Changes in Structure and Antigenicity of HIV-1 Env Trimers Resulting from Removal of a Conserved CD4 Binding Site-Proximal Glycan

ABSTRACT

The envelope glycoprotein (Env) is the major target for HIV-1 broadly neutralizing antibodies (bNAbs). One of the mechanisms that HIV has evolved to escape the host's immune response is to mask conserved epitopes on Env with dense glycosylation. Previous studies have shown that the removal of a particular conserved glycan at N197 increases the neutralization sensitivity of the virus to antibodies targeting the CD4 binding site (CD4bs), making it a site of significant interest from the perspective of vaccine design. At present, the structural consequences that result from the removal of the N197 glycan have not been characterized. Using native-like SOSIP trimers, we examine the effects on antigenicity and local structural dynamics resulting from the removal of this glycan. A large increase in the binding of CD4bs and V3-targeting antibodies is observed for the N197Q mutant in trimeric Env, while no changes are observed with monomeric gp120. While the overall structure and thermostability are not altered, a subtle increase in the flexibility of the variable loops at the trimeric interface of adjacent protomers is evident in the N197Q mutant by hydrogen-deuterium exchange mass spectrometry. Structural modeling of the glycan chains suggests that the spatial occupancy of the N197 glycan leads to steric clashes with CD4bs antibodies in the Env trimer but not monomeric gp120. Our results indicate that the removal of the N197 glycan enhances the exposure of relevant bNAb epitopes on Env with a minimal impact on the overall trimeric structure. These findings present a simple modification for enhancing trimeric Env immunogens in vaccines.

IMPORTANCE The HIV-1 Env glycoprotein presents a dense patchwork of host cell-derived N-linked glycans. This so-called glycan shield is considered to be a major protective mechanism against immune recognition. While the positions of many N-linked glycans are isolate specific, some are highly conserved and are believed to play key functional roles. In this study, we examine the conserved, CD4 binding site-proximal N197 glycan and demonstrate that its removal both facilitates neutralizing antibody access to the CD4 binding site and modestly impacts the structural dynamics at the trimer crown without drastically altering global Env trimer stability. This indicates that surgical glycosylation site modification may be an effective way of sculpting epitope presentation in Env-based vaccines.

INTRODUCTION

The trimeric HIV-1 envelope glycoprotein (Env), a trimer composed of gp120/gp41 heterodimers, is the primary antigenic feature on the virus and the sole target for neutralizing antibodies (1). Despite the extensive genetic diversity that exists among circulating HIV-1 variants, broadly neutralizing antibodies (bNAbs) capable of neutralizing a diverse panel of viral isolates have been identified in rare HIV-infected individuals. Elicitation of such bNAbs is thought to be a critical requirement for an effective HIV-1 vaccine (2, 3), but to date, HIV vaccine efforts have resulted in limited, narrow neutralization activity and have failed to elicit bNAbs (4–7).

Nearly 50% of the molecular mass of Env is contributed by host cell-derived N-linked glycans, and this dense glycan shield is considered to be a major protective mechanism against immune recognition (8, 9). Glycans play important roles in Env folding, viral assembly and infectivity, and modulating the immune response (10–12). Although glycans typically attenuate antigenicity by occluding polypeptide epitopes, several conserved glycans are actually targets for potent HIV-1 bNAbs (13–18). Specific glycans within variable loops, within conserved C2-C4 regions, and within gp41 were found to affect HIV-1 sensitivity to neutralizing antibodies (12, 15, 19). One of the most conserved epitopes on Env, the CD4 binding site (CD4bs), is a recessed pocket on gp120 surrounded by glycans (20). Removal of the glycans peripheral to the CD4 binding site results in increased sensitivity to neutralization (21–24).

Recent studies have intensified interest in glycan modification as a means of sculpting epitope accessibility in Env-based HIV-1 vaccine immunogens. For example, it has been demonstrated that the removal of the glycosylation site at N276 in the gp120 D-loop can increase the reactivity of germ line precursor forms of CD4bs bNAbs such as VRC01 (25–28). By and large, however, the consequence of glycosylation site modification on trimer structure and stability has not been directly probed. While glycans are often considered to be flexible decorations on a more structurally defined protein substrate, recent structures have revealed significant ordered density for glycan chains (29, 30) as well as interactions between glycan chains themselves and with the protein surface. It is conceivable that the removal of specific glycans can impact the presentation of others and may impact the stability of the trimer by removing favorable interactions.

The N197 glycosylation site is highly conserved (>92%) among HIV-1 Env sequences. It is located at the base of the V2 loop, proximal to the CD4bs (12). Removal of the N197 sequon was found to enhance the neutralization sensitivity of diverse HIV-1 isolates to CD4bs- and V3-specific antibodies and, more importantly, to induce higher levels of neutralizing antibodies with greater breadth in macaques (12, 21, 23, 24, 31). However, in the context of monomeric gp120, the removal of N197 had no significant effect on the binding of CD4 or neutralizing antibodies (24). Direct probing of the role of the N197 glycan in the context of trimeric Env has been hampered due to the difficulty in obtaining homogeneous preparations of native-like Env trimers for analysis (32, 33). Recently, a stabilized soluble Env trimer (“SOSIP”) was developed, which has structural and antigenic characteristics as well as a glycosylation pattern similar to those of native HIV-1 Env from virions (34–36). Although the stabilizing mutations are likely to alter some of the properties of the SOSIP Env trimer in solution, it presents the only available reagent for studying trimeric Env, and several studies have shown that antibody binding to SOSIPs correlates strongly with neutralization potency (34).

Here we used the SOSIP trimers to examine the structural and antigenic effects of N197 glycan removal in a native-like Env trimer. Biolayer interferometry (BLI) was used to measure binding kinetics for several bNAbs to track antigenic changes in gp120 monomers and the native-like trimer. Hydrogen-deuterium exchange mass spectrometry (HDX-MS) was applied to probe local structural dynamic changes resulting from the removal of N197. All-atom modeling of the glycosylated Env trimer was used to visualize glycan occlusion of various epitopes. Overall, this study shows that in the context of an Env trimer, the N197 glycan limits the accessibility of neutralizing antibodies targeting the CD4bs while also influencing the local dynamics of variable loops at the interface of adjacent protomers. Additionally, we find that global trimer stability was not significantly impacted by the removal of the glycan.

MATERIALS AND METHODS

Expression and purification of trimeric envelope glycoproteins.

The expression vector for trimeric BG505 SOSIP.664 Env and full-length furin expression vectors were kindly supplied by John P. Moore (34, 37–40). To remove the N-glycan at residue 197 of Env, the original asparagine (N) residue was replaced with glutamine (Q) by using a QuikChange II XL site-directed mutagenesis kit (Stratagene). The env gene for full-length PVO.4 (41) was synthesized by GeneArt Gene Synthesis Service (Invitrogen) and cloned into expression vector pPPI4 by using PstI and NotI sites (38). Mutations A501C, T605C, and I559P; six arginine residues replacing the REKR furin cleavage sequence; and a stop codon after residue 664 were introduced into the PVO.4 env gene to create the PVO.4 SOSIP vector. Mutation N197Q was then introduced to generate a PVO.4 SOSIP N197Q expression vector.

Env trimers were expressed in HEK293F cells by transient transfection using FreeStyle Max 293 (Life Technologies, Thermo Fisher Scientific). Six days after transfection, the supernatant was collected and passed through a 0.45-μm filter (Fisher Scientific). A cocktail of protease inhibitors (Roche Diagnostics) was added, and the supernatant was loaded onto a Galanthus nivalis lectin (GNA; Vector Laboratories) affinity column. The eluted protein from the GNA column was buffer exchanged into a solution containing 100 mM NaCl and 20 mM Tris-Cl (pH 8.0) and then loaded onto HiTrap DEAE Sepharose FF (GE Healthcare). The flowthrough from the DEAE column was collected, buffer exchanged into a solution containing 2 M (NH4)₂SO₄ and 0.1 M NaH₂PO₄ (pH 7.0) (buffer A), and loaded onto a proPac HIC-10 column (Thermo Scientific) equilibrated with buffer A. Trimeric Env was eluted from the hydrophobic interaction chromatography (HIC) column with a linear gradient from 0 to 100% buffer B (0.1 M NaH₂PO₄, pH 7.0). Subsequent purification of the Env trimer was done by gel filtration using a HiLoad 16/60 Superdex 200 column (GE Healthcare) equilibrated with 1× phosphate-buffered saline (PBS) (pH 7.4) (1 mM EDTA, 0.02% NaN₃) (89). Fractions (500 μl each) were collected and analyzed by using SDS-PAGE and blue native PAGE (BN-PAGE).

Dynamic light scattering (DLS).

The hydrodynamic radius (R_h) of Env trimer samples (1 mg/ml in PBS) was measured by using a Dynapro Nanostar instrument (Wyatt Technology, Santa Barbara, CA) from 20°C to 90°C in 1°C increments. A total of 5 acquisitions of 5 s were collected at each temperature. Dynamics software (Wyatt Technologies) was used to calculate the onset temperature at which R_h begins to change significantly and the midpoint of the transition curve (T_m).

Small-angle X-ray scattering.

Small-angle X-ray scattering (SAXS) measurements were conducted on Beam Line 4-2 at the Stanford Synchrotron Radiation Laboratory with online size exclusion chromatography (SEC) as described previously (42, 43). Fifty microliters of the BG505 trimer (2 mg/ml) was injected over a precision S200 column (GE Healthcare) and eluted with PBS (pH 7.4) with a solution containing 1 mM EDTA and 0.02% sodium azide. The first 100 data points (before the void volume) were averaged as buffer scattering data and subtracted from the corresponding protein scattering data. SAXS patterns, the radius of gyration (R_g), the maximal particle dimension (D_max), and the pairwise distance distribution histogram [P(r) plot] were analyzed by using the ATSAS software suite (44, 45).

Negative-stain electron microscopy (EM).

A 3-μl aliquot of the BG505 SOSIP.664 wild type (WT) or the N197Q mutant, diluted to 20 μg/ml in PBS, was applied for 60 s to glow-discharged C-Flat, 300-mesh, Cu grids (Electron Microscopy Sciences) and stained for an additional 60 s by using Nano-W (Nanoprobes). Data were collected by using an FEI Tecnai T12 transmission electron microscope operating at 120 keV. Images were collected by using a Gatan Ultrascan 4000 charge-coupled-device (CCD) camera at a magnification of ×52,000 and with a 1-μm defocus corresponding to a pixel size of 2.07 Å. Particle picking and classification were performed by using the EMAN2.1 image processing suite (46). In short, particles were selected by using interactive particle picking from 98 micrographs. A 2-fold binned, phase-flipped, contrast transfer function-corrected stack of ∼9,600 particles were created and subjected to reference-free two-dimensional (2D) classification to generate 50 2D classes. From these classes, a representative panel was chosen for illustrative purposes.

Biolayer interferometry.

An Octet RED 96 biolayer interferometry system (FortéBio, Menlo Park, CA) was used to measure the binding of each Env trimer to different antibodies. Antibodies 17b (47, 48); A32 (49); VRC01 (50); NIH45-46 (51); VRC03 (50); F105 (52); b12 (53); PG9 and PG16 (16); PGT121, PGT126, and PGT145 (54); 2G12 (55); and CD4-IgG2 (56) were obtained from the NIH AIDS Reagents Program. Anti-human IgG Fc capture (AHC) biosensors were hydrated for at least 10 min prior to antibody loading. Human anti-HIV antibody (10 μg/ml) in binding buffer (1× PBS [pH 7.4], 1% bovine serum albumin [BSA], 0.03% Tween 20, 0.02% NaN₃) was then bound to the AHC tips for 240 s and washed in binding buffer for 60 s to reach a stable baseline. The AHC antibody tips were then dipped into Env trimer samples prepared with a 2-fold dilution series. After association for several minutes, the tips were then placed back into the baseline buffer to allow dissociation to occur. According to different antibodies, the times for loading of antibody, association, and dissociation were optimized. Data Analysis 7.0 (FortéBio) was used to analyze changes in the refractive index, and the association and dissociation curves were fit by using a 1:1 binding model. The final association rate (k_a) and dissociation rate (k_d) rate constants were used to calculate the ensemble dissociation constants (K_D = k_d/k_a).

Hydrogen-deuterium exchange mass spectrometry.

Hydrogen-deuterium exchange was initiated by diluting 100 μg/ml Env trimer protein 10-fold into phosphate buffer (10 mM phosphate, 144 mM NaCl [pH 7.4]) with 85% D₂O (Cambridge Isotope Labs). Hydrogen-deuterium exchange was carried out at room temperature for 3 s, 1 min, 30 min, and 20 h. At each time point, the solution was quenched by 1:1 dilution into prechilled 200 mM Tris(2-carboxyethyl)phosphine hydrochloride (TCEP) in 0.02% formic acid for a final pH of 2.5. The samples were frozen in liquid nitrogen and stored at −80°C. A zero-time-point sample and a fully deuterated sample were prepared as previously described (57). HDX measurements were performed by liquid chromatography-mass spectrometry (LC-MS) on a Synapt G2-Si instrument coupled with Waters HDX manager, keeping all loops, lines, and columns, at 1°C (Waters, Milford, MA). Samples were loaded over a 2.1- by 50-mm column filled with POROS-immobilized pepsin (58) at 150 μl/min; captured on a 2.1- by 5-mm, 1.7-μm ethyl-bridged hybrid (BEH) C₁₈ trap column (Waters) for 3 min; and subsequently resolved over a 1- by 100-mm, 1.7-μm ethyl-bridged hybrid (BEH) C₁₈ column (Waters). Optimized settings were used to circumvent source-induced loss of deuterium (59). Deuterium uptake was analyzed with HX-Express v2 (60), and percent exchange is reported relative to the fully deuterated and zero time points.

Modeling of fully glycosylated monomeric gp120 and SOSIP trimers.

Models of glycosylated BG505 SOSIP.664 trimers and gp120 monomers were generated by threading the BG505 sequence into the available crystal structure of the SOSIP.664 trimer (PDB accession number 4TVP) (61) by using MODELLER (62). In monomeric gp120, the N/C-terminal tails, the V1/V2 loop, and the V3 loop are highly dynamic (43, 57). For this reason, residues 32 to 41, 495 to 511, 114 to 205, and 299 to 326 were left unrestrained, with the exception of disulfide bond connectivities during modeling. Glycan chains (Man₉ structures) were attached and treated as flexible residues during simulated annealing, as described previously (42). Forty models were generated for the trimer, and 120 models were generated for the gp120 monomers, using 5 rounds of refinement. Each round of refinement included simulated annealing from 1,300 K to 300 K in 8 steps with 1,000 iterations at each step. Each protomer was aligned to cocrystal structures of CD4-gp120 (PDB accession number 3JWD) (63), VRC01-gp120 (PDB accession number 3NGB) (64), VRC03-gp120 (PDB accession number 3SE8) (65), F105-gp120 (PDB accession number 3HI1) (66), b12-gp120 (PDB accession number 2NY7) (67), and PG9-V1/V2 (PDB accession number 3U4E) (68). Steric clashes (>1-Å overlap) between the various ligands and specific glycan chains in each model were assessed with PyMOL (69). The relative positions of PGT122 Fab and 2G12 Fab₂ were obtained from data reported under PBD accession number 4TVP (61) and Electron Microscopy Data Bank accession number EMD-5982 (70), respectively.

RESULTS

Removal of the N197 glycan does not alter the global structure of SOSIP trimers.

To examine the effects of N197 glycan removal within a native trimer, the N197Q mutation was introduced into the BG505 SOSIP construct, and both WT and N197Q trimers were transiently expressed in HEK293F cells. The two Env trimers were purified by a combination of lectin affinity chromatography, anion exchange chromatography, HIC, and a final SEC step to generate pure trimer preparations (89). The SEC elution profiles of the BG505 SOSIP.664 WT and the N197Q mutant showed the expected trimer peaks with identical elution volumes (Fig. 1A). The purity and homogeneity of the purified trimers were confirmed by SDS-PAGE and BN-PAGE (Fig. 1B and C).

FIG 1.

Purification of WT and N197Q SOSIP trimers. (A) Size exclusion profile of purified fully glycosylated WT and N197Q BG505 SOSIP expressed in HEK293F cells. Abs, absorbance. (B) SDS-PAGE analysis of purified BG505 (left) and PVO.4 (right) WT and N197Q mutant trimers under reducing and nonreducing conditions. DTT, dithiothreitol. (C) BN-PAGE analysis of WT and N197Q mutant trimers. (D) Sequence alignment of BG505 and PVO.4 at the V1/V2 region of Env. Protein alignment was performed by using Vector NTI AlignX, and the color scheme represents amino acid identity (yellow) and blocks of similarity (green) (88).

We next tested whether the removal of the N197 glycan affects thermal stability using dynamic light scattering (DLS) to monitor the hydrodynamic radius (R_h) with increasing temperatures. The R_h of the WT trimer was constant at 6.9 ± 0.04 nm up to 66°C (Fig. 2A). Between 66°C and 73°C, the R_h shifted to 9.1 nm, likely corresponding to a partial unfolding event. The R_h remained at 9.1 nm until ∼78°C, at which point it increased further, presumably due to further unfolding resulting in protein aggregation. The midpoint melting temperature (68.5°C) is consistent with the thermal denaturation midpoint of BG505 SOSIP (68.1°C) observed by differential scanning calorimetry (34). The N197Q mutant showed a very similar R_h and thermal profile, with a slightly lower midpoint unfolding temperature (67.6°C) (Fig. 2A). The nature of the aggregation appears different in the mutant, as seen in the change in the apparent R_h at temperatures above 78°C; however, the temperature of the transition was very similar. Therefore, the N197 glycan deletion appears to have only a very minor thermodynamic destabilizing effect on the trimer.

FIG 2.

Biochemical and biophysical comparison of the purified BG505 SOSIP.664 WT and the BG505 SOSIP.664 N197Q mutant. (A) Thermal stability of the BG505 SOSIP.664 WT and the BG505 SOSIP.664 N197Q mutant analyzed by DLS. The BG505 SOSIP.664 N197Q mutant presents slightly lower onset and midpoint temperatures of melting, indicating slightly less stability than the wild-type trimer. (B) SAXS parameters for the BG505 SOSIP.664 WT and the BG505 SOSIP.664 N197Q mutant. Radii of gyration (Rg) from Guinier approximation and from the P(r) plot calculated by the program GNOM were in good agreement. (C) 2D class averages of EM reconstruction of the BG505 SOSIP.664 WT and the N197Q mutant. The two display 3-fold symmetry and are consistent with trimeric, native-like morphology.

SAXS and electron microscopy (EM) were used to examine any significant effect of the N197Q mutation on the overall morphology of BG505 SOSIP trimers. The SAXS profiles of both constructs were nearly superimposable, indicating that the structures are highly similar (Fig. 2B). However, the N197Q mutant showed a subtle but significant increase in the radius of gyration (R_g) (Fig. 2B, inset). Real-space analysis also showed a significantly higher R_g and suggested a slightly higher maximal particle dimension (D_max) value. Class averages from negative-stain EM showed that the overall trimeric structure was not detectably altered by the removal of the N197 glycan (Fig. 2C), and both of the trimers appeared homogeneous and well ordered, as seen with high-quality trimers examined previously (34, 71).

Effects of the N197 glycan on the structural dynamics of Env trimers determined by HDX-MS.

To see if the removal of N197 affects structural flexibility, we compared the local structural dynamics of WT and N197Q SOSIP trimers by hydrogen-deuterium exchange with mass spectrometry (HDX-MS). A total of 50 peptic peptides were monitored, covering 83% of the BG505 SOSIP sequence (see Fig. S1 in the supplemental material). Due to heterogeneous glycosylation, we were unable to observe portions of V1/V2, V3, V4, and the gp41 loop. The overall exchange profiles for the WT and N197Q trimers were remarkably similar and consistent with data from previous HDX-MS studies of BG505 SOSIP trimers (Fig. 3A) (35). No differences in exchange within the V3 loop, the core of gp120, or gp41 were observed (Fig. 3B). A small but reproducible difference was observed at the base of the V1/V2 loop right at the trimeric interface of adjacent protomers. In the N197Q mutant, this region is slightly less protected, indicating that it is more flexible. Other subtle changes were observed at the bridging sheet, specifically at residues 112 to 128 and 416 to 426, which are also slightly more accessible in the N197Q mutant. Overall, the comparisons show that both trimers adopt stable trimeric conformations, but the N197Q mutation imparts a slight increase in structural flexibility around the quaternary interactions within V1/V2 and the bridging sheet segments.

FIG 3.

Local dynamic changes detected by HDX in the presence and absence of the N197 glycan. (A) Differences in deuterium exchange are plotted on the trimer crystal structure (PDB accession number 4TVP) (61). V1/V2 base and β20 bridging sheets are slightly less protected (red) in the BG505 SOSIP.664 N197Q trimer. The N197 glycan base is shown in yellow. (B) Individual deuterium uptake plots of selected peptides with error bars from replicate experiments. All uptake plots are shown in Fig. S1 in the supplemental material.

Removal of the N197 glycan increases binding of CD4 and CD4bs bNAbs.

Having shown that the N197 glycan-deleted trimer retained a proper native-like trimeric conformation, we sought to investigate how the removal of this glycan results in the antigenic differences that were observed in previous studies (23, 24, 31). We compared the binding kinetics and affinities of CD4-IgG2 and various CD4bs antibodies for WT and N197Q mutant BG505 SOSIP trimers as well as the corresponding gp120 monomers by biolayer interferometry (BLI) (Fig. 4; see also Fig. S2 in the supplemental material). The N197Q trimer showed stronger binding to CD4-IgG2, with a higher (1.6-fold) association rate and a dramatically lower dissociation rate (860-fold lower). Similarly, the CD4bs bNAbs VRC01, NIH45-46, and VRC03 all showed higher association rates (1.8-, 1.5-, and 3.5-fold) and lower dissociation rates than with the WT (Table 1). In the context of gp120 monomers, the N197Q mutant did not affect the binding kinetics of CD4-IgG2 and had only minor effects on the binding of CD4bs bNAbs. The only notable effect was a 2-fold increase in the association rate for VRC03 (Table 1).

FIG 4.

Biolayer interferometry to assess antibody binding kinetics. Sensograms show the BG505 SOSIP.664 WT or the BG505 SOSIP.664 N197Q mutant and the WT BG505 gp120 monomer or the N197Q mutant BG505 gp120 monomer for binding to and unbinding from immobilized CD4 and antibodies targeting different epitopes on Env. The full concentration series of the sensograms are shown in Fig. S2 in the supplemental material. All analyses were performed at least in duplicate.

TABLE 1.

Octet biolayer interferometry binding parameters for various antibodies and Env constructs^a

Antibody and parameter	Mean value for construct ± SD
	BG505 SOSIP.664		BG505 gp120		PVO.4 SOSIP.664
	WT	N197Q	WT	N197Q	WT	N197Q
CD4bs antibodies
CD4-IgG2
ka (M−1 s−1)	(6.6 ± 3) × 104	(1.08 ± 9) × 105	(7.6 ± 2) × 104	(7.0 ± 1) × 104	(1.58 ± 9) × 103	(5.9 ± 8) × 104
kd (s−1)	(8.6 ± 6) × 10−5	<1.0 × 10−7b	(5.2 ± 1) × 10−3	(5.4 ± 1) × 10−3	(8 ± 3) × 10−6	(2.8 ± 2) × 10−5
kD (nM)	1.3 ± 2	<1.0 × 10−3b	68 ± 3	77 ± 3	5 ± 1	(4.8 ± 1) × 10−1
NIH45-46
ka (M−1 s−1)	(3.05 ± 3) × 104	(5.56 ± 8) × 104
kd (s−1)	(10.0 ± 5) × 10−5	(1.37 ± 5) × 10−5
kD (nM)	3.3 ± 2	(2.5 ± 1) × 10−1
VRC01
ka (M−1 s−1)	(6 ± 1) × 104	(9 ± 1) × 104	(9.6 ± 3) × 104	(1.10 ± 3) × 105	(1.97 ± 7) × 104	(5.7 ± 3) × 104
kd (s−1)	(1.9 ± 3) × 10−4	<1.0 × 10−7	(3.6 ± 3) × 10−4	(3.16 ± 2) × 10−4	(4 ± 1) × 10−5	(4.2 ± 4) × 10−5
KD (nM)	3 ± 1	<1.0 × 10−3	3.8 ± 2	2.86 ± 1	2.0 ± 6	(7 ± 1) × 10−1
VRC03
ka (M−1 s−1)	(3.93 ± 4) × 104	(1.4 ± 2) × 105	(1.49 ± 8) × 104	(3.1 ± 9) × 104
kd (s−1)	<1.0 × 10−7	<1.0 × 10−7	(6.3 ± 2) × 10−3	(7.2 ± 7) × 10−3
KD (nM)	<1.0 × 10−3	<1.0 × 10−3	423 ± 35	238 ± 53
F105
ka (M−1 s−1)	(2.8 ± 6) × 103	(7 ± 1) × 103	(3.26 ± 3) × 104	(2.72 ± 4) × 104	(2.7 ± 5) × 102	(3.6 ± 1) × 103
kd (s−1)	(2.3 ± 4) × 10−4	(2.46 ± 7) × 10−4	(1.04 ± 5) × 10−3	(1.01 ± 2) × 10−3	(4.14 ± 6) × 10−5	(4.11 ± 2) × 10−4
KD (nM)	84 ± 4	36 ± 6	31.0 ± 6	37 ± 1	156 ± 6	114 ± 5
b12
ka (M−1 s−1)	NA	(5.14 ± 5) × 104	(7.0 ± 1) × 104	(1.18 ± 3) × 105	(3.2 ± 8) × 104	(5 ± 1) × 104
kd (s−1)	NA	(1.88 ± 5) × 10−4	(1.82 ± 4) × 10−3	(3.5 ± 1) × 10−3	(2.4 ± 6) × 10−3	(1.0 ± 4) × 10−3
KD (nM)	NA	3.6 ± 1	25.8 ± 9)	30 ± 2	73.3 ± 9	23 ± 2
V3 specific antibody V3-3074
ka (M−1 s−1)	(1.27 ± 2) × 104	(3.28 ± 3) × 104	(8.0 ± 5) × 104	(6.7 ± 5) × 104	(1.4 ± 2) × 105	(1.33 ± 4) × 105
kd (s−1)	(1.39 ± 8) × 10−4	(4 ± 1) × 10−6	<1.0 × 10−7	<1.0 × 10−7	(7 ± 1) × 10−5	<1.0 × 10−7
KD (nM)	11.0 ± 8	(1.2 ± 5) × 10−1	<1.0 × 10−3	<1.0 × 10−3	(4.88 ± 4) × 10−1	<1.0 × 10−3
V1/V2-glycan antibodies
PG9
ka (M−1 s−1)	(3.9 ± 2) × 104	(2.88 ± 9) × 104	(2.17 ± 1) × 104	(2.91 ± 3) × 104	(1.54 ± 4) × 105	(1.7 ± 2) × 105
kd (s−1)	(2.09 ± 6) × 10−4	(1.27 ± 2) × 10−4	(3.21 ± 7) × 10−4	(3.3 ± 1) × 10−4	(3.03 ± 2) × 10−4	(5.0 ± 6) × 10−4
KD (nM)	5.4 ± 1	4.42 ± 7	14.8 ± 4	11.4 ± 5	2.30 ± 1	2.97 ± 4
PG16
ka (M−1 s−1)	(9.8 ± 1) × 104	(7.42 ± 6) × 104
kd (s−1)	(1.25 ± 5) × 10−3	(4.8 ± 2) × 10−4
KD (nM)	12.7 ± 7	6.4 ± 4
PGT145
ka (M−1 s−1)	(1.6 ± 2) × 105	(1.4 ± 4) × 105			(2.8 ± 2) × 105	(2.6 ± 4) × 105
kd (s−1)	(4.9 ± 1) × 10−4	(3.22 ± 4) × 10−4			(4.0 ± 3) × 10−4	(3.8 ± 7) × 10−4
KD (nM)	3.2 ± 5	2.3 ± 6			1.4 ± 2	1.45 ± 3
V3-glycan antibodies
PGT121
ka (M−1 s−1)	(1.214 ± 4) × 105	(1.571 ± 8) × 105
kd (s−1)	<1.0 × 10−7	<1.0 × 10−7
KD (nM)	<1.0 × 10−3	<1.0 × 10−3
PG9T126
ka (M−1 s−1)	(1.10 ± 9) × 106	(1.1 ± 2) × 106
kd (s−1)	<1.0 × 10−7	<1.0 × 10−7
KD (nM)	<1.0 × 10−3	<1.0 × 10−3
Mannose-dependent neutralizing antibody 2G12
ka (M−1 s−1)	(1.040 ± 2) × 106	(1.006 ± 6) × 106
kd (s−1)	<1.0 × 10−7	<1.0 × 10−7
KD (nM)	<1.0 × 10−3	<1.0 × 10−3

^aSensograms are shown in Fig. S2 and S3 in the supplemental material. k_a, rate of association; k_d, rate of dissociation; K_D, equilibrium dissociation constant. Values are means ± standard deviations. NA, not applicable (due to weak binding, kinetics could not be fit).

^bDue to slow dissociation, dissociation rates or binding constants could not be fit reliably.

b12 and F105 also target the CD4bs but lack breadth and bind poorly to the BG505 SOSIP trimer (34). Accordingly, we detected weak binding of the BG505 SOSIP WT trimer to F105 and virtually no binding to b12 (Fig. 4). The removal of N197 significantly increased the rate of association with F105 and yielded observable binding to b12, with a K_D of 3.6 nM (Table 1). With gp120 monomers, the removal of N197 resulted in a minor increase in association (1.6-fold) but also a 2-fold higher rate of dissociation from b12, resulting in similar overall K_D values between the WT and the N197Q mutant. No difference in the association with or affinity for F105 was observed in the context of gp120 monomers.

To assess whether the removal of N197 has similar antigenic phenotypes in other Env isolates, we generated and purified WT and N197Q SOSIP trimers from isolate PVO.4, a clade B tier 3 virus (72). PVO.4 SOSIP produced native-like trimers that could be purified to homogeneity (Fig. 1B and C) but with lower yields (Verkerke et al., unpublished), which limited the number of antibodies examined in this study. Similarly to BG505 SOSIP trimers, the removal of the N197 glycan in PVO.4 trimers resulted in stronger binding to CD4-IgG2 by an order of magnitude. The N197Q mutant also showed a 2.9-fold increase in the association rate for CD4bs antibody VRC01 (Table 1; see also Fig. S3 in the supplemental material). Therefore, the removal of the N197 glycan within a trimer appears to generally enhance the binding of ligands targeting the CD4 binding site, predominantly by increasing the association rates.

Antigenic changes beyond the CD4 binding site after removal of the N197 glycan.

The crown of the V3 loop is buried under the N197 glycan from an adjacent protomer (37), and V3-specific antibodies are generally weakly neutralizing, although some can still bind SOSIP.664 trimers (34). By BLI, we observed significant binding of antibody V3-3074, which targets the tip of the V3 loop (73), to BG505 SOSIP WT trimers, with a K_D of 11 nM. The removal of N197 increased the overall level of binding, with an association rate more than twice that of the WT trimer and a much lower dissociation rate, resulting in a K_D of 0.1 nM (Table 1). A similar increase in affinity was observed with PVO.4 SOSIP trimers but without the increased association rate. V3-3074 showed stronger and faster binding to gp120 monomers, but there were no obvious differences in binding or association rates between WT and N197Q gp120 monomers.

Antibodies PG9, PG16, and PGT145 target quaternary epitopes in V1/V2 involving N-linked glycans (68, 74, 75). Both WT and N197Q BG505 trimers displayed strong binding to PG9/PG16/PGT145 with no obvious differences in association rates (Fig. 4 and Table 1). The only notable change was a 2-fold-lower dissociation rate for PG16 in the N197Q mutant. The PVO.4 N197Q trimer displayed the same binding to PG9 as that of the WT. The removal of N197 had no effect on BG505 gp120 monomers. Overall, the removal of the N197 glycan has little effect on the binding of V1/V2-directed antibodies.

CD4-induced (CD4i) antibodies, such as 17b and A32, are largely nonneutralizing, and both antibodies are unable to bind to BG505 SOSIP trimers by an enzyme-linked immunosorbent assay (ELISA) in the absence of CD4 (34, 48). By BLI, the binding of the BG505 WT SOSIP trimer to 17b or A32 was too weak to measure, even at a higher trimer concentration of 500 μM (Fig. 4). No increase in binding was observed with the N197Q mutant, suggesting that no CD4-induced conformational changes result from the removal of the N197 glycan. Finally, we compared the binding of glycan-dependent bNAbs 2G12, which targets a conserved glycan patch on gp120 (70), and PGT121 and PGT126, which recognize the base of V3 (54). All of these antibodies bound to both the WT and N97Q trimers with identical affinities and kinetics (Table 1).

Steric occlusion of the CD4 binding site by the N197 glycan.

As the removal of N197 has little effect the overall structure of the envelope trimer, we investigated whether the effect of the N197 glycan on antigenicity is due to steric occlusion of neutralizing epitopes. To define the spatial position of the N197 glycan relative to the epitopes of neutralizing antibodies, an ensemble of fully glycosylated BG505 trimer models was generated (see Materials and Methods). It is readily apparent that the N197 glycan position partially occludes the CD4bs in the Env trimer (Fig. 5A and C).

FIG 5.

Position of the N197 glycan in the Env trimer. (A and B) The spatial occupancy from the ensemble of glycosylated Env models shows the positions of the N197, N276, N301, N363, N386, and N463 glycans, which were found to limit access to the CD4bs (yellow surface) on trimeric Env (A) and monomeric gp120 (B). (C) Position of the N197 glycan relevant to various epitopes on the trimer. The binding surface is indicated in dashed lines for CD4, b12, and VRC01. The positions of other antibodies relevant to this study are shown, including PG9 (blue), PGT122 (which binds a similar epitope as PGT121) (orange), and 2G12 (yellow).

For a more quantitative assessment of how various glycans peripheral to the CD4bs might occlude various antibodies, we analyzed the occurrence of steric clashes between glycans and the position of CD4 or various bNAbs in the 40 fully glycosylated models. Glycans at N197, N276, N301, N363, N386, and N463 were identified to potentially clash with the position of 2-domain soluble CD4 (sCD4) when it is bound (Fig. 5A). Of these glycans, the N197 glycan was most likely to clash with CD4, as seen in 70% of the models (Tables 2 and 3). The nonneutralizing CD4bs antibodies F105 and b12 clashed with the N197 glycan in 100% of the models as well as with the N301 glycan from the neighboring protomer. With antibodies VRC01 and VRC03, the major clashes were seen with N276 and N463, occurring in >97% of the models, but clashes with N197 still occurred in 73% of the models. In contrast, the N197 glycan clashed with the position of PG9 in only 30% of the models (Fig. 5C).

TABLE 2.

Steric clashes between various ligands and glycans on the Env trimer as percentage of models in the ensemble^a

Glycan	% of models with steric clash on Env trimer
	sCD4	VCR01	VRC03	F105	b12	PG9
N197	70	73	78	100	100	30
N276	43	98	100	22	17
N301b	11	43	42	100	100
N363	56	64	57	32	41
N386	12	11	7	23	46
N463	24	98	97	7	7

^aSee also Fig. 5.

^bClashes from the N301 glycan were predominantly from the neighboring protomer.

TABLE 3.

Steric clashes between various ligands and glycans on gp120 monomers as a percentage of models in the ensemble^a

Glycan	% of models with steric clash on gp120 monomer
	sCD4	VCR01	VRC03	F105	b12
N197	1	2	2	13	11
N276	53	99	98	20	19
N301	0	0	0	9	3
N363	68	74	59	58	60
N386	3	3	1	12	49
N463	33	98	98	19	13

^aSee also Fig. 5.

To compare how glycan clashes differ in the context of a gp120 monomer, we generated an ensemble of BG505 gp120 models. The N/C-terminal regions, bridging sheet, and V1/V2 and V3 loops were largely unrestrained, as they have been shown to be highly disordered in solution (43, 57). For glycans at N276, N363, N386, and N463, the clashes are similar to what was observed with the trimers (Fig. 5B). N301, which in the trimers resulted in clashes with CD4 from an adjacent protomer, showed no clashes with CD4 in monomeric gp120. Interestingly, unlike in the trimer, N197 glycan clashes with sCD4 and VRC01 in gp120 are very rare (1 to 2% of models). These observations indicate that the N197 glycan occludes the CD4 binding site only within the context of a native Env trimer.

DISCUSSION

Effects of N197 glycan removal are observed only in the context of an Env trimer.

The N197 glycan was found to modulate tier 2 cross-reactivity (76), and removal of N197 has been shown to increase neutralization sensitivity to CD4bs and V3-specific antibodies (12, 22–24). Promising results have shown that N197-deleted Env elicited a stronger neutralizing antibody response in vivo (23), indicating improved immunogenicity. However, when tested with monomeric gp120, the N197 deletion had negligible effects on CD4 binding (24). In light of our binding data, it is clear that the N197 glycan has significant effects on the interactions with CD4 and CD4bs antibodies only in the context of a native-like Env trimer (Fig. 6). The removal of N197 led to significantly increased binding of CD4 and CD4bs antibodies. At the same time, the binding of bNAbs targeting other epitopes (2G12, PGT121, and PGT126) was unaffected, indicating that the N197Q trimers were otherwise antigenically similar to the WT. Computational modeling of the glycosylated Env trimer suggests that the spatial occupancy of the N197 glycan impedes access to the CD4bs. This steric occlusion explains why the removal of this particular glycan results in a large increase in the association rates with CD4bs-targeting ligands. Beyond the enhanced association rate, we also observed a slower dissociation of CD4 and CD4bs antibodies in the N197Q trimer. The entropic penalty of positioning glycans away from the CD4bs and restricting their spatial sampling may result in a lower net free energy of binding, which in turn would result in a higher rate of dissociation of ligands from the N197 glycan-deleted trimer. Therefore, the presence of the N197 glycan may not only limit epitope access (resulting in slow antibody association) but also destabilize certain Env-antibody complexes (resulting in faster antibody dissociation).

FIG 6.

Effects of N197 glycan removal in the Env trimer or gp120 monomer. In the trimer (top), the N197 glycan partially occludes the CD4 binding site. Removal of the N197 glycan increases the exposure of the CD4 binding site while also making the V1/V2 loop more flexible and increasing the accessibility of V3. In gp120 monomers (bottom), without tertiary and quaternary structural constrains, the V1/V2/V3 loops are relatively disordered. Since the N197 glycan does not occlude the CD4 binding site, its removal has no significant influence on the binding of CD4bs or V3-specific antibodies.

With gp120 monomers, there were no significant differences in binding to CD4bs antibodies or V3-specific antibodies between the WT and the N197Q mutant. Without the quaternary constraints of a trimer, gp120 monomers are relatively dynamic molecules with a high degree of flexibility at V1/V2 and V3 (43, 57, 77). In our models of glycosylated monomeric gp120, the N197 glycan was predominantly oriented away from the CD4 binding site, and clashes with most CD4bs antibodies were seldom observed. However, with F105 and b12, there were still some minor clashes with the N197 glycan, which can explain the slightly higher on-rate of binding of b12 to the gp120 N197Q mutant (Table 1). As the V3 loop is readily exposed in gp120 monomers, it is not surprising that the removal of the N197 glycan does not affect the binding of V3-specific antibodies. The stark differences in the phenotypes between trimeric Env and monomeric gp120 highlight the importance of evaluating Env modifications in the context of a native-like trimer to assess full effects on antigenicity.

The only structural effect on the trimer observed from the N197 deletion was a subtle increase in the dynamics of the base of the V1/V2 loop and the bridging sheet near the trimeric interface by HDX-MS (Fig. 6). The increased flexibility at V1/V2 could explain the slightly larger radius of gyration of N197Q trimers as observed by SAXS (Fig. 2B). This can also explain the slight decrease in the association rate for antibodies PG9 and PG16 targeting V1/V2 in the N197Q mutant (Table 1), as the epitope would be less conformationally restrained. Since the V3 loop is buried under V1/V2 in the trimer, increased flexibility at the crown of the trimer can also result in increased V3 exposure, providing a rationale for why some V3-specific antibodies are able to bind faster to the N197 glycan-deleted trimer. The subtle effects on the conformational dynamics of the trimer by N197 glycan removal would have largely been invisible by many other structural techniques (i.e., X-ray crystallography), illustrating the utility of HDX-MS for probing all of the structural effects that arise from Env modifications. We note that the full scale of structural effects imparted by the N197Q mutant may be underrepresented in this study due to the stabilizing mutations in the SOSIP trimer, and there may be other consequences in the context of full-length virus-associated Env that we cannot detect here.

Two recent reports touched upon the role of the N197 glycan in trimeric Env. A detailed glycan analysis of BG505 SOSIP trimers reported that N197 removal reduced bNAb recognition of V1/V2 apex-targeting antibodies, including PGT145 and PG9, as reflected by changes in the mutant's neutralization sensitivity (78). It was suggested that changes at N197 influence the nearby glycans at the trimer apex. The recent crystal structure of the glycosylated Env SOSIP revealed ordered density across the surface of the trimer crystal structure (29). For N197 in a clade G SOSIP trimer (X1193.c1; G459C mutant), the only ordered density that was observed was the core Man₁GlcNac₂, suggesting that it is relatively free of interactions with adjacent Env structural elements. However, it is possible that even without stable structural constrains, the absence of N197 increases the exposure of nearby glycans to glycosidases and glycosyltransferases, resulting in different glycosylation profiles. It should be noted that in the crystal structure, the trimer was complexed with 2 Fabs and the Fv portion of VRC01, and such interactions could perturb the native configuration of glycan chains, including N197.

From the observable glycopeptides used for HDX-MS comparison, we observed no apparent differences in the glycosylation patterns between the WT and the N197Q mutant. However, we could not monitor the glycosylation patterns within V1/V2 or the C-terminal base of V3, which is where the major antigenic effects of glycan alterations are expected to occur. Thus, while our study addresses how a targeted glycan deletion affects the steric occlusion of epitopes and the conformational dynamics of the trimer, the influence on the glycosylation patterns of nearby glycans remains to be directly addressed.

Isolate-specific effects of the N197 glycan.

Like with BG505, the N197Q PVO.4 trimer displayed stronger binding to CD4bs antibodies than the WT, with a higher association rate for CD4, VRC01, and F105. However, the association with b12 was not significantly faster, and the stronger binding affinity was due mostly to a lower dissociation rate. This is a distinct effect from what was observed for the BG505 trimer, which displayed detectable binding to b12 only after the removal of the N197 glycan. PVO.4 has a much longer V1/V2 loop than BG505, with 12 extra amino acids and additional glycosylation sites at the base of V1 (Fig. 1D). b12 makes contacts with a portion of V1/V2, and binding leads to a partial opening of the V1/V2 loop (79, 80). The longer V1/V2 loop might provide more steric hindrance for b12 accessing the Env trimer, even without the N197 glycan, explaining why the association of the N197Q trimer with b12 was not much faster than that of the wild type. The extra amino acids in the V1/V2 loop may also introduce additional contacts with b12 that are affected by the presence of the N197 glycan, explaining the changes in the dissociation rates. As with b12, the rates of association of the PVO.4 N197Q mutant and the WT with V3-3074 were similar, and both exhibited higher association rates than the BG505 trimers. It appears that the V3-3074 epitope is intrinsically more exposed in the PVO.4 SOSIP, and as a result, the removal of N197 has a negligible effect on the accessibility of V3.

Based upon the available data, we infer that the functional effects of the N197 glycan vary between different isolates in large part due to the large differences in V1/V2 sequences. This may provide a rationale for the isolate-dependent effects of N197 removal that were observed in previous studies. For example, the removal of the N197 sequon in some isolates has been reported to produce an enhancement in CD4-independent entry (81). In other cases, the magnitude of changes in antigenicity resulting from N197 modification differed among isolates (24). This was attributed to differences in V1/V2, but no obvious specific characteristic (i.e., length or number of glycans) correlated with the level of neutralization changes with N197 removal. Several studies have shown that isolates vary in the level of “openness” of Env (80, 82). Tier 1 or laboratory-adapted isolates are very susceptible to neutralization and are described as having more “open” conformations, while difficult-to-neutralize isolates have more “closed” Env conformations. We hypothesize that the level of openness will likely govern how much of an effect N197 removal will have, and much of this is determined by specific characteristics of V1/V2 and V3 that have yet to be understood in detail.

In general, the BLI measurements of antibody binding correlated well with the observed changes in neutralization; namely, CD4 binding site and V3 binding and neutralization sensitivities increase when the N197 glycan is removed. For example, isolate PVO.4 lacking N197 was ∼23-fold more susceptible to neutralization by VRC01 and 3 orders of magnitude more susceptible to V3-targeting 447-52D than the WT (24). From our binding assay, the N197Q SOSIP trimer from isolate PVO.4 showed 3-fold-stronger binding to VRC01 and 100-fold-stronger binding to 447-52D than the WT. Pritchard et al. reported a 10-fold increase in neutralization sensitivity to PGV04 (an antibody with a binding footprint very similar to that of VRC01) resulting from N197 glycan deletion in BG505 (36). In accord, we observe that the N197Q mutant trimer has an ∼1,000-fold increase in affinity for VRC01 compared to the WT. The relatively few differences between data from neutralization assays and binding may be due to the SOSIP-based stabilization of the soluble trimer constructs that we have available for analysis as well as the presence of the membrane-proximal external region (MPER), the transmembrane anchor, and the cytoplasmic tail on full-length Env on pseudotyped viruses used in neutralization assays.

Significance of the N197 glycan in vaccine design.

The structural and antigenic features of Env are a major focus of HIV-1 vaccine design. Conserved glycosylation sites are of particular interest since many bNAbs recognize conserved carbohydrates in combination with other protein elements (8, 15, 16, 54, 68, 74, 83–85). On the other hand, glycans also mask conserved polypeptide epitopes, and deletion of glycans to better expose neutralizing epitopes is another strategy in vaccine design (86). The BG505 SOSIP trimer is a promising immunogen candidate with stable native-like trimeric characters, but so far, it has been capable of eliciting only autologous tier 2 neutralizing antibodies, with tier 1 cross-reactivity biased to the V3 epitope (87). Our results give direct evidence for a specific modification for enhancing the binding of CD4bs antibodies while maintaining other relevant bNAb epitopes (e.g., PG9 and PGT122), without disrupting the overall quaternary structure of the trimer. Trimeric Env constructs modified by the deletion of one single glycan, like N197, are likely to provide improvements to the current trimer Env-based vaccine approaches, particularly for the elicitation of CD4bs antibodies.