Antibody 2G12 Recognizes Di-Mannose Equivalently in Domain- and Nondomain-Exchanged Forms but Only Binds the HIV-1 Glycan Shield if Domain Exchanged

Abstract

The broadly neutralizing anti-human immunodeficiency virus type 1 (HIV-1) antibody 2G12 targets the high-mannose cluster on the glycan shield of HIV-1. 2G12 has a unique V_H domain-exchanged structure, with a multivalent binding surface that includes two primary glycan binding sites. The high-mannose cluster is an attractive target for HIV-1 vaccine design, but so far, no carbohydrate immunogen has elicited 2G12-like antibodies. Important questions remain as to how this domain exchange arose in 2G12 and how this unusual event conferred unexpected reactivity against the glycan shield of HIV-1. In order to address these questions, we generated a nondomain-exchanged variant of 2G12 to produce a conventional Y/T-shaped antibody through a single amino acid substitution (2G12 I19R) and showed that, as for the 2G12 wild type (2G12 WT), this antibody is able to recognize the same Manα1,2Man motif on recombinant gp120, Candida albicans, and synthetic glycoconjugates. However, the nondomain-exchanged variant of 2G12 is unable to bind the cluster of mannose moieties on the surface of HIV-1. Crystallographic analysis of 2G12 I19R in complex with Manα1,2Man revealed an adaptable hinge between V_H and C_H1 that enables the V_H and V_L domains to assemble in such a way that the configuration of the primary binding site and its interaction with disaccharide are remarkably similar in the nondomain-exchanged and domain-exchanged forms. Together with data that suggest that very few substitutions are required for domain exchange, the results suggest potential mechanisms for the evolution of domain-exchanged antibodies and immunization strategies for eliciting such antibodies.

The broadly neutralizing anti-human immunodeficiency virus type 1 (HIV-1) human monoclonal antibody 2G12 recognizes a highly conserved cluster of oligomannose residues on the glycan shield of the HIV-1 envelope glycoprotein gp120 (9, 10, 36, 39, 44, 45). The antibody binds terminal Manα1,2Man-linked sugars of high-mannose glycans (Man_8-9GlcNAc₂) with nanomolar affinity using a unique domain-exchanged structure in which the variable domains of the heavy chains swap to form a multivalent binding surface that includes two conventional antigen-combining sites and a third potential noncanonical binding site at the novel V_H/V_H′ interface (10). gp120 is one of the most heavily glycosylated proteins identified to date, with approximately 50% of its mass arising from host-derived N-linked glycans (24). These glycans play an important role in shielding the virus from the host immune system (34). Carbohydrates are generally poorly immunogenic, and the dense covering of glycans is often referred to as the “silent face” (52). The oligomannose glycans on gp120 in particular are closely packed, forming a tight cluster, and the unique domain-exchanged structure of 2G12 has been proposed as a means to recognize this cluster (10).

The attraction of 2G12 as a template for HIV-1 vaccine design has recently been highlighted in a study that showed the antibody can protect macaques against simian-human immunodeficiency virus (SHIV) challenge at remarkably low serum neutralizing titers (18, 30, 43). When using 2G12 as a template for design of a carbohydrate immunogen, some important considerations must be taken into account. First, 2G12 is unusual in its specificity (targeting host cell-derived glycan motifs presented in a “nonself” arrangement), and although the 2G12 epitope is common to many HIV-1 envelopes, 2G12-like antibodies are rarely elicited (5, 38). Second, due to inherently weak carbohydrate-protein interactions (49, 50), it can be assumed that in order for a carbohydrate-specific antibody to achieve the affinity required to neutralize HIV-1, the avidity of the interaction must be enhanced by both Fab arms of the IgG-contacting glycan motifs simultaneously on the HIV-1 envelope. Third, the unique domain-exchanged structure of 2G12 has not been described for any other antibody (10). These considerations raise a number of questions. Which antigen or sequence of antigens elicited 2G12? Is domain exchange the only solution for recognition of highly clustered oligomannoses? If so, can domain exchange be elicited by immunization with clustered oligomannose motifs (38)?

Efforts to design immunogens that elicit responses to the glycan shield of HIV-1 and neutralize the virus have to date been unsuccessful (2, 3, 14, 20, 21, 28, 29, 32, 46-48). Immunogen design strategies that mimic the 2G12 epitope have focused on both chemical and biochemical methods to generate multivalent and clustered displays of both high-mannose sugars (Man_8-9GlcNAc₂) (13, 15, 20, 21, 27-29, 32, 47) and truncated versions of such sugars (Man₉ and Man₄ linked via a 5-carbon linker) (3, 46). These constructs typically bind 2G12 with a lower affinity (on the order of 1 to 3 logs) than recombinant gp120. Although mannose-specific antibodies have been elicited by these immunogens, no HIV-1-neutralizing activities have been described. In a study by Luallen et al., antibodies against recombinant gp120 were generated by immunization with yeast cells that had been mutated to display only Man₈GlcNAc₂ glycans (27, 29). However, no neutralization activity against the corresponding pseudovirus was noted. It was proposed that this was due to either the low abundance of the gp120-specific antibodies in the serum or the antibodies elicited being against carbohydrate epitopes that differed from the 2G12 epitope (27, 29).

To gain a better understanding of the importance of domain exchange for glycan recognition and how 2G12 may have been induced, we analyzed the binding characteristics of a nondomain-exchanged (conventional Y/T-shaped) 2G12 variant antibody. This variant was generated by a single point mutation, I19R, that disrupts the V_H/V_H′ interface. We show that the mutant is still able to recognize the Manα1,2Man motif arrayed on yeast, synthetic glycoconjugates, and recombinant gp120 in enzyme-linked immunosorbent assay (ELISA) format but is unable to recognize the discrete, dense mannose clusters found on the surface of the HIV-1 envelope (as measured by neutralization activity and binding to HIV-1-transfected cells). We further show that a major conformational change in the elbow region between V_H and C_H1 in this nondomain-exchanged variant of 2G12 allows the variable domains to assemble in similar orientations with respect to each other, as in the 2G12 wild type (WT), with an identical primary binding site, although with dramatically different orientations with respect to the constant domains. Thus, we conclude that 2G12 recognizes Manα1,2Man motifs in an identical manner in both conventional and domain-exchanged configurations, and the 2G12 specificity for Manα1,2Man likely first arose in a conventional IgG predecessor of 2G12. Subsequent domain exchange was the key event that then enabled high-affinity recognition of the tight oligomannose clusters on HIV-1.

MATERIALS AND METHODS

Antibody expression and purification.

The 2G12 WT variable genes were cloned into full-length IgG1 expression vectors pγ1HC and pκLC (42). The I19R substitution was introduced by QuikChange site-directed mutagenesis (Stratagene, La Jolla, CA). The constructs and mutation were verified by sequence analysis (Eton Bioscience, San Diego, CA). The 2G12 I19R mutant antibody was transiently expressed in the FreeStyle 293 expression system (Invitrogen, Carlsbad, CA), purified using affinity chromatography (protein A-Sepharose Fast Flow; GE Healthcare, United Kingdom), and its purity and integrity were confirmed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE).

Assessment of domain exchange.

2G12 I19R IgG at a concentration of 4 mg/ml was digested with Lys-C (1:1,600 [wt/wt], IgG/Lys-C; Roche) for 4 h at 37°C, and Fabs were purified using protein A affinity chromatography.

Enzyme-linked immunosorbent assays.

For ELISA, Costar high-binding microtiter 96-well plates (type 3690; Corning Life Sciences, Lowell, MA) were used. All steps were performed at room temperature, and all volumes used were 50 μl unless otherwise stated.

gp120 ELISA.

Wells were coated with 50 ng of gp120_JR-FL (Progenics, Tarrytown, NY), kif-gp120_BAL (40), gp120_BAL, or gp120_JR-CSF overnight at 4°C. Plates were washed four times with phosphate-buffered saline (PBS)-0.05% (vol/vol) Tween 20 (Sigma), blocked for 1 h with 100 μl 5% nonfat milk-PBS-0.05% Tween 20, and washed after each step five times with PBS-0.05% Tween. Serial dilutions of mutant 2G12 antibodies in 5% nonfat milk-PBS-0.05% Tween 20 were incubated for 2 h. Unbound antibody was removed by washing four times as described above. Binding was detected with goat anti-human Fcγ alkaline phosphatase (AP) conjugate (1:1,000; Jackson ImmunoResearch, West Grove, PA) and p-nitrophenol phosphate substrate (Sigma) at 405 nm.

Precomplexing of primary antibody.

The antibody was first incubated with goat-anti-human Fcγ AP (2:1, primary/secondary ratio) for 15 min at 4°C prior to addition to the ELISA plates.

Glycoconjugate ELISA.

Wells were coated with 250 ng Qβ-Man₄, Qβ-Man₈, Qβ-Man₉, Qβ-Man₈GlcNAc₂, or keyhole limpet hemocyanin (KLH)-Man₉D at 4°C overnight. The remaining steps were carried out as described above.

Virus neutralization assays.

HIV-1 enveloped pseudovirus capable of single-round infection was generated by cotransfection of HEK 293T cells with HIV-1 Env-expressing plasmid (pSVIII) and pSG3ΔEnv as previously described (25). Virus was harvested after 72 h. Serial dilutions of 2G12 I19R IgG were incubated with pseudovirus for 1 h at 37°C before being added to TZM-bl cells (2 × 10⁴ cells). The antibody dilution causing 50% reduction in luciferase reporter gene expression after 3 days was determined by regression analysis. Kifunensine was used at a concentration of 25 μM and was added to the HEK 293T cells at the time of transfection.

Cell surface binding to 293T cells expressing the JR-CSF HIV-1 trimer.

293T cells displaying the HIV-1 trimer were prepared by transient transfection with gp160 JR-CSF (the pSVIII plasmid) and a plasmid carrying the tat gene using FuGene by following the manufacturer's instructions (Roche) at a ratio of 1:3 (envelope/ptat). After 3 days, the cells were washed, and various concentrations of 2G12 variants were added to HIV-1 gp160-transfected 293T cells and incubated for 1 h at room temperature. The cells were washed and blocked with fetal bovine serum (FBS) (30 min, 4°C) before being stained with goat anti-human Fcγ phycoerythrin substrate (Jackson) at a dilution of 1:200 for 30 min at 4°C. Binding was analyzed by flow cytometry, and binding curves were generated by plotting the mean fluorescence intensity of antigen as a function of antibody concentration. A FACSArray plate reader (BD Biosciences) was used for flow cytometry, and FlowJo software (version 6.4.7; Tree Star) was used for data analysis.

Binding to Candida albicans and Saccharomyces cerevisiae.

Heat-killed yeast was incubated with serial dilutions of 2G12 mutants at room temperature for 1 h. Cells were washed, stained with goat anti-human Fcγ phycoerythrin substrate (Jackson) at a dilution of 1:200 for 30 min at 4°C, and washed again. Binding was analyzed by flow cytometry, and binding curves were generated by plotting the mean fluorescence intensity of antigen as a function of antibody concentration. A FACSArray plate reader (BD Biosciences) was used for flow cytometry, and FlowJo software (version 6.4.7; Tree Star) was used for data analysis.

SPR spectroscopy.

Surface plasmon resonance (SPR) was carried out using a Biacore 2000 (GE Healthcare). A CM5 chip was activated using an EDC-NHS ester mixture, and Kif-gp120_BaL (50 μg/ml in PBS) was immobilized, aiming for 5,000 response units (RU) (actual, 4,952 RU). Measurements were carried out in HBS-EP buffer (GE Healthcare) at 25°C. The chip was blocked with ethanolamine-HCl (1 M, pH 8.5). Titrations of antibodies of the 2G12 WT, 2G12 I19R, and precomplexed 2G12 I19R were flown over the surface at a flow rate of 30 μl/min for 180 s. Antibody injections were followed by a dissociation phase of 600 s. The surface was regenerated between each measurement using glycine (0.2 M, pH 2.2, 1 min).

Crystallographic analysis.

2G12 Fab fragments were prepared as previously described (10) and concentrated to 20 mg/ml. Manα1,2Man in solid form was added to the Fab solution to saturation. For crystallization, equal volumes (0.3 μl) of 2G12 I19R Fab/Manα1,2Man and reservoir solution containing 2 M ammonium sulfate and 0.1 M sodium citrate, pH 5.5, were mixed in sitting-drop, vapor-diffusion experiments. The cocrystals were cryoprotected with 20% glycerol. Data were collected at 100 K at Stanford Synchrotron Radiation Lightsource (SSRL) beam line 9.2 (λ of ∼1.0 Å) and were indexed, integrated, and scaled using HKL2000 (33). Data collection and refinement statistics are summarized in Table 1.

TABLE 1.

Data collection and refinement statistics

Parameter	Value(s)a
Data collection statistics
Space group	P1
Cell dimensions
a, b, c (Å)	39.3, 56.6, 57.7
α, β, γ (°)	92.9, 105.0, 92.7
Resolution range (Å)	50-1.9 (1.97-1.90)
Rsym (%)	7.4 (13.5)
Avg I/σ(I)	9.4 (6.6)
No. of observations	86,795
No. of unique reflections	32,889 (3,033)
Completeness (%)	89.0 (82.4)
Redundancy	2.6 (2.4)
Refinement statistics
Resolution range (Å)	19.8-1.9 (1.96-1.90)
No. of reflections (test)b	32,317 (1,633)
Rcryst/Rfree (%)	21.2/24.3
No. of atoms
Protein	3,280
Ligand	23
Water	195
Avg B values (Å2)
Protein	45.9
Ligand	52.3
Water	51.9
RMS deviations
Bond length (Å)	0.009
Bond angle (°)	1.12
Ramachandran statistics (%)c
Most favored regions	92.5
Additional allowed regions	7.0
Disallowed regionsd	0.5

^aThe values in parentheses represent the highest-resolution shell.

^bTest set of reflections for calculation of R_free.

^cCalculated using PROCHECK (23).

^dIncludes residue L51, which commonly exists in a γ turn but is incorrectly flagged by PROCHECK as an outlier. L30 is also designated as disallowed by PROCHECK but displays a good fit to the corresponding electron density.

The 2G12 I19R Fab/Manα1,2Man structure was solved by molecular replacement, using the 1.75-Å structure of 2G12 Fab/Manα1,2Man (Protein Data Bank accession number 1OP3) (10) as the starting model for Phaser (1). The asymmetric unit consisted of one 2G12 I19R Fab/Manα1,2Man complex. Model building was performed using COOT (17) and refined with Refmac5 (31) and Buster (7). Fo-Fc simulated annealing omit maps were calculated using CNS (version 1.2) (8).

Protein structure accession number.

The coordinates and structure factors for 2G12 I19R Fab/Manα1,2Man have been deposited in the Protein Data Bank under accession number 3OAU.

RESULTS

Expression of 2G12 variant IgG 2G12 I19R.

First, a nondomain-exchanged version of 2G12 IgG was produced. We previously showed that replacing isoleucine at position 19 of the heavy chain with arginine (as found in the 2G12 germ line) (19) resulted in a completely nondomain-exchanged antibody (as determined by size exclusion chromatography of 2G12 I19R Fab) (19). I19, located at the interface of the two heavy chains of the domain-exchanged dimer, resides in a small hydrophobic pocket, and substitution with a basic arginine residue might be expected to disrupt V_H/V_H′ interactions.

The 2G12 wild-type heavy and light chains were cloned into the IgG1 expression vectors pγ1HC and pκLC, respectively (42), and a point mutation was introduced by site-directed mutagenesis. The variant IgG was transiently expressed in 293F cells and purified by protein A affinity chromatography, and its size was confirmed by SDS-PAGE. The IgG was digested using Lys-C to generate Fab fragments, and size exclusion chromatography indicated that 2G12 I19R Fab was a monomer (data not shown) (19).

Recognition of mannose motifs specific to HIV-1.

The specificity and affinity of 2G12 I19R was first assessed by measuring its binding to the envelope glycoprotein gp120 by ELISA. Wells were coated with recombinant gp120_JR-CSF, and serial dilutions of the antibody were added. 2G12 I19R showed no detectable binding (Fig. 1A). However, when the avidity of the antibody was increased by first precomplexing with an anti-Fc antibody to form a bivalent IgG (i.e., with four binding sites), binding to gp120_JR-CSF was observed. This result indicated that the primary binding site of variant I19R was still able to recognize Manα1,2Man-linked glycans. Under these conditions, the variant bound several recombinant gp120 strains, including kif-gp120_BaL displaying only Man₉GlcNAc₂ glycans (arising from expression in the presence of the endoplasmic reticulum-mannosidase I inhibitor kifunensine) (16, 40) (Fig. 1B and C). Typically, the precomplexed 2G12 I19R variant bound gp120 with a 100-fold-higher concentration for 50% binding than the 2G12 WT (see Table S1 in supplemental material elsewhere [http://www.scripps.edu/ims/burton/supplemental/Doores_JVI_2010B.pdf]). The monovalent 2G12 I19R Fab fragment did not show significant binding to recombinant gp120 (Fig. 1D), even when precomplexed with anti-human F(ab′)₂ to form a bivalent Fab complex, likely due to lowered avidity.

FIG. 1.

Binding of 2G12 I19R to recombinant gp120 by ELISA (A through D) and to the HIV-1 envelope trimer (E through F), as assessed by neutralization and by flow cytometry. (A) gp120_JR-CSF; (B) gp120_BaL; (C) Kif-gp120_BaL; (D) Fab binding to gp120_JR-CSF; (E) neutralization of HIV-1 JR-FL; (F) neutralization of HIV-1 JR-CSF; (G) binding to 293T cells displaying the JR-CSF trimer. PC = antibody was first precomplexed with a goat anti-human Fc-specific antibody before addition to the ELISA plate; Ab = antibody; MFI = mean fluorescent intensity.

We next compared the abilities of 2G12 and variant I19R to neutralize HIV-1. The neutralization activities of 2G12 I19R, precomplexed 2G12 I19R, and 2G12 WT IgGs were tested with both JR-FL and JR-CSF pseudoviruses (Fig. 1E and F). 2G12 I19R was unable to neutralize either virus when uncomplexed or complexed with anti-Fc antibody at concentrations below 40 μg/ml, while 2G12 WT neutralized both pseudoviruses, with 50% inhibitory concentrations (IC₅₀s) in the 1 to 10 μg/ml range. 2G12 I19R when uncomplexed or complexed was also unable to neutralize virus prepared in the presence of kifunensine, which displays only Man₉GlcNAc₂ glycans (see Fig. S1 in supplemental material elsewhere [http://www.scripps.edu/ims/burton/supplemental/Doores_JVI_2010B.pdf]).

Due to the discrepancy between gp120 binding and neutralization activity of 2G12 I19R, we wanted to assess whether this mutant could still bind gp120 in the context of the trimer, despite its inability to neutralize the virus. Therefore, we measured the binding of 2G12 I19R both complexed and uncomplexed to HEK-293T cells expressing the JR-CSF HIV-1 envelope trimer by flow cytometry. As no detectable binding was observed at concentrations below 100 μg/ml (Fig. 1G), the data suggested that 2G12 I19R is only able to recognize the mannose epitope in the context of recombinant gp120 when arrayed on an ELISA plate and not in the context of the HIV-1 trimer.

In order to measure the affinity to gp120 of the nondomain-exchanged 2G12 I19R variant compared to that of 2G12 WT, SPR experiments were carried out. 2G12 WT and both the precomplexed and nonprecomplexed I19R variant were flowed across kif-gp120_BaL immobilized on a CM5 chip. Binding of 2G12 I19R was observed only when first precomplexed (see Fig. S2 in supplemental material elsewhere [http://www.scripps.edu/ims/burton/supplemental/Doores_JVI_2010B.pdf]), as observed in the ELISA format. We were unable to reliably model these data with first-order or bivalent binding kinetics, as these models do not accurately describe these systems, and also the K_d (dissociation constant) could not be estimated from the equilibrium plateau, as saturation was not achieved at the concentrations studied.

Recognition of mannose motifs on different chemical array formats.

The ability of 2G12 I19R to bind Manα1,2Man displayed in various forms on various carriers was measured by ELISA. We first examined truncated forms of Man₉GlcNAc₂ arrayed on Qβ virus-like particles, in particular, Man₄, Man₈, Man₉, and Man₈GlcNAc₂ (2). Each particle displays approximately 360 glycans on its surface attached via a 5-carbon linker to lysine residues (2). 2G12 has previously been shown to bind these particles with nanomolar affinity (2). The 2G12 I19R mutant only bound these glycoconjugates when first precomplexed with the anti-Fc antibody, as with gp120 (Fig. 2A to D). More highly clustered arrangements of high-mannose glycans have also been chemically synthesized in the form of glycodendrons, in which nine oligomannose glycans are linked to an AB₃ dendrimeric skeleton and form homogeneous discrete clusters (48). These glycodendron clusters show somewhat higher affinity for the 2G12 WT than the Qβ glycoconjugates (48). 2G12 I19R bound only when precomplexed with anti-Fc antibody (Fig. 2E).

FIG. 2.

Binding of 2G12 I19R to synthetic glycoconjugates by ELISA (A through E) and to the yeast strains Candida albicans and Saccharomyces cerevisiae (F) by flow cytometry. (A) Qβ-Man₄; (B) Qβ-Man₈; (C) Qβ-Man₉; (D) Qβ-Man₈GlcNAc₂; (E) bovine serum albumin (BSA)-Man₉D; (F) Candida albicans and Saccharomyces cerevisiae. PC = antibody was first precomplexed with a goat anti-human Fc-specific antibody before addition to the ELISA plate.

Recognition of mannose motifs found on other pathogens.

We were interested to see if the nondomain-exchanged 2G12 mutant was able to recognize Manα1,2Man motifs that are found on other pathogens. Dunlop et al. showed that 2G12 recognizes Candida albicans, a yeast strain with a polysaccharide coat consisting of a repeating Manα1,6Man backbone and Manα1,2Man-linked branches (15). We measured the affinity of 2G12 I19R and 2G12 WT for heat-killed C. albicans using flow cytometry (Fig. 2F). 2G12 I19R bound with an affinity that was around 10-fold lower than that of the 2G12 WT (1,200 nM and 130 nM, respectively). Interestingly, 2G12 I19R bound without the need for initial precomplexing. This is likely due to the extended multiple displays of Manα1,2Man residues on C. albicans compared to those of recombinant monomeric gp120 that would allow simultaneous binding of both Fab arms of 2G12 I19R IgG. The mutant, along with 2G12 WT, did not bind Saccharomyces cerevisiae as the negative control, in which the termini of the Manα1,2Man branches are capped with Manα1,3Man-linked residues, confirming the specificity of the primary binding site for Manα1,2Man.

Crystallographic analysis of the 2G12 I19R mutant.

To investigate the mechanism of Manα1,2Man binding by 2G12 I19R, a crystal structure of 2G12 I19R Fab in complex with the disaccharide was solved to 1.9-Å resolution (Table 1). The asymmetric unit contains one Fab molecule in monomeric form (Fig. 3A). The individual variable (V_H and V_L) and constant (C_H1 and C_L) domains of 2G12 I19R are structurally similar to those found in 2G12 WT (the root mean square [RMS] deviation for least squares quadratic [LSQ]-matched Cα atoms in each domain is <0.6 Å) (10). However, the juxtapositioning of the variable versus constant regions is dramatically different (Fig. 3C and D). This difference is accommodated by an alternate conformation of the elbow region that is stabilized by “ball-and-socket” contacts between V_H and C_H1 that are found in other Fab structures but are absent in the 2G12 WT (10) (Fig. 4A and B). The V_H versus V_L domains and C_H1 versus C_L domains also show slight rigid body shifts with respect to each other compared to those in the 2G12 WT (the RMS deviation for Cα atoms in V_L when V_H in 2G12 I19R and the 2G12 WT are superimposed is 1.4 Å, and the RMS deviation for Cα atoms in C_L when the C_H1 domains in each antibody are superimposed is 1.2 Å) (10). Nevertheless, the density for Manα1,2Man at the antigen-combining site is extremely well defined (Fig. 4C), and the Fab contacts with the disaccharide in the antigen binding pocket are identical to those found in the 2G12 WT (10) (Fig. 4D).

FIG. 3.

Crystal structure of the nondomain-exchanged 2G12 I19R Fab monomer compared to that of the domain-exchanged 2G12 WT Fab dimer. (A) The crystal structure of the 2G12 I19R Fab monomer (light chain shown in green; heavy chain shown in purple). (B) The crystal structure of a monomer of the domain-exchanged 2G12 WT extracted from the crystal structure of the 2G12 WT dimer (10) (PDB accession number 1OP3) to show the similar conformations of V_L-V_H and V_L-V_H′ in nondomain-exchanged and domain-exchanged configurations, respectively. The WT monomer does not exist by itself in the 2G12 WT crystal but exists only in the context of the domain-swapped dimer. (C) 2G12 I19R Fab monomer (light chain shown in green; heavy chain shown in purple) and the 2G12 WT Fab dimer (now shown in gray), with the V_H of the former superimposed on the V_H′ of the latter. V_L, V_H, C_L, and C_H1 in 2G12 I19R Fab are labeled. Pro^H113 in 2G12 I19R Fab is represented by an orange sphere, and Pro^H113 and Pro^H113′ in 2G12 WT Fab are represented by yellow spheres and labeled. (D) Same as shown in panel C but rotated by 90^o. The domains in 2G12 WT Fab (V_H′, C_H1, V_L, and C_L) that correspond to those in 2G12 I19R Fab (V_H, C_H1, V_L, and C_L) only are shown for clarity. All figures were made using PyMOL (12).

FIG. 4.

Hinge loop and “ball-and-socket” contacts in 2G12 I19R Fab and the antibody-combining site of 2G12 I19R Fab in complex with Manα1,2Man. (A) Fo-Fc simulated annealing omit map of the elbow linker region (or hinge loop) between V_H and C_H1 in 2G12 I19R Fab (shown in stick format), contoured at 3σ. (B) The residues involved in the “ball-and-socket” interaction in 2G12 I19R Fab are shown in stick format, with the location of Pro^H113 shown in orange. (C) Fo-Fc simulated annealing omit map of Manα1,2Man bound to 2G12 I19R Fab, contoured at 3σ. (D) 2G12 I19R Fab in complex with Manα1,2Man (Fab shown in cyan; disaccharide shown with yellow carbons and red oxygens) and 2G12 WT Fab in complex with Manα1,2Man (10) (PDB accession number 1OP3) (Fab shown in gray; disaccharide not shown) superimposed on the Fab contact residues in CDRH3. All contact residues are represented in stick format, with direct hydrogen bonds to the disaccharide shown as dashed lines. CDRH and CDRL, complementary-determining region of heavy and light chains, respectively.

Consequently, the structure of 2G12 I19R Fab in complex with Manα1,2Man reveals that, when the high-affinity V_H/V_H′ interface in the 2G12 WT is disrupted by the single germ line substitution I19R, the elbow region between V_H and C_H1 can adopt a completely different conformation in the nondomain-exchanged antibody to allow the V_H and V_L domains to assemble in such a way that only a slight rigid body shift in the relative position of these variable domains is evident, compared to their assembly in the 2G12 WT. The antibody-combining site otherwise remains intact, with all interactions to the disaccharide maintained as in the 2G12 WT. This highlights not only the importance of the V_H/V_H′ interface in stabilizing the 2G12 WT structure but also the significance of the flexibility in the unusual hinge loop that enables 2G12 to adopt either a monomeric or dimeric conformation, depending on the affinity of the V_H/V_H′ interface, while maintaining the same fine specificity at the primary binding site for Manα1,2Man. This conformational malleability may have been the key feature in allowing/promoting domain exchange in 2G12 and its subsequent anti-HIV-1 activity and also suggests that affinity maturation to evolve the unusual domain-swapped 2G12 structure may have begun with a fully competent binding site for Manα1,2Man.

DISCUSSION

The data presented here strongly suggest a necessity for domain exchange for recognition of the tightly clustered assembly of oligomannose glycans on HIV-1. Introduction of a single amino acid substitution resulted in a conventional Y/T-shaped antibody (2G12 I19R). This variant of 2G12 can still bind Manα1,2Man motifs in different presentation formats, as it maintains an essentially unchanged primary binding site to the domain-swapped 2G12. The Manα1,2Man specificity of 2G12 I19R was illustrated through binding to recombinant gp120 (Fig. 1), synthetic Qβ glycoconjugates displaying Man₄, Man₈, Man₉, and Man₈GlcNAc₂, glycodendrons (Fig. 2), and the yeast pathogen C. albicans (Fig. 2), as well as through cocrystallization studies (Fig. 4). Although the nondomain-exchanged 2G12 I19R variant binds gp120 when arrayed on an ELISA plate (Fig. 1), it is unable to recognize Manα1,2Man motifs in the context of the HIV-1 envelope trimer and is, therefore, unable to bind cells expressing the HIV-1 envelope (Fig. 1G) or to neutralize 2G12-sensitive HIV-1 pseudoviruses (Fig. 1E and F).

Crystallographic analysis of 2G12 I19R Fab/Manα1,2Man revealed that when the V_H/V_H′ interface is disrupted by the single I19R germ line substitution, 2G12 is unable to adopt a domain-exchanged conformation but adopts the conventional configuration of the light and heavy chains via a major conformational change in the hinge loop between V_H and C_H1 compared to that of 2G12 WT (Fig. 3). The conventional and domain-exchanged configurations of 2G12 display dramatically different juxtapositioning of their variable versus constant regions and slight rigid body shifts in their V_H versus V_L and C_H1 versus C_L domains but yet still maintain the same configuration of their antigen-combining sites (Fig. 3 and 4). Moreover, the Manα1,2Man contact residues in these conventional and domain-exchanged configurations of 2G12 are identical (Fig. 4D). Although these two antibodies have almost identical Fab binding units, the differences in their IgG structures lead to remarkably different binding characteristics.

As discussed above, due to inherently weak carbohydrate-protein interactions (49, 50), it can be assumed that, in order for a carbohydrate-specific antibody to achieve the affinity required to neutralize HIV-1, the avidity of the interaction must be enhanced by both of the Fab arms of the IgG-contacting glycan motifs simultaneously on the HIV-1 envelope. From the models of the envelope trimer, it is thought that the diameter of the mannose patch is approximately 70 to 85 Å (26, 41, 53). The average distance between Fab arms in a conventional antibody is 120 to 170 Å (6, 35, 37), and it is, therefore, unlikely that both Fab arms of a conventional IgG could contact the same mannose cluster on a single monomer subunit; in fact, an almost parallel orientation of the Fab arms would be required, which may be energetically disfavored because of steric clashes (10). Therefore, in order for both Fab arms of a conventional IgG to contact mannose glycans simultaneously, they must likely either span adjacent envelope spikes or contact two monomer units within one trimer. The first situation is highly unlikely due to the sparse covering of envelope spikes on the surface of HIV-1 (54), but the second may be possible, as it is thought that the average distance between oligomannose clusters on neighboring monomer units within a trimer is 100 to 150 Å (26, 41, 53). However, we have shown that 2G12 I19R does not neutralize HIV-1 pseudoviruses or bind HIV-1-transfected cells, suggesting that, in the context of the HIV-1 envelope trimer, bivalent binding to a single spike does not occur. Although multiple 2G12 epitopes are displayed on pseudovirus made in the presence of the glycosidase inhibitor kifunensine, no neutralization by 2G12 I19R was observed, even when precomplexed, presumably due to an incompatible orientation of Man₉GlcNAc₂ glycans relative to the two 2G12 I19R Fab arms. As both Fab arms of the I19R variant antibody appear to be unable to bind oligomannose residues on a single monomer or on adjacent monomers in a single trimer (even when precomplexed with the anti-Fc antibody), the variant is unable to neutralize HIV-1. The variant is, however, able to bind gp120 when arrayed on an ELISA plate or CM5 chip when it is precomplexed with an anti-Fc antibody. This mode of binding likely arises from the simultaneous binding of mannose epitopes on neighboring immobilized gp120 molecules in an arrangement not found on HIV-1.

For the domain-exchanged 2G12 WT, the affinity for gp120 is greatly enhanced compared to that of the Fab unit, which has no detectable binding. The antigen-combining sites are approximately 35 Å apart (10), which is close enough to allow bivalent binding of the Fab arms to mannose residues in the same oligomannose cluster on one gp120 molecule, as evidenced by binding to monomeric gp120. The alignment and proximity of the binding sites is fixed in the rigid, domain-exchanged 2G12 WT structure, and therefore, bivalent binding to gp120 occurs with a low to negligible entropy change compared to that of conventional Y/T-shaped antibodies (22). In addition, it is possible that the V_H/V_H′ interface of the 2G12 WT may form favorable interactions with glycans on gp120 (10), further contributing to the increased apparent affinity of 2G12 WT IgG over the Fab unit by itself. Overall, these factors result in the nanomolar binding affinity of 2G12 to gp120. These data suggest that domain exchange is required to recognize the tightly clustered assembly of oligomannose motifs on gp120 and to neutralize HIV-1 via binding to its glycan shield (10, 38). The importance of enhanced avidity was recently highlighted by West et al., who demonstrated that a 2G12 dimeric IgG has enhanced neutralization activity (51).

Although 2G12 I19R does not appear to be able to bind bivalently to gp120, it does, nevertheless, appear to bind bivalently to the appropriately spaced Manα1,2Man epitopes on C. albicans. The polysaccharide coating of C. albicans consists of multiple repeating Manα1,2Man units attached to a repeating Manα1,6Man backbone that form a continuous array and, hence, is covered by a very large number of epitopes. This continuous array is thought to differ significantly from the discrete mannose clusters on HIV-1 and synthetic Man₉ glycodendrons. 2G12 I19R binds C. albicans with a lower affinity than the 2G12 WT (1,200 nM and 130 nM, respectively) but, importantly, without the need for initial precomplexing (as is required for gp120). The large array of repeating mannose epitopes allows binding of both of the Fab arms of 2G12 I19R simultaneously, therefore increasing the avidity of the interaction. The differences in apparent binding affinities of 2G12 WT and 2G12 I19R to C. albicans may be a reflection of the differing entropies of binding of the antibodies. As discussed above, the fixed and rigid arrangement of the Fab binding sites in domain-exchanged 2G12 WT imparts a lower entropy cost for bivalent binding than the flexible Y/T-shaped IgG structure (22). The micromolar affinity of 2G12 I19R for C. albicans suggests that the low affinity of the Fab unit can be enhanced by bivalent binding and, therefore, supports the hypothesis that lack of neutralization activity arises from an incompatible arrangement of the high-mannose glycans on HIV-1.

These data suggest the hypothesis that a conventional antibody with specificity for Manα1,2Man may have been initially elicited against a pathogen displaying a large number of repeating glycans, such as C. albicans (15, 38). In the donor, subsequently infected or coinfected with HIV-1, the tightly clustered mannose glycans on the virus may have then favored the selection of somatic variants of antibody able to provide at least a fraction of the total population of antibodies in domain-exchanged form. Further antigenic stimulation by the oligomannose cluster on the virus would then drive selection of antibody variants with increasing proportion of domain exchange until 2G12, essentially completely domain exchanged, evolved. Although the affinity of 2G12 I19R to gp120 is not measurable (<6 μM) in our assays, this antibody may still be selected in vivo, as the affinity required for B-cell selection can be very low (4, 11). This mechanism is consistent with our observations that domain exchange appears to be fixed following antibody folding inside the cell; no equilibrium is established between the domain-exchanged and nondomain-exchanged forms of 2G12 (19). Further, we have also shown that very few residues are required to initiate domain exchange in a germ line version of 2G12 (19). These results provide a potential mechanism for the generation of antibodies that recognize clustered antigens.

The data presented here confirm that domain exchange is key to recognition of the tight oligomannose clusters on HIV-1 (10) and is, therefore, of significant relevance to carbohydrate HIV-1 vaccine design. The data may also give insights into why the antibodies generated from immunization with both synthetic and biosynthetic glycoconjugates that mimic the HIV-1 glycan shield are able to bind recombinant gp120 but are unable to neutralize HIV-1 pseudovirus (14, 27, 29). For example, studies by Luallen et al. (29) and Dunlop et al. (14) in which rabbits were immunized with mutant yeast cells displaying Man₈GlcNAc₂ and Manα1,2Man motifs, respectively, generated anti-mannose antibodies that bound recombinant gp120 but did not neutralize HIV-1. These results could be explained by the generation of conventional (Y/T-shaped) antibodies in which both Fab arms were unable to bind mannose motifs on different subunits of the trimer simultaneously (29) but could bind neighboring immobilized gp120 molecules in an arrangement not found on virions. This further suggests that gp120 binding alone is not a sufficient screening tool to probe the immune sera derived from studies of the type described above.

We propose, therefore, that the major shortcoming of the carbohydrate HIV-1 immunogens tested thus far has been their inability to elicit domain-exchanged antibodies. Even if a higher-affinity conventional mannose-specific antibody was elicited, it seems unlikely that this antibody would have the correct geometry for both of the Fab arms to bind simultaneously to one gp120 monomer on an envelope spike. Further understanding of how this unique domain-exchanged antibody was elicited by the immune system may aid in the design of immunogens and immunization protocols that may lead to the selection of antibodies that favor domain exchange. One could envisage an immunization protocol where anti-mannose antibodies are first elicited by immunization with an extended array of mannose motifs, followed by immunization with discrete mannose clusters arrayed on a synthetic platform, as on gp120, to drive domain exchange of the mannose-specific antibodies.