate of centripetal movement of actin and the rate of lamellipodial protrusion, so the components that control the rate of protrusion by regulating actin dynamics must be localized at the leading edge8. The depolymerization of actin from the meshwork seems to be tightly controlled. Depending on the cell type, the average lifetime of actin filaments in the lamellipodium is very short — around 20 seconds to 2 minutes7,8. The rate of filament loss is correlated with cell speed: rapidly moving cells have more labile actin filaments in their lamellipodia, whereas filaments in slowly moving cells are more stable9. But most importantly, in all cells, the turnover of actin filaments is at least two orders of magnitude faster than the turnover of pure actin filaments in solution, indicating that other proteins inside the cell must be actively disassembling the filaments in the lamellipodia. Such cell-based experiments established organizational principles for making a functional lamellipodium

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 1 1

REVIEWS that could not have been predicted from in vitro properties of actin polymerization. First, actin filaments are nucleated and grow primarily at the leading edge, immediately adjacent to the plasma membrane. Second, filaments are crosslinked into a coherent meshwork that either remains stationary with respect to the substrate as the cell moves forward (in rapidly moving cells) or moves rearward towards the cell body (in stationary or slowly moving cells). And last, actin filaments in the bulk of the meshwork, away from the leading edge, depolymerize rapidly so that steady-state, selforganized movement can be maintained. Cell motility requires that these three processes be properly coordinated in space and time. In a lamellipodium, there must be modulatory factors that govern these phenomena. This raises several specific molecular questions. What factors catalyse nucleation and elongation of actin filaments at the leading edge? How are filament nucleation and elongation suppressed elsewhere? What is causing the older filaments to depolymerize so rapidly? What holds the meshwork together, and is it important for motility that this meshwork be coherent? Technical problems have impeded attempts to answer these questions. Genetic systems have been of limited use in identifying the full complement of components that make up the machinery necessary for actin-based motility. Yeast, unfortunately, do not crawl. Genetic and reverse genetic approaches in model metazoans and in Dictyostelium successfully defined the cellular functions of some of the individual components of the motility apparatus. However, because the actin polymerization machinery necessary for amoeboid motility is so critical for other aspects of cellular behaviour, many cells with lesions in important cytoskeletal loci are inviable and therefore difficult to evaluate for motility phenotypes. In addition, functional redundancy is rampant in the actin cytoskeleton, so many null mutants in genes that encode interesting proteins have no detectable phenotype. Biochemical reconstitution of amoeboid motility has been hindered by the need for an intact cell plasma membrane, which must serve to localize filament nucleation10 and might contribute to force generation11. A decade ago, the identification of a genetically manipulable model system that could mimic the actin filament dynamics of lamellipodial protrusion without the requirement for a plasma membrane was desperately needed to understand actin-based motility at the molecular level. Actin-based motility of bacterial pathogens

In the late 1980s, several research groups found that Factin is responsible for the intracellular movement of two unrelated bacterial pathogens, Listeria monocytogenes12,13 and Shigella flexneri14,15, which live within the cytoplasm of the host cell. Because L. monocytogenes is less virulent and easier to handle experimentally than S. flexneri, most laboratories investigating this form of actin-based motility have chosen to focus on L. monocytogenes, and we will concentrate on the L. monocytogenes model system in this review.

112

Figure 3 | Reconstitution of Listeria monocytogenes motility in a cytoplasmic extract. Top row | Phasecontrast images showing the position of the bacterium. Adjacent frames are separated by ten-second intervals. Bottom row | Fluorescent signal arising from rhodamineâ&#x20AC;&#x201C;actin added as a tracer. The fluorescent images were captured less than one second after the corresponding phase image. The online movie shows a polystyrene bead coated with ActA moving in a similar extract. Movie online

Actin-based motility is essential in the L. monocytogenes life cycle. In a natural food-borne infection, the bacteria induce phagocytosis by the epithelial cells lining the small intestine, an event that can be replicated in the laboratory in a wide variety of tissue-culture lines. The bacteria then secrete a pore-forming toxin (listeriolysin O) that degrades the enclosing membrane, and escape into the cytoplasm of the host cell. After a few hours of growth and division, host-cell actin filaments begin to form a dense cloud on the surface of the bacteria. Subsequently, the actin cloud becomes polarized into a comet tail made up of an oriented, crosslinked network of actin filaments with their barbed ends pointing towards the bacterium12. Bacteria associated with comet tails move very rapidly within the host cytoplasm, at rates of up to 1 Âľm per second16 (FIG. 2) (Movie 2). Finally, the infection spreads as the bacteria push their way into neighbouring cells through plasma membrane projections17,18. The L. monocytogenes comet tail resembles a simplified lamellipodium, and the bacterial surface imitates the plasma membrane at the leading edge. Labelled exogenous actin monomers preferentially incorporate into the actin tail near the bacterial surface in living and permeabilized cells19,20, recapitulating the behaviour in lamellipodia, where new incorporation is primarily at the leading edge. Fluorescence photoactivation experiments reveal that filaments in the tail of a moving bacterium remain stationary and only the bacterium moves forward21, as in rapidly moving lamellipodia. The rate of filament depolymerization in the comet tail is independent of either position in the tail or bacterial speed, and the filaments have very short half-lives, of the order of 30 seconds21. In contrast to amoeboid motility, bacterial actinbased motility does not require the host cell plasma membrane and could therefore be reconstituted in a cell-free cytoplasmic extract, which has facilitated biochemical approaches to study the regulation of actin www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

ÂŠ 2000 Macmillan Magazines Ltd

REVIEWS

Box 2 | Special features of the lamellipodium In general, it is thought that the dynamic behaviour of the actin-binding proteins described here is comparable between comet tails and lamellipodia. However, there are important differences. For example, ActA is not found in eukaryotic systems. The search for the eukaryotic equivalent of ActA led to the characterization of a new protein family called WASP/Scar. Wiskott–Aldrich syndrome protein (WASP) is expressed only in human haematopoietic cells and contains a GTPase binding domain69 that binds the small GTPases, Cdc42 and Rac, known to be involved in regulating the triggering of actin polymerization in fibroblasts70. Its close relative N-WASP is expressed widely in vertebrate cells71, and causes filopodial formation when co-expressed with Cdc42 in cultured cells72. The more distantly related protein Scar (for suppressor of cyclic AMP receptor mutation) was discovered in Dictyostelium, where its deletion causes cytoskeletal defects73. WASP and Scar interact with the p21 subunit of Arp2/374 and, like ActA, Scar activates Arp2/3 to nucleate actin filaments75. Finally, polystyrene beads coated with WASP are capable of forming actin comet tails and moving in cytoplasmic extracts, in a manner apparently identical to the movement of L. monocytogenes or ActA-coated beads76. Much of what is known about cell motility and lamellipodial protrusion has come from descriptive observations. Two of the most visually striking behaviours of lamellipodia include ruffling and rearward flux. Ruffling is a phenomenon where the protruding leading edge detaches from the substrate and folds back on the dorsal surface of the lamellipodium. Rearward flux, described in the section ‘Actin dynamics in locomoting cells’, requires myosin77. Neither of these characteristic behaviours can be investigated using L. monocytogenes as a model system.

dynamics22 (FIG. 3). The exclusive localization of actin filament growth close to the bacterial surface indicated that factors either secreted by the bacterium or expressed on the bacterial surface must trigger actin polymerization. Because L. monocytogenes continue to move in cells in the presence of drugs that inhibit bacterial protein synthesis20, the key factor was probably a stable protein on the surface of the bacterium rather than a factor secreted continually by the bacterium. The bacterial factors involved in actin polymerization are required for the spread of bacteria from cell to cell. To identify the gene(s) required for actin assembly, screens were designed to identify mutant bacteria deficient in the ability to spread from cell to cell, but capable of normal initial cell invasion, membrane lysis and bacterial division. The only gene ever isolated in such L. monocytogenes screens is actA23,24. Furthermore, ActA confers actin-based motility on normally immotile bacteria, for example if actA is expressed in the non-pathogenic strain Listeria innocua25, or if purified ActA protein is attached asymmetrically to Streptococcus pneumoniae26. Polystyrene beads coated with purified ActA protein have been shown to form comet tails and move in cytoplasmic extracts, proving that no other bacterial surface components are required for motility27 (Movie 3). Dissection of the comet tail and lamellipodium

A flurry of experimental work followed the identification of ActA and the reconstitution of motility in cytoplasmic extracts. Today, researchers in the field largely agree on the identities and the functions of all the main molecules required for regulating actin filament dynamics in the self-organized motile system of the comet tail. The similarities, and differences, between the simplified bacterial comet tail and the far more complicated, dynamic lamellopodial structure have given great insight into the organization and regulation of these systems at many levels. Factors catalysing nucleation and elongation. Although ActA is sufficient to cause polymerization at the bacterial surface, it does not interact directly with actin to form

a comet tail, indicating that it probably interacts with other host-cell factors. As ActA is a surface protein, initial efforts focused on host-cell factors that localize to the bacterial surface and not throughout the comet tail. Two proteins that fulfil this criterion were initially identified by immunofluorescence surveys: the G-actinbinding protein profilin22 and an F-actin associated protein called vasodilator-stimulated phosphoprotein (VASP)28. Both proteins require ActA to associate with the bacterial surface. VASP binds directly to ActA28, and profilin binds to VASP29. It seemed unlikely that these factors were responsible for localized actin nucleation, however, as cytoplasmic extracts depleted of profilin could still support nucleation of actin clouds by L. monocytogenes22,30. Furthermore, VASP binds to F-actin29 but does not show nucleating activity, and profilin significantly inhibits nucleation31. Profilin can serve as a nucleotide exchange factor for actin32, and it can also lower the effective critical concentration for actin polymerization in a cellular environment33. So, it was concluded that the combination of VASP and profilin may accelerate filament elongation at the bacterial surface, but neither one is the nucleator. Systematic ActA deletion studies carried out in bacteria indicated that VASP and profilin are localized by a central polyproline-rich region in the middle of the protein, and that four consensus FPPPP repeats act in an additive fashion to bind multiple molecules of VASP34. Bacteria containing an ActA construct that lacks the VASP-binding domain still mediate actin nucleation, although they move more slowly than wild-type bacteria in both cytoplasmic extracts35 and in infected cells34. Interestingly, the rate of motility is linearly related to the number of proline-rich repeats present34. Immunolocalization shows that the central proline-rich domain recruits VASP36, and biochemical experiments show that the consensus motif is sufficient for VASP binding37. Subsequently, the VASP-related proteins mammalian Enabled (Mena) and Ena/VASP-like protein (Evl) were found to act interchangeably with VASP in bringing profilin to the bacterial surface and in

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 1 3

REVIEWS

b P

C C C C C

Proline-rich repeats

C C

P ActA

Actin/Arp2/3 interaction

ActA

P VASP

Bacterium

VASP Bacterium

P Profilin

Actin

Arp2/3 complex

Capping protein

C Cofilin

α-Actinin

Figure 4 | Diagram of the molecular components required for actin-based motility of Listeria monocytogenes. a | Interactions between host-cell proteins and ActA at the bacterial surface. Two domains of ActA are required for normal motility. The amino-terminal domain activates actin filament nucleation through Arp2/3. The central proline-rich domain binds VASP and profilin interacts with VASP, enhancing filament elongation. b | Host protein functions throughout the comet tail. In addition to the factors that act at the bacterial surface, capping protein binds to the barbed end of actin filaments to prevent elongation of older filaments, α-actinin crosslinks filaments to stabilize the tail structure, and ADF/cofilin disassembles old filaments. (VASP, vasodilator-stimulated phosphoprotein.)

enhancing the speed of actin-based motility38,39. Profilin tagged with green fluorescent protein (GFP–profilin) associates with moving bacteria in infected cells and, strikingly, the concentration of GFP–profilin at the bacterial surface is closely correlated with bacterial speed40. This indicates that the prolinerich repeats, VASP and profilin may act together as an accelerator for bacterial movement. VASP and profilin may also have independent effects on filament elongation. Experiments using human platelet extracts show that movement of L. monocytogenes is still enhanced in the presence of a mutant profilin that does not bind proline-rich sequences and therefore cannot associate with VASP. Conversely, VASP still accelerates L. monocytogenes motility in profilindepleted extracts39. Surprisingly, overexpression of members of the Ena/VASP family in mammalian cells causes them to move at less than half the speed of wildtype cells, and removal of Ena/VASP proteins by sequestration to the mitochondrial surface causes cells to move faster than wild-type cells41. This observation indicates that enhancement of the rate of filament elon-

gation by VASP may not directly translate into an increase in crawling speed in mammalian cells, possibly highlighting an important difference between lamellipodia and L. monocytogenes motility (for other differences, see BOX 2). But regardless of the nature of the accelerator, what turns the ignition key by catalysing F-actin nucleation in the first place? The answer to this question came from a biochemical study. Platelet extracts, rich in actin-associated cytoskeletal proteins and easy to obtain in large quantities, were fractionated, and the fractions examined using a visual assay to determine which could support the formation of F-actin clouds around L. monocytogenes. The most purified active fraction contained a tightly associated protein complex consisting of seven polypeptide chains42. This complex, named Arp2/3 after two of its members (actin related proteins 2 and 3), had initially been isolated as a profilin-binding complex by affinity chromatography of Acanthamoeba castellanii cytosol43, but its function in the regulation of actin dynamics had previously been unclear. The nucleating activity of Arp2/3 can be measured in vitro44, but it is

Table 1 | Actin-based motility in other systems System

Structure formed

Host factors

Special features

References

Shigella flexneri

Comet tail similar to Listeria

N-WASP, Arp2/3, profilin

• Bacterial factor IcsA (VirG) has no homology with ActA • Actin dynamics are identical to L. monocytogenes

14, 15, 25, 50, 80, 81

Rickettsia spp.

Comet tail, made of long twisted bundles

• Bacterial factor not identified • Tails are straight • Dynamics are distinct from L. monocytogenes and S. flexneri

82–85

Enteropathogenic Escherichia coli

Pedestal

WASP, Arp2/3

• Signalling from bacterium occurs across host cell membrane • Bacterial factors intimin and Tir required

86, 87

Vaccinia virus

Comet tail similar to Listeria

N-WASP, Nck, WIP, • Intracellular enveloped form moves Src-family tyrosine kinase • Actin-based movement may contribute to viral budding

Vesicles

Transient comet tail similar to Listeria

PtdIns(4,5)P2, Cdc42, N-WASP, Arp2/3

88–90

• Endosomal rocketing induced by phorbol esters, metal ions 91–95 • May occur normally with nascent endosomes

Nck, non-catalytic region of tyrosine kinase; PtdIns(4,5)P2, phosphatidylinositol-4,5-bisphosphate; N-WASP, neural Wiskott–Aldrich syndrome protein; WASP, Wiskott–Aldrich syndrome protein; WIP, Wiskott–Aldrich syndrome protein interacting protein.

114

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS

Plasma membrane N-WASP

N-WASP P

P P

P N -W AS

N -W AS

N-WASP

Actin

P Profilin

Arp2/3 complex

C C

N-WASP

N-W AS P

N-WASP

VASP

N-WASP P

NWA SP

Capping protein

C Cofilin

Filamin

Figure 5 | Functions of similar proteins in the lamellipodium. N-WASP activates Arp2/3 to nucleate actin filaments. VASP and profilin, which are localized to the leading edge, facilitate elongation. Capping protein caps the barbed ends of older filaments. Filamin crosslinks filaments into an actin network. Finally, ADF/cofilin accelerates depolymerization throughout the lamellipodium, except for a cofilin-free zone at the immediate leading edge (reviewed in REF. 78). The localization of N-WASP is not well known. Here we show it at the leading edge, binding and activating Arp2/3 in the cytoplasm, and associated with Arp2/3 at filament branches. (VASP, vasodilator-stimulated phosphoprotein; N-WASP; neural Wiskott–Aldrich syndrome protein.)

markedly activated by the presence of ActA (REF. 45). The amino-terminal domain of ActA, which was implicated by the deletion studies as being sufficient to cause actin nucleation inside of cells, is also sufficient for full activation of Arp2/3 (REF. 46). Arp2/3 can also bind to the side of a pre-existing actin filament and initiate nucleation of a new filament at that location, creating a branch at a 70° angle from the original filament44. Such branches are found throughout the lamellipodium47 with Arp2/3 at the branch points. Arp2/3 is localized to the leading edge of several cell types42,43,48, and it is found throughout the actin comet tail associated with L. monocytogenes49. So, activated Arp2/3 is responsible for the nucleation of actin polymerization at the bacterial surface (FIG. 4a). Activation of Arp2/3 by ActA is essential for L. monocytogenes motility. No full-length homologues of ActA exist in mammalian cells, so how is Arp2/3 activated in lamellipodia? Similarly, IcsA (VirG), the bacterial surface protein required for S. flexneri actin-based motility, does not interact directly with Arp2/3, indicating that some other factor mediates the activation of Arp2/3 in this system. Recently, it has been found that Wiscott–Aldrich syndrome protein (WASP) and its relatives, N-WASP and Scar (for suppressor of cyclic AMP receptor mutation), have a function similar to ActA, activating Arp2/3 downstream of signalling through small GTPases (BOX 2). In the case of S. flexneri motility, N-WASP binds IcsA (VirG) and activates Arp2/3 at the bacterial surface50 (TABLE 1). Factors suppressing nucleation and elongation. New actin filaments are continuously nucleated and elongated exclusively at the bacterial surface or at the leading edge of the cell, suggesting that some mechanism exists to prevent the continuing elongation of old filaments.

The simplest mechanism to achieve this would involve capping the growing barbed ends of the older filaments. Several proteins with barbed-end F-actin capping activity are known, including capping protein (also known as CapZ) and gelsolin. Biochemical studies indicate that ActA may suppress capping close to the bacterial surface30, although ActA probably exerts this effect indirectly. Capping protein is strongly associated with comet tails51. Gelsolin is localized throughout the tail and, paradoxically, is enriched at the bacterial surface52, but it may be inactive there. The combination of barbed-end capping suppression at the bacterial surface, exclusive localization of the elongation enhancers VASP and profilin at the bacterial surface, and potent activation of Arp2/3 by ActA, seems to be sufficient to enable nucleation and elongation only at the front of the comet tail. Factors causing filament depolymerization. An important feature of both lamellipodial and bacterial motility dynamics is the rapid depolymerization of the actin meshwork far from the leading edge and the bacterial surface, suggesting the existence of depolymerizing factors. Another functional study using L. monocytogenes elucidated the specific role of the protein that controls filament depolymerization. This protein, called ADF (actin depolymerizing factor) or cofilin, was first identified in biochemical assays as a factor that accelerates actin depolymerization. ADF/cofilin binds cooperatively to the sides of actin filaments and increases the twist of Factin53, destabilizing the filament structure and causing an increase in the rate of spontaneous filament breakage54 and a significant acceleration of subunit dissociation from the filament pointed end55. It has a higher affinity for ADP-containing filaments, and so preferentially accelerates the turnover of old filaments after nucleotide hydrolysis has occurred, rather than the newest filaments, which still contain ATP. This combination of activities makes cofilin the leading candidate as the factor responsible for the 10–100-fold higher actin turnover rate in cells compared with the turnover rate of pure actin. Immunodepletion of ADF/cofilin from cytoplasmic extracts supporting L. monocytogenes motility alters the morphology of the comet tails, making them five times longer than normal56. Conversely, addition of excess exogenous ADF/cofilin to extracts causes shortening of the actin tail and increases the rate of bacterial motility55. ADF/cofilin is localized throughout the L. monocytogenes comet tail56, consistent with the previous finding that depolymerization occurs uniformly everywhere in the tail21. Experiments done in intact cells also show that ADF/cofilin is important for acceleration of actin turnover57. ADF/cofilin is localized to the leading edge and ruffling membrane of motile cells58. Rapidly moving keratocytes show an ADF/cofilin-free zone at the very leading edge of the lamellipodium47. This narrow ADF/cofilin-free margin may be the geometrical correlate of the hydrolysis kinetics of ATP; the newest ATPcontaining filaments at the front are briefly protected from rapid disassembly by ADF/cofilin. The spatial separation between nucleation and elongation at the front,

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 1 5

REVIEWS

Table 2 | Similarities between lamellipodia and comet tails Protein

Function

Localization in lamellipodia

Localization in comet tails

F-actin*

Cell shape and cytoplasmic structure

Enriched in stress fibres and focal contacts

Throughout tail16

Arp2/3*

Nucleation Filament crosslinking Pointed-end capping

Leading edge and throughout lamellipodium 47,49 Filament branches47

Bacterial surface and throughout tail42

Capping protein* Barbed-end capping

Cytoplasm and cell–cell contacts96

Throughout tail51

Gelsolin‡

Barbed-end capping, severing

Cytoplasm and focal contacts97

Bacterial surface and throughout tail52,56

VASP

Binds F-actin and profilin, binds prolinerich region of ActA

Leading edge98 and focal contacts29

Bacterial surface34

Profilin

Actin monomer binding

Leading edge, focal contacts40

Bacterial surface22

ADF/cofilin*

Depolymerizes ADP filaments Lamellipodium58 (1 µm from edge47) Throughout tail56

α-actinin

Crosslinking

Focal contacts and stress fibres99

Throughout tail16

Lamellipodium

Comet tail

* Absolutely required for motility in the reconstituted system. ‡ Can be substituted for capping protein.

and disassembly further back, helps to maintain the steady-state organization of the motile F-actin meshwork in both lamellipodia and comet tails. Factors that crosslink filaments. Because Arp2/3 frequently binds to the side of a pre-existing filament as it nucleates the growth of a new filament, the meshwork forming at the front of the comet tail or the leading edge of the lamellipodium is effectively crosslinked at birth, in a dendritic web44,47. In addition, numerous F-actin crosslinking proteins are found throughout the comet tail, including fimbrin59 and α-actinin16. Microinjection of a dominant-negative fragment of α-actinin, which inhibits crosslinking by the endogenous protein, causes L. monocytogenes in infected cells to stop moving60. This observation indicates that strong crosslinking is important for movement through the highly viscous cytoplasm of a living cell, although its mechanical contribution may be less important in cytoplasmic extracts (see below). Fimbrin and α-actinin, which crosslink F-actin to form tight parallel bundles, are not generally found in lamellipodia. Instead, lamellipodia are enriched in a different type of crosslinking protein, filamin (also called ABP-280), which tends to crosslink filaments at right angles to form a web. Mutant melanoma cells that fail to express filamin show very poor motility61. These complementary results in comet tails and in lamellipodia seem to indicate that crosslinking is indeed important for mechanical stability of a protrusive self-organized actin structure, but that the nature of the crosslinker required is probably different depending on the details of filament organization in each case. Establishment of a purified system. Using the molecular information provided by the studies detailed above, we might hypothesize that L. monocytogenes motility could be reconstituted in vitro with a mixture of the following host proteins: actin, Arp2/3,VASP, profilin, capping protein, ADF/cofilin and α-actinin, along with a steady supply of ATP. This impressive feat has recently been accom-

116

plished62 (FIG. 4b). Of these proteins, only actin, Arp2/3, ADF/cofilin and capping protein are absolutely required for motility.VASP and profilin increase the rate of movement, and α-actinin stabilizes the tail. The success of this purified system definitively establishes the minimal components necessary for this type of self-organized actin-based motility. In addition, it clearly shows that this form of force generation does not require a myosin motor62, and that actin polymerization alone can act as a bona fide molecular motor63. Moreover, the functional conservation between the lamellipodium and the comet tail is striking (TABLE 2, FIG. 5), proving that L. monocytogenes is an excellent tiny cell biologist. Open questions

The establishment of a purified protein motility system that includes all of the bacterial and host components necessary and sufficient for actin-based movement62 represents a satisfying culmination of the past decade of molecular research in this field. Work from numerous laboratories has brought us to the point where we now know most of the critical molecules involved in actin regulation at the leading edge of cells. However, despite this list of molecules, we still cannot formulate the set of rules needed to generate a motile cell, or a lamellipodium, or even a comet tail. The purified system has confirmed the supposition that actin polymerization alone must produce the force necessary for motility, but provides no further information about the actual microscopic mechanism of force generation. It will, however, provide a useful experimental system for investigating some of the open mechanistic questions about cell movement. Translating the minimal requirements for bacterial actin-based motility to a mechanism of cell motility or even the protrusion of a lamellipodium requires a significant increase in complexity, the most important difference being the presence of a plasma membrane. Further understanding of cell motility will also require knowledge of how the regulation of cell adhesion and nuclear translocation are integrated with lamellipodial protruwww.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS

BROWNIAN RATCHET MODEL (ELASTIC)

A proposed model for actinbased motility in which actin filaments are thought to flex away from the bacterial surface to allow addition of monomer at the end of the filament. When the filament flexes back, it is one subunit longer and pushes the bacterium forward that distance. BULK ELASTIC MODEL

A proposed model for actinbased motility, which treats the actin comet tail as a cohesive elastic gel that responds elastically to deformation. This indicates that the energy from actin polymerization may be stored as elastic energy in the actin gel to produce force that propels the bacterium forward.

sion to move the entire cell. But even within the comparatively simple system of the actin comet tail, many basic mechanistic questions remain to be addressed. We suggest that the answers to the following questions will begin to reveal the rules needed to organize an actin comet tail, and may lead to insights into how to address the more difficult questions of whole-cell locomotion. How is force generated? In theory, motile force for actinbased motility can come solely from the chemical potential of actin polymerization63,64. Several models for the mechanism by which this conversion takes place have been proposed, including BROWNIAN RATCHET-TYPE 65 and BULK ELASTIC MODELS66. Current data are MODELS inconclusive about which of these models, if any, are correct, and have been limited in large part by the natural variations in the extract and cultured-cell preparations used to assay motility. With the purified protein system, force generation by the self-organizing actin polymerization machinery should now be amenable to detailed biophysical experimentation analogous to the studies that defined the protein conformational changes responsible for force generation by the discrete motor proteins myosin and kinesin1. How is movement initiated? Actin-associated bacteria in cells seem to exist in two states: moving with a comet tail, or stationary with a uniform cloud12. The same two states are seen in cytoplasmic extracts, both for bacteria and for ActA-coated polystyrene beads27. These two states can readily interconvert in a classic bistable system. What makes a bacterium or a bead start moving? Stochastic modelling based on the elastic Brownian ratchet mechanism for force generation has suggested that this symmetry breaking event might be caused by a form of dynamic positive feedback. Small variations in the polymerization rate on one side of the bacterium versus the other side can be amplified to cause largescale symmetry breaking67. The quantitative theoretical predictions of this model could be confirmed or refuted using the purified system.

Links

What causes variations in speed? Within a population of genetically identical bacteria moving in a single host-cell type, there are wide variations in average speed from one bacterium to another17,21,34. This wide variation is not due to sampling error, but rather to

Vale, R. D. & Milligan, R. A. The way things move: looking under the hood of molecular motor proteins. Science 288, 88–95 (2000). Euteneuer, U. & Schliwa, M. Persistent, directional motility of cells and cytoplasmic fragments in the absence of microtubules. Nature 310, 58–61 (1984). Verkhovsky, A. B., Svitkina, T. M. & Borisy, G. G. Selfpolarization and directional motility of cytoplasm. Curr. Biol. 9, 11–20 (1999).

the fact that some individual bacteria are intrinsically faster than other individual bacteria (P.A.G and J.A.T., unpublished observations). What is responsible for these intrinsic differences? Changes in ActA surface density do not affect speed27,34, so the source of variation must lie elsewhere. Even within the trajectory of an individual bacterium, there are significant variations in speed over time, often over an order of magnitude within a period of a few minutes. Are variations in the subcellular environment responsible for this? If so, can this property be used to map out the positions of biochemically distinct microenvironments within a living cell? What causes curvature in trajectories? It is extremely rare to find a bacterium that moves in a straight line; most bacterial tails are gently curved (FIG. 2, FIG. 3). What causes these curves? Is there a correlation between variations in curvature and variations in speed? Some bacterial strains carrying point mutations in ActA show curvature behaviours that are very different from wild-type bacteria, including one mutant that makes tighter smooth curves and one mutant that can ‘skid’, making occasional sharp turns at apparently random intervals (P. Lauer, S. Rafelski, D. Portnoy and J.A.T, unpublished observations). How do these mutations affect interactions with the host-cell proteins that govern actin self-organization and movement? These mechanistic questions, and many others, can now be definitively addressed in the purified protein system62, exploiting the ability of ActA to confer motility on artificial particles whose geometry can be controlled27. So, with a set of proteins in hand, and a simple system in which to study these interactions, the time is ripe for a powerful convergence of molecular, biochemical and physical techniques on a single area: the organization and control of the actin cytoskeleton. This review has detailed an experimental success story, in which an unusual bacterial system has been successfully exploited as a robust, reproducible proxy for eukaryotic actin-based amoeboid motility. Using immunofluorescence, biochemical and genetic techniques, proteins putatively involved in motility were identified and found to be necessary and sufficient for motility, as shown by a minimal reconstituted system. As a whole, this experimental journey can serve as a lesson on how to approach molecular mastery of a dynamic, self-organizing cellular system. Now that the molecular basis of this type of motility is fairly well understood, attention can be focused on the biophysical mechanisms68, bringing us one step closer to a detailed understanding of the beautiful and complex process of amoeboid motility.

Wang, Y. L. Exchange of actin subunits at the leading edge of living fibroblasts: possible role of treadmilling. J. Cell Biol. 101, 597–602 (1985). These photobleaching studies show that actin monomers are added at the leading edge and that the actin meshwork translocates backward towards the centre of the cell in a stationary lamellipodium. Okabe, S. & Hirokawa, N. Actin dynamics in growth cones. J. Neurosci. 11, 1918–1929 (1991).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

Forscher, P. & Smith, S. J. Actions of cytochalasins on the organization of actin filaments and microtubules in a neuronal growth cone. J. Cell Biol. 107, 1505–1516 (1988). Theriot, J. A. & Mitchison, T. J. Actin microfilament dynamics in locomoting cells. Nature 352, 126–131 (1991). Shows that the rate of cell motility is directly related to the rate of actin filament assembly at the leading

VOLUME 1 | NOVEMBER 2000 | 1 1 7

REVIEWS

10.

11. 12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

118

edge and that filaments further back in the lamellipodium remain stationary as the rapidly moving cell translocates over them. Theriot, J. A. & Mitchison, T. J. Comparison of actin and cell surface dynamics in motile fibroblasts. J. Cell Biol. 119, 367–377 (1992). Zigmond, S. H. Recent quantitative studies of actin filament turnover during cell locomotion. Cell Motil. Cytoskeleton 25, 309–316 (1993). Shariff, A. & Luna, E. J. Diacylglycerol-stimulated formation of actin nucleation sites at plasma membranes. Science 256, 245–247 (1992). Oster, G. F. & Perelson, A. S. The physics of cell motility. J. Cell Sci. 8, S35–S54 (1987). Tilney, L. G. & Portnoy, D. A. Actin filaments and the growth, movement, and spread of the intracellular bacterial parasite, Listeria monocytogenes. J. Cell Biol. 109, 1597–1608 (1989). This is the initial publication showing that the bacterial pathogen, Listeria monocytogenes, associates with host cytoplasmic actin, suggesting that these bacteria move within and between cells in some manner involving the actin cytoskeleton. Mounier, J., Ryter, A., Coquis-Rondon, M. & Sansonetti, P. J. Intracellular and cell-to-cell spread of Listeria monocytogenes involves interaction with F-actin in the enterocytelike cell line Caco- 2. Infect. Immunol. 58, 1048–1058 (1990). Makino, S., Sasakawa, C., Kamata, K., Kurata, T. & Yoshikawa, M. A genetic determinant required for continuous reinfection of adjacent cells on large plasmid in S. flexneri 2a. Cell 46, 551–555 (1986). Bernardini, M. L., Mounier, J., d’Hauteville, H., CoquisRondon, M. & Sansonetti, P. J. Identification of icsA, a plasmid locus of Shigella flexneri that governs bacterial intra- and intercellular spread through interaction with Factin. Proc. Natl Acad. Sci. USA 86, 3867–3871 (1989). Dabiri, G. A., Sanger, J. M., Portnoy, D. A. & Southwick, F. S. Listeria monocytogenes moves rapidly through the hostcell cytoplasm by inducing directional actin assembly. Proc. Natl Acad. Sci. USA 87, 6068–6072 (1990). Robbins, J. R. et al. Listeria monocytogenes exploits normal host cell processes to spread from cell to cell. J. Cell Biol. 146, 1333–1350 (1999). Theriot, J. A. The cell biology of infection by intracellular bacterial pathogens. Annu. Rev. Cell. Dev. Biol. 11, 213–239 (1995). Sanger, J. M., Sanger, J. W. & Southwick, F. S. Host cell actin assembly is necessary and likely to provide the propulsive force for intracellular movement of Listeria monocytogenes. Infect. Immunol. 60, 3609–3619 (1992). Shows that actin monomers add to the tail only at the bacterial surface whereas α–actinin and tropomyosin are found throughout the tail, and suggests that actin polymerization may provide the force for motility. Tilney, L. G., DeRosier, D. J., Weber, A. & Tilney, M. S. How Listeria exploits host cell actin to form its own cytoskeleton. II. Nucleation, actin filament polarity, filament assembly, and evidence for a pointed end capper. J. Cell Biol. 118, 83–93 (1992). Theriot, J. A., Mitchison, T. J., Tilney, L. G. & Portnoy, D. A. The rate of actin-based motility of intracellular Listeria monocytogenes equals the rate of actin polymerization. Nature 357, 257–260 (1992). Shows that the rate of bacterial motility is the same as the rate of actin polymerization and that actin depolymerization is independent of position in the comet tail. Theriot, J. A., Rosenblatt, J., Portnoy, D. A., GoldschmidtClermont, P. J. & Mitchison, T. J. Involvement of profilin in the actin-based motility of L. monocytogenes in cells and in cell-free extracts. Cell 76, 505–517 (1994). Kocks, C. et al. L. monocytogenes-induced actin assembly requires the actA gene product, a surface protein. Cell 68, 521–531 (1992). Shows that the bacterial surface protein ActA is required for actin-based motility of Listeria monocytogenes. Domann, E. et al. A novel bacterial virulence gene in Listeria monocytogenes required for host cell microfilament interaction with homology to the proline-rich region of vinculin. EMBO J. 11, 1981–1990 (1992). Kocks, C. et al. The unrelated surface proteins ActA of Listeria monocytogenes and IcsA of Shigella flexneri are sufficient to confer actin-based motility on Listeria innocua and Escherichia coli respectively. Mol. Microbiol. 18, 413–423 (1995). Smith, G. A., Portnoy, D. A. & Theriot, J. A. Asymmetric distribution of the Listeria monocytogenes ActA protein is

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41. 42.

43.

44.

45.

required and sufficient to direct actin-based motility. Mol. Microbiol. 17, 945–951 (1995). Cameron, L. A., Footer, M. J., van Oudenaarden, A. & Theriot, J. A. Motility of ActA protein-coated microspheres driven by actin polymerization. Proc. Natl Acad. Sci. USA 96, 4908–4913 (1999). Chakraborty, T. et al. A focal adhesion factor directly linking intracellularly motile Listeria monocytogenes and Listeria ivanovii to the actin-based cytoskeleton of mammalian cells. EMBO J. 14, 1314–1321 (1995). This was the first identification of a host-cell cytoskeletal protein (VASP) able to interact directly with a bacterial virulence factor (ActA). Reinhard, M. et al. The proline-rich focal adhesion and microfilament protein VASP is a ligand for profilins. EMBO J. 14, 1583–1589 (1995). Marchand, J. B. et al. Actin-based movement of Listeria monocytogenes: actin assembly results from the local maintenance of uncapped filament barbed ends at the bacterium surface. J. Cell Biol. 130, 331–343 (1995). Carlsson, L., Nystrom, L. E., Sundkvist, I., Markey, F. & Lindberg, U. Actin polymerizability is influenced by profilin, a low molecular weight protein in non-muscle cells. J. Mol. Biol. 115, 465–483 (1977). Goldschmidt-Clermont, P. J. et al. The control of actin nucleotide exchange by thymosin β 4 and profilin. A potential regulatory mechanism for actin polymerization in cells. Mol. Biol. Cell 3, 1015–1024 (1992). Pantaloni, D. & Carlier, M. F. How profilin promotes actin filament assembly in the presence of thymosin β 4. Cell 75, 1007–1014 (1993). Smith, G. A., Theriot, J. A. & Portnoy, D. A. The tandem repeat domain in the Listeria monocytogenes ActA protein controls the rate of actin-based motility, the percentage of moving bacteria, and the localization of vasodilatorstimulated phosphoprotein and profilin. J. Cell Biol. 135, 647–660 (1996). Lasa, I., David, V., Gouin, E., Marchand, J. B. & Cossart, P. The amino-terminal part of ActA is critical for the actinbased motility of Listeria monocytogenes; the central proline-rich region acts as a stimulator. Mol. Microbiol. 18, 425–436 (1995). Pistor, S., Chakraborty, T., Walter, U. & Wehland, J. The bacterial actin nucleator protein ActA of Listeria monocytogenes contains multiple binding sites for host microfilament proteins. Curr. Biol. 5, 517–525 (1995). Niebuhr, K. et al. A novel proline-rich motif present in ActA of Listeria monocytogenes and cytoskeletal proteins is the ligand for the EVH1 domain, a protein module present in the Ena/VASP family. EMBO J. 16, 5433–5444 (1997). Gertler, F. B., Niebuhr, K., Reinhard, M., Wehland, J. & Soriano, P. Mena, a relative of VASP and Drosophila Enabled, is implicated in the control of microfilament dynamics. Cell 87, 227–239 (1996). Laurent, V. et al. Role of proteins of the Ena/VASP family in actin-based motility of Listeria monocytogenes. J. Cell Biol. 144, 1245–1258 (1999). Geese, M. et al. Accumulation of profilin II at the surface of Listeria is concomitant with the onset of motility and correlates with bacterial speed. J. Cell Sci. 113, 1415–1426 (2000). Bear, J. E. et al. Negative regulation of fibroblast motility by Ena/VASP proteins. Cell 101, 717–728 (2000). Welch, M. D., Iwamatsu, A. & Mitchison, T. J. Actin polymerization is induced by Arp2/3 protein complex at the surface of Listeria monocytogenes. Nature 385, 265–269 (1997). Using a visual assay to detect nucleation of an actin cloud at the surface of the bacterial pathogen, L. monocytogenes, a seven-member protein complex responsible for actin filament nucleation was isolated. Machesky, L. M., Atkinson, S. J., Ampe, C., Vandekerckhove, J. & Pollard, T. D. Purification of a cortical complex containing two unconventional actins from Acanthamoeba by affinity chromatography on profilinagarose. J. Cell Biol. 127, 107–115 (1994). Mullins, R. D., Heuser, J. A. & Pollard, T. D. The interaction of Arp2/3 complex with actin: nucleation, high affinity pointed end capping, and formation of branching networks of filaments. Proc. Natl Acad. Sci. USA 95, 6181–6186 (1998). Shows that Arp2/3 has a high affinity for the pointed end of actin filaments in addition to side binding, and leads to a model for actin filament network branching that is the current model for actin assembly at the leading edge. Welch, M. D., Rosenblatt, J., Skoble, J., Portnoy, D. A. & Mitchison, T. J. Interaction of human Arp2/3 complex and the Listeria monocytogenes ActA protein in actin filament

nucleation. Science 281, 105–108 (1998). 46. Skoble, J., Portnoy, D. A. & Welch, M. D. Three regions within ActA promote Arp2/3 complex-mediated actin nucleation and Listeria monocytogenes motility. J. Cell Biol. 150, 527–537 (2000). 47. Svitkina, T. M. & Borisy, G. G. Arp2/3 complex and actin depolymerizing factor/cofilin in dendritic organization and treadmilling of actin filament array in lamellipodia. J. Cell Biol. 145, 1009–1026 (1999). This is a beautiful study using correlative light and electron microscopy to compare protein localization and structure of the actin network in the lamellipodia of fast moving keratocytes and slower moving fibroblasts. 48. Mullins, R. D., Stafford, W. F. & Pollard, T. D. Structure, subunit topology, and actin-binding activity of the Arp2/3 complex from Acanthamoeba. J. Cell Biol. 136, 331–343 (1997). 49. Welch, M. D., DePace, A. H., Verma, S., Iwamatsu, A. & Mitchison, T. J. The human Arp2/3 complex is composed of evolutionarily conserved subunits and is localized to cellular regions of dynamic actin filament assembly. J. Cell Biol. 138, 375–384 (1997). 50. Suzuki, T., Miki, H., Takenawa, T. & Sasakawa, C. Neural Wiskott-Aldrich syndrome protein is implicated in the actin-based motility of Shigella flexneri. EMBO J. 17, 2767–2776 (1998). Shows that N–WASP, a known activator of Arp2/3, mediates actin tail formation at the pole of S. flexneri through the bacterial surface protein VirG (IcsA), linking bacteria to the other host cell factors that help assemble the comet tail. 51. David, V. et al. Identification of cofilin, coronin, Rac and capZ in actin tails using a Listeria affinity approach. J. Cell Sci. 111, 2877–2884 (1998). 52. Laine, R. O. et al. Gelsolin, a protein that caps the barbed ends and severs actin filaments, enhances the actin-based motility of Listeria monocytogenes in host cells. Infect. Immunol. 66, 3775–3782 (1998). 53. McGough, A., Pope, B., Chiu, W. & Weeds, A. Cofilin changes the twist of F-actin: implications for actin filament dynamics and cellular function. J. Cell Biol. 138, 771–781 (1997). 54. Maciver, S. K., Zot, H. G. & Pollard, T. D. Characterization of actin filament severing by actophorin from Acanthamoeba castellanii. J. Cell Biol. 115, 1611–1620 (1991). 55. Carlier, M. F. et al. Actin depolymerizing factor (ADF/cofilin) enhances the rate of filament turnover: implication in actinbased motility. J. Cell Biol. 136, 1307–1322 (1997). Shows that the actin depolymerization rate in vitro is directly affected by ADF/cofilin in a concentrationdependent fashion and that addition of ADF to cell extracts increases turnover rate in L. monocytogenes actin tails. 56. Rosenblatt, J., Agnew, B. J., Abe, H., Bamburg, J. R. & Mitchison, T. J. Xenopus actin depolymerizing factor/cofilin (XAC) is responsible for the turnover of actin filaments in Listeria monocytogenes tails. J. Cell Biol. 136, 1323–1332 (1997). 57. Lappalainen, P. & Drubin, D. G. Cofilin promotes rapid actin filament turnover in vivo. Nature 388, 78–82 (1997). Using a yeast genetic study, provides in vivo evidence that cofilin, an actin depolymerizing factor, does in fact enhance filament depolymerization in cells. 58. Bamburg, J. R. & Bray, D. Distribution and cellular localization of actin depolymerizing factor. J. Cell Biol. 105, 2817–2825 (1987). 59. Prevost, M. C. et al. Unipolar reorganization of F-actin layer at bacterial division and bundling of actin filaments by plastin correlate with movement of Shigella flexneri within HeLa cells. Infect. Immunol. 60, 4088–4099 (1992). 60. Dold, F. G., Sanger, J. M. & Sanger, J. W. Intact α-actinin molecules are needed for both the assembly of actin into the tails and the locomotion of Listeria monocytogenes inside infected cells. Cell Motil. Cytoskeleton 28, 97–107 (1994). 61. Cunningham, C. C. et al. Actin-binding protein requirement for cortical stability and efficient locomotion. Science 255, 325–327 (1992). 62. Loisel, T. P., Boujemaa, R., Pantaloni, D. & Carlier, M. F. Reconstitution of actin-based motility of Listeria and Shigella using pure proteins. Nature 401, 613–616 (1999). Identifies the minimal set of proteins required for the reconstitution of actin-based motility with purified components for both L. monocytogenes and S. flexneri motility, and demonstrates that no myosin motor is required. 63. Theriot, J. A. The polymerization motor. Traffic 1, 19–28

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS (2000). 64. Hill, T. L. & Kirschner, M. W. Bioenergetics and kinetics of microtubule and actin filament assembly-disassembly. Int. Rev. Cytol. 78, 1–125 (1982). 65. Mogilner, A. & Oster, G. Cell motility driven by actin polymerization. Biophys. J. 71, 3030–3045 (1996). 66. Gerbal, F. et al. Measurement of the elasticity of the actin tail of Listeria monocytogenes. Eur. Biophys. J. 29, 134–140 (2000). 67. van Oudenaarden, A. & Theriot, J. A. Cooperative symmetry-breaking by actin polymerization in a model for cell motility. Nature Cell Biol. 1, 493–499 (1999). 68. Kuo, S. C. & McGrath, J. L. Steps and fluctuations of Listeria monocytogenes during actin–based motility. (submitted). 69. Symons, M. et al. Wiskott–Aldrich syndrome protein, a novel effector for the GTPase CDC42Hs, is implicated in actin polymerization. Cell 84, 723–734 (1996). 70. Hall, A. Rho GTPases and the actin cytoskeleton. Science 279, 509–514 (1998). 71. Miki, H., Miura, K. & Takenawa, T. N-WASP, a novel actindepolymerizing protein, regulates the cortical cytoskeletal rearrangement in a PIP2-dependent manner downstream of tyrosine kinases. EMBO J. 15, 5326–5335 (1996). 72. Miki, H., Sasaki, T., Takai, Y. & Takenawa, T. Induction of filopodium formation by a WASP-related actindepolymerizing protein N-WASP. Nature 391, 93–96 (1998). 73. Bear, J. E., Rawls, J. F. & Saxe, C. L., SCAR, a WASPrelated protein, isolated as a suppressor of receptor defects in late Dictyostelium development. J. Cell Biol. 142, 1325–1335 (1998). 74. Machesky, L. M. & Insall, R. H. Scar1 and the related Wiskott-Aldrich syndrome protein, WASP, regulate the actin cytoskeleton through the Arp2/3 complex. Curr. Biol. 8, 1347–1356 (1998). 75. Machesky, L. M. et al. Scar, a WASp-related protein, activates nucleation of actin filaments by the Arp2/3 complex. Proc. Natl Acad. Sci. USA 96, 3739–3744 (1999). 76. Yarar, D., To, W., Abo, A. & Welch, M. D. The WiskottAldrich syndrome protein directs actin-based motility by stimulating actin nucleation with the Arp2/3 complex. Curr. Biol. 9, 555–558 (1999). 77. Lin, C. H., Espreafico, E. M., Mooseker, M. S. & Forscher,

78.

79.

80.

81.

82.

83.

84.

85.

86.

87.

88.

P. Myosin drives retrograde F-actin flow in neuronal growth cones. Neuron 16, 769–782 (1996). Pollard, T. D., Blanchoin, L. & Mullins, R. D. Molecular mechanisms controlling actin filament dynamics in nonmuscle cells. Annu. Rev. Biophys. Biomol. Struct. 29, 545–576 (2000). Pantaloni, D., Boujemaa, R., Didry, D., Gounon, P. & Carlier, M. F. The Arp2/3 complex branches filament barbed ends: functional antagonism with capping proteins. Nature Cell Biol. 2, 385–391 (2000). Goldberg, M. B. & Theriot, J. A. Shigella flexneri surface protein IcsA is sufficient to direct actin- based motility. Proc. Natl Acad. Sci. USA 92, 6572–6576 (1995). Mimuro, H. et al. Profilin is required for sustaining efficient intra- and intercellular spreading of Shigella flexneri. J. Biol. Chem. 275, 28893–28901 (2000). Heinzen, R. A., Hayes, S. F., Peacock, M. G. & Hackstadt, T. Directional actin polymerization associated with spotted fever group Rickettsia infection of Vero cells. Infect. Immunol. 61, 1926–1935 (1993). Teysseire, N., Chiche-Portiche, C. & Raoult, D. Intracellular movements of Rickettsia conorii and R. typhi based on actin polymerization. Res. Microbiol. 143, 821–829 (1992). Heinzen, R. A., Grieshaber, S. S., Van Kirk, L. S. & Devin, C. J. Dynamics of actin-based movement by Rickettsia rickettsii in vero cells. Infect. Immunol. 67, 4201–4207 (1999). Gouin, E. et al. A comparative study of the actin-based motilities of the pathogenic bacteria Listeria monocytogenes, Shigella flexneri and Rickettsia conorii. J. Cell Sci. 112, 1697–1708 (1999). Kenny, B. et al. Enteropathogenic E. coli (EPEC) transfers its receptor for intimate adherence into mammalian cells. Cell 91, 511–520 (1997). Kalman, D. et al. Enteropathogenic E. coli acts through WASP and Arp2/3 complex to form actin pedestals. Nature Cell Biol. 1, 389–391 (1999). Cudmore, S., Cossart, P., Griffiths, G. & Way, M. Actinbased motility of vaccinia virus. Nature 378, 636–638 (1995). Demonstrates that a virus moves by actin-based motility in host cells, expanding the realm of this type of motility to another system which has a different shape and possibly a new way to form a polarized actin tail.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

89. Frischknecht, F. et al. Actin-based motility of vaccinia virus mimics receptor tyrosine kinase signalling. Nature 401, 926–929 (1999). 90. Moreau, V. et al. A complex of N-WASP and WIP integrates signalling cascades that lead to actin polymerization. Nature Cell Biol. 2, 441–448 (2000). 91. Ma, L., Cantley, L. C., Janmey, P. A. & Kirschner, M. W. Corequirement of specific phosphoinositides and small GTP-binding protein Cdc42 in inducing actin assembly in Xenopus egg extracts. J. Cell Biol. 140, 1125–1136 (1998). 92. Rohatgi, R. et al. The interaction between N-WASP and the Arp2/3 complex links Cdc42-dependent signals to actin assembly. Cell 97, 221–231 (1999). 93. Taunton, J. et al. Actin-dependent propulsion of endosomes and lysosomes by recruitment of N-WASP. J. Cell Biol. 148, 519–530 (2000). 94. 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). 95. Rozelle, A. L. et al. Phosphatidylinositol 4,5-bisphosphate induces actin-based movement of raft-enriched vesicles through WASP-Arp2/3. Curr. Biol. 10, 311–320 (2000). 96. Schafer, D. A., Mooseker, M. S. & Cooper, J. A. Localization of capping protein in chicken epithelial cells by immunofluorescence and biochemical fractionation. J. Cell Biol. 118, 335–346 (1992). 97. Cooper, J. A., Loftus, D. J., Frieden, C., Bryan, J. & Elson, E. L. Localization and mobility of gelsolin in cells. J. Cell Biol. 106, 1229–1240 (1988). 98. Rottner, K., Behrendt, B., Small, J. V. & Wehland, J. VASP dynamics during lamellipodia protrusion. Nature Cell Biol. 1, 321–322 (1999). 99. Lazarides, E. & Burridge, K. α–actinin: immunofluorescent localization of a muscle structural protein in nonmuscle cells. Cell 6, 289–298 (1975).

Acknowledgements We thank Rachael Ream for digital time-lapse used to make Figure 1 and Movie 1. We would like to apologize to those researchers whose work could not be cited due to space limitations. J.A.T. is supported by grants from the National Institutes of Health and a Fellowship in Science and Engineering from the David and Lucile Packard Foundation.

VOLUME 1 | NOVEMBER 2000 | 1 1 9

REVIEWS

APOPTOSIS IN NEURODEGENERATIVE DISORDERS Mark P. Mattson Neuronal death underlies the symptoms of many human neurological disorders, including Alzheimer’s, Parkinson’s and Huntington’s diseases, stroke, and amyotrophic lateral sclerosis. The identification of specific genetic and environmental factors responsible for these diseases has bolstered evidence for a shared pathway of neuronal death — apoptosis — involving oxidative stress, perturbed calcium homeostasis, mitochondrial dysfunction and activation of cysteine proteases called caspases. These death cascades are counteracted by survival signals, which suppress oxyradicals and stabilize calcium homeostasis and mitochondrial function. With the identification of mechanisms that either promote or prevent neuronal apoptosis come new approaches for preventing and treating neurodegenerative disorders. SYNAPTOGENESIS

The process of formation of synapses, the sites where neurons communicate through release of neurotransmitters from the presynaptic terminal and activation of receptors on the postsynaptic neuron. NEUROTROPHIC FACTORS

Proteins produced by neurons and glial cells that promote neuron survival and growth.

Laboratory of Neurosciences, National Institute on Aging, Gerontology Research Center, 5,600 Nathan Shock Drive, Baltimore, Maryland 21224, USA. e-mail: mattsonm@grc.nia.nih.gov

120

For proliferative tissues to maintain a constant size and function properly, the older cells must die to make way for new cells. Such ‘programmed’ death involves a stereotyped sequence of biochemical and morphological changes that allows the cell to die without adversely affecting its neighbours — a process called apoptosis (FIG. 1). In many diseases, aberrant regulation of apoptosis is the central abnormality. For example, resistance of cells to apoptosis is thought to be responsible for many types of cancer, and identification of the molecular alterations responsible for such cell immortalization is an important area in cancer research1. Although details of the definition of apoptosis vary among investigators, there is general agreement that apoptosis is a cell death process involving caspase activation and a lack of cell swelling with maintenance of organelle (mitochondria and endoplasmic reticulum) integrity. Apoptosis occurs within a tissue in a ‘spotty’ pattern such that healthy and dying cells are intermingled. This contrasts with another form of cell death, called necrosis, in which cellular organelles swell and the plasma membrane lyses, resulting in massive death of groups of cells throughout a tissue. As described below, the criteria for apoptosis are fulfilled in many neurological disorders in which neuronal death is a central feature. In contrast to the rapid turnover of cells in proliferative tissues, neurons commonly survive for the entire

lifetime of the organism — this enduring nature of neurons is necessary for maintaining the function of those cells within neuronal circuits. For example, motor neurons must maintain connections to skeletal muscles, and long-term memories probably require the continued survival of the neurons in the regions of the brain in which those memories are encoded. During development of the central and peripheral nervous systems, many neurons undergo apoptosis during a time window that coincides with the process of SYNAPTOGENESIS2. Signals that determine whether or not developing neurons live or die may include competition for a limited supply of target-derived NEUROTROPHIC FACTORS and activation of receptors for the excitatory neurotransmitter glutamate3. Initial overproduction of neurons, followed by death of some, is probably an adaptive process that provides enough neurons to form nerve cell circuits that are precisely matched to their functional specifications4. Accordingly, the decision as to which neurons die is made by cellular signal transduction pathways that are ‘tuned’ to the functionality of neuronal circuits. Unfortunately, many people experience excessive death of one or more populations of neurons as the result of disease or injury. For example, death of hippocampal and cortical neurons is responsible for the symptoms of Alzheimer’s disease; death of midbrain neurons that use the neurotransmitter dopamine underwww.nature.com/reviews/molcellbio

| OCTOBER 2000 | VOLUME 1

REVIEWS

Death signals: • Oxidative stress • Glutamate • Decreased growth-factors • Genetic mutation

Phosphatidylserine Cytochrome c Apaf-1 Caspase-9

Oxygen radicals/Ca2+

Blebs

Death substrates

Par-4

Proteolysis Endonucleases

Caspase-3

Bax, Bad PTP Nucleus Mitochondrion

Initiation phase

Cytochrome c Oxyradicals/Ca2+

Effector phase

Degradation phase

Figure 1 | Morphological and biochemical features of apoptosis. During the initiation phase of apoptosis, the death signal activates an intracellular cascade of events that may involve increases in levels of oxyradicals and Ca2+, production of Par-4 and translocation of pro-apoptotic Bcl-2 family members (Bax and Bad) to the mitochondrial membrane. Certain caspases (caspase8, for example) can also act early in the cell death process before, or independently of, mitochondrial changes. The effector phase of apoptosis involves increased mitochondrial Ca2+ and oxyradical levels, the formation of permeability transition pores (PTP) in the mitochondrial membrane, and release of cytochrome c into the cytosol. Cytochrome c forms a complex with apoptotic protease-activating factor 1 (Apaf-1) and caspase-9. Activated caspase-9, in turn, activates caspase-3, which begins the degradation phase of apoptosis in which various caspase and other enzyme substrates are cleaved, resulting in characteristic changes in the plasma membrane (blebbing and exposure of phosphatidylserine on the cell surface, which is a signal that stimulates cell phagocytosis by macrophages/microglia). Finally, the nuclear chromatin becomes condensed and fragmented.

lies Parkinson’s disease; Huntington’s disease involves the death of neurons in the striatum, which control body movements; and death of lower motor neurons manifests as amyotrophic lateral sclerosis (FIG. 2). The number of people with such neurodegenerative disorders is rapidly increasing as the average lifespan gets longer. Neuronal death cascades

METABOLIC STRESS

Conditions in which levels of glucose, oxygen and other molecules required for ATP (energy) production are decreased.

Many signals can initiate or ‘trigger’ apoptosis in neurons (FIG. 3). The best-studied signal is lack of neurotrophic factor support, which may trigger apoptosis during development of the nervous system and possibly in neurodegenerative disorders2–4. Most neurons in the mammalian central nervous system possess receptors for another trigger of apoptosis — the excitatory neurotransmitter glutamate. Overactivation of glutamate receptors can induce apoptosis by a mechanism involving calcium influx5,6, and such ‘excitotoxicity’ may occur in acute neurodegenerative conditions such as stroke, trauma and severe epileptic seizures, as well as in Alzheimer’s disease and motor system disorders7,8. A third trigger of neuronal death is increased oxidative stress, in which free radicals (such as the superoxide anion radical and the hydroxyl radical) damage cellular lipids, proteins and nucleic acids by attacking chemical bonds in those molecules9,10. METABOLIC STRESS, as occurs after a stroke or during ageing, may also initiate neuronal apoptosis. Finally, environmental toxins can induce neuronal apoptosis, and several such toxins can induce patterns of brain damage and behavioural phenotypes remarkably similar to Parkinson’s and Huntington’s diseases11,12. The genetic and environmental factors that trigger neuronal apoptosis may be different in various physio-

logical and pathological settings (TABLE 1), but many of the subsequent biochemical events that execute the cell death process are highly conserved. One classification of this death programme is based on compelling evidence that mitochondrial changes are pivotal in the cell death decision in many cases13. Mitochondria in cells undergoing apoptosis show increased oxyradical production, opening of pores in their membranes and release of cytochrome c (FIG. 1). These changes are central to the cell death process because agents such as manganese superoxide dismutase and cyclosporin A, which act directly on mitochondria to suppress oxidative stress and membrane pore formation, also prevent neuronal death in experimental models14. The biochemical alterations that occur during the early stages of apoptosis may induce mitochondrial dysfunction either directly or indirectly (TABLE 2). The B-cell lymphoma-2 (Bcl-2) family of proteins includes both pro- and anti-apoptotic members15. The best-studied anti-apoptotic members in neurons are Bcl-2 and BclxL; pro-apoptotic members include Bcl-2-associated Xprotein (Bax) and Bcl-associated death promoter (Bad). Overexpression of Bcl-2 in cell cultures and in transgenic mice increases resistance of neurons to death induced by excitotoxic, metabolic and oxidative insults relevant to Alzheimer’s disease, stroke and other disorders16,17. Conversely, neurons lacking Bax are protected against apoptosis18. The mechanism by which Bcl-2 proteins control cell death is not clear, but it may involve interactions among family members and association of the proteins with mitochondria, resulting in altered ion movements across mitochondrial membranes19. Further mechanisms that can regulate the early stages of apoptosis involve caspases, the prostate apoptosis

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 2 1

REVIEWS response-4 (Par-4) protein and telomerase. Caspases are evolutionarily conserved cysteine proteases central to apoptosis of many cell types20. Some caspases are activated during the early phase of apoptosis. A prominent example is the activation of caspase-8 in neurons in response to ligation of ‘death receptors’ such as Fas and the p75 neurotrophin receptor21. These upstream caspases can then activate ‘effector’ caspases such as caspase-3 either directly or indirectly, and may thereby elicit apoptosis independently of mitochondrial alterations. Effector caspases can also be activated in response to mitochondrial changes and cytochrome c release; these caspases can then activate a DNase that cleaves DNA into oligonucleosome-sized fragments22. Caspases can also cleave various substrate proteins that may coordinate the cell death process, including enzymes such as poly-ADP-ribose polymerase and ataxia telangiectasia mutated (ATM) kinase; ion channels including subunits of the α-amino3-hydroxy-5-methyl-4-isoxazole propionate (AMPA) subtype of neuronal glutamate receptor; and cytoskeletal proteins such as actin and spectrin (for a review, see REF. 20). Par-4 was identified as being upregulated in prostate tumour cells undergoing apoptosis, but is now known to be essential in developmental and pathological neuronal death23,24. Levels of Par-4 increase rapidly in response to various apoptotic stimuli through enhanced translation of Par-4 messenger RNA. A LEUCINE ZIPPER domain in the carboxyl terminus of Par-4 is essential for its pro-apoptotic function, and interactions of Par-4 with other proteins, including protein kinase Cζ and Bcl-2, through this zipper may be central to the mechanism by which Par-4 induces mitochondrial dysfunction. Telomerase adds a six-base DNA sequence (TTAGGG) to the ends of chromosomes, preventing their shortening and protecting them during chromosome segregation in mitotic cells25. Telomerase consists of a catalytic reverse-transcriptase subunit (TERT), an RNA template and several associated regulatory proteins. Telomerase activity is increased during cell immortalization and transformation (and is therefore strongly implicated in the pathogenesis of many cancers), and in many tissues including the brain during development, but is downregulated in all somatic tissues during late embryonic and early postnatal development. Telomerase activity and expression of TERT are associated with increased resistance of neurons to apoptosis in experimental models of developmental neuronal death and neurodegenerative disorders26,27. The anti-apoptotic action of TERT in neurons is exerted at an early step in the cell death pathway before mitochondrial alterations and caspase activation. These new findings indicate that telomerase may be important in neural development and injury responses. Cell survival mechanisms

LEUCINE ZIPPER

A leucine-rich domain within a protein that binds to other proteins with a similar domain.

122

Much of the structural and functional complexity of the nervous system arises because neurons do not divide. The persistence of neurons throughout life allows the nervous system to maintain continuous function over long distances and to encode enduring memories. Therefore considerable pressure was placed on the ner-

vous system during evolution to develop mechanisms that guard against neuronal death. The marked symptoms of neurodegenerative disorders emphasize the importance of mechanisms that promote neuron survival and plasticity (FIG. 3, BOX 1). There are several prominent anti-apoptotic signalling pathways3. Neurotrophic factors have been identified that can protect neurons against apoptosis by activating receptors linked through kinase cascades to production of cellsurvival-promoting proteins. For example, brain-derived neurotrophic factor (BDNF), nerve growth factor (NGF) and basic fibroblast growth factor (bFGF) can prevent death of cultured neurons, in part by stimulating production of antioxidant enzymes, Bcl-2 family members and proteins involved in regulation of calcium homeostasis3,28. Cytokines such as tumour necrosis factor-α (TNF-α), ciliary neurotrophic factor (CNTF) and leukaemia inhibitory factor (LIF) can prevent neuronal death in experimental models of natural neuronal death and neurodegenerative disorders29–31. Several neurotrophic factors and cytokines use a a

Cerebral cortex

Hippocampus

Striatum

Figure 2 | Brain regions in which neurodegenerative conditions are typified by selective apoptosis of neurons. a | In Alzheimer’s disease, neurons in the hippocampus and certain regions of the cerebral cortex degenerate. b | In Huntington’s disease, neurons in the striatum die. In Parkinson’s disease, dopamine neurons in the substantia nigra (not shown) undergo apoptosis, and in stroke, the neurons that die are those supplied by an occluded or ruptured blood vessel.

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS survival pathway involving the transcription factor NFκB (REF. 31). Activation of NF-κB can protect cultured neurons against death induced by diverse stimuli, including trophic-factor withdrawal and exposure to excitotoxic, oxidative and metabolic insults. Studies in mice that lack the p50 subunit of NF-κB show that NFκB is also anti-apoptotic in the brain in vivo32. Gene targets that mediate the survival-promoting action of NFκB may include manganese superoxide dismutase, Bcl-2 and inhibitor of apoptosis proteins (IAPs). However, NF-κB activation in MICROGLIA can promote neuronal apoptosis by inducing production of oxyradicals and 33 EXCITOTOXINS , so NF-κB may either prevent or promote neuronal death depending on specific conditions.

MICROGLIA

Phagocytic immune cells in the brain that engulf and remove cells that have undergone apoptosis. EXCITOTOXINS

Compounds such as glutamate, kainic acid and N-methyl-Daspartate that can kill neurons by activating excitatory aminoacid (glutamate) receptors.

Life cascade IAPs

Bcl-2 Bcl-xL

Ca2+ PTP

Caspases

Axon

Mn-SOD Ca2+ Ca2+

Mitochondrion

Kinases and transcription factors Life cascade Glutamate R Ca2+

NTF

Death cascade Bcl-2 p53

Bax Bad

Apaf-1 Caspases

Death cascade

Apoptosis and neurodegenerative disorders

PTP

Par-4

Dendrite ROS Ca2+ Cytochrome c Ca2+

Mitochondrion

In addition to extracellular signal-mediated neuroprotection pathways, several intracellular signalling pathways have been identified that can protect neurons against apoptosis. For example, stress can induce the expression of neurotrophic factors and heat-shock proteins34,35. The neurotrophic factors, in turn, act in an autocrine or paracrine manner to activate cell surface receptor-mediated kinase signalling pathways that ultimately induce expression of genes encoding survival-promoting proteins such as antioxidant enzymes. Heat-shock proteins act as ‘chaperones’ for many proteins, thereby maintaining protein stability. They may also interact directly with caspases, inhibiting their activation. Calcium is arguably the most versatile and important intracellular messenger in neurons36. Interestingly, although calcium may often promote neuronal death, it can also activate pathways that promote survival. For example, calcium can promote survival through a pathway involving activation of protein kinase B (PKB/Akt) by calcium/calmodulin-dependent protein kinase37. Calcium is a prominent regulator of cellular responses to stress, activating transcription through the cyclic-AMP response element-binding protein (CREB), which can promote neuron survival in experimental models of developmental cell death38. Calcium can also activate a rapid neuroprotective signalling pathway in which the calcium-activated actin-severing protein gelsolin induces actin depolymerization, resulting in suppression of calcium influx through membrane NMDA (N-methyl-Daspartate) receptors and voltage-dependent calcium channels39. This may occur through intermediary actinbinding proteins that interact with NMDA receptor and calcium channel proteins. Finally, signals such as calcium and secreted amyloid precursor protein-α (sAPP-α), which increase cyclic GMP production, can induce activation of potassium channels and the transcription factor NF-κB, and thereby increase resistance of neurons to excitotoxic apoptosis40.

Figure 3 | Roles for altered synaptic signalling in neurodegenerative disorders. Ageand disease-related stressors promote excessive activation of apoptotic (death) biochemical cascades in synaptic terminals and neurites. For example, overactivation of glutamate receptors under conditions of reduced energy availability or increased oxidative stress (from reactive oxygen species, ROS) results in Ca2+ influx into postsynaptic regions of dendrites. Ca2+ entering the cytoplasm through plasma membrane channels and endoplasmic reticulum (ER) channels induces apoptotic cascades (lower left) that involve Par-4, pro-apoptotic Bcl-2 family members (Bax and Bad), and/or p53. These factors act on mitochondria to induce Ca2+ influx, oxidative stress, opening of permeability transition pores (PTP) and release of cytochrome c, which forms a complex with apoptotic protease-activating factor 1 (Apaf-1). This results in caspase activation and execution of the cell death process. Anti-apoptotic (life) signalling pathways are also concentrated in synaptic compartments (upper right). For example, activation of receptors (R) for neurotrophic factors (NTF) in axon terminals stimulates kinase cascades and transcription factors and increased production of survival-promoting proteins such as Bcl-2, Bcl-xL and manganese superoxide dismutase (Mn-SOD) (which act at the level of mitochondria) and inhibitor of apoptosis proteins (IAPs) (which inhibit caspases).

For each of the disorders described below, analyses of post-mortem tissue from patients and studies of experimental animal and cell-culture models have implicated neuronal apoptosis. Studies of the pathogenic mechanisms of genetic mutations that cause early-onset autosomal dominant forms of Alzheimer’s and Huntington’s diseases and amyotrophic lateral sclerosis have been particularly valuable in implicating apoptosis in age-related neurological disorders41–43. Nevertheless, it is very difficult to demonstrate apoptosis convincingly in the brains of patients. This is because apoptosis usually occurs quite rapidly over several hours to a day, making it hard to ‘catch’ many cells showing classic features of apoptosis at any one time. In addition, experiments cannot be done in humans to establish whether blocking a step in the apoptotic biochemical cascade can prevent neuronal death. For these reasons, much of the evidence supporting an apoptotic mode of neuronal death comes from studies of animal and cell-culture models of neurodegenerative disorders.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 2 3

REVIEWS

LIMBIC STRUCTURES

Brain structures such as the hippocampus, amygdala and septum that function in learning and memory, and in emotions. NEURITES

Generic name for processes (axons and dendrites) elaborated from neuronal cell bodies. SYNAPTOSOME

A structure consisting of preand postsynaptic terminals prepared from homogenized brain tissue with cellular fractionation techniques.

Table 1 | Factors that may modulate apoptosis in neurodegenerative disorders Disorder

Genetic factors

Environmental factors

Alzheimer’s

APP, presenilin mutations, ApoE

Head trauma, low education, calorie intake

Parkinson’s

α-synuclein, parkin mutations

Head trauma, toxins, calorie intake

Huntington’s61,62,66

Poly-CAG expansions in huntingtin

ALS77

Cu/Zn-SOD mutations

Toxins, autoimmune response

Stroke83,84

Cadasil mutations

Smoking, dietary calories and fat

47–51

In most of the neurodegenerative disorders considered here, the disease process lasts for years or even decades. It is therefore likely that, although individual neurons may be dysfunctional for extended periods in these disorders, they may die rapidly once the apoptotic cascade is fully activated. In this view, the progressive deficits that occur in chronic neurodegenerative disorders are the result of progressive attrition of individual neurons. Although the following section presents some of the salient evidence supporting the involvement of apoptosis in a select group of neurodegenerative disorders, it should be noted that this is a controversial area of research with many gaps to be filled. Alzheimer’s disease. Alzheimer’s disease is characterized by progressive impairment of cognition and emotional disturbances that are strongly correlated with synaptic degeneration and death of neurons in LIMBIC STRUCTURES, such as the hippocampus and the amygdala, and associated regions of the cerebral cortex (FIG. 2). Degenerating neurons show aggregates of hyperphosphorylated tau protein, and evidence of excessive calcium-mediated proteolysis and oxidative stress44 (FIG. 4). A defining feature of Alzheimer’s disease is accumula-

Box 1 | Synaptic signalling in neuronal death and survival Components of signalling pathways that initiate or prevent apoptosis are highly concentrated in synaptic terminals — the sites of intercellular communication between neurons. For example, receptors for glutamate are located in postsynaptic regions of dendrites and receptors for neurotrophic factors are in both pre- and postsynaptic terminals. Much of the biochemical machinery involved in apoptosis can be activated in synaptic terminals, where it can alter synaptic function and promote localized degeneration of synapses and NEURITES105,106 (FIG. 3). For example, Par-4 production, mitochondrial alterations, caspase activation and release into the cytosol of factors that may cause nuclear apoptosis can be induced in SYNAPTOSOME preparations and neurites of cultured brain neurons by insults that induce apoptosis in intact neurons105,106. Caspase-mediated cleavage of synaptic proteins may control the process of neuronal apoptosis. For example, AMPA receptor subunits are selectively degraded in hippocampal neurons after exposure to an apoptotic dose of glutamate, resulting in decreased calcium influx, thereby preventing excitotoxic necrosis6. The latter mechanism might also allow neurons to ‘withdraw’ from participation in neuronal circuits, permitting them to recover from potentially lethal conditions. Apoptotic pathways may also function in synaptic plasticity, particularly under conditions of stress and injury. Studies showing that TNF-α and NF-κB activation modify long-term depression and potentiation of synaptic transmission in the hippocampus107 provide further evidence that anti-apoptotic signalling can modulate synaptic plasticity. Finally, changes in mitochondrial membrane permeability in synaptic terminals have been associated with impaired synaptic plasticity in the hippocampus93, suggesting a role for apoptotic mitochondrial alterations in synaptic function.

124

tion of amyloid plaques formed by aggregates of amyloid-β peptide — a 40–42 amino-acid fragment generated by proteolytic processing of the amyloid precursor protein (APP)44,45. Increased DNA damage and caspase activity, and alterations in expression of apoptosis-related genes such as Bcl-2 family members, Par4 and DNA damage response genes have been found in neurons associated with amyloid deposits in the brains of Alzheimer’s patients25,46,47. Expression-profile analysis of thousands of genes in brain tissue samples from Alzheimer’s patients and age-matched control patients revealed a marked decrease in expression of an anti-apoptotic gene called NCKAP1 (for NCKassociated protein 1)48. Exposure of cultured neurons to amyloid-β can induce apoptosis directly49, and can greatly increase their vulnerability to death induced by conditions such as increased oxidative stress and reduced energy availability that are known to occur in the brain during ageing10. The mechanism by which amyloid-β sensitizes neurons to death involves membrane lipid peroxidation. This impairs the function of membrane ionmotive ATPases and glucose and glutamate transporters, resulting in membrane depolarization, ATP depletion, excessive calcium influx and mitochondrial dysfunction. Accordingly, antioxidants that suppress lipid peroxidation and drugs that stabilize cellular calcium homeostasis can protect neurons against amyloidβ-induced apoptosis10. Moreover, neurotrophic factors and cytokines known to prevent neuronal apoptosis can protect neurons against amyloid-β-induced death3. Further evidence for the involvement of apoptotic cascades in Alzheimer’s disease comes from studies showing that APP is a substrate for caspase-3 (REF. 50). In addition to promoting apoptosis by generating amyloid-β, β- and/or γ-secretase cleavage of APP produces a membrane-associated carboxy-terminal fragment of APP that can induce apoptosis, possibly by a pathway involving an APP-binding protein called APP-BP1, which may drive neurons into a mitotic cycle that ends in apoptosis51. Caspase-mediated cleavage of APP can release a carboxy-terminal peptide called C31 that is a potent inducer of apoptosis52. Mutations in three genes, each inherited in an autosomal dominant manner, can cause early-onset inherited forms of Alzheimer’s disease — one gene encodes APP, and the other two genes encode presenilins 1 and 2. APP mutations seem to cause Alzheimer’s disease by altering proteolytic processing of APP such that levels of amyloid-β are increased and levels of sAPP-α are www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS

Table 2 | Examples of proteins that promote or suppress neuronal apoptosis Pro-apoptotic Caspases

Cleavage of various enzyme, cytoskeletal and ion-channel substrates

Bax, Bad

Pore formation in mitochondrial membrane; cytochrome c release

Glutamate receptor proteins Calcium influx; activation of kinase and proteases Fas

Initiates death cascade involving caspase-8

Par-4

Mitochondrial dysfunction; suppression of survival signals (NF-κB)

p53

Transcription of death genes; enhancement of Bax actions

Anti-apoptotic

SUBSTANTIA NIGRA

A part of the midbrain that contains dopamine-producing neurons whose axons innervate the striatum and thereby control body movements. MITOCHONDRIAL COMPLEX I

A group of proteins located at the inner mitochondrial membrane that function very early in the electron transport chain. LEWY BODIES

Eosinophilic, cytoplasmic neuronal inclusions that contain aggregates of the proteins α-synuclein and ubiquitin. LYMPHOBLASTS

Bone-marrow-derived cells that give rise to lymphocytes.

Bcl-2, Bcl-xL

Stabilize mitochondrial function; suppress oxidative stress

IAPs

Caspase inhibition

Trophic factors/cytokines

Induced expression of antioxidant enzymes, calcium regulating proteins, IAPs, Bcl-2; phosphorylation of Akt and other substrates

Telomerase

Prevents telomere shortening; modulates DNA-damage responses

Anti-oxidant enzymes

Suppress oxidative stress

Protein kinase Cζ

Stimulates survival-gene expression (NF-κB)

Calcium-binding proteins

Stabilize calcium homeostasis

decreased (FIG. 5). Presenilin mutations may promote neuronal degeneration by enhancing γ-secretase cleavage of APP, thereby increasing production of neurotoxic amyloid-β (1–42)45. When mutant presenilin-1 protein is expressed in cultured cells and in transgenic and knock-in mice, neurons become susceptible to death induced by various insults, including trophic-factor withdrawal, exposure to amyloid-β or glutamate, and energy deprivation53. Mutant presenilin-1 acts at an early step before Par-4 production, mitochondrial dysfunction and caspase activation. Instead, calcium homeostasis in the endoplasmic reticulum (ER) is disturbed such that more calcium is released when neurons are exposed to potentially damaging oxidative and metabolic insults54. Agents that suppress ER calcium release, including dantrolene and xestospongin, can counteract the endangering effects of the mutations54, indicating that enhanced calcium release is central to the pathogenic action of mutant presenilin-1. Presenilin-2 may also facilitate apoptosis, although the underlying mechanism has not been established55. Motor system disorders. These disorders include Parkinson’s disease, Huntington’s disease and amyotrophic lateral sclerosis. Patients with Parkinson’s disease show profound motor dysfunction owing to degeneration of dopamine neurons in their SUBSTANTIA NIGRA. Although the cause of Parkinson’s disease is unknown, increased oxidative stress and mitochondrial dysfunction in dopamine neurons are central to the disease56. There is also a deficit in MITOCHONDRIAL COMPLEX I, which may arise from, or contribute to, increased cellular oxidative stress. Both environmental and genetic factors may sensitize dopamine neurons to age-related increases in oxidative stress and energy deficits56,57. Environmental toxins are implicated — monkeys and people exposed to the toxin 1-methyl-4-phenyl-1,2,3,6-

tetrahydropyridine (MPTP) show Parkinson’s-like symptoms. Analyses of brain tissue from patients with Parkinson’s disease implicate apoptosis-related DNA damage and gene activation in the death of dopamine neurons56. Moreover, levels of Par-4 are selectively increased in dopamine neurons of the substantia nigra before their death, and suppression of Par-4 expression protects dopamine neurons against death12. Caspase-1 inhibition, drugs that suppress macromolecular synthesis, and neurotrophic factors, such as glial cell-derived neurotrophic factor (GDNF), can protect dopamine neurons in Parkinson’s disease models58,59. Mutations in α-synuclein, a component of the brain lesions called LEWY BODIES, are responsible for a small percentage of Parkinson’s disease cases57, and expression of mutant αsynuclein in cultured cells promotes apoptosis60. Huntington’s disease is an inherited disorder in which neurons in the striatum degenerate, resulting in uncontrolled body movements. It is caused by expansions of a trinucleotide (CAG) sequence in the huntingtin gene, producing a protein containing increased polyglutamine repeats61. Studies of patients with Huntington’s disease, and of rodents given the mitochondrial toxin 3-nitropropionic acid, indicate that impaired mitochondrial function and excitotoxic death may be central to the disease11. How does mutant huntingtin promote selective degeneration of striatal neurons? Although this is not known, activation of an apoptotic programme is implicated. Studies of LYMPHOBLASTS from patients with Huntington’s disease have revealed increased stressinduced apoptosis associated with mitochondrial dysfunction and increased caspase-3 activation62, suggesting an adverse effect of mutant huntingtin that is not limited to neurons. Caspase-8 is redistributed to an insoluble fraction in striatal tissue from patients, and expression of mutant huntingtin in cultured cells induces caspase-

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 2 5

REVIEWS

INTRANEURONAL INCLUSIONS

Aggregates of proteins that accumulate in neurons within the cytoplasm or nucleus. INFARCT

Brain tissue surrounding the site of a stroke in which cells die. ISCHAEMIC PENUMBRA

A region of tissue surrounding the necrotic core of an ischaemic infarct in which neurons die primarily by apoptosis.

8-dependent apoptosis63. In addition, huntingtin can be cleaved by caspases, which may promote protein aggregation and neurotoxicity64. Transgenic mouse models of Huntington’s disease recapitulate certain aspects of the human disease, including intracellular inclusions of huntingtin and degeneration of striatal neurons with several features of apoptosis65. Inhibition of caspase-1 was reported to slow disease progression in one of these models66. Expression of mutant huntingtin in the brains of adult rats using viral vectors results in the formation of INTRA67 NEURONAL INCLUSIONS and cell death . However, the formation of nuclear inclusions containing huntingtin may not be required for apoptosis; in fact, such inclusions may be part of a cytoprotective response68. Moreover, wild-type — but not mutant — huntingtin can protect cells against apoptosis by suppressing cell death before mitochondrial dysfunction69. People with amyotrophic lateral sclerosis (ALS) suffer progressive paralysis resulting from degeneration of motor neurons in the spinal cord. Most cases of ALS are sporadic, but some are inherited. This selective degeneration of motor neurons involves increased oxidative stress, overactivation of glutamate receptors and cellular calcium overload70. Production of autoantibodies against voltage-dependent calcium channels may contribute to the pathogenesis of ALS71. Mutations in the antioxidant enzyme Cu/Zn-superoxide dismutase (SOD) are responsible for some inherited cases of ALS, and expression of genes containing these mutations in transgenic mice results in spinal cord pathology remarkably similar to that of patients with ALS72,73. The mutations do not decrease antioxidant activity of the enzyme, but result in the gain of an adverse pro-apoptotic activity that may involve increased peroxidase activity. Through interac-

Figure 4 | Brain tissue section from the hippocampus of a patient who died with Alzheimer’s disease. Examination reveals two prominent abnormalities. First, abnormal, silver-stained (black) accumulations are present in degenerating neurons (arrow). Molecular analysis of such ‘neurofibrillary tangles’ reveals that they are composed of filamentous aggregates of the microtubule-associated protein tau. Second, spherical accumulations of amyloid are present and are often associated with degenerated neurites (arrowhead); the plaques contain aggregates of amyloid-β peptide.

126

tions with hydrogen peroxide or superoxide anion, the mutant enzyme may induce oxidative damage to membranes and disturbances in mitochondrial function that make neurons vulnerable to excitotoxic apoptosis74. In people with ALS, DNA starts to fragment between the nucleosomes (a sign of nuclear apoptosis) in neurons within the spinal-cord anterior horn and motor cortex. The DNA damage is associated with increased mitochondrial localization of Bax and decreased association of Bcl-2 (REF. 75). Levels of Bax, but not Bcl-2, are increased in spinal-cord motor neurons of ALS patients, and a similar expression pattern of Bcl-2 family members is observed in Cu/Zn-SOD mutant mice. The involvement of apoptosis in ALS is further suggested by the fact that overexpression of Bcl-2 and administration of caspase inhibitors delays degeneration and death of motor neurons in Cu/Zn-SOD mutant mice76,77. Stroke. Focal ischaemic stroke, a leading cause of death and disability worldwide, results in brain damage characterized by an INFARCT with a necrotic core, in which all cells die rapidly, and a surrounding ISCHAEMIC PENUMBRA, in which neurons die over days to weeks78. Metabolic compromise, overactivation of glutamate receptors, calcium overload and increased oxyradical production occur in neurons subjected to ischaemia. In addition, complex cytokine cascades involving microglial cells and the cerebrovasculature may be important in promoting or preventing neuronal death after stroke79. Cells in the ischaemic penumbra show DNA damage and activation of the DNA damage-responsive proteins PARP and Ku80. In rodent stroke models, neurons in the ischaemic penumbra show morphological and molecular changes consistent with apoptosis, including caspase activation, expression of pro-apoptotic genes and release of cytochrome c (REF. 78). Signalling pathways involving hydrolysis of membrane phospholipids are implicated in neuronal apoptosis in stroke. Cleavage of membrane sphingomyelin by acidic sphingomyelinase (ASMase) generates the lipid mediator ceramide. Focal cerebral ischaemia in mice induces large increases in ASMase activity and ceramide levels, and production of inflammatory cytokines80. If mice lack ASMase or are given a drug that inhibits production of ceramide, cytokine production is suppressed, brain damage is decreased, and symptoms are improved80. Mice lacking phospholipase-A2 show decreased brain damage after focal cerebral ischaemia, suggesting an important function for one or more lipid mediators generated by this enzyme in ischaemic neuronal injury81. Mice lacking specific caspases or given caspase inhibitors show reduced brain damage after stroke82,83. In addition, delivery of neurotrophic factors known to prevent neuronal apoptosis can prevent neuronal death after stroke — bFGF, BDNF, NGF and sAPP-α are particularly effective3. A pivotal role for mitochondrial alterations in stroke-induced neuron death is suggested by studies showing that lack of mitochondrial Mn-SOD exacerbates focal ischaemic brain injury84, whereas overexpression of Mn-SOD has the opposite effect85. In www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS a

addition, treatment of rats with cyclosporin A decreases the size of the ischaemic infarct14.

APP

Cytoplasm

Cell A γ Aβ

Extracellular space Aβ

sAPPα Aβ Aβ Aβ

Aβ Aβ Aβ Glucose

Cell B MLP

K+

cGMP

Na+ Ca2+

PKG

ACh Glutamate

NF-κB

Aβ Ca2+

APP Aβ

Ins(1,4,5)P3

PS-1 Ins(1,4,5)P3 RYR

Ca2+

OLIGODENDROCYTES

A specific type of glial cell, which forms myelin membranes that insulate axons of neurons and thereby increase impulse conduction velocity.

Ca2+

ROS

Ca2+ Mitochondria

Figure 5 | Mechanisms underlying the pro-apoptotic actions of altered APP processing and presenilin-1 mutations. a | The amyloid precursor protein (APP) can be proteolytically processed in two main ways. Cleavage of APP within the amyloid-β (Aβ) sequence by α-secretase (α) releases a secreted form of APP (sAPPα) from the cell surface. Secreted APPα activates a putative receptor (R) linked to cyclic-GMP production and activation of cGMP-dependent protein kinase (PKG). PKG can then promote opening of K+ channels, resulting in membrane hyperpolarization, and can also activate the transcription factor NF-κB; these effects of sAPPα are believed to mediate its neuron-survivalpromoting properties. A second pathway of APP processing involves cleavages at the N- and C-termini of Aβ by β-secretase (β) and γ-secretase (γ), respectively. This releases Aβ from cells, which, under suitable conditions (high concentration, oxidizing environment), begins to self-aggregate. Under these conditions, Aβ induces membrane lipid peroxidation (MLP), which impairs the function of membrane ion-motive ATPases (Na+ and Ca2+ pumps) and glucose transporters. Neurons are thus vulnerable to apoptosis. b | Presenilin-1 (PS-1) is an integral membrane protein located primarily in the endoplasmic reticulum (ER). Mutations in PS-1 perturb ER Ca2+ homeostasis, resulting in increased release of Ca2+ through ryanodine receptors (RYRs) and inositol-1,4,5 trisphosphate (Ins(1,4,5)P3) receptors. The enhanced Ca2+ release triggers further Ca2+ influx through Ca2+ release channels in the plasma membrane, and this altered Ca2+ homeostasis makes neurons vulnerable to apoptosis and excitotoxicity, and alters APP processing in a manner that increases Aβ production.

Traumatic brain and spinal cord injury. Brain and spinal cord injuries account for most deaths and permanent disabilities in people under 40 years old. Trauma initiates biochemical and molecular events involving many of the same neurodegenerative cascades and neuroprotective signalling mechanisms that occur in the chronic neurodegenerative diseases described above. Histological and immunochemical analyses of the brains from patients who died after traumatic brain injury (TBI) have found apoptosis-related changes in neurons, including the presence of DNA strand breaks, caspase activation and increased Bax and p53 expression86. Sensory, motor and cognitive deficits after TBI in mice are strongly correlated with the numbers of neurons showing apoptotic nuclear damage87. Such neuronal deaths are associated with increased expression of p53 (REF. 88) and of the death receptor Fas and its ligand89. Caspases are also thought to be involved — caspase-3 activity increases markedly in the cerebral cortex of rats in response to TBI, and intraventricular administration of the caspase-3 inhibitor z-DEVD-fmk before injury reduces cell death and improves symptoms, indicating a central function for caspases in this brain injury model90. In addition, mice expressing a dominant-negative inhibitor of caspase-1 show reduced brain damage and free radical production after TBI91. Intraventricular infusion of NGF in rats, beginning 24 hours after TBI, resulted in improved learning and memory, and decreased death of neurons compared with control rats92. Cyclosporin A protects against synaptic dysfunction and cell death in rodent models of TBI, consistent with a key role for mitochondrial membrane permeability in the neurodegenerative process93. Apoptosis, as demonstrated by nuclear DNA fragmentation and caspase activation, was a prominent feature in the spinal cords of 14 out of 15 people who had died three hours to two months after traumatic spinal cord injury (SCI) — apoptosis of OLIGODENDROCYTES in the injury centre and adjacent white matter tracts was particularly prominent94. In rodents, SCI results in neuronal apoptosis, which can be prevented by glutamate-receptor antagonists95. After SCI in rats96, caspase activation occurs in neurons at the injury site within hours, and in oligodendrocytes adjacent to, and distant from, the injury site over a period of days. Studies of SCI in rats and monkeys show apoptosis of oligodendrocytes involving a progressive inflammation-like process97. Thus, apoptosis of both neurons and oligodendrocytes may contribute greatly to the paralysis of patients with SCI. Prospects for treatment and prevention

Most translational research into neurodegenerative diseases is focused on developing drugs that inhibit neuronal dysfunction and death early in the disease process. Better understanding of the molecular and cellular underpinnings of neuronal apoptosis has led to the identification of specific drug targets. One

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 2 7

REVIEWS approach is to block the apoptotic trigger; for example, by suppressing amyloid-β production in Alzheimer’s disease or glutamate-receptor activation in stroke. Other approaches target early premitochondrial alterations, such as drugs that scavenge free radicals, block calcium influx into neurons or inhibit the activity of Par-4. Activation of anti-apoptotic pathways by treatment with neurotrophic factors is another approach. In several cases, work has progressed to clinical trials in patients; for example, bFGF in stroke, BDNF and insulin-like growth factor-1 (IGF-1) in ALS, and vitamin E in Alzheimer’s disease98–101. The next wave of trials will probably include caspase inhibitors, anti-inflammatory drugs and agents that stabilize mitochondrial function, such as cyclosporin A and creatine, which suppresses mitochondrial oxyradical production and prevents ATP depletion. However, there is always the possibility that such therapies will have serious side effects. Although this is an exciting era in the field of neurodegenerative disorders, with genetic and molecular biological approaches rapidly advancing our understanding of disease pathogenesis, there are no truly effective treatments for any of the disorders described above. We therefore need to develop methods of preventing or reducing risk for neurodegenerative diseases. One such approach is available now that, on the basis of existing data, is likely to be effective. Dietary restriction (reduced calorie intake with maintenance

7. 8.

9. 10.

11.

12.

13.

128

Wyllie, A. H. Apoptosis and carcinogenesis. Eur. J. Cell Biol. 73, 189–197 (1997). Provides a historical overview of the morphological and biochemical features of apoptosis. Oppenheim, R. W. Cell death during development of the nervous system. Annu. Rev. Neurosci. 14, 453–501 (1991). Mattson, M. P. & Lindvall, O. in The Aging Brain Vol. 2 (eds Mattson, M. P. & Geddes, J. W.) 299–345 (JAI Press, Greenwich, Connecticut, 1997). McKay, S. E., Purcell, A. L. & Carew, T. J. Regulation of synaptic function by neurotrophic factors in vertebrates and invertebrates: implications for development and learning. Learn. Mem. 6, 193–215 (1999). Ankarcrona, M. et al. Glutamate-induced neuronal death: a succession of necrosis or apoptosis depending on mitochondrial function. Neuron 15, 961–973 (1995). Glazner, G. W., Chan, S. L., Lu, C. & Mattson, M. P. Caspase-mediated degradation of AMPA receptor subunits: a mechanism for preventing excitotoxic necrosis and ensuring apoptosis. J. Neurosci. 20, 3641–3649 (2000). Choi, D. W. Excitotoxic cell death. J. Neurobiol. 23, 1261–1276 (1992). Wong, P. C., Rothstein, J. D. & Price, D. L. The genetic and molecular mechanisms of motor neuron disease. Curr. Opin. Neurobiol. 8, 791–799 (1998). Sastry, P. S. & Rao, K. S. Apoptosis and the nervous system. J. Neurochem. 74, 1–20 (2000). Mattson, M. P. Modification of ion homeostasis by lipid peroxidation: roles in neuronal degeneration and adaptive plasticity. Trends Neurosci. 21, 53–57 (1998). Beal, M. F. Aging, energy, and oxidative stress in neurodegenerative diseases. Ann. Neurol. 38, 357–366 (1995). Duan, W., Zhang, Z., Gash, D. M. & Mattson, M. P. Participation of prostate apoptosis response-4 in degeneration of dopaminergic neurons in models of Parkinson’s disease. Ann. Neurol. 46, 587–597 (1999). Kroemer, G., Dallaporta, B. & Resche-Rigon, M. The mitochondrial death/life regulator in apoptosis and necrosis. Annu. Rev. Physiol. 60, 619–642 (1998). Detailed review of the role of mitochondria in

of micronutrient nutrition) is known to extend the lifespan of all mammalian species examined, and reduces development of various age-related diseases. Epidemiological data indicate that low calorie intake is associated with reduced risk for Parkinson’s disease102. Importantly, the neurons of rodents maintained on dietary restriction are more resistant to apoptosis and show improved symptoms in experimental models of neurodegenerative disorders and stroke34,35,103. The mechanism by which dietary restriction benefits neurons may involve decreased levels of mitochondrial oxyradical production, or a mild metabolic stress response, in which neurons respond to the stress of reduced energy availability by increasing the production of stress proteins and neurotrophic factors34,35,103. Developing further means of increasing neuronal resistance through dietary and perhaps behavioural104 manipulations will provide an important complement to drugs that may delay neurodegeneration in patients who are already showing symptoms of disease. Links DATABASE LINKS Bcl-2 | Bcl-xL | Bax | Bad | Par-4 |

apoptosis. 14. Matsumoto, S., Friberg, H., Ferrand-Drake, M. & Wieloch, T. Blockade of the mitochondrial permeability transition pore diminishes infarct size in the rat after transient middle cerebral artery occlusion. J. Cereb. Blood Flow Metab. 19, 736–741 (1999). 15. Pellegrini, M. & Strasser, A. A portrait of the Bcl-2 protein family: life, death, and the whole picture. J. Clin. Immunol. 19, 365–377 (1999). 16. Martinou, J. C. et al. Overexpression of Bcl-2 in transgenic mice protects neurons from naturally occurring cell death and experimental ischemia. Neuron 13, 1017–1030 (1994). Direct evidence that Bcl-2 can prevent death of neurons in vivo during development and in an experimental stroke model. 17. Guo, Q. et al. Alzheimer’s presenilin mutation sensitizes neural cells to apoptosis induced by trophic factor withdrawal and amyloid β-peptide: involvement of calcium and oxyradicals. J. Neurosci. 17, 4212–4222 (1998). 18. White, F. A., Keller-Peck, C. R., Knudson, C. M., Korsmeyer, S. J. & Snider, W. D. Widespread elimination of naturally occurring neuronal death in Bax-deficient mice. J. Neurosci. 18, 1428–1439 (1998). Direct evidence that the pro-apoptotic protein Bax is essential in natural cell death during development of the nervous system. 19. Green, D. R. & Reed, J. C. Mitochondria and apoptosis. Science 28, 1309–1312 (1998). 20. Chan, S. L. & Mattson, M. P. Caspase and calpain substrates: roles in synaptic plasticity and cell death. J. Neurosci. Res. 58, 167–190 (1999). 21. Raoul, C., Pettmann, B. & Henderson, C. E. Active killing of neurons during development and following stress: a role for p75(NTR) and Fas? Curr. Opin. Neurobiol. 10, 111–117 (2000). 22. Nagata, S. Apoptotic DNA fragmentation. Exp. Cell Res. 256, 12–18 (2000). 23. Guo, Q. et al. Par-4 is a mediator of neuronal degeneration associated with the pathogenesis of Alzheimer disease. Nature Med. 4, 957–962 (1998). Initial description and characterization of the involvement of Par-4 in neuronal apoptosis in

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

experimental models of developmental cell death and Alzheimer’s disease. Mattson, M. P., Duan, W., Chan, S. L. & Camandola, S. Par-4: an emerging pivotal player in neuronal apoptosis and neurodegenerative disorders. J. Mol. Neurosci. 13, 17–30 (1999). Liu, J. P. Studies of the molecular mechanisms in the regulation of telomerase activity. FASEB J. 13, 2091–2104 (1999). Fu, W., Killen, M., Pandita, T. & Mattson, M. P. The catalytic subunit of telomerase is expressed in developing brain neurons and serves a cell survival-promoting function. J. Mol. Neurosci. 14, 3–15 (2000). Fu, W., Begley, J. G., Killen, M. W. & Mattson, M. P. Antiapoptotic role of telomerase in pheochromocytoma cells. J. Biol. Chem. 274, 7264–7271 (1999). Tamatani, M., Ogawa, S., Nunez, G. & Tohyama, M. Growth factors prevent changes in Bcl-2 and Bax expression and neuronal apoptosis induced by nitric oxide. Cell Death Differ. 5, 911–919 (1998). Hagg, T. & Varon, S. Ciliary neurotrophic factor prevents degeneration of adult rat substantia nigra dopaminergic neurons in vivo. Proc. Natl Acad. Sci. USA 90, 6315–6319 (1993). Middleton, G. et al. Cytokine-induced nuclear factor kappa B activation promotes the survival of developing neurons. J. Cell Biol. 148, 325–332 (2000). Mattson, M. P. & Camandola, S. NF-κB in neurodegenerative disorders. J. Clin. Invest. 61,134–139 (2000). Yu, Z. F., Zhou, D., Bruce-Keller, A. J., Kindy, M. S. & Mattson, M. P. Lack of the p50 subunit of NF-κB increases the vulnerability of hippocampal neurons to excitotoxic injury. J. Neurosci. 19, 8856–8865 (1999). Aschner, M., Allen, J. W., Kimelberg, H. K., LoPachin, R. M. & Streit, W. J. Glial cells in neurotoxicity development. Annu. Rev. Pharmacol. Toxicol. 39, 151–173 (1999). Yu, Z. F. & Mattson, M. P. Dietary restriction and 2deoxyglucose administration reduce focal ischemic brain damage and improve behavioral outcome: evidence for a preconditioning mechanism. J. Neurosci. Res. 5, 830–839 (1999). Lee, J., Duan, W., Long, J. M., Ingram, D. K. & Mattson,

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS

36.

37.

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

55.

56. 57.

58.

M. P. Dietary restriction increases survival of newlygenerated neural cells and induces BDNF expression in the dentate gyrus of rats. J. Mol. Neurosci. (in the press). Berridge, M. J., Lipp, P. & Bootman, M. D. The versatility and universality of calcium signalling. Nature Rev. Mol. Cell Biol. 1, 11–21 (2000). Yano, S., Tokumitsu, H. & Soderling, T. R. Calcium promotes cell survival through CaM-K kinase activation of the protein-kinase-B pathway. Nature 396, 584–587 (1998). Hu, S. C., Chrivia, J. & Ghosh, A. Regulation of CBPmediated transcription by neuronal calcium signaling. Neuron 22, 799–808 (1999). Furukawa, K. et al. The actin-severing protein gelsolin modulates calcium channel and NMDA receptor activities and vulnerability to excitotoxicity in hippocampal neurons. J. Neurosci. 17, 8178–8186 (1997). Furukawa, K., Barger, S. W., Blalock, E. M. & Mattson, M. P. Activation of K+ channels and suppression of neuronal activity by secreted β-amyloid-precursor protein. Nature 379, 74–78 (1996). Guenette, S. Y, & Tanzi, R. E. Progress toward valid transgenic mouse models for Alzheimer’s disease. Neurobiol. Aging 20, 201–211 (1999). Sathasivam, K. et al. Transgenic models of Huntington’s disease. Phil. Trans. R. Soc. Lond. B 354, 963–969 (1999). Borchelt, D. R., Wong, P. C., Sisodia, S. S. & Price, D. L. Transgenic mouse models of Alzheimer’s disease and amyotrophic lateral sclerosis. Brain Pathol. 8, 735–757 (1998). Cummings, J. L., Vinters, H. V., Cole, G. M. & Khachaturian, Z. S. Alzheimer’s disease: etiologies, pathophysiology, cognitive reserve, and treatment opportunities. Neurology 51, S2–S17 (1998). Haass, C. & De Strooper, B. The presenilins in Alzheimer’s disease — proteolysis holds the key. Science 286, 916–919 (1999). Su, J. H., Anderson, A. J., Cummings, B. & Cotman, C. W. Immunocytochemical evidence for apoptosis in Alzheimer’s disease. Neuroreport 5, 2529–2533 (1994). Masliah, E., Mallory, M., Alford, M., Tanaka, S. & Hansen, L. A. Caspase dependent DNA fragmentation might be associated with excitotoxicity in Alzheimer disease. J. Neuropathol. Exp. Neurol. 57, 1041–1052 (1998). Suzuki, T. et al. Molecular cloning of a novel apoptosisrelated gene, human Nap1 (NCKAP1), and its possible relation to Alzheimer disease. Genomics 63, 246–254 (2000). Loo, D., Copani, A., Pike, C., Whittemore, E., Walencewicz, A. & Cotman, C. W. Apoptosis is induced by β-amyloid in cultured central nervous system neurons. Proc. Natl Acad. Sci. USA 90, 7951–7955 (1993). Weidemann, A. et al. Proteolytic processing of the Alzheimer’s disease amyloid precursor protein within its cytoplasmic domain by caspase-like proteases. J. Biol. Chem. 274, 5823–5829 (1999). Chen, Y., McPhie, D. L., Hirschberg, J. & Neve, R. L. The amyloid precursor protein-binding protein APP-BP1 drives the cell cycle through the S–M checkpoint and causes apoptosis in neurons. J. Biol. Chem. 275, 8929–8935 (2000). Lu, D. C. et al. A second cytotoxic proteolytic peptide derived from amyloid precursor protein. Nature Med. 6, 397–404 (2000). Guo, Q. et al. Neurotrophic factors (activity-dependent neurotrophic factor (ADNF) and basic fibroblast growth factor (bFGF)) interrupt excitotoxic neurodegenerative cascades promoted by a presenilin-1 mutation. Proc. Natl Acad. Sci. USA 96, 4125–4130 (1999). Uses knock-in mice to identify the gain-of-function action of mutant presenilin-1 as a pro-apoptotic activity that results from perturbed cellular calcium homeostasis and can be suppressed by neurotrophic factors. Mattson, M. P. et al. Calcium signaling in the ER: its role in neuronal plasticity and neurodegenerative disorders. Trends Neurosci. 23, 222–229 (2000). Vito, P., Lacana, E. & d’Adamio, L. Interfering with apoptosis: Ca2+-binding protein ALG-2 and Alzheimer’s disease gene ALG-3. Science 271, 521–525 (1996). Jenner, P. & Olanow, C. W. Understanding cell death in Parkinson’s disease. Ann. Neurol. 44, S72–S84 (1998). Polymeropoulos, M. H. Autosomal dominant Parkinson’s disease and alpha-synuclein. Ann. Neurol. 44, S63–S64 (1998). Klevenyi, P. et al. Transgenic mice expressing a dominant negative mutant interleukin-1β converting enzyme show resistance to MPTP neurotoxicity. Neuroreport 10, 635–638 (1999).

59. Gash, D. M. et al. Functional recovery in parkinsonian monkeys treated with GDNF. Nature 380, 252–255 (1996). 60. El-Agnaf, O. M. et al. Aggregates from mutant and wildtype α-synuclein proteins and NAC peptide induce apoptotic cell death in human neuroblastoma cells by formation of β-sheet and amyloid-like filaments. FEBS Lett. 440, 71–75 (1998). 61. Brandt, J. et al. Trinucleotide repeat length and clinical progression in Huntington’s disease. Neurology 46, 527–531 (1996). 62. Sawa, A. et al. Increased apoptosis of Huntington disease lymphoblasts associated with repeat length-dependent mitochondrial depolarization. Nature Med. 5, 1194–1198 (1999). 63. Sanchez, I. et al. Caspase-8 is required for cell death induced by expanded polyglutamine repeats. Neuron 22, 623–633 (1999). 64. Wellington, C. L. et al. Inhibiting caspase cleavage of huntingtin reduces toxicity and aggregate formation in neuronal and non-neuronal cells. J. Biol. Chem. 275, 19831–19838 (2000). 65. Leavitt, B. R., Wellington, C. I. & Hayden, M. R. Recent insights into the molecular pathogenesis of Huntington disease. Semin. Neurol. 19, 385–395 (1999). 66. Ona, V. O. et al. Inhibition of caspase-1 slows disease progression in a mouse model of Huntington’s disease. Nature 399, 263–267 (1999). 67. Senut, M. C., Suhr, S. T., Kaspar, B. & Gage, F. H. Intraneuronal aggregate formation and cell death after viral expression of expanded polyglutamine tracts in the adult rat brain. J. Neurosci. 20, 219–229 (2000). 68. Kim, M. et al. Mutant huntingtin expression in clonal striatal cells: dissociation of inclusion formation and neuronal survival by caspase inhibition. J. Neurosci. 19, 964–973 (1999). Provides insight into the pathogenic action of polyglutamine-repeat huntingtin in neuronal apoptosis. 69. Rigamonti, D. et al. Wild-type huntingtin protects from apoptosis upstream of caspase-3. J. Neurosci. 20, 3705–3713 (2000). 70. Cookson, M. R. & Shaw, P. J. Oxidative stress and motor neurone disease. Brain Pathol. 9, 165–186 (1999). 71. Smith, R. G. et al. Autoimmunity and ALS. Neurology 47, S40–S45 (1996). 72. Gurney, M. E. et al. Motor neuron degeneration in mice that express a human Cu,Zn superoxide dismutase mutation. Science 264, 1772–1775 (1994). 73. Wong, P. C. et al. An adverse property of a familial ALSlinked SOD1 mutation causes motor neuron disease characterized by vacuolar degeneration of mitochondria. Neuron 14, 1105–1116 (1995). 74. Kruman, I., Pedersen, W. A. & Mattson, M. P. ALS-linked Cu/Zn-SOD mutation increases vulnerability of motor neurons to excitotoxicity by a mechanism involving increased oxidative stress and perturbed calcium homeostasis. Exp. Neurol. 160, 28–39 (1999). 75. Martin, L. J., Price, A. C., Kaiser, A., Shaikh, A. Y. & Liu, Z. Mechanisms for neuronal degeneration in amyotrophic lateral sclerosis and in models of motor neuron death. Int. J. Mol. Med. 5, 3–13 (2000). 76. Li, M. et al. Functional role of caspase-1 and caspase-3 in an ALS transgenic mouse model. Science 288, 335–339 (2000). 77. Kostic, V., Jackson-Lewis, V., de Bilbao, F., DuboisDauphin, M. & Przedborski, S. Bcl-2: prolonging life in a transgenic mouse model of familial amyotrophic lateral sclerosis. Science 277, 559–562 (1997). 78. Dirnagl, U., Iadecola, C. & Moskowitz, M. A. Pathobiology of ischaemic stroke: an integrated view. Trends Neurosci. 22, 391–397 (1999). Details molecular and cellular mechanisms in the neuronal death in ischaemic stroke. 79. Lipton, P. Ischemic cell death in brain neurons. Physiol. Rev. 79, 1431–568 (1999). 80. Yu, Z. et al. Pivotal role for acidic sphingomyelinase in cerebral ischemia-induced ceramide and cytokine production, and neuronal death. J. Mol. Neurosci. (in the press). 81. Bonventre, J. V. et al. Reduced fertility and postischaemic brain injury in mice deficient in cytosolic phospholipase A2. Nature 390, 622–625 (1997). 82. Schielke, G. P., Yang, G. Y., Shivers, B. D. & Betz, A. L. Reduced ischemic brain injury in interleukin-1β converting enzyme-deficient mice. J. Cereb. Blood Flow Metab. 18, 180–185 (1998). 83. Hara, H. et al. Inhibition of interleukin 1β converting enzyme family proteases reduces ischemic and excitotoxic neuronal damage. Proc. Natl Acad. Sci. USA 94, 2007–2012 (1997).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

84. Murakami, K. et al. Mitochondrial susceptibility to oxidative stress exacerbates cerebral infarction that follows permanent focal cerebral ischemia in mutant mice with manganese superoxide dismutase deficiency. J. Neurosci. 18, 205–213 (1998). 85. Keller, J. N. et al. Mitochondrial manganese superoxide dismutase prevents neural apoptosis and reduces ischemic brain injury: suppression of peroxynitrite production, lipid peroxidation, and mitochondrial dysfunction. J. Neurosci. 18, 687–697 (1998). 86. Clark, R. S. et al. Increases in Bcl-2 and cleavage of caspase-1 and caspase-3 in human brain after head injury. FASEB J. 13, 813–821 (1999). 87. Fox, G. B., Fan, L., Levasseur, R. A. & Faden, A. I. Sustained sensory/motor and cognitive deficits with neuronal apoptosis following controlled cortical impact brain injury in the mouse. J. Neurotrauma 15, 599–614 (1998). 88. Napieralski, J. A., Raghupathi, R. & McIntosh, T. K. The tumor-suppressor gene, p53, is induced in injured brain regions following experimental traumatic brain injury. Mol. Brain Res. 71, 78–86 (1999). 89. Beer, R. et al. Expression of Fas and Fas ligand after experimental traumatic brain injury in the rat. J. Cereb. Blood Flow Metab. 20, 669–677 (2000). 90. Yakovlev, A. G. et al. Activation of CPP32-like caspases contributes to neuronal apoptosis and neurological dysfunction after traumatic brain injury. J. Neurosci. 17, 7415–7424 (1997). 91. Fink, K. B. et al. Reduction of post-traumatic brain injury and free radical production by inhibition of the caspase-1 cascade. Neuroscience 94, 1213–1218 (1999). 92. Sinson, G., Perri, B. R., Trojanowski, J. Q., Flamm, E. S. & McIntosh, T. K. Improvement of cognitive deficits and decreased cholinergic neuronal cell loss and apoptotic cell death following neurotrophin infusion after experimental traumatic brain injury. J. Neurosurg. 86, 511–518 (1997). 93. Albensi, B. C., Sullivan, P. G., Thompson, M. B., Scheff, S. W. & Mattson, M. P. Cyclosporine ameliorates traumatic brain injury-induced alterations of hippocampal synaptic plasticity. Exp. Neurol. 162, 385–389 (2000). 94. Emery, E. et al. Apoptosis after traumatic human spinal cord injury. J. Neurosurg. 89, 911–920 (1998). 95. Wada, S. et al. Apoptosis following spinal cord injury in rats and preventative effect of N-methyl-D-aspartate receptor antagonist. J. Neurosurg. 91, 98–104 (1999). 96. Springer, J. E., Azbill, R. D. & Knapp, P. E. Activation of the caspase-3 apoptotic cascade in traumatic spinal cord injury. Nature Med. 5, 943–946 (1999). 97. Crowe, M. J., Bresnahan, J. C., Shuman, S. L., Masters, J. N. & Beattie, M. S. Apoptosis and delayed degeneration after spinal cord injury in rats and monkeys. Nature Med. 3, 73–76 (1997). 98. Ay, H., Ay, I., Koroshetz, W. J. & Finklestein, S. P. Potential usefulness of basic fibroblast growth factor as a treatment for stroke. Cerebrovasc. Dis. 9, 131–135 (1999). 99. The BDNF Study Group. A controlled trial of recombinant methionyl human BDNF in ALS: The BDNF Study Group (Phase III). Neurology 52, 1427–1433 (1999). 100. Borasio, G. D. et al. A placebo-controlled trial of insulin-like growth factor-I in amyotrophic lateral sclerosis. European ALS/IGF-I Study Group. Neurology 51, 583–586 (1998). 101. Grundman, M. Vitamin E and Alzheimer disease: the basis for additional clinical trials. Am. J. Clin. Nutr. 71, 630S–636S (2000). 102. Logroscino, G. et al. Dietary lipids and antioxidants in Parkinson’s disease: a population-based, case-control study. Ann. Neurol. 39, 89–94 (1996). 103. Duan, W. & Mattson, M. P. Dietary restriction and 2deoxyglucose administration improve behavioral outcome and reduce degeneration of dopaminergic neurons in models of Parkinson’s disease. J. Neurosci. Res. 57, 195–206 (1999). 104. Ohlsson, A. L. & Johansson, B. B. Environment influences functional outcome of cerebral infarction in rats. Stroke 26, 644–649 (1995). 105. Mattson, M. P. & Duan, W. Apoptotic biochemical cascades in synaptic compartments: roles in adaptive plasticity and neurodegenerative disorders. J. Neurosci. Res. 58, 152–166 (1999). Reviews the evidence for and implications of apoptosis-related mechanisms in synaptic remodelling and neuronal cell death. 106. Ivins, K. J., Bui, E. T. & Cotman, C. W. Beta-amyloid induces local neurite degeneration in cultured hippocampal neurons: evidence for neuritic apoptosis. Neurobiol. Dis. 5, 365–378 (1998). 107. Albensi, B. C. & Mattson, M. P. Evidence for the nvolvement TNF and NF-κB in hippocampal synaptic plasticity. Synapse 35, 151–159 (2000).

VOLUME 1 | NOVEMBER 2000 | 1 2 9

REVIEWS

GRABBING THE CAT BY THE TAIL: MANIPULATING MOLECULES ONE BY ONE Carlos Bustamante*‡§, Jed C. Macosko‡ and Gijs J. L. Wuite§ Methods for manipulating single molecules are yielding new information about both the forces that hold biomolecules together and the mechanics of molecular motors. We describe here the physical principles behind these methods, and discuss their capabilities and current limitations.

OPTICAL TWEEZERS

Focused photon fields. PIEZO-ELECTRIC

Describes a device that expands or contracts as a voltage is applied to an internal crystal.

*Howard Hughes Medical Institute and ‡Departments of Molecular and Cell Biology and §Physics, University of California, Berkeley, California 94720, USA. e-mails: carlos@alice.berkeley.edu; jed@alice.berkeley.edu; gwuite@nat.vu.nl

130

“I know a man who grabbed a cat by the tail and learned 40 per cent more about cats than the man who didn’t.” Mark Twain

Until quite recently, scientists could only investigate chemical processes on a bulk level. The forces and stresses that molecules exert on each other or develop in the course of reactions were not directly measurable. During the past few years, this situation has changed rapidly thanks to the development of methods for manipulating single molecules1–4. Methods such as OPTI5 CAL TWEEZERS and scanning force microscopy (SFM) are making it possible to follow, in real-time and at a singlemolecule level, the movements, forces and strains that develop during the course of a reaction. These methods can be used to measure directly the forces that hold together molecular structures. They can also be used to exert external forces to modify the extent and even alter the fate of reactions in the hope of discovering rules that govern the inter-conversion of mechanical and chemical energy in these processes. This area of research can rightly be called ‘mechanochemistry’, and includes biochemical processes as diverse as protein folding6, DNA elasticity7–9, the protein-induced bending of DNA10, the stress-induced catalysis of enzymes11, the behaviour of molecular motors12–15, and even the ubiquitous processes of protein–protein recognition16. Here we focus on the current capabilities and limitations of single-molecule manipulation methods, and provide guidelines for choosing the most appropriate method for a given problem.

Choosing the appropriate method

All single-molecule manipulation methods require two basic elements: a probe, which is usually of microscopic dimensions, that can generate or detect forces and displacements; and a way to spatially locate the molecules. As summarized in TABLE 1, the relevant force ranges, minimum displacements, probe stiffness, applications and practical advantages of each technique vary significantly. Mechanical transducers

Mechanical force transducers apply or sense forces through the displacement of a bendable beam. The most common examples are SFM cantilevers5 (FIG. 1) and microneedles12 (FIG. 2). The spatial control of transducers can be accomplished efficiently by PIEZO-ELECTRIC positioners (FIG. 1a). Mechanical transducers have been used to investigate systems ranging from protein unfolding6 and cell motility17 to forces generated by motor proteins12. Mechanical transducers possess a linear response over a broad range of displacement and forces. Two important factors determine how mechanical transducers interact with single molecules: their size and stiffness. The effect of these parameters is described below and in BOX 1. SFM cantilevers. Microfabricated cantilevers are available in a wide variety of sizes, shapes and materials. SFM devices are also commercially available, and some are specifically designed for manipulating single molecules18 (FIG. 1b). The advantages of SFM are its high spatial range and sensitivity, its throughput (the ability to study many single molecules on a surface) and versatility. www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS For example, SFM can be used both as an imaging instrument and as a manipulation device, as first shown by Müller et al.19 and further exploited by Oesterhelt and colleagues20 (see below). SFM cantilevers have stiffness (κ) ranging from 10–3 to 100 N m–1. Stiffer cantilevers have lower sensitivities, as force is always detected by measuring a displacement that is inversely proportional to the stiffness. They are nonetheless useful when a given process (conformational changes, for example) requires the application of high forces. Although stiffer cantilevers experience correspondingly large force fluctuations owing to thermal motion (for a typical cantilever with κ = 0.06 N m–1, the root-mean-square force fluctuation is ~16 pN; BOX 1), the signal-to-noise ratio of the measurement is independent of the stiffness of the cantilever2,21. In addition, drift over time caused by the thermal expansion and nonlinear voltage response of the piezo-electric crystals can further compromise the control of force on the sample. This is not a serious limitation, however, as ways to compensate for drift have been developed22. A promising development in SFM methodology is the fabrication of smaller, but still soft, cantilevers23. Their small physical dimensions allow them to have higher sensitivity and faster response times24. Being soft, these small cantilevers allow high spatial resolution without a subsequent increase in force fluctuations. This higher spatial resolution stems from the distribution of thermal fluctuations over a broader frequency range, thus decreasing the noise at biologically relevant frequencies (BOX 1). The scanning force microscope has been used successfully to study the mechanism of unfolding in proteins. Fernandez and co-workers6 unfolded a protein made of repeating immunoglobulin-type domains by pulling it with an SFM cantilever. To obtain refolding rates, they allowed the protein to refold for a varying amount of time before it was re-extended. These authors also determined how the unfolding force varies with pulling speed. Extrapolation of these data to zero force yielded a pulling speed of 0.013 nm s-1, which, when divided by the extension required to unfold one

HYDRODYNAMIC FIELD

A force field resulting from the momentum imparted by molecules in a flowing aqueous solution. PHOTON FIELD

A force field resulting from the momentum imparted by photons in a beam of light.

domain (28.4 nm), matches the unfolding rate determined by chemical denaturation experiments (4.9 ×10–4 s–1). So, mechanical unfolding experiments measure the same interactions as their chemical analogues, and have the potential to follow secondary-structureunfolding events. Microneedles. Because of their dimensions (typically 50–500 µm long and 0.1–1 µm thick), glass microneedles are usually softer than cantilevers (TABLE 1). This property gives them an advantage over SFM cantilevers for probing delicate biological systems. Microneedles are not commercially available, however, and the devices to detect their displacement are less standardized than in SFM. Two general approaches for displacement detection have been reported: imaging the microneedle itself 12,25–28; and using a chemically etched optical fibre that projects light from its tip onto a photodiode29–31 as a microneedle. This latter method has been used to measure the stretching of twisted DNA31 and to study the binding of RecA to stretched DNA30. Microneedles have also been used successfully to quantify the force that myosin exerts on F-actin12,25,27. In the first of these experiments, fluorescently labelled Factin filaments were attached to a microneedle coated with monomeric myosin (FIG. 2). This filament was brought into contact with a myosin-coated surface. The subsequent bending of the microneedle corresponded to the combined force (9.6 pN) exerted by no more than 53 interacting myosin heads (at least 0.2 pN per myosin head). This average force is comparable to that exerted in a muscle during contraction32 (~1 pN). More recent studies, using complex experimental geometries (FIG. 3a), have determined the step size of myosin on actin to be 5.3 nm (REF. 25), and the force generated during the actual power stroke to be ~3–5 pN (REFS 21,27). External field manipulators

External fields provide another approach to the manipulation of single molecules. Examples are HYDRODYNAMIC, magnetic and PHOTON fields. Unlike mechanical transducers, external fields act on molecules from a distance. These fields can be used to exert forces on

Table 1 | Overview of single-molecule manipulation methods Methods

Fmin–max (N)§ –11

–7

Xmin (m)§ –10

Stiffness (N m–1)

Applications

Practical advantages 6,64

Cantilevers*

10 –10

0.001–100

Protein/polysaccharides Bond strength65,66

High spatial resolution Commercially available

Microneedles*

10–12–10–10

10–9

10–6–1

Myosin motor force12 DNA/titin strength26,28

Good operator control Soft spring constant

Flow field‡

10–13–10–9

10–8

n.a.

DNA dynamics38 RNA polymerase36

Rapid buffer exchange Simplicity of design

Magnetic field‡

10–14–10–11

10–8

n.a.

DNA entropic elasticity8 Topoisomerase activity41

Specificity to magnets Ability to induce torque

Photon field‡

10–13–10–10

10–9

10–10–10–3

Protein motors13,14 Protein unfolding52

Specific manipulation High force resolution

*Mechanical transducers: probes are bendable beams; spatial location is by beam deflection. ‡External field manipulators: probes are microscopic beads; spatial location is by bead displacement. §These numbers represent only empirical, not absolute limits. (Fmin–max, force range; Xmin, minimum displacement.)

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 3 1

REVIEWS

Laser

4 Lens 5 Position-sensitive detector

3 4

Cantilever Sample

1 2 3

200 pN

1 Piezo-electric scanner

20 nm

Figure 1 | Applications of the scanning force microscope (SFM). a | The principal SFM components. Laser light is focused onto the back of a cantilever that ends with a nanometre-scale tip. The reflection and corresponding position of the tip is detected by a position-sensitive photodiode. A piezo-electric scanner moves the sample in all directions, enabling the tip to scan topography or to extend molecules attached to the surface. b | Diagrams and force curves showing the mechanical unfolding of repeating immunoglobulin-like domains6,64. As the distance between the surface and tip increases (from state 1 to state 2), the molecule extends and generates a restoring force that bends the cantilever. When a domain unfolds (state 3), the free length of the protein increases, relaxing the force on the cantilever. Further extension again results in a restoring force (state 4). The last peak represents the final extension of the unfolded molecule before detachment from the SFM tip (state 5).

molecules by acting either on the molecules themselves, or through ‘handles’ such as glass beads, polystyrene beads or metallic particles attached to the molecules. The various external fields give different degrees of control over the magnitude and stability of the applied forces. Flow fields. Flow fields exert forces on objects through the transfer of momentum from the fluid to the object (FIG. 3b). In LAMINAR FLOW, the drag force between a moving liquid with viscosity η and velocity v and a stationary bead handle of radius r can be calculated using Bendable microneedle

Coverslip

Actin filament LAMINAR FLOW

A flow of molecules in which neighbouring molecules have linearly dependent velocities, that is, not a turbulent flow. STOKES’S LAW

Fdrag = 6πrηv

132

Myosin molecules

Figure 2 | Using a microneedle to measure the force of myosin acting on actin. A bendable microneedle coated in myosin heads (not shown) catches an actin filament. This filament is brought into contact with a glass coverslip coated in myosin molecules. In the presence of ATP, the myosin drags the actin filament across the coverslip and generates a force on the microneedle, which is observable by videofluorescence microscopy.

STOKES’S LAW. For spheres of any size in water, Stokes’s law remains valid for forces up to ~10 nN, beyond which turbulence becomes a factor. Forces up to ~10 nN can therefore be applied reliably. The advantages of using flow fields for single-molecule experiments include the fact that the liquids surrounding the tethered macromolecule can be easily replaced. This feature is important in many single-molecule studies of enzymes, which require varying buffer conditions. Moreover, the flow can be used to introduce new beads or biomolecules. To calculate the drag force, the size of the bead handles and the actual flow velocity must be known. Furthermore, an accurate drag-force calculation requires the bead handle to be stationary in the flow, such as in the case of a bead tethered to a surface by a piece of DNA (FIG. 3b). In addition, as single-molecule experiments are almost always carried out inside a microchamber, force determination should also take into account the modification of Stokes’s law, owing to the coupling, through water, between the bead and the boundaries of the microchamber8,33,34. Finally, it should be kept in mind that, because the drag coefficient of an object scales largely with its longest dimension35, often the friction coefficient of a long polymer such as DNA for example is comparable to that of the bead handle and cannot be neglected. The flow-field manipulation technique was demonstrated in the earliest single-molecule study of DNA elasticity8. In that study, biotinylated DNA was attached by one end to a streptavidin-coated glass surface and its other end was attached to a magnetic bead (FIG. 3b). With this set-up, different tensions were applied to single

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS

RADIATION PRESSURE

The pressure on an object that arises from photon collisions rather than from bombarding molecules.

DNA molecules by changing the flow rate in one direction relative to an orthogonal magnetic force. These results have led to a precise description of DNA elasticity7,9. A recent study of transcriptional pausing and arrest of RNA polymerase (BOX 2) used computer-controlled fluid flow, which applied force to sub-piconewton resolution on active single RNA polymerases36. The flow-field technique has also been used by Chu and co-

Box 1 | Balancing signal, thermal noise and time resolution Amplitude 2 Hz –1

Every object in solution is Signal bombarded constantly by surrounding molecules. As a result, a spring-like device such as a cantilever, a microneedle or Area=kBT/κ a bead in an optical trap Area=kBT/κ experiences a mean-square ωc2 Frequency B ω c1 displacement noise, <∆x2>, proportional to the temperature, T, and inversely related to spring stiffness, κ (<∆x2> = kBT/κ, where kB is the Boltzmann constant). This is the so-called equipartition theorem. If the device is linear, its corresponding mean-square force noise is <∆F 2> = κkBT. Stiff mechanical transducers such as cantilevers (κ = 0.06 N m–1) experience larger force fluctuations but smaller displacement fluctuations than do soft transducers such as a bead in an optical trap (κ = 10–4 N m–1). As it happens, fluctuations are not spread out uniformly over all frequencies. The spectrum of fluctuations of an object is determined by the proportionality that exists between its ability to absorb thermal energy and its ability to dissipate it by friction. This result is embodied in the ‘fluctuation-dissipation theorem’: ∆x 2 (ω )

2kBT

(1)

(ω c2 +ω 2)

In EQN 1, <∆x2(ω)>eq is the mean-square amplitude of fluctuations per unit frequency of the device at frequency ω, and γ is the friction coefficient of the device. The ‘corner frequency’, ωc = κ/γ, is the frequency above which the system cannot respond to an external stimulus. The corner frequency sets a limit to the rate at which processes can be observed and measured experimentally. A 1-µm diameter bead in a typical optical trap has a corner frequency of 1,000 Hz, whereas a commercial cantilever, 100 µm long and 10 µm wide, has a corner frequency of 6,000 Hz and thus data can be gathered six times faster. The figure shows the effect of increased corner frequency on fluctuation distribution. For the same stiffness κ and bandwidth B, the signal-to-noise ratio of the measurement will be higher for the transducer having the larger corner frequency ωc2 (that is, the transducer with the smaller dimensions) than for ωc1. Measurements are often performed in a narrow band (bandwidth B) around the frequency of the signal (see figure). Assume that a molecule attached to a transducer can generate a force F. Then, the signal-to-noise ratio (SNR) of the measurement, for B << ωc, is given by EQN 2 : S/N = F/ √2ykBTB

(2)

In general, therefore, the SNR can be increased by decreasing the bandwidth of the measurement (that is, by averaging the signal over longer times). Decreasing the bandwidth reduces the time resolution of the measurements, however, so this approach is limited by the frequency of the biological event of interest. Note that this ratio is independent of stiffness2,21. Thus, a soft transducer is not a more sensitive detector than a stiff one: as κ decreases, the noise increases exactly as fast as the sensitivity. The physical meaning of the SNR expression is the following: the total area under the curves in the figure is a constant equal to kBT/κ. As the noise spreads over a larger frequency range for a transducer with a higher corner frequency, the fraction of total noise observed in a given bandwidth (shaded areas) can be reduced by increasing ωc = κ/γ, that is, by decreasing γ (see figure). Thus, the SNR and the time resolution of force measurements can only be improved by reducing the dimensions of the transducer18,62.

workers37–39 to characterize the rheological properties of individual DNA molecules. Magnetic fields. Magnetic fields can be used to manipulate and apply forces to biomolecules that are tethered to magnetic particles, as most biomolecules have zero magnetic susceptibility. Very stable and small forces can then be created with magnetic fields from either permanent magnets or electromagnets. The forces generated with permanent magnets on small magnetic beads (r < 3 µm) are usually below 10 pN (REFS 8,40). A drawback of magnetic fields is that the magnetic forces often have to be measured indirectly. For example, the forces acting on magnetic beads can be calibrated by determining the velocity attained by the beads in liquid in a given field and by using Stokes’s law8. However, an elegant and direct measurement of magnetic force has been implemented by Strick et al.40,41 by using the equipartition theorem (BOX 1). Magnetic force has been used to apply torsional stress to individual DNA molecules. Strick et al.40,41 used the fact that magnetic beads have a preferred magnetization axis that makes them orientate with an external field and rotate when the field rotates. In this way, individual DNA molecules, torsionally constrained between a coated glass surface and a magnetic bead (FIG. 3c), were under- or overwound to determine the force–extension behaviour of supercoiled DNA. More recently, this same set-up was used to investigate the relaxation of a supercoiled DNA by single topoisomerase II molecules (BOX 2). Photon fields. Optical tweezers developed rapidly after their power was shown by Ashkin and colleagues42 in the 1980s. Optical tweezers rely on forces imparted to matter by scattering, emission and absorption of light. The RADIATION PRESSURE, which stems from the momentum change as light refracts off an object43 (FIG. 4a), allows objects (a bead, for example) to be held in a focused laser beam, with which it is possible to generate a spring-like force. As with mechanical transducers, the stiffness of the optical trap is an important parameter for force and position resolution (BOX 1). In general, the spring stiffness is much smaller than that of a cantilever (TABLE 1). The force exerted on a refractive object depends on the power of the laser, the dimensions of the object and the difference in the refractive index between the object and its surrounding medium44,45. To apply forces in the piconewton range with tens of milliwatt power, beads are required that are at least one wavelength in diameter because for smaller beads the trapping force scales as the third power of the bead radius2. Such beads can then be attached to macromolecules46,47. There are several unique advantages of optical tweezers over other external field techniques. First, the radiation pressure will only trap a bead near the focus of the laser beam, so the photon field does not simultaneously affect other beads. Second, the momentum transfer between the trapped object and the laser beam can easily be calibrated against displacement and force, thus providing a method of direct, high-resolution force and

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 3 3

REVIEWS

NANOTECHNOLOGY

Any technological development that exceeds standard lower size limits of modern microfabrication techniques (hundreds of nanometres or less).

position measurement (<1 pN and <10 nm, respectively). Last, the applicable force range for photon fields (10–13–10–10 N) is highly relevant for biological systems. A major disadvantage of optical tweezers is laser damage to active biological systems. The negative effect of laser exposure on the working lifetime of biological complexes has been shown in RNA polymerase experi-

Box 2 | What can single-molecule manipulation tell us about biology?

Count

10 Two recent studies have demonstrated the ability of 8 single-molecule techniques to 5.5±1.3 bp s–1 6 elucidate new aspects of enzyme kinetics. First, Davenport et al.36, 4 by using a single RNA 9.1±1 bp s–1 polymerase (RNAP) molecule 2 moving along a DNA strand 0 attached to a bead in a flow field, 0 2 4 6 8 10 12 determined that RNAP can Average peak rate (bp s–1) operate in at least two modes, one slow and one fast. The figure illustrates the averaged peak rates of single RNAP molecules, showing that they can be in a slow or a fast transcription state. Molecules in the slower mode paused more readily than did the faster molecules, and a high correlation between the pausing and the complete stopping of individual molecules was observed. Together, these results imply that a temporary pause in transcription is the first kinetic step towards a complete halt in activity. Furthermore, they indicate that identical enzymes can exist in different microstates with distinct functions, opening the possibility of yet another level of control over gene expression in the cell. Second, Strick et al.41 examined the activity of type II topoisomerase (topo II) on DNA attached to magnetic beads and supercoiled by the application of a rotating magnetic field (FIG. 3c). Topo II was observed to unwind DNA, which was detected as discrete 90-nm jumps in the DNA length. This observation supports the claim that topo II catalyses the relaxation of two supercoils per molecule of ATP hydrolysed63. However, mechanically extending the DNA strand did not accelerate topo II catalysis. In fact, at saturating ATP concentrations, topo II activity decreased by a factor of three as the force was raised from 0.3 pN to 5 pN, suggesting that the rate-limiting step of the reaction is directly affected by the applied force.

ments. Exposure to laser light of a wavelength most compatible with biomolecules (λ = 835 nm) still decreased the active lifetime of this enzyme from more than 300 s, when not exposed to laser light, to 89 s at 90 mW of laser power48. Also, the data throughput is relatively low because, unlike in flow fields, only one molecule is handled at a time. Manipulation of biological systems with optical tweezers began with relatively large objects, such as bacteria, yeast and mammalian cells49,50. But optical trapping, combined with microsphere handles linked to molecules of interest, has quickly become a principal tool for manipulating and measuring the forceproducing properties of various molecular motors, including kinesin moving along microtubules13, actomyosin complexes28, RNA polymerase36,48 (BOX 2) and DNA polymerase11 (FIG. 4b). Moreover, optical trapping has also been used both to measure the elastic properties of DNA47 and to characterize the mechanical unfolding of proteins52,53. What next?

The future of biomolecular manipulation depends on three factors: the integration and further development of single-molecule techniques; progress in the field of NANOTECHNOLOGY; and the use of high-throughput systems such as MICROFLUIDICS. These factors will facilitate the application of single-molecule methods to more complex problems, in particular to in vivo systems. Already the power of integrating SFM imaging and pulling has been demonstrated in a study of bacteriorhodopsin20. Oesterhelt et al.20 first imaged a crystalline region of membrane-embedded bacteriorhodopsin, then pulled on the last two helices, F and G, of the seven bacteriorhodopsin transmembrane helices. As these two helices unfolded and left the membrane, the polypeptide chain extended and began pulling on the next helix pair. In this manner, the whole

Microneedle

Magnetic bead

Rotating magnetic field

Evanescent field Magnetic bead Flow force

DNA Actin filaments

Glass

Myosin

Magnetic force

Dig–anti-dig connector

θ Glass Biotin–avidin connector Biotin–avidin connectors

Laser in

Laser out

Figure 3 | Geometries of typical single-molecule experiments. a | A single myosin head attached to a microneedle moves along an actin filament. The motion of the myosin is observed with a laser by total internal reflection fluorescence microscopy, while the forces are detected by observing the displacement of the microneedle25. b | DNA tethering a magnetic bead to a point on the glass slide8. A magnetic force, which can be determined by means of Stokes’s law (see text), is applied perpendicularly to a flow force. The latter, and therefore the total resultant force acting on the molecule, can be determined from the known magnetic force and the angle between the DNA and the magnetic field, θ. The combined magnetic and flow fields can be used to stretch the DNA more than that achievable with the magnetic field alone. c | A rotating magnetic field is used to under- or overwind double-stranded DNA tethered between the glass slide and a magnetic bead40,41. The resulting supercoils (plectonemes) can be studied by measuring the displacement of the bead perpendicular to the glass slide as a function of the magnetic force.

134

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS

b F DNA

Polystyrene bead

dsDNA

DNAP

ssDNA RNAP Pipette Laser light

Figure 4 | Geometries of typical optical-trap single-molecule experiments. a | An optical trap measuring the force generated during transcription by a single RNA polymerase molecule (RNAP)54. During transcription, an RNAP bound nonspecifically to a glass slide must thread the template and do work against a load applied by the optical trap through a polystyrene bead attached at the end of the DNA molecule. b | A single-stranded DNA molecule, bound with a primer, connects a bead fixed at the end of a micropipette and a bead held in the optical trap. A feedback circuit is used to keep the DNA molecule at a fixed tension, F. As the DNAP converts single-stranded DNA into double-stranded DNA, keeping the tension constant requires the pipette to adjust its position relative to the optical trap by an amount proportional to the movement of the enzyme over the template11.

MICROFLUIDICS

Microscopic channels etched into a surface by modern microfabrication techniques for the purpose of transporting small amounts of solution from one place to another.

protein was pulled out of the membrane, helix pair by helix pair, revealing details of the attractive forces between helices and the membrane. Finally, the same region was imaged to verify that only one bacteriorhodopsin molecule had been extracted from the membrane. Although this study illustrates the advantage of combining single-molecule techniques, the possibilities abound. For instance, combining single-molecule fluorescence with optical tweezers will make it possible to observe spectroscopic signals in response to mechanically induced changes. Nanotechnology has yet to be effectively applied to single-molecule methods. Attaching nanotubes to SFM tips for improved imaging resolution has been

Bustamante, C., Smith, S., Liphardt, J. & Smith, D. Singlemolecule studies of DNA mechanics. Curr. Opin. Struct. Biol. 10, 279–285 (2000). 2. Svoboda, K. & Block, S. M. Biological applications of optical forces. Annu. Rev. Biophys. Biomol. Struct. 23, 247–285 (1994). 3. Ludwig, M. et al. AFM, a tool for single-molecule experiments. Appl. Phys. Mater. Sci. Process. 68, 173–176 (1999). 4. Mehta, A. D., Rief, M. & Spudich, J. A. Biomechanics, one molecule at a time. J. Biol. Chem. 274, 14517–14520 (1999). 5. Sarid, D. in Scanning Force Microscopy: With Applications to Electric, Magnetic, and Atomic Forces I–XI, 253 (Oxford Univ. Press, New York, 1991). 6. Carrion-Vazquez, M. et al. Mechanical and chemical unfolding of a single protein: a comparison. Proc. Natl Acad. Sci. USA 96, 3694–3699 (1999). 7. Bustamante, C., Marko, J. F., Siggia, E. D. & Smith, S. Entropic elasticity of lambda-phage DNA. Science 265, 1599–1600 (1994). 8. Smith, S. B., Finzi, L. & Bustamante, C. Direct mechanical measurements of the elasticity of single DNA molecules by using magnetic beads. Science 258, 1122–1126 (1992). This first single-molecule study of DNA elasticity demonstrates the combined use of magnetic and flow fields. 9. Marko, J. F. & Siggia, E. D. Stretching DNA. Macromolecules 28, 8759–8770 (1995). 10. Erie, D. A., Yang, G., Schultz, H. C. & Bustamante, C. DNA bending by Cro protein in specific and nonspecific

11.

12.

13.

14.

15.

16.

17.

demonstrated54,55, but not fully exploited. Carbon nanotubes have also been used as mechanical tweezers capable of grabbing polystyrene particles56. In fact, these atomically thin tubes may prove to be an ideal building block for the next generation of single-molecule manipulation devices. Ultimately, the goal of single-molecule manipulation is to access the machinery of a living cell. Although the task of characterizing molecular machines and organelles seems daunting, there has been exciting progress. Researchers have used microneedles to probe cells during cytokinesis57,58, and have adapted forcemapping atomic-force microscopy to study the activity of actin under the membrane of living cells17,59. Magnetic beads may also prove useful, in vivo, as the magnetic field will not interfere with other cellular processes. Finally, motor proteins for cell motility are perfect targets for in vivo manipulation and single-molecule studies have already yielded information about their force, efficiency and regulation60,61. In the near future, scientists may come to see each cell as an individual with its own set of molecular machinery. By using methods for manipulating single molecules, biologists will be able to investigate the nature of molecular machines one by one, and infer from their behaviour those properties common to the population and those corresponding to specific substates. Indeed, what Mark Twain observed with cats may be equally true of biomolecules. Links FURTHER INFORMATION SFM overview | SFM in depth | Microneedle research page | Movies of flow fields stretching DNA | Theory of optical tweezers | Building optical tweezers | Background of optical tweezers | Microfluidics applications

complexes: implications for protein site recognition and specificity. Science 266, 1562–1566 (1994). Wuite, G. J., Smith, S. B., Young, M., Keller, D. & Bustamante, C. Single-molecule studies of the effect of template tension on T7 DNA polymerase activity. Nature 404, 103–106 (2000). Kishino, A. & Yanagida, T. Force measurements by micromanipulation of a single actin filament by glass needles. Nature 334, 74–76 (1988). The basic premise of microneedle manipulation is illustrated here in one of the earliest measurements of the force from biological molecules using a mechanical transducer. Svoboda, K., Schmidt, C. F., Schnapp, B. J. & Block, S. M. Direct observation of kinesin stepping by optical trapping interferometry. Nature 365, 721–727 (1993). Howard, J., Hudspeth, A. J. & Vale, R. D. Movement of microtubules by single kinesin molecules. Nature 342, 154–158 (1989). Ishijima, A., Doi, T., Sakurada, K. & Yanagida, T. Subpiconewton force fluctuations of actomyosin in vitro. Nature 352, 301–306 (1991). Nakajima, H. et al. Scanning force microscopy of the interaction events between a single molecule of heavy meromyosin and actin. Biochem. Biophys. Res. Commun. 234, 178–182 (1997). Rotsch, C., Jacobson, K. & Radmacher, M. Dimensional and mechanical dynamics of active and stable edges in motile fibroblasts investigated by using atomic force microscopy. Proc. Natl Acad. Sci. USA 96, 921–926 (1999).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

18. Viani, M. B. et al. Small cantilevers for force spectroscopy of single molecules. J. Appl. Phys. 86, 2258–2262 (1999). 19. Müller, D. J., Baumeister, W. & Engel, A. Controlled unzipping of a bacterial surface layer with atomic force microscopy. Proc. Natl Acad. Sci. USA 96, 13170–13174 (1999). 20. Oesterhelt, F. et al. Unfolding pathways of individual bacteriorhodopsins. Science 288, 143–146 (2000). In this study of a membrane protein, the authors present an elegant example of combining force and imaging SFM. 21. Finer, J. T., Simmons, R. M. & Spudich, J. A. Single myosin molecule mechanics: piconewton forces and nanometre steps. Nature 368, 113–119 (1994). 22. Thomson, N.H. et al. Protein tracking and detection of protein motion using atomic force microscopy. Biophys. J. 70, 2421–2431 (1996). 23. Almqvist, N. et al. Methods for fabricating and characterizing a new generation of biomimetic materials. Mater. Sci. Eng. C 7, 37–43 (1999). 24. Viani, M. B. et al. Fast imaging and fast force spectroscopy of single biopolymers with a new atomic force microscope designed for small cantilevers. Rev. Sci. Instrum. 70, 4300–4303 (1999). 25. Kitamura, K., Tokunaga, M., Iwane, A. H. & Yanagida, T. A single myosin head moves along an actin filament with regular steps of 5.3 nanometres. Nature 397, 129–134 (1999). 26. Kellermayer, M. S. & Granzier, H. L. Elastic properties of single titin molecules made visible through fluorescent Factin binding. Biochem. Biophys. Res. Commun. 221,

VOLUME 1 | NOVEMBER 2000 | 1 3 5

REVIEWS

491–497 (1996). 27. Ishijima, A. et al. Multiple- and single-molecule analysis of the actomyosin motor by nanometer-piconewton manipulation with a microneedle: unitary steps and forces. Biophys. J. 70, 383–400 (1996). 28. Essevaz-Roulet, B., Bockelmann, U. & Heslot, F. Mechanical separation of the complementary strands of DNA. Proc. Natl Acad. Sci USA 94, 11935–11940 (1997). 29. Cluzel, P. et al. DNA: an extensible molecule. Science 271, 792–794 (1996). 30. Leger, J. F., Robert, J., Bourdieu, L., Chatenay, D. & Marko, J. F. RecA binding to a single double-stranded DNA molecule: A possible role of DNA conformational fluctuations. Proc. Natl Acad. Sci. USA 95, 12295–12299 (1998). 31. Leger, J. F. et al. Structural transitions of a twisted and stretched DNA molecule. Phys. Rev. Lett. 83, 1066–1069 (1999). 32. Oosawa, F. Actin–actin bond strength and the conformational change of F-actin. Biorheology 14, 11–19 (1977). 33. Lorentz, H. A. Abhändlungen über Theoretische Physik (Teubner, Leipzig, 1907). 34. Happel, J. B., H. Low Reynolds Number Hydrodynamics with Special Applications to Particulate Media (Prentice Hall, Englewood Cliffs, New Jersey, 1991). 35. Berg, H. C. Random Walks in Biology (Princeton Univ. Press, New Jersey, 1983). 36. Davenport, R. J., Wuite, G. J., Landick, R. & Bustamante, C. Single-molecule study of transcriptional pausing and arrest by E. coli RNA polymerase. Science 287, 2497–2500 (2000). 37. Smith, D. E., Babcock, H. P. & Chu, S. Single-polymer dynamics in steady shear flow. Science 283, 1724–1727 (1999). 38. Perkins, T. T., Smith, D. E., Larson, R. G. & Chu, S. Stretching of a single tethered polymer in a uniform flow. Science 268, 83–87 (1995). This paper describes how flow fields can be used to stretch DNA and reveal new facets of singlemolecule polymer rheology. 39. Perkins, T. T., Quake, S. R., Smith, D. E. & Chu, S. Relaxation of a single DNA molecule observed by optical microscopy. Science 264, 822–826 (1994). 40. Strick, T. R., Allemand, J. F., Bensimon, D., Bensimon, A. & Croquette, V. The elasticity of a single supercoiled DNA molecule. Science 271, 1835–1837 (1996). 41. Strick, T. R., Croquette, V. & Bensimon, D. Single-molecule

136

42.

43. 44.

45.

46. 47.

48. 49.

50.

51.

52.

53.

54.

analysis of DNA uncoiling by a type II topoisomerase. Nature 404, 901–904 (2000). Here, to investigate the activity of individual topoisomerase molecules, magnetic force is used to twist a molecule of DNA and supercoil it. Ashkin, A., Dziedzic, J., Bjorkholm, J. & Chu, S. Observation of a single-beam gradient force optical trap for dielectric particles. Optical Lett. 11, 288–290 (1986). Gordon, J. P. Radiation forces and momenta in dielectric media. Phys. Rev. A 8, 14–21 (1973). Ashkin, A. Forces of a single-beam gradient laser trap on a dielectric sphere in the ray optics regime. Biophys. J. 61, 569–582 (1992). Wright, W. H., Sonek, G. J. & Berns, M. W. Parametric study of the forces on microspheres held by optical tweezers. Appl. Optics 33, 1735–1748 (1994). Chu, S. Laser manipulation of atoms and particles. Science 253, 861–866 (1991). Smith, S. B., Cui, Y. & Bustamante, C. Overstretching BDNA: the elastic response of individual double-stranded and single-stranded DNA molecules. Science 271, 795–799 (1996). Yin, H. et al. Transcription against an applied force. Science 270, 1653–1657 (1995). Kuo, S. C. & Sheetz, M. P. Force of single kinesin molecules measured with optical tweezers. Science 260, 232–234 (1993). Ashkin, A. & Dziedzic, J. M. Optical trapping and manipulation of viruses and bacteria. Science 235, 1517–1520 (1987). This landmark paper demonstrates the power of optical traps to manipulate microscopic objects. Wang, M. D. et al. Force and velocity measured for single molecules of RNA polymerase. Science 282, 902–907 (1998). By using an optical trap, the authors reveal aspects of transcription on a single molecule level. Kellermayer, M. S., Smith, S. B., Granzier, H. L. & Bustamante, C. Folding–unfolding transitions in single titin molecules characterized with laser tweezers. Science 276, 1112–1116 (1997); erratum 277, 1117 (1997). Tskhovrebova, L., Trinick, J., Sleep, J. A. & Simmons, R. M. Elasticity and unfolding of single molecules of the giant muscle protein titin. Nature 387, 308–312 (1997). Wong, S. S., Joselevich, E., Woolley, A. T., Cheung, C. L. & Lieber, C. M. Covalently functionalized nanotubes as nanometre-sized probes in chemistry and biology. Nature 394, 52–55 (1998).

55. Cheung, C. L., Hafner, J. H. & Lieber, C. M. Carbon nanotube atomic force microscopy tips: direct growth by chemical vapor deposition and application to highresolution imaging. Proc. Natl Acad. Sci. USA 97, 3809–3813 (2000). 56. Kim, P. & Lieber, C. M. Nanotube nanotweezers. Science 286, 2148–2150 (1999). 57. Skibbens, R. V. & Salmon, E. D. Micromanipulation of chromosomes in mitotic vertebrate tissue cells: tension controls the state of kinetochore movement. Exp. Cell Res. 235, 314–324 (1997). 58. Roy, P., Petroll, W. M., Cavanagh, H. D., Chuong, C. J. & Jester, J. V. An in vitro force measurement assay to study the early mechanical interaction between corneal fibroblasts and collagen matrix. Exp. Cell Res. 232, 106–117 (1997). 59. Rotsch, C., Braet, F., Wisse, E. & Radmacher, M. AFM imaging and elasticity measurements on living rat liver macrophages. Cell Biol. Int. 21, 685–696 (1997). 60. Muramoto, K. et al. High-speed rotation and speed stability of the sodium-driven flagellar motor in Vibrio alginolyticus. J. Mol. Biol. 251, 50–58 (1995). 61. Muramoto, K. et al.Rotational fluctuation of the sodiumdriven flagellar motor of Vibrio alginolyticus induced by binding of inhibitors. J. Mol. Biol. 259, 687–695 (1996). 62. Chand, A., Viani, M. B., Schaffer, T. E. & Hansma, P. K. Microfabricated small metal cantilevers with silicon tip for atomic force microscopy. J. Microelectromech. Sys. 9, 112–116 (2000). 63. Osheroff, N., Shelton, E. R. & Brutlag, D. L. DNA topoisomerase II from Drosophila melanogaster. Relaxation of supercoiled DNA. J. Biol. Chem. 258, 9536–9543 (1983). 64. Fisher, T. E., Oberhauser, A. F., Carrion-Vazquez, M., Marszalek, P. E. & Fernandez, J. M. The study of protein mechanics with the atomic force microscope. Trends Biochem. Sci. 24, 379–384 (1999); erratum 25, 6 (2000). 65. Florin, E. L., Moy, V. T. & Gaub, H. E. Adhesion forces between individual ligand–receptor pairs. Science 264, 415–417 (1994). 66 Lee, G. U., Chrisey, L. A. & Colton, R. J. Direct measurement of the forces between complementary strands of DNA. Science 266, 771–773 (1994).

Acknowledgements We thank S. Smith and J. Choy for their helpful comments. This work was supported in part by grants from the NIH and the NSF (to C.B.).

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS

NUCLEAR COMPARTMENTALIZATION AND GENE ACTIVITY Claire Francastel*, Dirk Schübeler*, David I. K. Martin‡ and Mark Groudine*§ The regulated expression of genes during development and differentiation is influenced by the availability of regulatory proteins and accessibility of the DNA to the transcriptional apparatus. There is growing evidence that the transcriptional activity of genes is influenced by nuclear organization, which itself changes during differentiation. How do these changes in nuclear organization help to establish specific patterns of gene expression?

HETEROCHROMATIN

A condensed form of chromatin; the degree of compaction is similar to that of mitotic chromosomes. It is usually found around the centromere. INTERPHASE

The period between two mitotic divisions.

*Fred Hutchinson Cancer Research Center, 1,100 Fairview Avenue North, Seattle, Washington 98109, USA. ‡The Victor Chang Cardiac Research Institute, 384 Victoria Street Darlinghurst, Sydney, New South Wales 2010, Australia. §Department of Radiation Oncology, University of Washington School of Medicine, Seattle, Washington 98195, USA. Correspondence to M.G. e-mail: markg@fhcrc.org

Cellular differentiation is the process by which a cell acquires a new phenotype to accomplish specific functions, and it is accompanied by activation of a specific subset of genes and silencing of the remainder. As genes are silenced, the extent of chromatin condensation increases, and extended regions of DNA are packaged in a transcriptionally inactive form often referred to as 1 HETEROCHROMATIN . The amount and distribution of condensed chromatin is similar in terminally differentiated cells of the same lineage, but it varies in the nuclei of different cell types, indicating that nuclear organization may be cell-type specific2 (FIG. 1).

These observations have led to the idea that the topological organization of the INTERPHASE nucleus is related to the differentiated state of the cell, and that this spatial organization is involved in the establishment of the tissue-specific pattern of gene expression during cellular differentiation. Clearly, many overlapping pathways are involved in regulating gene expression3–6. Much research is focused on determining the proteins (trans-acting factors) and DNA sequences (cis-acting elements) involved in the dynamic localization of genes within the nucleus, and we are just beginning to understand the link between the structure and

Figure 1 | Patterns of chromatin condensation in haematopoietic cells. The electron micrographs show haematopoietic cells from normal human bone marrow82, and illustrate the distinct patterns of chromatin condensation in distinct terminally differentiated cells. a | Proerythroblast (immature red blood cell). The nucleus contains almost no heterochromatin. b | Late erythroblast. Heterochromatin in the nucleus is predominant and appears as darkly staining regions. c | The nucleus of a monocyte (right) appears less condensed than that of a granulocyte (left). (Figure adapted with permission from REF. 82 © Harcourt Brace, Madrid, Spain.)

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 3 7

REVIEWS

Figure 2 | Chromatin condensation accompanies cellular differentiation. Electron micrographs of Caenorhabditis elegans pharyngeal cells a | before and b | after commitment to the pharyngeal lineage. a | Cells are committed to give rise to pharyngeal cells, but have not yet undergone terminal differentiation. The nucleus seems to contain almost exclusively euchromatin, with a thin rim of heterochromatin at the nuclear periphery. b | In terminally differentiated cells, the nucleus appears more compact and condensed chromatin forms clumps inside the nucleus. (Images courtesy of J. Priess, Fred Hutchinson Cancer Research Center, Seattle, USA.)

function of chromatin and large-scale changes in the nucleus itself. We propose that tissue-specific ENHANCERS prevent active genes from being included in regions of transcriptionally inactive condensed chromatin that form during cell differentiation. This allows a subset of genes to be active in the appropriate lineage, whereas the remainder are silenced, and ensures, for example, that globin genes are expressed only in red blood cells, and immunoglobulin genes only in B cells. Gene potentiation during differentiation ENHANCERS

Increase transcription of a linked promoter if placed in either orientation, upstream or downstream. EUCHROMATIN

Chromatin that appears less compact than mitotic chromosomes. Active genes are contained within euchromatin.

Modulation of cell- and stage-specific gene expression during differentiation is thought to involve inactivation of those genes that do not need to be expressed in a given lineage. Conversely, those sets of genes that need to be activated are placed or maintained in a decondensed — or ‘potentiated’ — state in precursor cells of that lineage. Potentiation is characterized by an ‘open’ chromatin structure, meaning that a locus is accessible to the tissuespecific factors required for its appropriate expression7. Haematopoiesis has been used as a model system to

Box 1 | Nucleosome structure and function The nucleosome is the fundamental subunit of all chromatin, giving it the characteristic ‘beads-on-a-string’ appearance. Each nucleosome is composed of about 200 base pairs of DNA coiled roughly twice around an octamer of ‘core’ histone proteins — two molecules each of histones H2A, H2B, H3 and H4. The ‘linker’ histone H1 can be associated with each nucleosome; it seems to mediate the packing of adjacent nucleosomes during chromatin condensation. The structure and function of chromatin can be modulated by acetylation (and deacetylation) of histones. Acetylation is a covalent modification of lysine residues — addition of an acetyl group — in the histone tails. Acetylated histones are found in regions of decondensed chromatin, whereas deacetylated histones are found in regions of condensed chromatin. Acetylation is mediated by histone acetyltransferases (HATs), and deacetylation by histone deacetylases (HDACs). Because changes in histone acetylation are associated with changes in chromatin structures, these enzymes are important cofactors in the regulation of gene expression.

138

study the putative coupling of chromatin condensation and gene repression during the differentiation of somatic cells. The haematopoietic programme is initiated from a stem cell with self-renewal and differentiation capacities, which can give rise to several distinct lineages. In multipotent haematopoietic stem cells, many genes relevant to different lineage fates are transcribed before the decision to commit to any one lineage8,9. In other words, before the commitment decision, a general ‘permissive’ state exists in which genes specific to several lineages may be in an open chromatin configuration. Moreover, the subsequent activation of genes for any one lineage is probably coincident with closing of those gene domains associated with other potential cell fates. Indeed, this relationship may be a general feature of gene regulation during cellular differentiation. For example, microscopic analysis of the developing pharynx in the nematode worm Caenorhabditis elegans reveals a uniform structure of EUCHROMATIN in the nuclei of those cells committed to the pharyngeal cell fate (FIG. 2a). However, differentiation of these cells is associated with activation of pharyngeal-specific genes and extensive formation of heterochromatin, indicating that other domains may be inactivated (FIG. 2b). So gross changes in chromatin structure accompany cellular differentiation. Chromatin condensation and gene silencing correlate with a change in the nucleosomal structure characterized by the deacetylation of core histones10–13 and accumulation of linker histones14,15(BOX 1). For example, differentiation of embryonic stem cells is accompanied by a progressive deacetylation of histone H4 in the heterochromatin around centromeres16. Conversely, acetylation of histone tails marks ‘open’ (nuclease-sensitive) domains and is also involved in promoter activation (FIG. 3). Although gene potentiation is necessary for transcription, it is not in itself sufficient for full activation of a promoter. For example, the α-globin locus resides in a region that is constitutively open, but this gene is not transcribed in non-erythroid cells17. In haematopoietic cell lines, the nuclease sensitivity of potentially active genes — for example the β-globin locus and c-myc — correlates with their level of histone H4 acetylation, regardless of their transcriptional activity18,19. Therefore, chromatin opening and histone acetylation could represent the first step of a two-step mechanism to achieve gene expression — gene potentiation before transcriptional activation of the promoters. Whether activation of globin genes in normal differentiating red blood cells is a multistep process is not clear, but there is evidence that it might be. Studies in mouse erythroleukaemia (MEL) cells, comparing the mouse β-globin locus before and after induction of differentiation, reveal that it can exist in two distinct structural states. The first is a potentiated chromatin structure in cells that are committed to the erythroid lineage but not terminally differentiated. The second is an activated state seen in terminally differentiating cells20. In the potentiated state, the β-globin locus is more sensitive to nuclease digestion than in non-erythroid cells21, where the nucleosomes are arranged in a regular www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS (phased) pattern along the β-globin gene. In MEL cells, however, phasing is disrupted both before and after induction of differentiation, and this correlates with an increased generalized sensitivity to DNaseI (REF. 22). Moreover, in uninduced cells, the non-transcribed β-globin gene contains a hypersensitive site in the intervening sequence II, indicating the binding of a non-histone protein. In induced cells, however, transcription of the β-globin gene is associated with the appearance of a hypersensitive site at the promoter and an increase in general sensitivity to nucleases21,23. How are tissue-specific genes maintained in an open chromatin configuration before promoter activation? Although the molecular mechanisms behind this dynamic alteration of chromatin are not clear, a facet of chromatin opening and gene activation that has become apparent only in recent years deserves consideration — namely that active genes are localized in nuclear compartments that are permissive for transcription.

Figure 3 | Chromatin structure and gene expression. a | Organization of chromatin into the tightly condensed 30-nm fibre and ‘beads-on-a-string’ 10-nm fibre. b | Region of DNA containing an actively transcribed gene. The region encompassing the transcribed gene probably adopts the 10-nm configuration, whereas sequences up- and downstream are organized into a 30-nm fibre. Transcription involves changes in the chromatin structure that can be detected by digestion with the endonuclease DNaseI. The transcribed regions also contain hypersensitive (HS) sites, which may be associated with a small stretch of DNA devoid of nucleosomes or simply reflect a different nucleosomal structure (positioning or histone modification). They usually correlate with non-histone proteins bound to the DNA. Regions of general DNaseI sensitivity correlate with acetylation of the histone tails, defining a structural and functional ‘domain’ for gene activity that is characteristic of ‘open’ chromatin. This domain is embedded in regions of higher chromatin compaction that are more resistant to DNaseI digestion. This compacted structure correlates with the presence of deacetylated histones (blue wavy lines), and is characteristic of ‘closed’ chromatin. (Figure adapted from REF. 83.)

Changes in nuclear structure and differentiation

Although the nucleosomal organization of chromosomes has been the focus of much study, the higherorder structure of chromosomes in the interphase nucleus is only poorly understood. Recent advances in fluorescence in situ hybridization (FISH) have allowed individual genes in the interphase nucleus to be visualized and topographically analysed (FIG. 4), contributing to the increasing awareness that spatial context within the nucleus is important in modulating gene expression24–26. Two main levels of nuclear organization can be distinguished. First, chromosomes and genes are confined to discrete nuclear zones within the nuclear volume, referred to as ‘chromosome territories’27–29. These territories occupy non-random positions in the interphase nucleus that are specific to certain cell types30–32. Second, many nuclear functions — such as replication, RNA processing and transcription — take place in welldefined compartments in the nucleus33. How might this higher-order organization regulate gene expression? One view is that active and inactive regions of the genome, as well as protein factors involved in activation or repression of gene expression, are compartmentalized within the nucleus. Nucleasesensitive regions of chromatin and sites of active transcription are located in clusters scattered throughout the nucleus33–35. Whereas chromosomes with few genes are often associated with the nuclear periphery, generich chromosomes reside in a more internal nuclear position36. Genes on the same chromosome are further compartmentalized into distinct active and inactive domains within a chromosome territory: non-coding sequences are found in the interior or randomly distributed in the chromosome territory37,38, whereas potentially active genes, regardless of their transcriptional activity, are preferentially located at the periphery of chromosome territories. The surface area of the chromosomes interacts with the nuclear space between the chromosome territories, the so-called interchromosomal domain, where gene transcription and messenger RNA splicing are thought to occur38,39. Moreover, transcriptionally active gene clusters can be found on large chromatin loops extending outwards from the surface of the chromosome territory, indicating that genomic sequences may be recruited to environments that are permissive for transcription40. If active transcription is concentrated in a fraction of the nucleus, it is conceivable that RNA polymerase II and specific transcription factors would also accumulate in these nuclear compartments for active transcription. Indeed, a significant fraction of active transcription has been shown to colocalize with RNA polymerase II or with basal transcription factors41. Furthermore, nuclear regions containing transcription factors, as well as TATA-box binding protein (TBP), RNA polymerase II and nascent RNA, have been described42. However, a large fraction of transcription-factor molecules are not associated with sites of active transcription, nor with RNA polymerase II; the function of these sites is not yet clear38. One possibility is that they are storage sites from which proteins can be recruited. In

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 3 9

REVIEWS

METHYL-DNA-BINDING DOMAIN PROTEINS

Proteins that bind specifically to methylated DNA through a methyl-DNA-binding domain. Some of these proteins are involved in transcriptional repression of methylated DNA. SATELLITES

Relatively short DNA sequences that are highly repeated in long tandem arrays. CONSTITUTIVE HETEROCHROMATIN

The fraction of heterochromatin that stays compact through the cell cycle. It is mainly composed of repetitive sequences (satellite DNA; see above), and is concentrated in characteristic regions such as centromeres. CHROMOCENTRES

Aggregates of constitutive heterochromatin from different chromosomes.

this case, they may be specific for particular transcription factors, as most of the factors investigated occupy nonoverlapping sites. Interestingly, a subset of transcriptionfactor-rich zones colocalize in the nuclei of neurons in the hippocampus, and this may reflect their coordinated action on genes involved in neuronal excitability43. Factors involved in gene silencing have also been shown to colocalize with each other and with constitutive heterochromatin. These factors include proteins associated with heterochromatin (heterochromatin protein 1 (HP1), SUV39H1 and Ikaros)44–46, METHYL-DNA-BINDING DOMAIN PROTEINS (MBDs) and histone deacetylases (HDACs). For example, MBD proteins associate with major SATELLITES in embryonic stem cells47,48, and HDACs (BOX 1) colocalize with centromeric DNA49. Current evidence supports the idea that the eukaryotic nucleus is functionally divided into heterochromatin compartments that repress transcription, and compartments in which transcription is permitted. How are these discrete compartments and the specific spatial relationship between them established to produce a precise pattern of gene expression? Some evidence indicates that centromeres and CONSTITUTIVE HETEROCHROMATIN provide a structural framework for this nuclear architecture, and that changes in this architecture are involved in cellular diversification. During interphase, the constitutive heterochromatin,

Figure 4 | Fluorescence in situ hybridization to study gene localization. a | Chromosome from a metaphase spread. Centromeric sequences are detected with a Texasred-conjugated probe (red), and a transgene — integrated in the genome of a human erythroleukaemia cell line — is localized with a fluorescein isothiocyanate-conjugated probe (green). The transgene is integrated on the long arm of the chromosome distant from its centromere. b | Fluoresence in situ hybridization experiments in interphase cells reveal that the transgene is in close proximity to centromeric heterochromatin in silent cells (bottom cell), but it is found away from this compartment in expressing cells (upper cell). Silencing of the transgene is therefore accompanied by its relocalization, in the interphase nucleus, close to the heterochromatin compartment70.

140

which is found mainly around the centromeres, stays condensed. The position of the centromeres in interphase nuclei seems to be cell-type specific, and it can change during the cell cycle, during differentiation, or with cellular transformation30,50–52. In mammalian cells, centromeres tend to cluster in so-called CHROMOCENTRES. The specific combinations of centromeres found in chromocentres are different in fibroblasts and haematopoietic cells52. Moreover, they also seem to be distinct in mature cells of different haematopoietic lineages. Just as the organization and location of centromeres change with various cellular states, so a dynamic positioning of chromosomes, specific chromosomal domains or trans-acting factors has also been observed during cell-cycle progression, differentiation and malignant transformation51–58. For example, during nerve growth factor (NGF)-induced differentiation of rat pheochromocytoma cells, changes in gene expression correlate with redistribution of DNaseI hypersensitive domains34, and, in hormone-induced differentiation of pre-adipocytes, the trans-acting factor C/EBP is relocated into discrete nuclear foci59. Nuclear compartmentalization

Dynamic nuclear architecture could reflect changes in the physical position of tissue-specific genes during cell differentiation, which affect the accessibility of genes to regulatory factors and the transcriptional machinery. This, in turn, may facilitate nuclear functions such as gene transcription or silencing. Experimental evidence supports the idea that positioning of a gene near heterochromatin compartments promotes gene silencing. For example, in Drosophila melanogaster the insertion of a block of heterochromatin into one allele of the euchromatic brown gene results in the association of both alleles with centromeric heterochromatin, leading to transcriptional inactivation of the wild-type allele60–62. During differentiation of mouse lymphocytes, gene silencing is heritable and associated with repositioning of genes close to centromeres45,63. In addition, the human β-globin gene domain, in its native context, can be found in distinct locations in the interphase nucleus relative to the heterochromatin compartment19. Its positioning in the erythroid nucleus correlates with the chromatin and general acetylation configurations of the locus, but not with transcription of the β-globin genes19. Together with the fact that suppression of transgene silencing and maintenance of its open chromatin configuration require distance away from centromeric heterochromatin, these results indicate that stably inherited chromatin opening of a locus might be mediated by its sequestration in a permissive compartment. These findings also argue against the idea that nuclear relocation is the consequence of transcription. Other silencing systems — such as the Sir-dependent, telomeric and mating-type-locus silencing in yeast — are associated with compartmentalization of the silenced gene. Telomeric silencing occurs at the nuclear envelope in compartments where telomeres are clustered and the concentration of Sir proteins is high64–68. www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS

Figure 5 | Activation at the β-globin locus: a multistep process? Chromatin opening and transcriptional activation at the βglobin locus can be dissociated80,81, indicating that they are achieved through distinct mechanisms. First, an open chromatin configuration, marked by a locus-wide acetylation, is mediated by positioning of the locus away from centromeric heterochromatin. The resulting open chromatin structure favours the binding of tissue-specific trans-acting factors to enhancers and the promoter, concomitant with a local hyperacetylation of histone H3 and promoter activation. Although sequences outside the locus control region (LCR) are sufficient to promote the first step, gene activation requires an intact LCR. The demonstration that the β-globin locus can adopt distinct structural correlates, together with the demonstration that erythroid differentiation in a model system is a multistep process, supports the idea that activation of globin genes in normal differentiating red blood cells is also a multistep process. The configuration of the chromatin structure in a multipotent stem cell, as well as the structure and localization of the β-globin locus in precursor cells, remains to be determined. (HDACs, histone deacetylases; HP1, heterochromatin protein 1; Pc, Polycomb, HATs, histone acetyltransferases.)

SILENCER ELEMENTS

Cis-acting elements that are involved in silencing, most probably by directly recruiting repressive proteins.

Furthermore, positioning of a locus to the nuclear periphery can provoke Sir-mediated silencing69. In centromeric silencing, by contrast, there is no direct evidence that association with centromeric heterochromatin is the cause of silencing, nor is there evidence that localization away from heterochromatin is required for transcriptional activity. For example, an active transgene may be located close to centromeric heterochromatin. However, transcription and the open chromatin structure of a transgene are unstable in this compartment70. Conversely, a silent transgene can be located away from centromeric heterochromatin70, and transient gene silencing during differentiation of B-cell lines is not associated with relocation close to centromeres63. So although there is evidence that positioning to specific compartments may affect transcriptional activity or gene silencing, such compartmentalization does not seem to be an absolute requirement. In general, however, stably inherited gene activity and open chromatin configurations seem to be associated with positioning away from centromeric heterochromatin, whereas stably inherited gene silencing requires positioning close to centromeric heterochromatin. If there are indeed specific mechanisms to place and keep genes in a silent compartment, the potentiated state would be determined by the positioning of a gene in an active compartment of the nucleus. This, in turn, would allow tissue-specific transcription factors to induce the precise programmes of stable gene expression typical of differentiated cells.

Role of cis-acting elements

Finally we can ask how modifications of the chromatin structure are achieved over broad regions71, and what sequences are involved in mediating an open or closed structure. Sequences that mediate silencing have been described in yeast and Drosophila, and the observations indicate that the formation of an inactive structure may involve assembly of proteins that initially bind to SILENCER ELEMENTS. Such a repressive structure may then propagate along the chromatin. Mating-type silencing in yeast involves the nucleation of a repressive chromatin structure at silencer elements, with subsequent spreading of this structure. Interestingly, silencing at the mating-type loci is associated with both a silenced chromatin structure and localization of silenced regions near the nuclear periphery. This indicates that nuclear compartments and the spreading of specific chromatin structures are not incompatible phenomena but that both define the silenced state66. Telomeric silencing in yeast involves similar spreading of silenced chromatin, mediated by Sir proteins, and is associated with histone hypoacetylation25,72. Similar nucleation and spreading effects are thought to account for silencing of genes located in or near heterochromatin in higher eukaryotes and for Polycomb-mediated gene silencing in Drosophila — a system required for maintaining the inactive state of genes containing a Polycomb response element (PRE). So far, the PRE is the only clear example of a cis-element responsible for silencing in higher eukaryotes73,74. Numerous studies have shown that enhancers can

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 4 1

REVIEWS

LOCUS CONTROL REGION

Defined by its ability, in transgenic assays, to confer high-level, tissue-specific expression on a linked promoter, at all integration sites. FACULTATIVE HETEROCHROMATIN

Fraction of chromatin that is condensed and inactive in a given cell lineage, which may be decondensed and active in another. ANEUPLOIDY

The ploidy of a cell refers to the number of sets of chromosomes that it contains. Aneuploid karyotypes are those whose chromosome complements are not a simple multiple of the haploid set.

counteract such silencing events75–77. The demonstration that an enhancer is sufficient for both localization of a transgene away from centromeric heterochromatin and suppression of transgene silencing indicates that enhancers may maintain gene expression by preventing localization close to the repressive heterochromatic compartment70 (FIG. 5). The situation in a multigene endogenous locus seems to be more complex when compared with the simplified structure of a transgene. The human β-globin LOCUS CONTROL REGION (LCR), although essential for transcriptional activation, is not required to relocate the native locus away from heterochromatin19. The apparent paradox between results obtained with the native locus and those with transgenes can be explained by the existence of numerous factor-binding sites throughout the native locus, which probably alter subnuclear location and chromatin structure. So cis-acting elements other than the LCR, but with similar function, may maintain the β-globin locus in an open chromatin/acetylated configuration, localized away from centromeres. There are various means by which genetic regulatory elements, such as enhancers and LCRs, could cause a gene to localize away from heterochromatin. Initial transcription-factor binding to the enhancer might promote recruitment of a locus to a nuclear compartment enriched in chromatin remodelling complexes and histone acetyltransferases (HATs; for example, HAT1), as well as other elements of the transcriptional machinery. Just as there are transcription-factor-rich sites, there are regions of the nucleus that are devoid of, or contain low concentrations of, such factors. These areas might be the default location of a gene or transgene if transcription factors do not bind to its enhancer elements. Alternatively, the lack of a suitable transcription factor could result in failure of a gene to be targeted to a transcription-rich region of the nucleus. The instability of gene expression associated with localization near heterochromatin may result from this low concentration of

John, B. The Biology of Heterochromatin (ed. Verma, R. S.) (Cambridge Univ. Press, Cambridge, 1988). 2. Leitch, A. R. Higher levels of organization in the interphase nucleus of cycling and differentiated cells. Microbiol. Mol. Biol. Rev. 64, 138–152 (2000). 3. Felsenfeld, G. Chromatin as an essential part of the transcriptional mechanism. Nature 355, 219–224 (1992). 4. Gregory, P. D. & Horz, W. Chromatin and transcription — how transcription factors battle with a repressive chromatin environment. Eur. J. Biochem. 251, 9–18 (1998). 5. Kadonaga, J. T. Eukaryotic transcription: An interlaced network of transcription factors and chromatin-modifying machines. Cell 92, 307–313 (1998). 6. Wolffe, A. P. Transcription: In tune with the histones. Cell 77, 13–16 (1994). 7. Kramer, J. A., McCarrey, J. R., Djakiew, D. & Krawetz, S. A. Differentiation: The selective potentiation of chromatin domains. Development 125, 4749–4755 (1998). 8. Jimenez, G., Griffiths, S. D., Ford, A. M., Greaves, M. F. & Enver, T. Activation of the β-globin locus control region precedes commitment to the erythroid lineage. Proc. Natl Acad. Sci. USA 89, 10618–10622 (1992). 9. Hu, M. et al. Multilineage gene expression precedes commitment in the hemopoietic system. Genes Dev. 11, 774–785 (1997). 10. Lee, D. Y., Hayes, J. J., Pruss, D. & Wolffe, A. P. A positive role for histone acetylation in transcription factor access to

142

positive transcriptional regulators in the heterochromatic compartment, and from a high concentration of negative regulators. Perhaps the simplest mechanism for activation by regulatory elements is one in which transcriptional activators bound to regulatory elements disrupt local interactions between a locus and heterochromatin, permitting the gene to move into the active compartment. Future directions

The chromosomal and nuclear position of a gene can influence its activity, and the position of a gene within the nucleus can be dictated by the cis-acting sequences linked to it. Tissue-specific enhancers may prevent a linked gene from being included in FACULTATIVE HETEROCHROMATIN, which forms during cell differentiation, therefore allowing it to be active in the appropriate lineage. If the precise topological conformation of genes within the nucleus is disrupted, the result may be phenotypic changes or disease. For example, cancer involves disruption of stable patterns of gene expression, and can be accompanied by a global disorganization of chromosomal positioning in the nucleus78. Chromosomal abnormalities and ANEUPLOIDY, which often affect nuclear architecture, can influence gene expression not only of the affected chromosome, but also of nearby chromosomal regions79. So investigations into links between structural and functional aspects of interphase chromosomes promise to reveal fundamental aspects of biology, as well as insights into the development of disease. Links DATABASE LINKS Histone H4 | α-globin | β-globin | c-

nucleosomal DNA. Cell 72, 73–84 (1993). 11. Wolffe, A. P. & Pruss, D. Targeting chromatin disruption: Transcription regulators that acetylate histones. Cell 84, 817–819 (1996). 12. Turner, B. M. Histone acetylation as an epigenetic determinant of long-term transcriptional competence. Cell. Mol. Life Sci. 54, 21–31 (1998). 13. Wolffe, A. P. Packaging principle: How DNA methylation and histone acetylation control the transcriptional activity of chromatin. J. Exp. Zool. 282, 239–244 (1998). 14. Crane-Robinson, C. How do linker histones mediate differential gene expression? Bioessays 21, 367–371 (1999). 15. Wolffe, A. P., Khochbin, S. & Dimitrov, S. What do linker histones do in chromatin? Bioessays 19, 249–255 (1997). 16. Keohane, A. M., O’Neill L, P., Belyaev, N. D., Lavender, J. S. & Turner, B. M. X-inactivation and histone H4 acetylation in embryonic stem cells. Dev. Biol. 180, 618–630 (1996). 17. Vyas, P. et al. Cis-acting sequences regulating expression of the human α-globin cluster lie within constitutively open chromatin. Cell 69, 781–793 (1992). 18. O’Neill, L. P. & Turner, B. M. Histone H4 acetylation distinguishes coding regions of the human genome from heterochromatin in a differentiation-dependent but transcription-independent manner. EMBO J. 14, 3946–3957 (1995). 19. Schübeler, D. et al. Nuclear localization and histone

20.

21.

22.

23.

24. 25.

acetylation: A pathway for chromatin opening and transcriptional activation of the human β-globin locus. Genes Dev. 14, 940–950 (2000). Indicates that positioning away from the heterochromatin compartment may mediate open chromatin configuration and histone acetylation of the human globin locus. Reddy, P. M. & Shen, C. K. Erythroid differentiation of mouse erythroleukemia cells results in reorganization of protein–DNA complexes in the mouse β-major globin promoter but not its distal enhancer. Mol. Cell. Biol. 13, 1093–1103 (1993). Cohen, R. B. & Sheffery, M. Nucleosome disruption precedes transcription and is largely limited to the transcribed domain of globin genes in murine erythroleukemia cells. J. Mol. Biol. 182, 109–129 (1985). Benezra, R., Cantor, C. R. & Axel, R. Nucleosomes are phased along the mouse β-major globin gene in erythroid and nonerythroid cells. Cell 44, 697–704 (1986). Sheffery, M., Rifkind, R. A. & Marks, P. A. Murine erythroleukemia cell differentiation: DNase I hypersensitivity and DNA methylation near the globin genes. Proc. Natl Acad. Sci. USA 79, 1180–1184 (1982). Lamond, A. I. & Earnshaw, W. C. Structure and function in the nucleus. Science 280, 547–553 (1998). Cockell, M. & Gasser, S. M. Nuclear compartments and gene regulation. Curr. Opin. Genet. Dev. 9, 199–205 (1999).

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

REVIEWS 26. Marshall, W. F., Fung, J. C. & Sedat, J. W. Deconstructing the nucleus: Global architecture from local interactions. Curr. Opin. Genet. Dev. 7, 259–263 (1997). 27. Munkel, C. et al. Compartmentalization of interphase chromosomes observed in simulation and experiment. J. Mol. Biol. 285, 1053–1065 (1999). 28. Manuelidis, L. A view of interphase chromosomes. Science 250, 1533–1540 (1990). 29. Zink, D. et al. Structure and dynamics of human interphase chromosome territories in vivo. Hum. Genet. 102, 241–251 (1998). 30. Manuelidis, L. Different central nervous system cell types display distinct and nonrandom arrangements of satellite DNA sequences. Proc. Natl Acad. Sci. USA 81, 3123–3127 (1984). One of the first papers showing that the spatial organization of centromeres is non-random and cell-type specific, and indicating that this could represent specific functional capacities. 31. Manuelidis, L. & Borden, J. Reproducible compartmentalization of individual chromosome domains in human CNS cells revealed by in situ hybridization and three-dimensional reconstruction. Chromosoma 96, 397–410 (1988). 32. Haaf, T. & Schmid, M. Chromosome topology in mammalian interphase nuclei. Exp. Cell Res. 192, 325–332 (1991). 33. Sadoni, N. et al. Nuclear organization of mammalian genomes. Polar chromosome territories build up functionally distinct higher order compartments. J. Cell Biol. 146, 1211–1226 (1999). 34. Park, P. C. & De Boni, U. Transposition of DNase hypersensitive chromatin to the nuclear periphery coincides temporally with nerve growth factor-induced upregulation of gene expression in PC12 cells. Proc. Natl Acad. Sci. USA 93, 11646–11651 (1996). 35. de Graaf, A. et al. Three-dimensional distribution of DNase I-sensitive chromatin regions in interphase nuclei of embryonal carcinoma cells. Eur. J. Cell Biol. 52, 135–141 (1990). 36. Croft, J. A. et al. Differences in the localization and morphology of chromosomes in the human nucleus. J. Cell Biol. 145, 1119–1131 (1999). Analyses the influence of the cell cycle and the inhibition of transcription on the morphology and distribution of specific human chromosome territories. 37. Kurz, A. et al. Active and inactive genes localize preferentially in the periphery of chromosome territories. J. Cell Biol. 135, 1195–1205 (1996). 38. Verschure, P. J., van Der Kraan, I., Manders, E. M. & van Driel, R. Spatial relationship between transcription sites and chromosome territories. J. Cell Biol. 147, 13–24 (1999). Shows that transcription is compartmentalized, and that sites of active transcription are distinct from sites of inactive chromatin. 39. Zirbel, R. M., Mathieu, U. R., Kurz, A., Cremer, T. & Lichter, P. Evidence for a nuclear compartment of transcription and splicing located at chromosome domain boundaries. Chromosome Res. 1, 93–106 (1993). 40. Volpi, E. V. et al. Large-scale chromatin organization of the major histocompatibility complex and other regions of human chromosome 6 and its response to interferon in interphase nuclei. J. Cell Sci. 113, 1565–1576 (2000). This paper finds a correlation between the formation of large external chromatin loops, outside a chromosome territory, and the transcriptional activity of a gene cluster. 41. Grande, M. A., van der Kraan, I., de Jong, L. & van Driel, R. Nuclear distribution of transcription factors in relation to sites of transcription and RNA polymerase II. J. Cell Sci. 110, 1781–1791 (1997). 42. Pombo, A. et al. Regional and temporal specialization in the nucleus: A transcriptionally active nuclear domain rich in PTF, Oct1 and PIKA antigens associates with specific chromosomes early in the cell cycle. EMBO J. 17, 1768–1778 (1998). 43. van Steensel, B. et al. Partial colocalization of glucocorticoid and mineralocorticoid receptors in discrete

44.

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

55.

56.

57. 58.

59.

60.

61.

62.

63.

compartments in nuclei of rat hippocampus neurons. J. Cell Sci. 109, 787–792 (1996). Aagaard, L. et al. Functional mammalian homologues of the Drosophila PEV-modifier Su(var)3-9 encode centromere-associated proteins which complex with the heterochromatin component M31. EMBO J. 18, 1923–1938 (1999). Brown, K. E. et al. Association of transcriptionally silent genes with Ikaros complexes at centromeric heterochromatin. Cell 91, 845–854 (1997). Ikaros, a protein associated with heterochromatin foci in the nucleus of mouse B lymphocytes, is also associated with transcriptionally inactive genes, indicating a different mechanism for gene silencing. Wreggett, K. A. et al. A mammalian homologue of Drosophila heterochromatin protein 1 (HP1) is a component of constitutive heterochromatin. Cytogenet. Cell Genet. 66, 99–103 (1994). Hendrich, B. & Bird, A. Identification and characterization of a family of mammalian methyl-CpG binding proteins. Mol. Cell. Biol. 18, 6538–6547 (1998). Lewis, J. D. et al. Purification, sequence, and cellular localization of a novel chromosomal protein that binds to methylated DNA. Cell 69, 905–914 (1992). Kim, J. et al. Ikaros DNA-binding proteins direct formation of chromatin remodeling complexes in lymphocytes. Immunity 10, 345–355 (1999). Ikaros can recruit histone deacetylases and chromatin remodelling complexes to regions of heterochromatin in the nucleus of mouse T lymphocytes. Shelby, R. D., Hahn, K. M. & Sullivan, K. F. Dynamic elastic behavior of α-satellite DNA domains visualized in situ in living human cells. J. Cell Biol. 135, 545–557 (1996). Vourc’h, C., Taruscio, D., Boyle, A. L. & Ward, D. C. Cell cycle-dependent distribution of telomeres, centromeres, and chromosome-specific subsatellite domains in the interphase nucleus of mouse lymphocytes. Exp. Cell Res. 205, 142–151 (1993). Alcobia, I., Dilao, R. & Parreira, L. Spatial associations of centromeres in the nuclei of hematopoietic cells: Evidence for cell-type-specific organizational patterns. Blood 95, 1608–1615 (2000). Manuelidis, L. Indications of centromere movement during interphase and differentiation. Ann. NY Acad. Sci. 450, 205–221 (1985). Minc, E., Allory, Y., Worman, H. J., Courvalin, J. C. & Buendia, B. Localization and phosphorylation of HP1 proteins during the cell cycle in mammalian cells. Chromosoma 108, 220–234 (1999). Kozubek, S. et al. Distribution of ABL and BCR genes in cell nuclei of normal and irradiated lymphocytes. Blood 89, 4537–4545 (1997). Neves, H., Ramos, C., da Silva, M. G., Parreira, A. & Parreira, L. The nuclear topography of ABL, BCR, PML, and RARα genes: Evidence for gene proximity in specific phases of the cell cycle and stages of hematopoietic differentiation. Blood 93, 1197–1207 (1999). Bartova, E. et al. Nuclear topography of the c-myc gene in human leukemic cells. Gene 244, 1–11 (2000). Bridger, J. M., Boyle, S., Kill, I. R. & Bickmore, W. A. Remodelling of nuclear architecture in quiescent and senescent human fibroblasts. Curr. Biol. 10, 149–152 (2000). Tang, Q. Q. & Lane, M. D. Activation and centromeric localization of CCAAT/enhancer-binding proteins during the mitotic clonal expansion of adipocyte differentiation. Genes Dev. 13, 2231–2241 (1999). Dernburg, A. F. et al. Perturbation of nuclear architecture by long-distance chromosome interactions. Cell 85, 745–759 (1996). Csink, A. K. & Henikoff, S. Large-scale chromosomal movements during interphase progression in Drosophila. J. Cell Biol. 143, 13–22 (1998). Csink, A. K. & Henikoff, S. Genetic modification of heterochromatic association and nuclear organization in Drosophila. Nature 381, 529–531 (1996). Brown, K. E., Baxter, J., Graf, D., Merkenschlager, M. & Fisher, A. G. Dynamic repositioning of genes in the nucleus of lymphocytes preparing for cell division. Mol.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

Cell 3, 207–217 (1999). 64. Aparicio, O. M., Billington, B. L. & Gottschling, D. E. Modifiers of position effect are shared between telomeric and silent mating-type loci in S. cerevisiae. Cell 66, 1279–1287 (1991). 65. Chien, C. T., Buck, S., Sternglanz, R. & Shore, D. Targeting of SIR1 protein establishes transcriptional silencing at HM loci and telomeres in yeast. Cell 75, 531–541 (1993). 66. Gotta, M. & Gasser, S. M. Nuclear organization and transcriptional silencing in yeast. Experientia 52, 1136–1147 (1996). 67. Laroche, T. et al. Mutation of yeast Ku genes disrupts the subnuclear organization of telomeres. Curr. Biol. 8, 653–656 (1998). 68. Maillet, L. et al. Evidence for silencing compartments within the yeast nucleus: A role for telomere proximity and Sir protein concentration in silencer-mediated repression. Genes Dev. 10, 1796–1811 (1996). 69. Andrulis, E. D., Neiman, A. M., Zappulla, D. C. & Sternglanz, R. Perinuclear localization of chromatin facilitates transcriptional silencing Nature 394, 592–595 (1998). 70. Francastel, C., Walters, M. C., Groudine, M. & Martin, D. I. A functional enhancer suppresses silencing of a transgene and prevents its localization close to centrometric heterochromatin. Cell 99, 259–269 (1999). Stable transgene expression requires positioning away from the heterochromatin compartment and the binding of transcription factors to the enhancer. 71. Bulger, M. & Groudine, M. Looping versus linking: Toward a model for long-distance gene activation. Genes Dev. 13, 2465–2477 (1999). 72. Lustig, A. J. Mechanisms of silencing in Saccharomyces cerevisiae. Curr. Opin. Genet. Dev. 8, 233–239 (1998). 73. Moehrle, A. & Paro, R. Spreading the silence: epigenetic transcriptional regulation during Drosophila development. Dev. Genet. 15, 478–484 (1994). 74. Loo, S. & Rine, J. Silencing and heritable domains of gene expression. Annu. Rev. Cell Dev. Biol. 11, 519–548 (1995). 75. Aparicio, O. M. & Gottschling, D. E. Overcoming telomeric silencing: a trans-activator competes to establish gene expression in a cell cycle-dependent way. Genes Dev. 8, 1133–1146 (1994). 76. Festenstein, R. et al. Locus control region function and heterochromatin-induced position effect variegation. Science 271, 1123–1125 (1996). 77. Walters, M. C. et al. Transcriptional enhancers act in cis to suppress position-effect variagation. Genes Dev. 10, 185–195 (1996). 78. Linares-Cruz, G. et al. p21WAF-1 reorganizes the nucleus in tumor suppression. Proc. Natl Acad. Sci. USA 95, 1131–1135 (1998). Global disorganization of nuclear architecture can be associated with cancer, but is reversible during tumour suppression. 79. Qumsiyeh, M. B. Impact of rearrangements on function and position of chromosomes in the interphase nucleus and on human genetic disorders. Chromosome Res. 3, 455–465 (1995). 80. Epner, E. et al. The β-globin LCR is not necessary for an open chromatin structure or developmentally regulated transcription of the native mouse β-globin locus. Mol. Cell 2, 447–455 (1998). 81. Reik, A. et al. The locus control region is necessary for gene expression in the human β–globin locus but not the maintenance of an open chromatin structure in erythroid cells. Mol. Cell. Biol. 18, 5992–6000 (1998). 82. Rozman, C., Woessner, S., Feliu, E., Lafuente, R. & Berga, L. Cell Ultrastructure for Hematologists (ed. Ediciones Doyma, S. A.) (Barcelona, Spain, 1993). 83. Alberts, B. et al. Molecular Biology of the Cell 3rd edn (Garland, New York, 1994).

Acknowledgements The authors thank Matthew Lorincz and Bas van Steensel for their useful comments on this manuscript. This work was supported by a special fellowship to C.F. from the Leukemia and Lymphoma Society, a fellowship from the Deutsche Forschungsgemeinschaft to D.S., and NIH grants to M.G.

VOLUME 1 | NOVEMBER 2000 | 1 4 3

PERSPECTIVES TIMELINE

The meteoric rise of regulated intracellular proteolysis R. John Mayer It is often the case in biology that research into breaking things down lags behind research into synthesizing them, and this is certainly true for intracellular proteolysis. Now that we recognize that intracellular proteolysis, triggered by attaching multiple copies of a small protein called ubiquitin to target proteins, is fundamental to life, it is hard to believe that 20 years ago this field was little more than a backwater of biochemistry studied by a handful of laboratories. Among the few were Avram Hershko, Aaron Ciechanover and Alexander Varshavsky, who were recently awarded the Albert Lasker award for basic medical research for discovering the importance of protein degradation in cellular physiology. This Timeline traces how they and their collaborators triggered the rapid movement of ubiquitin-mediated proteolysis to centre stage.

tide bond, to the carboxy-terminal glycine of another ubiquitin. This reaction can be continued to form a multi-ubiquitin chain, which is the signal for degradation of the target protein by the 26S proteasome. Until 1942, biochemists accepted the dogma, laid down by Folin in 1905 on the basis of urinary metabolites, that poorly understood macromolecules known as ‘endogenous proteins’ were stable. Folin might have understood the error of this supposition had he read Lewis Carroll’s Through the Looking Glass1, in which the Red Queen (FIG. 2) states:“Now, here, you see, it takes all the running you can do, to keep in the same place. If you want to get somewhere else, you must run twice as fast as that.” Biological systems are

Discovery of ubiquitin

Our story begins only two years before Blobel made his incisive remarks at the Ciba

Plasma membrane Cytoplasmic proteins Autophagy

The concept of intracellular proteolysis is a very modern one. Cells have two main mechanisms for intracellular protein degradation: the endosome–lysosome system, which degrades proteins internalized by endocytosis, and the non-lysosomal system, which degrades proteins from the nucleus, cytosol and endoplasmic reticulum (FIG. 1). Most non-lysosomal intracellular proteolysis is carried out by the ubiquitin/26S proteasome system, in which an ε-amino group of a lysine residue in a target protein is covalently conjugated with the carboxy-terminal glycine residue of ubiquitin. Lysine 48 of the first ubiquitin can then link, through an isopep-

constantly turning over molecules to maintain a steady state, a fact that became apparent for proteins with the discovery of 15N. By measuring the uptake of this label into proteins, Schoenheimer2 (working at Harvard) discovered that proteins are continuously synthesized and degraded in cells (see TIMELINE). But what are the mechanisms of synthesis and degradation? The discovery of the pathways by which proteins are degraded in cells was held back because more emphasis was placed on protein synthesis than on protein degradation — not least because it is much easier to measure the incorporation of radiolabelled amino acids into newly synthesized proteins than to measure the loss of a label in a pulse–chase analysis. It was not until the end of the 1970s that such general approaches were recognized as being limited: a comment at a Ciba Foundation Symposium by Günter Blobel3 spoke volumes:“We have heard many studies here measuring trichloroacetic acid soluble counts, but this is a rather prehistoric method of looking at proteins now!”

Endosome–lysosome system Ubiquitin– proteasome system

Cooperation? Endosome/ lysosome ER proteins Autophagosome

Nuclear proteins Nucleus

Mitochondrion

Mitochondrial proteolytic system

Figure 1 | The main proteolytic pathways in eukaryotic cells. The endosome–lysosome pathway (green) degrades extracellular and cell-surface proteins, such as receptors and their ligands. Intracellular organelles also enter this pathway through double-membraned autophagosomes. The ubiquitin–proteasome pathway (red) degrades proteins from the cytoplasm, nucleus and endoplasmic reticulum (ER). Evidence is now emerging that the two pathways cooperate. Finally, the mitochondrion has its own proteolytic system (blue), similar to that in prokaryotes.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 4 5

PERSPECTIVES

Timeline | Regulated intracellular proteolysis Folin states that ‘endogenous proteins’ are stable.

1905

Schoenheimer uses 15 N to show continuous protein turnover in cells.

1912

1942

Lewy discovers his ‘bodies’ in the brain stems of some patients with Parkinson’s disease.

Hershko and colleagues identify enzymes of the ubiquitin–protein ligase system.

1978

Hershko and Ciechanover discover the process of ubiquitylation.

Foundation. Working at the Technion University in Haifa, Avram Hershko and his graduate students, including Aaron Ciechanover (FIG. 3a), first showed that a small protein ubiquitously expressed in all cell types, which was subsequently found to be ubiquitin4, is a component of an ATP-dependent in vitro proteolytic system in rabbit reticulocyte extracts. They found that ubiquitin is covalently linked to protein substrates and that it was necessary for proteolysis in their system. The linkage of one protein to another through an isopeptide bond was unusual but elicited little interest, except from the small number of workers in the field. Subsequently, Hershko and colleagues discovered that the addition of ubiquitin to target proteins (ubiquitylation) requires three types of enzyme, termed E1, E2 and E3. First, a ubiquitin-activating enzyme (E1) forms a thioester with the carboxy-terminal glycine residue of ubiquitin

1983

Rechsteiner and colleagues partially purify the 26S proteasome.

1984

1986

Varshavsky, Ciechanover and Finley discover that ubiquitylation is essential for viability and is necessary for cell-cycle progression.

Kirschner and colleagues discover that a cyclin is degraded by the ubiquitin pathway.

1988

1991

Lowe, Landon and Mayer discover that Lewy bodies are full of ubiquitylated proteins.

in a reaction that requires ATP. The activated ubiquitin is then transferred from E1 to a ubiquitin-conjugating enzyme (E2) by transesterification. Finally, a ubiquitin–protein ligase (E3) catalyses the transfer of ubiquitin to the ε-amino group of a lysine residue in a target protein to produce a ubiquitin–protein conjugate5. Furthermore, Hershko et al.6 showed that the activity of the ubiquitin pathway is greatly increased in cells making abnor-

“Now, here, you see, it takes all the running you can do, to keep in the same place. If you want to get somewhere else, you must run twice as fast as that.” mal proteins. It was independently shown that there are multiple E2s (REF. 7) and E3s, and that E3s are the arbiters of substrate selection for ubiquitylation. Specific recognition of proteins for degradation lies at the heart of the ubiquitin pathway, and we are only just beginning to appreciate how different combinations of E2 and E3 enzymes can impose exquisite specificity on substrate selection5. We now realize that mutations in proteins preventing recognition by E3s, mutations in E3s or sequestration of E3s by viruses8 can cause cancer (BOX 1).

The future — resolve the mechanism by which the ubiquitin–proteasome system and the endosome–lysosome system coordinately degrade membrane proteins. Uncover evidence for further versatility of the isopeptide bond in cellular physiology. Develop therapies based on the ubiquitin–proteasome system.

1997

2000

Several groups discover the combinatorial control and specificity of SCF ubiquitin protein ligases.

because histone H2A was found to be ubiquitylated. Varshavsky and his graduate student Daniel Finley, in collaboration with Aaron Ciechanover (who was by now an established researcher), used cells expressing a temperature-sensitive E1 to show that ubiquitin is required for protein degradation in living cells, and that ubiquitin conjugation is essential for cell viability. At the non-permissive temperature, cells were arrested in the G2 phase of the cell cycle, showing that ubiquitin-dependent proteolysis is also required for cell-cycle progression9,10. The dual role of protein phosphorylation and proteolysis in regulating the levels and activities of cyclins is now well understood11, but at the time this observation was quite unusual. Ubiquitylation then began to invade other areas of research (BOX 1). Varshavsky and colleagues purified some of the individual E2 enzymes of the yeast Saccharomyces cerevisiae, and showed that a genetically defined component of DNA repair pathways, termed RAD6, and another protein with weak homology to it (CDC34), are E2 enzymes12,13. These findings indicated that E2s have key roles in DNA repair and the cell cycle. It was at this time that the cell-cycle community became really interested in the ubiquitin pathway, particularly when Kirschner’s laboratory (then in San

Ubiquitylation is essential for life

Figure 2 | Lewis Carroll’s Red Queen. The Red Queen and Alice find themselves running to stand still, a process familiar to cells — steady state (standing still) is achieved by a constant turnover of molecules with short half-lives.

146

It took years before the importance of these biochemical findings was fully appreciated. The turning point came in 1984 through the work of Alexander Varshavsky (FIG. 3b), then at the Massachusetts Institute of Technology in Boston. He studied chromatin and gene expression and became interested in ubiquitin

Figure 3 | The fathers of the field of regulated intracellular proteolysis. a | Avram Hershko (left) with Aaron Ciechanover (right). b | Alexander Varshavsky.

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

PERSPECTIVES Francisco) found that a cyclin was degraded by the ubiquitin pathway (TIMELINE)14. Functions for ubiquitylation in key cellular processes continue to emerge at an alarming rate, as anyone who tries to keep up with the literature surrounding this field knows only too well (BOX 1). Piecing together the 26S proteasome

By 1984, it was known that ubiquitylated proteins are targeted for degradation by an ATPdependent protease15, but the identity of the protease was a mystery. Marty Rechsteiner and co-workers16,17 in Salt Lake City discovered and partially purified the 26S proteasome by monitoring its ability to degrade ubiquitin–protein conjugates. Rechsteiner et al.18 also suggested that the 20S proteasome shares subunits with the 26S enzyme. But when the Haifa laboratory showed that the large protease is formed through the association of three multisubunit components19 and that one of these, CF3, is indeed identical to the 20S proteasome20,21, several independent strands of research began to intertwine the ubiquitin-dependent proteolysis pathway and the proteasome. The 20S proteasome corresponds to a previously named 19S ‘prosome’ particle22,23. The association of a 19S regulator with the 20S core to form the 26S particle was shown in several laboratories. The 19S regulator was shown to contain six ATPases together with a cohort of other proteins, including the first receptor for multi-ubiquitylated proteins24. The essential and non-redundant functional roles of the ATPases, together with the activity of the 26S proteasome in controlling the cell cycle, was first shown by using elegant yeast genetics25,26. A seminal advance was made in Daniel Finley’s laboratory by using yeast lacking a proteasomal receptor for multi-ubiquitylated proteins. Ion-exchange chromatography of these mutant proteasomes resulted in separation of the so-called 19S ‘bases’ and ‘lids’: the former contain the six ATPases plus two non-ATPase subunits; the latter contain non-ATPases27. The ATPases have chaperone activity28 and are presumed to be involved in unfolding proteins for entry into the 20S core, which contains the proteolytic activities. Crystallography has revealed the elegant three-dimensional structure of the 20S core29,30 and emphasizes that the catalytic sites are compartmentalized within the cylindrical particle, necessitating that unfolded proteins must be inserted into the central chamber for proteolysis to occur. Ubiquitin and disease

Evidence for malfunction of the ubiquitin– proteasome system is now beginning to

emerge in various unrelated human diseases, from cancer to neurological disorders (BOX 1). But perhaps the earliest observations were in relation to human chronic neurodegenerative diseases, where ubiquitin immunocytochem-

istry not only revealed ubiquitylated proteins in intraneuronal inclusions in all the main neurodegenerative diseases (for example, Alzheimer’s, Parkinson’s and Huntington’s diseases, and amyotrophic lateral sclerosis)31,

Box 1 | Ubiquitylation in health and disease Process

Substrate (X)

Signal transduction

β-catenin

SCF-TrCP

EGF receptor

c-Cbl

Transcription

HIF IκB

pVHL SCF-TrCP

Cell cycle

p53

MDM2; E6-AP* SCF; APC SCF

Cyclins CDK inhibitors

Cancer Gankyrin

26S proteasome

Ub-X

EBV

Antigen processing

MHC class I antigens

Parkin

E6-AP‡

Juvenile-onset familial Parkinsonism

ER quality control

Misfolded proteins

Angelman's syndrome

CMV

Alzheimer's disease

*Induced by papilloma viruses. ‡Mutated in Angelman's syndrome.

Studies over the past ten years or so have revealed that ubiquitylation is involved in the regulation of a host of cellular processes, and is disrupted in several diseases. The selected examples below give a flavour of this system’s versatility, and also hint at the disastrous consequences when ubiquitylation and protein degradation malfunction. Signal transduction The multiple protein-dependent steps in signal transduction pathways offer excellent targets for regulated proteolysis. Examples are numerous and include receptor tyrosine-kinase dependent pathways involving an E3 enzyme36 and the Hedgehog and Wnt/Wingless pathways37. Transcription Transcription factors are short-lived proteins that are eliminated by the ubiquitin system5. Examples abound, but the degradation of IκB has been intensively studied since the discovery that degradation of IκB activates NF-κB-dependent gene expression38. Cell cycle Progression through the cell cycle requires cyclin-dependent kinases together with activators (cyclins) and inhibitors (for example, p27). Ubiquitin-regulated degradation of these and other cell-cycle proteins ensures temporal control of the cycle39. MHC class I antigen processing Peptides derived from proteins bind to MHC class I molecules to trigger the cytotoxic lymphocyte response — proteasomes are a major producer of small peptides for this purpose40. The cytokine interferon-γ induces a specific set of proteolytic and regulatory subunits41 that produce better peptides for class I presentation42. ER quality control Proteins regularly become misfolded in the endoplasmic reticulum (ER). In yeast, misfolded proteins can be retrogradely extracted from the ER and degraded by the 26S proteasome43. ER proteins can be similarly extracted in higher eukaryotes44. The ER degradation system, together with the unfolded protein response, controls the well-being of the ER lumen and the secretory pathway. Cancer Several oncogenes and tumour suppressors influence the ubiquitin–proteasome pathway; for example the von Hippel–Lindau tumour suppressor (pVHL) is a component of a ubiquitin protein ligase45, and the liver oncoprotein gankyrin interacts with a regulatory ATPase of the 26S proteasome46. Neurological disorders Apart from the accumulation of proteins in the main neurodegenerative diseases, a mutated E3 is associated with Angelman’s syndrome — a developmental neurological disorder47. Viral infection Viral proteins subvert the proteasome by binding to subunits of both the 20S core and 19S regulator48 to minimize the MHC class I response and promote viral replication.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 4 7

PERSPECTIVES but also led to the identification of a new neurological illness — dementia with Lewy bodies — which is the second most common cause of cognitive decline in the elderly after Alzheimer’s disease32. Astoundingly, 76 years elapsed between the discovery of Lewy bodies and the observation that they contain ubiquitylated proteins.

1. 2. 3. 4.

5. 6.

The isopeptide bond does it all

There were early indications that protein ubiquitylation is related to the endosome–lysosome proteolytic system33. This link has recently been strengthened by the remarkable findings of Yoshinori Ohsumi and colleagues in Okazaki34, which show that autophagy, an important lysosome-requiring process for cellular protein degradation in adaptation to starvation, is controlled through conjugation of novel proteins by a set of enzymes similar to those of the ubiquitin pathway (FIG. 1). This work emphasizes the fact that the ubiquitous isopeptide bond has been used throughout evolution to control the two main systems of intracellular proteolysis. With ubiquitin helping to control the endocytic pathway35, it is conceivable that the main protein degradation pathways in cells are controlled by isopeptide bond formation — a triumph for evolution indeed! As proteomics blooms, it is salutary to think that different forms of post-translational modification increasingly emerge as controllers of biological processes. Which modifications are the front runners? Does it matter? It is already clear that phosphorylation works hand in glove with proteolysis to regulate many aspects of the chemistry of life. The current state of the intracellular proteolysis literature indicates that comprehension of cell physiology may soon be dramatically changed — selective degradation of regulatory proteins is at the heart of life. Our perception of cell biology, medicine and pharmaceutical intervention is at a point equivalent to the discovery of plate tectonics for geology. R. John Mayer is in the Laboratory for Intracellular Proteolysis, School of Biomedical Sciences, University of Nottingham Medical School, Queen’s Medical Centre, Nottingham, NG7 2UH, UK. e-mail: john.mayer@nottingham.ac.uk

Links

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

FURTHER INFORMATION Regulatory subunits

148

24.

25.

26.

Carroll, L. Alice’s Adventures in Wonderland. Through the looking glass (Penguin Books, London, 1973). Schoenheimer, R. The Dynamic State of Body Constituents (Harvard Univ. Press, Boston, 1942). Blobel, G. in Ciba Foundation Symposium Vol. 75, 398 (Excerpta Medica, Amsterdam, 1980). Wilkinson, K. D., Urban, M. K. & Haas, A. L. Ubiquitin is the ATP-dependent proteolysis factor I of rabbit reticulocytes. J. Biol. Chem. 255, 7529–7532 (1980). Hershko, A. & Ciechanover, A. The ubiquitin system. Annu. Rev. Biochem. 67, 425–480 (1998). Hershko, A., Eytan, E., Ciechanover, A. & Haas, A. L. Immunochemical analysis of the turnover of ubiquitin–protein conjugates in intact cells. Relationship to the breakdown of abnormal proteins. J. Biol. Chem. 257, 13964–13970 (1982). Pickart, C. M. & Rose, I. A. Functional heterogeneity of ubiquitin carrier proteins. J. Biol. Chem. 260, 1573–1581 (1985). Huibregtse, J., Scheffner, M. & Howley, P. M. Cloning and expression of the cDNA for E6-AP, a protein that mediates the interaction of the human papillomavirus E6 oncoprotein with p53. Mol. Cell. Biol. 13, 775–784 (1993). Finley, D., Ciechanover, A. & Varshavsky, A. Thermolability of ubiquitin-activating enzyme from the mammalian cell cycle mutant ts85. Cell 37, 43–55 (1984). Ciechanover, A., Finley, D. & Varshavsky, A. Ubiquitin dependence of selective protein degradation demonstrated in the mammalian cell cycle mutant ts85. Cell 37, 57–66 (1984). Zachariae, W. & Nasmyth, K. Whose end is destruction: cell division and the anaphase promoting complex. Genes Dev. 13, 2039–2058 (1999). Jentsch, S., McGrath, J. & Varshavsky, A. The yeast DNA repair gene RAD6 encodes a ubiquitin-conjugating enzyme. Nature 329, 131–134 (1987). Goebl, M. G. et al. The yeast cell cycle gene CDC34 encodes a ubiquitin-conjugating enzyme. Science 241, 1331–1335 (1988). Glotzer, M., Murray, A. W. & Kirschner, M. W. Cyclin is degraded by the ubiquitin pathway. Nature 349, 132–138 (1991). Hershko, A., Leshinsky, E., Ganoth, D. & Heller, H. ATPdependent degradation of ubiquitin–protein conjugates. Proc. Natl Acad. Sci. USA 81, 1619–1623 (1984). Hough, R., Pratt, G. & Rechsteiner, M. Ubiquitin–lysozyme conjugates. Identification and characterization of an ATP-dependent protease from rabbit reticulocyte lysates. J. Biol. Chem. 261, 2400–2408 (1986). Hough, R., Pratt, G. & Rechsteiner, M. Purification of two high molecular weight proteases from rabbit reticulocyte lysate. J. Biol. Chem. 262, 8303–8313 (1987). Hough, R., Pratt, G. & Rechsteiner, M. in Ubiquitin (ed. Rechsteiner, M.) 101–134 (Plenum Press, New York, 1988). Ganoth, D., Leshinsky, E., Eytan, E. & Hershko, A. A multicomponent system that degrades proteins conjugated to ubiquitin. Resolution of factors and evidence for ATP-dependent complex formation. J. Biol. Chem. 263, 12412–12419 (1988). Eytan, E., Ganoth, D., Armon, T. & Hershko, A. ATPdependent incorporation of 20S protease into the 26S complex that degrades proteins conjugated to ubiquitin. Proc. Natl Acad. Sci. USA 86, 7751–7755 (1989). Armon, T., Ganoth, D. & Hershko, A. Assembly of the 26S complex that degrades proteins ligated to ubiquitin is accompanied by the formation of ATPase activity. J. Biol. Chem. 265, 20723–20726 (1990). Arrigo, A.-P., Tanaka, K., Goldberg, A. L. & Welch, W. J. Identity of the 19S ‘prosome’ particle with the large multifunctional protease complex of mammalian cells (the proteasome). Nature 331, 192–194 (1988). Falkenburg, P. E. et al. Drosophila small cytoplasmic 19S ribonucleoprotein is homologous to the rat multicatalytic proteinase. Nature 331, 190–192 (1988). Dubiel, W. & Rechsteiner, M. The 19S regulatory complex of the 26S proteasome. Adv. Mol. Cell Biol. 27, 129–163 (1998). Gordon, C., McGurk, G., Dillon, P., Rosen, C. & Hastie, N. Defective mitosis due to a mutation in the gene for a fission yeast 26S protease subunit. Nature 366, 355–357 (1993). Ghislain, M., Udvardy, A. & Mann, C. S. cerevisiae 26S protease mutants arrest cell division in G2/metaphase. Nature 366, 358–361 (1993).

27. Glickman, M. H. et al. A subcomplex of the proteasome regulatory particle required for ubiquitin conjugate degradation and related to the COP-9-signalasome and eIF3. Cell 94, 615–623 (1998). 28. Braun, B. C. et al. The base of the proteasome regulatory particle exhibits chaperone-like activity. Nature Cell Biol. 1, 221–226 (1999). 29. Lowe, J. et al. Crystal structure of the 20S proteasome from the archeon T. acidophilum at 3.4 Å resolution. Science 268, 533–539 (1995). 30. Groll, M. et al. Structure of 20S proteasome from yeast at 2.4 Å resolution. Nature 386, 463–471 (1997). 31. Lowe, J., Mayer, R. J. & Landon, M. Ubiquitin in neurodegenerative diseases. Brain Pathol. 3, 55–65 (1993). 32. McKeith, I. G. et al. Clinical and pathological diagnosis of dementia with Lewy bodies (DLB). Report of the CDLB international workshop. Neurology 47, 1113–1124 (1996). 33. Doherty, F. J. et al. Ubiquitin–protein conjugates accumulate in the lysosomal system of fibroblasts treated with cysteine protease inhibitors. Biochem. J. 263, 47–55 (1989). 34. Mizushima, N. et al. A protein conjugation system essential for autophagy. Nature 395, 395–398 (1998). 35. Hicke, L. Gettin’ down with ubiquitin: turning off cellsurface receptors, transporters and channels. Trends Cell Biol. 9, 107–112 (1999). 36. Joazeiro, C. A. P. et al. The tyrosine kinase negative regulator c-Cbl as a RING-type E2-dependent ubiquitin protein ligase. Science 286, 309–312 (1999). 37. Jiang, J. & Struhl, G. Regulation of the Hedgehog and Wingless signalling pathways by the F-box/WD40repeat protein Slimb. Nature 391, 493–496 (1998). 38. Palombella, V. J., Rando, O. J., Goldberg, A. L. & Maniatis, T. The ubiquitin–proteasome pathway is required for processing the NFκB1 precursor protein and the activation of NFκB. Cell 78, 773–785 (1994). 39. Huibregtse, J., King, R. W., Deshaies, R. J., Peters, J.M. & Kirschner, M. W. How proteolysis drives the cell cycle. Science 274, 1652–1659 (1996). 40. Groettrup, M., Soza, A., Kuckelkorn, U. & Kloetzel, P. M. Peptide antigen production by the proteasome: complexity provides efficiency. Immunol. Today 17, 429–435 (1996). 41. Gaczynska, M., Rock, K. L. & Goldberg, A. L. Interferon and expression of MHC genes regulate peptide hydrolysis by proteasomes. Nature 365, 264-267 (1993). 42. Dick, T. P. et al. Coordinated dual cleavages induced by the proteasome regulator PA 28 lead to dominant MHC ligands. Cell 86, 253–256 (1996). 43. Hiller, M. M., Finger, A., Schweiger, M. & Wolf, D. H. ER degradation of a misfolded luminal protein by the cytosolic ubiquitin–proteasome pathway. Science 273, 1725–1728 (1996). 44. Plemper, R. K. & Wolf, D. H. Retrograde protein translocation: ERADication of secretory proteins in health and disease. Trends Biochem. Sci. 24, 266–270 (1999). 45. 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). 46. Higashitsuji, H. et al. Reduced stability of retinoblastoma protein by gankyrin, an oncogenic ankyrin-repeat protein overexpressed in hepatomas. Nature Med. 6, 96–99 (2000). 47. Kishino, T., Lalande, M. & Wagstaf, J. UBE3A/E6-AP mutations cause Angelman’s syndrome. Nature Genet. 15, 70–73 (1997). 48. Ferrell, K., Wilkinson, R. M., Dubiel, W. & Gordon, C. Regulatory subunit interactions of the 26S proteasome, a complex problem. Trends Biochem. Sci. 25, 83–88 (2000).

Acknowledgements I thank Avram Hershko and Aaron Ciechanover (Haifa), Alex Varshavsky (Pasadena), Wolfgang Dubiel (Berlin), Dieter Wolf (Stuttgart), Mark Hochstrasser (Chicago), Peter Zwickl (Martinsreid), Cecile Pickart (Baltimore), Keith Wilkinson (Atlanta), Alan Weissman (Washington), Ron Hay (St Andrews), and Simon Dawson, Michael Landon, Andy Alban and Rob Layfield (Nottingham) for help with this article; Rohan Baker (Canberra) for critically reviewing the manuscript; and the MRC, BBSRC, Wellcome Trust and EU Framework IV for support of some of the quoted work. Numerous pivotal contributions have been omitted due to space constraints; many thanks to the ‘unsung heroes’.

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

PERSPECTIVES

TIMELINE

Biological machines: from mills to molecules Marco Piccolino Although scientific progress is usually represented as being linear, it may, in fact, have a cyclical character — some discoveries may be forgotten or lost (at least temporarily), and themes may reappear through the centuries. Consider, for example, the concept of ‘molecular machines’, from the exciting phase of research that flourished in the seventeenth century, to the idea of machines that is at centre stage in modern cell biology.

More than three centuries ago, the birth of modern life sciences was marked by the idea that body function is based on organic

machines whose performance can be explained by similar laws to those operating in man-made machines. In the seventeenth century, this concept was used not only to explain functions that obviously reflected those of mechanical devices (such as skeletal and articular motion or the action of muscles), but also for other operations — digestion, sensation, fermentation and production of blood, for example1–3. To account for these more delicate operations of animal economy, body machines were thought to involve tiny components that could escape detection by the naked eye. This view derived, in part, from a recurrence of the physicists’ view that the Universe is composed of atoms. In Greek classical science this view was advocated by Democritus, and in the seventeenth century by the French philosopher and scientist Pierre Gassendi. As Marcello Malpighi (FIG. 1), one of the greatest seventeenth-century life scientists1,2, put it4: “Nature, in order to carry out the marvellous operations in animals and plants, has been pleased to construct their organized bodies with a very large number of machines, which are of necessity made up of extremely minute parts so shaped and situated, such as to form a marvellous organ, the composition of which are usually invisible to the naked eye, without the aid of the microscope.” The rise of machines

Figure 1 | Marcello Malpighi (from the Opera Postuma, 1798 Venetian in Folio edition). Malpighi, a prominent scientist in the seventeenth century, was one of the first to attribute body function to an organized series of minute ‘organic machines’. The concept underlying his metaphor of the ‘angel and the mill’ prompted anatomical investigations, which laid the foundation for modern microscopic anatomy. (Image courtesy of the library G. Romiti of the Anatomical Institute of the University of Pisa.)

Until the sixteenth and seventeenth centuries, most progress in the life sciences and medicine elaborated on classical doctrines, dating from Hippocrates, Aristotle and Galen (BOX 1, overleaf). But in the seventeenth century, interest in experimental studies exploded because, as had happened in astronomy and physics, new investigations cast doubt on the infallibility of the Ancients. In particular, the discovery of blood circulation, published in 1628 by William Harvey5 — and the subject of some debate at the moment6–8 — questioned the very foundation of classical physiology on which the whole body structure was interpreted. In the wave of the scientific revolution promoted by Galileo9, a conceptual break-

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

through was the idea that body function can be explained by similar physical laws to those that account for the action of artificial machines. This idea was elaborated on philosophical grounds by René Descartes10, and developed into a scientific manifesto by investigators such as Giovanni Alfonso Borelli and Malpighi11,12. As a result, physiology no longer needed to depend on metaphysical theories for the interpretation of body functions. Instead, like astronomy and new physics, it could become a ‘true’ science — an investigation that combined experimental study with the application of the ‘laws of mathematics and geometry’ to body machines. One result of this new scientific attitude was that scientists were discouraged from searching for the ultimate causes of ‘vital processes’. This was vividly expressed by Malpighi, in a beautiful passage from his Opera postuma4: “The way our soul uses the body in operating is ineffable, yet it is certain that in the operations of growth, sensation and motion the soul is forced in conformity with the machine on which it is acting, just as a clock or a mill is moved in the same way by a pendulum or lead or stone, or by an animal or by a man; indeed if an angel moved it, he would produce the same motion with changes of positions as the animals or agents do. Hence, even though I did not know how the angel operates, if on the other hand I did know the precise structure of the mill, I would understand this motion and action, and if the mill were out of order, I would try to repair the wheels or the damage to their structure without bothering to investigate how the angel moving them operated.”

To know how a machine operates, you need to know its structure. So the idea of ‘organic machines’ prompted anatomical investigations — both classical, descriptive macroscopic anatomy, and a new,‘subtle anatomy’, based on the use of newly invented techniques (some of which were the precursors of modern histological methods). It is no surprise, then, that the basis of the modern microscopic anatomy of animals and plants emerged in the seventeenth century, owing to the work of Malpighi and many others4,12–17. As had happened with Galileo’s astronomical observations, this new investigative attitude was not due simply to the availability of new technology, but also to the new cultural climate. Decline and fall

The climate changed in the eighteenth century, as interest in microscopic studies dwin-

VOLUME 1 | NOVEMBER 2000 | 1 4 9

PERSPECTIVES

Box 1 | The four humours Although anatomy was a part of classical medicine, it was not used to investigate body function. According to the doctrine of four humours — blood, yellow bile or choler, phlegm, and black bile or melancholy — the body and its organs were conceived as the stage where the humours interacted (depending on astronomical, atmospheric, climatic or other influences). Health and good temper resulted when humours were in correct proportions and mixing was appropriate (‘to mix’ in Latin is temperare). Conversely, diseases, or bad tempers, were produced when one humour was in excess or the mixture was inappropriate. This idea hardly favoured anatomical investigation and certainly did not promote the study of the structure of organs and tissues. Indeed, many organs, including the liver and lung, were considered to consist of effused blood (parenchyma), and so were thought to be devoid of a real internal structure. Similarly, small animals, such as insects, were thought to lack an internal structural organization.

dled. This was partly due to the apparent failure of the investigative programme, based on mechanistic explanations of body function, that had dominated the seventeenth century. Although many discoveries were made during that time, such as the structures of blood capillaries and alveoli in the lungs, the possibilities of explaining life processes on a simple, mechanical basis were limited. So the idea of mechanical body machines was largely abandoned. Instead, interest moved towards new forces — particularly electricity which, together with the study of gas (‘airs’) and chemistry, took centre stage. Electricity was particularly attractive as a principle for explaining vital processes because its application could produce movements in paralytic limbs or in animal preparations. As well as muscle contraction it was easy to evoke an electrical mechanism for nerve conduction, owing to the easy and rapid propagation of electricity, which seemed to match the rapid flow of sensation along nerve fibres18.

es. Different kinds of stimuli (chemical, mechanical or electrical) could excite muscle irritability, which was normally brought about by the action of a nerve. But Haller believed that the nerve’s influence was not the real cause of the muscle contraction — instead, it acted only as a stimulating (or exciting) factor that activated the intrinsic irritability of muscles. Haller’s ideas spread over Europe, causing lively discussions and dividing physiologists into ‘Hallerians’ or ‘anti-Hallerians’. Hallerians claimed that irritability had the

Machines revisited

Sensibility and irritability

It was a conceptual advance in the second half of the eighteenth century that sowed the seeds for the modern idea of machines. This came from the idea of ‘irritability’, conceived by the Swiss physiologist Albrecht von Haller (FIG. 2) in 1752. On the basis of animal experiments, Haller concluded that ‘sensibility’ (the ability to perceive a stimulus) and ‘irritability’ (the ability to react to that stimulus with a contraction) were different properties of living tissue, pertaining typically to nerves (sensibility) and muscles (irritability)19. Haller confessed that he could not ascertain the mechanism of irritability, but suggested that it depended on an essential constituent of living tissue (the gluten). He distinguished irritability — a vital property — from elasticity, which has purely physical properties and is unrelated to vital process-

150

same function in living matter as the Newtonian idea of gravity had in the inorganic world. In fact, Haller’s reluctance to propose a mechanistic explanation for irritability paralleled Newton’s aversion to proposing hypotheses about the mechanism of gravity. The point to emerge from these discussions was that an organism’s response to a stimulus is not a purely physical consequence of that stimulus, but that it reflects the organism’s internal organization. In other words, the response (in the case of irritability, contraction of a muscle) is what the organism is prepared — we would now say ‘programmed’ — to produce. The energy of the response is ‘enclosed’ in the organism, and does not come from the energy of the stimulus. So what really matters is the information encoded by the stimulus. In the first half of the nineteenth century, a similar idea was behind the development, by Johannes Müller, of the doctrine of ‘specific nervous energy’, according to which the sensation aroused by the stimulation of a sensory structure does not depend on the characteristics of the stimulus, but on the type of structure stimulated20.

Figure 2 | Albrecht von Haller (from the first edition of his Elementa physiologiae42). Haller, of Swiss origin, was a leading figure in eighteenthcentury physiology. He conceived the idea of ‘sensibility’ and ‘irritability’ to explain the body’s reaction to stimulus. In his formulation of the concept of irritability to account for muscle contraction, he first acknowledged, although in an implicit way, the importance of information flow in biological systems. (Image courtesy of the library G. Romiti of the Anatomical Institute of the University of Pisa.)

Haller’s ideas laid the foundation for the development of another fundamental idea in the nineteenth century: that of the ‘internal milieu’, developed by Claude Bernard21 in 1865. Bernard attempted to found medicine as a true science, based on the laws of physics and chemistry. He studied the characteristics of living organisms that seemed to elude physico-chemical principles, such as their relative independence of the conditions of the external environment (milieu cosmique). He attributed these characteristics to the organizational complexity of organisms, and often referred to the body or its working components as ‘machines’ (although his machines were more operational than structural devices). For example, he discovered the liver’s ability to synthesize sugar not because he studied the morphological structure of this organ, but because he used chemical analysis to follow the fate of blood sugar passing through the liver. This typifies the study of body machines in the eighteenth and nineteenth centuries — the emphasis was not on knowing the minute structures responsible for physiological responses such as contractions or sensations, but rather on studying their operation. In part, this was due to the lack of knowledge about the organization of living tissues. For instance, cellular theory was www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

PERSPECTIVES

4H+ 4H+ a

b b γ

δ β

ε β

Figure 3 | The metaphor of a ‘machine’, applied to living organisms. Compare an old, manually operated hydraulic machine (left) to the rotary ATP synthase of modern molecular biology (right). Both machines are reversible with minor readjustment. In the molecular machine, electrochemical energy in a proton gradient is normally used to produce rotary movement and ATP, but the machine can also work in reverse to produce an electrochemical gradient at the expense of ATP (figure adapted from REF. 43). In the man-made machine, the hydraulic potential energy could be converted into mechanical work that the man could use (from the ‘Stanzino delle Matematiche’ Museo degli Uffizi, Florence; © by ‘Ministero Affari Culturali’ of Italy).

developed only around 1839 by Matthias Schleiden22 and Theodor Schwann 23, and much time elapsed before there was any real knowledge about genetic laws, the existence and structure of membranes, the functions of proteins and enzymes, and the existence of hormones and other chemical messengers. In the absence of adequate knowledge, attempts to devise mechanistic hypotheses of biological phenomena were likely to fail. First-generation biochemists

Another reason for the lack of interest in the minute organization of body structure was the growing importance of chemistry in biological studies in the eighteenth century. For example, the discovery by Antoine Laurent Lavoisier, Pierre Simon de Laplace24 and Lazzaro Spallanzani 25, that a process akin to combustion occurs in living tissues, had great biological relevance26. In the following century it became increasingly evi-

dent that many functions of living organisms depend on chemical reactions. A chemical reaction typically occurs in a solution, and involves particles that move by diffusion and collide randomly with one another. Similarly, within an organism, chemical reactions seemed to require a liquid medium, and did not depend on the existence of particular structures. So it is not by chance that Claude Bernard developed his idea of a liquid internal milieu just when biologists were becoming interested in chemistry. Interest focused on those reactions that could, potentially, produce the energy necessary for life. In his book Reflections on Muscle 27, Andrew Huxley remarked that the relative lack of interest in the structural details of biological processes partly explains why the observations made around 1880 by Theodor Engelmann28 — of characteristic changes in the dimensions of band

Box 2 | Modern molecular machines Today, biology is revealing the importance of ‘molecular machines’ and of other highly organized molecular structures that carry out the complex physico-chemical processes on which life is based. There are many diverse molecular machines: • The photosynthetic system and complex devices that produce ATP. • DNA replication and protein translation apparatus. • Enzymatic cascade of phototransduction. • The integrated membrane system, involving ionic pumps and channels, that produces ionic gradients and generates electric differences across membranes; this underlies the production of electric signals in nerve fibres. • Machines that convert chemical energy into mechanical energy during muscle contraction or flagellar motion. • Finely integrated metabolic cycles and networks, including the system involved in antigen recognition and antibody production, the integrated system of hormones, extracellular molecules and intracellular messengers that are connected by many control pathways.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

patterns in muscle fibres during contraction — did not immediately result in an anticipation of the ‘sliding theory’ that was eventually formulated in the 1950s. Many scientists were neither interested nor confident in microscopy, preferring to visualize muscle contraction as the result of shortening of a muscular protein, fuelled by a chemical process akin to those being discovered in fermentation reactions27,29. The first generation of biological chemists were more interested in breaking down the cellular and subcellular components of living tissues to make them amenable to chemical analysis than in adjusting their chemical techniques to the complex components of biological materials. On the other hand, a biochemical approach, combined with physiological and clinical investigation, was fundamental in developing the concept of a ‘hormone’ at the turn of the nineteenth century (the word was introduced by Bayliss and Starling in 1905). It soon became clear that hormones, together with nervous reactions, were essential for regulating body function and maintaining stability of the internal milieu. This led Walter Cannon, in 1925, to propose the idea of ‘homeostasis’30. Through the study of hormones and other chemical messengers it became clear that, besides being involved in metabolic processes and in other chemical actions, molecules can carry important information in biological systems. Moreover, these molecules might relay information through specific receptor and effector systems. The idea of catalysis also emerged through chemistry. Biological materials were found to have specific and highly efficient catalysts, termed ‘enzymes’ by Willy Kühne31 in 1877. The study of enzymes (and of other proteins, as well as large molecules such as nucleic acids) was, in fact, behind the resurgence of interest in the idea of ‘minute machines’ during the twentieth century32. Twentieth-century machines

It became increasingly clear that the function of enzymes depends not only on their elementary chemical composition, but also on the configuration of their components. For example, effective interactions between enzymes, substrates and cofactors depend on the spatial arrangement of the interacting elements. This insight led to interest in the structure of complex molecules. It was also evident that the function of enzymes and other biological molecules could be regulated through specific control mechanisms. For instance, in 1963, Jacques

VOLUME 1 | NOVEMBER 2000 | 1 5 1

PERSPECTIVES Monod, Jean Pierre Changeux and François Jacob33 introduced the concept of allosteric regulation — that enzymatic action can be regulated by chemical signals acting on sites other than the enzyme’s catalytic site. This has since provided a reference for interpreting mechanisms involving molecules and systems that differ from those based on typical enzymatic actions; for instance, ligand–receptor interactions and various modulatory actions. An important advance has been the recognition that complex receptor assemblies are linked to secondmessenger systems through specialized proteins34,35 and that there is a flux of biological information. This information is carried by specific messengers, which act on systems that recognize them and develop specific responses. Through this complex flux of information, different mechanisms can be organized in more complex systems, resulting in highly integrated and efficient processes. The concept of information flux is also central to one of the biggest advances of twentieth-century biology — recognition of the molecular mechanisms responsible for transmitting genetic information and protein synthesis. These mechanisms involve the coding of genetic information by nucleic acids; transmission of this information through complex molecular devices that work at high rates with few errors; transcription of this information; translation into an amino-acid sequence; and finally, post-translational editing of this sequence32,36. Although these devices carry out basically chemical reactions, these reactions can no longer be considered as purely chemical processes due to unrestricted encounter-limited diffusion in a liquid medium. In fact, cellular compartments can hardly be considered typical liquid media. The idea now is that ‘structure’ is fundamental to the operation of modern molecular devices: for example, take the threedimensional arrangement of individual molecules; the spatial arrangement of proteins in sequential operations; and the arrangement of different proteins in a given process with respect to the membranes surrounding subcellular organelles or the cell as a whole. Given the importance of structure, modern biological pathways fully deserve the names “molecular and supramolecular machines”36,37. Ancient versus modern machines

To an extent, these extraordinary biological machines (BOX 2) realize the dream of the seventeenth-century scientists — a dream that

152

led Malpighi to suppose, more than three centuries ago, that “machines will be eventually found not only unknown to us but also unimaginable by our mind”38. If we consider that basically the same molecular device underlies ATP synthesis and bacterial flagellar motion, we see that modern biological machines correspond to the uniformity of nature pictured by Malpighi when he said4: “In its things Nature operates by necessity always in a uniform way. . . . Even though they appear disparate, the things of Nature are not so disconnected that one cannot observe a concatenation and uniformity in operating.”

However, in the importance of information flow, modern biological (and non-biological) machines differ from old machines, and surpass the expectations of the early life scientists. The old biological machines were supposed, at a minute level, to be “. . . made up of cords, filaments, beams, levers, tissues, fluids coursing here and there, cisterns, canals, filters, sieves and similar mechanisms”4. Besides the “fluids coursing here and there”, energy — rather than information — was thought to circulate through such components. No feedback mechanisms or control processes were predicted. The lack of an adequate concept of ‘information’ explains other difficulties encountered by early life scientists. For instance, it was impossible to come up with a reasonable theory of body development and the transmission of hereditary characteristics14,39,40. Some modern molecular devices, such as the rotary mechanism involved in ATP synthesis, may visually resemble the artificial machines that inspired the scientific revolution more than three centuries ago (FIG. 3). However, as well as the intrinsic regulatory mechanisms in what Paul Boyer called these new “splendid molecular machines”41, there are the regulatory actions based on information flux which, in this case, control the various phases of energetic metabolism culminating in ATP synthesis. With reference to Malpighi’s metaphor of the angel and the mill, we could perhaps say that, besides trying to understand the mechanisms of these biological wheels, modern scientists have started to picture how, by controlling a flux of signals through information networks, the angel regulates the complex machine of the living mill. Marco Piccolino is in the Dipartimento di Biologia, Università di Ferrara, Via Borsari 46, 44100 Ferrara, Italy. e-mail: marco.piccolino@unife.it

1. 2. 3. 4. 5. 6. 7.

8. 9. 10. 11. 12. 13. 14. 15. 16.

17. 18.

19.

20. 21. 22. 23.

24.

25. 26. 27. 28.

29. 30. 31.

32. 33.

Adelmann, H. B. Marcello Malpighi and the Evolution of Embryology — 5 Vols (Cornell, Ithaca, 1966). Belloni, L. Opere Scelte di Marcello Malpighi (UTET, Torino, 1967). Piccolino, M. Marcello Malpighi and the difficult birth of modern life sciences. Endeavour 23, 175–179 (1999). Malpighi, M. Opera Postuma (Churchill, London, 1697). Harvey, W. Exercitatio Anatomica de Motu Cordis et Sanguinis in Animalibus (Fitzeri, Frankfurt, 1628). Chau, P. -L. Ancient Chinese had their fingers on the pulse. Nature 404, 431 (2000).[nature contents page] Cheng, T. O. Did the Greeks beat Chinese on blood circulation… . Nature 405, 993 (2000).[nature contents page] Prioreschi, P. …or was ‘blood as the river of life’ just poetic? Nature 405, 993 (2000).[nature contents page] Galileo, G. Discorsi e Dimostrazioni Matematiche Intorno a Due Nuove Scienze (Elzeviri, Leyden, 1638). Descartes, R. De Homine (Leffen & Moyardum, Leyden, 1662). Borelli, G. A. De Motu Animalium — 2 Vols (Bernabò, Rome, 1680). Malpighi, M. Opera Omnia (Scott & Wells, London, 1686). Stensen, N. Observationes Anatomicae (Chouët, Leyden, 1662). Swammerdam, J. Bybel der Natuur (Severinus, Amsterdam, 1737). Leeuwenhoek, A. Opera Omnia (Langerak, Leyden, 1722). Croone, W. De Ratione Motus Musculorum (Hayes, London, 1664) Translated by P. Maquet in On the Reason of the Movement of the Muscles (American Philosophical Society, Philadelphia, 2000) Mayow, J. Tractatus Quinque Medico-Physici (Theatro Sheldoniano, Oxford, 1674). Galvani, L. De viribus electricitatis in motu musculari commentarius. Bon. Sci. Art. Inst. Acad. Comm. 7, 363–418 (1791). Haller, A. De partibus corporis humani sensibilibus et irritabilibus. Comm. Soc. Reg. Scient. Gottingensis 2, 114–158 (1753). Müller, J. Handbuch der Physiologie des Menschen (Hölscher, Coblenz, 1844). Bernard, C. Introduction à l’Étude de la Médecine Expérimentale (Baillière, Paris, 1865). Schleiden, M. J. Beiträge zur Phytogenesis. Arch. Anat. Physiol. Wiss. Med. 13, 137–176 (1838). Schwann, T. Mikroskopische Untersuchungen über die Übereinstimmung in der Struktur und dem Wachstum der Tiere und Pflanzen (Reimer, Berlin, 1839). Lavoisier, A. L. & de Laplace, P. S. in Oeuvres de Lavoisier Vol. I 528–530 (Imprimerie Royale, 1780, printed in 1864, Paris). Spallanzani, L. in Rapports de l’Air avec les Êtres Organisés (ed. Senebier, J.) (Paschoud, Genève, 1807). Keilin, D. The History of Cell Respiration and Cytochrome (Cambridge Univ. Press, Cambridge, 1966). Huxley, A. F. Reflections on Muscle (Liverpool Univ. Press, Liverpool, 1980). Engelmann, T. W. Mikrometrische Untersuchungen an contrahirten Muskelfasern. Arch. Ges. Physiol. 23, 571–590 (1880). Needham, D. M. Machina Carnis (Cambridge Univ. Press, Cambridge, 1971). Cannon, W. B. Organization for Physiological Homeostatics. Physiol. Rev. 9, 399–431 (1925). Kühne, W. Verh. Ueber da Verhalten Verschiedener Organisirter und Sog. Ungeformter Fermente. Naturhist.medic. Vereins Heidelb. 1, 190–193 (1877). Alberts, B. et al. Molecular Biology of the Cell 3rd edu (Garland, New York, 1994). Monod, J., Changeux, J. P. & Jacob, F. Allosteric proteins and cellular control systems. J. Mol. Biol. 6, 306–329 (1963).

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

PERSPECTIVES 34. Sutherland, E. W. Studies on the mechanism of hormone action. Science 177, 401–408 (1972). 35. Udrisar, D. & Rodbell, M. Microsomal and cytosolic fractions of guinea pig hepatocytes contain 100-kilodalton GTP-binding proteins reactive with antisera against alpha subunits of stimulatory and inhibitory heterotrimeric GTPbinding proteins. Proc. Natl Acad. Sci. USA 87, 6321–6325 (1990). 36. Alberts, B. The cell as a collection of protein machines: preparing the next generation of molecular biologists. Cell 92, 291–294 (1998). 37. Mitchell, P. & Moyle, J. Chemiosmotic hypothesis of oxidative phosphorylation. Nature 213, 137–139 (1967). 38. Malpighi, M. The Viscerum Structura (Montii, Bologna, 1666). 39. Malpighi, M. Dissertatio Epistolica de Formatione Pulli in

Ovo (Martyn, London, 1673). 40. Bonnet, C. Considérations sur les Corps Organisés (Rey, Amsterdam, 1762). 41. Boyer, P. D. The ATP synthase — a splendid molecular machine. Annu. Rev. Biochem. 66, 717–749 (1997). 42. Haller, A. Elementa Physiologiae Corporis Humani (Bousquet, Lausanne, 1757). 43. Rastogi, V. K. & Girvin, M. E. Structural changes linked to proton translocation by subunit c of the ATP synthase. Nature 402, 263–268 (1999).

Acknowledgements This article has benefited from discussions with A. Cattaneo of the International School for Advanced Studies (S.I.S.S.A.) of Trieste, and has been made possible by bibliographical help from L. Iannucci of the University of Pisa. I also thank L. Galli-Resta, A. Pignatelli and B. Pelucchi for critically reading the manuscript.

transport represent the movement of cytoskeletal and cytosolic proteins at much slower rates, and the nature of the carrier structures for these proteins is not known. Proteins that associate with neurofilaments and microtubules move in slow component ‘a’ at average rates of roughly 0.3–3 mm day–1 (~0.004-0.04 µm s–1), and proteins that associate with microfilaments, as well as many other cytosolic proteins, are transported in slow component ‘b’ at average rates of roughly 2–8 mm day–1 (~0.02–0.09 µm s–1) (TABLE 1). No movement en masse

OPINION

Slow axonal transport: stop and go traffic in the axon Anthony Brown Efforts to observe the slow axonal transport of cytoskeletal polymers during the past decade have yielded conflicting results, and this has generated considerable controversy. The movement of neurofilaments has now been seen, and it is rapid, infrequent and highly asynchronous. This motile behaviour could explain why slow axonal transport has eluded observation for so long.

Neurons communicate with other cells by extending cytoplasmic processes called axons and dendrites. Remarkably, axons can attain lengths of one metre or more, although they lack ribosomes and Golgi complexes. Axonal proteins and Golgi-derived vesicles are formed in the neuronal cell body and are shipped along the axon by a process called axonal transport. This movement is essential for the growth and survival of axons, and continues throughout the life of the nerve cell. Studies on axonal transport in laboratory animals with radioisotopic pulse labelling have shown that there are hundreds of axonally transported proteins, but that these proteins move at a small number of discrete rates, which can be categorized as either fast or slow. Each discrete rate component represents the movement of a largely distinct subset of proteins that are transported together throughout their journey along the axon. To explain these observations, Lasek and colleagues proposed the structural hypothesis of axonal transport, which postulates that all axonal proteins move by association with, or as integral parts of, subcellular carrier structures1. According to this hypothesis, each rate component represents

the movement of a unique type of macromolecular structure (TABLE 1). The fast components of axonal transport are now known to represent the anterograde and retrograde movement of distinct types of membranous organelles along microtubules at average rates of roughly 50–400 mm day–1 (~0.5–5 µm s–1), propelled by the action of molecular motor proteins2. Membranous organelles can therefore be considered to be the carrier structures for fast axonal transport. In contrast, the slow components of axonal

In radioisotopic pulse-labelling experiments, slow components ‘a’ and ‘b’ form unimodal asymmetrical waves, often loosely described as ‘bell-shaped’, which spread as they move along the axon towards the axon tip (FIG. 1). Each wave represents the concerted movement of many distinct proteins whose individual waveforms coincide. Early studies on slow axonal transport stressed the coherence of these transport waves but not the spreading, and this gave rise to the idea that cytoskeletal and cytosolic proteins move along the axon en masse, that is, in a slow and synchronous manner1. The expectation of a slow and synchronous movement has had a profound influence on the design of experiments aimed at detecting slow axonal transport. For example, many studies have used fluorescence photobleaching or photoactivation strategies in which fluorescent or caged fluorescent cytoskeletal proteins are injected into nerve cells and then a popula-

Table 1 | The moving structures of axonal transport* Rate class

Average rate

Moving structures

Composition (selected examples)

Fast anterograde

200–400 mm day–1 (≈2–5 µm s–1)

Golgi-derived vesicles and tubules (secretory pathway)

Synaptic vesicle proteins, kinesin, enzymes of neurotransmitter metabolism

Bi-directional

50–100 mm day–1 (≈0.5–1 µms–1)

Mitochondria

Cytochromes, enzymes of oxidative phosphorylation

Fast retrograde

200–400 mm day–1 (≈2–5 µm s–1)

Endosomes, lysosomes Internalized membrane (endocytic pathway) receptors, neurotrophins, active lysosomal hydrolases

Fast components

Slow components Slow component ‘a’ 0.3–3 mm day–1

Neurofilaments, microtubules‡

Slow component ‘b’ 2–8 Microfilaments, supramolecular mm day–1 (≈0.02–0.09 µm s–1) complexes of the cytosolic matrix

Neurofilament proteins, tubulin, spectrin, tau proteins Actin, clathrin, dynein, dynactin, glycolytic enzymes

*Data compiled from REFS 1,41,44. ‡ In some neurons, microtubule proteins are transported in slow component ‘b’ as well as slow component ‘a’.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 5 3

PERSPECTIVES action. Science 177, 401–408 (1972). 35. Udrisar, D. & Rodbell, M. Microsomal and cytosolic fractions of guinea pig hepatocytes contain 100-kilodalton GTP-binding proteins reactive with antisera against alpha subunits of stimulatory and inhibitory heterotrimeric GTPbinding proteins. Proc. Natl Acad. Sci. USA 87, 6321–6325 (1990). 36. Alberts, B. The cell as a collection of protein machines: preparing the next generation of molecular biologists. Cell 92, 291–294 (1998). 37. Mitchell, P. & Moyle, J. Chemiosmotic hypothesis of oxidative phosphorylation. Nature 213, 137–139 (1967). 38. Malpighi, M. The Viscerum Structura (Montii, Bologna, 1666). 39. Malpighi, M. Dissertatio Epistolica de Formatione Pulli in Ovo (Martyn, London, 1673).

40. Bonnet, C. Considérations sur les Corps Organisés (Rey, Amsterdam, 1762). 41. Boyer, P. D. The ATP synthase — a splendid molecular machine. Annu. Rev. Biochem. 66, 717–749 (1997). 42. Haller, A. Elementa Physiologiae Corporis Humani (Bousquet, Lausanne, 1757). 43. Rastogi, V. K. & Girvin, M. E. Structural changes linked to proton translocation by subunit c of the ATP synthase. Nature 402, 263–268 (1999).

Acknowledgements This article has benefited from discussions with A. Cattaneo of the International School for Advanced Studies (S.I.S.S.A.) of Trieste, and has been made possible by bibliographical help from L. LIannucci of the University of Pisa. I also thank L. Galli-Resta, A. Pignatelli and B. Pelucchi for critically reading the manuscript.

OPINION

Table 1 | The moving structures of axonal transport* Rate class

Average rate

Moving structures

Composition (selected examples)

Fast anterograde

200–400 mm day–1 (≈2–5 µm s–1)

Golgi-derived vesicles and tubules (secretory pathway)

Synaptic vesicle proteins, kinesin, enzymes of neurotransmitter metabolism

Bi-directional

50–100 mm day–1 (≈0.5–1 µms–1)

Mitochondria

Cytochromes, enzymes of oxidative phosphorylation

Fast retrograde

200–400 mm day–1 (≈2–5 µm s–1)

Endosomes, lysosomes Internalized membrane (endocytic pathway) receptors, neurotrophins, active lysosomal hydrolases

Fast components

Slow components Slow component ‘a’ 0.3–3 mm day–1

Neurofilaments, microtubules‡

Slow component ‘b’ 2–8 Microfilaments, supramolecular mm day–1 (≈0.02–0.09 µm s–1) complexes of the cytosolic matrix

Neurofilament proteins, tubulin, spectrin, tau proteins Actin, clathrin, dynein, dynactin, glycolytic enzymes

*Data compiled from REFS 1,41,44. ‡ In some neurons, microtubule proteins are transported in slow component ‘b’ as well as slow component ‘a’.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 1 | NOVEMBER 2000 | 1 5 3

PERSPECTIVES

Figure 1 | Kinetics of slow axonal transport. Diagram illustrating the kinetics of slow axonal transport, as revealed by radioisotopic pulse labelling. a | Radioactive amino acids injected into the vicinity of the neuronal cell body produce a transient pulse of newly synthesized radioactive proteins, which b,c | move together along the axon by axonal transport. After a time interval ranging from hours to months, the animal is killed, the nerve is excised and sliced into segments, and each segment is analysed biochemically to identify the radioactive proteins. The pulse-labelled proteins form an asymmetrical wave (red) that spreads as it moves along the axon.

154

in axons has been a vexing problem, but the most likely explanation is that cytoskeletal proteins do not move en masse in axons after all. Neurofilaments move in fits and starts

A recent breakthrough in the study of slow axonal transport has come from observations on neurofilament proteins, tagged with green fluorescent protein (GFP), in cultured rat sympathetic neurons14,15. These cultured neurons contain relatively few neurofilaments and frequently show discontinuities in their axonal neurofilament array, resulting in short segments of axon that lack neurofilaments14. Time-lapse imaging of these naturally occurring gaps in the axonal neurofilament array has enabled the observation of axonal transport without the need for photobleaching or photoactivation approaches. Contrary to expectations, neurofilaments move rapidly, with peak rates as high as 3 µm s–1, and these movements are frequently interrupted by prolonged pauses14,15 (FIG. 2). The average velocity excluding the pauses is about 0.2–0.3 µm s–1. Assuming an average transport rate in the range of 0.3–3 mm day–1 (TABLE 1), we can estimate that individual neurofilaments spend 83–99% of their time pausing during their journey down the axon. Radioisotopic pulse labelling studies in the mouse optic nerve led Nixon and Logvinenko17 to propose almost 15 years ago that there are two kinetically distinct populations of neurofilament proteins in axons, one that moves and one that is stationary. According to this model, neurofilament polymers or oligomers exchange between the moving and stationary phase as they move along the axon18. On the other hand, Lasek and colleagues16,19 have challenged this hypothesis, arguing that there is a single population of neurofilaments in axons that all move relentlessly, but at a broad range of rates. In principle, the alternating movements and pauses observed for GFP-tagged neurofilaments in cultured neurons14,15 could be regarded as transitions between two distinct moving and stationary phases, or simply as the intermittent movements of a single population of neurofilaments that move at a broad and continuous range of rates. Further studies will be required to distinguish between these two possibilities. Re-evaluating previous approaches

Why have previous attempts6,7 to observe the axonal transport of neurofilament proteins using fluorescence photobleaching failed to reveal movement? One possible explanation is that those studies were not capable of detecting the rapid movement of single cytoskeletal polymers. For example, it is important to note

that the extent of bleaching in the photobleaching studies on neurofilament proteins was only partial, reducing the fluorescence intensity in the axon to 20–50% of its initial value6. If the residual unbleached fluorescence in the bleached region exceeded the fluorescence intensity of a single neurofilament, then it is likely that the movement of neurofilaments across the bleached zone could have gone unnoticed. This could also apply to the photobleaching studies on actin and tubulin3–5,9,12,13. For example, in the study of Okabe and Hirokawa12 on tubulin, photobleaching reduced the fluorescence intensity in the axon to 10–40% of its initial value and, in other similar studies, researchers have estimated that as much as 10–20% of the total tubulin in the axon could have moved through the photobleached gaps without detection4. Similar detection limits have also been estimated for the fluorescence photoactivation technique8. The ability of the photobleaching experiments to detect the movement of cytoskeletal polymers may also have been hampered by the short length of the bleached regions (3–5 µm), and the relatively long time-lapse intervals (typically five minutes or more). These considerations suggest that the photobleaching and photoactivation strategies should be capable of detecting the slow axonal transport of cytoskeletal proteins if they could be optimized to enable the detection of single rapidly moving polymers. Anterograde

Time (s)

tion of these proteins is marked by bleaching or activating the fluorescence in a narrow band across the axon [see supplementary figure online]. In these experiments, a slow and synchronous movement should be manifested as a slow translocation of the marked zone towards the axon tip. However, most studies on tubulin, actin and neurofilament proteins using one or both of these techniques showed that the marked zone does not move3–9. Although gradual recovery of the fluorescence was observed after photobleaching, it had no obvious directionality and was therefore attributed to exchange between the bleached polymers and diffusible fluorescent subunits. Directional movement of the photobleached or photoactivated zone was observed in cultured frog neurons10–12, but it probably resulted from stretching of the axon owing to the rapid growth and poor adhesion of these neurons on laminin substrates11,13. The repeated failure of so many efforts to demonstrate slow synchronous movement of cytoskeletal proteins

Figure 2 | A neurofilament on the move. Timelapse images of a neurofilament moving through a naturally occurring gap in the axonal neurofilament array of a cultured nerve cell. The neurofilaments were visualized using green fluorescent protein (GFP)-tagged neurofilament protein M. The fluorescence images are shown in inverted contrast for greater clarity. Scale bar= 5 µm. (Figure adapted from REF. 14.) (See movie online).

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1

PERSPECTIVES Why so slow?

The rate of slow axonal transport in radioisotopic pulse labelling experiments is generally quoted as the rate of movement of the wave peak, but the spreading of the transport wave indicates that the radiolabelled proteins actually move at a broad range of rates16,19(FIG. 1). The motile behaviour of GFP-tagged neurofilaments described above suggests a model for slow axonal transport that can account for the slow rate and the spreading of the transport wave. Consider a pulse of radioactive neurofilament proteins that assemble into filaments in the neuronal cell body20. Let us assume that each neurofilament moves rapidly along the axon but that the overall rate of movement is slow because the filaments spend a large proportion of their time pausing. By chance, or perhaps due to intrinsic differences, some filaments move more frequently than others, and this causes the population to spread out as it moves along the axon. The frequency with which filaments move, or the amount of time that they spend pausing, could be determined simply by proximity to the transport machinery or substrate, or by local variations in the resistance to movement, or by some regulatory process. Neurofilaments that move most often will end up at the leading edge of the transport wave19, whereas neurofilaments that move least often will end up at the trailing edge16. According to this hypothesis, the transport wave represents the distribution of many thousands of neurofilaments whose individual movements and pauses are summed over the days, weeks or months that they spend travelling down the axon (FIG. 1). Polymers as carrier structures

The mechanism of slow axonal transport has been debated for almost 15 years, and most of the controversy has focused on the structural form in which the cytoskeletal subunit proteins move21,22. Some studies have concluded that cytoskeletal proteins move as assembled polymers23–28, and some have concluded that they move as unassembled subunits28–32, but none has been conclusive. The observations on GFP-tagged neurofilaments described above14,15 have shown unequivocally that neurofilament polymers do move in axons. Whether actin and tubulin also move in the form of assembled polymers in axons remains to be determined, although there is clearly precedent for such movements in non-neuronal cells (for example, REFS 33,34). Microtubule polymers have been observed to move in growth cones and developing axonal branches of cultured neurons35, whereas experiments using fluorescence-speckle microscopy have not detected any

Proximal

Distal

–

Microtubule

Retrograde motor (minus-end-directed)

Neurofilament

Anterograde motor (plus-end-directed)

Figure 3 | A model for the movement of neurofilaments in axons. In this model, neurofilaments are considered to move bidirectionally along microtubules through the action of a plus-end-directed motor such as a kinesin-related protein, and a minus-end-directed motor such as dynein. Note that axonal microtubules are all orientated with their plus-ends distal, towards the axon tip. Only a small fraction of the axonal neurofilaments move at any one point in time.

movement36. If the motility of microtubules is as rapid and infrequent as for neurofilaments, then it is possible that their movement might have gone undetected using the speckling technique. Slow axonal transport represents the movement of a myriad of other cytosolic proteins in addition to cytoskeletal proteins (TABLE 1). One attractive hypothesis is that these cytosolic proteins are transported by forming physical associations with moving cytoskeletal polymers1. The relatively simple protein composition of slow component ‘a’ indicates that neurofilaments and microtubules could be the sole carrier structures for this rate component; all of the proteins that move in slow component ‘a’ are either integral parts of these cytoskeletal polymers or are known to associate with these polymers in vivo. In contrast, the protein composition of slow component ‘b’ is extremely complex and includes more than 200 proteins, many of which are traditionally described as ‘soluble’1. The presence of actin indicates that microfilaments could function as carrier structures for this rate component. However, given the large number of diverse proteins in slow component ‘b’, it is likely that the carrier structures for this rate component are complex and may comprise several macromolecular complexes that move by direct or indirect association with the moving microfilaments. Motors and substrates

To understand the mechanism of slow axonal transport, we must identify not only the structural forms in which cytoskeletal and

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

cytosolic proteins move, but also the motors that move them, and the substrates that they interact with. Potential substrates for motordriven movements of cytoskeletal polymers in axons include the plasma membrane, the endoplasmic reticulum and other cytoskeletal filaments21. In principle, neurofilaments could move by direct interaction with molecular motors, or they could ride ‘piggyback’ by attachment to other moving structures. Evidence for a direct interaction has come from a recent report that neurofilaments purified from bovine spinal cord can move rapidly along microtubules in an ATPdependent manner in vitro37, at peak rates of up to 1 µm s–1. A similar motile mechanism has also been described for vimentin filaments and their precursors along microtubules in non-neuronal cells38. Slow axonal transport has generally been assumed to be exclusively anterograde, moving towards the axon tip, but in the observations on GFP-tagged neurofilaments described above, about 20–30% of the observed filaments actually moved in a retrograde direction, towards the cell body14,15. Similarly, in the in vitro study described above, neurofilaments were observed to move towards the minus as well as the plus ends of microtubules37. One possible explanation is that the retrogradely moving neurofilaments represent a distinct population, as proposed by Griffin and colleagues39 on the basis of their studies on the redistribution of cytoskeletal proteins in transected peripheral nerves. Alternatively, the retrograde movements could represent transient reversals of

VOLUME 1 | NOVEMBER 2000 | 1 5 5

PERSPECTIVES filaments that all eventually move in a net anterograde direction. If there is a distinct population of retrogradely moving neurofilaments in axons, previous studies indicate that it represents no more than 5% of the total transported neurofilament protein40. Bidirectional movement of neurofilaments along microtubules could be achieved by a plus-end-directed motor such as kinesin or a kinesin-related protein, and a minus-enddirected motor such as cytoplasmic dynein (FIG. 3). Dynein, dynactin and several putative kinesin-related proteins have been identified in neurofilament preparations by immunoblotting, and the retrograde movement of neurofilaments on microtubules in vitro can be partly inhibited by pharmacological inhibitors of dynein and by monoclonal antibodies specific for dynein intermediate chains37. A potential role for dynein as a slow axonal transport motor is also indicated by the fact that a substantial proportion of the dynein and dynactin in axons moves in slow component ‘b’41. Less is known about the potential roles of kinesin and kinesin-related proteins in slow axonal transport. Yabe et al.42 have reported that conventional kinesin, which is a known vesicle transport motor, associates with axonally transported neurofilaments, whereas Elluru et al.43 detected little or no conventional kinesin in either of the slow components. Further investigation of the axonal transport of kinesin and kinesin-related proteins is clearly required. Some questions for the future

The motile behaviour of neurofilaments in axons lends support to a general model for slow axonal transport characterized by the rapid, infrequent and highly asynchronous movement of cytoskeletal polymers and their associated proteins. Proof of this model will require identification of both the structural forms in which other cytoskeletal and cytosolic proteins move in slow axonal transport, and of the kinetics of their movement. For example, do actin and tubulin also move along axons as assembled polymers, and do cytoskeletal polymers serve as the carrier structures for cytosolic proteins? The answers to these questions may shed light on fundamental organizational principles of the cytoplasm that are applicable to all eukaryotic cells. Many questions also remain regarding the axonal transport of neurofilaments. For example, do these cytoskeletal polymers associate directly with motor proteins or do they ride piggyback on other moving structures? And what is the significance of the retrogradely moving neurofilaments in axons? There is still much to be learned, but our abil-

156

ity to observe the slow axonal transport of neurofilament polymers in living axons now permits, for the first time, direct analysis of the molecular mechanism of this remarkable, and once intractable, motile phenomenon. Anthony Brown is at the Neuroscience Program, Department of Biological Sciences, Ohio University, Athens, Ohio 45701, USA. e-mail: browna1@ohio.edu

Links FURTHER INFORMATION Movies of moving

neurofilaments | The Brown lab page 1.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

Lasek, R. J., Garner, J. A. & Brady, S. T. Axonal transport of the cytoplasmic matrix. J. Cell Biol. 99, S212–S221 (1984). Hirokawa, N. Kinesin and dynein superfamily proteins and the mechanism of organelle transport. Science 279, 519–526 (1998). Lim, S.-S., Sammak, P. J. & Borisy, G. G. Progressive and spatially differentiated stability of microtubules in developing neuronal cells. J. Cell Biol. 109, 253–263 (1989). Lim, S.-S., Edson, K. J., Letourneau, P. C. & Borisy, G. G. A test of microtubule translocation during neurite elongation. J. Cell Biol. 111, 123–130 (1990). Okabe, S. & Hirokawa, N. Turnover of fluorescently labelled tubulin and actin in the axon. Nature 343, 479–482 (1990). Okabe, S., Miyasaka, H. & Hirokawa, N. Dynamics of the neuronal intermediate filaments. J. Cell Biol. 121, 375–386 (1993). Takeda, S., Okabe, S., Funakoshi, T. & Hirokawa, N. Differential dynamics of neurofilament-H protein and neurofilament-L protein in neurons. J. Cell Biol. 127, 173–185 (1994). Sabry, J., O’Connor, T. P. & Kirschner, M. W. Axonal transport of tubulin in Ti1 pioneer neurons in situ. Neuron 14, 1247–1256 (1995). Takeda, S., Funakoshi, T. & Hirokawa, N. Tubulin dynamics in neuronal axons of living zebrafish embryos. Neuron 14, 1257–1264 (1995). Reinsch, S. S., Mitchison, T. J. & Kirschner, M. W. Microtubule polymer assembly and transport during axonal elongation. J. Cell Biol. 115, 365–379 (1991). Okabe, S. & Hirokawa, N. Differential behavior of photoactivated microtubules in growing axons of mouse and frog neurons. J. Cell Biol. 117, 105–120 (1992). Okabe, S. & Hirokawa, N. Do photobleached fluorescent microtubules move? Re-evaluation of fluorescence laser photobleaching both in vitro and in growing Xenopus axons. J. Cell Biol. 120, 1177–1186 (1993). Chang, S. H., Rodionov, V. I., Borisy, G. G. & Popov, S. V. Transport and turnover of microtubules in frog neurons depend on the pattern of axonal growth. J. Neurosci. 18, 821–829 (1998). Wang, L., Ho, C.-L., Sun, D., Liem, R. K. H. & Brown, A. Rapid movement of axonal neurofilaments interrupted by prolonged pauses. Nature Cell Biol. 2, 137–141 (2000). Roy, S. et al. Neurofilaments are transported rapidly but intermittently in axons: implications for slow axonal transport. J. Neurosci. 20, 6849–6861 (2000). Lasek, R. J., Paggi, P. & Katz, M. J. Slow axonal transport mechanisms move neurofilaments relentlessly in mouse optic axons. J. Cell Biol. 117, 607–616 (1992). Nixon, R. A. & Logvinenko, K. B. Multiple fates of newly synthesized neurofilament proteins: Evidence for a stationary neurofilament network distributed nonuniformly along axons of retinal ganglion cells. J. Cell Biol. 102, 647–659 (1986). Nixon, R. A. Dynamic behavior and organization of cytoskeletal proteins in neurons: reconciling old and new findings. Bioessays 20, 798–807 (1998). Lasek, R. J., Paggi, P. & Katz, M. J. The maximum rate of neurofilament transport in axons: a view of molecular transport mechanisms continuously engaged. Brain Res. 616, 58–64 (1993). Black, M. M., Keyser, P. & Sobel, E. Interval between the synthesis and assembly of cytoskeletal proteins in cultured neurons. J. Neurosci. 6, 1004–1012 (1986). Baas, P. W. & Brown, A. Slow axonal transport: the polymer transport model. Trends Cell Biol. 7, 380–384

(1997). 22. Hirokawa, N., Terada, S., Funakoshi, T. & Takeda, S. Slow axonal transport: the subunit transport model. Trends Cell Biol. 7, 384–388 (1997). 23. Terasaki, M., Schmidek, A., Galbraith, J. A., Gallant, P. E. & Reese, T. S. Transport of cytoskeletal elements in the squid giant axon. Proc. Natl Acad. Sci. USA 92, 11500–11503 (1995). 24. Ahmad, F. J. & Baas, P. W. Microtubules released from the neuronal centrosome are transported into the axon. J. Cell Sci. 108, 2761–2769 (1995). 25. Yu, W., Schwei, M. J. & Baas, P. W. Microtubule transport and assembly during axon growth. J. Cell Biol. 133, 151–157 (1996). 26. Slaughter, T., Wang, J. & Black, M. M. Microtubule transport from the cell body into the axons of growing neurons. J. Neurosci. 17, 5807–5819 (1997). 27. Ahmad, F. J., Echeverri, C. J., Vallee, R. B. & Baas, P. W. Cytoplasmic dynein and dynactin are required for the transport of microtubules into the axon. J. Cell Biol. 140, 391–401 (1998). 28. Galbraith, J. A., Reese, T. S., Schlief, M. L. & Gallant, P. E. Slow transport of unpolymerized tubulin and polymerized neurofilament in the squid giant axon. Proc. Natl Acad. Sci. USA 96, 11589–11594 (1999). 29. Terada, S., Nakata, T., Peterson, A. C. & Hirokawa, N. Visualization of slow axonal transport in vivo. Science 273, 784–788 (1996). 30. Funakoshi, T., Takeda, S. & Hirokawa, N. Active transport of photoactivated tubulin molecules in growing axons revealed by new electron microscopic analyses. J. Cell Biol. 133, 1347–1354 (1996). 31. Miller, K. W. & Joshi, H. C. Tubulin transport in neurons. J. Cell Biol. 133, 1355–1366 (1996). 32. Yabe, J. T., Pimenta, A. & Shea, T. B. Kinesin-mediated transport of neurofilament protein oligomers in growing axons. J. Cell Sci. 112, 3799–3814 (1999). 33. Cao, L.-G. & Wang, Y.-L. Mechanism of the formation of contractile ring in dividing cultured animal cells. I. Recruitment of preexisting actin filaments into the cleavage furrow. J. Cell Biol. 110, 1089–1095 (1990). 34. Keating, T. J., Peloquin, J. G., Rodionov, V. I., Momcilovic, D. & Borisy, G. G. Microtubule release from the centrosome. Proc. Natl Acad. Sci. USA 94, 5078–5083 (1997). 35. Dent, E. W., Callaway, J. L., Szebenyi, G., Baas, P. W. & Kalil, K. Reorganization and movement of microtubules in axonal growth cones and developing interstitial branches. J. Neurosci. 19, 8894–8908 (1999). 36. Chang, S., Svitkina, T. M., Borisy, G. G. & Popov, S. V. Speckle microscopic evaluation of microtubule transport in growing nerve processes. Nature Cell Biol. 1, 399–403 (1999). 37. Shah, J. V., Flanagan, L. A., Janmey, P. A. & Leterrier, J.-F. Bidirectional translocation of neurofilaments along microtubules mediated in part by dynein/dynactin. Mol. Biol. Cell (In the press). 38. Prahlad, V., Yoon, M., Moir, R. D., Vale, R. D. & Goldman, R. D. Rapid movements of vimentin on microtubule tracks: Kinesin-dependent assembly of intermediate filament networks. J. Cell Biol. 143, 159–170 (1998). 39. Glass, J. D. & Griffin, J. W. Retrograde transport of radiolabeled cytoskeletal proteins in transected nerves. J. Neurosci. 14, 3915–3921 (1994). 40. Koehnle, T. J. & Brown, A. Slow axonal transport of neurofilament protein in cultured neurons. J. Cell Biol. 144, 447–458 (1999). 41. Susalka, S. J., Hancock, W. O. & Pfister, K. K. Distinct cytoplasmic dynein complexes are transported by different mechanisms in axons. Biochim. Biophys. Acta 1496, 76–88 (2000). 42. Yabe, J. T., Jung, C. W., Chan, W. K. H. & Shea, T. B. Phospho-dependent association of neurofilament proteins with kinesin in situ. Cell Motil. Cytoskeleton 45, 249–262 (2000). 43. Elluru, R. G., Bloom, G. S. & Brady, S. T. Fast axonal transport of kinesin in the rat visual system: Functionality of kinesin heavy chain isoforms. Mol. Biol. Cell 6, 21–40 (1995). 44. Dahlstrom, A. B., Czernik, A. J. & Li, J. Y. Organelles in fast axonal transport — what molecules do they carry in anterograde vs retrograde directions, as observed in mammalian systems. Mol. Neurobiol. 6, 157–177 (1992).

Acknowledgements The author thanks Ray Lasek and Peter Baas for stimulating discussions.

www.nature.com/reviews/molcellbio

| NOVEMBER 2000 | VOLUME 1