Crystal Structure of a Neutralizing Human IgG Against HIV-1: A Template for Vaccine Design

See allHide authors and affiliations

Science  10 Aug 2001:
Vol. 293, Issue 5532, pp. 1155-1159
DOI: 10.1126/science.1061692


We present the crystal structure at 2.7 angstrom resolution of the human antibody IgG1 b12. Antibody b12 recognizes the CD4-binding site of human immunodeficiency virus–1 (HIV-1) gp120 and is one of only two known antibodies against gp120 capable of broad and potent neutralization of primary HIV-1 isolates. A key feature of the antibody-combining site is the protruding, finger-like long CDR H3 that can penetrate the recessed CD4-binding site of gp120. A docking model of b12 and gp120 reveals severe structural constraints that explain the extraordinary challenge in eliciting effective neutralizing antibodies similar to b12. The structure, together with mutagenesis studies, provides a rationale for the extensive cross-reactivity of b12 and a valuable framework for the design of HIV-1 vaccines capable of eliciting b12-like activity.

HIV-1 vaccine development is greatly hindered by the extreme difficulty in eliciting a neutralizing antibody response to the virus (1–3). However, three human monoclonal antibodies have been identified that can efficiently neutralize a broad array of primary isolates of HIV-1 in vitro (4) and can protect against viral challenge in vivo (5–9). Antibody 2F5 (10) reacts with gp41, whereas 2G12 (11) and b12 (12) react with independent epitopes on gp120. Elucidation of the epitopes recognized by these antibodies may offer valuable insights into the design of antigens capable of eliciting a protective antibody response.

Antibody b12 was identified from a combinatorial phage display library developed from bone marrow donated by a 31-year-old homosexual male who had been seropositive, but without symptoms, for 6 years (12). This antibody recognizes a highly conserved epitope overlapping the CD4-binding region of gp120, which accounts for its broad recognition of different HIV-1 isolates. Antibody b12 neutralizes about 75% of clade B primary viruses and a similar, or somewhat lesser, proportion of other clades (12, 13). In addition, b12 can protect hu-PBL-SCID mice (5) and macaques (7) from viral challenge. This combination of potency and broad specificity suggests that the b12 epitope on gp120 may be a particularly effective target for vaccine design.

The b12 IgG1κ was expressed in CHO cells, purified, and crystallized as previously described (14). The crystal structure of the intact IgG1 was determined at 2.7 Å resolution through an exhaustive molecular replacement (MR) search using more than 100 individual Fc and Fab search models (14). The highly mobile hinge regions connecting the Fabs to the Fc domains were interpretable after extensive rebuilding, refinement, and density modification (Table 1). Only three residues of the upper hinge of one heavy chain, seven residues of a frequently disordered surface loop of one Fab CH1 domain (residues 128 to 135), and three COOH-terminal residues from one Fc are disordered.

Table 1

Structure determination of IgG1 b12. Data to 2.7 Å resolution were collected from a single cryocooled crystal at SSRL Beamline 7-1 on a MAR (MAR-Research) 30-cm area detector and processed with DENZO and SCALEPACK (38). Although theR sym for the highest resolution shell is relatively high (62.6%), this shell contains useful information, because the I/σ is 2.4 and because inclusion of these reflections clearly improved the electron density maps. The structure was determined by molecular replacement (MR) using the software package AMoRe within the CCP4 suite (39) as described (14). Initial R cryst andR free values after MR and substitution of the b12 sequence were 0.37 and 0.44, respectively. The structure was refined in CNS (40) using a maximum likelihood target function and an initial overall anisotropic temperature factor correction. Multiple cycles of rebuilding in TOM/FRODO (41), refinement, as well as density modification in DM (42) by using a perturbation γ correction, dramatically clarified density and allowed construction of approximately 65 residues in the hinge, carbohydrate, and solvent-exposed regions. The overall averageB value is 90 Å2, reflecting the high solvent content of the crystals and interdomain flexibility of the IgG molecule. However, the electron density, with exception of the flexible hinge region and some solvent exposed loops, is absolutely clear and completely unambiguous. Water molecules were only built into highly ordered regions of the structure. B values for protein atoms in the vicinity of the incorporated water molecules ranged from 30 to 65 Å2.

View this table:

The IgG structure is highly asymmetric (Fig. 1) and can be considered a “snapshot” of the broad range of conformations available in solution. The overall shape is between a Y and a T, with a 143° angle between the major axes of the two Fabs (15). The IgG spans 171 Å from the apex of one antigen-binding site to the other. The Fc region is twisted nearly perpendicularly to the planes of the Fabs and shifted some 32 Å from the central dyad relating the two Fabs, so that it packs into the space beneath only one of the Fabs (Fig. 1).

Figure 1

Cα trace of the intact human IgG1 b12. The residues are colored by B value, with lower than average colored blue, average colored white, and higher than average colored red. (A) Front view showing the reach of the two Fabs and packing of the Fc domain underneath the right Fab. The hinge regions and one Fc CH2 are characterized by the highest B values and greatest mobility. (B) Side view of the IgG [rotated 90° from (A)] demonstrating the near-perpendicular twist of the Fc relative to the Fabs. Carbohydrate chains form the contact between the CH2 domains and are illustrated in ball-and-stick. Figure generated with Bobscript (44) and Raster3D (45). See (46) for an interactive image.

The hinge regions form extended structures with some conformational variation in torsion angles between the two chains, reflecting the different relative placement and environment of the two Fabs relative to the Fc domain. One upper hinge forms a spiral arrangement similar to a partially unwound helix; the other forms an extended turn as the polypeptide chain reverses direction to connect the Fab to the Fc. The core hinge region contains two adjacent pairs of cysteine residues, but only one disulfide is observed in the electron density maps. The broken disulfide may be dynamic or may be a result of radiation damage, but probably has no functional importance, as only a single hinge disulfide is necessary for complement-mediated lysis and antibody-dependent cell-mediated cytotoxicity and phagocytosis (16).

Carbohydrate sequencing of b12 reveals two biantennary-branched oligosaccharide chains, but with branching fucose and terminal galactose residues in incomplete occupancy (14). This asymmetry is reflected in the electron density maps where one chain has both terminal galactose residues but lacks the fucose; the other incorporates the fucose but lacks the terminal galactose on the 1,3 arm.

Antibody b12 has a long CDR H3 (18 amino acids) that rises 15 Å above the surface of the antigen-binding site with a Trp residue at its apex (Fig. 2). A patch of acidic residues (17) along one face of this loop may help maintain the vertical projection through charge repulsion. This extended CDR H3 finger-like loop would allow the antibody to probe the recessed CD4-binding site of gp120; all members of a panel of 32 antibodies against the CD4-binding site developed from phage display have a similar length CDR H3 (18). To validate this notion, we designed a synthetic peptide to mimic the crystal structure of CDR H3. The peptide by itself is capable of viral neutralization when coupled to bovine serum albumin (BSA) (19) (Fig. 3).

Figure 2

Prominence of the CDR H3 finger-like loop in IgG1 b12. (A) Side view of the antigen-binding site of the Fv portion of b12. Contributions to the surface from each CDR are indicated. CDR H3 projects 15 Å above the other CDRs. (B) Stereo view of CDR H3 with surrounding final 2F obsF calc electron density contoured at 2.0 σ. Trp100 is presented at the apex of the loop, and five intra-loop hydrogen bonds are indicated by dotted blue lines. The CDR H3 is unlikely to deform upon gp120 binding because of the presence of aromatic residues at the base and charge repulsion of the acidic patch on the inner face. In fact, four crystallographically distinct structures of the CDR H3 are identical [two on this symmetric IgG and two solved as Fab fragments in complex with a peptide (47)]. Molecular surfaces were calculated by using MSMS (48) and a 1.5 Å probe and visualized by using PMV (49). Density figure generated by using Bobscript (44) and Raster3D (45).

Figure 3

Neutralization of HIV-1MN and HIV-1HxB2 by IgG1 b12 and by a synthetic CDR H3 peptide. The peptide was designed as a CDR H3 loop mimic and coupled to BSA. A control peptide Gβ (American Peptide Co., Inc.) does not neutralize when coupled to BSA. HIV-1MN and HIV-1HxB2 were neutralized by using H9 target cells and detection of p24 in ELISA as a reporter assay as described in (5). Filled circles, IgG1 b12 (MN); open circles, IgG1 b12 (HxB2); filled squares, CDR H3-BSA (MN); open squares, CDR H3-BSA (HxB2); filled diamonds, Gβ-BSA (MN); open diamonds, Gβ-BSA (HxB2).

Traditional views suggest that antigen-binding sites for antibodies against proteins are relatively flat. However, extended H3 loops are frequently seen in human antibodies directed against pathogens (20–22) that would allow them to access canyons and clefts on the viral surface (23). It is noteworthy that mouse antibodies do not normally exhibit such long CDR H3 loops, and indeed, few murine antibodies against the CD4-binding site have been described relative to the corresponding plethora of human antibodies of this specificity.

Ideally, we would prefer to dock our b12 structure onto an envelope gp120 trimer, as this is probably the relevant structure for neutralization. However, the only available gp120 structure is that of its monomeric core as a complex with CD4 and a Fab fragment (24). One interpretation of thermodynamic studies on recombinant monomeric gp120 is that CD4 binding might be accompanied by some structural rearrangement in gp120 (25). However, because structures of core gp120 from a primary isolate and a T cell line–adapted virus can be superimposed (24,26) and because earlier studies indicated that b12 and CD4 are sensitive to the same mutations in gp120, it is reasonable to use the CD4-bound core gp120 structure for generation of a docking model that can be tested by mutagenesis.

One hundred computational docking experiments were performed in parallel by using AutoDock (27). Each experiment used a 126 × 126 × 126 Å grid centered on gp120, each using 250,000 energy evaluations, and different, randomly selected initial orientations and translations of the b12 Fv. The best docking model has an energy of –24.7 kcal/mol, which is comparable to those obtained when docking CD4 onto gp120 in which AutoDock recreated the crystallized CD4-gp120 complex (28). This computational docking (Fig. 4A) arrives at a solution essentially identical to that first obtained manually, by using physical models generated from the b12 and gp120 coordinates.

Figure 4

Model of b12 interacting with gp120. (A) Comparison of Fab b12 docking onto gp120 with the CD4-gp120 crystal structure (24). Fab b12 is drawn as a Cα trace in yellow and gp120 in gray, showing the high complementarity of the gp120 and b12 structures, with their corresponding molecular surfaces outlined in the background. The placement of b12 overlaps with that of CD4 (red). (B) Close-up view of the docking. The b12 antibody presents Trp100 at the apex of the extended CDR H3 into the same pocket on gp120 that would be occupied by Phe43 of CD4 (red). Several gp120 residues (Ser365, Asp368, Ile371, Tyr384, and Val430) that are important to b12 binding are illustrated in gray.

The crystal structure of the gp120 complex (24,26) demonstrates that CD4 inserts a loop terminating in Phe43 into a polar pocket in gp120 in order to achieve complementarity (Fig. 4B). An antibody is twice as wide as CD4 (two immunoglobulin domains in width versus one). Thus, the imprint of a Fab onto the neutralizing face of gp120 will be extremely limited by geometric fit. The opening of the CD4-binding face of gp120 between the V1/V2 loop stem and constant region 4 (C4) is 35 Å wide, whereas the width of the Fab combining site is 31 Å. For gp120 in the envelope trimer, space for b12 binding in the lateral direction is bounded by the trimer interface and the 2G12 epitope, as b12 can bind native trimeric spikes (29), and also bind gp120 concurrently with antibody 2G12 (30). The docking shows that b12 may fit snugly onto gp120 by binding an epitope extending from the V1/V2 loop stem across the neutralizing face, with Trp100 at the tip of the H3 loop penetrating the Phe43 pocket (Fig. 4B). Trp100 is a crucial residue for b12 binding to gp120, as it is exclusively selected when CDR H3 residues are randomized (31). Further, mutation to Phe diminishes binding by 50% under conditions for which mutation to Ala, Val, or Ser diminishes binding by 85% (32).

IgG1 b12 and gp120 demonstrate complementary contact surfaces, like fingers fitting into a glove (Fig. 4A). The protruding ridge formed by Ser364 through Asp368 fits into a cleft between CDRs H3 and H2, and the protruding D loop of gp120 fits into a depression formed between CDRs H3, L1, and L3. In the docking model, approximately 2070 Å2 of solvent-accessible surface is buried in the b12-gp120 interface (1030 Å2 on gp120 and 1040 Å2 on b12). The b12 interaction surface, like the footprint of CD4, is centered on the outer domain of gp120 with some additional contact extending to the V1/V2 loop stem, but with minimal contact to the inner domain.

From this docking, a list of likely contact residues was generated. Alanine mutation of seventeen of these and additional neighboring residues supports the docking model (Fig. 5). In fact, several mutations enhance b12 binding, suggesting that b12, like CD4, may recognize gp120 through many main-chain contacts, allowing b12 to be relatively insensitive to side-chain variation.

Figure 5

Validation of the docking model of b12 onto gp120. Results of point mutations suggested by docking of b12 onto gp120 were mapped onto the gp120 surface. Mutation of residues in blue enhances b12 binding (>200% affinity relative to wild type); mutations in red diminish b12 binding (<50% relative affinity); and mutations in black have no significant effect (between 50 and 200% relative affinity). Most mutations are in the outer domain of gp120; see (46) for full characterization of mutants. The docking footprint of b12 on gp120 is outlined with a dashed black line. Variants of HIV-1 clade B isolates that escape neutralization by b12 have been generated in vitro (50) and in vivo (51). A common feature of these neutralization escape variants is the mutation of a proline at position 369, which strongly reduces b12 binding to gp120 in clade B isolates (50). Hence, Pro369 has been indicated with hatched red lines in this figure. Figure generated in PMV (49).

The ability of b12 to neutralize primary viruses is associated with its ability to bind trimeric, as well as monomeric, gp120, whereas the other nonneutralizing CD4-binding site antibodies principally bind only monomeric gp120. Antibody b12 is also unique in its sensitivity to mutations associated with the V1/V2 loop (29) and, in particular, to changes in the V2 stem structure (33). In our model, the b12 Fab fits onto gp120 by contacting the inside face of the V1/V2 loop stem. We suggest that this mode of interaction angles the rest of the antibody bulk away from the trimer interface in an arrangement that permits attachment to the oligomeric spikes on the viral surface. Thus, the likely reason that antibody b12 is capable of potent neutralization of a broad array of isolates is that its interaction with the conserved CD4 epitope is mediated through many main-chain contacts and is angled in such a way that the antibody can access its epitope on the native viral surface.

Trimeric envelope spikes have three equivalent CD4-binding sites. Although the 170 Å reach of the IgG indicates that it could bivalently span two different spikes (34), b12 would probably not be able to bind simultaneously to two binding sites on the same trimeric spike. Nevertheless, the large mass (150 kD) of the IgG molecule compared with that of the envelope trimer is likely to block attachment of the virus to the target cell and/or fusion of viral and cell membranes, even at coating densities well below three IgG molecules per trimeric spike (35).

In conclusion, the structure of an intact human antibody capable of neutralizing a broad range of primary HIV-1 isolates illuminates the surface topography of the antigen-binding site. The highly extended CDR H3 is capable of accessing the vulnerable CD4-binding site of gp120 and may provide leads for antiviral compounds or peptides. Fine mapping of the b12 epitope also facilitates the design of minimized gp120 cores or peptidomimetics. Such structural information should provide new possibilities in the global effort to design an effective HIV-1 vaccine.

  • * To whom correspondence should be addressed. E-mail: wilson{at}, burton{at}


View Abstract

Navigate This Article