HIV-1 fitness cost associated with escape from the VRC01 class of CD4 binding site neutralizing antibodies.

ABSTRACT

INTRODUCTION

The genetic diversity of globally circulating HIV-1 poses a significant challenge to active and passive vaccination approaches. The majority of this diversity is found in the viral env gene that encodes two glycoproteins, gp120 and gp41, which noncovalently associate and trimerize to form the Env viral spike. Antibodies directed against Env have the capacity to neutralize HIV-1, and the majority of HIV-infected people mount a neutralizing antibody response against the infecting virus. However, the virus is able to evade this autologous antibody response by generating epitope-specific mutations, lengthening flexible variable loops, and shifting glycans (1,–9). The virus and B cells continue to mutate in response to each other (10, 11), and this coevolution leads to various levels of cross-reactive serum neutralizing activity among HIV-1-infected individuals and, in a limited number of cases, the development of broad and potent neutralizing antibody responses (12,–17). However, even among donors with broadly neutralizing sera, the circulating plasma virus generally continues to escape from autologous serum neutralization, thus allowing persistent viral replication (10, 13, 18,–20).

Over the past 5 years, advances in B-cell culture (21,–25) and sorting technologies (26,–29) and the ability to recover antibody genes from single B cells (30,–32) have led to the isolation of many potent and broadly reactive HIV-1 monoclonal antibodies (MAbs). These broadly neutralizing antibodies (bNAbs) target epitopes on the Env viral spike that have been defined as “sites of vulnerability” (33, 34) and comprise the CD4 receptor binding site (CD4bs), the V1V2 regions, the N332 glycan supersite, the membrane-proximal external region (MPER) in gp41 (21, 23, 35, 36), and, recently, a site that bridges gp120 and gp41 on the native Env trimer (37,–39). Antibodies that target the functionally conserved CD4 receptor binding site on gp120, such as the VRC01 class of bNAbs, are of particular interest because viral attachment to CD4 on a target cell is a required first step in the viral entry process (40, 41). Structural studies have revealed that VRC01 class bNAbs partially mimic CD4 in their interaction with gp120, and thus, the antibody epitope contains residues that overlap those of CD4 (35, 42, 43).

Since bNAbs mainly target conserved epitopes on the viral spike, it is possible that escape from such bNAbs could result in impaired viral function or replication. Diminished viral replication after escape from T-cell immune pressure has been well documented for CD8 responses to the conserved Gag protein (44,–46); however, it is less clear if the same is true for Env-targeting antibodies. It has been observed that mutations in CD4bs residues at the end of loop V5 that abrogate binding to anticore antibodies subsequently decrease viral fusion as well as viral infectivity (47), and there is evidence that in vivo antibody escape mutations in env can affect replication capacity (4, 14, 48). Sather and colleagues studied one donor, VC10042, whose sera contained broadly neutralizing activity directed against the CD4bs and demonstrated that viral escape from this response negatively impacted viral replication in some autologous viral strains (19). These data suggest that viral escape from a polyclonal response could decrease viral replicative fitness, but the impact of viral escape from a specific bNAb, such as VRC01, has not been fully addressed.

Our laboratory previously demonstrated that the majority of plasma-derived Envs cloned from donor 45 (from whom the bNAb VRC01 was isolated) were resistant to VRC01, while several archival proviral Envs were VRC01 sensitive. These data and reagents provided an opportunity to study the potential fitness cost to viral replication after escape from VRC01 in vivo. We first studied several heterologous viruses (not from donor 45) that were highly resistant to VRC01 and to other members of the VRC01 class. We established that mutations restoring full VRC01 neutralization sensitivity across the class of antibodies could improve the efficiency of CD4 usage and increase viral replication. Similarly, the introduction of specific loop V5 resistance mutations into a sensitive infectious molecular clone (IMC) resulted in impaired viral replication, although other resistance mutations in loop D had a minimal impact on viral function. Finally, our study of autologous escape from VRC01 using the donor 45 viruses demonstrates that initial escape mutations likely conferred a replication cost, but viral replication could be restored with compensatory mutations. These data provide a detailed analysis of the pathway and associated viral fitness cost for escape from a broadly neutralizing MAb targeting the CD4bs in vivo.

MATERIALS AND METHODS

 

Replication assay with an NL4.3 proviral cassette.

All infectious molecular clones were derived by subcloned env genes into a replication-competent NL4.3 backbone that was previously used to evaluate the replication of diverse Envs and was provided by Cynthia Derdeyn (Emory University, Atlanta, GA) (4, 55, 56). This system is amenable to accepting diverse env genes and facilitates the substitution of virtually the entire env coding region (only 37 and 8 amino acids at the N and C termini, respectively, are derived from NL4.3). CD4 T cells (CD8-depleted PBMC) were cultured in complete RPMI medium for 3 days in the presence of 20 μg/ml phytohemagglutinin (PHA) for activation prior to infection. Infected cultures were maintained in complete RPMI medium supplemented with 20 U/ml recombinant human interleukin-2 (IL-2; Roche Diagnostics) for up to 10 days. Every 2 to 3 days, half of the supernatant was collected for p24 analysis (Perkin-Elmer, Waltham, MA), and this volume was replaced with fresh complete medium supplemented with IL-2. Replication of for each cloned IMC was assessed in 2 to 3 independent experiments. For one experiment, shown in Fig. 1, the wild-type (WT) Env T278-50 IMC did not replicate at all, and so data from this experiment were not included in further analyses. For paired wild-type and mutant IMCs, the mean log₁₀ area under the curve was calculated for replication curves, and statistical significance was assessed by using a paired Student t test.

 

Open in a separate window

FIG 1

Resistance to VRC01 class antibodies can affect CD4 usage and replication. (A) The contact region sequences of five previously described matched pairs of naturally circulating VRC01-resistant viral Envs and their mutated VRC01-sensitive variants are aligned to HXB2 within the three loops comprising contact sites for VRC01 (loop D, CD4 binding loop, and β23-loop V5). HXB2 numbering is indicated, as are the gp120 contacts for CD4 and VRC01, in blue and green, respectively. Open circles indicate main-chain contacts, stars indicate side-chain contacts, and filled circles indicate both main-chain and side-chain contacts. Residues in red boldface type denote mutations that confer VRC01 sensitivity. (B) Neutralization sensitivity of these five paired Env pseudoviruses to six VRC01 class antibodies. Each symbol is the IC₅₀ of the antibody (specified by color) against the indicated resistant wild-type (WT) and sensitive mutant (M) Env pair. (C to E) The impact of mutations on receptor usage was measured by sCD4 neutralization (C), infectivity in Affinofile cells as CD4 expression decreased (D), and in vitro replication using the same five paired Env pseudoviruses reconstructed into fully infectious molecular clones (E). The mean values ± standard errors of the means are plotted, representing data for two (C) or three (D and E) independent experiments. For one experiment in panel E, the wild-type T278-50 Env did not replicate at all. Thus, the data shown are for two experiments where some viral replication could be measured. (F) Quantified results from data shown in panels C to E. The mean IC₅₀ calculated from data in panel C and the fold difference between the mutant and wild type are shown, as is the percent infectivity in low-CD4-expressing cells calculated from data in panel D. Mean AUCs from panel E are indicated, and significant differences between wild-type and mutant replication were calculated by using a paired t test and are noted with asterisks.

RESULTS

 

Mutations in loop D or β23-loop V5 confer resistance to the VRC01 class of antibodies and impact viral replication.

Because in some cases VRC01 resistance was associated with a replicative cost, possibly from less efficient CD4 receptor usage, we sought to further define the amino acid mutations associated with VRC01 class resistance and their impact on CD4-mediated entry and viral replication. Several previous studies indicated that specific amino acid residues in loop D and β23-loop V5 were strongly associated with resistance to the VRC01 class of antibodies (18, 59,–61). For example, after infusion of the VRC01 class antibody NIH45-46 into HIV-1-infected humanized mice, Klein and colleagues documented mutations at positions 276, 278, 279, 280, and 281 within loop D and at positions 458 and 459 in β23 (62). Based on these in vivo escape data and knowledge of critical contact sites from cocrystal structures (42, 59, 63), we constructed a series of mutations singly and in combination in a YU2 Env backbone (Fig. 2A) and assessed these viral variants for antibody and sCD4 neutralization sensitivity, efficiency of CD4-mediated cell entry, and viral replication. We also assessed the frequency of these mutations among currently circulating viruses. An analysis of 5,140 group M Env sequences from the HIV database (http://www.hiv.lanl.gov/) demonstrated that loop D consensus residues N/D279 and N280 were >95% conserved, as were β23 residues R456 and G458 (Fig. 2B). Furthermore, the VRC01 resistance mutations in loop D and β23 highlighted in Fig. 2A occur in <1% of all viruses and never all together, suggesting that these residues are disfavored. Among the four single mutations tested, N279K conferred complete resistance to all tested VRC01 class antibodies, although the other single mutations resulted in partial resistance (Fig. 2C). The double mutation (N279K-R456W) and the quadruple mutations resulted in complete resistance to all six tested antibodies.

 

FIG 2

Impact of resistance mutations in loop D and β23-loop V5. (A) Mutations in loop D and β23-loop V5 previously associated with VRC01 resistance were made, individually and in combination, in a YU2 backbone (mutations indicated in blue boldface type). HXB2 numbering is indicated, as are the gp120 contacts for CD4 and VRC01, in blue and green, respectively. Open circles indicate main-chain contacts, stars indicate side-chain contacts, and filled circles indicate both main-chain and side-chain contacts. (B) Conservation of residues 279, 280, 456, and 458 within 5,140 circulating HIV-1 group M sequences from the HIV sequence database. (C) Neutralization sensitivities of wild-type and mutant Env pseudoviruses to six VRC01 class antibodies. Each symbol is the IC₅₀ of the antibody (specified by color) against the indicated Env. (D to G) Efficiency of CD4 binding was assessed by neutralization sensitivity to sCD4 (D), infectivity in Affinofile cells expressing low levels of CD4 induced by 0.63 ng doxycycline (E), and in vitro replication assays using infectious molecular clones with single-mutation Envs (F) and combination-mutation Envs (G). Wild-type YU2 Env is shown in black, single mutations in loop D are shown as open blue symbols, mutations in β23 are shown as open red symbols, and combination mutations are shown as closed blue symbols. Values represent the means (± standard errors of the means) of data from two (F and G) or three (D and E) independent experiments.

Because these mutated residues in loop D and β23 are CD4 contact sites (Fig. 2A), we next tested the effect of these mutations on CD4 binding. All mutations singly and in combination increased viral resistance to sCD4 (Fig. 2D). The N280D and G458D single mutations led to a >10-fold reduction in sCD4 neutralization, while the N279K mutation in loop D and R456W in β23 led to smaller effects. These observations were confirmed by using the Affinofile cell line to test the infectivity of these mutants in low-CD4-expressing cells (Fig. 2E). The percentage of infected cells was greatly reduced for all mutations except N279K and R456W. We next generated IMCs in the NL4.3 backbone with these mutations in YU2 Env and examined their in vitro replication in PHA-activated CD4 cells by measuring p24 in a 10-day time course. We tested the YU2 WT and N297K Envs as IMCs in the TZM-bl assay to confirm that the neutralization profile remained the same. Similar to our above-described data, we did not observe a direct correlation between an impact on CD4 binding and in vitro viral replication. Despite the measured decrease in CD4 binding for N280D in neutralization and Affinofile cell assays, both the N279K and N280D single amino acid mutations in loop D demonstrated replication kinetics similar to those of the wild-type YU2 virus (Fig. 2F). In contrast, the single mutations in β23 (R456W or G458D) resulted in a marked diminution of viral replication. Not surprisingly, the mutant virus with all four mutations in combination (which is not found in globally circulating viruses) did not replicate at all (Fig. 2G). Remarkably, the double mutant (N279K-R456W) demonstrated wild-type levels of replication, suggesting that the N279K mutation restored the replication capacity of the R456W mutant. These data indicate that although the 279K mutation is disfavored among circulating viral isolates (279N/D comprises 97% of circulating Env sequences), this mutation could serve a dual role as both a resistance mutation as well as a compensatory mutation for R456W (Fig. 2G).

 

Autologous virus from donor 45 escapes VRC01 by mutations in loop D.

We found that mutations in loop D can confer resistance to the VRC01 class of antibodies with no detectable replicative fitness cost to the YU2 IMC. We therefore hypothesized that this domain would be the preferred escape pathway of a virus under immune pressure from VRC01. To test this hypothesis, we analyzed viruses that developed over time in the VRC01 donor (donor 45). Envs isolated from this donor at three different time points (2001, 2006, and 2009) were previously cloned and tested for sensitivity to VRC01 (18). A neighbor-joining tree of these autologous cloned Envs confirmed previously reported data that two archival proviral sequences from PBMC DNA cluster together (45_01dH5 and 45_01dG5) (Fig. 3A). Representative clones from different time points (both PBMC and plasma derived) were chosen for further examination (Fig. 3A, asterisks). The two archival proviral Envs were sensitive to VRC01 neutralization, while the plasma-derived Envs were VRC01 resistant (Fig. 3B). One proviral Env (45_01dH1), which phylogenetically branched between the sensitive and resistant Envs from 2001, showed an intermediate neutralization phenotype. Despite having various sensitivities to VRC01, analysis of CD4 receptor usage showed that there was no detectable difference in neutralization sensitivity to sCD4 between archival and plasma-derived viruses (Fig. 3C) or in the efficiency of entry into cells expressing low levels of CD4 (Fig. 3D). Furthermore, the IMCs generated in the NL4.3 backbone bearing these Envs from donor 45 replicated in activated CD4 cells, and there was no overall difference between the VRC01-sensitive and VRC01-resistant Env variants (Fig. 3E). These data suggest that the predominant circulating VRC01-resistant viruses in donor 45 plasma evolved to both escape VRC01 neutralization and maintain efficient CD4 entry and replicative capacity. To further elucidate how the virus preserved CD4 usage while avoiding VRC01 neutralization, we examined the sequence changes over time in these autologous viruses. A sequence alignment of the main CD4 and VRC01 contact sites in these longitudinal Envs revealed sequence differences between the VRC01-sensitive, -intermediate, and -resistant Envs from 2001 to 2009. Putative escape mutations were defined as polymorphisms within the VRC01 epitope that were fixed in the resistant Envs compared to the sensitive ones, and residues that fit these criteria were K278, D279, A281, and S365 (Fig. 4A). Notably, the first observed amino acid change occurred at position 278 and resulted in a change from lysine to either serine or threonine and thus inserted an N-linked glycosylation sequon at position 276 in all subsequent intermediate and resistant Envs. Structural studies suggest that mature variants of the VRC01 light chain have evolved to avoid a clash with this glycan at position 276 and that its removal is associated with increased virus neutralization by VRC01 class antibodies (42, 43, 63,–65). These putative initial escape mutations were tested singly and in combination by placing them into the archival sensitive Envs of 45_01dG5 (Fig. 4B) and 45_01dH5 (Fig. 4C). These sensitive Envs became more resistant to VRC01 when the K278T (adding glycan at position 276) and/or D279E mutation, but not S365T, was introduced. The A281H mutation increased resistance only in 45_01dH5 Env. The intermediate 45_01dH1 Env naturally has glycan at position 276, and its removal (by a T278K mutation) increased VRC01 sensitivity almost 10-fold (Fig. 4D), confirming the potential role of this glycan in early escape from VRC01. Introduction of the D279E escape mutation into this intermediate Env rendered it completely resistant to VRC01 at 50 μg/ml, while, interestingly, neither the A281H nor the S365T mutation affected VRC01 sensitivity. Overall, these data indicate that the addition of glycan 276, followed by the D279 mutation in loop D, was likely the VRC01 escape pathway that evolved in this donor.

 

FIG 3

Phylogenetic and phenotypic characteristics of autologous donor 45 Env sequences. (A) Neighbor-joining tree of gp160 env gene sequences cloned from donor 45. PBMC-derived sequences from 2001 and plasma-derived sequences from 2001, 2006, and 2009 are shown. (B) Neutralization sensitivities of representative autologous Envs to VRC01. Archival Envs derived from 2001 PBMC are shown in shades of red, while longitudinal plasma-derived Envs are shown in gray. (C and D) Efficiency of CD4 binding was assessed by neutralization sensitivity to sCD4 (C) and infectivity in Affinofile cells expressing low levels of CD4 (D). (E) Replication kinetics of infectious molecular clones with VRC01-sensitive (red) and -resistant (gray) Envs were measured in PBMC-derived CD4⁺ T cells. (F) Average p24 antigen values for all VRC01-sensitive and VRC01-resistant viruses. Mean values (± standard errors of the means) are shown in panels B to F, representing two independent experiments.

 

Open in a separate window

FIG 4

VRC01 escape pathway for donor 45 Env sequences. (A) Alignment of CD4bs motifs from longitudinal Envs cloned from donor 45, including Envs from 2001 archival provirus (red), 2001 plasma (green), 2006 plasma (blue), and 2009 plasma (purple). HXB2 numbering is indicated, as are the gp120 contacts for CD4 and VRC01, in blue and green, respectively. Open circles indicate main-chain contacts, stars indicate side-chain contacts, and filled circles indicate both main-chain and side-chain contacts. Residues 278, 279, 281, and 365 are shown in boldface type, and sequence changes that represent putative VRC01 escape mutations are highlighted in blue. (B to D) VRC01 escape mutations were tested in the backbone of the two archival VRC01-sensitive Envs of 45_01dG5 (B) and 45_01dH5 (C) (maroon) as well as in the backbone of VRC01-intermediate Env 45_01dH1 (red) (D). Mutations made singly are shown as open blue symbols, while closed blue symbols indicate combination mutations. Mean IC₅₀s are plotted, representing data from at least two independent experiments.

 

Initial VRC01 escape mutations exert a fitness cost to viral replication, but compensatory mutations arise concomitantly.

To test the replication cost of these initial VRC01 escape mutations, we introduced the glycan 276 (K278T) and D279E mutations into two archival VRC01-sensitive viruses (45_01dG5 and 45_01dH5). The addition of glycan 276 alone or glycan 276 plus D279E resulted in a marked diminution of replicative capacity for both Envs (Fig. 5A and ​andB).B). Introduction of the D279E mutation into the intermediately sensitive Env (45_01dH1) that naturally contains glycan 276 also markedly reduced replicative capacity, and, as expected, removal of glycan 276 in this Env (T278K) improved replication (Fig. 5C). We confirmed that the neutralization phenotype was retained in both wild-type and mutant IMCs with the added glycan 276. These replication data confirmed the negative impact on viral replication of both the glycan 276 and D279E mutations. Interestingly, the longitudinal sequence alignment of donor 45 Envs revealed that amino acid changes at position 279 predominantly occurred concurrently with changes at position 281 (Fig. 4A). Because the A281H polymorphism did not affect VRC01 neutralization sensitivity when placed into the intermediate 45_01dH1 Env (Fig. 4D), there was the possibility that this change was compensatory. Remarkably, introduction of the A281H mutation in conjunction with the glycan 276 and D279E mutations (45_01dH1 D279E-A281H virus) restored viral replication to wild-type levels (Fig. 5C). These data suggest that an initial fitness cost to VRC01 escape within loop D can be overcome by compensatory mutations. Thus, viruses circulating in donor 45 appear to have first developed a partial resistance mutation by the insertion of a glycan at position 276 and next became fully resistant to VRC01 by the addition of the D279E residue change (Fig. 5D). The replication capacity of this fully escaped virus, however, was decreased, and therefore, the A281H mutation arose concomitantly to compensate and increase replicative capacity.

 

FIG 5

Initial VRC01 escape mutations confer a replication cost that is overcome with compensatory mutations. (A to C) VRC01 escape mutations (glycan 276 and D279E) were introduced sequentially into three VRC01-sensitive Envs, and replicative capacity in vitro was assessed. Infectious molecular clones expressing wild-type VRC01-sensitive Envs of 45_01dG5 (A) and 45-01dH5 (B) are shown in maroon, while the VRC01-intermediate Env of 45_01dH1 (C) is shown in red. Escape mutations (addition of glycan 276 [K278T] and subsequently D279E) are shown as open blue symbols, while the reversion mutation (removal of glycan 276 [T278K] in 45_01dH1) is shown in maroon, and compensatory mutation A281H is shown in solid blue. Mean values (± standard errors of the means) are graphed, representing data from two independent experiments. (D) Schematic pathway of viral escape from antibody VRC01 in donor 45. The two archival Envs lack the sequon for the N-linked glycan at position 276 and are VRC01 sensitive. Upon the introduction of this glycan 276 mutation (K278T), there is partial VRC01 resistance, and viral replication is markedly reduced. An additional single mutation in loop D (D279E) further reduced viral replication but resulted in high-level VRC01 resistance. A compensatory mutation in loop D (A281H) restored viral replication, while VRC01 resistance was maintained.

DISCUSSION

Broadly neutralizing antibodies that target conserved viral epitopes may induce escape mutations in the virus that affect viral replicative fitness, as has been described for CD8 T cells (44,–46). Structural and functional studies of the VRC01 class of antibodies indicate that they closely mimic the CD4 interaction with gp120 and bind to a conserved epitope that overlaps contact residues of the CD4 receptor (26, 42, 43, 49, 66). To understand if escape from the VRC01 class of antibodies can affect the Env affinity for CD4, the efficiency of viral entry, or replicative fitness, we studied the consequences of VRC01 class resistance among naturally circulating primary isolates on each of these viral functions. We found that naturally circulating VRC01 class-resistant Envs or Envs with introduced resistance mutations can have deficient interactions with CD4 and decreased replicative fitness. We next identified and characterized the initial in vivo VRC01 escape mutations (K278T and D279E) in the autologous donor 45 viruses and found that escape from VRC01 resulted in impaired viral replication, although subsequent compensatory mutations (for example, A281H) restored fitness. It is possible that the mutations that we introduced into these Envs could affect their overall neutralization phenotype. We therefore tested the reversion and escape mutations, and we saw no overall differences in neutralization by non-CD4bs antibodies PG9, PGT128, 10E8, and 17b. Thus, our results demonstrate that VRC01 class resistance mutations are associated with decreased receptor interactions and defective viral replication and that the autologous VRC01-sensitive viruses are unusual and lack a highly conserved glycan. These observations suggest that this study of viral escape from VRC01 may generate insights for immunogen design and/or passive treatment.

Because the VRC01 class of antibodies bind to gp120 in a similar manner, we used previously described matched pairs of five wild-type VRC01-resistant viruses and their reverted-to-sensitive partners (49) to test whether these reversions restored neutralization sensitivity to other VRC01 class antibodies. We observed that reversion to VRC01 sensitivity was associated with increased sensitivity to other VRC01 class antibodies, but fine epitope differences between antibodies may explain slight variations (43, 67). Furthermore, reversion to VRC01 class sensitivity resulted in increased sensitivity to sCD4 neutralization and improved viral entry for three of five viruses tested and improved viral replication for two viral pairs. Notably, all five VRC01 class-resistant viruses were circulating strains and, therefore, presumably biologically fit. Our data, however, establish that their replication fitness could be improved; therefore, we demonstrated that the CD4bs in these VRC01-resistant strains may suboptimally interact with the CD4 receptor, implying that the replicative ability of these viruses may be inherently limited.

This observation that VRC01 class resistance can have a biological impact on viral receptor interactions led us to test specific point mutations in the sensitive YU2 virus strain, which has been well characterized in the in vivo humanized mouse model. By using this model, Nussenzweig and colleagues demonstrated that treatment of YU2-infected mice with VRC01 class antibody 3BNC117 leads to a temporary reduction in viral loads, followed by a rebound associated with resistance mutations (61, 62). Based on those reports as well as studies by the Bjorkman group and by our group on residues associated with VRC01 resistance (49, 59, 60), we performed site-directed mutagenesis to introduce known VRC01 class resistance mutations into the VRC01-sensitive YU2 Env. We observed that specific VRC01 resistance mutations within the D loop and β23-loop V5 region of Env could substantially affect replication fitness and that, in some cases, this effect could be attributed to less efficient CD4 usage, especially when multiple resistance mutations were present. Notably, the key loop D and β23-loop V5 residues associated with VRC01 class antibody resistance are highly uncommon in nature, generally accounting for <1% of circulating HIV-1 strain sequences, and this observation lends further support to the hypothesis that these mutations have a detrimental impact on viral fitness. Our data demonstrate, however, that mutations in loop D were better tolerated than those in the β23-loop V5 domain in YU2 Env. Interestingly, we observed that at least one of these residues (N279K) could play a dual role in neutralization resistance and replication compensation simultaneously; i.e., the N279K mutation restored replicative fitness in an Env with the R456W mutation. Thus, certain pathways of escape from VRC01 class MAbs may be more advantageous than others because of the various replicative fitness costs to the virus.

The autologous Envs from donor 45 (from whom VRC01 was isolated) were previously cloned and shown to be under selection pressure from VRC01 (18). Therefore, we studied the functional consequences of in vivo escape from the broadly neutralizing MAb VRC01 during HIV-1 infection. When full autologous gp160 sequences were tested for replication as infectious molecular clones, there was no overall difference in replication between VRC01-sensitive and -resistant Envs. However, to understand if initial escape mutations exerted an effect on viral replication in the absence of compensatory mutations, we examined the effect of single mutations that arose longitudinally in Env sequences from donor 45. Specifically, we mapped the putative initial escape mutations within loop D, which is a critical part of the known epitope of VRC01. We tested these mutations in the backbone of early proviral VRC01-sensitive Envs of 45_01dG5 and 45_01dH5 and determined that two early mutations increased resistance to VRC01 and were escape mutations (K278T and D279E). Our data suggest a VRC01 escape pathway in donor 45 whereby the K278T mutation is the first escape mutation detected and inserts a putative N-linked glycan (glycan 276) in loop D. This addition of a glycan within the CD4bs increases resistance to VRC01 and, when followed by the D279E mutation in Env of 45_01dH1, leads to the virus becoming completely resistant. It is interesting to note that changes at position 279 in the longitudinal viral quasispecies were generally accompanied by changes at residue 281. When these specific escape mutations were tested in molecular infectious clones, the reason for this linkage became clear. The addition of glycan 276 impaired replication in all three Envs tested, and the subsequent addition of the second resistance mutation D279E completely knocked down viral replication in the 45-01dH1 Env unless accompanied by A281H, which restored viral replication to wild-type levels. Thus, the A281H mutation, which did not affect VRC01 neutralization, is a compensatory mutation in loop D. Strikingly, this wild-type 45_01dH1 Env naturally contains glycan 276 but replicates to much higher levels when this glycan is removed. These data suggest that even with compensatory mutations, there were competing pressures on viral evolution between replicative fitness and VRC01 escape, leading to the suboptimal replication of certain viral species. This donor was a long-term nonprogressor with a low viral load, and one explanation for this phenomenon may be that VRC01 escape mutations in certain viral quasispecies constrained viral replication to a relatively low level.

Thus, the major determinants of VRC01 escape and compensation in donor 45 were found within four residues of loop D (HXB2 positions 278 to 281), although other factors, such as loop length and glycosylation, etc., likely contributed to full VRC01 resistance. Overall, our data suggest that loop D tolerates sequence changes for neutralization escape and/or replication compensation when the CD4bs is being targeted by antibodies. This observation is somewhat surprising given how conserved loop D remains across circulating subtypes of HIV-1. We have previously reported, however, that the CD4bs is fairly immunogenic (51), and perhaps the need to shield the CD4bs with glycan 276 from this common antibody response is the reason for its high level of conservation. These data suggest that in antibody-based passive treatment of viremia, the use of a broadly neutralizing antibody from the VRC01 class in conjunction with a non-class CD4bs antibody that is not affected by changes in loop D may help minimize the possibility of viral escape.

There remain questions in the field as to the antigenic nature of the viruses that stimulated the broadly neutralizing VRC01 class lineages. Previously reported studies of VH gene-reverted versions of VRC01 class antibodies revealed a lack of binding to a multitude of HIV-1 Envs (64, 65). Those studies revealed that glycan 276 in loop D as well as glycans 460 and 463 in loop V5 play a role in preventing the inferred naive B cell receptors (BCRs) of the VRC01 class of antibodies from recognizing these gp120 proteins. Interestingly, Env of 45_01dG5 lacks putative glycans at positions 460 and 276. Furthermore, both of the early proviral VRC01-sensitive viral sequences from this donor (45_01dG5 and 45_01dH5) lack the putative glycosylation sequon at positions 276 to 278 to encode glycan 276, and this fact is striking because the glycan is 95% conserved in circulating HIV-1 viruses. Paradoxically, the removal of glycan 276 increases VRC01 sensitivity (Fig. 4D) (68) but is also part of the VRC01 epitope (42, 43), and our data may explain how this antibody lineage developed these characteristics. We speculate that these unusual glycan 276-lacking early viruses in donor 45 could bind and activate the naive VRC01 BCR. The subsequent addition of this glycan, in an attempt to escape from the VRC01 response, may have induced the VRC01 lineage to affinity mature toward incorporating glycan 276 into its epitope. This coevolution would therefore allow VRC01 to become capable of broadly neutralizing the majority of HIV-1 strains that contain this glycan. We postulate that donor 45 provides another example of autologous viral escape by insertion of a glycan that broadens a neutralizing epitope (13) and that other donors with VRC01 class responses may have antibody lineages that were similarly activated by viruses lacking this glycan. Furthermore, these observations have implications for vaccine design. If VRC01 class precursors can be triggered only by unusual viruses lacking glycan 276, sequential immunization with donor 45 env genes may be the solution to activating VRC01 class antibodies. Thus, this study demonstrates that there can be a replication cost to escape from broadly neutralizing antibodies and that by studying this viral evolution, we may learn unique properties of these viruses with implications for immunogen design.

ACKNOWLEDGMENTS

REFERENCES