Nature Reviews - Molecular Cell Biology - October 2000 vol1 January 2001 by Nature Magazine

toc_r

Page 1 of 2

CONTENTS

January 2001 Vol 2 No 1

1 | In this issue doi:10.1038/35048018

Highlights PDF

[1461K]

3 | CELL DIVISION Not all are born equal doi:10.1038/35048023

4 | WEB WATCH The great divide doi:10.1038/35048003

4 | PLANT DEVELOPMENT Stomatal waxing and waning doi:10.1038/35048026

4 | PROTEIN DEGRADATION Parkin finds a partner and a victim doi:10.1038/35048029

5 | MEIOSIS Synapsis spoilt doi:10.1038/35048031

5 | PROTEIN–PROTEIN INTERACTIONS Phosphothreonine lego doi:10.1038/35048033

6 | CELL CYCLE Two's company doi:10.1038/35048036

6 | APOPTOSIS Bax to Bak doi:10.1038/35048038

7 | CYTOSKELETON Molecular chauffeurs doi:10.1038/35048041

7 | IN BRIEF CELL CYCLE | DEVELOPMENT | PRIONS | CELL SIGNALLING doi:10.1038/35048044

8 | IN BRIEF CELL DIVISION | CELL SIGNALLING | PROTEININTERACTION MAPPING | NUCLEAR TRANSPORT

11 | CELL DIVISION ASYMMETRIC CELL DIVISION DURING ANIMAL DEVELOPMENT Juergen A. Knoblich doi:10.1038/35048085 [2025K]

21 | CELL DIVISION MITOTIC KINASES AS REGULATORS OF CELL DIVISION AND ITS CHECKPOINTS Erich A. Nigg doi:10.1038/35048096 [1093K]

33 | CELL DIVISION PLANT CELL DIVISION: BUILDING WALLS IN THE RIGHT PLACES Laurie G. Smith doi:10.1038/35048050 [3291K]

40 | CELL DIVISION MICROTUBULE-ASSOCIATED PROTEINS IN PLANTS — WHY WE NEED A MAP Clive Lloyd & Patrick Hussey doi:10.1038/35048005 [930K]

48 | CELL DIVISION NOT BEING THE WRONG SIZE Richard H. Gomer doi:10.1038/35048058 [4745K]

55 | PROTEIN INTERACTION MAPS FOR MODEL ORGANISMS Albertha J. M. Walhout & Marc Vidal doi:10.1038/35048107 [427K]

63 | OPINION BREAKING THE MITOCHONDRIAL BARRIER Jean-Claude Martinou & Douglas R. Green doi:10.1038/35048069 [675K]

67 | OPINION

http://www.nature.com.ezproxy.library.wisc.edu/nrm/journal/v2/n1/toc_r.html

12/3/2006

toc_r

Page 2 of 2

doi:10.1038/35048046

8 | CELL DIVISION Dodging death at division? doi:10.1038/35048000

9 | TECHNIQUE Fluorescent timer doi:10.1038/35048015

9 | MEMBRANE DYNAMICS Variation on a theme doi:10.1038/35048012

9 | WEB WATCH Smart by name . . . doi:10.1038/35048048

THE MITOCHONDRION IN APOPTOSIS: HOW PANDORA'S BOX OPENS Naoufal Zamzami & Guido Kroemer doi:10.1038/35048073 [207K]

72 | TIMELINE WALTHER FLEMMING: PIONEER OF MITOSIS RESEARCH Neidhard Paweletz doi:10.1038/35048077 [527K]

77 | NatureView doi:10.1038/35048083

NATURE REVIEWS | MOLECULAR CELL BIOLOGY

http://www.nature.com.ezproxy.library.wisc.edu/nrm/journal/v2/n1/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

Not all are born equal Contrary to the Marxist belief, not being equal can actually be a good thing. Asymmetric cell divisions generate diversity, and an avalanche of papers now shed some light on this process. During asymmetric cell division, the precursor cell is polarized to segregate cell fate determinants predominantly into just one daughter cell, and the mitotic spindle is orientated along the appropriate axis before cytokinesis to ensure this. In the Drosophila central nervous system, neuroblasts divide asymmetrically along the apical–basal axis. Petronczki and Knoblich, and Wodarz et al. show that this process is similar to the first cell division in Caenorhabditis elegans. In neuroblasts the apical PDZ domain protein Bazooka is in a complex with DmPAR-6 (shown in red in the picture) and an atypical protein kinase C, and this complex controls apical–basal polarity, necessary for the correct basal localization of cell fate determinants (such as the Notch antagonist Numb and the transcription factor Prospero) and asymmetric cell division. Ohshiro et al. and Peng et al. show that the tumour-suppressor genes lgl (lethal giant larvae) and dlg (discs large) are also essential to position cell fate determinants at the basal cortex of neuroblasts, independently of the Bazooka complex. As for the position of the spindle, this is known to be controlled by the Bazooka complex and the apical protein Inscuteable with its partner Pins. In the Drosophila peripheral ner-

vous system, a series of asymmetric divisions generates the external sensory organ from a single precursor cell. Roegiers et al. and Bellaïche et al. find that, as the precursor pI cell divides, Numb and its partner PON localize to the anterior pole of the cell before the spindle rotates to position itself along the anterior–posterior axis. Numb localization and correct spindle orientation depend on frizzled and flamingo, two genes involved in planar polarity. Roegiers et al. also show that division of the pIIb cell is similar to neuroblast division. They propose that this difference within the same lineage could arise because pIIb cells express Inscuteable, whereas pI cells do not. The Bazooka complex polarity cue (acting through Inscuteable) could be dominant over the Frizzled

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

cue, leading to an apical–basal polarization. But in the absence of Inscuteable the spindle orientates along an anterior–posterior axis, which could be specified by Frizzled. Raluca Gagescu References and links ORIGINAL RESEARCH PAPERS Petronczki, M. &

Knoblich, J. A. DmPAR-6 directs epithelial polarity and asymmetric cell division of neuroblasts in Drosophila. Nature Cell Biol. 3, 43–49 (2001) | Bellaïche, Y. et al. Frizzled regulates the localization of cell-fate determinants and mitotic spindle rotation during asymmetric cell division. Nature Cell Biol. 3, 50–57 (2001) | Roegiers, F. et al. Two types of asymmetric divisions in the Drosophila sensory organ precursor cell lineage. Nature Cell Biol. 3, 58–67 (2001) | Ohshiro, T. et al. Role of cortical tumour-suppressor proteins in asymmetric division of Drosophila neuroblast. Nature 408, 593–596 (2000) | Peng, C.-Y. et al. The tumour-suppressor genes lgl and dlg regulate basal protein targeting in Drosophila neuroblasts. Nature 408, 596–600 (2000) | Wodarz, A. et al. Drosophila atypical protein kinase C associates with Bazooka and controls polarity of epithelia and neuroblasts. J. Cell Biol. 150, 1361–1374 (2000)

VOLUME 2 | JANUARY 2001 | 3

HIGHLIGHTS

WEB WATCH The great divide Some subjects always get left until the end at conferences: apoptosis, dephosphorylation and cytokinesis are classic examples. This web page, hosted by Julie Canman at the University of North Carolina, was formed by a group who were determined to fight back. In their words: “We are tired of cytokinesis being an afterthought in journal clubs, and of hearing groans every time the word cytokinesis is mentioned. To ensure proper representation of our cause, we have founded the Cytokinetic Mafia”. First and foremost, the cytokinetic mafia is a journal club. Papers are chosen by mafia members on a monthly basis; you can view their choices, but I was disappointed that members’ discussions of the chosen papers aren’t made available. Even so, this is a good place to visit if you want to know what’s catching the cytokinesis enthusiast’s eye this month. Not surprisingly, aurora kinases are de rigueur. The mafia also highlights a relevant book each week, but again there’s no evaluation of them. If it’s movies you’re after, then look no further. The movies page features dividing cells from several species, from bacteria to kangaroo rats. The asymmetric cell divisions in Caenorhabditis elegans are definitely worth a look, although you’ll need to be patient about download times. Last, but not least, the cytokinetic mafia is a community: you’ll find lists of members, with contact details, and a comprehensive index of labs with an interest in cytokinesis. The mafia encourages new members and welcomes additions to the movie page from members and non-members alike. It’s supposed to be a secret club though, so I’m half expecting to find a horse’s head in my bed.

P L A N T D E V E LO P M E N T

Stomatal waxing and waning For plants, unlike animals, development is a continuing process. Whereas an adult animal’s gross anatomy is pretty much fixed and determined by its genetic make-up, a plant’s body is in permanent flux from the production of new organs, such as leaves and flowers. Plants use this plasticity to adapt to the different and changing environments in which they find themselves. In the 7 December issue of Nature, Alistair Hetherington and colleagues report that they have begun to dissect the pathway by which one such adaptation is controlled. Perhaps the most obvious way in which plants react to environmental conditions is their phytochrome-controlled response to light. However, much subtler effects also exist. One of these is a decrease in the number of stomatal pores on leaves grown under CO2-rich conditions. Stomata are the main route for gas exchange in plants, but there are trade-offs to be made. A plant’s demands to take up CO2 as a raw material for photosynthesis must be balanced against its need to preserve water loss through transpiration. When CO2 is plentiful, the pressure to

ized waxes found in the plant’s extracellular matrix. The authors showed that this is the source of HIC’s developmental effects by identifying mutations of other genes involved in wax production, which also produce altered stomatal densities. By changing the make-up of the leaf’s extracellular matrix, HIC probably controls its physical properties. The regular spacing of stomata across the leaf ’s surface has led to the suggestion that stomata inhibit the production of other stomata in their immediate neighbourhood — perhaps by producing a diffusible suppressor of stomatal development. HIC’s effect on the extracellular

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

Parkin finds a partner and a victim

Cath Brooksbank

conserve water is accommodated by a reduction in stomatal density. This inverse correlation has existed for over 400 million years — indeed, by counting fossil plant stomata ambient conditions can be estimated, providing clues to the causes of global extinctions. But what controls this ancient developmental response? While investigating the results of a screen to detect tissue-specific expression in Arabidopsis, Hetherington and colleagues isolated a gene, which they named HIC, only expressed in stomatal guard cells. hic mutants produced leaves with more (rather than fewer) stomata in response to elevated CO2. But functioning of the stomata was entirely normal — for example, the figure shows thermal imaging of hic (right) and abil-1 (bottom left) plants. The abil-1 plants are colder (more blue) due to increased heat loss through stomata that do not close efficiently. Sequencing of the HIC gene showed it to be homologous to KCS1, which encodes an Arabidopsis 3ketoacyl CoA synthase. KCS enzymes form part of microsomal fatty acid elongase complexes, which are involved in synthesizing the special-

Oliver Sacks’ Awakenings familiarized many with the devastating symptoms of Parkinson’s disease, and the fact that — dramatic though its initial effects may be — L-3,4dihydroxyphenylalanine (L-DOPA) provides only temporary relief. We need a drug that targets the cause, not the consequences, of the disease. One rare cause, leading to autosomal recessive juvenile parkinsonism (ARJP), is mutation of parkin, an E3 enzyme that catalyses the transfer of ubiquitin from an E2 enzyme to proteins

destined for proteasomal degradation. But what is the identity of the protein(s) whose death warrant is signed by parkin? Yi Zhang and colleagues, reporting in Proceedings of the National Academy of Sciences, have tracked down a substrate. But first, they had to identify the E2 that parkin collaborates with. Database searches for parkin homologues uncovered two that interact with UBCH7 and UBCH8, so the authors reasoned that these E2s might interact with parkin. Co-immunoprecipitations

showed that UBCH8 binds particularly tightly to parkin using its second RING finger domain, and that mutations that cause ARJP reduce or eliminate the interaction. In vitro, UBCH8 or UBCH7 allowed parkin to ubiquitylate itself — a property common among E3s. But a yeast two-hybrid screen also found another target for parkin — cell division control related protein 1 (CDCrel-1) — a protein implicated in blocking synaptic release. Wild-type parkin speeded up turnover of CDCrel-1 whereas

www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

HIGHLIGHTS

P R OT E I N – P R OT E I N I N T E R A C T I O N S

MEIOSIS

Synapsis spoilt

matrix could increase its permeability, thereby enlarging the suppressor’s area of influence. Whatever the role of HIC, showing its involvement in CO2-induced changes in stomatal density is the first step in revealing how plants adjust their development to cope with global environmental change. Christopher Surridge Senior Editor, Nature References and links ORIGINAL RESEARCH PAPER Gray, J. E. et al.

The HIC signalling pathway links CO2 perception to stomatal development. Nature 408, 713–716 (2000) FURTHER READING Smith, H. Phytochromes and light signal perception by plants — an emerging synthesis. Nature 407, 585–591 (2000) WEB SITE The Arabidopsis Information Resource

parkin mutants didn’t. Does parkin have other substrates, and is one of CDCrel1’s normal functions to block dopamine release? If it is, blockage of CDCrel-1’s breakdown in ARJP would provide a satisfactory explanation and a promising therapeutic target for this form of Parkinson’s. Cath Brooksbank References and links ORIGINAL RESEARCH PAPER Zhang, Y. et

al. Parkin functions as an E2-dependent ubiquitin protein ligase and promotes the degradation of the synaptic vesicle-associated protein, CDCrel-1. Proc. Natl Acad. Sci. USA 97, 13354–13359 (2000) FURTHER READING Shimura, H. et al. Familial Parkinson disease gene product, parkin, is a ubiquitin-protein ligase. Nature Genet. 25, 302–305 (2000)

Synapsis — the process by which paired homologous chromosomes are brought into close alignment — is a general feature of meiosis. But whereas S. cerevisiae requires a protein called Spo11p to initiate synapsis, C. elegans and Drosophila do not. Because Spo11p and its homologues generate double-stranded DNA breaks (DSBs) — which initiate meiotic recombination — one model for synapsis is that, in worms and flies at least, it occurs in specialized ‘pairing centres’, without the need for meiotic recombination. Could a dependence on pairing centres have emerged with increased genome complexity? And, if so, might mammals rely on such centres? Two papers in Molecular Cell report that, surprisingly, they don’t. Both groups have knocked out the mouse Spo11 gene, and they find that both male and female mice have severe gonadal abnormalities from defective meiosis. Moreover, two ‘markers’ for meiotic recombination — Rad51 and Dmc1, which load onto single-stranded DNA ends at the sites of DSBs — do not form characteristic foci in the Spo11–/– mice. In other words, synapsis in mice seems to depend on meiotic recombination. Romanienko and CameriniOtero also propose a second function for Spo11 in mice. They localized the Spo11 protein in meiotic chromosome spreads, and saw that it decorated the lengths of synapsed homologues during pachytene. This was unexpected — Spo11 is thought to catalyse the formation of DSBs at an earlier stage, leptotene. But, because pachytene is the stage at which homologous pairs become fully synapsed, the authors suggest that Spo11 could also have a structural function in stabilizing synapsis.

Phosphothreonine lego Protein phosphorylation can work in two ways: like lego, allowing other proteins to snap into place; or like a switch, triggering a conformational change. Not long ago, dogma decreed that phosphotyrosine was a lego builder whereas phosphoserine and phosphothreonine were conformational switchers. A paper by Daniel Durocher and colleagues in Molecular Cell provides definitive evidence that serine/threonine phosphorylation can build lego too. In 1999 the same team discovered that the forkhead-associated (FHA) domain — first identified in forkhead transcription factors but now known to be present in many other proteins — is a phosphopeptide recognition motif. They have now used a peptide library to find the optimal binding sequences for several FHA domains, including the two (FHA1 and FHA2) in yeast Rad53p, which binds phosphorylated Rad9p in response to DNA damage. They found that, just like the phosphotyrosine-binding SH2 domain, FHA domains recognize phosphothreonine in the context of flanking amino acids, so individual FHA domains bind specifically to different peptides. Peptide binding depends absolutely on phosphothreonine (its replacement with phosphoserine blocks binding), but the surrounding residues generate specificity. For example, Rad53p’s FHA1 domain strongly prefers aspartate three residues downstream of the phosphothreonine (the +3 position), whereas the FHA2 domain prefers isoleucine. Other residues between the –3 and +3 positions also influence binding affinity but, although there are some exceptions, in general the strongest selection depends on an FHA domain’s preference at the +3 position. The crystal structure of Rad53p’s FHA1 domain bound to a longer peptide containing its optimal binding sequence revealed why the +3 residue is so important (see picture). Phosphothreonine and Arg+3 make the largest number of direct contacts with the FHA domain. Another intriguing observation is that the FHA domain’s fold is identical to that of the MH2 domain in SMAD transcription factors — proof that you can teach an old fold new tricks. But perhaps the most exciting prospect is that this work allows us to trawl the sequence databases for specific FHAdomain-binding proteins. This should help us to build better models of DNA-damage response pathways and other processes that involve proteins containing FHA domains. Cath Brooksbank

References and links

ORIGINAL RESEARCH PAPER Durocher, D. et al. The molecular basis of FHA domain:

phosphopeptide binding specificity and implications for phospho-dependent signalling mechanisms. Mol. Cell 6, 1169–1182 (2000)

FHA

MH2 +3 Phosphothreonine

Alison Mitchell References and links ORIGINAL RESEARCH PAPERS Romanienko, P. J. & Camerini-Otero, R. D. The mouse Spo11 gene is required for meiotic chromosome synapsis. Mol. Cell 6, 975–987 (2000) | Baudat, F. et al. Chromosome synapsis defects and sexually dimorphic meiotic progression in mice lacking Spo11. Mol. Cell 6, 989–998 (2000)

Courtesy of Steve Smerdon and Mike Yaffe, Massachusetts Institute of Technology, Cambridge, Massachusetts, USA.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 5

HIGHLIGHTS

Two’s company It’s not always the case that you can’t have too much of a good thing. For example, although two centrosomes are essential for assembly of the bipolar spindle during mitosis, more than that can lead to genome instability. It’s important, then, that centrosomes are duplicated only once per cell cycle. But how is this regulated? In January’s Nature Cell Biology, Steven Reed

A P O P TO S I S

Bax to Bak There’s a fine line between life and death — a tightrope walked by the Bcl-2 family of proand anti-apoptotic proteins. Tip the balance too far one way and, as discussed in two new papers, the cell slides helplessly to its death. Members of the Bcl-2 family fall into three subfamilies. On one side of the death equation are the anti-apoptotic proteins Bcl-2 and Bcl-xL; on the other side are the pro-apoptotic members, including the Bax subfamily (Bax and Bak) and the ‘BH3-only’ proteins (such as Bid and Bad). But how do the interactions between these various proteins control apoptosis? Reporting in Cell, Nico Tjandra and colleagues discuss the regulation of Bax. They have used NMR to solve its solution structure — a structure, say the authors, that is strikingly similar to that of Bcl-xL. Both contain nine α-helices, with the first eight (α1–α8) occupying almost identical positions despite low sequence similarity. But whereas Bcl-xL (right-hand figure) contains a hydrophobic pocket that can accommodate another protein (here, the Bak BH3 peptide; yellow), in Bax this pocket is occupied by its own α9 helix (green). How, then, does Bax interact with other members of the Bcl-2 family? The authors believe that the answer lies in a conformational change. Early during apoptosis, Bax translocates from the cytosol to the mitochondria. Here it inserts into the outer mitochondrial membrane (OMM), where it is involved in the release of cytochrome c and apoptosis. The authors propose that a conformational change, which allows Bax to insert into the OMM, also disengages the α9 helix from the hydrophobic pocket. This would expose the pocket, allowing it to bind other proteins.

and co-workers propose a model for duplication of the spindle pole body (SPB; the yeast equivalent of the centrosome). Based on the premise that DNA replication — which also must occur only once per cell cycle — is coordinated with cell-cycle progression, the authors asked whether cyclin/CDK activities might activate duplication of the SPB and inhibit reduplication until completion of the cycle. According to Reed and colleagues’ model, the three G1 cyclins (Clns 1, 2 and 3) in budding yeast are involved in controlling SPB duplication. Subsequent maturation, an essential step that must be completed before

One candidate that might slip into this pocket is Bid — the subject of a report in Genes and Development by Stanley Korsmeyer and colleagues. They have studied tBid, a truncated, physiologically active form of Bid, which is involved in the release of cytochrome c from mitochondria. It could do this either by forming a pore through which cytochrome c can escape across the OMM, or by activating another mitochrondrial protein with the same net effect. Korsmeyer and co-workers favour the second possibility. They show that, for apoptosis to occur, tBid’s BH3 domain (which is required for dimerization) must be present on the cytoplasmic face of the mitochondria. This indicates that tBid acts by binding other proteins, and the authors reveal at least one of its partners to be Bak. Not only do Bak and tBid interact physically, but this association is required for the release of cytochrome c. Finally, on binding tBid, Bak undergoes a conformational change and forms oligomers — indicating,

SPBs can reduplicate, is directed either by the two S-phase B cyclins, Clb5 and Clb6, or by one of four mitotic B cyclins (Clbs 1, 2, 3 and 4). Finally, the four mitotic B cyclins can block SPB reduplication until mitosis has been completed or under checkpoint-arrest conditions. It seems that a fine balance between the positive and negative effects of cell-cycle proteins does, indeed, regulate SPB duplication. Alison Mitchell References and links ORIGINAL RESEARCH PAPER Haase, S. B., Winey, M. & Reed, S. I. Multi-step control of spindle pole body duplication by cyclin-dependent kinase. Nature Cell Biol. 3, 38–42 (2001)

speculate the authors, the possible formation of a cytochrome c -permeant pore. The mitochondrial preparation used by Korsmeyer and colleagues did not contain abundant Bax, and the authors are now repeating these experiments in cells that contain both Bax and Bak. Indeed, the idea that Bid acts as a death ligand fits well with the proposed opening of Bax’s hydrophobic pocket to accommodate other members of the Bcl-2 family. However, the functions of Bcl-2 family members in apoptosis remain controversial, and it’s likely to be some time before all of their molecular balancing acts are revealed. Alison Mitchell References and links ORIGINAL RESEARCH PAPERS Wei, M. C. et al. tBID, a

membrane-targeted death ligand, oligomerizes BAK to release cytochrome c. Genes Dev. 14, 2060–2071 (2000) | Suzuki, M., Youle, R. J. & Tjandra, N. Structure of Bax: coregulation of dimer formation and intracellular localization. Cell 103, 645–654 (2000) FURTHER READING Zha, J. et al. Posttranslational Nmyristoylation of BID as a molecular switch for targeting mitochondria and apoptosis. Science 290, 1761–1765 (2000)

Nico Tjandra, NIH

CE LL CYCLE

www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

HIGHLIGHTS

IN BRIEF CE LL CYCLE

Targeted disruption of the three Rb-related genes leads to loss of G1 control and immortalization. Sage, J. et al. Genes Dev. 14, 3037–3050 (2000) The Kobal Collection

C Y TO S K E L E TO N

Molecular chauffeurs Nerves rely on microtubule-dependent transport along the axon to ensure that vesicles containing synaptic proteins are delivered to the synapse. Their vehicle — the microtubule-binding motor kinesin — is well characterized. However, the identity of the molecule that mediates the interaction between kinesin and its cargo in axons has remained elusive. Reporting in Cell and Neuron, Lawrence Goldstein and colleagues now show that there might be more than one chauffeur for the job. To screen for proteins that might mediate this interaction, Bowman et al. used the genetically amenable Drosophila melanogaster. They mutagenized flies and screened larvae for behavioural phenotypes previously characterized in mutants lacking subunits of kinesin-1. Through this, they identified a candidate — Sunday driver (SYD). Their suspicions were confirmed when they found large accumulations of axonal membrane-bound cargos in the nerves of the larvae, indicating problems in axonal transport. The syd gene product contains two predicted protein–protein interaction domains. BLAST analysis showed syd homologues in the worm, mouse and humans. So might SYD be the longsought factor on vesicles that interacts with kinesin-1? To test this, the authors asked if syd is expressed in the right places. Expression of syd fused to green fluorescent protein in epithelial cells shows that it localizes to the Golgi, and to vesicles of the secretory pathway together with kinesin-1. Furthermore, expression analysis on western blots of mouse brain and extracts from nervous tissue showed that SYD is proba-

bly present in axons. Next, Bowman et al. tested whether SYD can bind to kinesin-1. They confirmed this using the yeast two-hybrid system and co-immunoprecipitation. To show that this interaction is direct, they checked that the recombinant purified proteins bound together in a GST-pulldown assay. And they narrowed down the region responsible for this interaction to the tetratricopeptide (TPR) domain of the kinesin-1 light chain (KLC). Does SYD mediate binding of all cargos with kinesin-1? Kamal et al. suggest not. They show that at least one cargo protein — the β-amyloid precursor protein (APP) — interacts with KLC directly. Intriguingly, APP binds to the same region as SYD. So one possibility is that different cargos compete for interaction with KLC, ensuring their safe delivery to the synapse. But is it important for SYD to reach the synapse? Two other labs have also recently shown that the mouse syd gene product acts as a scaffold for the mitogen-activated protein kinase cascade, indicating that once SYD has reached the synapse, it might have other jobs beyond chauffeuring. Alison Schuldt References and links ORIGINAL RESEARCH PAPERS Bowman, A. B. et al. Kinesin-dependent axonal transport is mediated by the Sunday driver (SYD) protein. Cell 103, 583–594 (2000) | Kamal, A. et al. Axonal transport of amyloid precursor protein is mediated by direct binding to the kinesin light chain subunit of kinesin-1. Neuron 28, 449–459 (2000) FURTHER READING Ito, M. et al. JSAP1, a novel Jun N-terminal protein kinase (JNK)-binding protein that functions as a scaffold factor in the JNK signalling pathway. Mol. Cell. Biol. 19, 7539–7548 (1999) | Kelkar, N. et al. Interaction of a mitogen-activated protein kinase signalling module with the neuronal protein JIP3. Mol. Cell. Biol. 20, 1030–1043 (2000)

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

Retinoblastoma and its close relatives p130 and p107 have overlapping functions in controlling the cell cycle. To investigate the functions of each protein in vivo, the authors generated single, double and triple-knockout mouse embryonic fibroblasts. The triple knockouts have a shortened cell cycle, show characteristics of transformed cells, and do not undergo G1 arrest in response to several factors. These results, say the authors,“further the link between loss of cell cycle control and tumorigenesis”. D E V E LO P M E N T

DNA methylation in Drosophila melanogaster. Lyko, F., Ramsahoyer, B. H. & Jaenisch, R. Nature 408, 538–539 (2000)

Inactive regions of many eukaryotic genomes contain methylated cytosine residues, but the fruitfly was thought to be an exception until recently. Jaenisch and colleagues now reveal that this methylation is restricted to the early stages of embryonic development. The authors show that methylation decreases during later stages of development, and they predict that this could be due to reduced expression of the enzyme responsible, DNA methyltransferase. PRIONS

Binding of disease-associated prion protein to plasminogen. Fischer, M. B. et al. Nature 408, 479–483 (2000)

PrPSc is associated with transmissible spongiform encephalopathies, but deposition of this protein alone is not sufficient to damage the brain, indicating that it interacts with other cellular factors to cause disease. Fischer et al. now show that blood plasminogen binds to PrPSc but not to normal PrPC. Plasminogen is therefore the first molecule that can reliably discriminate between normal and pathological prion proteins, making it a useful tool for diagnostic purposes. CELL SIGNALLING

InsP4 facilitates store-operated calcium influx by inhibition of InsP3 5-phosphatase. Hermosura, M. C. et al. Nature 408, 735–740 (2000)

Is inositol-1,3,4,5-tetrakisphosphate (Ins(1,3,4,5)P4), like inositol-1,4,5-trisphosphate (Ins(1,4,5)P3), a calcium-releasing second messenger? This study reveals that low Ins(1,3,4,5)P4 concentrations collaborate with Ins(1,4,5)P3 by inhibiting the enzyme that inactivates Ins(1,4,5)P3, whereas higher concentrations have the opposite effect by blocking the Ins(1,4,5)P3 receptor. By sacrificing a proportion of the Ins(1,4,5)P3 pool to make Ins(1,3,4,5)P4, cells can turn calcium signals up or down, according to their needs.

VOLUME 2 | JANUARY 2001 | 7

HIGHLIGHTS

IN BRIEF

CELL DIVISION

Dodging death at division?

CELL DIVISION

Role of the p53-homologue p73 in E2F1-induced apoptosis. Stiewe, T. & Pützer, B. M. Nature Genet. 26, 464–469 (2000)

The transcription factor E2F1 can suppress tumorigenesis by prolonging the half-life of p53, through induction of p14ARF. Stiewe and Pützer now report a p53-independent mechanism of action for E2F: expression of the TP53 homologue TP73 is directly activated by E2F and can lead to the activation of proapoptotic genes in a p53-independent manner, providing an antitumorigenic ‘safety catch’ in the absence of functional p53. CELL SIGNALLING

Quantitative imaging of lateral ErbB1 receptor signal propagation in the plasma membrane. Verveer, P. J. et al. Science 290, 1567–1570 (2000)

Ligand-driven ErbB1 activation is believed to occur through the formation of stable receptor dimers in which receptors crossphosphorylate each other. Verveer et al. use an ingenious imaging technique based on FRET and fluorescence-lifetime imaging microscopy to measure the activation of ErbB1 in living cells. They find that receptor dimers are in fact transient. After focal stimulation of ErbB1, receptor phosphorylation rapidly propagates over the entire cell surface in a ligand-independent manner, leading to the full activation of all receptors. P R OT E I N - I N T E R A C T I O N M A P P I N G

A network of protein–protein interactions in yeast. Schwikowski, B., Uetz, P. & Fields, S. Nature Biotechnol. 18, 1257–1261 (2000)

Proteins that associate with one another are likely to have similar functions. So if the function of one protein is known, those of its partners can be predicted. But first, a map of protein interactions must be built, and Fields and colleagues have done this in the yeast Saccharomyces cerevisiae. They analysed 2,709 published interactions and built up a network containing 2,358 interactions among over 1,500 proteins. Based on the functions of interacting partners, they then assigned possible functions to 364 previously uncharacterized proteins. N U CLE A R TRA N S PO RT

Vesicular stomatitis virus matrix protein inhibits host cell gene expression by targeting the nucleoporin Nup98.

Survivin is having an identity crisis. Is it an inhibitor of apoptosis (IAP), as suggested by its baculoviral IAP repeat (BIR) domain? Or is it necessary for cell division, like its closest relations in yeast and worms? Reporting in Proceedings of the National Academy of Sciences, Daniel O’Connor and colleagues present evidence that these two functions need not be mutually exclusive, whereas Anthony Uren and co-workers come down firmly on the side of a mitotic function for survivin. O’Connor et al. found that survivin is unique among IAPs in that it has a consensus sequence for phosphorylation by the cyclin-dependent kinase CDC2, and is phosphorylated by CDC2 in vitro and in vivo. They could immunoprecipitate survivin phosphorylated on threonine 34 (T34) only from cells undergoing mitosis. Survivin could also be coimmunoprecipitated with CDC2, and this interaction doesn’t depend on phosphorylation at T34 because it worked just as well when T34 was mutated to alanine (T34A). This suggests that the T34A mutant might act as a dominant-negative inhibitor. Sure enough, when T34A was overexpressed, mitotic cells died by apoptosis; but how does survivin prevent death during mitosis? The apoptotic protease caspase-9 could also be found in survivin immunoprecipitates, but the T34A mutant didn’t associate with caspase-9, suggesting that this interaction, whether direct or indirect, requires phosphorylation of T34. Uren and colleagues used antibodies to track survivin’s behaviour thoughout the cell cycle, and found that survivin’s movements closely mimicked those of a group of proteins known as chromosome passenger proteins. These hitch a ride on the centromeres to the spindle equator, where they remain until sister chromatids separate (see picture). Survivin bound to centromeres along the same axis as the inner centromere protein INCENP, which is needed to localize Aurora1 kinase to centromeres. To get a handle on what survivin might be doing at centromeres, the authors knocked it out in mice. At first glance, knockout embryos looked normal until embryonic day 4.5, but they then looked irregular, with cells that failed to separate and disorganized mitotic spindles. By day 5.5, the knockout embryos contained an average of only 13 nuclei, compared with around 200 in wild-type embryos. Is survivin a death defier, an orchestrator of division or something in between? We’re still far from an answer but, whatever the final verdict, one thing is clear: dividing cells can’t manage without it. Cath Brooksbank References and links ORIGINAL RESEARCH PAPERS O’Connor, D. S. et al. Regulation of apoptosis at cell division by p34cdc2

phosphorylation of survivin. Proc. Natl Acad. Sci. USA 97, 13103–13107 (2000) | Uren, A. G. et al. Survivin and the inner centromere protein INCENP show similar cell-cycle localization and gene knockout phenotype. Curr. Biol. 10, 1319–1328 (2000) FURTHER READING Reed, J. C. & Bischoff J. R. BIRinging chromosomes through cell division — and survivin' the experience. Cell 102, 545–548 (2000)

Prophase

Metaphase

Anaphase

von Kobbe, C. et al. Mol. Cell 6, 1243–1252 (2000)

Vesicular stomatitis virus is an RNA virus that causes acute infections in many mammalian hosts. Of particular importance during infection is the viral matrix protein (M) which has pleiotropic effects, shutting off transcription and inhibiting nuclear export of certain RNA species. In this paper, von Kobbe et al. identify the cellular target of M as the nucleoporin Nup98, indicating that M specifically blocks nuclear export of RNAs, the inhibition of transcription being a secondary effect.

Localization of survivin (green; appears pale blue when colocalized with DNA and yellow when colocalized with tubulin), tubulin (red) and DNA (blue) at different stages of the cell cycle. Courtesy of Lee Wong and K. H. Andy Choo, The Murdoch Children's Research Institute, Parkville, Victoria, Australia.

| JANUARY 2001 | VOLUME 2

www.nature.com/reviews/molcellbio

HIGHLIGHTS

WEB WATCH

TECHNIQUE

Fluorescent timer There are many ways to monitor the onset of gene expression, but so far it has been impossible to detect its downregulation. This problem might have been solved now, as Terskikh and colleagues report in Science a simple method to follow promoter activity. Last year, a red fluorescent protein (drFP583) was identified in tropical corals, further increasing the wide spectrum of possibilities to light up cells in different colours. Not satisfied with just one colour, Terskikh and colleagues introduced random mutations into drFP583, and found one mutant (called E5) that changes its fluorescence from green to red in a time-dependent manner. As E5 switches from green to red fluorescence over time, it can be used as a timer for gene expression. During the first hours of activity of a promoter, green fluorescence is predominant, whereas sustained activity of the promoter leads to a mixture of green and red fluorescence. A few hours after the promoter is turned off, only red fluorescence remains. Terskikh and colleagues verified these predictions in three experimental systems. First they monitored up- and down-regulation of E5 expression in Tet-on and Tet-off mammalian expression systems. Then they followed the activity of a heat-shock promoter during heat-induced

stress of Caenorhabditis elegans. Last, they traced the expression of a homeobox gene involved in the patterning of anterior structures in Xenopus laevis. In all cases, green fluorescence correctly indicated the onset of gene expression and was replaced with red fluorescence when expression ceased. So after decades of blue-stained embryos, we’ll now have to get used to seeing gene expression in green and red. Raluca Gagescu References and links ORIGINAL RESEARCH PAPER Terskikh, A. et al. “Fluorescent timer”:

protein that changes color with time. Science 290, 1585–1588 (2000)

M E M B R A N E DY N A M I C S

Variation on a theme When cells are starving, they can eat almost anything, even their own proteins and organelles. This desperate act — called autophagy — involves the engulfment of cytosol and organelles by a membrane that folds back onto itself, giving rise to an autophagosome. Screens in the yeast Saccharomyces cerevisiae identified many Apg mutants defective in autophagy, and their characterization is progressing rapidly. About two years ago, Ohsumi and colleagues characterized a ubiquitin-like protein modification necessary for autophagy. Two papers from the same group now describe a second, more unusual, ubiquitin-like modification required for this process. Apg8 is essential for autophagy in yeast, and starvation increases its transcription. Its weak homology with the ubiquitin-like protein

Apg12p suggests that Apg8p might also be a ubiquitin-like protein. Kirisako and colleagues showed that the carboxyl terminus of newly synthesized Apg8p is proteolytically cleaved by the cysteine protease Apg4p. The cleaved form of Apg8p is covalently modified on its terminal glycine and, as a consequence, becomes tightly membrane associated. Ichimura and colleagues determined the

nature of this modification and found that it’s not a protein, but the lipid phosphatidylethanolamine (PE). The sequence of events leading to the conjugation of Apg8p to PE is reminiscent of ubiquitylation, as Apg8p is first bound to the E1-type enzyme Apg7p, and then to the E2-type enzyme Apg3p, before being finally transferred to PE. The reaction is reversible, and Apg4p can cleave the amide bond. So what is the function of ‘Apg8ylation’? The authors speculate that Apg8p might be part of the fusion machinery involved in the formation of autophagosomes. Or could it be that, much like ubiquitin tags proteins for degradation in the proteasome, Apg8p tags PE-containing membranes for degradation through autophagy? Raluca Gagescu References and links ORIGINAL RESEARCH PAPERS Ichimura, Y. et

Autophagic bodies in yeast cell. M. Baba, Y. Ohsumi

al. A ubiquitin-like system mediates protein lipidation. Nature 408, 488–492 (2000) | Kirisako, T. et al. The reversible modification regulates the membrane-binding state of Apg8/Aut7 essential for autophagy and the cytoplasm to vacuole targeting pathway. J. Cell Biol. 151, 263–275 (2000)

Smart by name… Modular protein domains are nature’s solution for building versatile proteins from ready-made building blocks, but keeping track of how they’ve been shuffled is no easy task. Simple Modular Architecture Research Tool (SMART ), written by Peer Bork’s group at the European Molecular Biology Laboratory in Heidelberg, is a powerful tool for putting domains in context. The simplest way to familiarize yourself with SMART’s database of over 400 characterized domains is to click on an entry in the list. This categorizes the occurrence of each domain according to evolutionary distribution and cellular location, as well as including information on structure and function. You can also do sequence alignments. More specific searches can be done on two levels: sequence or architecture. You can simply paste in a sequence or type in accession numbers from all the commonly used protein databases. If you already know what modular domains your protein of interest contains, you can use the architecture tool to search for proteins that contain the same combination of domains. You can also narrow your search down to a particular organism or group of organisms. SMART can also represent groups of proteins as ‘beads on a string’ cartoons — an effective way to compare the architecture of a group of proteins. All the ‘beads’ in these cartoons are clickable, taking you through to that domain’s page. Data entry is a bit clumsy (for instance, the Boolean operators only work if you type them in upper case), but it’s worth perservering; even if you don’t have a specific query, half an hour spent playing with SMART will reveal nature’s uncanny capacity to reuse the same module in many contexts.

Cath Brooksbank

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 9

REVIEWS ASYMMETRIC CELL DIVISION DURING ANIMAL DEVELOPMENT Juergen A. Knoblich Although most cells produce two equal daughters during mitosis, some can divide asymmetrically by segregating protein determinants into one of their two daughter cells. Interesting parallels exist between such asymmetric divisions and the polarity established in epithelial cells, and heterotrimeric G proteins might connect these aspects of cell polarity. The discovery of asymmetrically segregating proteins in vertebrates indicates that the results obtained in invertebrate model organisms might also apply to mammalian stem cells. ZYGOTE

Embryo at the one-cell stage between fertilization and the first cell division. NEUROBLAST

Primary Drosophila neural precursor that gives rise to neural cells only. Divides asymmetrically in a stem-celllike mode into another neuroblast and a ganglion mother cell. SEVEN-TRANSMEMBRANE RECEPTOR

A class of receptors that contains seven membranespanning helices and usually transmits signals to the inside of a cell by activating heterotrimeric G proteins.

Research Institute of Molecular Pathology (IMP), Dr Bohr Gasse 7, A-1030 Vienna, Austria. e-mail:knoblich@nt.imp. univie.ac.at

CELL DIVISION

Asymmetric cell divisions provide a mechanism for placing specific cell types at defined positions in a developing organism1, and a general concept for the orientation of such divisions has begun to emerge. First, an axis of polarity is established in the mother cell and coordinated with the general body plan. Then, cell-fate determinants are localized asymmetrically along this axis, and during mitosis the mitotic spindle is orientated along this axis so that cytokinesis creates two daughter cells containing different concentrations of these determinants (FIG. 1). Our view of the molecular mechanisms that direct asymmetric cell division in animals is derived mainly from experiments done in the fruitfly Drosophila melanogaster and the nematode worm Caenorhabditis elegans. There are striking parallels between these two organisms, and each has two distinct mechanisms to orientate asymmetric cell divisions either cell autonomously or in response to extracellular signals. The first cell division of the C. elegans ZYGOTE and the asymmetric division of Drosophila NEUROBLASTS are both examples of the first mechanism, and homologous protein machineries are at the heart of this mechanism in both systems. Asymmetric cell divisions in the Drosophila peripheral nervous system and in certain C. elegans cells use the second mechanism and are orientated by extracellular signals transmitted through SEVEN-TRANSMEMBRANE RECEPTORS. Recent experiments indicate that HETEROTRIMERIC GUANINENUCLEOTIDE-BINDING (G) proteins participate in both mechanisms. In this review, I speculate on a role for G proteins in integrating the extrinsic and intrinsic pathways to

establish the axis of polarity, review the machineries that segregate determinants into one daughter cell and, finally, discuss the potential for transfer of these results to vertebrates and their implications for research on stem cells. The Par proteins in C. elegans

The worm is an ideal model system for studying asymmetric cell division because, during its development, all somatic cells are generated in a stereotyped lineage2. Even the first division of the C. elegans zygote is asymmetric, producing a larger anterior AB cell and a smaller posterior P1 cell, which express different proteins and give rise to different cell types3–5 (FIG. 2a). The zinc-finger proteins MEX-5 and MEX-6 localize asymmetrically to the anterior half of the cytoplasm in the zygote and segregate into the AB cell6 (FIG. 2b). In their absence, several proteins that are normally restricted to AB (GLP-1 and MEX-3) or P1 (SKN-1, PAL-1, MEX-1 and POS-1) are found in both daughter cells, so cell fate in the progeny of AB and P1 is incorrectly specified. Thus, MEX-5 and MEX-6 have a key function in establishing asymmetric protein expression in AB and P1, even though their precise mode of action is not yet clear. Asymmetric distribution of MEX-5 and MEX-6 requires a set of proteins, collectively referred to as the Par proteins (for ‘partitioning defective’)3,5,7. Like MEX5/6, Par proteins are required for differential gene expression in AB and P1 but they also affect size asymmetry between AB and P1, and the orientation of the

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 1 1

REVIEWS second cell division. The PDZ DOMAIN proteins PAR-3 and PAR-6 and the atypical protein kinase C PKC-3, colocalize at the anterior cell cortex before and during the first mitosis and segregate into the AB cell8–10. In par-3, par-6 or pkc-3 mutants, the mitotic spindle fails to move towards the posterior during the first cell division and AB and P1 are the same size7,10,11. In wild-type embryos, the second cell division in AB is perpendicular to the first division, whereas P1 divides along the anterior–posterior axis. In par-3, par-6 or pkc-3 mutants, however, both daughter cells behave like P1 and divide along the anterior–posterior axis10–12. In other words, in the absence of PAR-3, PAR-6 or PKC-3, both daughter cells behave like P1, indicating that these proteins define the anterior domain of the cell cortex. The posterior domain is defined by the RING FINGER D R HETEROTRIMERIC G PROTEIN

A protein complex of three proteins (Gα, Gβ and Gγ). Whereas Gβ and Gγ form a tight complex, Gα is part of the complex in its inactive, GDPbound, form but dissociates in its active, GTP-bound, form. Both Gα and Gβγ can transmit downstream signals after activation.

P L V

PDZ DOMAIN

Protein interaction domain that often occurs in scaffolding proteins and is named after the founding members of this protein family (Psd-95, discslarge and ZO-1). RING FINGER

Protein domain consisting of two loops held together at their base by cysteine and histidine residues that complex two zinc ions. Many ring fingers function in protein degradation by facilitating protein ubiquitination. GANGLION MOTHER CELL

(GMC) Small daughter cell of a neuroblast. Divides only once more to produce two neurons. COILED-COIL DOMAINS

A protein domain that forms a bundle of two or three α-helices. Whereas short coiledcoil domains are involved in protein interactions, long coiled-coil domains forming long rods occur in structural or motor proteins. SOP CELL

Sensory organ precursor cell that gives rise to all cells in a Drosophila sensory organ.

Figure 1 | General concept for the orientation of asymmetric cell divisions. Before cell division, an axis of polarity is established and coordinated with the general body plan. During mitosis, both spindle orientation and asymmetric localization of determinants follow this axis of polarity, so that different amounts of these determinants are inherited by the two daughter cells during cytokinesis. After mitosis, different concentrations of determinants induce different cell fates in the two daughter cells. (L, left; R, right; D, dorsal; V, ventral; A, anterior; P, posterior).

protein PAR-2 and the serine/threonine kinase PAR-1, both of which localize to the posterior cell cortex and segregate into P1 (REFS 13,14). In par-2 mutants, both daughter cells of the zygote behave like AB and divide perpendicularly to the first division12. In par-1 mutants, by contrast, both AB and P1 divide with normal orientation, but the P-granules — large ribonucleoprotein particles that normally segregate into P1 and are degraded in AB — do not segregate and then disappear from both daughters7, indicating that PAR-1 and PAR-2 might define the posterior cell cortex of the zygote. Thus, the axis of polarity in the C. elegans zygote is established by a genetic cascade of Par proteins that subdivides the cell cortex into anterior and posterior domains. Asymmetric division in Drosophila neuroblasts

The function of Par proteins in cell polarity has been evolutionarily conserved. Drosophila neuroblasts, the precursors of the central nervous system, arise from polarized epithelial cells during development. When cells are specified to become neuroblasts, they leave the epithelium and move to a more basal position (a process called delamination). Shortly after division, they divide asymmetrically along the apical–basal axis, giving rise to a small basal and a larger apical daughter cell (FIG. 3a). The basal daughter cell is called the GANGLION MOTHER CELL (GMC) and divides only once more to produce two differentiating neurons, whereas the apical daughter cell retains neuroblast characteristics and continues to divide like a stem cell. Candidate cell-fate determinants that are asymmetrically segregated include the proteins Prospero, Staufen, Miranda, Numb and PON (for Partner of Numb), which localize asymmetrically into a crescent-shaped pattern at the basal neuroblast cortex during mitosis and segregate into the GMC15–24 (FIG. 3b). Prospero is a transcription factor that enters the GMC nucleus after mitosis and is required to turn on transcription of GMCspecific genes and turn off neuroblast-specific genes in the GMC15–17,25,26. Prospero RNA, like the protein, is segregated into the GMC, and this segregation is mediated by the Staufen protein, which binds to the 3′ untranslated region of the Prospero RNA18–20. Miranda is a COILED-COIL protein that binds to both Prospero and Staufen and is required for their translocation to the cell cortex and for their asymmetric segregation during mitosis21,22. The cortical protein Numb, which contains a phosphotyrosine-binding domain, acts as a segregating determinant and has been well characterized in other tissues23,27–32. In sensory organ precursor cells (SOP CELLS), for example, Numb functions by repressing Notch signalling by means of an unknown mechanism33,34. Although Numb’s function in specifying the GMC’s fate remains to be demonstrated, it is transported into the GMC during metaphase with the aid of the coiled-coil protein PON24. Basal protein crescents in neuroblasts always overlie one of the two spindle poles. Although crescent formation and spindle orientation are independent events, they probably use the same spatial cue16. The core component of the protein machinery that establishes this www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS

a ABa ABp AB

EMS

b MEX-5/6 PAR-3, PAR-6, PKC-3 PAR-1, PAR-2 AB

Figure 2 | Asymmetric cell division in the C. elegans zygote. a | In C. elegans, the first division of the zygote is asymmetric and produces a larger anterior AB and a smaller posterior P1 cell. Whereas the AB cell divides perpendicularly to the first division, the P1 cell divides parallel to the first division and gives rise to an anterior EMS and a posterior P1 cell. (Mitoses in AB and P1 actually do not happen at the same time, and the positions of the four daughter cells differ as a result of constraints imposed by the eggshell). b | MEX-5 and MEX-6 localize to the anterior cytoplasm before mitosis, segregate into the AB cell and establish the correct gene-expression pattern. The Par proteins establish the axis of polarity: PAR-3, PAR-6 and PKC-3 co-localize at the anterior cortex and establish the anterior membrane domain, whereas PAR-1 and PAR-2 co-localize posteriorly.

cue35,36 is a protein called Inscuteable, which is not present in epithelial cells but orientates several independent events during asymmetric cell division in neuroblasts (FIG. 4). During neuroblast delamination, Inscuteable accumulates in a stalk that extends into the epithelial cell layer. This stalk is later retracted and, in fully delaminated neuroblasts, Inscuteable forms a crescent at the apical cell cortex. This crescent is maintained during most of the first mitosis until it disappears in telophase. In the absence of Inscuteable, mitotic spindles in neuroblasts fail to rotate into an apical–basal orientation37, and neuroblasts divide in random orientations35. Numb, Prospero and Miranda still localize asymmetrically, but their crescents form at random positions around the cell cortex and are no longer correlated with one of the spindle poles35. Par proteins in Drosophila

ORTHOLOGUE

The functional equivalent of a protein in another species. TIGHT JUNCTIONS

Connections between individual cells in an epithelium that form a diffusion barrier between the two surfaces of an epithelium. FRIZZLED

A protein family of seventransmembrane receptors. Its founding member, frizzled, was identified as a so-called tissue polarity mutation in Drosophila that causes defects in the orientation of bristles and hairs. Frizzled proteins function as receptors for wingless and its vertebrate homologues, the Wnt proteins.

The apical localization of Inscuteable requires the Bazooka protein38,39, which connects asymmetric cell divisions in neuroblasts to the axis of epithelial apical–basal polarity. In contrast to Inscuteable, Bazooka is expressed in epithelial cells, where it localizes to the apical cell cortex and is required for epithelial apical–basal polarity40,41 (FIG. 4a). Apical Bazooka localization is maintained when neuroblasts delaminate from the epithelium. In delaminated neuroblasts, Inscuteable and Bazooka localize together at the apical cell cortex (FIG. 4a)38,39 and Bazooka can bind to Inscuteable in vivo and in vitro. In inscuteable mutants, Bazooka crescents in neuroblasts can still be detected, although they become much weaker. By contrast, in the absence of Bazooka, Inscuteable relocalizes into the cytoplasm, leading to a misorientation of neuroblast divisions similar to that observed in inscuteable mutants (FIG. 4b). Bazooka is the Drosophila ORTHOLOGUE of C. elegans PAR-3 (REF. 40), indicating that the PAR-6/PAR-3/PKC-3 complex is functionally conserved. Indeed, PAR-6 and PKC-3 have been shown to direct cell polarity in

Drosophila epithelial cells and neuroblasts42,43. Like Bazooka, the Drosophila PAR-6 homologue DmPAR-6 (REF. 42) and the PKC-3 homologue DaPKC (Drosophila atypical protein kinase C)43 localize apically in epithelial cells and neuroblasts. The three proteins bind to each other and, in the absence of any one protein, the other two relocalize into the cytoplasm. In neuroblasts, Inscuteable is integrated into this complex by binding directly to Bazooka. All three proteins are required for Inscuteable localization, spindle orientation and basal localization of Numb, Miranda and Prospero. However, subtle differences between the inscuteable and bazooka phenotypes suggest that localization of Inscuteable is not the only function of the apical Bazooka–DmPAR-6–DaPKC complex (FIG. 4b). In inscuteable mutants, Numb and Miranda crescents are merely mislocalized35; however, they do not form at all in the absence of either Bazooka or DmPAR-6 (REFS 38,39,42). So Bazooka, DmPAR-6 or DaPKC could interact directly with the Numb and Miranda localization machinery. The fact that Miranda contains six PKC consensus phosphorylation sites44 is consistent with a functional connection — not involving Inscuteable — between DaPKC and Miranda. Homologues of PAR-3, PAR-6 and PKC-3 have also been found in vertebrates45–49. They localize to TIGHT JUNCTIONS in epithelial cells and, like their invertebrate counterparts, are thought to be involved in epithelial polarity45,50. Like the Drosophila proteins, mammalian PAR-3, PAR-6 and the atypical protein kinase C (aPKC) PKC-ζ interact as a complex, the formation of which might regulate the activity of aPKC47. PAR-6 binds to PAR-3, but also to the GTPases Cdc42 and Rac1, which regulate the actin cytoskeleton45–48. The interaction is specific for the activated, GTP-bound, form of Cdc42 and Rac1, and occurs through a ‘CRIB domain’ in PAR-6 — a motif that binds activated Cdc42. Thus, the PAR-3–PAR6–aPKC complex might recruit active Cdc42 to the apical cell cortex of epithelial cells, thereby polarizing the actin cytoskeleton. Because the inactivation of cdc42 in C. elegans causes a defect in asymmetry during the first cell division51, the function of the complex during asymmetric cell division might follow a similar mechanism. Wingless receptors and cell polarity

Whereas the Par proteins provide a cell-intrinsic polarization mechanism, signalling through the seven-transmembrane receptor FRIZZLED can establish the axis of polarity in response to an extracellular signal. In C. elegans, Par proteins are not required for all asymmetric cell divisions. The P1 cell — the posterior daughter cell of the first division — produces an anterior EMS cell and a posterior P2 cell. Asymmetric division of the EMS cell into an anterior, mesoderm-producing MS cell and a posterior, endoderm-producing E cell involves a different mechanism52. Both correct spindle orientation and the generation of asymmetric cell fates during this division seem to be directed by a signal from the neighbouring P2 cell: isolated EMS cells fail to orientate their spindle and therefore make two mesoderm-producing MS-like daughters. However, contact

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 1 3

REVIEWS

Neuroblast

Neuroblast GMC

Neurons

b Miranda

PON

Staufen

Prospero RNA

Prospero

Numb

Figure 3 | Asymmetric cell division in Drosophila neuroblasts. a | The two different daughter cells of asymmetric cell division in Drosophila neuroblasts: the larger daughter retains neuroblast characteristics and continues to divide several times, whereas the smaller daughter cell (the ganglion mother cell, GMC) divides once more into two differentiating neurons. b | In dividing neuroblasts, several proteins localize into a basal cortical crescent and segregate into the GMC. The determinant Prospero is transported by the adaptor Miranda. Asymmetric segregation of Prospero RNA is mediated by the RNA-binding protein Staufen and helps in, but is not essential for, Prospero protein segregation. The adaptor protein Partner of Numb (PON) facilitates, but is not essential for, segregation of the determinant Numb.

with a cultured P2 cell early in the cell cycle induces spindle rotation towards the contact point and polarizes the cell so that an endoderm-producing E cell is born next to the P2 cell52. Several genes are required for this signalling event. In embryos that are mutant for any of the five mom (‘more mesoderm’) genes, both EMS daughters become MS cells; pop-1 mutants, by contrast, produce two E cells and no MS cell53. The mom and pop genes are members of the wingless signalling pathway: mom-2 encodes a wingless homologue, mom-5 is a Frizzled-like wingless receptor and pop-1 is a TCF/Lef-like transcription factor that acts as a nuclear target of this signalling event. The mom-2 gene is required in the P2 cell and mom-5 is required in the EMS cell (FIG. 5a), suggesting that a wingless-dependent signal from P2 to EMS orientates the axis of polarity in this cell54. Frizzled also acts in asymmetric cell division in Drosophila SOP cells; in a wild-type organism these divide asymmetrically along the anterior–posterior axis55. They produce an anterior pIIa cell, which gives rise to the external structures of the organ, and a posterior pIIb cell, which produces the internal cells. Numb is the key for generating asymmetry during this division23,27. Numb localizes into a crescent along the posterior cell cortex, and segregates into the pIIb cell. In the absence of Numb, pIIb is completely transformed

into pIIa and, conversely, when Numb is overexpressed, two pIIb cells are generated. However, unlike in neuroblasts, Numb localization in SOP cells does not depend on inscuteable35,56 — instead, the orientation of SOP cell division is connected to a phenomenon called planar polarity. Planar polarity describes the polarization of epithelial cells within the epithelial plane, for example along the anterior–posterior axis. In Drosophila it is responsible for the directed outgrowth of wing hairs and the correct orientation of photoreceptor clusters in the eye57,58. Planar polarity involves signalling through the wingless receptor Frizzled59 (FIG. 5b), although the signal-transduction pathway differs from the classical wingless pathway60 — in fact, it might not even involve wingless61. In frizzled mutants55,62, SOP cells divide with random orientation. However, in most of these SOP cells, the Numb crescents still form over one of the two spindle poles, and the fates of the daughter cells are correctly specified55,63, suggesting that the axis of polarity is established but assumes a random orientation. Tissue-polarity mutants are therefore not functionally equivalent to inscuteable mutants — other mechanisms must coordinate spindle orientation and Numb localization in SOP cells. However, the results from flies and worms indicate that Frizzled-dependent extracellular signalling provides a second mechanism of orientating asymmetric cell divisions that has been conserved during evolution. Heterotrimeric G proteins

Both the intrinsic (Inscuteable and Par-protein-dependent) and extrinsic (Frizzled-dependent) mechanisms can orientate asymmetric cell divisions. But how do these pathways interact with the machineries for spindle orientation and segregation of protein determinants? Heterotrimeric G proteins could be downstream targets in both mechanisms. In C. elegans, embryos mutant for the Gβ subunit GPB-1 or the Gγ subunit GPC-2, or double mutant for the Gα subunits GOA-1 and GPA16, have defects in various aspects of spindle positioning and orientation64. Par proteins are correctly localized in these mutants, indicating that the axis of polarity is set up but is not interpreted. In Gβ mutants, mitotic spindles are misorientated in both the P1 cell (intrinsic pathway) and the EMS cell (extrinsic pathway)65, indicating that both the extrinsic and the intrinsic pathways require heterotrimeric G proteins. Heterotrimeric G proteins usually function downstream of seven-transmembrane receptors in transducing extracellular signals. The extrinsic pathway involves the seven-transmembrane receptor Frizzled. Although the Gprotein coupling of Frizzled receptors in Drosophila and C. elegans has not been shown, some of the Frizzleddependent responses in zebrafish embryos can be blocked by pertussis toxin, an inhibitor of heterotrimeric G proteins66. So G proteins could function downstream of Frizzled receptors in the extrinsic pathway. It is more difficult to imagine how heterotrimeric G proteins function in the intrinsic pathway. Experiments on Drosophila neuroblasts indicate a potential mechanism for receptor-independent G-protein activation in www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS a Neuroblast GMC Polarized epithelium

Delaminating neuroblast

Interphase

Metaphase Bazooka Insc

Pins Numb, Miranda etc.

Wild type

baz –

insc – or pins–

Figure 4 | Orientation of division in Drosophila neuroblasts. a | Drosophila neuroblasts delaminate from a polarized epithelium and divide asymmetrically along the apical–basal axis shortly after delamination to produce a larger apical neuroblast and a smaller basal ganglion mother cell (GMC). In epithelia, Bazooka (green) localizes apically, whereas Pins (blue) is distributed around the cortex. In neuroblasts, Inscuteable expression starts and recruits Pins to the apical cell cortex. During mitosis, several proteins including Numb and Miranda localize into a basal crescent and segregate into the GMC. b | In inscuteable mutants, epithelial cells are normal but spindle orientation and determinant localization become random. pins mutants have the same phenotype, but in bazooka mutants epithelial cells lose their polarity and determinants in neuroblasts do not localize asymmetrically at all.

these cells. In a search for proteins that bind Inscuteable, the protein Pins (for Partner of Inscuteable) was identified67,68. In neuroblasts, Pins has an Inscuteable-dependent apical localization67–69 (FIG. 4a). In epithelial cells, Pins is localized around the cell cortex but it can be recruited apically by the ectopic expression of Inscuteable, indicating that Inscuteable might be a molecular link between Pins and apically localized Bazooka. pins mutant neuroblasts have an inscuteable-like phenotype: mitotic spindles are misorientated, and Numb and Miranda crescents form at incorrect positions. Inscuteable localization is established but not maintained in these mutants. Inscuteable therefore seems to function by recruiting Pins to the apical neuroblast cell cortex, and apically localized Pins is needed for maintaining the axis of polarity and orientating asymmetric cell division in neuroblasts. Pins contains several socalled GoLoco domains, which bind to heterotrimeric G-protein α-subunits70. Indeed, a Gα protein has also been found in the Inscuteable–Pins complex67,69, suggesting that Pins might link Inscuteable to G-protein signalling. The rat homologue of Pins, AGS-3, was identified on the basis of its ability to activate heterotrimeric G proteins in the absence of any extracellular signal71. It is therefore possible that Inscuteable and Pins polarize neuroblasts by activating a heterotrimeric G-protein signalling cascade at the apical cell cortex. Heterotrimeric G proteins are involved in cell polarity in other systems. In the slime mould Dictyostelium cell polarization in response to cyclic AMP is mediated by a G-protein-coupled receptor whose activation recruits the Pleckstrin homology domain protein CRAC (cytoplasmic regulator of adenylyl cyclase) to the site of activation72 (FIG. 6a). In yeast cells, α-factor acts as a pheromone and induces cell polarization towards its source (FIG. 6b). The α-factor binds to a G-protein-cou-

pled receptor and causes the release of the Gβ/γ subunits from the intracellular domain of the receptor73. Free Gβ/γ subunits recruit Cdc24, an exchange factor for Cdc42, and thereby locally activate this small GTPase at the site of receptor activation, which in turn is thought to polarize the actin cytoskeleton74 (FIG. 6c). Thus, cell polarization by heterotrimeric G proteins is a a

b Wild type EMS

MOM-5

frizzled –

MOM-2

Posterior

Figure 5 | Frizzled receptors direct asymmetric cell division in response to extracellular signals. a | The C. elegans EMS cell requires a signal from the neighbouring P2 cell to orientate its asymmetric division. MOM-2, a winglesslike molecule, provides the signal that is transmitted through the Frizzled homologue MOM-5 to the EMS cell. b | In Drosophila, tissue polarity orientates asymmetric cell division in SOP cells. In wild-type flies, epidermal hairs (and bristles) all point to the posterior end. In frizzled mutants, planar polarity is lost and hairs assume a random orientation. Whereas sensory organ precursor (SOP) cells divide along the anterior–posterior axis and Numb localizes to the anterior cortex in wild-type cells, SOP divisions become randomly orientated in frizzled mutants. However, coordination between Numb crescents and spindle orientation is still maintained.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 1 5

REVIEWS

a cAMP CRAC–GFP

α-factor Cdc24–GFP α-factor

Cdc42–GTP

Gα–GDP

Gβγ

Gα–GTP

Cdc42–GDP

Gβγ Cdc24

Figure 6 | Function of heterotrimeric G proteins in cell polarity. a | In Dictyostelium, cells polarize in response to an extracellular source of cAMP. cAMP receptor and heterotrimeric G-protein activation can be shown by a fusion of the downstream target CRAC (cytoplasmic regulator of adenylyl cyclase) with green fluorescent protein (GFP). b | Yeast cells polarize in response to α-factor, which binds to its seven-transmembrane receptor and triggers G-protein activation. Cdc24–GFP is released from the nucleus and localizes towards the pheromone source. c | Mechanism of αfactor action: receptor activation causes release of Gβγ. Free Gβγ recruits Cdc24, a GDP/GTP exchange factor, and causes local activation of Cdc42, a regulator of the actin cytoskeleton.

common phenomenon and the asymmetric activation of a G-protein signalling cascade could be the common readout of the extrinsic and intrinsic pathways for orientating asymmetric cell divisions. Segregation of determinants

IMAGINAL DISCS

Epithelial sacs in Drosophila larvae that give rise to the tissues that later form the adult fly.

Although we are beginning to understand how asymmetric cell divisions are orientated, we still do not know the molecular machineries that localize determinants and segregate them during mitosis. Video time-lapse imaging of fusion proteins with green fluorescent protein (GFP) has permitted the live observation of determinant segregation in both worms and flies, and actindependent transport seems to be a common theme in both organisms. In C. elegans, the RNA-binding protein PIE-1 segregates into the posterior P1 cell during the first division75. It continues to segregate with the germ line during later divisions and is required for suppressing transcription in the zygote76. The PIE-1 protein concentrates in the posterior half of the cytoplasm before division in an actin-dependent process77. Residual PIE-1 in the anterior daughter cell is then degraded after mitosis, and this degradation is directed by a separate part of the protein. So both asymmetric protein localization and local degradation ensure the proper segregation of PIE-1 into one daughter cell. In Drosophila neuroblasts and SOP cells, PON–GFP moves along the cell cortex and the mitotic spindle is later aligned with the PON crescent56,63,78. Myosindependent transport along the cortical actin cytoskeleton is thought to be responsible for this basal localiza-

tion. As well as actin and myosin, the proteins Discslarge (DLG) and Lethal (2) giant-larvae (LGL) might also be part of the PON and Miranda transport machinery79,80. In neuroblasts that lack either of these proteins, Miranda, Prospero, PON and Numb fail to form a basal crescent even though mitotic spindles are correctly orientated. Both LGL and DLG are also required for apical–basal polarity in epithelial cells81,82. The basal-localization defect is not just a consequence of defects in the epithelium because Bazooka and Inscuteable are still localized apically in lgl and dlg mutant neuroblasts79,80. Thus, LGL and DLG are dispensable for setting up the axis of polarity, but they are required for asymmetric protein localization along this axis. Both LGL and DLG are components of the submembrane cytoskeleton83,84, and they might facilitate the transport of determinants along the cell cortex. Epistasis experiments indicate that DLG’s function in neuroblasts is to anchor LGL at the cell cortex79,80. LGL, by contrast, might be directly linked to myosin because it binds to the cytoplasmic myosin heavy chain85. Genetic interactions indicate that it actually inhibits myosin function in neuroblasts79,80. In vitro it can inhibit the self-assembly of myosin filaments86, thereby locking cytoplasmic myosin in a noncontractile form. Although this does not explain how determinants are transported, a model in which the local inhibition of myosin motility results in directional transport becomes feasible.Yeast and mammalian LGL homologues have been implicated in vesicle fusion87, but no such connection has been made in Drosophila. Indeed, disruption of the Golgi does not inhibit basal protein localization in Drosophila neuroblasts79, suggesting that LGL does not target vesicle fusion in neuroblasts. Both lgl and dlg were originally identified as tumoursuppressor genes. In both mutants, IMAGINAL DISC cells and brain neuroblasts in larvae overproliferate. Whereas the overproliferation of imaginal discs might be an indirect consequence of epithelial polarity defects, brain neuroblasts do not arise from epithelial cells. Given that lgldefective GMCs divide more times than normal and do not differentiate into neurons, perhaps the failure to segregate determinants during asymmetric cell divisions causes tumours in lgl mutants, at least in the brain. Asymmetric division in mammalian stem cells?

Most mechanistic insights into asymmetric cell division come from invertebrate experiments. In vertebrates, discoveries in work on stem cells have revealed enormous flexibility among the progeny of individual cells (BOX 1). Many different cell fates can be induced by changing growth factors in the culture medium, suggesting that lineage restrictions and intrinsic asymmetries have only minor functions. However, video time-lapse microscopy shows that cortical progenitor cells divide in stereotyped lineages — even in culture, where directional extrinsic signals can be largely excluded88. Although there is no clear genetic evidence for intrinsically asymmetric cell divisions in vertebrates, proteins that segregate asymmetrically during mitosis have been identified, particularly in the developing nervous system. In the developing mammalian cortex, cell divisions www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS Notch maintains the undifferentiated state in other cell types, so it is possible that apically localized Notch segregates into the apical, future ependymal, cell and maintains stem cell fate.

Box 1 | Stem cells

Conclusions Differentiation

Stem cell

Transient amplifying pool

Stem cells constitute a population of cells that continues to divide in adult organisms and produces cells for tissue regeneration. Stem cells can self-renew (they produce both differentiating daughters and daughters that maintain stem-cell identity) and are pluripotent (they can give rise to all cell types in a given organ). Most stem cells actually divide very slowly and differentiating cells are produced from a partly lineage-restricted, rapidly dividing, cell population, the so-called transient amplifying pool. Stem cells are usually identified by tissue culture procedures: under certain culture conditions, individual cells from a tissue can produce all cell types present in this tissue. If the progeny of such a cell contain at least one cell that also fulfils these pluripotency criteria, the mother cell could self-renew and is by definition a stem cell. No evidence for asymmetric cell divisions has yet been obtained in such cultures, but the observation of putative stem cells in intact tissues has revealed several examples for asymmetrically segregating proteins that might function as determinants (see main text).

EPENDYMAL CELLS

A layer of cells that line the large fluid-filled cavities in the brain, the ventricles.

are confined to the so-called ventricular zone. Neural precursors divide in the ventricular zone, and daughter cells either stay in this zone and continue to divide, or move away from it to differentiate. Experiments in the ferret cortex indicate that differentiating neurons are produced primarily by divisions perpendicular to the ventricular surface; parallel divisions give rise to two stem cells89. Asymmetric protein segregation has been implicated in these perpendicular divisions. One of the two mouse Numb homologues is expressed in the ventricular zone and localizes asymmetrically to the apical cell cortex during mitosis90 (FIG. 7a). However, in contrast to Drosophila this asymmetric apical localization occurs irrespective of spindle orientation and, during horizontal divisions, the Numb crescent is bisected by the cleavage plane and the protein segregates into both daughter cells (FIG. 7B). A similar lack of correlation with the mitotic spindle occurs during chicken neurogenesis, although here Numb localizes to the basal â&#x20AC;&#x201D; and not the apical â&#x20AC;&#x201D; cell cortex91. Given that Numb knockout mice have cortical defects92, the function of Numb as a segregating determinant is unlikely to be restricted to Drosophila. Asymmetric protein localization is also observed in stem cells of the adult mammalian brain. Whereas most adult neurons are quiescent, neurons in the nasal epithelium and certain hippocampal neurons are replaced from stem cells during adulthood. Careful lineage analysis has traced back the stem cells to EPENDYMAL 93 CELLS in the ventricular zone or astrocytes in the sub94 ventricular zone of the adult hippocampus. Although Numb localization has not been seen, the Notch receptor is localized to the apical cortex in ependymal cells93.

Since the initial discovery of asymmetrically segregating determinants in somatic cells23, our understanding of asymmetric cell division has increased significantly. We know that Par proteins can subdivide the cell cortex into different domains before mitosis and can generate an axis of polarity that is used during mitosis to orientate the spindle and segregate determinants asymmetrically into one daughter cell. Alternatively, asymmetric cell divisions can be orientated by external signals transmitted by seven-transmembrane receptors. Although G proteins might be involved in both processes, it is not clear how these pathways interact with the spindle-orientation and determinant-segregation machineries. In fact, the biggest challenge in this area is to determine the molecular mechanisms by which determinants are asymmetrically localized and segregated. It will also be crucial to determine how the determinant-segregation and spindle-orientation machineries interact with the axis of polarity. Other questions include the orientation of asymmetric cell divisions relative to previous divisions. In C. elegans, for example, the P1 cell orientates its spindle towards a cortical site that remains at the position of cytokinesis95. What is the molecular nature of this site? In Drosophila neuroblasts, Par proteins and Inscuteable are degraded during mitosis35,38,39, but how is the second mitosis orientated parallel to the first division? In Drosophila SOP cells, the second division of the pIIa cell is orientated perpendicular to the first division, even in frizzled mutants55, suggesting that an unknown mechanism orientates the axis of polarity in pIIa. Finally, it will be crucial to transfer our knowledge a Stem cell

Differentiation

Two stem cells

Figure 7 | A potential mechanism for asymmetric cell division in vertebrate neural precursor cells. a | During asymmetric division, spindle orientation and determinant localization are coordinated, giving rise to a differentiating cell and a stem cell. b | During symmetrical division, spindle orientation and determinant localization are not coordinated. Determinants segregate equally, giving rise to two equal (stem) cells.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 1 7

ÂŠ 2001 Macmillan Magazines Ltd

REVIEWS

Table 1 | Proteins involved in asymmetric cell division Protein

Structure

Localization

Function

Invertebrate homologue

Vertebrate homologue

Key references

Caenorhabditis elegans MEX-5 / MEX-6

Zn-finger

Anterior cytoplasm, then AB.

Differential gene expression in AB and P1.

–

PAR-1

Serine-Threonine kinase

Zygote: posterior cortex, then P1.

AB and P1 asymmetry P-granule localization.

DmPAR-1 96

mPar-197 MARK

7,14

PAR-2

Ring finger protein

Zygote: posterior cortex, then P1.

AB and P1 asymmetry; spindle orientation in P1.

–

13,98

PAR-3

PDZ-domain protein

Zygote: anterior cortex, then AB.

AB and P1 asymmetry; inhibition of spindle orientation in AB.

Bazooka (Dm)

ASIP50

7,8

PAR-6

PDZ-domain protein

Zygote: anterior cortex, then AB.

AB and P1 asymmetry; inhibition of spindle orientation in AB.

DmPAR-6

hPAR-6 mPar-645–48

PKC-3

Atypical protein kinase C

Zygote: anterior cortex, then AB.

AB and P1 asymmetry; inhibition of spindle orientation in AB.

DaPKC

PKC-ζ PKC-λ

PIE-1

Zn-finger

Posterior cytoplasm, Repression of zygotic gene then P1. transcription in germ line.

–

75–77

MOM-2

Wnt-like secreted protein

Not known.

Required in P2 for asymmetry; Wingless and (perhaps) spindle orientation in EMS.

Wnts

53,54

MOM-5

Frizzled-like seventransmembrane receptor

Not known.

Required in EMS for spindle orientation and asymmetry.

Frizzled

53,54

POP-1

HMG domain transcription factor

Not known.

Required in P2 for asymmetry.

Pangolin

TCF/Lef-1

53,54

Downregulation of neuroblastspecific and induction of GMCspecific genes.

Yes

Prox-1100, No asymmetric localization

15–17

Drosophila melanogaster 99

Prospero

Homeodomain transcription factor

Neuroblast: basal crescent, then in GMC.

Staufen

RNA-binding protein

Neuroblast: basal Oogenesis: posterior transport ? crescent, then GMC. of oskar RNA and anterior anchoring of bicoid RNA; neuroblasts: transport of Prospero RNA into GMC.

Yes

18–20

Miranda

Coiled-coil protein

Neuroblast: basal Transport of Prospero and crescent, then GMC. Staufen into GMC.

–

21,22

Numb

PTB-domain protein

Neuroblast: basal Inhibitor of Notch signalling in crescent, then GMC. one daughter cell (various tissues).

Yes

mNumb

23,27

PON

Coiled-coil protein

Neuroblast: basal Efficient Numb localization. crescent, then GMC.

–

Inscuteable

No homology

Neuroblast: apical crescent.

Orientation of neuroblast divisions.

–

Bazooka

PDZ-domain protein

Apical cortex in epithelial cells and neuroblasts.

Epithelial polarity, Inscuteable localization, orientation of neuroblast divisions.

PAR-3

ASIP50

38,39

DmPAR-6

PDZ-domain protein

Apical cortex in epithelial cells and neuroblasts.

Epithelial polarity, Inscuteable localization, orientation of neuroblast divisions.

PAR-6

hPAR-6 mPar-645-48

DaPKC

Atypical protein kinase C

Apical cortex in epithelial cells and neuroblasts.

Epithelial polarity, Inscuteable localization, orientation of neuroblast divisions.

PKC-3

PKC-ζ PKC-λ

Pins

TPR repeat and GoLoco domains (Gα binding)

Cortical in epithelial Orientation of neuroblast cells, apical crescent divisions, maintenance of in neuroblasts. Inscuteable localization.

Yes

LGN, AGS3

67–69

Frizzled

Seventransmembrane receptor

Not known.

MOM-5

mFrizzled

55,56,63

Tissue-polarity, anterior– posterior orientation of SOP division.

(GMC, ganglion mother cell; PTB; phosphotyrosine-binding; SOP, sensory organ precusor; TPR, tetratrico-peptide repeats.)

www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS from flies and worms to vertebrates. Many components, including most of the Par proteins, Pins, G proteins and the determinant Numb are conserved in vertebrates (TABLE 1), and the tools for their analysis in higher organisms are available. We should therefore know soon whether asymmetric cell division is common to all multicellular organisms, and whether the molecules identified so far have a similar function in the development of our own bodies.

4. 5. 6.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

Horvitz, H. R. & Herskowitz, I. Mechanisms of asymmetric cell division: two Bs or not two Bs, that is the question. Cell 68, 237–255 (1992). Sulston, J. E., Schierenberg, E., White, J. G. & Thomson, J. N. The embryonic cell lineage of the nematode Caenorhabditis elegans. Dev. Biol. 100, 64–119 (1983). Guo, S. & Kemphues, K. J. Molecular genetics of asymmetric cleavage in the early Caenorhabditis elegans embryo. Curr. Opin. Genet. Dev. 6, 408–415 (1996). Kemphues, K. PARsing embryonic polarity. Cell 101, 345–348 (2000). Rose, L. S. & Kemphues, K. J. Early patterning of the C. elegans embryo. Annu. Rev. Genet. 32, 521–545 (1998). Schubert, C. M., Lin, R., de Vries, C. J., Plasterk, R. H. & Priess, J. R. MEX-5 and MEX-6 function to establish soma/germline asymmetry in early C. elegans embryos. Mol. Cell 5, 671–682 (2000). Kemphues, K. J., Priess, J. R., Morton, D. G. & Cheng, N. S. Identification of genes required for cytoplasmic localization in early C. elegans embryos. Cell 52, 311–320 (1988). This report describes the identification of PAR genes, which turned out to be a conserved machinery for orientating asymmetric cell division. Etemad-Moghadam, B., Guo, S. & Kemphues, K. J. Asymmetrically distributed PAR-3 protein contributes to cell polarity and spindle alignment in early C. elegans embryos. Cell 83, 743–752 (1995). Hung, T. J. & Kemphues, K. J. PAR-6 is a conserved PDZ domain-containing protein that colocalizes with PAR-3 in Caenorhabditis elegans embryos. Development 126, 127–135 (1999). Tabuse, Y. et al. Atypical protein kinase C cooperates with PAR-3 to establish embryonic polarity in Caenorhabditis elegans. Development 125, 3607–3614 (1998). Watts, J. L. et al. par-6, a gene involved in the establishment of asymmetry in early C. elegans embryos, mediates the asymmetric localization of PAR-3. Development 122, 3133–3140 (1996). Cheng, N. N., Kirby, C. M. & Kemphues, K. J. Control of cleavage spindle orientation in Caenorhabditis elegans: the role of the genes par-2 and par-3. Genetics 139, 549–559 (1995). Boyd, L., Guo, S., Levitan, D., Stinchcomb, D. T. & Kemphues, K. J. PAR-2 is asymmetrically distributed and promotes association of P granules and PAR-1 with the cortex in C. elegans embryos. Development 122, 3075–3084 (1996). Guo, S. & Kemphues, K. J. par-1, a gene required for establishing polarity in C. elegans embryos, encodes a putative Ser/Thr kinase that is asymmetrically distributed. Cell 81, 611–620 (1995). This report describes the cloning of PAR-1, the first asymmetrically segregating protein in C. elegans. Hirata, J., Nakagoshi, H., Nabeshima, Y. & Matsuzaki, F. Asymmetric segregation of the homeodomain protein Prospero during Drosophila development. Nature 377, 627–630 (1995). Knoblich, J. A., Jan, L. Y. & Jan, Y. N. Asymmetric segregation of Numb and Prospero during cell division. Nature 377, 624–627 (1995). Spana, E. P. & Doe, C. Q. The Prospero transcription factor is asymmetrically localized to the cell cortex during neuroblast mitosis in Drosophila. Development 121, 3187–3195 (1995). Li, P., Yang, X., Wasser, M., Cai, Y. & Chia, W. Inscuteable and Staufen mediate asymmetric localization and segregation of Prospero RNA during Drosophila neuroblast cell divisions. Cell 90, 437–447 (1997). Broadus, J., Fuerstenberg, S. & Doe, C. Q. Staufendependent localization of Prospero mRNA contributes to neuroblast daughter-cell fate. Nature 391, 792–795 (1998). Schuldt, A. J. et al. Miranda mediates asymmetric protein

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

and RNA localization in the developing nervous system. Genes Dev. 12, 1847–1857 (1998). Shen, C. P., Jan, L. Y. & Jan, Y. N. Miranda is required for the asymmetric localization of Prospero during mitosis in Drosophila. Cell 90, 449–458 (1997). Ikeshima-Kataoka, H., Skeath, J. B., Nabeshima, Y., Doe, C. Q. & Matsuzaki, F. Miranda directs Prospero to a daughter cell during Drosophila asymmetric divisions. Nature 390, 625–629 (1997). Rhyu, M. S., Jan, L. Y. & Jan, Y. N. Asymmetric distribution of Numb protein during division of the sensory organ precursor cell confers distinct fates to daughter cells. Cell 76, 477–491 (1994). The first report describing an asymmetrically segregating determinant during somatic cell division. Lu, B., Rothenberg, M., Jan, L. Y. & Jan, Y. N. Partner of Numb colocalizes with Numb during mitosis and directs Numb asymmetric localization in Drosophila neural and muscle progenitors. Cell 95, 225–235 (1998). Vaessin, H. et al. prospero is expressed in neuronal precursors and encodes a nuclear protein that is involved in the control of axonal outgrowth in Drosophila. Cell 67, 941–953 (1991). Doe, C. Q., Chu-LaGraff, Q., Wright, D. M. & Scott, M. P. The prospero gene specifies cell fates in the Drosophila central nervous system. Cell 65, 451–464 (1991). Uemura, T., Shepherd, S., Ackerman, L., Jan, L. Y. & Jan, Y. N. numb, a gene required in determination of cell fate during sensory organ formation in Drosophila embryos. Cell 58, 349–360 (1989). Carmena, A., Murugasu-Oei, B., Menon, D., Jimenez, F. & Chia, W. Inscuteable and numb mediate asymmetric muscle progenitor cell divisions during Drosophila myogenesis. Genes Dev. 12, 304–315 (1998). Buescher, M. et al. Binary sibling neuronal cell fate decisions in the Drosophila embryonic central nervous system are nonstochastic and require inscuteablemediated asymmetry of ganglion mother cells. Genes Dev. 12, 1858–1870 (1998). Ruiz Gomez, M. & Bate, M. Segregation of myogenic lineages in Drosophila requires numb. Development 124, 4857–4866 (1997). Park, M., Yaich, L. E. & Bodmer, R. Mesodermal cell fate decisions in Drosophila are under the control of the lineage genes numb, notch, and sanpodo. Mech. Dev. 75, 117–126 (1998). Spana, E. P., Kopczynski, C., Goodman, C. S. & Doe, C. Q. Asymmetric localization of numb autonomously determines sibling neuron identity in the Drosophila CNS. Development 121, 3489–3494 (1995). Guo, M., Jan, L. Y. & Jan, Y. N. Control of daughter cell fates during asymmetric division: interaction of Numb and Notch. Neuron 17, 27–41 (1996). Frise, E., Knoblich, J. A., Younger-Shepherd, S., Jan, L. Y. & Jan, Y. N. The Drosophila Numb protein inhibits signaling of the Notch receptor during cell–cell interaction in sensory organ lineage. Proc. Natl Acad. Sci. USA 93, 11925–11932 (1996). Kraut, R., Chia, W., Jan, L. Y., Jan, Y. N. & Knoblich, J. A. Role of inscuteable in orienting asymmetric cell divisions in Drosophila. Nature 383, 50–55 (1996). Describes the first protein that couples determinant segregation to spindle orientation and seems to establish the axis of polarity. Kraut, R. & Campos-Ortega, J. A. inscuteable, a neural precursor gene of Drosophila, encodes a candidate for a cytoskeleton adaptor protein. Dev. Biol. 174, 65–81 (1996). Kaltschmidt, J. A., Davidson, C. M., Brown, N. H. & Brand, A. H. Rotation and asymmetry of the mitotic spindle direct asymmetric cell division in the developing central nervous

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

system. Nature Cell Biol. 2, 7–12 (2000). 38. Schober, M., Schaefer, M. & Knoblich, J. A. Bazooka recruits Inscuteable to orient asymmetric cell divisions in Drosophila neuroblasts. Nature 402, 548–551 (1999). 39. Wodarz, A., Ramrath, A., Kuchinke, U. & Knust, E. Bazooka provides an apical cue for Inscuteable localization in Drosophila neuroblasts. Nature 402, 544–547 (1999). References 38 and 39 reveal a potential mechanism that orients asymmetric cell division along one of the body axis, in this case the apical–basal axis. 40. Kuchinke, U., Grawe, F. & Knust, E. Control of spindle orientation in Drosophila by the Par-3-related PDZ-domain protein Bazooka. Curr. Biol. 8, 1357–1365 (1998). 41. Muller, H. A. & Wieschaus, E. armadillo, bazooka, and stardust are critical for early stages in formation of the zonula adherens and maintenance of the polarized blastoderm epithelium in Drosophila. J. Cell Biol. 134, 149–163 (1996). 42. Petronczki, M. & Knoblich, J. A. DmPAR-6 directs epithelial polarity and asymmetric cell division of neuroblasts in Drosophila. Nature Cell Biol. 3, 43–49 (2001). 43. Wodarz, A., Ramrath, A., Grimm, A. & Knust, E. Drosophila atypical protein kinase C associates with Bazooka and controls polarity of epithelia and neuroblasts. J. Cell Biol. 150, 1361–1374 (2000). 44. Fuerstenberg, S., Peng, C. Y., Alvarez-Ortiz, P., Hor, T. & Doe, C. Q. Identification of Miranda protein domains regulating asymmetric cortical localization, cargo binding, and cortical release. Mol. Cell. Neurosci. 12, 325–339 (1998). 45. Joberty, G., Petersen, C., Gao, L. & Macara, I. G. The cellpolarity protein Par6 links Par3 and atypical protein kinase C to Cdc42. Nature Cell Biol. 2, 531–539 (2000). 46. Johansson, A., Driessens, M. & Aspenstrom, P. The mammalian homologue of the Caenorhabditis elegans polarity protein PAR-6 is a binding partner for the Rho GTPases Cdc42 and Rac1. J. Cell Sci. 113, 3267–3275 (2000). 47. Lin, D. et al. A mammalian PAR-3–PAR-6 complex implicated in Cdc42/Rac1 and aPKC signalling and cell polarity. Nature Cell Biol. 2, 540–547 (2000). 48. Qiu, R. G., Abo, A. & Martin, G. S. A human homolog of the C. elegans polarity determinant Par-6 links Rac and Cdc42 to PKCζ signaling and cell transformation. Curr. Biol. 10, 697–707 (2000). 49. Brazil, D. B. & Hemmings, B. A. Cell polarity: Scaffold proteins par excellence. Curr. Biol. 10, R592–R594 (2000). 50. Izumi, Y. et al. An atypical PKC directly associates and colocalizes at the epithelial tight junction with ASIP, a mammalian homologue of Caenorhabditis elegans polarity protein PAR-3. J. Cell Biol. 143, 95–106 (1998). 51. Jantsch-Plunger, V. et al. CYK-4. A rho family GTPase activating protein (GAP) required for central spindle formation and cytokinesis. J. Cell Biol. 149, 1391–1404 (2000). 52. Goldstein, B. Cell contacts orient some cell division axes in the Caenorhabditis elegans embryo. J. Cell Biol. 129, 1071–1080 (1995). 53. Thorpe, C. J., Schlesinger, A., Carter, J. C. & Bowerman, B. Wnt signaling polarizes an early C. elegans blastomere to distinguish endoderm from mesoderm. Cell 90, 695–705 (1997). 54. Schlesinger, A., Shelton, C. A., Maloof, J. N., Meneghini, M. & Bowerman, B. Wnt pathway components orient a mitotic spindle in the early Caenorhabditis elegans embryo without requiring gene transcription in the responding cell. Genes Dev. 13, 2028–2038 (1999). 55. Gho, M. & Schweisguth, F. Frizzled signalling controls orientation of asymmetric sense organ precursor cell divisions in Drosophila. Nature 393, 178–181 (1998). This report describes a connection between

VOLUME 2 | JANUARY 2001 | 1 9

REVIEWS

56.

57. 58.

59.

60.

61.

62.

63.

64.

65.

66.

67.

68.

69.

70.

71.

72.

asymmetric cell division and planar polarity and shows the function of Frizzled in establishing this connection. Roegiers, F., Younger-Shepherd, S., Jan, L. Y. & Jan, Y. N. The sensory organ precursor cell lineage in Drosophila contains two types of asymmetric divisions. Nature Cell Biol. 3, 58–67 (2001). Bray, S. Planar polarity: out of joint? Curr. Biol. 10, R155–R158 (2000). Shulman, J. M., Perrimon, N. & Axelrod, J. D. Frizzled signaling and the developmental control of cell polarity. Trends Genet. 14, 452–458 (1998). Vinson, C. R., Conover, S. & Adler, P. N. A Drosophila tissue polarity locus encodes a protein containing seven potential transmembrane domains. Nature 338, 263–264 (1989). Strutt, D. I., Weber, U. & Mlodzik, M. The role of RhoA in tissue polarity and Frizzled signalling. Nature 387, 292–295 (1997). Rulifson, E. J., Wu, C. H. & Nusse, R. Pathway specificity by the bifunctional receptor frizzled is determined by affinity for wingless. Mol. Cell 6, 117–126 (2000). 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). Bellaiche, Y., Gho, M., Kaltschmidt, J. A., Brand, A. H. & Schweisguth, F. Frizzled regulates the localisation of cellfate determinants and mitotic spindle rotation during asymmetric cell division. Nature Cell Biol. 3, 50–57 (2001). Gotta, M. & Ahringer, J. Distinct roles for Gα and Gβγ in regulating spindle position and orientation in early C. elegans embryos. Nature Cell Biol. (in the press). Zwaal, R. R. et al. G proteins are required for spatial orientation of early cell cleavages in C. elegans embryos. Cell 86, 619–629 (1996). Sheldahl, L. C., Park, M., Malbon, C. C. & Moon, R. T. Protein kinase C is differentially stimulated by Wnt and Frizzled homologs in a G-protein-dependent manner. Curr. Biol. 9, 695–698 (1999). Schaefer, M., Shevchenko, A. & Knoblich, J. A. A protein complex containing inscuteable and the Gα-binding protein Pins orients asymmetric cell divisions in Drosophila. Curr. Biol. 10, 353–362 (2000). Yu, F., Morin, X., Cai, Y., Yang, X. & Chia, W. Analysis of partner of inscuteable, a novel player of Drosophila asymmetric divisions, reveals two distinct steps in inscuteable apical localization. Cell 100, 399–409 (2000). Parmentier, M. L. et al. Rapsynoid/Partner of Inscuteable controls asymmetric division of larval neuroblasts in Drosophila. J. Neurosci. (Online) 20, RC84 (2000). Siderovski, D. P., Diverse-Pierluissi, M. & De Vries, L. The GoLoco motif: a Gαi/o binding motif and potential guaninenucleotide exchange factor. Trends Biochem. Sci. 24, 340–341 (1999). Takesono, A. et al. Receptor-independent activators of heterotrimeric G-protein signaling pathways. J. Biol. Chem. 274, 33202–33205 (1999). Parent, C. A., Blacklock, B. J., Froehlich, W. M., Murphy, D. B. & Devreotes, P. N. G protein signaling events are

73. 74.

75.

76.

77.

78.

79.

80.

81.

82.

83.

84.

85.

86.

87.

activated at the leading edge of chemotactic cells. Cell 95, 81–91 (1998). Chant, J. Cell polarity in yeast. Annu. Rev. Cell. Dev. Biol. 15, 365–391 (1999). Nern, A. & Arkowitz, R. A. A Cdc24p–Far1p–Gβγ protein complex required for yeast orientation during mating. J. Cell Biol. 144, 1187–1202 (1999). Mello, C. C. et al. The PIE-1 protein and germline specification in C. elegans embryos. Nature 382, 710–712 (1996). Seydoux, G. et al. Repression of gene expression in the embryonic germ lineage of C. elegans. Nature 382, 713–716 (1996). Reese, K. J., Dunn, M. A., Waddle, J. A. & Seydoux, G. Asymmetric segregation of PIE-1 in C. elegans is mediated by two complementary mechanisms that act through separate PIE-1 protein domains. Mol. Cell 6, 445–455 (2000). Lu, B., Ackerman, L., Jan, L. Y. & Jan, Y. N. Modes of protein movement that lead to the asymmetric localization of partner of Numb during Drosophila neuroblast division. Mol. Cell 4, 883–891 (1999). Peng, C.-Y., Manning, L., Albertson, R. & Doe, C. Q. The tumour suppressor genes lgl and dlg regulate basal protein targeting in Drosophila neuroblasts. Nature 408, 596–600 (2000). Ohshiro, T., Yagami, T., Zhang, C. & Matsuzaki, F. Role of cortical tumour suppressor proteins in asymmetric division of Drosophila neuroblast. Nature 408, 593–596 (2000). References 79 and 80 describe the first genes that are required for basal, but not apical, protein localization in Drosophila neuroblasts, suggesting that they are components of the transport machinery. Woods, D. F., Hough, C., Peel, D., Callaini, G. & Bryant, P. J. Dlg protein is required for junction structure, cell polarity, and proliferation control in Drosophila epithelia. J. Cell Biol. 134, 1469–1482 (1996). Bilder, D., Li, M. & Perrimon, N. Cooperative regulation of cell polarity and growth by Drosophila tumor suppressors. Science 289, 113–116 (2000). Strand, D., Raska, I. & Mechler, B. M. The Drosophila lethal(2)giant larvae tumor suppressor protein is a component of the cytoskeleton. J. Cell Biol. 127, 1345–1360 (1994). Woods, D. F. & Bryant, P. J. The discs-large tumor suppressor gene of Drosophila encodes a guanylate kinase homolog localized at septate junctions. Cell 66, 451–464 (1991). Strand, D. et al. The Drosophila lethal(2)giant larvae tumor suppressor protein forms homo-oligomers and is associated with nonmuscle myosin II heavy chain. J. Cell Biol. 127, 1361–1373 (1994). De Lorenzo, C., Mechler, B. M. & Bryant, P. J. What is Drosophila telling us about cancer? Cancer Metastasis Rev. 18, 295–311 (1999). Kagami, M., Toh-e, A. & Matsui, Y. Sro7p, a Saccharomyces cerevisiae counterpart of the tumor suppressor l(2)gl protein, is related to myosins in function. Genetics 149, 1717–1727 (1998).

88. Qian, X., Goderie, S. K., Shen, Q., Stern, J. H. & Temple, S. Intrinsic programs of patterned cell lineages in isolated vertebrate CNS ventricular zone cells. Development 125, 3143–3152 (1998). 89. Chenn, A. & McConnell, S. K. Cleavage orientation and the asymmetric inheritance of Notch1 immunoreactivity in mammalian neurogenesis. Cell 82, 631–641 (1995). 90. Zhong, W., Feder, J. N., Jiang, M. M., Jan, L. Y. & Jan, Y. N. Asymmetric localization of a mammalian numb homolog during mouse cortical neurogenesis. Neuron 17, 43–53 (1996). 91. Wakamatsu, Y., Maynard, T. M., Jones, S. U. & Weston, J. A. NUMB localizes in the basal cortex of mitotic avian neuroepithelial cells and modulates neuronal differentiation by binding to NOTCH-1. Neuron 23, 71–81 (1999). 92. Zhong, W. et al. Mouse numb is an essential gene involved in cortical neurogenesis. Proc. Natl Acad. Sci. USA 97, 6844–6849 (2000). 93. Johansson, C. B. et al. Identification of a neural stem cell in the adult mammalian central nervous system. Cell 96, 25–34 (1999). 94. Doetsch, F., Caille, I., Lim, D. A., Garcia-Verdugo, J. M. & Alvarez-Buylla, A. Subventricular zone astrocytes are neural stem cells in the adult mammalian brain. Cell 97, 703–716 (1999). 95. Hyman, A. A. Centrosome movement in the early divisions of Caenorhabditis elegans: a cortical site determining centrosome position. J. Cell Biol. 109, 1185–1193 (1989). Laser ablation of microtubules reveals a mechanism for orientation of mitotic spindles: astral microtubules attach to a cortical site and pull the spindle towards the site. 96. Shulman, J. M., Benton, R. & St Johnston, D. The Drosophila homolog of C. elegans PAR-1 organizes the oocyte cytoskeleton and directs oskar mRNA localization to the posterior pole. Cell 101, 377–388 (2000). 97. Bohm, H., Brinkmann, V., Drab, M., Henske, A. & Kurzchalia, T. V. Mammalian homologues of C. elegans PAR-1 are asymmetrically localized in epithelial cells and may influence their polarity. Curr. Biol. 7, 603–606 (1997). 98. Levitan, D. J., Boyd, L., Mello, C. C., Kemphues, K. J. & Stinchcomb, D. T. par-2, a gene required for blastomere asymmetry in Caenorhabditis elegans, encodes zinc-finger and ATP-binding motifs. Proc. Natl Acad. Sci. USA 91, 6108–6112 (1994). 99. Burglin, T. R. A Caenorhabditis elegans prospero homologue defines a novel domain. Trends Biochem. Sci. 19, 70–71 (1994). 100. Oliver, G. et al. Prox 1, a prospero-related homeobox gene expressed during mouse development. Mech. Dev. 44, 3–16 (1993).

Acknowledgements I thank all members of my laboratory, in particular M. Petronczki and M. Schaefer, for their contributions to the ideas presented here; M. Glotzer for valuable discussions that had considerable influence on this review; and J. -M. Peters and M. Glotzer for comments on the manuscript. Work in my laboratory is funded by Boehringer Ingelheim.

www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS

MITOTIC KINASES AS REGULATORS OF CELL DIVISION AND ITS CHECKPOINTS Erich A. Nigg Mitosis and cytokinesis are undoubtedly the most spectacular parts of the cell cycle. Errors in the choreography of these processes can lead to aneuploidy or genetic instability, fostering cell death or disease. Here, I give an overview of the many mitotic kinases that regulate cell division and the fidelity of chromosome transmission. CELL DIVISION OMNIS CELLULA E CELLULA

All cells are derived from cells. MITOSIS

The process of nuclear division. CYTOKINESIS

The process of cytoplasmic division. SISTER CHROMATIDS

Duplicated chromosomes. CENTROSOME

The main microtubuleorganizing centre of animal cells. MITOTIC SPINDLE

A highly dynamic bipolar array of microtubules that forms during mitosis or meiosis and serves to move the duplicated chromosomes apart.

Max-Planck-Institute for Biochemistry, Department of Cell Biology, Am Klopferspitz 18a, D-82152 Martinsried, Germany. e-mail: nigg@biochem.mpg.de

Ever since Rudolf Virchow (1821–1902) proclaimed his famous “OMNIS CELLULA E CELLULA”, the challenge has been to understand how cells divide and how they faithfully transmit genetic information from one cell generation to the next. In a typical somatic cell cycle, M phase comprises MITOSIS and CYTOKINESIS. The main purpose of mitosis is to segregate SISTER CHROMATIDS into two nascent cells, such that each daughter cell inherits one complete set of chromosomes. In addition, each daughter cell must receive one CENTROSOME and the appropriate complements of cytoplasm and organelles. Mitosis is usually divided into five distinct stages: prophase, prometaphase, metaphase, anaphase and telophase (for a brief description of these phases, see BOX 1). Cytokinesis, the process of cell cleavage, occurs at the end of mitosis, and its regulation is linked intimately to mitotic progression. Although the exact temporal and spatial organization of mitosis and cytokinesis differ between animals, plants and fungi, chromosome segregation invariably requires the assembly of the MITOTIC SPINDLE, whereas cytokinesis depends on the concerted action of the spindle, the actomyosin cytoskeleton and the cell cortex. The regulation of M-phase progression relies predominantly on two post-translational mechanisms: protein phosphorylation and proteolysis. These are intimately intertwined in that the proteolytic machinery is controlled by phosphorylation, whereas several mitotic kinases are downregulated by degradation. The most prominent mitotic kinase is the cyclin-

dependent kinase 1 (Cdk1), the founding member of the Cdk family of cell-cycle regulators. Recent studies have, however, brought to light additional mitotic kinases. These include members of the Polo family, the aurora family and the NIMA (never in mitosis A) family, as well as kinases implicated in mitotic CHECKPOINTS, mitotic exit and cytokinesis (TABLE 1 and BOX 1). Here, I review our current understanding of these new mitotic kinases and their cooperation with Cdk1 in the regulation of mitosis and cytokinesis. Whereas the main text describes progression through M phase in chronological order, separate boxes provide brief summaries on selected kinase families. The review is written mainly from the perspective of cell division in humans, but much of our current thinking reflects extrapolations from pioneering work done in yeast and other genetically tractable organisms. No correct M phase without proper S phase

The error-free segregation of sister chromatids during mitosis depends on the proper execution of two events during the preceding S phase. These are the replication of chromosomal DNA, with the concomitant establishment of sister-chromatid cohesion, and the duplication of centrosomes. To keep the ploidy of an organism constant, it is essential that both chromosomes and centrosomes are duplicated only once in every cell cycle. Recent work has revealed a first glimpse of how DNA replication and centrosome duplication are coordinated in a somatic cell cycle. Both processes depend on the phos-

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 2 1

REVIEWS

Box 1 | A primer on the chronology of M-phase events Cdk1 activation Plk1 Aurora-A/B ? Nek2

Cdk1

Cdk1 Prophase Prometaphase

DNA structure checkpoints Interphase

Spindle assembly checkpoint

Spindlepositioning checkpoint

Cytokinesis

Metaphase

? Cdk1 Plk1 Bub1/BubR1 TTK ?

? MEN kinases Sid kinases Plk1 ? Aurora-B ? Telophase

CHECKPOINT

A point where the cell division cycle can be halted until conditions are suitable for the cell to proceed to the next stage. SPINDLE POLE BODY

The yeast equivalent of the centrosome. NUCLEAR LAMINA

A nuclear membraneassociated protein structure made up of lamin intermediatefilament proteins. KINESIN

Microtubule-based molecular motor, most often directed towards the plus end of microtubules.

Aurora-B ? Plk1 ? Cdk1 inactivation

The principal events typical of animal cell division can briefly be summarized as follows. During ‘prophase’, interphase chromatin condenses into well-defined chromosomes and previously duplicated centrosomes migrate apart, thereby defining the poles of the future spindle apparatus. Concomitantly, centrosomes begin nucleating highly dynamic microtubules that probe space in all directions, and the nuclear envelope breaks down. During ‘prometaphase’, microtubules are captured by kinetochores (specialized proteinaceous structures associated with centromere DNA on mitotic chromosomes). Although monopolar attachments of chromosomes are unstable, the eventual interaction of paired sister chromatids with microtubules emanating from opposite poles results in a stable, bipolar attachment. Chromosomes then congress to an equatorial plane, the metaphase plate, where they continue to oscillate throughout ‘metaphase’, suggesting that a balance of forces keeps them under tension. After all the chromosomes have undergone a proper bipolar attachment, a sudden loss in sister-chromatid cohesion triggers the onset of ‘anaphase’. Sister chromatids are then pulled towards the poles (anaphase A) and the poles themselves separate further towards the cell cortex (anaphase B). Once the chromosomes have arrived at the poles, nuclear envelopes reform around the daughter chromosomes, and chromatin decondensation begins (‘telophase’). Finally, an actomyosin-based contractile ring is formed and ‘cytokinesis’ is completed. The figure summarizes the stages of M phase. It also indicates where the major checkpoints exert quality control over mitotic progression and where mitotic kinases are thought to act. The insert illustrates a Ptk2 cell in metaphase; DNA is shown in blue (DAPI staining), microtubules in green and spindle poles (γ-tubulin) in orange. (Picture kindly provided by P. Meraldi.)

phorylation of the retinoblastoma gene product and the release of E2F transcription factors1, and both require the activity of Cdk2, in association with either cyclin A or cyclin E1–3. Another kinase, Mps1p, has been implicated in the duplication of the SPINDLE POLE BODY in Saccharomyces cerevisiae4. Whether homologues of this kinase control centrosome duplication in metazoan organisms is not known at present.

NUCLEAR ENVELOPE

Cdk1, the maestro of M phase

Double membrane that surrounds the nucleus. The outer membrane is continuous with the endoplasmic reticulum.

Of the ensemble of kinases involved in the orchestration of M-phase events, Cdk1 remains the unchallenged conductor. The regulation of Cdk1–cyclin complexes is comparatively well understood (summarized in FIG. 1).

Anaphase

Activation of mammalian Cdk1 depends on dephosphorylation of two neighbouring residues in the ATP-binding site (threonine 14 and tyrosine 15). This occurs at the G2/M transition when the activity of the dual-specificity phosphatase Cdc25C towards Cdk1 exceeds that of the opposing kinases Wee1 and Myt1. In turn, these enzymes are controlled by DNA structure checkpoints, which delay the onset of mitosis in the presence of unreplicated or damaged DNA (see below). Activated Cdk1–cyclin complexes then phosphorylate numerous substrates, for example nuclear lamins, KINESIN-related motors and other microtubule-binding proteins, condensins and Golgi matrix components, and these events are important for NUCLEAR ENVELOPE breakwww.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS

Table 1 | Mitotic kinases* Mammalian members

Founding members

Comments

The Cdk family

Cdk1 (Cdc2)

Cdc2p (Sp)/Cdc28p (Sc)

Mammalian Cdk1 functions in association with both A- and B-type cyclins

The Polo family

Plk1

Polo (Dm)

The vertebrate Polo families comprise additional members (see BOX 2)

The Aurora family

Aurora-A Aurora-B Aurora-C

Aurora (Dm)/Ipl1p (Sc)

The NIMA family

Nek2

NIMA (An)

Mitotic checkpoint

Bub1 BubR1 TTK/Esk

Bub1p (Sc)

MEN/SIN kinases

The Aurora nomenclature is explained in TABLE 2

The vertebrate Nek families comprise additional members (see BOX 3). Whether NIMA and Nek2 represent bona fide functional homologues is not known at present

Mps1p (Sc)

TTK and Esk are the names given to putative human and mouse homologues, respectively, of budding yeast Mps1p

Multiple yeast kinases (Sc and Sp) (see FIG. 5)

Several metazoan kinases (Ndr/LATS family members) are structurally related to a yeast SIN/MEN kinase (budding yeast Dbf2p/Mob1p and fission yeast Sid2p/Mob1p), but no functional homologies have yet been shown

*This Table is not meant to be exhaustive, and kinases with widely pleiotropic functions, such as MAP kinases and PKA (the cAMPdependent kinase), are only mentioned in passing. This should not detract from the fact that both MAP kinases and PKA probably have important roles in the regulation of M phase, at least in some cell types. Furthermore, several of the ‘mitotic’ kinases listed here are highly expressed in the germ line, implying probable functions also in meiosis. (Sc, Saccharomyces cerevisiae; Sp, Schizosaccharomyces pombe; An, Aspergillus nidulans; Dm, Drosophila melanogaster.)

down, centrosome separation, spindle assembly, chromosome condensation and Golgi fragmentation, respectively5–8. Furthermore, Cdk1–cyclin complexes contribute to regulate the anaphase-promoting complex/cyclosome (APC/C), the core component of the ubiquitin-dependent proteolytic machinery that controls the timely degradation of critical mitotic regulators, in particular inhibitors of anaphase onset (securins) and cyclins9. On cyclin destruction, Cdk1 is inactivated, setting the stage for mitotic exit and cytokinesis. Cdk1 inactivation also allows the reformation of pre-initiation complexes at origins of replication, thereby licensing cellular chromatin for the next round of replication10. In the following sections, major M-phase events will be discussed in more detail, with particular emphasis on the intervention of mitotic kinases at various stages. γ-TUBULIN RING COMPLEXES

Ring-like multiprotein structures implicated in microtubule nucleation. DYNEIN

Microtubule-based molecular motor that moves towards the minus end of microtubules. RNA-MEDIATED INTERFERENCE

Process by which an introduced double-stranded RNA specifically silences the expression of genes through degradation of their cognate mRNA.

Early mitotic events

Centrosome separation and activation. In most cell types, duplicated centrosomes remain closely paired and continue to function as a single microtubuleorganizing centre during G2. After G2, however, they separate and migrate apart. Concomitantly, they recruit additional γ-TUBULIN RING COMPLEXES, and this maturation event sets the stage for increased microtubule nucleation activity. As inferred from antibody microinjection studies in human cells11 and Xenopus embryos12, centrosome maturation requires the action of Polo-like kinases (Plks; BOX 2). Consistent with this view, Drosophila Polo is likely to regulate a micro-

tubule-associated protein, termed Asp (for ‘abnormal spindle’), whose function is to hold γ-tubulin ring complexes at the mitotic centrosome13. The separation of centrosomes seems to be regulated by several kinases. At an early step, the NIMA-family member Nek2 (BOX 3) is thought to phosphorylate the centrosomal protein, C-Nap1, thereby causing the dissolution of a dynamic structure that tethers duplicated centrosomes to each other14. A type 1 phosphatase interacts with both Nek2 and C-Nap1, and cell-cycle-regulated inhibition of this phosphatase may contribute to cause an abrupt increase in C-Nap1 phosphorylation at the G2/M transition15. At a later step, several kinesin-related motor proteins (KRPs) and cytoplasmic DYNEIN are required for centrosome separation. Prominent among these motors is the KRP Eg5, whose recruitment to centrosomes depends on the phosphorylation of a highly conserved carboxyterminal motif by Cdk1–cyclin B16,17. A role in centrosome separation has also been postulated for aurora-A family members (BOX 4; TABLE 2). This was originally inferred from the phenotype of aurora mutants in Drosophila18, and supported by the finding that vertebrate A-type aurora kinases also localize to centrosomes, spindle poles and spindle microtubules19,20. However, RNA-MEDIATED INTERFERENCE (RNAi) with aurora-A (AIR-1) in Caenorhabditis elegans did not prevent centrosome separation, although both spindle formation and centrosomal morphology were abnormal21. A better understanding of the role of

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 2 3

REVIEWS

Spindle assembly checkpoint DNA structure checkpoints Cdc25C

APC/C Degradation of: · securin · mitotic cyclins

Cdk1– cyclin

Cdk1 Wee1/Myt1

Spindle assembly Chromosome condensation Nuclear envelope breakdown

G1 Spindle disassembly Chromosome decondensation Nuclear envelope reformation Cytokinesis Formation of pre-replication complexes

Figure 1 | A narrative of mitotic progression from the Cdk1 perspective. Entry into mitosis results from the activation of Cdk1–cyclin complexes. In mammalian cells, this depends primarily on the dephosphorylation of Cdk1, which occurs when the activity of the phosphatase Cdc25C exceeds that of the kinases Wee1 and Myt1. By contrast, exit from mitosis depends on the inactivation of Cdk1–cyclin complexes. This occurs as a consequence of cyclin destruction, which in turn results from the activation of the APC/C ubiquitin ligase. This simple scheme provides an account of mitotic regulation from the perspective of Cdk1, but it does not consider spatial aspects of mitotic control109, and it ignores the important contribution of several additional mitotic kinases.

A-type aurora kinases in spindle assembly will require the identification of both physiological substrates and upstream regulators. The Xenopus KRP Eg5 is one candidate substrate, but the molecular consequences of this phosphorylation remain unknown22. Nuclear envelope breakdown. In organisms undergoing open mitosis, nuclear envelope breakdown (NEBD) occurs shortly after centrosome separation. During interphase, the nuclear envelope is stabilized by a karyoskeletal structure known as the NUCLEAR LAMINA, but at the onset of mitosis, this structure disassembles as a consequence of lamin hyperphosphorylation. Although lamins can be phosphorylated by many kinases in vitro, the predominant kinase triggering mitotic lamina depolymerization in vivo is almost certainly Cdk1–cyclin B5. Lamina disassembly reduces nuclear envelope stability but is not in itself sufficient to cause NEBD. The additional requirements for NEBD remain poorly understood, although phosphorylation probably has an important role.

CATASTROPHE RATE

The frequency of transitions between rapid growth and shrinkage of microtubules.

Chromosome condensation. Chromosome condensation is accompanied by extensive phosphorylation of both histones and non-histone proteins. Histone modifications, including phosphorylation, acetylation and methylation, have long been correlated with changes in chromatin condensation states. The linker histone H1 is an excellent substrate of Cdk1–cyclin B, but despite extensive study, the significance of this phosphorylation remains unknown. More recently, phosphoryla-

tion of the core histone H3 (at serine 10) has also attracted great interest. This modification is highly conserved, correlates with chromosome condensation during mitosis and meiosis, and is required for proper chromosome segregation in at least some organisms (for example, the protozoan Tetrahymena). Of the several histone H3/serine 10 kinases described, two are of particular interest from the perspective of mitosis. Genetic and biochemical data concur to indicate that aurora family members, in particular Ipl1p of S. cerevisiae and the B-type aurora AIR-2 of C. elegans, can control histone H3 phosphorylation in opposition to a type 1 phosphatase (Glc7p in S. cerevisiae)23. However, studies done in Aspergillus nidulans suggest that NIMA is another candidate histone H3 kinase24. This discrepancy illustrates the notorious difficulty in unequivocally assigning kinases to their physiological substrates, and it will be interesting to determine whether aurora and/or NIMA-related kinases (Neks) phosphorylate histone H3 in vertebrates, or whether yet other histone H3 kinases await discovery. Prominent among the trans-acting factors involved in chromosome condensation are topoisomerase II and a multiprotein complex known as condensin, and both are regulated by phosphorylation. In the case of the fivemember condensin complex, there is evidence that Cdk1–cyclin B regulates its DNA supercoiling activity in Xenopus extracts6 and its cell-cycle-regulated nuclear accumulation in Schizosaccharomyces pombe25. Spindle dynamics and chromosome movements

Spindle assembly and mitotic movements rely on three parameters: the inherent dynamic properties of microtubule polymers (particularly dynamic instability and treadmilling); a balance of microtubule stabilizing and destabilizing accessory proteins; and the action of microtubule-dependent motors of the dynein and kinesin families. Dynamic instability is particularly important at the onset of mitosis, when the CATASTROPHE RATE increases markedly. This transition can be triggered in vitro by several kinases, including Cdk1–cyclin A and mitogen-activated kinase (MAP kinase), but the substrates involved remain unknown8. Microtubule dynamics are extensively regulated by microtubule-associated proteins and microtubule-destabilizing proteins, and most of these are controlled by phosphorylation. A good example is stathmin (also known as oncoprotein 18), whose microtubule-destabilizing activity is turned off during mitosis by sequential phosphorylation involving Cdk1–cyclin B and an as yet unidentified kinase26. Studies on stathmin phosphorylation also illustrate the importance of spatial regulation in spindle assembly27. In Xenopus extracts, chromatin-induced spindle formation seems to depend on gradients of differentially phosphorylated microtubule-associated proteins, and these gradients in turn are likely to arise from the action of immobilized kinases and diffusible phosphatases (or vice versa)27. Although this attractive model awaits rigorous proof, it highlights the importance of the mutual localization of kinases, phosphatases and their www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS

Box 2 | The Polo kinase family Named after the Drosophila Prometaphase Metaphase Plx1 polo gene, polo-like kinases (centrosome) (Plks) are conserved in all eukaryotes. Whereas only a Plx1 single Plk exists in S. cerevisiae (centrosome) (Cdc5p), and presumably S. pombe (Plo1p), at least three Plks are expressed in mammals. Of these, Plk1 (Plx1 in Xenopus) functions Plx1 during mitosis. Information (kinetochore) about the two other mammalian Plks, Snk (Plk2) and Fnk/Prk (Plk3), remains scarce, and there is evidence Telophase/cytokinesis that their functions may not Late anaphase be limited to cell-cycle control94. All known Plks bear the catalytic domain at the amino terminus, and they feature a conspicuously Plx1 conserved sequence motif, the Plx1 (midzone) polo-box, in the carboxy(midbody) terminal domain. This motif has been proposed to function in targeting Plks to subcellular compartments95 or to mediate interactions with other proteins94. Alternatively, the polo-box may constitute part of an autoregulatory domain62. As shown in the figures for Xenopus Plx1, Plks associate transiently with several mitotic structures96–98, including spindle poles, kinetochores, the central spindle midzone and the midbody (Plx1 is stained in red, DNA in blue, and microtubules in green; all pictures kindly provided by P. Meraldi). Concomitantly, kinase activity peaks during M phase, and this results from activating phosphorylation(s). With Xenopus xPlkk1 and the structurally related human kinase, SLK, two candidate Plkactivating kinases have been described recently99,100, but additional regulators of Plks almost certainly await identification. Important issues also remain unresolved with regard to the ubiquitin-dependent degradation of Plks at the end of M phase. Whereas budding yeast Cdc5p features a D-box in the amino terminus and seems to be targeted by APC/CCdc20 (REFS 37,38), this destruction box has not been conserved during evolution. Instead, mammalian Plk1 may constitute a substrate of APC/CCdh1 and thus be degraded primarily during G1 (REF. 39).

KINETOCHORE

Specialized assembly of proteins that binds to a region of the chromosome called the centromere.

substrates in the control of mitotic reactions. Throughout mitosis, microtubule–kinetochore interactions are highly dynamic. They rely on tethering proteins such as cytoplasmic dynein and CENPE28, and may be regulated by phosphorylation. This latter point is shown most convincingly in S. cerevisiae, where the aurora kinase Ipl1p phosphorylates a KINETOCHORE protein, Ndc10p, thereby reducing its ability to bind microtubules29. Recent findings indicate that metazoan B-type aurora kinases localize to centromeres/kinetochores, presumably through interactions with INCENP proteins30. This raises the possibility that aurora kinases regulate kinetochore function also in multicellular organisms. No mammalian homologue of Ndc10p has yet been identified, but in view of the evidence implicating aurora family members in the phosphorylation of histone H3, the centromere-associated histone H3 variant CENP-A is an attractive candidate substrate. Spindle assembly and function throughout mitosis depend on several distinct KRPs and cytoplasmic dynein. However, although coordination between different motor activities would seem critical for the cor-

rect execution of chromosome segregation, little is known about how individual motors are targeted to particular structures and how their local activities are controlled. As exemplified by studies on Eg5 (REFS 16,17), phosphorylation is certainly involved in the spatial and temporal control of motor activity, and this field holds considerable promise for future research. Anaphase onset and mitotic exit

Anaphase begins shortly after all chromosomes have undergone proper bipolar attachment to the spindle. Its onset is characterized by the simultaneous separation of all sister chromatids and results from a loss of sister-chromatid cohesion rather than an increase in forces moving towards the pole. Studies in yeast have revealed that sister-chromatid separation depends on the degradation of an inhibitor, a so-called securin, by ubiquitin-dependent proteolysis31,32. This inhibitor prevents a protease, termed separase, from abolishing sister-chromatid cohesion by cutting a component of a multiprotein complex known as cohesin33. Although this mechanism has undoubtedly been conserved during evolution, the situation is more

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 2 5

REVIEWS

Box 3 | The NIMA kinase family NIMA-related kinases (Neks) are named after the NIMA (never in mitosis A) gene product of Aspergillus nidulans101. Studies in this filamentous fungus suggested that NIMA cooperates with Cdk1 at the G2/M transition and this prompted extensive searches for NIMA homologues in other organisms. However, whether bona fide functional homologues of NIMA exist outside of filamentous fungi remains unclear. Structural relatives of NIMA have been identified in S. cerevisiae and S. pombe101,102, but in contrast to NIMA, these genes are not essential for viability. To what extent they functionally resemble NIMA remains to be determined. The mammalian genome carries at least seven NIMA-related kinases, called Nek1–Nek7 (REF. 103). Of these, Nek2 represents the closest structural relative of NIMA, but, except for putative coiled-coil structures, the non-catalytic domains of Nek2 and NIMA bear no resemblance104. As described in the main text, one important function of Nek2 relates to the control of centrosome structure during the mitotic cell cycle105. However, Nek2 is also highly expressed in the germ line, suggesting that it may have additional roles. NIMA has been implicated in chromosome condensation and proposed to phosphorylate histone H3 (REF. 24). The available evidence argues against such a function for Nek2, but it remains possible that one of the other mammalian Neks contributes to chromosome condensation. Importantly, there is no reason to assume that all Neks have a role in cell-cycle control. The structural similarity between different NIMA family members is largely confined to the catalytic domain, suggesting that different Neks may function in widely different physiological contexts.

CENTROMERE

A region of a eukaryotic chromosome that is attached to the mitotic spindle.

complicated in vertebrates, where different mechanisms seem to destroy cohesion at chromosome arms and CENTROMERES, respectively. The bulk of cohesin is in fact already removed from chromosome arms during prophase, perhaps to permit the extensive chromosome condensation typical of vertebrate mitosis. Importantly, this first wave of cohesin removal does not depend on APC/C and instead requires phosphorylation of cohesin34,35. One kinase able to phosphorylate cohesin is Cdk1, but other mitotic kinases may also be involved. The small amount of cohesin remaining at vertebrate centromeres is then removed at the metaphase-anaphase transition. This second step is dependent on APC/C, and presumably follows the securin–separase pathway described for sister-chromatid separation in yeast36. APC/C is responsible not only for the destruction of anaphase onset inhibitors but also of other proteins, notably mitotic cyclins and several mitotic kinases (FIG. 2). In addition to Cdk1–cyclin, these include Plks37–39, NIMA family members40,41 and aurora kinases42. Importantly, however, the degradation of different substrates occurs at different times, implying that there is exquisite regulation of APC/C. In a typical somatic cell, two forms of APC/C are activated sequentially by the association of two distinct WD40 repeat proteins known as Cdc20 and Cdh1, respectively (for alternative names see legend to FIG. 2). Whereas APC/CCdc20 is active at the metaphase–anaphase transition, APC/CCdh1 is turned on later in mitosis but then remains active throughout the subsequent G1 phase43. When assayed in vitro, APC/CCdc20 and APC/CCdh1 display partly distinct substrate specificities39,41, but it is important to bear in mind that Cdh1 is not expressed in early Xenopus and Drosophila embryos, when cell cycles are extremely rapid and essentially comprise alternating S and M phases44,45. The onset of Cdh1 expression during development then correlates with the establishment of G1 phases, suggesting that the sequential activation of APC/CCdc20 and APC/CCdh1 may be more important for temporal aspects of cellcycle control than substrate selection. Mitotic kinases regulate the two forms of APC/C in opposite fashion

and thus play a key role in establishing the temporal order of APC/C activity: phosphorylation of APC/C core subunits (and perhaps Cdc20) is required for activation of APC/CCdc20, whereas phosphorylation of Cdh1 prevents the activation of APC/CCdh1 (REF. 9). The kinases Cdk1, Plk1 and BubR1 have all been implicated in the activation of APC/CCdc20 (REFS 9,37,38,45–50), and protein kinase A (PKA) has been described as a negative regulator48, but which of these kinases acts directly on APC/C subunits in vivo remains to be established. Cdc20 itself is also phosphorylated and has been detected in a complex with aurora-A51, but the role of Cdc20 phosphorylation is not clear9. With regard to APC/CCdh1, it is striking that the inactivity of this complex correlates with Cdh1 phosphorylation from the onset of S phase until late mitosis, suggesting that Cdh1 is sequentially inactivated by cyclin E, cyclin A and cyclin B-dependent Cdk complexes9. In budding yeast, the activating dephosphorylation of Cdh1 depends on a phosphatase, Cdc14p52,53, that is activated only after the silencing of a spindlepositioning checkpoint (see below). Once activated, APC/CCdh1 promotes mitotic exit by causing the degradation of B-type cyclins and other ubiquitylation substrates. To what extent this model can be extended to mammalian cells remains to be seen. G2/M- and M-phase checkpoints

Surveillance mechanisms, the so-called checkpoint pathways, ensure the proper order and correct execution of cell-cycle events54. Checkpoints are thought to monitor passage through M phase at several stages55,56 (FIG. 3). Some M-phase checkpoints are well established, but others exist primarily in the realm of speculation at present. Best understood are the ‘DNA structure checkpoints’ that arrest cells at the G2/M transition in response to unreplicated DNA or DNA damage, and the ‘spindle assembly checkpoint’ that prevents anaphase onset as long as chromosomal kinetochores do not show a correct bipolar attachment. In budding yeast, a third checkpoint, the ‘spindle-positioning checkpoint’, links Cdk1 inactivation and mitotic exit to the proper orientation of the mitotic spindle. Whether www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS

Box 4 | The aurora kinase family The founding members of the aurora kinase family are Ipl1p from S. cerevisiae and aurora from Drosophila melanogaster19,20. Whereas Ipl1p is the only representative of this family in yeast, two aurora-related kinases are present in Drosophila and C. elegans, and at least three in mammals (TABLE 2). These kinases share similar catalytic domains located in the carboxyl terminus, but their amino-terminal extensions are of variable lengths with little or no similarity. Unfortunately, the nomenclature used to describe Ipl1/aurora-related kinases is highly confusing19,20. This is particularly true with regard to mammals, where orthologues in man, mouse and rat have been given distinct names. For the sake of clarity, the vertebrate aurora kinases are here referred to as aurora-A, -B and -C (for other names, see TABLE 2). Both aurora-A and -B are expressed in proliferating cells and overexpressed in tumour cells. During the cell cycle, the activity of aurora-A peaks before that of aurora-B. Furthermore, the two kinases display strikingly distinct subcellular localizations (TABLE 2). Whereas aurora-A is associated predominantly with centrosomes and the spindle apparatus from prophase through telophase, aurora-B is prominent at the midzone during anaphase and in postmitotic bridges during telophase19,20. It is remarkable that all these mitotic structures also carry mitotic Plks (BOX 2), suggesting that a deliberate search for functional interactions between aurora kinases and Plks might be fruitful. Aurora-C has not yet been studied extensively. It is highly expressed in testis, but can also be detected in tumour cell lines, where it localizes to spindle poles from anaphase to cytokinesis106. Only very little information is currently available on the regulation and substrates of aurora kinases at present (TABLE 2). Similar to Plks, aurora kinases are regulated by APC/C-dependent proteolysis42 and by phosphorylation107,108. However, conflicting data have been reported with regard to the relative timing of activation of aurora and Cdk1 during Xenopus oocyte maturation, and the upstream regulatory enzymes (kinases and phosphatases) remain unknown. Similarly, only few candidate substrates have so far been identified (see TABLE 2). Table 2 | Nomenclature and properties of aurora family kinases Nomenclature guide* Mammals

Other names

Xenopus

C. elegans

Drosophila

S. cerevisiae

Aurora-A

Aurora-2, HsAIRK1, ARK1, Aik, BTAK, STK-15 (human); ARK1, Ayk1, IAK1 (mouse)

Eg2

AIR-1

aurora

Ipl1p‡

Aurora-B

Aurora-1, HsAIRK2, ARK2, Aik2, AIM-1, STK-12 (human); ARK2, STK-1 (mouse); AIM-1 (rat)

AIRK2

AIR-2

IAL

Aurora-C

Aurora-3, HsAIRK3, AIE2, Aik3, STK-13 (human); AIE1 (mouse)

Properties Family member

Localization

Regulation

Putative substrates

Aurora-A

Centrosome Spindle MTs

Phosphorylation Degradation

Kinesin-related motor Eg5 CPEB§

Aurora-B

Kinetochores Spindle midzone

Association with INCENP Phosphorylation? Degradation Functional interaction with Bir-domain proteins (survivin?)

Histone H3?||

Aurora-C

Centrosome

Yeast Ipl1p

Histone H3 Kinetochore protein Ndc10p

*The aurora-A, -B, -C nomenclature has been approved by many scientists working in the aurora field. For the sake of clarity, the future use of this unifying nomenclature is recommended. ‡ Ipl1p is the only aurora family member in budding yeast and cannot be attributed to the any particular subfamily. § CPEB is involved in regulating polyadenylation and translation of c-mos mRNA during Xenopus oocyte maturation115. (CPEB, cytoplasmic polyadenylation-element-binding protein.) || Shown for C. elegans AIR-2, which most probably represents a B-type aurora kinase23.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 2 7

REVIEWS a corresponding checkpoint also exists in mammalian cells is not known. The DNA structure checkpoints. The three enzymes that control the activation of mammalian Cdk1, the phosphatase Cdc25C and the kinases Wee1 and Myt1 (FIG. 1), are themselves phosphorylated by multiple kinases, albeit with different consequences. On the one hand, Cdc25C is inhibited by kinases (Chk1, Chk2) that operate in DNA structure checkpoint signalling, whereas Wee1 and Myt1 are upregulated by the same pathways54,57. On the other hand, Cdk1–cyclin B is able to activate Cdc25C and inactivate Wee1, thereby creating a positive feedback loop. Plk1 also activates Cdc25C12,58, and it may downregulate Wee1 and Myt1. Whether this occurs as part of the feedback loop involving Cdk1–cyclin B or, alternatively, contributes to the initial activation of Cdc25C remains a subject of debate59–62. In view of the many roles emerging for Plks in mitotic progression, it is also interesting that mammalian Plk1 is inhibited on DNA damage checkpoint activation63, extending earlier work in yeast64,65. The spindle assembly checkpoint. Genetic studies in yeast, as well as laser-ablation and micromanipulation studies in animal cells, have identified a checkpoint Pi Inactive Cdc20

Spindle assembly checkpoint

Mad2

Cdh1

Cdc14 phosphatase

Cdk–cyclin

Spindlepositioning checkpoint ?

PKA ? Cdc20 Cdk1 Plk1 BubR1

Cdh1

? Pi ? ?

APC/C

APC/C Pi

Active Metaphase/ anaphase transition

Targets: · Securin (D-box) · B-type cyclins (D-box) · A-type cyclins · Plk1 · Nek2 (KEN-box) · Aurora-A · etc.

Late mitosis and G1 phase

Figure 2 | Phosphorylation differentially regulates two forms of APC/C. APC/C acts on different substrates at different times, implying extensive regulation. This involves the binding of adaptor proteins (Cdc20 and Cdh1, respectively), as well as the phosphorylation and dephosphorylation of both APC/C core subunits and adaptor proteins. Note that Cdc20 is also known as Fizzy (Drosophila), p55CDC (mammals) and Slp1p (S. pombe), whereas synonyms for Cdh1 are Hct1p (S. cerevisiae), Fizzy-related (Drosophila) and Ste9p/Srw1 (S. pombe). Substrates of APC/C can be classified depending on whether they bear D-box or KEN-box consensus motifs, which seem to favour recognition by APC/CCdc20 or APC/CCdh1, respectively39,41. Much remains to be learned, however, about the precise mechanisms that regulate ubiquitin-dependent degradation of individual substrates. For example, it remains mysterious how cyclin A is degraded during prometaphase — a time when securin degradation by APC/CCdc20 is blocked by the spindle assembly checkpoint.

that delays sister-chromatid separation until all chromosomes are properly aligned on the spindle (FIG. 4). This checkpoint monitors the attachment of microtubules to kinetochores and/or the generation of tension that results from bipolar attachment of sister chromatids. Hence, it is also referred to as the kinetochore attachment checkpoint66,67. Yeast mutant screens have identified six gene products involved in this checkpoint, specifically the dualspecificity kinase Mps1p, the kinase Bub1p and its partner Bub3p, and the three proteins Mad1p, Mad2p and Mad3p55. Subsequently, homologues of several Mad and Bub proteins have been shown to associate preferentially with unattached kinetochores in animal cells, confirming earlier cytological evidence for a critical role of phosphorylation in the generation of an anaphase-inhibitory signal at kinetochores28,66,67. Studies done in animal cells also indicate that the spindle assembly checkpoint is not merely activated in response to spindle damage, but contributes to the timing of anaphase onset in every cell division68,69. According to a current model, structural changes induced by microtubule attachment (and/or tension) are translated, through phosphorylation, into a biochemical signal. In vertebrates, this mechano-chemical coupling was proposed to involve a molecular interaction between the KRP CENP-E and the kinase BubR1 (REFS 50,70,71). How this regulates the kinetochore association of Mad proteins is unknown. However, unattached kinetochores are thought to function as sites of continuous assembly and release of Mad2–Cdc20 complexes that prevent the activation of APC/CCdc20. On attachment of the last kinetochore, the production of inhibitory Mad2–Cdc20 complexes ceases, allowing Cdc20 to dissociate from Mad2 and activate APC/C; as a result, securin is degraded and anaphase ensues (FIG. 4). Although attractive, this model leaves many important questions unresolved. In particular, many kinases are localized to centromeres/kinetochores, and it remains to be explained how these kinases interact with each other. This would seem critical for understanding both checkpoint activation and checkpoint silencing. It is also not understood at present why mammalian cells express two Bub1 family members, (Bub1 and BubR1). These are unlikely to be redundant, as both Bub1 and BubR1 are required for checkpoint signalling50,68,72. Both kinases are recruited to unattached kinetochores in association with Bub3, a WD-repeat-containing substrate73,74. Whether Bub3 functions as a regulatory subunit or a downstream effector is not known. Similarly, the precise functions of Mps1p family kinases remain to be uncovered. Epistasis experiments suggest that Mps1p and Bub1p cooperate to generate an anaphase-inhibitory signal, and this may involve the phosphorylation of Mad1p75. Overexpression of Mps1p imposes an M-phase arrest, thereby mimicking checkpoint activation75, and the same is true of the fission yeast homologue Mph1p76. However, no functional studies have yet been reported on the putative mammalian members of the Mps1/Mph1 family, human TTK and mouse Esk. www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS

DNA damage

Absence of centrosome separation ?

Unreplicated DNA

C Unattached kinetochores/ misaligned chromosomes

D Faulty spindle position (?)

Cytokinesis G2

Prophase

Metaphase

Anaphase

Telophase

Figure 3 | Checkpoints ordering progression through M phase. The ‘DNA structure checkpoints’ (A) delay the G2/M transition in response to DNA damage or unreplicated DNA54,57, whereas the ‘spindle assembly checkpoint’ (C) monitors microtubule attachment to kinetochores (or tension that results from the proper bipolar attachment of sister chromatids) and delays anaphase onset until all chromosomes are properly aligned55. Recent data argue that a ‘spindlepositioning checkpoint’ (D) links M-phase exit to correct spindle orientation, but whether this checkpoint operates in organisms other than S. cerevisiae remains to be determined77. Additional work will also be required to corroborate the idea that another checkpoint, perhaps monitoring centrosome separation (B), operates at the onset of mitosis in mammalian cells11,110.

The spindle-positioning checkpoint. Intuitively, it seems plausible that a spindle-positioning checkpoint might enforce the correct orientation of the elongating spindle to ensure that cleavage occurs in the right plane and only after complete separation of sister chromatids. Recently, evidence for such a checkpoint has been obtained in S. cerevisiae. Its silencing requires that a spindle pole body associates productively with the cortex of the budding cell, thus establishing a dependency between correct spindle positioning and mitotic exit77. The first identified component of this pathway was Bub2p, a spindle-poleCheckpoint activated BubR1/Bub3 Bub1/Bub3 ? CENP-E Mad2 Mad1 TTK/Mps1 ? MAPK ? Plk1 ? Aurora-B ?

Securin Separase

Mad2*

APC/CCdc20

Securin degradation (Ub-dependent)

Separase

Checkpoint silenced · Checkpoint proteins displaced · Production of Mad2* ceases (?)

Cohesin cleavage

Anaphase onset

Cohesin phosphorylation ?

Figure 4 | The spindle assembly checkpoint. One model holds that interactions between the kinesin-related protein CENP-E and BubR1 translate structural information (the presence or absence of appropriate microtubule–kinetochore interactions) into a chemical signal (phosphorylation of as yet unidentified substrates). These events are believed to regulate both the recruitment of Mad1–Mad2 complexes to unattached kinetochores and the release of ‘conformationally altered Mad2’ (represented by Mad2*). Mad2* then blocks a productive interaction between Cdc20 and APC/C, thereby preventing the degradation of securin and the cleavage of cohesin by separase. On attachment of the last kinetochore, the production of Mad2* ceases and activation of APC/CCdc20 ensues. The checkpoint is depicted here as an essentially linear pathway, but this should not detract from the importance of multiprotein complexes and spatial organization in the spindle55,111,112. In addition to Bub1 and BubR1, activated MAP kinase113,114, Plk1 (REFS 96,97) and aurora-B30 have also been detected at kinetochores, suggesting that these enzymes may function in either checkpoint signalling (MAPK ?) or silencing (Plk1 and/or aurora-B ?). (Ub, ubiquitin.)

associated subunit of a two-component GTPase-activating protein (GAP). This GAP downregulates the activity of a small GTPase (Tem1p) that in turn functions at an early step in a pathway controlling mitotic exit (FIG. 5). Downstream of active Tem1p, several kinases cooperate in a so-called mitotic exit network (MEN)77 to activate the Cdc14p phosphatase. Cdc14p then acts not only as an activator of APC/CCdh1, but also dephosphorylates the Cdk1-inhibitor Sic1p (causing its stabilization) and the transcription factor Swi5p (enhancing the production of Sic1p), thereby causing the inactivation of budding yeast Cdk1 by three complementary mechanisms52 (FIG. 5). Interestingly, gene products homologous to most components of the MEN pathway have also been identified in S. pombe (FIG. 5). However, as the corresponding genes were identified in studies on septation, a process akin to cytokinesis in animal cells78,79, the term SIN (septation initiation network) was coined. At present, there is no evidence that the SIN is part of a checkpoint controlling Cdk1 inactivation. Taken at face value, this would indicate that apparently homologous gene products control partly distinct processes in the two yeasts. The study of a corresponding pathway in a metazoan organism might help clarify this situation. Signalling cytokinesis

When considering kinases in relation to cytokinesis, it is useful to distinguish signalling pathways that determine the timing and positioning of contractile ring assembly from mechanical aspects of cleavage furrow ingression. Progress in understanding kinase function in relation to the latter process is beyond the scope of this article. With regard to the coordination of cytokinesis with mitotic progression, it remains unclear whether a kinase cascade similar to the SIN pathway operates in metazoan organisms (see legend to FIG. 5). However, both Plks and B-type aurora kinases have been implicated in the control of cytokinesis. The case for Plk was originally built on data from S. pombe, where septation is impaired in the absence of Plo1p, whereas overexpression of the kinase triggers additional rounds of septation from any point in the cell cycle80. It was concluded that Plo1p functions high up in the SIN pathway79,81, and data from Drosophila and mammalian cells support the idea that Plks are upstream regulators of cytokinesis62,81,82. Relevant substrates in Plk mutants have not been identified, but it is possible that the observed cytokinesis defects result from the mislocalization and/or impaired function of the KRP MKLP-1/Pavarotti83,84 or the cytoskeletonassociated protein Mid1p85. It should also be emphasized that Cdc5p is required for generating a mitotic exit signal through the MEN pathway77, which seems to imply a rather indirect link between this budding yeast Plk and cytokinesis. A role for B-type aurora kinases in cytokinesis is supported by the finding that overexpression of a catalytically inactive aurora-B disrupts cleavage furrow formation in mammalian cells86, and by the chromosome segregation and cytokinesis defects observed in C. elegans embryos in which the B-type aurora (AIR-

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 2 9

REVIEWS

S. pombe SIN

Plo1p Cdc5p ?

S. cerevisiae MEN

G-protein switch Bub2p Cdc16p Bfa1p Byr4p Spg1p–GTP Spg1p–GDP Tem1p–GDP Tem1p–GTP Lte1p ?

Kinase (Cdc7p)

Kinase (Cdc15p)

Kinase (Sid1p/Cdc14p)

Kinase (Sid2p/Mob1p)

Kinase (Dbf2p/Mob1p)

Cytokinesis

Phosphatase (Cdc14p) Sic1p

Swi5p APC/CCdh1 Cdk1 inactivation Mitotic exit

Figure 5 | Mitotic kinases regulating mitotic exit and cytokinesis. Diagram comparing the septation initiation network (SIN) and the mitotic exit network (MEN), two recently described regulatory networks in S. pombe and S. cerevisiae, respectively. At present, these pathways are based primarily on results from genetic (epistasis) and cytological experiments. For the sake of simplicity, they are depicted in a linear fashion, but biochemical data on direct interactions between the known components remain scarce and additional factors almost certainly await discovery. Interestingly, the two yeast networks comprise several homologous gene products, many of which are kinases or phosphatases. However, although the genomes of C. elegans, Drosophila and mammals harbour candidate homologues of some of these yeast genes (for example, budding yeast CDC14 and DBF2/MOB1), the existence of homologues of other components (notably the fission yeast Sid1p kinase) is less obvious. Thus, although it is likely that functionally similar networks exist in metazoan organisms, this remains to be proved. (Note that fission yeast Cdc14p is not a phosphatase but a subunit of the Sid1p kinase. The use of the same acronym for the two proteins is purely accidental.)

2) was suppressed by RNAi87. A similar phenotype was observed on elimination of the protein Bir1 by RNAi, suggesting that B-type aurora and Bir1 interact either directly or indirectly88. This is intriguing because survivin, a potential mammalian Bir1 homologue, has been implicated in the protection of cells from apoptosis. If survivin indeed plays such a role, further exploration of a functional interaction between kinetochores, B-type aurora kinases and survivin might provide insight into the important connection between aneuploidy, apoptosis and tumorigenesis. Mitotic kinases in cancer

Chromosome rearrangements and aneuploidy are hallmarks of most human cancers, and severe karyotypic anomalies generally correlate with poor prognosis. Considering the central role of phosphorylation

in mitotic checkpoints, spindle function and chromosome segregation, it is not surprising that several mitotic kinases have been implicated in tumorigenesis. For example, aurora kinases map to chromosomal regions that are frequently altered in tumours (auroraA, 20q13.2–q.13.3; aurora-B, 17p13; aurora-C, 19q13.3-ter). Aurora-A is overexpressed in many primary tumours and is able to transform cells in culture89, suggesting that it is a relevant gene on the 20q13.2 amplicon20. Furthermore, ectopic expression of aurora-A causes centrosome amplification and aneuploidy in cultured cells, although the underlying mechanism remains to be unravelled90. Similarly, Plk1 can transform rodent cells91 and is frequently overexpressed in tumours, making it a potential marker for diagnostic or prognostic purposes92. Finally, dominantly acting Bub1 mutations were identified in some colorectal tumour cells lines, fuelling the speculation that mitotic checkpoints are commonly inactivated in aneuploid tumours72. The known genes for spindle assembly checkpoint are only rarely mutated in tumours93, however, implying that the molecular origins of aneuploidy await discovery and that the search for suspects must go on. Conclusions and perspectives

www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS 1.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

24.

Meraldi, P., Lukas, J., Fry, A. M., Bartek, J. & Nigg, E. A. Centrosome duplication in mammalian somatic cells requires E2F and Cdk2–cyclin A. Nature Cell Biol. 1, 88–93 (1999). Matsumoto, Y., Hayashi, K. & Nishida, E. Cyclindependent kinase 2 (Cdk2) is required for centrosome duplication in mammalian cells. Curr. Biol. 9, 429–432 (1999). Hinchcliffe, E. H., Li, C., Thompson, E. A., Maller, J. L. & Sluder, G. Requirement of Cdk2-cyclin E activity for repeated centrosome reproduction in Xenopus egg extracts. Science 283, 851–854 (1999). Lauze, E. et al. Yeast spindle pole body duplication gene MPS1 encodes an essential dual specificity protein kinase. EMBO J. 14, 1655–1663 (1995). Nigg, E. A. Cyclin-dependent protein kinases: key regulators of the eukaryotic cell cycle. BioEssays 17, 471480 (1995). Kimura, K., Hirano, M., Kobayashi, R. & Hirano, T. Phosphorylation and activation of 13S condensin by Cdc2 in vitro. Science 282, 487–490 (1998). Lowe, M. et al. Cdc2 kinase directly phosphorylates the cis-Golgi matrix protein GM130 and is required for Golgi fragmentation in mitosis. Cell 94, 783–793 (1998). Andersen, S. S. Balanced regulation of microtubule dynamics during the cell cycle: a contemporary view. BioEssays 21, 53–60 (1999). Kramer, E. R., Scheuringer, N., Podtelejnikov, A. V., Mann, M. & Peters, J. M. Mitotic regulation of the APC activator proteins CDC20 and CDH1. Mol. Biol. Cell 11, 1555–1569 (2000). A careful biochemical study on the role of phosphorylation in controlling the activities of the Cdc20 and Cdh1 forms of APC/C. Noton, E. & Diffley, J. F. CDK inactivation is the only essential function of the APC/C and the mitotic exit network proteins for origin resetting during mitosis. Mol. Cell 5, 85–95 (2000). Lane, H. A. & Nigg, E. A. Antibody microinjection reveals an essential role for human polo-like kinase 1 (Plk1) in the functional maturation of mitotic centrosomes. J. Cell Biol. 135, 1701–1713 (1996). Qian, Y. W., Erikson, E., Li, C. & Maller, J. L. Activated polo-like kinase Plx1 is required at multiple points during mitosis in Xenopus laevis. Mol. Cell Biol. 18, 4262–4271 (1998). do Carmo-Avides M. & Glover, D. M. Abnormal spindle protein, Asp, and the integrity of mitotic centrosomal microtubule organizing centers. Science 283, 1733–1735 (1999). Fry, A. M. et al. C-Nap1, a novel centrosomal coiled-coil protein and candidate substrate of the cell cycle-regulated protein kinase Nek2. J. Cell Biol. 141, 1563–1574 (1998). Helps, N. R., Luo, X., Barker, H. M. & Cohen, P. T. NIMArelated kinase 2 (Nek2), a cell-cycle-regulated protein kinase localized to centrosomes, is complexed to protein phosphatase 1. Biochem. J. 349, 509–518 (2000). Sawin, K. E. & Mitchison, T. J. Mutations in the kinesin-like protein Eg5 disrupting localization to the mitotic spindle. Proc. Natl Acad. Sci. USA 92, 4289–4293 (1995). Blangy, A. et al. Phosphorylation by p34cdc2 regulates spindle association of human Eg5, a kinesin-related motor essential for bipolar spindle formation in vivo. Cell 83, 1159–1169 (1995). Glover, D. M., Leibowitz, M. H., McLean, D. A. & Parry, H. Mutations in aurora prevent centrosome separation leading to the formation of monopolar spindles. Cell 81, 95–105 (1995). Giet, R. & Prigent, C. Aurora/Ipl1p-related kinases, a new oncogenic family of mitotic serine–threonine kinases. J. Cell Sci. 112, 3591–3601 (1999). Bischoff, J. R. & Plowman, G. D. The Aurora/Ipl1p kinase family: regulators of chromosome segregation and cytokinesis. Trends Cell Biol. 9, 454–459 (1999). Schumacher, J. M., Ashcroft, N., Donovan, P. J. & Golden, A. A highly conserved centrosomal kinase, AIR-1, is required for accurate cell cycle progression and segregation of developmental factors in Caenorhabditis elegans embryos. Development 125, 4391–4402 (1998). Giet, R., Uzbekov, R., Cubizolles, F., Le Guellec, K. & Prigent, C. The Xenopus laevis aurora-related protein kinase pEg2 associates with and phosphorylates the kinesin-related protein XlEg5. J. Biol. Chem. 274, 15005–15013 (1999). Hsu, J. Y. et al. Mitotic phosphorylation of histone H3 is governed by Ipl1/aurora kinase and Glc7/PP1 phosphatase in budding yeast and nematodes. Cell 102, 279–291 (2000). De Souza, C. P., Osmani, A. H., Wu, L. P., Spotts, J. L. & Osmani, S. A. Mitotic histone H3 phosphorylation by the

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

42.

43. 44.

45.

46.

NIMA kinase in Aspergillus nidulans. Cell 102, 293–302 (2000). Sutani, T. et al. Fission yeast condensin complex: essential roles of non-SMC subunits for condensation and Cdc2 phosphorylation of Cut3/SMC4. Genes Dev. 13, 2271–2283 (1999). Larsson, N., Marklund, U., Gradin, H. M., Brattsand, G. & Gullberg, M. Control of microtubule dynamics by oncoprotein 18: dissection of the regulatory role of multisite phosphorylation during mitosis. Mol. Cell Biol. 17, 5530–5539 (1997). Andersen, S. S. et al. Mitotic chromatin regulates phosphorylation of Stathmin/Op18. Nature 389, 640–643 (1997). Rieder, C. L. & Salmon, E. D. The vertebrate cell kinetochore and its roles during mitosis. Trends Cell Biol. 8, 310–318 (1998). Biggins, S. et al. The conserved protein kinase Ipl1 regulates microtubule binding to kinetochores in budding yeast. Genes Dev. 13, 532–544 (1999). Adams, R. R. et al. INCENP binds the aurora-related kinase AIRK2 and is required to target it to chromosomes, the central spindle and cleavage furrow. Curr. Biol. 10, 1075–1078 (2000). This paper identifies the ‘chromosomal passenger protein’ INCENP as a partner of aurora-B, and shows that INCENP is required for correct targeting of the kinase. Nasmyth, K., Peters, J. M. & Uhlmann, F. Splitting the chromosome: cutting the ties that bind sister chromatids. Science 288, 1379–1385 (2000). Yanagida, M. Cell cycle mechanisms of sister chromatid separation; roles of Cut1/separin and Cut2/securin. Genes Cells 5, 1–8 (2000). Uhlmann, F., Wernic, D., Poupart, M. A., Koonin, E. V. & Nasmyth, K. Cleavage of cohesin by the CD clan protease separin triggers anaphase in yeast. Cell 103, 375–386 (2000). This important paper identifies the Esp1p gene product as a protease (separase) and elegantly shows that cohesin cleavage by separase is sufficient for sister chromatid separation in budding yeast. Sumara, I., Vorlaufer, E., Gieffers, C., Peters, B. H. & Peters, J. M. Characterization of vertebrate cohesin complexes and their regulation in prophase. J. Cell Biol. 151, 749–762 (2000). Losada, A., Yokochi, T., Kobayashi, R. & Hirano, T. Identification and characterization of SA/Scc3p subunits in the Xenopus and human cohesin complexes. J. Cell Biol. 150, 405–416 (2000). Waizenegger, I. C., Hauf, S., Meinke, A. & Peters, J. M. Two distinct pathways remove mammalian cohesin from chromosome arms in prophase and from centromeres in anaphase. Cell 103, 399–410 (2000). Shirayama, M., Zachariae, W., Ciosk, R. & Nasmyth, K. The Polo-like kinase Cdc5p and the WD-repeat protein Cdc20p/fizzy are regulators and substrates of the anaphase promoting complex in Saccharomyces cerevisiae. EMBO J. 17, 1336–1349 (1998). Charles, J. F. et al. The Polo-related kinase Cdc5 activates and is destroyed by the mitotic cyclin destruction machinery in S. cerevisiae. Curr. Biol. 8, 497–507 (1998). Fang, G., Yu, H. & Kirschner, M. W. Direct binding of CDC20 protein family members activates the anaphasepromoting complex in mitosis and G1. Mol. Cell 2, 163–171 (1998). Ye, X. S., Fincher, R. R., Tang, A., Osmani, A. H. & Osmani, S. A. Regulation of the anaphase-promoting complex/cyclosome by bimAAPC3 and proteolysis of NIMA. Mol. Biol. Cell 9, 3019–3030 (1998). Pfleger, C. M. & Kirschner, M. W. The KEN box: an APC recognition signal distinct from the D box targeted by cdh1. Genes Dev. 14, 655–665 (2000). Honda, K. et al. Degradation of human Aurora2 protein kinase by the anaphase-promoting complex-ubiquitinproteasome pathway. Oncogene 19, 2812–2819 (2000). Morgan, D. O. Regulation of the APC and the exit from mitosis. Nature Cell Biol. 1, E47–E53 (1999). Sigrist, S. J. & Lehner, C. F. Drosophila fizzy-related downregulates mitotic cyclins and is required for cell proliferation arrest and entry into endocycles. Cell 90, 671–681 (1997). Lorca, T. et al. Fizzy is required for activation of the APC/cyclosome in Xenopus egg extracts. EMBO J. 17, 3565–3575 (1998). Descombes, P. & Nigg, E. A. The polo-like kinase Plx1 is required for M phase exit and destruction of mitotic regulators in Xenopus egg extracts. EMBO J. 17, 1328–1335 (1998).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

47. Patra, D. & Dunphy, W. G. Xe-p9, a Xenopus Suc1/Cks protein, is essential for the Cdc2-dependent phosphorylation of the anaphase-promoting complex at mitosis. Genes Dev. 12, 2549–2559 (1998). 48. Kotani, S. et al. PKA and MPF-activated polo-like kinase regulate anaphase-promoting complex activity and mitosis progression. Mol. Cell 1, 371–380 (1998). 49. Kotani, S., Tanaka, H., Yasuda, H. & Todokoro, K. Regulation of APC activity by phosphorylation and regulatory factors. J. Cell Biol. 146, 791–800 (1999). 50. Chan, G. K., Jablonski, S. A., Sudakin, V., Hittle, J. C. & Yen, T. J. Human BUBR1 is a mitotic checkpoint kinase that monitors CENP-E functions at kinetochores and binds the cyclosome/APC. J. Cell Biol. 146, 941–954 (1999). 51. Farruggio, D. C., Townsley, F. M. & Ruderman, J. V. Cdc20 associates with the kinase aurora2/Aik. Proc. Natl Acad. Sci. USA 96, 7306–7311 (1999). 52. Visintin, R. et al. The phosphatase Cdc14 triggers mitotic exit by reversal of Cdk-dependent phosphorylation. Mol. Cell 2, 709–718 (1998). 53. Jaspersen, S. L., Charles, J. F. & Morgan, D. O. Inhibitory phosphorylation of the APC regulator Hct1 is controlled by the kinase Cdc28 and the phosphatase Cdc14. Curr. Biol. 9, 227–236 (1999). 54. Elledge, S. J. Cell cycle checkpoints: preventing an identity crisis. Science 274, 1664–1672 (1996). 55. Hardwick, K. G. The spindle checkpoint. Trends Genet. 14, 1–4 (1998). 56. Burke, D. J. Complexity in the spindle checkpoint. Curr. Opin. Genet. Dev. 10, 26–31 (2000). 57. Russell, P. Checkpoints on the road to mitosis. Trends Biochem. Sci. 23, 399–402 (1998). 58. Kumagai, A. & Dunphy, W. G. Purification and molecular cloning of Plx1, a Cdc25-regulatory kinase from Xenopus egg extracts. Science 273, 1377–1380 (1996). This paper identifies the Cdc25 phosphatase as a likely physiological substrate of Plx1 and thereby implicates Plks in controlling Cdk1 activity. 59. Abrieu, A. et al. The Polo-like kinase Plx1 is a component of the MPF amplification loop at the G2/M-phase transition of the cell cycle in Xenopus eggs. J. Cell Sci. 111, 1751–1757 (1998). 60. Karaiskou, A., Jessus, C., Brassac, T. & Ozon, R. Phosphatase 2A and polo kinase, two antagonistic regulators of cdc25 activation and MPF auto-amplification. J. Cell Sci. 112, 3747–3756 (1999). 61. Qian, Y. W., Erikson, E. & Maller, J. L. Mitotic effects of a constitutively active mutant of the Xenopus polo-like kinase Plx1. Mol. Cell Biol. 19, 8625–8632 (1999). 62. Nigg, E. A. Polo-like kinases: positive regulators of cell division from start to finish. Curr. Opin. Cell Biol. 10, 776–783 (1998). 63. Smits, V. A. et al. Polo-like kinase-1 is a target of the DNA damage checkpoint. Nature Cell Biol. 2, 672–676 (2000). 64. Sanchez, Y. et al. Control of the DNA damage checkpoint by chk1 and rad53 protein kinases through distinct mechanisms. Science 286, 1166–1171 (1999). 65. Toczyski, D. P., Galgoczy, D. J. & Hartwell, L. H. CDC5 and CKII control adaptation to the yeast DNA damage checkpoint. Cell 90, 1097–1106 (1997). 66. Gorbsky, G. J. Cell cycle checkpoints: arresting progress in mitosis. BioEssays 19, 193–197 (1997). 67. Nicklas, R. B. How cells get the right chromosomes. Science 275, 632–637 (1997). 68. Taylor, S. S. & McKeon, F. Kinetochore localization of murine Bub1 is required for normal mitotic timing and checkpoint response to spindle damage. Cell 89, 727–735 (1997). This paper implicates the murine Bub1 kinase in the timing of anaphase onset during normal mitosis. 69. Basu, J. et al. Mutations in the essential spindle checkpoint gene bub1 cause chromosome missegregation and fail to block apoptosis in Drosophila. J. Cell Biol. 146, 13–28 (1999). 70. Yao, X., Abrieu, A., Zheng, Y., Sullivan, K. F. & Cleveland, D. W. CENP-E forms a link between attachment of spindle microtubules to kinetochores and the mitotic checkpoint. Nature Cell Biol. 2, 484–491 (2000). 71. Abrieu, A., Kahana, J. A., Wood, K. W. & Cleveland, D. W. CENP-E as an essential component of the mitotic checkpoint in vitro. Cell 102, 817–826 (2000). 72. Cahill, D. P. et al. Mutations of mitotic checkpoint genes in human cancers. Nature 392, 300-303 (1998). 73. Basu, J. et al. Localization of the Drosophila checkpoint control protein Bub3 to the kinetochore requires Bub1 but not Zw10 or Rod. Chromosoma 107, 376–385 (1998). 74. Taylor, S. S., Ha, E. & McKeon, F. The human homologue of Bub3 is required for kinetochore localization of Bub1 and a Mad3/Bub1-related protein kinase. J. Cell Biol. 142, 1–11 (1998).

VOLUME 2 | JANUARY 2001 | 3 1

REVIEWS 75. Hardwick, K. G., Weiss, E., Luca, F. C., Winey, M. & Murray, A. W. Activation of the budding yeast spindle assembly checkpoint without mitotic spindle disruption. Science 273, 953–956 (1996). 76. He, X., Jones, M. H., Winey, M. & Sazer, S. Mph1, a member of the Mps1-like family of dual specificity protein kinases, is required for the spindle checkpoint in S. pombe. J. Cell Sci. 111, 1635–1647 (1998). 77. Hoyt, M. A. Exit from mitosis: spindle pole power. Cell 102, 267–270 (2000). 78. Le, G., X, Utzig, S. & Simanis, V. Controlling septation in fission yeast: finding the middle, and timing it right. Curr. Genet. 35, 571–584 (1999). 79. Balasubramanian, M. K., McCollum, D. & Surana, U. Tying the knot: linking cytokinesis to the nuclear cycle. J. Cell Sci. 113, 1503–1513 (2000). 80. Ohkura, H., Hagan, I. M. & Glover, D. M. The conserved ikinase plo1, required to form a bipolar spindle, the actin ring, and septum, can drive septum formation in G1 and G2 cells. Genes Dev. 9, 1059–1073 (1995). 81. Glover, D. M., Hagan, I. M. & Tavares, A. A. Polo-like kinases: a team that plays throughout mitosis. Genes Dev. 12, 3777–3787 (1998). 82. Carmena, M. et al. Drosophila polo kinase is required for cytokinesis. J. Cell Biol. 143, 659–671 (1998). 83. Lee, K. S., Yuan, Y. L., Kuriyama, R. & Erikson, R. L. Plk is an M-phase-specific protein kinase and interacts with a kinesin-like protein, CHO1/MKLP-1. Mol. Cell Biol. 15, 7143–7151 (1995). 84. Adams, R. R., Tavares, A. A., Salzberg, A., Bellen, H. J. & Glover, D. M. pavarotti encodes a kinesin-like protein required to organize the central spindle and contractile ring for cytokinesis. Genes Dev. 12, 1483–1494 (1998). 85. Bahler, J. et al. Role of polo kinase and Mid1p in determining the site of cell division in fission yeast. J. Cell Biol. 143, 1603–1616 (1998). 86. Terada, Y. et al. AIM-1: a mammalian midbody-associated protein required for cytokinesis. EMBO J. 17, 667–676 (1998). This study characterizes mammalian aurora-B and provides evidence that this kinase has a role in cytokinesis. 87. Schumacher, J. M., Golden, A. & Donovan, P. J. AIR-2: An Aurora/Ipl1-related protein kinase associated with chromosomes and midbody microtubules is required for polar body extrusion and cytokinesis in Caenorhabditis elegans embryos. J. Cell Biol. 143, 1635–1646 (1998). 88. Speliotes, E. K., Uren, A., Vaux, D. & Horvitz, H. R. The survivin-like C. elegans BIR-1 protein acts with the Auroralike kinase AIR-2 to affect chromosomes and the spindle midzone. Mol. Cell 6, 211–223 (2000). 89. Bischoff, J. R. et al. A homologue of Drosophila aurora kinase is oncogenic and amplified in human colorectal cancers. EMBO J. 17, 3052–3065 (1998). This paper provides evidence for a causal

90.

91.

92.

93.

94.

95.

96.

97.

98.

99.

100.

101.

102.

103.

relationship between aurora-A overexpression and tumorigenesis. Zhou, H. et al. Tumour amplified kinase STK15/BTAK induces centrosome amplification, aneuploidy and transformation. Nature Genet. 20, 189–193 (1998). Smith, M. R. et al. Malignant transformation of mammalian cells initiated by constitutive expression of the polo-like kinase. Biochem. Biophys. Res. Commun. 234, 397–405 (1997). Knecht, R. et al. Prognostic significance of polo-like kinase (PLK) expression in squamous cell carcinomas of the head and neck. Cancer Res. 59, 2794–2797 (1999). Cahill, D. P. et al. Characterization of MAD2B and other mitotic spindle checkpoint genes. Genomics 58, 181–187 (1999). Kauselmann, G. et al. The polo-like protein kinases Fnk and Snk associate with a Ca2+- and integrin-binding protein and are regulated dynamically with synaptic plasticity. EMBO J. 18, 5528–5539 (1999). Lee, K. S., Grenfell, T. Z., Yarm, F. R. & Erikson, R. L. Mutation of the polo-box disrupts localization and mitotic functions of the mammalian polo kinase Plk. Proc. Natl Acad. Sci. USA 95, 9301–9306 (1998). Arnaud, L., Pines, J. & Nigg, E. A. GFP tagging reveals human Polo-like kinase 1 at the kinetochore/centromere region of mitotic chromosomes. Chromosoma 107, 424–429 (1998). Wianny, F., Tavares, A., Evans, M. J., Glover, D. M. & Zernicka-Goetz, M. Mouse polo-like kinase 1 associates with the acentriolar spindle poles, meiotic chromosomes and spindle midzone during oocyte maturation. Chromosoma 107, 430–439 (1998). Golsteyn, R. M., Mundt, K. E., Fry, A. M. & Nigg, E. A. Cell cycle regulation of the activity and subcellular localization of Plk1, a human protein kinase implicated in mitotic spindle function. J. Cell Biol. 129, 1617–1628 (1995). Qian, Y. W., Erikson, E. & Maller, J. L. Purification and cloning of a protein kinase that phosphorylates and activates the polo-like kinase Plx1. Science 282, 1701–1704 (1998). An impressive biochemical study that identifies the first candidate upstream regulator of a Plk. Ellinger-Ziegelbauer, H. et al. Ste20-like kinase (SLK), a regulatory kinase for polo-like kinase (Plk) during the G2/M transition in somatic cells. Genes Cells 5, 491–498 (2000). Osmani, S. A. & Ye, X. S. Cell cycle regulation in Aspergillus by two protein kinases. Biochem. J. 317, 633–641 (1996). Krien, M. J. et al. A NIMA homologue promotes chromatin condensation in fission yeast. J. Cell Sci. 111, 967–976 (1998). Kandli, M., Feige, E., Chen, A., Kilfin, G. & Motro, B. Isolation and characterization of two evolutionarily conserved murine kinases (Nek6 and nek7) related to the

104.

105.

106.

107.

108.

109. 110.

111.

112.

113.

114.

115.

fungal mitotic regulator, NIMA. Genomics 68, 187–196 (2000). Fry, A. M., Schultz, S. J., Bartek, J. & Nigg, E. A. Substrate specificity and cell cycle regulation of the Nek2 protein kinase, a potential human homolog of the mitotic regulator NIMA of Aspergillus nidulans. J. Biol. Chem. 270, 12899–12905 (1995). Fry, A. M., Meraldi, P. & Nigg, E. A. A centrosomal function for the human Nek2 protein kinase, a member of the NIMA family of cell cycle regulators. EMBO J. 17, 470–481 (1998). This study implicates Nek2 in the regulation of the centrosome, thereby providing first insights into the role of a mammalian NIMA kinase family member. Kimura, M., Matsuda, Y., Yoshioka, T. & Okano, Y. Cell cycle-dependent expression and centrosome localization of a third human aurora/Ipl1-related protein kinase, AIK3. J. Biol. Chem. 274, 7334–7340 (1999). Andresson, T. & Ruderman, J. V. The kinase Eg2 is a component of the Xenopus oocyte progesterone-activated signaling pathway. EMBO J. 17, 5627–5637 (1998). Frank-Vaillant, M. et al. Progesterone regulates the accumulation and the activation of Eg2 kinase in Xenopus oocytes. J. Cell Sci. 113, 1127–1138 (2000). Pines, J. Four-dimensional control of the cell cycle. Nature Cell Biol. 1, E73–E79 (1999). Scolnick, D. M. & Halazonetis, T. D. Chfr defines a mitotic stress checkpoint that delays entry into metaphase. Nature 406, 430–435 (2000). Rieder, C. L. et al. Mitosis in vertebrate somatic cells with two spindles: implications for the metaphase/anaphase transition checkpoint and cleavage. Proc. Natl Acad. Sci. USA 94, 5107–5112 (1997). Hardwick, K. G., Johnston, R. C., Smith, D. L. & Murray, A. W. MAD3 encodes a novel component of the spindle checkpoint which interacts with bub3p, cdc20p, and mad2p. J. Cell Biol. 148, 871–882 (2000). Shapiro, P. S. et al. Activation of the MKK/ERK pathway during somatic cell mitosis: direct interactions of active ERK with kinetochores and regulation of the mitotic 3F3/2 phosphoantigen. J. Cell Biol. 142, 1533–1545 (1998). Zecevic, M. et al. Active MAP kinase in mitosis: localization at kinetochores and association with the motor protein CENP-E. J. Cell Biol. 142, 1547–1558 (1998). Mendez, R. et al. Phosphorylation of CPE binding factor by Eg2 regulates translation of c-mos mRNA. Nature 404, 302–307 (2000).

Acknowledgements I thank F. Barr, P. Duncan, W. Earnshaw, A. Fry, P. Meraldi, J. Pines, H. Silljé and S. Wheatley for helpful comments on the manuscript, and P. Meraldi for generously providing immunofluorescence figures. My apologies go to all authors whose primary work could not be cited because of space constraints.

www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS

PLANT CELL DIVISION: BUILDING WALLS IN THE RIGHT PLACES Laurie G. Smith Plant cells are surrounded by walls that define their shapes and fix their positions with tissues. Consequently, establishment of a plant’s cellular framework during development depends largely on the positions in which new walls are formed during cytokinesis. Experiments using various approaches are now building on classical studies to shed light on the mechanisms underlying the spatial control of cytokinesis. CELL DIVISION PROPHASE

An early stage of the cell cycle, during which the chromosomes become condensed in preparation for mitosis. TRANSVACUOLAR CYTOPLASMIC STRANDS

Strands of cytoplasm passing through the vacuole that link the cytoplasm around the nucleus with the cortical cytoplasm just inside the plasma membrane. MICROTUBULE

A hollow tube, 25 nm in diameter, formed by the lateral association of 13 protofilaments, which are themselves polymers of α- and β-tubulin subunits.

Section of Cell and Developmental Biology, University of California at San Diego, 9500 Gilman Drive, La Jolla, California 92093-0116, USA. e-mail: lsmith@biomail.ucsd.edu

Over 100 years ago, plant biologists recognized that the orientation of division for most cells could be predicted by their shapes. In 1863, Hofmeister noted that new cell walls are usually formed in a plane perpendicular to the main axis of cell expansion — that is, perpendicular to the long axis of the mother cell1. In 1888, Errera formulated the rule that the plane of division for most plant cells corresponds to the shortest path that will halve the volume of the parental cell2. Experimental evidence supporting the idea of a direct link between cell shape and the plane of division has come from studies in which shapes were altered by application of a compressive force. Spherical cells within multicellular clumps of callus3, or isolated by suspension of single cells in semi-solid medium4, divide in random orientations. However, when compressed into oval shapes, cells become strongly biased towards division in the plane perpendicular to the long axis of the oval (FIG. 1). Over 60 years ago, Sinnott and Block reported that the plane of division in highly vacuolated plant cells is predicted during PROPHASE by a plate-like arrangement of TRANSVACUOLAR CYTOPLASMIC STRANDS that radiate away from the nucleus, linking the cytoplasm surrounding the nucleus to the cortical cytoplasm5 (FIG. 2). Subsequently, visualization of the cytoskeleton revealed that during prophase, cortical MICROTUBULES in highly vacuolated and densely cytoplasmic cells alike are organized into a beltlike arrangement circumscribing the future division plane called a PREPROPHASE BAND (PPB)6–8 (FIG. 3). More recent studies have shown that F-ACTIN is also a compo-

Spherical Division

Compressed Division

Figure 1 | Effects of deformation on the plane of cell division. Spherical cells with or without intact cell walls suspended in an agar block divide in random orientations. Compression of the agar block deforms cells into oval shapes, causing about 75% to divide in a plane parallel to the compressive force, perpendicular to the long axis of the cell. Most of the remaining cells divide in a plane parallel to the long axis4.

nent of the PPB9–11 (FIG. 3). So plant cells seem to select a division plane early in the cell cycle (before mitosis), which is marked by the location of the PPB, and also by the phragmosome in highly vacuolated cells. CYTOKINESIS in plant cells is achieved through the construction of a new cell wall between daughter nuclei after mitosis. This process is directed by a cytoskeletal structure called the PHRAGMOPLAST, which is made up in part by two interdigitated discs of parallel microtubules (FIG. 3). Microtubules of the phragmoplast are thought to guide the movement of Golgi-derived vesicles containing cell-wall materials to the equator of the phragmoplast, where these vesicles fuse together, gradually coalescing to form a new cell wall8,12 (BOX 1). Actin fila-

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 3 3

REVIEWS Transvacuolar cytoplasmic strands

Phragmosome

Division

Figure 2 | Selection of the division plane in an elongated, highly vacuolated cell18–20. At interphase, transvacuolar cytoplasmic strands radiate from the nucleus in all directions, linking the perinuclear cytoplasm to the cortical cytoplasm. As the cell prepares for division, the nucleus migrates to its centre. Subsequently, most of the cytoplasmic strands gather in the future division plane to form a phragmosome in the plane corresponding to the shortest path that halves the volume of the mother cell as predicted by Errera’s rule2.

ments are also present in the phragmoplast, mostly lying parallel to the microtubules (FIG. 3), but their functions are not well understood13. The phragmoplast arises between daughter nuclei from the remnants of the mitotic spindle, initially in isolation from the parental wall and plasma membrane, and then expands radially to complete the formation of the new cell wall. The site at which the new cell wall will become attached to the parental wall seems to be governed by an interaction

between the phragmoplast and a specialized cortical site, the ‘division site’, which is left behind when the PPB is disassembled upon entry into mitosis8. In this review, research is discussed that contributes to our understanding of the selection and establishment of division sites before mitosis, and guidance of phragmoplasts and associated new cell walls to the division site during cytokinesis. Exciting advances that have recently been made in understanding the process of cytokinesis itself (how new cell walls are formed) are not discussed here, but have recently been reviewed elsewhere12–16. Selection of the division plane

How might cells read their shapes to select division planes that follow Hofmeister’s and Errera’s rules? As illustrated in FIG. 2, transvacuolar cytoplasmic strands radiate from the nucleus in all directions before prophase in highly vacuolated cells. As a cell enters prophase, these strands become rearranged to form a phragmosome (FIG. 2). Experiments showing that premitotic transvacuolar cytoplasmic strands are under tension17 support the idea that, as a cell enters prophase, these strands might pull the nucleus to the centre of the cell and then adopt an arrangement that minimizes their lengths, thereby forming a phragmosome18–20. If the gathering of phragmosomal strands into a single

Box 1 | Plant cell walls

PREPROPHASE BAND

(PPB) A cytoskeletal array composed of F-actin and microtubules that is found in the cell cortex of plant cells during prophase. Its position predicts the future location of the new cell wall. F-ACTIN

(Filamentous actin). A flexible, helical polymer of G-actin (globular actin) monomers that is 5–9 nm in diameter. CYTOKINESIS

The division of one cell into two at the conclusion of the cell cycle. PHRAGMOPLAST

A cytoskeletal structure composed of microtubules and actin filaments that guides the formation of a new cell wall during cytokinesis (see FIG. 3).

The extracellular matrix of plant cells is commonly called the cell wall. It is a dense matrix that comprises mainly polysaccharides, but also Middle contains proteins and lipids. If lamella Pectin the cell wall is stripped off by enzymatic treatment, the wallless cell (called a ‘protoplast’) assumes a spherical shape. In Primary cell wall the presence of the wall, the protoplast assumes the shape defined by the wall that Cellulose surrounds it, much as if it were a balloon squeezed into a box. Plasma membrane The network of cell walls Hemicellulose provides a structural framework for the plant. The 50 nm walls of fully grown portions of the plant body are tough and rigid ‘secondary’ walls that are highly resistant to both tension and compression — this is what allows plants to attain a large size without having a skeleton. By contrast, the ‘primary’ wall surrounding cells that are still dividing and/or expanding are not as rigid, and can stretch to allow cells to grow. Primary walls contain three main classes of polysaccharide. The first is cellulose, an unbranched polymer of glucose subunits that is synthesized and deposited into the wall by an enzyme complex (cellulose synthase) in the plasma membrane. Individual cellulose polymers associate into bundles called ‘microfibrils’. Cellulose microfibrils are crosslinked together by two other classes of polysaccharide — hemicelluloses and pectins — which are both branched polysaccharides of varying composition synthesized in the Golgi and deposited into the wall through secretion. During cytokinesis, when a new cell wall is initially formed, its composition is different — it contains mainly callose rather than cellulose, a different polymer of glucose subunits that is also synthesized by an enzyme complex in the plasma membrane (callose synthase). After completion of the new cell wall, flattening and rigidification of the wall is associated with replacement of callose with cellulose. After a cell stops growing, more components are added to form thicker and more rigid secondary walls — in woody tissues, for example, deposition of lignin makes cell walls extremely tough and rigid. (For reviews, see REFS 12, 52 and 53. Figure modified from REF. 54.)

www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS

STOMATAL COMPLEX

A group of cells in the shoot epidermis, which functions as a regulated aperture that controls the exchange of gases and moisture between internal tissues and the air. GUARD MOTHER CELL

A precursor cell that divides to form a pair of guard cells (see ‘stomatal complex’). SUBSIDIARY MOTHER CELL

A precursor cell that divides to form at least one subsidiary cell. SUBSIDIARY CELLS

Cells flanking the guard cells of the stomatal complex. They contribute to the regulation of stomatal opening and closing by exchanging ions with the adjacent guard cells. GUARD CELLS

plane somehow promotes the formation of a cortical PPB in the same plane, then this could largely explain the relationship between cell shape and division plane in highly vacuolated cells. Although phragmosomes are not observed in small, densely cytoplasmic cells, a functional counterpart of some kind might be involved in division plane selection in these cells. Migration of the nucleus into the division plane is generally not necessary, as the nucleus occupies most of the cell’s volume. However, microtubules undergo a rearrangement similar to that observed for phragmosomal strands (which contain both microtubules and actin filaments) in highly vacuolated cells. Before prophase in small cells, cytoplasmic microtubules radiate from the nuclear surface in all directions, and then coalesce into a plate-like arrangement as the cortical PPB forms in the same plane21. These nucleus-radiating microtubules might have a role in shapedependent division plane selection in densely cytoplasmic cells similar to that proposed for phragmosomal strands in highly vacuolated cells.

Although most cells select a division plane that can be predicted by their shapes, some do not, and asymmetrically dividing cells are a conspicuous exception to the general rule. In the example illustrated in FIG. 4, STOMATAL COMPLEXES in grasses are formed through a series of asymmetric divisions. An asymmetric transverse division forms a GUARD MOTHER CELL (GMC), then subsequent asymmetric divisions in the nearest neighbours of the GMC (SUBSIDIARY MOTHER CELLS or SMCs) produce SUBSIDIARY CELLS. Finally, the GMC itself divides longitudinally to produce a pair of GUARD CELLS. These asymmetric division planes are coordinated with polarity of the mother cell, which seems to override the influence of cell shape on selection of the division plane. As illustrated for SMCs (FIG. 4), polarization of the mother cell is indicated by migration of the premitotic nucleus to an asymmetric location, accumulation of a dense patch of cortical actin at that site, and formation of an asymmetric PPB22,23. Throughout mitosis, one end of the mitotic spindle remains associated with the actin patch; during cytokinesis, one of the

The pair of cells in the centre of a stomatal complex that flank the stomatal pore. Transvacuolar cytoplasmic strands containing actin filaments

PPB

Cortical actindepleted zone Early phragmoplast Prophase

Mitosis

Early cytokinesis

Cortical actindepleted zone

Cell plate vesicles Nucleus

Late phragmoplast

F-actin Microtubule Early cell plate Late cytokinesis

Daughter cells (interphase)

Figure 3 | Cytoskeletal organization in dividing plant cells. Drawings for prophase and interphase cells represent projections of a three-dimensional view showing both the cell surface and internal features. Drawings for cells in mitosis and cytokinesis represent mid-plane, cross-sectional views showing only the outlines of the cell cortex. At all stages of the cell cycle, cytoplasmic strands containing actin filaments span the vacuole, linking the cytoplasm surrounding the nucleus with the cortical cytoplasm. During prophase, a cortical preprophase band (PPB) of microtubules circumscribes the future plane of cell division. Actin filaments are distributed throughout the cell cortex during prophase, but some are clearly aligned together with the microtubules of the PPB. When the microtubule PPB is disassembled on entry into mitosis, the actin component of the PPB also disappears, leaving behind an actin depleted zone in the cell cortex that persists and marks the division site throughout mitosis and cytokinesis. After completion of mitosis, a phragmoplast of microtubules and actin filaments is initiated between daughter nuclei, which guides the movement of Golgi-derived vesicles containing cell wall materials to the cell plate. As cytokinesis proceeds, the phragmoplast expands centrifugally until it fuses with the parental plasma membrane and cell wall at the cortical division site previously occupied by the PPB.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 3 5

REVIEWS

GMC

SMC

PPB Actin patch

Early phragmoplast

Spindle

Guard cells Late phragmoplast

Subsidiary cells Actin filaments Microtubules

Wild-type stomate Guard cell pair

Subsidiary pair

Nucleus

Figure 4 | Stomatal development in maize34. In wild-type plants, stomata are formed through a series of asymmetric divisions, beginning with the division that forms a guard mother cell (GMC) and followed by division of subsidiary mother cells (SMCs) to form subsidiary cells flanking the GMC. Subsequently, symmetric division of the GMC forms a pair of guard cells. (PPB, preprophase band.)

CYTOCHALASIN

A fungal compound that specifically interferes with actin polymerization. ZYGOTE

A single diploid cell formed by the fusion of haploid female and male gametes. GUANINE NUCLEOTIDE EXCHANGE FACTOR

(GEF). A protein that facilitates the exchange of GDP (guanine diphosphate) for GTP (guanine triphosphate) in the nucleotidebinding pocket of a GTPbinding protein. ADP-RIBOSYLATION FACTOR (ARF) G PROTEIN

A GTP-binding protein that regulates the ribosylation of ADP (adenosine diphosphate).

daughter nuclei remains in continuous contact with this patch. The prophase nucleus and spindle are resistant to displacement by centrifugation24,25, indicating that they are attached to the actin patch. Moreover, CYTOCHALASIN treatments have shown that this attachment depends on actin26. So it seems that the actin patch is an adhesion site for the dividing nucleus that retains it in the asymmetric division plane throughout the cell cycle. This adhesion site might be functionally related to a specialized cortical site identified in some asymmetrically dividing animal cells, which interacts with spindle microtubules in an actin-dependent manner to pull the spindle into the appropriate position before mitosis27,28. However, it is not known whether any of the same proteins are involved in these apparently similar processes in plant and animal cells. Molecular mechanisms governing the polarization of mother cells before asymmetric division are not well understood. However, several findings indicate an essential role for the cell wall. Similar to the SMC division described above, the first embryonic division in ZYGOTES of the brown algae Fucus is an asymmetric division that is coordinated with the axis of polarity. Fixation of this polar axis requires actin29 and an intact cell wall30, and depends on secretion31. Likewise, polarization of the Arabidopsis zygote before its asymmetric division requires the GNOM/EMB30 gene32, which encodes a membrane-associated GUANINE NUCLEOTIDE EXCHANGE FACTOR for ADP-RIBOSYLATION FACTOR G PROTEIN (an ARF GEF). GNOM is related to ARF GEFs involved in trafficking through the secretory pathway in yeast33. So these findings might reflect a requirement for asymmetric secretion of one or more molecules into the walls of polarized cells. Such molecules could then remain anchored at the site of deposition and interact, directly or indirectly, with the cortical actin cytoskeleton to maintain cell polarity in asymmetrically dividing cells. During the formation of stomatal complexes in maize, the Brick (Brk1) and Pangloss1 (Pan1) genes mediate polarization and establishment of the asymmetric division plane in SMCs. As shown in FIG. 5, SMCs often divide symmetrically after having completely failed to polarize in brk1 mutants, or having only partially polarized in pan1 mutants34. Therefore, these genes might encode new components of the molecular

machinery involved in establishing or maintaining polarity in asymmetrically dividing cells; cloning and molecular analysis of these genes is underway. Establishment of the cortical division site

As illustrated in FIG. 3, the site at which a new cell wall will become attached to the parental wall is predicted during prophase by the location of the PPB8,21. During cytokinesis, the cortical site previously occupied by the PPB seems to attract the expanding phragmoplast. For example, if dividing Tradescantia stamen hair cells are centrifuged during mitosis to displace the spindle, the ensuing phragmoplast can migrate back to the former PPB site to attach the new cell wall there as usual35,36. Similarly, in some plant cells, rotation of the spindle during mitosis results in the initiation of a phragmoplast that is out of alignment with the plane defined by the PPB, but compensatory movements of the phragmoplast during cytokinesis ensure that the new cell wall attaches at the former PPB site37,38. Interestingly, if prolonged centrifugation of Tradescantia stamen hair cells39 or maize SMCs40 is used to force the new cell wall to become attached somewhere other than the former PPB site, these mislocalized walls remain wrinkled (instead of flattening as they usually do after attaching to the parental walls). All of these observations point to the existence of a specialized cortical site left behind when the PPB is disassembled, which guides the expanding phragmoplast and allows proper maturation of the new cell wall when it becomes attached there. In TONNEAU (TON) mutants of Arabidopsis, PPBs do not form and irregular cell shapes might indicate that cell divisions are aberrantly orientated41. This analysis shows that PPBs are not essential for cytokinesis per se, but supports the conclusion that PPBs are important in the orientation of new cell walls. An attractive hypothesis is that the PPB guides the localized deposition of as-yet-unknown components of the division site into the cell cortex, plasma membrane or cell wall, which are retained at this site after the PPB is disassembled39. Guiding the phragmoplast to the division site

Treatment of dividing plant cells with cytochalasins has long been known to cause new cell walls to be misorienwww.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

ÂŠ 2001 Macmillan Magazines Ltd

REVIEWS

CELL CORTEX

A thin layer of cytoplasm immediately adjacent to the inner surface of the plasma membrane. LEAF PRIMORDIUM

A newly initiated organ that will eventually give rise to a mature leaf.

tated in both symmetrically and asymmetrically dividing plant cells. This indicates a requirement for actin in the spatial control of cytokinesis20,23,42. Because actin filaments are found at many locations in the cell, throughout the cell cycle (FIG. 3), there are various ways in which actin could contribute to the orientation of new cell walls. During mitosis and early cytokinesis, the dividing nucleus is suspended in the division plane by transvacuolar strands containing F-actin (FIG. 3). These strands can often be seen to contact the CELL CORTEX preferentially at, or near, the division site20,43. Analyses of the effects of cytochalasins on dividing cells indicate that the F-actin in these strands might tether the dividing nucleus in the division plane and helps to guide the phragmoplast towards the division site as cytokinesis proceeds44. Cortical F-actin might also be important for guidance of the phragmoplast to the division site. At the end of prophase, the F-actin component of the PPB is disassembled along with the microtubule component. But cortical F-actin is retained elsewhere, leaving an ‘actindepleted zone’ (ADZ) in the cell cortex at the former PPB site (FIG. 3). In many plant cell types, the ADZ persists and marks the division site throughout mitosis and cytokinesis45–47. The ADZ might be directly involved in positioning the phragmoplast with respect to the division site once it comes into contact with the cell cortex — for example, by somehow ‘repelling’ the edges of the

brk1– brk1–

GMC

pan1–

dcd1– or dcd2 – SMC

pan1–

dcd1– or dcd2 –

Figure 5 | Effects of four mutations that disrupt the normal orientation of SMC divisions in maize34. The Brk1 and Pan1 genes are required for normal polarization of subsidiary mother cells (SMCs) before division. Failure of premitotic SMCs to polarize in brk1 and pan1 mutants is indicated by the absence of an actin patch flanking the guard mother cell (GMC) (in brk1 mutants), failure of the nucleus to migrate to its normal asymmetric position adjacent to the GMC (in brk1 and pan1 mutants), or formation of a symmetric preprophase band instead of an asymmetric one (in pan1 mutants). The ensuing SMC divisions in brk1 and pan1 mutants are usually transverse or oblique. In discordia 1 (dcd1) or dcd2 mutants, SMCs polarize normally, but the Dcd genes are required later for guidance of the phragmoplast to the asymmetric division site. Many SMCs divide normally in each of the mutants illustrated, represented by the normal sequence of events illustrated for the left SMC in the brk1 mutant.

phragmoplast from actin-rich cortical regions. Alternatively, organization of the cortex into actin-rich and actin-depleted zones might be important for retention of other molecules (which directly interact with the phragmoplast) at the division site. Drug studies indicate that the acto-myosin system mediates dynamic interactions between the phragmoplast and the division site at both early and late stages of cytokinesis. For instance, treatment of Tradescantia stamen hair cells undergoing cytokinesis with butanedione monoxime (BDM; a specific inhibitor of myosin ATPase activity) causes cell plates to become misoriented at early stages of cytokinesis without altering the cytoskeletal organization of the phragmoplast. In addition, both BDM and a specific inhibitor of myosin II (ML-7, which inhibits myosin light-chain kinase activity) delay lateral expansion of the cell plate at later stages of cytokinesis. This causes phragmoplasts and associated cell plates to adopt various abnormal shapes48. Thus, acto-myosin-based interactions between the phragmoplast and cell cortex apparently contribute both to the proper orientation and to the lateral expansion of the cell plate. For further progress in understanding the function of myosins in cytokinesis, we need to know the localization and functions of individual plant myosins in dividing cells. Another avenue for exploring the spatial control of cytokinesis in plant cells is the identification and analysis of mutations that disrupt the orientation of cell division. In wild-type maize LEAF PRIMORDIA, most cells are rectangular and divide either transversely or longitudinally relative to the long axis of the mother cell (FIG. 6a). In leaf primordia of tangled1 (tan1) mutants, many cells divide transversely, but few divide longitudinally. Instead, most cells divide in abnormal orientations, forming crooked or curved new cell walls49. Analysis of the cytoskeleton in dividing cells of tan1 mutant leaves has shown that the microtubule and actin arrays associated with cell division are structurally normal, but are often misorientated. In particular, most of the phragmoplasts fail to be guided to the cortical site formerly occupied by a PPB38 (FIG. 6b). So the Tan1 gene is required for phragmoplast guidance and/or maintenance of the division site. Tan1 encodes a highly basic protein that binds directly to microtubules in vitro, and antibodies against TAN1 preferentially label microtubule structures in dividing cells (that is, the PPB, the spindle and the phragmoplast) (FIG. 6e–i)50. So the TAN1 protein associates with the cytoskeletal structures that are misorientated in tan1 mutant cells, although its function in orientating these structures is not known. An intriguing possibility is that TAN1 participates in an interaction between phragmoplast microtubules and the actin network that links the dividing nucleus to the cortical division site early in cytokinesis. TAN1 could also (or alternatively) be involved in a direct interaction between the phragmoplast and the cortical division site later in cytokinesis. The Tan1 gene is required for spatial control of cytokinesis during most — if not all — leaf cell divisions. By contrast, the Discordia genes (Dcd1 and

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 3 7

REVIEWS

PPB Spindle Phragmoplast Cell wall Wild type

Tangled

Figure 6 | Function and localization of TANGLED1 (TAN1) in dividing cells of maize leaf primordia31,41,42. a | In wild-type primordia, preprophase bands (PPBs) are orientated either transversely or longitudinally, predicting future planes of cell division. Many spindles are transverse or longitudinal, but most rotate to become obliquely orientated. The orientations of early phragmoplasts match those of the preceding spindles. By the end of the cell cycle, however, the oblique angles of most early phragmoplasts have been corrected so that the final division planes are transverse or longitudinal. b | In tan1 mutant primordia, longitudinal PPBs are rarely formed; virtually all PPBs are transverse or slightly oblique. As in wild type, spindle orientations are variable. Unlike wild type, however, most phragmoplasts in tan1 mutant cells fail to be guided to the former PPB site. c–i | Labelling of leaf primordium cells with a monoclonal antibody against TAN1 (c,d,f,h) and with anti-β-tubulin (e,g,i). c–e show whole, isolated cells from leaf primordium squashes, whereas f–i show 3-µm tissue sections (white brackets indicate cell outlines). c | An interphase cell labelled with anti-TAN1. A PPB is labelled with anti-TAN1 in d and with anti-β-tubulin in e. A spindle is labelled with TAN75 in f and with anti-β-tubulin in g. A phragmoplast is labelled with TAN75 in h and with anti-β-tubulin in i. The results show that anti-TAN1 labels cells at all stages of the cell cycle in a punctate manner, suggesting that TAN1 is associated with particles of some sort in the cell. In interphase cells, there is no evidence for an association with microtubules, which are restricted to the cell cortex. In dividing cells, however, anti-TAN1 preferentially labels microtubule-containing structures (PPBs, spindles, and phragmoplasts). Scale bar in c = 10 µm and applies to c–i. (Figure modified from REFS 15,50.)

Dcd2) are required only for the asymmetric divisions involved in the formation of stomatal complexes and other specialized cell types in the maize leaf epidermis51. As illustrated in FIG. 5, SMCs in dcd1 mutants become polarized and form asymmetrically localized PPBs, as in wild-type cells. During mitosis, spindles are localized normally, with one pole contacting the actin patch, and phragmoplasts are initiated in a normal location. However, SMCs and other asymmetrically dividing cells often divide aberrantly in dcd mutants because phragmoplasts are not guided to the asymmetric division site (FIG. 5). Mutations in the Dcd genes mimic the effects of the anti-actin drug cytochalasin D on the SMC division. Moreover, dcd mutants are hypersensitive to cytochalasin D: low concentrations of this drug, which have no observable effect on SMC divisions in wild-type leaves, increase the frequency of abnormal divisions in dcd1 leaves twofold to threefold51. These observations indicate that the Dcd genes are required for an actin-dependent process that guides the phragmoplast to the division site in asymmetrically dividing cells. Further understanding of the role of Dcd genes in the control of asymmetric cell divisions awaits the molecular analysis of these genes.

Future directions

Building on decades of earlier work that characterized the spatial control of cytokinesis in plant cells, various approaches are now being taken to advance our understanding of these processes in molecular terms. Identification of molecules involved has so far come mainly from drug studies, and more recently, isolation of mutants specifically disrupting the placement of new cell walls. The recent availability of complete Arabidopsis and rice genome sequences will greatly facilitate cloning of the corresponding genes, paving the way for molecular analysis of their products. Moreover, the availability of complete plant genome sequences now also permits the identification of entire families of genes encoding proteins implicated in the spatial control of cytokinesis, such as myosins, whose functions are not well understood. Well-established biochemical and cell biological approaches (such as microinjection into dividing Tradescantia stamen hair cells) can be taken to analyse the functions of these proteins. In addition, powerful tools for reverse genetics in maize and Arabidopsis, in which insertions disrupting the functions of genes of interest can now be routinely isolated, have opened new avenues for functional analysis of relevant proteins. Moreover, the ability to visualize the localization of proteins fused to www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS GFP in living cells undergoing cell division is another valuable new addition to the collection of tools for analysis of these proteins. So, the stage is now set for important advances to be made in our understanding of the spatial control of cytokinesis in plant cells.

2. 3.

9. 10.

11.

12.

13.

14.

15. 16.

17.

18.

19.

20.

Hofmeister, W. Zusatze und Berichtigungen zu den 1851 veröffentlichen Untersuchungengen der Entwicklung höherer Kryptogamen. Jahrbucher für Wissenschaft und Botanik 3, 259–293 (1863). Errera, L. Über Zellformen und Siefenblasen. Botanisches Centralblatt 34, 395–399 (1888). Lintilhac, P. M. & Vesecky, T. B. Stress-induced alignment of division plane in plant tissues grown in vitro. Nature 307, 363–364 (1984). Lynch, T. M. & Lintilhac, P. M. Mechanical signals in plant development: a new method for single cell studies. Dev. Biol. 181, 246–256 (1997). Reports a simple and fascinating experiment showing that compression of wall-less protoplasts causes them to divide primarily in a plane parallel to the compressive force. Sinnott, E. W. & Bloch, R. Cytoplasmic behavior during division of vacuolate plant cells. Proc. Natl Acad. Sci. USA 26, 223–227 (1940). Pickett-Heaps, J. D. & Northcote, D. H. Organization of microtubules and endoplasmic reticulum during mitosis and cytokinesis in wheat meristems. J. Cell Sci. 1, 109–120 (1966). Venverloo, C. J., Hovenkamp, P. H., Weeda, A. J. & Libbenga, K. R. Cell division in Nautilocalyx explants. I. Phragmosome, preprophase band and plane of division. Z. Pflanzenphysiol. 100, 161–174 (1980). Gunning, B. E. S. in The Cytoskeleton in Plant Growth and Development (ed. Lloyd, C. W.) 229–292 (Academic Press, London, 1982). A classic review — a goldmine of information related to cytokinesis and its spatial regulation in plant cells accumulated before 1982. Palevitz, B. A. Actin in the preprophase band of Allium cepa. J. Cell Biol. 104, 1515–1519 (1987). Traas, J. A. et al. An actin network is present in the cytoplasm throughout the cell cycle of carrot cells and associates with the nucleus. J. Cell Biol. 105, 387–395 (1987). Kakimoto, T. & Shibaoka, H. Actin filaments and microtubules in the preprophase band and phragmoplast of tobacco cells. Protoplasma 140, 151–156 (1987). Staehelin, L. A. & Hepler, P. K. Cytokinesis in higher plants. Cell 84, 821–824 (1996). An excellent, brief review of cytokinesis in plant cells. Otegui, M. & Staehelin, L. A. Cytokinesis in flowering plants: More than one way to divide a cell. Curr. Opin. Plant Biol. 3, 493–502 (2000). A thorough and current review of the important advances that have been made in recent years in understanding how new cell walls are formed during cytokinesis. Heese, M., Ulrike, M. & Jürgens, G. Cytokinesis in flowering plants: Cellular processes and developmental integration. Curr. Opin. Plant Biol. 1, 486–491 (1998). Smith, L. G. Divide and conquer: Cytokinesis in plant cells. Curr. Opin. Plant Biol. 2, 447–453 (1999). Sylvester, A. W. Division decisions and the spatial regulation of cytokinesis. Curr. Opin. Plant Biol. 3, 58–66 (2000). Goodbody, K. C., Venverloo, C. J. & Lloyd, C. W. Laser microsurgery demonstrates that cytoplasmic strands anchoring the nucleus across the vacuole of pre-mitotic plant cells are under tension. Implications for division plane alignment. Development 113, 931–939 (1991). Flanders, D. J., Rawlins, D. J., Shaw, P. J. & Lloyd, C. W. Nucleus-associated microtubules help determine the division plane of plant epidermal cells: avoidance of fourway junctions and the role of cell geometry. J. Cell Biol. 110, 1111–1122 (1990). Lloyd, C. W. How does the cytoskeleton read the laws of geometry in aligning the division plane of plant cells? Development (Suppl.) 1, 55–65 (1991). An interesting review summarizing a series of studies published by Clive Lloyd and his colleages during the period 1988–1991, leading to a new model for how cell shape and mechanical forces influence the choice of division plane. Lloyd, C. W. & Traas, J. A. The role of F-actin in

Links DATABASE LINKS GNOM ENCYCLOPEDIA OF LIFE SCIENCES Plant cell wall

biosynthesis | Plant cell walls

determining the division plane of carrot suspension cells. Drug studies. Development 102, 211–221 (1988). 21. Wick, S. M. in The Cytoskeletal Basis of Plant Growth and Form (ed. Lloyd, C. W.) 231–244 (Academic Press, London, 1991). A review summarizing a wealth of information regarding preprophase bands and the spatial regulation of cytokinesis. 22. Cho, S. -O. & Wick, S. M. Microtubule orientation during stomatal differentiation in grasses. J. Cell Sci. 92, 581–594 (1989). 23. Cho, S.-O. & Wick, S. M. Distribution and function of actin in the developing stomatal complex of winter rye (Secale cereale cv. Puma). Protoplasma 157, 154–164 (1990). 24. Pickett-Heaps, J. D. Proprophase microtubules and stomatal differentiation: some effects of centrifugation on symmetrical and asymmetrical cell division. J. Ultrastruct. Res. 27, 24–44 (1969). 25. Pickett-Heaps, J. D., Gunning, B. E. S., Brown, R. C., Lemmon, B. E. & Cleary, A. L. The cytoplast concept in dividing plant cells: cytoplasmic domains and the evolution of spatially organized cell division. Am. J. Bot. 86, 153–172 (1999). 26. Kennard, J. L. & Cleary, A. L. Pre-mitotic nuclear migration in subsidiary mother cells of Tradescantia occurs in G1 of the cell cycle and requires F-actin. Cell Motil. Cytoskel. 36, 55–67 (1997). 27. Hyman, A. A. Centrosome movement in the early divisions of Caenorhabditis elegans: a cortical site determining centrosome position. J. Cell Biol. 109, 1185–1193 (1989). 28. Knoblich, J. A. Asymmetric cell division during animal development. Nature Rev. Mol. Cell Biol. 2, 11–20 (2001). 29. Quatrano, R. S. Separation of processes associated with differentiation of two-celled Fucus embryos. Dev. Biol. 30, 209–312 (1973). 30. Kropf, D. L., Kloareg, B. & Quatrano, R. S. Cell wall is required for fixation of the embryonic axis in Fucus zygotes. Science 239, 187–190 (1988). 31. Shaw, S. L. & Quatrano, R. S. The role of targeted secretion in the establishment of cell polarity and the orientation of the division plane in Fucus zygotes. Development 122, 2623–2630 (1996). 32. Vroemen, C. W., Langeveld, S., Mayer, U., Ripper, G. & Jürgens, G. Pattern formation in the Arabidopsis embryo revealed by position-specific lipid transfer protein gene expression. Plant Cell 8, 783–791 (1996). 33. Steinmann, T. et al. Coordinated polar localization of auxin efflux carrier PIN1 by GNOM ARF GEF. Science 286, 316–318 (1999). 34. Gallagher, K. & Smith, L. G. Roles for polarity and nuclear determinants in specifying daughter cell fates following an asymmetric division in the maize leaf. Curr. Biol. 10, 1229–1232 (2000). 35. Ota, T. The role of cytoplasm in cytokinesis of plant cells. Cytologia 26, 428–447 (1961). A classic study, the first to show the ability of a displaced phragmoplast to find the previously established division site. 36. Gunning, B. E. S. & Wick, S. M. Preprophase bands, phragmoplasts, and spatial control of cytokinesis. J. Cell Sci. 2, S157–S179 (1985). 37. Palevitz, B. A. Division plane determination in guard mother cells of Allium: Video time-lapse analysis of nuclear movements and phragmoplast rotation in the cortex. Dev. Biol. 117, 644–654 (1986). An early demonstration of phragmoplast reorientation during cytokinesis ensuring the attachment of a new cell wall at the former PPB site. 38. Cleary, A. L. & Smith, L. G. The tangled1 gene is required for spatial control of cytoskeletal arrays associated with cell division during maize leaf development. Plant Cell 10, 1875–1888 (1998). Analysis of the maize tangled1 mutant at the level of the cytoskeleton showing the failure of most phragmoplasts to position the new cell wall at the former PPB site.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

39. Mineyuki, Y. & Gunning, B. E. S. A role for preprophase bands of microtubules in maturation of new cell walls, and a general proposal on the function of preprophase band sites in cell division in higher plants. J. Cell Sci. 97, 527–537 (1990). A classic study showing that when new cell walls are forced to attach somewhere other than the former PPB site, they fail to mature normally. This leads to the proposal that the PPB directs the deposition of cell wall maturation factors at the division site during prophase, which are transferred to the new wall if it attaches at this site. 40. Galatis, B., Apostolakos, P. & Katsaros, C. Experimental studies on the function of the cortical cytoplasmic zone of the preprophase microtubule band. Protoplasma 122, 11–26 (1984). 41. Traas, J. A. et al. Normal differentiation patterns in plants lacking microtubular preprophase bands. Nature 375, 676–677 (1995). Analysis of the TONNEAU/FASS mutations of Arabidopsis showing that cells divide (in a disorganized manner) without forming PPBs. 42. Palevitz, B. A. & Hepler, P. K. The control of the plane of division during stomatal differentiation in Allium. II. Drug studies. Chromosoma 46, 327–341 (1974). 43. Valster, A. H. & Hepler, P. K. Caffeine inhibition of cytokinesis: effect on the phragmoplast cytoskeleton in living Tradescantia stamen hair cells. Protoplasma 196, 155–166 (1997). 44. Wick, S. M. Spatial aspects of cytokinesis in plant cells. Curr. Opin. Cell Biol. 3, 253–260 (1991). 45. Cleary, A. L., Gunning, B. E. S., Wasteneys, G. O. & Hepler, P. K. Microtubule and F-actin dynamics at the division site in living Tradescantia stamen hair cells. J. Cell Sci. 103, 977–988 (1992). The first description of an actin-depleted zone in the cell cortex of dividing plant cells that marks the former PPB site throughout mitosis and cytokinesis. 46. Liu, B. & Palevitz, B. A. Organization of cortical microfilaments in dividing root cells. Cell Motil. Cytoskeleton 23, 252–264 (1992). 47. Cleary, A. L. F-actin redistributions at the division site in living Tradescantia stomatal complexes as revealed by microinjection of rhodamine-phalloidin. Protoplasma 185, 152–165 (1995). 48. Molchan, T. M., Valster, A. H., Vos, J. W. & Hepler, P. K. Actomyosin promotes cell plate alignment and late lateral expansion in plant cells. Mol. Biol. Cell 10 (Suppl.):15a (1999). 49. Smith, L. G., Hake, S. C. & Sylvester, A. W. The tangled1 mutation alters cell division orientations throughout maize leaf development without altering leaf shape. Development 122, 481–489 (1996). 50. Smith, L. G., Gerttula, S., Han, S. & Levy, J. TANGLED1: A microtubule binding protein required for spatial control of cytokinesis in maize. J. Cell Biol.(in the press). Molecular analysis of the Tangled1 gene and protein suggesting that TAN1 protein participates in the orientation of cytoskeletal structures in dividing cells through an association with microtubules 51. Gallagher, K. & Smith, L. G. discordia mutations specifically misorient asymmetric cell divisions during development of the maize leaf epidermis. Development 126, 4623–4633 (1999). Analysis of two mutants that disrupt the spatial regulation of asymmetric divisions in the maize leaf epidermis by interfering with phragmoplast guidance. 52. Carpita, N. C. & Gibeaut, D. M. Structural models of primary cell walls in flowering plants: consistency of molecular structure with the physical properties of the walls during growth. Plant J. 3, 1–30 (1993). 53. Cosgrove, D. J. Assembly and enlargement of the primary cell wall in plants. Annu. Rev. Cell Dev. Biol. 13, 171–201 (1997). 54. Alberts, B. et al. (eds) Molecular Biology of the Cell 3rd edn 1002 (Garland, New York, 1994).

VOLUME 2 | JANUARY 2001 | 3 9

REVIEWS

MICROTUBULE-ASSOCIATED PROTEINS IN PLANTS — WHY WE NEED A MAP Clive Lloyd* and Patrick Hussey‡ Plants have four main microtubule assemblies. Three are involved in arranging when and where the cell wall is laid down and have no direct homologues in animals. Microtubuleassociated proteins are important components of these assemblies, and we are now starting to uncover what these proteins are and how they might work. CELL DIVISION CELLULOSE

A tough, inelastic fibre wrapped in layers (lamellae) within the plant cell wall. Composed of β-1,4-linked glucosyl residues. MERISTEM

A zone (for example, the very apex of the shoot) containing undifferentiated cells that continue to divide, providing cells for further growth and differentiation. MICROTUBULE-ASSOCIATED PROTEIN

A protein that, in the loosest sense, binds to microtubules. More stringently, a protein that co-purifies with microtubules in vitro.

*Department of Cell Biology, John Innes Centre, Colney, Norwich NR4 7UH, UK. ‡Department of Biological Sciences, University of Durham, South Road, Durham DH1 3LE, UK. e-mail: clive.lloyd@bbsrc.ac.uk

There are many differences between plant and animal cells. Both types of cell are surrounded by an extracellular matrix, but whereas some motile animal cells can break free of the matrix and migrate, the immobilized plant cell is boxed in by a complex wall containing layers of tough CELLULOSE microfibrils. Despite its toughness, the wall can be loosened to allow plant cells to expand — often to many times their original size — as shoots extend up towards the sunlight and roots grow down into the ground. It is this capacity for constant extension and consolidation of the growth axis that distinguishes plant growth. Unlike animal morphogenesis, which is once and for all, plant morphogenesis continues beyond embryogenesis; it is a reiterative process in which new developmental modules (for example, leaves, floral organs or roots) are produced by localized cell division within MERISTEMS. In the meristem, the three-dimensional geometry of the division products is determined by orientation of the cross-walls between sister cells. However, outside the meristem, once division has stopped, the direction in which the cell walls yield to turgor pressure is assumed to be a key factor in determining tissue morphology. In both dividing and non-dividing cells, microtubules are involved in these architectural processes. Microtubules are hollow ‘scaffolding rods’ with a fastgrowing end and a slow-growing end, reflecting the way the αβ-tubulin heterodimer is polymerized into the tubule. This intrinsic polarity of microtubules can

be read by the motor proteins that transport materials to one end or the other. Motor proteins and other MICROTUBULE-ASSOCIATED PROTEINS (MAPs) increase the versatility of the individual microtubules, transforming unitary tubules into architecturally complex assemblies. These assemblies are intimately involved in the processes of plant cellular morphogenesis, and it is against this background of cell division and expansion that we should see the plant microtubule cycle and the role of plant MAPs. What is a MAP?

Warmed to body heat in the absence of calcium, tubulin subunits can be induced to self-assemble into microtubules, which can be centrifuged to form a microtubule pellet. The pellet can then be disassembled on ice and, after re-centrifuging to clear the solution, the tubulin can be put through further assembly/disassembly cycles. In this way, proteins that specifically associate with the assembled microtubules will be enriched in the pellet. One classical definition of a MAP is therefore a protein that is enriched by rounds of assembly/disassembly. This definition arose from work on animal brains, which contain a high proportion of tubulin because the long interconnecting neuronal cells are supported by microtubules. But non-neuronal cells have a smaller proportion of tubulin, and in plants it is difficult to concentrate this protein against a background of hydrolases and cross-linking phenolics that fill the plant vacuole. www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS the spindle remnant — the mid-body — during cytokinesis). In higher plants, the situation is more complicated; here the microtubule cycle has four discrete stages, described below (also see FIG. 1), some of which have no direct parallels in animal cells4. The challenge is to find out how MAPs help make these plant-specific arrays.

Figure 1 | The plant microtubule cycle. a | The interphase cortical array. Transverse cortical microtubules (green) in an elongated tobacco BY-2 suspension cell are shown. Chromatin is stained blue. b | The preprophase band. Formed in G2, this gradually narrows to form a ring that predicts the division plane. c | The plant spindle at anaphase. d | The phragmoplast. Note the dark midline, where the two circlets of microtubules meet end on. e | The metaphase plant spindle. Compare this with f | the metaphase animal spindle. Note that the plant poles tend to be broader, there are no obvious spindle pole bodies and there are no astral microtubules. (Scale bar = 7.5 µm.)

In plants, the affinity purification of MAPs on Taxolstabilized neuronal microtubules has been used as an alternative strategy. However, this bypasses the classical test of enrichment through rounds of tubulin assembly/disassembly. Under these circumstances, it is important to show that the putative microtubule-interacting protein does bind to microtubules in the cell. By using this alternative MAP criterion, several proteins known to have alternative functions have been identified as plant MAPs. For example, Richard Cyr and colleagues have established the biochemical effects of the protein-translation factor EF-1 (on tubulin and microtubules), and they have shown that this protein is located on microtubules in cells1. Initiation factor-(iso)4F, another part of the protein-translation machinery, also binds to tubulin and bundles microtubules in vitro, as well as annealing microtubules end to end2. And heat shock protein 90 — a molecular chaperone — also binds tubulin in vitro and decorates all microtubule arrays (see below) except the spindle3. These studies illustrate the limitation of applying rigid biochemical definitions to complex cellular processes. As integrators of cellular space, microtubules probably organize a range of metabolic activities (for example, providing a matrix for metabolic activities, localization of organelles and transport of cargo) and so there could be many physiologically relevant microtubule-interacting proteins. However, in this review we concentrate on crossbridging structural MAPs and microtubule motor proteins (TABLE 1). CYTOPLASMIC ARRAY

The microtubules that spread through the cytoplasm of nondividing cells.

When and where do MAPs act?

In animal cells, microtubules form a CYTOPLASMIC ARRAY during interphase or the spindle during mitosis (and

The interphase cortical array. This is the first plant microtubule structure with no obvious similarities in animal cells. Here, we concentrate on the array in uniformly expanding tissue cells, as microtubules are arranged differently in tip-growing cells, such as root hairs and pollen tubes. In animal cells, the slow-growing minus ends of microtubules are usually anchored at discrete microtubule-organizing centres (MTOCs) where they are nucleated and from which they radiate. Higher land plants do not have discrete MTOCs; γtubulin (the seed for microtubule growth) is dispersed around the cell5, so it is difficult to see where cortical microtubules originate. After division, microtubules radiate from the nucleus towards the cortex. In most cells, this seems to be a transient stage and it is possible that disorganized microtubule nucleating material may be active at the nuclear surface for only a brief period6. As a consequence, later in interphase, the cortical microtubules in most cells are not reported to maintain contact with the nucleus. Such microtubules have both ends free, they wind around the inner face of the plasma membrane in staggered relay, and they are maintained in an approximately parallel alignment by inter-microtubule bridges. The cortical microtubules are also attached to the membrane. The organization of the cortical array may therefore depend more on the interconnecting MAPs than on so-called ‘organizing’ centres. The transverse cortical microtubules in an elongated tobacco BY-2 suspension cell are shown in FIG. 1a. Microtubules at the cytoplasmic face of the plasma membrane mirror the alignment of cellulose microfibrils on the external face of the plasma membrane. Somehow — and this is one of the mysteries of plant cell biology — cellulose-synthesizing enzymes embedded in the plasma membrane are thought to spin out microfibrils as they move across the face of the membrane, tracking the guidelines provided by the microtubules. Tough cellulose microfibrils wrapped transversely around the cell should therefore prevent the cell from bulging sideways, channelling the swelling force of turgor into elongation instead. However, as reviewed in REF. 7, there are important exceptions to this ‘rule’ and the exact relationship between microtubules, cellulose microfibrils and growth polarity is still not universally agreed. What is clear is that microtubules are not fixed elements and that they can reorientate between transverse and longitudinal configurations as the direction of growth changes in response to intrinsic factors such as the hormone gibberellic acid8 and environmental factors such as gravity9. In the cortical microtubule array, we shall describe one set of structural MAPs that crossbridges microtubules and thereby helps organize the cortical array, although there is likely to be another type of linkage that

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 4 1

REVIEWS (compare the metaphase plant spindle in FIG. 1e with the animal spindle in FIG. 1f). As in animal mitosis, microtubule motors are likely to be involved in organizing the spindle and aiding the separation of chromatids but, because of the different polar structure, the details of MAP function may be different.

PREPROPHASE BAND

A cortical ring of microtubules that usually predicts the site where the outgrowing cell plate will fuse with the mother cell wall. CELL PLATE

The flattened disk of immature cell wall deposited by fusion of Golgi vesicles in the plane where the two half sets of phragmoplast microtubules overlap. It contains a polysaccharide, callose, composed of chains of β-1,3linked glucosyl residues.

The phragmoplast. This, the fourth structure, usually appears during late anaphase, when the sister chromatids have separated to opposite ends of the mitotic apparatus. A bundle of microtubules appears within the central spindle but it grows outwards to become a double ring of microtubules whose fast-growing plus ends interdigitate at the midline — the line of overlap where the new cross wall is deposited11. In FIG. 1d, the ring-like PHRAGMOPLAST, which has been flattened during processing, shows in places the dark midline where the two circlets of microtubules meet end on. Enclosed within the phragmoplastic ring is the new cross wall (unstained), which will eventually grow outwards to seal the sister nuclei (stained blue) into separate cells. The direction in which the phragmoplast/cell plate grows obviously determines the spatial relationship between the two daughter cells and this morphogenetic aspect is described in the accompanying review by Laurie Smith on page 33 of this issue.

PHRAGMOPLAST

The cytokinetic apparatus of plants, composed of two rings of interdigitating microtubules and actin filaments. These guide Golgi vesicles to the line of overlap where they fuse to form the new cell wall. This structure evolves outwards, in contrast to the in-pulling of the animal contractile ring.

The identification of plant MAPs Figure 2 | Microtubule crossbridges. Carrot MAP65 forms 25–30 nm crossbridges between neuronal microtubules. MAP65, biochemically isolated from carrot cells, bundles brain microtubules in vitro18. The filamentous MAPs reproduce the inter-microtubule spacing observed in the cortex of plant cells19. In maintaining the local parallelism of the cortical microtubules these MAPs are also likely to contribute to the overall collinear alignment of the array (see FIG. 1a). (Scale bar = 50 nm.)

maintains contact with the plasma membrane10. Both classes of MAP would seem to be sufficiently plastic to allow reorientation of the array. The preprophase band. This is another structure for which there are no parallels in animal cells. The evenly distributed microtubules of G0/G1 are replaced in G2 by a cortical band — the PREPROPHASE BAND (PPB) — that eventually narrows to a ring (FIG. 1b). Although the PPB depolymerizes by metaphase, its memory seems to linger on because later, during cytokinesis, the new cross wall (the CELL PLATE) will join the mother cell wall at the former PPB site. MAPs could well be involved in the accumulation of the microtubules into a band. The plant spindle. As in animal cells, a spindle of microtubules engages and moves chromosomes at metaphase and anaphase. In prophase, the nuclear envelope is surrounded by a cage of microtubules — the so-called ‘prophase spindle’ — which is reorganized as the metaphase spindle forms. The anaphase spindle in FIG. 1c has a similar overall organization to the animal spindle, except the poles tend to be broader, there are no obvious spindle pole bodies, and there are no astral microtubules

Owing to the difficulty of concentrating the low proportion of plant tubulin to the level that supports selfassembly and the co-purification of MAPs, most investigators have used the microtubule-stabilizing agent Taxol, either to stimulate polymerization of endogenous tubulin or to stabilize brain microtubules that can be used as an affinity matrix. In this way, various presumptive MAPs have been isolated from carrot12 and maize cells13 and shown to induce microtubule bundling. Brain MAPs (high-molecular-weight MAP2 and low-molecular-weight tau) also cause the bundling of microtubules, and the expectation that one of the plant proteins could be a homologue of the brain MAPs was raised by the demonstration of crossreactivity with an anti-tau antibody13. Another potential parallel between plant and brain MAPs was provided by the crossreactivity of a 90-kDa microtubule-bundling protein from the freshwater alga Dichotomosiphon tuberosus with antibodies against vertebrate MAP4 — the widely occurring relative of brain MAP2 and tau14. Furthermore, a heat-stable fraction of tobacco BY-2 cells contains a 70-kDa protein that crossreacts with antibodies against MAP4 (REF. 15). (Neuronal MAPs are also heat stable; indeed, this provides a basis for their rapid purification.) Any or all of these proteins may, when sequenced, turn out to be plant relatives of the animal models. However, it now seems that the most extensively studied plant MAPs — the 65-kDa MAPs, a group of up to four gel bands of similar molecular mass — show little or no homology to animal MAPs. The 65-kDa MAPs were first isolated by Jiang and Sonobe16 from tobacco BY-2 suspension cells, and found to decorate all four plant microtubule arrays and to bundle microtubules in vitro. www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS

Figure 3 | MAPs in the midline of the phragmoplast. a | A phragmoplast in a carrot suspension cell stained with anti-tubulin antibodies (green); the separated chromosomal material is stained blue. Note the diminished staining in the midline where the two opposed sets of microtubules meet. b | The cell plate forms where the two half-sets of phragmoplast microtubules overlap. This region of a carrot cell is stained green by antibodies against the BimC kinesin, DcKRP120-2 (REF. 33). c | A dividing tobacco BY-2 cell in which antibodies against NtMAP65-1 (green) decorate the midline20. The midline remains unstained by antitubulin antibodies (red). (Scale bar = 10 µm.)

Chan et al.17 then used detergent-extracted cytoskeletons from carrots to isolate 65-kDa MAPs that were antigenically related to the tobacco proteins. Biochemically purified 65-kDa MAPs induce the bundling of neuronal microtubules18 and form regular filamentous crossbridges (FIG. 2) that reproduce in vitro the 25–30-nm inter-microtubule spacing observed when rapidly frozen plant cells are examined under the electron microscope19. Because anti-MAP65 antibodies decorate the cortical microtubules, it seems that these are the elusive crosslinks responsible for the parallel arrangement of microtubules in the cortical array.

KINESIN

A dimeric protein that uses the energy of ATP hydrolysis to process along a microtubule. It has a microtubule-binding domain at its amino-terminal head and moves cargo to the fast-growing plus end of the microtubule. It is unrelated to another, larger motor, cytoplasmic dynein, which moves to the minus ends (see also KRP). KINESIN-RELATED PROTEIN

The kinesin superfamily contains a range of motor proteins that share similarities to classical ‘vesicle’ kinesin in the conserved microtubulebinding motor region. The motor can be at the amino terminus, in the middle, or at the carboxyl terminus (in which case the protein usually migrates to the minus end of the microtubule). KRPs usually function in cell division.

Cloning a plant structural MAP. Although the carrot 65-kDa MAPs show the same regular spacing along the microtubules as brain MAPs18, the two sets of proteins are thought to be unrelated. Smertenko et al.20 used antibodies against the tobacco 65-kDa MAPs to isolate a complementary DNA from a Nicotiana tabacum expression library. Each of the animal MAPs 2, 4 and tau have three or four tandem repeats that contain the ‘PGGG’ motif. These repeats are believed to be involved in binding to microtubules. However, the proteins

encoded by the tobacco gene family (NtMAP65-1) contain no such repeats and would therefore seem to represent a novel class of MAP. By immunofluorescence, antibodies against the bacterially expressed tobacco protein differentially stain microtubules throughout the cell cycle. In interphase, subsets of microtubules in the cortical array are stained. At late anaphase, the most interesting staining is the midzone of the central spindle. It is in this region that the phragmoplast forms, with the staining ending up in the midline of the cytokinetic structure (FIGS 3, 4). Normally the midline does not stain with anti-tubulin antibodies (FIG. 3a); instead it forms a black line down the middle of the fluorescent structure, usually attributed to exclusion of antibodies from a congested cytoplasmic site. This line is where plus-end-directed microtubule motor proteins would be expected to accumulate. However, NtMAP65-1 (FIG. 3c) contains no known motor domains20, and it may either interact with a protein targeted to that locus or have its own targeting sequence. In their study, Smertenko et al.20 showed that although the 65-kDa tobacco protein, produced by overexpression in bacteria, bound to microtubules and stimulated tubulin polymerization, it did not cause bundling of microtubules. Recent studies (A. Smertenko and P.H., unpublished observations) indicate that higher concentrations of expressed protein do cause bundling, although further work is required to see whether the protein forms 25–30-nm crossbridges. Kinesins in the phragmoplast

The motor protein KINESIN was first isolated from nervous tissue by a method based on its ATP-dependent movement along microtubules21. KINESIN-RELATED PROTEINS (KRPs) have now been found in many eukaryotes, and certain subclasses of this superfamily are involved in mitosis22. All kinesins share principles of construction, containing a globular ATP- and microtubule-binding motor domain (head) with common motifs, and a variable tail region that may bind various cargoes. Generally

Table 1 | Plant structural and motor MAPs that have been sequenced and immunolocalized MAP

Type

Species

Location

MAP65, NtMAP65-1 (REFS 16–18, 20)

Structural

Tobacco, carrot

IMTA, PPB, spindle, phragmoplast

TKRP125 (REFS 23, 24, 29)

KinN

Tobacco, carrot

IMTA (in G2), PPB, spindle, phragmoplast

AtAKRP1 (REF. 28)

KinN

Arabidopsis, rice

Anaphase to phragmoplast

DcKRP120-2 (REF. 29)

KinN

Carrot

IMTA, and PPB (weakly), spindle, phragmoplast

KatA (REFS 35,36)

KinC

Arabidopsis, tobacco

Spindle, phragmoplast

KatB/C (REF. 34)

KinC

Arabidopsis

Spindle? phragmoplast?

KCBP (REFS 38-40,42, 62)

KinC

Arabidopsis tobacco, potato, Haemanthus

PPB, spindle, phragmoplast

KinN, kinesin-related protein with the motor domain in the N-terminus; KinC, kinesin-related protein with motor domain in the C-terminus; IMTA, interphase microtubule array; PPB, preprophase band.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 4 3

REVIEWS

Box 1 | Proteins that accumulate in the centre of the phragmoplast • Several proteins may be involved in targeting precursors to the cell plate in the correct spatio-temporal manner. • In addition to the structural MAP NtMAP65-1 (REF. 20), two plus-end-directed motors have been found located where the microtubules overlap, AtPAKRP1 (REF. 28) and DcKRP120-2 (REF. 29). • KNOLLE is an Arabidopsis syntaxin-related protein involved in the selective fusion of membrane vesicles during exocytosis55. It is expressed only in dividing cells and concentrates in the midline of the phragmoplast. Golgi vesicles can be routed to diverse membrane compartments, but the KNOLLE syntaxin seems to target vesicles specifically to the cell plate rather than to the mature mother cell wall, which has a different composition. • Phragmoplastin also concentrates in the cell plate56. This plant protein shows homology to animal dynamin — a GTPase involved in pinching off vesicles from the plasma membrane during receptor-mediated endocytosis. Overexpression causes cell plates to become skewed. Interfering with microtubule stability prevents the redistribution of this protein, suggesting cytoskeletal involvement. • Centrin is a small, Ca2+-binding protein associated with microtubule-organizing centres57. It concentrates at the forming cell plate and may be involved in the Ca2+-mediated fusion of vesicles. • In alfalfa, a mitogen-activated protein kinase (MMK3), present throughout the cell cycle, becomes active during cytokinesis in a microtubule-dependent manner58. It, too, concentrates at the cell plate. The function of this signalling kinase is not known, but is suggested either to regulate the activity of plus-end motors or the fusion of vesicles at the cell plate. • Actin filaments are also found among the phragmoplast microtubules59. They share the same polarity as the phragmoplast microtubules (faster-growing barbed ends to the midline), providing another directional cue for the guidance of material to the cell plate60. • KORRIGAN is an Arabidopsis endo-1,4-β-glucanase, essential for cytokinesis. It concentrates in the cell plate, where it is thought to be involved in cell-wall biochemistry61.

speaking, kinesins with an amino-terminal motor walk towards the plus end of the microtubule, whereas those with a carboxy-terminal motor travel to the minus end. As yet, it is difficult to assign functions to many of the plant MAPs, but most of the KRPs described below have been shown to be involved in formation of the phragmoplast (BOX 1). Although its construction initially mirrors that of the spindle (found in both plants and animals), the phragmoplast is formed differently and has the morphogenetically important role of constructing the new cross wall.

KINETOCHORE

A structure that connects each chromatid to the spindle microtubules, which shorten as pairs of chromatids are separated to opposite poles.

Plus-end-directed KRPs. Ten years ago, Asada et al.23 synchronized tobacco BY-2 cells to obtain a high proportion in cytokinesis, then they permeabilized the cells with glycerol and added fluorescent tubulin. The fluorescence was incorporated into the line of overlap between the growing ends of the interdigitating half sets of phragmoplast microtubules (FIG. 4c). If a pulse of unlabelled tubulin was added, the fluorescent line split into two and migrated away from the midline, towards the minus ends of the microtubules. This result indicated that soluble tubulin is added at overlapping, fastgrowing microtubule plus ends and subtracted from the minus end in a process known as ‘treadmilling’, such that fluorescently marked tubulin moves away from the site of addition at the midline (but see below). Next,

Asada and Shibaoka24 isolated two proteins with masses of 120 and 125 kDa from the tobacco phragmoplasts. This protein fraction was shown to contain plus-enddirected microtubule motor activity. One of the proteins, designated TKRP125, is a BimC-type KRP. BimC (‘blocked in mitosis’) mutants of the fungus Aspergillus nidulans fail to separate the spindle poles25. Members of the BimC subfamily are thought to form bipolar homotetramers (FIG. 4d); they crossbridge antiparallel microtubules and, by walking towards the plus ends, cause the microtubules to slide apart, forming two equal but oppositely directed sets26. In the elegant permeabilized-cell system mentioned above27, antibodies against TKRP125 inhibit the splitting of the fluorescently labelled midline. This suggests that there are two microtubule-based functions at work: first, the treadmilling of subunits from the equatorial region to the distal minus ends; and second, the sliding apart of the microtubule half-sets. The sliding filament action of BimC motors involved in the separation of anti-parallel microtubules, first described in fungal and animal mitosis, would therefore seem to be retained for cytokinetic duty in plants. Recently, Lee and Liu28 detected a non-BimC KRP in the midline of the phragmoplast. They showed that AtPAKRP1 (Arabidopsis thaliana phragmoplast-associated KRP-1) localizes to the spindle from anaphase onwards, decorating the interzonal microtubules that pass between the poles, but not to the KINETOCHORE microtubules that attach to the chromosomes. Then, in cytokinesis, the protein begins to concentrate at the midline. Other studies have identified a 120-kDa BimC-type KRP in carrot cells, in addition to a TKRP125 homologue29. As well as staining the mitotic spindle, antibodies against this novel BimC kinesin, DcKRP120-2, stain the phragmoplast and, in late cytokinesis, heavily decorate the midline in some cells (FIG. 3b). Despite its resemblance in initial stages to the mitotic apparatus, the phragmoplast can, under certain circumstances, form long after mitosis when it is unlikely to be simply a reincarnated spindle30. The different ways of forming a phragmoplast have been recently reviewed by Otegui and Staehelin31. If the phragmoplast is formed by microtubules radiating from the surfaces of the daughter nuclei, with plus ends distal (FIG. 4a), then crossbridging BimC motors could have several functions in cytokinesis. First, by recognizing plus ends, they could lock the two opposed sets of microtubules together (perhaps in conjunction with structural MAPs, such as NtMAP651). Next, by ‘reeling in’ the perinuclear minus ends of the microtubules they could position the nuclei (this would be especially important in multinucleate syncitial tissue where radiating microtubules seem to define the domain-like spacing between nuclei30). Then, by sliding the two sets of anti-parallel phragmoplast microtubules apart, BimC-type KRPs would maintain a minimal line of overlap (FIG. 4d), constantly editing the line along which the cell plate is to be constructed. If the microtubule half-sets were to interpenetrate completely, the line of overlap might be blurred and the fidelity of cytokinesis lost. www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS

+ + + ++++ ++ +

+ Chromatin

Bipolar kinesin

+ +

Cell plate

Cell wall

Figure 4 | Plant cytokinesis. a | Microtubules, possibly nucleated from the surface of the opposed chromatin masses at late anaphase, interlock at their plus ends to form the phragmoplast. b | The two anti-parallel sets of microtubules form a double-ringed structure. c | Tubulin subunits add on to the fast-growing plus ends and are subtracted from the opposite minus ends. Marked subunits therefore ‘treadmill’ through the polymer. d | BimC kinesinrelated proteins, either as homotetramers or some other complexed form, may crosslink antiparallel microtubules. By migrating towards the plus ends (small arrows), the KRPs would cause the microtubules to slide apart (large arrows). e | Golgi vesicles (small arrows) move with the aid of ‘vesicle kinesins’ to the midline of the phragmoplast. There they fuse, forming the central disk of cell plate material, which will eventually fuse with the mother wall (large arrow) at the cortical site predicted by the preprophase band. f | Consolidation of the Golgi material in the central disk drives the cell plate, and the ring of phragmoplast microtubules at its boundary, centrifugally outwards until it fuses with the mother wall.

Minus-end-directed KRPs. There is ample evidence that plus-end-directed motors function in plant cytokinesis, but what of the motor proteins at the minus ends? By using common motifs as primers for the polymerase chain reaction, the kat (for ‘kinesins from Arabidopsis thaliana’) genes were the first plant KRP genes to be identified32. The family contains five members (A to E), all of which encode 80–90-kDa proteins with minusend-directed motor activity33. The KatB and KatC proteins accumulate in mitotically synchronized tobacco BY-2 cells, indicating these forms may have a function in cell division34. In support of this conclusion, antibodies against KatA peptides stain spindles and phragmoplasts in Arabidopsis seedlings and tobacco BY-2

suspension cells35,36, and so three members of this family are active in division, although their precise roles await clarification. Although motor activity for the Kat proteins has not been reconstituted in vitro, Liu et al.36 did a cellular assay that indicated minus-end-directed activity. By lysing cells and adding ATP, they showed that KatA quickly moves to the spindle poles during mitosis and towards the daughter nuclei in cytokinesis. In both cases, KatA moves away from the zone of overlap between the opposed microtubule half sets of the spindle and phragmoplast respectively — that is, towards the minus ends. In 1999, KatD was characterized and shown to differ considerably from the previously described Kat kinesins37. Secondary-structure predictions for KatA, B and C indicate that they have a long central region — the so-called stalk — containing α-helical coiled-coils that can promote dimerization. KatD does not contain such a region and the authors speculated that it may function as a monomer. Unlike KatA, B and C, which have the motor in the carboxyl terminus, KatD has the putative motor region in the centre of the molecule. Furthermore, unlike the other Kats, expression of KatD is restricted to floral organs. Although the function of KatD is unknown, some other members of the central motor domain subfamily are known to destabilize microtubules. The kinesin-like calmodulin-binding protein, KCBP, first discovered in plants, is unusual in that it has the chimeric properties of a calcium-dependent calmodulin-binding protein and a microtubule motor domain that moves towards the minus ends of microtubules38. Binding of Ca2+/calmodulin substantially reduces the affinity with which KCBP binds to microtubules39. Changes in cytosolic concentrations of Ca2+ at key transitional points are thought to change the affinity of the motor for the microtubules, resulting in changes in microtubule–microtubule crosslinking and cell organization40. As a minus-end-directed motor, KCBP would be antagonistic to the plus-end-directed motors. There is discussion about how this may work for the better-studied mitosis of animal cells (and a KCBP homologue has recently been discovered in sea urchins41), but how might this antagonism function in plant cytokinesis? Although convergence of the minus ends of microtubules is appropriate during mitosis for focusing the movement of chromatids to the two separate poles, convergence could have a negative effect on cytokinesis. KCBP does accumulate at the poles of the anaphase spindle at the time when the broad, barrelshaped spindle poles begin to converge42, but this converging activity is suggested to be switched off during cytokinesis as the phragmoplast grows out beyond the initial confines of the spindle40. Consistent with this, Vos et al.40 observed wavy, misaligned or absent cell plates following microinjection of antibodies that activate KCBP by interfering with the Ca2+/calmodulin regulation. It is possible that although polar convergence is beneficial for mitotic function, it may have a negative effect on the cytokinetic apparatus. KCBP was first isolated from Arabidopsis as a kinesinlike calmodulin-binding protein43, and the following

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 4 5

REVIEWS

TRICHOME

Hair-like epidermal cell.

year the gene was identified as the zwichel mutant of Arabidopsis44. The fact that mutant Arabidopsis plants can develop normally without a functional KCBP gene suggests functional redundancy — the only apparent defect being the abnormal morphogenesis of hair-like TRICHOMES on the surfaces of leaves. Recently, Cai et al.45 biochemically isolated a 90-kDa kinesin-like microtubule motor protein from tobacco pollen tubes. This protein binds organelles found in the cortex of these tip-growing cells and may be important for the translocation of materials for growth. Prospects

The plant MAP field is still in its infancy. Some of the structural MAPs have been identified, and it will be interesting to find out the roles of the individual members of the MAP65, whether they have cell-cycle-related functions and, if so, what signalling factors control their expression. Not only will this be important for understanding cell-cycle progression but, because of the central role that the balance between division and expansion has in continuing plant growth, also in plant physiology. How do factors that influence plant growth, such as endogenous plant growth hormones and environmental factors such as light and gravity, affect the cytoskeleton? By which signalling pathways are the effects of plant MAPs modulated as they are likely to be when microtubules reorientate as the direction of growth changes? Although the 65-kDa plant MAPs seem to be different from animal MAPs, there could still be homologues of the better-known animal MAPs to be discovered. The arrival of the complete sequence of the Drosophila genome has allowed putative homologues of mammalian tau protein to be identified and a range of cytoskeletal proteins to be catalogued46. The Arabidopsis genome sequencing project is now complete47–50 and it is possible to do similar homology searches. The excellent kinesin homepage now lists several dozen putative kinesins in Arabidopsis. Useful though they are, the problem with homology-based approaches is that novel proteins and those with weak homology may not be recognized as having MAP function. MAP65 (REF. 20), for instance, is different from the ‘classical’ animal MAPs. In vitro studies of the interaction between microtubules, MAPs and other accessory proteins should therefore continue to be important. Another problem in testing the function of putative MAPs is that there is likely to be functional redundancy between members of multigene families. The genome of

Moore, R. C. & Cyr, R. J. Association between elongation factor-1a and microtubules in vivo is domain dependent and conditional. Cell Motil. Cytoskeleton 45, 279–272 (2000). Hugdahl, J. D., Bokros, C. L. & Morejohn, L. C. End-to-end annealing of plant microtubules by the p86 subunit of eukaryotic initiation factor-(iso)4F. Plant Cell 7, 2129–2138 (1995). Freudenrich, A. & Nick, P. Microtubular organization in tobacco cells: heat-shock protein 90 can bind to tubulin in vitro. Bot. Acta 111, 273–279 (1998). Lloyd, C. W. (ed.) The Cytoskeletal Basis of Plant Growth and Form (Academic, London, 1991).

5. 6.

the yeast Saccharomyces cerevisiae, for example, has six kinesin-related genes and a single dynein heavy chain (dynein forms part of a large complex that possesses minus-end-directed motor activity), but none is essential for cell viability. Cottingham et al.51 showed that yeast can function with only two motors: one plus-enddirected BimC-type kinesin and one minus-end motor, underlining the functional overlap between different kinesins. Nevertheless, the full description of plant MAPs is likely to affect all areas of plant biology, not just mitosis and cytokinesis. In tissue cells, new plasma membrane is presumably inserted uniformly along the long side walls as they expand, and it will be interesting to search for interphase kinesins involved in membrane processing. Some cells, such as root hairs and pollen tubes, grow only at the tip, and there are likely to be specific kinesins that target growth materials to the apex. There may also be interphase KRPs involved in microtubule–microtubule interactions. In that case, we can speculate that intermicrotubule sliding may be involved in reorganizing microtubules and, perhaps, even in the movement of membrane complexes involved in cell wall biosynthesis. Another area of interest will be in identifying the proteins such as the kinases, phosphatases and accessory proteins that regulate activity of the various kinesins during the cell cycle. As for the other MAPs, one of the biggest challenges is to characterize the membrane complex by which cortical microtubules attach to the plasma membrane. In the post-genomic phase, it will be important to develop cell biological and biochemical assays to test the function of plant MAPs. Much of the microtubule cycle seems to be dedicated to the coordinated and orientated deposition of the cell wall; unfortunately, this same cell wall has hindered the development of cellular assays. Nevertheless, with the ability to microinject materials into cells52,53, the use of model systems to study cytokinesis23, and the advent of fusion proteins tagged with green-fluorescent protein for studying living processes54, we are embarking on an exciting time in the study of plant MAPs. Links FURTHER INFORMATION The kinesin homepage |

Mitosis world ENCYCLOPEDIA OF LIFE SCIENCES Plant cells: Mitosis,

cytokinesis and cell plate formation

Joshi, H. C. & Palevitz, B. A. γ-Tubulin and microtubule organization in plants. Trends Cell Biol. 6, 41–44 (1996). Lambert, A.-M. & Lloyd, C. W. in Microtubules (eds Hyams, J. S. & Lloyd, C. W.) 327–341 (Alan R. Liss, New York, 1994). Wasteneys, G. O. The cytoskeleton and growth polarity. Curr. Opin. Plant Biol. 3, 503–511 (2000). Thought-provoking review discussing the complex relationship between cortical microtubules, cellulose microfibrils and the direction of cell expansion. Lloyd, C. W., Shaw, P. J., Warn, R .M. & Yuan, M. Gibberellic acid-induced reorientation of cortical microtubules in living plant cells. J. Microscopy 181,

140–144 (1996). Himmelspach, R., Wymer, C. L., Lloyd, C. W. & Nick, P. Gravity-induced reorientation of cortical microtubules observed in vivo. Plant J. 18, 449–453 (1999). Gravity-stimulated maize coleoptiles undergo a bending response in which epidermal cells on the lower surface elongate more than cells on the upper. The stages in microtubule reorientation are seen in living cells on the slower-growing upper side. 10. Marc, J., Sharkey, D. E., Durso, N. A., Zhang, M. & Cyr, R. J. Isolation of a 90-kDa microtubule-associated protein from tobacco membranes. Plant Cell 8, 2127–2138 (1996).

www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS 11. Euteneuer, U. & McIntosh, R. Polarity of phragmoplast and midbody microtubules. J. Cell Biol. 87, 509–515 (1980). Classic paper showing that opposing sets of phragmoplast microtubules meet at their fastgrowing plus ends. 12. Cyr, R. J. & Palevitz, B. A. Microtubule-binding proteins from carrot. I. Initial characterization and microtubule bundling. Planta 177, 245–260 (1989). 13. Vantard, M., Schellenbaum, P., Fellous, A. & Lambert, A.-M. Characterization of maize microtubule-associated proteins, one of which is immunologically related to tau. Biochemistry 30, 9334–9340 (1991). 14. Maekawa, T., Ogihara, S., Murofushi, H. & Nagai, R. Green algal microtubule-associated protein with a molecular weight of 90kDa which bundles microtubules. Protoplasma 158, 10–18 (1990). 15. Higashiyama, T., Sonobe, S., Murofushi, H. & Hasezawa, S. Identification of a novel 70kDa protein in cultured tobacco cells that is immunologically related to MAP4. Cytologia 61, 229–233 (1996). 16. Jiang, C.-J. & Sonobe, S. Identification and preliminary characterization of a 65kDa higher-plant microtubuleassociated protein. J. Cell Sci. 105, 891–901 (1993). Isolation of a group of approximately 65-kDa tobacco proteins that bundle microtubules in vitro. 17. Chan, J., Rutten, T. & Lloyd, C. W. Isolation of microtubuleassociated proteins from carrot cytoskeletons: a 120kDa MAP decorates all four microtubule arrays and the nucleus. Plant J. 10, 251–259 (1996). 18. Chan, J., Jensen, C. G., Jensen, L. C. W., Bush, M. & Lloyd, C. W. The 65-kDa carrot microtubule-associated protein forms regularly arranged filamentous cross-bridges between microtubules. Proc. Natl Acad. Sci. USA 96, 14931–14936 (1999). MAP65 forms inter-microtubule bridges in vitro of the length observed in plant cells. 19. Lancelle, S. A., Callaham, D. A. & Hepler, P. K. A method for rapid freeze fixation of plants. Protoplasma 131, 153–165 (1986). Some of the clearest images of inter-microtubule bridges in the plant cortical array. 20. Smertenko, A. et al. Sequencing a plant MAP-65 cDNA reveals a novel class of microtubule–associated protein. Nature Cell Biol. 2, 750–753 (2000). First sequencing of a plant structural MAP. 21. Vale, R. D., Reese, T. S. & Sheetz, M. P. Identification of a novel force generating protein, kinesin, involved in microtubule-based motility. Cell 42, 39–50 (1985). 22. Vale, R. D. & Fletterick, R. J. The design plan of kinesin motors. Annu. Rev. Cell. Dev. Biol. 13, 745–777 (1997). 23. Asada, T., Sonobe, S. & Shibaoka, H. Microtubule translocation in the cytokinetic apparatus of cultured tobacco cells. Nature 350, 238–241 (1991). Elegant experimental system showing the behaviour of tubulin subunits in the phragmoplast, suggesting that the two circlets of microtubules are kept apart by anti-parallel sliding. 24. Asada, T. & Shibaoka, H. Isolation of polypeptides with microtubule-translocating activity from phragmoplasts of tobacco BY-2 cells. J. Cell Sci. 107, 2249–2257 (1994). 25. Enos, A. P. & Morris, N. R. Mutation of a gene that encodes a kinesin-like protein blocks nuclear division in A. nidulans. Cell 60, 1019–1027 (1990). 26. Sharp, D. J., Yu, K. R., Sisson, J. C., Sullivan, W. & Scholey, J. M. Antagonistic microtubule-sliding motors position mitotic centrosomes in Drosophila early embryos. Nature Cell Biol. 1, 51–54 (1999). 27. Asada, T., Kuriyama, R. & Shibaoka, H. TKRP125, a kinesin-related protein involved in the centrosomeindependent organization of the cytokinetic apparatus in tobacco BY-2 cells. J. Cell Sci. 110, 179–189 (1997). Cloning of a BimC-type kinesin found in the phragmoplast shows that these mitotic motors also function in cytokinesis in plants. 28. Lee, Y.-R. & Liu, B. Identification of a phragmoplastassociated kinesin-related protein in higher plants. Curr. Biol. 10, 797–800 (2000).

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

42.

43.

Identification of a novel kinesin that is specific to the phragmoplast. Barroso, C. et al. Two kinesin-related proteins associated with the cold-stable cytoskeleton of carrot cells: characterization of a novel kinesin, DcKRP120-2. Plant J. 24, 1–12 (2000). In addition to the TKRP125 kinesin (reference 27), another BimC-type kinesin seems to function in plant cytokinesis. van Lammeren, A. A. M. Structure and function of the microtubular cytoskeleton during endosperm development in wheat: an immunofluorescent study. Protoplasma 146, 18–27 (1988). Otegui, M. & Staehelin, L. A. Cytokinesis in flowering plants: more than one way to divide a cell. Curr. Opin. Plant Biol. 3, 493–502 (2000). Mitsui, H., Yamaguchi-Shinozaki, K., Shinozaki, K., Nishikawa, K. & Takahashi, H. Identification of a gene family (kat) encoding kinesin-like proteins in Arabidopsis thaliana and the characterization of secondary structure of KatA. Mol. Gen. Genet. 238, 362–368 (1993). The first molecular identification of plant kinesins. Mitsui, H. et al. Sequencing and characterization of the kinesin-related genes katB and katC of Arabidopsis thaliana. Plant Mol. Biol. 25, 865–876 (1994). Mitsui, H., Hasezawa, S., Nagata, T. & Takahashi, H. Cell cycle-dependent accumulation of a kinesin-like protein, KatB/C, in synchronized tobacco BY-2 cells. Plant Mol. Biol. 30, 177–181 (1996). Shows that homologues of kinesin-related proteins first identified in the model plant Arabidopsis are expressed in a cell-cycle-dependent manner in the highly synchronizable tobacco cell suspension. Liu, B. & Palevitz, B. A. Localization of a kinesin-like protein in generative cells of tobacco. Protoplasma 195, 78–89 (1996). Liu, B., Cyr, R. J. & Palevitz, B. A. A kinesin-like protein, KatAp, in the cells of Arabidopsis and other plants. Plant Cell 8, 119–132 (1996). Tamura, K., Nakatani, K., Mitsui, H., Ohashi, Y. & Takahashi, H. Characterization of katD, a kinesin-like protein gene specifically expressed in floral tissues of Arabidopsis thaliana. Gene 230, 23–32 (1999). Song, H., Golovkin, M., Reddy, A. S. N. & Endow, S. A. In vitro motility of AtKCBP, a calmodulin-binding kinesin protein of Arabidopsis. Proc. Natl Acad. Sci. USA 94, 322–327 (1997). Deavours, B. E., Reddy, A. S. N. & Walker, R. A. Ca2+/calmodulin regulation of Arabidopsis kinesin-like calmodulin-binding protein. Cell Motil. Cytoskeleton 40, 408–416 (1998). Vos, J. W., Safadi, F., Reddy, A. S. N. & Hepler, P. K. The kinesin-like calmodulin binding protein is differentially involved in cell division. Plant Cell 12, 979–990 (2000). Antibody microinjection experiments indicate how calcium and the minus-end-directed kinesin-like calmodulin protein, KCBP, may interact at key transition points during plant mitosis and cytokinesis. Rogers, G. C., Hart, C. L., Wedaman, K. P. & Scholey, J. M. Identification of kinesin-C, a calmodulin-binding carboxy-terminal kinesin in animal (Strongylocentrotus purpuratus) cells. J. Mol. Biol. 294, 1–8 (1999). Smirnova, E. A., Reddy, A. S. N., Bowser, J. & Bajer, A. S. Minus end-directed kinesin-like motor protein, Kcbp, localizes to the anaphase spindle poles in Haemanthus endosperm. Cell Motil. Cytoskeleton 41, 271–280 (1998). Reddy, A. S. N., Safadi, F., Narasimhulu, S. B., Golovkin, M. & Hu, X. A novel plant calmodulin-binding protein with a kinesin heavy chain motor domain. J. Biol. Chem. 271, 7052–7060 (1996). First description of a novel plant kinesin with calcium-regulated calmodulin-binding activities. Other studies from this group indicate how it may be physiologically regulated. This protein was subsequently detected in animal cells (see reference

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

41). 44. Oppenheimer, D. G. et al. Essential role of a kinesin-like protein in Arabidopsis trichome morphogenesis. Proc. Natl Acad. Sci. USA 94, 6261–6266 (1997). 45. Cai, G. et al. Identification and characterization of a novel microtubule-based motor associated with membranous organelles in tobacco pollen tubes. Plant Cell 12, 1719–1736 (2000). Biochemical isolation of a kinesin. Found in tipgrowing pollen tubes, it may be involved in directional vesicle transport. 46. Goldstein, L. S. B. & Gunawardena, S. Flying through the Drosophila cytoskeletal genome. J. Cell Biol. 150, F63–F68 (2000). 47. Arabidopsis Genome Initiative. Analysis of the genome sequence of the flowering plant Arabidopsis thaliana. Nature 408, 796–815 (2000). 48. Theologis, A. et al. Sequence and analysis of chromosome 1 of the plant Arabidopsis thaliana. Nature 408, 816–820 (2000). 49. Salanoubat, M. et al. Sequence and analysis of chromosome 3 of the plant Arabidopsis thaliana. Nature 408, 820–822 (2000). 50. Tabata, S. et al. Sequence and analysis of chromosome 5 of the plant Arabidopsis thaliana. Nature 408, 823–826 (2000). 51. Cottingham, F. R., Gheber, L., Miller, D. L. & Hoyt, M. A. Novel roles for Saccharomyces cerevisiae mitotic spindle motors. J. Cell Biol. 147, 335–349 (1999). 52. Hush, J. M., Wadsworth, P., Callaham, D. A. & Hepler, P. K. Quantification of microtubule dynamics in living plant cells using fluorescence redistribution after photobleaching. J. Cell Sci. 107, 775–784 (1994). 53. Yuan, M., Shaw, P. J., Warn, R. M. & Lloyd, C. W. Dynamic reorientation of cortical microtubules, from transverse to longitudinal in living plant cells. Proc. Natl Acad. Sci. USA 91, 6050–6053 (1994). 54. Granger, C. L. & Cyr, R. J. Microtubule reorganization in tobacco BY-2 cells stably expressing GFP-MBD. Planta 210, 502–509 (2000). 55. Lauber, M. H. et al. The Arabidopsis KNOLLE protein is a cytokinesis-specific snytaxin. J. Cell Biol. 139, 1485–1493 (1997). The direction and timing of membrane flow are key elements in plant cytokinesis. This and the following references 56–62 describe a range of non-MAP proteins likely to be important in regulating the formation of the cell plate. 56. Gu, X. & Verma, D. P. S. Dynamics of phragmoplastin in living cells during cell plate formation and uncoupling of cell elongation from the plane of cell division. Plant Cell 9, 157–169 (1997). 57. Del Vecchio, A. J. et al. Centrin homologues in higher plants are prominently associated with the developing cell plate. Protoplasma 196, 224–234 (1997). 58. Bogre, L. et al. A MAP kinase is activated late in plant mitosis and becomes localized to the plane of cell division. Plant Cell 11, 101–113 (1999). 59. Clayton, L. & Lloyd, C. W. Actin organization during the cell cycle in meristematic plant cells: Actin is present in the cytokinetic phragmoplast. Exp. Cell Res. 156, 231–238 (1985). 60. Kakimoto, T. & Shibaoka, H. Cytoskeletal ultrastructure of phragmoplast-nuclei complexes isolated from cultured tobacco cells. Protoplasma S2, 95–103 (1988). 61. Zuo, J. et al. KORRIGAN, an Arabidopsis endo-1,4-βglucanase, localizes to the cell plate by polarized targeting and is essential for cytokinesis. Plant Cell 12, 1137–1152 (2000). 62. Bowser, J. & Reddy, A. S. N. Localization of a kinesin-like calmodulin–binding protein in dividing cells of Arabidopsis and tobacco. Plant J. 12, 1429–1437 (1997).

Acknowledgements We thank Jordi Chan, Sandra McCutcheon, Consuelo Barroso, Andrey Smertenko and Christian Roghi for the images.

VOLUME 2 | JANUARY 2001 | 4 7

REVIEWS

NOT BEING THE WRONG SIZE Richard H. Gomer Size regulation is a never-ending problem. Many of us worry that parts of ourselves are too big whereas other parts are too small. How organisms — and their tissues — are programmed to be a specific size, how this size is maintained, and what might cause something to become the wrong size, are key problems in developmental biology. But what are the mechanisms that regulate the size of multicellular structures? CELL DIVISION TRACHEAE

The air tubes that form the respiratory system of an insect. MIDBLASTULA TRANSITION

Marks the initiation of zygotic gene transcription and the end of the embryo’s dependency on maternal mRNA. The midblastula transition also marks a lengthening of the cell cycle.

Howard Hughes Medical Institute and Department of Biochemistry and Cell Biology MS-140, Rice University, 6,100 South Main Street, Houston, Texas 77005-1892, USA. e-mail: richard@bioc.rice.edu

In his essay On Being the Right Size, J. B. S. Haldane1 pointed out that several factors affect the evolution of animal size. Upper limits are generated by physical constraints, such as the strength of bones, the ability of the heart to pump blood up to the brain, or the power needed to keep a flying animal airborne. For insects, the limited diffusion of oxygen from TRACHEAE limits size. Lower limits on the size of animals are generated by factors such as the surface-to-volume ratio, which affects the ability of animals to retain heat — large animals, with a relatively small surface-to-volume ratio, lose a lower percentage of body heat. So Arctic animals (polar bears, seals and walruses) tend to be large. For animals that are within the size limits, the nature of the ecological niche tends to favour a certain fixed size of the animals within a species2.

Ploidy is clearly correlated with cell size5. For instance, tetraploid newts have much larger cells than haploid newts. Interestingly, however, all of these newts are the same size6. Moreover, altering cell size or the cell division rate (by interfering with cyclins or other cell-cycle regulatory proteins) in wing segments of the fruitfly Drosophila melanogaster can result in a larger number of smaller cells or a smaller number of larger cells. However, these changes do not affect the eventual size of the segment, suggesting that the tissue volume (the product of cell size and cell number) is regulated, rather than being a function of either cell size or number alone7,8. These experiments indicate the existence of mechanisms that regulate the size of an organism or structure irrespective of the number of cells.

Cell size and number determine total size

Establishing an initial number of cells

The size of an organism is determined by the average cell size multiplied by the number of cells. These parameters are controlled by regulating the growth, division, and death rates of cells. Much is known about the signaltransduction machinery within a cell that controls cell division and cell death3,4, but little is known about how this machinery is then used to regulate body or tissue size, or even what regulates cell size. One hypothesis is that, because there is an upper limit to transcription and translation rates, there is a maximal protein-production rate from a gene. Given that proteins have a limited lifetime, there then is a maximal amount of protein that can be supported by a single gene and, in turn, a maximal amount of protein and protein-associated cell mass that can be supported by a single genome.

If cell size is constant, then tissue size can be controlled by regulating the number of cells. At least two types of mechanism seem to regulate cell number. One mechanism allows cells to divide a certain number of times, or for a specified amount of time, to generate some number of cells (BOX 1). This is typified by the nematode Caenorhabditis elegans, in which a mechanism counts the number of cell divisions, and produces a specific number of cells, each with a predetermined lineage9. Another example of a counting mechanism is the MIDBLASTULA TRANSITION in the African clawed frog Xenopus laevis, where the egg rapidly divides 12 times to form 212 cells, after which cell division slows down. The egg starts with a fixed amount of an intracellular factor and titrates this against the amount of DNA. When there is a www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS

OLIGODENDROCYTE

A supporting cell in the nervous system that forms a myelin sheath around axons.

Box 1 | Even cells have deadlines Cell number can be established by having cells divide for a predetermined number of divisions or for a predetermined time. For instance, a simple way to obtain roughly 2n cells is to use a timer. If the cells divide every m hours and a timer allows divisions to proceed for k hours, then there will be k/m cycles of division or ~2k/m cells. Such a mechanism seems to regulate the number of oligodendrocyte precursors in the rat optic nerve. Oligodendrocytes are the glial cells that form a myelin sheath around nerve cells. Both in the nerve and in tissue culture, the oligodendrocyte precursors divide for a fixed amount of time73–75. Changing the culture temperature alters the cell division period m without significantly altering the time k when cell division stops, suggesting that the mechanism involves a timer rather than Divide a certain number of times, or for a fixed time something counting the number of cell divisions, as in the case of the Xenopus midblastula transition. One component of the timer seems to be the cyclin inhibitor p27kip1, which slowly accumulates in the dividing oligodendrocyte precursors76. Another component is a transcriptional regulatory protein whose levels decrease with time, indicating that many timers might be used to ensure that cell division will eventually stop74. Although the timer mechanism could, theoretically, be cell-autonomous, extracellular signals are necessary for its operation: plateletderived growth factor is required for the cell division, and thyroid hormone is required for the timer to stop cell division77,78. A likely possibility is that these factors are used to coordinate the growth and development of oligodendrocyte precursors with that of other cell types. A similar timer seems to regulate the number of cardiac myocytes, indicating that a timer mechanism might be used to regulate the number of many cell types79.

212 increase in the amount of DNA, all of the factor is bound to the DNA. There is, then, no free factor, allowing the cells to ‘know’ that 12 rounds of replication and division have taken place10,11. By contrast, a reasonably specific number of cells can also be generated using an intrinsic timer that stops cell division after some interval, resulting in a fairly specific number of divisions assuming a roughly constant cell-cycle time. This, apparently, is used by OLIGODENDROCYTE precursors and cardiac myocytes to establish their number (BOX 1). Counting the number of cells

Another mechanism to regulate cell number involves a feedback pathway, which monitors the number of cells

and induces cell growth when there are too few cells. This requires that there be some way to sense the number of cells in a group, and one method is to secrete a factor that they simultaneously sense (BOX 2). Examples are the A- and C-factors that are secreted by high density, starving Myxococcus xanthus12–14; the autoinducer produced by the light organ symbiont Vibrio fischeri15; the extracellular differentiation factor A, which stimulates spore formation of Bacillus subtilis16,17; and the conditioned-medium factor that coordinates development in the social amoeba Dictyostelium discoideum18 (TABLE 1). If the cells are in an enclosed space, any such secreted factor will reflect cell density and, hence, cell number. If the size of the enclosed space (that is, the

Box 2 | Secreted factors A simple mechanism to sense the number N of type y cells in a tissue or organism would be to have these cells secrete a specific factor Y. If each type y cell secretes Y at a rate of φ molecules per second, the average lifetime of the Y molecules is T seconds, and the tissue or body volume is V, then the concentration of Y will be NφT/V. This means that there would be a roughly linear relationship between the number of type y cells and the concentration of Y. If the Y factor represses growth of the y cells above a certain concentration, this would then regulate the number of y cells for a given total body volume. If the body volume increased or decreased, the number of y cells would change accordingly to keep the number of y cells fairly constant. A key requirement of this mechanism is that the Y factor is unstable, with a limited lifetime T; molecules with properties of a Y factor are indeed unstable (BOX 4). If there is a decrease (by wounding, for instance) in the number of y cells, the remaining cells will sense their absence as a decrease in Y serum concentration within time T. In the figure, a factor (red dots) secreted by the cells (yellow) binds to receptors (green) and inhibits cell growth. When there is the correct number of cells, a high concentration of the factor causes growth to stop.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 4 9

REVIEWS blood volume) is fixed, then the cell density is a measure of the number of cells. If the cells are in a group that is small compared with the rest of the organism, and the factor can diffuse away from the secreting cells, the concentration of the factor can be used to sense the number of cells in the group19,20. Variations of this mechanism can also be used to sense and regulate cell number (BOX 3). Pathways that regulate organism size

Some genes, when mutated, affect body size. But we do not understand the complete mechanisms that they disrupt to do this. The proteins that regulate body size can be classified as hormones (and proteins involved in the synthesis or secretion of hormones), compo-

Box 3 | Cells can act as a sink rather than a source A simple mechanism can be proposed to regulate the number of cells in a tissue or organism, based on the idea of a single cell secreting a growth factor X at a rate of φ molecules per second. The cells bind X subject to the standard equilibrium conditions of EQN 1: R B= K T ( D F +1)

(1)

where B is the number of bound molecules on a cell, RT is the total number of receptors per cell, KD is the binding constant, and F is the free concentration of X. If the cells internalize or otherwise destroy X at a rate of r molecules per second per cell, and we assume that this degradation rate is proportional to the number of molecules bound to the cell (for instance 1% of the X molecules bound to a cell get internalized every second), then: r=kB

(2)

where k is the internalization/degradation rate per bound molecule of X. If N is the total number of cells whose growth rate is regulated by X, then F will reach an equilibrium when the secretion rate is equal to the total internalization/degradation rate:

φ =Nr

(3)

Combining (EQN 2) and (EQN 3) to get B=φ/Nk, and substituting into (EQN 1), we obtain (EQN 4):

φ (KD F +1) N= kRT

(4)

Possible examples where limiting amounts of a secreted factor control size are the regulation of body size by growth hormone21, or the regulation of oligodendrocyte precursors, where limiting amounts of secreted platelet-derived growth factor regulate the number of cells80. Later in development, the number of neurons that the oligodendrocytes can touch becomes the limiting factor that regulates the number of oligodendrocytes81,82. We do not know what regulates the number of neurons. In the figure, a limiting amount of a factor secreted by a different cell or population of cells (indicated by the dark blue cell) is required for the growth of cells. When there is the correct number of cells, a low concentration of the factor then causes growth to stop.

A li i i 50

nents of signal-transduction pathways, and downstream effectors that regulate growth and cell division. In mammals, levels of growth hormone regulate body size21. Children who lack growth hormone have severe growth defects, and treatment of these children with growth hormone can correct these deficiencies. Moreover, short children with normal levels of growth hormone (presumably the children are short owing to a decreased sensitivity of their cells to growth hormone) can be treated with extra growth hormone to increase their height22,23. Studies on mammalian cells and C. elegans show that the insulin/insulin-like growth factor receptor, insulin receptor substrates, phosphatidylinositol-3-OH kinase (PI(3)K), and S6 kinase (an enzyme that phosphorylates the S6 protein of ribosomes) form a signal-transduction pathway that regulates protein synthesis and thus cell growth24,25. In Drosophila, mutations in genes for the insulin receptor, CHICO (a homologue of insulinreceptor substrates), PI(3)K and S6 kinase decrease body size26–29. In mammals, insulin allows cells to sense the amount of available nutrients. When Drosophila larvae are grown on a low-nutrient medium, the resulting adult flies are small, so one possibility is that the Drosophila insulin pathway coordinates the growth of all tissues to be commensurate with the available nutrients. Overexpression of PI(3)K in the appropriate primordia can increase the size of wings and eyes in flies grown on rich media30. PI(3)K regulates tissue size in different ways in different tissues. In the flies that overexpress PI(3)K, the increased wing size was due to increases in both cell number and size, whereas the increase in eye size was due to increased cell size28. Interestingly, a similar pathway seems to regulate lifespan in C. elegans, suggesting a possible molecular basis for the negative correlation between body size and lifespan within a species31,32. Cell growth and division are, in general, tightly coupled. So as well as growth regulators, such as S6 kinase, cell-cycle regulators also affect body size. In mice, disruption of the cyclin-dependent kinase inhibitors p27 or p18 results in increased cell proliferation and increased body size33–36. Mutations of cdk4 in Drosophila result in a decreased size of the animal37. By contrast, overexpression of the cyclin D–cdk4 complex in some Drosophila cell types causes cell proliferation without a change in cell size, whereas expression of the same complex in other tissues causes cell growth without a change in cell number38. One possible explanation is that cyclin D–cdk4 induces cell growth, and that whereas proliferating cells can compensate for growth by dividing, post-mitotic cells can no longer compensate. In mammals, cyclin D seems to be involved in a similar size regulatory pathway, controlling the regrowth of kidney renal tubule cells after injury39. As with animals, little is known about the regulation of plant size. Various factors, such as rain, wind or even touching the plant, can regulate its size by altering cell size without notably changing the total number of cells40. Overexpression of cyclin D in tobacco plants increases the number of cells and plant size

h www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS

without altering cell size41. As with any overexpression study, however, it is not clear if cyclin D is part of the normal pathway that regulates plant size.

Muscle size is regulated by a feedback loop involving a secreted factor that is used to sense the number of cells that secrete the factor. Myostatin is a polypeptide belonging to the platelet-derived growth factor superfamily that is made by — and secreted from — MYOBLASTS. So, as the percentage of the body occupied by muscle increases, the serum concentration of myostatin increases47. Myostatin negatively regulates myoblast proliferation, and this negative feedback maintains the amount of muscle in the body48. In support of this hypothesis, mutations in myostatin result in abnormally large muscles in mice and cattle49,50. Thyroid size seems to be regulated by a more complex negative-feedback loop. Thyroid cells secrete T3 thyroid hormone, which inhibits the release of thyroid-stimulating hormone (TSH) from the pituitary gland. TSH stimulates growth of the thyroid so, if the thyroid is damaged, the resulting lowered levels of T3 allow more TSH to be released, which in turn stimulates thyroid growth51–54. The percentage of adipose tissue in the body is regulated by an even more complex negative-feedback loop that involves neuronal pathways. The serum concentration of leptin — a polypeptide factor secreted by adipose cells — indicates the percentage of adipose tissue in the body55. Leptin acts on various targets in the hypothalamus, resulting in decreased appetite through unknown behavioural pathways in the brain56. A decreased appetite then results in a decreased accumulation of adipose tissue, completing the feedback loop.

Secreted factors that regulate tissue size

Breaking tissues into subgroups

There are at least two main types of size-regulation mechanism in tissues. The first regulates the size of the entire tissue. A good example is how the size of the liver is maintained and how it regenerates to the correct size after a part of it has been removed42. The second mechanism breaks a large primordium into groups of a specific size. Examples include the formation of teeth, SOMITES or segments, and the development of hair, feather or leaf buds43–45. In mammals there is evidence that tissue size is regulated by secreted diffusible factors (BOX 4). For instance, when fragments of the spleen are transplanted into various sites in a host animal, these fragments grow until their integrated mass is equivalent to that of an individual, normal spleen46. This might indicate — but does not prove — that some factor mediates signalling between the different spleen fragments.

As described above, many biological structures are formed by a primordial tissue or group of cells divided into subgroups. So the size at which the group breaks up, and the mechanism that regulates this, affects the size of the subgroups. In the Drosophila egg, gradients of morphogens specify subregions of the egg, and thus the size of the subgroups within the egg57. Alan Turing developed an elegant mathematical model, using two different diffusible factors, that can establish the spacing of subgroups even when the size of the original group is effectively infinite and there are therefore no external gradients, or if there is a finite group and no external gradients (BOX 5). Other mechanisms can also generate subgroups from a population of cells. For instance, Dictyostelium cells use only one diffusible factor to regulate a rearrangement of cells into groups. Dictyostelium is one of the simplest eukaryotic sys-

Table 1 | Systems where a cell type secretes a specific factor Cell

Factor

Effect of factor

Factors with a direct feedback effect on secretory cells Vibrio fischeri or Homoserine lactone Agrobacterium derivatives tumefaciens bacteria

At high cell density: luciferase production (Vf)101; transfer of Ti plasmid among cells (At)102.

Starving Myxococcus Mixture of amino acids Permits aggregation when there is a xanthus bacteria high density of starving cells103. Bacillus subtilis bacteria

Small polypeptide

Permits uptake of extracellular DNA when there is a high density of cells104.

Starving Dictyostelium CMF glycoprotein

Permits aggregation when there is a high density of starving cells105–108.

Aggregating Dictyostelium

Counting factor

Regulates group size by decreasing cell–cell adhesion when there are too many cells in a stream62,63,72.

Muscle

Myostatin

Regulates muscle mass by inhibiting muscle proliferation47–50.

Factors with an indirect feedback effect on the secretory cells Thyroid

Decreases TSH release; TSH stimulates thyroid growth51–54.

Adipose

Leptin

Decreases appetite55.

(CMF, conditioned medium factor; TSH, thyroid stimulating hormone.)

SOMITE

A group of cells that breaks off from a column of mesoderm cells in a vertebrate embryo; the group then forms a segment of the backbone and associated structures. MYOBLAST

An embryonic cell that becomes a muscle cell or part of a muscle cell.

Box 4 | Early observations of secreted factors that regulate tissue size Many years ago, researchers postulated that a tissue could secrete factors that negatively regulate the growth of that tissue (BOX 1). These factors were named chalones after the Greek chalon, meaning to slacken83–87. The first experiments were based on the observation that, after part of the kidney or liver of a mammal (such as a rat) is removed, there is a burst of mitosis in the remaining tissue. Saetren88 found a heat-labile, non-dialysable, unstable factor from kidney that inhibited the mitosis rate of regenerating kidney but not regenerating liver; a similar factor in liver specifically inhibited mitoses in regenerating liver. Eight years later, Bullough and Lawrence83 found a heat-labile, non-dialysable, unstable factor secreted by epidermis that specifically inhibited the proliferation of epidermal cells; extracts from other tissues did not have this activity89,90. The identity of these factors is still not known.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 5 1

REVIEWS

Box 5 | Turing patterns Alan Turing is best known for two inventions, one theoretical and the other practical. To determine whether mathematics could be mechanized, he developed the idea of the Turing machine. With this idea Turing showed that one could not build a machine that would identify all undecidable propositions. During the Second World War, he helped to design machines that allowed the British to read much of the encrypted German radio traffic, and this arguably prevented the Germans from defeating the British91. As well as these extraordinary contributions, Turing also developed a theoretical mechanism92 that could potentially explain much of differentiation, patterning and size regulation in biological systems. The essential idea is that a field of identical cells can break symmetry and develop patterns of stripes or spots of differentiated cells if the following conditions are met. First, the cells secrete an inducer that has a short diffusion range, potentiates its own secretion and increases the secretion of an inhibitor. Second, the inhibitor has a long diffusion range and inhibits the secretion of the inducer. A slight instability will tend to get amplified, causing a stripe or spot of cells to differentiate into cells that secrete large amounts of inducer, surrounded by cells that respond to the inhibitor and secrete small quantities of inducer. The type of pattern that forms depends on parameters such as the secretion and breakdown rates of the two signals and their diffusion coefficients93. Although this model is appealing owing to its simplicity and generality, the existence of such a mechanism has yet to be proven. In several systems, however, there have been suggestions that a Turing mechanism regulates the size and spacing of structures94–96.

tems for the study of cell-number counting. When starved, these unicellular soil amoebae stream together and form a fruiting body consisting of a mass of spore cells supported by a stalk, 1–2 mm high58,59. If the fruiting body is too large it will fall over, so there is a strong selective pressure to have an upper limit on the number of cells in a fruiting body60,61. One way in which Dictyostelium cells regulate the size of the fruiting body is by using a secreted factor to sense the number of cells streaming in to a group. If the group is too large (as indicated by a high concentration of the factor), the streams break up. The secreted ‘counting factor’ is a 450-kDa complex of polypeptides62,63. One of the polypeptides is a 40-kDa protein named countin. In transformants with a disrupted countin gene, there is no detectable secretion of the counting factor. The aggregation streams therefore do not break up, resulting in huge aggregates and huge fruiting bodies. Addition of purified counting factor to cells causes the streaming cells to break up into abnor-

mally small groups64,65. So, Dictyostelium cells use a secreted signal to regulate the break up of a tissue into subgroups. Linking adhesion and motility to group size

In Dictyostelium and, possibly, other systems a secreted factor that decreases cell–cell adhesion or increases random cell motility can regulate the rearrangement of a set of cells into groups, thereby regulating the size of a group of cells (FIG. 1). Formation of a multicellular structure requires cell–cell adhesion, and many morphogenetic processes seem to involve adhesion molecules66–69. A group of cells — or any other objects, such as molecules or animals within a herd (BOX 6) — will tend to disperse if their adhesion decreases or their random motility increases. Increasing cell–cell adhesion during Dictyostelium development results in the formation of unbroken streams and large aggregates, whereas decreasing adhesion causes broken streams and formation of many small aggregates70–72. The Dictyostelium counting factor seems to regulate group Box 6 | Similar size-regulatory mechanisms

Figure 1 | Division of groups into subgroups. If a secreted factor inhibits cell–cell adhesion and/or increases random cell motility, a group of cells can break into subgroups. If the group is too large (left panel), a decreased adhesion and/or an increased motility causes the group to disperse. When the sources of the factor (that is, the cells) are dispersed (middle panel), there is a resulting decrease in the concentration of the factor. This lower concentration of the factor allows the adhesion to increase, which results in regrouping of the cells (usually into clumps that are smaller than the original group, right panel). With the right parameters, this can then regulate the size and number of the subgroups.

If general physical mechanisms exist to regulate the size of a group of cells, might similar mechanisms regulate the size of social groups of animals? There are models for the size of herds or schools of fish based on the affinity of individual animals for one another, and the propensity of a large herd or school to fragment97–100. These models tend to replicate the observed size distributions of, for instance, schools of tuna or herds of buffalo. The models have similarities to models for cell number regulation in Dictyostelium. In the animals, gregariousness is the equivalent of cell–cell adhesion, and a higher desire by animals to stay together — or a higher cell–cell adhesion in the developing Dictyostelium cells — increases group size. Different animal velocities are the equivalent of random motility in the cells, and a dispersion of velocities has been observed to cause a herd to fragment (the faster-moving animals form a group that pulls away from the slower animals).

www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS size by decreasing cell–cell adhesion72. Computer simulations of an irregular ribbon of cells indicate that, with a given range of random cell motility, if the cell–cell adhesion is constantly high then the stream stays intact. If adhesion is constantly low, then the stream disperses. But if the adhesion is low for a while and then increases, the stream tends to break up into groups72. There is a rough correlation between the amount of adhesion and group size. If adhesion is decreased by a secreted factor, the feedback loop (the concentration of the factor letting the cells know how big the group is) causes the group size to become remarkably uniform72. Future directions

We are just beginning to understand how size is regulated and how cell number is counted. But many experiments remain to be done. Although we can artificially manipulate cell number or size and see a compensation, such that the tissue or organism size stays constant, we are only beginning to understand the basic mechanisms involved. We need to work out the mechanisms that count the number of cell divisions, and this will probably be done by genetics for the total cell number in C. elegans, and by biochemistry for the DNA titration in the Xenopus midblastula transition. We may need to use techniques such as proteomics to uncover the timer mechanisms that allow cells to grow and divide for a fixed amount of time — such as those used by oligodendrocyte precursors and cardiac myocytes.

Haldane, J. B. S. in Possible Worlds and Other Papers 20–28 (Harper & Brothers, New York, 1928). 2. Colinvaux, P. Why Big Fierce Animals Are Rare. An Ecologist’s Perspective (Princeton University Press, Princeton, New Jersey, 1978). 3. Su, T. T. The regulation of cell growth and proliferation during organogenesis. In Vivo 14, 141–148 (2000). 4. Conlon, I. & Raff, M. Size control in animal development. Cell 96, 235–244 (1999). 5. Galitski, T., Saldanha, A., Styles, C., Lander, E. & Fink, G. Ploidy regulation of gene expression. Science 285, 251–254 (1999). 6. Fankhauser, G. The effects of changes in chromosome number on amphibian development. Quart. Rev. Biol. 20, 20–78 (1945). 7. Weigmann, K., Cohen, S. M. & Lehner, C. F. Cell cycle progression, growth, and patterning in imaginal discs despite inhibition of cell division after inactivation of Drosophila Cdc2 kinase. Development 124, 3555–3563 (1997). 8. Neufeld, T. P., de la Cruz, A. F., Johnston, L. A. & Edgar, B. A. Coordination of growth and cell division in the Drosophila wing. Cell 93, 1183–1193 (1998). 9. Sulston, J., Schierenberg, E., White, J. & Thomson, J. The embryonic cell lineage of the nematode Caenorhabditis elegans. Dev. Biol. 100, 64–119 (1983). 10. Newport, J. & Kirschner, M. A major developmental transition in early Xenopus embryos: I. Characterization and timing of cellular changes at the midblastula stage. Cell 30, 675–686 (1982). 11. Newport, J. & Kirschner, M. A major developmental transition in early Xenopus embryos: II. Control of the onset of transcription. Cell 30, 687–696 (1982). References 10 and 11 show that a cell division counter can function by titrating a fixed amount of a compound with DNA for the dividing cells. 12. Kim, S. K. & Kaiser, D. Cell motility is required for the transmission of C-factor, an intercellular signal that coordinates fruiting body morphogenesis of Myxococcus xanthus. Genes Dev. 4, 896–905 (1990).

We also need to identify the various secreted factors that mediate cell-number sensing (BOX 4). As some of these molecules are unstable, this will require very rapid protein-purification methods, the identification of ways to inhibit factor breakdown, or a genetic approach. Studies on the Drosophila insulin receptor/CHICO/PI(3)K signal-transduction pathway indicate that it regulates tissue size. The nature of the signal is unknown, although it is likely to be an insulin/insulin-like growth factor homologue. Conversely, for myostatin, leptin and the Dictyostelium counting factor, the signals are known, but their signaltransduction pathways remain to be worked out. There are many practical uses for manipulating tissue size. Thousands of years of agricultural genetics have been aimed at altering tissue size in crop plants, to make a desired part larger at the expense of unwanted parts. And growth factor is now being used to increase height in humans. Finally, the development of pharmacological agonists or antagonists of other factors that are involved in counting cell number might allow us to regulate the size of specific tissues for cosmetic reasons or, in the case of leptin and adipose tissue, for health reasons. Links DATABASE LINKS CHICO | PI(3)K | S6 kinase | p27 | p18 | cdk4 | cyclin D | TSH | p27kip1 FURTHER INFORMATION Raff lab | Leevers lab | Gomer lab | Edgar lab | Dictyostelium | C. elegans | Drosophila

13. Kim, S. K. & Kaiser, D. C-factor: Cell–cell signaling protein required for fruiting body morphogenesis of M. xanthus. Cell 61, 19–26 (1990). 14. Kaplan, H. B. & Plamann, L. A Myxococcus xanthus cell density-sensing system required for multicellular development. FEMS Microbiol. Lett. 139, 89–95 (1996). 15. Eberhard, A. et al. Structural identification of autoinducer of Photobacterium fischeri luciferase. Biochemistry 20, 2444–2449 (1981). 16. Grossman, A. D. & Losick, R. Extracellular control of spore formation in Bacillus subtilis. Proc. Natl Acad. Sci. USA 85, 4369–4373 (1988). 17. Kaplan, H. B. Cell–cell interactions that direct fruiting body development in Myxococcus xanthus. Curr. Opin. Genet. Dev. 1, 363–369 (1991). 18. Mehdy, M. C. & Firtel, R. A. A secreted factor and cyclic AMP jointly regulate cell-type–specific gene expression in Dictyostelium discoideum. Mol. Cell. Biol. 5, 705–713 (1985). 19. Yuen, I. S. & Gomer, R. H. Cell density-sensing in Dictyostelium by means of the accumulation rate, diffusion coefficient and activity threshold of a protein secreted by starved cells. J. Theor. Biol. 167, 273–282 (1994). 20. Clarke, M. & Gomer, R. H. PSF and CMF, autocrine factors that regulate gene expression during growth and early development of Dictyostelium. Experientia 51, 1124–1134 (1995). 21. Palmiter, R. D., Norstedt, G., Gelinas, R. E., Hammer, R. E. & Brinster, R. L. Metallothionein-human GH fusion genes stimulate growth of mice. Science 222, 809–814 (1983). A major advance in regulating the size of an animal. 22. Voss, L. D. Growth hormone therapy for the short normal child: who needs it and who wants it? The case against growth hormone therapy. J. Pediatr. 136, 103–106 (2000). 23. Sandberg, D. E. Should short children who are not deficient in growth hormone be treated? West. J. Med. 172, 186–189 (2000). 24. Weinkove, D. & Leevers, S. J. The genetic control of organ growth: insights from Drosophila. Curr. Opin. Genet. Dev. 10, 75–80 (2000).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

25. Coelho, C. M. & Leevers, S. J. Do growth and cell division rates determine cell size in multicellular organisms? J. Cell Sci. 113, 2927–2934 (2000). 26. Chen, C., Jack, J. & Garofalo, R. S. The Drosophila insulin receptor is required for normal growth. Endocrinology 137, 846–856 (1996). 27. Bohni, R. et al. Autonomous control of cell and organ size by CHICO, a Drosophila homolog of vertebrate IRS1-4. Cell 97, 865–875 (1999). 28. Leevers, S. J., Weinkove, D., MacDougall, L. K., Hafen, E. & Waterfield, M. D. The Drosophila phosphoinositide 3kinase DP110 promotes cell growth. EMBO J. 15, 6584–6594 (1996). 29. Montagne, J. et al. Drosophila S6 kinase; a regulator of cell size. Science 285, 2126–2129 (1999). 30. Weinkove, D., Neufeld, T., Twardzik, T., Waterfield, M. & Leevers, S. Regulation of imaginal disc cell size, cell number and organ size by Drosophila class I(A) phosphoinositide 3-kinase and its adapter. Curr. Biol. 9, 1019–1029 (1999). 31. Lin, K., Dorman, J. B., Rodan, A. & Kenyon, C. daf-16: An HNF-3/forkhead family member that can function to double the life-span of Caenorhabditis elegans. Science 278, 1319–1322 (1997). 32. Wolkow, C., Kimura, K., Lee, M. & Ruvkun, G. Regulation of C. elegans life-span by insulin-like signaling in the nervous system. Science 290, 147–150 (2000). 33. Nakayama, K. et al. Mice lacking p27Kip1 display increased body size, multiple organ hyperplasia, retinal dysplasia, and pituitary tumors. Cell 85, 707–720 (1996). 34. Franklin, D. et al. CDK inhibitors p18INK4c and p27Kip1 mediate two separate pathways to collaboratively suppress pituitary tumorigenesis. Genes Dev. 12, 2899–2911 (1998). 35. Kiyokawa, H. et al. Enhanced growth of mice lacking the cyclin-dependent kinase inhibitor function of p27kip1. Cell 85, 721–732 (1996). 36. Fero, M. L. et al. A syndrome of multiorgan hyperplasia with features of gigantism, tumorigenesis, and female sterility in p27Kip1-deficient mice. Cell 85, 733–744 (1996).

VOLUME 2 | JANUARY 2001 | 5 3

REVIEWS

37. Meyer, C. et al. Drosophila cdk4 is required for normal growth and is dispensable for cell cycle progression. EMBO J. 19, 4533–4542 (2000). 38. Datar, S., Jacobs, H., de La Cruz, A., Lehner, C. & Edgar, B. The Drosophila cyclin D–cdk4 complex promotes cellular growth. EMBO J. 19, 4543–4554 (2000). 39. Preisig, P. A cell cycle-dependent mechanism of renal tubule epithelial cell hypertrophy. Kidney Int. 56, 1193–1198 (1999). 40. Braam, J. & Davis, R. W. Rain-, wind-, and touch-induced expression of calmodulin and calmodulin-related genes in Arabidopsis. Cell 60, 357–364 (1990). 41. Cockcroft, C. E., den Boer, B. G., Healy, J. M. & Murray, J. A. Cyclin D control of growth rate in plants. Nature 405, 575–579 (2000). 42. Micalopoulos, G. K. & DeFrances, M. C. Liver regeneration. Science 276, 60–66 (1997). 43. Meir, S. Development of the chick embryo mesoblast. Dev. Biol. 73, 25–45 (1979). 44. Jaing, T.-X., Jung, H.-S., Widelitz, R. B. & Chuong, C.-M. Self-organization of periodic patterns by dissociated feather mesenchymal cells and the regulation of size, number and spacing of primordia. Development 126, 4997–5009 (1999). 45. Rawls, A., Wilson–Rawls, J. & Olson, E. N. Genetic regulation of somite formation. Curr. Top. Dev. Biol. 47, 131–154 (2000). 46. Metcalf, D. Restricted growth capacity of multiple spleen grafts. Transplantation 2, 387–392 (1964). 47. Lee, S. & McPherron, A. Myostatin and the control of skeletal muscle mass. Curr. Opin. Genet. Dev. 9, 604–607 (1999). 48. Thomas, M. et al. Myostatin, a negative regulator of muscle growth, functions by inhibiting myoblast proliferation. J. Biol. Chem. 275, 40235–40243 (2000). 49. McPherron, A. & Lee, S. Double muscling in cattle due to mutations in the myostatin gene. Proc. Natl Acad. Sci. USA 94, 12457–12461 (1997). 50. McPherron, A. C., Lawler, A. M. & Lee, S. -J. Regulation of skeletal muscle mass in mice by a new TGF-β superfamily member. Nature 387, 83–90 (1997). An excellent example of a secreted-factor directly inhibiting the growth of the secreted cells. 51. De Groot, L. J. in Control of the Thyroid Gland (eds Ekholm, R., Kohn, L. D. & Wollman, S. H.) 5–10 (Plenum, New York, 1989). 52. Larsen, P. R. in Control of the Thyroid Gland (eds Ekholm, R., Kohn, L. D. & Wollman, S. H.) 11–26 (Plenum, New York, 1989). 53. Eggo, M. & Burrow, G. N. in Control of the Thyroid Gland (eds Ekholm, R., Kohn, L. D. & Wollman, S. H.) 327–339 (Plenum, New York, 1989). 54. Dumont, J. E. et al. in Control of the Thyroid Gland (eds Ekholm, R., Kohn, L. D. & Wollman, S. H.) 357–372 (Plenum, New York, 1989). 55. Schwartz, M., Woods, S., Porte, D. J., Seeley, R. & Baskin, D. Central nervous system control of food intake. Nature 404, 661–671 (2000). 56. Ahima, R. S. Leptin and the neuroendocrinology of fasting. Front. Horm. Res. 26, 42–56 (2000). 57. St. Johnston, D. & Nusslein-Volhard, C. The origin of pattern and polarity in the Drosophila embryo. Cell 68, 201–219 (1992). Lucid description of the Drosophila patterning mechanism. 58. Loomis, W. F. Dictyostelium discoideum: A Developmental System (Academic, New York, 1975). 59. Loomis, W. F. Development of Dictyostelium discoideum. (Academic, New York, 1982). 60. Bonner, J. T. & Hoffman, M. E. Evidence for a substance responsible for spacing pattern of aggregation and fruiting bodies in the cellular slime mold. J. Embryol. Exp. Morphol. 11, 571–589 (1963). 61. Kopachik, W. J. Size regulation in Dictyostelium. J. Embryol. Exp. Morphol. 68, 23–35 (1982). 62. Brock, D. A. & Gomer, R. H. A cell-counting factor regulating structure size in Dictyostelium. Genes Dev. 13, 1960–1969 (1999). 63. Brown, J. M. & Firtel, R. A. Just the right size: Cell counting in Dictyostelium. Trends Genet. 16, 191–193 (2000).

64. Spann, T. P., Brock, D. A., Lindsey, D. F., Wood, S. A. & Gomer, R. H. Mutagenesis and gene identification in Dictyostelium by shotgun antisense. Proc. Natl Acad. Sci. USA 93, 5003–5007 (1996). 65. Brock, D. A. et al. A Dictyostelium mutant with defective aggregate size determination. Development 122, 2569–2578 (1996). 66. Goodman, C. The likeness of being: phylogenetically conserved molecular mechanisms of growth cone guidance. Cell 78, 353–356 (1994). 67. Radice, G. et al. Developmental defects in mouse embryos lacking N-cadherin. Dev. Biol. 181, 64–78 (1997). 68. Myat, M. & Andrew, D. Organ shape in the Drosophila salivary gland is controlled by regulated, sequential internalization of the primordia. Development 127, 679–691 (2000). 69. 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). 70. Kamboj, R. K., Lam, T. Y. & Siu, C. H. Regulation of slug size by the cell adhesion molecule gp80 in Dictyostelium discoideum. Cell Reg. 1, 715–729 (1990). 71. Siu, C. H. & Kamboj, R. K. Cell–cell adhesion and morphogenesis in Dictyostelium discoideum. Dev. Genet. 11, 377–387 (1990). 72. Roisin-Bouffay, C., Jang, W. & Gomer, R. H. A precise group size in Dictyostelium is generated by a cell-counting factor modulating cell–cell adhesion. Mol. Cell 6, 953–959 (2000). 73. Gao, F., Apperly, J. & Raff, M. Cell-intrinsic timers and thyroid hormone regulate the probability of cell-cycle withdrawal and differentiation of oligodendrocyte precursor cells. Dev. Biol. 197, 54–66 (1998). 74. Kondo, T. & Raff, M. Basic helix-loop-helix proteins and the timing of oligodendrocyte differentiation. Development 127, 2989–2998 (2000). 75. Durand, B. & Raff, M. A cell-intrinsic timer that operates during oligodendrocyte development. BioEssays 22, 64–71 (2000). Elegant description of a timer mechanism regulating how long a group of cells can continue dividing. 76. Durand, B., Fero, N. L., Roberts, J. M. & Raff, M. C. p27Kip1 alters the response of cells to nitrogen and is part of a cellintrinsic timer that arrests the cell cycle and initiates differentiation. Curr. Biol. 8, 431–440 (1998). 77. Raff, M., Lillien, L., Richardson, W., Burnem, J. & Noble, M. Platelet-derived growth factor from astrocytes drives the clock that times oligodendrocyte development in culture. Nature 333, 562–565 (1988). 78. Barres, B., Lazar, M. & Raff, M. A novel role for thyroid hormone, glucocorticoids and retinoic acid in timing oligodendrocyte development. Development 120, 1097–1108 (1994). 79. Burton, P. B. J., Raff, M. C., Kerr, P., Yacoub, M. H. & Barton, P. J. R. An intrinsic timer that controls cell-cycle withdrawal in cultured cardiac myocytes. Dev. Biol. 216, 659–670 (1999). 80. Calver, A. et al. Oligodendrocyte population dynamics and the role of PDGF in vivo. Neuron 20, 869–882 1998). 81. Burne, J. F., Staple, J. K. & Raff, M. C. Glial cells are increased proportionally in transgenic optic nerves with increased numbers of axons. J. Neurosci. 16, 2064–2073 (1996). 82. Barres, B. A. & Raff, M. C. Axonal control of oligodendrocyte development. J. Cell Biol. 147, 1123–1128 (1999). 83. Bullough, W. S. & Laurence, E. B. Mitotic control by internal secretion: The role of the chalone–adrenalin complex. Exp. Cell Res. 33, 176–194 (1964). 84. Bullough, W. S. Mitotic and function homeostasis: a speculative review. Cancer Res. 25, 1683–1727 (1965). 85. Boldingh, W. & Laurence, E. Extraction, purification and preliminary characterization of the epidermal chalone: A tissue specific mitotic inhibitor obtained from vertebrate skin. Eur. J. Biochem. 5, 191–198 (1968). 86. Sassier, P. & Bergeron, M. Specific inhibition of cell proliferation in the mouse intestine by an aqueous extract of rabbit small intestine. Cell Tissue Kinet. 10, 223–231

(1977). 87. Barfod, N. M. Isolation and partial identification of eight endogenous G1 inhibitors of JB–1 ascites tumor cell proliferation. Cancer Res. 42, 2420–2425 (1982). 88. Saetren, H. A principle of autoregulation of growth. Production of organ specific mitose-inhibitors in kidney and liver. Exp. Cell Res. 11, 229–232 (1956). 89. Bullough, W. S., Hewett, C. L. & Laurence, E. B. The epidermal chalone: A preliminary attempt at isolation. Exp. Cell Res. 36, 192–200 (1964). 90. Richter, K. et al. Epidermal G1-chalone and transforming growth factor-β are two different endogenous inhibitors of epidermal cell proliferation. J. Cell Physiol. 142, 496–504 (1990). 91. Hodges, A. Alan Turing: The Enigma (Simon & Schuster, New York, 1983). 92. Turing, A. M. The chemical basis of morphogenesis. Phil. Trans. R. Soc. (Lond.) 237, 37–72 (1952). A tour de force of theoretical biology. Turing’s genius, ability to explain things simply, and kind personality are evident in this paper. 93. Meinhardt, H. Models of Biological Pattern Formation (Academic, London, 1982). 94. McNally, J. G. & Cox, E. C. Geometry and spatial patterns in Polysphondylium pallidum. Dev. Genet. 9, 663–672 (1988). 95. Sawai, S., Maeda, Y. & Swada, Y. Spontaneous symmetry breaking Turing-type pattern formation in a confined Dictyostelium cell mass. Phys. Rev. Lett. 85, 2212–2215 (2000). 96. Smith, K. M., Gee, L. & Bode, H. R. HyAlx, an aristalessrelated gene, is involved in tentacle formation in hydra. Development 127, 4743–4752 (2000). 97. Gueron, S., Levin, S. A. & Rubenstein, D. I. The dynamics of herds: From individuals to aggregations. J. Theor. Biol. 182, 85–98 (1996). 98. Okubo, A. in Advances in Biophysics (eds Kotani, M. & Noda, H.) 1–87 (Japan Sci. Soc., Tokyo, 1986). 99. Flierl, G., Grunbaum, D., Levins, S. & Olson, D. From individuals to aggregations: the interplay between behavior and physics. J. Theor. Biol. 196, 397–454 (1999). 100. Bonabeau, E., Dagorn, L. & Freon, P. Scaling in animal group-size distributions. Proc. Natl Acad. Sci. USA 96, 4472–4477 (1999). 101. Fuqua, W. C., Winans, S. C. & Greenberg, E. P. Quorum sensing in bacteria: the LuxR-LuxI family of cell densityresponsive transcriptional regulators. J. Bacteriol. 176, 269–275 (1994). 102. Piper, K. R., von Bodiman, S. B. & Farrand, S. K. Conjugation factor of Agrobacterium tumefaciens regulated Ti plasmid transfer by autoinduction. Nature 362, 448–450 (1993). 103. Kuspa, A., Plamann, L. & Kaiser, D. Identification of heatstable A-factor from Myxococcus xanthus. J. Bacteriol. 174, 3319–3326 (1992). 104. Magnuson, T., Solomon, J. & Grossman, A. D. Biochemical and genetic characterization of a competence pheromone from B. subtilis. Cell 77, 207–216 (1994). 105. Jain, R., Yuen, I. S., Taphouse, C. R. & Gomer, R. H. A density-sensing factor controls development in Dictyostelium. Genes Dev. 6, 390–400 (1992). 106. Jain, R. & Gomer, R. H. A developmentally regulated cell surface receptor for a density-sensing factor in Dictyostelium. J. Biol. Chem. 269, 9128–9136 (1994). 107. Van Haastert, P. J. M., Bishop, J. D. & Gomer, R. H. The cell density factor CMF regulates the chemoattractant receptor cAR1 in Dictyostelium. J. Cell Biol. 134, 1543–1549 (1996). 108. Brazill, D. T., Lindsey, D. F., Bishop, J. D. & Gomer, R. H. Cell density sensing mediated by a G protein-coupled receptor activating phospholipase C. J. Biol. Chem. 273, 8161–8168 (1998).

Acknowledgments I thank K. Beckingham, D. Bell-Pedersen and J. Braam for helpful suggestions, D. Hatton for assistance with the manuscript and figures, and Sheila Herman for preparation of FIG. 1. R.H.G. is an Investigator of the Howard Hughes Medical Institute.

www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS

PROTEIN INTERACTION MAPS FOR MODEL ORGANISMS Albertha J. M. Walhout and Marc Vidal Until recently, classical genetics and biochemistry were the main techniques used to investigate how organisms develop, reproduce, behave and age. But with the availability of complete genome sequences new approaches are emerging. Complete sets of proteins — ‘proteomes’ — can be predicted from genome sequences and used to characterize protein functions globally. For example, through the large-scale identification of physical protein–protein interactions, comprehensive protein interaction maps are being generated. And these maps might help us to understand the processes that control the biology of living organisms. ORTHOLOGUES

Genes that belong to different organisms and that have a similar function. EXPRESSION-PROFILING EXPERIMENTS

One of the first functional genomic approaches developed, in which expression levels of large sets of transcripts are compared under different experimental conditions, or during development of an organism in a large-scale, highthroughput fashion.

Dana-Farber Cancer Institute and Department of Genetics, Harvard Medical School, 44 Binney Street, Boston, Massachusetts 02115, USA. e-mail: marc_vidal@dfci. harvard.edu

Complete genome sequences are available for several model organisms, including Escherichia coli, Saccharomyces cerevisiae, Caenorhabditis elegans and Drosophila melanogaster1–4. In addition, the human genome sequence is near completion and a first version of its annotation is anticipated in the near future5. For S. cerevisiae, C. elegans and Drosophila, this information led to the prediction of ~6,000, ~19,000 and ~13,600 open reading frames (ORFs), respectively. The number of predicted genes that are encoded by the human genome remains unclear, but estimates range between 28,000 and 120,000 (REFS 6–8). The complete sequence of a genome does not necessarily lead to clear insights into the development and biology of the corresponding organism. This is mainly because the molecular function of many predicted gene products has remained experimentally uncharacterized. For example, about 95% (~18,000) of C. elegans predicted ORFs are uncharacterized, except by searches for potential ORTHOLOGUES in comparative genomic strategies9,10. So an important challenge of the ‘post-genomic era’ is to develop strategies that use complete genome sequences to accelerate functional predictions for the genes and their products. Gene or protein function analysis using conventional approaches — that is, with one or a few proteins — can be relatively slow. These approaches are generally not applicable to the tens of thousands of

uncharacterized proteins predicted from genome sequences. So several projects aimed at characterizing gene or protein function on a genome-wide scale have been initiated. Such projects collectively define the field of ‘functional genomics’. In general, they are based on conventional functional assays that have been modified to allow high-throughput or even automated settings in which many genes or proteins can be analysed simultaneously. One of the first functional genomic strategies developed uses DNA chips or microarrays to analyse the EXPRESSION PROFILE of genes whose expression varies during development or under particular experimental conditions11–13. This approach was applied initially to unicellular organisms and cell lines and is now being extended to more complex multicellular organisms such as Drosophila and C. elegans14–16. Another strategy is based on systematic analysis of phenotypes conferred by gene knockouts. Large-scale gene knockout projects have already provided insight into the biological functions of many predicted yeast ORFs17,18. Loss-of-function phenotypes can also be characterized in C. elegans by using RNA-mediated interference (RNAi). This technique involves introduction of double-stranded RNA (dsRNA) into the animal, which can be done either directly by injection or soaking, or indirectly by feeding worms with bacteria expressing the dsRNA19. The RNAi method is rapid and reliable, and has pro-

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 5 5

REVIEWS a Genetic pathways

b Pathway scaffolding

Z X

c Enzymatic reactions

d Molecular machines P

Figure 1 | Protein interactions are crucial in many aspects of biological function. a | Genetic interactions often correlate with physical interactions between the corresponding gene-products. b | Signal-transduction pathways can be tethered to a subcellular location by scaffolding proteins. c | Many enzyme–protein substrate interactions are surprisingly stable and can therefore be detected in protein interaction assays. d | Protein interactions form the basis for the function of large molecular machines.

vided a wealth of functional data20,21. A slightly less developed strategy is based on the large-scale localization of transcripts or proteins in whole animals or within cells. This approach, for the most part, involves in-frame fusions of ORFs of interest and their promoters to the coding sequence of green fluorescent protein (GFP) or β-galactosidase (β-Gal)22–24. Transgenic animals or cells are then generated with these constructs and localization of GFP or β-Gal examined. Here, we focus on another functional genomic approach that consists of the systematic identification of physical protein–protein interactions (hereafter referred to as protein interactions) with the aim of generating comprehensive protein interaction maps. Protein interaction maps

PROTEOME

The complete set of (predicted) proteins, by analogy to genome (the complete genetic material). EPISTASIS ANALYSIS

Epistasis is the masking of a phenotype caused by a mutation in one gene by a mutation in another gene. Epistasis analysis can therefore be used to dissect the order in which genes in a genetic pathway act. SPLICEOSOME

Molecular machine involved in gene splicing.

Many proteins mediate their biological function through protein interactions, so the systematic identification of such interactions for a given PROTEOME has been proposed as a potentially informative functional genomic strategy25–27. As well as their well-described role in the assembly of a cell’s structural compartments such as the cytoskeleton and nuclear pore, protein interactions are crucial for many other aspects of cellular biology. First, genetic interactions often correlate with physical interactions between the corresponding gene products (FIG. 1a). For example, classical EPISTASIS ANALYSES showed that, in a C. elegans pathway that regulates apoptosis, the ced-3 caspase gene acts downstream of ced-4, which itself acts downstream of ced-9 (REFS 28,29). Subsequently, the corresponding proteins CED-3, CED-4 and CED-9 were shown to interact physically30. Second, protein interactions are required to tether the components of signal-

transduction pathways physically (FIG. 1b). A protein providing such a scaffold is the yeast protein Ste5p, which interacts with components of a mitogen-activated protein kinase (MAPK) cascade involved in pheromone signalling31. Third, enzyme–protein substrate interactions are important for catalysis, and are often found to be more stable than previously presumed (FIG. 1c). For example, a cyclin-dependent kinase physically interacts with its substrate p107, a homologue of the retinoblastoma protein pRb32. Last, protein interactions are crucial for the integrity of multicomponent enzymatic machines such as RNA polymerases or the SPLICEOSOME33 (FIG. 1d). So protein interaction mapping projects have been initiated under the assumption that identification of interaction partners for proteins of unknown function can provide insight into their biological function. Comprehensive protein interaction maps should complement conventional genetic and biochemical experiments by providing new hypotheses that can be tested back into the model of interest. The strategies used so far to generate protein interaction maps can be classified into classical and reverse proteomics (FIG. 2). The distinction between these two approaches is similar to the difference between forward and reverse genetics. In forward genetics, the starting point is usually a genetic screen using the mutagenized organism itself to identify mutants with a phenotype of interest. Similarly, in classical proteomics, the starting material is the organism itself, in the form of a protein extract. In reverse genetics, experiments are designed using the (complete) genome sequence of the organism as the starting point. An example of a reverse genetic method in C. elegans is RNAi, in which the gene for which a loss-of-function phenotype is to be examined is chosen a priori. Analogously, in reverse proteomics, the genome sequence is used to predict the corresponding proteome and, with that information in hand, proteomic experiments are designed. Reverse proteomics includes both in silico (that is, computer-based) and experimental approaches. Below, we describe these strategies and discuss one reverse proteomic approach — the yeast two-hybrid system — in greater detail. Proteomics

The discovery of protein interactions has traditionally been based on biochemical or immunochemical purification of proteins of interest, together with their potential interacting partners (for example, see REF. 34) (TABLE 1). These approaches are limited to one or a few proteins at a time, mainly because of the relative difficulty of assigning the identity of each potential partner (for example, see REF. 35). The availability of complete predicted proteomes, together with ever-improving mass spectrometry techniques for protein identification, has allowed these methods to be used for interaction mapping experiments on large protein complexes36,37. In such classical proteomic approaches, the starting material is a protein extract, highly enriched for the complex of interest (TABLE 1). For example, the composition of the yeast www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS nuclear pore complex has recently been determined in this way38. Despite the advantage of using endogenous protein complexes, it is difficult to imagine how these methods will be applicable to complete proteomes. This is mainly because it remains technically challenging to standardize both the biochemical purification conditions and the large-scale production of antibodies (in the case of immunochemical purifications). Alternative proteomic approaches have been developed that could allow standardization of protein interaction mapping strategies. For example, as part of a reverse proteomic project, every predicted yeast ORF has been fused to glutathione-S-transferase (GST)39. Initially, this was used in a biochemical genomic approach to identify enzymatic activities associated with ORF products39. However, it should also be feasible to use such GST-fusion proteins for large-scale protein interaction mapping projects, because it should be more convenient to standardize conditions for complex purification (TABLE 1). A limitation of classical proteomic approaches is that information about the individual interactions within a given protein complex cannot readily be obtained. In other words, although it can be concluded from such analyses that proteins X, Y and Z are part of a complex, it cannot be concluded which pairs among X/Y, X/Z or Y/Z occur within the complex. Moreover, relatively less abundant partners are often hard to detect. These limitations can potentially be overcome by using reverse proteomic approaches (BOX 1). Reverse proteomics: in silico

Once the sequence of a genome is available, it can be annotated by using algorithms that predict the exis-

Genome

Classical proteomics

Reverse proteomics

Transcriptome

Proteome

Figure 2 | Two directions for proteomics in protein interaction mapping. In classical proteomics, the starting material is the organism of interest. Protein complexes are isolated and analysed, and complete genome sequences are used to identify complex components. In reverse proteomics, the starting point is the DNA sequence of the genome of an organism. First, the transcriptome (complete set of transcripts) and proteome (complete set of proteins) are predicted in silico and subsequently this information is used to generate reagents for their analysis.

tence of genes as well as their intron/exon structure when relevant. The resulting information consists of a complete set of potential protein-encoding ORFs40–42. In the first reverse proteomic approach discussed here, potential functional and/or physical interactions are predicted using in silico methods (TABLE 1). This can be done by systematically comparing the proteins of two or more predicted proteomes using specific evolutionary criteria. The ‘Rosetta stone’ method is based on the observation that, during evolution, spe-

Table 1 | Different proteomic strategies Approach

General classification

Biochemical Current Maps Method amenable preparation of levels of available? to biological bait protein standardization verification?

Complex purification

Proteomics

Not required

Low

Yes

Not directly

Co-immunoprecipitation Proteomics

Not required

Low

Not directly

GST pull-downs

Proteomics

Not required

Medium

Not directly

Rosetta stone

Reverse proteomics: Not required in silico

Very high

Yes

Not directly

Phylogenetic profiling

Reverse proteomics: Not required in silico

Very high

Yes

Not directly

Interologues

Reverse proteomics: Not required in silico

High

Yes

Not directly

Yeast two-hybrid

Reverse proteomics: Not required experimental

High

Yes

Yes (by genetics)

Phage display

Reverse proteomics: Required experimental

Medium

Not directly

DHFR two-hybrid

Reverse proteomics: Not required experimental

High

Not directly

FRET

Reverse proteomics: Not required experimental

Medium

Yes

Directly (by imaging)

(DHFR, dihydrofolate reductase; FRET, fluorescence resonance energy transfer; GST, glutathione-S-transferase.)

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 5 7

REVIEWS

Reverse proteomics: yeast two-hybrid

Box 1 | Reverse proteomic strategies In reverse proteomic approaches, the starting point is the genome sequence of an organism of interest, not a protein extract (FIG. 2). As mentioned in the main text, conventional proteomic strategies offer little information about distinct protein interaction pairs. In addition, the characterization of low-abundance proteins is relatively limited. By using genome sequences, complete sets of ORFs can be predicted and subsequently cloned into appropriate expression vectors51,52,72. This allows both the systematic testing of interactions between pairs of individual proteins and the characterization of proteins which, back in the organism of choice, are expressed at very low levels. The main advantage of reverse proteomic approaches that use large sets of ORFs, is that, in theory, ORFs can be systematically fused to appropriate tags. This allows complete standardization of the conditions used for protein interaction detection. The yeast two-hybrid system was the first method developed that allows completely standardized interaction screens (see below). However, it should be remembered that other methods present similar characteristics and thus could potentially be used for the generation of complementary protein interaction maps. This is the case for phage display73 and the dihydrofolate reductase (DHFR) reconstitution method74 for example.

TWO-HYBRID MATRIX EXPERIMENTS

The systematic testing of different protein pairs for a two-hybrid interaction phenotype.

ciation events can result in pairs of genes that fuse together or, conversely, in single genes that split into two separate ones. The assumption is that the proteins encoded by pairs of separate genes might functionally interact in one organism if their orthologues are known to be part of a single protein in another organism43–45. An example of such a potential interaction is that between the E. coli proteins gyrase B and gyrase A. The corresponding Rosetta stone protein is yeast topoisomerase II44. The increasing number of available genome sequences should help in identifying many such potential interactions46. Another method — ‘phylogenetic profiling’ — predicts functional links by examining the conserved existence of particular proteins in predicted proteomes47. This method has not yet been validated experimentally and is based on the assumption that there is strong selective pressure for proteins that functionally interact with each other to be inherited together during speciation events. So, even in the absence of any experimental characterization, proteins can be hypothesized to interact with each other physically or functionally if they tend to be present in the same proteomes. In silico predictions of protein interactions can also be obtained from available information on conserved interactions. In this strategy, proteins that physically interact in one organism are assumed to co-evolve in such a way that their respective orthologues maintain the ability to interact in another organism. Such pairs of evolutionarily conserved interactions have been termed interologues48, examples of which can be found in the MAPK-dependent signalling modules used by many organisms. The three methods described above can be used together to strengthen the hypotheses generated by each of them individually. In addition, systematic searches of keywords in databases that store functional information, such as SWISS-PROT, might allow increasingly accurate predictions of protein interactions and, hence, protein functions45.

The detection of protein interactions using the twohybrid system is based on the functional reconstitution of a transcription factor — for example, the yeast protein Gal4p. Through an interaction between two proteins, say X and Y, fusion proteins of a DNA-binding domain (DB–X) and an activation domain (AD–Y) are tethered to a reporter gene49. Subsequent activation of reporter-gene expression allows selection of yeast cells expressing an interacting protein pair. The two-hybrid system is one of the most widely used reverse proteomic approaches for generating comprehensive protein interaction maps. The advantage of it is that, unlike most proteomic approaches, it does not require protein manipulation because the proteins to be tested are expressed by the yeast cells (TABLE 1) . Standardization allows three large-scale two-hybrid approaches. First, in MATRIX EXPERIMENTS, pairs of proteins are systematically tested for a twohybrid interaction phenotype48,50. In this setting, yeast cells of opposite mating types are used to generate diploid strains that contain different DB–X/AD–Y combinations. These strains are subjected to selective conditions such that only the cells in which X and Y interact will grow. Second, in array experiments the potential interactions of a single DB–X fusion protein are examined against the complete proteome, in the form of individual yeast strains expressing AD–Y ORFs 51. In such an experiment, DB–X-expressing yeast cells are mated with yeast cells expressing each of the AD–Y fusion proteins. Last, different DB–X expressing yeast strains can be pooled and used in mass-mating experiments with a pool of yeast cells expressing AD–Y ORFs51,52. The two-hybrid approach has been used on a relatively small scale for yeast to generate protein interaction maps for components of the splicing machinery53,54 and RNA polymerase III55. Several laboratories have started proteome-wide analyses of protein interactions in S. cerevisiae51,52. So far, hundreds of potential protein interactions have been identified in the different yeast projects56. These interactions should be viewed as hypotheses until they have been validated in yeast using classical biochemical or genetic methods (see below). For C. elegans, relatively few potential protein interactions are available48, and more large-scale protein interaction mapping projects are underway. However, although small, the C. elegans protein interaction map is the first step towards a comprehensive map for a multicellular organism. In addition, the approach taken has allowed the validity of the two-hybrid system as a tool for generating protein interaction maps to be tested. The biological system of choice was vulval development, in which many protein interactions are crucial. This allowed the rate of false negatives in this version of the two-hybrid system to be determined. About 50% of previously reported interactions could be detected, underlining the fact that the two-hybrid system is suitable to obtain an initial coverage for a protein interaction map, but also that it should be complemented by www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

REVIEWS

CONTIG

Usually refers to an overlapping set of DNA fragments. Here, we use contig in the context of interaction clusters in which contiguous series of interactions can be found. COMPLEMENTATION GROUP

Independent mutations in a single locus that fail to complement the phenotypes they confer. CLUSTERING ANALYSIS

By looking for interaction partners that have been found with multiple baits, interaction contigs can be identified.

other proteomic strategies. As well as the overlap with known biological interactions, roughly 150 potential interactions were found that can now be examined using standard genetic approaches. Using maps to generate hypotheses

Although potentially useful, the interactions detected by protein interaction mapping projects are hypothetical until they are validated in the relevant organism. Some interactions will be missed (false negatives) and some will be irrelevant (false positives; BOX 2). By compiling data from different proteomic projects, we can estimate that between two and ten interaction partners will be found for each predicted protein in a proteome44,56. Because each interaction should, in theory, be found twice — that is, X potentially interacts with Y, therefore Y binds X — as many as 30,000 (that is (6,000 × 10)/2) and 100,000 (that is (20,000 × 10)/2) potential interactions can be expected for yeast and C. elegans, respectively. With so many potential protein interactions, interaction-classification strategies need to be developed according to their potential biological significance, with the assumption that such classifications will allow increasingly meaningful hypotheses to be formulated. There are several possible classification strategies. First, protein interactions detected previously in vivo or by another interaction-detection method are more likely to represent genuine interactions. For example, by searching the yeast proteome database, YPD57, 109 yeast two-hybrid interactions were found to have been reported previously51. Second, the identification of interologues should help in the identification of bona fide interactions48. A third method of classification involves integrating genetic data available in the literature. For example, if mutations in the genes that encode the bait and its potential partner confer similar or oppo-

Box 2 | False negatives and false positives False negatives represent interactions that cannot be detected under the conditions used and false positives represent spurious interactions that are meaningless in vivo. False negatives might arise in different ways. For example, as mentioned in the main text, in classical proteomic strategies, transient or unstable interactions are difficult to retrieve. In addition, interactions can be missed using computer-based methods because primary protein sequence information might not be sufficient to predict particular protein interactions. One limitation of the two-hybrid system is that some proteins do not function well as DB–X baits in yeast. For example, they might activate transcription irrespective of the presence of an AD–Y interaction partner75 or fail to localize to the yeast nucleus. For such proteins, alternative two-hybrid approaches, such as the Sos recruitment system76, are more likely to identify putative interaction partners. A second limitation is that many protein interactions depend on posttranslational modifications of either partner, which might not occur in yeast cells. False positives can also arise in different ways. First, certain proteins can behave as spurious interactors in biochemical assays and in the context of the two-hybrid system. Moreover, some protein domains can be found in many proteins, so they behave as spurious interactors in computational analyses44. As more data become available, it will become easier to identify such domains and proteins, and omit them from the analyses. Finally, false positives can occur if interactions detected in the two-hybrid system or predicted in silico are irrelevant in the biological system of interest. For example, members of a potential interaction pair might never ‘meet in real life’ because they are expressed in different cells or localized in different cellular compartments.

site phenotypes, the interaction is more likely to be relevant in vivo. The usefulness of these classification strategies is exemplified by the genetic and protein interactions that were found in the LET-60/Ras pathway in C. elegans48 (FIG. 3a) and in other systems. After two-hybrid analyses using the nematode proteins, most of these physical interactions were retrieved (FIG. 3b). These results confirm the overlap between genetic and physical interactions in this pathway, and show the conservation of these interactions throughout evolution. Another way to classify potential protein interactions relates to the observation that, in transcription or DNA replication for example, several proteins act as part of large molecular machines33 (FIG. 1d). In such multiprotein complexes, protein X interacts with Y, which interacts with Z and so on. These ‘interaction CONTIGS’ might be circular, as for molecular machines, or linear, in the case of genetic pathways (FIG. 3b). The search for such contigs leads to interaction clusters and is applicable to protein interactions predicted in silico44 or using the yeast twohybrid system48,51,52. In large-scale two-hybrid screens, several potential interaction clusters have been identified. They were recognized when several of the same interaction partners were selected with multiple baits that function in the same genetic pathway. An example of this is the synthetic multivulva (synMuv) pathway in C. elegans. This pathway inhibits the LET-60/Ras inductive pathway, and mutations in both a class A and class B synMuv gene are required to obtain a synMuv phenotype. Different COMPLEMENTATION GROUPS have been identified in forward genetic screens, and several of the corresponding genes have been cloned58 (FIG. 4a). Subsequently, additional class A and B synMuv genes have been identified using reverse genetic methods59,60. Some gene products have been characterized in more detail, but the molecular function of most of the synMuv gene products remains unknown. Using two-hybrid assays followed by CLUSTERING ANALYSES, several potential functional links were made between synMuv gene products and predicted proteins (FIG. 4b)48. For example, EGR-1 (MTA-1) was found to interact with two synMuv proteins, LIN-53 (RbAp48) and LIN-36. Previously, lin-53 (RbAp48) and lin-36 have been shown to interact genetically 61. In addition, egr-1 has been reported to function as a synMuv A gene59, and MTA-1 (the mammalian orthologue of EGR-1) is a component of the NuRD complex, which also contains RbAp48 and is involved in transcriptional repression62. So, the hypothesis is that EGR-1 might be involved in transcriptional repression in C. elegans, in a complex that contains LIN-53(RbAp48) and, potentially, other components of this genetic pathway such as LIN-37 and LIN-36 (FIG. 4b). These observations also indicate that the class A and class B synMuv pathways might be physically associated. Interpreting maps

After putative protein interactions have been ordered according to potential biological relevance, the next challenge is to interpret the resulting protein interac-

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 5 9

REVIEWS tion maps. For example, although different potential binding partners of a given protein might have been identified, it does not necessarily mean that they all act in a multiprotein complex â&#x20AC;&#x201D; different interacting partners might be required for different biological functions. This is exemplified by the existence of many signalling pathways centring on the small GTPase Ras. Numerous downstream effectors of Ras, which govern specific downstream outcomes in response to different stimuli, have been identified in many model organisms63,64. Different binding partners for a given protein might also be required in distinct cellular compartments. For example, the Notch signalling pathway in C. elegans and other systems65 is activated through binding of a ligand such as LAG-2 to the Notch receptor (LIN-12 or GLP-1). As the ligand and receptor are both transmembrane proteins, this interaction occurs at the plasma membrane. However, the Notch receptor is cleaved after activation and translocates to the nucleus, where it is thought to act as a transcription factor. This function requires physical interactions with other partners, such as LAG-1 (Su(H)) and LAG-366. Finally, different binding partners might function with a protein of interest in different parts of an animal or at different times during development. Information from cellular and subcellular protein localization studies should help to interpret the many potential protein interactions identified. Validation of potential protein interactions

One validation approach for novel interactions is to co-purify or co-immunoprecipitate the partner proa

teins. However, both methods require antibodies against both partner proteins. Because such reagents are not available for every predicted protein, this approach is not yet feasible on a large scale. Furthermore, it might be difficult to use such an approach in model organisms such as C. elegans if the interaction occurs in only one or a few cells. Another method for validating potential protein interactions is fluorescence resonance energy transfer (FRET), which can be done either in vitro or in living cells using derivatives of GFP. For example, this method has been used to examine protein interactions in nuclear transport in yeast67. So far, this method has been amenable only to unicellular organisms or cell lines, and we do not know how well it will translate into multicellular organisms. Potential protein interactions might also be validated genetically in yeast and C. elegans. One approach would be to examine loss-of-function phenotypes of the genes that encode the interaction partners if this information is not yet available in the literature. If a potential protein interaction is relevant in vivo, the loss-of-function phenotypes should be similar for an activating interaction or opposite for an inhibitory interaction. Another way to validate potential protein interactions is to examine the effect of a loss of protein interaction in vivo. This can be done in two ways. First, the product of a genetically identified mutant allele can be tested in the two-hybrid system for the loss of protein interactions to any of its potential partners48,66. To show the relevance of an interaction, compensatory mutations in the partner protein that restore the interaction can then be generated and tested for functional comb

LIN-3

LET-23 LIN-7

Anchor cell Vulvar precursor cell

LET-23 SEM-5

LIN-2

LIN-7 LIN-2

LIN-10 LET-60 SUR-8

GAP-1 SUR-2 LIN-39

SEM-5

LIN-45

SLI-1

LIN-10 LIN-10

LET-60 SUR-8

GAP-1 SUR-2 LIN-39

SUR-5 LIN-25 UNC-101

PTP-2 KSR-1 SLI-1

PTP-2 KSR-1 MPK-1

LIN-45

MEK-2

LIN-1 LIN-31

MPK-1 MEK-2

LIN-1 LIN-31

Figure 3 | Genetic and physical interactions overlap in the LET-60/Ras pathway involved in vulval development in Caenorhabditis elegans. a | The LET-60/Ras pathway is important in vulval development in C. elegans77. The anchor cell expresses an EGF-like molecule (LIN-3) that activates the LET-23 (EGFR) protein expressed at the surface of the vulval precursor cell. Several genetic interactions and a few physical interactions have been reported in the literature for the LET60/Ras pathway in C. elegans. Touching circles indicate protein interactions that were reported previously in the nematode. b | The interactions that could be detected in the two-hybrid system are indicated in red. Two-hybrid analyses confirm that many physical interactions overlap with genetic interactions in the LET-60/Ras pathway. Furthermore, most of these physical interactions were found previously with orthologous gene products in other model organisms. This suggests that the conservation of protein interactions throughout evolution, or interologues, is a valid approach for the classification of potential protein interactions.

www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

ÂŠ 2001 Macmillan Magazines Ltd

REVIEWS

a Class A

Class B

LIN-15A

LIN-53

EGR-1

LIN-35 HDA-1

LIN-9 LIN-15B LIN-36 LIN-37

b LIN-15A

LIN-36 LIN-37 EGR-1 LIN-53

LIN-9 LIN-15B

mutant phenotype68. Second, missense alleles of the interaction partners can be generated de novo69,70. The ability of such alleles to rescue loss-of-function phenotypes conferred by the corresponding genes should be tested in vivo. Many genes confer a lethal phenotype when knocked out by RNAi, so a final validation approach is to identify specific molecules that dissociate the interaction71. Phenotypes associated with a protein interaction would then have to be examined by exposing the organism to the interaction-dissociating molecules. Summary

LIN-35 HDA-1

Comprehensive genome sequencing projects started for a few model organisms and for humans in the late 1980s. The usefulness of such projects has only recently been fully realized. Protein interaction mapping projects are now in their infancy and, so far, they have not drastically modified the way we formulate and test biological hypotheses. However, as the coverage of potential interactions increases and reaches some level of saturation, it is assumed that protein interaction maps will modify how we address biological problems. Here, we have attempted to point out early observations that indicate that this assumption might be correct. We anticipate that large-scale protein interaction mapping projects will be extended from yeast and nematodes to other model organisms and even humans in the near future.

Figure 4 | Clustering analysis of interactions involving synthetic multivulva proteins in Caenorhabditis elegans suggests the existence of a multiprotein complex. a | The synthetic multivulva (synMuv) genes in C. elegans constitute two redundant pathways. A mutation in both a class A and a class B synMuv gene is required to obtain a multivulva phenotype. Three synMuv class B proteins, LIN35(Rb), LIN-53(RbAp48) and HDA-1, have been analysed in more detail and were shown to interact with each other in vitro78. However, most of the gene products have not been characterized at the functional level. b | Two-hybrid analyses using the synMuv proteins and subsequent protein interaction clustering provided functional links between synMuv proteins and several novel predicted proteins. Touching red (synMuv class B) and blue (synMuv class A) circles indicate potential interactions found in the two-hybrid system. Through clustering analyses, several synMuv class B (green and red circles) proteins could be linked to the synMuv class A (blue circles) proteins EGR-1 and LIN-15A, suggesting that the two pathways are physically linked.

DATABASE LINKS CED-4 | CED-9 | Ste5p | topoisomerase

plementation of the mutant partner in vivo. This approach has been used to show the relevance of the interaction between LET-23 (EGFR) and LIN-7 in C. elegans: mutations in LIN-7 that restore binding to a mutant LET-23 receptor restored a specific let-23

II | Gal4p | LIN-12 | GLP-1 | LAG-1 | LET-23 Yeast proteome database | Yeast two-hybrid protein interaction map | In silico protein interaction map | Nematode two-hybrid protein interaction map ENCYCLOPEDIA OF LIFE SCIENCES Protein quaternary structure: subunit–subunit interactions

Links

FURTHER INFORMATION

Blattner, F. R. et al. The complete genome sequence of Escherichia coli K12. Science 277, 1453–1474 (1997). Goffeau, A. et al. The yeast genome directory. Nature 387, S1–S105 (1997). 3. The C. elegans sequencing consortium. Genome sequence of the nematode C. elegans: a platform for investigating biology. Science 282, 2012–2018 (1998). 4. Adams, M. D. et al. The genome sequence of Drosophila melanogaster. Science 287, 2185–2195 (2000). 5. Pennisi, E. Finally, the book of life and instructions for navigating it. Science 288, 2304–2307 (2000). 6. Ewing, B. & Green, P. Analysis of expressed sequence tags indicates 35,000 human genes. Nature Genet. 25, 232–234 (2000). 7. Roest-Crollius, H. et al. Estimate of human gene number provided by genome-wide analysis using Tetraodon nigroviridis DNA sequence. Nature Genet. 25, 235–238 (2000). 8. Liang, F. et al. Gene index analysis of the human genome estimates approximately 120,000 genes. Nature Genet. 25, 239–240 (2000). 9. Chervitz, S. A. et al. Comparison of the complete protein sets of worm and yeast: orthology and divergence. Science 282, 2022–2027 (1998). 10. Rubin, G. M. et al. Comparative genomics of the eukaryotes. Science 287, 2204–2215 (2000). References 9 and 10 are an excellent source for 2.

comparative genomics in worms, yeast and flies. 11. Schena, M., Shalon, D., Davis, R. W. & Brown, P. O. Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science 270, 467–470 (1995). 12. Lockhart, D. J. et al. Expression monitoring by hybridization to high-density oligonucleotide arrays. Nature Biotechnol. 14, 1675–1680 (1996). 13. Chee, M. et al. Accessing genetic information with highdensity DNA arrays. Science 274, 610–614 (1996). 14. White, K. P., Rifkin, S. A., Hurban, P. & Hogness, D. S. Microarray analysis of Drosophila development during metamorphosis. Science 286, 2179–2184 (1999). 15. Reinke, V. et al. A global profile of germline gene expression in C. elegans. Mol. Cell 6, 605–616 (2000). 16. Hill, A. A., Hunter, C. P., Tsung, B. T., Tucker-Kellogg, G. & Brown, E. L. Genomic analysis of gene expression in C. elegans. Science 290, 809–812 (2000). 17. Winzeler, E. A. et al. Functional characterization of the S. cerevisiae genome by gene deletion and parallel analysis. Science 285, 901–906 (1999). 18. Ross-Macdonald, P. et al. Large-scale analysis of the yeast genome by transposon tagging and gene disruption. Nature 402, 413–418 (1999). 19. Fire, A. et al. Potent and specific genetic interference by double-stranded RNA in Caenorhabditis elegans. Nature 391, 806–811 (1998).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

20. Fraser, A. G. et al. Functional genomics analysis of C. elegans chromosome I by systematic RNA interference. Nature 408, 325–330 (2000). 21. Gonczy, P. et al. Functional genomics analysis of cell division in C. elegans using RNAi of genes on chromosome III. Nature 408, 331–336 (2000). 22. Lynch, A. S., Briggs, D. & Hope, I. A. Developmental expression pattern screen for genes predicted in the C. elegans genome sequencing project. Nature Genet. 11, 309–313 (1995). 23. Ding, D. Q. et al. Large-scale screening of intracellular protein localization in living fission yeast cells by the use of a GFP-fusion domain DNA library. Genes Cells 5, 169–190 (2000). 24. Jansen, G. et al. The complete family of genes encoding G proteins of Caenorhabditis elegans. Nature Genet. 21, 414–419 (1999). 25. Bartel, P. L., Roecklein, J. A., SenGupta, D. & Fields, S. A protein linkage map of Escherichia coli bacteriophage T7. Nature Genet. 12, 72–77 (1996). 26. Lander, E. S. The new genomics: global views of biology. Science 274, 536–539 (1996). 27. Walhout, A. J. M., Endoh, H., Thierry-Mieg, N., Wong, W. & Vidal, M. A model of elegance. Am. J. Hum. Genet. 63, 955–961 (1998). 28. Hengartner, M. O. in C. elegans II (eds Riddle, D. L., Blumenthal, T., Meyer, B. & Priess, J. R.) 383–415 (Cold

VOLUME 2 | JANUARY 2001 | 6 1

REVIEWS

29.

30.

31.

32.

33.

34.

35.

36. 37. 38.

39.

40. 41.

42.

43.

44.

45.

Spring Harbor Laboratory Press, Cold Spring Harbor, 1997). Yang, X., Chang, H. Y. & Baltimore, D. Essential role of CED-4 oligomerization in CED-3 activation and apoptosis. Science 281, 1355–1357 (1998). Chinnaiyan, A. M., O’Rourke, K., Lane, B. R. & Dixit, V. M. Interaction of CED-4 with CED-3 and CED-9: a molecular framework for cell death. Science 275, 1122–11126 (1997). Choi, K. Y., Satterberg, B., Lyons, D. M. & Elion, E. A. Ste5 tethers multiple protein kinases in the MAP kinase cascade required for mating in S. cerevisiae. Cell 78, 499–512 (1994). Zhu, L., Harlow, E. & Dynlacht, B. D. p107 uses a p21CIP1-related domain to bind cyclin/cdk2 and regulate interactions with E2F. Genes Dev. 9, 1740–1752 (1995). Alberts, B. The cell as a collection of protein machines: preparing the next generation of molecular biologists. Cell 92, 291–294 (1998). Harlow, E., Whyte, P., Franza, B. R. J. & Schley, C. Association of adenovirus early-region 1A proteins with cellular polypeptides. Mol. Cell. Biol. 6, 1579–1589 (1986). Whyte, P. et al. Association between an oncogene and an anti-oncogene: the adenovirus E1A proteins bind to the retinoblastoma gene product. Nature 334, 124–129 (1988). Pandey, A. & Mann, M. Proteomics to study genes and genomes. Nature 405, 837–846 (2000). Yates, J. R. Mass spectrometry. From genomics to proteomics. Trends Genet. 16, 5–8 (2000). Rout, M. P. et al. The yeast nuclear pore complex: composition, architecture and transport mechanism. J. Cell Biol. 148, 635–651 (2000). This paper describes a comprehensive approach for identifying components of a large protein complex and describes validation strategies to strengthen these observations obtained by classical proteomic strategies. Martzen, M. R. et al. A biochemical genomics approach for identifying genes by the activity of their products. Science 286, 1153–1155 (1999). This paper describes a biochemical genomic approach in yeast by identifying enzymes using the yeast proteome fused to glutathione-S-transferase (GST). Burge, C. & Karlin, S. Prediction of complete structures in human genomic DNA. J. Mol. Biol. 268, 78–94 (1997). Birney, E. & Durbin, R. Using GeneWise in the Drosophila annotation experiment. Genome Res. 10, 547–548 (2000). Reese, M. G., Kulp, D., Tammana, H. & Haussler, D. Genie — gene finding in Drosophila melanogaster. Genome Res. 10, 529–538 (2000). Enright, A. J., Iliopoulos, I., Kyrpides, N. C. & Ouzonis, C. A. Protein interaction maps for complete genomes based on gene fusion events. Nature 402, 86–89 (1999). Marcotte, E. M. et al. Detecting protein function and protein–protein interactions from genome sequences. Science 285, 751–753 (1999). Marcotte, E. M., Pellegrini, M., Thompson, M. J., Yeates, T. O. & Eisenberg, D. A combined algorithm for genomewide prediction of protein function. Nature 402, 83–86 (1999). References 43–45 were among the first to describe in silico reverse proteomic strategies.

46. Xenarios, I. et al. DIP: the database of interacting proteins. Nucleic Acids Res. 28, 289–291 (2000). 47. Pellegrini, M., Marcotte, E. M., Thompson, M. J., Eisenberg, D. & Yeates, T. O. Assigning protein functions by comparative genome analysis: protein phylogenetic profiles. Proc. Natl Acad. Sci. USA 96, 4285–4288 (1999). 48. Walhout, A. J. M. et al. Protein interaction mapping in C. elegans using proteins involved in vulval development. Science 287, 116–122 (2000). This paper describes the first, small-scale, worm protein interaction map using the yeast two-hybrid system. 49. Fields, S. & Song, O. A novel genetic system to detect protein–protein interactions. Nature 340, 245–246 (1989). 50. Finley, R. L. Jr & Brent, R. Interaction mating reveals binary and ternary connections between Drosophila cell cycle regulators. Proc. Natl Acad. Sci. USA 91, 12980–12984 (1994). 51. Uetz, P. et al. A comprehensive analysis of protein–protein interactions in Saccharomyces cerevisiae. Nature 403, 623–627 (2000). 52. Ito, T. et al. Toward a protein–protein interaction map of the budding yeast: a comprehensive system to examine two-hybrid interactions in all possible combinations between yeast proteins. Proc. Natl Acad. Sci. USA 97, 1143–1147 (2000). References 51 and 52 describe large-scale protein interaction maps for yeast using the yeast twohybrid system. 53. Fromont-Racine, M., Rain, J. C. & Legrain, P. Toward a functional analysis of the yeast genome through exhaustive two-hybrid screens. Nature Genet. 16, 277–282 (1997). 54. Fromont-Racine, M. et al. Genome-wide protein interaction screens reveal functional networks involving Sm-like proteins. Yeast 17, 95–110 (2000). 55. Flores, A. et al. A protein–protein interaction map of yeast RNA polymerase III. Proc. Natl Acad. USA 96, 7815–7820 (1999). 56. Walhout, A. J. M., Boulton, S. J. & Vidal, M. Yeast twohybrid systems and protein interaction mapping projects for yeast and worm. Yeast 17, 88–94 (2000). 57. Garrels, J. I. YPD — A database for the proteins of Saccharomyces cerevisiae. Nucleic Acids Res. 24, 46–49 (1996). 58. Ferguson, E. L. & Horvitz, H. R. The multivulva phenotype of certain Caenorhabditis elegans mutants results from defects in two functionally redundant pathways. Genetics 123, 109–121 (1989). 59. Solari, F. & Ahringer, J. NURD-complex genes antagonise Ras-induced vulval development in Caenorhabditis elegans. Curr. Biol. 10, 223–226 (2000). 60. Hsieh, J. et al. The RING finger/B-box factor TAM-1 and a retinoblastoma-like protein LIN-35 modulate contextdependent gene silencing in Caenorhabditis elegans. Genes Dev. 13, 2958–2970 (1999). 61. Thomas, J. H. & Horvitz, H. R. The C. elegans gene lin-36 acts cell autonomously in the lin-35 Rb pathway. Development 126, 3449–3459 (1999). 62. Xue, Y. et al. NURD, a novel complex with both ATPdependent chromatin-remodeling and histone deacetylase activities. Mol. Cell 2, 851–861 (1998). 63. Wittinghofer, A. & Nassar, N. How Ras-related proteins talk to their effectors. Trends Biochem. Sci. 21, 488–491 (1996).

64. Tan, P. B. & Kim, S. K. Signaling specificity: the RTK/RAS/MAP kinase pathway in metazoans. Trends Genet. 15, 145–149 (1999). 65. Greenwald, I. LIN-12/Notch signaling: lessons from worms and flies. Genes Dev. 12, 1751–1762 (1998). 66. Petcherski, A. G. & Kimble, J. LAG-3 is a putative transcriptional activator in the C. elegans Notch pathway. Nature 405, 364–368 (2000). 67. Damelin, M. & Silver, P. A. Mapping interactions between nuclear transport factors in living cells reveals pathways through the nuclear pore complex. Mol. Cell 5, 133–140 (2000). This paper describes the use of fluorescence resonance energy transfer (FRET) for protein interaction mapping in vivo in yeast. 68. Kaech, S. M., Whitfield, C. W. & Kim, S. K. The LIN2/LIN-7/LIN-10 complex mediates basolateral membrane localization of the C. elegans EGF receptor LET-23 in vulval epithelial cells. Cell 94, 761–771 (1998). 69. Vidal, M., Brachmann, R., Fattaey, A., Harlow, E. & Boeke, J. D. Reverse two-hybrid and one-hybrid systems to detect dissocation of protein–protein and DNA–protein interactions. Proc. Natl Acad. Sci. USA 93, 10315–10320 (1996). 70. Vidal, M., Braun, P., Chen, E., Boeke, J. D. & Harlow, E. Genetic characterization of a mammalian protein–protein interaction domain by using a yeast reverse two-hybrid system. Proc. Natl Acad. Sci. USA 93, 10321–10326 (1996). 71. Vidal, M. & Endoh, H. Prospects for drug screening using the reverse two-hybrid system. Trends Biotechnol. 17, 374–381 (1999). 72. Walhout, A. J. M. et al. GATEWAY recombinational cloning: application to the cloning of large numbers of open reading frames or ORFeomes. Methods Enzymol. 328, 575–592 (2000). 73. Zozulya, S., Lioubin, M., Hill, R. J., Abram, C. & Gishizky, M. L. Mapping signal transduction pathways by phage display. Nature Biotechnol. 17, 1193–1198 (1999). 74. Pelletier, J. N., Arndt, K. M., Pluckthun, A. & Michnick, S. W. An in vivo library-versus-library selection of optimized protein–protein interactions. Nature Biotechnol. 17, 683–690 (1999). 75. Walhout, A. J. M. & Vidal, M. A genetic strategy to eliminate self-activator baits prior to high-throughput yeast two-hybrid screens. Genome Res. 9, 1128–1134 (1999). 76. Aronheim, A., Zandi, E., Hennemann, H., Elledge, S. J. & Karin, M. Isolation of an AP-1 repressor by a novel method for detecting protein–protein interactions. Mol. Cell. Biol. 17, 3094–3102 (1997). 77. Sternberg, P. W. & Han, M. Genetics of RAS signalling in C. elegans. Trends Genet. 14, 466–472 (1998). 78. Lu, X. & Horvitz, H. R. lin-35 and lin-53, two genes that antagonize a C. elegans Ras pathway, encode proteins similar to Rb and its binding protein RbAp48. Cell 95, 981–991 (1998).

Acknowledgements We thank S. Boulton, L. Matthews and J. Dekker for critical reading of the manuscript. The work from this laboratory is supported by grants from the NHGRI, the NCI and the MGRI awarded to M.V.

www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

PERSPECTIVES OPINION

Breaking the mitochondrial barrier Jean-Claude Martinou and Douglas R. Green Pro- and anti-apoptotic members of the Bcl-2 family control the permeability of the outer mitochondrial membrane. They could do this either by forming autonomous pores in the membrane or by collaborating with components of the permeability transition pore. Here we discuss why we favour the first of these possibilities.

The mitochondrion acts as a pivotal decision centre in many types of apoptotic responses: it releases death-promoting factors from its intermembrane space into the cytosol and it produces ATP. One of the factors released by the mitochondrion during apoptosis is cytochrome c, which normally shuttles electrons between protein complexes of the respiratory chain. Once released into the cytosol, cytochrome c binds to Apaf-1 (apoptotic protease-activating factor 1) and, in the presence of dATP or ATP, recruits and activates procaspase 9 to form a complex called the apoptosome1. Activated caspase 9 can, in turn, activate other caspases that finally dismantle the cell. Apoptosis therefore depends on ATP. If mitochondria are damaged and fail to produce ATP early during apoptosis, the apoptosome cannot form, caspase 9 is not activated and cells die by necrosis2,3 — this underscores the ‘pro-apoptotic’ role of functional mitochondria. The outer mitochondrial membrane (OMM) becomes permeable to apoptogenic factors such as cytochrome c, Smac/DIABLO4,5 (which is another caspase activator) and AIF6 (apoptosis-inducing factor); this might mediate the cleavage of DNA during caspase-independent cell death). Permeabilization of the

OMM is controlled by members of the Bcl-2 family (for a review see REF. 1). The anti-apoptotic members such as Bcl-2 or Bcl-xL inhibit the release of mitochondrial apoptogenic factors, whereas the pro-apoptotic members (for example Bax and Bak) trigger this release (BOX 1). But how do Bcl-2 family members control permeabilization of the OMM while preserving mitochondrial function, at least temporarily, and how do they favour apoptosis rather than necrosis? The point of action

Anti-apoptotic members of the Bcl-2 family are often anchored to membranes (such as the mitochondria, endoplasmic reticulum or nuclear envelope) through their hydrophobic carboxy-terminal domain. By contrast, many pro-apoptotic members including Bax and the ‘BH3-only’ proteins (where BH3 stands for Bcl-2 homology domain 3) Bid, Bad and Bim (BOX 1) are normally found in the cytosol or are loosely associated with membranes. After a death signal, these proteins translocate into the OMM where they act1,7. How do the pro-apoptotic proteins work? Data from several groups indicate that, in response to environmental cues, post-translational modifications render the BH3 domains of BH3-only proteins accessible to other members of the Bcl-2 family. For example, Bid is cleaved by caspase 8 and then binds to Bax or Bak, resulting in exposure of the amino terminus of these proteins. Moreover, in the case of Bax, the carboxyl terminus would be disengaged from the BH3 domain that it normally masks8. These conformational changes would be followed by insertion of

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

Bax and/or Bak into the OMM, where these proteins would homo-oligomerize and form channels1,9 (FIG. 1). Anti-apoptotic proteins such as Bcl-2, Bcl-xL and E1B19K can interact with Bax or Bak to prevent their conformational changes or their oligomerization and insertion into the OMM1,10. We do not know whether other BH3-only proteins also trigger changes in Bax or Bak, nor whether they induce permeabilization of the OMM in a manner akin to Bid. Bad is activated by dephosphorylation, which allows it to release from 14-3-3 proteins, although it is then thought to heterodimerize with Bcl-xL to block the function of this anti-apoptotic protein1. Similarly, activation of Bim induces its dissociation from the cytoskeletal dynein complex, although it is also believed to function by antagonizing Bcl-2 and Bcl-xL11. Permeabilization of the OMM

We still do not know precisely how cytochrome c and other proteins are released Box 1 | The Bcl-2 family Bcl-2 was discovered at the chromosomal breakpoint of t(14;18)-bearing human B-cell lymphomas. The Bcl-2 family contains many members that can be subdivided into antiand pro-apoptotic proteins. These proteins have up to four Bcl-2 homology (BH) domains, which correspond to α-helical segments. Anti-apoptotic proteins such as Bcl-2 and Bcl-xL have four BH domains, whereas the pro-apoptotic members seem to lack BH4. The pro-apoptotic Bcl-2-family members can be subdivided into two groups: the Bax subfamily (Bax, Bak and Bok), members of which contain BH1, BH2 and BH3; and the ‘BH3-only’ proteins (such as Bid, Bad and Bim), which have only the BH3 domain. This domain is required for the killing activity of the pro-apoptotic proteins. Many Bcl-2 family members also contain a carboxy-terminal hydrophobic domain, which allows these proteins to be anchored in intracellular membranes.

VOLUME 2 | JANUARY 2001 | 6 3

PERSPECTIVES

Apoptotic signal

Cytosol Bax

tBid

Bcl-2 Bcl-xL

19K

E1B

Bak Cytochrome c OMM

IMM

Matrix

Figure 1 | Activation of Bax and Bak by BH3-only proteins. Members of the Bax subfamily are activated by BH3-only proteins (where BH3 stands for Bcl-2 homology domain 3) in either the cytosol or the outer mitochondrial membrane (OMM). After interacting with cleaved Bid (tBid) or another BH3-only protein in the cytosol or at the mitochondrial surface, Bax undergoes a conformational change44, inserts in the OMM, oligomerizes, and releases cytochrome c and other apoptogenic factors from the intermembrane mitochondrial space26. Cleaved Bid can also translocate and insert into the OMM. Here it activates Bak, which is constitutively anchored to this membrane through its carboxy-terminal domain9. Like Bax, Bak undergoes a conformational change, oligomerizes, and permeabilizes the OMM. Activation of Bak by cleaved Bid has been described in Fas-mediated apoptosis of hepatocytes9. The anti-apoptotic proteins Bcl-2, Bcl-xL and E1B19K inhibit the conformational change and/or the insertion and oligomerization of Bax1,10.

from the mitochondrial intermembrane space. But there are two prevailing models: the formation of autonomous channels by members of the Bax subfamily; and the nonspecific rupture of the OMM by swelling of the mitochondrial matrix and expansion of the inner membrane. Kinetic studies of cytochrome c release during apoptosis in HeLa cells indicate that, after an initial and variable lag period, all the cytochrome c is released from all the mitochondria within a very short time12. This ‘all or nothing’ effect underscores the importance of this step as a decision-making process. The channel models

According to these models, large channels or pores in the OMM conduct soluble proteins out of the mitochondria. The theory is based on protein crystallography, which revealed a structural similarity between Bcl-xL and the pore-forming domain of diphtheria toxin and bacterial colicins13. A similar structure has been discovered recently for Bax8 and can be predicted for other members of the Bcl-2 family, such as Bcl-2 and Bak. Moreover, despite sequence homology limited to its BH3 domain, the three-dimensional struc-

ture of Bid is remarkably similar to that of Bcl-xL14,15, indicating that other BH3-only proteins might also have such a structure. In a similar way to the bacterial toxins that they resemble, several members of the Bcl-2 family — including Bcl-xL, Bcl-2, Bax and cleaved Bid — form functional ion channels in synthetic lipid vesicles and planar lipid membranes1. These channels show multiconductance levels, are voltage- and pH-dependent, and are poorly ion selective (except for the Bcl-2 and Bcl-xL channels, which prefer monovalent cations). But do these proteins also form channels in mitochondrial membranes? As yet, there is only indirect evidence for this. Addition of Bax, Bak or Bid directly to isolated mitochondria from liver or HeLa cells results in permeabilization of the OMM, but the ultrastructure and volume homeostasis of mitochondria are preserved. Moreover, these mitochondria maintain their membrane potential (∆ψm), protein-import capacity and oxygen consumption1,9,16,17. As well as the apoptogenic factors, many other proteins readily escape from the mitochondria during apoptosis. These proteins are often very large (take Smac/DIABLO, for

example, which behaves as a 100 kDa protein as it is an end-to-end dimer18). So how can a channel formed by a 20 kDa protein let so many large proteins pass through? The answer might lie in the ability of Bax to oligomerize. Indeed, only Bax oligomers — and not monomers — are pro-apoptotic19,20. We think that this property might be specific to members of the Bax subfamily, although we cannot exclude the idea that, after cleavage of their amino terminus, antiapoptotic proteins might also oligomerize and become pro-apoptotic21. Saito et al.22 recently reported that, in liposomes, a Bax tetramer can form a channel with an estimated pore size of around 22 Å that can transport cytochrome c. In theory, however, this channel would not be big enough to conduct larger proteins, indicating that further oligomerization or other proteins might be required to enlarge the pore. There are many examples of small bacterial toxins multimerizing to form huge channels. For example, pneumolysin, a toxin produced by Streptococcus pneumoniae, oligomerizes after binding to cholesterol in cell membranes, forming a ring of 30–50 subunits with a size of 350–450 Å that allows the passage of large proteins23. Similarly, Bax and Bax-like molecules could form autonomous multimeric megachannels in the OMM to conduct large proteins out of the intermembrane space. However, no one has yet seen such channels. The permeability transition pore

The second model for permeabilization of the OMM is the idea that it is ruptured through swelling of the mitochondrial matrix and expansion of the inner membrane. The structure behind this is the permeability transition pore (PTP), formed from a complex of the voltage-dependent anion channel (VDAC), the adenine nucleotide translocator (ANT) and cyclophilin D, plus several other proteins, at contact sites between the mitochondrial outer and inner membranes24,25 (see the article by Zamzami and Kroemer on page 67 of this issue). In vitro, the PTP opens under conditions of oxidative stress, high Ca2+ or low ATP concentrations. This allows low-molecularweight solutes (up to ~1.5 kDa) to diffuse across the inner mitochondrial membrane (IMM), resulting in mitochondrial swelling. Unfolding of the inner membrane upon matrix swelling can break the OMM, and swollen mitochondria have been observed during tissue ischaemia and reperfusion, indicating that opening of the PTP might be important in the pathogenesis of necrosis www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

PERSPECTIVES after ischaemia and reperfusion. This idea was proposed by Crompton and Costi more than ten years ago, and has been largely confirmed by studies on acute pathologies involving impaired Ca2+ metabolism and oxidative stress25. These data have prompted many laboratories, including ours, to test whether the PTP could also be involved in apoptosis by recruiting members of the Bcl-2 family. Although we and others failed to detect any interaction between components of the PTP and Bax26, Bak9, Bcl-2 or Bcl-xL (Martinou et al. unpublished data), other groups have reported physical and functional links of these proteins with either the ANT or the VDAC. For example, using systems reconstituted into liposomes, Bax and the ANT have been shown to interact and cooperate to generate high-conductance channels with high opening probabilities, an effect prevented by Bcl-2 (REF. 27). This interaction is proposed to account for induction of the PTP by pro-apoptotic members of the Bcl-2 family and inhibition of the PTP by anti-apoptotic family members. According to other studies, however, it is the VDAC rather than the ANT that interacts with Bax28. Whereas Bax or the VDAC alone could not trigger the efflux of cytochrome c incorporated into liposomes, these proteins allowed diffusion of cytochrome c when both were incorporated into proteoliposomes. Bax might alter the physical properties of the VDAC, making the VDAC channel permeable to cytochrome c 29; Bcl-2, on the other hand, through its BH4 domain, could maintain the VDAC in a closed configuration30. The ‘enlarged’ VDAC in this model will allow the release of cytochrome c, but not larger proteins. Although not completely compatible with the experimental observation that larger proteins are released during OMM permeabilization, this model has common features with the channel models discussed above. However, data on the opening and enlargement of the VDAC seem to contradict those obtained from cells undergoing apoptosis in response to growth-factor deprivation31. In these cells, closure of the VDAC leads to hyperpolarization of the IMM and subsequent damage to the OMM owing to defective exchange of ATP between the mitochondria and cytosol. Bcl-xL prevents this VDAC closure and OMM damage. Studies showing that creatine phosphate accumulates in the intermembrane space after the withdrawal of growth factors further support the idea that the VDAC closes under these conditions32, although the possibility that the VDAC subsequently opens or otherwise participates in OMM permeability cannot be excluded.

Whether the PTP opens or the VDAC closes, each of these models (except the ‘VDAC pore’ model) postulates that permeability of the OMM is caused by swelling of the matrix, leading to rupture of the OMM as the IMM expands. Such a rupture is compatible with the observed release of high-molecular-weight proteins. However, ruptures have not been seen in Xenopus mitochondria after OMM permeabilization17, and studies in mammalian a

cells have shown that discontinuities in the OMM during apoptosis can be prevented with caspase inhibitors33. The search for the ‘holes’ through which proteins exit the intermembrane space goes on. Sources of conflict

The main advantage of the ‘PTP model’ and rupture of the OMM over the ‘channel model’ is that it more easily explains the release of b

Cell death signal

'BH3-only' proteins

Cell death signal

Ceramide, Ca2+ and other PTP openers, but neither Bax nor Bak

Bcl-2 Bax/Bak

Bcl-2

Bcl-xL

Open PTP PTP Open

Competence to die by apoptosis? Yes

Moderate ∆Ψm and ATP loss

Severe ∆Ψm and ATP loss

Caspase activation

Apoptosis

Caspase-independent cell death/necrosis

Figure 2 | Two models for permeabilization of the outer mitochondrial membrane. a | Bax-subfamily members. After a death signal, BH3-only proteins (where BH3 stands for Bcl-2 homology domain 3) activate members of the Bax subfamily, leading to their insertion in the outer mitochondrial membrane (OMM), where they form large channels. Bcl-2 and Bcl-xL inhibit the conformational change and/or oligomerization of Bax or Bak. These channels allow the release of apoptogenic factors from the mitochondrial intermembrane space without altering the function of mitochondria. If cells are competent to die, then cytochrome c and ATP participate in formation of the apoptosome, resulting in activation of caspase 9 and apoptosis. When cells are not competent to die (because they cannot activate the apoptosome and caspases), an alternative pathway is engaged. Owing to the release of cytochrome c, mitochondria cease to produce ATP. This leads to further mitochondrial damage, energy catastrophe and, finally, to a caspase-independent death that resembles necrosis. This model is in line with the nature of the death observed in Apaf-1 knockout mice3, in which interdigital cells cannot activate the apoptosome and so undergo necrosis rather than apoptosis. b | Opening of the permeability transition pore (PTP). Death stimuli activate pathways involving classical PTP openers such as Ca2+ and ceramide, but not members of the Bax subfamily. This leads to opening of the PTP, disruption of the mitochondrial membrane potential (∆Ψm) and loss of ATP. Solutes and water enter the mitochondrial matrix, mitochondria swell and their OMM ruptures, allowing the release of proteins from the intermembrane space. ‘Strong’ cell death stimuli usually lead to necrosis due to irreversible mitochondrial damage and energetic catastrophe. But ‘mild’ death stimuli may damage a subpopulation of mitochondria only, leading to apoptosis if the level of ATP produced by spared mitochondria is enough to activate the apoptosome. Anti-apoptotic proteins such as Bcl-2 and Bcl-xL could prevent opening of the PTP through a mechanism that remains to be determined. This model would be in line with the fact that anti-apoptotic and pro-apoptotic members of the Bcl-2 family can act independently45 and that Bcl-2 can prevent necrosis under certain circumstances46. It and would explain all experiments describing a drop in ∆ψm preceding permeabilization of the OMM and caspase activation34.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 6 5

PERSPECTIVES many proteins from the intermembrane space of mitochondria. Moreover, involvement of the PTP is supported by many reports that the mitochondrial membrane potential (∆ψm) collapses before activation of caspases and apoptosis (for a review see REF. 34). Despite these data, the cooperation of Bax or Bcl-2 with the PTP is far from being generally accepted. Why? Using genetic analysis of the mitochondrial response of yeast to Bax, many groups have failed to validate the requirement for either the VDAC35,36 or the ANT37,38 for Bax killing in yeast, as originally described27,28. Even the strongest argument in favour of the PTP — which relies on the adverse effect of cyclosporin A on Bax activity27,39 and its protective action in many apoptotic responses34 — must be taken with caution. Indeed, the effect of cyclosporin A might not be a specific indicator of PTP involvement. As well as cyclophilin D in the mitochondrial matrix, human cells contain at least eight other cyclophilins outside mitochondria, the functions of which have yet to be defined25. Moreover, contradictory results as to the effect of cyclosporin A on the inhibition of Bax-induced cytochrome c release from isolated mitochondria have been reported1,40. And in many apoptotic responses, the release of cytochrome c and other proteins precedes the disruption of ∆ψm and opening of the PTP1,12,40. But maybe the main criticism of the ‘PTP model’ is that it does not seem to be compatible with the ability of mitochondria to remain functional long enough to allow caspase activation and apoptosis. How, then, can we reconcile all these data? Our opinion

As discussed above, two mechanisms for permeabilization of the OMM during apoptosis emerge from the literature. One involves members of the Bax subfamily, which act independently of the PTP (FIG. 2a); the other involves the PTP in a manner that can be regulated by anti-apoptotic members of the Bcl2 family, but not by Bax subfamily members (FIG. 2b). These two mechanisms are not mutually exclusive, and could occur sequentially or concurrently in certain apoptotic pathways. Moreover, the mechanisms that lead to OMM permeabilization might not be the same in all tissues. In the first model (FIG. 2a), permeabilization of the OMM by members of the Bax subfamily leads to caspase activation or to opening of the PTP, depending on the competence of cells to die by apoptosis. ‘Competence to die by apoptosis’ reflects the

ability of a given cell to activate the apoptosome in response to apoptogenic factors. After microinjection of cytochrome c into the cytosol, competent cells activate the apoptosome and undergo apoptosis41. But some cells, such as sympathetic neurons cultured in the presence of nerve growth factor, do not activate caspases and survive despite a cytochrome-c-flooded cytosol42. Although one could argue that, in addition to cytochrome c, Smac/DIABLO is required43 to activate caspases, we think that ‘competence to die by apoptosis’ might extend beyond the simple requirement for mitochondrial apoptogenic factors and might reflect a property of many cells exposed to survival and death stimuli at the same time. After permeabilization of the OMM, cells that are not competent to die by apoptosis are condemned to a caspase-independent process that resembles necrosis. Indeed, once their OMM is permeabilized, the mitochondria stop working because the cytochrome c depletion can impair electron transport and ATP production. Loss of mitochondrial energy production leads inevitably to death, with kinetics that depend on the ability of cells to produce ATP through anaerobic glycolysis. This model explains all apoptotic responses involving the release of apoptogenic factors before mitochondrial dysfunction. A second model suggests that, in many apoptotic responses, a drop in ∆ψm precedes permeabilization of the OMM (FIG. 2b). We postulate that in these apoptotic pathways, mitochondrial dysfunction is due not to members of the Bax subfamily, but rather to classical PTP openers (ceramide, elevated concentrations of Ca2+ and so on). But how can opening of the PTP result in apoptosis rather than necrosis? Depending on the intensity of the death stimulus, cells can undergo either necrosis or apoptosis. For example, exposing neurons to high levels of glutamate results in necrosis, whereas low levels activate the apoptosis pathway25. In some cells, ‘mild death signals’ might alter only a subpopulation of the mitochondria, or be responsible for a transient opening (flickering) of the PTP in some mitochondria. Spared mitochondria can produce enough ATP to activate caspases, allowing cells to undergo what Crompton25 has called “accidental apoptosis”. Anti-apoptotic members of the Bcl-2 family can counteract this type of cell death by preventing opening of the PTP through a mechanism that remains unclear. In conclusion, dangerous molecules are safely stored in the intermembrane space of mitochondria. The safety is guaranteed by the physical barrier of the OMM, disruption of

which is likely to lead to cell death. At least two mechanisms can make this membrane leaky: activation of Bax-subfamily members and opening of the PTP, although only Bax subfamily members can trigger permeabilization of the OMM without mitochondrial dysfunction. Because apoptosis depends on ATP, activation of Bax or Bax-like proteins independently of the PTP is the best guarantee for death of a cell by apoptosis. Jean-Claude Martinou is in the Departement de Biologie Cellulaire, Sciences III, 30 quai Ernest Ansermet, 1211 Genève 4, Switzerland. (Previous address: Serono Pharmaceutical Research Institute, 14, Chemin des Aulx, 1228 Plan les Ouates, Geneva, Switzerland.) e-mail: JeanClaude.Martinou@cellbio.unige.ch. Douglas R. Green is at the La Jolla Institute for Allergy and Immunology, 10355 Science Center Drive, San Diego, California 92121, USA. e-mail: Dgreen5240@aol.com.

Links DATABASE LINKS Apaf-1 | Smac | AIF | Bcl-2 |

Bcl-xL | Bax | Bak | Bid | Bad | Bim | cyclophilin D | Bok ENCYCLOPEDIA OF LIFE SCIENCES Apoptosis: molecular mechanisms 1.

10.

11.

12.

13.

14.

15.

Desagher, S. & Martinou, J.-C. Mitochondria as the central control point of apoptosis. Trends Cell Biol. 10, 369–377 (2000). Nicotera, P. & Leist, M. Energy supply and the shape of death in neurons and lymphoid cells. Cell Death Differ. 4, 435–442 (1997). Chautan, M., Chazal, G., Cecconi, F., Gruss, P. & Golstein, P. Interdigital cell death can occur through a necrotic and caspase-independent pathway. Curr. Biol. 9, 967–970 (1999). Du, C., Fang, M., Li, Y., Li, L. & Wang, X. Smac, a mitochondrial protein that promotes cytochrome cdependent caspase activation by eliminating IAP inhibition. Cell 102, 33–42 (2000). Verhagen, A. M. et al. Identification of DIABLO, a mammalian protein that promotes apoptosis by binding to and antagonizing IAP proteins. Cell 102, 43–53 (2000). Susin, S. A. et al. Molecular characterization of mitochondrial apoptosis-inducing factor. Nature 397, 441–446 (1999). Hsu, Y.-T., Wolter, K. G. & Youle, R. J. Cytosol-tomembrane redistribution of Bax and Bcl-xL during apoptosis. Proc. Natl Acad. Sci. USA 94, 3668–3672 (1997). Suzuki, M., Youle, R. J. & Tjandra, N. Structure of Bax: Coregulation of dimer formation and intracellular localization. Cell 103, 645–654 (2000). Wei, M. C. et al. tBID, a membrane-targeted death ligand, oligomerizes BAK to release cytochrome c. Genes Dev. 14, 2061–2071 (2000). Perez, D. & White, E. TNF-α signals apoptosis through a Bid-dependent conformational change in Bax that is inhibited by E1B19K. Mol. Cell 6, 53–63 (2000). Puthalakath, H., Huang, D. C. S., O’Reilly, L. A., King, S. M. & Strasser, A. The proapoptotic activity of the Bcl-2 family member Bim is regulated by interaction with the dynein motor complex. Mol. Cell 3, 287–296 (1999). Goldstein, J. C., Waterhouse, N. J., Juin, P., Evan, G. I. & Green, D. R. The coordinate release of cytochrome c during apoptosis is rapid, complete and kinetically invariant. Nature Cell Biol. 2, 156–162 (2000). Muchmore, S. W. et al. X-ray and NMR structure of human Bcl-xL, an inhibitor of programmed cell death. Nature 381, 335–341 (1996). Chou, J. J., Honglin, L., Salvesen, G. S., Yuan, J. & Wagner, G. Solution structure of Bid, an intracellular amplifier of apoptotic signaling. Cell 96, 615–624 (1999). McDonnell, J. M., Fushman, D., Milliman, C. L., Korsmeyer, S. J. & Cowburn, D. Solution structure of the proapoptotic molecule Bid: a structural basis for apoptotic

www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

PERSPECTIVES agonists and antagonists. Cell 96, 625–634 (1999). 16. Shimizu, S. & Tsujimoto, Y. Proapoptotic BH3-only Bcl-2 family members induce cytochrome c release, but not mitochondrial membrane potential loss, and do not directly modulate voltage-dependent anion channel activity. Proc. Natl Acad. Sci. USA 97, 577–582 (2000). 17. von Ahsen, O. et al. Preservation of mitochondrial structure and function after Bid- or Bax-mediated cytochrome c release. J. Cell Biol. 150, 1027–1036 (2000). 18. Chai, J. C. et al. Structural and biochemical basis of apoptotic activation by Smac/DIABLO. Nature 406, 855–862 (2000). 19. Antonsson, B., Montessuit, S., Lauper, S., Eskes, R. & Martinou, J.-C. Bax oligomerization is required for channel-forming activity in liposomes and to trigger cytochrome c release from mitochondria. Biochem. J. 345, 271–278 (2000). 20. Gross, A., Jockel, J., Wei, M. C. & Korsmeyer, S. J. Enforced dimerization of BAX results in its translocation, mitochondrial dysfunction and apoptosis. EMBO J. 17, 3878–3885 (1998). 21. Cheng, E. H.-Y. et al. Conversion of Bcl-2 to a Bax-like death effector by caspases. Science 278, 1966–1968 (1997). 22. Saito, M., Korsmeyer, S. J. & Schlesinger, P. H. Baxdependent transport of cytochrome c reconstitued in pure liposomes. Nature Cell Biol. 2, 553–555 (2000). 23. Gilbert, R. J. C. et al. Two structural transitions in membrane pore formation by pneumolysin, the poreforming toxin of Streptococcus pneumoniae. Cell 97, 647–655 (1999). 24. Bernardi, P., Scorrano, L., Colonna, R., Petronilli, V. & Di Lisa, F. Mitochondria and cell death. Eur. J. Biochem. 264, 687–701 (1999). 25. Crompton, M. The mitochondrial permeability transition pore and its role in cell death. Biochem. J. 341, 233–249 (1999). 26. Eskes, R., Desagher, S., Antonsson, B. & Martinou, J.-C. Bid induces the oligomerization and insertion of Bax into the outer mitochondrial membrane. Mol. Cell. Biol. 20, 929–935 (2000). 27. Marzo, I. et al. Bax and adenine nucleotide translocator cooperate in the mitochondrial control of apoptosis. Science 281, 2027–2031 (1998). 28. Shimizu, S., Narita, M. & Tsujimoto, Y. Bcl-2 family proteins regulate the release of apoptogenic cytochrome c by the mitochondrial channel VDAC. Nature 399, 483–487 (1999). 29. Shimizu, S., Ide, T., Yanagida, T. & Tsujimoto, Y. Electrophysiological study of a novel large pore formed by Bax and the voltage-dependent anion channel that is permeable to cytochrome c. J. Biol. Chem. 275, 12321–12325 (2000). 30. Shimizu, S., Konishi, A., Kodoma, T. & Tsujimoto, Y. BH4 domain of antiapoptotic Bcl-2 family members closes voltage-dependent anion channel and inhibits apoptotic mitochondrial changes and cell death. Proc. Natl Acad. Sci. USA 97, 3100–3105 (2000). 31. Vander Heiden, M. G., Chandel, N. S., Schumacker, P. T. & Thompson, C. B. Bcl-xL prevents cell death following growth factor withdrawal by facilitating mitochondrial ATP/ADP exchange. Mol. Cell 3, 159–167 (1999). 32. Vander Heiden, M. G. et al. Outer mitochondrial membrane permeability can regulate coupled respiration and cell survival. Proc. Natl Acad. Sci. USA 97, 4666–4671 (2000). 33. Zhuang, J., Dinsdale, D. & Cohen, G. M. Apoptosis, in human monocytic THP.1 cells, results in the release of cytochrome c from mitochondria prior to their ultracondensation, formation of outer membrane discontinuities and reduction in inner membrane potential. Cell Death Differ. 5, 953–962 (1998). 34. Kroemer, G., Dallaporta, B. & Resche-Rigon, M. The mitochondrial death/life regulator in apoptosis and necrosis. Annu. Rev. Physiol. 60, 619–642 (1998). 35. Gross, A. et al. Biochemical and genetic analysis of the mitochondrial response of yeast to Bax and Bcl-xL. Mol. Cell. Biol. 20, 3125–3136 (2000). 36. Priault, M., Chaudhuri, B., Clow, A., Camougrand, N. & Manon, S. Investigation of Bax-induced release of cytochrome c from yeast mitochondria. Eur. J. Biochem. 260, 684–691 (1999). 37. Kissova, I. et al. The cytotoxic action of Bax on yeast cells does not require mitochondrial ADP/ATP carrier but may be related to its import to the mitochondria. FEBS Lett. 471, 113–118 (2000). 38. Priault, M., Camougrand, N., Chaudhuri, B., Schaeffer, J. & Manon, S. Comparison of the effects of Bax-expression in yeast under fermentative and respiratory conditions: investigation of the role of adenine nucleotides carrier and

cytochrome c. FEBS Lett. 456, 232–238 (1999). 39. Pastorino, J. G., Chen, S.-T., Tafani, M., Snyder, J. W. & Farber, J. L. The overexpression of Bax produces cell death upon induction of the mitochondrial permeability transition. J. Biol. Chem. 273, 7770–7775 (1998). 40. Bossy-Wetzel, E., Newmeyer, D. D. & Green, D. R. Mitochondrial cytochrome c release in apoptosis occurs upstream of DEVD-specific caspase activation and independently of mitochondrial transmembrane depolarization. EMBO J. 17, 37–49 (1998). 41. Zhivotovsky, B., Orrenius, S., Brustugun, O. T. & Doskeland, S. O. Injected cytochrome c induces apoptosis. Nature 391, 449–450 (1998). 42. Deshmukh, M. & Johnson, E. M. Evidence of a novel event during neuronal death: development of competence-to-die in response to cytoplasmic

cytochrome c. Neuron 21, 653–655 (1998). 43. Green, D. Apoptotic pathways: papers wraps stone blunts scissors. Cell 102, 1–4 (2000). 44. Desagher, S. et al. Bid-induced conformational change of Bax is responsible for mitochondrial cytochrome c release during apoptosis. J. Cell Biol. 144, 891–901 (1999). 45. Knudson, C. M. & Korsmeyer, S. J. Bcl-2 and Bax function independently to regulate cell death. Nature Genet. 16, 358–363 (1997). 46. Kane, D. J., Ord, T., Anton, R. & Bredesen, D. E. Expression of Bcl-2 inhibits necrotic neuronal cell death. J. Neurosci. Res. 40, 269–275 (1995).

Acknowledgements. We thank B. Antonsson, S. Desagher, M. Koco–Vilbois, K. Maundrell, S. Montessuit, O. Terradillos and X. Roucou for critical reading of the manuscript.

OPINION

The mitochondrion in apoptosis: how Pandora’s box opens Naoufal Zamzami and Guido Kroemer There is widespread agreement that mitochondria have a function in apoptosis, but the mechanisms behind their involvement remain controversial. Here we suggest that opening of a multiprotein complex called the mitochondrial permeability transition pore complex is sufficient (and, usually, necessary) for triggering apoptosis.

As with Pandora’s box, the mitochondrion is full of potentially harmful proteins and biochemical reaction centres. Loss of subcellular and submitochondrial compartmentalization, then, may liberate a flood of toxic compounds, such as reactive oxygen species and proteins that can activate catabolic hydrolases. In most

pathways leading to apoptosis, permeabilization of the inner and outer mitochondrial membranes (IMM and OMM, respectively) is critical1,2 (BOX 1). This culminates in the release of certain inactive caspase precursors (procaspases), cytochrome c (a caspase activator), Smac (second mitochondria-derived activator of caspase, also known as DIABLO, which is a caspase co-activator) and apoptosis-inducing factor (AIF; a nuclease activator), as well as a plethora of other proteins from the intermembrane space. All of these proteins must pass through the OMM3, but the IMM is also affected (although its permeabilization may be transient, and is partial for solutes up to 1.5 kDa, allowing proteins in the mitochondrial matrix to be retained).

Box 1 | Evidence for the involvement of mitochondria in controlling cell death • Mitochondrial membrane permeabilization (MMP), which can affect both the inner and outer mitochondrial membranes, precedes the signs of necrotic or apoptotic cell death, including the apoptosis-specific activation of caspases. • MMP is a more accurate predictive parameter for cell death than caspase activation (which often is not required for cell death to occur and, in contrast, may also participate in positive signalling). • Many pro-apoptotic proteins and second messengers act on mitochondria to induce MMP. Such signalling molecules include the pro-apoptotic members of the Bcl-2 family, phosphatases and kinases acting on Bcl-2-like proteins, as well as transcription factors (for example, p53 and TR3/Nur-77/NGFI-B). • Bcl-2 and related anti-apoptotic proteins are present in mitochondrial membranes and prevent apoptosis by suppressing MMP. • MMP-mediated release of caspase and nuclease activators is required for full-blown apoptotic cell death. Such activators are normally sequestered in the intermembrane space and include cytochrome c, Smac/DIABLO and apoptosis-inducing factor (AIF).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 6 7

Acknowledgements. We thank B. Antonsson, S. Desagher, M. Koco–Vilbois, K. Maundrell, S. Montessuit, O. Terradillos and X. Roucou for critical reading of the manuscript.

OPINION

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 6 7

PERSPECTIVES Many proteins from the Bcl-2/Bax family regulate apoptosis when locally present in mitochondrial membranes (see BOX 1 in the article by Martinou and Green on page 63). As a rule, close homologues of Bcl-2 (for example, Bcl-xL, Bcl-W, Mcl1 and A1) reside in the mitochondria and stabilize the barrier function of the mitochondrial membranes. Pro-apoptotic proteins can shuttle between a non-mitochondrial localization (the cytosol for Bax, Bad and Bid; microtubules for Bim and mitochondria). Upon induction of apoptosis, these proteins insert into mitochondrial membranes and cause them to be permeabilized.

The controversy surrounds the mode of action of these Bcl-2 family members2,4,5. Do they induce permeabilization of mitochondrial membranes in an autonomous fashion, without the need for interactions with sessile mitochondrial proteins? Or do they act on a multiprotein complex called the permeability transition pore complex (PTPC)? The PTPC regulates cell death

The PTPC — or mitochondrial megachannel — is formed at the contact site between the IMM and OMM (FIG. 1). Its core components are the adenine nucleotide translocator (ANT, found in the IMM) and the volt-

Box 2 | The mitochondrial permeability transition

Definition. The mitochondrial permeability transition involves a sudden (and initially reversible) increase in permeability of the IMM to solutes up to 1.5 kDa. It is commonly defined by its inhibition by cyclosporin A or derivatives of this compound that bind to mitochondrial cyclophilin (peptidyl-prolyl-cis–trans-isomerase) such as N-methyl-Val-4-cyclosporin A. Cyclosporin A-mediated inhibition of the permeability transition is transient (lasting 60 min). Regulation. The permeability transition pore complex (PTPC) functions as a sensor for: • Voltage: The PTPC decodes voltage changes into variations of the probability (the ‘gating potential’) at which pore opening occurs. Pore agonists shift the gating potential to more negative values (physiological = 200 mV, negative inside), favouring pore opening, whereas pore antagonists favour its closure. • Divalent cations: Matrix Ca2+ increases the probability of pore opening. Matrix Mg2+ or Mn2+, and external divalent metal ions including Ca2+ all decrease the probability of pore opening. • Matrix pH: The permeability transition pore is closed at neutral or acidic pH owing to reversible protonation of histidine residues and/or inhibition of the interaction between matrix cyclophilin and the ANT. Alkalinization is permissive for pore opening with a maximum effect at a matrix pH of ~7.3. • Thiol oxidation: Oxidation (disulphide formation) of a critical mitochondrial dithiol (presumably cysteine 56 of the ANT dimer) increases the probability of pore opening. The redox status of this dithiol is in equilibrium with that of matrix glutathione. • Oxidation/reduction state of pyridine nucleotides (NADH/NAD+ and NADPH/NADP+). Oxidation of pyridine nucleotides favours permeability transition. • ANT ligands: The endogenous ANT ligand ADP as well as bongkrekate inhibit permeability transition. Atractyloside, another ANT ligand, induces permeability transition. • Metabolites: Glucose and creatine inhibit permeability transition, presumably through their action on hexokinase and creatine kinase. Ubiquinone O (coenzyme Q) also inhibits permeability transition. Long-chain fatty acids, ceramide and ganglioside GD3 favour permeability transition. • Anti- and pro-apoptotic members of the Bcl-2 family.

Metabolic consequences. Full-blown permeability transition causes uncoupling of the respiratory chain with collapse of the electrochemical proton gradient ∆Ψm and cessation of ATP synthesis, matrix Ca2+ outflow, depletion of reduced glutathione, depletion of NADPH, hypergeneration of superoxide anion, and mitochondrial release of intermembrane proteins. Several of the consequences of permeability transition themselves favour opening of the permeability transition pore, implying that permeability transition is a self-amplifying process. Physiological function. Periodic reversible opening of the permeability transition pore allows for the release of Ca2+ from the mitochondrial matrix, thereby participating in Ca2+ homeostasis and/or the generation of Ca2+ waves (Ca2+-induced Ca2+-release). A role in neuronal plasticity has been suggested. The ANT/VDAC couple (and its interacting proteins hexokinase and creatine kinase) may also participate in regulating ATP/ADP transport/synthesis. Irreversible permeability transition triggers mitochondrial autophagy (a process by which cells digest parts of their cytoplasm), apoptosis or necrosis.

age-dependent anion channel (VDAC, located in the OMM). The VDAC is normally permeable to solutes of up to 5 kDa, thereby allowing the free exchange of respiratory-chain substrates such as NADH, FADH and ATP/ADP between the mitochondrial intermembrane space and the cytosol. In contrast, the IMM is almost impermeable — a feature that is essential for generation of the electrochemical proton gradient (∆ψm) used for oxidative phosphorylation. But the permeability of both membranes may suddenly increase in vitro — in isolated mitochondria, for instance, after the addition of atractyloside (which is a ligand of the ANT) — and this ‘permeability transition’6 is mediated by the PTPC (BOX 2). As well as the abundant IMM and OMM proteins — ANT and VDAC respectively — the PTPC contains the peripheral benzodiazepine receptor (located in the OMM), creatine kinase (located in the intermembrane space), hexokinase II (tethered to the VDAC on the cytosolic face of the OMM), cyclophilin D (located in the mitochondrial matrix), as well as Bax/Bcl-2-like proteins (FIG. 1). Direct interactions have been shown for the ANT and cyclophilin D, as well as between the ANT and VDAC6,7. As a consequence of such interactions, changes in the conformation of the ANT (which are indirectly affected by cyclosporin A, which affects the cyclophilin D–ANT interaction) may indirectly impinge on the function of the VDAC, or vice versa. So the PTPC may simultaneously control permeability of the OMM and the IMM. Alternatively, initial permeabilization of the IMM may cause the mitochondrial matrix to swell (through osmosis), leading to rupture of the OMM. (The IMM, with its folded christae, does not burst as its surface area is greater than that of the OMM.) Ultrastructural evidence for matrix swelling or rupture of the OMM has been obtained in many models of cell death, including hepatocyte apoptosis induced by injection of an antibody recognizing the CD95 ‘death receptor’ in the plasma membrane8. This particular phenotype is deficient in knockout mice that lack either of the two pro-apoptotic proteins from the Bcl-2 family (Bid or Bak) or overexpress a Bcl-2 transgene in the liver9. However, mitochondrial swelling and membrane rupture is not a universal feature of apoptosis10–13, suggesting alternative possibilities of membrane permeabilization, including the formation of pores in the OMM. Many reports show that apoptosis correlates with signs of a permeability transition, such as loss of the mitochondrial transmembrane potential (∆ψm); that induction of a www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

PERSPECTIVES permeability transition is enough to trigger cell death; and that, by inhibiting this permeability transition, cell death can be prevented (for reviews see REFS 6,14). For example, cyclosporin A (and N-methyl-4-Valcyclosporin A, which is a non-immunosuppressive derivative that acts on cyclophilin D) can inhibit apoptosis induced in vivo by brain trauma15, by ischaemia reperfusion damage6, or by injection of an antibody against CD95 (which mainly affects hepatocytes)8. Moreover, bongkrekate, an ANT ligand that inhibits the PTPC, prevents apoptosis induced by diverse stimuli including glucocorticoids (in thymocytes)14, excitotoxins (in neurons)16 and tumour necrosis factor (in hepatocytes)17. However, such pharmacological inhibitors are not universally cytoprotective, implying that permeabilization of the mitochondrial membrane may involve individual PTPC components (for example, the VDAC without the ANT/cyclophilin D) or occur in a completely PTPC-independent fashion. Several viral or bacterial proteins that modulate apoptosis also interact with components of the PTPC: for example, Vpr from HIV-1 (which binds the ANT and is pro-apoptotic)18; vMIA/UL37 from cytomegalovirus (which binds the ANT and is anti-apoptotic)19; the hepatitis virus B X-protein (which binds the VDAC and is pro-apoptotic)20; and porin B from Neisseria meningidis (which binds the VDAC and is anti-apoptotic)21. Bcl-2 and Bax interact with the PTPC

Co-immunoprecipitation studies indicate that Bcl-2 and Bax can interact — directly or indirectly — with the VDAC in the OMM22. Bcl-2, Bcl-xL, Bax and Bak interact directly with the ANT, as shown by three independent methods: co-purification, coimmunoprecipitation and yeast two-hybrid screening23. The two-hybrid screen revealed that a short stretch of human ANT2 (amino acids 105–156) suffices for the interaction with Bcl-2-related proteins23. The apoptosisregulatory potential of this portion of the ANT has been confirmed by deletion mapping: overexpression of full-length mouse Ant1 (which has 95% amino-acid identity with human ANT2) or its amino-terminal half (amino acids 1–141) induces apoptosis, but expression of an even shorter truncation mutation (amino acids 1–101) does not. So the Bcl-2/Bax-binding site (amino acids 105–156) overlaps with the apoptosis-regulatory region of ANT (amino acids 102–141), as well as with its Vpr-binding consensus motif (WXXF; amino acids 109–113). Interestingly, a substitution in the

Glucose-6phosphate + ADP

Glucose + ATP

Kinases Phosphatases pH change Caspases

HK II

HVB-X Porin B

BH3 peptides

Bax

Bcl-2

PBR OMM

VDAC

Creatine + ATP mtCK Creatine-P + ADP

vMIA Vpr

Atractyloside/palmitate Bongkrekate/ATP/ADP

ANT

∆ψm IMM

Cardiolipin Ca2+/oxidation?

Cyclosporin A Cyp-D

Figure 1 | Hypothetical molecular architecture of the permeability transition pore complex and its regulation. The permeability transition pore complex (PTPC) involves several transmembrane proteins: the adenine nucleotide translocator (ANT), the voltage-dependent anion channel (VDAC) and the peripheral benzodiazepine receptor (PBR). It also involves members of the Bax/Bcl-2 family, as well as associated proteins such as hexokinase II (HK II), mitochondrial creatine kinase (mtCK) and the peptidyl–prolyl isomerase cyclophilin D (Cyp-D). The VDAC functions as a nonspecific pore, allowing diffusion of solutes up to 5 kDa. The ANT is responsible for exchange of ATP and ADP on the inner membrane. The exact function of the PBR is unknown. Agents or metabolites labelled in green facilitate PTPC opening; agents in red inhibit pore opening. Proteins or peptides carrying the Bcl-2 homology region-3 (BH3) motif may act on either Bax or Bcl-2 (or their homologues) in the outer mitochondrial membrane. Ca2+ has been postulated to act on ANT-associated cardiolipin molecules; cyclosporin A acts on cyclophilin D. HIV-1 Vpr, the viral mitochondrial inhibitor of apoptosis (vMIA), hepatitis virus B X-protein (HVB-X), and Neisseria meningitidis porin B also act on the PTPC. (OMM, outer mitochondrial membrane; IMM, inner mitochondrial membrane; ∆ψm electrochemical proton gradient.)

highly conserved alanine residue 114 in ANT1 causes a dominant human mitochondriopathy24. Whether this pathology is related to a deficient control of apoptosis, however, remains unclear. There is still controversy about where the Bcl-2-like proteins act. Most authors suggest that these proteins are found mainly in the OMM, although a few reports indicate that Bcl-2 (REFS 25,26) and Bax23 are localized to the IMM. Pro-apoptotic stimuli may cause Bax to translocate from the OMM to the IMM23 or cause Bcl-2 to move from the IMM to the OMM26, using as-yetunknown mechanisms. It is also conceivable that protruding domains of Bcl-2 and Bax, which are anchored to the OMM, bind to the IMM within the OMM–IMM contact sites. Indeed, Bcl-2 is particularly enriched at these contact sites.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

Bcl-2 and Bax act on the PTPC

The biochemical features of pore-forming proteins can be studied by reconstituting them into synthetic lipid bilayers, either in proteoliposomes or in planar membranes. In response to various pro-apoptotic agents, including atractyloside23, Ca2+, the thiol crosslinker diamide27 and the HIV-1 protein Vpr18, ANT proteoliposomes become permeabilized to hydrophilic compounds with a relative molecular mass of less than 1.5 kDa. Proteoliposomes containing both the ANT and recombinant Bax show higher permeabilization responses to atractyloside23or Vpr18 than do proteoliposomes containing either ANT or Bax alone. In contrast, Bcl-2 prevents the ANT-mediated permeabilization of liposomes responding to atractyloside23 or Vpr18. Analogous results have been obtained for VDAC-containing proteoliposomes (which

VOLUME 2 | JANUARY 2001 | 6 9

PERSPECTIVES

Bcl-2

Bax

Nonspecific Autonomous channel formation sets off local ion imbalances and indirectly stabilizes OMM/IMM

Specific

Nonspecific

Cooperative channel formation within the PTPC or with PTPC components (ANT, VDAC), in OMM and/or IMM

Membrane insertion and oligomerization leads to formation of protein-permeable conduits in OMM

MMP and cell death

Figure 2 | Alternative hypotheses for the modus operandi of Bcl-2/Bax-like proteins. Although Bax and Bcl-2 heterodimerize and inhibit each other, both proteins can act independently in regulating apoptosis. According to one hypothesis (left), Bcl-2 forms protein-impermeable channels, thereby preventing local ion imbalances and non-physiological closure of the voltage-dependent anion channel (VDAC) in the outer mitochondrial membrane (OMM). Bcl-2 and Bax may also interact with the permeability transition pore complex (PTPC) and regulate channel formation by this two-membranespanning protein complex, regulating the permeability of the inner mitochondrial membrane (IMM) and OMM (middle). Bax-like proteins have also been proposed to oligomerize in the OMM, upon interaction with a truncated form of Bid known as tBid, thereby causing the formation of cytochrome c-permeable conduits. (ANT, adenine nucleotide translocator; MMP, mitochondrial membrane permeabilization.)

are permeable to sucrose)22. Bcl-2, as well as peptides derived from the BH4 domain of Bcl-2 (where BH4 stands for Bcl-2 homology domain 4, and is a domain that is missing in pro-apoptotic Bcl-2-like proteins), reduces the permeability of such VDAC proteoliposomes to sucrose, whereas Bax enhances their permeability22. Bax/VDAC liposomes (but not liposomes containing VDAC or Bax alone) are permeable to cytochrome c (14.5 kDa), but impermeable to a 50 kDa protein22.

nel is qualitatively different from that formed by Bax alone. With a combination of ANT and Bcl-2 (in a molar ratio of 1:1), atractyloside-induced ion movements are no longer seen, indicating the closure of both the ANT and the Bcl-2 channels28. Electrophysiological experiments also indicate cooperative pore formation by the VDAC plus Bax, and inhibition of VDAC channel formation by Bcl-2 (REF. 29). The VDAC and Bax cooperatively create a large pore, with conductance levels fourfold and

Bax, ANT and VDAC channels

Bax and the ANT (or VDAC) can cooperate to generate channels with higher conductance levels as well as higher opening probabilities than either of the two compounds alone. Moreover, Bcl-2 and the ANT (or VDAC) mutually inhibit the formation of ion channels. Single-channel current measurements involving proteins reconstituted into planar lipid bilayers indicate that the ANT can form large channels with multiple subconductance states (70 to 600 pS) in response to Ca2+. The ANT can also form small channels (30 pS) in response to atractyloside28. A mixture of Bax and the ANT (in a molar ratio of 1:4) has a higher probability of atractyloside-induced pore opening than the ANT alone and shows two conductance levels (30 and 80 pS), as well as cation specificity. At low concentrations (1 nM), Bax does not yield any pronounced macroscopic conductance, unless combined with ANT treated with atractyloside28. So channel formation is more efficient for the combination of the ANT, atractyloside and Bax than for combinations of the ANT plus atractyloside, the ANT and Bax, or atractyloside and Bax. Moreover, the channel formed by Bax at high concentrations is anion specific. So, the ANT/Bax chan-

“The intricate relationship between Bcl-2/Bax-like proteins and mitochondrial proteins adds another level of complexity to regulation of cell death.” tenfold greater than those of the VDAC and Bax channels, respectively. Although the VDAC and Bax channels both show ion selectivity and voltage-dependent modulation of their activity, the VDAC–Bax channel has neither of these properties. Cytochrome c reportedly passes through a single VDAC–Bax channel, but not through the VDAC or Bax channels in a planar lipid bilayer29. Anti-apoptotic Bcl-xL and its BH4 oligopeptide reportedly close the VDAC channel30. Studies in isolated mitochondria

When incorporated into mitochondrial membranes, Bcl-2 and its anti-apoptotic homologues enhance the resistance of these

mitochondria to a permeability transition. Similarly, microinjection of Bcl-2 into the cytoplasm of intact cells prevents both permeabilization of their mitochondrial membranes and nuclear apoptosis induced by the ANT ligand atractyloside31. If added to purified mitochondria in vitro, recombinant Bax or Bak induce membrane permeabilization, and this is inhibited by various PTPC inhibitors, including Koenig’s polyanion, cyclosporin A, N-methyl-4-Val-cyclosporin A and bongkrekate23,31–33. These are controversial findings, however, as some groups10,11 have not detected any effect of cyclosporin A. Oligomycin, an inhibitor of the F0–F1-ATPase (which, on theoretical grounds, might indirectly affect the ANT), also inhibits mitochondrial membrane permeabilization induced by Bax31–33. At low doses, Bax causes signs of outermembrane permeabilization (that is, release of cytochrome c) but not of inner-membrane permeabilization (for instance, matrix swelling and loss of the ∆ψm). But higher doses of Bax affect both the OMM and the IMM33, and cyclosporin A prevents the effects of Bax on both membranes23,31–33. Studies in intact cells

Similar inhibitory profiles to those seen in vitro have been obtained when Bax is microinjected into the cytoplasm of intact cells. In control cells, Bax causes loss of the ∆Ψm and nuclear apoptosis. Our group has found23 that treatment with cyclosporin A, N-methyl-4-Val-cyclosporin A or bongkrekate abolishes both the mitochondrial and the nuclear signs of Bax-induced apoptosis. However, other data10 show that cyclosporin A does not inhibit apoptosis (and the associated mitochondrial membrane permeabilization) induced by transfection with the Bax gene. Proteolytic activation of Bid by caspase-8 (yielding a truncated form of Bid known as tBid) normally leads to the permeabilization of mitochondrial membranes. This permeabilization (and subsequent apoptosis) can be prevented by the ANT-targeted viral mitochondrial inhibitor of apoptosis (vMIA)19. Killing by another pro-apoptotic protein from the Bcl-2 family, BNIP3, is prevented by cyclosporin A and bongkrekate34. Obviously, however, it may be argued that pharmacological inhibition experiments are not truly conclusive — drugs may affect not only components of the PTPC, but also other molecules in the cell — and that genetic interventions on components of the PTPC would be more informative. In yeast cells, several mitochondrial defects www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

PERSPECTIVES reduce Bax-induced killing, including deletion of the three ANT isoforms23,35, and mutants inactivating the F0–F1-ATPase35,36. These data do not support the idea that Bax kills cells through simple lysis, indicating that Bcl-2-like proteins might modulate cell death through a functional interaction with the PTPC.

cell type. Whatever the answer to this conundrum, it seems clear that the intricate relationship between Bcl-2/Bax-like proteins and mitochondrial proteins adds another level of complexity to regulation of cell death. Naoufal Zamzami and Guido Kroemer are at the Centre National de la Recherche Scientifique, UMR1599, Institut Gustave Roussy, 39 rue Camille-Desmoulins, F-94805 Villejuif, France. Correspondence to G.K. e-mail: kroemer@igr.fr

Do Bax/Bcl-2 always act on the PTPC?

Recombinant Bax, as well as tBid (but not Bcl-2), has been reported to destabilize lipid bilayers without causing the appearance of ion channels37,38. Alternatively, Bcl-2, Bax and tBid may form channels, with defined levels of conductivity and variable requirements of pH or the local presence of acidic phospholipids39,40. When added to liposomes, Bax can insert into the membranes, oligomerize (presumably to tetramers) and form cytochrome c-permeant conduits with an estimated pore size of around 30 Å (REF. 40). It has been proposed that Bid aids the insertion or oligomerization of Bax12 or Bak13, which would form pores without any interaction with VDAC or ANT. Moreover, recombinant tBid (in contrast to Bax or Bak), added to isolated mitochondria, has been reported41 to cause an exclusive permeabilization of the OMM that is not affected by PTPC inhibitors. All these data may be interpreted to mean that proteins from the Bcl-2/Bax family can permeabilize membranes through a nonspecific effect — that is, without the need for interactions with other proteins from the PTPC. This interpretation is supported by the observation that, during apoptosis, cytochrome c can be released from mitochondria that have an apparently normal ultrastructure and conserve an inner transmembrane potential11,12. How can we reconcile the evidence for these nonspecific effects of Bax/Bcl-2 with the many reports that suggest specific, PTPCdependent effects? One possibility would be to assume that, at physiological concentrations, Bax and Bcl-2 act through these specific effects; for instance, by modulating the probability of channel formation by other proteins. But at higher concentrations, Bax would become able to oligomerize and to exert autonomous, nonspecific effects, involving either channel formation or membrane destabilization. This implies a hierarchic superposition of two levels of regulation. Alternatively, the two pathways could come into action as a function of the local abundance of PTPC components, their isoforms, or PTPC inhibitors/activators (FIG. 2). Indeed, a similar mechanism has been proposed for CD95induced signalling, in which the same primary signal can trigger cell death through completely distinct mechanisms, as a function of the

Links DATABASE LINKS Smac | Bcl-2 | Bcl-xL | Bcl-W |

Mcl1 | A1 | Bax | Bad | Bid | Bim | creatine kinase | hexokinase II | cyclophilin D | CD95 | ANT2 | mouse Ant1 | BNIP3 ENCYCLOPEDIA OF LIFE SCIENCES Apoptosis: molecular mechanisms 1. 2. 3.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

Green, D. R. & Reed, J. C. Mitochondria and apoptosis. Science 281, 1309–1312 (1998). Kroemer, G. & Reed, J. C. Mitochondrial control of cell death. Nature Med. 6, 513–519 (2000). Budijardjo, I., Oliver, H., Lutter, M., Luo, X. & Wang, X. Biochemical pathways of caspase activation during apoptosis. Annu. Rev. Cell Dev. Biol. 15, 269–290 (1999). Vander Heiden, M. G. & Thompson, C. B. Bcl-2 proteins: Inhibitors of apoptosis or regulators of mitochondrial homeostasis? Nature Cell Biol. 1, E209–E216 (1999). Gross, A., McDonnell, J. M. & Korsmeyer, S. J. Bcl-2 family members and the mitochondria in apoptosis. Genes Dev. 13, 1899–1911 (1999). Crompton, M. The mitochondrial permeability transition pore and its role in cell death. Biochem. J. 341, 233–249 (1999). Woodfield, K., Ruck, A., Brdiczka, D. & Halestrap, A. P. Direct demonstration of a specific interaction between cyclophilin-D and the adenine nucleotide translocase confirms their role in the mitochondrial permeability transition. Biochem. J. 336, 287–290 (1998). Feldmann, G. et al. Opening of the mitochondrial permeability transition pore causes matrix expansion and outer membrane rupture in Fas-mediated hepatic apoptosis in mice. Hepatology 31, 674–683 (2000). Yin, X.-M. et al. Bid-deficient mice are resistant to Fasinduced hepatocellular apoptosis. Nature 400, 886–891 (1999). Eskes, R. et al. Bax-induced cytochrome c release from mitochondria is independent of the permeability transition pore but highly dependent on Mg2+ ions. J. Cell Biol. 143, 217–224 (1998). von Ahsen, O. et al. Preservation of mitochondrial structure and function after Bid- or Bax-mediated cytochrome c release. J. Cell Biol. 150, 1027–1036 (2000). Eskes, R., Desagher, S., Antonsson, B. & Martinou, J. C. Bid induces the oligomerization and insertion of Bax into the outer mitochondrial membrane. Mol. Cell. Biol. 20, 929–935 (2000). Wei, M. C. et al. tBID, a membrane-targeted death ligand, oligomerizes BAK to release cytochrome c. Genes Dev. 14, 2060–2071 (2000). Kroemer, G., Dallaporta, B. & Resche-Rigon, M. The mitochondrial death/life regulator in apoptosis and necrosis. Annu. Rev. Physiol. 60, 619–642 (1998). Sullivan, P. G., Thompson, M. B. & Scheff, S. W. Cyclosporin A attenuates acute mitochondrial dysfunction following traumatic brain injury. Exp. Neurol. 160, 226–234 (1999). Budd, S. L., Tenneti, L., Lishnak, T. & Lipton, S. A. Mitochondrial and extramitochondrial apoptotic signaling pathways in cerebrocortical neurons. Proc. Natl Acad. Sci. USA 97, 6161–6166 (2000). Tafain, M., Schneider, T. G., Pastorino, J. G. & Farber, J. L. Cytochrome c-dependent activation of caspase-3 by tumor necrosis factor requires induction of the mitochondrial permeability transition. Am. J. Pathol. 156, 2111–2121 (2000). Jacotot, E. et al. The HIV-1 viral protein R induces apoptosis via a direct effect on the mitochondrial permeability transition pore. J. Exp. Med. 191, 33–45 (2000). Goldmacher, V. S. et al. A cytomegalovirus-encoded

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

mitochondria-localized inhibitor of apoptosis structurally unrelated to Bcl-2. Proc. Natl Acad. Sci. USA 96, 12536–12541 (1999). Rahmani, Z., Huh, K. W., Lasher, R. & Siddiqui, A. Hepatitis B virus X protein colocalizes to mitochondria with a human voltage-dependent anion channel, HVDAC3, and alters its transmembrane potential. J. Virol. 74, 2840–2846 (2000). Massari, P., Ho, Y. & Wetzler, L. M. Neisseria meningitidis porin PorB interacts with mitochondria and protects cells from apoptosis. Proc. Natl Acad. Sci. USA 97, 9070–9075 (2000). Shimizu, S., Narita, M. & Tsujimoto, Y. Bcl-2 family proteins regulate the release of apoptogenic cytochrome c by the mitochondrial channel VDAC. Nature 399, 483–487 (1999). Marzo, I. et al. Bax and adenine nucleotide translocator cooperate in the mitochondrial control of apoptosis. Science 281, 2027–2031 (1998). Kaukonen, J. et al. Role of adenine nucleotide translocator 1 in mtDNA maintenance. Science 289, 782–785 (2000). Motoyama, S. et al. Bcl-2 is located predominantly in the inner membrane and crista of mitochondria in rat liver. Biochem. Biophys. Res. Commun. 249, 628–636 (1998). Gotow, T. et al. Selective localization of Bcl-2 to the inner mitochondrial and smooth endoplasmic reticulum membranes in mammalian cells. Cell. Death Differ. 7, 666–674 (2000). Costantini, P. et al. Oxidation of a critical thiol residue of the adenine nucleotide translocator enforces Bcl-2independent permeability transition pore opening and apoptosis. Oncogene 19, 307–314 (2000). Brenner, C. et al. Bcl-2 and Bax regulate the channel activity of the mitochondrial adenine nucleotide translocator. Oncogene 19, 329–336 (2000). Shimizu, S., Ide, T., Yanagida, T. & Tsujimoti, Y. Electrophysiological study of a novel large pore formed by Bax and the voltage-dependent anion channel that is permeable to cytochrome c. J. Biol. Chem. 275, 12321–12325 (2000). Shimizu, S., Konishi, A., Kodama, T. & Tsujimoto, Y. BH4 domain of antiapoptotic Bcl-2 family members closes voltage-dependent anion channel and inhibits apoptotic mitochondrial changes and cell death. Proc. Natl Acad. Sci. USA 97, 3100–3105 (2000). Jürgensmeier, J. M. et al. Bax directly induces release of cytochrome c from isolated mitochondria. Proc. Natl Acad. Sci. USA 95, 4997–5002 (1998). Narita, M. et al. Bax interacts with the permeability transition pore to induce permeability transition and cytochrome c release in isolated mitochondria. Proc. Natl Acad. Sci. USA 95, 14681–14686 (1998). Pastorino, J. G. et al. Functional consequences of sustained or transient activation by Bax of the mitochondrial permeability transition pore. J. Biol. Chem. 274, 31734–31739 (1999). Vande Velde, C. et al. BIP3 and genetic control of necrosis-like cell death through the mitochondrial permeability transition pore. Mol. Cell. Biol. 20, 5454–5468 (2000). Harris, M. H., Vander Heiden, M. G., Kron, S. J. & Thompson, C. B. Role of oxidative phosphorylation in Bax toxicity. Mol. Cell. Biol. 20, 3590–3596 (2000). Matsuyama, S., Xu, Q., Velours, J. & Reed, J. C. Mitochondrial F0F1-ATPase proton pump is required for function of proapoptotic protein Bax in yeast and mammalian cells. Mol. Cell 1, 327–336 (1998). Basañez, G. et al. Bax, but not Bcl-xL, decreases the lifetime of planar phospholipid bilayer membranes at subnanomolar concentrations. Proc. Natl Acad. Sci. USA 96, 5492–5497 (1999). Kudla, G. et al. The destabilization of lipid membrane induced by the C-terminal fragment of caspase 8cleaved Bid is inhibited by the N-terminal fragment. J. Biol. Chem. 275, 22713–22718 (2000). Schendel, S. L. et al. Ion channel activity of the BH3 only bcl-2 family member, BID. J. Biol. Chem. 274, 21932–21936 (1999). Saito, M., Korsmeyer, S. J. & Schlesinger, P. H. Baxdependent transport of cytochrome c reconstituted in pure liposomes. Nature Cell Biol. 2, 553–555 (2000). Shimizu, S. & Tsujimoto, Y. Proapoptotic BH3-only Bcl-2 family members induce cytochrome c release, but not mitochondrial membrane potential loss, and do not directly modulate voltage-dependent anion channel activity. Proc. Natl Acad. Sci. USA 97, 577–582 (2000).

Acknowledgements The authors’ work is supported by a special grant from the Ligue Nationale contre le Cancer, as well as grants from ANRS, FRM and EC (to G.K.).

VOLUME 2 | JANUARY 2001 | 7 1

PERSPECTIVES

TIMELINE

Walther Flemming: pioneer of mitosis research Neidhard Paweletz The German anatomist Walther Flemming began his pioneering studies of mitosis almost 150 years ago. What were his achievements, and where have his discoveries led?

Browsing through the latest issues of cell and molecular biology journals, it is striking how many cover pages show images of dividing cells. This reflects the fact that research into cell division is at the forefront of the field. But what are the origins of this discipline? It began in the seventeenth century, when Hooke1, van Leeuwenhoek2 and others discovered the cellula as a building block of many organisms. Then, in the first half of the nineteenth century, Schleiden3 and Schwann4 established the ‘cell theory’, according to which all organisms are composed of tiny units, the cells. Schleiden and Schwann assumed that cells are formed de novo from an intercellular substance in some kind of crystallization (‘free cell formation’) — an assumption that misled many scientists and inhibited research into cell division for almost three decades. For example, in 1875 Strasburger5 published a comprehensive book Ueber Zellbildung und Zelltheilung (“About cell formation and cell division”) in which he defended free cell formation. However, he had abandoned this idea by the time the third edition of his book was published in 1880. By the 1870s, some scientists (such as Dumortier6, von Mohl7, Remak8 and others) had shown that cells multiply by binary fission. At this time, Strasburger’s colleague (and competitor) Walther Flemming (FIG. 1) was beginning detailed studies on dividing cells in different organs and organisms, mainly from the animal kingdom. Flemming’s studies were not hampered by the idea of free cell formation, which he no longer believed in, and they eventually led to a solid foundation for modern cellular and molecular biology. Flemming’s career

Walther Flemming was born on 21 April 1843, in Sachsenberg/Mecklenburg in Germany. His father, Carl Friedrich, was a

Figure 1 | Portrait of Walther Flemming. A welldocumented appreciation of Flemming’s work is given in The Birth of the Cell by Henry Harris36. (Image provided by the Science Photo Library.)

famous psychiatrist and neurologist. Flemming grew up in Sachsenberg as a shy but intelligent boy. Although his favourite topics were literature and philology, he decided to study medicine. He began his studies at the University of Göttingen and continued in Tübingen, Berlin and Rostock. During his training in the clinic at Rostock, Flemming studied histological and zoological preparations under the guidance of Franz Eilhard Schulze (who was himself strongly influenced

by his mentor, Max Schultze, one of the first ‘cell biologists’). From Schulze, Flemming learned constructive criticism, the cautious evaluation of results and the avoidance of speculation — all of which were characteristic of his later scientific work. Other features of his research included careful observation, frequent controls and a thorough evaluation of all results. Flemming was also influenced by Rudolf Virchow, one of his academic teachers, and Max Schultze’s students Wilhelm Kühne and Gustav Schwalbe, who implanted in him the idea of the cell as the fundamental, autonomous unit of life. For short periods Flemming assisted in anatomy and histology in Würzburg and Amsterdam until, in 1870, he was offered the position of Prosektor (leader of dissections and anatomical preparations) in Rostock. He also taught histology and comparative anatomy, and his students were enthusiastic about his talent for drawing, which brought cells, organs or organisms to life on the blackboard. Indeed, all of his later publications were illustrated by fine detailed drawings that aided understanding (FIG. 2). At the end of 1870 he presented his Habilitation thesis about connective substances and the vessel wall in molluscs, to become Privatdozent (academic teacher). In February 1872 the head of anatomy at Rostock, Wilhelm Henke, asked Flemming to go with him to the German University of Prague, where Flemming was responsible for all histological lectures, seminars and courses. Here, in the same institute as Johannes Evangelista Purkinje, who was considered the father of histology, Flemming began his detailed investigations into cell division. Since the German revolution of 1848, nationalism had been growing all over Europe, and Czech students passionately demanded a Czech University in Prague. So the climate became increasingly hostile until most German professors preferred to return to Germany. Although Flemming was not called to the Chair at Königsberg (East

Box 1 | Cytoplasm and mitochondria One of Flemming’s favourite topics was the structure and function of the protoplasm. During his careful observations, particularly in the 1880s (REF. 37), he used optimal fixations and different staining procedures to show that the protoplasm has a mainly filamentous appearance; this contradicted the widely accepted proposal by Carl Frommann and Karl Heitzmann of a granular and reticular substructure. Flemming defended his Filartheorie (“theory of a filamentous structure”) vigorously, and surrendered only when he was too ill and weak. In 1898, however, Carl Benda used a special fixation and staining method to show elongated corpuscles in the protoplasm. He termed these mitochondria because of their tendency to form chains. Flemming’s assistant Friedrich Meves38 later showed, shortly after Flemming’s death, that Flemming was not completely wrong — Meves identified Flemming’s ‘filaments’ and Benda’s mitochondria as one and the same.

www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

PERSPECTIVES

Figure 2 | Illustration from Zellsubstanz, Kern und Zelltheilung22.

Prussia), as he had hoped, he was recruited to the vacant Chair of Anatomy at Kiel (Schleswig-Holstein). Almost all the medical faculty voted for Flemming. However, during negotiations, one faculty member strongly recommended Friedrich Merkel, an anatomist from Rostock, who was the sonin-law of a well-known German anatomist. Nevertheless, Flemming took up the position in February 1876. Although the Christiana Albertina University in Kiel was very small, the old institute of anatomy was not big enough for the 70 or so medical students. There was not enough money to buy new microscopes and other equipment, and, at the beginning, Flemming took charge of all lectures, seminars and courses without assistance. He had to do battle with the university’s administration; these struggles for resources were a heavy burden for Flemming, a peace-loving man whose students loved him for his cordiality and benevolence. In his late forties, Flemming developed a severe neurological disease from which he did not recover. At the turn of the century, his illness became so severe that he had to retire and, on 4 August 1905, he died aged 62 in Kiel9. By this time, however, Flemming’s institute had become a leading centre for research into histology, cytology, comparative anatomy and, in particular, mitosis. Initial studies

When Flemming began his research, cell biology was just beginning to boom (TIMELINE). In 1833, even before Schleiden and Schwann had presented their cell theory3,4, Robert Brown10 had described an ovoid in the cell as the “nucleus”, and Dumortier6 and von Mohl7 had discovered binary fission of the nucleus and cell. Remak8 gave the first descriptions of the changes that occur in the nucleus, and Purkinje11 underlined its importance and the requirement for this organelle throughout the life of a cell. But in 1868, at the beginning of his career, Flemming — whose knowledge of histology was derived mainly from zoological

objects — was interested in the sensory organs of molluscs. He also studied adipose tissue, and clearly stated its character as connective tissue; before this, adipose tissue had been considered to be a separate organ. Flemming also analysed lipid droplets as

products of cellular metabolism. In addition, he was interested in the involution of adipose tissue, and studied the fine structure of the fibres of connective tissue and their swelling during treatment with acids. At a time when the focus of Flemming’s interest was still the behaviour of individual cells, research into the process of cell division had already begun. In 1873, Schneider12 sketched the important steps of cell division. He saw the transformation of the nucleus into rod-like structures (Stäbchen), which assembled in the centre of the cell (at what we now know as the metaphase plate). At a stage that we now call anaphase, two groups of Stäbchen could be seen in the elongated cell. Between 1874 and 1876, Flemming described these steps in more detail13–15. Whereas Schneider12 had postulated that the nucleus undergoes deformation during cell

Progressive phase

Regressive phase

Mother nucleus

Daughter nuclei

Scaffold of the resting nuclei (interphase nuclei)

Scaffold of the resting nucleus (interphase nucleus)

Skein of fine threads (prophase) Spirem

Skein of fine threads (reconstruction phase) Dispirem

Thickenings of the threads and loosening of the skein (late prophase)

Condensation on the skein (telophase)

Star-like configuration of threads (prometaphase) Aster

Two star-like arrangements of the threads (anaphase) Dyaster

Equatorial plate (metaphase) Metakinese

Figure 3 | The progressive and regressive phases of cell division. Mitosis starts with the skein-like form of the nuclear threads (prophase), which changes into the aster (star-like configuration of the threads at prometaphase). This stage moves into the equatorial plate (metaphase), which then immediately forms the double star (anaphase). When the threads have reached the position of the daughter-cell nucleus, the double skein (telophase) can be observed. (Images reproduced from REF. 22).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | JANUARY 2001 | 7 3

PERSPECTIVES

Timeline | The origins of research into mitosis Dumortier and von Mohl discover that cell multiplication occurs by binary fission.

Hooke discovers that cork is composed of little chambers, which he calls cellula (cells).

1665

1682

1832/35

Van Leeuvenhoek observes levende dierkens (small living animals) in infusions of organic matter.

Remak recognizes deformations of the nucleus as preparation for division.

1833

Brown sees ovoid bodies in cells and coins the term ‘nucleus’.

multiplication, Flemming showed that the scaffold and network within the nucleus transformed into ‘threads’, which then separated into two groups. These two groups, in turn, formed two skeins, from which the scaffold of the nuclei reappeared. By carefully studying wounds and scars, Flemming and his students found an accumulation of dividing cells in these tissues, and concluded that the regeneration of tissues and organs occurs by cell division16. At that time, no general repertoire of histological methods existed — indeed, one of the first monographs on histological methods, by Alfred Fischer17, was not published until the end of the nineteenth century. In this book, many studies of fixed cells were considered to be based on artefacts, so Flemming had to spend a long time designing methods to facilitate his observations18,19. He experimented with various acids to find an appropriate fixative for preserving the fine structure that he had seen in the living cells and finally used a mixture of chromic, osmic and glacial acetic acids, which was soon adopted by colleagues and known as ‘Flemming’s solution’. He tested haematein and haematoxylin for their usefulness as dyes, and also found that the addition of very low concentrations of picric, acetic or formic acid to the medium best brought out the structures of the nuclear scaffold and the fine structure of the protoplasm (cytoplasm; BOX 1). Nuclear division

In 1878 and 1879, Flemming published two important papers20,21, in the second of which he coined the term ‘indirect nuclear division’ because he had observed that a transformation of the nuclear content had to take place before fission could occur. (A cleavage of the nucleus and protoplasm — which, until then, had been generally assumed — was called ‘direct nuclear

1835

Schneider shows mitotic figures in spermatogenesis of platyhelminths.

1838/39

Schleiden and Schwann state that all plants and all animals are composed of cells. De novo formation of cells from intercellular substance.

1873

Strasburger presents the detailed drawings of dividing plant cells but still sticks to de novo cell formation.

1874–1876

1875

Flemming gives first descriptions of cell division in animals.

division’.) The methods that Flemming had developed allowed him to recognize a fibrous scaffold in the nucleus, which could easily be stained and was therefore named Chromatin (‘stainable material’). Some other structures remained unstained and were therefore termed Achromatin. These results led, in 1882, to the publication of Flemming’s comprehensive book Zellsubstanz, Kern und Zelltheilung (‘Cell substance, nucleus and cell division’)22, which became the foundation for all further research into mitosis. Although Schleicher23 had proposed the name Karyokinesis for this process, Flemming decided to use a more exact term, and he called the observed alterations within the nucleus Karyomitosis (meaning threadlike metamorphosis of the nucleus). He christened the arrangements of the nuclear threads Mitosen. Only afterwards, in 1888, did Heinrich Wilhelm Waldeyer24 coin the term Chromosomen (‘chromosomes’, meaning stainable bodies) for Flemming’s nuclear threads. Flemming described the processes in the nucleus as we know them today, and he made a distinction between the ‘progressive’ and ‘regressive’ phases of cell division (FIG. 3). The progressive phase started with the appearance of the threads in the nucleus of the mother cell and continued as far as the arrangement of the threads in the centre of the cell. The regressive phase, by contrast, began with the separation of the threads into two groups and ended with the reappearance of the daughter nuclei. Although Flemming had the correct idea that the chromatin network in the ‘resting’ nucleus transforms into the threads (chromosomes) — thereby representing continuity of the nuclear material — he did not have the techniques or equipment to prove this. The objective lenses of his microscope were composed of lenses with different refractive indices, but these lenses contained many aber-

Flemming decides to name cell division indirekte Zellteilung (indirect cell division) to distinguish it from direkte Zellteilung (direct cell division), which is less frequent. The scaffold in the nucleus is the Chromatin.

1876

Bütschli detects fine filaments especially at the poles; the spindle is recognized.

1879

1882

Flemming (temporarily) summarizes his results in a book. The term mitosis for indirect nuclear division is born.

rations — in particular, the chromatic aberration often delivered structures with coloured halos. Moreover, the illumination was not yet very bright and depended strongly on the intensity of the daylight. The microscopes had no sophisticated condenser systems, so it was not possible to produce a pseudo-phase-contrast image. But Flemming’s drawings clearly showed correct images of the spindle apparatus, for example. In 1891, Flemming published a paper25 describing the remnants of the spindle just before complete cleavage. He called this the Mittelkörper or midbody and considered it to be an equivalent of the cell plate in plant cells. Otto Bütschli had shown earlier26 that a fibrillar structure becomes visible, which he called the pole aster. Edouard van Beneden27 and, almost simultaneously but independently, Theodor Boveri28 had found a tiny structure at the pole, which they both termed the Polkörperchen (polar body), but they had assumed that this formed de novo during cell division. Also in 1891, in a sensational paper 29, Flemming showed unequivocally that this body is not formed anew but persists, and he coined the term Zentralkörperchen (central body) or Zentriol (centriole). He was convinced that the filamentous structure of the spindle in mitosis was responsible for transport of the threads, but again he could not prove this. His delicate observations on the behaviour of spindle fibres were later confirmed by electron microscopy. Division during development

In his attempts to present a general interpretation of mitosis that was valid for all organisms, Flemming also studied division during the development of spermatozoa; he described this in a lecture in 1888 (REF. 30). Although Flemming failed to recognize the www.nature.com/reviews/molcellbio

| JANUARY 2001 | VOLUME 2

PERSPECTIVES

Waldeyer observes the stainability of the nuclear ‘threads’ during division. Thay are named Chromasomen (stainable bodies).

Flemming proves the presence of the small polar body in the ‘resting’ cell as well as in the dividing cell and names it Zentriole.

Pernice recognizes colchicine as a mitosis inhibitor39.

8. 9.

10.

1883-87

1888

1889

1891

1905

11. 12.

Van Beneden and Boveri discover the centrosphere during division. Boveri calls it the centrosome.

Flemming studies division in spermatogonia.

Flemming discovers the ‘midbody’.

Farmer and Moore study the ‘reduction divisions’ of gametes and call these divisions maiosis.

13.

14.

15.

differences between the division of somatic cells and that of gametes, as he reported in his paper of 1882 (REF. 31) he had already observed the paired nature of the chromosomes in the early stages of spermatozoan development. In 1905, Farmer and Moore32 reported the first descriptions of maiosis. Strasburger8 assumed that the rod-like structures (chromosomes) were transversely split, and this was a source of strong controversy between him and Flemming. Flemming insisted — and could prove — that, in Metakinese or earlier, the threads were split longitudinally. He had already assumed31,33,34 that one half of this longitudinally split pair was destined for one daughter cell, whereas the second half went to the other daughter — a prediction that has turned out to be correct. Consequences of Flemming’s findings

A host of papers appeared over the two or three decades after Flemming published his spectacular book on mitosis22. But research into mitosis then slowed down until around the 1920s, once Alfred Fischer’s book17 had warned about the danger of studying artefacts caused by fixation and staining. For example, for some time the spindle fibres had been considered to be coagulation artefacts produced by fixation. In the mid-1920s, Karl Belar experimented with dividing spermatocytes to find out the mechanics of chromosome transport and, in 1929, he proposed the stem body hypothesis35. A few years before the Second World War, a new age of mitosis research began. This was interrupted by the war — especially by the holocaust and the emigration of many Jewish scientists from Germany. Flemming could not have foreseen the variety of disciplines that have come out of his work. First, of course, are the fields that are closely connected with the original

mitosis research. Chromosome structure and function has become a special branch of this, leading to investigations of kinetochores and telomeres for example, and even to the discovery of the function of the nucleolus. The combination of mitosis research with breeding experiments to explain Mendelian inheritance finally resulted in genetics and cytogenetics, which, in turn, led to gene manipulation, gene therapy, mutation research and the deciphering of the genetic code. The spindle is still a structure of interest. Research is being done into its behaviour during division, into its function as an apparatus for transporting chromosomes, microtubules, tubulin, microtubule-associated proteins and motor proteins, and into ciliary movements, the centrosome (centriole) and mitotic poisons (used as cytostatic agents). Other fields include the ‘uncontrolled’ growth of cancer, and cell-cycle regulation. Last, Flemming’s research has also led indirectly to studies into programmed cell death, which starts with drastic changes in nuclear structure and cell-cycle regulation. Neidhard Paweletz is at Wilhelmsfelder Strasse 47/1, D-69118 Heidelberg, Germany. e-mail: 100.272955@germanynet.de

16.

17. 18.

19.

20.

21.

22. 23.

24.

25. 26.

27. 28. 29.

30. 31.

32.

Links 33.

ENCYCLOPEDIA OF LIFE SCIENCES Theodor

Ambrose Hubert Schwann | Matthias Jacob Schleiden 1. 2. 3. 4.

5. 6.

Hooke, R. Micrographia (London, 1665). van Leeuwenhoek, A. Letter no. 35, March 3, 1682. Schleiden, M. J. Beiträge zur Phytogenesis. Müller´s Arch. Anat. Physiol. Wiss. Med. 136–176 (1838). Schwann, T. Mikroskopische Untersuchungen über die Übereinstimmung in der Struktur und dem Wachsthum der Thiere und Pflanzen (Verlag der Sander´schen Buchhandlung, Berlin, 1839). Strasburger, E. Über Zellbildung und Zelltheilung (Hermann Dabis, Jena, 1875). Dumortier, B. C. Nova Acta Phys. -Med. Acad. Caesar.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

34.

35. 36. 37. 38. 39.

Leopold–Carolinae Nat. Curios (part 1) 16, 217 (1832). (Cited in reference 36.) von Mohl, H. Über die Vermehrung der Pflanzenzellen durch Theilung (Inaugural dissertation, Tübingen 1835). Remak, R. Untersuchungen über die Entwickelung der Wirbelthiere (G. Reimer, Berlin 1855). Peters, G. Walther Flemming (1843–1905): sein Leben und sein Werk. Inaugural dissertation (doctoral thesis), Kiel (1967). Brown, R. Observations on the organs and mode of fecundation in Orchidae and Asclepiadeae. Trans. Linn. Soc. (Lond.) 16, 685–743 (1833). Purkinje, J. E. Symbolae ad Ovi Avium Historiam ante Incubationem (Leopold Voss, Leipzig, 1830). Schneider, A. Untersuchungen über Plathelminthen. Jahrb. Oberhess. Ges. Naturwiss. Heilk. 14, 69–81 (1873). Flemming, W. Über die ersten Entwicklungserscheinungen am Ei der Teichmuschel. Arch. Mikrosk. Anat. 10, 257–292 (1874). Flemming, W. Studien in der Entwicklungsgeschichte der Najaden. Sitzungsber. Kaiserl. Akad. Wiss. 71, 81–212 (1875). Flemming, W. Beobachtungen über die Beschaffenheit des Zellkerns. Arch. mikrosk. Anat. 13, 693–717 (1876). Flemming, W. Studien über Regeneration der Gewebe. Aus dem Anatomischen Institut Kiel 4–22, 23–42, 60–65, 76–102 (Bonn, 1885). Fischer, A. Fixierung, Färbung und Bau des Protoplasmas (Fischer, Jena, 1899). Flemming, W. Über das E. Hermannsche Kernfärbungsverfahren. Arch. mikrosk. Anat. 19, 317–330 (1881). Flemming, W. Über die Wirkung von Chrom–Osmium–Essigsäure auf Zellkerne. Arch. mikrosk. Anat. 45, 162–166 (1895). Flemming, W. Zur Kenntnis der Zelle und ihrer Theilungserscheinungen. Schr. naturwiss. Verein Schleswig-Holstein 3, 23–27 (1878). Flemming, W. Ueber das Verhalten des Kerns bei der Zellteilung und über die Bedeutung mehrkerniger Zellen. Arch. Pathol. Anat. 77, 1–28 (1879). Flemming, W. Zellsubstanz, Kern und Zelltheilung (F. C. W. Vogel, Leipzig, 1882). Schleicher, W. Die Knorpelzellteilung. Ein Beitrag zur Lehre der Teilung von Gewebszellen. Arch. Mikrosk. Anat. 16, 248–300 (1879). Waldeyer, H. W. Über Karyokinese und ihre Beziehungen zu den Befruchtungsvorgängen. Arch. Mikrosk. Anat. 32, 1–22 (1888). Flemming, W. Neue Beiträge zur Kenntnis der Zelle. Arch. Mikrosk. Anat. 37, 685–751 (1891). Bütschli, O. Studien über die ersten Entwicklungsvorgänge der Eizelle, der Zellteilung und die Conjugation der Infusorien. Abh. Senckenbergische Naturf. Ges. 10, 213–452 (1876). van Beneden, E. Récherches sur la maturation de l´oeuf et la fécondation. Arch. Biol. 4, 265–638 (1883). Boveri, T. Zellen–Studien (Gustav Fischer, Jena, 1887–1900). Flemming, W. Attraktionssphären und Zentralkörperchen in Gewebs– und Wanderzellen. Anat. Anz. 6, 78–86 (1891). Flemming, W. Über die Entwicklung der Spermatosomen bei Salamandra (Kiel, 1888). Flemming, W. Beiträge zur Kenntnis der Zelle und ihrer Lebenserscheinungen. Arch. Mikr. Anat. 20, 1–86 (1882). Farmer, J. B. & Moore, J. E. S. On the maiotic phase (reduction-divisions) in animals and plants. Q. J. Microsc. Sci. 48, 489–557 (1905). Flemming, W. Beiträge zur Kenntnis der Zelle und ihrer Lebenserscheinungen. Arch. Mikrosk. Anat. 16, 302–436 (1879). Flemming, W. Beiträge zur Kenntnis der Zelle und ihrer Lebenserscheinungen. Arch. Mikrosk. Anat. 18, 151–259 (1880). Belar, K. Beiträge zur Kausalanalyse der Mitose. Roux’ Arch. Entw. Mech. Org. 118, 359–484 (1929). Harris, H. The Birth of the Cell (Yale Univ. Press, New Haven, 1999). Flemming, W. Über Zellstrukturen. Anat. Anz. 16, 2–12 (1899). Meves, F. Die Kondriomikonten im Verhältnis zur Filarmasse Flemmings. Anat. Anz. 31, 561–569 (1907). Pernice, B. Sicil. Med. 1, 265 (1889). (Cited in Eigsti, O. J. & Dustin, P. Jr Colchicine in Agriculture, Medicine, Biology, and Chemistry (Iowa State Coll. Press, Ames, Iowa, 1955).)

VOLUME 2 | JANUARY 2001 | 7 5