Plasticity and Epitope Exposure of the HIV-1 Envelope Trimer

ABSTRACT

We recently showed that mutations in the HIV-1 envelope (Env) destabilize the V3 loop, rendering neutralization-resistant viruses sensitive to V3-directed monoclonal antibodies (MAbs). Here, we investigated the propagation of this effect on other Env epitopes, with special emphasis on V2 loop exposure. Wild-type JR-FL and 19 mutant JR-FL pseudoviruses were tested for neutralization sensitivity to 21 MAbs specific for epitopes in V2, the CD4 binding site (CD4bs), and the CD4-induced (CD4i) region. Certain glycan mutants, mutations in the gp120 hydrophobic core, and mutations in residues involved in intraprotomer interactions exposed epitopes in the V2i region (which overlies the α4β7 integrin binding site) and the V3 crown, suggesting general destabilization of the distal region of the trimer apex. In contrast, other glycan mutants, mutations affecting interprotomer interactions, and mutations affecting the CD4bs exposed V3 but not V2i epitopes. These data indicate for the first time that V3 can move independently of V2, with V3 pivoting out from its “tucked” position in the trimer while apparently leaving the V2 apex intact. Notably, none of the mutations exposed V2 epitopes without also exposing V3, suggesting that movement of V2 releases V3. Most mutations increased sensitivity to CD4bs-directed MAbs without exposure of the CD4i epitope, implying these mutations facilitate the trimers' maintenance of an intermediate energy state between open and closed conformations. Taken together, these data indicate that several transient Env epitopes can be rendered more accessible to antibodies (Abs) via specific mutations, and this may facilitate the design of V1V2-targeting immunogens.

IMPORTANCE Many epitopes of the HIV envelope (Env) spike are relatively inaccessible to antibodies (Abs) compared to their exposure in the open Env conformation induced by receptor binding. However, the reduced infection rate that resulted from the vaccine used in the RV144 HIV-1 vaccine trial was correlated with the elicitation of V2- and V3-directed antibodies. Previously, we identified various mechanisms responsible for destabilizing the V3 loop; here, we determined, via mutation of numerous Env residues, which of these elements maintain the V1V2 loop in an inaccessible state and which expose V1V2 and/or V3 epitopes. Notably, our data indicate that V3 can move independently of V2, but none of the mutations studied expose V2 epitopes without also exposing V3. Additionally, V1V2 can be rendered more accessible to Abs via specific mutations, facilitating the development of engineered V2 immunogens.

INTRODUCTION

Numerous studies of the trimeric human immunodeficiency virus type 1 (HIV-1) envelope (Env) surface protein have demonstrated that the molecule exhibits significant flexibility and can assume several conformations regardless of whether it is bound by CD4 (1–4). When in the ground state, not bound by CD4, V3 is preferentially masked, and many epitopes of the Env spike are generally inaccessible to antibodies (Abs) relative to their accessibility in the open conformation induced by receptor binding (4–7). The dynamic movement of the Env trimer, with transient exposure of many parts of the molecule, along with gp120 shedding, helps to explain the robust immunogenicity of V3 and the presence of V3-specific Abs in the sera of essentially all HIV-infected individuals despite its occlusion in the ground state of the Env (8–10).

In a previous study, we investigated Env spike plasticity and the various mechanisms that keep V3 occluded when the Env trimer is in the closed ground state (11). We examined the effects of individual mutations in several regions of the gp120 monomer that forced the trimer into a more open state, rendering it highly interactive with V3-directed monoclonal Abs (MAbs). We determined that several interactions were responsible for the packing of the V3 loop in the Env spike and that many of these interactions, if disrupted, could “pop out” V3. These interactions included quaternary interprotomer interactions; hydrophobic packing of the core of gp120 in the bridging sheet; intraprotomer interactions of V3 with adjacent regions of gp120; and limiting flexibility of the V3 region, which can be mediated by glycans (11). Indeed, most primary HIV isolates are typically insensitive to V3 Abs, but many isolates that are sensitive to V3 Abs possess these types of mutations that release V3 (11). Whereas our previous study revealed the alterations that exposed V3 and rendered viruses susceptible to neutralization by V3 MAbs, here, we aimed to further understand the plasticity of Env and whether mutations that release V3 would propagate changes in Env structure and stability affecting exposure of V2 and other epitopes in the Env trimer.

Exposure of epitopes in the second variable loop (V2) of gp120 was of particular interest, since the RV144 clinical HIV vaccine trial data showed that vaccine-induced anti-V2 serum Abs were correlated with a reduced risk of infection (12–14). These data suggested that the V1V2 loop might be an important epitope for vaccine targeting and renewed interest in V2 MAbs. Given the generation and characterization of numerous V2-specific MAbs from infected patients, we sought to employ a large panel of these MAbs with specificities for various epitopes within V2 to determine whether the same structural features that affected V3 MAb potency affected the activities of V2 MAbs.

As V3 is highly flexible relative to other Env domains and V3 is among the most immunogenic regions of the Env glycoproteins (8–10, 15–17), we hypothesized that release of the V3 loop in various mutants from its occluded position in the Env spike occurs independently of substantial conformational alterations in the V1V2 loop. We also sought to determine if the same mutations that released V3 would affect the exposure of various epitopes presented by the V2 region of gp120, using the JR-FL pseudovirus, which exhibits a “closed” trimer conformation and is therefore relatively resistant to neutralization by V2i and V3 MAbs (18–23). The effects of these mutations were subsequently tested, as well, to determine how they affected the potency of MAbs specific for the CD4 binding site (CD4bs) and the CD4-induced (CD4i) epitopes. To address these questions, we measured the relative changes in neutralization sensitivity of wild-type (WT) and mutant JR-FL pseudoviruses to a large panel of human MAbs. The information provided by this study speaks to the overall plasticity of Env. Additionally, the data reveal how V2 Abs can impose immune pressure on the virus and how this information may be used to better elicit V2-specific Abs as part of a vaccine regimen.

RESULTS

Variable effects of JR-FL glycan mutations on neutralization by V2i MAbs.

The wild-type JR-FL pseudovirus is highly resistant to neutralization in the TZM-bl assay by virtually all V2i, V2p, and V2q MAbs tested, which is why it was selected for analysis (Table 1). Since glycosylation of Env shields and stabilizes Env from Abs of many specificities (5, 7, 24, 25), we first tested the effect on the neutralization sensitivity of JR-FL and its variants from which selected gp120 glycans had been deleted (Fig. 1 to 3). We had previously determined that the removal of the glycan in the N160K mutant destabilized V3 by significantly enhancing the flexibility of that region, causing the mutant to become highly sensitive to many V3 MAbs (Table 1) (11). We therefore tested the N160K mutant against the V2 MAb panel but found that the pseudovirus remained highly resistant to neutralization by essentially all V2 MAbs (Table 1). PGT145 neutralization was lost, as would be predicted by this MAb's binding requirements (Fig. 4) (26).

TABLE 1.

Sensitivities of mutant pseudoviruses to V2- and V3-specific MAbs^a

^aThe titles at the top of the table indicate the types of interactions in the Env trimer affected by the mutations listed directly below. JR-FL WT and mutant pseudoviruses (PsV) are shown in the next row across the table. Mutants shaded in dark gray indicate neutralizing sensitivity to V2 and V3 MAbs compared to the WT, while mutants shaded in light gray show neutralization sensitivity to V3 MAbs but not V2 MAbs. IC₅₀s are shown and were calculated by titration of MAbs in the TZM-bl assay, as detailed in Materials and Methods. “>50” (blue shading) indicates that 50% neutralization was not achieved even at the highest MAb concentration tested, 50 μg/ml. Increasing neutralization sensitivity is indicated by color coding, as shown in the color legend. The black boxes indicate that experiments were not done. The boxes with partial white shading indicate decreased neutralization sensitivity. The solid horizontal lines separate classes of MAbs, whereas the dashed horizontal lines indicate subdivisions with the V2i family of MAbs.

FIG 1.

Mutated residues of the JR-FL-JB pseudoviruses used in neutralization assays. (A) The various regions of gp120 are shown above the residues and positions in the wild-type virus that were mutated for this study. (B) Diagram of the positions of the mutations made in gp120. The V1V2 domain is shown in green, the gp120 core in magenta, and the V3 loop in gray ribbon representation. The image is based on PDB entry 4TVP, the crystal structure of the BG505 SOSIP trimer complexed with MAbs PGT122 and 35O22 (27). The numbering is based on the HxB2 reference sequence.

FIG 2.

Closeup of the interface between the V1V2 domain (green) and the gp120 core (magenta). Extensive intraprotomer interactions between the region in V1V2 that includes the putative integrin binding motif (¹⁷⁹LDI¹⁸¹) and the R419-to-I423 segment in C4 of gp120 are shown. The image is based on PDB entry 4TVP.

FIG 3.

Pseudovirus infectivity. (A) Pseudoviruses were titrated on TZM-bl cells as described previously (61). Replicate titration curves were used to calculate the area under the concentration/time curve (AUC) values using GraphPad Prism. The values were divided by 10⁷. The asterisks indicate increasing degrees of significant (*, 0.05 > P > 0.01; **, 0.01 > P > 0.001; ***, P < 0.001) changes from the WT as determined by 2-tailed Student's t test. (B) Using the titration curves, the appropriate pseudovirus dilutions needed to achieve ∼150,000 RLU were interpolated.

FIG 4.

Neutralization curves of MAb PGT145 against mutant pseudoviruses. Neutralization was performed using the TZM-bl assay, and the percent neutralization was calculated as detailed in Materials and Methods. The curves are colored according to the type of mutation: glycan deletion, lavender; gp120 core disruption, cyan; interprotomer disruption, pink; intraprotomer disruption, tan; CD4bs abrogation, green. The wild type is colored black.

Removal of the glycan at position 332 in the N332K mutant was previously shown to have a relatively small effect on JR-FL V3 MAb sensitivity, as it is part of the secondary and tertiary structures of the outer domain and does not significantly disturb the space that is potentially accessible for local V3 flexing (11). Similarly, the N332K mutant was found to have essentially no effect on sensitivity to any of the V2 MAbs, with the exceptions of V2i MAb 21a9, which became only moderately effective against N332K, exhibiting a 50% inhibitory concentration (IC₅₀) neutralization value of 18 μg/ml (Table 1).

In contrast to the minimal effects elicited by the elimination of glycans at 160 and 332, we found that the elimination of the N301 glycan had significant effects on sensitivity to V2i MAbs. We had previously demonstrated through modeling that the N301 glycan plays a critical role in limiting the flexibility of the crown and stem of V3 (roughly between residues 300 and 328), thus affecting the packing of V3, and that removal of this glycan in the N301Y and N301D mutants resulted in profound increases in sensitivity to V3 MAbs (Table 1) (11). When the V2i MAb panel was tested against these mutants, it was found that the relatively flexible trimer apex of the pseudoviruses allowed access to V2i MAbs compared to JR-FL WT. Notably, this increased sensitivity was not as pronounced as was previously shown for V3 MAbs, as the IC₅₀s for V2i MAbs generally ranged from 2 to 28 μg/ml, values that were on average more than an order of magnitude higher than those for V3 MAbs (<0.05 to 11 μg/ml) (Table 1) (11). However, one group of V2i MAbs (21a9, 18a4, 20a12, and 20a24) was distinct in barely achieving 50% neutralization of N301Y (IC₅₀ = 47 to 49 μg/ml), and MAb 2297 remained unable to neutralize either of the N301 mutants (Table 1). Sample titration curves of the neutralizing activities of three V2i MAbs active against the N301Y mutant JR-FL are shown in Fig. 5A. The V2p and V2q MAb potencies appeared unaffected by the N301 glycan mutants of JR-FL, with the exception of PGT145, whose potency against N301Y was ablated (Fig. 4 and Table 1).

FIG 5.

Neutralization curves of V2 MAbs against V2-destabilizing mutant pseudoviruses. Neutralization was performed using the TZM-bl assay, and the percent neutralization was calculated as detailed in Materials and Methods.

Profound effects of mutations in the gp120 hydrophobic core on neutralization sensitivity to V2i-specific MAbs.

Residues A204 and M434 in the V2 and C4 regions of JR-FL, respectively, form a part of the hydrophobic core underlying the “prebridging sheet” in each Env monomer, and we previously predicted from modeling that alteration of this core would elicit generalized Env-destabilizing effects (11). Indeed, V3 MAb sensitivity profoundly increased for pseudoviruses with either of these mutations, consistent with previous studies of a virus isolated from a patient with a core mutant at residue 424 that exhibited increased V3 MAb sensitivity and with mutants of simian-human immunodeficiency virus (SHIV) (27, 28). Similarly, it was predicted that the structural instability induced by mutations in the hydrophobic core would propagate structural changes throughout gp120, and indeed, it was found that both the A204E and M434G pseudovirus mutants became highly sensitive to V2i MAb neutralization, with IC₅₀s ranging from 0.1 to 2.0 μg/ml (A204E) and 0.2 to 15.0 μg/ml (M434G) for most V2i MAbs tested, although both mutants remained relatively resistant to neutralization by the two V2i MAbs (830A and 2297) with distinct characteristics, as described in Materials and Methods (Table 1 and Fig. 5B and C). While the neutralizing activity of V2p MAbs was unaffected by mutations affecting core stability, the activity of V2q MAbs was impacted. The M434G mutant was rendered sensitive to PG9 and neutralized with an IC₅₀ of 2.9 μg/ml; in contrast, neutralization by PGT145 was ablated (Table 1 and Fig. 4 and 5B).

Destabilization of intraprotomer interactions induces sensitivity to V2i-specific MAbs.

Within each protomer, the hydrophobic face of V3 also reacts with the gp120 core via the V1V2 stem. Previously, we found that mutation of the V2 stem residue L125 dramatically increased sensitivity to V3 MAbs (11). The L125F mutant was also found to be sensitized to V2i MAb neutralization, although the effect was less striking than with V3 MAbs: the IC₅₀s for 7 of 15 V2i MAbs ranged from 1 to 9.9 μg/ml, while for V3 MAbs, the IC₅₀s were consistently ≤0.8 μg/ml (Table 1 and Fig. 5D) (11).

The intraprotomer interface between V1V2 and the gp120 core shown in Fig. 2 was analyzed to identify side chains that contribute significantly to the interactions between the two domains. Because mutations on the V1V2 domain side could affect binding of V1V2 MAbs, we focused on the gp120 side of the interface, specifically residues I420, K421, and Q422 in the C4 region. Two mutations to alanine were chosen to eliminate stabilizing interactions: I420A eliminates lipophilic contacts, and Q422A eliminates several hydrogen bonds. Two other mutations were designed to create steric clashes (repulsion): K421N and Q422P. Specific substitutions were selected using multiple alignments of 4,173 gp120 sequences from the LANL database (2014 data set; https://www.hiv.lanl.gov/content/sequence/HIV/COMPENDIUM/2014/hiv1prot.pdf) to identify rare but naturally observed substitutions. Thus, four mutations not previously tested for V3 release were identified and predicted to disrupt the intraprotomer interface between V1V2 and the gp120 core. Since the V1V2 domain normally forms one of the walls of the pocket in which the V3 loop sits, when V1V2 is released from its interaction with the gp120 core, V3 would also be predicted to be released. Residues 420, 421, and 422 are directly in contact with the V2 YKLDV kink/integrin binding motif of the same protomer (29, 30). I420A and Q422A mutations rendered these mutant pseudoviruses susceptible to neutralization by most V2i MAbs with IC₅₀s ranging from 1.3 to 12 μg/ml (Table 1 and Fig. 5E). Replacement of these C4 residues with more repulsive substitutions (K421N and Q422P) yielded greater release of the V1V2 loop: V2i MAbs exhibited neutralization of these pseudoviruses, with most IC₅₀s between 0.7 and 5.5 μg/ml (Table 1 and Fig. 5F). All V3 MAbs tested with these novel mutants exhibited increases in neutralization potency by 2 or 3 orders of magnitude (Table 1). In contrast to the enhanced sensitivity of intraprotomer mutants to V2i MAbs, these mutations resulted in a loss of neutralization ability by the V2q MAb PGT145, consistent with the disruption of the trimer association domain (Table 1 and Fig. 4). As with the WT and all the tested mutants of JR-FL, V2p MAbs remained unable to neutralize (Table 1).

Mutation of residues critical for interprotomer interactions produces little effect on neutralization sensitivity to V2i-specific MAbs.

To test for the effects of mutations that disturb interprotomer interactions within the Env spike, several mutants were made that had previously been shown to disrupt the stabilizing interactions between the V1V2 stem and the tip of V3 from the neighboring protomer (11, 31). JR-FL has an aspartate at position 197 (D197); however, numerous strains possess a glycosylated asparagine at this position, which has been shown to have a stabilizing effect on the trimer (11, 31). Restoration of the glycan at position 197 was previously shown not to affect V3 sensitivity, while the D197H and D197Q mutants were highly sensitive to V3 neutralization (Table 1) (11). This suggested that D197 itself forms critical contacts that stabilize V3. In contrast to these data, the present study found none of the D197 mutants tested to be sensitive to V2i neutralization (Table 1). Moreover, disruption of the I184 residue, which putatively contributes to the quaternary interaction of the V1V2 regions of neighboring protomers (Fig. 6 shows the van der Waals interactions between the only partially solvent-accessible surfaces of I184 and L165 from the neighboring protomer), had no effect on V2i neutralization (Table 1) despite being previously shown to render JR-FL exceptionally sensitive to V3 MAbs (11).

FIG 6.

I184 lipophilic contact between protomers. This illustration was generated from PDB entry 4TVP, the X-ray structure of SOSIP BG505 (29). The dot mesh shading represents the solvent-accessible surfaces (SAS) (the van der Waals surfaces expanded by the water molecule radius) of the two residues from adjacent protomers (73, 74). Where the SAS overlap, the two residues exclude solvent access to each other.

Most V2p and V2q MAbs were also unaffected by mutation of residues critical for interprotomer interactions, with the exception of PGT145, whose neutralizing IC₅₀s for D197N decreased by an order of magnitude, again suggesting the requirement for a closed trimer apex in order for the MAb to bind and function (Table 1 and Fig. 4).

Mutation of the CD4bs does not affect neutralization by V2 MAbs.

Next, we investigated the effect of a mutation in the CD4bs on V1V2 loop exposure. Previously, we had determined that mutation of the CD4bs moderately affected V3 stability, as the N280E JR-FL mutant became sensitive to most V3 MAbs at concentrations ranging from 2 to 28 μg/ml whereas WT JR-FL was resistant to most V3 MAbs (Table 1) (11). Here, we tested this CD4bs mutant for V2 MAb-mediated neutralization sensitivity and found the N280E mutant pseudovirus to be unchanged in its sensitivity to any of the V2 MAbs tested.

V1V2 domain motion simulations in the trimer context.

To gain insight into how V2i MAbs might acquire access to their epitopes, we performed stochastic Monte Carlo simulations of the V1V2 domain movements within the context of the trimer. The X-ray structures of complexes of V2i MAbs with the V1V2 domain showed that the internal tertiary β-barrel structure of the domain is largely preserved between the closed Env trimer state and the Ab-bound state (7). Therefore, we set up simulations in such a way as to specifically explore the movements that involve the flexible V1V2 stems, radial movements of whole protomers with respect to the trimer axis, and V3 “flickering” (with two discrete V3 states) (Fig. 7).

FIG 7.

Types of stochastic movements in the Monte Carlo simulations of the radial movements of the protomer cores, as well as the mobility of the V3 and V1V2 domains within the Env trimer.

Ensembles of conformations were generated for the WT and the K421N mutant, which affects the packing of the intraprotomer core. In order to detect events of V2i exposure, each conformation was analyzed for compatibility with binding of V2i MAbs 697 and 830A: the Fab fragments were placed in their bound positions with respect to the V1V2 domain, and the number of backbone atoms that would overlap other parts of the spike was determined. Lack of overlap would thus be indicative of unobstructed epitope access. The original closed conformation exhibited overlap of 55 to 60 atoms, i.e., epitopes were clearly highly sterically obstructed in the closed WT conformation. In order to visualize how the system sampled states with different degrees of epitope exposure, we plotted the points representing two overlaps for each conformation (Fig. 8). For WT simulations, low-energy conformations largely clustered around the original experimental crystallographic structure. Some higher-energy states had reduced overlaps (between 5 and 50) (Fig. 8), but few if any had unobstructed V2i epitope access (<5 overlaps). In contrast, simulations of the K421N mutant showed that the low-energy states now sampled a much broader range of conformations, all the way down to fully V2i-exposed states with <5 overlaps (Fig. 8). The ensembles also showed relatively limited correlation between accesses for V2i MAbs 830A and 697, i.e., most of the conformers that allowed access for one antibody were not compatible with the other, and vice versa. This is likely related to the significantly different angles of access of the two MAbs with respect to the V1V2 domain (32).

FIG 8.

Motion simulations of V2i MAb access to the V1V2 domain. Ensembles of conformations from the simulations of JR-FL WT and from the K421N mutant Env trimers were analyzed to determine binding accessibility for two V2i-specific MAbs: 697 and 830A. Each dot represents a conformation, colored according to its energy (E) relative to the lowest energy conformer in the ensemble. The x and y coordinates show the numbers of atoms that would overlap other parts of the spike if the MAb were bound to its epitope on the V1V2 domain. Access to an epitope is unobstructed if the overlap is close to zero.

Mutations affecting V2 and V3 exposure enhance access to CD4bs- but not CD4i-specific MAbs.

Mutations that affected V3 or V2 and V3 epitope exposure also potently enhanced the availability of epitopes recognized by CD4bs MAbs b12, 654, and 559. They included glycan mutations, particularly removal of the N301 glycan, as well as mutations affecting the gp120 core and interprotomer and intraprotomer interactions (Table 2). The potency of these CD4bs MAbs increased by 2 to 3 orders of magnitude, even inducing neutralization in the case of MAbs 559 and 654, which do not neutralize WT JR-FL. VRC01 similarly displayed increased potency against many of the mutants, though sensitivity to this MAb was reduced ∼10-fold in the cases of the A204E and M434G core mutants and the more disruptive K421N and Q422P intraprotomer mutants (Table 2). In contrast, most of the same mutations had little effect on the neutralization potencies of four CD4i MAbs.

TABLE 2.

Sensitivities of mutant pseudoviruses to CD4bs and CD4i MAbs^a

^aSee footnote a to Table 1 for an explanation of the colors and symbols.

Mutant JR-FL variants that were not expected to have structurally disrupted Env spikes were included as controls in all of the MAb neutralization experiments; they included the C4 mutations N448Q and T450A and the V2 mutation V182Q. The neutralization sensitivities of these mutants to all V2 MAbs were unchanged from those of the WT (data not shown). Similarly, the mutant pseudoviruses, as well as the WT, were not neutralized by the negative-control anti-parvovirus human MAb 1418.

DISCUSSION

The data presented here demonstrate that selected individual mutations in gp120 expose both V2 and V3 epitopes that are generally occluded in circulating viruses. Thus, mutations that eliminate the N301 glycan, that disrupt the gp120 hydrophobic core, and that alter intraprotomer packing expose both V2 and V3, suggesting general destabilization of the distal region of the trimer. Interestingly, a complementary study showed that mutants in V1V2 render epitopes elsewhere in the trimer more accessible, also highlighting the dynamic nature of the molecule (33). In contrast, several mutants studied exposed V3 but had no effect on V2 epitope accessibility. Thus, mutants that eliminated the N160K glycan, disrupted the CD4bs, or interfered with interprotomer contacts exposed V3 only. These data indicate for the first time that V3 can move independently of V2, pivoting out from its “tucked” position in the trimer while apparently leaving the V2 apex intact. These data are consistent with previous work by Upadhyay et al. showing that engagement of the CD4 receptor with soluble CD4 prior to subsequent incubation with MAbs and pseudovirus strongly affected V3 MAb potency but had only a mild effect on V2 neutralization (1). Similarly, production of pseudoviruses in the presence of kifunensine, which inhibits complex glycan processing, resulted in increased neutralization potency of V3 MAbs but little change in the potency of V2 MAbs (1). It is noteworthy, however, that Lyumkis et al. described the V3 loop as being situated directly beneath the V1V2 hairpin, using their cryo-electron microscopy (EM) structure of the Env trimer in complex with MAb PGV04, and suggested it was unlikely that there could be disruption of V3 without V1V2 reorganization (34). The data described above clearly show that V3 is not “locked” in a cryptic configuration, as predicted by the cryo-EM technique, which provides a depiction of the most prevalent trimer structure at only one time point. Thus, MAb neutralization of the mutants studied demonstrated exposure of V3 alone or V2 and V3, but none exposed V2 epitopes without also exposing V3, indicating that the movement of V2 releases V3, but not vice versa.

These studies, along with many others cited, indicate that intricate contacts throughout the trimer are required to maintain the metastable, closed state of the Env trimer and that disruption of any one of these contacts can release V3 or both V1V2 and V3. Moreover, the closed state of the unliganded Env trimer is only one configuration that can be adopted by the unliganded Env spike. Indeed, several studies have now shown that the unliganded trimer undergoes shifts between a closed ground state (state 1), a functional intermediate (state 2), and an open conformation (state 3) in which the chemokine receptor binding site (CD4i epitope) is exposed (5, 6, 29, 33). Importantly, the SOSIP structure, such as that published by Lyumpkis et al. bound to PGV04, illustrates the most prevalent state of the trimer rather than the only possible state. The data presented in our paper underscore the dynamic nature of Env; there is no single unbound conformation of the trimer. Our V1V2 mobility simulations were performed using a SOSIP trimer in order to show how fluctuations around that predominant state explain the observed data. Studies by both Spurrier et al. (35) and Herschhorn et al. (33) identified “restricting” amino acid residues in the hydrophobic core of V2; these lead to lowered activation barriers between the various states of the Env trimer. In the current study, we identify additional residues within, as well as outside, V2 that behave similarly, causing conformational changes that release previously occluded epitopes and rendering viruses sensitive to the neutralizing effects of MAbs to which they were previously resistant.

The spontaneous transitions between these states of the trimer may provide an explanation for how V2i Abs, which were identified as the only independent correlate of reduced infection risk in the RV144 clinical vaccine trial, could play a protective role and impose immunologic pressure on infecting and circulating viruses (16, 36). In the Herschhorn et al. study (33) of the tier 2 JR-FL virus, the majority of viruses (53%) were present in the closed state 1, while 19% were present in the intermediate state 2 and 28% were present in the open state 3. The majority of the viruses with a closed trimer are resistant to V3 and V2i MAbs under standard neutralizing assay conditions when the V3 and V2i epitopes are occluded. However, assuming that a significant portion of the viruses are present in vivo in, or can spontaneously transition into, the intermediate or open state, these viruses should be sensitive to the V3 and V2i Abs. In fact, in the data shown above, the mutant JR-FL viruses appear to be in the intermediate state, i.e., with V2i and/or V3 epitopes accessible to MAbs but with an occluded CD4i site, as indicated by their general resistance to CD4i MAbs (Tables 1 and 2). These findings appear to be highly relevant in explaining how V2i and V3 Abs were effective in RV144, reducing the rate of infection and imposing immune pressure on the virus in vaccinees (13, 15, 16, 36): if a significant proportion of the trimers on circulating viruses or infected cells are, like JR-FL, in the intermediate or open conformation, then these viruses and infected cells could be affected by the vaccine-induced V2i and V3 Abs.

Generally, the various mutants studied here did not induce exposure of V2q epitopes in JR-FL, which is resistant to MAbs 2909, PG9, and PG16 (Table 1). This was not unexpected, since these MAbs do not bind JR-FL well, as key residues in the C β-strand of V2 that are required for engagement are not present in the JR-FL sequence (37). In order for the antibody to recognize the epitope it should (i) be in the right conformation, (ii) be exposed, and (iii) contain correct residues. One can only detect changes in the first two prerequisites on the condition that the third is satisfied. Our data do confirm that these V2q epitopes cannot be efficiently exposed merely by opening the trimer and exposing V2. PG9 is somewhat unusual within this class, as it binds to selected gp120 monomers (38, 39). Consistent with this feature of PG9, the M434G mutant tested here was sensitive to PG9 neutralization, suggesting that the significant structural displacements induced by disrupting the hydrophobic core of gp120 allowed unusual contacts with the MAb to form. In contrast, JR-FL is moderately sensitive to PGT145 neutralization, and this MAb is trimer dependent, requiring an intact, closed trimer apex for binding (26, 34, 37). The loss of PGT145 neutralizing activity for the core mutants highlights the significant reorganization of Env, where no closed trimer contacts could be formed. Similarly, all trimer-opening mutations abolished PGT145 neutralization, including the intraprotomer and the N160 and N301 glycan mutations. The D197Q interprotomer mutant appears to have also ablated the trimer contacts required for PGT145 binding, though no V2i epitopes were made available. Intriguingly, replacement of D197 with a glycosylated asparagine, which is present in most strains at this position and has been shown to have a stabilizing effect on the trimer, seems to significantly enhance PGT145 neutralization, suggesting the N197 glycan either contributes directly to or helps stabilize the PGT145 epitope (11, 31).

For expanded studies, we created a novel family of mutants expected to expose the V1V2 loop. They were identified by analyzing the crystal structures of BG505.SOSIP and identifying residues forming the interface between the V1V2 domain and the gp120 core (29, 30). In the C4 region of gp120, the I420, K421, and Q422 side chains interact extensively with the V1V2 domain, in particular with the ¹⁷⁷YKLDV¹⁸¹ motif, which forms a “kink” connecting the C and C′ strands of the domain (Fig. 1 and 2) (32). As expected, mutation of these C4 residues caused the pseudovirus mutants to be susceptible to neutralization by V2i MAbs (Table 1). In addition, mutants carrying the C4 mutations were also generally rendered more sensitive to CD4bs-specific MAbs. Indeed, two CD4bs MAbs, 654 and 559, that do not neutralize most primary isolates and did not neutralize WT JF-FL could effectively neutralize the C4 intraprotomer mutants (Table 2). While structural data for complexes of these MAbs with gp120 are not available, it is likely that poor neutralization of primary isolates is due to the restriction of access to the CD4bs by the adjacent protomer or by the V1V2 domain and/or its glycans in the context of the tightly closed trimer. Stringent requirements for the angle of approach of broadly neutralizing MAbs are illustrated by crystal structures of SOSIP with the CD4bs-specific MAbs PGV04 and VRC01 (34, 40). However, mutation-induced liberation of V1V2 movement and increased separation of the protomers in mutant spikes would relax these restrictions, allowing binding and neutralization by normally nonneutralizing CD4bs Abs. The CD4bs was also rendered accessible by the more globally destabilizing mutations to N301, A204, and M434 (Table 2).

Interestingly, MAb VRC01 displayed increased potency for N301Y by more than 2 orders of magnitude (Table 2). The crystal structure of the SOSIP_JR-FL-VRC01 complex demonstrates that substantial contacts exist between the back of the VRC01 heavy-chain Fab and the N301 glycan of the adjacent gp120 protomer (40). Our results suggest that this interaction is not favorable and that VRC01 has to push aside the N301 glycan to access its epitope on the closed trimer. In contrast to the several mutations that increased sensitivity to neutralization by various MAbs, the potency of VRC01 was decreased by mutations in the core (A204E and M343G) and by intraprotomer mutations at K421N and Q422P (Table 2), documenting the unique nature of this CD4bs MAb and its dependence on the metastable packing of the Env spike.

As opposed to the effects of Env mutations on the various classes of MAbs studied here, most mutations had weak or no effect on the neutralization potencies of four CD4i MAbs. We had previously suggested that the A204E mutation promotes transition from a mixed β-sheet conformation to an all antiparallel conformation of the bridging sheet/coreceptor binding site (11). CD4i MAbs 17b and E51 target this site; emergence of neutralization by these MAbs is in agreement with the expected structural effects of A204E (41–43). The specificity of MAb A32, while phenomenologically classified as CD4i, in fact targets a cluster A epitope located in the inner domain of gp120, far from the coreceptor binding site, and requires a much larger opening of the trimer (44–46). Accordingly, none of our mutant viruses were neutralized by A32 (Table 2). Generally, these results indicate that exposure of many V1V2 and V3 epitopes requires only a limited degree of trimer opening, significantly smaller than the major transition induced by CD4 binding (4, 47).

While Abs to V2q are present in a very small proportion of HIV-infected individuals (48, 49) and the V2p class of Abs has not yet been studied extensively in natural infection, Abs directed at the V2i epitope of gp120 are found in greater than 80% of naturally infected subjects, indicating that V1V2 is highly immunogenic (references 22 and 50–52 and M. K. Gorny, unpublished data), and many human MAbs specific for this highly conformational region of V2 have been generated and characterized in detail (18, 22, 32, 35, 53, 54). Indeed, the present study was mainly focused on reactivity to V2i MAbs and the mutations that would change the exposure of the V2i epitope. Many V2q-specific MAbs do not bind well to JR-FL, as the key residues in the C β-strand of V2 that are required for engagement are not present in the JR-FL sequence. Our data confirm that the V2q epitope is not present in WT JR-FL and cannot be efficiently exposed or formed by opening the trimer and exposing V2 and V3.

MAbs with V2i-directed specificities are now sufficiently plentiful and well studied to allow recognition of subtle differences in this Ab category. For example, in contrast to the effect of the intraprotomer mutants on most V2i MAbs, these mutants did not render V2i MAbs 830A and 2297 more potent (Table 1). Similarly, MAbs 21a9, 18a4, 20a12, and 20a24 were generally much less effective against the N301D, N301Y, and L125 mutants than were the nine V2i MAbs shown at the top of Table 1. A further example of the subtle differences within the V2i MAb family is revealed by examining the disparate activities of MAbs 697 and 830A. MAbs 697 and 830A have distinct angles of approach to the V1V2 domain (22, 32). It seems (and simulations give some support to this) that the epitope of MAb 697 is in general more easily accessed by V1V2 domain movements than is that of MAb 830A. One can speculate that M434G introduces more flexibility in the bridging sheet region, facilitating the flexion of the V1V2 stem and allowing V1V2 rotation that exposes the 830A epitope.

Overall, these data speak to the plasticity of the HIV-1 Env. Beyond illuminating the fine points of the structural dynamics of HIV Env, the data provide a road map for creating Env constructs and engineered immunogens designed to target the V1V2 loop. In addition, the data address how Env plasticity provides a plausible explanation for the efficacy of V2 Abs in reducing infection rates of HIV, SHIV, and simian immunodeficiency virus (SIV) (13, 36, 55–60) and the potential value of vaccines that target V1V2.

MATERIALS AND METHODS

Envelope clones and site-directed mutagenesis.

Mutant plasmids of pCAGGS_JR-FL.JB gp160 (originally supplied by John Mascola) were created using QuikChange II XL or QuikChange Lightning kits (Stratagene, La Jolla, CA, USA) according to the manufacturer's instructions or by Genscript (Piscataway, NJ). All constructs were sequenced completely to confirm their identities before use. Mutations were selected from those that had been shown previously to release V3 from its occluded state in the Env trimer, and control mutants that had been shown not to affect the exposure of V3 were used (11). These mutations were in various regions of gp120: V1 (L125F), V2 (N160K, V182Q, I184G, and D197H), C2 (A204E and N280E), V3 (N301D and N301Y), C3 (N332K), and C4 (M434G, N448Q, and T450A) (Fig. 1). Additionally, mutations in the C4 region not previously tested for V3 release and predicted to expose the V2i epitope were also tested: I420A, K421N, Q422P, and Q422A. These sites are directly in contact with the V2 YKLDV kink/integrin binding motif (Fig. 2) (32).

Pseudovirus production and neutralization assays.

Pseudoviruses were produced as previously described (61) in 293T cells transfected with 2 μg of envelope plasmid and 10 μg of an env-deficient subtype B backbone plasmid, pSG3Δenv (NIH AIDS Research and Reference Reagent Program). Pseudoviruses in supernatants were harvested at 48 h posttransfection. Pseudovirus was titrated in TZM-bl cells as previously described (61). The titrations found that, compared to the WT, 5 mutants exhibited increased infectivity, 6 mutants exhibited decreased infectivity, and 7 exhibited no change in infectivity (Fig. 3A). These differences in infectivity were normalized by using a dilution of pseudoviral supernatant shown by titration to yield 150,000 relative light units (RLU) (Fig. 3B). Neutralization assays were performed in TZM-bl cells as previously described. Briefly, after incubation of titrated MAb and pseudovirus for 1 h, cells were added, and the plates were incubated for 48 h and read using Britelite Plus reagent (Perkin-Elmer) according to the manufacturer's protocol on a Biotek Cytation3 imaging reader. The percent neutralization was determined by comparison to pseudovirus-only control wells. The average percent neutralization at each MAb concentration from at least two independent experiments was used to determine the IC₅₀ of the MAb in GraphPad Prism.

Human monoclonal antibodies.

A large panel of human HIV MAbs made in house or obtained from the AIDS Reagent Repository were tested in this study, including representatives of all three classes of V2 MAbs. V2i-specific MAbs included 697-D, 1393A, 1361, 1357, 2158, 4-150, 4-182, 4-324, 4-366, 21a9, 18a4, 20a12, 20a24, 2297, and 830A (produced in house) (18–22). As previously described, V2i Abs recognize a conformation-dependent epitope on V2 that includes the α4β7 integrin binding site and are commonly elicited during infection (22, 35). It is noteworthy that MAb 2297 was previously shown to have a slightly different pattern of binding reactivity based on competition with other V2i MAbs for binding to V1V2-gp70 and was unable to neutralize any pseudoviruses, including tier 1A and 1B pseudoviruses (22). In addition, the structure of the paratope of MAb 830A is different than that of other V2i MAbs, like 697 (22) and 2158 (35); 830A has a narrower neutralizing breadth against tier 1 viruses than other V2i MAbs, and it was shown to have a different angle of approach to the V1V2 region of gp120 than the well-characterized V2i MAb 697 (32). These characteristics of MAbs 2297 and 830A may account for their patterns of reactivity, which distinguish them from other V2i MAbs. In addition, it is noteworthy that V2i MAbs 21a9, 18a4, 20a12, and 20a24 were selected from patient samples with a scaffolded V1V2 (V1V2-gp70) rather than with gp120 (reference 22 and Liuzhe Li, Lily Liu, Susan Zolla-Pazner, Xiang-Peng Kong, and Miroslaw K. Gorny, unpublished data).

The V2p-specific MAbs CH58 and CH59 and the V2q (or V2 apex)-specific MAbs 2909, PG9, PG16, and PGT145 (20, 62–67) were prepared in house or obtained through the AIDS Reagent Repository. V2p Abs are glycan independent and bind to peptides representing a V2 linear epitope on the C strand of V1V2 (65). V2q Abs target a quaternary epitope, including regions of V2 as represented in the Env trimer; these V2q MAbs are glycan dependent, are rarely elicited by infection, are extensively hypermutated, and preferentially bind to trimeric Env (35, 66). It should be noted that the specificity of PG9 for quaternary epitopes is not absolute, while the PG145 epitope exists only on the trimer, and PG16 binding is greatly enhanced on the trimer (38). CD4bs MAbs b12, 654, 559, and VRC01, as well as CD4i MAbs 17b, A32, E51, and 48d, were also tested (64, 68–71). The human anti-parvovirus MAb 1418 was used as a negative control, and CD4 Ig, obtained from the AIDS Reagent Repository, served as a positive control (72).

V1V2 mobility simulations.

Structure preparation, Monte Carlo simulations, and analysis of the resulting conformational ensembles were performed in the ICM molecular-modeling package (Molsoft, San Diego, CA). As a starting point for the simulations, the X-ray structure of the JR-FL SOSIP.559 trimer (PDB accession no. 5FYK) was used (40). Stochastic Monte Carlo perturbations were applied to the internal variables of the two loops comprising the V1V2 stem (amino acid residues V119 to L124 and N197 to A204), to the radial positions of the gp120 protomer cores with respect to the 3-fold symmetry axis of the trimer, and to a V3 conformation. Flickering of V3 was sampled in a discrete manner, i.e., a single Monte Carlo step would switch the entire loop between the tucked-in conformation observed in the X-ray structure and a fully exposed (popped-out) conformation. (A representative conformation was taken from the V3 simulations we previously reported [11].) Radial movements were restrained by a soft (0.25-kcal/Ų) harmonic constraint, as well as hard boundaries beyond −2 and +18 Å. Ensembles of conformations were collected from 100 independent simulation runs performed for both WT and K421N mutant structures. Antibody epitope access for each conformation in the ensemble was evaluated by Fab overlap analysis: the X-ray structure of the Fab-V1V2 complex was superimposed onto each of the V1V2 domains in the given trimer conformation from the simulations, and the number of Fab backbone atoms within a strong clash cutoff (1.5 Å) of the Env spike backbone was determined.