march 2001 by Nature Magazine

toc_r

Page 1 of 2

CONTENTS

March 2001 Vol 2 No 3

159 | In this issue doi:10.1038/35056541

Highlights PDF

[1188K]

161 | CELL DIVISION Location, location, location doi:10.1038/35056546

162 | WEB WATCH What's in it for me? doi:10.1038/35056549

162 | CELL DIVISION Don't shoot now doi:10.1038/35056506

162 | PLANT DEVELOPMENT De-pipping the Pippin doi:10.1038/35056536

163 | APOPTOSIS Silent but deadly doi:10.1038/35056500

164 | CELL SIGNALLING Ready, steady, go! doi:10.1038/35056514

164 | CELL CYCLE A complex twist for BRCA2 doi:10.1038/35056553

165 | TRANSCRIPTION Dial S for silence doi:10.1038/35056533

165 | TELOMERES Double delivery

169 | UBIQUITIN AND PROTEASOMES THEMES AND VARIATIONS ON UBIQUITYLATION Allan M. Weissman doi:10.1038/35056563 [1685K]

179 | UBIQUITIN AND PROTEASOMES ANTIGEN PROCESSING BY THE PROTEASOME Peter-M. Kloetzel doi:10.1038/35056572 [4047K]

188 | UBIQUITIN AND PROTEASOMES REGULATION OF CELLULAR POLYAMINES BY ANTIZYME Philip Coffino doi:10.1038/35056508 [592K]

195 | UBIQUITIN AND PROTEASOMES PROTEIN REGULATION BY MONOUBIQUITIN Linda Hicke doi:10.1038/35056583 [1183K]

202 | UBIQUITIN AND PROTEASOMES SUMO, UBIQUITIN'S MYSTERIOUS COUSIN Stefan M端ller, Carsten Hoege, George Pyrowolakis & Stefan Jentsch doi:10.1038/35056591 [603K]

211 | UBIQUITIN AND PROTEASOMES MOLECULAR DISSECTION OF AUTOPHAGY: TWO UBIQUITIN-LIKE SYSTEMS Yoshinori Ohsumi doi:10.1038/35056522 [1176K]

doi:10.1038/35056555

165 | WEB WATCH First among equals doi:10.1038/35056539

166 | MITOCHONDRIA

217 | TIMELINE BIOCHEMISTRY AND MOLECULAR BIOLOGY TEACHING OVER THE PAST

http://www.nature.com.ezproxy.library.wisc.edu/nrm/journal/v2/n3/toc_r.html

12/3/2006

toc_r

Page 2 of 2

Mitoskeleton, you said? doi:10.1038/35056557

50 YEARS Edward J. Wood doi:10.1038/35056600

166 | CELLULAR MICROBIOLOGY Lethal injection doi:10.1038/35056519

166 | ENDOCYTOSIS Tent pegs for clathrin doi:10.1038/35056559

[514K]

221 | OPINION THE FUTURE OF EDUCATION IN THE MOLECULAR LIFE SCIENCES Ellis Bell doi:10.1038/35056610

167 | PHOSPHORYLATION The key to staying faithful doi:10.1038/35056503

[605K]

226 | NatureView doi:10.1038/35056528

167 | IN BRIEF TECHNOLOGY | CELL SIGNALLING | DNA RECOMBINATION | DNA REPAIR doi:10.1038/35056561

NATURE REVIEWS | MOLECULAR CELL BIOLOGY

http://www.nature.com.ezproxy.library.wisc.edu/nrm/journal/v2/n3/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

Location, location, location About two years ago, the small GTPase Ran, well known as the regulator of nuclear transport, surprised many scientists by also taking centre stage in mitotic spindle assembly. Several groups have now independently discovered part of the molecular mechanism through which Ran might regulate this process. Two papers in Nature Cell Biology discuss which particular aspect of spindle assembly is controlled by Ran. Carazo-Salas and colleagues report that Ran regulates the frequency of transition from shrinkage to growth of microtubules, as well as the capacity of centrosomes to nucleate microtubules. Wilde and co-workers confirm the first finding, and add that Ran might also be able to regulate the balance of microtubule-motor activities, in particular that of Eg5. Two papers in Cell and one in Science go into more mechanistic detail, showing that Ran–GTP acts by releasing some microtubule-associated proteins (MAPs) from sequestration by importins (cargo receptors involved in nuclear transport), thereby allowing MAPs to carry out their functions in spindle assembly. Although the three groups attacked the problem from different angles, their conclusions are remarkably similar. The Karsenti/Mattaj task force find that importin-β sequesters the MAP TPX2, which they propose to be the Ran effector for spindle assembly. The Zheng and the Heald/Weis groups, on the other hand, show that importin-β sequesters the MAP

Mitotic spindle in the presence (left) and absence (right) of functional Ran–GTP. Reproduced with permission from Nature Cell Biology.

NuMA, but both agree that importins probably sequester more than one factor involved in spindle assembly. The Karsenti/Mattaj and the Heald/ Weis groups show that importin-β binds the respective MAPs indirectly, through importin-α, but the Zheng and the Heald/Weis groups suggest that importin-β might also be able to interact directly with some MAPs. The emerging model is that Ran–GTP probably acts just as it does during nuclear transport, by promoting the release of proteins from importins at a specific cellular location. During interphase, Ran–GTP dissociates cargo from importins only in the nucleus, conferring directionality to nuclear transport. In mitosis, Ran–GTP is concentrated around chromatin owing to the tight association of RCC1 (its nucleotide exchange

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

factor) to chromatin. So Ran–GTP only dissociates spindle assembly effectors from importins in a small perimeter around chromatin, thereby ensuring that they build the spindle in the right place. Raluca Gagescu References and links ORIGINAL RESEARCH PAPERS Carazo-Salas, R. E. et al. Ran–GTP coordinates regulation of microtubule nucleation and dynamics during mitotic-spindle assembly. Nature Cell Biol. 3, 228–234 (2001) | Wilde, A. et al. Ran stimulates spindle assembly by altering microtubule dynamics and the balance of motor activities. Nature Cell Biol. 3, 221–227 (2001) | Gruss, O. J. et al. Ran induces spindle assembly by reversing the inhibitory effect of importin α on TPX2 activity. Cell 104, 83–93 (2001) | Nachury, M. V. et al. Importin β is a mitotic target of the small GTPase Ran in spindle assembly. Cell 104, 95–106 (2001) | Wiese, C. et al. Role of importin-β in coupling Ran to downstream targets in microtubule assembly. Science 291, 653–656 (2001) FURTHER READING Dasso, M. Running on Ran: nuclear transport and the mitotic spindle. Cell 104, 321–324 (2001)

VOLUME 2 | MARCH 2001 | 1 6 1

HIGHLIGHTS

WEB WATCH What’s in it for me? Thanks to the efforts of the Human Genome Project, the draft human genome is out there for every molecular biologist to use, but we’re not all gene jockeys. The NCBI’s draft human genome sequence: an introduction gets inside the head of a molecular biologist who has never had to navigate his/her own DNA before, and poses the questions that s/he might ask. For one question — ‘is there only one copy of ALD (the adrenoleukodystrophy gene) in the genome?’ — a detailed tour of the draft genome is provided, with a subway map that explains how the draft human genome data intersect with the NCBI’s other resources. Each question takes you to the type of search that you need to perform, complete with a ‘help’ window explaining what it’s just done and where you can go from there. So, whether you’re interested in intron–exon structure, single nucleotide polymorphisms, homologous sequences or protein domains and their structures, the tour tells you which train to catch and where you need to change lines. One stop along this tour has been significantly refurbished. By now, all readers of Nature Reviews journals will be familiar with the NCBI’s LocusLink database as all our articles carry extensive links to it. Each LocusLink entry now provides links to the Gene Ontology database (providing information on molecular function, biological process and cellular component), Proteome’s databases of protein function (you have to register, but access is free to not-for-profit organizations), the Mouse Genome Database and Flybase. The wealth of functional data provided by these links make this the Grand Central Station of the NCBI’s tour.

CELL DIVISION

Don’t shoot now In hostage situations, the armed police show up but they might not be needed — provided that the negotiator does a good job. Likewise, the cell sometimes needs to prepare for a disaster that might never happen. In the 30 January issue of Proceedings of the National Academy of Sciences, Vanesa Gottifredi and colleagues describe such a case: when DNA synthesis is blocked, p53 is on call but is prevented from going in with guns blazing. In response to agents that might lead to DNA damage, such as γirradiation, p53 becomes phosphorylated. This stabilizes it by

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

De-pipping the Pippin Seedless fruit may be a pointless waste of resources for a plant, but the economic rewards of growing such crops are considerable. Whereas seedless grapes, bananas and oranges are common groceries, other fruits are less easy to produce. Now, reporting in Proceedings of the National Academy of Sciences, Jia-Long Yao and colleagues have identified muta-

Cath Brooksbank

162

preventing its interaction with the ubiquitin ligase MDM2, allowing it to orchestrate a transcription programme that induces cell-cycle arrest. It’s always been assumed that a stable p53 equals an active p53, but Gottifredi et al. now show that this isn’t true. Treatment with agents that block DNA replication, such as hydroxyurea or aphidicolin, stabilize p53 but many of its target genes, including p21WAF1 and MDM2, are not transcribed and, once the cell-cycle block is released, the cells continue to divide in the presence of high levels of p53. This is not due to a general shutdown of RNA synthesis because a p53 target gene, PIG3 is transcribed, as are other genes such as c-fos and cyclin E. Furthermore, p53 remains in the nucleus, so cellular localization

tions in a single gene that result in seedless apples. Conventional plant breeding has produced a small number of seedless apple varieties such as Wellington Bloomless, Spencer Seedless and Rae Ime, but these varieties produce undersized fruit of poor quality. The apples are not seedless because they cannot make seeded fruit — in fact, if hand pollinated, they produce fruit with twice the conventional number of seeds. Rather, their flowers are so stunted that they fail to attract insect pollinators. The blooms of Rae Ime have neither petals nor anthers, these organs being replaced by additional sepals and styles. But lack of pollination is not, on its own, enough to make seedless fruit — plants require both pollination and fertilization to trigger fruit development.

doesn’t hold the key to its inactivity. But perhaps most surrprising is the finding that when DNA synthesis is blocked, p53 fails to transcibe p21WAF1 and MDM2 even when the cells are blasted with γ-irradiation. What might be reining p53 in? The patterns of phosphorylation and acetylation of p53 after blocking DNA sythesis or γ-irradiation seem largely the same. Furthermore, ATM, the kinase that leads to p53’s stabilization in response to γirradiation, is functional and loss of ATM doesn’t prevent the accumulation of p53 when DNA synthesis is blocked. So an unidentified kinase or an alternative, kinase-independent mechanism must be responsible for p53’s stable but inactive state. Why do cells need to keep p53 in

Nevertheless, these stunted flowers reminded Yao and colleagues of a classic Arabidopsis thaliana mutation called pistillata, in which flowers also lack both petals and anthers. The PISTILLATA gene belongs to the so-called MADS-box family. By a scheme of overlapping expression, these genes direct the development of the four organs that make up a flower: carpels, petals, anthers and styles. Honma and Goto recently reported in Nature that, in the flower, the MADS-box gene products form multi-protein transcription factor complexes, the compositions of which determine their DNA-binding specificities. Inclusion of the PISTILLATA protein in a complex changes its transcriptional targets, converting organs that would otherwise become carpels and styles into petals and anthers, respectively. Taking the hint from Arabidopsis, Yao and colleagues first identified the apple homologue of PISTILLATA in Granny Smith apples; they share 64% identity at the amino-acid level. In Rae Ime, Wellington Bloomless and Spencer Seedless varieties, however, the authors found transposon insertions that disrupt the gene and prevent it from being transcribed. This study is the first hint that PISTILLATA is involved in parthenocarpy — the official name for fruit production without fertilization. www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

HIGHLIGHTS

check in response to the DNAsynthesis block? The authors speculate that it’s a survival mechanism: during S phase there are likely to be DNA strand breaks and stalled replication forks that will lead to p53 stabilization. At this point in the cell cycle E2F-1 levels are high. E2F-1 and p53 comprise a lethal concoction, so to prevent cells from apoptosing every time they divide, p53 must be kept in check. How this occurs now needs to be worked out. Cath Brooksbank References and links ORIGINAL RESEARCH PAPER Gottifredi, V. et

al. p53 accumulates but is functionally impaired when DNA synthesis is blocked. Proc. Natl Acad. Sci. USA 98, 1036–1041 (2001) FURTHER READING Takimoto, R. & El-Diery, W. S. DNA replication blockade impairs p53transactivation. Proc. Natl Acad. Sci. USA 98, 781–783 (2001)

Arabidopsis produces a simple type of fruit known as a silique, which develops from the ovaries alone (an organ in which PISTILLATA is never expressed). Apples, on the other hand, form pome fruits with seeds embedded in fleshy tissue derived from sepals, petals and anthers. Somehow PISTILLATA must block the development of pome tissue — a block that is relieved in normal apples by fertilization. Seedless fruit varieties are more attractive to consumers and, because they crop without the need for pollinators, they do not depend on insect species during flowering. The identification by Yao and colleagues of the source of seedlessness opens the way for producing seedless strains of commercial apple varieties, whether by conventional breeding or by genetic-manipulation techniques. It may also lead to pipless pears (another pome fruit) and, who knows, perhaps even the stoneless plum. Christopher Surridge Senior Editor, Nature References and links ORIGINAL RESEARCH PAPERS Yao, J.-L.,

Dong, Y.-H. & Morris, B. A. M. Parthenocarpic apple fruit production conferred by transposon insertion mutations in a MADS-box transcription factor. Proc. Natl Acad. Sci. USA 98, 1306–1311 (2001) | Honma, T. & Goto, K. Complexes of MADS-box proteins are sufficient to convert leaves into floral organs. Nature 409, 525–529 (2001) FURTHER READING Ng, M. & Yanofsky, M. F. Function and evolution of the plant MADS-box gene family. Nature Rev. Genet. 2, 186–196 (2001)

HUMAN GENOME A P O P TO S I S

Apoptotic molecular machinery: vastly increased complexity in vertebrates revealed by genome comparisons. A P O P TO S I S

Aravind, L. et al. Science 291, 1279–1284 (2001)

Silent but deadly Cancer cells find many ways to cheat death, but one puzzle is how they do so in malignant melanoma, where p53 — a key trigger of apoptosis — is often functional. The answer, revealed by Scott W. Lowe and colleagues in Nature, is that they switch off a gene further down the death pathway. Lowe and co-workers surveyed several tumour types for loss-of-function mutations in Apaf-1. This potential tumour-suppressor gene encodes the co-activator for caspase-9, a downstream effector of p53. In malignant melanoma samples (top panel in the figure) the authors found that the levels of Apaf-1 messenger RNA were reduced, and experiments with melanoma cell lines confirmed that in these ‘Apaf-1-negative’ cells the p53dependent response to chemotherapeutic drugs was compromised. Lowe and co-workers showed that several melanoma cell lines contain only one copy of Apaf-1, and they concluded that this allele is transcriptionally silenced — perhaps by DNA methylation. To test this possibility they added a methylation inhibitor to melanoma cells, and observed an 8–20-fold increase in the levels of Apaf-1 protein and RNA in the Apaf1-negative cells, but little effect on cells expressing normal levels of Apaf-1. These results, say the authors, “imply that Apaf-1 loss contributes to the aggressive nature and extreme chemoresistance of metastatic melanomas”. And, although other anti-apoptotic mutations might be involved, these data raise the possibility that such cancers may one day be treated with chemotherapeutic drugs. Alison Mitchell References and links ORIGINAL RESEARCH PAPER Soengas, M. et

al. Inactivation of the apoptosis efector Apaf-1 in malignant melanoma. Nature 409, 207–211 (2001) FURTHER READING Jones, P. Death and methylation. Nature 409, 141–144 (2001)

The apoptotic machinery evolved from signalling pathways present in the common ancestor of plants, animals and fungi. Analysis of the human, fly and nematode genomes now reveals an increase in both the number and complexity of apoptosis-related proteins in vertebrates. D N A R E PA I R

Human DNA repair genes. Wood, R. D. et al. Science 291, 1284–1289 (2001)

This catalogue groups 130 repair genes on the basis of function — for instance, base-excision repair, nucleotide-excision repair or mismatch repair — or sequence homology to known repair genes in other organisms. A strong message is the likelihood of clinical applications relating to human DNA repair genes. M E M B R A N E DY N A M I C S

A genomic perspective on membrane compartment organization. Bock, J. B. et al. Nature 409, 839–841 (2001)

The authors compared four protein families involved in membrane traffic (SNAREs, Rabs, coats and members of the Sec1 family) in four different organisms (yeast, worm, fly and human). There was no difference in the basic machinery between the unicellular yeast and multicellular flies or worms. But humans have about twice as many genes in each of these families. The final (?) count is 35 SNAREs, 60 Rabs and 53 coat complex subunits. C Y TO S K E L E TO N

Genomics, the cytoskeleton and motility. Pollard, T. D. Nature 409, 842–843 (2001)

The search for new cytoskeletal proteins yielded mixed results depending on the protein family. The author found seven highly divergent actin genes and seven new Arp genes. But the search yielded essentially no new myosins and kinesins in addition to the 40 or so known members of each of these families. One explanation is that it is much harder to assemble the genes of large multi-domain proteins. CE LL CYCLE

Can sequencing shed light on cell cycling? Murray, A. W. & Marks, D. Nature 409, 844–846 (2001)

The authors were disappointed to find only a few new cyclins and no new Cdks and spindle checkpoint proteins. No need to be disappointed — this could simply mean that they’ve done an excellent job in the past!

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 1 6 3

HIGHLIGHTS

CELL SIGNALLING

CE LL CYCLE

Ready, steady, go! Traffic lights regulate the movement of vehicles on roads by transmitting ‘stop’, ‘get ready’ and ‘go’ signals to drivers. Similarly, antigen-presenting cells use cytokines as stop and go signals for lymphocytes. But what’s the switch that changes the signal? In the January issue of Nature Immunology, Amy Weinmann and colleagues describe a two-part switch for regulating transcription of a cytokine gene: one signalling pathway leads to chromatin remodelling, and a second, independent pathway activates transcription. An important element in the initiation of inflammatory responses is the activation of macrophages, resulting in the production of pro-inflam-

A complex twist for BRCA2

matory cytokines such as interleukin 12 (IL-12), a heterodimeric protein comprising p40 and p35 subunits. Toll-like receptors (TLRs), which are expressed on macrophages, recognize microbial molecules and transmit signals that initiate transcription of cytokine genes; TLR4 recognizes the Gram-negative bacterial product lipopolysaccharide (LPS). TLRs use several signalling pathways, including the nuclear factor κB (NF-κB) and Jun N-terminal kinase pathways, to initiate gene transcription. Which of these pathways stimulates macrophages to produce IL-12? Using restriction enzyme accessibility assays, Weinmann and colleagues found that TLR4 signalling in response to LPS activation results in nucleosome remodelling at the p40 promoter. Curiously, although active NF-κB is essential for transcription of p40, remodelling was not dependent on NF-κB or another transcription factor, CCAAT enhancer-binding protein β. It seems that other TLR4inducible factors can stimulate remodelling, perhaps making the p40 promoter more accessible to transcription factors such as NF-κB. So chromatin remodelling — a previously unrecognized endpoint of TLR signalling — behaves like an amber signal that prepares the chromatin for NF-κB, the green light for transcription of p40. But what is the identity of the protein that recruits the remodelling complex, and what exactly is this complex? Further work in this area should enhance our understanding of TLR signalling and the regulatory mechanisms controlling induction of the inflammatory response. Elaine Bell Editor, Nature Reviews Immunology References and links

BRCA2 is one of the best-known genes associated with breast and ovarian cancer, yet we know remarkably little about its role in the cell. Despite its large size (390 kDa), BRCA2 contains no obvious homology to known sequences — a fact that has eluded many previous attempts to characterize its function. It has been linked to both transcriptional regulation and DNA repair, but we still know relatively little about how loss of BRCA2 disrupts these processes. Now, reporting in Cell, Marmorstein and colleagues have isolated the multiprotein complex that BRCA2 resides in, and they show that it binds directly to a DNA-binding protein, BRCA2-associated factor 35 (BRAF35). What’s more, they have identified a new function for BRCA2 — mediating timely progression into mitosis. To track down the elusive factors that interact with BRCA2, the authors used affinity purification. They found that BRCA2 is present in a 2 MDa complex. Within this complex, they identified BRAF35, an evolutionarily conserved protein that contains a high mobility group (HMG) domain. They then showed, using a pulldown assay, that BRCA2 and BRAF35 interact directly. Interestingly, BRAF35 is also present in a second complex that does not contain BRCA2. Because HMG domains are implicated in mediating binding at DNA junctions, they wondered whether BRAF35 might bind selectively to specific types of DNA. And they found, consistent with such a role, that BRAF35 is an architectural DNA-binding protein that binds cruciform DNA — the type of structure formed at junctions. Next, Marmostein and colleagues asked whether BRAF35 and BRCA2 are expressed together in vivo. Using in situ hybridization to mouse embryos, they found that both are expressed most highly in rapidly dividing cells. To look more closely at the subcellular localization, the authors examined mitotic HeLa cells using indirect immunofluorescence, and found that BRAF35 and BRCA2 both localize to mitotic chromosomes during initiation of chromosome condensation. This localization suggests a role for BRCA2 in cell-cycle progression — a function that they confirmed by antibodyinjection experiments. Here, injection of antibodies against either BRCA2 or BRAF35 resulted in a G2 delay in HeLa cells. So it seems that this complex might have a dual function in regulating both DNA repair and the cell cycle. Many questions have now been raised. For instance, can this complex bind to the DNA junctions that are formed during DNA repair? Does BRAF35 mediate interaction of BRCA2 with DNA for all BRCA2’s functions? Now that the identity of BRCA2’s cohorts has been unveiled, the answers to these questions are likely to come pouring in. Alison Schuldt

ORIGINAL RESEARCH PAPER Weinmann, A. S.

et al. Nucleosome remodelling at the IL-12 p40 promoter is a TLR-dependent, Rel-independent event. Nature Immunol. 2, 51–57 (2001) FURTHER READING Beutler, B. & Poltorak, A. Toll we meet again. Nature Immunol. 2, 9–10 (2001) | Aderem, A. & Ulevitch, R. J. Toll-like receptors in the induction of the innate immune response. Nature 406, 782–787 (2000) WEB SITE The Smale lab

164

References and links ORIGINAL RESEARCH PAPER Marmorstein, L. Y. et al. A human BRCA2 complex

containing a structural DNA binding component influences cell cycle progression. Cell 104, 247–257 (2001) FURTHER READING Chen, P. L. et al. The BRC repeats in BRCA2 are critical for RAD51 binding and resistance to methyl methanesulfonate treatment. Proc. Natl Acad. Sci. USA 95, 5287–5292 (1998) | Marmorstein, L. Y. et al. The BRCA2 gene product functionally interacts with p53 and RAD51. Proc. Natl Acad. Sci. USA 95, 13869–13874 (1998) | Milner, J. et al. Transcriptional activation functions in BRCA2. Nature 386, 772–773 (1997).

www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

HIGHLIGHTS

WEB WATCH

TRANSCRIPTION

Dial S for silence Somewhere between early S phase and G2/M lies a cell-cycle switch that allows gene expression to be silenced. In response to this event, the chromatin adopts an altered form — heterochromatin — that prevents the transcriptional machinery from reaching its target genes. DNA replication has long been thought to flick this switch, but two reports in Science suggest that it might not be required after all. Both groups studied the budding yeast HMRa mating-type locus, which is silenced through the formation of heterochromatin. Silencing requires the binding of several proteins to cis-acting elements called silencers. Among these proteins is the origin-recognition complex (ORC), which is involved in the initiation of DNA replication (although this function is separable from its role in silencing). ORC then recruits a Sir (silent information regulator) protein called Sir1, and this, in turn, facilitates the incorporation of further Sir proteins into the heterochromatin. To find out whether replication through the HMRa locus is needed to establish silencing, the two groups adopted a similar strategy — they compared the formation of heterochromatin at chromosomal HMRa with the same process on a non-repliT E LO M E R E S

Double delivery For long-term survival a cell must take care of its telomeres. It needs to protect them from the DNArepair apparatus — which might otherwise view them as doublestranded breaks — while making sure that they are completely replicated during cell division. Reports in Cell and Genes and Development by Vicki Lundblad and colleagues now show that these two functions are reconciled in yeast through a telomerebinding protein called Cdc13. Lundblad’s group previously

cating, extrachromosomal DNA ring. The extrachromosomal DNA (which was excised by site-specific recombination) could not replicate because it lacked the ORC-binding site. And, to circumvent the need for ORC in silencing, the authors had engineered either Gal4 (Kirchmaier and Rine) or LexA (Li and colleagues) DNA-binding sites into the extrachromosomal ring. Expression of Gal4–Sir1 or LexA–Sir1 fusion proteins then provided all the ingredients required for silencing. Using this system, both groups saw little difference in silencing of the two HMRa loci, with the startling implication that replication is not required to initiate this process. An obvious concern is whether the extrachromosomal ring can, in fact, replicate, but the authors did a number of controls to

show that this was not the case. They also showed, consistent with previous dogma, that passage through S phase is required for silencing. But if replication is not the switch, then what is? Just one of the many theories is that the activity of a silencing component — perhaps a Sir protein — is controlled in a cell-cycledependent manner, possibly by regulated synthesis of cofactors or degradation of inhibitors.

showed that Cdc13 can positively regulate telomere replication by recruiting the enzyme responsible — telomerase — to chromosome ends. But Cdc13 also negatively modulates telomere replication — an effect that occurs after the recruitment of telomerase, and depends on a protein known as Stn1. One explanation for this negative regulation is the recruitment, by Cdc13, of an end-protecting activity. The obvious candidate is Stn1, so the authors fused the DNAbinding domain of Cdc13 to Stn1. Expression of this construct rescued the lethality of a cdc13 null strain, suggesting that Stn1 is the arbiter of end protection and that it is

delivered to telomeres by Cdc13. The association of Cdc13 with both telomerase and Stn1 is blocked by a single mutation (cdc13-2), leading Lundblad and colleagues to describe how Cdc13 might regulate telomere replication. According to their model, telomerase is delivered to the DNA end in the first (positive) step. Then, in the second (negative) step, Stn1 binds an overlapping site on Cdc13, allowing it, in turn, to be recruited to the telomere.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

Alison Mitchell References and links ORIGINAL RESEARCH PAPERS Kirchmaier, A.

& Rine, J. DNA replication-independent silencing in S. cerevisiae. Science 291, 646–650 (2001) | Li, Y.-C., Cheng, T.-H. & Gartenberg, M. R. Establishment of transcriptional silencing in the absence of DNA replication. Science 291, 650–653 (2001) FURTHER READING Smith, J. S. & Boeke, J. D. Is S phase important for transcriptional silencing? Science 291, 608–609 (2001) WEB SITE Chromatin structure and function page

Alison Mitchell References and links ORIGINAL RESEARCH PAPERS Pennock, E.,

Buckley, K. & Lundblad, V. Cell 104, 387–396 (2001) | Chandra, A. et al. Cdc13 both positively and negatively regulates telomere replication. Genes Dev. 15, 404–414 (2001)

First among equals Bio OnlineTM, which began life as the Biological Research Network, International (BRNI) in 1992, claims to be recognized as “the first life sciences web site”. Almost ten years on, and the site shows few signs of old age. At the heart of Bio Online is a fresh-looking home page fronting links to a variety of sections. For the corporateminded, the ‘Industry news’ and ‘Industry reports and analyses’ cover company reports and business issues in the biotechnology sector. ‘Research news’, on the other hand, gives short summaries of recent research. Although the papers highlighted reflect the site’s slant towards biotechnology and medicine, the articles are readable and up to date, with links to useful web sites. Each month, ‘In focus’ describes a cutting-edge technology, including a round-up of hot papers in the area and short transcripts of online discussions with leaders in the field. Proteomics and bioinformatics are featured in February and March, respectively, although some of the previous topics are a little more off-beat. Right on track, however, is the ‘Research and education’ section, with extensive links to lab protocols and research tools. The ‘Career centre’ is also useful, featuring careers advice and a moderated career forum. Only the jobs database disappoints — most adverts are placed by US companies, few specific positions are offered, and a high proportion of those seem to be in sales rather than research. Those behind Bio Online claim that it is used by “researchers … from academia and from pharmaceutical and biotechnology companies”. This is a commercial site, though, and we suspect that researchers in industry will gain most from it.

Alison Mitchell

VOLUME 2 | MARCH 2001 | 1 6 5

HIGHLIGHTS

M I TO C H O N D R I A

Mitoskeleton, you said? Cells do not like simplicity. So instead of being spherical, mitochondria form tubular networks with complex inner membranes. How do cells build these complicated structures? To tackle this question, a few years ago, Rob Jensen’s laboratory screened for Saccharomyces cerevisiae mutants defective in mitochondrial shape, and isolated one mutant, mmm1, in which mitochondria were large and spherical. The Jensen group have now analysed the localization of an Mmm1–GFP chimera in living yeast cells. They found Mmm1 in spots on the surface of mitochondria, adjacent to a subset of mitochondrial DNA (mtDNA) nucleoids. In mmm1 mutants, mtDNA C E L L U L A R M I C R O B I O LO G Y

Lethal injection Many pathogenic bacteria have the nasty habit of injecting proteins into the cytoplasm of their host to interfere with its signalling pathways. Gram-negative bacteria use syringes — type III secretion systems — but Gram-positive bacteria are not equipped with such sophisticated tools. So how do they get their deadly messengers across the host’s plasma membrane? Madden and colleagues report in Cell that Streptococcus pyogenes might inject one of its virulence factors through pores formed by a bacterial cholesteroldependent cytolysin, streptolysin O (SLO).

collapses into one large structure, and cannot be properly segregated during mitochondrial division, which leads to its loss. But Mmm1 is an outer membrane protein, so how can it bind to mtDNA inside mitochondria? The authors suggest that Mmm1 functions in the formation of contact sites between the outer and inner membranes, and this is supported by the fact that inner membranes collapse in mmm1 mutants. So the authors’ model is that Mmm1 is part of a scaffold-like complex — the ‘mitoskeleton’ — that holds the outer and inner membranes together and is required for normal mitochondrial shape and mtDNA segregation. Raluca Gagescu References and links ORIGINAL RESEARCH PAPER Aiken Hobbs, A. E. et al.

Mmm1p, a mitochondrial outer membrane protein, is connected to mitochondrial DNA (mtDNA) nucleoids and required for mtDNA stability. J. Cell Biol. 152, 401–410 (2001)

Infection of keratinocytes by S. pyogenes is accompanied by modulation of the host’s proinflammatory response. Madden and colleagues noticed that a bacterial factor appeared in the host cytosol only when SLO was active. They identified the factor as SPN (S. pyogenes NADglycohydrolase), which can produce the second messenger cyclic ADP-ribose and could therefore be the effector that is responsible for modulating host cell signalling during infection. The functions of both SPN and SLO, as well as bacterial adherence, are required for cytotoxicity. The emerging model is that bacteria adhere and secrete SLO to form pores through which SPN can access the host’s cytoplasm to trigger cytotoxicity. Whether this occurs by diffusion or by an active process is not completely clear, but the authors present circumstantial evidence that transport occurs in a vectorial manner: coinfection of keratinocytes with strains that express either SPN or SLO alone does not trigger cytotoxicity, indicating that the action of the two proteins is tightly coordinated. Hence, transport of SPN into the host is probably not a random diffusion process. The ability of SLO to form 30-nm pores that allow the passage of fully folded proteins has been known for a long time and is widely used by cell biologists for selective permeabilization of plasma membranes. This study provides the first indication that cytolysins, including SLO, might do more than just make the host cell leaky. But important questions remain, the most obvious being how vectorial transport of SPN through the pore occurs — if, indeed, vectorial transport it is. Raluca Gagescu References and links ORIGINAL RESEARCH PAPER Madden, J. C., Ruiz, M. &

Caparon, M. Cell 104, 143–152 (2001)

166

E N D O C Y TO S I S

Tent pegs for clathrin Every construction — no matter how temporary — needs sound foundations. So how are the clathrin coats that surround endocytic vesicles tethered to the plasma membrane? Two papers in the 9 February issue of Science reveal the importance of a phospholipid, phosphatidylinositol-4,5-bisphosphate (PtdIns(4,5)P2), in this process. Several proteins involved in endocytosis, including epsin, clathrin assembly lymphoidmyeloid leukaemia protein (CALM) and its neuronal homologue AP180, contain an epsin amino-terminal homology (ENTH) domain. But what is its function? Itoh and colleagues couldn’t find any proteins that bind epsin’s ENTH domain, so they searched for phospholipid partners and discovered that PtdIns(4,5)P2-containing vesicles bind it. NMR spectroscopy of epsin’s ENTH domain bound to inositol-1,4,5-trisphosphate (Ins(1,4,5)P3) revealed that Ins(1,4,5)P3 binds to a positively charged pocket in an eight-helix bundle. Deletion of loop 1 or substitution of Arg63 or Lys76 for Ala in helix 3 almost abolished Ins(1,4,5)P3 binding, and overexpression of a mutant epsin lacking the ENTH domain, or a Lys76Ala mutant, blocked endocytosis. Ford and colleagues used X-ray crystallography to solve the structure of CALM’s amino terminus bound to a series of inositol phosphates and phosphoinositides. CALM’s amino terminus looks reassuringly similar to epsin’s ENTH domain but, surprisingly, PtdIns(4,5)P2 binds to a cluster of protruding lysines and a histidine not present in epsin’s ENTH domain. Both CALM and AP180 were specifically sedimented by liposomes or tubules containing PtdIns(4,5)P2, and mutation of the lysine cluster in either protein blocked sedimentation. Ford and colleagues could partially reconsitute clathrin-coat formation using PtdIns(4,5)P2-containing phospholipid monolayers, clathrin, AP180 and the adaptor protein AP2. Substitution of PtdIns(4,5)P2 for PtdIns, or omission of AP180, abolished the formation of clathrin lattices. Why do two similar domains bind PtdIns(4,5)P2 through different sites? Whatever the answer, the function of PtdIns(4,5)P2 as a tent-peg par excellance is now undisputed. Cath Brooksbank References and links ORIGINAL RESEARCH PAPERS Itoh, T. et al. Role of the

ENTH domain in phosphatidylinositol-4,5-bisphosphate binding and endocytosis. Science 291, 1047–1051 (2001) | Ford, M. G. J. et al. Simultaneous binding of PtdIns(4,5)P2 and clathrin by AP180 in the nucleation of clathrin lattices on membranes. Science 291, 1051–1055 (2001)

www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

HIGHLIGHTS

IN BRIEF T E C H N O LO G Y

A high signal-to-noise Ca2+ probe composed of a single green fluorescent protein. Nakai, J., Ohkura, M. & Imoto, K. Nature Biotechnol. 19, 137–141 (2001)

Currently available Ca2+ probes based on green fluorescent protein (GFP) have a low signal-to-noise ratio, limiting their use. Nagai et al. have built a GFP with a Ca2+–calmodulin-sensitive enzyme tagged on one end and calmodulin on the other. Ca2+-induced conformational changes cause a large increase in fluorescence. CELL SIGNALLING

Regulation of a novel human phospholipase C, PLC-ε, through membrane targeting by Ras. Song, C. et al. J. Biol. Chem. 276, 2752–2757 (2001)

P H O S P H O R Y L AT I O N

The key to staying faithful The cascades in which mitogen-activated protein kinases (MAPKs) play a part can have radically different outcomes, making promiscuity among MAPKs dangerous to the cell. Yet each pathway comprises closely related components. How do MAPKs remain faithful to their pathways? In the 1 February issue of EMBO Journal, Takuji Tanoue and colleagues explain one mechanism. Remarkably, this relies on the identity of just two amino acids. We know of three types of MAPKs, each with their own cascade. The extracellular-signal regulated kinases (ERKs) are activated by growth factors, generally leading to proliferation, whereas p38s and Jun-Nterminal kinases (JNKs) are activated by stress signals, usually causing cellcycle arrest or apoptosis. Each MAPK has to interact with the kinases that activate it (MAPKKs), the phosphatases that inactivate it (MKPs) and its substrates, which are also kinases (MAPKAPKs). A single acidic site outside the active site — the common docking or CD site — interacts with all these molecules, but it doesn’t explain the different binding specificities of the MAPKs. Could there be another site that regulates specificity? Mutation of the CD in p38 reduced, but didn’t completely prevent, binding of the p38-specific MAPKAPK 3pk, implying that another docking site exists. By searching for charged residues and systematically mutating them, the authors identified a pair of residues on p38, Glu 160 and Asp 161, that account for this residual binding. Mutation of the CD site and this second site, which they dubbed the ED site, markedly reduced the ability of p38 to phosphorylate 3pk. The corresponding residues in ERK2 are two threonine residues. Mutation of these to Glu and Asp enabled ERK2 to bind 3pk, and mutation of ERK2’s CD site to make it identical to p38’s improved the interaction further. Extending these studies to other MAPKAPKs and MKPs revealed that, although the CD is necessary for binding, the nature of the ED regulates specificity. Together, the two sites form a groove with two pins in it. Only if it can interact with both pins can a MAPKinteracting protein do its business. Cath Brooksbank References and links ORIGINAL RESEARCH PAPER Tanoue, T. et al. Identification of a docking groove on ERK and p38 MAP kinases that regulates the specificity of docking interactions. EMBO J. 20, 466–479 (2001) FURTHER READING Tanoue, T. et al. A conserved docking motif in MAP kinases common to substrates, activators and regulators. Nature Cell Biol. 2, 110–116 (2000) WEB SITE Mammalian MAPK signalling pathways

A novel bifunctional phospholipase C that is regulated by Gα12 and stimulates the Ras/MAP kinase pathway. Lopez, I. et al. J. Biol. Chem. 276, 2758–2765 (2001)

Phospholipase Cε: a novel Ras effector. Kelley, G. G. et al. EMBO J. 20, 743–754 (2001)

These papers report a new mammalian phospholipase C that is activated by Ras (as well as the α-subunit of a heterotrimeric G protein) but can also activate Ras through its guanine nucleotide exchange factor domain. D N A R E C O M B I N AT I O N

Branch migration and Holliday junction resolution catalyzed by activities from mammalian cells. Constantinou, A., Davies, A. A. & West, S. C. Cell 104, 259–268 (2001)

A key intermediate of homologous recombination and doublestranded-break repair is the Holliday junction. This dynamic structure can move (branch migrate) to generate stretches of heteroduplex DNA, and it is resolved by a junction-specific endonuclease. These reactions are well characterized in bacteria, and West and colleagues now describe analogous activities in mammalian cell-free extracts. This report highlights the conservation of this pathway from prokaryotes to mammals. D N A R E PA I R

XRCC1 stimulates human polynucleotide kinase activity at damaged DNA termini and accelerates DNA singlestrand break repair. Whitehouse, C. J. et al. Cell 104, 107–117 (2001)

XRCC1 is involved in repairing single-stranded DNA breaks, but little is known about its biochemical function. Now, using XRCC1 as bait in a yeast two-hybrid screen, the authors have identified a new partner for it — human polynucleotide kinase (PNK). XRCC1 stimulates the DNA kinase and phosphatase activities of PNK at damaged DNA termini, and this accelerates the repair reaction. It is, claim the authors,“a novel pathway for mammalian single-strand break repair”.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 1 6 7

REVIEWS THEMES AND VARIATIONS ON UBIQUITYLATION Allan M. Weissman Ubiquitylation — the conjugation of proteins with a small protein called ubiquitin — touches upon all aspects of eukaryotic biology, and its defective regulation is manifest in diseases that range from developmental abnormalities and autoimmunity to neurodegenerative diseases and cancer. A few years ago, we could only have dreamt of the complex arsenal of enzymes dedicated to ubiquitylation. Why has nature come up with so many ways of doing what seems to be such a simple job? U B I Q U I T I N A N D P R OT E A S O M E S

Laboratory of Immune Cell Biology, Center for Cancer Research, National Cancer Institute, Bethesda, Maryland 20892-1152, USA. e-mail: amw@nih. gov

Ubiquitin is a 76-amino-acid globular protein that is highly conserved throughout eukaryotes, with only three amino-acid changes from yeast to human. Its covalent conjugation to other proteins — ubiquitylation (or ubiquitination) — is essential for the degradation of proteins whose levels are regulated either constitutively or in response to changes in the cellular environment. Ubiquitin is integral to myriad processes such as cellcycle progression; organelle biogenesis; apoptosis; regulated cell proliferation; cellular differentiation; quality control in the endoplasmic reticulum; protein transport; inflammation; antigen processing; DNA repair; and stress responses. In this way, it resembles another posttranslational modification — phosphorylation — with which it is intimately intertwined. Phosphorylation can augment or inhibit ubiquitylation, by modifying either the protein destined to be ubiquitylated or the enzymes that catalyse the addition of ubiquitin. So what makes ubiquitin such a great multitasker? The classical view of ubiquitylation is that it targets proteins for degradation by a multisubunit, ATPdependent protease termed the proteasome (see the review by Peter Kloetzel on page 179 of this isssue for more information on the proteasome). In addition to its role in proteasomal degradation, ubiquitylation is also emerging as a signal that targets plasma membrane proteins for destruction in vacuoles and/or lysosomes (see the review by Linda Hicke on page 195 of this issue). Thus, ubiquitin targets proteins from topologically distinct locations to fundamentally different prote-

olytic structures. We are only just beginning to understand the functional diversity of the ubiquitin signal. Although targeting for degradation is undoubtedly one of its key tasks, other cellular functions not directly involving protein degradation, including regulation of translation, activation of transcription factors and kinases, and DNA repair, are controlled in one way or another by this seemingly simple protein. If ubiquitin can be attached to so many proteins, how is specificity generated? Moreover, how does ubiquitylation of one protein sentence it to destruction in proteasomes when, in another setting, modification with the same polypeptide leads to enhanced translation? We don’t have a full solution to this puzzle, but the pieces are falling into place. Specificity is generated largely by the enzymes that recognize substrates and mediate ubiquitylation (FIG. 1). But it is also evident that the fate of the ubiquitylated proteins is determined by the types of ubiquitin conjugate formed. For instance, a single ubiquitin tag does not target a protein for proteasomal degradation, whereas a chain of four or more does1. There are also subtly different ways of building a multi-ubiquitin chain — by using different lysine residues of ubiquitin — and these have functional consequences. In addition, intracellular location helps to determine the fate of ubiquitylated proteins: ubiquitylation in the nucleus might not have the same consequence as that in the cytosol, and ubiquitylation of a transmembrane protein at the endoplasmic reticulum (ER) membrane might have a different result from at the plasma membrane.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 1 6 9

REVIEWS

E2 E3

O Ub

C OH

AMP+PPi E1

ATP E1

S C

E2 E2

NH2 O

S C

E2 E3

S C

RING E3

O Ub

Ub HECT E3

SH S NH2

E2 E3

S C

E3 S

NH2

O C

Figure 1 | The ubiquitylation pathway. Free ubiquitin (Ub) is activated in an ATP-dependent manner with the formation of a thiol-ester linkage between E1 and the carboxyl terminus of ubiquitin. Ubiquitin is transferred to one of a number of different E2s. E2s associate with E3s, which might or might not have substrate already bound. For HECT domain E3s, ubiquitin is next transferred to the active-site cysteine of the HECT domain followed by transfer to substrate (S) (as shown) or to a substratebound multi-ubiquitin chain. For RING E3s, current evidence indicates that ubiquitin might be transferred directly from the E2 to the substrate.

RUB1

(Nedd8 in metazoans). A ubiquitin-like (UBL) protein that is activated by its own E1and E2-like molecules and modifies cullin family members. HECT

Stands for homologous to E6AP carboxyl terminus. The HECT domain is a ~350amino-acid domain, highly conserved among a family of E3 enzymes. RING FINGER

Defined structurally by two interleaved metal-coordinating sites. The consensus sequence for the RING finger is: CX2CX(9–39)CX(1–3)HX(2–3) C/HX2CX(4–48)CX2C. The cysteines and histidines represent metal-binding sites with the first, second, fifth and sixth of these binding one zinc ion and the third, fourth, seventh and eighth binding the second.

170

The concept that cellular proteins can be targeted for modification by small proteins, resulting in alteration in the fate or function of the targeted protein, extends beyond ubiquitin. A growing list of ubiquitin-like (UBL) proteins is being identified and characterized. As with ubiquitin, the active forms of UBLs include a glycine at the carboxyl terminus that forms an isopeptide bond with ε-amino groups of lysines on target proteins. UBLs include at least five distinct proteins that are related in sequence to ubiquitin as well as two that are not. Of the UBLs that are homologous to ubiquitin, the first to be characterized was a protein that resembles a ubiquitin dimer, known as the ubiquitin cross-reactive protein (UCRP) or ISG15 (REF. 2). Also in this group is Saccharomyces cerevisiase RUB1 (which stands for related to ubiquitin) — known as Nedd8 in metazoans (herein referred to as Rub1) and SUMO-1 (small ubiquitinrelated modifier; also known as Ubl1, Sentrin or PIC-1 (see the review by Stefan Jentsch and colleagues on page 202 of this issue and REFS 3, 4). Modification with Rub1 (rubylation) or with SUMO-1 (sumoylation) can have direct effects on ubiquitylation. Apg12 is a UBL that lacks amino-acid homology with ubiquitin. Apg12 is a central player in a fascinating story in which a multienzyme process that parallels ubiquitylation mediates autophagy (see the review by Yoshinori Ohsumi on page 211 of this issue). Although ubiquitin must now share the limelight on the protein modification stage with the UBLs, it alone has the remarkable ability to form a variety of different chains on target proteins — potentiating its capacity to generate a diverse array of signals. In addition to the UBLs, an increasing number of otherwise structurally unrelated proteins are being found to contain domains homologous to ubiquitin. These ubiquitin-domain proteins (UDPs) have varied cellular functions and, unlike the UBLs, are not known to be covalent modifiers of proteins. Included among the UDPs are ubiquitylation substrates as well as

enzymes involved in both the addition and the removal of ubiquitin from proteins. Some UDPs interact with proteasomes as well as with enzymes that are involved in mediating ubiquitylation. Thus, the presence of the ubiquitin domain in otherwise disparate proteins does not simply reflect conservation of a stable structural domain. Instead, the ubiquitin domain probably has an important function in regulating ubiquitin-mediated processes5 (reviewed in REF. 3). The ubiquitylation ‘toolkit’

Ubiquitylation is a multistep process (FIG. 1), involving at least three types of enzyme. First, a ubiquitin-activating enzyme (also known as E1) forms a thiol-ester bond with the carboxy-terminal glycine of ubiquitin in an ATP-dependent process. Then, a ubiquitin-conjugating enzyme or ubiquitin-carrier enzyme (UBC, also known as E2) accepts ubiquitin from the E1 by a transthiolation reaction, again involving the carboxyl terminus of ubiquitin. Finally, a ubiquitin protein ligase (E3) catalyses the transfer of ubiquitin from the E2 enzyme to the ε-amino group of a lysine residue on the substrate. Two distinct E3 families, containing conserved protein domains, have now been identified. HECT domain E3s form thiol-ester intermediates with ubiquitin as part of the process, leading to ubiquitylation of substrates (HECT domain stands for homologous to E6-AP carboxyl terminus, E6-AP being the founder member of this family)6. Members of the other class, RING FINGER E3s, are now believed to mediate the direct transfer of ubiquitin from E2 to substrate 7. There are more E2s than E1s, and more E3s than E2s so, at each step, the number of proteins that can potentially be involved increases, as does the specificity of binding to the next component. It is ultimately the E3, either alone or in combination with its bound E2, that determines the exquisite sensitivity of substrate recognition. www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS Before we discuss how these enzymes achieve their high substrate specificities, it is important to appreciate that ubiquitylation is a dynamic and reversible process. De-ubiquitylating enzymes (DUBs) cleave ubiquitin from proteins and from residual proteasome-associated peptides, and disassemble multi-ubiquitin chains. DUBs are also important for processing immature ubiquitin, which is encoded on multiple genes and translated as fusion proteins either with other ubiquitin molecules or as the amino-terminal component of two small ribosomal subunits8. These are processed by members of a subfamily of DUBs — the ubiquitin carboxy-terminal hydrolases — resulting in mature ubiquitin (FIG. 2)9.

BIR REPEAT

(Baculovirus inhibitor of apoptosis repeat). Cysteinebased motif of ~65 amino acids. Inhibitors of apoptosis (IAPs) contain several BIR domains. c-CBL

Multifunctional protein that modulates signalling through tyrosine-kinase-containing growth factor receptors and tyrosine-kinase-coupled receptors. Has RING-fingerdependent E3 activity.

E1: the ubiquitylation starter pack

E1 is the product of a single gene with two isoforms arising from alternative translation start sites10. Sequences contained within the amino-terminal region of the longer isoform, E1a, allow cell-cycle-regulated nuclear localization and phosphorylation, with an increase in nuclear distribution in G2 phase11,12. The finding that cells expressing a temperature-sensitive E1 undergo cell-cycle arrest provided the first evidence for the physiological significance of ubiquitylation13. The carboxy-terminal glycine of ubiquitin is essential for activation by E1, and glycines are also found at the carboxyl termini of UBLs (see the review by Stefan Jentsch and colleagues and REFS 3,4). It comes as no surprise, then, that conjugation of UBLs such as Rub1and SUMO-1 to target proteins also requires E1-like enzymes. These have sequence homology to E1, but the E1-like proteins for SUMO-1 and Rub1 are heterodimers, with subunits homologous to the aminoand carboxy-terminal halves of E1.

Small ribosomal subunits Polyubiquitin fusion protein Ub

Pro-protein processing

Ub Ub

Disassembly of degradation intermediates

Editing of ubiquitylated proteins

Residual peptide

DUB

Figure 2 | The many functions of de-ubiquitylating enzymes. Ubiquitin is synthesized as fusion proteins of ubiquitin (Ub) monomers (polyubiquitin) or with small ribosomal subunits, which are then processed by cleavage at the carboxy-terminal glycine. After the degradation of protein substrates, ubiquitin must be freed from residual peptides and disassembled. Deubiquitylating enzymes also reverse the activity of E3s, sequentially removing ubiquitin from substrates (S) . This might occur in specific cellular locations where ubiquitylation is occurring and at the proteasome (adapted from REF. 9).

E2s: take your partner

Enzyme diversity — implying specificity — becomes apparent in the E2s. Even the modest genome of S. cerevisiae encodes 13 E2-like products, termed Ubc1–13 (ONLINE TABLE 1), and there are at least 25 mammalian family members. But not all E2-like molecules form thiol-esters with ubiquitin: Ubc9 is dedicated to sumoylation, and Ubc12 functions in rubylation; the mammalian orthologues of these E2-like proteins behave similarly. The identifying characteristic of E2s is a 14–16-kDa core that is ~35% conserved among family members (FIG. 3). Whereas several E2s are limited to this core domain, others have significant amino- or carboxyterminal extensions. These might facilitate interactions with specific E3s14,15, or serve as membrane anchors juxtaposing them with specific E3s and substrates16. Most known E2s, including all of those in the S. cerevisiae genome, are less than 36 kDa, but there are notable exceptions. The most striking of these is the giant 528-kDa polytopic E2 BIR-REPEAT-containing ubiquitin-conjugating enzyme (BRUCE)17. The degree of identity among E2s indicates possible redundancy in function. Although there is evidence for this, in some cases quite homologous E2s show considerable differences in their abilities to function with E3s18,19. Beware that E2 nomenclature is not standardized across species. For example, S. cerevisiae Ubc2, Ubc6 and Ubc7 are not closely related to the human UBCH2, UBCH6 and UBCH7, respectively. Arthur Haas and Thomas Siepmann have tried to make sense of this confusion20. The crystal structures of several E2s have been solved, as has the crystal structure of one E2, UBCH7, bound to both the HECT domain of E6-AP and to the RING finger of c-CBL21, 22. These structures have provided insight into how E2s recognize the two types of E3. Intriguingly, these two E3 domains interact with almost identical regions on the E2, specifically loops designated L1 and L2 (FIG. 3). In addition, the E2 amino-terminal α-helix, also involved in interactions with E1 (REF. 20), has a minor part in E3 interactions. The involvement of this helix in interactions with both E1 and E3s indicates that E2s might dissociate from E3s to receive ubiquitin from E1. So how do E3s pair up with specific E2s? Aminoand carboxy-terminal E2 extensions are involved, but so are regions within the E2 core. UBCH7 has a phenylalanine at position 63 that provides a point of hydrophobic interaction between UBCH7’s L1 loop and regions in the E6-AP HECT domain and in the c-Cbl RING domain. The E2-interacting regions of these two E3s seem to be otherwise unrelated (FIG. 3). Notably, phenylalanine 63 of UBCH7 is conserved in a subset of E2s that interact with HECT E3s (REF. 23), and there is also evidence that, for other E2s, the amino acid in this position helps determine E2–E3 pairs22. Interestingly, Ubc3, Ubc7 and their orthologues have 12- and 13amino-acid insertions, respectively, between the active site and the region that corresponds to the L2 loop in other E2s. On the basis of the Ubc7 crystal structure24, and viewed in the context of the UBCH7–E3 crystal structures, it is tempting to speculate that these insertions restrict interactions of these E2s to specific E3s.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 1 7 1

REVIEWS

WW DOMAIN

Protein interaction domain found in the amino-terminal halves of many HECT E3s, and also in other proteins. Characterized by a pair of tryptophans 20–22 amino acids apart, and an invariant proline within a region of 40 amino acids. WW domains interact with proline-rich regions, including those with phosphoserine or phosphothreonine.

HECT E3s

The discovery of the HECT E3s was a direct consequence of the finding that oncogenic strains of human papillomavirus (HPV) encode isoforms of a protein called E6, which specifically inactivate the tumour suppressor protein p53 (ONLINE TABLE 2; FIG. 4). The breakthrough came when E6-associated protein (E6-AP) — a cellular partner for E6 — was identified. E6 serves as an adaptor between E6-AP and p53, allowing E6-AP to catalyse the ubiquitylation of p53 (FIG. 4)25. The characterization of E6-AP led to the identification of a family of proteins that are closely related to E6-AP in a ~350residue region at their carboxyl termini, the HECT domain. This includes a conserved cysteine that forms a covalent thiol-ester intermediate with ubiquitin6. The crystal structure of the E6-AP HECT domain with UBCH7 has a U-shaped appearance with the E2 at one end and the HECT carboxyl terminus at the other (FIG. 3a). The strikingly large distance of 41 Å between the catalytic cysteine of UBCH7 and that of E6-AP leaves much unanswered about how ubiquitin is transferred from E2 to E3. In addition to p53, physiological suba

strates have now been identified for E6-AP including a UDP, HHR23A (REF. 26), and mutations in the E6-AP gene, including those that effect the HECT domain, give rise to Angelman syndrome, a severe neurological disorder27 (ONLINE TABLE 2). Another feature shared by many HECT E3s, but not E6-AP, is the WW DOMAIN, which is involved in protein–protein interactions and undoubtedly has a role in targeting substrates for ubiquitylation. WW domains occur in groups of two to four in the aminoterminal halves of these proteins (ONLINE TABLE 2; FIG. 4). These tryptophan-based motifs form a hydrophobic pocket for proline-rich sequences as well as certain phosphoserine and phosphothreonine-containing sequences28, 29. Most WW domain HECT E3s also have an amino-terminal C2 domain that mediates translocation to the plasma membrane in response to increases in intracellular Ca2+. A function for the C2 domain in membrane translocation of a metazoan member of this family, Nedd4, is well established, with evidence to indicate that this domain might mediate interactions with lipid rafts30. b

E6-AP HECT domain

UBCH7

c-Cbl

UBCH7

C86

C820

C86 L2

Zn L2 L1 N

N c-Cbl RING domain

d H1

P62

UBCH7

P97

UBCH7

P62

A98

P97

A98

F63

E6-AP HECT domain

c-Cbl RING domain

Figure 3 | E2–E3 interactions. Model based on the crystal structure of a | UBCH7 (red) with the HECT domain of E6-AP (blue) and b | with c-Cbl (blue). The structure of UBCH7 is similar to that of other core E2s, which include an amino-terminal α-helix (H1), a 4–5-strand anti-parallel β-pleated sheet (arrows) and a second α-helix that, together with the β-pleated sheet, forms a hydrophobic core. The carboxy-terminal region of E2s is folded into a helix–loop–helix. The conserved catalytic cysteine (C86) is part of a consensus sequence that includes a histidine ten residues upstream of it20. c | Loops L1 and L2 of UBCH7 are involved in E3 interactions with both the HECT domain and the RING. Phe63 (F63) of UBCH7 inserts into a groove on both E3 enzymes. (Models courtesy of Nicola Pavletich, Lan Huang and Ning Zheng, Memorial Sloan-Kettering Cancer Center, New York, USA. Modified from REFS 21, 22.)

172

www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS a

ENaC

c p53

p53

HECT Ub Nedd4

Ub E6-AP

p19ARF MdmX RING Mdm2

F box

P P WD40

Skp1 Rbx1

Ub E2

SUMO-1

Ub Ub

PY E2

IκB

Cul1 Rub1 SCFβTRCP

Figure 4 | Representative E3–substrate interactions a | Association of Nedd4 with the epithelial sodium channel (ENaC) at the plasma membrane. The C2 domain mediates interactions with the membrane in a Ca2+-dependent manner. The WW domain interacts with PY domains on ENaC, which are deleted in Liddle syndrome. b | Ternary complex of p53 with human papillomavirus E6 and E6-AP. c | Mdm2 and p53. Both p53 and Mdm2 are substrates for modification by Mdm2 (BOX 1). MdmX blocks binding through the RING finger. p19ARF blocks by binding upstream of the RING and by revealing a nucleolar localization signal. d | SCFβTRCP as a prototypical cullin-containing E3. Modification of Cul1 with Rub1 increases activity and requires the RING finger protein Rbx1 (BOX 2).

Rsp5 is the only S. cerevisiae C2–WW domain HECT E3 and exemplifies the capacity of a single E3 to ubiquitylate distinct proteins in several cellular compartments. Rsp5 interacts with one of its substrates, the large subunit of RNA polymerase II (LsPolII), directly through its WW domains, which bind the proline-rich carboxyl terminus of LsPolII (REF. 29). A second function of Rsp5 is to activate two transcription factors, Spt23 and Mga2, by facilitating the ubiquitin- and proteasome-dependent cleavage of the soluble components of these proteins from their ER-membrane-bound precursors31. This concept of limited cleavage by proteasomes has a precedent in the maturation of a metazoan transcription factor, NF-κB32. Rsp5 is best known for its ability to ubiquitylate at least 13 plasma membrane transporters and receptors. Surprisingly, however, there is little evidence for direct interactions between these targets and Rsp5. This suggests that these interactions are indirect and perhaps facilitated by C2-mediated membrane targeting. In contrast to Rsp5, a direct WW domain–substrate interaction is important for the ubiquitylation of at least one membrane protein: Nedd4 binds and ubiquitylates subunits of the epithelial sodium channel (ENaC) through its WW domains, leading to downregulation of the number of active channels (FIG. 4a). Mutation of proline-rich regions on ENaC causes Liddle syndrome, an inherited form of hypertension in which ENaC activity is enhanced, presumably owing to the inability of Nedd4 to downregulate ENaC29,33,34 (also see the review by Linda Hicke.) F-BOX

A conserved ~50-residue region found in proteins that associate with Skp1 and potentially form the SCF E3s. There are over a hundred distinct members of this family. CULLIN FAMILY

Proteins with homology to Cul1, which was first shown to be involved in cell-cycle exit in Caenorhabditis elegans.

RING finger E3s

Unlike the HECT domain, the RING finger was described in the early 1990s, years before any suspicion of a role in ubiquitylation. RING fingers include eight metal-binding residues that coordinate two zinc ions, arranged in an interleaved pattern35. This distinguishes them from the tandem arrangement of metal-coordinating residues characteristic of zinc fingers. The realization that the RING finger has a general role in ubiquitylation has come about during the past two years from the convergence of a number of lines of investigation

that included: first, the discovery that a small RING finger protein, Rbx1 (Ring box protein-1; also known as ROC1 or Hrt1), was a requisite component of the multisubunit SCF (Skp1/Cul1/F-BOX protein) family E3s36–40; second, the finding that many otherwise unrelated RING finger proteins mediate ubiquitylation41; and last, the realization that all known or suspected E3s that are not HECT proteins include a RING finger15, 42–48. We do not know how many of the hundreds of RING finger proteins have the capacity to mediate ubiquitylation. Nonetheless, a sample of otherwise unrelated members of this family predicts that it will be a large percentage41 (ONLINE TABLE 3). So far, most RING finger proteins that have been shown to interact with E2s and to mediate ubiquitylation in in vitro systems lack defined substrates other than themselves. Prominent among these is the product of the breast and ovarian cancer susceptibility gene 1 ( BRCA1)41: mutations in this protein — including one in the RING finger — are found in familial forms of breast and ovarian cancer49. Ascertaining which E2-interacting RING finger proteins are bona fide E3s for heterologous substrates, and which are primarily substrates for regulated, E2-dependent,‘auto-ubiquitylation’ is an exciting challenge. Unlike the HECT-domain E3s, where roles for thiolester intermediates with ubiquitin are well established, there is little evidence to indicate the existence of similar intermediates between ubiquitin and RING finger proteins. So is the RING finger simply an E2-docking site that passively juxtaposes the carboxyl terminus of ubiquitin bound to E2 with lysines on substrates, or does the RING allosterically activate E2 bound to ubiquitin and thereby enhance transfer? Although there is some experimental evidence for a possible activating function15, a comparison of the structure of an E2 (UBCH7) bound to c-Cbl to that of E2s by themselves22 (FIG. 3) provides little support for this. So, at present there is no clear answer to this question. A convenient way to think about RING finger proteins is to divide them into single and multisubunit E3s. Single-subunit E3s contain the substrate recognition element and the RING finger on the same polypeptide.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 1 7 3

REVIEWS and UDP families. Mutations in Parkin’s RINGs are associated with juvenile Parkinson’s disease, and a synaptic-vesicle-associated protein (CDCrel-1) has been identified52,53 as a substrate for this E3. The compact RING finger is found in diverse, otherwise unrelated, proteins. It therefore follows that the sites of substrate interaction for RING proteins will be highly varied. For example, the interactions of c-Cbl depend on its atypical SH2 domain, and for the IAPs the BIR domain probably facilitates binding of some substrates. For Mdm2, interactions with p53 occur through its amino-terminal domain, whereas the RING is located at its carboxyl terminus. Mdm2 is illustrative of the complex regulation that can be a feature of single-subunit RING finger E3s (BOX 1).

Box 1 | Mdm2 and ubiquitylation As might be expected for a protein whose task is normally to destroy the ‘guardian of the genome’, there are numerous safety mechanisms that prevent Mdm2 from running amok. First, Mdm2 has intrinsic RING-finger-dependent E3 activity towards itself, as well as p53 (REFS 42,43). Second, phosphorylation of p53 blocks its interaction with Mdm2 (REFS 85, 86). Third, an Mdm2-binding protein, p19ARF, binds upstream of its RING finger and exposes a cryptic nucleolar localization signal co-linear with the Mdm2 RING finger. This sequesters Mdm2 away from p53, preventing the Mdm2mediated degradation of p53 (REFS 87, 88). Binding of p19ARF also inhibits the intrinsic activity of Mdm2 (REFS 89,90) in vitro. Last, Mdm2 activity is similarly inhibited by dimerization with a related RING finger protein, MdmX (REFS 91–93). Nevertheless, there is evidence that Mdm2 is also subject to positive regulation by modification with SUMO-1, which seems to enhance ubiquitylation of p53 by Mdm2 while diminishing auto-ubiquitylation94. Mdm2 also illustrates the substrate specificity of RING finger E3s. The p53 family member p73 binds Mdm2 but is stabilized by this interaction rather than targeted for degradation95–97. Substitution of a heterologous RING for that of Mdm2 reconstitutes auto-ubiquitylation and proteasomal targeting of the chimeric molecule; however it does not ubiquitylate p53 or target it for degradation42.

Multisubunit cullin-containing RING E3s

Exploration into the intricacies of the cell cycle led to the discovery of multisubunit SCF E3s (TABLE 1, FIG. 4) and to the discovery of the anaphase-promoting complex (APC) or cyclosome, which includes at least 12 distinct subunits. A missing link in the function of SCF E3s was provided in 1999 with the identification of a noncanonical RING finger protein, Rbx1, as a component of both SCF and the structurally related von Hippel– Lindau–Cul2/elongin B/elongin C (VHL–CBC) complex36–40. In retrospect, it became obvious that the small RING finger protein Apc11 functions in a similar capacity in the APC, and indeed this subunit has activity towards substrates in vitro48,54,107. An emerging concept is that the cullin family proteins intrinsic to these E3 complexes (Apc2 in the APC) interact with linker proteins that recruit the substrate-recognition components (see

Multisubunit E3s all include a small RING finger protein and a member of the CULLIN family of proteins as well as other subunits, some of which recognize substrates (FIG. 3). Single-subunit RING finger proteins include well-studied E3s such as the oncoprotein Mdm2, which ubiquitylates p53 (REFS 42, 43), the protooncoprotein c-Cbl, which ubiquitylates growth factor receptors44–46, and the inhibitors of apoptosis (IAPs)50, 51. Parkin is a RING finger E3 that has two RINGs at its carboxyl terminus separated by an IBR (in-between RING), a region common to proteins that have two RING fingers. Parkin also has an amino-terminal ubiquitin domain, making it a member of both the RING

Table 1 | Multisubunit, Cullin-containing RING E3s* SCF

VCB-CUL2

APC

Elongin B F-box protein Skp1

VHL

Rbx1 Ubc3 Cul1

Rbx1

Cdc20/Hct1 Ubc H5A

Cul2 Elongin C

Apc11 UbcII/ UbcX Apc2

APC subunits

RING

Rbx1 (Hrt1/Roc1)

Apc11

Cullin

Cdc53 (Cul1) ‡

Cul2 ‡

Apc2

Adaptors

Skp1

Elongin B: homologous to amino terminus of Skp1. Elongin C: a UDP.

Multiple APC subunits (pink), some with tetratricopeptide repeats. These presumably have adaptor functions.

Ubc3 (Cdc34)

UbcH5A, others?

Ubc11, UbcX

Substrate recognition

F-box proteins. These include VHL, possibly other SOCS those with WD40 repeats, box-containing proteins. leucine-rich domains and others.

Cdc20 (Fizzy) and Hct1 (Fizzy-related); both contain WD40 repeats.

Substrates (partial list)

Sic1, IκBα, β-catenin, G1 cyclins, CD4 bound to phosphorylated HIV Vpu, others.

Mitotic cyclins, Pds1, Cut2, Ase1, Scc1, Securin, others.

HIF1α

*See REFS 55–57 for comprehensive reviews on the multisubunit RING finger E3s. ‡ Modified with Rub1, which is mediated by Ubc12 and Rbx1 — evidence suggests that this increases E3 activity. Other cullin family members are similarly modified99–106 (reviewed in REF. 3).

174

www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS

Box 2 | Examples of de-ubiquitylating enzymes • Fat facets: De-ubiquitylating enzyme (DUB) implicated in Drosophila melanogaster photoreceptor development. • DUB1 and DUB2: Products of immediate-early response genes expressed in response to cytokines (IL-3, IL-5, and GM-CSF in the case of DUB1, and IL-2 in the case of DUB2). • BAP1: Binds to the amino-terminal RING finger domain of BRCA1. • UBPY: Human growth-regulated DUB. • Ap-UCH: Aplysia neuronal DUB implicated in long-term facilitation. • Ubp-M: Implicated in cell-cycle progression, phosphorylated in a cell-cycledependent manner. • Ubp4: Mutants defective in degradation of the yeast mating-type factor MATα2. • Ubp3: Regulation of gene silencing. • D-UBP-64E: Position effect variegation in Drosophila. • UCH-L1: Mutations associated with neurological disorders. • TRE-2: A mammalian proto-oncogene homologous to yeast Doa4. The latter is a proteasome-associated DUB that is implicated in removing ubiquitin from postproteolyic proteasome-bound peptides68. • Isopeptidase T: Disassembles free multi-ubiquitin chains beginning with the most proximal ubiquitin. A free carboxy-terminal glycine is required. In Alzheimer’s disease, a frameshift in translation results in the generation of ‘ubiquitin+1’, which lacks this glycine. Ubiquitin+1 can serve as a substrate for generation of isopeptidase-T-resistant multi-ubiquitin chains98. DUBs and their functions are reviewed in REFS 9, 67. REFS 55–57 for reviews on the multisubunit RING finger

IκBα Inhibitory subunit of the NFκB transcription factor. It is phosphorylated, ubiquitylated and degraded in response to stimuli that activate NF-κB. SOCS BOX

Suppressor of cytokine signalling box first identified in an inhibitor of Jak family kinases.

E3s) (ONLINE TABLE 2). SCF E3s recognize and ubiquitylate a diverse group of phosphoproteins, with substrate specificity conferred by members of the large family of F-box proteins58. SCF substrates are generally phosphoproteins, but phosphorylation is not an inherent requirement for ubiquitylation by SCF E3s, as shown through the use of engineered F-box proteins59. There are examples in which one F-box protein is responsible for recognizing several substrates, as is the case for β-transducin-repeat-containing protein (βTRCP). SCF βTRCP recognizes phosphorylated β-catenin and IκBα. Additionally, reminiscent of the E6-AP and p53 story, nascent forms of the HIV receptor CD4 are indirectly targeted for ubiquitylation in the ER membrane by SCFβTRCP owing to the binding of CD4 to HIV-encoded Vpu, which has phosphorylation sites akin to those of β-catenin and IκBα55. Some F-box proteins are themselves ubiquitylated and targeted for degradation. Whether this downregulates their levels or facilitates proteasomal targeting of associated phosphoproteins awaits determination. A recent provocative observation is that ubiquitylation of the transcription factor Met4 by SCFMet30 leads to the functional inactivation of Met4, but not to its proteolysis60. Architecturally related to the SCF E3s is the VHL–CBC complex (TABLE 1). In this complex, the adaptor Skp1 is replaced by the dimer of elongin B, which has homology to Skp1, and elongin C, which is a UDP. Notably, VHL mutants that fail to assemble with the CBC core are associated with the malignancies of von Hippel–Lindau disease61, 62. An important substrate for VHL–CBC is hypoxia-inducible transcription factor

1α (HIF1α), which positively regulates vascular endothelial growth factor (VEGF), providing an explanation for the highly vascular nature of the clear cell renal carcinomas seen in VHL disease63–65. Analogous to the F box, the VHL protein contains a suppressor of cytokine signalling (SOCS) box that interacts with the core of this E3. It might be that other SOCS-containing proteins can replace the VHL and allow for recognition of other specific substrates66. The most complicated of the multisubunit E3s is the APC. The first identified substrates for this E3 were mitotic cyclins, but the list of substrates is growing (TABLE 1). In S. cerevisiae, at least 12 essential APC components have been identified. Although the intricacies of the APC’s architecture are largely unknown, there are substantial parallels to the SCF and VHL–CBC E3s (TABLE 1). Phosphorylation and dephosphorylation are known to be important regulators of APC activity56. Several functions for DUBs

One lesson learned from studying phosphorylation is that the removal of phosphate groups can be as tightly regulated as their addition. Knowing that ubiquitylation is a reversible process, we might expect similarly tight controls for removal of ubiquitin. It should come as no surprise, then, that there are at least 19 yeast DUBs and substantially more in mammals. DUBs come in two flavours — ubiquitin carboxy-terminal hydrolases (UCHs) and ubiquitin-specific processing enzymes (UBPs) — both of which are thiol proteases. UCHs catalyse the removal of carboxy-terminal fusion proteins from ubiquitin (recall that ubiquitin is always translated as a fusion protein), with a preference for substrates in which ubiquitin is fused to small peptides. UBPs are generally larger, thought of as being responsible for removing ubiquitin from larger proteins, and are involved in the disassembly of multi-ubiquitin chains9,67. At the proteasome, DUBs cleave multi-ubiquitin chains from residual peptides68 and shorten proteinbound multi-ubiquitin chains by sequentially removing the terminal ubiquityl group69. This ‘proof-reading’ function ensures that highly ubiquitylated proteins preferentially remain associated with the proteasome. Another important function of DUBs is to prevent the accumulation of residual multi-ubiquitin chains at proteasomes. Failure to disassemble these chains has the potential to wreak havoc upon the normal movement of ubiquitylated proteins to and through the proteasome (FIG. 2). Moreover, DUBs are constitutively active in the removal of ubiquitin from substrates, as inhibition of proteasome function causes the accumulation of mostly non-ubiquitylated proteins. It is clear that DUBs have crucial cellular roles, but beyond general housekeeping, is there any evidence that certain DUBs have functions in specific cellular processes? It is early days, but there are examples in both yeast and higher eukaryotes (BOX 2). Different types of ubiquitin signal

The way in which ubiquitin is linked to proteins has the potential to alter their fate (FIG. 5). A single protein can be modified on one or more lysines with a single

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 1 7 5

REVIEWS

Ubiquitin

N Linkage

Lys 11

48 Many E2s

E2 ?

Many E3s

Substrate

Many substrates

Process

63 Ubc2, Ubc13

Rad5 Rad18

Substrates ?

L28

DNA repair

Translation

Ubc13

Ubc1, Ubc4, Ubc5

TRAF6

IκB kinase activation

Rsp5

Membrane proteins Endocytosis and transport

Proteasomemediated degradation

Figure 5 | Different functions for different ubiquitin linkages. Schematic representation of ubiquitin with lysines and roles of different linkages. Ubiquitin has several lysine residues and, in vivo, can form multi-ubiquitin chains linked through positions 11, 29, 48 and 63. The functions of Lys11 and Lys29-linked chains are unknown. Lys48-linked chains target proteins to the proteasome but might have other functions, and Lys63-linked chains have a range of fates.

ubiquitin (monoubiquitylation; see the review by Linda Hicke), with lysine-linked chains of ubiquitin (multi-ubiquitylation) or combinations of the two. As mentioned above, only multi-ubiquitin chains target proteins for proteasomal degradation, with multiubiquitin chains of four or more ubiquitin molecules linked through lysine 48 (K48) being adequate as a proteasome-targeting signal. Such chains are formed by isopeptide linkages between a lysine on the last ubiquitin of a growing chain with the carboxy-terminal glycine of a new ubiquitin molecule. How E3s mediate both the transfer of ubiquitin to a lysine on a substrate and also add ubiquitin to a growing end of a multiubiquitin chain of more than ten ubiquitins is poorly understood, but an accessory factor (E4) that facilitates the formation of multi-ubiquitin chains for one yeast E3 has been identified70. However, E4s are not generally required for the formation of multi-ubiquitin chains. The choice of lysine is also an important decision when building up a multi-ubiquitin chain because ubiquitin itself has seven conserved lysine residues, all of which are potential sites of isopeptide linkage to the carboxyl terminus of another ubiquitin. In vivo, K11, K29, K48 and K63 all can form ubiquitin–ubiquitin linkages. K48-linked multi-ubiquitin chains are potent targeting signals that lead to recognition and degradation of proteins by proteasomes. K63 linkages, however, are apparently not proteasome-targeting signals: instead, they are important for DNA repair71 and other functions. A specific E2, Ubc13, functions together with Mms2, a molecule that structurally resembles an E2 but lacks the canonical cysteine, to generate K63 chains72. These proteins are implicated in DNA repair together with Ubc2 (Rad6) and two RING finger proteins, Rad5 and Rad18 (REFS 73, 74). A recent and surprising observation is that a ribosomal subunit, L28, is a major substrate for modification with K63-linked chains. L28 ubiquitylation, which is most prominent during S phase of the cell cycle, is stimulated by irradiation, is reversible, and enhances translation75. Activation of K63 multi-ubiqui-

176

tin chain formation by a RING finger E3, TRAF6 (TNFreceptor-associated factor 6, where TNF stands for tumour necrosis factor), results in the activation of IκB kinase76. K63-linked chains also have roles in the endocytosis and targeting for vacuolar degradation of yeast transporters (reviewed in REFS 29, 33; see the review by Linda Hicke). Most proteins have many lysine residues, so can the position of a ubiquitin signal on a protein affect its fate in different ways? In some cases, specific lysines on proteins are ubiquitylation targets77,78, whereas in others there is little specificity79,80. Furthermore, there are now several examples where the amino termini of proteins, rather than lysines, can serve as ubiquitylation sites81. We do not know how or whether the site of ubiquitylation on a protein, like the nature of the multi-ubiquitin linkages, affects its eventual fate. Destination: proteasome

Alluring as these variations on a theme are, to our knowledge most ubiquitylated proteins have K48-linked chains and are recognized by 26S proteasomes. The catalytic component of this remarkable and highly complex structure is a cylindrical chamber of 28 subunits (the 20S core) that includes two copies each of subunits with trypsin, chymotrypsin, and peptidylglutamyl peptidaselike activities (see the review by Peter Kloetzel). The 20S core is capped at each end by a multisubunit regulatory complex, the 19S cap. This multisubunit cap fulfils several roles, including recognition of multi-ubiquitin chains and some UDPs, and also allows for the ubiquitin-independent proteasomal targeting of ornithine decarboxylase (see the review by Philip Coffino on page 188 of this issue). There is compelling evidence for the ubiquitinindependent, proteasome-dependent degradation of at least one other protein, p21Cip1 (REF. 82). Whether the 19S cap has a function in the proteasomal targeting of p21Cip1 remains unknown. A cap component that recognizes multi-ubiquitin chains — S5A — has been identified, but genetic evidence indicates that there might be other multi-ubiquitin recognition elements within the cap83. The 19S cap also contains DUBs and multiple ATPases. It should be appreciated that the proteasome is a dynamic structure that is modified, for example, in response to the inflammatory cytokine interferon-γ. Proteasomal degradation is not limited to ubiquitylated cytosolic and nuclear proteins. Ubiquitylation also targets ER lumenal and membrane proteins for degradation. Specific E2s and a yeast RING finger protein are among the proteins involved in this process33,84. Understanding how ubiquitylation and proteasomal degradation — processes that do not occur in the ER lumen — contribute to the retrograde movement of proteins out of the ER and their concomitant degradation is a topological puzzle that awaits resolution. From proteolysis to proteomics

Ubiquitin-mediated regulated protein degradation is essential to virtually all aspects of eukaryotic cell biology. We now know that HECT domains and a substantial number of RING fingers are E3 modules and that Fwww.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS box and possibly SOCS-box proteins are substraterecognition elements for multisubunit E3s. These insights have coincided with a massive increase in the rate at which deduced protein sequences are becoming available through the genome projects — a resource that we will need to define the complex network of components, substrates and regulators of the ubiquitylation system. This information should prove especially useful as we develop more sophisticated tools, such as protein arrays, to probe differences in cellular protein levels, allowing us to identify proteins that undergo accelerated or delayed degradation in disease in much the same way that we now use DNA microarrays to probe for differences in gene expression. It is now clear that ubiquitylation is much more than a proteasomal targeting signal. How it mediates responses to DNA damage, facilitates endosomal transport, and increases the efficiency of translation are all open questions, as is the role of UDPs in ubiquitin-mediated processes. The realization that UBLs are also conjugated to proteins using similar tools, and that some UBLs modulate ubiquitylation, makes it evident that cells have

4. 5.

10.

11.

12.

13.

14.

15.

Thrower, J. S., Hoffman, L., Rechsteiner, M. & Pickart, C. M. Recognition of the polyubiquitin proteolytic signal. EMBO J. 19, 94–102 (2000). Loeb, K. R. & Haas, A. L. Conjugates of ubiquitin crossreactive protein distribute in a cytoskeletal pattern. Mol. Cell. Biol. 14, 8408–8419 (1994). Jentsch, S. & Pyrowolakis, G. Ubiquitin and its kin: how close are the family ties? Trends Cell Biol. 10, 335–342 (2000). Yeh, E. T., Gong, L. & Kamitani, T. Ubiquitin-like proteins: new wines in new bottles. Gene 248, 1–14 (2000). Kleijnen, M. F. et al. The hPLIC proteins may provide a link between the ubiquitination machinery and the proteasome. Mol. Cell 6, 409–419 (2000). Huibregtse, J. M., Scheffner, M., Beaudenon, S. & Howley, P. M. A family of proteins structurally and functionally related to the E6-AP ubiquitin-protein ligase. Proc. Natl Acad. Sci. USA. 92, 2563–2567 (1995). Joazeiro, C. A. & Weissman, A. M. RING finger proteins: mediators of ubiquitin ligase activity. Cell 102, 549–552 (2000). Finley, D., Bartel, B. & Varshavsky, A. The tails of ubiquitin precursors are ribosomal proteins whose fusion to ubiquitin facilitates ribosome biogenesis. Nature 338, 394–401 (1989). Wilkinson, K. D. Ubiquitination and deubiquitination: targeting of proteins for degradation by the proteasome. Semin. Cell. Dev. Biol. 11, 141–148 (2000). Handley-Gearhart, P. M., Stephen, A. G., Trausch-Azar, J. S., Ciechanover, A. & Schwartz, A. L. Human ubiquitinactivating enzyme, E1. Indication of potential nuclear and cytoplasmic subpopulations using epitope-tagged cDNA constructs. J. Biol. Chem. 269, 33171–33178 (1994). Nagai, Y. et al. Ubiquitin-activating enzyme, E1, is phosphorylated in mammalian cells by the protein kinase Cdc2. J. Cell Sci. 108, 2145–2152 (1995). Grenfell, S. J., Trausch-Azar, J. S., Handley-Gearhart, P. M., Ciechanover, A. & Schwartz, A. L. Nuclear localization of the ubiquitin-activating enzyme, E1, is cell-cycledependent. Biochem. J. 300, 701–708 (1994). Finley, D., Ciechanover, A. & Varshavsky, A. Thermolability of ubiquitin-activating enzyme from the mammalian cell cycle mutant ts85. Cell 37, 43–55 (1984). This is a landmark study that provided the first evidence of a role for the ubiquitin-conjugating system in cellular function. Mathias, N., Steussy, C. N. & Goebl, M. G. An essential domain within Cdc34p is required for binding to a complex containing Cdc4p and Cdc53p in Saccharomyces cerevisiae. J. Biol. Chem. 273, 4040–4045 (1998). Xie, Y. & Varshavsky, A. The E2–E3 interaction in the N-end rule pathway: the RING-H2 finger of E3 is required for the synthesis of multiubiquitin chain. EMBO J. 18, 6832–6844 (1999).

16.

17.

18.

19.

20. 21.

22.

23.

24.

25.

26.

27.

This study establishes a role for the RING finger in the activity of the N-end rule E3. Sommer, T. & Jentsch, S. A protein translocation defect linked to ubiquitin conjugation at the endoplasmic reticulum. Nature 365, 176–179 (1993). This study describes Ubc6 and provided the first suggestion of a linkage between the ubiquitinconjugating system and the degradation of proteins from the endoplasmic reticulum. Hauser, H. P., Bardroff, M., Pyrowolakis, G. & Jentsch, S. A giant ubiquitin-conjugating enzyme related to IAP apoptosis inhibitors. J. Cell Biol. 141, 1415–1422 (1998). Gonen, H. et al. Identification of the ubiquitin carrier proteins, E2s, involved in signal-induced conjugation and subsequent degradation of IκBα. J. Biol. Chem. 274, 14823–14830 (1999). Schwarz, S. E., Rosa, J. L. & Scheffner, M. Characterization of human hect domain family members and their interaction with UbcH5 and UbcH7. J. Biol. Chem. 273, 12148–12154 (1998). Haas, A. L. & Siepmann, T. J. Pathways of ubiquitin conjugation. FASEB J. 11, 1257–1268 (1997). Huang, L. et al. Structure of an E6AP–UbcH7 complex: insights into ubiquitination by the E2-E3 enzyme cascade. Science 286, 1321–1326 (1999). Zheng, N., Wang, P., Jeffrey, P. D. & Pavletich, N. P. Structure of a c-Cbl–UbcH7 complex: RING domain function in ubiquitin-protein ligases. Cell 102, 533–539 (2000). References 21 and 22 describe the crystal structures of a HECT domain and a RING finger protein with an E2. Nuber, U. & Scheffner, M. Identification of determinants in E2 ubiquitin-conjugating enzymes required for hect E3 ubiquitin-protein ligase interaction. J. Biol. Chem. 274, 7576–7582 (1999). Cook, W. J., Martin, P. D., Edwards, B. F., Yamazaki, R. K. & Chau, V. Crystal structure of a class I ubiquitin conjugating enzyme (Ubc7) from Saccharomyces cerevisiae at 2.9 angstroms resolution. Biochemistry 36, 1621–1627 (1997). Scheffner, M., Huibregtse, J. M., Vierstra, R. D. & Howley, P. M. The HPV-16 E6 and E6-AP complex functions as a ubiquitin-protein ligase in the ubiquitination of p53. Cell 75, 495–505 (1993). This is a seminal paper describing the characterization of E6-AP, the first described member of the HECT domain family of E3s. Kumar, S., Talis, A. L. & Howley, P. M. Identification of HHR23A as a substrate for E6-associated protein-mediated ubiquitination. J. Biol. Chem. 274, 18785–18792 (1999). Kishino, T., Lalande, M. & Wagstaff, J. UBE3A/E6-AP mutations cause Angelman syndrome. Nature Genet. 15, 70–73 (1997).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

28. Kay, B. K., Williamson, M. P. & Sudol, M. The importance of being proline: the interaction of proline-rich motifs in signaling proteins with their cognate domains. FASEB J. 14, 231–241 (2000). 29. Rotin, D., Staub, O. & Haguenauer-Tsapis, R. Ubiquitination and endocytosis of plasma membrane proteins: role of Nedd4/Rsp5p family of ubiquitin-protein ligases. J. Membr. Biol. 176, 1–17 (2000). 30. Plant, P. J. et al. Apical membrane targeting of Nedd4 is mediated by an association of its C2 domain with annexin XIIIb. J. Cell Biol. 149, 1473–1484 (2000). 31. Hoppe, T. et al. Activation of a membrane-bound transcription factor by regulated ubiquitin/proteasomedependent processing. Cell 102, 577–586 (2000). 32. Orian, A. et al. Ubiquitin-mediated processing of NF-κB transcriptional activator precursor p105. Reconstitution of a cell-free system and identification of the ubiquitin-carrier protein, E2, and a novel ubiquitin-protein ligase, E3, involved in conjugation. J. Biol. Chem. 270, 21707–21714 (1995). 33. Bonifacino, J. S. & Weissman, A. M. Ubiquitin and the control of protein fate in the secretory and endocytic pathways. Annu. Rev. Cell. Dev. Biol. 14, 19–57 (1998). 34. Kamynina, E., Debonneville, C., Bens, M., Vandewalle, A. & Staub, O. A novel mouse Nedd4 protein suppresses the activity of the epithelial Na+ channel. FASEB J. 15, 204–214 (2001). 35. Freemont, P. S. RING for destruction? Curr. Biol. 10, R84–R87 (2000). 36. Kamura, T. et al. Rbx1, a component of the VHL tumor suppressor complex and SCF ubiquitin ligase. Science 284, 657–661 (1999). 37. Ohta, T., Michel, J. J., Schottelius, A. J. & Xiong, Y. ROC1, a homolog of APC11, represents a family of cullin partners with an associated ubiquitin ligase activity. Mol. Cell 3, 535–541 (1999). 38. Tan, P. et al. Recruitment of a ROC1-CUL1 ubiquitin ligase by Skp1 and HOS to catalyze the ubiquitination of IκBα. Mol. Cell 3, 527–533 (1999). 39. Skowyra, D. et al. Reconstitution of G1 cyclin ubiquitination with complexes containing SCFGrr1 and Rbx1. Science 284, 662–665 (1999). 40. Seol, J. H. et al. Cdc53/cullin and the essential hrt1 RINGH2 subunit of SCF define a ubiquitin ligase module that activates the E2 enzyme cdc34. Genes Dev. 13, 1614–1626 (1999). References 36–40 all describe the characterization of a small RING finger protein as an integral component of SCF E3s. 41. Lorick, K. L. et al. RING fingers mediate ubiquitinconjugating enzyme (E2)-dependent ubiquitination. Proc. Natl Acad. Sci. USA 96, 11364–11369 (1999). This study suggests a general role for RING fingers in ubiquitylation.

VOLUME 2 | MARCH 2001 | 1 7 7

REVIEWS

42. Fang, S., Jensen, J. P., Ludwig, R. L., Vousden, K. H. & Weissman, A. M. Mdm2 is a RING finger-dependent ubiquitin protein ligase for itself and p53. J. Biol. Chem. 275, 8945–8951 (2000). 43. Honda, R. & Yasuda, H. Activity of MDM2, a ubiquitin ligase, toward p53 or itself is dependent on the RING finger domain of the ligase. Oncogene 19, 1473–1476 (2000). 44. Waterman, H., Levkowitz, G., Alroy, I. & Yarden, Y. The RING finger of c-Cbl mediates desensitization of the epidermal growth factor receptor. J. Biol. Chem. 274, 22151–22154 (1999). 45. Joazeiro, C. A. et al. The tyrosine kinase negative regulator c-Cbl as a RING-type, E2- dependent ubiquitin-protein ligase. Science 286, 309–312 (1999). 46. Yokouchi, M. et al. Ligand-induced ubiquitination of the epidermal growth factor receptor involves the interaction of the c-Cbl RING finger and UbcH7. J. Biol. Chem. 274, 31707–31712 (1999). 47. Hu, G. & Fearon, E. R. Siah-1 N-terminal RING domain is required for proteolysis function, and C-terminal sequences regulate oligomerization and binding to target proteins. Mol. Cell. Biol. 19, 724–732 (1999). References 42–47 demonstrate a role for the RING finger in a variety of known and suspected E3s. 48. Zachariae, W. et al. Mass spectrometric analysis of the anaphase-promoting complex from yeast: identification of a subunit related to cullins. Science 279, 1216–1219 (1998). 49. Brzovic, P. S., Meza, J., King, M. C. & Klevit, R. E. The cancer-predisposing mutation C61G disrupts homodimer formation in the NH2-terminal BRCA1 RING finger domain. J. Biol. Chem. 273, 7795–7799 (1998). 50. Yang, Y., Fang, S., Jensen, J. P., Weissman, A. M. & Ashwell, J. D. Ubiquitin protein ligase activity of IAPs and their degradation in proteasomes in response to apoptotic stimuli. Science 288, 874–877 (2000). 51. Hwang, H. K. et al. The inhibitor of apoptosis, cIAP2, functions as a ubiquitin-protein ligase and promotes in vitro monoubiquitination of caspases 3 and 7. J. Biol. Chem. 275, 26661–26664 (2000). 52. Shimura, H. et al. Familial Parkinson’s disease gene product, Parkin, in a ubiquitin-protein ligase. Nature Genet. 25, 302–305 (2000). 53. 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). 54. Leverson, J. D. et al. The APC11 RING-H2 finger mediates E2-dependent ubiquitination. Mol. Biol. Cell 11, 2315–2325 (2000). 55. Deshaies, R. J. SCF and Cullin/Ring H2-based ubiquitin ligases. Annu. Rev. Cell. Dev. Biol. 15, 435–467 (1999). 56. Page, A. M. & Hieter, P. The anaphase-promoting complex: new subunits and regulators. Annu. Rev. Biochem. 68, 583–609 (1999). 57. Tyers, M. & Jorgensen, P. Proteolysis and the cell cycle: with this RING I do thee destroy. Curr. Opin. Genet. Dev. 10, 54–64 (2000). 58. Bai, C. et al. SKP1 connects cell cycle regulators to the ubiquitin proteolysis machinery through a novel motif, the F-box. Cell 86, 263–274 (1996). 59. Zhou, P., Bogacki, R., McReynolds, L. & Howley, P. M. Harnessing the ubiquitination machinery to target the degradation of specific cellular proteins. Mol. Cell 6, 751–756 (2000). 60. Kaiser, P., Flick, K., Wittenberg, C. & Reed, S. I. Regulation of transcription by ubiquitination without proteolysis: Cdc34/SCFMet30-mediated inactivation of the transcription factor Met4. Cell 102, 303–314 (2000). 61. Lisztwan, J., Imbert, G., Wirbelauer, C., Gstaiger, M. & Krek, W. The von Hippel-Lindau tumor suppressor protein is a component of an E3 ubiquitin-protein ligase activity. Genes Dev. 13, 1822–1833 (1999). 62. Iwai, K. et al. Identification of the von Hippel-lindau tumorsuppressor protein as part of an active E3 ubiquitin ligase complex. Proc. Natl Acad. Sci. USA 96, 12436–12441 (1999). 63. Ohh, M. et al. Ubiquitination of hypoxia-inducible factor requires direct binding to the β-domain of the von HippelLindau protein. Nature Cell Biol. 2, 423–427 (2000). 64. Cockman, M. E. et al. Hypoxia inducible factor-α binding and ubiquitylation by the von Hippel-Lindau tumor

178

65.

66.

67.

68.

69.

70. 71.

72.

73.

74.

75.

76.

77.

78.

79.

80.

81.

82.

83.

84.

85.

suppressor protein. J. Biol. Chem. 275, 25733–25741 (2000). Kamura, T. et al. Activation of HIF1α ubiquitination by a reconstituted von hippel-lindau (VHL) tumor suppressor complex. Proc. Natl Acad. Sci. USA 97, 10430–10435 (2000). References 61–65 establish the VHL–CBC complex as an E3 and show that HIF1α is a substrate. Kamura, T. et al. The Elongin BC complex interacts with the conserved SOCS-box motif present in members of the SOCS, ras, WD-40 repeat, and ankyrin repeat families. Genes Dev. 12, 3872–3881 (1998). Chung, C. H. & Baek, S. H. Deubiquitinating enzymes: their diversity and emerging roles. Biochem. Biophys. Res. Commun. 266, 633–640 (1999). Papa, F. R. & Hochstrasser, M. The yeast DOA4 gene encodes a deubiquitinating enzyme related to a product of the human tre-2 oncogene. Nature 366, 313–319 (1993). Lam, Y. A., Xu, W., DeMartino, G. N. & Cohen, R. E. Editing of ubiquitin conjugates by an isopeptidase in the 26S proteasome. Nature 385, 737–740 (1997). Koegl, M. et al. A novel ubiquitination factor, E4, is involved in multiubiquitin chain assembly. Cell 96, 635–644 (1999). Spence, J., Sadis, S., Haas, A. L. & Finley, D. A ubiquitin mutant with specific defects in DNA repair and multiubiquitination. Mol. Cell. Biol. 15, 1265–1273 (1995). Hofmann, R. M. & Pickart, C. M. Noncanonical MMS2encoded ubiquitin-conjugating enzyme functions in assembly of novel polyubiquitin chains for DNA repair. Cell 96, 645–653 (1999). Bailly, V., Lauder, S., Prakash, S. & Prakash, L. Yeast DNA repair proteins Rad6 and Rad18 form a heterodimer that has ubiquitin conjugating, DNA binding, and ATP hydrolytic activities. J. Biol. Chem. 272, 23360–23365 (1997). Ulrich, H. D. & Jentsch, S. Two RING finger proteins mediate cooperation between ubiquitin-conjugating enzymes in DNA repair. EMBO J. 19, 3388–3397 (2000). Spence, J. et al. Cell cycle-regulated modification of the ribosome by a variant multiubiquitin chain. Cell 102, 67–76 (2000). This is provocative study that demonstrates a role for K63-linked multi-ubiquitin chains in regulating translation. Deng, L. et al. Activation of the IκB kinase complex by TRAF6 requires a dimeric ubiquitin-conjugating enzyme complex and a unique polyubiquitin chain. Cell 103, 351–361 (2000). Chau, V. et al. A multiubiquitin chain is confined to specific lysine in a targeted short-lived protein. Science 243, 1576–1583 (1989). Baldi, L., Brown, K., Franzoso, G. & Siebenlist, U. Critical role for lysines 21 and 22 in signal-induced, ubiquitinmediated proteolysis of IκBα. J. Biol. Chem. 271, 376–379 (1996). Hou, D., Cenciarelli, C., Jensen, J. P., Nguyen, H. B. & Weissman, A. M. Activation-dependent ubiquitination a T cell antigen receptor subunit on multiple intracellular lysines. J. Biol. Chem. 269, 14244–14247 (1994). Treier, M., Staszewski, L. M. & Bohmann, D. Ubiquitindependent c-Jun degradation in vivo is mediated by the delta domain. Cell 78, 787–798 (1994). Breitschopf, K., Bengal, E., Ziv, T., Admon, A. & Ciechanover, A. A novel site for ubiquitination: the Nterminal residue, and not internal lysines of MyoD, is essential for conjugation and degradation of the protein. EMBO J. 17, 5964–5973 (1998). Sheaff, R. J. et al. Proteasomal turnover of p21Cip1 does not require p21Cip1 ubiquitination. Mol. Cell 5, 403–410 (2000). This study provides strong evidence for ubiquitinindependent proteasomal degradation of a protein that is known to ubiquitylated. van Nocker, S. et al. The multiubiquitin-chain-binding protein Mcb1 is a component of the 26S proteasome in Saccharomyces cerevisiae and plays a nonessential, substrate-specific role in protein turnover. Mol. Cell. Biol. 16, 6020–6028 (1996). Brodsky, J. L. & McCracken, A. A. ER protein quality control and proteasome-mediated protein degradation. Semin. Cell Dev. Biol. 10, 507–513 (1999). Unger, T. et al. Mutations in serines 15 and 20 of human p53 impair its apoptotic activity. Oncogene 18, 3205–3212 (1999).

86. Shieh, S. Y., Taya, Y. & Prives, C. DNA damage-inducible phosphorylation of p53 at N-terminal sites including a novel site, Ser20, requires tetramerization. EMBO J. 18, 1815–1823 (1999). 87. Lohrum, M. A., Ashcroft, M., Kubbutat, M. H. & Vousden, K. H. Identification of a cryptic nucleolar-localization signal in MDM2. Nature Cell Biol. 2, 179–181 (2000). 88. Weber, J. D. et al. Cooperative signals governing ARFmdm2 interaction and nucleolar localization of the complex. Mol. Cell. Biol. 20, 2517–2528 (2000). 89. Midgley, C. A. et al. An N-terminal p14ARF peptide blocks Mdm2-dependent ubiquitination in vitro and can activate p53 in vivo. Oncogene 19, 2312–2323 (2000). 90. Honda, R. & Yasuda, H. Association of p19ARF with Mdm2 inhibits ubiquitin ligase activity of Mdm2 for tumor suppressor p53. EMBO J. 18, 22–27 (1999). 91. Jackson, M. W. & Berberich, S. J. MdmX protects p53 from Mdm2-mediated degradation. Mol. Cell. Biol. 20, 1001–1007 (2000). 92. Sharp, D. A., Kratowicz, S. A., Sank, M. J. & George, D. L. Stabilization of the MDM2 oncoprotein by interaction with the structurally related MDMX protein. J. Biol. Chem. 274, 38189–38196 (1999). 93. Tanimura, S. et al. MDM2 interacts with MDMX through their RING finger domains. FEBS Lett. 447, 5–9 (1999). 94. Buschmann, T., Fuchs, S. Y., Lee, C.-G., Pan, Z.-Q. & Ronai, Z. SUMO-1 modification of Mdm2 prevents its selfubiquitination and increases Mdm2 ability to ubiquitinate p53. Cell 101, 753–762 (2000). 95. Balint, E., Bates, S. & Vousden, K. H. Mdm2 binds p73α without targeting degradation. Oncogene 18, 3923–3929 (1999). 96. Dobbelstein, M., Wienzek, S., Konig, C. & Roth, J. Inactivation of the p53-homologue p73 by the mdm2oncoprotein. Oncogene 18, 2101–2106 (1999). 97. Zeng, X. et al. MDM2 suppresses p73 function without promoting p73 degradation. Mol. Cell. Biol. 19, 3257–3266 (1999). 98. Lam, Y. A. et al. Inhibition of the ubiquitin-proteasome system in Alzheimer’s disease. Proc. Natl Acad. Sci. USA. 97, 9902–9906 (2000). 99. Morimoto, M., Nishida, T., Honda, R. & Yasuda, H. Modification of cullin-1 by ubiquitin-like protein Nedd8 enhances the activity of SCFskp2 toward p27kip1. Biochem. Biophys. Res. Commun. 270, 1093–1096 (2000). 100. Osaka, F. et al. Covalent modifier NEDD8 is essential for SCF ubiquitin-ligase in fission yeast. EMBO J. 19, 3475–3484 (2000). 101. Podust, V. N. et al. A Nedd8 conjugation pathway is essential for proteolytic targeting of p27Kip1 by ubiquitination. Proc. Natl Acad. Sci. USA 97, 4579–4584 (2000). 102. Read, M. A. et al. Nedd8 modification of cul-1 activates SCF(beta(TrCP))-dependent ubiquitination of IkappaBalpha. Mol. Cell. Biol. 20, 2326–2333 (2000). 103. Wada, H., Yeh, E. T. & Kamitani, T. A dominant-negative UBC12 mutant sequesters NEDD8 and inhibits NEDD8 conjugation in vivo. J. Biol. Chem. 275, 17008–17015 (2000). 104. Wu, K., Chen, A. & Pan, Z. Q. Conjugation of Nedd8 to CUL1 enhances the ability of the ROC1-CUL1 complex to promote ubiquitin polymerization. J. Biol. Chem. 275, 32317–32324 (2000). 105. Liakopoulos, D., Doenges, G., Matuschewski, K. & Jentsch, S. A novel protein modification pathway related to the ubiquitin system. EMBO J. 17, 2208–2214 (1998). 106. Liakopoulos, D., Busgen, T., Brychzy, A., Jentsch, S. & Pause, A. Conjugation of the ubiquitin-like protein NEDD8 to cullin-2 is linked to von Hippel-Lindau tumor suppressor function. Proc. Natl Acad. Sci. USA 96, 5510–5515 (1999). 107. Gmachl, M., Gieffers, C., Podtelejnikov, A. V., Mann, M. & Peters, J. M. The RING-H2 finger protein APC11 and the E2 enzyme UBC4 are sufficient to ubiquitinate substrates of the anaphase-promoting complex. Proc. Natl Acad. Sci. USA 97, 8973–8978 (2000).

Acknowledgements I am grateful to the members of my laboratory for countless invaluable discussions. My apologies to colleagues whose important contributions to the field have been cited only indirectly because of space limitations.

www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS

ANTIGEN PROCESSING BY THE PROTEASOME Peter-M. Kloetzel The proteasome is an essential part of our immune surveillance mechanisms: by generating peptides from intracellular antigens it provides peptides that are then ‘presented’ to T cells. But proteasomes — the waste-disposal units of the cell — typically do not generate peptides for antigen presentation with high efficiency. How, then, does the proteasome adapt to serve the immune system well? U B I Q U I T I N A N D P R OT E A S O M E S SELF

Endogenous peptides derived from the organism’s own protein pool. CYTOTOXIC T CELLS

T cells that can kill other cells. They are important in host defence against most viral pathogens. MULTI-UBIQUITIN TAG

Ubiquitin is a small protein that can form multimeric chains. Multi-ubiquitin chains, which are covalently bound to a substrate, target this substrate to the 26S proteasome for degradation. TRIPLE-A FAMILY

ATPases associated with a variety of cellular activities. They contain an ATP-binding site with two conserved motifs known as Walker A and Walker B.

As part of their immune surveillance system, vertebrate cells display intracellular antigens — from intracellular pathogens (especially viruses) and from SELF — at the cell surface to distinguish between infected and uninfected cells. This task falls to proteins of the class I major histocompatibility complex (MHC), which bind antigenic peptides of defined length (usually 8–10 amino-acid residues long) and sequence. The peptide–MHC pair can then be recognized by specific Tcell receptors on CYTOTOXIC T CELLS. But what generates these peptides? And how? The proteasome — an ATPdependent, multisubunit protease — is the central proteolytic machinery in the cell, and one of its many jobs is to generate antigenic peptides1,2. The problem is that ‘typical’ proteasomes are not always well suited to this task and the machinery has to be tuned for higher efficiency. To process antigens more efficiently, the cell replaces some of its proteasomal subunits with more appropriate subunits, forming a so-called immunoproteasome. How does this immunoproteasome achieve such mechanical perfection? Proteasome components

Institut für Biochemie, Medical Faculty, Charité, Humboldt University, Monbijoustrasse 2, 10117 Berlin, Germany. e-mail: pm.kloetzel@charite.de

To understand the intricacies of immunoproteasomes, we first need to appreciate the proteasome’s structure. The proteolytic active sites of the proteasome are housed in a cylindrical enzymatic chamber — the 20S core — which has a 19S regulator complex at either end (FIG. 1). The proteasome recognizes most of its substrates by a MULTI-UBIQUITIN TAG (see the review by Allan Weissman on page 169 of this issue for information on

how this is added), although there are exceptions (see the review by Philip Coffino on page 188 of this issue). One of the 19S regulator’s functions is to recognize this multi-ubiquitin signal, so that it does not destroy intracellular proteins indiscriminately1,3,4. This complex of one 20S and two 19S units forms the 26S proteasome. The 19S regulator has two multisubunit components, the ‘base’ and the ‘lid’. The base, which is composed of six ATPases of the TRIPLE-A FAMILY plus two non-ATPase subunits3, binds to the 20S catalytic core. The ATPases have a CHAPERONE-like activity and are believed to help unfold substrates and channel them into the 20S core, thereby controlling access of substrates to the proteases within5–7. The residual eight different nonATPase subunits of the 19S regulator form the lid, the exact function of which remains a mystery. The 20S particle comprises four stacked rings, each containing seven evolutionarily related, but not identical, subunits. These fall into two categories, α and β, on the basis of sequence similarity. The outer rings contain the α-subunits (α1–α7), which form the ‘gates’ through which substrates enter and products are released. The two inner rings contain the β-subunits (β1–β7), three of which — β1, β2 and β5 — harbour the six active sites (two copies of each; see FIG. 1). The catalytically active residue is a single threonine located at the amino termini of the three β-subunits8–10, and characterizes the proteasome as a member of the family of AMINO-TERMINAL NUCLEOPHILE (NTN) HYDROLASES. The active-site threonines of the β-subunits are preceded by pro-sequences with different sequences and lengths. During assembly of the 20S proteasome, these pro-sequences are removed by a

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 1 7 9

REVIEWS

ATPases

Base

α7′

β1 β7′

α6′ β6′ α5′ α4′

β5′

β2 β3

α1 α2

α7 α6

α3

β4

α5

α4

20S core

α7′

19S regulator

β1 β7′

α6′ β6′ α5′ α4′

Lid

β5′

β2

α1 α2

β3

α6

α3

β4

α4

26S proteasome

CHAPERONE

Proteins that support the folding of other proteins. AMINO-TERMINAL NUCLEOPHILE (NTN)HYDROLASE

Enzyme family that shares an amino-terminal nucleophile as a single active-site residue, which can be threonine, serine or cysteine. ANCHOR RESIDUE

Residues within an epitope that bind, via their side chains, into the pockets of the MHC molecule lining the peptidebinding groove of the MHC class I molecule. HAPLOTYPE

A linked set of genes associated with the haploid genome. Mostly used in connection with genes of the MHC complex. TRANSPORTER ASSOCIATED WITH ANTIGEN PROCESSING

(TAP). ATP-binding cassette protein involved in the transport of peptides from the cytosol to the endoplasmic reticulum.

180

Figure 1 | Proteasome composition. The 20S proteasome consists of 28 (14 different) subunits, 21–31 kDa in molecular mass, which are arranged as four heptameric staggered rings. The two outer rings contain the α subunits (α1–α7, green), the amino termini of which protrude into the central opening of the cylinder, as shown by X-ray structure analysis of the yeast proteasome8. The two inner rings contain two copies of the βsubunits (β1–β7, blue and red), three of which (β1, β2 and β5; red) harbour the six active sites. The 19S subunit comprises two substructures. The base, which attaches to the two αrings, is composed of six ATPase (purple) and two nonATPase subunits (orange). The lid, which contains up to ten non-ATPase subunits (orange), is responsible for substrate binding and has poorly understood regulatory functions.

two-step mechanism: in the first step, neighbouring active sites cleave within the pro-sequences, whereas the second step is autocatalytic, generating the active-site threonine and a functional 20S complex11,12. As shown by the use of short fluorogenic peptide substrates, each of the three β-subunits has a different preference for cleaving after acidic, basic or hydrophobic residues, respectively. However, the cleavage preference of mammalian proteasomes is not so obvious when larger, more physiological substrates are analysed13,14. Antigenic-peptide quality

The MHC class I genes are highly polymorphic, with several hundred alleles. The protein product of each allele binds a unique set of peptides with an average length of 8–10 amino acids15. The final generation of peptides loaded onto MHC class I proteins is a multistep process (FIG. 2), each step contributing to the selectivity and efficiency of the immune response. Peptide binding to the MHC class I protein occurs

through the interaction of the peptide’s side chains with pockets in the peptide-binding groove of the MHC protein16. The specificity of this interaction is mediated by ANCHOR RESIDUES. One of these is always located at the carboxyl terminus of the peptide, fixing the peptide to the edge of the MHC-peptide-binding groove. The other one or two anchor residues are usually located near the amino terminus of the peptide17. To allow peptide binding, the anchor residue positions must be occupied by specific residues defined by the MHC class I HAPLOTYPE. The fact that the carboxy-terminal anchor positions are enriched for specific hydrophobic or basic residues, and that most peptides bound to MHC class I molecules have a defined length of 8–10 residues, suggests the existence of a proteolytic machinery with a certain predisposition for generating such peptides. The proteasome fulfils these requirements and is therefore regarded as the main machinery for the generation of antigenic peptides, but in specific cases other proteases (such as tripeptidyl peptidase II, furin and thimet oligopeptidase) might also contribute to the MHC class I peptide pool18. However, owing to their selective cleavage specificity, these proteases can generate only a limited subset of peptides that are appropriate for MHC binding, and therefore cannot be important for the production of antigenic peptides or substitute for proteasome function, even when expressed together. ‘Hide and seek’ with substrates

The availability of peptides that bind MHC class I molecules with high affinity is crucial for the establishment of cellular immunogenicity. The prevalent view is that cellular proteins that are targeted for degradation as part of normal cellular protein turnover are the main source of antigenic peptides. However, most of these peptides derive from proteins that are compartmentalized and/or metabolically stable. How, then, does the proteasome gain access to these proteins? A possible solution is offered by an old observation that up to 40% of newly synthesized proteins are degraded within a minute of synthesis. Indeed, a large proportion of polypeptides might never attain their native structure owing to errors in translation or defects in post-translational protein folding19. This observation sat on the shelf for 18 years until last year when these ‘non-functional proteins’, also known as defective ribosomal products (DRiPs), were found to be rapidly ubiquitylated and degraded by proteasomes20,21. DRiPs could therefore represent an important source of MHC class I peptides (FIG. 3). A processing mechanism linked to translation would guarantee that the cellular immune system has access to its substrates before they reach their final destination within or outside the cell, and two observations support the existence of such a coupling mechanism. The first is that the activity of the TRANSPORTER ASSOCIATED WITH ANTIGEN PROCESSING (TAP) — which transports antigenic peptides from the cytoplasm into the ER — depends on continuous protein synthesis, implying that MHC class I peptides originate predominantly from newly synthesized proteins22. The second is that proteasomes have been www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS

Multi-ubiquitin chain

Peptides

Antigen

e Unfolded MHC class I heavy chain

ERp57

TAP complex

β2M

Fully folded MHC class I molecule

β2M Calreticulin

BiP

Golgi Vesicular transport Antigen-presenting cell

f TCR complex

T cell T-cell proliferation and activation of cytotoxic activity

Figure 2 | Antigen processing and presentation. a | Most substrates that harbour MHC class I epitopes are conjugated with a multi-ubiquitin chain, which targets it to the 26S proteasome for degradation80. b | Proteasomal protein substrate processing results in the generation of peptides of 8–11 residues in length, which are c | transported into the endoplasmic reticulum (ER) by the transporter associated with antigen presentation (TAP) complex81,82. d | Within the ER, peptides bind to and stabilize the MHC class I heterodimers, which comprise a heavy chain and a β2 microglobulin molecule. e | Importantly, the assembly of MHC molecules is coordinated by molecular chaperones such as BiP, calreticulin and ERp57; only fully assembled, peptide-loaded MHC molecules are translocated, through the Golgi apparatus, to the cell surface83–85. f | By binding specifically to a given MHC allomorph — an MHC molecule encoded by a specific haplotype — loaded with a unique peptide, cytotoxic T cells with complementary T-cell receptors (TCRs) are stimulated to proliferate and destroy the infected target cell.

CYTOKINES

Originally used to describe a group of immunomodulatory growth factors, the term cytokine is now used to describe a diverse group of soluble proteins that modulate the activities of cells and tissues. COOPERATIVE

The incorporation of immunosubunits into the 20S proteasome is cooperative because the efficiency of their incorporation is strongly dependent on each other. EPITOPE

A short peptide derived from a protein antigen. It binds to an MHC molecule (or an antibody) and is recognized by a specific T-cell clone (or B-cell clone).

found in close association with polyribosomes23, making coordinated translational processing of proteins and generation of peptides an attractive hypothesis. Immunoproteasome formation

One of the characteristics of the MHC class I antigen presentation pathway (FIG. 2) is that several of its components are induced by the CYTOKINE interferon-γ (IFN-γ). These include the MHC class I heavy chain, the TAP proteins, several of the 20S proteasome subunits and the proteasome activator PA28 — a protein complex that can influence proteasome behaviour by substituting for one of the two 19S regulator complexes (see later)24. Three of the 20S proteasome’s β-subunits — all with proteolytic activity— are IFN-γ inducible: β1i (LMP2), β5i (LMP7) and β2i (MECL-1). Collectively, these are referred to as the immunosubunits, and their incorporation into the 20S core requires its de novo assembly25,26. As a result, new 20S complexes are

formed in which the constitutive proteases, β1 (δ), β2 (ζ) and β5 (MB1), are replaced by the three immunosubunits (FIG. 4). Assembly and maturation of the 20S core is a precisely ordered multistep event. It is assisted by chaperone proteins (FIG. 4) including hsc70 and the proteasome maturation protein (POMP), which form part of a high molecular weight ‘proteasome precursor complex’27. This process seems to be evolutionarily conserved because yeast have a chaperone, Ump1, that is thought to be a functional homologue of vertebrate POMP28,29. Incorporation of the immunosubunits is COOPERA30–32 : β2i is incorporated only if β1i is present, whereas TIVE β1i incorporation is independent of β2i (FIG. 4). The β5i subunit seems to influence and support the kinetics of immunoproteasome formation. This cooperativity implies that the immunosubunits are incorporated together, guaranteeing that a defined population of proteasome complexes with altered cleavage properties are formed. However, the situation is probably more complex than this: IFN-γ not only induces the formation of pure immunoproteasomes, but also that of 20S complexes in which immunosubunits and constitutive subunits exist together26. Cytokine induction therefore seems to increase the functional diversity of the proteasome pool in cells, allowing them to produce a wider range of peptides for presentation to the immune system. Manipulating peptide quality

Although the IFN-γ-inducible subunits seem to optimize MHC class I antigen presentation, the extent of their role, and the mechanisms that are involved, are not fully resolved. An indirect way of getting at this problem is by looking at how IFN-γ affects the quality of peptides generated from known viral EPITOPES. For example, when HeLa cells are infected with vaccinia virus expressing the hepatitis B virus (HBV) core antigen, they present one epitope only after stimulation by IFN-γ33. Analysis of 20S proteasome digests of synthetic peptides harbouring the epitope show that only immunoproteasomes liberate this epitope efficiently. How do the immunosubunits manipulate peptide quality? One emerging concept is that incorporation of immunosubunits might result in subtle structural changes of the whole 20S complex and thus, in turn, influence its processing properties. For example, in the case of the HBV epitope discussed above, although the concerted presence of all three immunosubunits is essential for epitope generation, an inactive β5i subunit (in which Thr1 is mutated to Ala — T1A) also supports epitope generation. This result indicates that incorporation of β5i influences the structural architecture of 20S proteasomes and consequently affects the cleavage-site preferences of the β1i and β2i active sites. An inactive β1i subunit carrying a T1A mutation behaves similarly34. These experiments require more corroboration, but the idea that immunosubunits alter proteasome structure is also supported by the altered chromatographic properties of immunoproteasomes, caused by changes in their surface charge35.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 1 8 1

REVIEWS

Proteasome

Ribosome 5'

5' 3'

Emerging protein Ubiquitin

Peptides Multi-ubiquitin chain

Figure 3 | The DRiP model. A relatively large percentage of nascent proteins never reach a functional state owing to errors in translation or folding. These defective ribosomal products (DRiPs) are rapidly ubiquitylated during translation and are degraded by proteasomes. DRiPs could therefore represent an important source of MHC class I peptides.

Studying immunoproteasomes by inducing them with IFN-γ is not ideal because this cytokine stimulates the expression of many genes that influence antigen presentation. A cleaner approach is to place the three immunosubunits behind a constitutive or a tetracyclineregulated promoter36–38. So far, the effect of immunoproteasomes on production of 14 MHC class I epitopes — mostly viral — has been analysed in this way (TABLE 1). The emerging picture is not straightforward to interpret: the liberation of some epitopes is greatly improved by — or absolutely dependent on — the presence of immunoproteasomes, but generation of others is not affected at all, and for two epitopes it might even be negatively influenced by immunosubunits (TABLE 1). How can these different effects be explained? In some cases, slight differences in cleavage activities, due to the presence of immunosubunits, might be responsible. Alternatively, as pointed out above, immunosubunits might induce subtle conformational changes to the 20S proteasome, with striking consequences for the generation of certain epitopes33. How might conformational changes affect proteasome activity? One idea is that immunosubunits could α-ring matrix

Early assembly intermediate

Transient assembly intermediate

β2i

β1i β2i

β5i β5i β2i

20S proteasome

Large assembly intermediate

C2 Autocatalytic activation of active sites

β1i

POMP

Figure 4 | Formation of the immunoproteasome. Proteasome formation is characterized by the initial cooperative formation of a heptameric α-ring matrix that provides the docking sites for a defined set of β-subunits. Subsequent structural rearrangement allows the residual subunits to bind. The pro-sequences of the β-subunits (red tails) have specific functions during this process and are removed by cis-and trans-autocatalysis. This last step in proteasome maturation probably occurs in the completely assembled cylinder concomitant with the final activation of the 20S proteasome. (C, chaperone-like proteins; POMP, proteasome maturation protein.)

182

alter substrate channelling within the proteasome, thereby changing the cleavage sequence of a substrate. Another possibility is that subtle structural changes might alter the properties of the substrate-binding cleft that extends from the active sites of the immunosubunits along the surface of the cavity towards the lumen of the barrel. Combined with the slightly altered active site properties, these effects might alter substrate susceptibility to proteolysis, or cleavage-site usage, thereby changing both the quality and the quantity of peptide products (FIG. 5). This model does not exclude the possibility that generation of certain epitopes will remain unchanged, as the binding affinity and mode of binding to the substrate-binding sites will depend on the amino-acid sequence of the processing intermediate and especially on the sequence in and around the MHC class I epitope. The proteasome activator PA28

Another IFN-γ-inducible component of the proteasome system that affects antigen presentation is the proteasome activator PA28, also known as the 11S regulator39,40. This 180–200 kDa complex is composed of two IFN-γ-induced subunits — PA28α and PA28β — that probably form an α3β4 complex41. PA28 attaches in an ATP-independent way to the outer α-rings of the 20S proteasome and was initially identified by its ability to enhance the turnover of a small, fluorogenic peptide substrate in vitro. The fact that the expression of both of its subunits is controlled by IFN-γ42 suggested that its in vivo function is connected with proteasomal antigen processing. Cell systems that express PA28 independently of IFN-γ showed that PA28 enhances the presentation of several viral antigens without increasing overall protein turnover or turnover of viral protein substrates37,43. Furthermore, this enhanced peptide presentation is independent of the presence of immunosubunits in the 20S proteasome43,44, excluding the possibility that PA28 might exert its function by increasing formation of the immunoproteasome45. But, like the immunosubunits, PA28 does not affect presentation of all MHC class I epitopes equally (TABLE 1). Earlier models suggested that PA28 somehow enhances dual cleavage events, but this could only be shown for selected substrates46,47. By contrast, recent kinetic studies imply that binding of PA28 to the 20S core increases substrate affinity without changing the maximal activity of the enzyme complex, and that PA28 activates the proteasome by enhancing either the uptake of substrates or the release of peptide products48. In a heteromeric 19S–20S–PA28 proteasome complex that is formed upon IFN-γ stimulation49, the 19S regulator is the rate-limiting component with regard to substrate binding and transport, implying that PA28 facilitates product release. But how does it do this? Controlling the gates

The crystal structure of the Saccharomyces cerevisiae 20S proteasome reveals that its gates are closed in the absence of regulatory particles8 (FIG. 6). The reason for this is that the amino-terminal ‘tails’ of the α-subunits www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS

Table 1 | Effects of immunoproteasomes and PA28 on MHC class I antigen processing Antigen

Epitope

Effect of immunosubunit

Effect of PA28

Reference

Ovalbumin

257–264

No effect

HY antigen

Influenza A nucleoprotein

366–374

Murine cytomegalovirus pp89

168–176

JAK1 tyrosine kinase

355–363

Influenza A/PR/8 nucleoprotein

146–154

43 39

Hepatitis B virus core antigen

141–151

Adenovirus E1A

234–243

No effect

Adenovirus E1B

192–200

LCMV nucleoprotein

118–126

No effect

Moloney murine leukaemia Virus gag pr75

75–83

Moloney murine leukaemia Virus env gp70

189–196

No effect

RU1 tumour antigen

34–42

–

Melan-A tumour antigen

26–35

–

*Sijts, A. et al. unpublished observations. (ND, not determined.)

project into the gate and block entry to the catalytic cavity. X-ray structure analysis of mutant proteasomes, containing a truncated α3 subunit lacking the first ten residues, showed that the α3 amino terminus is crucial for closing the gate because it interacts with the amino termini of all the other α-subunits. In fact, mutating Asp9 to Ala results in activated 20S proteasomes, indicating that a single aspartyl residue is crucial for maintaining the closed conformation50. In vitro, this latent or inactive state51 can be reversed with low concentrations of sodium dodecyl sulphate, lipids or heat treatment, indicating that these agents a

might open the gate52. But what opens the gate in vivo? Attachment of either the 19S regulator or PA28 to the two outer α-rings results in strong proteolytic activation of the latent 20S proteasome, making these good candidates for regulators of the gate. In the case of the 19S regulator, the gating mechanism is likely to be ATPdependent because heat shock locks 20S proteasomes in their latent state and impairs full activation of the 26S proteasome by ATP53. But how is the gate opened? The recently obtained X-ray structure of PA28 bound to the 20S proteasome54 shows that attachment of PA28 causes the amino-terminal tails of the α-subunits to flip b

Figure 5 | How immunoproteasomes affect peptide quality. Model view of how conformational changes might induce altered substrate binding and consequently lead to subtle changes in cleavage-site usage. a | The human proteasome subunit β2 (ζ) and b | the immunosubunit β2i (MECL-1) were modelled into the corresponding site of the yeast proteasome on the basis of the X-ray structure of the yeast 20S proteasome. The 35-residue hepatitis B virus (HBV) substrate (TABLE 1) was fitted into the active-site cavities formed by both types of subunit. The model illustrates that the incorporation of the β2i immunosubunits allows a good fit of the HBV substrate along the substrate-binding cavity, supporting cleavage between the second and third arginine. The γ1 oxygens of the active site are shown in red. The substrate-binding cavity is towards the left. (Image courtesy of C. Gille and C. Frömmel, Institute of Biochemistry, Charité, Berlin).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 1 8 3

REVIEWS upwards into the hollow core of PA28, thereby completely opening the gate (FIG. 6). Whether the 19S regulator induces a similar conformational change awaits solution of the 26S proteasome’s structure. Effect of gating on peptide quality

In vivo, the rate-limiting step in protein degradation by the 26S proteasome is not the width of the gate but substrate binding to the 19S regulator55, as well as its channelling into the 20S core. So, if attachment of PA28 opens the gate to its maximum width without changing turnover of cellular proteins, how does it influence pepa

Without PA28

tide quality, if at all? Although experimental data dealing with the qualitative aspects of gating are scarce, our knowledge of proteasome kinetics and the requirements for antigen processing allows us to make some predictions that can be verified experimentally. Because product exit from the proteasome is not active but occurs by diffusion, the width of the gate will determine the diffusion rate. Therefore, in a situation in which a gate is not fully open, the retention time of a processing intermediate in the catalytic chamber will be increased and the exit of larger peptide products will be hindered. In consequence, if a gate is not fully open the generation, accumulation and release of smaller peptide fragments through the gate will be favoured. By contrast, an open gate will decrease the retention time of a processing intermediate and will facilitate the diffusion of larger peptide fragments (FIG. 6). In other words, just like in a food processor, the length of time that the ingredients are processed affects the quality of the product. This view is supported by detailed studies of fragment generation by the yeast proteasome, which show that fragment size is not determined by the distance between the different active sites as had been proposed previously56,57. This gating-induced size control is important for antigen presentation because a relatively large percentage of antigenic peptides are not generated as 8 or 10-residue products that bind MHC class I molecules, but as precursor peptides of 10–12 residues that require amino-terminal trimming before they can bind to the MHC class I protein58. The generation of these precursor peptides is important because some processed peptides have residues close to their amino termini (Pro, for example) that impair their TAPmediated transport into the endoplasmic reticulum (ER)59. By creating products with an amino-terminal extension, the proteasome allows these epitopes to pass unhindered into the ER, where they can be trimmed to the optimum size for binding to MHC class I molecules. The enzymes responsible for this trimming are amino exopeptidases, most of which are localized in the ER. In some rare cases, cytosolic trimming might also take place60. Peptide processing

Bound to PA28

Figure 6 | The proteasome gating mechanism. a | X-ray structure analysis has revealed how the gate of the 20S proteasome is kept closed by the amino-terminal tails of the α-subunits. b | Deletion of the first ten amino acids from the α3 amino-terminal tail, which contacts all the others, causes the tails to become disordered, thereby opening the gate. c | Effect of the PA28 subunit on the gate. By binding a pocket in the interfaces between the α-subunits, d | PA28 puts pressure on a reverse turn at the end of each α-subunit’s amino-terminal tail. This causes the tails to point towards PA28 instead of obscuring the gate. (Reproduced with permission from REFS 50 AND 54 © (2000) Macmillan Magazines Ltd).

184

One of the intrinsic properties of the mammalian proteasome is its ability to generate the carboxy-terminal anchor residue of a MHC class I peptide with high fidelity. At first sight, this seems to be because the proteasome tends to cleave after hydrophobic or basic residues — residues that are favoured as carboxy-terminal anchor residues. However, the efficiency of epitope generation also depends on the sequence environment of the P1 residue (FIG. 7). Studies on a number of ovalbumin epitopes and a mouse cytomegalovirus epitope (pp89) showed that their generation strongly depends on their flanking sequences. For example, when an efficiently liberated epitope is transferred to the sequence environment of an epitope whose generation is less efficient, the transferred epitope adopts the characteristics of the new sequence environment61,62. Small amino www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS a N

Carboxy-terminal cleavage site P9 P8 P7 P6 P5 P4 P3 P2 P1

P1′

Influence of position on cleavage

b MHC class I peptide-binding groove

P9 P8 P7 P6 P5

proteasomal cleavage sites in a protein with surprisingly high fidelity. Both programs use parameters that are based on experimentally generated peptide fragments, and the main difference between them is the different data sets used to establish them. A third program combines the established kinetic properties of the 20S proteasome with proteasomal cleavage-site usage71. These programs are already proving to be useful, although larger data sets will improve them further. By combining one of these programs66 with a program that determines epitopes on the basis of haplotype-specific anchor residues72, it was possible to screen the entire proteome of CHLAMYDIA TRACHOMATIS to identify unknown HLA-B27-restricted epitopes, without having to resort to lengthy and difficult biochemical analysis (Kuon, W. et al., manuscript in preparation).

P3 P2 P1 C

Viral interference with proteasomes

Carboxy-terminal anchor residue

Figure 7 | How antigenic peptides achieve optimal binding to MHC class I proteins. a | Residues on both sides of the carboxy-terminal anchor site influence cleavage by the proteasome. The P1 residue represents the carboxy-terminal anchor residue, behind which proteasomal cleavage takes place. Residues at positions P8–P6 influence carboxy-terminal cleavage to different degrees. b | The MHC class I binding groove with antigenic peptide fitted into the pocket.

CHLAMYDIA TRACHOMATIS

A Gram-negative bacterial pathogen.

acids such as glycine or alanine at the P1′ position support cleavage after the carboxy-terminal anchor residue56, and naturally occurring mutations in the P1′ position impair the generation of several epitopes62–65. In addition to residues that flank the epitope, residues within the epitope at positions P4 to P7, which are some distance from the carboxyl terminus, also determine the efficiency of carboxy-terminal cleavage56,66. Interestingly, Pro residues within the epitope also seem to be involved in determining the carboxyterminal cleavage position67. Moreover, residues around the second anchor residue, which is usually near the amino terminus of the epitope, strongly influence the efficiency of the correct carboxy-terminal cleavage and, to some extent, also that at the amino terminus68. In all cases analysed so far, the generation of the correct amino terminus of an epitope is less well defined. This might be because when MHC-class I epitopes are generated as precursor peptides, the correct amino terminus is generated by post-proteasomal trimming by amino-exopeptidases60,69. On the basis of the cleavage sites used by the yeast 20S proteasome70 or those used by the constitutive mammalian 20S proteasome66, two groups have developed first-generation computer programs that predict

As the proteasome is the central source of antigenic peptides, it is not surprising that viruses can interfere with the proteolytic activity of the proteasome. However, as functional proteasomes are vital for cell survival — and consequently for viral propagation — viral proteins modulate proteasome activity rather than completely blocking it. Direct functional evidence for viral interference with the proteasome system is only just beginning to emerge, in striking contrast to the large amount of data implying that viruses interfere directly with peptide transport into the ER or at the level of MHC-protein routing73. Biochemical evidence for a direct interaction with proteasome subunits has been shown for the human Tcell leukaemia virus type 1 (HTLV-1) Tax protein, HBV pX protein, HIV Nef and Tat, and type 5 adenovirus (Ad5) E1A74–79. Although the proteasome has been identified as a target for all these viral proteins, information about the functional consequences of such interactions is still limited (TABLE 2). For most of these proteins, evidence that they also exert these effects in vivo is still missing. Our group has obtained evidence that Ad5 infection can affect both proteasome activity and antigen presentation at the cell surface, by upregulating the level of an endogenous protein called PI31 (31 kDa proteasome inhibitor). This delays immunoproteasome formation and affects presentation of the immunoproteasomedependent epitope E1B (Sijts, A. and Zaiss, D., manuscript in preparation). Future challenges

We are making good progress towards understanding how the proteasome orchestrates the processing of intracellular antigens, but many gaps in our knowledge remain to be filled. First, X-ray structures of vertebrate proteasomes — both the constitutive proteasome and the immunoproteasome — are needed. Their availability will greatly improve our chances of working out how the structural differences between constitutive proteasomes and immunoproteasomes correlate with differences in peptide quality. Second, further detailed analysis of how gating affects the quality of antigenic peptides

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 1 8 5

REVIEWS

Table 2 | Viral interference with the proteasome system Virus

Protein

Interacting subunit

Functional consequence

References

HTLV-1

TAX*

N3,C9

NF-κB activation

74 75, 76

HBV

pX*

NF-κB activation

HIV

Nef

HIV

TAT‡

α-ring

Competes with PA28 and upregulates ubiquitindependent degradation

Adenovirus type 5

E1A§

Adenovirus type 5

PI31 upregulation

*Tax and pX have been hypothesized to trigger NF-κB activation, which relies on the proteasome to destroy its inhibitory subunit IκB. ‡ HIV Tat binds the α-ring of the proteasome and impairs binding of PA28 but not of the 19S regulator, suggesting that Tat influences the quality of antigenic peptides generated after an HIV infection without impairing the proteolytic activity of the proteasome system. § The effect of adenovirus type 5 E1A is more complex. It interacts directly with the 19S S4 ATPase subunit and upregulates PI31. || Unpublished observations. (ND, not determined.)

generated in vivo is required. This will tell us how important a regulated mechanism of gating for antigen presentation is. Third, more sophisticated inducible cell systems that allow simultaneous analysis of the effects of PA28 and the immunosubunits, without interference from other IFN-γ-inducible proteins, will resolve the question of whether their effects on antigen processing are complementary. Fourth, reconstituting 26S proteasomes that can process protein substrates in vitro will allow us to analyse their peptide products biochemically. This requires the development of more efficient in vitro ubiquitylation systems that will permit the production of larger amounts of ubiquitin-conjugated substrates for biochemical analysis. Last, a more detailed analysis of viral interference with the proteasome sys-

186

Coux, O., Tanaka, K. & Goldberg, A. L. Structure and functions of the 20S and 26S proteasomes. Annu. Rev. Biochem. 65, 801–847 (1996). This review gives a good basic introduction to the proteasome system, its components and its different functions. Rock, K. L. et al. Inhibitors of the proteasome block the degradation of most cell proteins and the generation of peptides presented on MHC class I molecules. Cell 27, 761–771(1994). With the aid of a proteasome-specific inhibitor, this paper describes the first experiments demonstrating that inhibition of proteasome activity impairs antigen presentation. Glickman, M H. et al. A subcomplex of the proteasome regulatory particle required for ubiquitin-conjugate degradation and related to the COP9-signalosome and eIF3. Cell 94, 615–623 (1998). Deveraux, Q., Ustrell, V., Pickart, C. & Rechsteiner, M. A 26S subunit that binds ubiquitin conjugates. J. Biol. Chem. 267, 22369–22377 (1994). Braun, B. C. et al. The base of the proteasome regulatory particle exhibits ATP-dependent chaperone-like activity. Nature Cell Biol. 1, 221–226 (1999). The first demonstration that the base has a chaperone-like function. A model is discussed in which the 19S regulatory particle is responsible for the binding, unfolding and channelling of substrates. Strickland, E., Hakala, K., Thomas, P. J. & DeMartino, G. N. Recognition of misfolded proteins by PA 700, the regulatory subcomplex of the 26S proteasome. J. Biol. Chem. 275, 5565–5572 (2000). Glickman, M. H. et al. Functional analysis of the proteasome regulatory particle. Mol. Biol. Reprod. 26, 21–28 (1999). Groll, M. et al. Structure of the 20S proteasome from yeast at 2.4 Å resolution. Nature 386, 463–471(1997).

10.

11.

12.

13.

14.

15.

16.

17.

tem will ultimately help to block those viral proteins that deter proteasome function. Together, these studies will help us to design and optimize immunodominant epitopes for polyepitope DNA vaccines. Links DATABASE LINKS IFNγ | TAP | PA28 | β1i | β5i | β2i | β1 | β2 | β5 | hsc70 | Ump1 | PA28α | PA28β | crystal structure of the Saccharomyces cerevisiae 20S proteasome | X-ray structure analysis of mutant proteasomes | X-ray structure of PA28 bound to the 20S proteasome FURTHER INFORMATION IMGT/HLA sequence database ENCYCLOPEDIA OF LIFE SCIENCES Antigen processing | Histocompatibility antigens | Protease complexes | Ubiquitin pathway

The first X-ray structure analysis of a eukaryotic 20S proteasome presents structural evidence that the central gate of the 20S proteasome is closed. The six active sites in the catalytic cavity are defined by co-crystallization with proteasome inhibitors. Löwe, J. et al. Crystal structure of the 20S proteasome from the archaeon T. acidophilum at 3.4 Å resolution. Science 268, 533–539 (1995). Fenteany, G. et al. Inhibition of proteasome activity and subunit specific amino terminal modification of lactacystin. Science 268, 726–731(1995). Chen, P. & Hochstrasser, M. Autocatalytic subunit processing couples active sites formation in the 20S proteasome to completion of assembly. Cell 86, 961–972 (1996). Schmidtke, G. et al. Analysis of proteasome biogenesis: The maturation of β-subunits is an ordered two step mechanism involving autocatalysis. EMBO J. 15, 6887–6898 (1996). Kisselev, A. F., Akopian, T. N., Woo, K. M. & Goldberg, A. L. The sizes of peptides generated from protein by mammalian 26 and 20S proteasomes. Implications for understanding the degradative mechanism and antigen presentation. J. Biol. Chem. 274, 3363–3371 (1999). Schmidtke, G. et al. Inactivation of a defined active site in the mouse 20S proteasome complex enhances inactivation MHC class I antigen presentation of a murine cytomegalovirus protein. J. Exp. Med. 10, 1641–1664 (1998). Rammensee, H. G., Friede, T. & Stevanovic, S. MHC ligands and peptide motifs: first listing. Immunogenetics 41, 178–228 (1995). Madden, D. R. The three dimensional structure of peptide MHC complexes. Annu. Rev. Immunol. 13, 587–622 (1995). Falk, K. et al. Identification of naturally processed viral nonapeptides allows their quantification in infected cells and suggests an allele-specific T cell epitope forecast.

J. Exp. Med. 174, 425–434 (1991). 18. Schwarz, K. et al. The selective proteasome inhibitors lactacystin and expoxmycin can be used to either up-or down-regulate antigen presentation at non-toxic doses. J. Immunol. 164, 6147–6157 (2000). 19. Wheatley, D. N., Grisola, S. & Hernandez-Yago, J. Significance of rapid degradation of newly synthesised proteins in mammalian cells: A working hypothesis. J. Theor. Biol. 98, 283–300 (1982). 20. Schubert, U. et al. Rapid degradation of a large fraction of newly synthesised proteins by proteasomes. Nature 404, 770–774 (2000). Schubert et al. present the first experimental evidence for the DriP (defective ribosomal products) model, which proposes that defective translation products are the main source of MHC class I antigens. 21. Turner, C. C. & Varshavsky, A. Detecting and measuring cotranslational protein degradation. Science 289, 2117–2120 (2000). 22. Reits, E. A. J., Vos, J. C., Grommé, M. & Neffjes, J. The major substrates for TAP in vivo are derived from newly synthesized proteins. Nature 404, 744–748 (2000). In support of the DRiP model, the authors present evidence that TAP peptide transport activity relies on de novo protein synthesis. 23. Kloetzel, P. M., Falkenburg, P. E., Hössl, P. & Glätzer, K. H. The 19S ring-type particles of Drosophila: Cytological and biochemical analysis of their intracellular association and distribution. Exp. Cell Res. 170, 204–213 (1987). 24. Rock, K. & Goldberg, A. L. Degradation of cell proteins and the generation of MHC class I presented peptides. Annu. Rev. Immunol. 17, 737–779 (1999). 25. Nandi, D., Woodwards, E., Ginsburg, D. B. & Monaco, J. J. Intermediates in the formation of mouse 20S proteasomes: implications for the assembly of precursor β-subunits EMBO J. 17, 5363–5375 (1997). 26. Frentzel, S., Pesold-Hurt, B., Seelig, A. & Kloetzel, P. M.

www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41. 42.

43.

44.

45.

20S proteasomes are assembled via distinct precursor complexes: Processing of LMP2 and LMP7 proproteins takes place in 13–16S preproteasome complexes. J. Mol. Biol. 236, 975–981 (1994). Schmidt, M. et al. Sequence information within the proteasomal prosequence mediates efficient integration of β-subunits into the 20S proteasome complex. J. Mol. Biol. 288, 99–110 (1999). Schmidtke, G. et al. Analysis of mammalian 20S proteasome biogenesis: the maturation of β-subunits is an ordered two step mechanism involving autocatalysis. EMBO J. 15, 6887–6898. (1996). Witt, E. et al. Characterisation of the newly identified human Ump1 homologue POMP and analysis of LMP7 (β5i) incorporation into 20S proteasomes. J. Mol. Biol. 30, 1–9 (2000). Groettrup, M., Standera, S., Stohwasser, R. & Kloetzel, P. M. The subunits MECL-1 and LMP2 are mutually required for incorporation into the 20S proteasome. Proc. Natl Acad. Sci. USA 94, 8970–8975 (1996). Griffin, T. A. et al. Immunoproteasome assembly: cooperative incorporation of inferon-γ (IFN-γ) inducible subunits. J. Exp. Med. 187, 97–104 (1998). Schmidt, M. & Kloetzel, P. M. Biogenesis of eukaryotic proteasomes: the complex maturation pathway of a complex enzyme. FASEB J. 11, 1235–1243 (1997). Sijts, A. et al. Structural features of immunoproteasomes determine the efficient generation of a viral CTL epitope. J. Exp. Med. 191, 503–513 (2000). Experimental evidence is presented that the β5i (LMP7) subunit influences the structure of the immunoproteasomes and affects the activity of the two other immunosubunits. Schmidtke, G. et al. Inactivation of a defined active site in the mouse 20S proteasome complexes enhances MHC class I presentation of a murine cytomegalovirus protein. J. Exp. Med. 187, 1641–1646 (1998). Dahlmann, B. et al. Different proteasomes subtypes in a single tissue exhibit different enzymatic properties. J. Mol. Biol. 303, 643–653 (2000). Sijts, A. et al. MHC class I antigen processing of an Adenovirus CTL epitope is linked to the levels of immunoproteasomes in infected cells. J. Immunol. 164, 450–456 (2000). This paper describes the establishment of tetracycline-regulated immunosubunit expression and presents evidence that only minor amounts of immunoproteasomes are required for efficient antigen presentation. van Hall, T. et al. Differential influence on CTL epitope presentation by controlled expression of either proteasome immuno-subunits or PA28. J. Exp. Med. 192, 483–492 (2000). Schwarz, K. et al. Overexpression of the proteasome subunits LMP2, LMP7 and MECL-1 but not PA28α/β enhances the presentation of an immunodominant lymphocyte choriomeningitis virus T cell epitope. J. Immunol. 165, 768–778 (2000). Dubiel, W., Pratt, G., Ferrell, K. & Rechsteiner, M. Purification of a 11S regulator of the multicatalytic proteinase. J. Biol. Chem. 267, 22369–22377 (1992). Ma, C. P., Slaugther, C. A. & DeMartino, G. N. Identification, purification and characterisation of a protein activator (PA28) of the 20S proteasome (macropain). J. Biol. Chem. 267, 10515–10523 (1992). Knowlton, J. R. et al. Structure of the proteasome activator REGα (PA28α). Nature 390, 639–643 (1997). Realini, C. et al. Molecular cloning and expression of a γinterferon inducible activator of the multicatalytic proteinase. J. Biol. Chem. 269, 20727–20732 (1994). Groettrup, M. et al. A role for the proteasome regulator PA28α in antigen presentation. Nature 381, 166–168 (1996). Schwarz, K. et al. The proteasome regulator PA28α/β can enhance antigen presentation without affecting 20S proteasome subunit composition. Eur. J. Immunol. (in the press). Preckel, T. et al. Impaired immunoproteasome assembly and immune response in PA28–/– mice. Science 286, 2162–2165 (1999).

46. Dick, T. P. et al. Coordinated dual cleavages induced by the proteasome regulator PA28 lead to dominant MHC ligands. Cell 86, 253–256 (1996). 47. Shimbara, N. et al. Double cleavage production of the CTL epitope by proteasomes and PA28: role of the flanking region. Genes Cells 2, 785–800 (1997). 48. Stohwasser, R. et al. Kinetic evidences for facilitation of peptide channelling by the proteasome activator PA28. Eur. J. Biochem. 276, 6221–6239 (2000). 49. Tanahashi, N. et al. Hybrid proteasomes: Induction by interferon-γ and contribution to ATP-dependent proteolysis. J. Biol. Chem. 275, 14336–14345 (2000). 50. Groll, M. et al. A gated channel into the proteasome core particle. Nature Struct. Biol. 7, 1062–1067 (2000). X-ray structure analysis shows that gating is most likely a regulated mechanism and that a single residue within the amino terminus of the α3 subunit is responsible for opening and closing the gate. 51. McGuire, M. J., Mc Cullough, M. L., Croall, D. E. & DeMartino, G. N. The high molecular weight multicatalytic proteinase, macropain, exists in a latent form in human erythrocytes. Biochim. Biophys. Acta 995, 181–186 (1989). 52. Dahlmann, B., Rutschmann, M., Kuehn, L. & Reinauer, H. Activation of the multicatalytic proteinase from skeletal muscle by fatty acid and sodium dodecyl sulphate. Biochem. J. 228, 171–177 (1985). 53. Kuckelkorn, U. et al. The effect of heat shock on 20S/26S proteasomes. Biol. Chem. 381, 1017–1024 (2000). 54. Whitby, F. G. et al. Structural basis for the activation of 20S proteasomes by 11S regulators. Nature 408, 115–120 (2000). Presents the first co-crystallization of the proteasome activator PA28 with the 20S proteasome and explains how PA28 facilitates the opening of the gate. 55. Thrower, J. S., Hofmann, L., Rechsteiner, M. & Pickart, C. M. Recognition of the polyubiquitin proteolytic signal. EMBO J. 19, 94–102 (2000). 56. Nussbaum, A. K. et al. Cleavage motifs of the yeast proteasome β-subunits deduced from digests of enolase 1. Proc. Natl Acad. Sci. USA 95, 12504–12509 (1998). 57. Wenzel, T., Eckerskorn, C., Lottspeich, F. & Baumeister, W. Existence of a molecular ruler in proteasomes suggestes by analysis of degradation products. FEBS Lett. 349, 205–209 (1994). 58. Lauvau, G. et al. Human transporters associated with antigen processing (TAPs) select epitope precursor peptides for processing in the endoplasmic reticulum and presentation to T cells. J. Exp. Med. 190, 1227–1240 (1999). 59. Neisig, A. et al. Major differences in transporter associated with antigen presentation (TAP)-dependent translocation of MHC class I presentable peptides and the effect of flanking sequences. J. Immunol. 154, 1273–1279 (1995). 60. Mo, X. Y. et al. Distinct proteolytic processes generate the C and N-termini of the MHC class I-binding peptides. J. Immunol. 163, 5851–5859 (1999). 61. Niedermann, G. et al. Contribution of proteasome mediated proteolysis to the hierarchy of epitopes presented by major histocompatibiliy complex class I molecules. Immunity 2, 289–299 (1995). 62. Del Val, M. et al. Efficient processing of an antigenic sequence for presentation by MHC class I molecules depends on its neighbouring residues in proteins. Cell 66, 1145–1153 (1991). This paper presents, for the first time, experiments showing that the efficiency of antigen processing depends on the sequence environment of the epitope. 63. Theobald, M. et al. A mutational hotspot in p53 protects cells from lysis by CTL specific for a flanking epitope. J. Exp. Med. 11, 1017–1020 (1998). 64. Beekman, N. J. et al. Abrogation of CTL epitope processing by single amino acid substitution flanking the C-terminal proteasome cleavage site. J. Immunol. 164, 1898–1905 (2000). 65. Ossendorp, F. et al. A single residue exchange within a viral CTL epitope alters proteasome-mediated degradation resulting in lack of antigen presentation.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

Immunity 5, 115–124 (1996). 66. Holzhütter, H. G., Frömmel, C. & Kloetzel, P. M. A theoretical approach towards the identification of cleavage determining amino acid motifs of the 20S proteasome. J. Mol. Biol. 286, 1251–1265 (1999). The first mathematical model that permits the identification of proteasomal cleavage sites in a substrate protein. 67. Shimbara, N. et al. Contribution of proline residue for efficient production of MHC class I ligands by proteasomes. J. Biol. Chem. 273, 23062–23071 (1998). 68. Kraft, R. et al. Influence of single amino acid exchanges in epitope generation by 20S proteasomes. J. Protein Chem. 17, 547–548 (1998). 69. Beninga, J., Rock, K. L. & Goldberg, A. L. Interferon-γ can stimulate post-proteasomal trimming of the N-terminus of an antigenic peptide by inducing leucine aminopeptidase. J. Biol. Chem. 273, 18734–18742 (1998). 70. Kuttler, C. et al. An algorithm for the prediction of proteasome cleavage. J. Mol. Biol. 5, 298, 417–429 (2000). 71. Holzhütter, H.-G. & Kloetzel, P.-M. A kinetic model of vertebrate 20S proteasome accounting for the generation of major proteolytic fragments from oligomeric peptide substrates. Biophys. J. 79, 1196–1205 (2000). 72. Rammensee, H. G. et al. SYFPEITHI: database for MHC ligands and peptide motifs. Immunogenetics 50, 313–316 (1999). 73. Yewdell, J. W. & Bennik, J. R. Mechanisms of viral interference with MHC class I antigen processing and presentation. Annu. Rev. Cell Dev. Biol. 15, 579–606 (1999). 74. Rousset, R., Desbois, C., Bantignies, F. & Jalinot, P. Effects of the NF-kB1/105 processing of the interaction between HTLV-1 transactivator Tax and the proteasome. Nature 381, 328–331 (1996). 75. Fischer, M., Runkel, L. & Schaller, H. HBx protein of hepatitis B virus interacts with the C-terminal portion of a novel human proteasome α-subunit. Virus Genes 10, 99–102 (1995). 76. Huang, J., Kwong, J., Sun, E. C. & Liasng, T. J. Proteasome complex as a potential cellular target for hepatitis B virus X protein. J. Virol. 70, 5582–5591 (1996). 77. Rossi, F. et al. HsN3 proteasomal subunits as a target for human immunodeficiency virus type 1 Nef protein. Virology 237, 33–45 (1997). 78. Turnell, A. S. et al. Regulation of 26S proteasome by adenovirus. EMBO J. 19, 4759–4773 (2000). 79. Seeger, M., Ferrell, K., Frank, R. & Dubiel, W. HIV-tat inhibits the 20S proteasome and its 11S regulatormediated activation. J. Biol. Chem. 272, 8145–8148 (1997). 80. Young, P. et al. Characterisation of two polyubiquitin binding sites in the 26S protease subunit 5a. J. Biol. Chem. 273, 5461–5467 (1998). 81. Elliot, T. Transporter associated with antigen processing. Adv. Immunol. 65, 47–109 (1997). 82. Momburg, F. & Hämmerling, G. J. Generation of TAP mediated trasnport of peptides for major histocampatibility complex class I molecules. Adv. Immunol. 68, 191–256 (1998). 83. Pamer, E. & Cresswell, P. Mechanism of MHC class I restricted antigen processing. Annu. Rev. Immunol. 16, 323–358 (1998). 84. Hughes, E. A. & Cresswell, P. The thiooxidoreductase ERp57 is a component of the MHC class I peptide loading complex. Curr. Biol. 8, 709–712 (1998). 85. Morrice, N. A. & Powis, S. J. A role for the thiol-dependent reductase ERp57 in the assembly of MHC class I molecules. Curr. Biol. 8, 713–716 (1998). 86. Van Kaer, L. et al. Altered peptidase and viral-specific T cell response in LMP2 mutant mice. Immunity 1, 533–541 (1994). 87. Fehling, H. J. et al. MHC class I expression in mice lacking the proteasome subunit LMP-7. Science 265, 1234–1237 (1994). 88. Morel, S. et al. Processing of some antigens by standard proteasome but not by the immunoproteasome result in poor presentation by dendritic cells. Immunity 12, 101–117 (2000).

VOLUME 2 | MARCH 2001 | 1 8 7

REVIEWS

REGULATION OF CELLULAR POLYAMINES BY ANTIZYME Philip Coffino Proteins that are degraded by the proteasome are first modified by a set of enzymes that attach multiple copies of ubiquitin to substrate lysines, but a tiny minority, including the polyamine-synthesizing enzyme ornithine decarboxylase, is handled differently. This enzyme is targeted for destruction by another protein — antizyme. Why does ornithine decarboxylase have its own dedicated destruction mechanism, how does it work, and is it the only protein to be targeted to the proteasome in this way? U B I Q U I T I N A N D P R OT E A S O M E S HYPUSINATION

A post-translational modification thought to be unique to eukaryotic translation initiation factor 5A. An amino-butyl group is attached to lysine and then becomes hydroxylated to form a hypusyl group.

Department of Microbiology and Immunology and Department of Medicine, University of California, San Francisco, San Francisco, California 94143-0414, USA. e-mail: pcoffin@itsa.ucsf.edu

188

Antizyme, a moniker for ‘anti-enzyme for ornithine decarboxylase’, first emerged as an inhibitor of a key enzyme in polyamine biosynthesis1. In the manner of other ascendant biological molecules, antizyme rose through the ranks, going from a mere activity to a protein, to a gene and finally attaining the status of a gene family (BOX 1). Antizyme thus achieved dignity as governor of a metabolic outpost, the polyamines. It now seems probable that antizyme is also involved in maintaining coordination between polyamine pools and cellular events such as cell growth that might impose special demands on those pools.

Although polyamines might be dispensable for rodent social life, cells cannot manage without them (BOX 2). Mutations that reduce polyamine biosynthesis or drugs that inhibit their biosynthesis halt cell growth and eventually cause cell death. In vitro studies have shown that spermidine promotes both transcription and translation. Suboptimal concentrations of Mg2+ diminish translation, an effect that can be partially compensated for by adjusting polyamine concentration and vice versa5. Perhaps one way to think of polyamines is as Mg2+-like ions, but ions that can be made to rise or fall through multilevel controls and that have more distributed charges that might facilitate binding to nucleic What are polyamines? acids6. Spermidine is known to have one vital and Polyamines are abundant multivalent organic unique biochemical function — it provides an aminocations2, largely bound in cells to RNA and DNA. The butyl group for a post-translational modification, structures of the most abundant natural polyamines, termed HYPUSINATION, which is limited to a single lysyl residue of one protein, the eukaryotic translation intiaalong with the enzymes that make them, are depicted tion factor 5A (eIF-5A)7. Both the structure of eIF-5A in FIG. 1. The direct product of orthinine decarboxylase (ODC) is the diamine putrescine. A series of and its modification — the work of two enzymes (FIG. 2) enzymes then convert this to the polyamines spermi— are highly conserved from yeast to humans, and even in archaea. In yeast, failure to synthesize the fully modidine and spermine. The trivial names putrescine, fied protein is lethal. The initial description of eIF-5A as spermidine and spermine recall the historical discova general translation factor is almost surely wrong, but ery of these compounds in putrefying meat and semiits genuine cellular role remains enigmatic8. Preventing nal fluid3. Rats bury their dead companions when bacterial decay has perfumed the cadaver with hypusination in eukaryotes or archaea with drug polyamines; and olfactory detection of these cominhibitors leads to cell growth arrest. More recently, pounds is sufficient to trigger this behaviour, because polyamines have been shown to gate the inward rectifier rats also bury bits of wood that are sprinkled with current of an ion channel9 and to have effects on chropolyamines4. matin condensation and transcriptional regulation10. © 2001 Macmillan Magazines Ltd

| MARCH 2001 | VOLUME 2

www.nature.com/reviews/molcellbio

REVIEWS

Box 1 | Unearthing antizyme against the odds The name of Shin-ichi Hayashi of Jikei University, Tokyo, is identified closely with antizyme. His stubborn efforts to describe it and to understand its behaviour and synthesis have prevailed against widespread initial scepticism. It is rare that one scientist creates a fruitful line of investigation that would otherwise have been long delayed. Shin-ichi Hayashi is such a scientist. Shortly after the discovery of orthinine decarboxylase (ODC)38, two curious properties of the enzyme were uncovered: first, it had a half-life of less than an hour39, unusually short for an enzyme. Second, polyamine treatment of cells caused a profound and rapid fall in ODC activity40. Canellakis and his colleagues found that the loss of ODC activity coincided with the appearance of an activity that was inhibitory for ODC1. This activity was called anti-enzyme for ODC, or more briefly, antizyme. The low concentrations of both proteins presented a formidable challenge in addressing their biochemical nature, physiological behaviour and the mechanism of their interaction. Shin-ichi Hayashi began his attack on the problem by purifying ODC to homogeneity41, a heroic effort at the time. He then used the pure enzyme to show that ODC and antizyme form a tight, 1:1 complex that is enzymatically inactive, but that can be dissociated to regenerate ODC activity. An important observation, and a hint of what was to follow, came with the finding that the antizyme:ODC ratio in the cell was correlated strongly with the degradation rate of ODC42. The definitive investigations of antizyme function were achieved only after Hayashi’s successful cloning of the antizyme complementary DNA43 and gene44. Using these tools, his group promptly showed that artificially forced expression of antizyme in cells caused ODC activity and protein to fall, owing to increased turnover of the enzyme45. The protease that carries out this destructive process proved to be the proteasome, a conclusion established definitively by showing that antizyme promotes the destruction of ODC by the 26S proteasome in an in vitro system that contains purified components16. These studies were confirmed when translational frameshifting was described as the mechanism that regulates antizyme expression14. Arginine

Although cells need polyamines, failure to limit their amount can be disastrous11. All of this implies the need for some governing agent.

MITOGENS

Compounds that induce cell division. FRAMESHIFT

All nucleic acids can be transcribed in three different frames, although generally only one frame is used. Frameshifting can be induced by removing or adding a nucleotide, or by frame slippage, to yield an RNA that encodes a different protein.

Inducing frameshifting

bADC (plants/bacteria)

Ornithine

Agmatine

ODC (animals/ eukaryotes)

How antizyme regulates polyamine levels

Polyamine metabolism is elaborately orchestrated3, and antizyme is central in this process (reviewed in REFS 12,13). Cells are not merely equipped to preclude polyamine starvation, but also to limit, maintain or crank up pools, depending on need. Cells might be expected to need more polyamines when they begin to grow, to enhance macromolecular synthesis. Furthermore, polyamines act as counter-ions that neutralize the negative charges of phosphates in RNA and DNA. Making and having more polynucleotides implies a need for more polyamines. Indeed, MITOGENS produce rapid transient increases in the activity of polyamine biosynthetic enzymes3. Three influences act on polyamine pool size: production, transport and catabolism. Antizyme has a negative influence on two of these — production and transport into the cell. If antizyme is to act as an effective regulator, its action must be responsive to the level of polyamines that are present in the cell or, more accurately, their free levels, as most are bound. How do polyamine levels modulate antizyme?

Arginase

Agmatinase

S-Adenosylmethionine SamDC

Putrescine NH3+

+H N 3

DecarboxyS-Adenosylmethionine

SpdS

PAO

N-Acetylspermidine SSAT

+H N 3

Spermidine + NH3+ N H H PAO

SpmS

N-Acetylspermine

+H N 3

Spermine + N H H

H H N +

NH3+

SSAT

Figure 1 | Biosynthesis and catabolism of polyamines. Ornithine decarboxylase (ODC) initiates the production of polyamines by producing the diamine putrescine. In some plants and bacteria arginine decarboxylase (bADC) initiates an alternative two-step pathway to putrescine. Next, an aminopropyl group derived from methionine by the decarboxylation of S-adenosylmethionine adds to putrescine to form spermidine and a second aminopropyl group adds to spermidine to form spermine by discrete synthases, spermidine synthase (SpdS) and spermine synthase (SpmS). The biosynthetic process can be reversed by a series of catabolic reactions that are initiated by spermidine/spermine acetylase (SSAT), a step that is rate-limiting for conversion of spermine to spermidine and of spermidine to putrescine. Acetylation is followed by the action of polyamine oxidase (PAO). The three key enzymes that control the rate of flux through this pathway, ODC, S-adenosylmethionine decarboxylase (Sam-DC) and SSAT, are each subject to multilevel control, with degradation being especially prominent. Entry to and exit from the cell also affect polyamine pools.

Antizyme production depends on polyamine levels through an unusual mechanism, one that uses translational FRAMESHIFTING14 (FIG. 3). Two overlapping open reading frames (ORFs) encode antizyme. The first (ORF1) is short and has AUG codons that can initiate translation; the second (ORF2) encodes most of the protein, but lacks a usefully placed initiation codon. ORF2 can be translated only by starting in ORF1, frameshifting forward one base just before the stop codon of ORF1 is read, and continuing translation in the +1 frame to the © 2001 Macmillan Magazines Ltd

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 1 8 9

REVIEWS

Box 2 | Polyamine functions • Act as counter-ions for negative charges on RNA and DNA. • Enhance biosynthesis of DNA, RNA and proteins. • Mediate gating of K+ rectifier channel. • Donate modifying groups in hypusination. • Act as odorants. • Function during transcriptional regulation. • Function as constituents of bisglutothionylspermidine (trypanothione), which might have some of the functional roles of glutathione in trypanosomes.

end of ORF2. Polyamines greatly increase the efficiency of the frameshifting process, tying the level of polyamines to the production of their negative regulator. Conserved sequence motifs and secondary structure in the antizyme messenger RNA near the frameshifting site provide necessary signals. However, it is unclear whether polyamines interact directly with these to promote frameshifting, or instead do so through additional unknown mediators, perhaps in the ribosome. In general, feedback regulation requires both a means for sensing a state and another for altering it. In the case of antizyme, sensing uses the unusual mechanism of translational frameshifting. The effector arm of the loop is also not conventional, and is the next subject for discussion.

patch on the surface of the αβ-barrel that forms one domain of the ODC monomer structure15. When antizyme and ODC come together, the carboxyl terminus of ODC becomes more accessible to antibody, indicating that a conformational change might be propagated to that part of the molecule18. Conversely, ODC can no longer be degraded by the proteasome if as few as five amino acids are truncated from the carboxyl terminus of ODC19. These observations imply that the last amino acids of ODC are critical for proteasomal degradation. In vivo, ODC is a substrate for the proteasome even if antizyme is absent, with a half-life of an hour or two; antizyme reduces this half-life to minutes. Within antizyme, the carboxy-terminal half of the 227-aminoacid protein is necessary and sufficient for high-affinity binding to ODC. Binding of the carboxy-terminal half of antizyme is enough to expose the ODC carboxyl terminus. It would be reasonable to conclude from the information provided so far that antizyme simply increases accessibility of the ODC carboxyl terminus, and that the proteasomes recognize this more or less efficiently, depending on its conformation. This model is not tenable, however, because the carboxy-terminal half of antizyme, although able to expose the ODC carboxyl terminus, does not accelerate ODC degradation20. eIF-5A precursor N Lysyl residue

ODC degradation by the proteasome

Because each active site is composed of elements from both polypeptide subunits, ODC is enzymatically active only as a homodimer. Monomers of ODC interact with each other only weakly, but antizyme has high affinity for the ODC monomer. The site of antizyme binding is placed so that it obstructs the ODC homodimer interface15. Consequently, two antizyme molecules convert one homodimer to a pair of enzymatically inactive antizyme–ODC heterodimers. But stoichiometric inhibition is only the beginning of the work that antizyme does on ODC. It also acts catalytically to direct the proteasome to degrade the enzyme16. The proteasome (see the review by Peter Kloetzel on page 179 of this issue) is the main neutral protease of the cell, eliminating proteins that have outlived their usefulness, or that might become a danger to the cellular community because they are mutant or misfolded. The proteasome conceals its proteolytic apparatus in the interior of a cylindrical chamber of nanometre dimensions. ODC is first admitted to this chamber17 and is then destroyed. Antizyme is released to participate in subsequent cycles of ODC destruction. All the elements of a feedback loop are in place: polyamines rise; polyamines induce production of more antizyme; antizyme inhibits and demolishes ODC; polyamine synthesis declines (FIG. 4). Domains important for degradation

Spermidine

NH NH3+

+ +H3N

+ N H H

NH3+

C eIF-5A intermediate N

Deoxyhypusine synthase

NH NH3+

N H

C Deoxyhypusine hydroxylase

eIF-5A N

NH N H

NH3+

Figure 2 | The two-step enzymatic reaction that results in hypusination of a single lysine in eIF-5A. In the first step, an amino-butyl group is transferred from spermidine to the free amine of a lysyl residue of the eukaryotic translation initiation factor 5A (eIF-5A) precursor. A different enzyme then adds a hydroxyl group. Both the protein and its modification are conserved and essential, but the function of eIF-5A is not known.

| MARCH 2001 | VOLUME 2

www.nature.com/reviews/molcellbio

REVIEWS a

– Polyamines

Met Met STOP AUG X Y Z X Y Z X Y Z AUG X Y Z X Y Z X Y Z UGA ORF1

3′ Protein products

+ Polyamines 2

Shift site AUG X Y Z X Y Z X Y Z AUG X Y Z X Y Z X Y Z U GAX YZX YZX YZX YZX YZX 5′ ORF1

1 UGA ORF1 STOP

ORF2 Protein products

Figure 3 | Motifs in vertebrate antizymes that underlie translational frameshifting. Two overlapping reading frames are present in the antizyme messenger RNA. a | ORF1 is short and contains two methionine AUG initiation codons (both are used). b | ORF2, which lacks an appropriately placed AUG, encodes the bulk of the protein. Translation begins in ORF1, but a +1 frameshift can occur just before the ORF1 UGA stop codon. Polyamines enhance the efficiency of frameshifting. If the reading frame shifts, translational termination in ORF1 is evaded and translation continues, decoding ORF2. c | Three elements contribute to polyamine-induced frameshifting in vertebrates. These are, depicted in the context of human antizyme 1 mRNA: (1) the ORF1 UGA stop codon; (2) a stem-loop structure that follows the ORF1 UGA, together with a region 3′ to the stem-loop that can base pair with a portion of the loop to form an RNA pseudoknot; (3) a region within the 3′ portion of ORF1. In invertebrates, there is no similar mRNA secondary structure 3′ to the ORF1 UGA, but the region remains important for stimulating frameshifting. REF. 28 provides extensive analysis of functional motifs within antizyme mRNAs.

The amino-terminal half of antizyme is contributing something critical to the process, perhaps proteasome recognition or activation. Delivering the amino-terminal half of antizyme to ODC in an unnatural fashion, as a fusion to ODC, also enhances the efficiency of proteasomal degradation, reinforcing the idea that this portion of antizyme augments a signal that is carried by the ODC carboxyl terminus.

the G1 phase of the cell cycle by inducing antizyme. This depletes cyclin-dependent kinases whose action is required for progression from G1 into the DNA synthetic phase of the cell cycle. As polyamines are not general cell growth inhibitors, prostate cells might prove to have (or lack) some special component that determines the unusual response of these cells. Checkpoints versus homeostasis

The diverse, conserved and elaborate mechanisms for An expanded repertoire for antizyme has been found in managing polyamine pools can obviously hold the prostate. The prostate is the only vertebrate organ in polyamine pools steady, as shown for the feedback loop which polyamines are made for export rather than for that involves antizyme and ODC. But when the demand internal cell use; most of the spermidine and spermine for polyamines goes up or down, something different is produced there end up in seminal fluid (where polyrequired. It is not clear how mechanisms that resist amines were first discovered in the seventeenth century change can be altered or subverted to instead promote as microscopic amorphous crystalloids). Zetter and colit. One possibility is that a transient change arises and is leagues, observing that prostatic carcinomas commonly then dampened, restoring the earlier state. The trangrow in situ for years before disseminating, hypothesized scriptional burst from the ODC gene that is observed that prostate cells are bathed in a growth inhibitory subwhen cells respond to a mitogen could represent an stance. Such an activity was found in prostate extracts example of this. Alternatively, the set point of the homeand, upon purification, proved to be spermine21. Experiostatic mechanism might become altered. As antizyme mental application of spermine to cultured prostate carframeshifting seems to act as the primary sensor of celcinoma cells showed that spermine induced G1 arrest in lular polyamines, this seems a good place to seek potenpoorly metastatic cells but not in highly malignant cells22. tial mechanisms for modulating sensitivity. Although uptake was similar in the two cell types, Just as polyamine pools must be coordinated with antizyme was induced in the spermine-sensitive but not pronounced changes in cell state, a reciprocal constraint also seems to be needed — cells ought not commit in the spermine-resistant cells. Candidate proteins were themselves to irrevocable change without first checking screened for antizyme interaction — the G1 CYCLIN D1 and its associated cyclin-dependent kinase, cdk4, were whether polyamine pools have been properly adjusted. Antizyme, the polyamine surrogate, might be expected each found to interact with antizyme23. In vitro, purified to participate in such a verification process. The proteasomes degraded cyclin D1 or cdk4 in reactions degradative effect of antizyme on the cell-cycle regulator dependent on antizyme and independent of ubiquitin. cyclin D1 and its associated kinase might represent an Furthermore, treatments that raised polyamine levels in instance of this in prostate cells. Here, polyamine pools cultured prostate cells reduced the steady-state level of influence and perhaps exert a veto against the committhe cyclin and the kinase. These results are consistent ment of a cell to DNA synthesis. with a model whereby polyamines hold prostate cells in © 2001 Macmillan Magazines Ltd A cell-cycle target for antizyme

CYCLINS

A family of proteins whose levels fluctuate throughout the cell cycle. By activating cyclindependent kinases, they help to regulate several stages of cell division.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 1 9 1

REVIEWS

Consequences of disrupting function

Disrupting polyamine homeostasis is difficult because biosynthesis, catabolism and import/export all affect the pool. Perturbing any one of these processes leaves important regulatory capacities intact. Despite this, pronounced phenotypes have been observed in transgenic mice with altered polyamine biosynthetic or catabolic enzymes. For example, spermidine/spermine acetyltransferase is the rate-limiting enzyme for polyamine catabolism, a sequence of conversions that lead from spermine to spermidine to putrescine (FIG. 1). Overexpressing this enzyme in mice caused an increase in putrescine, at the expense of spermidine and spermine, a distortion more extreme in some tissues than in others. These mice had a complex phenotype that included hair loss, dermal cysts and female ovarian hypofunction24. Male fertility was unaffected, but there was little distortion of the polyamine pool in the testes of these animals. Other transgenic mice that were targeted to overexpress ODC in KERATINOCYTES developed spontaneous and induced skin cancer at high rates25. What about transgenic mice with altered antizyme expression? Excess antizyme expression in keratinocytes reduced the frequency of skin-tumour induction, the opposite of the effect seen from an excess of ODC in these cells26. The effect of expressing antizyme from a heart-specific promoter has also been looked at27. In the hearts of normal mice, the β-adrenergic agonist isoproterenol induces hypertrophy, increases ODC and makes polyamine pools rise. In the transgenics with increased cardiac antizyme, the rises in ODC and polyamines were blocked, but HYPERTROPHY still took place. The effect of a more radical intervention, ablating antizyme expression in all tissues, has also been examined. Antizyme-knockout mice were viable, morphologically normal and fertile (S. Matsufuji and T. Noda, personal communication). However, they had high perinatal fatality, with about one-third dying in the days before and after full term. This is probably due to the unconstrained activity of ODC, because perinatal treatment with a specific inhibitor of that enzyme prevents death. Further analysis of these antizyme-knockout mice will be informative. Nevertheless, the recent discovery that vertebrates express more than one antizyme might complicate interpretation of the data.

KERATINOCYTES

Differentiated epithelial cells of the skin. HYPERTROPHY

An increase in the size of a tissue or organ that results from an increase in the size of cells present.

192

Polyamines

Putrescine

Ornithine mRNA

C AZ

ODC dimer (active)

ODC–AZ dimer (inactive)

Proteasome

ODC peptides

Figure 4 | A regulatory feedback mechanism stabilizing polyamine pools. Polyamines increase the production of antizyme (AZ). The carboxy-terminal half of antizyme interacts with ODC, generating AZ–ODC heterodimers at the expense of enzymatically active ODC homodimers20. A carboxyterminal domain of ODC that is occluded within the homodimer is exposed within the heterodimer, and is essential for subsequent degradation18. A domain within the amino-terminal portion of antizyme provides a function needed for efficient degradation of ODC by the proteasome20. The proteasome processes AZ–ODC, sequestering ODC17 and then degrading it to peptides but releasing antizyme, which participates in additional rounds of binding and degradation. Antizyme-mediated inhibition and destruction of ODC reduces synthesis of polyamines, the downstream products of the enzyme. Additionally, antizyme inhibits polyamine transport into the cell. Antizyme production is thus reduced, completing the regulatory circuit.

to a late stage in sperm production30,31. What might be the special biological roles of AZ2 and AZ3 that have assured their conservation as distinct isoforms in the vertebrate lineage? AZ2 has a known functional difference from AZ1, which could make it more suitable to act as a reversible inhibitor of ODC rather than as a Functions of antizyme isoforms destructive agent. In this way, AZ2 might store the There are at least three independently conserved enzyme for future use. The most distinctive characterantizyme isoforms among vertebrates, two of which istic of AZ3 is its expression at a very restricted time have been recently described28. So far, discussion has and place, suggesting that its appearance abruptly been restricted to the first-discovered antizyme, which alters polyamine pools late in sperm morphogenesis. will subsequently be termed AZ1. The second, AZ2, This accords with the observation that transgenic mice has, like AZ1, a wide tissue distribution in vertebrates, expressing ODC at a very high level in the testes have but its mRNA is less abundantly expressed. Although defects in spermatogenesis32. Here, excess ODC might structurally similar to AZ1, AZ2 has little or no ability be swamping the capacity of AZ3 to impose limits on to drive ODC proteasomal degradation in vitro, but is putrescine production. approximately equipotent as an ODC inhibitor or Antizyme is widely dispersed in eukaryotes, and is inhibitor of cellular polyamine uptake29. Further experfound in vertebrates, insects, nematodes and fungi28. iments are required to illuminate the comparative ODC binding activity (where tested) and a need for functional properties of these two antizymes in more translational frameshifting are both conserved. The detail. AZ3 is testis specific; its expression is restricted extraordinary preservation of both antizyme structure © 2001 Macmillan Magazines Ltd

| MARCH 2001 | VOLUME 2

www.nature.com/reviews/molcellbio

REVIEWS and the means for controlling its generation raises a question: why devote such a singularly complex mechanism to the regulation of polyamines, which represent, after all, but a small corner of biology? It might be that sensing polyamines is a difficult feat of bioengineering and that a solution, once found, was preserved. Having acquired the sensing trick, further functions for antizyme might then have accumulated in the realms of polyamine homeostasis and the scheduling of cell commitments. Future prospects

The general features that are required for reprogramming the translational reading frame of antizyme have become clear, but the devil is in the details. Two problems should be mentioned. First, vertebrates and lower organisms use trans-acting motifs in antizyme mRNA that are different, but produce similar responses, even when artificially expressed in highly diverged biological contexts28. What mechanisms unite these diverse structural approaches? Beyond antizyme and polyamines, how widely used is frameshifting as a control mechanism? Sequence-based genomic analysis is strongly ORF-orientated. Finding further examples of frameshifting might require looking more closely at ORF interstices, predicting and testing conditions that favour frameshifting. Second, we do not know the identity of the direct sensor of polyamine levels. This is probably RNA; but is it antizyme mRNA, ribosomal RNA or another? How is this interaction transmitted to alter translational coding? Is the set point of polyamine homeostasis fixed, or can it be reset? Proteins that are degraded by the proteasome are first modified by a set of enzymes that attach multiple copies of the 76 amino-acid ubiquitin to substrate lysines (see the Review by Allan Weissman on page 169 of this issue). A small minority, including ODC, is handled differently33. Why then should we take notice of a member of this marginal class? First, because exceptions test rules and might tell us something about the nature and limits of those rules; second, because antizyme delivers a homogeneous and structurally simple signal to its target ODC, unlike ubiquitylation, which needs a congeries of

2. 3. 4.

Heller, J. S., Fong, W. F. & Canellakis, E. S. Induction of a protein inhibitor to ornithine decarboxylase by the end products of its reaction. Proc. Natl Acad. Sci. USA 73, 1858–1862 (1976). Tabor, C. W. & Tabor, H. Polyamines. Annu. Rev. Biochem. 53, 749–790 (1984). Cohen, S. A Guide to the Polyamines (Oxford Univ. Press, New York, 1998). Pinel, J. P. J., Gorzalka, B. B. & Ladak, F. Cadaverine and putrescine initiate the burial of dead conspecifics by rats. Physiol. Behav. 27, 819–824 (1981). Jackson, R. J., Campbell, E. A., Herbert, P. & Hunt, T. The preparation and properties of gel-filtered rabbit-reticulocyte lysate protein-synthesis systems. Eur. J. Biochem. 131, 289–301 (1983). Watanabe, S., Kusama-Eguchi, K., Kobayashi, H. & Igarashi, K. Estimation of polyamine binding to macromolecules and ATP in bovine lymphocytes and rat liver. J. Biol. Chem. 266, 20803–20809 (1991). Park, M. H., Wolff, E. C. & Folk, J. E. Is hypusine essential for eukaryotic cell proliferation? Trends Biochem. Sci. 18, 475–479 (1993).

10.

11.

12.

13.

enzymes to mark the target with a high-copy and variegated series of ubiquitins; and last, because antizyme has subdomains with identified distinct functions. These last two characteristics could help us answer general questions about how substrates are delivered to the proteasome and what happens to them next. Antizyme modulates polyamine transport into cells34,35. Despite recent advances in understanding the constituents of the polyamine-specific transport machinery36, no information about this regulatory interaction has yet emerged. In this same category of long-standing mysteries is the antizyme inhibitor. This protein is structurally similar to ODC, but lacks residues that are critical for catalytic activity37. Antizyme inhibitor has a higher affinity for antizyme than does ODC, and so must act stoichiometrically to dampen antizyme activity. Its biological function is a mystery. Until recently, the wizardry of antizyme has encompassed a few cute tricks done on a small, polycationic stage. Broader vistas have now opened — in two directions. The first of these, exemplified by the discovery that AZ3 comes on suddenly during late sperm development, supports the conclusion that polyamine metabolism might often tolerate lax discipline, but must shape up at critical junctions. Identifying a series of such crucial events, especially in experimentally accessible systems, offers some hope of better determining additional biochemical functions that are fulfilled by the polyamines. The second new prospect is that antizyme might influence cell events, such as commitment to DNA synthesis, an effect that is not specific to polyamine metabolism. So far, the work on growth regulatory effects of antizyme in the prostate presents an isolated example. Genetic and biochemical approaches that aim to extend the ambit of antizyme to other potential targets can be readily imagined. Links DATABASE LINKS Antizyme | orthinine decarboxylase |

Lipowsky, G. et al. Exportin 4: a mediator of a novel nuclear export pathway in higher eukaryotes. EMBO J. 19, 4362–4371 (2000). Lopatin, A. N., Makhina, E. N. & Nichols, C. G. Potassium channel block by cytoplasmic polyamines as the mechanism of intrinsic rectification. Nature 372, 366–369 (1994). Pollard, K. J., Samuels, M. L., Crowley, K. A., Hansen, J. C. & Peterson, C. L. Functional interaction between GCN5 and polyamines: a new role for core histone acetylation. EMBO J. 18, 5622–5633 (1999). Poulin, R. L., Coward, J. K., Lakanen, J. R. & Pegg, A. E. Enhancement of the spermidine uptake system and lethal effects of spermidine overaccumulation in ornithine decarboxylase-overproducing L1210 cells under hyposmotic stress. J. Biol. Chem. 268, 4690–4698 (1993). Coffino, P. in Proteasomes: The World of Regulatory Proteolysis (eds Hilt, W. & Wolf, D.) 256–265 (Landes Bioscience, Georgetown, Texas, 2000). Hayashi, S., Murakami, Y. & Matsufuji, S. Ornithine decarboxylase antizyme: a novel type of regulatory protein. Trends Biochem. Sci. 21, 27–30 (1996).

14. Matsufuji, S. et al. Antizyme autoregulatory frameshifting in decoding mammalian ornithine decarboxylase antizyme. Cell 80, 51–60 (1995). Showed that translation of antizyme requires polyamine-stimulated frameshifting and defined the messenger RNA region that mediates that process. 15. Kern, A. D., Oliveira, M. A., Coffino, P. & Hackert, M. L. Structure of mammalian ornithine decarboxylase at 1.6 A resolution: stereochemical implications of PLP-dependent amino acid decarboxylases. Struct. Fold. Des. 7, 567–581 (1999). 16. Murakami, Y. et al. Ornithine decarboxylase is degraded by the 26S proteasome without ubiquitination. Nature 360, 597–599 (1992). Showed that purified 26S proteasomes degrade ODC, and that this process depends on antizyme but not ubiquitylation. 17. Murakami, Y., Matsufuji, S., Hayashi, S. I., Tanahashi, N. & Tanaka, K. ATP-dependent inactivation and sequestration of ornithine decarboxylase by the 26S proteasome are prerequisites for degradation. Mol. Cell. Biol. 19, 7216–7227 (1999).

VOLUME 2 | MARCH 2001 | 1 9 3

REVIEWS 18. Li, X. & Coffino, P. Degradation of ornithine decarboxylase: exposure of the C-terminal target by a polyamine-inducible inhibitory protein. Mol. Cell. Biol. 13, 2377–2383 (1993). 19. Ghoda, L., van Daalen Wetters, T., Macrae, M., Ascherman, D. & Coffino, P. Prevention of rapid intracellular degradation of ODC by a carboxyl-terminal truncation. Science 243, 1493–1495 (1989). Showed that a region of vertebrate ODC that is redundant for enzymatic activity is necessary for degradation. 20. Li, X. & Coffino, P. Distinct domains of antizyme required for binding and proteolysis of ornithine decarboxylase. Mol. Cell. Biol. 14, 87–92 (1994). Identified distinct regions of antizyme that mediate target binding and degradation. 21. Smith, R. C., Litwin, M. S., Lu, Y. & Zetter, B. R. Identification of an endogenous inhibitor of prostatic carcinoma cell growth. Nature Med. 1, 1040–1045 (1995). 22. Koike, C., Chao, D. T. & Zetter, B. R. Sensitivity to polyamine-induced growth arrest correlates with antizyme induction in prostate carcinoma cells. Cancer Res. 59, 6109–6112 (1999). Indicated a possible function for antizyme in the regulation of prostate cell growth. 23. Newman, R. M. et al. Antizyme targets cyclins and cyclin-dependent kinases for degradation. (submitted). 24. Pietila, M. et al. Activation of polyamine catabolism profoundly alters tissue polyamine pools and affects hair growth and female fertility in transgenic mice overexpressing spermidine/spermine N1-acetyltransferase. J. Biol. Chem. 272, 18746–18751 (1997). 25. Megosh, L. et al. Increased frequency of spontaneous skin tumors in transgenic mice which overexpresses ornithine decarboxylase. Cancer Res. 55, 4205–4209 (1995). 26. Feith, D. & Pegg, A. Targeted antizyme expression protects transgenic mice from skin carcinogenesis. Proc. Am. Assoc. Cancer Res. 41, 208 (2000). 27. MacKintosh, C., Feith, D., Shantz, L. & Pegg, A. Overexpression of antizyme in the hearts of transgenic

28.

29.

30.

31.

32.

33. 34.

35.

36.

mice prevents the isoprenaline-induced increase in cardiac ornithine decarboxylase activity, but does not prevent cardiac hypertrophy. Biochem. J. 350, 645–653 (2000). Ivanov, I., Gesteland, R. & Atkins, J. Antizyme expression: A subversion of triplet decoding, which is remarkably conserved by evolution, is a sensor for an autoregulatory circuit. Nucleic Acids Res. 28, 3185–3196 (2000). Zhu, C., Lang, D. W. & Coffino, P. Antizyme2 is a negative regulator of ornithine decarboxylase and polyamine transport. J. Biol. Chem. 274, 26425–26430 (1999). Ivanov, I., Rohrwasser, A., Terreros, D., Gesteland, R. & Atkins, J. Discovery of a spermatogenesis, stage specific, ornithine decarboxylase antizyme:antizyme 3. Proc. Natl Acad. Sci. USA 97, 4808–4813 (2000). Described a novel form of antizyme with a very restricted tissue distribution. Tosaka, Y. et al. Identification and characterization of testis specific ornithine decarboxylase antizyme (OANTIZYME-t) gene: Expression in haploid germ cells and polyamine-induced frameshifting. Genes to Cells 5, 265–276 (2000). Hakovirta, H. et al. Polyamines and regulation of spermatogenesis: selective stimulation of late spermatogonia in transgenic mice overexpressing the human ornithine decarboxylase gene. Mol. Endocrinol. 7, 1430–1436 (1993). Verma, R. & Deshaies, R. J. A proteasome howdunit: the case of the missing signal. Cell 101, 341–344 (2000). Mitchell, J. L. A., Judd, G. G., Bareyal-Leyser, A. & Ling, S. Y. Feedback repression of polyamine transport is mediated by antizyme in mammalian tissue-culture cells. Biochem. J. 299, 19–22 (1994). He, Y., Suzuki, T., Kashiwagi, K. & Igarashi, K. Antizyme delays the restoration by spermine of growth of polyaminedeficient cells through its negative regulation of polyamine transport. Biochem. Biophys. Res. Commun. 203, 608–614 (1994). Seiler, N., Delcros, J. G. & Moulinoux, J. P. Polyamine

37.

38.

39.

40.

41.

42.

43.

44.

45.

transport in mammalian cells. An update. Int. J. Biochem. Cell Biol. 28, 843–861 (1996). Murakami, Y., Ichiba, T., Matsufuji, S. & Hayashi, S. Cloning of antizyme inhibitor, a highly homologous protein to ornithine decarboxylase. J. Biol. Chem. 271, 3340–3342 (1996). Russell, D. & Snyder, S. H. Amine synthesis in rapidly growing tissues: ornithine decarboxylase activity in regenerating rat liver, chick embryo, and various tumors. Proc. Natl Acad. Sci. USA 60, 1420–1427 (1968). Russell, D. H. & Snyder, S. H. Amine synthesis in regenerating rat liver: extremely rapid turnover of ornithine decarboxylase. Mol. Pharmacol. 5, 253–262 (1969). Kay, J. E. & Lindsay, V. J. Control of ornithine decarboxylase activity in stimulated human lymphocytes by putrescine and spermidine. Biochem. J. 132, 791–796 (1973). Kameji, T., Murakami, Y., Fujita, K. & Hayashi, S. Purification and some properties of ornithine decarboxylase from rat liver. Biochim. Biophys. Acta 717, 111–117 (1982). Murakami, Y. & Hayashi, S. Role of antizyme in degradation of ornithine decarboxylase in HTC cells. Biochem. J. 226, 893–896 (1985). Matsufuji, S. et al. Antizyme analyses of ornithine decarboxylase antizyme mRNA with a cDNA cloned from rat liver. J. Biochem. 108, 365–371 (1990). Miyazak, Y., Matsufuji, S. & Hayashi, S. Cloning and characterization of a rat gene encoding ornithine decarboxylase antizyme. Gene 113, 191–197 (1992). Murakami, Y., Matsufuji, S., Miyantizymeaki, Y. & Hayashi, S. Destabilization of ornithine decarboxylase by transfected antizyme gene expression in hepatoma tissue culture cells. J. Biol. Chem. 267, 13138–13141 (1992).

Acknowledgements I am grateful to J. Atkins, D. Finley, S. Matsufuji, A. Pegg and B. Zetter for providing data before publication. The author’s laboratory is supported, in part, by a National Institutes of Health grant.

| MARCH 2001 | VOLUME 2

www.nature.com/reviews/molcellbio

REVIEWS

PROTEIN REGULATION BY MONOUBIQUITIN Linda Hicke Multi-ubiquitin chains at least four subunits long are required for efficient recognition and degradation of ubiquitylated proteins by the proteasome, but other functions of ubiquitin have been discovered that do not involve the proteasome. Some proteins are modified by a single ubiquitin or short ubiquitin chains. Instead of sending proteins to their death through the proteasome, monoubiquitylation regulates processes that range from membrane transport to transcriptional regulation. U B I Q U I T I N A N D P R OT E A S O M E S HISTONE

A family of small, highly conserved basic proteins, found in the chromatin of all eukaryotic cells, that associate with DNA to form a nucleosome. ENDOCYTOSIS

Internalization and transport of extracellular material and plasma membrane proteins from the cell surface to intracellular organelles known as endosomes. RETROVIRUS

RNA virus that uses reverse transcriptase to convert its RNA into DNA.

Department of Biochemistry, Molecular Biology and Cell Biology, Northwestern University Evanston, Illinois 60208, USA. e-mail: l-hicke@northwestern.edu

Ubiquitin is a versatile beast. It is a small protein that serves as a complex post-translational modification that is conjugated to lysine residues on a wide assortment of substrates (see the review by Allan Weissman on page 169 of this issue). Proteins can be modified by monoubiquitylation, the conjugation of a single ubiquitin molecule to one or several lysines. In addition, because ubiquitin itself carries lysines that act as sites of self-conjugation, multi-ubiquitin chains can form and be appended to proteins (FIG. 1). Multi-ubiquitin chains, in which the carboxy-terminal glycine of ubiquitin is linked to Lys48 of the preceding ubiquitin, mediate the best-characterized function of ubiquitin1,2 — to target proteins for destruction by the proteasome (see the review by Peter Kloetzel on page 179 of this issue; reviewed in REFS 3,4). However, other functions for ubiquitin are being discovered at a rapid rate, many of which are controlled by alternative types of ubiquitin modification5–9. Several proteins that are monoubiquitylated have previously been identified in vivo or in vitro, but only very recently has it become clear that monoubiquitin is a regulator of the location and activity of diverse cellular proteins. Monoubiquitylation is involved in at least three distinct cellular functions: HISTONE regulation, ENDOCYTOSIS and the budding of RETROVIRUSES from the plasma membrane. Histone regulation

Histones were found to be monoubiquitylated more than 20 years ago10–12, but only during the past year have

a K

b K

c K

d K

Figure 1 | Monoubiquitylation versus multiubiquitylation. Proteins might be (a, b) monoubiquitylated or (c, d) multi-ubiquitylated on one or more lysine (K) residues. (c) Multi-ubiquitin chains are most commonly linked through Lys48 of ubiquitin. Other links (d) occur and probably result in chains with different structures. Chains linked through Lys29, Lys48 and Lys63 occur in vivo in yeast65. Chains linked through other lysines have been synthesized in vitro and might occur in higher eukaryotic cells66.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 1 9 5

REVIEWS a H2A

H2B

H2A H4 TAF250 b

c ubac N Ub

N N H1

N C

Rad6

N N

NUCLEOSOME

The basic structural subunit of chromatin, which consists of ~200 base pairs of DNA and an octamer of histones. SPORULATION

Sexual reproduction in yeast and fungi. UBIQUITIN-CONJUGATING ENZYME (E2)

An enzyme that accepts ubiquitin from a ubiquitinactivating enzyme (E1) and, together with a ubiquitin ligase (E3), transfers it to a substrate protein. TAF250

A subunit of TFIID, where TAF stands for TBP-associated factor, and TBP stands for TATA-box-binding protein. TFIID

Transcription factor IID. A multisubunit general transcription factor, necessary for the transcription of all genes in eukaryotes. LYSOSOME

A membrane-bounded organelle with a low internal pH (4–5) that contains hydrolytic enzymes and that is the site of the degradation of proteins in both the biosynthetic and the endocytic pathways.

196

Figure 2 | Monoubiquitylation of histones. a | Structure of a nucleosome including the four core histones enwrapped by DNA (white); H2A (orange); H2B (red); H3 (blue); H4 (green). b,c | The nucleosome of core histones, represented as a grey ball, surrounded by a (light blue) left-handed DNA helix. Linker histone H1 is shown in maroon. b | The carboxy-terminal tails of core histones H2A and H2B are sites of ubiquitylation (Lys119 on H2A; Lys120 on H2B; Lys123 on yeast H2B). The carboxyl terminus of H2A is exposed and is shown in blue; the carboxyl terminus of H2B is buried in the nucleosome67. The Rad6/Ubc2 ubiquitin-conjugating enzyme ubiquitylates the core histones. Monoubiquitylation of the carboxy-terminal tails might alter the position of the histone to open chromatin structure. c | The amino termini of the core histones, which are sites of acetylation, protrude from the nucleosome structure67 (a) and are represented as blue tails. The TAF250 subunit of transcription factor IID ubiquitylates H1. A speculative model is shown in which the modification of one histone tail influences the modification of another62. The bromodomains of TAF250 bind to acetylated core histone tails, perhaps promoting H1 monoubiquitylation by the TAF250 ubac (ubiquitinactivating/conjugating) domain. The location of the H1 lysine that is conjugated to ubiquitin is not known.

functions been identified for histone ubiquitylation13,14. Histones H2A and H2B form part of the core structure of NUCLEOSOMES (FIG. 2a). They are modified by monoubiquitin or short ubiquitin chains on lysines in their carboxy-terminal tails (FIG. 2b). In mammalian cells, ~10% of H2A and ~1% of H2B are ubiquitylated but in the budding yeast Saccharomyces cerevisiae, H2B is the predominant ubiquitylated histone. Monoubiquitylation clearly has an important function in yeast, because yeast cells carrying a mutant H2B that lacks its ubiquitylation site do not SPORULATE, indicating that H2B monoubiquitylation is needed for meiosis. These cells also grow more slowly: their doubling time is 1.3 times as long as that of wild-type cells, so monoubiquitylation of H2B is necessary for normal

growth13. Further evidence that is consistent with a role for histone ubiquitylation in meiosis comes from the examination of mutants that are defective in the Rad6 UBIQUITIN-CONJUGATING ENZYME (E2), which catalyses the ubiquitylation of H2A and H2B (REFS 13,15) (see the review by Allan Weissman for details of the ubiquitylation process). Yeast rad6 mutants, like the H2B mutants discussed above, cannot sporulate16 and mouse knockouts lacking a Rad6 homologue (HR6B) are specifically defective in spermatogenesis, leading to male infertility17. Histones other than H2A and H2B are modified by ubiquitin. The H1 histone is not part of the core nucleosome structure, but is a ‘linker’ histone that binds to DNA, extending between neighbouring nucleosomes (FIG. 2c). It is monoubiquitylated in Drosophila melanogaster embryos14. This modification depends on a component of the general transcription machinery, TAF250 (a subunit of TFIID), which has a ubiquitin conjugation activity that is specific for H1 in vitro. Mutations in TAF250 that inhibit its ubiquitylation activity cause defects in transcriptional activation in Drosophila embryos14, indicating that TAF250-dependent H1 ubiquitylation might control gene expression. TAF250 seems to be an unusual component of the ubiquitylation machinery. Generally, a cascade of three enzymes participates in a ubiquitylation reaction: E1 activates ubiquitin in an ATP-dependent step; E2 accepts activated ubiquitin from the E1 and, with the help of an E3 (ubiquitin ligase), transfers ubiquitin to specific substrates. TAF250 carries domains that are homologous to E1s and E2s, and seems to have both of these activities fused in one polypeptide chain. This unusual type of ubiquitylation activity might be one way of targeting histones for monoubiquitylation rather than modification with multi-ubiquitin chains. Endocytosis

Monoubiquitylation also regulates the activity of proteins located at the plasma membrane. Many plasma membrane proteins are downregulated by internalization into the endocytic pathway. In Saccharomyces cerevisiae, most of these proteins require ubiquitylation of their cytoplasmic domains to be internalized (reviewed in REFS 18,19). In mammalian cells, several ion channels and signal-transducing receptors that undergo regulated internalization are ubiquitylated in response to an extracellular signal, and ubiquitylation regulates their endocytic transport18,20–25. Most proteins that are ubiquitylated at the plasma membrane are targeted for degradation in the LYSOSOME (for an exception, see REF. 26). But how is ubiquitin used to send membrane proteins to their deaths in this intracellular acid bath? In both yeast and mammalian cells, monoubiquitylation is sufficient to trigger internalization into primary endocytic vesicles6,27–29. For several yeast amino-acid permeases, di-ubiquitin chains linked through Lys63 further enhance basal internalization rates5,30. Internalization information is carried in the ubiquitin polypeptide itself because ubiquitin can stimulate internalization when it is fused in-frame to receptors that lack lysine www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS

EPS15

Epidermal growth factor receptor pathway substrate clone 15. Mammalian protein required for budding of clathrin-coated vesicles during endocytosis. CLATHRIN

The main component of the coat that is associated with clathrin-coated vesicles, which are involved in membrane transport both in the endocytic and biosynthetic pathways. DE-UBIQUITYLATING ENZYME

Enzyme that catalyses the cleavage of ubiquitin from multi-ubiquitin chains or protein conjugates. GROWTH HORMONE RECEPTOR

A signal transducing receptor of the tyrosine-kinase family. GAG

The protein of the nucleocapsid shell around the RNA of a retrovirus. POLYPROTEIN

A single polypeptide chain that is cleaved into several separate proteins. UBIQUITIN PROTEIN LIGASE (E3)

An enzyme that acts together with a ubiquitin-conjugating enzyme (E2) to couple the small protein ubiquitin to lysine residues on a target protein, marking that protein for destruction by the proteasome.

ubiquitylation sites or other internalization signals, or to heterologous proteins that are not normally internalized28,29,31 (FIG. 3a). Internalization is controlled by the fused ubiquitin moiety, rather than post-translationally conjugated ubiquitin, because a fused ubiquitin that lacks all its lysine residues, and therefore cannot conjugate with further ubiquitin molecules, also functions to promote internalization of chimeric proteins29,31. Monoubiquitin is not only required as an internalization signal on endocytic cargo; it might also control the activity of the endocytic machinery. EPS15, a protein that interacts with the CLATHRIN-based endocytic machinery, becomes monoubiquitylated upon stimulation of cells with epidermal growth factor32, a ligand that upregulates activity of the endocytic machinery33. Ubiquitylation is a reversible process, owing to the existence of DE-UBIQUITYLATING ENZYMES. Fat facets is one such de-ubiquitylating enzyme. Eyes of Drosophila fat facets mutants do not develop normally and have too many cells in the facets that make up the compound eye34. Genetic evidence strongly implicates an interaction of Fat facets with epsin, an endocytic protein that interacts with Eps15 (REF. 35), again indicating that ubiquitylation might regulate the activity of the endocytic machinery. More support for this idea comes from the finding that internalization of the GROWTH HORMONE RECEPTOR requires the ubiquitylation machinery, although ubiquitylation of the receptor itself is not necessary20. The targets of the ubiquitin machinery that are required for growth hormone receptor internalization are not known.

sequence that is similar to a tyrosine-based endocytosis signal. This sequence, YXXL, found in the L domain of the equine infectious anemia virus binds to the AP2 (adaptor protein 2) complex44. These observations indicate that either ubiquitin or tyrosine-based internalization signals might be able to co-opt the host cell endocytic machinery to promote virus budding. Monoubiquitin versus multi-ubiquitin signals

Do the characterized functions of monoubiquitin involve the proteasome? This is unlikely as efficient proteasome binding and degradation require a multia or

Rsp5 and Nedd4 Through PPXY or phospho-Ser/Thr? Internalization

Cbl Through phospho-Tyr

X Ub Ub

Ub Ub

Virus budding

Enveloped viruses bud from the plasma membrane of infected cells in the opposite direction to that of vesicles during endocytosis (reviewed in REF. 36), and evidence is accumulating that monoubiquitin is an essential part of this process (FIG. 4). The GAG POLYPROTEIN, common to all retroviruses, which associates with the cytoplasmic face of the plasma membrane, is necessary and sufficient for pinching off membrane-bounded particles during the process of budding. Within Gag, there is a small sequence called the late (L) domain that is essential for budding and particle release. In most cases, the L domain carries a proline-rich sequence, PPXY and/or P(T/S)AP, and this domain can act as an interaction motif with a UBIQUITIN PROTEIN LIGASE (E3) known as Nedd4 (J. Leis, personal communication and REF. 37). Gag itself is monoubiquitylated38 and ubiquitylation depends on its L domain37. Ubiquitylation of Gag is important for its function in virus budding because the depletion of intracellular ubiquitin levels inhibits budding37,39,40. Under these conditions, budding is partially restored by the inframe fusion of ubiquitin to a Gag protein39, indicating that monoubiquitin is sufficient to promote budding. The requirement of ubiquitin for virus budding might be more general because other enveloped viruses (vesicular stomatitis virus and Ebola) carry PPXY sequences in matrix proteins that seem to function analogously to the Gag L domain37,41â&#x20AC;&#x201C;43. Moreover, in at least one case it seems that the proline-rich L domain in Gag that mediates ubiquitylation can be replaced by a

Figure 3 | Roles of monoubiquitin during internalization of plasma membrane proteins into the endocytic pathway. a | Post-translationally conjugated ubiquitin or ubiquitin fused in-frame can induce the internalization of plasma membrane proteins. Ubiquitin itself carries a threedimensional internalization signal (shown in blue) and acts as a regulated endocytosis signal that can be appended to proteins destined for downregulation. Ubiquitin protein ligases that are known to modify plasma membrane proteins are indicated. Cbl is a proto-oncoprotein that recognizes and ubiquitylates activated, phosphorylated growth factor receptors by binding to phospho-tyrosines25,68. It is required to downregulate activated receptors by endocytosis and subsequent degradation in the lysosome. Nedd4 binds to and ubiquitylates the epithelial sodium channel that undergoes ubiquitin-dependent degradation in the lysosome21,69. PPXY motifs in channel subunits are required for Nedd4 interaction. The yeast homologue of Nedd4, Rsp5, is required for the ubiquitylation and internalization of several proteins19. b | Models for the role of ubiquitin in endocytosis. The ubiquitin internalization signal might recruit a binding protein that promotes endocytosis. This protein might localize endocytic cargo into subdomains of the plasma membrane competent for vesicle budding or might be part of the vesicle budding machinery. Ubiquitylated plasma membrane proteins might form multimers70,71 that provide a noncovalently linked multivalent binding surface on ubiquitin. The activity of at least one component of the endocytic machinery (shown schematically in yellow) seems to be regulated by monoubiquitylation. This monoubiquitylated protein and the ubiquitin internalization signal on endocytic cargo could act to assemble a complex that is necessary for the budding of primary endocytic vesicles.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 1 9 7

ÂŠ 2001 Macmillan Magazines Ltd

REVIEWS

X Nedd4

Nedd4

Figure 4 | Monoubiquitylation of Gag is required for retrovirus budding. The Gag polyproteins (ovals) associate with the cytoplasmic face of the plasma membrane as a multimeric lattice. The L domain of Gag (purple) is not required for membrane association but is required at a late step in viral budding. The L domain contains proline-rich sequences that recruit a ubiquitin protein ligase, Nedd4 (blue circle), to membrane-associated Gag. Monoubiquitylation of Gag is required for virus budding. In avian retroviruses, there are ~1,500 Gag proteins in the budded virion and less than 100 molecules of free ubiquitin (red circles). Very little or no conjugated ubiquitin is found associated with virus particles, indicating that Gag might be de-ubiquitylated before or immediately after budding. The fusion of ubiquitin inframe to a Gag protein (right) can overcome a defect in virus budding induced by low levels of free ubiquitin in the cell. The L domain is still required for budding in cells expressing a Gagubiquitin chimera, indicating that, like endocytosis, virus budding might require L-domainmediated ubiquitylation of a protein other than virus cargo.

SCANNING ALANINE MUTAGENESIS

A method for determining the function of every residue in a protein sequence by mutating each one to alanine. CULLIN

A family of proteins present in multisubunit ubiquitin ligases; they recruit RING-fingercontaining proteins to the ligase complex. SCF UBIQUITIN LIGASE

A multisubunit ubiquitin ligase that contains Skp1, a member of the cullin family (Cul1), and an F-box-containing protein (Skp2), as well as a RINGfinger-containing protein (Roc1/Rbx1). GATE-16

Protein with a ubiquitin fold, required for intra-Golgi transport and autophagy.

198

ubiquitin chain at least four subunits long in vitro2. Instead, monoubiquitylated proteins are stable or are degraded in the lysosome. For instance, monoubiquitylated histones are stable proteins and ubiquitylated, internalized plasma membrane proteins in yeast are degraded only after transport to the lysosome-like vacuole45. Yeast mutants that are defective in either regulatory or proteolytic subunits of the proteasome, including the known ubiquitin-binding subunit Rpn10, have no effect on plasma membrane protein internalization or degradation (REFS 45,46 and L.H., unpublished data). Although proteasome inhibitors repress retrovirus budding, this effect is due to the lower cellular levels of ubiquitin that are induced by the inhibitors and not to a direct role of the proteasome during budding37,39,40. Which parts of the ubiquitin molecule participate in multi-ubiquitin- or monoubiquitin-dependent functions? Despite the different targets of mono- versus multi-ubiquitin, there is overlap in the regions of the ubiquitin molecule that are required for these functions. Lys48-linked tetraubiquitin — the smallest multi-ubiquitin chain that is recognized by the proteasome — has a repeating hydrophobic patch on its surface that consists of Leu8, Ile44 and Val70 (REF. 47) (FIG. 5). This patch is required for binding to the proteasome and for the subsequent degradation of ubiquitylated substrates48. Surprisingly, SCANNING ALANINE MUTAGE31 NESIS of a receptor–ubiquitin chimera revealed that

the same patch is necessary for endocytosis: the substitution of Ile44 severely diminished internalization of the chimeric protein and substitution of Val70 or Leu8 had a smaller, but still significant effect on internalization. Furthermore, another patch of ubiquitin surface residues that surround Phe4 is specifically required for endocytosis (FIG. 5). Although Phe4 is exposed in a tetraubiquitin chain, Phe4 mutations do not affect proteasome binding or degradation49. How does the Leu8/Ile44/Val70 patch, which is important for both proteasome-mediated degradation and endocytosis28,31,48, bind its different targets — the 19S subunit of the proteasome and the endocytic machinery? For proteasome binding, the Leu8/Ile44/Val70 patch acts as a multivalent-binding surface on tetraubiquitin. Presumably, it acts differently on monoubiquitin (see the structure of monoubiquitin online) to bind proteins during endocytosis (see below), perhaps as part of an extended protein-binding site with the Phe4 patch. Whether Phe4 is required generally for cellular processes that involve monoubiquitin or whether Phe4 functions only during endocytosis is not yet known. A reversible regulatory modification

The ubiquitin-like proteins (UBLs) are a family of proteins that are distantly related in amino-acid sequence and share the same structural fold. Like ubiquitin, they are covalently conjugated to substrates by an isopeptide bond through their carboxyl termini50. (Some proteins carry a UBL domain in the context of a larger sequence; these proteins are not conjugated to other proteins, but their UBL domain might target them for interaction with the proteasome51,52.) Ubiquitin seems to be the only member of the UBL family that can self-conjugate to form chains. Might monoubiquitin regulate proteins in a manner that is similar to that of the UBLs, which seem to function as monomeric conjugates? The best-characterized UBLs are SUMO-1 (small ubiquitin-like modifier 1), Rub1 (related to ubiquitin; also known as Nedd8) and Apg12. Each of the UBLs modifies one or more substrates and, akin to phosphorylation and acetylation, modification with a UBL controls the activity, post-translational modification or localization of the target. Rub1 modifies members of the CULLIN family that are subunits of the SCF UBIQUITIN LIGASES, and rubylation — conjugation of proteins with Rub1 — is required for maximal SCF ligase activity50. Apg12 is a UBL whose conjugation to Apg5 is required for autophagy, the formation and fusion of membranebounded vesicles with the lysosome (see the review by Yoshinori Ohsumi on page 211 of this issue). It is not clear how Apg12 conjugation functions during autophagy but at least one of its roles seems to be to enhance a specific protein–protein interaction53. Apg8 and its mammalian homologue, GATE-16, are other UBLs that are required for autophagy. Apg8 acts in a novel and surprising way — it is conjugated through its carboxyterminal glycine to phosphatidylethanolamine by enzymes that resemble E1 and E2, and this modification regulates association of Apg8 with the membrane. One example has been described in which ubiquitin is also www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS

CHAPERONE

A protein that ensures the proper folding of other proteins.

modified with phospholipid (in membranes of baculovirus particles); however, unlike Apg8, this does not occur by conjugation through the carboxyl terminus of ubiquitin54. The UBL SUMO-1 is conjugated to many types of substrate (at least 20) (see the review by Stefan Jentsch and colleagues on page 202 of this issue). In several cases, sumoylation (conjugation of proteins to SUMO1) seems to regulate the location of proteins to sites in or around the nucleus. In others, SUMO-1 is conjugated to the same lysines that are also modified by ubiquitin, which apparently negatively regulates multi-ubiquitylation and thereby prevents degradation. Although the mechanisms by which different UBLs act remain to be discovered, it is clear that monomodification by UBLs can regulate substrate proteins in different ways. Monoubiquitin, just like the UBLs, seems to function as a reversible modification to regulate protein function — a role quite distinct from that of multi-ubiquitin. The type of ubiquitin modification that a protein receives is therefore crucial for its fate. Proteins might be monoubiquitylated on one or more lysine residues, or might be modified by multi-ubiquitin chains linked through Lys48, Lys63 or other ubiquitin lysines (FIG. 1). But how is ubiquitylation regulated so that proteins that acquire a monoubiquitin signal are not inappropriately multi-ubiquitylated and degraded? The answer to this question is not known. Specialized ubiquitylation machinery such as the unusual ubac (ubiquitin-activating/conjugating) domain that seems to act as a fused E1–E2 in TAF250 might catalyse some monoubiquitylation reactions. Some combinations of the standard ubiquitylation enzymes — E1s, E2s and E3s — seem to participate in both monoubiquitylation and the formation of multi-ubiquitin chains. For instance, the Rsp5 E3 enzyme is required not only to modify plasma membrane proteins with monoubiquitin or with di-ubiquitin chains that are linked through Lys63 (REFS 5,30,55,56), but also to catalyse the multi-ubiquitylation of RNA polymerase II (REF. 57). Therefore, it is likely that there are positive and negative regulators for the assembly of ubiquitin chains in the cell (TABLE 1). Several types of protein that regulate the length of ubiquitin chains during synthesis have recently been identified58,59. Some ubiquitylation reactions require the presence of a positive regulator, known as an E4, to progress from monoubiquitin or short ubiquitin chains to long multiubiquitin chains58. Conversely, several mechanisms negatively regulate the formation of multi-ubiquitin chains. First, a subset of the many de-ubiquitylating enzymes might trim ubiquitin chains on specific substrates to

Table 1 | Positive and negative regulators of ubiquitin chain assembly Protein

Substrate

Function in ubiquitin

E4 (Ufd2)

Multi-ubiquitin conjugates

Assembly of multi-ubiquitin chains58 (longer than di- or tri-ubiquitin)

Rad23

Ubiquitin conjugates

De-ubiquitylating enzymes

Multi-ubiquitylated proteins

Inhibits the assembly of long-chain multi-ubiquitin conjugates59 Edits the length of ubiquitin chains72

Tle44 Phe4 Val70

Leu8

Carboxyl terminus

Figure 5 | Three-dimensional structure of ubiquitin. Ubiquitin amino acids required for proteasome binding and/or endocytosis. Residues that are important for either function are shown in magenta. Residues that have a minor role in endocytosis are shown in pink.

present a monoubiquitin signal (it is possible that Fat facets could have this type of editing function). Second, a negative regulator of multi-ubiquitin chain formation, similar to the Rad23 protein that seems to protect substrates from proteasome-mediated degradation59, might promote monoubiquitylation. A negative regulator of chain assembly that limits the conjugation of ubiquitin to a single ubiquitin moiety would protect substrates from the proteasome by preventing formation of the necessary multi-ubiquitin chain recognition signal, thereby allowing monoubiquitin to carry out other regulatory functions. Mechanisms of regulation

The mechanism by which monoubiquitin regulates substrates is another mystery. Ubiquitin might alter the conformation or oligomeric state of a protein. Much of the cellular ubiquitin is translated as fusions to ribosomal proteins. This ubiquitin acts as a CHAPERONE to promote folding of the newly synthesized proteins before being cleaved60, indicating that monoubiquitin might influence the local threedimensional structure of a substrate. In actively transcribed chromatin, ubiquitylated H2A and H2B are enriched and H1 is either depleted or its interaction with the nucleosome is altered61. So, ubiquitin might affect histone structure to control how histones interact with each other or with DNA, and this might alter local chromatin organization to make promoter DNA more or less accessible to the general transcription machinery. Instead or in addition, ubiquitin could recruit or inhibit binding of regulatory proteins that interact with histones to regulate transcription. We do not know how monoubiquitin regulates other functions, such as endocytosis and virus budding. It is possible that ubiquitin interacts with lipids to regulate the localization of proteins to budding-competent plasma membrane subdomains or to regulate the formation or fission of a budding vesicle. Ubiquitin-dependent budding might also involve the interaction of ubiquitin with specific monoubiquitin-binding proteins that have

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 1 9 9

REVIEWS yet to be identified (FIG. 3b). An interaction between monoubiquitin and a particular protein could be specified by the location of the ubiquitylated substrate. In some cases, determinants from the substrate (including those resulting from other post-translational modifications such as phosphorylation and acetylation62) might contribute to an interaction. In others, the actual regulatory signal might be more complex than just a single ubiquitin molecule. For example, plasma membrane proteins and membrane-associated Gag proteins can form multimers that might allow the formation of a unique binding surface composed of non-covalently linked, but closely associated, ubiquitin monomers. Similarly, the monoubiquitin moieties attached to different histones in a nucleosome might interact to form a new binding site. These models could explain how monoubiquitin acts as a signal in the sea of ubiquitin that is present in the cell. Ubiquitin goes solo: future challenges

Monoubiquitin, then, like its UBL cousins, acts as a regulator of substrate–protein location and activity. The observation that most of the UBLs, unlike ubiquitin, act primarily as monomodifiers indicates that perhaps ubiquitin originated as one of several polypeptide modifications that regulated protein location and activity. Ubiquitin might have been singled out to form multi-ubiquitin chains, an ability that evolved to expand its regulatory activities. Or, the UBLs might have evolved from ubiquitin to capitalize on its novel conjugation mechanism and amplify the ways in which reversible polypeptide modifications could

Chau, V. et al. A multiubiquitin chain is confined to specific lysine in a targeted short-lived protein. Science 243, 1576–1583 (1989). 2. Thrower, J. S., Hoffman, L., Rechsteiner, M. & Pickart, C. M. Recognition of the polyubiquitin proteolytic signal. EMBO J. 19, 94–102 (2000). 3. Hochstrasser, M. Ubiquitin-dependent protein degradation. Annu. Rev. Genet. 30, 405–439 (1996). 4. Voges, D., Zwickl, P. & Baumeister, W. The 26S proteasome: a molecular machine designed for controlled proteolysis. Annu. Rev. Biochem. 68, 1015–1068 (1999). 5. Galan, J. M. & Haguenauer-Tsapis, R. Ubiquitin Lys63 is involved in ubiquitination and endocytosis of a yeast plasma membrane protein. EMBO J. 16, 5847–5854 (1997). 6. Terrell, J., Shih, S., Dunn, R. & Hicke, L. A function for monoubiquitination in the internalization of a G proteincoupled receptor. Mol. Cell 1, 193–202 (1998). This paper shows that monoubiquitylation on a single lysine residue is necessary and sufficient for rapid endocytosis of an activated signal-transducing receptor in yeast. 7. Hofmann, R. M. & Pickart, C. M. Noncanonical MMS2encoded ubiquitin-conjugating enzyme functions in assembly of novel polyubiquitin chains for DNA repair. Cell 96, 645–653 (1999). 8. Deng, L. et al. Activation of the IκB kinase complex by TRAF6 requires a dimeric ubiquitin-conjugating enzyme complex and a unique polyubiquitin chain. Cell 103, 351–361 (2000). 9. Spence, J. et al. Cell-cycle-regulated modification of the ribosome by a variant multiubiquitin chain. Cell 102, 67–76 (2000). 10. Busch, H. & Goldknopf, I. L. Ubiquitin-protein conjugates. Mol. Cell. Biochem. 40, 173–187 (1981). 11. van Holde, K. E. Chromatin (Springer, New York, 1988). 12. Spencer, V. A. & Davie, J. R. Role of covalent modifications of histones in regulating gene expression. Gene 240, 1–12 (1999).

200

control protein fate. In any case, monoubiquitin has joined the expanding cohort of regulatory ubiquitin chain lengths and linkages that mark proteins for distinct fates and functions. But, there is still a lot to learn in this rapidly expanding area. How does monoubiquitin act as a signal for different cellular functions at the cell surface, in the cytoplasm or in the nucleus? We need to know how monoubiquitin affects the structure, location and activity of modified proteins. In addition, new cellular functions of monoubiquitin are likely to arise in the near future. Monoubiquitylated proteins that have not been discussed here have been identified in vivo and in vitro (for example, see REFS 63,64) and others probably exist. Moreover, proteins that are primarily monoubiquitylated might not have been accurately identified as such because many antisera against ubiquitin efficiently recognize only multi-ubiquitin or multi-ubiquitylated conjugates. Identifying monoubiquitylated proteins, and working out how monoubiquitylation affects their function and location, will be quite a challenge. Links DATABASE LINKS H2A | H2B | Rad6 | H1 | TAF250 | Fat facets | epsin | Eps15 | growth hormone receptor | Nedd4 | AP2 | Rpn10 | Apg12 | SCF ligase | Apg5 | Apg8 | SUMO-1 | Rsp5 | Rad23 FURTHER INFORMATION Structure of monoubiquitin | Hicke lab ENCYCLOPEDIA OF LIFE SCIENCES Ubiquitin pathway | Lysosomal degradation of protein | Nucleosomes: detailed structure and mutations

13. Robzyk, K., Recht, J. & Osley, M. A. Rad6-dependent ubiquitination of histone H2B in yeast. Science 287, 501–504 (2000). Although histones H2A and H2B were known to be monoubiquitylated for more than two decades, this paper provided the first evidence that histone ubiquitylation was important for function. Yeast mutants that cannot ubiquitylate histone H2B grow more slowly than wild-type cells and do not sporulate. 14. Pham, A. D. & Sauer, F. Ubiquitin-activating/conjugating activity of TAF(II)250, a mediator of activation of gene expression in Drosophila. Science 289, 2357–2360 (2000). Drosophila histone H1 is monoubiquitylated by TAF250, an unusual multifunctional protein that seems to carry E1 and E2 activities in the same polypeptide. TAF250 ubiquitylation of H1 seems to be important for the proper regulation of transcriptional activity in Drosophila embryos. 15. Jentsch, S., McGrath, J. P. & Varshavsky, A. The yeast DNA repair gene RAD6 encodes a ubiquitin-conjugating enzyme. Nature 329, 131–134 (1987). 16. Prakash, L. The structure and function of RAD6 and RAD18 DNA repair genes of Saccharomyces cerevisiae. Genome 31, 597–600 (1989). 17. Roest, H. P. et al. Inactivation of the HR6B ubiquitinconjugating DNA repair enzyme in mice causes male sterility associated with chromatin modification. Cell 86, 799–810 (1996). 18. Hicke, L. Gettin’ down with ubiquitin: turning off cell surface receptors, transporters and channels. Trends Cell Biol. 9, 107–112 (1999). 19. Rotin, D., Staub, O. & Haguenauer-Tsapis, R. Ubiquitination and endocytosis of plasma membrane proteins: Role of Nedd4/Rsp5p family of ubiquitin-protein ligases. J. Membr. Biol. 176, 1–17 (2000). 20. Strous, G., van Kerkhof, P., Govers, R., Ciechanover, A. & Schwartz, A. L. The ubiquitin conjugation system is

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

required for ligand-induced endocytosis and degradation of the growth hormone receptor. EMBO J. 15, 3806–3812 (1996). Staub, O. et al. Regulation of stability and function of the epithelial Na+ channel (ENaC) by ubiquitination. EMBO J. 16, 6325–6336 (1997). Bonifacino, J. S. & Weissman, A. M. Ubiquitin and the control of protein fate in the secretory and endocytic pathways. Annu. Rev. Cell Dev. Biol. 14, 19–57 (1998). Levkowitz, G. et al. c-Cbl/Sli-1 regulates endocytic sorting and ubiquitination of the epidermal growth factor receptor. Genes Dev. 12, 3663–3674 (1998). Lee, P. S. et al. The Cbl protooncoprotein stimulates CSF-1 receptor multiubiquitination and endocytosis, and attenuates macrophage proliferation. EMBO J. 18, 3616–3628 (1999). Levkowitz, G. et al. Ubiquitin ligase activity and tyrosine phosphorylation underlie suppression of growth factor signaling by c-Cbl/Sli-1. Mol. Cell 4, 1029–1040 (1999). Jeffers, M., Taylor, G. A., Weidner, K. M., Omura, S. & Vande Woude, G. F. Degradation of the Met tyrosine kinase receptor by the ubiquitin–proteasome pathway. Mol. Cell. Biol. 17, 799–808 (1997). Lucero, P., Penalver, E., Vela, L. & Lagunas, R. Monoubiquitination is sufficient to signal internalization of the maltose transporter in Saccharomyces cerevisiae. J. Bacteriol. 182, 241–243 (2000). Nakatsu, F. et al. A di-leucine signal in the ubiquitin moiety: possible involvement in ubiquitin-mediated endocytosis. J. Biol. Chem. 275, 26213–26219 (2000). Roth, A. F. & Davis, N. G. Ubiquitination of the PEST-like endocytosis signal of the yeast a-factor receptor. J. Biol. Chem. 275, 8143–8153 (2000). Springael, J. Y., Galan, J. M., Haguenauer-Tsapis, R. & André, B. NH4+-induced down-regulation of the Saccharomyces cerevisiae Gap1p permease involves its ubiquitination with lysine-63-linked chains. J. Cell Sci. 112, 1375–1383 (1999).

www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS 31. Shih, S. C., Sloper-Mould, K. E. & Hicke, L. Monoubiquitin carries a novel internalization signal that is appended to activated receptors. EMBO J. 19, 187–198 (2000). 32. van Delft, S., Govers, R., Strous, G., Verkleij, A. & Van Bergen en Henegouwen, P. Epidermal growth factor induces ubiquitination of Eps15. J. Biol. Chem. 272, 14013–14016 (1997). 33. Wilde, A. et al. EGF receptor signaling stimulates SRC kinase phosphorylation of clathrin, influencing clathrin redistribution and EGF uptake. Cell 96, 677–687 (1999). 34. Fischer-Vize, J. A., Rubin, G. M. & Lehmann, R. The fat facets gene is required for Drosophila eye and embryo development. Development 116, 985–1000 (1992). 35. Cadavid, A. L., Ginzel, A. & Fischer, J. A. The function of the Drosophila Fat facets deubiquitinating enzyme in limiting photoreceptor cell number is intimately associated with endocytosis. Development 127, 1727–1736 (2000). 36. Garoff, H., Hewson, R. & Opstelten, D. J. E. Virus maturation by budding. Microbiol. Mol. Biol. Rev. 62, 1171–1190 (1998). 37. Strack, B., Calistri, A., Accola, M. A., Palú, G. & Göttlinger, H. G. A role for ubiquitin ligase recruitment in retrovirus release. Proc. Natl Acad. Sci. USA 97, 13063–13068 (2000). This paper shows that ubiquitin is important for virus budding and brings ubiquitin protein ligases into the picture. The proline-rich motifs in viral L domains that mediate interaction with ubiquitin protein ligases are required for Gag ubiquitylation and virus budding. 38. Ott, D. E. et al. Ubiquitin is covalently attached to the p6Gag proteins of human immunodeficiency virus type 1 and simian immunodeficiency virus and to the p12Gag protein of Moloney murine leukaemia virus. J. Virol. 72, 2962–2968 (1998). 39. Patnaik, A., Chau, V. & Wills, J. W. Ubiquitin is part of the retrovirus budding machinery. Proc. Natl Acad. Sci. USA 97, 13069–13074 (2000). Depletion of intracellular ubiquitin levels inhibits virus budding at a late stage and this defect can be partially restored by fusing ubiquitin in-frame to Gag. This provides evidence that ubiquitylation of Gag is required for release of virus particles from the host cell. 40. Schubert, U. et al. Proteasome inhibition interferes with Gag polyprotein processing, release and maturation of human immunodeficiency viruses. Proc. Natl Acad. Sci. USA 97, 13057–13062 (2000). 41. Craven, R. C., Harty, R. N., Paragas, J., Palese, P. & Wills, J. W. Late domain function identified in the vesicular stomatitis virus M protein by use of rhabdovirus-retrovirus chimeras. J. Virol. 73, 3359–3365 (1999). 42. Jayakar, H. R., Murti, K. G. & Whitt, M. A. Mutations in the PPPY motif of vesicular stomatitis virus matrix protein reduce virus budding by inhibiting a late step in virion release. J. Virol. 74, 9818–9827 (2000).

43. Harty, R. N., Brown, M. E., Wang, G., Huibregtse, J. & Hayes, F. P. A PPxY motif within the VP40 protein of Ebola virus interacts physically and functionally with a ubiquitin ligase: implications for filovirus budding. Proc. Natl Acad. Sci. USA 97, 13871–13876 (2000). 44. Puffer, B. A., Watkins, S. C. & Montelaro, R. C. Equine infectious anemia virus Gag polyprotein late domain specifically recruits cellular AP-2 adapter protein complexes during virion assembly. J. Virol. 72, 10218–10221 (1998). 45. Hicke, L. & Riezman, H. Ubiquitination of a yeast plasma membrane receptor signals its ligand-stimulated endocytosis. Cell 84, 277–287 (1996). 46. Galan, J. M., Moreau, V., André, B., Volland, C. & Haguenauer-Tsapis, R. Ubiquitination mediated by the Npi1p/Rsp5p ubiquitin-protein ligase is required for endocytosis of the yeast uracil permease. J. Biol. Chem. 271, 10946–10952 (1996). 47. Cook, W., Jeffrey, L., Kasperek, E. & Pickart, C. Structure of tetraubiquitin shows how multiubiquitin chains can be formed. J. Mol. Biol. 236, 601–609 (1994). 48. Beal, R., Deveraux, Q., Xia, G., Rechsteiner, M. & Pickart, C. Surface hydrophobic residues of multiubiquitin chains essential for proteolytic targeting. Proc. Natl Acad. Sci. USA 93, 861–866 (1996). 49. Sloper-Mould, K. E., Jemc, J., Pickart, C. M. & Hicke, L. Distinct functional surface regions on ubiquitin (submitted). 50. Hochstrasser, M. Evolution and function of ubiquitin-like protein-conjugation systems. Nature Cell Biol. 2, E153–E157 (2000). 51. Schauber, C. et al. Rad23 links DNA repair to the ubiquitin/proteasome pathway. Nature 391, 715–718 (1998). 52. Kleijnen, M. F. et al. The hPLIC proteins may provide a link between the ubiquitination machinery and the proteasome. Mol. Cell 6, 409–419 (2000). 53. Mizushima, N., Noda, T. & Ohsumi, Y. Apg16p is required for the function of the Apg12p –Apg5p conjugate in the yeast autophagy pathway. EMBO J. 18, 3888–3896 (1999). 54. Guarino, L. A., Smith, G. & Dong, W. Ubiquitin is attached to membranes of baculovirus particles by a novel type of phospholipid anchor. Cell 80, 301–309 (1995). 55. Medintz, I., Jiang, H. & Michels, C. A. The role of ubiquitin conjugation in glucose-induced proteolysis of Saccharomyces maltose permease. J. Biol. Chem. 273, 34454–34462 (1998). 56. Dunn, R. & Hicke, L. Domains of the Rsp5 ubiquitin protein ligase required for receptor-mediated and fluid-phase endocytosis. Mol. Biol. Cell 12, 421–435 (2001). 57. Huibregtse, J. M., Yang, J. G. & Beaudenon, S. L. The large subunit of RNA polymerase II is a substrate of the Rsp5 ubiquitin-protein ligase. Proc. Natl Acad. Sci. USA 94, 3656–3661 (1997). 58. Koegl, M. et al. A novel ubiquitination factor, E4, is involved in multiubiquitin chain assembly. Cell 96, 635–644 (1999).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

59. Ortolan, T. G. et al. The DNA repair protein Rad23 is a negative regulator of multi-ubiquitin chain assembly. Nature Cell Biol. 2, 601–608 (2000). 60. Finley, D., Bartel, B. & Varshavsky, A. The tails of ubiquitin precursors are ribosomal proteins whose fusion to ubiquitin facilitates ribosome biogenesis. Nature 338, 394–401 (1989). 61. Wolffe, A. Chromatin Structure and Function (Academic, San Diego, 1998). 62. Strahl, B. D. & Allis, C. D. The language of covalent histone modifications. Nature 403, 41–45 (2000). 63. Ball, E. et al. Arthrin, a myofibrillar protein of insect flight muscle, is an actin –ubiquitin conjugate. Cell 51, 221–228 (1987). 64. Huang, H. et al. The inhibitor of apoptosis, cIAP2, functions as a ubiquitin-protein ligase and promotes in vitro monoubiquitination of caspases 3 and 7. J. Biol. Chem. 275, 26661–26664 (2000). 65. Arnason, T. & Ellison, M. J. Stress resistance in Saccharomyces cerevisiae is strongly correlated with assembly of a novel type of multiubiquitin chain. Mol. Cell. Biol. 14, 7876–7883 (1994). 66. Baboshina, O. V. & Haas, A. L. Novel multiubiquitin chain linkages catalyzed by the conjugating enzymes E2EPF and RAD6 are recognized by 26S proteasome subunit 5. J. Biol. Chem. 271, 2823–2831 (1996). 67. Luger, K., Mader, A. W., Richmond, R. K., Sargent, D. F. & Richmond, T. J. Crystal structure of the nucleosome core particle at 2.8 Å resolution. Nature 389, 251–260 (1997). 68. Joazeiro, C. A. et al. The tyrosine kinase negative regulator c-Cbl as a RING-type, E2-dependent ubiquitin-protein ligase. Science 286, 309–312 (1999). 69. Staub, O. et al. WW domains of Nedd4 bind to the proline-rich PY motifs in the epithelial Na+ channel deleted in Liddle’s syndrome. EMBO J. 15, 2371–2380 (1996). 70. Overton, M. C. & Blumer, K. J. G-protein-coupled receptors function as oligomers in vivo. Curr. Biol. 10, 341–344 (2000). 71. Yesilaltay, A. & Jenness, D. D. Homo-oligomeric complexes of the yeast a-factor pheromone receptor are functional units of endocytosis. Mol. Biol. Cell 11, 2873–2884 (2000). 72. Wilkinson, K. D. & Hochstrasser, M. in Ubiquitin and the Biology of the Cell (eds Peters, J. M., Harris, J. R. & Finley, D.) 99–125 (Plenum, New York and London, 1998).

Acknowledgements I thank R. Lamb, J. Widom, J. Wills and members of my lab for advice and helpful discussions, and R. Lamb and K. Lee for critical comments on the manuscript. J. Wills, H. Göttlinger, U. Schubert and J. Leis generously communicated unpublished results. I apologize to my colleagues whose work or references were not included owing to space restrictions.

VOLUME 2 | MARCH 2001 | 2 0 1

REVIEWS

SUMO, UBIQUITIN’S MYSTERIOUS COUSIN Stefan Müller, Carsten Hoege, George Pyrowolakis and Stefan Jentsch Covalent modification of cellular proteins by the ubiquitin-like modifier SUMO regulates various cellular processes, such as nuclear transport, signal transduction, stress response and cellcycle progression. But, in contrast to ubiquitylation, sumoylation does not tag proteins for degradation, but seems to enhance their stability or modulate their subcellular compartmentalization. U B I Q U I T I N A N D P R OT E A S O M E S ISOPEPTIDE BOND

Any amide bond formed between a carboxyl group of one amino acid and an amino group of another where either group occupies a position other than α. 26S PROTEASOME

Large multisubunit protease complex that selectively degrades multi-ubiquitylated proteins. It contains a 20S particle that carries the catalytic activity and two regulatory 19S particles.

Max Planck Institute of Biochemistry, Department of Molecular Cell Biology, Am Klopferspitz 18a, 82152 Martinsried, Germany. Correspondence to S.M. e-mail: stmuelle@biochem.mpg.de

202

Post-translational modification of proteins is an important means to alter their function, activity or localization after their synthesis has been completed. In these cases, specific amino-acid residues of target proteins are chemically modified by various molecules, such as phosphate, acetate, lipids or sugars. Modification with ubiquitin represents a unique case because the modifier itself is a small polypeptide. Ubiquitin is usually attached to lysine side chains of target proteins, resulting in ‘branched’, ISOPEPTIDElinked ubiquitin–protein conjugates (see review by Allan Weissman on page 169 of this issue). Conjugation of ubiquitin (termed ubiquitylation or ubiquitination) has a well-established role in earmarking proteins for degradation by the 26S 1 PROTEASOME (see review by Peter Kloetzel on page 179 in this issue). During the past few years, several proteins have been discovered that have sequence similarity to ubiquitin. These ubiquitin-like proteins fall into two separate classes. Proteins of the first class, termed ‘ubiquitin-like modifiers’ (UBLs) function as modifiers in a manner analogous to that of ubiquitin. Examples are SUMO (small ubiquitin-related modifier), Rub1 (also called Nedd8), Apg8 and Apg12 (see review by Yoshinori Ohsumi on page 211 of this issue). Proteins of the second class, including parkin, RAD23 and DSK2, are designated ‘ubiquitin-domain proteins’ (UDPs). These proteins bear domains that are related in sequence to ubiquitin but are otherwise unrelated to each other. In contrast to UBLs, these proteins are not conjugated to other proteins2,3.

The UBL protein SUMO (also called sentrin) is present in all eukaryotic kingdoms4 and is highly conserved from yeast to humans. Whereas invertebrates have only a single SUMO gene, which has also been termed SMT3, three members of the SUMO family have been described in vertebrates: SUMO-1 and the close homologues SUMO-2 and SUMO-3 (REF. 5). SUMO-1, the founding member of the family, is also known as PIC-1, sentrin or GMP1 (REFS 6–8). For simplicity, we use the term SUMO here for all members of the SUMO family, including those from yeast and flies. Human SUMO-1, a 101-amino-acid polypeptide, shares ~50% sequence identity with the closely related SUMO-2/SUMO-3 and with the yeast Saccharomyces cerevisiae Smt3 protein (FIG. 1). Although the overall sequence identity between SUMO-1 and ubiquitin is only about 18%, structure determination by nuclear magnetic resonance (NMR) revealed that both share a common three-dimensional structure that is characterized by a tightly packed globular fold with β-sheets wrapped around one α-helix9. SUMO has a short amino-terminal extension that is absent in ubiquitin. The function of this extension is unknown, but interestingly, the sequences of these extensions are different for the distinct SUMO members (FIG. 1). After the initial observation that SUMO-1 can be attached to the mammalian RanGAP1 protein8,10 (see below), work in several laboratories has established that covalent modification of substrate proteins with SUMO represents a novel and widely used type of post-translational modification. But the functional significance of this modification remains a matter of debate. www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS

SUMO-1 SUMO-2 SUMO-3 SMT3 Ubiquitin

MSD----QEAKPSTEDLGDKKEGEYIKLKVIGQDSSEIHFKVKMTTHLKKLKESYC MAD----EKPKEGVKTENN----DHINLKVAGQDGSVVQFKIKRHTPLSKLMKAYC MSE----EKPKEGVKTEN-----DHINLKVAGQDGSVVQFKIKRHTSLSKLMKAYC MSDSEVNQEAKPEVKPEVKP--ETHINLKVS-DGSSEIFFKIKKTTPLRRLMEAFA MSD----QEAKPSTEDLGDKKEGEYMQIFVKTLTGKTITLEVEPSDTIENVKAKIQ

SUMO-1 SUMO-2 SUMO-3 SMT3 Ubiquitin

QRQGVPMNSLRFLFEGQRIADNHTPKELGMEEEDVIEVYQEQTGGHSTV ERQGLSMRQIRFRFDGQPINETDTPAQLEMEDEDTIDVFQQQTGGVY ERQGLSMRQIRFRFDGQPINETDTPAQLRMEDEDTIDVFQQQTGGVPESSLAGHSF KRQGKEMDSLRFLYDGIRIQADQTPEDLDMEDNDIIEAHREQIGGATY DKEGIPPDQQRLIFAGKQLEDGRTLSDYNIQKESTLHLVLRLRGG

Figure 1 | Sequence alignment of SUMO family members and ubiquitin. The sequences of human SUMO-1, -2, and -3 are compared to SMT3 from Saccharomyces cerevisiae and to human ubiquitin. Identities are shown in blue, similarities in yellow. Precursor processing occurs carboxy-terminal to the double-glycine motif (scissors symbol).

CYSTEINE PROTEASE

Protease that has a cysteine at the active site.

Pathway of SUMO conjugation

The pathway of sumoylation is mechanistically analogous to ubiquitylation, but SUMO conjugation requires a set of enzymes that is distinct from that acting on ubiquitin (FIG. 2). Ubiquitin is synthesized as an inactive precursor protein, which has to be processed by a specific protease (called ubiquitin carboxy-terminal hydrolase) to make the carboxy-terminal double-glycine motif available for conjugation. Subsequently, ubiquitylation proceeds through a three-step mechanism. Ubiquitin is initially activated by the ubiquitin-activating enzyme E1 and then transferred to a reactive cysteine residue of one of several E2s (ubiquitin-conjugating enzyme or UBC). In most cases at least one additional factor, called E3 or ubiquitin–protein ligase, is required for the formation of an isopeptide bond between ubiquitin and the target protein. E3s are considered to be largely responsible for substrate specificity. Usually, several ubiquitin molecules are conjugated to a substrate in the form of a so-called ‘multi-ubiquitin’ chain and the resulting multi-ubiquitylated proteins are the preferred substrates of the 26S proteasome. Multi-ubiquitylation sometimes requires another factor termed E4 (REF. 11). Like ubiquitin, all SUMO forms are initially made as inactive precursors. They mature by a carboxy-terminal proteolytic cleavage event, which yields the mature modifiers with exposed carboxy-terminal glycine residues. These residues are required for the formation of SUMO–protein conjugates, that is, the formation of an isopeptide bond between the carboxyl terminus of

Processing Modifier

Ubiquitin

ATP

ATP SUMO

SUMO de-conjugation

Conjugation

Carboxy-terminal Mature hydrolase

Activating enzyme E1

Conjugating enzyme E2

UBA1

UBCs

AOS1/ UBA2

SUMO with an ε-amino group of a lysine residue of a target protein. The processing reaction is catalysed by a group of CYSTEINE PROTEASES, termed ULPs (ubiquitin-like protein-processing enzyme) or SUMO-specific proteases (see below). A SUMO-specific E1 activity has been purified and characterized in yeast and humans12–14. The protein is a heterodimer composed of the proteins AOS1 (SUA1) and UBA2 (SUA2). Remarkably, UBA2 bears clear sequence similarity to the carboxy-terminal region of UBA1, the E1 enzyme for ubiquitin, whereas AOS1 is related to the amino-terminal part of UBA1. The UBA2 subunit bears the ‘active site’ cysteine residue required for the formation of SUMO–E1 enzyme thioesters, but both subunits are required for SUMO activation in vitro and in vivo. SUMO has a single E2-type-conjugating enzyme, UBC9, which is specific for SUMO and does not act on ubiquitin15–18. A structural comparison of UBC9 with ubiquitin-specific E2 enzymes revealed, in spite of an overall similarity, important differences between these enzymes19,20. In particular, the surface of UBC9, which is involved in SUMO binding, is mainly positively charged, whereas the corresponding regions in ubiquitin-specific E2s (for example, UBC4 and UBC7) have negative or neutral potentials. These differences are likely to have a role in modifier discrimination. Notably, the positively charged binding surface of UBC9 is highly complementary in its electrostatic potentials and hydrophobicity to the negatively charged surface of SUMO. Ubiquitin cannot bind to UBC9 because it has positive charges in this region. UBC9 has been shown to physically interact with almost all known SUMO substrates in yeast two-hybrid assays, possibly indicating that UBC9 can be sufficient for substrate recognition. It seems more attractive to speculate, however, that, in analogy to the ubiquitylation process, substrate specificity of the SUMO-conjugation pathway might require additional, as-yet-unidentified E3-type ligases. In contrast to ubiquitin, SUMO conjugation does not seem to lead to the formation of SUMO–SUMO chains. This is consistent with the observation that the typical branched-point lysine residues (K29, K48, K63) of ubiquitin are not present in SUMO. Moreover, no mixed modifier chains that contain both SUMO and ubiquitin or other UBLs have been reported.

Ligating enzyme E3

Substrate

S Many

UBC9

Figure 2 | Conjugation pathway of ubiquitin and the ubiquitin-like modifier SUMO. Ubiquitin and SUMO are synthesized as precursors and processed carboxy-terminally by hydrolases (vertical arrows) and are subsequently conjugated to proteins involving activating (E1) and conjugating (E2) enzymes that form thioesters (S) with the modifiers.

Sumoylation is a dynamic, reversible process. The cleavage of SUMO from its target proteins, here termed ‘desumoylation’, is catalysed by ULP proteases. In yeast, two of these enzymes — Ulp1 and Ulp2 — have been identified21–23. In vitro, both Ulp1 and Ulp2 can catalyse the carboxy-terminal processing of SUMO and both enzymes can remove SUMO from isopeptide-linked conjugates. The sequence similarity of the two enzymes is restricted to a 200-amino-acid sequence called the ULP domain, which harbours the catalytically active region. The three-dimensional structure of the ULP domain from Ulp1 has been determined in a complex with the S. cerevisiae SUMO (Smt3) precursor24 (FIG. 3). Interestingly, Ulp1 shares no sequence or structural similarity to de-ubiquitylating enzymes, although both belong to the cysteine protease superfamily.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 2 0 3

REVIEWS

from the nuclear substrate PML (promyelocytic leukaemia)27. However, other factors are likely to also be important, as the cytosolic SUSP-1 enzyme does not remove SUMO from RanGAP1 (REF. 28). Further studies of ULP-family members are expected to show that SUMO de-conjugation has important regulatory roles, especially in more complex organisms.

Main functions of SUMO

Figure 3 | Structure of the ULP-1–SMT3 complex. ULP-1 is shown in blue, SMT3 in purple. β-strands are numbered; α-helices are lettered. a | View looking onto one side of the complex. b | A perpendicular view of the complex looking onto the active site. SUMO does not undergo large conformational changes when bound to ULP1, but its carboxyl terminus adopts a more extended conformation in the complex that ends with a covalent bond between the terminal glycine residue and the catalytic cysteine residue (Cys580) of ULP1. The active site of ULP1 resides in a narrow cleft comprised by conserved amino acids that recognize the carboxyterminal Gly-Gly-X motif of the SUMO precursor by van der Waals contacts. The structure of the active-site cleft of ULP1 allows the access of even large SUMO-protein conjugates. This is in striking contrast to de-ubiquitylating enzymes, where a bulky loop at a similar region allows only the access of small or denatured ubiquitin conjugates.

SUMO CONJUGATE PATTERN

Pattern of bands corresponding to sumoylated substrates detectable on an immunoblot with an anti-SUMO antibody. EST

DNA sequence obtained by sequencing an end of a random complementary DNA clone from a cDNA library. NUCLEAR PORE COMPLEX

Large multiprotein complex that forms a channel in the nuclear envelope of an eukaryotic cell, joining the inner and outer nuclear membranes and allowing transport of proteins to and from the nucleus. BICOID

A segment polarity protein, discovered in Drosophila, that provides positional cues for the development of head and thoracic segments. PML NUCLEAR BODIES

One type of nuclear speckles of unknown function that contains several proteins, including the promyelocytic leukaemia protein PML. PML nuclear bodies are also called PODs (PML oncogenic domains) or ND10 (nuclear dots 10).

204

Like SUMO conjugation, SUMO de-conjugation is needed for viability in the budding yeast21–23. In this organism, only Ulp1, but not Ulp2, is essential for viability. SUMO conjugates accumulate in cells lacking either ULP1 or ULP2. Interestingly, the SUMO CONJUGATE PATTERN is distinct for each of the individual mutants, indicating that the two enzymes act on distinct substrates. Consistent with that, ULP2 overexpression cannot suppress the lethality of the ULP1 deletion. Surprisingly, however, inactivation of ULP2 partially rescues the defects caused by a ULP1 deficiency, and the double mutant accumulates fewer SUMO conjugates than either of the single mutants. An expressed sequence tag (EST) databank search, based on the conserved ULP domain from yeast Ulp1, identified a family of mammalian ULP homologues. In humans at least 7 ULPs, with sizes ranging from 238 to 1,112 amino acids, were identified25. The enzymes were termed SENPs or SUSPs (for sentrin/SUMO-specific proteases). It is not yet known whether SUSPs have a dual role as maturation enzymes for SUMO and as isopeptidases for substrate de-sumoylation. With respect to the large number of mammalian SUSPs, it is reasonable to speculate that distinct enzymes might have specialized functions. Alternatively, the diversification in the mammalian SUMO de-conjugation system might be explained in part by the fact that higher eukaryotes express distinct SUMO forms (that is, SUMO-1, -2, 3). Indeed, a recently identified SUSP enzyme from mouse, Smt3-specific isopeptidase 1 (SMT3IP1), has a higher in vitro cleavage activity towards SUMO-2 conjugates than towards SUMO-1 conjugates26. What determines the substrate specificity of ULPs in vivo is not known, but studies in mammalian cells indicate that ULPs have different intracellular localization26–28. For example, the mammalian enzyme SENP-1 that localizes to the nucleus does not act on cytoplasmic SUMO–RanGAP1 conjugates but it removes SUMO

SUMO has much fewer cellular substrates than ubiquitin but, intriguingly, several identified targets turned out to be important cellular regulators (TABLE 1; ONLINE FIG 1). But how the SUMO modification influences the function of the target proteins is not clear. In the following section, we focus on the main findings and describe two functional models for SUMO that are now being discussed in the field. An address tag for protein targeting? The first substrate identified as a target for SUMO-1 was the Ran GTPaseactivating protein RanGAP1, a component of the nuclear import machinery8,10. RanGAP1 is a key regulator of the Ras-like GTPase Ran, which controls nucleo-cytoplasmic transport29. Intriguingly, only the SUMO-1–RanGAP1 conjugate can stably interact with Ran-binding protein 2 (RanBP2), a protein located at the cytoplasmic face of the NUCLEAR PORE COMPLEX. This finding might indicate that sumoylation either targets RanGAP1 to the nuclear pore complex or, alternatively, stabilizes RanGAP1–RanBP2 complexes. The regions in RanGAP1 that determine sumoylation and RanBP2 binding, are non-overlapping. This indicates that SUMO does not directly mediate RanBP2 binding but rather induces a structural change in RanGAP1 that allows its binding to RanBP2 (REFS 30–32). Consistent with this, the free SUMO modifier does not directly interact with RanBP2. The yeast homologue of RanGAP1, Rna1, lacks the carboxy-terminal tail region of mammalian RanGAP1 that harbours the sumoylation site and does not seem to be sumoylated18. However, the ULP1 protein has been shown to interact with components of the yeast nuclear pore complex22,33,34, indicating that sumoylation in yeast might also have some function in certain nuclear transport pathways. Indirect evidence for a more general role of SUMO in nuclear import processes also comes from studies with a Drosophila melanogaster mutant that harbours a loss-of-function mutation in the UBC9 gene (semushi)35. In these mutants, nuclear import of the transcription factor BICOID is prevented, leading to defects in embryogenesis. It remains to be determined whether this is directly mediated by impaired sumoylation of the Drosophila RanGAP1 protein. Whereas RanGAP1 seems to be the main cytosolic substrate of SUMO, most sumoylated proteins are nuclear36. Most intriguingly, sumoylated substrates are often found in specific subnuclear protein complexes and preferential accumulation sites for sumoylated proteins are the so-called PML NUCLEAR BODIES (BOX 1). PML, the defining component of PML nuclear bodies, is a member of the family of RING-FINGER PROTEINS. It shows certain cell growth and tumour-suppressive properties and has www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS

Table 1 | Known substrates for SUMO Protein

Function

Role of sumoylation

References

RanGAP1 PML

Nuclear import Tumour supressor

Sp100 p53 p73 HIPK2 TEL c-Jun Androgen receptor IκBα

Chromatin remodelling (?) Tumour supressor p53 homologue Transcriptional corepression Transcriptional repression Transcriptional activation Transcriptional activation

Mediates interaction with RanBP2 Allows formation of NBs and recruitment of Daxx/p53 to NBs Mediates interaction with HP1 (?) Activates p53 transactivation and apoptosis Unknown Mediates the localization of HIPK2 to nuclear dots Mediates the localization of TEL to nuclear dots Slightly reduces transcriptional activity of c-Jun Reduces transcriptional activity of androgen receptor

Mammalian

Mdm2 Topo I

Signal transduction, NF-κB inhibition E3 ubiquitin ligase for p53

GLUT1

DNA replication, DNA repair DNA replication, DNA repair DNA helicase (RecQ family) Component of nuclear pore complex Glucose transport

GLUT4

Glucose transport

Topo II WRN RanBP2

Drosophila Ttk 69 Dorsal CaMK Yeast Septins Viral CMV IE1 CMV IE2 EBV BZLF1

Inhibits ubiquitylation of IκBα Blocks NF-κB activity Inhibits ubiquitylation of Mdm2 Acitvates the E3 function of Mdm2 Unknown, induced after DNA damage with camptothecin Unknown, induced after DNA damage with teniposide Unknown Unknown Unknown, GLUT1 protein levels are downregulated by UBC9 Unknown, GLUT4 protein levels are upregulated by UBC9

Transcriptional repression Signal transduction Calcium/calmodulindependent kinase

Unknown Activates nuclear import of Dorsal Unknown

Bud-neck formation

Regulates dynamics of the neck ring

Immediate-early viral regulator Immediate-early viral regulator Immediate-early viral regulator PML sumoylation HPV/BPV E1 DNA helicase Initiates viral replication

40,58 52–54 55 60 61,62 54 63 65 66 83 84 85 98 99 99

64 68 100

74,75

Unknown, correlates with the loss of PML sumoylation Decreases transactivation potential of IE2 Unknown, correlates with the loss of Regulates nuclear import of E1

8,10,30 –32 39–50

45 93 94 101

CMV, cytomegalovirus; EBV, Epstein–Barr virus; HPV/BPV, human papillomavirus/bovine papillomavirus; RanGAP1, Ran GTPase activating protein; IE, immediate-early; PML, promyelocytic leukaemia; RanBP2, Ran-binding protein 2; Topo, topoisomerase; Ttk, tramtrack.

RING-FINGER PROTEINS

A family of proteins structurally defined by the presence of the zinc-binding RING-finger motif. The RING consensus sequence is: CX2CX(9–39)CX(1–3)HX(2–3) C/HX2CX(4–48)CX2C. The cysteines and histidines represent metal binding sites. The first, second, fifth and sixth of these bind one zinc ion and the third, fourth, seventh and eighth bind the second.

a pro-apoptotic function. These activities are at least partially mediated by the potential of PML to act as a transcriptional co-activator in conjunction with the p53 tumour suppressor protein37,38. A fraction of PML undergoes sumoylation at three distinct lysine residues in the protein, and there is accumulating evidence that the sumoylation of PML regulates the assembly and/or stability of nuclear bodies39–43. Thus, the nuclear-bodycontaining detergent-insoluble cell fraction is highly enriched for SUMO–PML conjugates39. Disassembly of nuclear bodies by viral proteins (BOX 2) or during mitosis correlates with a complete loss of PML sumoylation44–46. Intriguingly, in PML-deficient mutant cell lines (PML–/–) that cannot form nuclear bodies properly, reassembly of these structures can be induced by transfection with

wild-type PML but not with a mutant version that lacks the sumoylation sites47. Although the function of PML nuclear bodies remains enigmatic, a number of examples show how effectively sumoylation of PML and subsequent recruitment of certain proteins to nuclear bodies can modulate transcriptional activity. Upon sumoylation of PML, the transcriptional corepressor Daxx relocalizes to nuclear bodies, where it seems to be stored in an inactive state48–50 (FIG. 4). Sumoylation of PML also directs p53 to nulcear bodies but this leads to a stimulation of the transcriptional and pro-apoptotic activity of p53 rather than an inhibition38. It has been suggested that recruitment of p53 to nuclear bodies could trigger activating modifications in p53, such as acetylation51. Furthermore, evidence

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 2 0 5

REVIEWS

Box 1 | PML nuclear bodies PML (promyelocytic leukaemia) nuclear bodies (also known as ND10 or PODs) are distinct subnuclear structures of higher eukaryotes, which appear as dense spherical particles, 0.3–0.5 µm in diameter, that are tightly associated with the nuclear matrix91,92. Normal cells have between 10–30 PML nuclear bodies per nucleus. They all contain the nuclear body core components PML and Sp100 and a number of additional proteins that transiently associate with nuclear bodies. Among these are the transcriptional repressor Daxx, the tumour suppressors Rb and p53 and the Bloom syndrome protein (BLM). The assembly of nuclear bodies is highly dynamic and sensitive to environmental stimuli, such as response to stress or interferon. Furthermore, the integrity of PML nuclear bodies is compromised in certain pathological situations such as acute promyelocytic leukaemia and upon infection by certain DNA viruses. Although nuclear bodies have been implicated in a number of cellular processes, including the regulation of programmed cell death, transcription or antigen presentation, their main biological function remains largely enigmatic. Many sumolyated proteins are found in PML nuclear bodies, indicating that SUMO directs proteins to nuclear bodies or, alternatively, that nuclear bodies might support sumoylation.

HP1 FAMILY

(Heterochromatin protein 1 family). A family of chromosomal non-histone proteins primarily associated with heterochromatin. HP1 proteins have been implicated in gene regulation, DNA replication and nuclear architecture.

is accumulating that nuclear bodies might stimulate SUMO conjugation, and that proteins that transiently associate with nuclear bodies include SUMO targets. Indeed, previous work has shown that p53 itself is a target for sumoylation52–54. The sumoylation sites in p53 are located within the carboxy-terminal region, which is known to regulate the DNA-binding activity of the protein. As a consequence, sumoylation of p53 moderately stimulates the transcriptional and apoptotic activity of the protein. It is attractive to speculate that the two findings are linked, that is, that the association of p53 with nuclear bodies is a prerequisite for p53 sumoylation. The significance of p53 sumoylation is underscored by the finding that the p53-related protein p73 is also covalently modified by SUMO at its extreme carboxy-terminal lysine residue55. Sumoylation of p73 does not notably alter its transcriptional properties but rather seems to be implicated in regulating its subcellular localization. Similarly to what is observed for PML, the SUMO–p73 conjugate is preferentially recovered in detergent-insoluble nuclear complexes. Whether these complexes represent PML nuclear bodies has not yet been determined. A further link between the SUMO-modification system and nuclear bodies is suggested by the finding that the Sp100 protein, another core component of PML

nuclear bodies, is also a major cellular substrate for sumoylation40. The precise function of Sp100 is unknown, but its interaction with chromosomal nonhistone proteins of the HP1 and HMG1/2 families points to a role in chromatin organization56,57. A Sp100 mutant protein that lacks the SUMO attachment site can still localize to nuclear bodies, indicating that sumoylation of Sp100 is not required for targeting to nuclear bodies58. However, in vitro, sumoylated Sp100 has a higher affinity for the HP1 protein, indicating that sumoylation of Sp100 might directly mediate this interaction59. In light of the recent findings on PML, it will be interesting to see whether sumoylation of Sp100 recruits HP1 to the nuclear bodies and modulates its repressive function. More recent data on the transcriptional regulators HIPK2 and TEL provide further evidence for a role of SUMO in regulating the subnuclear localization and transcriptional activity of proteins. HIPK2 is a serine/threonine kinase that physically interacts with HOMEODOMAIN TRANSCRIPTION FACTORS and acts as a transcriptional corepressor. Similarly to what is observed for PML, SUMO-1-conjugated HIPK2 forms are found in detergent-insoluble subnuclear complexes. The nature of these nuclear dots has not yet been determined, but the localization of HIPK2 to these dots depends on its sumoylation60. Similarly, TEL, an ETS-related transcriptional repressor, localizes to S-phase-specific nuclear dots in a SUMO-dependent manner, providing further evidence of a role for SUMO in regulating subnuclear localization61. Whether the modulation of nuclear localization by SUMO alters the repressive function of HIPK2 or TEL is unclear. However, overexpression of UBC9 relieves TEL-mediated repression, arguing for a modulation of TEL activity through sumoylation62. The regulation of transcription factors through sumoylation seems to be a recurring theme. Thus, in mammals the proto-oncogene c-Jun and the androgen receptor are sumoylated, and disruption of the SUMO-1 acceptor sites in these proteins enhances their transcriptional activity, indicating that sumoylation negatively regulates their transactivation potential54,63. In Drosophila, the transcriptional repressor Tramtrack 69 protein (Ttk69), which inhibits neuronal differentiation, was identified

HMG1/2 FAMILY

(High-mobility group 1/2 ). Large protein family of small non-histone components of chromatin that function in higher-order chromatin structure. HOMEODOMAIN TRANSCRIPTION FACTORS

Transcription factors with a 60-amino-acid DNA-binding domain comprised of three α-helices. ETS

Proto-oncogene family related to v-ets, one of the oncogenes of the acutely transforming avian erythroblastosis virus E26.

206

Box 2 | Viral proteins — regulators of sumoylation? The immediate-early proteins IE1 and IE2 from human cytomegalovirus and the BZLF1 (Z or Zebra) protein from Epstein–Barr virus are modified by SUMO45,93,94. Both IE1 and IE2 colocalize transiently in PML nuclear bodies at very early time points after infection. Subsequently, IE1 triggers the disaggregation of nuclear bodies, leading to a complete redistribution of nuclear body components into a diffuse nuclear pattern. Intriguingly, SUMO conjugation to IE1 is paralleled by the abrogation of PML sumoylation and thus correlates with the disassembly of nuclear bodies45. In striking analogy, sumoylation of BZLF1 is accompanied by the disappearance or desumoylation of SUMO–PML conjugates and the disruption of PML bodies94. Because PML sumoylation seems to regulate p53 activity, the virus-induced destruction of nuclear bodies and desumoylation of PML could be a viral strategy to overcome the negative effect of p53 on cell proliferation and apoptosis. Indeed, expression of BZLF1 can inactivate p53 function95. Intriguingly, viral and bacterial cysteine proteases have been identified that have sequence similarity to ubiquitinlike protein-processing enzymes (ULPs)21,96. One of these proteins, the Yersinia protein YopJ, seems indeed to interfere with the cellular SUMO-conjugation system97. So pathogens might have evolved a strategy to alter the sumoylation activity of the host for their own benefits.

www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS

Nuclear body Sumoylation PML

PML

PML Daxx

Daxx

Off

Repression of Daxx-regulated genes

On Relief of repression

Figure 4 | Sumoylation of PML modulates Daxxmediated transcriptional repression. Sumoylation of PML recruits Daxx to nuclear bodies and removes it from its target genes, thereby relieving Daxx-mediated repression of these genes.

as one of the most important substrates for SUMO conjugation64. SUMO and Ttk69 proteins localize together on POLYTENE CHROMOSOMES, indicating that the SUMOconjugated Ttk69 species is present at sites of Ttk69 action in vivo. It is unknown how SUMO regulates transcriptional processes mechanistically. However, with respect to the role of SUMO in complex formation, it is tempting to speculate that sumoylation can modulate the interaction of transcription factors with transcriptional coregulators.

POLYTENE CHROMOSOME

A giant chromosome formed by many replications of the DNA. The replicated DNA molecules tightly align side-by-side in parallel register, creating a nonmitotic chromosome that is visible by light microscopy. IκBα

Inhibitory subunit of the NF-κB transcription factor, which is phosphorylated, ubiquitylated and degraded in response to activating stimuli. MITOTIC SPINDLE

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

Highly conserved protein family first identified in yeast and more recently found in a wide range of animal cells. They are thought to function primarily in the control of cytokinesis, where they form a 10-nm filamentous ring that encircles the yeast bud neck.

...or an inhibitor of ubiquitylation? Studies on the NFκB inhibitor IκBα and the ubiquitin-ligase Mdm2 have uncovered an interesting functional link between the ubiquitylation and sumoylation systems65,66. The NF-κB transcription factor is kept inactive in the cytosol by binding to its inhibitor IκBα. Upon stimulation with effectors such as tumour necrosis factor (TNF), IκBα is phosphorylated, ubiquitylated and subsequently degraded by proteasomes67. This liberates NF-κB from its inhibitor, thereby allowing the transcription factor to enter the nucleus and activate its target genes. Intriguingly, SUMO can compete with ubiquitin on IκBα, as both modifiers target the same lysine residue in IκBα. The SUMO-modified pool of IκBα is protected from TNF-induced degradation, and sumoylation of IκBα inhibits NF-κB function. The NF-κB/IκBα system is an evolutionarily conserved pathway; its counterpart in Drosophila is the Dorsal/Cactus system, in which Dorsal is the vertebrate homologue of NF-κB and Cactus corresponds to IκBα. Surprisingly, however, it is Dorsal — not Cactus — that undergoes sumoylation68. In this case, sumoylation facilitates nuclear import of Dorsal and thus, in contrast to the mammalian system, activates its function. Hence, it seems that SUMO can regulate similar pathways in different organisms, but the target proteins might not necessarily be the same. Recent data on Mdm2 provide evidence that SUMO might have a more general role in protein stabilization. Mdm2 is a RING-finger protein that serves as an E3type ubiquitin ligase for p53 and for itself 69,70. Interestingly, the stability of Mdm2 itself is also regulated by the ubiquitin–proteasome system and it has been

proposed that SUMO interferes with the ubiquitylation of Mdm2 (REF. 66). Similarly to what was observed for IκBα, SUMO and ubiquitin were reported to act on the same lysine residue within the RING finger of Mdm2. Hence, sumoylation of Mdm2 would prevent ubiquitylation and subsequent degradation. In normal cells, most Mdm2 has mobility corresponding to a protein of 90 kDa, which could represent a SUMO–Mdm2 conjugate. Upon DNA damage, however, Mdm2 is presumed to be de-sumoylated and subsequently ubiquitylated and degraded. As a result of decreasing cellular Mdm2 levels, p53 is stabilized. Although this model seems appealing, a note of caution has been raised recently71. In particular, it is unclear why bacterially expressed, unsumoylated Mdm2 comigrates in gels (90 kDa) with the presumed SUMO–Mdm2 conjugate. Furthermore, the 90 kDa Mdm2 species is insensitive to ULP-mediated de-sumoylation in vitro (Müller, S. and Dejean, A., unpublished observations). Mass spectrometric analysis of the 90-kDa species is expected to settle these issues and to define the nature of this species. What does genetics tell us?

In the budding yeast S. cerevisiae, sumoylation is essential for viability. By contrast, SUMO from the yeast Schizosaccharomyces pombe is not essential for viability, but mutants defective in this pathway show strong growth defects72. Interestingly, temperature-sensitive mutants in the S. cerevisiae genes for the SUMO-activating enzyme (uba2-ts), SUMO-conjugating enzyme (ubc9-ts) and the de-sumoylation enzyme Ulp1 (ulp1ts) show strong cell-cycle defects12,21,73. They predominately arrest at the G2/M boundary of the cell division cycle as large-budded cells with replicated DNA and a short MITOTIC SPINDLE indicating that sumoylation of one or more targets in yeast is essential for cell-cycle progression at this phase. Mutants deficient in Ulp2 are viable but they show various phenotypic abnormalities, such as aberrant cell-cycle progression, hypersensitivity to DNA damage and chromosome mis-segregation22. ULP2 messenger RNA levels are upregulated during early sporulation, pointing to a role for de-conjugation of certain substrates during yeast meiosis. The main targets of SUMO in S. cerevisiae are the SEPTINS Cdc3, Cdc11 and Sep7, which form a 10-nm filamentous ring that encircles the yeast bud neck74,75. Proper assembly of the septin ring is monitored by an unknown mechanism at the morphogenesis checkpoint, which acts at the G2/M phase boundary of the cell cycle. Sumoylation of the septins occurs during mitosis before anaphase and the modifications disappear abruptly at cytokinesis. Disappearance of this neck ring is disturbed in a yeast mutant strain in which all the SUMO-conjugation sites of the septins are changed to arginine residues, arguing that the dynamics of the neck ring require sumoylation. However, the G2/M arrest of mutants with deficiencies in the SUMO conjugation pathway (uba2-ts, ubc9-ts) cannot be explained by the absence of sumoylated septins, because mutants that express septins lacking the SUMO conjugation sites show no growth defects.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 2 0 7

REVIEWS

HIGH-COPY SUPPRESSOR

Gene that suppresses a phenotype when expressed at high copy number. CENTROMERE

Region of a chromosome that is attached to the spindle during nuclear division. MINICHROMOSOME

An extrachromosomal plasmid DNA that contains a chromosomal origin of replication. SYNAPTONEMAL COMPLEX

Structure that holds paired chromosomes together during prophase I of meiosis and that promotes genetic recombination. WERNER SYNDROME

A rare autosomal recessive disorder, characterized by the early development of various age-related diseases. The gene responsible for Werner syndrome (WRN) encodes a DNA helicase homologous to Escherichia coli RecQ. BLOOM SYNDROME

A rare cancer-predisposing autosomal recessive disorder characterized by genomic instability, immunodeficiency and small stature. BLM, the gene mutated in Bloom syndrome, encodes a DNA helicase of the RecQ family.

A first hint of SUMO’s importance for the maintenance of genomic integrity came from genetic studies in yeast. Yeast SUMO was originally identified as a HIGHCOPY SUPPRESSOR of mutations in Mif2 (REF. 76). The MIF2 protein is part of a centromeric multiprotein complex and is required for proper segregation of chromosomes and for integrity of the mitotic spindle77. The interaction of Ubc9 with CENTROMERE proteins from S. cerevisiae supports the idea that sumoylation could regulate centromeric proteins. In the yeast S. pombe, deletion of the gene for either SUMO (pmt3+) or Ubc9 (hus5+) leads to aberrant mitosis, increased sensitivity to DNA-damaging agents and high-frequency loss of MINICHROMOSOMES, indicating that Pmt3 is also involved in processes such as chromosome segregation and DNA-damage repair78. However, a molecular understanding of these phenotypes requires the identification of the SUMO substrates. A role for SUMO in DNA-damage repair and recombination is further supported by the finding that the human homologues of yeast Rad51 and Rad52, physically interact with both Ubc9 and SUMO in yeast two-hybrid assays79,80. In addition, Ubc9 localizes together with Rad51 in SYNAPTONEMAL COMPLEXES during meiosis81. Whether Rad51 and Rad52 are modified by SUMO is, at present, not known. The Rad51/Rad52 pathway is involved in DNA recombination and the repair of DNA double-stranded breaks. Intriguingly, overexpression of SUMO downregulates homologous recombination after DNA double-stranded breaks and reduces cellular resistance to ionizing radiation82. Additional known SUMO substrates with links to DNA repair are human DNA topoisomerase I and II, which both become sumoylated upon DNA damage83,84 and the WERNER SYNDROME gene product, a DNA helicase of the RecQ family85. Furthermore, another member of the RecQ helicase family, the BLOOM SYNDROME protein (BLM), interacts with Ubc9 (REF. 85). Interestingly, some of the proteins mentioned above, such as Rad51 and Rad52, BLM and CENP-C (the human Mif2 homologue), are transiently associated with PML nuclear bodies48,86–89. However, conclusive data on whether PML nuclear bodies are implicated in the maintenance of genomic stability and repair are missing. From substrates to function

Studies on SUMO have entered an exciting phase. An astonishing number of important cellular regulators and functions are now known to be controlled by sumoylation. It seems that the reversible nature of this modification is a crucial factor in these processes. The findings described above implicate SUMO in the stabilization of proteins and/or their localization to subcellular complexes. Whether SUMO targets the proteins to these complexes or whether conjugation occurs within these complexes is a key issue that has to be settled. The available data on the most intensely studied SUMO substrates, PML and RanGAP1, suggest that their sumoylation promotes specific protein–protein interactions or stabilizes preformed protein complexes. The reversible nature of the SUMO modification might be well suited to dynamically regulate these protein

208

Sumoylation SUMO AOS1/ UBA2, UBC9

Degradation

Stabilization

ULPs

Complex formation

SUMO De-sumoylation

Figure 5 | Reversible modification of proteins by SUMO and its consequences. Sumoylation either prevents ubiquitylation followed by degradation or results in the formation of protein complexes. SUMO is depicted in green, ubiquitin in red. Sumoylation requires AOS1/UBA2 and UBC9 enzymes; de-sumoylation is catalysed by members of the ULP family.

complexes. A different aspect of sumoylation became evident from the findings on IκBα and Mdm2, in which SUMO seems to counteract the function of ubiquitin, thereby blocking ubiquitin/proteasomemediated proteolysis. Although theses findings are intriguing, SUMO clearly does not always function in this way. For example, the sumoylation and ubiquitylation sites within p53 seem to be different and sumoylation does not directly interfere with the ubiquitylation of p53 (REFS 53, 54). It seems possible that SUMO functions in different ways depending on its substrate. Could both suggested functions of SUMO be mechanistically linked (FIG. 5)? An attractive hypothesis is that proteins within large assemblies that are induced by sumoylation are more resistant to ubiquitin/proteasome-dependent degradation than their respective free counterparts. Consistent with this model is the observation that the abrogation of PML sumoylation by the Herpes simplex virus ICPO protein not only disassembles PML nuclear bodies, but subsequently induces the proteasome-dependent degradation of PML90. Given the pace of recent research on this enticing and rewarding field, it can be hoped that the near future will uncover more of SUMO’s secrets. Links DATABASE LINKS ubiquitin | RUB1 | Apg8 | Apg12 |

parkin | RAD23 | DSK2 | SMT3 | SUMO-1 | SUMO-2 | SUMO-3 | RanGAP1 | AOS1 | UBA2 | UBA1 | UBC9 | UBC4 | UBC7 | ULP1 | ULP2 | SMT3IP1 | SENP-1 | SUSP-1 | Ran | RanBP2 | RNA1 | semushi | PML | p53 | Daxx | p73 | Sp100| HP1 | HIPK2 | TEL | c-Jun | androgen receptor | Tramtrack 69 | NF-κB | Mdm2 | IκBα | TNF | Dorsal | Cactus | RING finger | CDC3 | CDC11 | SEP7 | MIF2 | RAD51 | RAD52 | DNA topisomerase I | DNA topoisomerase II | RecQ | BLM | CENP-C FURTHER INFORMATION Jentsch lab ENCYCLOPEDIA OF LIFE SCIENCES Nuclear–cytoplasmic transport | Ubiquitin pathway

www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS 1. 2.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

Hochstrasser, M. Ubiquitin-dependent protein degradation. Annu. Rev. Genet. 30, 405–439 (1996). Jentsch, S. & Pyrowolakis, G. Ubiquitin and its kin: how close are the family ties? Trends Cell Biol. 10, 335–342 (2000). Hochstrasser, M. Evolution and function of ubiquitin-like protein-conjugation systems. Nature Cell Biol. 2, E153–E157 (2000). Hanania, U., Furman-Matarasso, N., Ron, M. & Avni, A. Isolation of a novel SUMO protein from tomato that suppresses EIX-induced cell death. Plant J. 19, 533–541 (1999). Kamitani, T., Kito, K., Nguyen, H. P., Fukuda-Kamitani, T. & Yeh, E. T. Characterization of a second member of the sentrin family of ubiquitin-like proteins. J. Biol. Chem. 273, 11349–11353 (1998). Boddy, M. N., Howe, K., Etkin, L. D., Solomon, E. & Freemont, P. S. PIC1, a novel ubiquitin-like protein which interacts with the PML component of a multiprotein complex that is disrupted in acute promyelocytic leukaemia. Oncogene 13, 971–982 (1996). Okura, T. et al. Protection against Fas/APO-1- and tumor necrosis factor-mediated cell death by a novel protein, sentrin. J. Immunol. 157, 4277–4281 (1996). Matunis, M. J., Coutavas, E. & Blobel, G. A novel ubiquitinlike modification modulates the partitioning of the RanGTPase-activating protein RanGAP1 between the cytosol and the nuclear pore complex. J. Cell Biol. 135, 1457–1470 (1996). RanGAP1 is identified as the first substrate for SUMO and the role of sumoylation in regulating the localization of RanGAP1 is shown. Bayer, P. et al. Structure determination of the small ubiquitin-related modifier SUMO-1. J. Mol. Biol. 280, 275–286 (1998). Mahajan, R., Delphin, C., Guan, T., Gerace, L. & Melchior, F. A small ubiquitin-related polypeptide involved in targeting RanGAP1 to nuclear pore complex protein RanBP2. Cell 88, 97–107 (1997). This important study identifies RanGAP1 as a SUMO substrate and shows that sumoylation of RanGAP1 determines its interaction with RanBP2. Koegl, M. et al. A novel ubiquitination factor, E4, is involved in multiubiquitin chain assembly. Cell 96, 635–644 (1999). Johnson, E. S., Schwienhorst, I., Dohmen, R. J. & Blobel, G. The ubiquitin-like protein Smt3p is activated for conjugation to other proteins by an Aos1p/Uba2p heterodimer. EMBO J. 16, 5509–5519 (1997). Gong, L., Li, B., Millas, S. & Yeh, E. T. Molecular cloning and characterization of human AOS1 and UBA2, components of the sentrin-activating enzyme complex. FEBS Lett. 448, 185–189 (1999). Desterro, J. M., Rodriguez, M. S., Kemp, G. D. & Hay, R. T. Identification of the enzyme required for activation of the small ubiquitin-like protein SUMO-1. J. Biol. Chem. 274, 10618–10624 (1999). Desterro, J. M., Thomson, J. & Hay, R. T. Ubch9 conjugates SUMO but not ubiquitin. FEBS Lett. 417, 297–300 (1997). Gong, L., Kamitani, T., Fujise, K., Caskey, L. S. & Yeh, E. T. Preferential interaction of sentrin with a ubiquitinconjugating enzyme, Ubc9. J. Biol. Chem. 272, 28198–28201 (1997). Johnson, E. S. & Blobel, G. Ubc9p is the conjugating enzyme for the ubiquitin-like protein Smt3p. J. Biol. Chem. 272, 26799–26802 (1997). Schwarz, S. E., Matuschewski, K., Liakopoulos, D., Scheffner, M. & Jentsch, S. The ubiquitin-like proteins SMT3 and SUMO-1 are conjugated by the UBC9 E2 enzyme. Proc. Natl Acad. Sci. USA 95, 560–564 (1998). Liu, Q. et al. The binding interface between an E2 (UBC9) and a ubiquitin homologue (UBL1). J. Biol. Chem. 274, 16979–16987 (1999). Giraud, M. F., Desterro, J. M. & Naismith, J. H. Structure of ubiquitin-conjugating enzyme 9 displays significant differences with other ubiquitin-conjugating enzymes which may reflect its specificity for sumo rather than ubiquitin. Acta Crystallogr. D. Biol. Crystallogr. 54, 891–898 (1998). Li, S. J. & Hochstrasser, M. A new protease required for cell-cycle progression in yeast. Nature 398, 246–251 (1999). This paper describes the identification of the yeast Ulp1 protease as the first SUMO de-conjugating enzyme and shows that de-sumoylation is essential for viability in yeast. Li, S. J. & Hochstrasser, M. The yeast ULP2 (SMT4) gene encodes a novel protease specific for the ubiquitin-like Smt3 protein. Mol. Cell. Biol. 20, 2367–2377 (2000).

23. Schwienhorst, I., Johnson, E. S. & Dohmen, R. J. SUMO conjugation and deconjugation. Mol. Gen. Genet. 263, 771–786 (2000). 24. Mossessova, E. & Lima, C. D. Ulp1-SUMO crystal structure and genetic analysis reveal conserved interactions and a regulatory element essential for cell growth in yeast. Mol. Cell 5, 865–876 (2000). This important study provides insight into the mechanism of substrate recognition and catalysis by Ulp1. 25. Yeh, E. T., Gong, L. & Kamitani, T. Ubiquitin-like proteins: new wines in new bottles. Gene 248, 1–14 (2000). 26. Nishida, T., Tanaka, H. & Yasuda, H. A novel mammalian Smt3-specific isopeptidase 1 (SMT3IP1) localized in the nucleolus at interphase. Eur. J. Biochem. 267, 6423–6427 (2000). 27. Gong, L., Millas, S., Maul, G. G. & Yeh, E. T. Differential regulation of sentrinized proteins by a novel sentrinspecific protease. J. Biol. Chem. 275, 3355–3359 (2000). 28. Kim, K. I. et al. A new SUMO-1-specific protease, SUSP1, that is highly expressed in reproductive organs. J. Biol. Chem. 275, 14102–14106 (2000). 29. Görlich, D. & Kutay, U. Transport between the cell nucleus and the cytoplasm. Annu. Rev. Cell. Dev. Biol. 15, 607–660 (1999). 30. Saitoh, H., Pu, R., Cavenagh, M. & Dasso, M. RanBP2 associates with Ubc9p and a modified form of RanGAP1. Proc. Natl Acad. Sci. USA 94, 3736–3741 (1997). 31. Mahajan, R., Gerace, L. & Melchior, F. Molecular characterization of the SUMO-1 modification of RanGAP1 and its role in nuclear envelope association. J. Cell Biol. 140, 259–270 (1998). 32. Matunis, M. J., Wu, J. & Blobel, G. SUMO-1 modification and its role in targeting the Ran GTPase-activating protein, RanGAP1, to the nuclear pore complex. J. Cell Biol. 140, 499–509 (1998). 33. Rout, M. P. et al. The yeast nuclear pore complex: composition, architecture, and transport mechanism. J. Cell Biol. 148, 635–651 (2000). 34. Takahashi, Y., Mizoi, J., Toh, E. A. & Kikuchi, Y. Yeast Ulp1, an Smt3-specific protease, associates with nucleoporins. J. Biochem. 128, 723–725 (2000). 35. Epps, J. L. & Tanda, S. The Drosophila semushi mutation blocks nuclear import of bicoid during embryogenesis. Curr. Biol. 8, 1277–1280 (1998). 36. Kamitani, T., Nguyen, H. P. & Yeh, E. T. Preferential modification of nuclear proteins by a novel ubiquitin-like molecule. J. Biol. Chem. 272, 14001–14004 (1997). 37. Guo, A. et al. The function of PML in p53-dependent apoptosis. Nature Cell Biol. 2, 730–736 (2000). 38. Fogal, V. et al. Regulation of p53 activity in nuclear bodies by a specific PML isoform. EMBO J. 19, 6185–6195 (2000). 39. Muller, S., Matunis, M. J. & Dejean, A. Conjugation with the ubiquitin-related modifier SUMO-1 regulates the partitioning of PML within the nucleus. EMBO J. 17, 61–70 (1998). This study identifies PML as a substrate for SUMO and implicates sumoylation of PML in the regulation of its compartmentalization in nuclear bodies. 40. Sternsdorf, T., Jensen, K. & Will, H. Evidence for covalent modification of the nuclear dot-associated proteins PML and Sp100 by PIC1/SUMO-1. J. Cell Biol. 139, 1621–1634 (1997). 41. Kamitani, T., Nguyen, H. P., Kito, K., Fukuda-Kamitani, T. & Yeh, E. T. Covalent modification of PML by the sentrin family of ubiquitin-like proteins. J. Biol. Chem. 273, 3117–3120 (1998). 42. Kamitani, T. et al. Identification of three major sentrinization sites in PML. J. Biol. Chem. 273, 26675–26682 (1998). 43. Duprez, E. et al. SUMO-1 modification of the acute promyelocytic leukaemia protein PML: implications for nuclear localization. J. Cell Sci. 112, 381–393 (1999). 44. Everett, R. D. et al. The disruption of ND10 during herpes simplex virus infection correlates with the Vmw110- and proteasome-dependent loss of several PML isoforms. J. Virol. 72, 6581–6591 (1998). 45. Muller, S. & Dejean, A. Viral immediate-early proteins abrogate the modification by SUMO-1 of PML and Sp100 proteins, correlating with nuclear body disruption. J. Virol. 73, 5137–5143 (1999). 46. Everett, R. D., Lomonte, P., Sternsdorf, T., van Driel, R. & Orr, A. Cell cycle regulation of PML modification and ND10 composition. J. Cell Sci. 112, 4581–4588 (1999). 47. Zhong, S. et al. Role of SUMO-1-modified PML in nuclear body formation. Blood 95, 2748–2752 (2000). 48. Ishov, A. M. et al. PML is critical for ND10 formation and recruits the PML-interacting protein Daxx to this nuclear structure when modified by SUMO-1. J. Cell Biol. 147, 221–234 (1999).

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

49. Li, H. et al. Sequestration and inhibition of Daxx-mediated transcriptional repression by PML. Mol. Cell. Biol. 20, 1784–1796 (2000). 50. Lehembre, F., Muller, S., Pandolfi, P. P. & Dejean, A. Regulation of Pax3 transcriptional activity by SUMO-1modified PML. Oncogene 20, 1–9 (2001). 51. Pearson, M. et al. PML regulates p53 acetylation and premature senescence induced by oncogenic Ras. Nature 406, 207–210 (2000). 52. Gostissa, M. et al. Activation of p53 by conjugation to the ubiquitin-like protein SUMO-1. EMBO J. 18, 6462–6471 (1999). 53. Rodriguez, M. S. et al. SUMO-1 modification activates the transcriptional response of p53. EMBO J. 18, 6455–6461 (1999). 54. Muller, S. et al. c-Jun and p53 activity is modulated by SUMO-1 modification. J. Biol. Chem. 275, 13321–13329 (2000). 55. Minty, A., Dumont, X., Kaghad, M. & Caput, D. Covalent modification of p73α by SUMO-1. Two-hybrid screening with p73 identifies novel SUMO-1-interacting proteins and a SUMO-1 interaction motif. J. Biol. Chem. 275, 36316–36323 (2000). 56. Seeler, J. S., Marchio, A., Sitterlin, D., Transy, C. & Dejean, A. Interaction of SP100 with HP1 proteins: a link between the promyelocytic leukemia-associated nuclear bodies and the chromatin compartment. Proc. Natl Acad. Sci. USA 95, 7316–7321 (1998). 57. Lehming, N., Le Saux, A., Schuller, J. & Ptashne, M. Chromatin components as part of a putative transcriptional repressing complex. Proc. Natl Acad. Sci. USA 95, 7322–7326 (1998). 58. Sternsdorf, T., Jensen, K., Reich, B. & Will, H. The nuclear dot protein SP100, characterization of domains necessary for dimerization, subcellular localization, and modification by small ubiquitin-like modifiers. J. Biol. Chem. 274, 12555–12566 (1999). 59. Seeler, J. S. et al. Common properties of the nuclear body protein SP100 and the TIF1α chromatin factor: the role of SUMO modification. Mol. Cell. Biol. (in the press). 60. Kim, Y. H., Choi, C. Y. & Kim, Y. Covalent modification of the homeodomain-interacting protein kinase 2 (HIPK2) by the ubiquitin-like protein SUMO-1. Proc. Natl Acad. Sci. USA 96, 12350–12355 (1999). 61. Chakrabarti, S. R., Sood, R., Nandi, S. & Nucifora, G. Posttranslational modification of TEL and TEL/AML1 by SUMO-1 and cell-cycle-dependent assembly into nuclear bodies. Proc. Natl Acad. Sci. USA 97, 13281–13285 (2000). 62. Chakrabarti, S. R. et al. Modulation of TEL transcription activity by interaction with the ubiquitin-conjugating enzyme UBC9. Proc. Natl Acad. Sci. USA 96, 7467–7472 (1999). 63. Poukka, H., Karvonen, U., Janne, O. A. & Palvimo, J. J. Covalent modification of the androgen receptor by small ubiquitin-like modifier 1 (SUMO-1). Proc. Natl Acad. Sci. USA 97, 14145–14150 (2000). 64. Lehembre, F. et al. Covalent modification of the transcriptional repressor tramtrack by the ubiquitin-related protein Smt3 in Drosophila flies. Mol. Cell. Biol. 20, 1072–1082 (2000). 65. Desterro, J. M., Rodriguez, M. S. & Hay, R. T. SUMO-1 modification of IκBα inhibits NF-κB activation. Mol. Cell 2, 233–239 (1998). This report identifies IκBα as a target for SUMO and provides evidence for a role of sumoylation in counteracting the ubiquitylation of IκBα, thus leading to the model that SUMO acts as a protein stabilizer. 66. Buschmann, T., Fuchs, S. Y., Lee, C. G., Pan, Z. Q. & Ronai, Z. SUMO-1 modification of Mdm2 prevents its selfubiquitination and increases Mdm2 ability to ubiquitinate p53. Cell 101, 753–762 (2000). This study reports that SUMO and ubiquitin share an identical lysine residue within the RING finger of Mdm2 and suggest that sumoylation stabilizes Mdm2. 67. Israel, A. The IKK complex: an integrator of all signals that activate NF-κB? Trends Cell Biol. 10, 129–133 (2000). 68. Bhaskar, V., Valentine, S. A. & Courey, A. J. A functional interaction between dorsal and components of the Smt3 conjugation machinery. J. Biol. Chem. 275, 4033–4040 (2000). 69. Fang, S., Jensen, J. P., Ludwig, R. L., Vousden, K. H. & Weissman, A. M. Mdm2 is a RING finger-dependent ubiquitin protein ligase for itself and p53. J. Biol. Chem. 275, 8945–8951 (2000). 70. Honda, R. & Yasuda, H. Activity of MDM2, a ubiquitin ligase, toward p53 or itself is dependent on the RING finger domain of the ligase. Oncogene 19, 1473–1476 (2000).

VOLUME 2 | MARCH 2001 | 2 0 9

REVIEWS 71. Melchior, F. & Hengst, L. Mdm2–SUMO1: is bigger better? Nature Cell Biol. 2, E161–E163 (2000). 72. al-Khodairy, F., Enoch, T., Hagan, I. M. & Carr, A. M. The Schizosaccharomyces pombe hus5 gene encodes a ubiquitin conjugating enzyme required for normal mitosis. J. Cell Sci. 108, 475–486 (1995). 73. Seufert, W., Futcher, B. & Jentsch, S. Role of a ubiquitinconjugating enzyme in degradation of S- and M-phase cyclins. Nature 373, 78–81 (1995). 74. Takahashi, Y. et al. Smt3, a SUMO-1 homolog, is conjugated to Cdc3, a component of septin rings at the mother-bud neck in budding yeast. Biochem. Biophys. Res. Commun. 259, 582–587 (1999). 75. Johnson, E. S. & Blobel, G. Cell cycle-regulated attachment of the ubiquitin-related protein SUMO to the yeast septins. J. Cell Biol. 147, 981–994 (1999). This comprehensive study identifies septins as the major substrates for SUMO in yeast and suggests a role of septin sumoylation in cytokinesis. 76. Meluh, P. B. & Koshland, D. Evidence that the MIF2 gene of Saccharomyces cerevisiae encodes a centromere protein with homology to the mammalian centromere protein CENP-C. Mol. Biol. Cell 6, 793–807 (1995). This work initially identifies the yeast SMT3 gene among other genes as a suppressor of MIF2 mutations, suggesting a role of SUMO for the maintenance of genomic integrity. 77. Brown, M. T., Goetsch, L. & Hartwell, L. H. MIF2 is required for mitotic spindle integrity during anaphase spindle elongation in Saccharomyces cerevisiae. J. Cell Biol. 123, 387–403 (1993). 78. Tanaka, K. et al. Characterization of a fission yeast SUMO1 homologue, pmt3p, required for multiple nuclear events, including the control of telomere length and chromosome segregation. Mol. Cell. Biol. 19, 8660–8672 (1999). 79. Shen, Z., Pardington-Purtymun, P. E., Comeaux, J. C., Moyzis, R. K. & Chen, D. J. UBL1, a human ubiquitin-like protein associating with human RAD51/RAD52 proteins. Genomics 36, 271–279 (1996). 80. Shen, Z., Pardington-Purtymun, P. E., Comeaux, J. C., Moyzis, R. K. & Chen, D. J. Associations of UBE2I with

210

RAD52, UBL1, p53, and RAD51 proteins in a yeast twohybrid system. Genomics 37, 183–186 (1996). 81. Kovalenko, O. V. et al. Mammalian ubiquitin-conjugating enzyme Ubc9 interacts with Rad51 recombination protein and localizes in synaptonemal complexes. Proc. Natl Acad. Sci. USA 93, 2598–2563 (1996). 82. Li, W. et al. Regulation of double-strand break-induced mammalian homologous recombination by UBL1, a RAD51-interacting protein. Nucleic Acids Res. 28, 1145–1153 (2000). 83. Mao, Y., Sun, M., Desai, S. D. & Liu, L. F. SUMO-1 conjugation to topoisomerase I: a possible repair response to topoisomerase-mediated DNA damage. Proc. Natl Acad. Sci. USA 97, 4046–4051 (2000). 84. Mao, Y., Desai, S. D. & Liu, L. F. SUMO-1 conjugation to human DNA topoisomerase II isozymes. J. Biol. Chem. 275, 26066–26073 (2000). 85. Kawabe, Y. et al. Covalent modification of the Werner’s syndrome gene product with the ubiquitin-related protein, SUMO-1. J. Biol. Chem. 275, 20963–20966 (2000). 86. Yeager, T. R. et al. Telomerase-negative immortalized human cells contain a novel type of promyelocytic leukemia (PML) body. Cancer Res. 59, 4175–4179 (1999). 87. Zhong, S. et al. A role for PML and the nuclear body in genomic stability. Oncogene 18, 7941–7947 (1999). 88. Johnson, F. B. et al. Association of the Bloom syndrome protein with topoisomerase IIIα in somatic and meiotic cells. Cancer Res. 60, 1162–1167 (2000). 89. Everett, R. D. et al. A dynamic connection between centromeres and ND10 proteins. J. Cell Sci. 112, 3443–3454 (1999). 90. Everett, R. D. ICP0 induces the accumulation of colocalizing conjugated ubiquitin. J. Virol. 74, 9994–10005 (2000). 91. Zhong, S., Salomoni, P. & Pandolfi, P. P. The transcriptional role of PML and the nuclear body. Nature Cell Biol. 2, E85–E90 (2000). 92. Seeler, J. S. & Dejean, A. The PML nuclear bodies: actors or extras? Curr. Opin. Genet. Dev. 9, 362–367 (1999). 93. Hofmann, H., Floss, S. & Stamminger, T. Covalent modification of the transactivator protein IE2-p86 of

94.

95.

96.

97.

98.

99.

100.

101.

human cytomegalovirus by conjugation to the ubiquitinhomologous proteins SUMO-1 and hSMT3b. J. Virol. 74, 2510–2524 (2000). Adamson, A. L. & Kenney, S. The epstein-barr virus immediate-early protein BZLF1 is SUMO–1 modified and disrupts promyelocytic leukemia (PML) bodies. J. Virol. 74, 1224–1233 (2000). Zhang, Q., Gutsch, D. & Kenney, S. Functional and physical interaction between p53 and BZLF1: implications for Epstein-Barr virus latency. Mol. Cell. Biol. 14, 1929–1938 (1994). Andres, G., Alejo, A., Simon-Mateo, C. & Salas, M. L. African swine fever virus protease: A new viral member of the SUMO-1–specific protease family. J. Biol. Chem. 276, 780–787 (2000). Orth, K. et al. Disruption of signaling by yersinia effector YopJ, a ubiquitin-like protein protease. Science 290, 1594–1597 (2000). Saitoh, H. et al. Ubc9p and the conjugation of SUMO-1 to RanGAP1 and RanBP2. Curr. Biol. 8, 121–124 (1998). Giorgino, F. et al. The sentrin-conjugating enzyme mUbc9 interacts with GLUT4 and GLUT1 glucose transporters and regulates transporter levels in skeletal muscle cells. Proc. Natl Acad. Sci. USA 97, 1125–1130 (2000). Long, X. & Griffith, L. C. Identification and characterization of a SUMO–1 conjugation system that modifies neuronal CaMKII in Drosophila melanogaster. J. Biol. Chem. 275, 40765–40776 (2000). Rangasamy, D., Woytek, K., Khan, S. A. & Wilson, V. G. SUMO–1 modification of bovine papillomavirus E1 protein is required for intranuclear accumulation. J. Biol. Chem. 275, 37999–38004(2000).

Acknowledgements We wish to thank C. D. Lima for the permission to include the structural data on ULP1-SMT3 in this review and we are indebted to many colleagues for sharing unpublished results. S. M. would like to express his special thanks to Anne Dejean for stimulating discussions, encouragement and continuous support during his stay in her laboratory.

www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS

MOLECULAR DISSECTION OF AUTOPHAGY: TWO UBIQUITIN-LIKE SYSTEMS Yoshinori Ohsumi Recent analyses of the genes required for autophagy — intracellular bulk protein degradation — in yeast have revealed two ubiquitin-like systems, both of which are involved in the membrane dynamics of the process. Molecular dissection of these systems is now revealing some surprises. U B I Q U I T I N A N D P R OT E A S O M E S

Department of Cell Biology, National Institute for Basic Biology, Myodaiji, Okazaki, 444-8585, Japan. e-mail: yohsumi@nibb.ac.jp

Cellular activities are maintained as a balance between the synthesis and breakdown of various proteins. Since the ubiquitin/proteasome pathway was found1, shortlived proteins have attracted much attention. But more than 99% of cellular proteins have relatively long lives. It is well known that protein lifetimes range from a few minutes to more than ten days. Usually, a long lifetime is defined as a protein with a half-life of more than five hours. Most long-lived proteins are believed to be eventually degraded in a lytic compartment, the lysosome or vacuole (for a review, see REF. 2; the vacuole in yeast is equivalent to the mammalian lysosome). The degradative routes to this compartment are complex (BOX 1). Macroautophagy — hereafter referred to as autophagy — is the main route for sequestration of cytoplasm to the lytic compartment. It is defined as the process by which a portion of the cytosol is sequestered by a so-called isolation membrane. This results in the formation of a double-membrane structure called the autophagosome, which subsequently fuses with the lysosome/vacuole3 (FIG. 1). The inner membrane and its contents are then degraded for re-use. Depletion of various nutrients such as nitrogen, carbon, sulphate or phosphate triggers a similar membrane reorganization, indicating that, in yeast, autophagy might be a general physiological response against adverse circumstances4. In mammals, autophagy is also induced by nutrient starvation, but it might have further roles in response to certain physiological demands. Various aspects of membrane dynamics during formation of the autophago-

some are distinct from classical membrane transport, so they are probably based on novel principles. Moreover, although studies of lysosomal protein degradation have a long history, mammalian lysosomes are so dynamic and complicated that biochemical analyses are not easy. So the molecular basis of autophagy has remained hidden for many years. During the past decade, however, molecular biological approaches using the yeast Saccharomyces cerevisiae as a model system have begun to uncover the secrets of autophagy5,6. Discovery of autophagy in yeast

The yeast vacuole is an acidic compartment that contains hydrolytic enzymes. A key observation in the study of how a portion of cytoplasm is sequestered into the vacuole was made by light microscopy4. Yeast mutants lacking vacuolar proteases were incubated under starvation conditions, and morphological changes were observed. After 30–40 minutes of incubation, spherical bodies appeared in the vacuole. These gradually accumulated until, after 4–6 hours, they occupied the entire vacuole. Electron-microscopic analyses revealed that these structures (300–900 nm in diameter) are mostly single-membrane structures containing cytoplasm3,4 (FIG. 1). This indicates that starved yeast cells take up their own cytoplasm into vacuoles through these membrane structures, which are called autophagic bodies. Further electron microscopy revealed an organelle that was bound by a double membrane in the cytoplasm. This organelle, the autophagosome, immediately

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 2 1 1

REVIEWS

Genetic approaches to autophagy in yeast

Box 1 | Lysosomal/vacuolar degradation systems There are many routes by which cellular proteins can be a sequestered to the lysosome/vacuole to c b be degraded. a | Endocytosis: specific sets of plasma-membrane proteins are internalized and e transported to the lysosome/vacuole39. d b | Macroautophagy*: a Vacuole/lysosome portion of cytoplasm including an organelle is first enclosed in a double or multilamellar structure, the autophagosome. The autophagosome then fuses with the outer membrane of the lysosome/vacuole, and the inner-membrane structure is rapidly disintegrated2. c | Microautophagy: the lysosomal/vacuolar membrane invaginates and pinches off, resulting in sequestration of the cytosol or organelles to the lysosome/vacuole40. d | The vacuolar import and degradation (Vid) pathway: in yeast, specific enzymes such as fructose-1,6-bisphosphatase are selectively taken up into small membrane vesicles when they are no longer necessary. These vesicles then fuse with the vacuole and the enzymes are digested41. e | Direct transport. Various proteins containing the broad consensus sequence KFERQ are directly transported across the lysosomal membrane through heat-shock protein73 and LampIIa42. Recently, Dice et al.42 proposed this route as ‘chaperone-mediated autophagy’. It is also found in yeast. *Macroautophagy is generally a non-selective sequestration. But in some physiological situations, excess or unnecessary organelles are selectively degraded in a lysosome/vacuole. The best-studied case is peroxisome degradation, called pexophagy by Daniel Klionsky. In the yeast Pichia pastoris, peroxisomes are sequestered through either macroautophagy or microautophagy, depending on the nutrient conditions40.

fuses with the vacuole — the result is an autophagic body in the vacuole 3. Freeze-fracture electron microscopy showed that the autophagosome is a specialized membrane structure. The inner membrane of the autophagosome completely lacks any intramembrane particles, representing transmembrane proteins, but the outer membrane contains a low but significant number of particles7. The whole process in yeast is topologically the same as macroautophagy in mammals.

Taking advantage of yeast genetics, the first set of autophagy-defective (apg) mutants was obtained by a morphological criterion — lack of accumulation of autophagic bodies under starving conditions8. Mutants defective in degradation of cytosol, the aut (autophagy) mutants, were isolated by immunochemical detection of a cytosolic marker protein, fatty acid synthase, after starvation9. These screens identified 14 APG and 6 AUT genes that are necessary for autophagy. Further genes, APG16 and APG17, were obtained by two-hybrid screening using Apg12 and Apg1 as bait, respectively10,11. Besides them, cytosol-to-vacuole-targeting (cvt) mutants defective in the biogenesis of aminopeptidase I were obtained12 (BOX 2). These CVT genes partially overlap with the AUT and APG genes. And all apg mutants except apg17 are defective in the Cvt pathway. Most apg mutants have similar phenotypes, except apg6 and apg17 mutants. The apg6 mutation turns out to be allelic to the vps30 mutation, one of the genes required for vacuolar protein sorting. All of the original apg mutants cannot induce bulk protein degradation under conditions of starvation8. Although they grow normally in a rich medium, they cannot survive long-term starvation, and start to die after two days. Autophagy is required for adaptation to nutrient limitation, and is essential in wild-yeast populations, which are often depleted in nutrients. Homozygous diploids of each apg mutant fail to sporulate, indicating that autophagy might be needed to remodel cells during differentiation. Electron microscopy has shown that apg mutants do not accumulate autophagosomes in the cytoplasm. In addition, the APG genes are not involved in the step of fusion to the vacuole13,14. So all apg mutants have defects at or before formation of the autophagosome. Most of them might also have some role in microautophagy and/or pexophagy, which is selective degradation of peroxisomes in the vacuole15–19. When the APG genes were initially sequenced, almost all were found to be novel genes with no similarities to known proteins or motifs that might hint at their functions. However, within the past few years, analyses of these genes have progressed rapidly, and now it is clear that the APG gene products are involved in several stages of autophagy, with related

Figure 1 | Electron microscopic images of autophagy in yeast. a | A BJ926 cell that is deficient in vacuolar proteinases and has been incubated for three hours in a carbon-deficient medium. Many autophagic bodies have accumulated in the vacuole. (Magnification x 6,400.) b | An isolation membrane enclosing a portion of cytosol next to the vacuole. (Magnification x 43,000.) c | An autophagosome attached to the vacuolar membrane. (Magnification x 36,000; M. Baba and Y.O., unpublished.)

212

www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS

Box 2 | The cytoplasm-to-vacuole-targeting pathway One of the vacuolar resident hydrolases, aminopeptidase I (API), is synthesized as an inactive precursor in the cytosol, then transported to the vacuole. This system is called the cytoplasm-to-vacuole-targeting (Cvt) pathway. It is a constitutive and biosynthetic pathway, whereas macroautophagy is a starvation-induced, degradative pathway43. Electron microscopic analyses have revealed that the Cvt pathway is similar to macroautophagy44. The complex containing pro-API is selectively enclosed to form small double-membrane vesicles (Cvt vesicles), which fuse with the vacuole. Mutants that are defective in API maturation (cvt mutants) have been isolated. Some cvt mutants overlap with autophagy-defective (apg) and autophagy (aut) mutants, and all the original apg mutants are defective in the Cvt pathway45,46. Some Cvt proteins are specifically required for the Cvt pathway but not for autophagy, whereas the apg17 mutant has a specific defect in autophagy but not in the Cvt pathway15. Because both pathways mostly share machinery, analysis of either pathway contributes towards understanding the other. So far, a Cvt-like system has not been found in higher eukaryotes.

Vacuole

Cvt vesicle Growing

Cvt body Cvt complex Isolation

Nucleus

Fusion

Autophagic body

Nutrient starvation

Autophagosome

functions. In particular, two ubiquitin-like systems found among the APG gene products have expanded the world of ubiquitin-like proteins (UBLs; FIG. 2). The Apg12 conjugation system

E1 ENZYME

An enzyme that activates the carboxy-terminal glycine of the small protein ubiquitin, or ubiquitin-like proteins, allowing them to form a highenergy bond to a specific cysteine residue of the E1. E2 ENZYME

An enzyme that accepts ubiquitin or a ubiquitin-like protein from an E1 and transfers it to the substrate, mostly using an E3 enzyme.

Apg12 is a 186-amino-acid hydrophilic protein, with no apparent homology to ubiquitin20. It was, however, the first UBL that was discovered to act during autophagy. Here we refer to a protein as a UBL when it forms a conjugate with target molecules through the ubiquitylationlike system (see the review by Allan Weissmann on page 169 of this issue). A carboxy-terminal glycine residue of Apg12 forms a covalently bound complex with Lys149 at the centre of Apg5 (REF. 20). As in the ubiquitin system, the carboxy-terminal glycine of Apg12 is first activated by an E1 ENZYME — Apg7/Cvt2 (REF. 21). (Apg7 is homologous to the other E1 enzymes around its ATP-binding site, but not in its overall structure.) This ATP-dependent reaction results in the formation of a high-energy thioester bond with Cys306 of Apg7 (REF. 21). Subsequently, Apg12 is transferred to an E2 ENZYME, Apg10, forming a thioester bond with Cys133 (REF. 22). Known E2 enzymes resemble each other, but Apg10 shows no homology to any of them. The Apg12 system has unique characteristics compared with the other UBL systems. First, Apg12 is synthesized as a mature form — that is, the APG12 gene

encodes a protein that ends with a single glycine residue. (Contrast this with ubiquitin, which is synthesized as a precursor with carboxy-terminal fusion proteins after a double glycine). In addition, we have not been able to detect any proteolytic activity that processes genetically extended Apg12 after the terminal glycine (N. Mizushima, T. Noda and Y.O., unpublished observations). This might indicate that there is no processing enzyme that cleaves the isopeptide bond between Apg12 and Apg5 (the functional equivalent of a de-ubiquitylating enzyme). Second, Apg5 seems to be the sole target for Apg12 conjugation, whereas ubiquitin and other UBLs have many substrates. Third, after translation, Apg12 immediately conjugates with Apg5. Surprisingly, formation of this conjugate is constitutive and is not regulated by starvation. At steady state, almost all Apg12 molecules are conjugated to Apg5, and the free form of Apg12 is scarcely detected (N. Ishihara, N. Mizushima and Y.O., unpublished observations). These characteristics contradict the prevailing idea that UBLs transiently and reversibly modulate the fate or the function of their target proteins. Another protein, Apg16, was first found by a twohybrid screen that used Apg12 as the bait10. Apg16 binds to the Apg12–Apg5 conjugate, and also to free Apg5 (albeit less effectively) but not to free Apg12, indicating that the conjugation reaction might stabilize the Apg12–Apg5–Apg16 complex or stimulate binding to Apg16. Apg16 has a coiled-coil region through which it forms homo-oligomers. This allows Apg16 to crosslink two or more Apg12–Apg5 conjugates into a single large protein complex. Thus, conjugate formation might contribute to the formation of the large protein complex that is essential for autophagy. Apg8, another UBL in autophagy

The second UBL discovered to function during autophagy was Aut7/Apg8/Cvt5, a 117-amino-acid basic protein23–25. Epitope-tagging of the carboxyl terminus of Apg8 revealed that nascent Apg8 is processed at its carboxy-terminal region immediately after it has been synthesized by Aut2/Apg4 (REF. 26). Although Apg4 is not homologous to any known protein, it has turned out to be a new member of the cysteine protease (caspase) family26. By constructing various carboxy-terminal versions of Apg8, we have shown that Apg8 is processed at its very end, with a single amino acid being cleaved off to leave the glycine residue at the carboxyl terminus. This carboxy-terminal glycine is necessary for the function of Apg8 during autophagy. We have also found26 that Aut1/Apg3 physically interacts with Apg8 in an Apg7 (E1)-dependent manner. Detailed analysis revealed that Apg8 is conjugated with Apg7 and then with Apg3, through a thioester bond in both cases27. Notably, Apg3 shows a limited homology to Apg10, the E2 enzyme for Apg12, around its active cysteine residue. Both this cysteine residue and the carboxy-terminal glycine of Apg8 are necessary for this binding and, from these observations, we have concluded that Apg8 is another UBL.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 2 1 3

REVIEWS

Lipidation by the UBL system

Despite the above conclusion, the identity of the molecule that Apg8 targets was not clear because standard SDS (sodium dodecyl sulphate) polyacrylamide gel electrophoresis showed only a single band of Apg8 at the expected molecular size26. However, we then realized that Apg8 exists in two forms — either in a free/peripherally membrane-bound form, or in a tightly membrane-associated form26. Formation of the membranebound form depends on Apg4, Apg7, Apg3 and the carboxy-terminal glycine of Apg8. Moreover, in ∆apg4 cells that express Apg8 with glycine at its carboxyl terminus, the amount of the membrane-associated form is greatly enhanced. From these observations we deduced that Apg8 must form a conjugate with an unknown small hydrophobic molecule, and that Apg4 is necessary for deconjugating it from Apg8 (REF. 27). Mass spectrometry analyses of the purified, tightly membrane-associated form of Apg8 indicated a novel modification by a phosphoglycerolipid. An amino group of phosphatidylethanolamine is covalently attached to the carboxy-terminal glycine of Apg8 through an amide bond27, indicating that Apg8

De-ubiquitylating enzymes Ub a Ubiquitin system UbG

Ub Ub

AMP Ub GG ATP +PPi

UbG

Ub Ub UbG G

Targets

AMP ATP +PPi

Ap g1 2

b Apg12 system

Apg7

Apg10

Apg5

The homologues of Apg12 and Apg8

Apg12 Apg7

Ap g8

c Apg8 system

AMP G ATP +PPi

GR Apg4

Apg8

Apg4 Ap g8

Ap g8

Apg8 Apg7

Apg3

? PE

Apg7

Figure 2 | Two ubiquitin-like systems are required for autophagy. a | The ubiquitin system. b | The autophagy-defective 12 (Apg12) system. The carboxy-terminal glycine of Apg12 is activated by Apg7, an E1-like enzyme of the ubiquitin (Ub) system. Subsequently, Apg12 is transferred to Apg10, an E2-like conjugating enzyme. Finally, Apg12 forms a conjugate with Apg5 through an isopeptide bond. The Apg12–Apg5 conjugate forms a large protein complex with Apg16. Apg12 is synthesized as an active precursor and immediately forms a conjugate with Apg5. No deconjugation of Apg12–Apg5 has been observed. c | The Apg8 system. Nascent Apg8 is first processed to a glycine-exposed form by a protease, Apg4. Apg8 is also activated by Apg7 (an E1) then transferred to Apg3 (an E2). Finally, Apg8 forms a conjugate with phosphatidylethanolamine (PE). Apg8–PE is deconjugated by Apg4.

214

forms a conjugate with a common membrane phospholipid through ubiquitylation-like reactions. In vitro studies26 indicate that Apg4 cleaves the amide bond between Apg8 and phosphatidylethanolamine and, as a consequence, some Apg8 is released from the membrane. In this respect, Apg4 resembles the deconjugating enzymes of the ubiquitin or other UBL systems. The lipidation reaction is reversible, and the second function of Apg4 is also necessary for the normal progress of autophagy. These findings have expanded the concept of a UBL to include conjugation not only with target proteins, but also with other molecules such as lipids. There have been some long-standing arguments about the origin of the membrane that surrounds the autophagosome. Apg8 is the first molecule that has been shown to localize to the intermediate membrane structures that arise during formation of the autophagosome23,26. Morphological studies using Apg8 as a marker indicate that the autophagosome is not formed by pre-existing large membrane cisternae, but that its formation is accompanied by the assembly of small membranous sources23. Immunofluorescence microscopy has revealed that one or a few punctate Apg8-positive structures appear next to the vacuole upon starvation23. And, by immuno-electron microscopy, the Apg8 signals have been shown to be densely localized on the isolation membrane as it forms and on small structures adjacent to it23. Complete autophagosomes and autophagic bodies are also stained, but less densely, and in their lumens rather than on their membranes. This transient membrane association of Apg8 must be essential for formation of the autophagosome, and can be explained by the reversible lipidation reaction. According to this scheme, phosphatidylethanolamine may be delivered to the autophagosome through the Apg8 system. This hypothesis now needs to be tested.

The Apg12 and Apg8 systems are highly conserved among higher eukaryotes, with apparent homologues in the nematode, mammals and plants. Indeed, the human homologues of Apg12 and Apg5 have been shown to be conjugated in a similar manner to their yeast counterparts28. Apg12–Apg5 localizes to the isolation membrane throughout its elongation process. Targeting of the Apg5 gene in mouse embryonic stem cells revealed that Apg12–Apg5 is required for elongation of the isolation membrane29. Apg8 is the most conserved of the Apg proteins. Whereas in yeast it is derived from a single gene, in higher eukaryotes Apg8 consists of a multigene family. The first homologue of Apg8, MAP1-LC3, was identified as a light chain of the microtubule-associated protein 1 in rat30. Recently, LC3 has been shown to be proteolytically processed after the conserved glycine residue to produce an active form, the LC3-I form31. It is then further modified to a different form, LC3-II, that is localized to the autophagosomal membrane31. Although the features of this second modification are not yet known, LC3-II, in a similar way to Apg8, seems to behave like an integral www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

REVIEWS

NSF

N-ethylmaleimide-sensitive factor, an AAA-type ATPase essential for membrane fusion during vesicle transport. V-SNARE

A family of proteins on secretory vesicles. They form a complex with t-SNAREs on the target membrane during vesicle fusion.

membrane protein after the modification31. Interestingly, insertion of the LC3 gene into the genome of an animal virus (pestivirus), and its processing at the conserved glycine residue, is correlated with the cytotoxicity of the virus32. Another mammalian homologue of Apg8, GATE16 (Golgi-associated ATPase enhancer of 16 kDa), acts as a soluble transport factor. GATE-16 interacts with N-ethylmaleimide-sensitive factor (NSF) and significantly stimulates its ATPase activity33. It also interacts with the Golgi V-SNARE GOS-28, which acts during membrane fusion, in an NSF-dependent manner33. (Yeast Apg8 also interacts both physically and genetically with some v-SNARE proteins34.) Richard Olsen and colleagues have proposed that GATE-16 functions during the vesicular transport that occurs between the cisternae of the Golgi. The third mammalian homologue of Apg8 is the GABA A (γ-aminobutyric acid type A) receptor-associated protein (GABARAP), which promotes the clustering of GABAA receptors in combination with microtubules34,35. So, members of the Apg8 family seem to carry out diverse cellular activities, and it remains to be seen whether, like Apg8, all of them are modified with lipid. Significance of the Apg12 and Apg8 systems

Among 16 APG genes, then, eight are related to UBL systems. It is intriguing that in both of the main proteindegradation pathways — the ubiquitin/proteasome and autophagy pathways — UBL systems have crucial functions but completely different roles. Although Apg12 and Apg8 have no significant homology to ubiquitin or UBLs (except, of course, the all-important carboxy-terminal glycine residue), they show some limited homology with each other26. However, GATE-16 has been reported37 to have a ubiquitin fold with two additional amino-terminal helices, and Apg12 and Apg8 will also probably turn out to have three-dimensional structures similar to that of ubiquitin.

1. 2.

Hershko, A. & Ciechanover, A. The ubiquitin system. Annu. Rev. Biochem. 67, 425–479 (1998). Klionsky, D. J. & Ohsumi, Y. Vacuolar import of proteins and organelles from the cytoplasm. Annu. Rev. Cell Dev. Biol. 15, 1–32 (1999). Baba, M., Takeshige, K., Baba, N. & Ohsumi, Y. Ultrastructural analysis of the autophagic process in yeast: detection of autophagosomes and their characterization. J. Cell Biol. 124, 903–913 (1994). Takeshige, K., Baba, M., Tsuboi, S., Noda, T. & Ohsumi, Y. Autophagy in yeast demonstrated with proteinasedeficient mutants and conditions for its induction. J. Cell Biol. 119, 301–311 (1992). This is the first report that yeast induces autophagy that is quite similar to that in mammals under various starvation conditions. Kim, J. & Klionsky, D. J. Autophagy, the cytoplasm-tovacuole-targeting pathway, and pexophagy in yeast and mammalian cells. Annu. Rev. Biochem. 69, 303–342 (2000). Klionsky, D. J. & Emr, S. Autophagy as a regulated pathway of cellular degradation. Science 290, 1717–1721 (2000). Baba, M., Osumi, M. & Ohsumi, Y. Analysis of the membrane structures involved in autophagy in yeast by freeze-replica method. Cell Struct. Funct. 20, 465–471 (1995). Tsukada, M. & Ohsumi, Y. Isolation and characterization of autophagy-defective mutants of Saccharomyces

10.

11.

12.

13.

14.

15.

16.

Apg7 is the first E1 enzyme shown to activate two distinct UBLs. It assigns Apg12 and Apg8 to specific E2 molecules — Apg10 and Apg3, respectively. Although these E2 molecules contain small homologous regions around their active sites, their primary sequences are different both from one another and also from those of other known E2 enzymes. Hence, although the Apg12 and Apg8 systems are akin to one another, they belong to the most distantly related group to the ‘classical’ UBLs ever described. In addition, these two UBLs might not function as modifiers of target molecules. So there might be more UBL systems carrying out further divergent functions than suspected previously. In mutants of the Apg8 system, such as apg8, apg4 and apg3 cells, the Apg12–Apg5 conjugation is not affected20. Until the step when they form their respective E2 conjugates, the Apg8 and Apg12 UBL reactions proceed independently26. However, in mutants of the Apg12 system, such as apg5, apg10 and apg12 cells, formation of the Apg8–phosphatidylethanolamine complex is severely impaired38, indicating that the concerted function of two UBL systems is necessary for formation of the autophagosome. Where do these conjugations take place? What is the role of the Apg12 system on lipidation of Apg8 and on formation of the autophagosome? And what is the function of the Apg8–phosphatidylethanolamine conjugate? To answer these questions, we must work out the molecular mechanism of membrane formation during autophagy (and in general), and discover why eukaryotes have adopted such sophisticated systems. Links DATABASE LINKS APG16 | APG17 | Apg12 | Apg1 | apg6 | vps30 | Apg7/Cvt2 | Apg10 | Aut7/Apg8/Cvt5 | Aut2/Apg4 | Aut1 | Apg3 | Apg5 | MAP1-LC3 | GABARAP ENCYCLOPEDIA OF LIFE SCIENCES Lysosomal degradation of proteins

cerevisiae. FEBS Lett. 333, 169–174 (1993). Thumm, M. et al. Isolation of autophagocytosis mutants of Saccharomyces cerevisiae. FEBS Lett. 349, 275–280 (1994). Mizushima, N., Noda, T. & Ohsumi, Y. Apg16p is required for the function of the Apg12p–Apg5p conjugate in the yeast autophagy pathway. EMBO J. 18, 3888–3896 (1999). Kamada, Y. et al. Tor-mediated induction of autophagy via an Apg1 protein kinase complex. J. Cell Biol. 150, 1507–1513 (2000). Harding, T. M., Morano, K. A., Scott, S. V. & Klionsky, D. J. Isolation and characterization of yeast mutants in the cytoplasm to vacuole protein targeting pathway. J. Cell Biol. 131, 591–602 (1995). Noda, T. et al. Apg9p/Cvt7p is an integral membrane protein required for transport vesicle formation in the Cvt and autophagy pathways. J. Cell Biol. 148, 465–480 (2000). George, M. D. et al. Apg5p functions in the sequestration step in the cytoplasm-to-vacuole targeting and macroautophagy pathways. Mol. Biol. Cell 11, 969–982 (2000). Yuan, W., Stromhaug, P. E. & Dunn, W. A. Jr Glucoseinduced autophagy of peroxisomes in Pichia pastori requires a unique E1-like protein. Mol. Biol. Cell 10, 1353–1366 (1999). Kim, J., Dalton, V. M., Eggerton, K. P., Scott, S. V. & Klionsky, D. J. Apg7p/Cvt2p is required for the cytoplasm-

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

17.

18.

19.

20.

21.

22.

23.

24.

to-vacuole targeting, macroautophagy, and peroxisome degradation pathways. Mol. Biol. Cell 10, 1337–1351 (1999). Hutchins, M. U., Veenhuis, M. & Klionsky, D. J. Peroxisome degradation in Saccharomyces cerevisiae is dependent on machinery of macroautophagy and the Cvt pathway. J. Cell Sci. 112, 4079–4087 (1999). Muller, O. et al. Autophagic tubes. Vacuolar invaginations involved in lateral membrane sorting and inverse vesicle budding. J. Cell Biol. 151, 519–528 (2000). Suriapranata, I. et al. The breakdown of autophagic vesicles inside the vacuole depends on Aut4p. J. Cell Sci. 113, 4025–4033 (2000). Mizushima, N. et al. A protein conjugation system essential for autophagy. Nature 395, 395–398 (1998). This paper presented the first ubiquitin-like protein conjugation reaction, the Apg12 system, as being essential for autophagy. Tanida, I. et al. Apg7p/Cvt2p: A novel protein-activating enzyme essential for autophagy. Mol. Biol. Cell 10, 1367–1379 (1999). Shintani, T. et al. Apg10p, a novel protein-conjugating enzyme essential for autophagy in yeast. EMBO J. 18, 5234–5241 (1999). Kirisako, T. et al. Formation process of autophagosome is traced with Apg8/Aut7p in yeast. J. Cell Biol. 147, 435–446 (1999). Huang, W. P., Scott, S. V., Kim, J. & Klionsky, D. J. The itinerary of a vesicle component, Aut7p/Cvt5p, terminates

VOLUME 2 | MARCH 2001 | 2 1 5

REVIEWS

25.

26.

27.

28.

29.

30.

31.

in the yeast vacuole via the autophagy/Cvt pathways. J. Biol. Chem. 275, 5845–5851 (2000). Lang, T. et al. Aut2p and Aut7p, two novel microtubuleassociated proteins are essential for delivery of autophagic vesicles to the vacuole. EMBO J. 17, 3597–3607 (1998). 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, 263–276 (2000). This showed that Apg8 associates with the intermediate membranes of the autophagosome by serial modification reactions. Ichimura, Y. et al. A ubiquitin-like system mediates protein lipidation. Nature 408, 489–493 (2000). This paper described the Apg8 lipidation system, in which the processed form of Apg8 is activated by a ubiquitin-like system and finally forms a conjugate with the membrane phospholipid, phosphatidylethanolamine. Mizushima, N., Sugita, H., Yoshimori, T. & Ohsumi, Y. A new protein conjugation system in human. The counterpart of the yeast Apg12p conjugation system essential for autophagy. J. Biol. Chem. 273, 33889–33892 (1998). Mizushima, N. et al. Dissection of autophagosome formation using Apg5-deficient mouse embryonic stem cells. J.Cell Biol. 152, 657–667 (2001). Mann, S. S. & Hammarback, J. A. Molecular characterization of light chain 3. A microtubule binding subunit of MAP1A and MAP1B. J. Biol. Chem. 269, 11492–11497 (1994). Kabeya, Y. et al. LC3, a mammalian homologue of yeast

32.

33.

34.

35.

36.

37.

38.

Apg8p, is localized in autophagosome membranes after processing. EMBO J. 19, 5720–5728 (2000). Two UBL systems, Apg12 and Apg8, are conserved in higher eukaryotes and this is the first report of the functional homologue of Apg8 in mammals. Meyers, G., Stoll, D. & Gunn, M. Insertion of a sequence encoding light chain 3 of microtubule-associated proteins 1A and 1B in a pestivirus genome: connection with virus cytopathogenicity and induction of lethal disease in cattle. J. Virol. 72, 4139–4148 (1998). Sagiv, Y., Legesse-Miller, A., Porat, A. & Elazar, Z. GATE16, a membrane transport modulator, interacts with NSF and the Golgi v-SNARE GOS-28. EMBO J. 19, 1494–1504 (2000). Legesse-Miller, A., Sagiv, Y., Glozman, R. & Elazar, Z. Aut7p, a soluble autophagic factor, participates in multiple membrane trafficking processes. J. Biol. Chem. 275, 32966–32973 (2000). Chen, L., Wang, H., Vicini, S. & Olsen, R. W. The γaminobutyric acid type A (GABAA) receptor-associated protein (GABARAP) promotes GABAA receptor clustering and modulates the channel kinetics. Proc. Natl Acad. Sci. USA 97, 11557–11562 (2000). Wang, H., Bedford, F. K., Brandon, N. J., Moss, S. J. & Olsen, R. W. GABAA-receptor-associated protein links GABAA receptors and the cytoskeleton. Nature 397, 69–72 (1999). Paz, Y., Elazar, Z. & Fass, D. Structure of GATE-16, membrane transport modulator and mammalian ortholog of autophagocytosis factor Aut7p. J. Biol. Chem. 275, 25445–25450 (2000). Suzuki, K., Kirisako, T. & Ohsumi, Y. Localization of Apg5p and Apg8p/Aut7p reveals functional classes of APG

CORRECTIONS

genes and their interactions in autophagy. (submitted). 39. Wendland, B., Emr, S. D. & Riezman, H. Protein traffic in the yeast endocytic and vacuolar protein sorting pathways. Curr. Opin. Cell Biol. 10, 513–522 (1998). 40. Tuttle, D. L. & Dunn, W. A. Divergent modes of autophagy in the methylotrophic yeast Pichia pastoris. J. Cell Sci. 108, 25–35 (1995). 41. Huang, P. H. & Chiang, H. L. Identification of novel vesicles in the cytosol to vacuole protein degradation pathway. J. Cell Biol. 136, 803–810 (1997). 42. Cuervo, A. M. & Dice, J. F. Age-related decline in chaperone-mediated autophagy. J. Biol. Chem. 275, 31505–31513 (2000). 43. Scott, S. V., Baba, M., Ohsumi, Y. & Klionsky, D. J. Aminopeptidase I is targeted to the vacuole by a nonclassical vesicular mechanism. J. Cell Biol. 138, 37–44 (1997). 44. Baba, M., Osumi, M., Scott, S. V., Klionsky, D. J. & Ohsumi, Y. Two distinct pathways for targeting proteins from the cytoplasm to the vacuole/lysosome. J. Cell Biol. 139, 1687–1695 (1997). 45. Harding, T. M., Hefner-Gravink, A., Thumm, M. & Klionsky, D. J. Genetic and phenotypic overlap between autophagy and the cytoplasm to vacuole protein targeting pathway. J. Biol. Chem. 271, 17621–17624 (1996). 46. Scott, S. V. et al. Cytoplasm-to-vacuole targeting and autophagy employ the same machinery to deliver proteins to the yeast vacuole. Proc. Natl Acad. Sci. USA 93, 12304–12308 (1996).

Acknowledgements Thanks to T. Noda and N. Mizushima for critical discussion and reading of this manuscript.

ERRATUM

LIPID RAFTS AND SIGNAL TRANSDUCTION

THREE WAYS TO MAKE A VESICLE Kirchhausen, T.

RAB PROTEINS AS MEMBRANE ORGANIZERS

Simons, K. & Toomre, D.

Nature Reviews Molecular Cell Biology 1, 187–198 (2000).

Zerial, M. & McBride, H.

Nature Reviews Molecular Cell Biology 1, 31–39 (2000).

Some entries in Table 1 contained mistakes: Sec12p is a GEF (not a GAP) for Sar1p. β-COP (not β′-COP) has weak sequence identity to β-AP. ζ-COP (not ε-COP) has weak sequence identity to σ-AP. The online version of this Review has been corrected.

Nature Reviews Molecular Cell Biology 2, 107–117 (2001).

The last two sentences of the legend to Figure 1 should have read: 5 | One possible way of downregulating the signal may occur by binding of the cytosolic kinase Csk to the raft-associated protein CBP. Csk may then inactivate the Src-family kinases through phosphorylation57. The online version of this Review has been corrected.

216

The second sentence of the last paragraph on page 113 contained an error and should have read: Sec1p and related proteins, Vps33p and Vps45p, regulate SNARE pairing by sequestering syntaxin molecules82,83. The online version of this Review has been corrected.

www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

PERSPECTIVES factors, receptors and a vast network of signalling pathways.

TIMELINE

Biochemistry and molecular biology teaching over the past 50 years Edward J. Wood The revolution in molecular biology that began nearly 50 years ago has had an enormous impact in the lab, but it also triggered a quieter revolution in undergraduate teaching. How has it affected the content and teaching methodology of undergraduate courses, and is the training that students now receive in the molecular life sciences appropriate to their future career paths?

Our understanding of how living systems work has grown enormously during the past 150 years and shows no sign of diminishing. In this time-frame, biology has evolved from a descriptive to an analytical subject. The school-leaver intent on following a career in molecular or cellular biology typically begins with a degree in biochemistry or a related subject. How, since the birth of the undergraduate degree in biochemistry, have we equipped such students with the skills to tackle their rapidly evolving subject? Asking this question is worthwhile because, paradoxically, the increasing mass of information that biochemistry students now face might be hindering their future ability to analyse that information. Considering the past might help us to shape the future of education in the molecular life sciences. The expansion of knowledge

Biochemical education, like lichen, consists of two co-evolving organisms — the subject itself and teaching methodology. The early 1950s saw biochemistry give birth to molecular biology. James Watson and Francis Crick’s

paper on the molecular structure of nucleic acids was published1, and Fred Sanger completed the sequence of insulin2,3. Little was known of the three-dimensional structures of proteins (although Linus Pauling and Robert Corey had proposed the α-helix as a structure for peptides in the early 1950s) (REF. 4), and DNA sequencing came much later (1977 (REF. 5)). At about the same time, the methods for electron microscopy were being perfected, ribonucleoprotein particles (later called ribosomes) were shown to be the site of protein synthesis, and the sequence of metabolites in the pentose phosphate pathway was sorted out; Albert Lehninger had shown that oxidative phosphorylation was coupled to electron transport in the electron transport chain6, and Daniel Arnon and coworkers discovered photosynthetic phosphorylation7. Anyone teaching a course or writing a textbook at that time had a great deal to take on board: so many new things clamoured to be taught to students. Later discoveries have caused further paradigm shifts, with consequences for teachers and textbook writers, not to mention students. One could mention the extensive advances in X-ray crystallography and computing that have given us the threedimensional structures of thousands of proteins, the technical advances that allowed genes to be cloned and DNA to be sequenced (often more easily than proteins) and the polymerase chain reaction, which changed the way that much of molecular biology was done. Where there were hormones, we now have a plethora of growth

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

Evolving teaching methods

Biochemistry courses have been a feature of the medical curriculum since early in the twentieth century, but began to be offered as an independent subject for honours degrees in the late 1940s. In the preface to the first edition of Dynamic Aspects of Biochemistry (reprinted in REF. 8), Ernest Baldwin mentions that biochemistry with an emphasis on clinical problems (‘chemical physiology’) was taught to medical students, whereas ‘biochemistry’ was taught as an independent discipline. When he published Dynamic Aspects in 1946, he believed that there was no suitable textbook for the second course. Early biochemistry students therefore relied a great deal on teaching staff to guide them in their studies. Teaching in the 1950s — indeed right through to the 1980s — mostly involved lectures, practical classes and some teaching in small groups (seminars or tutorials). How have teaching methods changed or matured over the years? What’s the use of lectures? Lectures are an accepted way of presenting information, but whether the students have received and understood this information is another matter. As universities expanded from the 1960s onwards, classes became much larger, and lectures were perceived to be an efficient and economical way of teaching to the masses. However, educationalists such as Donald Bligh, author of the widely read book What’s the Use of Lectures9, were sceptical and had less faith in the value of lectures. In universities, teaching deals with: first, the transfer/acquisition of information; second, the promotion of thought; third, changes in attitudes; and last, behavioural skills. Bligh’s examination of the evidence led him to conclude that the lecture may be used appropriately to convey information, but could not be used effectively on its own to promote thought, or to change or develop attitudes, without variation of the

VOLUME 2 | MARCH 2001 | 2 1 7

PERSPECTIVES

Timeline | How biochemistry has been taught in the United Kingdom from 1950 to 2000

• Few universities offer a biochemistry course. • Lectures in small groups (perhaps 10 –15 students in final year). • Chemistry lectures and practical classes form a significant part of the course (perhaps one-third), as well as other courses such as physiology.

1950

• More universities offer biochemistry-related degree courses and classes are growing. • Less chemistry required in many university courses. • Some universities offer joint honours in two subjects. • Much examining done by multiple-choice tests (computer-marked) to handle larger classes. • Some universities offer time in industry as part of the course. • Some films available to illustrate and teach. • Contribution of information technology almost zero.

1960

1970

• Teaching does not evolve much although there is a rapid increase in knowledge and understanding of biological phenomena — DNA and protein primary structure, and membranes. • Training remains largely academic with little awareness of the needs of industry.

usual lecturing technique. His book examines the evidence and suggests changes in the method, or additional methods, to try to achieve objectives other than information transfer. Lectures have an important role but a much more limited one than they are often credited with, and indeed other methods can be used to transfer information just as well. A key aspect is to make sure that students are actively learning during lectures, rather than passively taking notes. There are ways of keeping the lecture interactive (although this can be difficult with large classes), but it might be more appropriate to try to think of other ways of teaching and engendering learning as alternatives to the lecture. Have things changed since the 1960s? Unfortunately the pressure is on to teach more students with a diminishing ‘unit of resource’ (the money per student that universities in the United Kingdom receive from the government for teaching), and researchers may even be encouraged not to ‘waste’ too much time on teaching. However, some things have certainly changed for the better: lecturers are now encouraged to state the objectives of their lectures, and the widespread use of computers to design slides and produce lecture notes helps to make presentations more effective as learning experiences. Although the traditional lecture still forms the basis of most undergraduate courses, new teaching methods have been devised, including group work (such as designing a poster together and problem-based sessions), reflection on how learning took place, peer marking and peer assessment, as well as the use of computers. Although these help to deal with diminishing staff-to-student ratios, they have

218

• • • • • •

Modularization of courses. Information technology and other transferable skills explicitly taught. Many more universities offer a year out in industry or abroad. Lectures in large groups (typically 100 in final year). Token number of chemistry lectures and often fewer practical classes. Wide choice of courses to complement core biochemistry courses, often including business studies and computing. • More specialized degrees offered (for example, medical biochemistry, plant biochemistry) as a solution to cope with the widening knowledge base. • Many universities run fewer practical classes, owing to the constraints of modular courses and increasing costs as class sizes rise. • Most courses have a final-year research project (one or two terms). • Some computer-based teaching (typically simulation of experiments).

1980

1990

• Major advances in the subject; molecular biology now forms a major part of the course. • This poses the problem of ‘what to leave out’; content that was previously a key part of the course has to be dropped. • Computer-based learning is introduced.

other benefits too: for example, collaborative projects teach students how most people work in the ‘real world’, give students responsibility, and encourage active learning. From glass pipette to Gilson. Training and practice in the lab, as well as learning the logic of experimental design and data interpretation, are vital for biochemistry undergraduates to develop appropriate skills. Students begin with simple experiments that familiarize them with equipment and the nature of biological materials. They then move on to planning experiments, doing them and interpreting their own data, and in most universities this culminates in an extended final-year research project. During the course, they will have used spectrophotometers, pH meters, electrophoresis appara-

Figure 1 | Frame from the program ‘Protein Purification’ in which students in a virtual lab attempt a purification procedure. This frame deals with gel filtration and students specify which fractions they wish to pool. They can test for enzyme activity and carry out one- and twodimensional gel electrophoresis. Reproduced by permission of Dr A. G. Booth, University of Leeds.

tus, radioactivity and plunger pipettes. By contrast, 50 years ago practical classes would have involved simple tests for carbohydrates, proteins and fats in biological materials, some enzyme assays, and later in the course the pleasures of Warburg manometry. In recent years, the rising costs that are associated with increased class sizes have driven many universities to cut down the number of practical classes but, on the positive side, many students — all of them in some universities — are encouraged to spend a whole year working in a laboratory outside the university, perhaps in an industrial or clinical context. There are also far more opportunities for undergraduates to do practical projects abroad. Practical classes put students under a great deal of stress: working with unfamiliar equipment under time constraints that prohibit repetition of experiments, and perhaps with an unfamiliar practical partner, hinder the production of reliable data. Small wonder, then, that understanding what they are doing and the biochemical principles involved takes second place to actually getting the experiment done in the session. How can our enlightened attitude to teaching be applied to practical classes? Although classes are now much larger, and individual attention by academic staff is less likely to be available, the processes of data acquisition and data-processing (on computer-generated data sets, for example) can now be separated so that these skills can be learned independently. It is now possible to repeat and design ‘experiments’ at leisure using computer simulations10 (FIG. 1). But it is important to realize that computer simulations are a useful addition tο — not a www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

PERSPECTIVES

Box 1 | Problem-based learning Problem-based learning has been used widely in medical courses, less so in biochemistry. A problem is posed to a small group of students with a tutor (possibly a non-expert, called a facilitator) and then the task is divided up. Each member of the group has to go away, find information and then communicate it to the rest of the group. This develops the important skills of searching for and presenting information, building students’ confidence in handling information. The process is reiterated until a solution to the problem can be arrived at. The experience is that this is effective, but takes even more staff time plus initial training (and cajoling) of staff to act as facilitators rather than teachers17. Academic staff worry that there does not seem to be any clear curriculum and that students might finish up knowing different things. But the most important feature of this method is that it accepts that it is impossible for any one person to remember enough of the vast body of knowledge that constitutes modern biochemistry and molecular biology to solve the problems that they might encounter in their work, and also that knowledge has no intrinsic value if it cannot be used (for example, to solve problems).

replacement for — practical experience. Face to face. Small-group teaching in the form of tutorials or seminars is an important educational tool but, of necessity, is falling out of fashion to some extent because it is labour-intensive of teaching staff. Tutorials started out as one-to-one sessions in many universities, but now tend to be one-to-ten or more. There is no doubt that such classes offer effective learning environments: students learn actively, having to explain concepts in their own words. Seminars offer presentations to small groups in an environment that allows students to ask and answer questions. But, despite their educational value, seminars tend to be perceived by academic staff and institutions as an inefficient method of teaching, tying up many hours of staff time. As small-group teaching becomes less feasible, are there other ways of encouraging active learning? One modern alternative is the computer tutorial (see below). Undergraduates doing final-year research projects also get quite a lot of one-to-one attention from their supervisors and research staff. Finally, problem-based learning encourages group work and gives students responsibility for their own learning (BOX 1). Textbooks

The textbook is the mainstay of academic courses, and many textbooks developed out of courses taught by research-active biochemists in their own universities. Apart from showing how attitudes to presentation have changed (BOX 2), old textbooks encapsulate the content of past biochemistry courses. A typical textbook of the late 1950s (REF. 11) started with the chemistry of the building blocks of life (carbohydrates, lipids, amino acids and proteins, nucleic acids and nucleoproteins) before moving on to enzymes and metabolism. Physiological chemistry — dealing with body fluids, specialized tissues, endocrinology and

nutrition — were likely to be given more prominence than genes and proteins. Textbooks took a while to catch on to the emergence of molecular biology. For example, the fourth edition (published in 1963 — ten years after Watson and Crick’s seminal discovery) of Baldwin’s Dynamic Aspects of

Biochemistry 8, like several other books of that time, speaks of the ‘hypothesis’ proposed by Watson and Crick that DNA is a double helix with complementary base pairs. It is clear that the author understands that this structure can explain replication, and indeed he goes on to explain the Meselson and Stahl experiment (FIG. 2). However, he seems to miss the point that this experiment resulted in a marked change in our understanding of genetics. In his 1965 book Molecular Biology of the Gene12, James Watson stated that most introductory biology courses already included a section on molecular biology but that he expected to see this section expand to a point where it was central in the planning of the rest of a biology programme, a prediction that did eventually come true.With the acceptance of the double-helical structure for DNA came a different attitude to how biochemistry textbooks (and courses) should be organized. For example, the authors of one 1958 textbook13 remarked that “during the five years since the preparation of

Box 2 | The evolution of textbook design 1950 Linear text, monochrome, few diagrams and photographs (for example, histology, electronmicrographs), no cell structures, some tables and graphs, almost no molecular structures of macromolecules (other than structural formulae of starch and glycogen). 1960 Increased length to cope with the expansion of biochemical knowledge. Design becoming more imaginative. Much more thought put into figures; many more photographs. Mahler and Cordes18 was a key advance in 1966: thorough, highly detailed, maths and chemistry not downgraded; good coverage of techniques. Irving Geis’s drawings of haemoglobin molecule. 1970 The first edition of Lehninger’s Biochemistry appears in 1970 (REF. 6). Two colours used, many diagrams and photographs, a few three-dimensional protein structures (myoglobin, haemoglobin, cytochrome c) but not detailed. Authors try to express the logic of biochemistry itself and also use their own logic in the way that they present the subject — to help students’ learning. Stryer’s full-colour Biochemistry appears in 1975 (REF. 19). Chapter-end problems now a common feature. 1980 Increasing imagination put into textbook design; second editions appear frequently, necessitated by the rapid development of biochemical knowledge. The first major molecular cell biology books appear (for example, Alberts et al. Molecular Biology of the Cell in 1983 (REF. 20)). Some authors decide that DNA and the gene should come first and traditional format of chemistry of biomolecules, metabolism, biochemical genetics starts to change. Less linear presentation with text boxes and other in-text features. 1990 Molecular cell biology texts (which actually have a great deal of biochemistry in them) proliferate. Textbooks begin to include supplementary material such as CD-ROMs, problems, instructors texts, study guides, OHP transparencies. Towards the end of the decade, web sites become available. 2000? See “the textbook is dead”21.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 2 1 9

PERSPECTIVES the first edition (1953), the rapid development of many areas of biochemistry rendered much of what had been written obsolete”. In this book, amino acids and proteins are the front runners, followed by nucleoproteins (and the Watson and Crick double helix), then came enzymes, oxidation and reduction, and metabolism, with the chemistry of the carbohydrates tucked in. Not surprisingly, the pace of updates has increased: contrast the speed at which PCR was incorporated into new textbooks in the 1990s with the time that it took to incorporate the double helix.

Chemistry and maths

In the 1950s, a substantial fraction of any biochemistry course — including the associated practical classes — was chemistry. In the United Kingdom and some other countries, this requirement for chemistry has gradually been pushed aside by teaching topics that are perceived to be more exciting and up to date, such as genes, proteins and pathways, it being assumed that sufficient chemistry and maths are being taught in high school. In the United States and many European countries this has not been the

pH 7.0

pH 12.0

case, the strong chemistry and maths component being retained in university biochemistry courses. This change in the United Kingdom was not entirely due to the explosion of information in the molecular biosciences. Apart from wanting to know about all the recent advances, biochemistry students felt that the chemistry they were being taught was not particularly relevant to them, and their disillusionment with the subject caused performance to diminish. Departments took students’ complaints about chemistry to heart and reduced the proportion of chemistry taught. But this poses a fundamental problem. A principal aim of biochemistry is to explain biological phenomena at the molecular level, in other words to provide a chemical explanation for life. How can someone with only a minimal grasp of chemistry hope to achieve this aim? There are similar concerns about maths and there is now the perception that the level of mathematical skills that students gain from high school is diminishing. Consequently, teaching about quantitative aspects of biochemistry becomes more difficult14. Another side to this problem is that many students other than career-track biochemists now take courses or modules in biochemistry. These students often deliberately take a ‘biological’ subject to avoid chemistry and maths which they perceive as ‘difficult’ subjects. Such students are in an even worse position when it comes to understanding the chemistry of life from a quantitative standpoint. Inevitably, for those teaching the courses, there is pressure to ‘dumb down’ these aspects.

Relative DNA concentration

What have computers done for us? 2

1.710

1.717

1.724

1.760

1.767

1.774

Density (g/cc)

Figure 2 | The changing face of textbooks and the diagrams in them. a | The Meselson and Stahl experiment according to Baldwin, Dynamic Aspects of Biochemistry fourth edition (1963) © Cambridge Univ. Press, and b | according to Mathews and Van Holde (1996), Biochemistry, second edition, © Benjamin Cummings, Redwood City, California.

220

No one can deny that computers have had an enormous effect on the way that we conduct research in modern-day biochemistry, but have they made similar waves in teaching? The question of staff time and pressure makes computer-aided learning seem attractive, and indeed computers have been used for marking multiple-choice tests since the 1970s, although their serious use in teaching, as opposed to assessment, started later. Many early teaching programs were really no better than crude textbooks combined with test questions. This way of teaching was also actually more expensive than textbooks, both for the material (computer versus book), and for the time and labour spent by staff in producing the programs. The commercialization of computerbased learning has improved it beyond all recognition, allowing students to educate themselves using tutorials and the World Wide Web (for an example, see the most www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

PERSPECTIVES recent addition of Lehninger’s Principles of Biochemistry 15 online). The advantage of (good) teaching programs is that the computer never tires, items can be taken at an individual student’s own pace, with reiteration if necessary, and their interactive nature keeps students learning actively, allowing them to check their progress. During the past few years, the Internet has grown in prominence as an educational medium. Students have become very adept at finding information on the Internet (although they are not quite so good at evaluating its quality). Teachers need to spend time in instructing students how to use the Internet constructively 16. ‘Webbased learning’ is now assuming great importance as a learning and teaching tool and one can only look towards this mode of learning increasing in the future.

Links FURTHER INFORMATION Biochemistry in 3D

— Lehninger’s Principles of Biochemistry | The textbook is dead | A. G. Booth’s protein purification program

9. 10.

11. 12.

1. 2.

6. 7. 8.

Watson, J. D. & Crick, F. C. Molecular structure of nucleic acids. Nature 171, 737–738 (1953). Sanger, F. & Tuppy, H. The amino-acid sequence of the phenylalanyl chain of insulin. Biochem. J. 49, 463–490 (1951). Sanger, F. & Thompson, E. O. P. The amino-acid sequence of the glycyl chain of insulin. Biochem. J. 53, 353−374 (1953). Pauling, L., Corey, R. B. & Branson, H. R. The structure of proteins: two hydrogen-bonded helical configurations of the polypeptide chain. Proc. Natl Acad. Sci. USA 37, 205−211 (1951). Sanger, F., Nicklen, S. & Coulson, A. R. DNA sequencing with chain-terminating inhibitors. Proc. Natl Acad. Sci. USA 74, 5463−5467 (1977). Lehninger, A. L. Biochemistry (Worth, New York, 1970). Arnon, D. I. Conversion of light into chemical energy in photosynthesis. Nature 184, 10−21 (1959). Baldwin, E. Dynamic Aspects of Biochemistry (Cambridge Univ. Press, Cambridge, 1946).

13. 14. 15. 16. 17. 18. 19. 20. 21.

Bligh, D. What’s the Use of Lectures? 2nd edn (Intellect Books, Exeter, 1998). Booth, A. G. & Wood, E. J. in Progress in Medicinal Chemistry Vol. 26 (eds Ellis, G. P. & West, G. B.) 323− 355 (Elsevier, Amsterdam, 1989). White, A., Handler, P. & Smith, E. L. Principles of Biochemistry 3rd edn (McGraw–Hill, New York, 1964). Watson, J. D. Molecular Biology of the Gene (Benjamin Cummings, New York, 1965). Fruton, J. S. & Simmonds, S. General Biochemistry 2nd edn (Wiley, New York, 1958). Dawes, E. A. Quantitative Problems in Biochemistry (Livingstone, Edinburgh, 1956). Nelson, D. L. & Cox, M. M. Lehninger’s Principles of Biochemistry (Worth, New York, 2000). Clegg, B. Mining the INTERNET (Kogan Page, London, 1999). Boud, D. & Felletti, G. (eds) The Challenge of Problembased Learning 2nd edn (Kogan Page, London, 1997). Mahler, H. R. & Cordes, E. H. Biological Chemistry (Harper and Row, New York, 1966). Stryer, L. Biochemistry (W. H. Freeman, New York, 1975). Alberts, B. et al. Molecular Biology of the Cell (Garland, New York, 1983). Kimball, J. ‘The textbook is dead’ or at least will be replaced within 5 years by the virtual Internet textbook. Science 291, 119 (2001). Also see www.ultranet.com/~jkimball/BiologyPages

Adapt or die

So where are we today? At face value, it is easy for university lecturers and instructors to bemoan the combination of increasing class sizes and diminishing resources. The challenge is to adopt teaching methods that not only allow teachers and students to cope with these pressures, but that are also better educationally: such methods already exist and are continuing to be developed (see the article by Ellis Bell on page 221 of this issue). We cannot continue to do things as we did them 30 years ago and, as biology tells us, if we fail to adapt we will not survive. Teaching will change from dealing with individuals to dealing with groups, we will teach transferable skills (how to find information and communicate it) as well as facts, and education will be student led rather than teacher led. Even though the number of biochemistry graduates continues to rise (Leeds University produced about 10 in 1958 compared with nearly 100 next summer), few will become academics, which was a strong career path in the past. Some will go into industry (normally after a Ph.D.), but many more will work in hospitals, become sales representatives, or pursue careers quite outside science, such as accountancy, patent law and banking. To match the changing fates of graduate biochemists, we need an explicit change of attitude: the degree will not be seen as the end of an education, but as the beginning, a preparation for life-long learning. Edward. J. Wood is at the School of Biochemistry & Molecular Biology, University of Leeds, Leeds LS2 9JT, UK e-mail: e.j.wood@leeds.ac.uk

OPINION

The future of education in the molecular life sciences Ellis Bell The changing landscape of education in biochemistry and molecular biology presents many challenges for the future, for students and educators alike. The exponential increase in knowledge, the genomics, proteomics and computing revolutions, and the merging of once separate fields in biology, chemistry, physics and mathematics, mean that we need to rethink how we should be preparing today’s science undergraduates for the future. What do we need to change, and how will we implement it?

When most of the teachers of today were learning their trade, none of these factors was a force in what we now call biochemistry and molecular biology, or a part of the education process. Education focused on knowledge, and an understanding of how knowledge had been obtained. No one thought of purifying and studying proteins for which the structure and function were unknown. In the world of education, lecturers lectured, techniques were taught in labs and students studied textbooks and occasionally original literature, but usually with the aim of learning facts. And, largely,

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

we are still in this situation today. Problembased learning1,2 has made inroads at medical schools, and is clearly part of the solution, but it is often still taught from the standpoint of assimilating and interpreting facts, and is commonly taught in special classes rather than being integrated throughout the curriculum. Exams continue to test factual knowledge, and this has, in many ways, contributed to the problems that we now face. This is summarized in the ‘quantum theory of education’ (FIG. 1): students learn quanta of information and are tested on their ability to recall these quanta accurately. A student that can emit the quantum of information successfully and accurately is rewarded with a good grade and moves on. The problem, as any atom that has emitted a quantum particle knows, is that once the particular particle is emitted it is gone forever until another quantum of the same kind collides with the atom. Current examination systems are geared towards testing the ability to emit quanta accurately because it is efficient and easy to measure. But is it the best way, either to examine or to encourage learning? Part of the purpose of education in biochemistry and mole-

VOLUME 2 | MARCH 2001 | 2 2 1

Links FURTHER INFORMATION Biochemistry in 3D

— Lehninger’s Principles of Biochemistry | The textbook is dead | A. G. Booth’s protein purification program

9. 10.

11. 12.

1. 2.

6. 7. 8.

13. 14. 15. 16. 17. 18. 19. 20. 21.

Adapt or die

OPINION

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 2 2 1

PERSPECTIVES

Professor instructs . . .

Student memorizes . . .

Student answers correctly on the exam . . .

Student forgets.

Figure 1 | The quantum theory of education. Courtesy of Emily Hadland.

cular biology is to train the next generation of scientists. Today’s students will bring about future revolutions in our understanding. While we can and should teach them about past revolutions, how do we prepare them to both ask and answer the questions that will take us to the next level of understanding of biochemistry and molecular biology? Furthermore, today’s students face an increasingly diverse job market3. How do we prepare them for those choices? Tackling the expansion of knowledge

There is now too much information around for anyone to know and memorize: the teacher does not know it all and the student certainly cannot be expected to remember it all. As educators, we should ask: what are the qualities that we need future molecular life scientists to have? By identifying these (BOX 1) we can then begin to think about how these qualities can be instilled into future undergraduates of biochemistry and molecular biology. A related issue is whether or not there is a necessary core of material that a student should be expected to know (BOX 2). Do students really need to know the structure of prostaglandin E2, the formal kinetic mechanism of glutamate dehydrogenase, the specificity of EcoR1 or the roles of the Grim and Reaper proteins in apoptosis, so long as they know how to access this information if they need it? An integrated education. Many of the characteristics of an ideal student are encompassed in the ‘research paradigm of teaching’. Likewise, many aspects of an integrated education go hand in hand with the research paradigm. Rather than just involving students in research at some point of their undergraduate career, the research paradigm fully integrates a research mindset from day one of an undergraduate education. An understanding

222

of the research paradigm of teaching can be best approached with a quote from Paul Saltman: “One can be no better a teacher than one is a learner at any point in one’s life”4. Teachers must, therefore, be active researchers. Research itself can be used as a teaching tool but all aspects of the research enterprise need to be integrated into the educational process. For example, all too frequently laboratory courses are separate from lecture material: this does not help students to integrate related issues of a course. Integration in the education of a student should be at all levels: within a course, between courses, within a degree programme and even between degree programmes. Presenting material from a variety of perspectives caters to the fact that not all individuals learn in the same way, but achieving this level of integration often requires the cooperation of many departments and programmes.

What educational strategies should be included in such an approach to education? In short, everything: lectures, laboratories, problem sets, self study, research, seminars, internships, reading the original literature, writing and, perhaps most importantly, increasing the involvement of the student in the process of education. This not only is the basis for life-long learning but also helps to give the student ownership of the process. It is also more satisfying for teachers: the best students to teach are those who are directly involved in their own education. Changing the examination system. Courses should move away from having too many inclass tests (another way the quantum theory of education is subversively supported) to using other methods of assessing the progress of students, such as laboratory reports that mimic original research papers, research reports and presentations, literature reports and topic papers. Students should have more opportunity to show proficiency at those things (BOX 1) that, as educators, we think they will need to make progress in the future. Merging disciplines

So what are the basics that students need to know and understand if they are to be biochemistry and molecular biology specialists of the future (BOX 2)? Let me dispel the idea that biochemistry and molecular biology are in any way separate disciplines: both areas ask questions about structure–function relationships at the molecular and atomic level and also at the cellular/intracellular level. The names of the main professional societies, such as the International Union of

Box 1 | The ideal education in biochemistry In an ideal world, biochemistry students should have the following skills by the time they have finished their undergraduate course: • Understanding of the fundamentals of chemistry and biology and the key principles of biochemistry and molecular biology. • Awareness of the major issues at the forefront of the discipline. • Ability to assess primary papers critically. • Good ‘quantitative’ skills, such as the ability to prepare reagents accurately and reproducibly for experiments. • Ability to dissect a problem into its key features. • Ability to design experiments and understand the limitations of what the experimental approach can and cannot tell you. • Ability to interpret experimental data and identify consistent and inconsistent components. • Ability to design follow-up experiments. • Ability to work safely and effectively in a laboratory. • Awareness of the available resources and how to use them, including the ability to collaborate with other researchers. • Ability to think in an integrated manner and look at problems from different perspectives.

www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

PERSPECTIVES

Box 2 | What do budding biochemists need to understand? • Fundamentals of macromolecular structure and function; how to determine macromolecular structure. • Modes of macromolecule–ligand interactions and approaches to study them. • Basic concepts of biocatalysis (including ribozymes and abzymes). • Enzyme kinetics and inhibition. • Basic mechanisms of the control of cellular processes and reactions. • Techniques of macromolecular isolation and characterization. • How to use sequence and structural databases. • Fundamentals of computational chemistry and molecular modelling. • The uses of techniques such as cloning, polymerase chain reaction, site-directed mutagenesis, phage display and array technology. • Cellular structure and compartmentalization including the functions of subcellular organelles. • Basic mechanisms of cell division and macromolecular synthesis. • Basics of tissue development and differentiation. • Basics of genetic disease. • Fundamental differences between prokaryotes and eukaryotes; molecular evolution. • Fundamentals of cellular and molecular immunology, and the fundamental lifestyles of pathogens such as bacteria and viruses.

Biochemistry and Molecular Biology, and the American Society for Biochemistry and Molecular Biology testify to this merger of disciplines. Furthermore, subjects such as cell biology, immunology and biophysics all fall under this interdisciplinary umbrella. These shifts need to be recognized at the undergraduate level: if students are to be well prepared to take on the challenges of research in the future they must approach problems from an interdisciplinary perspective. Perhaps most importantly of all, students in the molecular life sciences should be exposed to the principal issues of the subject and its applications, such as those related to genetically modified organisms, the BSE crisis, gene therapy and bioremediation. Molecular approaches are increasingly being applied in these areas but the breakthroughs of the twenty-first century will undoubtedly come from our understanding of the molecular basis and mechanisms of behaviour, communication (at all levels), learning and adaptation. Unless students are exposed to the important issues of these areas, it will be hard for them to contribute meaningfully to this area of progress. So how will this material be taught in the future? One of the keys to helping students integrate the basics of chemistry, physics and biology will be to either teach these subjects in a more integrated, problem-solving manner or ensure that once stand-alone, independent courses now can work together. This should be done from the most introductory courses through to the end of the curriculum4. Students need to see this integration right from the start otherwise ‘integration’ becomes

one more quantum that they learn. This does not mean that we should abandon teaching the fundamentals of chemistry, biology and physics in favour of the applications — far from it. Students should be constantly reminded of the fundamentals. In current courses, these are the first quanta transmitted to the students and the first quanta emitted by the students. In the courses of the future, students need to have the fundamentals constantly reinforced. To this end, the revolution in computer speeds and access will help. A group of scientists on both sides of the Atlantic are working towards the creation of a web site (see link online to Biochemistry Explorer to find out more about the site or to participate in its creation) that will act primarily as a teaching resource. Consider a

metabolic map in which every molecule is linked to information from its basic electronic structure to its chemical reactivity and biophysical behaviour, to its roles in metabolism, to key information about structure–function relationships. For macromolecules, there could be links to existing sequence and structural databases so that students would have virtually instant access to information. Education could then focus on teaching students how to ask appropriate questions, how to design experiments and interpret data and how to use the available resources rather than focusing on how to remember facts. Courses could also be more connected to each other: many courses lead to natural overlap and can successfully be interfaced at the level of lab work, problem sets and discussions to reinforce basic concepts. For example, integrating introductory biochemistry and physical chemistry really helps to emphasize the fundamental concepts behind structure–function relationships in macromolecules. Integrating elements of organic chemistry, physical chemistry and enzyme chemistry makes obvious sense as well, by helping students to think about enzymecatalysed reactions at the level of which electrons are involved in making and breaking bonds, and the kinetics of the steps. A taste of real research

Undergraduate courses should have investigative labs with realistic expectations about student involvement in experimental design, data analysis and data interpretation. Labs should not have expected or obvious outcomes so that students have to think about the reliability of the data and the possible explanations rather than just the one they know to expect. Student groups (FIG. 2) in labs could collect different parts of an

Figure 2 | Undergraduates getting a taste of real research.

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 2 2 3

PERSPECTIVES Educating the public

Box 3 | The HIV protease problem set Starting from a single HIV protease sequence, students can use web-based resources to search for homologous sequences and either align them or be provided with the alignments of all available HIV protease sequences and the three-dimensional structure of one HIV protease (see the picture). Give them the sequence of pepsin and ask them to identify its active site on the basis of what they have learned from the HIV protease. Ask them to think of the sorts of questions that this information can be used to provide insight into. Ask them to think about the role of each amino acid in the protein and how they might experimentally test these hypothesized roles. Ask them to do literature searches for drug-resistant mutations and map those onto the sequence and structure, to help familiarize them with databases and algorithms for sequence alignment. Ask them to think about the relevance of protein folding and function, and molecular evolution of the protease. This introduces them to the basics of enzyme kinetics and inhibitors, drugresistant mutants and the effect that a single enzyme-catalysed step has in the life of an organism. Ask them to examine the sequences between the protease sequence and the surrounding peptides. See what they make of the cleavage-site specificity of the protease. (See link online to the Protease Problem Set.)

overall data set so that they must work together on design and analysis/interpretation: this way, students learn to collaborate, and take other data and interpret it; they also learn to see how their data set fits into a bigger picture. In other words, they learn how research really operates. Using electronic resources

Students should be able to use the vast resources that the genomics revolution and computational speeds have made available. How can we integrate the use of such resources into the undergraduate education system? Think of the principles of biology and chemistry that are involved when one asks a student to examine a series of sequence alignments and asks questions about structure–function relationships of the protein. This is easily illustrated by the example of the HIV protease problem set (BOX 3). This problem can be used to teach the basic elements of biochemistry and molecular biology to nonscience students and to teach more advanced concepts of data mining to advanced biochemistry and molecular biology undergraduates. Combined with a laboratory sequence of experiments into protease kinetics and mechanism and protease inhibitors, a tremendous reinforcement of the material occurs. Go further and add in a computational chemistry problem, in which students dock inhibitors with a structure and calculate the effect that inhibitor binding might have on the structure and suddenly students are doing science at the level that current research approaches problems while they are learning the basics of protein structure–function relationships, enzyme kinetics and mechanisms.

224

Fostering responsibility

Programmes must set realistic expectations of students and the students themselves must be involved in every aspect of this integrated education. Students who learn material and experimental approaches in different contexts (and realize it) will be better prepared to use that material in different contexts in the future. Simply having students keep a portfolio of where they think they learn key skills and concepts will help them to learn more completely — and often has the added advantage of helping faculty to realize that maybe they are not teaching what they think they are teaching! Getting students to find out information for themselves (not just from the pages of a textbook) can be very beneficial. A colleague of mine last year introduced me to what she called ‘adopt an enzyme’6: students adopt a molecule (a gene, enzyme, metabolite, hormone or vitamin, depending on what part of the programme they are at). The students are responsible for finding out, over several weeks, all they can about their adopted molecule and reporting on it through lab books and class reports. Students learn not only to use available resources but also to marshal the material into a cogent whole and report on the salient features of what they have found. At the same time, they see and learn a lot of ‘old-fashioned’ biochemistry and molecular biology. An added incentive is to include an openended question on an exam that allows them to show what they learned about the molecule in question. Mini-projects that include student presentations (oral or poster) are another useful teaching tool: other faculty and the students themselves can help critique the outcomes and discuss improvements.

The problem of preparing students for the future, which has shaped this discussion on education, raises an issue that faces not just the science education community but also the public in general. The complexity of modern science and the fundamentals that underlie the issues of the future (such as the genomics revolution) are not well understood by the general public. As scientists are increasingly aware of this disconnection, why not use it to further the goals of the education system? Being able to teach a topic well is in a way the ultimate demonstration of understanding the topic. Preparing to teach a topic is an excellent way of thoroughly learning it. We can make use of this in our education of current students. Students could work in groups in service learning or outreach opportunities; the excitement of science can be brought to students in both the primary and secondary education systems by asking college students to explain science to school students. At Gustavus, the biochemistry club members run a ‘Science on Saturdays’ scheme where they demonstrate and involve 8–12 year olds in hands-on experiments. Other possibilities include having students work on a newsletter that explains nutrition to older people, or organize a public debate about the pros and cons of genetic engineering. Students will learn more science, understand social issues and will be part of the process that contributes to better general understanding of both science and the issues facing scientists. As part of this integrated whole of the students’ education should be opportunities for students to learn in various ways: working one on one with faculty on research projects, interacting with visiting seminar speakers (and making connections for summer research opportunities) or attending professional meetings to present the results of their own research. The most able students will take advantage of many of these opportunities to learn and understand, and the hope is that all students will find some niche in the system that lets them take full advantage of the opportunities presented. From ideas to implementation

How can we best implement some of the changes and innovations discussed here? Most importantly, their implementation will require energetic faculty who are willing to experiment with the new paradigms of education and to appropriately assess the outcomes of implementation. Are these innovations suitable for all students in the sciences? Hopefully, but this puts pressure on faculty time: in institutions where staff are dedicated www.nature.com/reviews/molcellbio

| MARCH 2001 | VOLUME 2

PERSPECTIVES

“One can be no better a teacher than one is a learner at any point in one’s life.” to teaching and do little or no research, the future is best served by having more researchactive faculty; in research institutions the converse is true: more research faculty need to become involved in the direct education of undergraduates. The solution lies in closer interactions between teaching institutions and research institutions, to mutual benefit. By investing more time in helping undergraduate education (for example, giving research seminars at local institutions, helping mentor summer research internships, involvement in reviewing undergraduate research projects), faculty at research universities will benefit from the better level of preparation of undergraduates entering graduate school. Their colleagues who are involved primarily in teaching will benefit from the more direct interactions with research faculty, particularly if collaborative research projects can be developed. School administrations, however, need

to face up to the fact that a faculty that is implementing the research paradigm of teaching has less time for didactic lectures and labs, and must give credit to faculty who invest some of their time and energy in research. Funding agencies, in particular, look favourably at such interactions as ways of increasing the research activity in primarily teaching institutions. The role of educators as mentors is also continually evolving and will become increasingly important as we individualize education for students. Not all students are destined for academic research; some are headed towards the biotech industry, and towards other careers such as patent law or government policy for which a solid understanding of modern science is essential. By adapting the research paradigm at the undergraduate level, for example to involve internships in industry or government, and by advising students of alternative graduate programmes that encompass those types of interests, we can serve the maximum number of students and ensure a more scientifically literate population as well as optimizing the education of future research scientists. Ellis Bell is at Gustavus Adolphous College, 800 West College Avenue, Saint Peter, Minnesota 56082, USA. e-mail: eb@gac.edu

Take a look online Online, all Nature Reviews articles are enhanced with hyperlinks, including the following:

Links Problem-based learning | “The Other Side of Life: Educating Young Scientists about Business” Deborah J. Ausman | Science and Technology Policy: Past and Prologue. A Companion to Science and Technology Indicators — 2000 | Beyond BIO 101 | Role of the Private Sector in Training the Next Generation of Biomedical Scientists | Then, Now and In the Next Decade | Shaping the Future of Undergraduate Science, Mathematics and Technology Education | “New Paradigms, Teaching in Context, and on a Need-to-Know Basis” A. Malcolm Campbell | Biochemistry explorer

FURTHER INFORMATION

4. 5.

january 2001 volume 2 no. 1 www.nature.com/reviews

Allen, D. & Duch, B. Thinking Towards Solutions: Problem Based Learning Activities for General Biology (Harcourt College Publishing, Chestnut Hill, Massachusetts, 1998). White, E. B. in To Improve the Academy Vol. 15 (ed. Richlin, L.) 75–91 (New Forums Press, Stillwater, Olkahoma, 1996). Science and Technology Policy: Past and Prologue. A Companion to Science and Technology Indicators — 2000’ (NSF, Jessop, Maryland, 2000). Saltman, P. Remarks made at the 1997 ASBMB Education Satellite Meeting in San Francisco. Howard Hughes Medical Institute. Beyond BIO 101 (Howard Hughes Medical Institute, Chevy Chase, Maryland, 1997). Wobbe, K. K. Three years of pet enzyme project: a group learning companion to introductory biochemistry. FASEB J. 12, 143 (1998).

december 2000 volume 1 no. 3 www.nature.com/reviews

GENETICS

MOLECULAR CELL BIOLOGY

december 2000 volume 1 no. 3 www.nature.com/reviews

NEUROSCIENCE

nature publishing group

Nature Reviews is a rapidly expanding division of the Nature Publishing Group. The series already includes Nature Reviews Genetics, Nature Reviews Neuroscience and Nature Reviews Molecular Cell Biology, and will soon be joined by Nature Reviews Immunology. Over the next few months we plan to recruit three skilled:

Art/Layout Editors • References are linked to PubMed abstracts.

• Links to other related online resources are provided.

Essential qualities include at least one year’s experience of magazine layout using QuarkXpress and previous experience with Photoshop and Illustrator drawing high quality scientific diagrams. The successful candidates will be solely responsible for producing illustrations, galley proofs, final page layout and continuous publication PDFs exclusively for one review journal. In addition, they will work with the Senior Art/Layout Editor to maintain a consistent style for this journal in keeping with the other review titles. They must be organized, self motivated and enjoy working as part of a team, sharing their expertise with others. Please send your application letter together with a copy of your CV and details of references to:

• Web watch articles are linked to the resources discussed in the text.

Mrs Luisa Estanislao, Office Manager, Nature Reviews, 4 Crinan Street, London N1 9XW

• Gene and protein names, inherited diseases and protein domains are linked to publicdomain databases such as LocusLink, Flybase, SGD, OMIM and InterPro.

For more information on the review journals: http://www.nature.com/reviews

NATURE REVIEWS | MOLECUL AR CELL BIOLOGY

VOLUME 2 | MARCH 2001 | 2 2 5