letters

812 nature structural biology • volume 9 number 11 • november 2002

Structure of the Tsg101 UEVdomain in complex withthe PTAP motif of the HIV-1p6 proteinOwen Pornillos1,2, Steven L. Alam1,2, Darrell R. Davis2,3

and Wesley I. Sundquist2

1These authors contributed equally to the structure determination.Departments of 2Biochemistry and 3Medicinal Chemistry, University ofUtah, Salt Lake City, Utah 84132, USA.

Published online 15 October 2002; doi:10.1038/nsb856

The structural proteins of HIV and Ebola display PTAP pep-tide motifs (termed ‘late domains’) that recruit the humanprotein Tsg101 to facilitate virus budding. Here we presentthe solution structure of the UEV (ubiquitin E2 variant)binding domain of Tsg101 in complex with a PTAP peptidethat spans the late domain of HIV-1 p6Gag. The UEV domainof Tsg101 resembles E2 ubiquitin-conjugating enzymes, andthe PTAP peptide binds in a bifurcated groove above the ves-tigial enzyme active site. Each PTAP residue makes impor-tant contacts, and the Ala 9-Pro 10 dipeptide binds in a deeppocket of the UEV domain that resembles the X-Pro bindingpockets of SH3 and WW domains. The structure reveals themolecular basis of HIV PTAP late domain function and rep-resents an attractive starting point for the design of novelinhibitors of virus budding.

To spread infection, non-lytic viruses must assemble and budfrom producer cells before re-entering uninfected cells. Cis- andtrans-acting factors required for efficient budding have beenidentified for several enveloped RNA viruses (see refs 1 and 2 forrecent reviews) and are perhaps best defined for the type-1human immunodeficiency virus (HIV-1). HIV-1 assembly andbudding are directed by the viral Gag protein, which is the majorstructural protein of the virus3. Efficient HIV-1 budding frommost cell types requires a conserved, cis-acting tetrapeptidemotif, P(T/S)AP (the ‘late domain’)4–6, which is located in the C-terminal p6 region of Gag (Fig. 1) and forms a docking site forthe cellular protein Tsg101 (tumor susceptibility gene 101)7–10.Tsg101, in turn, recruits additional cellular factors that helpfacilitate viral budding.

The small cellular protein ubiquitin also plays a role in thebudding of HIV-1 and other retro-, filo- and rhabdoviruses11–16.This role is independent of the well-characterized function ofubiquitin in targeting proteins for degradation by the protea-

some and, instead, seems related to the function of ubiquitin intargeting proteins through the endosomal pathway for eventualdegradation in the lysosome17,18. Intriguingly, HIV-1 Gag itself isubiquitylated at low levels19,20, although it is not yet clearwhether this is because Gag is the functional target for ubiquity-lation or simply because assembling Gag molecules are broughtinto close proximity to a promiscuous ubiquitin ligase whosefunctional target is a cellular protein(s).

The requirements for Tsg101 and ubiquitin in HIV buddingare consistent with a model in which the virus buds by usurp-ing cellular machinery that normally forms vesicles that budinto a late endosomal compartment called the multivesicularbody (MVB)8,21. This model rationalizes the requirement forTsg101 and ubiquitin in HIV budding because Tsg101 nor-mally helps sort ubiquitylated protein cargos into MVB vesi-cles22–24. The model is also attractive because the topologicalconstraints on vesicle formation in the MVB and envelopedvirus particle formation at the plasma membrane are similar— that is, in both cases the direction of budding is away fromthe cytoplasm and therefore requires that the membrane fission machinery function from the cytoplasmic face of thebudding vesicle13.

To bud from the cell surface, HIV-1 must redirect Tsg101from its normal sites of action on the late endosome to sites ofviral budding at the plasma membrane. Tsg101 is recruited tothe sites of virus assembly by a direct binding interactionbetween the N-terminal UEV (ubiquitin E2 variant) domain ofTsg101 and the p6Gag PTAP late domain7,8,10,25 (Fig. 1). UEVdomains show structural similarity to E2 ubiquitin ligases butare enzymatically inactive because they lack the active site Cysresidue that forms a transient thioester bond with ubiqui-tin25–30. Both UEV and E2 proteins are known to bind ubiqui-tin31, but the binding of Pro-rich peptide ligands has notpreviously been described for either class of proteins. It is,therefore, of interest to understand how the Tsg101 UEVdomain can bind the late domain of HIV-1 p6Gag. Moreover,Tsg101 recruitment is essential for viral replication, and anunderstanding of the molecular basis for the Tsg101 UEV–p6interaction may help in the structure-assisted development ofviral inhibitors. Therefore, we have determined the solutionstructure of the Tsg101 UEV domain in complex with its peptide binding site from the HIV-1 p6Gag protein.

Structure determinationA nine-amino acid ‘PTAP peptide’ from HIV-1NL4-3 p6Gag encom-passing the PTAP late domain (residues 5–13, PEPTAPPEE;Fig. 1) was selected for study because the peptide: (i) bindsTsg101 UEV in slow exchange on the NMR time scale at600 MHz, (ii) binds with the same affinity as intact p6 protein(Kd = 3 µM under the NMR solution conditions)25, and

Fig. 1 Schematic representations of human Tsg101 (390 amino acids) and HIV-1 Gag (500 amino acids). Abbreviations for putative domains andmotifs within Tsg101 are: UEV (ubiquitin E2 variant), PRD (proline-rich domain), coil (predicted coiled coil-forming region) and S-box (‘steadiness’box)54. Viral protease cleavage sites within the Gag polyprotein are denoted by vertical lines, and the names of the resulting proteins are given. Thelocation of the PTAP peptide (5PEPTAPPEE13) within p6Gag is indicated. ‘Myr’ represents a myristoyl modification at the Gag N-terminus.

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y

letters

nature structural biology • volume 9 number 11 • november 2002 813

(iii) causes the same chemical shift perturbations within Tsg101UEV as p6, indicating that it makes all relevant binding con-tacts25. Both the protein and peptide were isotopically labeledwith 13C and 15N, allowing the use of standard triple-resonance

experiments to assign both components and isotope-edited/fil-tered experiments to identify intermolecular NOEs (Fig. 2a,b).The structure of the complex was calculated using 1,955 NOErestraints (including 82 intermolecular NOEs), 90 hydrogen

Fig. 2 Determination of the Tsg101 UEV–PTAP peptide complex structure. a, Lower panel, overlaid 1H-15N HSQC spectra of the 15N-labeled PTAP pep-tide either free (black) or bound to Tsg101 UEV (green). The five peptide amide protons are labeled. Multiple peaks in the unbound peptide spec-trum reflect differing magnetic environments of minor Pro isomers. All Pro residues in the bound state are in trans conformations. Upper panel,corresponding region of a 2D 15N-filtered NOESY-HSQC spectrum of the 15N-PTAP peptide–Tsg101 UEV complex showing intermolecular, interresidueand intraresidue NOEs (red, green and black, respectively). b, Strips from 1H-13C NOESY-HSQC spectra of the Tsg101 UEV–PTAP peptide complex. Theleft panels (boxed in green) show 13C/15N-labeled PTAP peptide and unlabeled UEV domain. Intrapeptide NOEs are green and intermolecular NOEsare red. The right panels (boxed in black) show 13C/15N-labeled UEV domain and unlabeled PTAP peptide. Intramolecular (UEV–UEV) NOEs are black,and intermolecular NOEs are red. Resonance assignments are indicated at the bottom and sides of the figure, with peptide resonances labeled ingreen and UEV resonances labeled in black. c, Sequence and secondary structure of Tsg101 UEV. Residues that contact the PTAP peptide are red. d, Stereo view superposition of the 20 final structures (Cα trace) of Tsg101 UEV (gray) and PTAP peptide (dark green). e, Superposition of the free(yellow) and PTAP peptide-bound (gray) structures of the Tsg101 UEV domain. Tsg101 UEV residues surrounding the PTAP peptide (green)-bindingsite shift significantly to accommodate the peptide (residues 58–71, 91–107 and 139–144 shift by an average of 3.4 Å).

a b

c

d e

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y

letters

814 nature structural biology • volume 9 number 11 • november 2002

bonding restraints and 85 dihedral angle restraints (Table 1).The structure is well ordered throughout (Fig. 2d), and the 20lowest penalty structures superimpose over the mean coordinatepositions with r.m.s. deviations of 0.75 Å (backbone atoms) and1.04 Å (all heavy atoms).

Overview of the structureTsg101 UEV adopts the α/β/loop/α fold characteristic of E2ubiquitin ligases31 but, like other UEV domains, lacks the two C-terminal helices found in the E2 enzymes25–27. The PTAP pep-tide binds in a groove between the strand S2/S3 hairpin, the N-terminal third of the vestigial active site loop and the C-termi-nal residues of the Tsg101 UEV (Figs 2, 3). The analogous site inE2 enzymes is occupied by the C-terminal helix, and divergencefrom the canonical E2 fold allows the Tsg101 UEV to use this siteto mediate intermolecular interactions.

As the PTAP peptide binds, the binding groove of Tsg101UEV closes significantly, enveloping the central PTAP motif(Fig. 2e). The most significant change is that the flexible C-ter-minal residues of Tsg101 UEV become well ordered and wraparound one side of the PTAP motif. Other changes in Tsg101

UEV that accompany peptide binding include (i) the final turnof helix H4 rotates toward the peptide, facilitated by backbonetorsion angle changes at Phe 135 and Gly 136; (ii) the S2/S3hairpin shifts and twists to orient S3 perpendicular to the pep-tide, thereby allowing Tyr 68 and Asn 69 to interact with thepeptide from opposite sides of this strand; and (iii) the N-ter-minal third of the vestigial active site loop reorients andincreases in order, as judged by a 30% increase in intramolecu-lar NOEs in this region.

PTAP peptide recognitionThe central Ala 9–Pro 11 tripeptide of the bound PTAP peptideforms one turn of a left-handed type II polyproline helix, andthe flanking N- and C-termini adopt extended conformations.Peptide binding is mediated primarily by extensive intermolecu-lar contacts with the four central residues, Pro 7-Thr 8-Ala 9-Pro10 (discussed below), hydrogen bonds to the Pro 5 and Glu 6main chain carbonyl oxygens and modest van der Waal contactswith the side chains of Pro 5, Pro 11 and Glu 12 (Fig. 3). Thus,the overall structure is consistent with the observation that sin-gle point mutations in any of the four conserved PTAP residues

Fig. 3 Molecular recognition in the Tsg101 UEV–PTAP peptide complex. a, Summary of contacts53 between the Tsg101 UEV domain and the PTAP peptide(dark green). Hydrogen bonding interactions shown for Oγ of Thr 8 were notobserved in all calculated structures. b, Stereo view of the PTAP peptide (darkgreen) in its binding groove on Tsg101 UEV. The ‘Pro’ pocket, which binds Pro 7 ofPTAP (residues 7–10) from p6, and the ‘Ala-Pro’ pocket, which binds the Ala 9-Pro10 dipeptide, are circled in blue and magenta, respectively. PTAP peptide bindingburies a total surface area of 1,350 Å2. The ubiquitin-binding surface of Tsg101UEV is indicated in black.

a b

Fig. 4 Proline recognition by Tsg101 UEV. Expanded views of the a, Pro and b, Ala-Pro binding pockets, viewed along the binding groove from theN-terminal end of the peptide. c, Similarities between X-Pro proline recognition in the Tsg101 UEV (dark green), SH3 (yellow) and WW (orange)domains. In each case, ‘key’ Pro residues of bound ligands are sandwiched between two aromatic rings. The figure was created by superimposingkey ligand prolines: Pro 10 in the Tsg101 UEV–PTAP complex, Pro 3 in the dystrophin WW–β-dystroglycan PPXY complex34 and Pro 3 in the Grb2SH3–Sos PXXP complex35

a b c

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y

letters

nature structural biology • volume 9 number 11 • november 2002 815

can prevent Tsg101 UEV binding and viral budding, whereasmutations in the less conserved flanking residues have onlymodest effects on both processes5,8.

Sequence-specific recognition of the PTAP motif by theTsg101 UEV is achieved primarily within two distinct pocketsalong the peptide-binding groove (Figs 3b, 4). Pro 7 is cradled in

a shallow pocket lined by Tsg101 residue Pro 71 (bottom),and methyl groups from Thr 58, Thr 92 and Met 95 (sides)(Fig. 4a). The Ala 9-Pro 10 dipeptide binds in a deeperpocket, with Ala 9 contacting the methyl and methylenegroups from Tsg101 residues Ile 70, Met 95, Val 141 andSer 143 (Fig. 4b,c). The Pro 10 ring wedges between thearomatic rings of Tyr 63 and Tyr 68 (Fig. 4c), and is buttressed on one side by Pro 139 and Val 141 (Fig. 4b).

Understanding how proteins recognize proline-richbinding sites is of particular importance because these arethe most frequently occurring motifs in metazoan pro-teomes32,33. In the case of Tsg101 UEV, the Ala 9-Pro 10binding pocket seems to be the most important recognitionelement and is reminiscent of the ‘X-Pro’ pockets used byWW and SH3 domains for recognizing their Pro-rich lig-ands. ‘Key’ Pro residues in the ligands for WW and SH3domains also adopt type II polyproline helical conforma-tions and bind with the proline ring wedged between twoorthogonal aromatic rings32. The relative positions of thekey prolines and the two aromatic rings are highly con-served across the Tsg101 UEV–PTAP, dystrophinWW–PPXY34 and Grb2 SH3–PXXP35 complexes (Fig. 4c),revealing that all three domains have converged upon thesame strategy for recognizing proline, despite having dif-ferent protein folds.

The conserved Thr 8 residue of p6 PTAP traverses aridge that separates the Pro 7 and Ala 9-Pro 10 bindingpockets, thereby linking the two recognition elements(Fig. 3b). Unlike linker residues in SH3 and WW domainligands, however, Thr 8 contacts the UEV domain and isessential for binding8. The Thr 8 hydroxyl proton reso-nance was observed in NMR spectra (not shown), show-ing that it is protected from solvent exchange andsuggesting a hydrogen bonding interaction. Our dataindicate that Oγ of Thr 8 may donate a hydrogen bond tothe backbone carbonyl of Arg 144 and accept one from theOγ of Ser 143, although these interactions are notobserved in all calculated structures because the Ser 143and Arg 144 positions are not precisely defined by theNOE data.

Implications of the structureThe structure of the Tsg101 UEV–PTAP peptide complexagrees well with mutational analyses of the HIV-1 p6requirements for Tsg101 UEV binding and viral budding2,5.In particular, Ala/Gly mutations at each of the last three p6PTAP residues reduce UEV binding affinity by at least 35-fold8, and all of these mutations remove significant sidechain contacts in the complex. In contrast, the methylgroup of Thr 8 of the peptide does not make significantcontacts, consistent with the observation that both Thr andSer are found at this position in natural HIV-1 isolates36

and with our observation that PSAP and PTAP sequencesbind Tsg101 UEV equally well (not shown). The Pro 7 ringseems less critical than other side chains within the PTAPmotif, because the γ and δ atoms make only modest inter-molecular contacts. This is consistent with the observationthat a P7A mutation of p6 reduces binding of the Tsg101UEV by only three-fold, whereas a P7L mutation, whichcannot be accommodated in the binding site without dis-

Table 1 Structure statistics for the Tsg101 UEV–PTAP peptide complex

<TAD>1 <CNS>1

NOE distance restraints2 (Å) 1,955 1,955IntraUEV 1,765 1,765Sequential (|i – j| = 1) 594 594Medium range (2 ≤ |i – j| ≤ 5) 319 319Long range (|i – j| > 5) 479 479Intrapeptide 108 108Intermolecular3 82 82

Hydrogen bond distance restraints4 (Å) 90 90Dihedral angle restraints (°) 85 85Number of HNHA coupling constants 0 96DYANA target function (Å2) 1.39 ± 0.08 n.a.CNS energy (kcal mol–1) ∼ 10,000 ± 2,0005 213 ± 15Residual distance restraint violations

Number of violations ≥0.4 Å 0 0Sum of violations6 1.8 ± 0.4 21 ± 3Maximum violation (Å) 0.34

Residual dihedral angle restraint violationsNumber of violations ≥3° 1 0Sum of violations6 0.02 ± 0.01 0.20 ± 0.07Maximum violation (°) 3.15

van der Waals violationsNumber ≥0.6 Å 0 0Sum of violations6 5.2 ± 0.3 71 ± 6Maximum violation (Å) 0.58

Ramachandran statistics (%)7

Favored 66.1 74.1Allowed 26.3 22.2Generously allowed 4.5 2.4Disallowed 3.1 1.2

R.m.s. deviations to the average coordinates8 (Å)Tsg101 UEV residues 4–143

Backbone 0.75 ± 0.37 0.70 ± 0.35Heavy atoms 1.07 ± 0.56 1.01 ± 0.52

Tsg101 UEV secondary structures9

Backbone 0.61 ± 0.27 0.56 ± 0.25Heavy atoms 0.91 ± 0.44 0.85 ± 0.39

PTAP peptide residues 5–13Backbone 0.60 ± 0.31 0.66 ± 0.28Heavy atom 0.75 ± 0.51 0.83 ± 0.51

1<TAD> is the ensemble of 20 lowest-penalty structures (from 200 total) calcu-lated using DYANA48. <CNS> is the same ensemble after 2,000 steps (15 pseach) of simulated annealing at 300 K, 2,000 slow-cooling steps to 0 K and2,000 steps of restrained Powell minimization in Cartesian space (anneal.inpprotocol)49.2Only meaningful and non-redundant restraints as determined by the CALIBAfunction in DYANA48.3>75% of the 82 intermolecular NOEs are to the 7PTAP10 motif, emphasizingthe extensive contacts between Tsg101 UEV and this conserved tetrapeptide.4Two upper-limit distance restraints were used to define each hydrogen bond.5Energies for structures input into CNS49 (from DYANA48) were estimated with-in generate_easy.inp (CNS task file) after the first regularization withoutrestraints.6Violation energies from DYANA have units of Å or °, whereas energies fromCNS are in kcal mol–1.7Determined using PROCHECK-NMR50.8Superposition and overall r.m.s. deviations for the ensemble were calculatedusing MolMol51.9Helices: H1, residues 4–13; H2, 18–31; H3, 112–115; and H4, 124–138. Strands:S1, residues 36–44; S2, 48–63; S3, 66–76; and S4, 86–90 (Fig. 2c).

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y

letters

816 nature structural biology • volume 9 number 11 • november 2002

tortion, reduces Tsg101 UEV binding by 70-fold and blocks virusbudding5,8.

The Tsg101 UEV–PTAP peptide structure is also consistentwith mutational analyses of the Tsg101 UEV domain. Ala substi-tutions of Tsg101 UEV residues Tyr 63 and Met 95 reduce p6binding by 13- and 52-fold, respectively25, and are readilyexplained because Met 95 makes extensive intermolecularhydrophobic interactions with Pro 7 and Ala 9, whereas Tyr 63forms one side of the Pro 10 binding slot (Figs 3, 4). The Y63Aand M95A mutations of Tsg101 also inhibit HIV-1 budding andinfectivity, underscoring the importance of the Tsg101–PTAPcomplex in HIV replication (J.E. Garrus, U.K. van Schwedlerand W.I.S., unpub. data). Mutation of the Tsg101 UEV Thr 67-Tyr 68-Asn 69 tripepeptide to Ala 67-Ala 68-Ala 69 also dimin-ishes the p6–Tsg101 interaction in a yeast two hybrid system7,and this observation is again explained by the structure as bothTyr 68 and Asn 69 contact the PTAP peptide. Other mutations ofthe UEV of Tsg101 reported to reduce p6 binding — for exam-ple, Y110W, Y113V, W117A and K118A7 — are well removedfrom the PTAP binding site, however, and their effects probablyreflect indirect structural perturbations.

It is likely that the PTAP binding groove evolved to mediateTsg101 interactions with other cellular proteins and that virusesare simply mimicking a natural Tsg101 interaction. PotentialP(T/S)AP-binding motifs of Tsg101 UEV are located down-stream in Tsg101 itself and within Hrs, a protein that may recruitTsg101 to the endosomal membrane during vacuolar proteinsorting23,37,38. Tsg101 UEV can also bind P(T/S)AP-containingpeptides derived from Tsg101 and Hrs, although the biologicalrelevance of these interactions remains to be established (R. Fisher, O.P. and W.I.S., unpub. data).

Tsg101 and its yeast ortholog, Vps23p, also bind ubiqui-tin8,22,23,25, which targets proteins into vesicles that bud into thelate endosome to create multivesicular bodies22,23. Althoughubiquitin recognition is not understood in molecular detail,binding of ubiquitin maps to the concave Tsg101 β-sheet22,25

(Figs 2e, 3b). An extended groove on Tsg101 UEV connects theN-terminal end of the PTAP peptide-binding site to this ubiquitin-binding surface, and we speculate that this groovecould create a continuous polypeptide recognition surfaceupon monoubiquitylation of an upstream Lys residue(Fig. 3b). This model is consistent with the observation thatseveral potential binding partners of Tsg101 UEV aremonoubiquitylated11,22,23,39, and with the emerging roles forubiquitin in enhancing protein–protein interactions and stim-ulating productive complex formation and biological activityin several systems40.

Recruitment of Tsg101 is essential for HIV replication, makingthe PTAP peptide-binding site of Tsg101 UEV a potentiallyattractive target for new inhibitors of viral replication (althoughthe importance of this binding site for normal Tsg101 functionsand cell viability remains to be tested). Similarities in the mech-anisms of proline recognition by SH3 domains and by the UEVdomain of Tsg101 suggest that proline peptidomimetic strategiesdeveloped for SH3 domains may also apply to Tsg101 inhibitordevelopment. For example, Lim and co-workers41 have shownthat N-substituted amino acids can replace the key Pro residuesrecognized by X-Pro pockets. These ‘peptoids’ have been used asa basis for designing or screening small molecule ligands withenhanced affinity for specific SH3 domains over natural peptideligands41,42. The structure of the Tsg101 UEV–PTAP complextherefore suggests a plausible approach for the development ofnovel inhibitors of virus budding.

MethodsProtein and peptide samples. The UEV domain (residues 2–145)of Tsg101 was expressed and purified as described25. PTAP peptides(HIV-1NL4-3 p6 residues 5–13: HN-PEPTAPPEE-COOH) were synthe-sized (unlabeled peptide) or expressed for isotopic labeling asTrp∆LE peptide fusions, purified from inclusion bodies, cleavedwith CNBr and purified by reversed-phase HPLC43. The expectedTsg101 UEV and PTAP peptide masses were confirmed by massspectrometry.

NMR spectroscopy. NMR samples were 1:1 UEV domain:PTAP pep-tide, at ∼ 1.5 mM in 20 mM sodium phosphate, pH 5.5, 10 mM NaCland 10% (v/v) 2H2O. Spectra were recorded at 18 °C on a VarianINOVA 600 MHz spectrometer. Backbone and side chain resonanceassignments were obtained using standard triple-resonance NMRexperiments44 on mixed samples in which only one component was13C/15N labeled. Inter- and intramolecular NOEs were similarly iden-tified using X-edited/filtered spectra (X = 13C / 15N) on mixed sam-ples45–47. Spectra were processed using FELIX97 (MSI) and analyzedin SPARKY (T.D. Goddard and D.G. Kneller, University of California,San Francisco).

Structure determination. Structures were calculated usingDYANA48 and CNS49 as described for the uncomplexed Tsg101 UEVdomain25. Structures were analyzed using PROCHECK-NMR50,MolMol51 and InsightII (MSI) (Table 1), and displayed usingMolScript52, LIGPLOT53 and PyMOL (DeLano Scientific). Figures werecreated using the lowest penalty structure.

Coordinates. The coordinates, chemical shifts and NOE restraintshave been deposited in the Protein Data Bank (accession codes1M4Q (CNS ensemble) and 1M4P (DYANA ensemble)).

AcknowledgmentsWe thank B. Schackmann and S. Endicott for peptide synthesis and purification,D. Edwards for technical NMR support and E. Ross for computer support. Thiswork was supported by NIH funding to W.I.S. The Utah Biomolecular NMRFacility is supported by grants from the NIH and the NSF.

Competing interests statementThe authors declare that they have no competing financial interests.

Correspondence should be addressed to W.I.S. email: wes@biochem.utah.edu orD.R.D. email: davis@adenosine.pharm.utah.edu

Received 26 August, 2002; accepted 12 September, 2002.

1. Freed, E.O. J. Virol. 76, 4679–4687 (2002).2. Pornillos, O.P., Garrus, J.E. & Sundquist, W.I. Trends Cell Biol. in the press (2002).3. Göttlinger, H.G. AIDS 15, S13–S20 (2001).4. Göttlinger, H.G., Dorfman, T., Sodroski, J.G. & Haseltine, W.A. Proc. Natl. Acad.

Sci. USA 88, 3195–3199 (1991).5. Huang, M., Orenstein, J.M., Martin, M.A. & Freed, E.O. J. Virol. 69, 6810–6818

(1995).6. Demirov, D.G., Orenstein, J.M. & Freed, E.O. J. Virol. 76, 105–117 (2002).7. VerPlank, L. et al. Proc. Natl. Acad. Sci. USA 98, 7724–7729 (2001).8. Garrus, J.E. et al. Cell 107, 55–65 (2001).9. Martin-Serrano, J., Zang, T. & Bieniasz, P.D. Nature Med. 7, 1313–1319(2001).

10. Demirov, D.G., Ono, A., Orenstein, J.M. & Freed, E.O. Proc. Natl. Acad. Sci. USA99, 955–960 (2002).

11. Strack, B., Calistri, A., Accola, M.A., Palu, G. & Göttlinger, H.G. Proc. Natl. Acad.Sci. USA 97, 13063–13068 (2000).

12. Schubert, U. et al. Proc. Natl. Acad. Sci. USA 97, 13057–13062 (2000).13. Patnaik, A., Chau, V. & Wills, J.W. Proc. Natl. Acad. Sci. USA 97, 13069–13074

(2000).14. Vogt, V.M. Proc. Natl. Acad. Sci. USA 97, 12945–12947 (2000).15. Harty, R.N. et al. J. Virol. 75, 10623–10629 (2001).16. Harty, R.N., Brown, M.E., Wang, G., Huibregtse, J. & Hayes, F.P. Proc. Natl. Acad.

Sci. USA 97, 13871–13876 (2000).17. Hicke, L. Cell 106, 527–530 (2001).18. Strack, B., Calistri, A. & Göttlinger, H.G. J. Virol. 76, 5472–5479 (2002).19. Ott, D.E. et al. J. Virol. 72, 2962–2968 (1998).20. Ott, D.E., Coren, L.V., Chertova, E.N., Gagliardi, T.D. & Schubert, U. Virology 278,

111–121 (2000).21. Lemmon, S.K. & Traub, L.M. Curr. Opin. Cell Biol. 12, 457–466 (2000).22. Katzmann, D.J., Babst, M. & Emr, S.D. Cell 106, 145–155 (2001).23. Bishop, N., Horman, A. & Woodman, P. J. Cell Biol. 157, 91–102 (2002).24. Dupre, S., Volland, C. & Haguenauer-Tsapis, R. Curr. Biol. 11, R932–R934 (2001).25. Pornillos, O. et al. EMBO J. 21, 2397–2406 (2002).

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y

letters

nature structural biology • volume 9 number 11 • november 2002 817

26. VanDemark, A.P., Hofmann, R.M., Tsui, C., Pickart, C.M. & Wolberger, C. Cell 105,711–720 (2001).

27. Moraes, T.F. et al. Nature Struct. Biol. 8, 669–673 (2001).28. Koonin, E.V. & Abagyan, R.A. Nature Genet. 16, 330–331 (1997).29. Ponting, C.P., Cai, Y.D. & Bork, P. J. Mol. Med. 75, 467–469 (1997).30. Sancho, E. et al. Mol. Cell. Biol. 18, 576–589 (1998).31. Pickart, C.M. Annu. Rev. Biochem. 70, 503–533 (2001).32. Zarrinpar, A. & Lim, W.A. Nature Struct. Biol. 7, 611–613 (2000).33. Kay, B.K., Williamson, M.P. & Sudol, M. FASEB J. 14, 231–241 (2000).34. Huang, X. et al. Nature Struct. Biol. 7, 634–638 (2000).35. Wittekind, M. et al. J. Mol. Biol. 267, 933–952 (1997).36. Kuiken, C. et al. HIV Sequence Compendium (Los Alamos National Laboratory,

Los Alamos; 2000).37. Piper, R.C., Cooper, A.A., Yang, H. & Stevens, T.H. J. Cell Biol. 131, 603–617 (1995).38. Raiborg, C., Bache, K.G., Mehlum, A. & Stenmark, H. Biochem. Soc. Trans. 29,

472–475 (2001).39. Polo, S. et al. Nature 416, 451–455 (2002).

40. Pickart, C.M. Mol. Cell 8, 499–504 (2001).41. Nguyen, J.T., Turck, C.W., Cohen, F.E., Zuckermann, R.N. & Lim, W.A. Science 282,

2088–2092 (1998).42. Nguyen, J.T. et al. Chem. Biol. 7, 463–473 (2000).43. Dadlez, M. & Kim, P.S. Biochemistry 35, 16153–16164 (1996).44. Ferentz, A.E. & Wagner, G. Q. Rev. Biophys. 33, 29–65 (2000).45. Ogura, K., Terasawa, H. & Inagaki, F. J. Magn. Reson. B 112, 63–68 (1996).46. Ikura, M. & Bax, A. J. Am. Chem. Soc. 114, 2433–2440 (1992).47. Zwahlen, C. et al. J. Am. Chem. Soc. 119, 6711–6721 (1997).48. Güntert, P., Mumenthaler, C. & Wüthrich, K. J. Mol. Biol. 273, 283–298 (1997).49. Brünger, A.T. et al. Acta Crystallogr. D 54, 905–921 (1998).50. Laskowski, R.A., Rullmann, J.A., MacArthur, M.W., Kaptein, R. & Thornton, J.M. J.

Biomol. NMR 8, 477–486 (1996).51. Koradi, R., Billeter, M. & Wüthrich, K. J. Mol. Graph. 14, 51–55 (1996).52. Kraulis, P.J. J. Appl. Crystallogr. 24, 946–950 (1991).53. Wallace, A.C., Laskowski, R.A. & Thornton, J.M. Protein Eng. 8, 127–134 (1995).54. Feng, G.H., Lih, C.J. & Cohen, S.N. Cancer Res. 60, 1736–1741 (2000).

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rest

ruct

ura

lbio

log

y