news and views

nature structural biology • volume 5 number 8 • august 1998 671

The surface of the human immuno-deficiency virus (HIV) is coated withspikes representing trimers of the enve-lope glycoprotein (Env). Studies overmore than a decade have generated a pic-ture of how the external Env subunit(gp120) interacts first with theprimary receptor (CD4) and thenwith a co-receptor (a chemokinereceptor such as CCR5 andCXCR4) to trigger the fusionevent mediated by the Env trans-membrane subunit (gp41)1. Yetdespite intensive effort, gp120 hasproven refractory to high resolutionstructural analysis. Investigatorshave therefore been limited to car-toon-like depictions of Env inaction as a fusogenic machine, oras a moving target for neutralizingantibodies.

Now, thanks to collaborativeefforts of the groups headed byWayne Hendrickson at ColumbiaUniversity and Joseph Sodroski atthe Dana-Farber Cancer Center,the cartoon has been transformedinto a living picture in the form ofan X-ray crystallographic struc-ture of a ternary complex contain-ing a portion of HIV-1 gp120bound to CD4 and a neutralizingmonoclonal antibody2. And whata story this picture tells! With theenvelope's hermetic seal finallyopened, one can see in vivid atom-ic detail the gp120–CD4 inter-actions that previously had onlybeen imagined in schematic form.Moreover, the structure illumi-nates some of gp120's antigenic complexi-ty, revealing many features that underlyHIV's ability to escape surveillance by thehumoral immune system3. Finally, thestructure provides a framework for exten-sive mutagenesis analysis of the interac-tion between gp120 and co-receptor4. Thethree-dimensional snapshot directly con-firms earlier models, and entices the view-er with some intriguing surprises. This

new perspective is likely to have a pro-found impact on vaccine strategies aimedat generating humoral immunity, andmay also invite development of new drugstargeted at heretofore unknown siteswithin the gp120 molecule.

To solve the structure, the invest-igators had to overcome formidableobstacles. Obtaining sufficient quantitesof gp120 was not one of them, since themolecule can easily be produced byrecombinant protein expression technol-ogy. Recombinant gp120 is secreted as asoluble monomeric protein that displayscritical properties indicative of structuralintegrity, including binding to CD4,

CD4-induced binding to co-receptors,and reactivity with conformationallyrestricted antibodies. The difficulties liein properties of gp120 that hinder pro-duction of crystals suitable for X-ray dif-fraction analysis. Greater than half the

mass of the protein is carbohydrate(both high-mannose and com-plex), resulting in extensive molec-ular heterogeneity. Moreover, thepolypeptide sequence consists offive highly variable regions (V1–V5)interspersed with five relatively con-served regions; the first four vari-able regions form disulfide-bondedloops which are thought to be flexi-ble, thereby introducing conform-ational variability. These featureswere the major hurdles to crystal-lization.

The Hendrickson/Sodroski col-laboration was a perfect match tosurmount this hurdle. Severalyears earlier, Hendrickson led oneof two independent groups5,6 tosolve the structure of a solubleCD4 construct, consisting of the two N-terminal extracellulardomains (D1D2, with D1 contain-ing the gp120 binding site). Hesubsequently followed this with thecrystal structure of the entire CD4ectodomain (D1–D4)7. During thesame period, Sodroski and col-leagues were scrutinizing the gp120 structure using mutagenic,immunologic, and biochemicalapproaches. By analyzing numer-ous molecular variants of gp120and characterizing diverse mono-

clonal antibodies, they generated a modelin which immunodominant gp120regions are displayed on a ‘non-neutral-izing face’ and key functional regions aresequestered on a ‘neutralizing face’1.These earlier efforts provided the criticallaunching points for the structural solu-tions detailed in the new reports.

As detailed by Kwong et al.2, theresearch teams embarked upon a strategy

And the Best Picture is — the HIV gp120envelope, please!Edward A. Berger

X-ray crystallographic structure of the gp 120 core coupled with mutagenesis analyses reveal details ofreceptor interactions and multiple layers of immune evasion.

Fig. 1 Ternary complex between the gp120 core (red), CD4D1D2 (yellow) and Fab 17b light chain (blue) and heavy chain(purple). The side chain of CD4 Phe 43 projects into the gp120core. Reproduced with permission from ref. 2.

news and views

of removing problematic portions from thegp120 surface, and stabilizing the modifiedproteins with suitable ‘ligands’. Afterscreening nearly 20 modified protein/lig-and combinations, they succeeded in crys-tallizing a ternary complex containing atruncated deglycosylated HIV-1 gp120‘core’ (clade B, T cell line-adapted HXBc2strain) bound to D1D2 of CD4 and an Fabfragment of a neutralizing human mono-clonal antibody (17b) that had been shownpreviously to bind to a CD4-induced(CD4i) epitope8. The gp120 construct haddeletions at the N- and C-termini (52 and19 residues respectively), and the V1/V2and the V3 variable loops were replacedwith Gly-Ala-Gly substitutions. The pro-tein was expressed in insect cells and enzy-matically deglycosylated, leaving only thetwo core N-acetyl-glucosamine residues ateach site. The final deglycosylated core pro-tein retained >80% of the non-variableloop residues and <10% of the carbohy-drate; most importantly it displayed nearnormal bindng to CD4 and several confor-mation-dependent monoclonal antibod-ies9. Small needle crystals (30–40 µMcross-section) were obtained and the struc-ture was solved at 2.5 Å resolution by acombination of molecular replacement,isomorphous replacement, and densitymodification techniques, using diffractiondata collected at the BrookhavenLaboratory National Synchrotron LightSource. The ternary complex is shown inFig. 1

In the gp120 core structure (Fig. 2), thepolypeptide chain is folded into discrete

structural domains, each derived fromdiscontiguous regions of the primarysequence: a relatively conserved ‘inner’domain, a more variable ‘outer’ domain,and a conserved ‘bridging sheet’ com-posed of four antiparallel β-strands fromthe V1/V2 stem the fourth conservedregion. This overall design has no prece-dent amongst known protein structures,and is considered to be a prototype forEnvs of the primate immunodeficiencyviruses (genetically diverse HIV-1 strainsas well as with the more distantly relatedHIV-2 and simian immunodeficiencyvirus). Interpretations of the structure arerich with implications for the mechanismsof virus entry and immune evasion.

The gp120–CD4 interaction is particu-larly intricate2. The receptor binds withina recessed pocket in the gp120 coreformed at the interface of the outerdomain, the inner domain and the bridg-ing sheet (Fig. 1). Main chain atoms ofgp120 play prominent roles, contributingmore than half of the contacts with CD4.Particularly intriguing are two largeinterior cavities at the gp120–CD4 inter-face, created by topographical mismatch-es between the surfaces of the twoproteins. The smaller cavity extendsdeeply (~10 Å) into the hydrophobicinterior of gp120; it is plugged by thephenyl ring of CD4 Phe 43 (Fig. 1), aresidue previously implicated in gp120binding based on CD4 mutagenesis stud-ies1. At the cavity’s center is a mysterious,large spherical density that cannot beaccounted for by protein or any major

672 nature structural biology • volume 5 number 8 • august 1998

components involved in crystallization.Also puzzling is the highly conservednature of the hydrophobic gp120residues that line this cavity; the sidechains make minimal contacts with CD4,and substitutions would not appear tocause steric hindrance. The functionalsignificance of the conservation of theseresidues remains unclear. The larger cavi-ty is less buried on the surface of thegp120 core; it is water filled and linedwith hydrophilic residues from bothgp120 and CD4. This cavity is flanked byhighly conserved gp120 residues (Asp368, Glu 370, and Trp 427 at one end, Asp457 at the other) that make direct con-tacts with CD4; between the conservedgp120 residues is an ‘anti-hotspot’ ofvariable residues where substitutions canbe tolerated.

On the opposite face of the bridgingsheet, distinct from the CD4 binding site,lies the CD4i epitope for the neutralizing17b antibody2 (Fig. 1). The 17b Fab frag-ment makes contacts with all fourstrands at the base of the bridging sheet,suggesting that the 17b epitope requiresstructural integrity of this domain. CD4binding may play a role in stabilizing thebridging sheet structure, thereby con-tributing to the ‘CD4i nature’ of this epi-tope. While the ternary complex does notcontain any portions of co-receptor, crit-ical features of the gp120–co-receptorinteraction can be inferred from the crys-tal structure2 coupled with the mutation-al analysis described by Rizzuto et al.4. Inparticular, basic and polar residues in thebridging sheet that directly contact 17bare also shown to be critical for interac-tion with CCR5. This basic surface ofgp120 may interact electrostatically withthe acidic N-terminus of the co-receptorand possibly also with the acidic headgroups in the phospholipid bilayer of thetarget cell, presumably to trigger subse-quent conformational changes in Envinvolved in membrane fusion/virusentry. Also notable is that the nearby V3stem seems to favorably position the V3loop (absent from the gp120 core) forinteraction with the co-receptor, consis-tent with abundant evidence implicatingV3 in HIV entry, cytotropism and co-receptor binding1.

Features of the monomeric gp120 corediscerned from the crystal structure, cou-pled with requirements for surface expo-sure of glycosylation sites, variableregions, neutralizing epitopes, and recep-tor binding sites, led Wyatt et al. to pos-tulate a schematic model for thestructure of native assembled Env on the

Fig. 2 Core gp120showing the innerdomain, outer domain,and bridging sheet.Orientation is relatedto that in Fig. 1 by a 90°rotation about the ver-tical axis, with the C-terminal tail of CD4coming out of thepage. The overalldimension are approxi-mately 50 × 50 × 25 Å.The V1/V2 and V3stems are shown (thecorresponding loopsare deleted); the V4loop is poorly resolved.The N- and C-terminiare proposed to beoriented towards gp41on the viral surface,positioning the bridg-ing sheet and thestem of the V3 loopsuch that they facethe target cell.Reproduced with per-mission from ref. 2.

news and views

virion3. The surface of the trimeric spikeassumes a roughly hemispherical shape.The three gp120 subunits are packedtogether through their conserved, electro-neutral, unglycosylated inner domains,whereas the periphery contains the morevariable outer domains that are heavilymasked with complex oligosaccharides.The N- and C-termini of gp120 bothpoint toward the viral membrane wherethey can interact with gp41. The bridgingsheet, devoid of carbohydrate, is at thevertex of the trimer and is directed theopposite way toward the target cell with itsCD4 and co-receptor. The major neutral-ization epitopes are concentrated on thisface of the trimer, with variable structures(particularly V2 and V3) masking con-served neutralization epitopes.

In interpreting the crystal structure,Kwong et al.2 conclude that the gp120core captured in the CD4-bound statehas undergone profound conformationalchanges compared to free gp120. Indeedthe notion that CD4 binding inducesconformational changes that activate Envfor fusion has been a dominant theme formany years, and has been bolstered bythe recent demonstrations that CD4binding greatly enhances the affinity ofsoluble gp120 for co-receptor1. Severalfeatures of the gp120 core structure in theternary complex confirm and extend thisnotion. For one, certain features of thecomplex are difficult to reconcile if CD4is removed; these include the exposureand conservation of the hydrophobicgp120 residues in what would be the nowvacant Phe 43 cavity, and the likely insta-bility of the Phe 43 cavity–bridging sheetinteraction in the face of orientationalshifts between the inner and outer gp120domains. Moreover, the ‘CD4i nature’ ofthe 17b epitope initially observed withintact gp120 is retained in the core struc-ture, suggesting that even in the truncat-ed deglycosylated core this epitope ismasked or unstable prior to CD4 bind-ing. Finally, computer-based secondarystructure predictions match very wellthroughout most of the gp120 core struc-ture, except for major discrepancies atregions intimately involved in the CD4interaction. Based on unpublishedthermodynamic studies, Kwong et al.2

suggest that gp120 exists in an equilibri-um of conformational states, and that theCD4 binding process involves multiplesteps which culminate in insertion of Phe43 to stabilize the conformation seen inthe crystal structure.

The gp120 core structure and themodel for the trimer suggest mechanisms

nature structural biology • volume 5 number 8 • august 1998 673

for how Env carries out its mission in theface of immune surveillance2,3. For themembrane fusion reaction involved invirus entry, CD4 serves two functions toensure that Env is activated only at theright place and time — that is, at the sur-face of the newly encountered target cell.The positioning function follows fromthe inherent affinity of native Env forCD4, and the structural features of thetwo proteins on their respective surfacesthat orient the CD4-bound viral spike toface the membrane of the target cell. Thetiming function results from the CD4-induced conformational changes thatcreate/stabilize/unmask the co-receptorbinding site (bridging sheet plus the V3loop); the bound CD4 holds the trimericEnv in a transient metastable state, untilco-receptor binding triggers the finalconformational changes leading to mem-brane fusion. Viral entry is thus the finalstep in a highly choreographed cascade ofconformational changes induced by Env’ssequential interaction with CD4 followedby co-receptor.

How can this fusogenic machine func-tion over time in the infected individual,since it is so prominently displayed on thevirion surface and should therefore makean easy target for neutralizing antibodies?The crystal structure, combined with pre-vious antigenic analyses, provide muchinsight1–3. To begin with, much of thegp120 surface is simply unavailable toeither elicit or bind antibodies. The previ-ously noted ‘non-neutralizing face’ repre-sents the conserved gp120 surface of theinner domain that is occluded in thetrimeric structure; antibodies against thisregion can be elicited by free gp120 mole-cules shed by virions or infected cells, butthey are unable to access their epitopes onthe native trimer. The crystal structurealso reveals a surprising new immunologicfeature: a ‘silent face’ representing theouter domain of the gp120 core.Variability and extensive glycosylationmake this a futile target for antibodyattack. But what about the ‘neutralizingface’, for which Env-blocking antibodieshave been clearly demonstrated? Here, thedefenses are particularly varied and inge-nious. The regions of gp120 involved ininteraction with target cell receptors arecovered by the floppy V1/V2 and V3loops. Antibodies to these variable loopsdo arise in infected people, but their neu-tralizing activities tend to be highly strain-specific. Variation here provides a readymeans for immune escape, and is a drivingselective force for evolution of the viralquasi-species in the infected host. As for

the CD4 interaction site, its recessednature and dependence on specificarrangement of the domains in the gp120core may compromise its ability to elicit aneutralizing antibody response. Once pro-duced, a CD4BS antibody must bypass theoverlapping V1/V2 loop. The conservedpocket capped by CD4 Phe 43 is deeplyburied and probably sterically inaccessi-ble. Moreover, the CD4 interaction regioncloser to the gp120 surface can accomo-date variability at residues where the con-tacts are with main chain atoms, as well asat residues in the water-filled interfacialcavity surrounding the highly conservedcontacts. These variable residues assumeadded intrigue in view of the fact thattheir location and structure are entirelyunknown in the native Env trimer prior toCD4 binding. Finally, there are the CD4iepitopes that include determinants criticalfor co-receptor binding. Not surprisingly,these epitopes are inefficient at elicitingantibody responses since they are creat-ed/stabilized/exposed only after CD4binding. Even when present, such anti-bodies are only weakly neutralizing, pre-sumably because they have difficultyaccessing their (not yet existent?) epitopesin the native Env prior to CD4 binding.Once Env engages CD4, kinetic and stericproblems may prevail, since the antibodymust bind before the interaction withcoreceptor and subsequent fusion (whichpresumably occur rapidly), and the closeproximity of viral and target membranesmay impair antibody access.

So, the very features that make gp120 sodifficult to visualize by X-ray crystallogra-phy (flexible loops, extensive glycosyla-tion, requirement for bound ligand)appear also to hide it from the watchfuleye of the humoral immune system. Env’smolecular decoys and camouflages wouldseem to bode poorly for vaccine strategiesfocused on an antibody response. Yetinsights from the crystal structure providehints at a possible way out from thisdilemma. Perhaps approaches can bedevised to reveal to the humoral immunesystem certain neutralizing epitopes thatare normally hidden in the native Env(such as CD4i epitopes). True, if such epi-topes are created or exposed only afterCD4 binding, antibodies against themseem unlikely to provide much protectionfor the kinetic and steric reasons notedabove. But the suggestion that gp120exists in an equilbrium of conformationalstates2 may mean that a subset of suchantibodies will efficiently bind to theirepitopes even prior to Env interactionwith CD4.

The fragment of the type I insulin-likegrowth factor receptor (IGF-1R) forwhich the structure has just been deter-mined (residues 1–462; reported byGarrett et al. Nature 394, 395–399; 1998)is incapable of binding to its ligand,insulin-like growth factor. This, initially,might seem a disappointment, raising thequestion: what then is the relevance of thestructure? The answer is a great deal.

The fragment, which comprises abouthalf of the ectodomain of the IGF-1R, iscomposed of three domains: two right-handed β-helix motifs, L1 (1–150; top)and L2 (300–460; bottom); and an elong-ated cysteine-rich region (middle). A firstglance would suggest that the large notchin the structure with the prominentlyplaced loop (255–265) would be a likelyplace for ligand to bind. Such a presump-tion turns out to be correct.

Studies of chimeric receptors reveal thatresidues 1–68 of the insulin receptor (IR,which exhibits very considerable homologywith IGF-1R) confer insulin binding in thecontext of the IGF-1R. Conversely, residues191–290 of IGF-1R confer IGF binding inthe context of IR. The authors note that a

ligand bound in the notch could contactboth these regions. Mapping of the residuesthat are important for insulin binding inthe IR (and, by extension, IGF-1R giventheir similarity) reveal they cluster in threeareas: on the underside of the L1 domain(indicated by the arrow); in the N-terminalregion of the L2 domain (upper portion ofL2 in the picture), where a mutation caus-ing Rabson-Mendenhall syndrome andsevere insulin resistance is found; and with-in the central cys-rich region, whereresidues 223–274 are involved in determin-ing ligand specificity (including the255–265 loop). All three of these sites faceinto the notch, supporting the idea that thisis where IGF binds.

In the structure the notch is too wide(~30 Å between L1 and L2) for the threefaces to bind ligand simultaneously. Thepicture clearly shows that the position ofthe L2 domain is not constrained by interactions with the other twodomains, rather it is probably crystalpacking effects that dictate the overlylarge size of the notch. Thus, eventhough this fragment of the ectodomainis not sufficient to bind IGF, the notch

does bare the major specificity determi-nants necessary for the interaction withligand. GR

A notch for IGF

news and views

The gp120 core structure also presentsnew opportunities for rational drug design.The interfacial cavities that line the CD4interactive region are particularly inviting,though as yet it is unclear whether thesecavities also exist on native Env prior toCD4 binding. Only further structuralanalyses will tell.

As with most major scientific achieve-ments, the crystal structure described byHendrickson, Sodroski, and their col-leagues will be appreciated as much for thenew questions it raises as for the old ones itsolves. Particularly intriguing problems arethe structure of gp120 prior to CD4 bind-ing, the orientations of the V1/V2 and V3loops relative to the core structure, theinteractions between gp120 and gp41, thesubunit packing within the Env trimer, the

interaction between gp120 and co-receptor,comparison of pre-fusion and fusion-activestates, and the structural basis for the neu-tralization differences between laboratory-adapted and primary HIV strains.

Thus the first look inside the gp120 enve-lope has proven most revealing. As with allBest Pictures, this one will be viewed overand over, each time from a somewhat dif-ferent vantage point. Inspired by the gp120success, coupled with the recent crystallo-graphic solution of the fusion-active gp41core10,11, we can reasonably expect morerevealing structural sequels to follow. Nodoubt the insights gained will advance ourfundamental understanding of the HIVinfection process, and hopefully guide thedesign of novel protective and therapeuticstrategies.

674 nature structural biology • volume 5 number 8 • august 1998

Edward A. Berger is in the Laboratory ofViral Diseases, National Institute ofAllergy and Infectious Diseases, NationalInstitutes of Health, Building 4, Room 236,Bethesda, Maryland 20892, USA.

Correspondence should be addressed toemail: edward_berger@nih.gov

1. Wyatt, R. & Sodroski, J. Science 280, 1884–1888 (1998).2. Kwong, P. D et al. Nature 393, 648–659 (1998).3. Wyatt, R. et al. Nature 393, 705–711 (1998).4. Rizzuto, C. D. et al. Science 280, 1949–1953 (1998).5. Ryu, S.-E. et al. Nature 348, 419–426 (1990).6. Wang, J. et al. Nature 348, 411–418 (1990).7. Wu, H., Kwong, P. D. & Hendrickson, W. A. Nature

387, 527–530 (1997).8. Thali, M. et al. J. Virol. 67, 3978–3988 (1993).9. Binley, J. M. et al. AIDS Res. Hum. Retrovir. 14,

191–198 (1998).10. Chan, D. C., Fass, D., Berger, J. M. & Kim, P. S. Cell

89, 263–273 (1997).11. Weissenhorn, W., Dessen, A., Harrison, S. C., Skehel,

J. J. & Wiley, D. C. Nature 387, 426–430 (1997).

picture story