Dissection of the carbohydrate specificity of the broadly neutralizing anti-HIV-1 antibody 2G12

Abstract

Human antibody 2G12 neutralizes a broad range of HIV-1 isolates. Hence, molecular characterization of its epitope, which corresponds to a conserved cluster of oligomannoses on the viral envelope glycoprotein gp120, is a high priority in HIV vaccine design. A prior crystal structure of 2G12 in complex with Man₉GlcNAc₂ highlighted the central importance of the D1 arm in antibody binding. To characterize the specificity of 2G12 more precisely, we performed solution-phase ELISA, carbohydrate microarray analysis, and cocrystallized Fab 2G12 with four different oligomannose derivatives (Man₄, Man₅, Man₇, and Man₈) that compete with gp120 for binding to 2G12. Our combined studies reveal that 2G12 is capable of binding both the D1 and D3 arms of the Man₉GlcNAc₂ moiety, which would provide more flexibility to make the required multivalent interactions between the antibody and the gp120 oligomannose cluster than thought previously. These results have important consequences for the design of immunogens to elicit 2G12-like neutralizing antibodies as a component of an HIV vaccine.

There is widespread agreement that the most promising approach to contain the ongoing HIV pandemic is through the development of an effective vaccine (1, 2). However, HIV vaccine design has faced many difficulties including, prominently, the lack of an immunogen able to elicit broadly neutralizing antibodies (Abs). The feasibility of developing such an immunogen is suggested by a small panel of broadly neutralizing human monoclonal antibodies (mAbs) that have been isolated from seropositive donors (3). One of these Abs, 2G12, recognizes a conserved and unusually dense cluster of oligomannose residues on the “silent face” of gp120, the major envelope protein of HIV-1 (4, 5). Many lectins have been identified that can bind to envelope and act antivirally, but 2G12 remains the only known anti-carbohydrate protein that has been specifically elicited to HIV-1 in an adaptive immune response (6). 2G12 has an unusual architecture in which the arms of the IgG swap variable heavy domains, creating a domain-swapped dimer of Fabs (7). The crystal structure of Fab 2G12 complexed with Man₉GlcNAc₂ indicated that the conventional Ab-binding sites are occupied by the D1 arms of the Man₉GlcNAc₂ moieties (1; Fig. 1) (7). The terminal Manα1-2Man residues of the D1 branch account for 85% of the Fab contacts to Man₉GlcNAc₂, although the disaccharide by itself is 50-fold less potent in binding to 2G12 than Man₉GlcNAc₂ (7).

Fig. 1.

Chemical structure of Man₉GlcNAc₂ 1 and oligomannoses 2-9. The individual sugar residues of Man₉GlcNAc₂ are labeled in red. Elsewhere, all mannose residues of oligomannoses 2-9 are labeled to correspond with their structural equivalent on Man₉GlcNAc₂.

Based on these structural results, several initiatives have been launched to design novel immunogens that will elicit 2G12-like Abs (8-12). Recently, we described the design and synthesis of novel antigens, oligomannoses 2-6 (Fig. 1), that bind to mAb 2G12 (13). The Manα1-2Man-containing oligomannoses 4, 5, and 6 were identified as new epitope mimics that inhibit the binding of gp120 to mAb 2G12 as well as, or better than, Man₉GlcNAc₂ (13). Encouraged by this result, we now report the design and synthesis of three additional antigens, oligomannoses 7, 8, and 9 (Fig. 1), using the reactivity-based, one-pot synthesis method (14-18). The ligand specificity of 2G12 was further probed by studying the ability of these oligomannoses to (i) inhibit the binding of 2G12 to gp120 in solution-phase ELISA (4) and (ii) bind 2G12 in microtiter plate-based or glass-slide assays (19-24). We also determined the crystal structures of Fab 2G12 bound to four of these synthetic oligomannoses (4, 5, 7, and 8). Our biochemical, biophysical, and crystallographic results reveal that Fab 2G12 can recognize the terminal Manα1-2Man of both the D1 and D3 arms of Man₉GlcNAc₂. These data confirm that 2G12 is highly specific for terminal Manα1-2Man, but in the context of a broader range of linkages to the third position of the oligomannose moieties than previously thought, which may expedite development of a carbohydrate-based immunogen that could contribute to a HIV-1 vaccine.

Materials and Methods

Oligomannose Synthesis. Building blocks 10 and 13 (13) are known compounds. Experimental details for the synthesis of the key thioglycoside building blocks (12, 16, and 19); the protected Man₇ 14, Man₈ 17, and Man₉ 20; the unprotected Man₇ 7, Man₈ 8, and Man₉ 9; the remaining reaction intermediates 11, 15, and 18; all of the characterization data for 7-9, 11, 12, and 14-25; and Schemes 4-6 are reported separately in Supporting Text, which is published as supporting information on the PNAS web site.

ELISA. Microtiter plate wells (flat bottom, Costar type 3690; Corning) were coated with 50 ng per well gp120_JR-CSF overnight at 4°C. All subsequent steps were performed at room temperature. The wells then were washed four times with PBS/0.05% (vol/vol) Tween 20 (Sigma) by using a microplate washer (SkanWash 400, Molecular Devices) before blocking for 1 h with 3% (mass/vol) BSA. IgG 2G12, diluted to 0.5 μg/ml (25 ng per well) with 1% (mass/vol) BSA/0.02% (vol/vol) Tween 20/PBS (PBS-BT), then was added for 2 h to the antigen-coated wells in the presence of serially diluted oligomannoses starting at a concentration of 2 mM. Unbound Ab was removed by washing four times as described above. Bound 2G12 was detected with an alkaline phosphatase-conjugated goat anti-human IgG F(ab′)₂ Ab (Pierce) diluted 1:1,000 in PBS-BT. After 1 h, the wells were washed four times, and bound Ab was visualized with p-nitrophenol phosphate substrate (Sigma) and monitored at 405 nm.

Carbohydrate Microarray Analysis. Ninety-six-well N-hydroxysuccin-imide-coated microtiter plates (NoAb Biodiscoveries, Mississauga, ON, Canada) were treated with 200 μl of a 1 mM methanolic solution of the amino-functionalized disulfide and alkyne containing linker (see Scheme 3) containing 5% diisopropylethylamine and incubated overnight at 4°C. The microtiter plate then was washed with 2 × 200 μl of methanol and 2 × 200 μl of water. Next, 200-μl solutions of azido-functionalized oligomannose derivatives 21-25 at varying concentrations from 0 to 500 μM in 5% diisopropylethylamine/methanol were introduced. A sprinkle of copper(I) iodide was added, and the contents were allowed to react overnight at 4°C. The next day, the contents were removed, and the plates were washed with 2 × 200 μl of methanol and 2 × 200 μl of water. The plates then were blocked with 0.1% Tween 20 solution in Hepes buffer (pH 7.5) at 4°C for 1 h and then washed with 3 × 200 μl of Hepes buffer. Next, 200 μl of a 1 μg/ml solution of 2G12 Ab in 0.1% Tween 20/PBS buffer (pH 7.4) was added to the wells for a 1-h incubation at 4°C and then washed with 2 × 200 μl of PBS buffer (pH 7.4). At this point, 200 μl of FITC-tagged goat anti-human IgG Ab (10 μg/ml in PBS; Calbiochem) was added for 1 h at 4°C, and the wells then were washed with 2 × 200 μl of PBS buffer (pH 7.4). Detection of the FITC-tagged secondary Ab was performed in 200 μl of water by using a fluorescence plate reader. The resulting data yielded the oligomannose-2G12 binding isotherms, and Scatchard plot analysis was implemented in the determination of dissociation constants (24).

2G12 Purification, Crystallization, Structure Determination, and Analysis. Human mAb 2G12 (IgG1, κ) was produced by recombinant expression in CHO cells. Fab fragments were produced by digestion of the Ig with papain followed by purification on protein A and G columns, and then concentrated to ≈30 mg/ml. For each complex (Man₄, Man₅, Man₇, and Man₈), the solid sugar ligand was added to the Fab solution to saturation. For crystallization, 0.6 μl of protein plus sugar were mixed with an equal volume of reservoir solution. All crystals were grown by the sitting drop vapor diffusion method with a reservoir volume of 1 ml. Fab 2G12 plus Man₄ crystals were grown with a reservoir solution of 27% polyethylene glycol (PEG) 4000 and 0.05 M sodium acetate; Man₅ cocrystals were grown with 1.6 M Na/K phosphate (pH 6.8); Man₇ cocrystals were grown with 20% PEG 4000 and 0.2 M sodium tartrate; and Man₈ cocrystals were grown with 20% PEG 4000 and 0.2 M imidizole malate (pH 7.0). All crystals were cryoprotected with 25% glycerol. Data were collected at 100 K at the Advanced Light Source beamline 8.2.2 and Stanford Synchrotron Radiation Laboratory beamlines 9-2 and 11-1. All data were indexed, integrated, and scaled with hkl2000 (25) using all observations greater than -3.0σ.

The structures were determined by molecular replacement using the 1.75-Å structure of Fab 2G12 [Protein Data Bank ID code 1OP3 (8)] as the starting model for phaser (26). The Matthews' coefficients of the asymmetric unit suggested that the Fab 2G12 plus Man₄ data contained a single Fab plus sugar complex, whereas the asymmetric unit of the other complexes consisted of two Fab plus sugar complexes. The model building was performed with tom/frodo (27) and refined with cns (Version 1.1; ref. 28) and refmac by using tls refinement (29), using all measured data (with F > 0.0 σ). Tight noncrystallographic symmetry restraints were applied initially but were released gradually during refinement. An R_free test set (5%) was maintained throughout the refinement. Data collection and refinement results are summarized in Table 1.

Table 1. Summary of crystallographic data.

Parameter	Fab 2G12 + Man4	Fab 2G12 + Man5	Fab 2B12 + Man7	Fab 2G12 + Man8
Space group	C222	P212121	P212121	P212121
Unit cell dimensions, Å	a = 144.5, b = 148.3, c = 54.6	a = 44.9, b = 131.8, c = 170.3	a = 44.8, b = 131.1, c = 170.0	a = 45.2, b = 165.7, c = 169.6
Resolution,* Å	30-2.0 (2.05-2.0)	50-2.75 (2.85-2.75)	50-2.33 (2.39-2.33)	50-2.85 (2.93-2.85)
No. of observations	130,911	152,645	294,916	119,281
No. of unique reflections	37,831	32,006	43,875	33,657
Completeness,* %	94.3 (88.0)	89.3 (82.1)	99.5 (98.1)	99.5 (99.9)
Rsym,* %	8.5 (57.6)	8.5 (37.6)	5.3 (41.9)	10.2 (52.9)
Average I/σ†	16.4 (2.4)	23.5 (4.0)	45.1 (3.8)	13.0 (2.4)
Rcryst,* %	28.1 (33.7)	22.2 (34.8)	20.9 (34.7)	22.0 (44.6)
Rfree,* %	32.6 (40.9)	28.6 (51.6)	25.1 (40.8)	27.7 (48.8)
No. of refined atoms	3206/45/121	6468/79/—	6463/93/77	6447/88/—
Fab/ligand/water
values, Å2
Variable domain 1/2	46.2	33.7/48.4	47.8/53.3	44.4/45.2
Constant domain 1/2	72.7	49.3/43.9	67.8/56.5	82.0/68.5
Ligand	32.9	52.0	71.9	43.1
Ramachandran plot, %
Most favored	90.6	80.5	88.1	83.4
Additionally allowed	8.6	16.9	10.8	14.7
Generously allowed	0.0	1.9	0.7	0.8
Disallowed†	0.8	0.7	0.4	1.1
rms deviations
Bond lengths, Å/angles, °	0.016/1.8	0.019/2.0	0.017/1.7	0.018/1.9

Crystals of Fab 2G12 and Man₄ exhibited strong anisotropic diffraction beyond 2.75 Å, although electron density maps to 2.0 Å were superior and clearly interpretable. Constant domains generally have higher B values relative to the variable domains. Ramanchandran values were calculated by using procheck (41).

^*Numbers in parentheses are for the highest resolution shell.

^†Includes residue L51 of each light chain, which commonly exists in a γ turn in all Abs, but is flagged by procheck as an outlier. Other residues designated as disallowed by procheck have a good fit to the corresponding electron density.

Diffraction data for Fab 2G12 in complex with Man₄ suffered from severe anisotropy despite the 2.0-Å diffraction limit. Although we report on the measured data observed to 2.0 Å (I/σ > 2.0 and a completeness of 88% in the highest resolution shell of 2.05-2.00 Å), anisotropic diffraction, which is significant beyond 2.75 Å, leads to modest R values. However, the electron density is very well defined and more easily interpretable at higher resolution. Refinement of the Man₄ structure at a lower resolution of 2.75 Å yields slightly better R_crys and R_free values of 21.9% and 28.2%, respectively, but with significantly poorer-quality electron density maps. Thus, the higher-resolution structure is reported.

Potential H bonds and van der Waals contacts were evaluated by using the program contacsym (30). Buried molecular surface areas were measured by using the program ms (31).

Results and Discussion

Synthesis of Man₇ 7, Man₈ 8, and Man₉ 9. The synthesis of high mannose-type oligosaccharides, Man₉GlcNAc₂ and its analogues, has been extensively explored (11, 32-37) by using a convergent approach that is generally more concise and efficient. The one-pot programmable relative reactivity value (RRV) assay (14) revealed that thioglycoside disaccharide building block 10 (RRV = 790) is less reactive than compound 11 (RRV = 3638), 15 (RRV = 3137), and 18 (RRV = 4398). Thus, the trichloroacetimidate group, which is a more reactive leaving group than p-methylphenylthio, was implemented.

Synthesis of Man₇ 7 is described in Scheme 1 and was performed as described in ref. 38. Thioglycoside disaccharide building block 10 was converted to its trichloroacetimidate derivative, which was activated with tert-butyldimethylsilyl trifluoromethansulfonate (TBDMSOTf) for glycosylation with building block 11 to give trisaccharide building block 12 in good yield (75% over three steps). Convergent synthesis of Man₇ 7 in good yield (85%) was achieved by glycosylation of tetrasaccharide acceptor 13 with trisaccharide donor 12 by using the sodium iodide symporter/trifluoromethane-sulfonic acid promoting system in anhydrous CH₂Cl₂ at -25°C (39). Excellent Manα1-6Man selectivity was controlled by the presence of the tert-butyldimethylsilyl group at the 2-position of trisaccharide donor 12. Global deprotection of protected Man₇ 14 was achieved smoothly through desilylation with tetrabutylammonium fluoride/AcOH buffer (11), deacetylation, and hydrogenolysis to afford unprotected Man₇ 7 (60% in three steps).

Scheme 1.

By using a similar strategy, syntheses of Man₈ 8 and Man₉ 9 were performed as shown in Scheme 2. Thioglycoside disaccharide building block 10 was converted to its trichloroacetimidate derivative and glycosylated by using building block 15 (1.1 equivalent) or 18 (0.45 equivalent) to give the tetrasaccharide building block 16 (75% over three steps) and pentasaccharide building block 19 (65% over three steps). In the convergent synthesis of Man₈ 8 and Man₉ 9, the Manα1-6Man selectivity was controlled by implementing Seeberger's protocol (36). Thioglycoside tetrasaccharide 16 and pentasacharride 19 were converted to the corresponding trichloro-imidates, which were coupled to tetrasaccharide 13 to give protected Man₈ 17 (75% over three steps) and Man₉ 20 (75% over three steps). Global deprotection of protected Man₈ 17 and Man₉ 20 was achieved through deacetylation and hydrogenolysis to afford unprotected Man₈ 8 (60% in two steps) and Man₉ 9 (60% in two steps).

Scheme 2.

Solution-Phase ELISA Analysis of Oligomannose 1-9 Inhibition of 2G12 Binding. Man₉GlcNAc₂ 1 (7) and deprotected oligomannoses 2-6 (13) and 7-9 were evaluated for their ability to inhibit the interaction between 2G12 and gp120 in a solution-phase ELISA. These results (Fig. 2) confirmed that terminal Manα1-2Man is critical for binding. All of the oligomannoses that contain a Manα1-2Manα1-2Man motif (which corresponds to the D1 arm of Man₉GlcNAc₂) are capable of inhibiting 2G12 binding at similar levels to the intact Man₉GlcNAc₂ moiety. However, 2G12 does not readily recognize Manα1-2Manα1-3Man, because oligomannose 3 does not inhibit effectively at lower concentrations (15.8% at 0.5 mM). Oligomannose 5, which is similar to oligomannose 3, but contains the Manα1-2Manα1-6Man motif, is capable of inhibition (37.7% at 0.5 mM). These results suggest that 2G12 recognizes Manα1-2Man in the context of the D1 arm (Manα1-2Manα1-2Man) or the D3 arm (Manα1-2Manα1-6Man), but not the D2 arm (Manα1-2Manα1-3Man). Overall, many of the oligomannose derivatives can compete for binding of Man₉GlcNAc₂ and, therefore, may serve as building blocks for potential immunogens to elicit 2G12-like Abs.

Fig. 2.

Analysis of 2G12 Ab binding to oligomannoses. (Left) ELISA of Oligomannose inhibition (%) of 2G12 binding to gp120 coated in microtiter plates. Black and gray bars represent the level of inhibition at oligomannose concentrations of 0.5 and 2.0 mM, respectively. (Right) Covalent oligomannose arrays in microtiter plates for analysis of 2G12 binding.

Carbohydrate Microarray Analysis. Recently, we reported the study of a panel of carbohydrate epitopes for interaction with 2G12 by using covalent microtiter plates display (21, 23) by converting the amine-containing oligomannoses 4, 5, 7, 8, and 9 to the corresponding azide derivatives 21, 22, 23, 24, and 25 (40). These derivatives then were covalently attached to a microplate-immobilized cleavable linker via the Cu(I)-catalyzed 1,3-dipolar cycloaddition reaction (23). K_d values for the interaction of 2G12 with oligomannoses 4, 5, 7, 8, and 9 were determined by using a microtiter-based assay with detection through a fluorescent secondary Ab (23) (Fig. 2). The K_d values were as follows: oligomannose 4, 0.1 μM; 5, 0.1 μM; 7, 0.7 μM; 8, 1.3 μM; and 9, 1.0 μM. Significant enhancement of the binding affinity of these compounds in microarray studies may be explained by multivalent interactions of the oligomannoses with 2G12 that mimic the cluster of oligomannoses on the surface of gp120. The smaller oligomannose derivatives especially benefit from multivalent display. Thus, this system may be an effective model for studying binding events involving carbohydrates presented on a surface, such as that of a virus.

Crystal Structure of Fab 2G12 with Man₄. In the crystal structure of Fab 2G12 in complex with Man₄ (oligomannose 4), the Man₄ sugar is located in the combining site of the Ab and has extremely well defined electron density (Fig. 3A). As expected, the Manα1-2Man terminus of the sugar (consisting of mannoses D1 and C) binds end-on in a cleft in the binding site. An overlay with a previous structure of Fab 2G12 plus Man₉GlcNAc₂ (7) shows that Man₄ binds with a similar geometry to the D1 arm of Man₉GlcNAc₂ (Fig. 3B) that correlates with the biochemical data that indicate Man₄ binds as well as the entire Man₉GlcNAc₂ moiety (13). The crystal structure of the Fab 2G12-Man₄ complex is of higher resolution and allows better characterization of the interactions between Fab and sugar ligand. A total of 51 van der Waals interactions and 15 H bonds are formed on complex formation; 292 Ų of molecular surface is buried on Fab 2G12 and 294 Ų is buried on the sugar in the complex. Although the bulk of the interactions are with Manα1-2Man (mannoses D1 and C), three H bonds are made between 2G12 and mannose 4, two of which are water-mediated (Fig. 3C). Tyr^L94 H bonds to the O3 oxygen of mannose 4, whereas water molecules coordinate between the O5 of mannose D1 and the O3 of mannose 4, and between Tyr^H56 and the O4 of mannose 4.

Fig. 3.

Crystal structure of Fab 2G12 with Man₄. (A)2F_o - F_c electron density map of the Man₄ sugar (shown in ball-and-stick) contoured at 1.8σ. Mannose residues are labeled in red. The light and heavy chains of Fab 2G12 are shown in cyan and purple. (B) Overlap of the Man₄ sugar (shown in blue) with a previous Fab 2G12 structure (7) bound to Man₉GlcNAc₂ (shown in green). The Man₄ clearly adopts a conformation similar to the D1 arm of Man₉GlcNAc₂. (C) Additional H bonds between Fab 2G12 and mannose 4 of Man₄. Two of the H bonds are water mediated (water molecules shown as gray spheres). These interactions may explain in the preference for D1 arm binding in the context of the full Man₉GlcNAc₂ moiety. Figures were made by using molscript (42), bobscript (43), and raster3d (44).

Crystal Structure of Fab 2G12 with Man₅. Man₅ (oligomannose 5) is the structural equivalent of the D2 and D3 arms of Man₉GlcNAc₂. 2G12 binds to Man₅ with similar apparent affinity as Man₄ and Man₉GlcNAc₂ (13). The Manα1-2Man moieties (mannoses D3 and B), which correspond to the D3 arm of Man₉GlcNAc₂, are well ordered in the primary combining site (Fig. 4A). The other Manα1-2Man (mannoses D2 and A) branch, which corresponds to the D2 arm of Man₉GlcNAc₂, points away from the binding site, confirming biochemical studies that suggest 2G12 recognizes the D3 arm in preference to D2. 302 Ų of molecular surface is buried on Fab 2G12 and 307 Ų is buried on Man₅ upon complex formation. Unlike Man₄ (D1 arm) binding, no additional H bonds are made to sugars other than with the terminal Manα1-2Man.

Fig. 4.

2G12 interactions with Man₅ and Man₈. (A)2F_o - F_c electron density map of the Man₅ sugar (shown in ball-and-stick) bound at the primary combining site of 2G12 contoured at 0.8σ. (B) Crystal structure of the domain-swapped Fab 2G12 dimer with Man₈. The light chains are shown in cyan, and the heavy chains are shown in red and purple. One Man₈ sugar (shown in ball-and-stick) is bound at each primary combining site of the domain-swapped Fab dimer, although in one site the sugar is bound by the D1 arm, whereas in the other the sugar is bound by the D3 arm. The D1 and D3 “arms” of the Man₈ sugar are labeled. (C) 2F_o - F_c electron density map of the Man₈ sugar (shown in ball-and-stick) contoured at 1.5σ. The Man₈ sugar cross-links the crystal together, with a D1 arm bound in the primary combining site of one Fab molecule and the D3 arm bound in the primary combining site of a different, crystallographically related Fab molecule. Figures were made by using molscript (42), bobscript (43), and raster3d (44).

Crystal Structure of Fab 2G12 with Man₇. Man₇ (oligomannose 7) is the structural equivalent of the D1 and D2 arms of Man₉GlcNAc₂. One Man₇ moeity is bound to each Ab-combining site of the 2G12 domain-swapped Fab dimer, but only four of the seven sugars are clearly interpretable in the electron density maps. In general, sugars are extremely flexible and give rise to poorly defined electron density unless they are somehow “fixed,” either by direct binding along the length of the entire sugar chain or through the incorporation of the sugar into the crystal lattice. In this case, as opposed to the Man₈ (see below) and Man₉GlcNAc₂ (1) sugars that cross-link the crystals together, and hence give well-defined sugar electron density, no such cross-linking occurred.

However, the D1 arm, and not the D2 arm, can be clearly resolved in the Ab-combining site. This Man₇ moiety overlaps well with the corresponding Man₄ structure and the D1 arm of the Man₉GlcNAc₂ structure (8). The interactions between 2G12 and Man₇ are similar to those between 2G12 and Man₄, including the H bonds with mannose 4.

Crystal Structure of Fab 2G12 with Man₈. Man₈ (oligomannose 8) is nearly equivalent to the entire Man₉ moeity, with only the terminal sugar of the D2 arm missing. In the Man₈ crystal structure, the two primary combining sites of the domain-swapped dimer each interact with a Man₈ moeity but bind to different arms of the sugar. One binding site interacts with the D1 arm, whereas the other is occupied by the D3 arm (Fig. 4B). In this crystal lattice, the Man₈ sugar cross-links the crystal and is, therefore, well ordered, as indicated by its well-resolved electron density. Each Man₈ sugar moiety in the crystal has its D1 arm bound in one of the two combining sites of an Fab 2G12 dimer, while its D3 arm is bound in one of the combining sites of a crystallographically related Fab 2G12 dimer (Fig. 4C). Thus, the Man₈ sugar cross-links the crystal lattice together, which is very different from what was previously observed with Man₉GlcNAc₂, in which the D1 arms of two Man₉GlcNAc₂ interacted with each of the two combining sites of the domain-swapped dimer and the D2/D3 arms with the novel V_H/V_H′ surface of a crystallographically related Fab dimer (7).

The Man₈ D1 arm interactions with Fab 2G12 are similar to those observed in the Man₄, Man₇, and Man₉GlcNAc₂ structures. For interaction with the D3 arm, 336 Ų of molecular surface of Man₈ and 355 Ų of Fab 2G12 are buried during complex formation, with 69 van der Waals interactions and 12 H bonds. No additional interactions occur with the D3 arm beyond the terminal Manα1-2Man.

Conclusions

We have expanded on our previous synthesis and study of oligomannose derivatives using the reactivity-based one-pot self-condensation reaction. We have prepared three previously undescribed Manα1-2Man-containing oligomannose compounds, Man₇ 7, Man₈ 8, and Man₉ 9, that have reactivity with broadly neutralizing HIV-1 Ab 2G12. Overall, the carbohydrate specificity of 2G12 is less restrictive than originally believed (7). The combined biochemical, biophysical, and crystallographic evidence clearly indicates that 2G12 can bind to the Manα1-2Man at the termini of both the D1 and D3 arms of an oligomannose sugar. In the Man₄, Man₇, and Man₈ crystal structures, 2G12 interacts with the D1 arm, whereas in the Man₅ and Man₈ crystal structures, the D3 arm can also bind in the combining site (Fig. 5). Therefore, 2G12 can bind not only the D1 arms from two different N-linked oligomannoses on gp120 but also to both the D1 and D3 arms from different sugars within the oligomannose constellations on gp120. This mode of recognition would enhance binding to a cluster of oligomannose moieties and relax the constraint of an exact match of the oligomannose moieties with respect to the multivalent binding site of the Ab. Nevertheless, despite this increased potential for multivalent interaction, 2G12 is highly restricted to oligomannose cluster binding on gp120, because no significant binding to “self” proteins has been observed.

Fig. 5.

Overlap of bound oligomannoses in six cocrystal structures with 2G12. The Man₄, Man₇, Man₈, and previously described Man₉GlcNAc₂ structures (shown as ball-and-stick in various shades of gray) all show 2G12 binding to the Manα1-2Manα1-2Man motif. The Man₅ and Man₈ structures (shown as ball- and-stick in shades of red) also show 2G12 binding to the Manα1-2Manα1-6Man motif, although only the two terminal sugars make contact. Full details of the Manα1-2Man interactions can be found in ref. 7. The figure was made by using molscript (42) and raster3d (44).

The 2G12 Ab can neutralize a broad range of HIV-1 isolates. The results presented here reveal more precisely the carbohydrate specificity of this Ab. This deeper understanding of the 2G12-oligomannose interaction can now be applied to carbohydrate-based immunogen design, because the nature of the mannose building blocks needed to design a multivalent oligomannose presentation for immunization trials has been established.