Structure of HIV-1 gp120 with gp41-interactive region reveals layered envelope architecture and basis of conformational mobility

Abstract

The viral spike of HIV-1 is composed of three gp120 envelope glycoproteins attached noncovalently to three gp41 transmembrane molecules. Viral entry is initiated by binding to the CD4 receptor on the cell surface, which induces large conformational changes in gp120. These changes not only provide a model for receptor-triggered entry, but affect spike sensitivity to drug- and antibody-mediated neutralization. Although some of the details of the CD4-induced conformational change have been visualized by crystal structures and cryoelectron tomograms, the critical gp41-interactive region of gp120 was missing from previous atomic-level characterizations. Here we determine the crystal structure of an HIV-1 gp120 core with intact gp41-interactive region in its CD4-bound state, compare this structure to unliganded and antibody-bound forms to identify structurally invariant and plastic components, and use ligand-oriented cryoelectron tomograms to define component mobility in the viral spike context. Newly defined gp120 elements proximal to the gp41 interface complete a 7-stranded β-sandwich, which appeared invariant in conformation. Loop excursions emanating from the sandwich form three topologically separate—and structurally plastic—layers, topped off by the highly glycosylated gp120 outer domain. Crystal structures, cryoelectron tomograms, and interlayer chemistry were consistent with a mechanism in which the layers act as a shape-changing spacer, facilitating movement between outer domain and gp41-associated β-sandwich and providing for conformational diversity used in immune evasion. A “layered” gp120 architecture thus allows movement among alternative glycoprotein conformations required for virus entry and immune evasion, whereas a β-sandwich clamp maintains gp120–gp41 interaction and regulates gp41 transitions.

The viral spike (gp120/gp41) of HIV type 1 (HIV-1) uses substantial conformational changes to facilitate viral entry (reviewed in ref. 1). Receptor binding by gp120 triggers a series of conformational changes in gp41, which in the unliganded envelope spike possesses a high potential energy that will ultimately be used to fuse the viral and target cell membranes. Binding of the initial receptor, CD4, induces changes in gp120 conformation that allow high-affinity interaction with the coreceptor, CCR5 or CXCR4, and the formation of a gp41 prehairpin intermediate. Subsequent engagement of coreceptor is thought to promote additional conformational changes in gp41 that create an energetically stable six-helix bundle coincident with the fusion of viral and cell membranes.

The gp120 and gp41 glycoproteins are not linked by disulfide bonds, and the noncovalent association of these spike subunits presents significant challenges. The gp120 glycoprotein must be flexible to allow conformational change, yet retain sufficient contact with gp41 to maintain the integrity of the unliganded trimer and, after CD4 binding, to arrest gp41 transitions at a prehairpin intermediate stage. The shedding of gp120 from the trimer and functional inactivation of the CD4-bound spike are well documented consequences of missteps in this demanding process (2, 3).

Although some of the details for the CD4-induced change have been visualized by crystal structures (4, 5) and ligand-oriented cryoelectron tomograms (6), atomic-level descriptions have been missing for the viral spike as well as for the critical gp41-interactive region of gp120. How does gp120 alter its conformation upon CD4 binding without premature triggering of the metastable gp41-entry machinery, and what features allow for the observed diversity in gp120 structure required for immune evasion (7)? To define the molecular mechanism of spike-conformational mobility, we crystallized a core gp120 with intact gp41-interactive region—adding gp120 residues 31 to 82 and 493 to 511—as part of a ternary complex with 2-domain CD4 and the antigen-binding fragment (Fab) of the human antibody, 48d. Mapping of gp120 mutants with enhanced spike-dissociation phenotypes onto this structure highlighted a patch, including the newly defined N- and C-termini, which, in ligand-oriented cryoelectron tomograms, faced the expected location of gp41. Comparisons were made of the more complete gp120 structure to crystal structures of gp120 in other conformations: unliganded and Fab b12-bound structures of gp120 (4, 8), as well as Fab F105-bound and Fab b13-bound structures of gp120, which we recently determined (7). These structural comparisons provided data to model the conformational mobility of gp120 substructures in the monomeric context, and fittings to ligand-oriented cryoelectron tomograms of the HIV-1 viral spike illustrated the component-substructure mobility in the oligomeric context. Taken together, the results provide insight into how one conformational machine (gp120) controls another (gp41). They suggest a “layered” gp120 architecture, which allows for extensive conformational diversity, both to move in the prefusion viral spike without triggering the metastable gp41-fusion machinery and to fold into alternative conformations required to fulfill biological functions of receptor binding and immune evasion.

Results

Overall Structure of gp120 with Intact gp41-Interactive Region.

More than 5,000 crystallization trials involving 20 different gp120–ligand combinations were screened (Table S1). Crystals in several space groups could be grown from a deglycosylated HXBc2 gp120 core with intact gp41-interactive region, partially stabilized in the CD4-bound state by T257S and S375W substitutions on gp120 (9). Diffraction to 2.6 and 3.5 Å resolution was observed from orthorhombic and tetragonal crystals, respectively, of the same ternary complex (257/375-stabilized gp120 core, 2-domain CD4, and Fab of antibody 48d); anisotropy and incompleteness, however, reduced the effective resolution of the orthorhombic data to approximately 3 Å (Table S2). The structures, which were solved by molecular replacement (10), were refined to R-factors of 20.1% and 25.7% (R_free of 27.5% and 27.3%) for orthorhombic and tetragonal lattices, respectively (11, 12). Close structural similarity was observed for ternary complexes in both crystal forms, a consequence in part of related lattices (Fig. S1).

The new structure of gp120 allows a first look at the intact N and C termini of the mature protein as well as at the gp41-interactive region (Fig. 1). Many aspects were reminiscent of previously determined CD4-bound structures (5, 13–15), including orientations of inner domain, outer domain, and bridging sheet minidomain. CD4 reaches in from the side of gp120, and the site for coreceptor binding interacts with the CD4-induced antibody 48d (Fig. S2). Although more than 70 amino acids were added to the previous determined core, the termini remain clustered together, poised to interact with their gp41-transmembrane partner.

Fig. 1.

 Structure of an HIV-1 gp120 core with intact gp41-interactive region. (A) Ternary complex: ribbon diagram shows previously described core gp120 (gray), with newly identified intact N terminus (red) and C terminus (purple), bound by the membrane-distal two domains of CD4 (yellow) and the antigen-binding fragment (Fab) of antibody 48d (light and dark blue). In this orientation, the viral membrane (and hence the gp41-transmembrane glycoprotein) would be positioned toward the top of the page and the target cell toward the bottom. (B) Close-up: core gp120 with intact gp41-interaction region is shown from a 90° view about a vertical axis from A. Coloring is the same as in A, except that the outer domain is orange. (C) Topology: the inner domain of gp120 is shown with β-strands as arrows and α-helices as coils; close strand proximity is indicative of hydrogen bonding. Coloring is the same as in B, with the bridging sheet (β2, β3, β20, β21) and extensions to the outer domain shown with dotted lines. (D) Sequence: the sequences of the crystallized clade B isolate (HXBc2), the consensus for HIV-1 group M (CONSENSUS M), and SIVmac239 (SIVMM239 MAC) are shown, with secondary structure as arrows (β-strand) and coils (α-helices). The gg sequence refers to the dipeptide truncation, gly-gly, which was used to replace the V1/V2 variable region. Regions of disorder are marked with an x, and outer domain (residues 257–472) is labeled. Blue boxes highlight residues that are involved in gp41 interaction, as defined by mutagenesis studies (16–21). (E) Ribbon representation of gp120 as shown in B, with residues implicated by mutagenesis in the interaction with gp41 highlighted with blue spheres and sticks. (F) Placement of the gp120-CD4-Fab 17b complex in the electron density map derived from cryoelectron tomography (light gray) (6), with gp120 and residues involved in gp41 interactions colored as in E, and Fab 17b in light brown. (G) View of 90° rotation from F from the viral membrane.

Overall, the newly defined N terminus (residues 31–82) packs intimately with the previously determined core (5) (Fig. 1). Residues 43 to 63 (2 β-strands and connecting coil) extend approximately 40 Å toward the coreceptor-binding surface, and residues 64 to 82 (a parallel β-strand, a 3-turn α-helix, and loop regions) twist back toward the termini (Fig. 1 B and C). The originally described 2-strand, 2-helix bundle of the core (5) is now revealed to be a 4-strand, 3-helix bundle (Fig. 1C). Meanwhile, the original 5-stranded β-sandwich of the termini [3-antiparallel strands packing against 2-antiparallel strands (5)], which was subsequently suggested to include a sixth strand (13), becomes a 7-stranded β-sandwich (5 strands packed against 2). Together, these additions involve more than 1,000 Ų of interactive surface area between the newly defined termini and the previously determined core.

gp41-Interactive Region of gp120.

The dominant interactions of gp120 with gp41 have been defined previously by mutagenesis and implicate several discontinuous gp120 segments (16–21) (Fig. 1D). Mapping mutationally defined residues onto the newly defined crystal structure reveals that these segments form primarily a single surface, involving the termini and the 7-stranded β-sandwich (Fig. 1E). To place this surface into the context of the functional viral spike, we used the recently defined ligand-oriented cryoelectron tomograms from Subramaniam and colleagues (6) of BaL HIV-1 in complex with CD4 and the CD4-induced Fab 17b. To obtain an appropriate atomic-level model for the tomogram fitting, we superimposed the newly defined ternary complex (which contained the CD4-induced Fab 48d) with the gp120 component of the previously determined structure of core gp120-CD4-Fab 17b (5) (Fig. S3) and computationally swapped Fabs. Placement of the Fab-swapped model of CD4, 17b, and gp120 with intact gp41-interactive region generated a unique fit as defined by the program Chimera (22), with 85% of the atoms in the tomogram density (Fig. 1 F and G). The fitted structure defines an orientation for gp120 in the viral spike in which all 48 (16 × 3) sites of N-linked glycosylation face outward in a clash-free manner (Fig. S4). The only portion of the model that did not fit well with the tomogram density corresponded to the extended gp120 termini. These appeared structurally flexible, such that interactions with gp41 in the viral spike or with symmetry-related molecules in the crystal lattice could easily induce termini movement (extensive contacts between the N terminus of gp120 and domain 2 of CD4 were observed in both orthorhombic and tetragonal crystal forms; Fig. S1). In general, the mutationally defined gp41-interactive region faced the expected location of gp41, toward the center of the tomogram density (Fig. 1 F and G). Moreover, epitopes for nonneutralizing antibodies (Fig. S4) and conserved surface-exposed residues (Fig. S4) also map to regions that are expected to be buried in the trimer spike. Together, the results suggest that the primary anchor holding gp120 in the viral spike involves its termini and the newly defined 7-stranded β-sandwich.

gp120 Architecture: A Collection of Invariant and Plastic Structural Elements.

The unliganded [i.e., simian immunodeficiency virus (SIV)] to CD4-bound (i.e.,HIV-1) transition of gp120 in the monomeric context was previously suggested by Harrison and colleagues (4) to involve rearrangements of virtually all of the inner domain secondary structure. In the truncated SIV inner domain (4), only the termini-proximal β-sandwich remains a coherent ensemble of secondary structure, and even there, only 3 of the strands are well retained. The β-sandwich structure is more evident in antibody-bound forms of HIV-1 core gp120, e.g., a gp120 core modified to retain the CD4-bound state and then bound to antibody b12 (8) or to antibody b13 (7), and a gp120 core with intact V3 loop bound to antibody F105 (7) (Fig. 2). The presence of additional strands comprising the β-sandwich reinforces the view that it is a stable structural element. The invariance of the gp41-associated β-sandwich is validated not only through structural comparisons (as discussed later; see Table S3, for example, for quantified movements), but also by antigenic analysis: the binding of C11 antibody, for example, which recognizes the β-sandwich and extended termini, is not inhibited by CD4 and CD4 binding-site antibodies in antibody cross-competition studies (23).

Fig. 2.

“Layered” gp120 architecture: invariant β-sandwich and structurally plastic layers. (A) The newly defined 7-stranded β-sandwich is shown in ribbon diagram representation (red) with β-strands labeled and loop excursions to the layers depicted schematically. (B) Schematic representation of gp120 structure. (C) β-Sandwich superposition of the newly defined CD4-bound conformation of gp120 with unliganded and antibody-bound conformations, in two 90° orientations. Polypeptides are drawn in ribbon representations, with 7-stranded β-sandwich colored red, layer 1 in violet, layer 2 in cyan, layer 3 in blue, and the outer domain in light brown. (D) Different conformations of monomeric gp120 are shown in ribbon diagram representation. Note that layer 1 is present in only the CD4-bound form described here. Structural elements are colored the same as in C.

From the perspective of this central β-sandwich (Fig. 2), much of the new N terminus can be viewed as a loop excursion, emanating between strands and β0. Two other substantial excursions also occur, between β1 and β5 and between β7 and β25. In the newly defined gp120, each of these three excursions packs as a separate topological layer, each with an associated helix. The first or innermost layer ( to β0 excursion with α0 helix) packs against the second layer (β1 to β5 excursion with α1 helix), and the second layer packs against the third layer (β7 to β25 excursion with α5 helix), which in turn packs against the structurally invariant outer domain.

Quantification of Layer Diversity.

To quantify motions of different portions of gp120, we superimposed different substructures (sandwich, layers, and outer domain) and calculated rmsds of these components for gp120 in unliganded, CD4-bound, and antibody-bound conformations (Table S3). The termini-associated β-sandwich was the most fixed, and layer 2 the most conformationally diverse, with rmsds as high as 20 Å (Fig. 2C and Fig. S3). We also calculated movements of the center of mass of each structural element from the vantage point of the central β-sandwich (Fig. S3 and Table S4). Changes in center-of-mass distances averaged only 2.4 Å, but angular motions averaged 24°, suggesting that the primary flexibility of gp120 involves pivoting around the gp41-associated invariant β-sandwich.

When the inner domain architecture is viewed in the context of an invariant β-sandwich and mobile layers, the large conformational changes induced by CD4 or antibodies (b12, b13, and F105) can be viewed as a layer-constrained shuffling of secondary structure (Fig. 2D). Thus, CD4 binding involves the relative movement of the outer domain with respect to the termini-associated β-sandwich, and a repacking of the α1 helix (in layer 2) against the α5 helix (in layer 3). The b12-bound and b13-bound conformations retain many of the elements of the CD4-bound conformation, influenced perhaps by 8 substitutions—including 2-disulfide bonds—shown to stabilize the CD4-bound state (8). In particular, the relative orientations of termini-associated β-sandwich and outer domain shift by only 7.7° and 5.1°, respectively (Table S4). The relative orientation of layer 3 changes modestly as well. Nonetheless, the central α1-helix unfolds completely in the b12-bound conformation and partially for the b13-bound conformation, with layer 2 rmsds of 13 and 31 Å, respectively, relative to the CD4-bound state (Fig. 2 and Table S4). The F105-bound conformation, meanwhile, is remarkably similar to the CD4-bound state, despite the absence of any stabilizing mutations. Nonetheless, portions of layer 2 move up to 40 Å, with an overall rmsd for layer 2 of 15 Å. (The unliganded and antibody-bound conformations of the first layer have not yet been determined, but it seems likely that it will maintain the layered inner domain architecture, as structural rearrangements within the first layer are additionally constrained by the presence of a disulfide between helix α0 and strand . Moreover, because of the way layer 2 reaches between the strands of layer 1 to connect to the β-sandwich (Fig. 2A), layer 2 would have to “thread” itself through the layer 1 connections to the β-sandwich to switch positions topologically.) The biological implications—especially with regard to immune evasion—of the newly determined antibody-bound gp120 conformations are analyzed elsewhere (7).

Mobility of gp120 in the Functional Viral Spike.

A tight association of the extended gp120 termini and neighboring β-sandwich with gp41 would provide a means to hold gp41 in its metastable prefusion conformation. In such a model, extensive gp120 alterations could occur related to layer refolding and outer domain movement, but gp41 would be insulated from these by the fixed β-sandwich. To test the hypothesis that the gp41-associated β-sandwich of the inner domain remains immobile while the layers and outer domain move, we turned to the ligand-oriented cryoelectron tomograms of Subramaniam and colleagues (6). These tomograms capture images of BaL HIV-1 viral spikes with either Fab b12 or CD4 and Fab 17b. Although its resolution is only approximately 20 Å, the tomogram definition is sufficient with crystallographically defined-marker ligands to restrict potential orientations of gp120 (see, e.g., the high-quality fit of our newly determined gp120 with N/C within a CD4-antibody complex; Fig. 1 F and G). We used the fit gp120–CD4–Fab 17b ternary complex as a starting point, superimposed different structural elements of the gp120–Fab b12 binary complex, and analyzed the resultant fit to the Fab b12–BaL viral spike tomogram (Fig. S5).

Superposition of β-sandwiches was found to result in a fit superior to those obtained from superposition of either outer domain or layers (Table 1). These results presume that the ligand-induced conformations of monomeric gp120 will be retained in the context of the oligomeric viral spike and that ligand-induced constraints can be transmitted through the various gp120 domain interfaces to the β-sandwich. Relevant to this, we note that binding of CD4 constrains the orientations of layers 2, 3, and outer domain, whereas binding of b12 constrains layer 3 and outer domain. (The unliganded conformation by definition lacks ligand fixation, and indeed superposition of either β-sandwich or outer domain of the unliganded monomeric structure leads to clashes in the oligomeric context, although the tomogram density—without an orienting ligand—was not sufficiently resolved to define a clear solution; see fits in Fig. S5.)

Table 1.

Identification of fixed component of gp120 in the viral spike by using the simultaneous placement of atomic-level structures in cryoelectron tomograms of viral spike complexes with CD4-Fab 17b and Fab b12

Structural component superimposed	Atoms not in tomographic density (18,543 total)	Atoms in density, %
β–Sandwich	2,960	84.0
Layer 2	7,102	61.7
Layer 3	4,309	76.8
Outer domain	6,230	66.4

The range of gp120 mobility observed in the monomeric context is likely restricted in the viral spike. To gain insight into these restrictions, we simultaneously fit structures of CD4- and b12-bound structures of gp120 in cryoelectron tomograms of the HIV-1 BaL viral spike for CD4-Fab 17b from Subramaniam and colleagues (6), with the following procedure: (i) three copies of the CD4- gp120 structure with intact N and C termini along with swapped 17b Fab were fit to the cryoelectron tomogram of the viral spike bound by CD4 here as shown in Fig. 1; (ii) different structural components of gp120 (left column) were used to superimpose three copies of the crystal structures of gp120 bound by antibody b12 (8); (iii) this component-oriented trimer of Fab b12-gp120 was then fit to the cryoelectron tomogram of the viral spike with Fab b12 and the resultant fit calculated (middle and right columns; see Fig. S5 for details).

Overall, the results suggest that the orientation in the viral spike of the gp41-associated β-sandwich of gp120 is not significantly altered by b12 and CD4 binding. Although these observations provide evidence for segmental flexibility, they do not provide direct visualization for atomic level details of the intact unliganded envelope structure and its pathway of activation. Still, the results provide insight into how observed movements of the gp120 outer domain are likely buffered from gp41 by the flexible gp120 layers and a fixed β-sandwich clamp.

Conformational Diversity of gp120 Relative to Other Proteins.

We also assessed gp120 conformational diversity with superposition-independent means of quantification, as this affords a means to compare gp120 diversity to that of other proteins. One superposition-independent means of quantification involves the direct enumeration of the number of residues that alter secondary structure upon conformational change. A second superposition-independent metric involves difference distance-matrix analysis, which compares distances between all residue pairs in one conformation with analogous residue-pair distances in a second conformation (24).

We applied both these superposition-independent metrics to the various conformational states of gp120. Results from these were observed to group into 3 clusters (Fig. 3). One cluster involved changes in secondary structure of 80 to 110 residues. This cluster involved comparisons with the unliganded conformation of gp120, which was determined for SIV, and it seems likely that at least part of the structural differences in these comparisons relates to differences between HIV-1 and SIV. A second cluster involved comparisons between CD4-bound states of HIV-1 clade B gp120. In this second cluster, changes in secondary structure involved less than 20 residues (of more than 300), with maximal difference distances of approximately 10 Å. These small differences seem more typical of a well folded, mostly invariant protein with a few flexible loops. The second cluster shows that, when a particular conformation is induced (in this case by CD4 binding), gp120 may fold into a well defined revisited conformation.

Fig. 3.

Conformational diversity of gp120. Protein motions can be classified into categories of hinge, shear, and refold (Left) (35). These archetypes show distinctive signatures when analyzed by enumerating secondary structure changes (horizontal axis) and by difference–distance maximum (vertical axis). Hinge motions generate large difference distances, although only a few residues refold (e.g., calmodulin, hemolysin, and transglutaminase). Shear motions show small changes with both metrics (e.g., citrate synthase and calpain). Refold motions, by contrast, show large changes with both metrics. Refolding is perhaps best epitomized by the transmembrane components of type 1 fusion machines (e.g., “F1” of parainfluenza virus, “HA2” of influenza A virus, and “gp2” of Ebola virus). This analysis shows how gp120 fits into the refold category when moving from unliganded state or between different states induced by receptor and/or antibody. In this it differs from the N-terminal components of other type 1 fusion proteins (e.g., “HA1” of influenza virus). When fixed by CD4 binding, however, these metrics indicate that primarily a single state is induced. [Like gp120, the ubiquitin-conformational ensemble (29) also forms a cluster, although one of much smaller magnitude than that of gp120; to clarify presentation, only a single point for a relatively divergent pair of ubiquitin conformations is displayed, with the ubiquitin cluster extending from this point to the origin.] Labels are italicized representations of structures for which the sequence identity is less than 40% between the two conformations.

A third cluster also involved comparisons between HIV-1 clade B gp120s, but this time with different CD4-binding site ligands. In this third cluster, changes in secondary structure involved 40 to 50 residues and maximal difference distances of 20 to 50 Å. Structural differences of this magnitude are unusual, and were observed to occur even with antibody-bound variants of gp120 stabilized by multiple disulfides (e.g., with b12- and b13-bound variants of gp120). This third cluster emphasizes the potential conformational diversity of gp120, which was more akin to that observed with transmembrane-fusion machines (e.g., gp2 of Ebola virus or HA2 of influenza A virus) than to their N-terminal receptor–binding counterparts. Thus, for example, the N-terminal component of influenza A virus hemagglutinin (HA1) displays maximal difference distances of less than 12 Å, with less than 10 residues changing secondary structure upon receptor binding, antibody binding, or pH-induced conformational rearrangement (25–27) (Fig. 3).

Structural Features Associated with Layer Plasticity.

What features allow for the structural plasticity of gp120, particularly of the layers? It seems likely that, for layers to rearrange, structural elements between them cannot rigidly adhere. We thus compared the chemical interactions between layers with the interactions between related portions of the β-sandwich. In particular, we enumerated the number of unique hydrophobic interactions and hydrogen bonds in layers and β-sandwich—normalized by the total number of residues—for CD4-bound, unliganded, b12-bound, b13-bound, and F105-bound conformations of gp120 (Fig. S6). Between 1.5 and 3 times more hydrophobic interactions were observed in the β-sandwich compared with the layers for all 5 gp120 conformations (P = 0.025). Similarly, 3 to 4 times more hydrogen bonds were present in the β-sandwich compared with the layers (P = 0.0036; Fig. S6). The results indicate that the structurally plastic layers have significantly fewer interlayer interactions (hydrogen bonds and hydrophobic interactions) than the invariant β-sandwich. The reduction in interlayer interactions should affect layer 2 most strongly, as it is sandwiched between the other layers. In addition, layer 2 shows significant hydration, and this may enhance its structural plasticity. This hydration may be associated with the greater relative surface area (e.g., 103 Ų per amino acid for layer 2) versus the β-sandwich and outer domain (79 and 53 Ų per amino acid, respectively).

Structural Invariance and N-Linked Glycosylation.

Interestingly, most N-linked glycans reside on regions of gp120 that do not change secondary structure. In the gp120 core with intact N and C termini, N-linked glycan were not present in the layers, whereas 16 N-linked glycans could be found on the conformationally invariant β-sandwich and outer domain. Although glycan can reside on structurally mobile and/or flexible elements, such as the V4 and V5 loops of the outer domain, these elements do not refold (e.g., from extended loop to a β-strand or α-helix). The segregation of glycan to regions that do not change secondary structure suggests that the presence of glycan may not be compatible with the structural plasticity and refolding observed in the layers. We note parenthetically that N-linked glycan must remain exposed and clash-free in all protein conformations and, moreover, the side chain of the Ser or Thr residue in the glycan sequon (Asn-X-Ser/Thr) needs to satisfy requirements of hydrogen-bonding donor/acceptor (28).

Discussion

The mechanism for conformational diversity revealed here enables gp120 to accomplish functions related to virus entry as well as immune evasion. In terms of entry, HIV-1 receptor–triggered entry appears to involve 2 conformational machines: the standard type 1 fusion machinery of the transmembrane–gp41 component and a unique receptor-triggered gp120 component, which uses conformational changes to fulfill competing requirements of viral entry and immune evasion. Very few proteins exhibit the range of conformational diversity observed with gp120; the conformational diversity of ubiquitin (29), for example, is substantially less than that of gp120 (Fig. 3).

In terms of immune evasion, the diversity of conformations allows for an effective decoy strategy, which hides the site of CD4 attachment (7). Moreover, spike flexibility appears to relate to neutralization resistance, with variable loop interactions making the spike less able to deform in neutralization-resistant isolates (13). Finally, the sparsity of N-linked glycan in structurally plastic regions observed here suggests a tradeoff between protection from antibody recognition afforded by glycan [e.g., the glycan shield (30, 31)] versus that associated with refolding [e.g., conformational masking (32)].

The architecture of gp120 thus includes structurally invariant, although potentially mobile, elements (β-sandwich, outer domain, and perhaps V1/V2 loops) and structurally plastic elements (layers 1–3), which are able to refold into alternative conformations (Fig. 4). Proceeding outward from gp41 at the center of the viral spike, we first encounter the gp120 termini and invariant 7-stranded β-sandwich. These elements appear to maintain the gp120–gp41 interaction and to regulate gp41 transitions. Excursions from this sandwich form three mobile and structurally plastic layers, which appear to be the source of observed gp120 conformational diversity. Emanating from layer 2 are the V1/V2 loops, and from layer 3 is the highly glycosylated outer domain. Thus, the layers link 3 structural elements, each of which has approximately half its molecular weight in N-linked glycosylation—β-sandwich (4 sites of N-linked glycosylation in the HXBc2 strain of HIV-1), V1/V2 loops (6 sites), and outer domain (14 sites). Appropriate orienting of these glycan-bearing elements by the layers likely allows for formation of the previously described glycan shield (30, 31), which prevents recognition by most potentially neutralizing antibodies (Fig. 4). Subsequent receptor-induced refolding of the layers provides the means by which to rearrange gp120 into alternative conformations required to negotiate the conformational transitions necessary for viral entry.

Fig. 4.

 Mechanism for gp120 mobility, gp41 fixation, and immune evasion. β-Sandwich clamp and embracing N and C termini of gp120 are shown holding gp41 in a metastable state. The layers, meanwhile, can refold to position the relative orientations of three highly glycosylated components: β-sandwich, V1/V2 loops, and outer domain. In the unliganded state (Left), the glycan-bearing elements arrange to form an effective “glycan shield” that prevents recognition by most antibodies. In the CD4-bound state (Right), the layers organize with outer domain to form the high-affinity binding sites for CD4 and coreceptor.

Materials and Methods

Constructs containing HIV-1 gp120 were expressed in Drosophila Schneider 2 cells (33). The secreted gp120 was purified by antibody-affinity chromatography, deglycosylated to the protein-proximal N-acetylglucosamine, complexed with various ligands (2-domain CD4 and Fabs of various gp120-reactive antibodies), further purified by gel filtration, and screened for crystallization. X-ray data were collected at a wavelength of 1.00 Å with the third generation undulator beam-line (ID-22; SER-CAT) at the Advanced Photon Source. Structures of orthorhombic (P2₁2₁2₁) and tetragonal (P4₁2₁2) forms of core gp120 with intact gp41-interactive regions in complex with 2-domain CD4 and Fab 48d were determined by molecular replacement (10) starting from the published models of Fab 48d and of HXBc2 core gp120 with CD4 (5, 34). These structures were refined (11, 12), fit into previously described cryoelectron tomograms (6), subjected to modeling of N-linked glycosylation, and analyzed in the context of other gp120 structures (4, 7, 8, 14) for structural diversity and for the chemical basis of their plasticity. Additional methodological details are given in SI Materials and Methods.