Mechanistic Analysis of the Broad Antiretroviral Resistance Conferred by HIV-1 Envelope Glycoprotein Mutations

Although combination antiretroviral (ARV) therapy is highly effective in controlling the progression of HIV disease, drug resistance can be a major obstacle. Recent findings suggest that resistance can develop without ARV target gene mutations.

IMPORTANCE Although combination antiretroviral (ARV) therapy is highly effective in controlling the progression of HIV disease, drug resistance can be a major obstacle. Recent findings suggest that resistance can develop without ARV target gene mutations. We previously reported that mutations in the HIV-1 envelope glycoprotein (Env) confer resistance to an integrase inhibitor. Here, we investigated the mechanism of Env-mediated drug resistance and the possible contribution of Env to virological failure in vivo. We demonstrate that Env mutations can reduce sensitivity to major classes of ARVs in multiple viral isolates and define the effect of the Env mutations on Env subunit interactions. We observed that many Env mutations accumulated in individuals failing integrase inhibitor therapy despite a low frequency of resistance mutations in integrase. Our findings suggest that broad-based Env-mediated drug resistance may impact therapeutic strategies and provide clues toward understanding how ARV-treated individuals fail therapy without acquiring mutations in drug target genes.
KEYWORDS antiretroviral drugs, envelope glycoprotein, virus transmission, gp41, human immunodeficiency virus compensatory mutations Env-Y61H and P81S in gp120 and the Env-A556T mutation in gp41. These Env mutations decrease the sensitivity of the virus to the highly potent INSTI dolutegravir (DTG). Moreover, propagation of HIV-1 in the presence of DTG led to the selection of the DTG-escape mutant Env-A539V in the absence of any mutations in IN. These four Env mutations cluster within the C1 domain of gp120 and the HR1 domain of gp41 ( Fig. 1), which have been shown through mutagenesis and structural studies to be critical for the stability of gp120-gp41 association (50)(51)(52). These results suggest that Env mutations that increase cell-cell spread may alter the stability of the gp120-gp41 interaction and demonstrate that mutations in Env can confer resistance to an ARV, at least in a subtype B, CXCR4-tropic laboratory-adapted HIV-1 strain. Whether Env-mediated drug resistance occurs in clinical HIV-1 isolates is unclear, because Env from laboratory-adapted and primary HIV-1 strains differs in significant functional and structural aspects. Laboratory-adapted isolates often sample the open Env conformation and are more sensitive to NAbs and soluble CD4 (sCD4) than primary isolates (17,53,54). sCD4 inhibits HIV-1 entry not only by competing with CD4 on the target cell but also by inducing the shedding of gp120 from the surfaces of viral particles. Env complexes from laboratoryadapted isolates are more prone to sCD4-induced gp120 shedding than primary isolates (55)(56)(57), indicating that Env stability differs between laboratory-adapted and primary isolates.
In the present study, we examined whether our previously described Env mutations confer resistance not only to DTG but also to other classes of ARVs. To evaluate the FIG 1 The location of previously reported Env mutations that overcome blocks to virus replication (48). (A) Schematic of the HIV-1 Env-coding region with the position of mutations indicated using NL4-3 (and HXB2, in parentheses) numbering. Labeled domains are defined as follows: C1 to C5, constant regions 1 to 5; V1 to V5, variable regions 1 to 5; FP, fusion peptide; HR1/HR2, heptad repeat 1/2; DSL, disulfide loop; MPER, membrane proximal external region; TM, transmembrane; CT, cytoplasmic tail. (B) A Env structure of subtype B JR-FL SOSIP.664 (PDB accession number 5FYK [137]) with the position of the Env mutations highlighted. Most of gp120 is shown in white, with gp120 V1/ V2 and V3 loops colored in light yellow and light green, respectively. gp41 is colored in tan. Y61, P81, A539, and A556 are highlighted in green, yellow, red, and blue, respectively. Structural model was generated using UCSF Chimera software (138).
impact of the Env mutations on the conformation and stability of the Env complex, we examined their effects on sensitivity to NAbs and on gp120 shedding. To further characterize the mechanism of Env-mediated drug resistance, we obtained Env mutants through in vitro selection experiments by propagating clinically relevant HIV-1 clones in the presence of ARVs and analyzed the phenotypes of the selected Env mutants. Finally, to address the possible in vivo relevance of Env-mediated drug resistance, we analyzed viral sequences obtained from the plasma of individuals failing a raltegravir (RAL)-containing regimen. The results indicate that Env mutations confer resistance across a broad range of ARVs in vitro and that the NL4-3 Env mutations decrease gp120 shedding and alter the sensitivity of Env to NAbs. Finally, we show that in some individuals failing a RAL-containing regimen in the absence of resistance mutations in IN, changes arise at the highly conserved gp120-gp41 interface, as observed during INSTI escape in vitro.

RESULTS
The A539V mutation in gp41 provides a replication advantage over WT NL4-3 in the presence of multiple classes of ARVs. We previously reported that the Env-A539V mutation provides resistance to the INSTI DTG (48). To examine whether Env-A539V also provides a selective replication advantage in the presence of other ARVs, the SupT1 T-cell line was transfected with the WT NL4-3 molecular clone or the Env-A539V derivative in the absence or presence of various concentrations (0.1 to 3,000 nM) of a panel of ARVs. Replication kinetics were monitored by measuring the reverse transcriptase (RT) activity in the cell culture medium. Consistent with our previous report (48), Env-A539V exhibited faster replication kinetics than the wild type (WT) in the absence of drugs (Fig. 2). Whereas replication of WT NL4-3 was largely inhibited in the presence of 3.0 nM DTG, Env-A539V was able to replicate, albeit with delayed kinetics ( Fig. 2A). By calculating the DTG 50% inhibitory concentration (IC 50 ) based on the RT activity at the peak of replication in the absence of drugs, we found that Env-A539V showed 8.1-fold resistance to DTG relative to that of the WT (Fig. 2I). We also examined the sensitivity of Env-A539V to another INSTI, raltegravir (RAL). Env-A539V was able to replicate in the presence of 10 nM RAL and showed 5.4-fold resistance relative to that of WT NL4-3 ( Fig. 2B and I). We also measured replication kinetics of Env-A539V in the presence of the NNRTI efavirenz (EFV) and the NRTI emtricitabine (FTC). Whereas replication of WT NL4-3 was strongly inhibited in the presence of 3.0 nM EFV or 30 nM FTC, Env-A539V still replicated at these concentrations ( Fig. 2C and D). IC 50 calculations showed that Env-A539V exhibited 28-and 5.4-fold resistance against EFV and FTC, respectively (Fig. 2I). Next, we examined the impact of PIs on the replication kinetics of Env-A539V. While replication of WT NL4-3 was strongly inhibited in the presence of 3.0 nM nelfinavir (NFV), the Env-A539V mutant replicated and exhibited 24-fold resistance ( Fig. 2E and I). We also examined the sensitivity of Env-A539V to the allosteric integrase inhibitor (ALLINI) BI-224436, which induces aberrant IN multimerization and impairs the interaction between IN and the cellular cofactor LEDGF/p75 (58). As shown in Fig. 2F, Env-A539V caused a small but statistically significant reduction in the sensitivity to BI-224436 relative to that of WT NL4-3 (2.3-fold resistance) (Fig. 2I). These data demonstrate that Env mutations can overcome inhibition imposed by multiple classes of ARVs targeting not only postentry steps but also the maturation step of the HIV-1 replication cycle.
It has been suggested that entry inhibitors (Ent-Is) are effective in the context of cell-cell transmission (59,60). In addition, several studies have shown that Env mutations selected in the presence of Ent-Is altered viral sensitivity to other anti-Env agents (61)(62)(63)(64). These findings raise the possibility that Env mutations selected in the presence of ARVs might alter the viral sensitivity to Env-targeted inhibitors. To examine this question, we measured replication kinetics in the presence of the fusion inhibitor T-20 and BMS-378806, which blocks CD4-induced conformational changes in Env (19,(65)(66)(67). Interestingly, these Ent-Is efficiently suppressed replication of both WT NL4-3 and Env-A539V with no statistically significant differences in antiviral IC 50 s in a multicycle spreading infection ( Fig. 2G to I). These observations suggest that Ent-Is can suppress the replication of Env mutants selected under the pressure of ARVs targeting the viral enzymes.
The Env-A539V mutation does not alter drug sensitivity in cell-free infection. We previously proposed that Env mutations that confer resistance to ARVs do so by enhancing the efficiency of cell-cell transfer (48). Based on this hypothesis, the Env mutations would not be predicted to confer resistance in the context of cell-free infectivity. To test this hypothesis, we measured the single-round infectivity of WT NL4-3 and the Env-A539V mutant in the TZM-bl indicator cell line. As we reported previously (48), Env-A539V showed comparable cell-free infectivity relative to that of WT NL4-3 in the absence of inhibitor ( Fig. 3A and Table 1). Consistent with the hypothesis that the reduced ARV sensitivity of the Env-A539V mutant is conferred at the level of cell-cell  50 were calculated compared to that for the WT. Data from at least two independent experiments are shown as means 6 standard errors (SEs). ***, P , 0.001; **, P , 0.01; *, P , 0.05 by unpaired t test. The EFV and NFV data are shown on a separate bar graph to avoid compression of the y axis. transmission, WT and Env-A539V infectivity was reduced to the same extent by the inhibitors (Fig. 3B to I and Table 1). Interestingly, Env-A539V did not alter susceptibility to either T-20 or BMS-378806, suggesting that Ent-Is efficiently suppress both cell-free infection and cell-cell transmission of HIV-1. We also measured cell-cell fusion activity of Env-A539V by coculture of TZM-bl cells with transfected 293T cells in the presence of a cocktail of EFV and DTG (200 nM each) to prevent infection of the TZM-bl cells (see Fig. S1 in the supplemental material). In contrast to the Y61H, P81S, and A556T mutations, which markedly suppressed cell-cell fusion activity (48), the Env-A539V mutation only slightly decreased cell-cell fusion compared to that of the WT.
The Env-A539V mutation increases viral resistance to EFV when coupled with the RT-Y188L mutation. To examine whether mutations in Env can increase the level of resistance conferred by drug target gene mutations, we tested EFV resistance of the RT-Y188L mutant (68) in the context of either WT or A539V Env. As shown in Fig. 4A (left), RT-Y188L showed a modest delay in replication compared to that of the WT in the absence of EFV, indicating that this RT mutation confers a small fitness defect. This replication defect was rescued by Env-A539V, consistent with our earlier finding that Env mutations can rescue the replication defect conferred by a mutation in IN (48). In the presence of 1.0 mM EFV, the Env-A539V mutation was able to rescue the replication kinetics of the RT-Y188L mutant at early time points (Fig. 4A, right), a result that was confirmed over a broad range of EFV concentrations ( Fig. 4B and D). In contrast to the results for the multicycle spreading infection (Fig. 4B), the Env-A539V mutation did not increase EFV resistance of RT-Y188L in a single-cycle infectivity assay ( Fig. 4C and D). Taken together, these results demonstrate that, in the context of a spreading infection in which virus transmission can take place via cell-cell transfer, an Env mutation can significantly increase the resistance conferred by a mutation in an ARV target gene.
NL4-3 Env mutations that overcome blocks to HIV-1 replication in spreading infections stabilize the gp120-gp41 interaction. Our previously reported Env mutations, which we selected for their ability to overcome blocks to virus replication in multicycle spreading infections, are located in the C1 domain of gp120 and the HR1 region of gp41 (Fig. 1B). In the crystal structure of the HIV-1 Env SOSIP trimer, which has recently been identified as adopting the state 2 conformation (69-72), these mutations are clustered at the gp120-gp41 interface. These domains of Env have been shown in mutagenesis and structural studies to be critical for the stability of the gp120-gp41 association (50-52, 62, 73). To provide clues regarding the mechanism by which Env mutations enhance the capacity of HIV-1 to spread via cell-cell transfer, we examined their impact on gp120-gp41 association. It is known that sCD4 induces gp120 shedding from virus particles (57); we therefore incubated purified virions with sCD4 at 37°C for 2 h and then measured the amount of particle-associated gp120 and p24 by Western blotting. As expected, levels of gp120 associated with WT NL4-3 particles were decreased in a dose-dependent fashion by sCD4 treatment (0.3 to 10 mg/ml). In contrast, the mutations in both gp120 (Env-Y61H and P81S) and gp41 (Env-A539V and A556T) showed significantly reduced sCD4-induced gp120 shedding (Fig. 5A). To quantify this effect, we calculated the 50% effective concentration (EC 50 ) of sCD4-induced shedding. While the EC 50 of sCD4-induced gp120 shedding for WT NL4-3 was 0.99 mg/ ml sCD4, the EC 50 s for the Env-P81S and the other Env mutants were 3.5 and .10 mg/ ml, respectively (Fig. 5B). To examine the interaction between sCD4 and a representative Env mutant, we measured the sensitivity of Env-A539V infectivity to sCD4 (Fig. 5C). Env-A539V showed a 3.7-fold resistance to sCD4 relative to that of the WT. Moreover, the IC 50 of sCD4 against Env-A539V was 1.1 mg/ml (Table 1), which is ;10-fold less than the EC 50 of sCD4-induced gp120 shedding; this result suggests that sCD4 can bind to mutant Env at lower concentrations than needed for gp120 shedding, leading to inhibition of viral entry. To directly examine the interaction of Env with CD4, we measured the binding of a fusion protein (CD4-IgG2 [74]) comprising human IgG2 in which the Fv portions of both heavy and light chains have been replaced by the V1 and V2 domains of human CD4 (see Fig. S2). We observed comparable binding of WT and A539V Env to CD4-IgG2, suggesting that the Env-A539V mutation does not affect the binding affinity of Env for CD4. These results also suggest that sCD4 inhibits WT infection by inducing gp120 shedding and competing with gp120 binding to CD4 on target cells, whereas sCD4 inhibits infection of Env-A539V primarily by competing with CD4 binding on target cells. To further characterize the intrinsic stability of mutant Envs, we also performed a time-dependent gp120 shedding assay by incubating viruses at 37°C for up to 5 days (Fig. 5D). The Env mutations significantly reduced timedependent gp120 shedding (Fig. 5E); whereas the half-life of WT NL4-3 gp120 shedding was 2.6 days, that of the Env mutants was .5 days. These data suggest that the NL4-3 Env mutations stabilize the gp120-gp41 interaction.
The Env-A539V mutation reduces sensitivity to NAbs that recognize the CD4induced Env conformation. To probe the impact of the Env-A539V mutation on Env conformation, we examined the sensitivity of Env-A539V to a panel of NAbs that recognize different Env conformations (Fig. 6). We used three groups of NAbs. The first group ( Fig. 6A to E), which included VRC01 (specific for the CD4 binding site [CD4bs]), PGT145, PG16 (specific for the V2 apex), 2G12, and PGT121 (V3-glycan specific), preferentially bind to the unliganded closed Env conformation (75)(76)(77)(78). The second group  Concentrated viruses were incubated with sCD4 at the indicated concentrations at 37°C for 0 or 2 h. Incubated viruses were subsequently purified through a 20% sucrose cushion, and viral proteins were detected by Western blotting. (A) A representative gel for the sCD4-induced gp120 shedding assay is shown. (B) The ratio of gp120 to p24 was quantified and plotted. Data from at least two independent experiments are shown as means 6 SEs. (C) TZM-bl cells were exposed to 100 TCID 50 of the indicated viruses incubated with sCD4 prior to infection for 0 or 2 h. Luciferase activity was measured at 48 h postinfection. Data from three independent experiments are shown as means 6 SEs. ***, P , 0.001; **, P , 0.01; *, P , 0.05 by unpaired t test, with asterisks for 0 h or 2 h incubation time points indicated in black or gray, respectively. (D and E) Concentrated viruses were incubated at 37°C for the indicated times. Incubated viruses were subsequently purified through a 20% sucrose cushion, and viral proteins were detected by Western blotting. (D) A representative gel for the timedependent gp120 shedding assay is shown. (E) The ratio of gp120 to p24 was quantified and plotted. Data from at least three independent experiments are shown as means 6 SEs. ***, P , 0.001; **, P , 0.01; *, P , 0.05 by unpaired t test.
( Fig. 6F) included 35O22, which is specific for the gp120-gp41 interface and shows no conformational preference (79). The third group ( Fig. 6G to J) included F105 (CD4bs specific), 17b (specific for the CD4-induced epitope), 447-52D (V3 loop), and 10E8 (gp41 membrane-proximal external region [MPER]), which preferentially target the CD4-bound Env conformation (80)(81)(82)(83). Although the Env-A539V mutation is not located within the epitopes of the NAbs used in this analysis, it altered the sensitivity to several NAbs, with reduced sensitivity to F105 (3.1-fold), 447-52D (28-fold), 17b (27fold), and 10E8 (1.8-fold) and increased (33-fold) sensitivity to PGT145 (Table 1). To further probe the effect of the Env-A539V mutation on Env conformation, we compared NAb binding to Env on 293T cells by flow cytometry. As indicated in Fig. 6M and N, FIG 6 Sensitivity or binding of Env-A539V to NAbs recognizing different Env conformations, anti-CD4 Ab, and coreceptor antagonist. TZM-bl cells were exposed to 100 TCID 50 of WT NL4-3 or Env-A539V viruses in the presence of various concentrations of NAbs (A to J), anti-CD4 Ab (K), and CXCR4 antagonist (L); luciferase activity was measured at 48 h postinfection. Data from three independent experiments are shown as means 6 SEs. 293T cells transfected with the green fluorescent protein (GFP)-encoding reporter clone pBR43IeG expressing WT NL4-3 Env or the Env-A539V mutant were preincubated with 17b (M), 10E8 (N), PG16 (O), or PGT145 (P) at 4°C for 1 h. KFS is an Env-defective mutant. The cells were washed, and APC-conjugated anti-human IgG was used to detect bound Ab. APC signals were normalized by GFP signal to calculate the Ab binding efficiency. Data are shown as means 6 SEs from three independent experiments. ***, P , 0.001; **, P , 0.01; *, P , 0.05 by unpaired t test or one-way ANOVA and Tukey's multiple-comparison test.
17b and 10E8 bound less efficiently to Env-A539V than WT Env. In contrast, PGT145 and PG16 bound the mutant Env to a similar extent as WT NL4-3 Env ( Fig. 6O and P). These data suggest that the Env-A539V mutation stabilizes the closed Env conformation. We also examined sensitivity to SIM.4 (an anti-CD4 Ab) (84) and AMD3100 (a CXCR4 antagonist) (85) to evaluate the impact of the Env-A539V mutation on Env function in HIV-1 entry. As shown in Fig. 6K and L, Env-A539V showed comparable sensitivity to SIM.4 and AMD3100 as WT NL4-3, suggesting that this mutation does not alter CD4 or CXCR4 dependency.
Env mutations confer ARV resistance in the context of CCR5-tropic HIV-1. To examine whether Env-mediated resistance to ARVs occurs in strains of HIV-1 other than the CXCR4-tropic lab-adapted strain NL4-3, we propagated the CCR5-tropic clone NL(AD8) (86) in the SupT1huR5 T-cell line, which expresses high levels of human CCR5, in the presence of DTG. Treatment of transfected SupT1huR5 cultures with 6.0 nM DTG markedly delayed NL(AD8) replication, but at 31 days posttransfection, we observed a peak of replication (Fig. 7A). Sequencing of the putative escape mutant indicated the absence of any mutations in IN but revealed the presence of an Env-N654K mutation in gp41 HR2 (Fig. 7B). To examine whether the Env-N654K mutation in NL(AD8) confers resistance to DTG, we introduced the mutation in WT NL(AD8) and examined replication kinetics in the presence of DTG (Fig. 7C). Interestingly, this mutant exhibited faster replication kinetics than WT and still replicated in the presence of 300 nM DTG (Fig. 7C, left). IC 50 calculations indicate that the Env-N654K mutation confers 30-fold resistance to DTG relative to that of the WT NL(AD8) (Fig. 7D). As observed with several of the Env mutations selected in our previous study (48), the Env-N654K mutation impairs singlecycle infectivity (Fig. 7E) and does not confer DTG resistance in the context of cell-free virus ( Fig. 7F and Table 2). These results are all consistent with the hypothesis that, as with the mutations selected in the context of NL4-3, the NL(AD8) Env-N654K mutation confers DTG resistance by enhancing the efficiency of cell-cell transfer.
To examine whether the Env-A539V mutation, which was originally selected in the context of NL4-3 (48), would also confer ARV resistance in the context of another viral isolate, we introduced this mutation into the NL(AD8) molecular clone. In contrast to the phenotype of this mutant in NL4-3, the NL(AD8) Env-A539V mutant exhibited delayed replication in the absence of DTG and an increase (3.1-fold) in DTG sensitivity in multicycle spreading infections. Moreover, the cell-free infectivity of the NL(AD8) Env-A539V mutant was reduced relative to that of WT NL(AD8) (Fig. 7C to E). These data demonstrate that the impact of Env mutations on replication kinetics and drug sensitivity can be HIV-1 strain dependent.
We also propagated the subtype C, CCR5-tropic transmitted-founder isolate K3016 (87) in the presence of DTG, EFV, and rilpivirine (RPV) (Fig. 8A). We identified an Env-T541I mutation in gp41 HR1, which is located at the same position as NL4-3 Env-A539V (Fig. 8B), in the presence of DTG. We also identified an Env-E621V mutation in the gp41 disulfide loop (DSL) region in the presence of both FTC and RPV. To examine whether the Env-T541I and E621V mutations in K3016 confer resistance to ARVs, we introduced these mutations into WT K3016 and examined replication kinetics in the presence of DTG, EFV, NFV, and T-20 ( Fig. 8C to J). These three Env mutants exhibited faster-than-WT replication kinetics in spreading infections but reduced cell-free infectivity relative to that of the WT. These mutants also showed reduced sensitivity to DTG, NFV, and EFV but not to T-20 in spreading infections ( Fig. 8K and L). Again, these Env mutants did not alter sensitivity to DTG in the context of cell-free virus (Fig. 8M). These results indicate that Env-mediated drug resistance, which is linked to enhanced cellcell transmission, can occur in clinically relevant HIV-1 strains independent of coreceptor usage, subtype, or ARV classes.
Next, to examine the impact of the Env-N654K mutation on the conformation and neutralization properties of NL(AD8) Env, we examined the sensitivity of this mutant to a panel of NAbs ( Fig. 9 and Table 2). Compared to NL4-3, NL(AD8) is more susceptible to NAbs, such as PGT121, PG16, and PGT145, that preferentially recognize the closed Env conformation (see Fig. S3 and Tables 1 and 2). Conversely, NL(AD8) is resistant to 17b ( Fig. S3 and Tables 1 and 2) and F105 (Table 1). These results suggest that NL(AD8) tends to sample the closed Env conformation relative to NL4-3. NL(AD8) Env-N654K increased sensitivity to 35O22 (56-fold) and 10E8 (9.1-fold) (Fig. 9E and G and Table 2).  [137]), highlighting the location of Env mutations selected in the context of NL4-3 (48) and the Env-N654K mutation selected in NL (AD8). Env amino acid positions are indicated using the NL4-3 numbering. Most of gp120 is shown in white, with gp120 V1/V2 and V3 loops colored in light yellow and light green, respectively. Env-Y61, P81, and A556 are highlighted in cyan; Env-A539 and N654 are highlighted in red and purple, respectively. The structural model was generated using the UCSF Chimera software. These results suggest that the Env-N654K mutation in NL(AD8) has different impacts on Env conformation than the Env-A539V mutation in NL4-3. We also examined sensitivity to several entry inhibitors (Fig. 9H to L and Table 2). Env-N654K slightly increased sensitivity to SIM.4, maraviroc (MVC; a CCR5 antagonist), and T-20. Moreover, Env-N654K increased the sensitivity to sCD4. Unlike NL4-3 (Fig. S3J), incubation time did not affect the susceptibility of Env-N654K to sCD4 (Fig. 9H and Fig. S3J). This observation suggests that sCD4-induced gp120 shedding is minimal in the NL(AD8) strain, consistent with previous reports for non-lab-adapted isolates (55)(56)(57). Indeed, we did not observe sCD4-induced gp120 shedding from either WT or Env-mutant particles at sCD4 concentrations up to 10 mg/ml (see Fig. S4A, C, and E). We also examined timedependent gp120 shedding ( Fig. S4B and D). In contrast to that for NL4-3 ( Fig. 5D and E, and Fig. S4F), only minimal gp120 shedding was observed for NL(AD8) Env over a 5day incubation period at 37°C, and there was no significant difference in shedding between WT and Env-N654K NL(AD8) Env ( Fig. S4B and D).
Sequence analysis of IN-and Env-coding regions of viruses from individuals failing a RAL-containing regimen. Our results indicate that Env-mediated drug resistance can occur in clinically relevant HIV-1 strains in vitro. However, whether Env mutations contribute to drug resistance in vivo is unclear. To address this question, we used single-genome sequencing to compare viral sequences obtained from the plasma of participants failing a RAL-containing regimen in the SELECT study (5). The aim of the SELECT study was to examine whether RAL with boosted lopinavir (LPV) would be noninferior to boosted LPV with NRTIs. By 48 weeks of treatment, 10.3% of 258 participants treated with the RAL-containing regimen experienced virological failure, defined here as the inability to achieve a viral load ,400 copies/ml at two consecutive time points or after 24 weeks of treatment. By the end of the follow-up in this study, 10 participants developed new resistance mutations in IN, including IN-N155H. In the present study, we compared changes in the sequences of IN-and Env-coding regions from five participants who experienced viral rebound (Table 3). These participants were infected with subtype C CCR5-tropic strains. Each participant had therapeutic plasma concentrations of RAL (.33 ng/ml) at the time of virological failure. No resistance mutations in IN were detected at day 0 (Table 3). While we identified the IN-N155H mutation in participant identifier (PID) A4 with high frequency (73%), PID A1, A3, and A5 had only a low  frequency of resistance mutations in IN (Table 3), suggesting that virological failure may have occurred in these participants as a result of mutations elsewhere in the HIV-1 genome. Evaluation of the Env-coding region before treatment and during virologic failure indicated the presence of many changes in gp120 and gp41 that emerged or were selected with failure. Most of the changes in gp120 were located in highly variable regions, such as the variable loops (V1/V2, V4, and V5 loops), and the C3 region (data not shown). Because these regions contain immunodominant epitopes of subtype C Env, mutations in these regions may have been driven by immune pressure (88,89). Although we did not identify mutations at the same positions as those observed in our in vitro studies, we observed an accumulation of mutations in the gp120 C1 domain and gp41 ectodomain in PID A1, A3, A4, and A5 ( Fig. 10A and B). The frequency of most of the mutations increased significantly with viral rebound. Although we identified a number of Env mutations in PID A2, the frequency of mutations before RAL therapy was similar to that after viral rebound, although the RAL-containing regimen decreased viral load before viral rebound (data not shown and Table 3). In many cases, amino acid residues were replaced with more conserved residues; however, we did identify changes to residues with very low prevalence, such as R577K (gp41 HR1) and L602R (gp41 HR2). Some mutated positions in the gp41 ectodomain were identified in several participants (Fig. 10A and B). The observation of changes arising in highly Broad Drug Resistance Mediated by HIV-1 Env ® conserved positions at the gp120-gp41 interface-a region in which, based on our data, mutations that confer ARV resistance in vitro cluster-in the absence of drug resistance mutations in IN, suggests the possibility that these gp41 changes may have contributed to virological failure in a subset of participants in the study. Further analysis will be needed to explore this hypothesis in more detail.

DISCUSSION
In this study, we demonstrate that the NL4-3 Env-A539V mutant, which was selected in the presence of DTG, confers broad resistance to multiple classes of ARVs in multicycle spreading infections. This Env mutation also increases resistance to ARVs when coupled with ARV target gene mutations, again in the context of spreading infections. However, this Env mutation does not alter ARV susceptibility in cell-free infection. The results demonstrate that Env mutations clustered at the gp120-gp41 interface can confer broad resistance by enhancing the ability of HIV-1 to spread via a cell-cell route of transmission. Interestingly, NL4-3 Env mutants exhibiting an enhanced ability to spread via cell-cell transmission have more stable and closed Env conformations than WT NL4-3. By propagating CCR5-tropic HIV-1 strains NL(AD8) and K3016 in the presence of DTG, we obtained gp41 ectodomain mutations Env-N654K and Env-T541I, respectively. We also identified Env-E621V by propagating K3016 in the presence of RT inhibitors. We demonstrate that these Env mutations reduce susceptibility to several ARVs in multicycle spreading infection, as observed for the NL4-3 Env mutants, indicating that Env-mediated drug resistance can occur in non-laboratoryadapted HIV-1 strains. However, the effects of the Env mutations on replication kinetics and Env structure are HIV-1 strain dependent. Finally, we performed single-genome sequencing analysis of the IN-and Env-coding regions of viruses from infected individuals failing a RAL-containing regimen. We observed that many mutations accumulated in the Env-but not IN-coding region in most study participants. While most of the gp120 changes that arose in vivo were located in highly variable regions, a number of changes in the gp41 ectodomain were observed at the highly conserved gp120-gp41 interface, as observed in vitro.
Several studies have demonstrated that cell-cell transfer is associated with a reduced susceptibility to ARVs relative to that with cell-free infection, but high concentrations of ARVs, or their use in combination, can block both cell-free and cell-cell infection (32)(33)(34). We previously proposed a model, consistent with other findings (33), whereby Env mutations that enhance cell-cell transmission increase the MOI following viral transfer across the VS; concentrations of DTG that are sufficient to inhibit cell-free infection by these mutants are therefore insufficient to block their cell-cell transmission   48). Based on this model, the selected Env mutations would be predicted to confer resistance to multiple ARVs independent of their mode of action and should confer resistance in the context of a spreading, but not cell-free, infection. Indeed, the NL4-3 Env-A539V mutant selected in the presence of DTG exhibits broad resistance to a number ARVs, including INSTIs, NRTIs, NNRTIs, and PIs, in spreading replication assays but not in cell-free infectivity assays. It is possible that additional mutations may accumulate during spreading replication of Env-A539V in the presence of ARVs; Env mutations may thus be the starting point on the path to acquisition of high-level drug resistance. It is also possible that mutations in Env could arise that confer drug resistance by greatly increasing the efficiency of cell-free infection. However, we have not observed such mutations in any of our selection experiments; rather, the mutations we have identified in this and our previous study (48) either greatly impair cell-free infectivity or have no major effect on this mode of transfer. Taken together, these observations suggest that the broad resistance conferred by the HIV-1 Env mutations will affect susceptibility not only to the currently approved ARVs but also potentially to next-generation drugs. It is noteworthy that other studies have also observed the selection of Env mutations upon propagation of HIV-1 in the presence of various inhibitors of virus replication (13,(90)(91)(92).
In contrast to the other classes of ARVs, the Env mutants we identified in NL4-3 and K3016 are sensitive to two types of Ent-Is: a fusion inhibitor (T-20) and a compound that blocks CD4-induced conformational changes in Env (BMS-378806) (19,(65)(66)(67). Previous studies have shown that Ent-Is targeting CD4 or a coreceptor are effective in blocking cell-cell transmission (34,60,93). While some CD4bs-targeting agents inefficiently inhibit cell-cell transmission, anti-gp41 MPER antibodies (Abs) and T-20 are effective (60,94). These observations suggest that accessibility of epitopes or time of action may be important factors for the efficacy of Ent-Is in blocking cell-cell transfer. Interestingly, the Env-A539V mutation has been reported to emerge in the presence of low concentrations of fusion inhibitors in vitro; however, this mutation does not confer resistance to the fusion inhibitor and reverted back to WT at a high inhibitor concentration (95). Overall, our findings imply that including Env-targeted inhibitors in a cART regimen may help prevent the emergence of the type of broadly resistance-conferring Env mutations described here and previously (48).
Virological failure has been observed in patients on PI-containing regimens in the absence of PI-resistance mutations in PR or in the vicinity of Gag cleavage sites (7,(96)(97)(98)(99). Likewise, recent studies have demonstrated that viral rebound can occur in individuals on INSTI-containing therapies without the emergence of INSTI resistance mutations in IN (5,6). In some studies (100,101), there are concerns about whether the infected individuals remained adherent (i.e., whether suppressive concentrations of the inhibitors were maintained), but in other studies (e.g., reference 5), the maintenance of adequate plasma drug concentrations was experimentally verified. In some cases in which target gene mutations were identified in the treated individuals experiencing virological failure, the observed resistance mutations conferred only low-level resistance (3)(4)(5)(6)(7)(8)(9)(10). These findings suggest that mutations outside the pol-coding region (and not around Gag cleavage sites, in the case of PIs) also contribute to virological failure in association with the target gene mutations. Mutations in the cytoplasmic tail of gp41 were previously proposed to contribute to PI resistance in vivo (13), potentially reflecting the role of virion maturation in Env-mediated viral entry (102,103). Mutations in the gp41 HR region have also been observed in association with PI failure, although their contribution to virological failure is unclear (9). In addition to mutations in IN, it has also been proposed that mutations in the 39 PPT contribute to INSTI failure Mutations that changed conserved positions to less conserved residues are indicated with a superscript lowercase "a." Frequencies were determined for subtype C sequences (n = 5,923) retrieved from the Los Alamos HIV databse. Fischer's exact test was performed to determine statistical significance. ***, P , 0.001; **, P , 0.01; *, P , 0.05; ns, not significant; N.D., not determined. Broad Drug Resistance Mediated by HIV-1 Env ® both in vitro and in vivo (12,104). We demonstrate here that the Env-A539V mutation rescues the defect in virus replication imposed by the RT-Y188L mutation (68) at early time points and increases the resistance of this RT mutant to EFV in multicycle spreading but not cell-free infection. These observations indicate that Env mutations can increase resistance conferred by mutations in the ARV target gene by enhancing viral cell-cell transmission. Mathematical modeling suggests that cell-cell transmission increases the probability that NAb resistance mutations will emerge, relative to cellfree infection, as a result of reduced NAb potency in the context of cell-cell transmission (94). As shown in Fig. 2, Env-A539V provides a replication advantage over WT NL4-3 in the presence of ARVs, suggesting that mutations in Env such as those described here may facilitate the acquisition by the virus of resistance mutations in ARV target genes. According to this "stepping stone" model, by facilitating virus replication in the presence of ARVs, the Env mutations would facilitate the emergence of high-level resistance mutations in ARV target genes. Additional studies will be needed to investigate the role of HIV-1 Env in the development of high-level drug resistance.
We observed that Env mutations in the NL4-3 strain reduce sCD4-induced and time-dependent shedding of gp120 from viral particles. In addition, the Env-A539V mutation decreases sensitivity to NAbs preferring the CD4-bound conformation, whereas this mutant is more susceptible to PGT145, which recognizes the closed Env conformation. These observations suggest that the NL4-3 mutant Envs are more closed and stable on viral particles than WT NL4-3 Env. Flow cytometry analysis using Envexpressing cells indicates that the NAb-binding properties of the mutants largely parallel their neutralization properties, suggesting that mutant Envs on the surfaces of infected cells exhibit similar conformational dynamics to those on viral particles. To examine the impact of the Env-A539V mutation on the HIV-1 entry step, we also analyzed the sensitivity of this mutant to an anti-CD4 Ab (SIM.4) (84) and a CXCR4 antagonist (AMD3100) (85). Because the NL4-3 Env-A539V mutation does not alter sensitivity to either entry inhibitor, it is likely that the enhancement of cell-cell transfer is due to the increased probability of gp120-CD4 interactions rather than the increased affinity of gp120 for CD4 or CXCR4. Indeed, Env-A539V does not affect CD4-IgG2 binding. Our findings suggest that decreased gp120 shedding may contribute to enhanced Env-CD4 interactions at cell-cell contact sites. Further structural analysis will be needed to understand how the Env mutations stabilize the gp120-gp41 interaction in the unliganded Env conformation, because current structural models using HIV-1 Env SOSIP trimers represent the state-2 conformation (69)(70)(71)(72).
We performed resistance analyses using the more clinically relevant, relative to NL4-3, HIV-1 strains NL(AD8) and K3016. We introduced the Env-A539V mutation into NL(AD8) and observed a different phenotype from that observed with this mutation in NL4-3. Whereas NL4-3 Env-A539V exhibited faster-than-WT replication kinetics, NL (AD8) Env-A539V exhibited impaired replication kinetics and was more susceptible to DTG than WT NL(AD8). These results imply that the Env-A539V mutation will not emerge in the context of the NL(AD8) strain. However, DTG selection experiments led to the emergence of an Env-T541I mutation, which is located at the same position as NL4-3 Env-A539V, in the context of the subtype C, CCR5-tropic transmitted founder isolate K3016. Whereas an Ala at Env-541 (539 in NL4-3) is highly conserved in subtypes B and C (98.00% and 98.65%, respectively), Thr, Ile, and Val are rare residues at this position in both subtypes. As observed with the NL4-3 Env mutants, K3016 Env-T541I exhibited faster-than-WT replication and conferred resistance to ARVs in spreading infections. These results highlight the context dependence of resistance-conferring Env mutations, consistent with the suggestion that residues in the gp41 ectodomain are functionally linked with gp120 and modulate conformational dynamics of Env in a strain-dependent manner (52,62,105,106). As shown in Fig. S3 and S4 in the supplemental material, the stability of the gp120-gp41 interaction and the conformational dynamics of NL(AD8) Env are different from those of NL4-3 Env, suggesting that the differences in intra-and interprotomer interactions of Env are likely to affect the viral phenotypes of the Env mutants in terms of their fitness and ability to confer broad ARV resistance.
Propagation of NL(AD8) in the presence of DTG led to the selection of the Env-N654K mutation in gp41 HR2. Like several of the selected NL4-3 Env mutations (this report and in reference 48), the NL(AD8) Env-N654K mutation confers resistance to DTG and is highly fit in multicycle spreading infections but is also profoundly defective in cell-free infectivity. These observations suggest that this mutant spreads predominantly via a cell-cell route, as observed for the NL4-3 Env mutants. However, the effects of the NL(AD8) Env-N654K mutation on Env stability and conformation appear to be different from those of the NL4-3 Env mutants. This mutation increases the sensitivity to the anti-CD4 Ab, MVC, and to sCD4, T-20, and an anti-MPER Ab, suggesting that Env-N654K may be in a more open conformation than WT NL(AD8). As discussed above, WT AD8 is a primary isolate and prefers to adopt the closed Env conformation. It may have evolved in vivo while avoiding the N654 change to maintain the closed state. An Asn at Env-654 is highly conserved (99.76% in subtype B), and previous studies have shown that substitutions at this position decrease the affinity of HR2 for HR1, resulting in decreased cell-free infectivity and fusogenicity (107)(108)(109). Several lines of evidence suggest that syncytium formation is tightly regulated at the site of cell-cell contact by viral and cellular factors (102,103,110). Although the roles of syncytium formation in HIV-1 replication are still debated, the formation of large syncytia may negatively impact HIV-1 replication (110,111). Several studies have reported that mutations in Env arose in infected SupT1 cells and improved replication in the absence of syncytium formation (112)(113)(114)(115). As we reported previously, Env-Y61H, P81S, and A556T in the NL4-3 strain enhance cell-cell transmission while markedly impairing cell-free infectivity and fusogenicity, indicating that for these mutants, reduced fusogenicity is associated with enhanced cell-cell transmission capacity (48). In contrast, we show here that the Env-A539V mutant exhibits only a small impairment in fusogenicity in a cellcell fusion assay. Additional studies will be needed to elucidate the precise mechanism of action of the Env-N654K mutation.
To explore the possibility that Env-mediated ARV resistance may occur in vivo, we performed single-genome sequence analysis using samples from the SELECT study (5), one of several studies that examined whether INSTI monotherapy would effectively suppress viral loads in animal models and human participants (6,116,117). In many cases, virological failure occurred in either the absence of mutations in IN or in association with low-frequency resistance mutations in IN (118). Consistent with these observations, our single-genome sequence analysis revealed that four of the five participants we analyzed had only low-frequency resistance mutations in the IN-coding region. These results suggest that virological failure occurred in these four participants as a result of mutations outside the Pol-coding region. As expected, we observed that many Env mutations accumulated in less-conserved regions of Env, such as gp120 C3 and variable loops, which are predominant targets for the Ab response (88,89). More interestingly, we also found that several mutations in the gp41 ectodomain accumulated at the highly conserved gp120-gp41 interface, as observed in our in vitro selection experiments. As shown in our in vitro studies, it is unlikely that the identical mutations observed in vitro would emerge in vivo, because the effects of the Env mutations on viral fitness and drug resistance are virus strain dependent. It is interesting to note that viruses from PID A4, which had the IN-N155H mutation with high frequency, had many changes in the gp120 C1 domain and the gp41 ectodomain compared to viruses isolated from the other participants. A limitation of the current study is the relatively small number of INSTI failure samples analyzed. Larger numbers of samples will be needed to identify specific mutational patterns associated with virological failure. In addition, other regions outside IN, such as the 39 PPT, may also be associated with virological failure (12,104). Further analysis will be needed to understand the impacts of Env mutations on viral susceptibility to cART and their contribution to virological failure. Nevertheless, our findings are consistent with the possibility that HIV-1 Env simultaneously evolves to escape from both NAbs and ARVs in association with ARV target gene mutations in vivo.
In summary, the findings of this study demonstrate that mutations in Env can markedly reduce the susceptibility of HIV-1 to a broad panel of ARVs in vitro. These findings offer mechanistic insights into Env-mediated drug resistance in vitro and provide new clues to understand how HIV-1 develops high-level resistance to ARVs in vivo.
Virus replication kinetics assays. Virus replication was monitored in SupT1 cells as previously described with minor modifications (48). Briefly, SupT1 cells were incubated with pNL4-3 clones (1.0 mg/ 1.0 Â 10 6 cells) in the presence of 700 mg/ml DEAE-dextran at 37°C for 15 min. Transfected cells (1.5 Â 10 5 cells) were plated in 96-well flat-bottom plates and incubated at 37°C in the presence of various concentrations of ARVs. For CCR5-tropic viruses, we used SupT1huR5 cells, which express high levels of human CCR5. Aliquots of supernatants were collected to monitor RT activity, and cells were split 1:3 every other day with fresh drug and medium. As described previously (48), IC 50 values were calculated based on RT activity at the peak of replication of each virus; specifically, the IC 50 was defined as the amount of inhibitor required to reduce RT levels by 50% at the peak of virus replication. To identify the selected mutations in the Pol/Env-coding region, genomic DNA was extracted from infected cells by using a DNeasy Blood and Tissue minikit (Qiagen), and then the Pol-and Env-coding regions were amplified by PrimeSTAR GXL DNA polymerase (TaKaRa) and sequenced (Psomagen) using previously described primers (48).
Single-round infection assays. For cell-free infectivity assays, TZM-bl cells (1.0 Â 10 4 cells) in 96-well plates were exposed to RT-normalized virus stocks produced in HeLa cells. For drug sensitivity assays, 100 TCID 50 of virus produced in 293T cells was exposed to TZM-bl cells (1.0 Â 10 4 cells) in the presence of various concentrations of drugs or neutralizing antibodies. For the sCD4 sensitivity assay, prior to the infection, the viruses were incubated with various concentrations of sCD4 at 37°C for 0 or 2 h. For assays using PIs or ALLINIs, viruses were produced in the presence of serial dilutions of inhibitors as described previously (127). Briefly, 293T cells were transfected with WT or mutant molecular clones as indicated above. Six hours posttransfection, 1.5 Â 10 5 transfected cells were incubated with a 3-fold serial dilution of the inhibitors at 37°C for 48 h. Virus-containing supernatants were harvested and used to infect TZMbl cells. At 48 h postinfection, luciferase activity was measured using the Britelite plus reporter gene assay system and Wallac MicroBeta plate reader (PerkinElmer).
Cell-cell fusion assay. The HEK293T cells were transfected with HIV-1 proviral DNA as described above. At 24 h posttransfection, serial dilutions of the transfected HEK293T cells were cocultured with 1.0 Â 10 4 TZM-bl cells in the presence of a cocktail of EFV and DTG (200 nM each). At 24 h postcoculture, luciferase activity was measured as described above.
Western blotting. Viral proteins were separated by SDS-PAGE and transferred to polyvinylidene disulfide (PVDF) membranes (Merck Millipore). After blocking the membranes with 5% skim milk, they were probed with primary antibodies for 1 h and then incubated for 1 h with species-specific horseradish peroxidase-conjugated secondary antibody. After the final washes, bands were detected by chemiluminescence with a Sapphire Biomolecular imager (Azure Biosystems). Quantification was performed using Image Studio Lite (LI-COR Biosciences) software. The p24 protein was detected with anti-HIV Ig at a final concentration of 5.0 mg/ml in 5% skim milk. gp160 and gp120 were detected with the 16H3 monoclonal antibody at a final concentration of 0.5 mg/ml in 5% skim milk.
Antibody binding assay. HEK293T cells were transfected with pBR43IeG constructs as indicated above. At 24 h posttransfection, the cells were detached with 5 mM EDTA-phosphate-buffered saline (PBS) and then washed with PBS. Cells (2.0 Â 10 5 ) were incubated with the specified antibodies at a final concentration of 1.0 mg/ml at 4°C. After 1 h of incubation, the cells were washed with PBS. The cells were then washed with PBS and incubated with allophycocyanin (APC)-conjugated F(ab9)2 fragment donkey anti-human IgG antibody (Jackson ImmunoResearch Laboratories; catalog no. 709-136-149) at a final concentration of 62.5 mg/ml. The cells were washed with PBS and fixed with 4% paraformaldehyde (PFA; Boston Biosciences). Fixed cells were analyzed with a BD LSR-II or FACSCalibur (BD Biosciences) flow cytometer. Data were analyzed by FCS Express Cytometry software 7 (De novo Software).
Ethics Statement. Plasma samples were obtained prior to and after virologic failure from donors on combination ART of ritonavir (RTV)-boosted lopinavir (100 mg RTV, 400 mg LPV) plus 400 mg RAL twice a day. Samples were collected by the AIDS Clinical Trial Group (ACTG) study A5273, a randomized, openlabel phase-3 noninferiority study at 15 research sites in nine resource-limited countries (three sites in India and South Africa, two in Malawi and Peru, and one each in Brazil, Kenya, Tanzania, Thailand, and Zimbabwe) (5). The primary endpoint was time to confirmed virologic failure (two measurements of HIV-1 RNA viral load .400 copies/ml). This trial is registered with ClinicalTrials.gov, NCT01352715. Entry and exclusion criteria are listed in the protocol (5). The study was approved by ethics committees at each site and written informed consent was obtained from each participant.
Plasma viral RNA extraction and cDNA synthesis. Viral RNA was extracted from plasma containing 200 to 20,000 copies of HIV-1 RNA. Plasma was first centrifuged at 5,300 Â g for 10 min at 4°C to remove cellular debris. The supernatant was then centrifuged at 21,000 Â g for 1 h at 4°C. Viral RNA was extracted from the viral pellet as previously described (128) and resuspended in 40 ml of RNase-free 5 mM Tris-HCl (pH 8.0). Five microliters of each of 10 mM deoxynucleoside triphosphates (dNTPs) and 10 mM oligo(dT) were added prior to denaturing the RNA at 65°C for 10 min. cDNA was synthesized as previously described (129) except at 50°C for 1 h followed by 55°C for another hour. Two microliters RNase H (2 U/reaction) was added, and the RNA was digested at 37°C for 20 min.
Sequence analysis. Sanger sequencing data were analyzed using MEGA (136) to align and translate sequences. Sequences were analyzed for known INSTI resistance mutations using HIV Stanford HIV Drug Resistance database (https://hivdb.stanford.edu/).
Statistical analyses. Two-tailed unpaired t tests and Fischer's exact tests were performed using GraphPad Prism 8 (GraphPad). A P value of ,0.05 was considered statistically significant. GraphPad Prism was also used to assess statistical significance by one-way analysis of variance (ANOVA) and Tukey's multiple-comparison test.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.