Events Calendar Post and Go
| Change Password | Change User Info | CiteTrack Alerts | Access Rights | Subscription Help | Sign Out
HelpSubscriptionsFeedbackSign In

Abstract of this Article
PDF Version of this Article
Supporting Online Material
Download to Citation Manager
Alert me when:
new articles cite this article
Search for similar articles in:
  Science Online
Search Medline for articles by:
Calarese, D. A. || Wilson, I. A.
This article appears in the following Subject Collections:

Antibody Domain Exchange Is an Immunological Solution to Carbohydrate Cluster Recognition

Daniel A. Calarese,1 Christopher N. Scanlan,2,5 Michael B. Zwick,2 Songpon Deechongkit,3 Yusuke Mimura,5 Renate Kunert,6 Ping Zhu,7 Mark R. Wormald,5 Robyn L. Stanfield,1 Kenneth H. Roux,7 Jeffery W. Kelly,3,4 Pauline M. Rudd,5 Raymond A. Dwek,5 Hermann Katinger,6 Dennis R. Burton,1,2* Ian A. Wilson1,4*

Human antibody 2G12 neutralizes a broad range of human immunodeficiency virus type 1 (HIV-1) isolates by binding an unusually dense cluster of carbohydrate moieties on the "silent" face of the gp120 envelope glycoprotein. Crystal structures of Fab 2G12 and its complexes with the disaccharide Man{alpha}1-2Man and with the oligosaccharide Man9GlcNAc2 revealed that two Fabs assemble into an interlocked VH domain-swapped dimer. Further biochemical, biophysical, and mutagenesis data strongly support a Fab-dimerized antibody as the prevalent form that recognizes gp120. The extraordinary configuration of this antibody provides an extended surface, with newly described binding sites, for multivalent interaction with a conserved cluster of oligomannose type sugars on the surface of gp120. The unique interdigitation of Fab domains within an antibody uncovers a previously unappreciated mechanism for high-affinity recognition of carbohydrate or other repeating epitopes on cell or microbial surfaces.

1 Department of Molecular Biology, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA 92037, USA.
2 Department of Immunology, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA 92037, USA.
3 Department of Chemistry, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA 92037, USA.
4 Skaggs Institute for Chemical Biology, The Scripps Research Institute, 10550 North Torrey Pines Road, La Jolla, CA 92037, USA.
5 The Glycobiology Institute, Department of Biochemistry, University of Oxford, Oxford OX1 3QU, UK.
6 Institute for Applied Microbiology, University of Agriculture, Vienna, Austria.
7 Department of Biological Science and Institute of Molecular Biophysics, Florida State University, Tallahassee, FL 32306, USA.

* To whom correspondence should be addressed. E-mail:,

Neutralizing antibodies are elicited by most, if not all, successful vaccines (1, 2). However, immunogens that are able to elicit neutralizing antibodies to a broad range of primary HIV-1 isolates have not been found. Nevertheless, a few rare, broadly neutralizing monoclonal antibodies that have been isolated from patients protect against viral challenge in animal models (3). Their epitopes include regions on gp41 [2F5 (46) and 4E10 (6, 7)], the CD4-binding site of gp120 [b12 (810)], and part of the carbohydrate-masked "silent" face of gp120 [2G12 (11, 12)]. A molecular understanding of the binding of these broadly neutralizing antibodies to their cognate envelope epitopes should facilitate rational HIV-1 vaccine design.

In this research article, we focus on the broadly neutralizing human antibody 2G12, which binds with nanomolar affinity to gp120. This antibody recognizes terminal Man{alpha}1-2Man–linked moieties (13, 14), contributed by oligomannose-type sugars that form a cluster on the silent face of gp120 (15). This face is designated "silent" first, because the oligosaccharides shield potential antigenic epitopes; second, because oligosaccharides attached to the viral coat proteins are processed by the host and are, therefore, unlikely to be immunogenic; and third, because glycosylated proteins are synthesized as a collection of glycoforms in which multiple sugars can be present at a single site, which dilutes any potential antigenic response (16). Furthermore, carbohydrate-protein interactions are usually much weaker (in the micro-molar range) than protein-protein interactions (1719) and restrict antibodies from approaching their expected range of nanomolar binding affinities. Nevertheless, antibody 2G12 binds with high affinity to carbohydrate epitopes on gp120.

Crystal structures of Fab 2G12. Crystal structures were determined for the unliganded Fab 2G12 at 2.2 Å resolution, for the Fab bound to oligosaccharide Man9GlcNAc2 at 3.0 Å, and for the Fab bound to disaccharide Man{alpha}1-2Man at 1.75 Å (table S1) (20). The asymmetric unit in each crystal form contains two Fab molecules that form a tightly packed dimer via a three-dimensional swap (21) of their VH domains (Fig. 1, A and B). This dimeric assembly is not observed in over 250 Fab structures deposited in the Protein Data Bank. Although the variable (VH and VL) and constant regions (CH1 and CL) are each structurally similar to their corresponding domains in other Fab molecules, the VH domain exchange in 2G12 is accommodated by twisting the variable regions with respect to the constant region when compared with their standard orientation in a Fab (fig. S1). The VH domain-exchanged dimer lacks the highly conserved ball-and-socket joint (22) between VH and CH1 that is believed to play a key role in the flexibility of the variable domains with respect to the constant domains, although the ball-and-socket residues are present (fig. S2).

 Fig. 1. Novel architecture of antibody 2G12 and structural factors that promote the Fab VH/VH' domain exchange. (A) Monomer of Fab 2G12 in the crystal showing that the VH clearly separates from its normal interaction with the VL. The light and heavy chains are shown in cyan and red, respectively. The monomer does not exist in the crystal, but only in the context of the domain-swapped dimer. (B) Structure of the two domain-swapped Fab molecules, as they assemble in the crystal. Both light chains are shown in cyan, with the heavy chains from Fab 1 and Fab 2 shown in red and purple. The distance between the two combining sites is indicated. (C) Elbow region between the constant heavy and variable heavy domains illustrates the domain exchange. The linker region between VH' and CH1' is shown as a ball-and-stick scheme with corresponding 2FobsFcalc electron density contoured at 1.5 {sigma}. The location of the unusual ProH113 that may favor the domain exchange is shown. (D) Close-up view (ball-and-stick scheme) of the VH/VH' interface between the variable heavy domains. Potential hydrogen bonds are shown with dashed black lines. We generated the figures using programs Bobscript (64), Molscript (65), and Raster3D (66). [View Larger Version of this Image (72K GIF file)]

The VH domains within the dimer are related by a noncrystallographic two-fold symmetry axis of 178.5°, such that the two Fabs are arranged side-by-side with their respective combining sites facing in the same direction and separated by approximately 35 Å. Analysis of the Fab 2G12 structure reveals three factors, resulting from somatic mutations, that appear to promote domain exchange: weakening of the VH/VL interface contacts (closed interface), an unusual sequence and structure of the elbow region connecting the VH and CH1 domains (hinge loop), and the creation of a favorable VH/VH' interface (open interface) (23). Thus, all of the key factors that have been suggested to promote protein domain exchange and favor stabilization of the dimer over the monomer are found here (21).

The VH/VL interface in 2G12 is perturbed by the absence of a highly conserved interaction between antibody VH and VL domains that is also found in {alpha}ß T cell receptors (24). GlnL38 and GlnH39 [94% and 97% conserved, respectively (25) (fig. S3] usually hydrogen bond to each other at the base of the combining site, but, in 2G12, position H39 is a rarely observed arginine residue (0.7%) that is too distant (almost 4 Å) from GlnL38 to interact.

Comparison of Fab 2G12 with other Fab structures shows that the connection between VH and CH1 is also unusual, such that the VH domain pivots around residue ProH113 to promote domain swapping (Fig. 1C). A proline residue in the elbow region (or hinge loop) at H113 is relatively uncommon, as it occurs in only 1.8% of known sequences, and this is the first Fab structure described with proline at this position. Usually, serine is the most prevalent residue (95.2%) at this location. Proline residues occur in connecting hinge loops in many other domain-swapped or oligomerizing proteins (21, 26), and the unique {phi} constraints that a proline residue imposes on the peptide backbone appear to facilitate domain interchange. This conformation of the hinge loop appears to be stabilized in the domain-exchanged structure by hydrophobic interactions between ProH113 and ValH84 (27). ValH84 occurs in only 5% of known antibody sequences and is sometimes found with ProH113 in other antibodies [analysis performed with Kabatman (28)].

The complementarity of the VH/VH' interface [shape complementarity (Sc) coefficient 0.73 (29)] is illustrated by an extensive hydrogen bonding and salt bridge network (Fig. 1D) (30) with a total of 10 hydrogen bonds and 136 van der Waals interactions (31). Of the hydrophilic interface residues, only ArgH57 is uncommon (1.4%). At the bottom of this interface, an extensive hydrophobic patch is created by IleH19, IleH19', PheH77, PheH77', TyrH79, and TyrH79' (32). In addition, {pi}-stacking interactions occur between residues TyrH79 and TyrH79'. IleH19 and PheH77 are rare at these positions (0.11% and 0.19%, respectively) and arise from somatic mutation. A total of 1245 Å2 of molecular surface is buried at this VH/VH' interface (29). This area is substantial compared with the standard VH/VL interface in antibodies, which in 2G12 buries 1690 Å2 of molecular surface.

Oligomeric state of 2G12. To determine whether the 2G12 Fab dimer exists in solution, we examined the Fab oligomeric state by sedimentation equilibrium analytical ultracentrifugation (fig. S4) and by gel filtration (Fig. 2A) (20). In both experiments, Fab 2G12 exists predominantly (80 to 100%) as a dimer in solution. We next examined the configuration of the intact IgG1 2G12 to rule out the possibility that the Fab is only capable of domain swapping when untethered from the Fc fragment of the IgG (33). The sedimentation coefficient (at 20°C in water, S20,W) of 2G12 is unusually high when compared with other HIV-1–specific human IgG1s and previously published values for human IgG1s (Fig. 2B) (20), consistent with a more compact linear configuration, as opposed to a Y- or T-shape of the typical antibody molecule. Furthermore, negative-stain electron microscopy of unbound 2G12 IgG1 (20) and 2G12 IgG1 bound to SOS gp140 [a covalently constrained gp120/gp41 molecule (34)] provided clear images of the antibody in an unusual, extended linear conformation (not the normal Y- or T-shaped configuration) (Fig. 2C). It is noteworthy that these data are all consistent with Fab domain exchange in 2G12, whether as Fab fragments, or as the intact IgG1 or when the IgG1 is complexed to gp120.

 Fig. 2. Biophysical evidence for a domain-exchanged dimer of 2G12. (A) Gel filtration of Fab 2G12 and b12 from papain digests. Protein concentration shown on the y axis was measured by UV absorbance. Fab 2G12 elutes from the column at a molecular mass of ~100 kD, whereas a control Fab (b12) elutes at ~50 kD. The molecular masses suggest that Fab 2G12 exists almost entirely as a dimer in solution, whereas Fab b12 is present as the expected monomer. The completeness of the papain digests and the molecular masses of the Fab monomers were confirmed by SDS–polyacrylamide gel electrophoresis (67). (B) Sedimentation coefficients of IgG1 2G12 relative to other human IgG1 molecules (b6, b12, and 2F5, all antibodies against HIV-1). The y axis shows the relative concentration (measured by UV absorbance) of the protein at that particular S value. The S20,W value of 2G12 IgG1 was demonstrably higher (7.39) than other human IgG1 molecules, which had S20,W values between 6.50 and 6.89. Previously published S20,W values for human IgG1 molecules are around 6.6 (68, 69). Thus, 2G12 is an outlier, in agreement with the streamlined structure of the IgG1 that would arise from the domain-swapped dimer of its Fabs. (C) Negative-stain electron microscopy of unbound b12 IgG1 (top row), unbound 2G12 IgG1 (middle row), and 2G12 IgG1 bound to SOS gp 140 [bottom row (34)] shows that the 2G12 IgG1 (bound and unbound) adopts an unusual linear shape as opposed to a classic Y shape such as that of b12. This linear shape corresponds to a parallel arrangement and intimate association of Fabs of the IgG1 and supports the domain-swapped Fab dimer as the functionally relevant form of the antibody. [View Larger Version of this Image (78K GIF file)]

Carbohydrate specificity and binding sites of 2G12. 2G12 recognizes Man9GlcNAc2 moieties (Fig. 3A) covalently attached to gp120 (13, 14). To explore the binding specificity, we cocrystallized Fab 2G12 with Man9GlcNAc2. Although the cocrystals were highly anisotropic and diffracted only to modest resolution (3 Å), the electron density for the Man9GlcNAc2 is unusually well defined for a carbohydrate ligand (Fig. 3B), albeit with an increase in B values (as expected) farther from the protein surface. The two branching points (at sugars 3 and 4', Fig. 3A) of the Man9GlcNAc2 are clearly visible, which permits an unambiguous interpretation of the electron density. We also cocrystallized 2G12 with the disaccharide Man{alpha}1-2Man, which also inhibits binding of 2G12 to gp120 (13). In this high-resolution structure (1.75 Å), the Man{alpha}1-2Man density is extremely well defined, and its conformation is within its preferred range (Fig. 4A) (35). These two cocrystal structures indicate that the 2G12 domain-exchanged dimer contains multiple, distinct binding sites for carbohydrate: two correspond to the normal antibody combining site and two to novel sites within the VH/VH' interface generated in the domain exchange (Fig. 3C).

 Fig. 3. Interactions of the Fab 2G12 dimer with Man9GlcNAc2. (A) Chemical structure of Man9GlcNAc2. Red sugars make contacts with Fab 2G12 at the primary binding site (conventional combining pocket), whereas blue sugars contact Fab 2G12 at the secondary binding site (the unusual VH/VH' interface). N-Acetylglucosamine residues are colored in purple. (B) Man9GlcNAc2 bound to the primary binding site of Fab 2G12, with corresponding 2FobsFcalc electron density contoured at 1.6 {sigma}. (C) Overall structure of the Fab 2G12 dimer bound to Man9GlcNAc2. A total of four Man9GlcNAc2 moieties are bound to each Fab dimer. The Man9GlcNAc2 moieties are colored corresponding to those in (A). For stereo and orthogonal views of (C), see fig. S5. (D) The relative apparent binding affinities of Fab 2G12 mutants are indicated on the structure. Results are shown relative to wild-type Fab 2G12 binding of gp120JR-FL (100%). Residues that are black indicate that an alanine substitution at that position resulted in no significant effect (50% to 200% relative to wild type) on the apparent binding affinity of 2G12 for gp120JR-FL, whereas residues in red (labeled) indicate an alanine substitution at that position resulted in more than a 50% decrease in the apparent binding affinity of 2G12 for gp120JR-FL. For more detailed affinity values, see table S2. We generated the figures using programs Bobscript (64), Molscript (65), and Raster3D (66) [View Larger Version of this Image (73K GIF file)]

 Fig. 4. The antibody–combining site interactions with the disaccharide Man{alpha}1-2Man. (A) The 2FobsFcalc electron density for Man{alpha}1-2Man is contoured at 1.7{sigma} and the CDR loops are labeled. (B) The combining site showing Fab atoms within hydrogen bonding distance (dotted lines) of Man{alpha}1-2Man. The Fab heavy chain and light chain are shown in purple and cyan, respectively. We generated the figures using programs Bobscript (64), Molscript (65), and Raster3D (66). [View Larger Version of this Image (31K GIF file)]

In the primary combining site of the Man9GlcNAc2 complex, 2G12 contacts four sugars (3, 4, C, and D1) in the D1 arm, with the majority of contacts (85%) with the terminal Man{alpha}1-2Man disaccharide (Fig. 3, A, B, and C). In the disaccharide complex, Man{alpha}1-2Man occupies only the two conventional combining site pockets, which are separated by about 35 Å; this binding mode suggests that these are the higher-affinity sites for this mannose linkage. The 2G12 contact residues with the disaccharide in the antigen-binding pocket are L93 to L94 (CDR L3), H31 to H33 (CDR H1), H52a (CDR H2), and H95 to H100D (CDR H3). A total of 226 Å2 of molecular surface from Fab 2G12 and 220 Å2 of molecular surface from Man{alpha}1-2Man is buried during complex formation, with 12 hydrogen bonds and 48 van der Waals interactions in each antigen-binding site (Fig. 4B).

Competition studies confirm that the interaction of the D1 and C mannose residues (Man{alpha}1-2Man) with the primary binding site cannot account for the large increase in affinity observed when 2G12 binds to Man9GlcNAc2 (Fig. 5) (20). Additional antibody contacts with the D1 arm 3 and 4 mannose residues in the primary combining site presumably provide extra favorable interactions with Man9GlcNAc2. AspH100B, which is oriented differently in the Man9GlcNAc2 complex, hydrogen bonds to the branching mannose residue 3, whereas TyrL94 hydrogen bonds to mannose residue 4. The buried surface area is larger, ranging from 350 to 450 Å2 for the Fab and 330 to 450 Å2 for Man9 GlcNAc2 in the two primary binding sites.

 Fig. 5. Inhibition of 2G12 binding to HIV-1 gp120. The median inhibitory concentration (IC50 values) of mannose relative to the IC50 value of different carbohydrates. Man9GlcNAc2 inhibits binding of monoclonal antibody 2G12 to gp120JR-FL better than mannose (by more than 200 times) and better than the disaccharide Man{alpha}1-2Man (by more than 50 times). Surprisingly, fructose is a better inhibitor than mannose. The structure of fructose, when docked into the primary combining site, can mimic positions of four of the oxygen atoms of mannose and can also potentially make additional hydrogen bonding interactions compared with mannose. No other simple sugars or mannose disaccharides with other linkages inhibit 2G12 binding to gp120. [View Larger Version of this Image (24K GIF file)]

The specificity of the primary combining site of 2G12 for Man{alpha}1-2Man at the tip of the D1 arm of Man9GlcNAc2 stems from a combination of several factors. First, the primary combining site forms a deep pocket that can only accommodate terminal sugar residues. Second, the highly complementary geometry of the binding site enables formation of hydrogen bonds specific to Man{alpha}1-2Man. Last, the additional interactions with the mannose 3 and mannose 4 sugars provide specificity for the Man{alpha}1-2Man linkage at the tip of the D1 arm.

The secondary binding site is formed by the VH/VH' domain swap, which creates a surface not previously described in antibody structures. The D2 arms of the symmetry-related Man9GlcNAc2 residues in the crystal interact with this composite VH/VH' surface, demonstrating that it can provide two additional sugar-binding sites (Fig. 3C). The VH/VH' interface interactions are mainly with the central mannose A of the D2 arm, but contacts are also made with the D2 and 4' sugars. Furthermore, the carbohydrate chain runs parallel to the antibody surface in a relatively shallow binding site. Hence, it is not clear whether this secondary binding site is as restricted for the D2 arm as compared with the highly specific D1 arm interaction in the primary binding site. Given that the D1 arm of the same Man9GlcNAc2 occupies the higher-affinity primary combining site of a crystallographically related Fab 2G12 molecule, it is possible that the secondary binding site could also interact with free D1 or D3 arms, but these interactions are not observed here because of crystal packing.

The two independent Man9GlcNAc2 moieties in the asymmetric unit differ slightly in their interaction with the VH/VH' interface, but a total of 280 to 310 Å2 of molecular surface from Fab 2G12 and 250 to 290 Å2 from Man9GlcNAc2 is buried during complex formation. Eight to nine hydrogen bonds and 22 to 26 van der Waals contacts are made in each VH/VH' interface–binding site. Although these secondary binding site interactions arise from the juxtaposition of four Fab-carbohydrate complexes in the crystal lattice, these additional binding sites are a consequence of the unique assembly of the domain-exchanged Fabs and are likely to emulate the high-affinity interaction of the antibody with the dense array of oligomannose sugars on the surface of gp120.

Mutagenesis of 2G12. Mutagenesis of Fab 2G12 was carried out to investigate the role of domain exchange and multivalent interactions in the binding of 2G12 to gp120. Residues in 2G12 that were suspected to play a role in domain exchange, as well as in ligand binding, were replaced by alanine residues and assayed for binding to gp120JR-FL (Fig. 3D; table S2) (20). In some instances, where the germline residues or somatic mutations involved were rare, reverse mutations to the residue encoded by the closest germline gene were introduced. Alanine substitution of many primary combining site residues abolished 2G12 binding to gp120JR-FL. Mutations in sites believed to facilitate domain exchange (VH/VH' interface residues IleH19, ArgH57, PheH77, and TyrH80 and VH/CH1 "elbow" residues ValH84 and ProH113), all promote loss of 2G12 binding to gp120JR-FL. In addition, alanine substitutions of many of the residues involved in binding the D2 arm of Man9GlcNAc2 in the secondary binding site decreased gp120 binding and provided further evidence that this VH/VH' interface plays a role in multivalent binding of 2G12 to gp120.

Biological significance of the 2G12 domain-swapped dimer. Biochemical, biophysical and crystallographic evidence indicates that the VH domains of antibody 2G12 exchange in its two Fab regions so as to form an extensive multivalent binding surface composed of the two conventional combining sites and a homodimeric VH/VH' interface. The 2G12 VH/VH' interface is composed of many conserved germline-encoded residues, but with three uncommon mutations (IleH19, ArgH57, and PheH77) that appear to promote stabilization of this interaction. The proline at position H113 also appears to promote the VH/VH' domain exchange, and the unusual extended conformation of the hinge peptide in the elbow region appears to be stabilized by hydrophobic interactions between ProH113 and ValH84. Analysis of the Kabat antibody-sequence databases with Kabatman (28) yielded no other heavy-chain sequences with the exact combination of IleH19, ArgH57, PheH77, and ProH113, presumably because they arise from independent somatic mutation events. However, one could certainly envision that other combinations of mutations could promote domain exchange and favorable VH/VH' interactions.

Recognition of carbohydrates on HIV-1 by an antibody poses a series of problems for which the structure of 2G12 represents an elegant immunological solution. First of all, an antibody response to the "self" carbohydrates on HIV-1 would appear to be excluded by tolerance mechanisms. However, the dense cluster of oligomannose residues found on the silent face of gp120 has not been described for any other mammalian glycoprotein (36). Thus, gp120 could be distinguished from self glycoproteins if the antibody response was dependent on clustered oligomannose moieties. Second, recognition of a dense cluster of carbohydrates is problematic for a conventional Y-or T-shaped IgG molecule. Geometrical constraints suggest that a single antibody-combining site can bind only to carbohydrate residues from one oligomannose chain. Recognition of two oligomannose chains can be achieved only by bivalent antibody binding. It is conceivable that an IgG molecule could bivalently recognize two oligomannose chains 35 Å apart at their tips, as suggested for gp120 below (Fig. 6), but this would require a near parallel orientation of the two Fab arms that would be energetically disfavored. In contrast, the 2G12 domain-exchanged structure is well adapted for recognition of two oligomannose chains at a spacing of about 35 Å. In this intertwined arrangement, no entropic penalty need be paid for bivalent attachment to the Fab arms, as in a conventional antibody, and indeed, previous studies have shown that 2G12 binds with low entropy to gp120 (37). In addition, the VH/VH' interface provides a potential increase in valency of the antibody-antigen interaction through antigen-binding regions that can facilitate binding to an array of carbohydrates. This oligomeric structure of 2G12 can account for the unusually high-affinity (nanomolar) recognition for a carbohydrate antigen by providing a virtually continuous surface for multivalent recognition with interaction sites that match the geometrical spacing of the carbohydrate cluster on gp120.

 Fig. 6. Model of 2G12 glycan recognition of gp120. On the basis of our model, three separate Man9GlcNAc2 moieties, shown in red (two in the primary combining sites and one in the VH/VH' interface), potentially mediate the binding of 2G12 to gp120. The glycans at the primary combining sites originate from Asn332 and Asn392 (labeled) in gp120, whereas the carbohydrate located at the VH/VH' interface would arise from Asn339 (labeled). N-linked glycans occurring at Asn332 and Asn392 have previously been implicated as critical for 2G12 binding (13). The N-linked glycan at Asn 339 is not as critical for 2G12 binding, although this glycan could potentially interact with the VH/VH' interface. Figure was generated using programs Molscript (65) and Raster3D (66). [View Larger Version of this Image (57K GIF file)]

Comparison with other lectins. The proposed mode of binding of 2G12 is reminiscent of one of the suggested mechanisms of multivalent recognition by animal lectins, such as serum mannose-binding protein, in which avidity can be optimized by matching the appropriate geometrical arrangement of the binding sites in the lectin oligomer with the spacing of carbohydrate epitopes on the pathogen. Furthermore, as for 2G12, the specificity of mannose-binding proteins is achieved through multivalent interactions, as opposed to recognition by a single high-affinity site (17, 38).

A C-type lectin, DC-SIGN (dendritic cell–specific intercellular adhesion molecule–3 grabbing nonintegrin) also binds carbohydrates on the envelope of HIV and facilitates viral infection of CD4+ T cells (39). DC-SIGN differs from 2G12 in that it binds to an internal core feature of high-mannose oligosaccharides, as opposed to the terminal mannoses (13, 14, 40, 41). It is interesting that HIV-1 may have evolved to display oligomannose clusters on its surface, in part to enhance binding to DC-SIGN by increased avidity through multiple interactions (42), and 2G12 then is able to exploit this Achilles' heel through its own brand of multivalent recognition.

The 2G12 antibody can also be compared with cyanovirin (13, 14, 41), a cyanobacterial lectin that neutralizes HIV-1 by binding carbohydrate on the surface of gp120 (4345). Crystal structures of cyanovirin have shown that it is also capable of binding Man{alpha}1-2Man at the end of the D1 arm of Man9GlcNAc2 (46). Coincidentally, cyanovirin also can exist as a domain-swapped dimer (47) with four possible binding sites (48). However, previous studies on cyanovirin have proposed that high-affinity binding to gp120 is achieved by interaction with only one oligomannose rather than a constellation of oligomannose moieties, as for 2G12 (49).

Recognition of HIV-1 gp120 by 2G12. From the crystal structures of 2G12 in complex with Man9GlcNAc2 and the gp120 core structure (50), we can now approximate how 2G12 might bind to gp120. Previous mutagenesis studies have implicated N-linked glycans at positions 295, 332, and 392 in gp120 as being most critical for 2G12 binding (13). Taking into account the conformational space available to the Man9GlcNAc2-Asn by virtue of both the N-glycosidic linkage and the sugar residues, we propose that the Fab 2G12 dimer most likely binds gp120 at the N-linked glycans at positions 332 and 392 in the primary combining sites, with a potential interaction with the N-linked glycan at position 339 at the VH/VH' interface (Fig. 6) (51). On the basis of our model, the glycan at position 295 plays an indirect role in 2G12 binding by preventing further processing of the glycan at 332 and by maintaining its oligomannose structure as one that is recognized by 2G12.

The structures of Fab 2G12 complexed with Man9GlcNAc2 and Man{alpha}1-2Man are also provocative templates for innovative HIV-1 vaccine design (52). Immunogens designed to mimic the unique cluster of oligomannose sugars binding to antibody 2G12 can now be tested for their ability to elicit a 2G12-like immune response. The 2G12 structure further provides a scaffold for engineering high-affinity antibodies to other molecular clusters in general, not only carbohydrates, as might be found on pathogens and tumor cells, but also other naturally occurring or synthetic clusters.

Reference and Notes

1. R. M. Zinkernagel et al., Adv. Immunol. 79, 1 (2001).[ISI][Medline]
2. D. R. Burton, Nat. Rev. Immunol. 2, 706 (2002).[CrossRef][ISI][Medline]
3. F. Ferrantelli, R. M. Ruprecht, Curr. Opin. Immunol. 14, 495 (2002).[CrossRef][ISI][Medline]
4. A. J. Conley et al., Proc. Natl. Acad. Sci. U.S.A. 91, 3348 (1994).[Abstract]
5. C. E. Parker et al., J. Virol. 75, 10906 (2001).[Abstract/Free Full Text]
6. M. B. Zwick et al., J. Virol. 75, 10892 (2001).[Abstract/Free Full Text]
7. G. Stiegler et al., AIDS Res. Hum. Retroviruses 17, 1757 (2001).[CrossRef][ISI][Medline]
8. D. R. Burton et al., Science 266, 1024 (1994).[ISI][Medline]
9. P. Roben et al., J. Virol. 68, 4821 (1994).[Abstract]
10. E. O. Saphire et al., Science 293, 1155 (2001).[Abstract/Free Full Text]
11. A. Trkola et al., J. Virol. 69, 6609 (1995).[Abstract]
12. A. Trkola et al., J. Virol. 70, 1100 (1996).[Abstract]
13. C. N. Scanlan et al., J. Virol. 76, 7306 (2002).[Abstract/Free Full Text]
14. R. W. Sanders et al., J. Virol. 76, 7293 (2002).[Abstract/Free Full Text]
15. J. P. Moore, J. Sodroski, J. Virol. 70, 1863 (1996).[Abstract]
16. P. M. Rudd, R. A. Dwek, Crit. Rev. Biochem. Mol. Biol. 32, 1 (1997).[ISI][Medline]
17. W. I. Weis, Curr. Opin. Struct. Biol. 7, 624 (1997).[CrossRef][ISI][Medline]
18. I. A. Wilson, R. L. Stanfield, Nature Struct. Biol. 2, 433 (1995).[ISI][Medline]
19. R. Thomas et al., J. Biol. Chem. 277, 2059 (2002).[Abstract/Free Full Text]
20. Materials and methods are available as supporting material on Science Online.
21. Y. Liu, D. Eisenberg, Protein Sci. 11, 1285 (2002).[Abstract/Free Full Text]
22. A. M. Lesk, C. Chothia, Nature 335, 188 (1988).[ISI][Medline]
23. The "closed interface" refers to the interface between the swapped domain and the main domain that exists in both the monomer and the domain-swapped oligomer. The "hinge loop" is the segment of polypeptide chain that links the swapped domain and the main domain and adopts different conformations in the monomer and the domain-swapped oligomer. The "open interface" exists only in the domain-swapped oligomer, but not in the monomer. The interplay between these three factors (destabilization of the closed interface, conformational shift in the hinge loop, and an energetically favorable open interface) can promote domain swapping (21).
24. K. C. Garcia et al., Science 274, 209 (1996).[Abstract/Free Full Text]
25. All measurements of residue occurrence are made by using the Kabat sequence database (53) at For a detailed sequence analysis of Fab 2G12, see fig. S3.
26. M. Bergdoll, M. H. Remy, C. Cagnon, J. M. Masson, P. Dumas, Structure 5, 391 (1997).[ISI][Medline]
27. Alanine is the most common residue at this position (58%).
28. A. C. Martin, Proteins 25, 130 (1996).[CrossRef][ISI][Medline]
29. We calculated Sc coefficients (54) and buried molecular surface using the programs SC (55) and MS (56), in which a 1.7 Å probe radius and standard van der Waals radii were used (57). The Sc coefficients here represent a tightly packed interface typical of those found in oligomeric protein structures [which have Sc coefficients that range from 0.70 to 0.76 (54)].
30. AsthisVH/VH' interface isfound in all three independent crystal structures of Fab 2G12, all measurements and analysis described here will use the highest resolution structure (1.75 Å) of Fab 2G12 complexed with Man{alpha}1-2Man.
31. S. Sheriff, W. A. Hendrickson, J. L. Smith, J. Mol. Biol. 197, 273 (1987).[ISI][Medline]
32. Residues that are marked with a prime symbol (') are from the second Fab molecule of the domain-exchanged Fab dimer.
33. Previous studies have shown that truncation of some proteins can lead to artificial domain swaps, which do not or cannot occur in the native protein, as, for example, domain 5 of TrkA, TrkB, and TrkC (58). Also, artificial domain swaps in engineered Fv fragments have been identified through variation of the length of the linker region between VH and VL, as, for example, in diabodies (59) and triabodies (60), in which the natural VH/VL pairing is perturbed because of the shortness of the linker connection.
34. N. Schülke et al., J. Virol. 76, 7760 (2002).[Abstract/Free Full Text]
35. M. R. Wormald et al., Chem. Rev. 102, 371 (2002).[CrossRef][ISI][Medline]
36. A. J. Petrescu, M. R. Wormald, unpublished data.
37. P. D. Kwong et al., Nature 420, 678 (2002).[CrossRef][ISI][Medline]
38. W. I. Weis, K. Drickamer, Annu. Rev. Biochem. 65, 441 (1996).[CrossRef][ISI][Medline]
39. T. B. Geijtenbeek et al., Cell 100, 587 (2000).[ISI][Medline]
40. H. Feinberg, D. A. Mitchell, K. Drickamer, W. I. Weis, Science 294, 2163 (2001).[Abstract/Free Full Text]
41. P. W. Hong et al., J. Virol. 76, 12855 (2002).[Abstract/Free Full Text]
42. D. A. Mitchell, A. J. Fadden, K. Drickamer, J. Biol. Chem. 276, 28939 (2001).[Abstract/Free Full Text]
43. M. R. Boyd et al., Antimicrob. Agents Chemother. 41, 1521 (1997).[Abstract]
44. M. T. Esser et al., J. Virol. 73, 4360 (1999).[Abstract/Free Full Text]
45. B. R. O'Keefe et al., Mol. Pharmacol. 58, 982 (2000).[Abstract/Free Full Text]
46. I. Botos et al., J. Biol. Chem. 277, 34336 (2002).[Abstract/Free Full Text]
47. F. Yang et al., J. Mol. Biol. 288, 403 (1999).[CrossRef][ISI][Medline]
48. L. G. Barrientos, A. M. Gronenborn, Biochem. Biophys. Res. Commun. 298, 598 (2002).[CrossRef][ISI][Medline]
49. L. C. Chang, C. A. Bewley, J. Mol. Biol. 318, 1 (2002).[CrossRef][ISI][Medline]
50. Gp120 coordinates represent the 2.2 Å crystal structure of core gp120 from the HxB2 strain of HIV-1 complexed to CD4 and Fab 17b (61).
51. A family of 18 oligomannose structures obtained from a molecular dynamics simulation (62) was superimposed on the primary binding sites in the Man9GlcNAc2 + 2G12 complex structure at the mannose residues D1 and C (fig. S6). The majority of the structures could be superimposed without altering the conformation of either the glycan or the protein. In order to model the complex between 2G12 and gp120, the range of Asn epitopes consistent with 2G12 binding, on the basis of the molecular dynamics, was matched against the Asn positions and orientations observed in the gp120 crystal structure (61). A nearly perfect match (within 0.5 Å) could be obtained by a pair of glycans and the residues Asn332 and Asn392, as the glycans attached to both residues were strongly implicated in 2G12 binding. There is no other pair of N-linked glycan sites on gp120 that fit the defined epitope. It is interesting that this model also places the N-linked glycan at position 339 close to the VH/VH' interface of the 2G12 Fab dimer. Although this glycan is not as critical for binding 2G12 as glycans at 295, 332, or 392, it could interact with the secondary, perhaps lower-affinity, binding site at the VH/VH' interface. The interaction between the glycan at Asn339 and the VH/VH' interface was modeled in a manner similar to that described above.
52. The design of multivalent carbohydrate-based immunogens as vaccines has been proposed for targeting cancer cells (63).
53. E. A. Kabat, T. T. Wu, H. M. Perry, K. S. Gottesman, C. Foeller, Sequences of Proteins of Immunological Interest (U.S. Department of Health and Human Services, Government Printing Office, Washington, DC, 5th ed., 1991).
54. M. C. Lawrence, P. M. Colman, J. Mol. Biol. 234, 946 (1993).[CrossRef][ISI][Medline]
55. CCP4, Acta Crystallogr. D50, 760 (1994).
56. M. L. Connelly, J. Mol. Graphics 11, 139 (1993).[CrossRef][ISI][Medline]
57. B. R. Gelin, M. Karplus, Biochemistry 18, 1256 (1979).[ISI][Medline]
58. M. H. Ultsch et al., J. Mol. Biol. 290, 149 (1999).[CrossRef][ISI][Medline]
59. O. Perisic, P. A. Webb, P. Holliger, G. Winter, R. L. Williams, Structure 2, 1217 (1994).[ISI][Medline]
60. X. Y. Pei, P. Holliger, A. G. Murzin, R. L. Williams, Proc. Natl. Acad. Sci. U.S.A. 94, 9637 (1997).[Abstract/Free Full Text]
61. P. D. Kwong et al., Structure Fold Des. 8, 1329 (2000).[Medline]
62. R. J. Woods et al., Eur. J. Biochem. 258, 372 (1998).[Abstract]
63. G. Ragupathi et al., Proc. Natl. Acad. Sci. U.S.A. 99, 13699 (2002).[Abstract/Free Full Text]
64. R. M. Esnouf, Acta Crystallogr. D55, 938 (1999).
65. P. J. Kraulis, J. Appl. Crystallogr. 24, 946 (1991).[CrossRef][ISI]
66. E. A. Merritt, D. J. Bacon, Methods Enzymol. 277, 505 (1997).[ISI]
67. C. N. Scanlon et al., unpublished data.
68. M. L. Phillips, M. H. Tao, S. L. Morrison, V. N. Schumaker, Mol. Immunol. 31, 1201 (1994).[ISI][Medline]
69. B. Carrasco et al., Biophys. Chem. 93, 181 (2001).[CrossRef][ISI][Medline]
70. Supported by NIH Grants GM46192 (I.A.W.), AI33292 (D.R.B.), and a grant from IAVI (International Aids Vaccine Initiative) through the Neutralizing Antibody Consortium. We are grateful to the staff of the Stanford Synchrotron Radiation Laboratory (SSRL) Beamline 11-1 for assistance at the beamline; E. O. Saphire (The Scripps Research Institute) for the initial model of the carbohydrate on gp120; A. Heine (The Scripps Research Institute), R. Aguilar (The Scripps Research Institute), S. Church (The Scripps Research Institute) and B. Matthews (Oxford Glycobiology Institute, UK) for expert technical assistance; R. Lerner (The Scripps Research Institute), J. Binley (The Scripps Research Institute), and M. Crispin (Oxford Glycobiology Institute, UK) for helpful discussion and comments. This is publication 15508-MB from The Scripps Research Institute. Coordinates have been deposited in the Protein Data Bank (codes 1OM3, 1OP3, and 1OP5) and will be released immediately upon publication.

Supporting Online Material

Materials and Methods

Tables S1 and S2

Figs. S1 to S6


7 February 2003; accepted 6 May 2003
Include this information when citing this paper.

Abstract of this Article
PDF Version of this Article
Supporting Online Material
Download to Citation Manager
Alert me when:
new articles cite this article
Search for similar articles in:
  Science Online
Search Medline for articles by:
Calarese, D. A. || Wilson, I. A.
This article appears in the following Subject Collections:

Volume 300, Number 5628, Issue of 27 Jun 2003, pp. 2065-2071.
Copyright © 2003 by The American Association for the Advancement of Science. All rights reserved. Next Wave