Structural Basis of the gp120 Superantigen-Binding Site on Human Immunoglobulins¹

Superantigens (SAgs)⁵ represent a family of molecules, produced by bacteria, Mycoplasma, and viruses, that are capable of stimulating powerfully T and B lymphocytes in an unconventional fashion. T cell SAgs do not require Ag processing, and their interaction with responsive lymphocytes occurs through contacts with lateral surfaces of both the MHC and the TCR. Typically, they activate T cells that express the TCR β-chain (1). Similarly, B cell SAgs target lymphocytes that express a heavy chain using a specific variable (V_H) gene family, independently of the diversity (D_H) and junctional (J_H) gene segments, and in vitro can deliver an activation signal to stimulate their selective differentiation and Ig secretion (2, 3, 4). The best characterized B cell SAg is protein A from Staphylococcus aureus (SpA), which interacts with most, if not all, nonmutated Igs from the V_H3 gene family, but not Igs from other V_H families (5, 6, 7, 8). This V_H-specific binding activity is distinct from the Fc-binding site of SpA (9, 10) and is independent from D_H, J_H, and light chain usage. Structural studies of V_H3⁺ Igs binding to SpA have shown that the binding site is highly conformational and affected by the sequence of three different V_H subdomains, framework region 1 (FR1), complementarity-determining region 2 (CDR2), and FR3. However, the distribution of actual contact sites is unclear, although position 57 in the CDR2 and a lysine at position 75 of the FR3 have been suggested to be important for SpA binding (6, 11, 12, 13).

In addition to this bacterial SAg, there is evidence that viruses can produce B cell SAgs. In contrast to its interaction with immunization-induced Abs through conventional Ag binding which requires the tertiary structure of the heavy and light chain variable regions, the envelope glycoprotein gp120 of HIV-1 reacts also with normal human nonimmune Igs and B cells expressing members of the V_H3 gene family, independently of the light chain isotype, and of D_H and J_H usage. This unconventional interaction results in activation and differentiation of nonimmune V_H3⁺B cells (14). Since the V_H3 gene family dominates (>50%) the human Ab expressed repertoire (15, 16, 17), the SAg properties of gp120 may induce alterations in the B cell repertoire with potential pathologic consequences in HIV-infected individuals (3). For instance, a progressive depletion of V_H3⁺ B cells was seen during late stages of HIV infection (18, 19). However, the molecular mechanisms underlying this B cell-specific depletion remain the focus of investigation. Recently, the SAg residues engaged in binding to V_H3⁺ Igs have been identified (20, 21). Yet, the precise topology of interaction between gp120-encoded residues and Ig residues remains unknown. Also unresolved is the repertoire of V_H3⁺ Igs that can interact with gp120. Here, we have used human mAbs of known V_HD_HJ_H sequences to determine the spectrum of V_H3⁺ Igs capable of gp120 binding and to delineate the structural basis of this interaction by means of molecular modeling.

Recombinant HIV-1_MN gp120 was obtained through the AIDS Research and Reference Reagent Program (Division of AIDS, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, MD) from MicrogeneSys (West Haven, CT). Human monoclonal IgG and IgM expressing different V_H gene family genes were obtained from different sources. Their molecular characteristics are summarized in Table I.

Direct binding was assessed by ELISA. Polystyrene microtiter plates (Maxisorp F96, Nunc Laboratories, Glostrup, Denmark) were incubated overnight at 4°C with 100 μl of gp120_MN (50 ng/well) diluted in borate-buffered saline (pH 8.4). Plates were washed with PBS (pH 7.2) containing 0.1% Tween 20 (Prolabo, Fontenay-sous-Bois, France) (PBS-Tween), and nonspecific sites of the wells were blocked with PBS containing 1% (w/v) of BSA (PBS-BSA) for 2 h at 37°C. A 100-μl volume of the test Ab diluted in PBS-BSA was added to each well, and, after 90 min of incubation at 37°C, the plates were washed again with PBS-Tween. Bound Abs were revealed with an alkaline phosphatase-labeled anti-Ig conjugate. After a further incubation for 1 h at 37°C, the plates were washed with PBS-Tween, and a volume of 100 μl per well of the alkaline phosphatase substrate p-nitrophenyl phosphate (Sigma, St. Louis, MO) (1 mg/ml) diluted in 0.05 M carbonate buffer, pH 9.5, containing 2 mM MgCl₂ was added. Optical density was recorded at 405 nm with a multiscan automatic plate reader (MR 5000, Dynatech Labs, Chantilly, VA).

SpA expresses binding sites for both V_H3⁺ Igs and the constant region domain of IgG. Iodine monochloride modification of SpA selectively inactivates IgG-binding activity of protein A, without affecting its Fab-mediated binding (40, 41). We treated SpA (Sigma) with iodine chloride to inactivate its Fc-binding site (Mod-SpA). The Mod-SpA was then biotinylated. Both forms of SpA were used in an ELISA assay as described above.

Ig amino acid (aa) sequences were obtained from the current GenBank/EMBL data base. They were aligned using the Clustal V algorithm (D. Higgins, European Molecular Biology Laboratory) using default settings. The aa were classified in four groups: nonpolar (Ala, Val, Leu, Ile, Pro, Phe, Met, and Gly); polar (Trp, Ser, Thr, Cys, Tyr, Asn, and Gln); acidic (Asp and Glu); and basic (Lys, Arg, and His). First, the 3-23 germline gene-encoded, gp120 positive, 18/2 Ab was used as a template for sequence comparison of gp120-positive Igs. The 3-23 gene is supposed to be closest to the consensus sequence of V_H3 germline configuration genes. Then, the sequences of gp120-negative Igs were compared with those of gp120-positive Igs. Substitutions leading to expression of an aa from a different aa group were considered nonconservative. For each aa position, the percentage of replacement changes was calculated by dividing the number of nonconservative substitutions by the number of Abs tested. When an aa was unknown, or when there was a deletion, the Ab was not counted.

The structure of the 18/2 Ab was modeled using the coordinates from the crystal structure of a highly sequence homologous human Ig. Calculations were conducted and visualized on a SGI graphics work station using the software InsightII/95.0 (MSI, Cambridge, U.K.). A homology-based 3D model of 18/2 was obtained using the HOMOLOGY facility of InsightII. The resulting preliminary model was refined by performing a side chain rotamer search to alleviate unfavorable interatomic contacts whenever they were present. The solvent accessible molecular surface of 18/2 Ab was calculated using a probe radius of 1.4 Å. Atomic radii used are: C 1.55 Å, H 1.10 Å, O 1.35 Å, S 1.81 Å, and N 1.40 Å.

Two distinct types of Fab-mediated binding interactions with the envelope glycoprotein gp120 of HIV-1 have been described. The first occurs with immunization-induced Abs through conventional Ag binding and requires the tertiary structure of the heavy and light chain variable regions. The second targets normal human nonimmune Igs expressing members of the V_H3 gene family, independently of the light chain isotype, and of D_H and J_H usage (14). To examine the structural basis of the interaction between human Igs derived from the preimmune repertoire and gp120, we selected a panel of 40 nonimmune, human monoclonal Igs expressing V_H segments from the V_H1, V_H3, V_H4, and V_H5 gene families, and derived from patients and cell lines not associated with HIV infection.

To define the extent of the repertoire used by human Igs binding to HIV-1 gp120, we tested the Ig panel for binding to gp120_MN by ELISA. None of V_H1⁺, V_H4⁺, and V_H5⁺ Igs bound to gp120, confirming previous findings that the binding of gp120 to human Igs is restricted to Abs using V_H3 gene segments (14). However, among the 26 V_H3⁺ Igs, only 27% (four IgM and three IgG) bound gp120 (Table I). To see whether this low proportion of gp120 binders was due to a selection bias in the panel used, we tested reactivity of all the Igs with SpA, the prototypic SAg for human V_H3⁺ B cells (5, 6, 7, 8). Comparative binding to Mod-SpA showed that 77% (20 of 26) of V_H3⁺ Igs bound to SpA, indicating that gp120 interacts with only a subset of V_H3⁺ Igs that can convey SpA binding. However, among the gp120 binders, RSP-4, a V_H3⁺ Ig, did not bind Mod-SpA, suggesting that there may be unusual examples in which a somatic mutation interferes primarily with the SpA binding with little or no effects on the gp120 binding.

The observation that all SpA⁺ V_H3⁺ Igs are not positive for gp120 and that all gp120 binders do not interact with SpA suggests that these two SAgs might use different sites to contact V_H3⁺ Igs. This possibility was tested in an experiment designed to probe whether SpA and gp120 target the same sites on V_H3⁺ Igs. A purified V_H3⁺ IgG (SAU) with binding specificity to SpA and gp120 (21) was first incubated with gp120-coated wells; then, biotinylated Mod-SpA was added at different concentrations. Formation of gp120-IgG-SpA-biotin complexes was revealed by an alkaline phosphatase-labeled Extravidin conjugate (Sigma). In the absence of biotinylated Mod-SpA, no IgG binding to gp120 was seen (Fig. 1). In the presence of biotinylated Mod-SpA, there was a dose-dependent binding activity of IgG to gp120, suggesting that gp120 and SpA may use different Ig motifs for SAg binding. However, we cannot rule out the possibility that the same, or overlapping, V_H sites are involved or that gp120 has more rigorous conformational requirements for binding. Also, subtle conformational variations from a prototypic Ig structure might be permissive for SpA binding, while they are unfavorable for gp120 binding. Finally, the bivalency of the IgG molecule could allow one V_H region to bind gp120 on the plate, leaving the other V_H available for Mod-SpA binding.

To characterize the variable gene products that impart gp120 binding, we analyzed in detail the sequences of the V_H3⁺ Igs. It was clear that gp120-positive Igs used a variety of D_H and J_H gene segments, together with diverse V_L and J_L gene elements (Table I). With regard to V_H gene usage, only three V_H3⁺ germline gene donors were encountered, namely 3-23, 3-30, and 3-73 genes, and in all cases the expressed genes were highly homologous (96–100%) to germline configuration. By contrast, the V_H3 genes utilized by gp120-negative Igs were, generally, more divergent from the corresponding germline homologues, encompassing between 2 and 15 aa substitutions (data not shown). It seems therefore that the repertoire of gp120 SAg Ig binders is restricted to a limited number of V_H3 genes and that gp120 binding is sensitive to somatic mutations. These characteristics contrast with the diversity of the V_H3 genes that impart SpA binding. The positive Igs were derived from eight V_H3 germline genes (3-7, 3-11, 3-21, 3-23, 3-30, 3-30.3, 3-33, and 3-73), and the 3-23 and 3-30 genes were the most frequently encountered. This finding is in agreement with previous results which were interpreted as evidence of hierarchy in the SpA binding affinity conveyed by different V_H3 genes, in which V3-23 was predominant (6).

The aa sequences of the V_H3⁺ Igs were analyzed to determine whether there was a common motif that could account for gpl20 binding activity. To enhance the significance of the sequence comparisons, three Igs, TaaO, NANUC1, and NANUC2, with binding specificity for gp120 (42) were added to the Ig panel. NANUC1 and NANUC2 are human Igs encoded by the 3-30 V_H gene (43). TaaO is a mouse Ig belonging to Ig clan 3, which is a group of structurally related Ig V_H gene families from different species, the greatest homologies being in the FR1 and FR3 (44). Since Igs expressing the 3-23 V_H gene are the most frequently expressed in the human repertoire (see Discussion), we used the 3-23 germline gene-encoded 18/2 Ab as a template to align the aa sequences of the Ig panel (Fig. 2). First, we calculated the percentage of nonconservative aa substitutions in the sequences of the nine gp120-positive proteins, the six Igs represented in Table I and the three Igs described above. This sequence comparison revealed aa conservations, mainly in the FR regions. With the exception of FR2 which is inaccessible to the solvent (45), some of these conserved residues likely play a role in gp120 recognition by V_H3⁺ proteins. Reciprocally, the substituted positions probably do not interfere with gp120 binding ability (Fig. 3).

Since 19 V_H3⁺ Igs were found to be gp120 nonbinders, comparison of their sequences with those of the V_H3⁺ Ig gp120 binders was made to identify specific regions of the positive Igs that might account for SAg binding. By superimposing the conserved residues identified among the gp120 positive V_H3⁺ Igs and the nonconservative substitutions found in sequence comparison with V_H3⁺ Ig gp120 nonbinders (Fig. 3), it became apparent that the regions critical for gp120-Ig interaction might lie in the FR1 (aa positions 10, 13, 19, and 23), in the H1 (aa positions 28 and 32), in the H2 (aa position 54), and in the FR3 (aa positions, 59, 64, 65, 75, 79, 81, 82a, 83, and 85).

To define the surface residues that may participate in gp120 SAg binding, we constructed a model for the structure of 18/2 Ab based on known structures of Ig domains. Within the limitations of such models, certain inferences can be drawn from the basic similarities of highly homologous Igs to predict residue exposures and spatial orientations. The heavy chain of a crystallized human IgM Fv fragment (Protein Database code 1igm) was selected for modeling (46). The potential putative contact sites are shown in Fig. 4. The side chains of all the residues identified and the backbone of the three glycines are solvent exposed. Moreover, most of the residues are located on the face of the Ab opposite to the interface between the heavy and light chains. On that side, they form essentially two major clusters extending from residue Lys⁷⁵ to residue Lys⁶⁴. The larger is composed of Lys⁷⁵, Ala²³, Tyr⁷⁹, Arg¹⁹, Gln⁸¹, and Asn^82a and has a linear ridge-like shape. The second cluster encompasses residues Tyr⁵⁹, Gly⁶⁵, and Lys⁶⁴, which are in close proximity and exhibit a triangular configuration. Residue Gln⁵⁴ on one hand, and residues Gln¹³ and Arg⁸³ on the other, are located on either side of this patch. The remaining residues fall rather away from this area, i.e., in the conventional Ag binding site (Thr²⁸ and Tyr³² of H1), or in the lower right hand of the figure (Gly¹⁰ of FR1). Overall, the convex shaped area contains polar residues such as Gln⁸¹ and Asn^82a and is bordered by four positively charged residues arranged perpendicularly in two pairs (Lys⁷⁵-Arg¹⁹, and Lys⁶⁴-Arg⁸³). The interatomic distances between the C_β atoms of Lys⁷⁵ and Lys⁶⁴ and of Arg¹⁹ and Arg⁸³ approximate 26 and 17 Å, respectively. Thus, the gp120-binding site of Igs consists of nonsequential residues, which can be modeled to fold in space into a solvent-exposed accessible compact region.

In this article, we addressed two issues regarding the structural and functional basis of gp120 SAg binding to human Igs. First, we probed the repertoire of Ig V_H genes that can impart binding to this SAg. Second, we pinpointed aa residues of the Ig molecule that contact the gp120 SAg.

This study showed that gp120 SAg binding is restricted to V_H3⁺ Igs and that it requires contribution from a limited number of V_H3⁺ gene products. Among the 10 V_H3⁺ genes expressed by the Ab panel tested, only 27% (3-23, 3-30, and 3-73 V_H genes) were used by gp120 binders (40% of V_H3⁺ IgM and 18% of V_H3⁺ IgG). In agreement with this low proportion of reactive Igs, previous studies showed that only ∼25% of V_H3⁺ human Igs and B cells bind gp120 (14, 42). Since in these studies, the 3-15 V_H3 gene product was also identified by an antiidiotypic reagent to encode gp120 binding, it appears that, in toto, only four V_H genes, 3-23, 3-30, 3-73, and 3-15, have been found to be responsible for gp120 SAg binding. This restriction contrasts with the high frequency of SpA binding among V_H3⁺ Igs found here (80% of V_H3⁺ IgM and 75% of V_H3⁺ IgG) and elsewhere (85% of V_H3⁺ IgM, and 65% of V_H3⁺ IgG) (6). We also noted that gp120 may be recognized by V_H3⁺ Ig that do not interact with SpA (4). It is of further interest that some SpA-negative, V_H3⁺ IgG were found to bind pFv, another well-characterized SAg for B cells, and that pFv is also recognized by V_H6⁺ IgM that do not interact with SpA (47). Thus, the three prototypic SAgs for human V_H3⁺ B cells, SpA, gp120, and pFv, seem to target overlapping subsets of the Ab repertoire (4).

A possible explanation of the relatively low frequency of gp120 Ig binders is that the SAg-Ig interaction is highly sensitive to mutation. We found that the V_H genes used by gp120-positive Igs exhibit >96% aa similarity to V3-30 germline gene (Table II) and that V3-30 gene-encoded Igs with <96% similarity lose this capacity, thus suggesting that hypermutation may have an adverse effect on gp120 binding. This inverse correlation between the degree of somatic mutation and gp120 binding is similar to the previously reported binding pattern described for SpA (22). However, gp120 binding may be even more sensitive, as most V3-30 and V3-23 gene products tested showed binding activity for SpA, whether or not gp120 binding activity was retained (Table II). It is also possible that certain combinations of D, J_H, V_L, and J_L gene elements create an unfavorable conformation of the variable domain and abrogate gp120 binding. In parallel studies, it was found that protein Fv interacts with a greater set of V_H3⁺ IgG Abs than does SpA, suggesting that hypermutation more commonly has an adverse effect on SpA binding than upon protein Fv binding (47). These observations imply that the positions and/or the nature of the mutations involved in the contact sites on Igs of these B cell SAgs are different.

Since in this study we used the recombinant gp120_MN, it is relevant to emphasize the influence of the variability of clinical gp120s on SAg binding. Using gp120s from clades of various geographical origins, previous studies showed that SAg binding activity is well-conserved (21), a feature reminiscent of the conservation of the CD4⁻and the chemokine receptor-binding sites of gp120 among different viral clades. It is remarkable that sequence-divergent gp120s from both laboratory and field isolates exhibit the SAg property, suggesting that SAg binding is a complex process which depends on interactions with motifs that are influenced by the tertiary and quaternary structures of the gp120 and Ig proteins.

The finding that only a limited number of V_H3⁺ genes (3-23, 3-30, 3-73, and 3-15) can impart gp120 binding deserves consideration with regard to the functional consequences of SAg interaction on the B cell repertoire. On the surface, it might seem that such a limited number of V_H genes would result in minimal repertoire triggering. However, three lines of evidence suggest that Igs encoded by this restricted set of V_H genes have the potential to impact significantly the human B cell repertoire. First, studies of human V_H gene polymorphisms showed that the 3-23 V_H gene is present in all individuals tested to date and that some subjects carry up to six copies (48). Secondly, its is recognized that the 3-23 V_H gene comprises ∼25% of V_H3 rearrangements expressed in B cell clones (49, 50, 51, 52). Finally, a recent study of human V_H gene utilization throughout ontogeny revealed that 3-23, 3-30, and 3-15 V_H genes are expressed at the pre-B, the immature B, and the mature B cell stages of normal adults and that 3-23 and 3-30 V_H genes are overexpressed at all stages of B cell development (53). Given this overexpression, it may be concluded that gp120 targets a sizable fraction of the human germline-encoded B cell repertoire. As a result, ∼25% of human B cells bear V_H3⁺ Igs with a binding capacity to gp120 (42).

In addition to showing that SAgs for human B cells target distinct subsets of the Ab repertoire, our results indicate that the V_H domain of a single Ig molecule can confer distinct reactivities with the viral SAg gp120 and the bacterial SAg SpA. Reciprocally, we found that some VH3⁺ Abs bind both gp120 and SpA. These observations imply that, apparently, these two SAgs are recognized by V_H3⁺ Igs, through distinct sites or via overlapping contact residues. This property is reminiscent of parallel characteristics of T cell SAgs, where TCR binding was shown to differ for a bacterial SAg, staphylococcal enterotoxin E, and the retroviral SAg MTV-9 (54). In addition, a viral and a bacterial SAg were found to interact with distinct sites of the Vβ domain of a single TCR (55). When two naturally occurring alleles of a TCR Vβ gene were tested for staphylococcal enterotoxin E binding, their functional reactivities also differed (54). It seems therefore that, as is the case for T cell SAgs, there are distinct recognition sites for viral and bacterial SAgs for B cells. Nevertheless, it has been shown that Ig binding to protein Fv, a B cell SAg of endogenous origin, can be abrogated by SpA, or a single domain of SpA (22, 47).

Comparison of the V_H3⁺ sequences from gp120 Ig binders and nonbinders provides a basis to delineate the interaction site of gp120 with the Ig molecule. Since FR2 is not accessible to the solvent and, hence, is unlikely to be involved in gp120 binding (45), possible determinants that could be critical for gp120 binding were found in the FR1 (positions 10–23), the H1 (positions 28 and 32), the H2 (position 54), and the FR3 (positions 59–85), as there were nonconservative changes in these regions. Remarkably, 13 of the 16 aa positions identified map outside the conventional Ag-binding site, in the FR1 (residues 10, 13, 19, and 23), and the FR3 regions (residues 59, 64, 65, 75, 79, 81, 82a, 83, and 85). In addition, three positions localized in the H1 and H2 loops were identified that might influence gp120 binding.

Localization of the residues in Ig variable regions that contact SpA has been also the focus of interest. Sequence analyses of SpA-positive human Igs led to the suggestion that residues in FR1 (position 28), the carboxyl-terminal end of CDR2 (positions 57 and 59), and FR3 (positions 75, 76, 80, 82a, and 84) are critical for SpA binding (6, 11, 12). Recent experiments where regions of an SpA-positive human-derived Ab were exchanged with those from an SpA-nonreactive mouse Ab provided further evidence that the three regions, FR1, CDR2, and FR3, are required to interact simultaneously with SpA for binding to occur (13). Here, we have identified 16 aa positions localized in the FR1, H1, H2, and FR3 which correlate with gp120 binding. Among these, only residues 28, 75, and 82a overlap with the SpA binding site which encompasses at least eight residues.

Well-established similarities between the structures of the Ig variable domains and those of the TCR invite comparisons of their mode of interaction with SAgs. It may therefore be valid to extend this analogy to infer structural attributes of a B cell SAg from the recent three-dimensional structure characterization of T cell SAgs. Crystal structures of bacterial T cell SAgs, S. aureus enterotoxins C2 and C3, show that binding occurs via the CDR1, CDR2, FR3, and involves 12 contact residues of the TCR Vβ chain and that residues implicated in TCR binding sit in a depression at the center of the surface while secondary structural elements protrude on either side (56). In other studies of T cell SAgs, the interaction with the TCR was also shown to involve the β hairpin loop of the Vβ FR3 region, which is structurally homologous to the FR3 region of the Ig molecule (55). Our data are consistent with a model in which the putative binding site for gp120 comprises 16 residues located in the solvent-exposed regions and strands of V_H3⁺ Igs, primarily outside the conventional binding site. In this model, the points of contact between the two molecules represent a virtually continuous area, flanked by four basic side chains, and surrounded by aa residues 54 and 13 on either side. If charge and shape complementarity are to be observed, the contact surface presented by gp120 should be overall concave and acidic. Furthermore, the similarity in structures with the TCR would allow the V_H FR3 region (corresponding to the hypervariable loop 4 in Vβ) to make direct contacts with gp120 residues critical for binding (20, 21).

Recently, the x-ray crystallographic structure of gp120 has been elucidated at 2.5 Å resolution. It shows that a broadly neutralizing human mAb directed to the CD4-induced gp120 epitope uses essentially its heavy chain for binding. Additionally, the interaction between this Ab and gp120 involves a hydrophobic central region flanked by charged regions (57, 58). Our findings revealed that the potential putative contact sites of Ig SAg binding form essentially two major clusters and contain nonsequential polar residues bordered by positively charged residues. It will be important to determine the atomic structure of a complex of gp120 complexed to a human V_H3⁺ Ig exhibiting the B cell SAg activity.

In a recent report (61), murine repertoire analyses provided the first evidence that in vivo experimental challenge with a form of SpA, the prototypic B cell superantigen, can induce specific large scale B cell clonal specific tolerance by deletional and nondeletional mechanisms. These findings represent independent support of a role for envelope protein, gp120, in the induction of superantigen-induced B cell clonal defects during clinical HIV infection.

We are indebted to Drs. M. Ohlin (Lund, Sweden), J. Rauch (Montreal, Canada), P. Lafaye (Paris, France), C. Newkirk (Montreal, Canada), C. Ravirajan (London, U.K.), M. Ehrenstein (London, U.K.), and P. Lambin (Paris) for the mAbs. We thank the AIDS National Institute of Allergy and Infectious Diseases, Research and Reference Reagent Program, Division of AIDS, National Institutes of Health, from MicrogeneSys for gift of recombinant proteins.