Subtype-specific Conservation of Isoleucine 309 in the Envelope V3 Domain is Linked to Immune Evasion in Subtype C HIV-1 Infection

Abstract

The V3 region of the HIV-1 envelope (Env) glycoprotein gp120 is a key functional domain yet it exhibits distinct mutational patterns across subtypes. Here an invariant residue (Ile 309) was replaced with Leu in 7 subtype C patient-derived Envs from recent infection and 4 related neutralizing antibody escape variants that emerged later. For these 11 Envs, I309L did not alter replication in primary CD4 T cells; however, replication in monocyte-derived macrophages was enhanced. Infection of cell lines with low CD4 or CCR5 revealed that I309L enhanced utilization of CD4 but did not affect the ability to use CCR5. This CD4-enhanced phenotype tracked with sensitivity to sCD4, indicating increased exposure of the CD4 binding site. The results suggest that Ile 309 preserves a V3-mediated masking function that occludes the CD4 binding site. The findings point to an immune evasion strategy in subtype C Env to protect this vulnerable immune target.

Introduction

HIV-1 group M is a collection of genetically diverse viruses that have been classified into 9 major subtypes as well as multiple circulating and unique recombinant forms (for a complete listing see http://www.hiv.lanl.gov). The envelope (env) gene is a major source of this genetic diversity, varying by approximately 10% within an individual, 20% within a subtype, and 35% between subtypes (Korber et al., 2001). The product of the env gene is a 160kDa polyprotein precursor that is proteolytically processed into individual subunits, gp41 and gp120, which associate non-covalently to form trimeric ‘spikes’ on the surface of the virion. The gp120 subunit protrudes from the virion surface, and contains the binding sites for the CD4 receptor and the coreceptors CCR5 or CXCR4. The gp120 is also the major target for neutralizing antibodies (Nab) but its genetic variability poses a significant obstacle for vaccine-induced protection (Binley et al., 2008; Gorny, 2004; Gray et al., 2009; Scheid et al., 2009). Much of what is currently known about the organization of gp120 is based on crystal structures of a truncated, de-glycosylated, CD4-bound subtype B core or a truncated, glycosylated, unliganded SIV core (Chen et al., 2005; Kwong et al., 2000; Kwong et al., 1998). The structure and position of the five ‘hyper-variable’ domains (V1-V5) on gp120 have been difficult to determine because of their conformational flexibility; it is, therefore, not fully understood how these domains could influence the overall conformation and immunogenicity of the native protein. Consequently, the positions, inter-molecular interactions, and genetic diversity of the hyper-variable domains could lead to subtle but important structural differences, particularly between viral subtypes (Gnanakaran et al., 2007; Gray et al., 2007; Lynch et al., 2009; Patel, Hoffman, and Swanstrom, 2008; Rong et al., 2007b).

Of the five hyper-variable domains, V3 is relatively conserved (Huang et al., 2005) and does not exhibit the dramatic insertions, deletions, and shifts in potential N-linked glycosylation sites that are characteristic of the V1V2 and V4 domains. Perhaps this reflects that the V3 domain participates directly in coreceptor binding, which is a critical step in viral entry (Cardozo et al., 2007; Cormier, 2002; Trkola et al., 1996). Nevertheless, the amino acid sequence of V3 and its mutational pattern exhibit differences across subtypes (Felsovalyi et al., 2006; Gaschen et al., 2002; Korber et al., 1994; Patel, Hoffman, and Swanstrom, 2008). One striking example is that subtype A and C V3 domains contain a highly conserved GPGQ amino acid motif at the crown, while GPGR is predominant in subtype B Envs (Korber et al., 1994; Stanfield et al., 2006). Subtype D Envs, on the other hand, carry a mixture of residues at the R/Q position (www.hiv.lanl.gov). The subtype B V3 domain facilitates a switch in tropism, from CCR5 to CXCR4 usage in about 50% of patients (Connor et al., 1997; Richman and Bozzette, 1994; Schuitemaker et al., 1992; Tersmette et al., 1989), whereas CXCR4 usage among subtype C viruses is infrequent, even in advanced stage patients (Choge et al., 2006; Cilliers et al., 2003; Coetzer et al., 2006; Isaacman-Beck et al., 2009; Morris et al., 2001; Sullivan et al., 2008). Consistent with possible functional constraints, the subtype C V3 domain exhibits less sequence variation compared to subtype B (Gaschen et al., 2002; Gilbert, Novitsky, and Essex, 2005; Korber et al., 1994; Patel, Hoffman, and Swanstrom, 2008; Rong et al., 2007b; Stanfield et al., 2006). Even during escape from autologous Nab in subtype C HIV-1 infection, the V3 domain remains remarkably conserved amid ongoing sequence evolution in other Env regions (Moore et al., 2009; Rong et al., 2009). It has recently been shown that distinct mutational patterns in subtype B and C lead to conformational differences as well (Patel, Hoffman, and Swanstrom, 2008). Interestingly, position 309 in V3 exhibits extreme conservation as Ile in subtype C but in subtype B, Leu, Met, and Val also occur with relative frequency (Patel, Hoffman, and Swanstrom, 2008). Thus, while lineage-specific genetic differences in the V3 domain have been firmly established, their underlying biology is not clearly understood.

Here we have begun to explore the biological basis for conservation of V3 sequence by creating an I309L substitution in a panel of eleven diverse, patient-derived subtype C Envs that includes recently transmitted viruses and defined autologous Nab escape variants. We evaluated changes in both function and antigenicity of the wildtype and mutated Envs and uncovered evidence suggesting that in subtype C, I309 participates in immune evasion by masking the CD4bs and possibly other vulnerable sites, even though this residue is not critical for interactions with CCR5.

Materials and Methods

Env clones

Details of the ZEHRP cohort, sample collection, and processing have been described previously (Derdeyn et al., 2004; McKenna et al., 1997; Trask et al., 2002). The Envs studied here were derived from seven newly infected subjects from this cohort (185F, 153M, 205F, 221M, 109F, 106F, and 55F). The Emory University Institutional Review Board, and the University of Zambia School of Medicine Research Ethics Committee approved informed consent and human subjects protocols. None of the subjects received antiretroviral therapy during the evaluation period. PCR amplification and cloning of the Envs has been described (Derdeyn et al., 2004; Haaland et al., 2009; Rong et al., 2009). The env genes were cloned into one of the CMV-driven expression plasmids pCR3.1 or pcDNA 3.1/V5-His TOPO (Invitrogen), which were then used to generate viral pseudotypes. All Envs were derived from plasma or PBMC DNA according to protocols previously described (Li et al., 2006) and are subtype C. Nucleotide sequences (either V1-V4 or full-length) have been deposited into Genbank under the following accession numbers: 55 FPB4a AY423973, 106 FPB9 AY424163, 109 FPB32 AY424141, 205FPB 12MAY05ENV5.1 GQ485442, 205FPL27MAR03ENV5.2 GQ485418, 185FPL 10JUL04ENV1.1 GQ485388, 185FPB 24AUG02 ENV3.1 GQ 485314, 153MPL 13MAR02 ENV5.1 HM068596, 153MPL 13MAR04 ENV4.1 HM068597, 221MPL 7MAR03 ENV2.1 HM068598, 221MPL 26FEB05 ENV3.1 HM068599. All amino acid positions are based on HXB2 gp160 numbering.

PCR-based site mutagenesis

To generate I309L mutations, PCR-directed site mutagenesis was performed using two overlapping primers that contained the mutated sequence for each Env using a strategy similar to that described in (Rong et al., 2007a). Briefly, each env gene (plus Rev and partial Nef coding sequences) was amplified using primer sequences similar to the following, which were used with 185F 0-Month EnvPB3.1 (substituted nucleotide is underlined and HXB2 locations are provided): 7124-7162 (5′-CAA TAA TAC AAG GAA AAG TGT GAG ACT AGG ACC-3′) and 7134-7167 (5′-GTC CTG GTC CTA GTC TCA CAC TTT TCC-3′). The amplification conditions were: 1 cycle of 95°C for 30 sec; 18 cycles of 95°C for 30 sec; 45°C for 1 min. (optimal annealing temperature was determined for each primer set), and 68°C for 9 min.; storage at 4°C. The 50 μl PCR reactions contained 100ng of each primer, 10 ng of the plasmid template, 0.2 mM dNTP, and 1× reaction buffer. pfuUltra II Fusion Hotstart DNA polymerase (Stratagene) was used to generate the PCR amplicons, which were digested with DpnI to remove contaminating template DNA, and then transformed into maximum efficiency XL2-Blue ultracompetent cells (>5 × 109 cfu/μg DNA; Stratagene) so that the DNA volume did not exceed 5% of the cell volume. The entire transformation was plated onto LB-Ampicillin agar plates, generally resulting in 10 to 50 colonies.

Colonies were inoculated into LB-Ampicillin broth for overnight cultures and the plasmid was prepared using the QIAprep Spin Miniprep Kit. Each plasmid was screened for biological function as previously described (Derdeyn et al., 2004; Li et al., 2006). Briefly, 600 ng of Env DNA was co-transfected into 293T cells along with 1200 ng of an Env-deficient subtype B proviral plasmid, pSG3ΔEnv, using Fugene-6 according to the manufacturer's instructions (Hoffman-La Roche). Seventy-two hours later, the transfection supernatant was transferred to JC53-BL (TZM-bl) indicator cells. At 48 hours post-infection, each well was scored positive or negative for blue foci using b-gal staining. For clones that produced functional Env pseudotypes, the plasmids were re-transfected into 293T cells on a larger scale to produce a working pseudotype virus stock. Transfection supernatants were collected at 48 hours post-transfection, clarified by low speed centrifugation for 20 minutes, aliquoted into 0.5 ml or less portions, and stored at −80°C. The titer of each pseudotyped virus stock was determined by infecting JC53-BL cells with 5-fold serial dilutions of virus as described previously (Derdeyn et al., 2004; Li et al., 2006). All env sequences were confirmed by nucleotide sequencing.

Virus neutralization and inhibition assays

Neutralization assays using sCD4, monoclonal 17b, and a pool of anti-V3 monoclonal antibodies derived from subtype C infected Indian patients were performed using viral pseudotypes to infect JC53-BL (Tzm-bl) indicator cells using a luciferase readout as described previously (Derdeyn et al., 2004; Li et al., 2006; Rong et al., 2007a; Rong et al., 2007b; Rong et al., 2009). A human IgG isotype control was used in all monoclonal antibody experiments. Briefly, 2000 IU of pseudovirus was incubated for 1 hour in DMEM + 10% FBS (Hyclone) + 40μg/ml DEAE-Dextran with serial dilutions of sCD4 or monoclonal antibody and 100μl was then added to the indicator cells for a 48 hour infection before being lysed and evaluated for luciferase activity. To pre-trigger Envs, the pseudovirus was incubated for one hour with either its IC50 of sCD4 or 100nM sCD4 if 50% inhibition was not achieved. After incubation for another hour with serial dilutions of 17b, 100μl of sCD4-17b-virus was added to JC-53BL cells for a 48 hour infection. The 17b monoclonal antibody (contributed by Dr. James Robinson) (Thali et al., 1993) and sCD4-183 (2-domain) (contributed by Progenics Pharmceuticals) (Garlick et al., 1990) were obtained from the NIH AIDS Research and Reference and Reagent Program, Division of AIDS, NIAID, NIH. The anti-CD4 antibody B13.8.2 was kindly provided by Dr. Quentin Sattentau (University of Oxford) (Sattentau, 1995). A pool of three monoclonal antibodies against the V3 domain isolated from Indian patients infected with subtype C viruses was contributed by Dr. Susan Zolla-Pazner.

Replication in CD4 and monocyte-derived macrophages (MDM) using an NL4.3 proviral cassette

A panel of Envs was subcloned into a replication competent NL4.3 backbone that has been described previously (Lohrengel et al., 2005; Neumann et al., 2005). This system is amenable to accepting diverse env genes, and facilitates substitution of virtually the entire coding region (only 36 and 6 amino acids at the N and C terminus respectively are derived from NL4.3). We have previously shown that patient-derived Envs retain their entry phenotype in comparison with the primary isolate, indicating that the NL4.3 backbone is neutral to Env entry properties (Neumann et al., 2005). Peripheral blood mononuclear cells (PBMC) were isolated from the whole blood of a normal, seronegative donor by ficoll-hypaque centrifugation. CD8-depleted PBMC cultures were prepared by negative selection using Dynabeads (Invitrogen). The CD4-enriched PBMC were cultured in complete RPMI for 3 days in the presence of 3mg/ml phytohemagglutinin (PHA) for activation prior to infection. Infected cultures were maintained in complete RPMI supplemented with 30 U/ml recombinant human IL-2 (Roche) for up to 10 days. Every two days, 200μl of supernatant was collected for p24 analysis (Perkin-Elmer), and this volume was replaced with fresh complete media with IL-2. MDM were prepared from whole blood as described in (Salazar-Gonzalez et al., 2009). Briefly, 3×10⁶ PBMC per well were subject to adherence for 2 hours in a 24-well plate, and non-adherent cells were aspirated off. The remaining cells were incubated in DMEM with 10% Giant Tumor Cell-Conditioned Media (GCT) (Irvine Scientific), 10% Human AB Serum (Sigma) and 50ng/ml recombinant human Macrophage Colony-Stimulating Factor (rhMCSF) (R&D Systems). After 3 days, the cells were washed 3 times with PBS and media was replaced with 800μl of Macrophage media (DMEM + 10% FBS (Hyclone) + 10% GCT + 50ng/ml rhMCSF). On day 6, 100,000 IU of virus was added in 300μl serum-free DMEM for 2 hours, and 500μl of Macrophage media was added for overnight infection. Every 3 days for 13 days after infection, 250μl of supernatant was collected for p24 analysis, and 1.5 ml of media was removed and then replaced with fresh media. Subtype B Envs NL4.3 and YU-2 and subtype C Env MJ-4 were cloned into the NL4.3 backbone and used as controls for these experiments. The infectious subtype C proviral clone MJ4 (contributed by Drs. Thumbi Ndung'u, Boris Renjifo, and Max Essex) (Ndung'u, Renjifo, and Essex, 2001) was obtained from the NIH AIDS Research and Reference and Reagent Program, Division of AIDS, NIAID, NIH.

Receptor-dependent pseudovirus entry assay

Pseudovirus entry assays were performed as described above except that a panel of HeLa cells (provided by Drs. Emily Platt and David Kabat, Oregon Health and Science University) stably transduced to express varying levels CD4 and CCR5 were used instead of JC53-BL cells and the Env-deficient subtype B proviral plasmid pNL4.3ΔEnv, kindly provided by Dr. Ron Collman, expressing a luciferase reporter gene was used instead of pSG3DEnv (which does not express a reporter gene) (Kabat et al., 1994; Platt et al., 1998). Briefly, transfection supernatants were normalized to p24 and equivalent amounts were added per well in each experiment to infect each cell line. The cell lines used were JC.53 (1.5×10⁵ CD4 and 1.3×10⁵ CCR5 molecules/cell), JC.10 (1.5×10⁵ CD4 and 2.0×10³ CCR5 molecules/cell) and RC.49 (1.0×10⁴ CD4 and 8.5×10⁴ CCR5 molecules/cell).

Statistical analysis

To compare groups, a non-parametric Wilcoxon signed rank test was used for paired comparisons. All analyses were performed using a two-tailed p-value in Graphpad Prism 4.0c, and p-values ≤ 0.05 were considered statistically significant.

Results

I309L was created in a representative panel of diverse subtype C Envs

The major goal of this study was to investigate the biological effects of mutating Ile 309 in Envs of HIV-1 subtype C, in which this residue is 99% conserved. We chose Leu because this amino acid occurs naturally, albeit in less than 1% of subtype C Envs, and has been associated with structural changes in V3 (Stanfield et al., 2006). We therefore selected a panel of seven Envs that were each cloned during acute/early infection from a different subject enrolled in the Zambia-Emory HIV Research Project (Table 1). The newly transmitted Envs were cloned within 129 days of the last seronegative test, with Envs from 185F, 221M, and 205F being within an estimated 48 days of infection. To broaden the relevance of this study, we also included four Envs cloned at a time point between 23 and 25 months after infection that had developed resistance against the contemporaneous autologous neutralizing antibody (Nab) pool ((Rong et al., 2009) and data not shown) (Table 1). These later Envs allowed us to evaluate the effects of the I309L mutation specifically within Envs known to have adapted to Nab immune pressure during natural infection. Supplemental Fig. 1 demonstrates that all of the Envs cluster phylogenetically as subtype C, and that their genetic diversity is representative of subtype C variants circulating across multiple geographic regions. This figure also shows that the newly transmitted Envs from subjects 205F, 153M, 185F, and 221M each cluster with the Nab escape variant from the same patient. The I309L mutation was created in the V3 domain of each of the 11 Envs (Fig. 1) and used in the biological studies described in the following sections.

Table 1. Subject Information.

Subject ID	Sample date	SGA	Days since seroneg.	Estimated days since infection	Nab escape variants
106F	8-Jun-98	no	129	ND	NT
55F	13-Aug-98	no	90	ND	NT
109F	16-Mar-00	no	96	ND	NT
153M	13-Mar-02	yes	88	ND	24-months
185F	17-Aug-02	yes	11	33	23-months
221M	7-Mar-03	yes	100	31	23-months
205F	27-Mar-03	yes	26	48	25-months

ND = not determined

NT = not tested

SGA = single genome amplification

Figure 1. Alignment of V3 sequences from subject Envs.

Amino acid sequences of the V3 domain from the 11 patient-derived Envs used in the mutagenesis experiments were aligned using SeqPublish (Los Alamos Database) and are shown in comparison to the consensus sequence for subtype C. The 7 Envs from an acute/early time point of infection are shown in red and the four Nab escape variants from a longitudinal time point two years post-infection are shown in blue. The consensus subtype C sequence is shown in black. Dashes indicate residues that are the same as the subtype C consensus sequence. A green box designates the I309 residue that was mutated to an L309 in each Env using site-directed mutagenesis.

The I309L mutation does not decrease replication in primary CD4 T cells

Purifying selection imposed upon I309 could indicate that changes at this position are not tolerated because they decrease replication capacity. We therefore evaluated whether the I309L change would be detrimental to replication in activated CD4 T cells. Ten of the eleven Envs, with and without I309L, were transferred into a replication competent NL4.3 proviral backbone and used to infect PBMC enriched for CD4 T cells by CD8 depletion. A replication competent virus containing185F 23-month EnvPL1.1 could not be generated using this system due to internal restriction sites. Supernatant was collected every two days and viral p24 was quantified for a measure of viral replication (Fig. 2A and Suppl. Fig. 2). The subtype C provirus MJ4 was used as a positive control for replication in CD4 T cells (Ndung'u, Renjifo, and Essex, 2001).

Figure 2. Replication of virus containing a wildtype and I309L mutated Envs in PBMC.

10 wildtype and 10 I309L mutated Envs were placed into a replication competent NL4.3 backbone and used to infect CD8-depleted human PBMC; representative graphs of 3 of these pairs are shown in panel (A). Viral p24 antigen production in the supernatant was measured by ELISA and is plotted (pg/ml) on the vertical axis on a log10 scale while days post infection are on the horizontal axis. Mock-infected (squares) and positive control virus NL4.3 (triangles) are shown in black. Wildtype Envs are blue and I309L mutant Envs are green. The error bars represent two independent experiments using two different donors. Total p24 at day 10 of infection is shown as a dot plot (B). The residue at position 309 is indicated on the horizontal axis, and a comparison of p24 antigen levels between I309 Envs (blue) and L309 (green) Envs was performed using Graphpad Prism. Each dot represents the mean p24 for one Env. The p-value was determined using a paired Wilcoxon test. The p24 level for the positive control virus NL4.3 is represented by a black arrow.

The acute/early subtype C Envs mediated replication with variable magnitude and kinetics, but day 10 represented the peak p24 level for the majority of the subtype C Envs (Fig 2A and Suppl. Fig. 2). Therefore, this time point was used to compare the overall replication between paired sets of wildtype and I309L mutant Envs. Each wildtype Env and its I309L mutant displayed comparable replication kinetics and p24 levels (Fig 2A and Suppl. Fig. 2), and there was no statistical difference in day 10 p24 levels between of the two groups (Fig. 2B). To ensure that the I309L mutation had not reverted, the V3 domain of each mutant virus was sequenced using virion-associated RNA in supernatant collected at day 10 during one experiment (data not shown). This result demonstrates that I309L-containing Envs are replication competent in activated CD4 T cells, and suggests that the strong conservation of I309 is not due to a detrimental effect on viral replication, at least as measured by this in vitro assay.

The I309L mutation leads to a moderate enhancement of replication in monocyte-derived macrophages

In addition to CD4 T cells, human monocyte-derived macrophages (MDM) can also serve as a target cell for HIV-1 in vivo. We therefore evaluated whether I309L altered in vitro replication in MDM cultures. For these experiments, the subtype B Env YU-2 was used as a positive control for replication in MDM, and NL4.3 was used as a negative control (Fig 3A and Suppl. Fig. 3) (Li et al., 1992). None of the subtype C Envs replicated as efficiently as the ‘control’ strain YU-2, which is consistent with our previous finding that subtype C Envs, while CCR5-tropic, do not infect MDM with high efficiency (Isaacman-Beck et al., 2009). Nevertheless, using p24 level at day 13 post-infection for comparison, the I309L Envs overall replicated to moderately higher levels than their wildtype counterparts (Fig. 3B; p=0.02). This enhanced ability to replicate in MDM cultures could be linked to augmented usage of low amounts of CD4 found on these cells, as compared to levels found on T-cells (Bannert et al., 2000; Collman et al., 1990; Duenas-Decamp et al., 2009; Thomas et al., 2007).

Figure 3. Replication of I309 and I309L Env viruses in MDM.

10 wildtype and 10 I309L mutated Envs were placed into a replication competent NL4.3 backbone and used to infect CD8-depleted human PBMC; representative graphs of 3 of these pairs are shown in panel (A). Viral p24 antigen production in the supernatant was measured by ELISA and is plotted (pg/ml) on the vertical axis on a log10 scale while days post infection are on the horizontal axis. Mock-infected is shown in black squares, negative control virus NL4.3 (triangles) and positive control virus YU-2 (diamonds) are shown in red, wildtype I309 Envs are blue and I309L mutant Envs are green. The error bars represent two independent experiments from two different donors. Total p24 at day 13 of infection is shown as a dot plot (B). The residue at position 309 is indicated on the horizontal axis, and a comparison of p24 antigen levels between I309 Envs (blue) and L309 (green) Envs was performed using Graphpad Prism. Each dot represents the mean p24 for one Env. The p-value was determined using a paired Wilcoxon test. The p24 level for the positive control viruses YU-2 (shown as a red dot) and the negative control virus NL4.3 are represented by black arrows.

The I309L mutation confers increased entry into a cell line expressing low CD4

We next investigated whether differences in infection efficiency would be observed in HeLa-based cell lines that express known quantities of CD4 and CCR5. JC.53 cells, which express high levels of CCR5 and CD4, were used to determine relative infectivity in two different cell lines: JC.10, which express low CCR5/high CD4, and RC.49, which express high CCR5/low CD4 (see methods) (Platt et al., 1998). These cell lines were infected with pseudoviruses created by expressing the wildtype or I309L mutant Envs with an HIV-1 env-deficient backbone that encodes luciferase. Under the condition of high CD4 and limited CCR5, there was no statistically significant difference between the wildtype and I309L Envs in ability to enter the JC.10 cells (Fig 4A). There was one pair of Envs from subject 221 for which the I309L did appear to decrease infectivity, although this effect may be due to an inherent property of the V3 loop (perhaps the D321, see Fig. 1) in these two Envs. Overall, however, infectivity of JC.10 cells was decreased by approximately 10-fold for all Envs, compared to the JC.53 cells (high CCR5, high CD4), suggesting that the wildtype and I309L Envs were equally dependent on CCR5 for entry into JC.10 cells. Dependence on CCR5 was confirmed using the HI-J cell line, which expresses CD4 and endogenous CXCR4 but no CCR5 (data not shown). When CD4 levels were reduced, however, the I309L containing Envs consistently had higher relative infectivity in the RC.49 cells, ranging from a 1.1- to an 18.6-fold increase over the matched wildtype Env (Fig. 4B; p=0.003). It should be noted that the infectivity of all Envs for RC.49 cells was generally about 1% of that achieved on the JC.53 cells. Thus, all of the Envs were highly dependent on CD4 levels, consistent with other studies by our group using patient-derived subtype C Envs from acute/early and chronic infection (Alexander et al.). The I309L mutation did not, therefore, “pre-trigger” the coreceptor binding domain, leading to CD4 independence, as has been described for subtype B Envs carrying non-consensus residues at position 309 (Decker et al., 2005; Zhang et al., 2002). Thus, although both I309L and wildtype Envs were inefficient at infecting cells with limiting levels of CD4, the I309L Envs consistently had higher relative infectivity, suggesting more efficient binding to CD4 and/or increased exposure of the CD4bs.

Figure 4. Lower dependence on CD4 but not CCR5 levels for entry by I309L Envs.

The infectivity of the 11 wildtype and 11 I309L mutated pseudotyped Envs were measured in HeLa cell-lines JC.10 with high CD4 and low CCR5 (A) and RC.49 expressing low CD4 and high CCR5 (B). The relative infectivity compared to JC.53 cells, which express high CD4 and high CCR5, was determined by provirally-encoded luciferase and is shown on the vertical axis. The I309 wildtype Envs are blue while the I309L mutated Envs are green. Each point represents the mean relative infectivity of an individual Env pseudovirus from 2 independent experiments. A paired Wilcoxon test was used to determine the p-value.

The I309L mutation confers increased sensitivity to sCD4

Patient-derived virus Envs are typically resistant to inhibition by the soluble form of the CD4 receptor, sCD4, reflecting the cryptic nature of the CD4 binding site (CD4bs) (Ashkenazi et al., 1991; Bures et al., 2002; Daar et al., 1990; Moore et al., 1993; Moore et al., 1992; Pugach et al., 2004; Willey, Martin, and Peden, 1994; Wu et al., 2009). However, the ability to infect MDM and RC.49 cells has been associated with increased sensitivity to sCD4 inhibition (Duenas-Decamp et al., 2009; Peters et al., 2008; Peters et al., 2006). Fig. 5 demonstrates that all of the wildtype Envs were relatively resistant to sCD4 out to 100 nM in a single round pseudovirus assay. The I309L substitution, however, led to an almost universal increase in sCD4 sensitivity, although the magnitude varied between Envs (Fig 5A). For 9 of the 11 Envs, I309L resulted in an increase in sensitivity compared against wildtype at the highest concentration of sCD4 (100nM). Overall, viral infectivity was significantly lower with the I309L mutation at both 100nM and 20nM concentrations of sCD4 (Fig. 5B; p=0.001 and p=0.005, respectively), suggesting a consistent increase in exposure of the CD4bs in the mutant Envs as compared to the wildtype Envs. Binding to CD4 was further assessed by incubating the pseudotyped Envs with an anti-CD4 monoclonal antibody B13.8.2, and subsequently calculating the concentration of antibody necessary to reduce viral infectivity by 50% (IC50). The I309L Envs required higher concentrations of anti-CD4 to inhibit infection (with a higher IC50) than wildtype, confirming their more efficient utilization of the CD4 receptor (Fig. 5C; p=0.01).

Figure 5. I309L mutation increases sCD4 sensitivity in multiple subtype C Envs.

The parent and wildtype Envs were assessed for sensitivity to sCD4 using pseudoviruses (A). Mutant I309L Envs are shown by green lines and wildtype Envs are shown by blue lines. Each line on the graph represents an individual Env pseudovirus. Percent viral infectivity compared to no sCD4 is shown on the vertical axis and was calculated from luciferase units by dividing virus-infected wells in the presence of inhibitor by the virus-infected well in the absence of inhibitor. The concentration of sCD4 (in nanomolar) is plotted on the horizontal axis on a log10 scale. Error bars represent the standard error of the mean of at least 2 independent experiments. Vertical point plots of the mean percent viral infectivity for each Env pseudovirus at 100nM and 20 nM sCD4 respectively were generated using Graphpad Prism (B) while the IC50 (μg/ml) of pseudovirus in the presence of anti-CD4 binding antibody B13.8.2 is shown in (C). I309 Envs in blue were compared to L309 Envs in green. A paired Wilcoxon test was performed to determine the p-value.

The I309L mutation moderately increased sensitivity to monoclonal antibodies directed against V3 and a CD4-induced epitope

If I309L induces conformationally based changes in gp120 that increase exposure of the CD4bs, we hypothesized that other Env domains could also be affected. Patient derived Envs, and in particular those of subtype C, are typically refractory to neutralization by anti-V3 antibodies (Binley et al., 2004; Davis et al., 2009; Keele et al., 2008; Moore et al., 2008; Salazar-Gonzalez et al., 2009; Wu et al., 2008). Therefore, a pool of three anti-V3 directed monoclonal antibodies from Indian patients infected with subtype C viruses were used to assess changes in susceptibility to V3-targeted neutralization. In 6 of the 11 Envs, there was an increase in sensitivity to anti-V3 mediated neutralization with the I309L mutation (Fig. 6A), suggesting a local unmasking of the V3 domain. Taken as a whole, when infectivity at the highest concentration of antibody (10μg/ml) was compared in a pairwise manner, there was a statistically significant difference between wildtype and mutant Envs (Fig. 6B; p=0.04).

Figure 6. I309L increases neutralization by anti-V3 antibodies.

Neutralization sensitivity to a pool of three monoclonal antibodies against V3 derived from subtype C infected subjects in India was assessed for each mutant I309L and wildtype Env (A). The data is shown as the percent viral infectivity (luciferase units in the presence of antibody divided by in the presence of control anti-parvovirus antibody) on the y-axis. The concentration of the monoclonal antibodies is shown on the x-axis on a log-10 scale. Mutant I309L Envs are shown by green lines and wildtype Envs are shown by blue lines. Each line on the graph represents an individual Env pseudovirus. Error bars represent the standard error of the mean of at least 2 independent experiments. A vertical point plot of the mean percent viral infectivity at 10μg/ml of the monoclonal pool was generated using Graphpad Prism (B). I309 Envs in blue were compared to L309 Envs in green. A paired Wilcoxon test was performed to generate the p-value shown.

We next investigated whether the I309L Envs were also more susceptible to sCD4-triggering of the coreceptor binding site, which contains an epitope recognized by the monoclonal antibody 17b (Thali et al., 1993). The Envs were pre-incubated with their IC50 of sCD4, or 100nM if an IC50 was not determined, and then neutralization by 17b was evaluated. Four of the 11 Envs showed a moderate increase in neutralization sensitivity to 17b after incubation with sCD4 (Fig. 7A). Overall the difference between the wildtype and I309L Envs reached the borderline of statistical significance in a paired analysis (Fig. 7B; p=0.05), suggesting that this mutation did also affect efficiency with which the coreceptor binding site was formed, albeit in a context dependent manner.

Figure 7. I309L slightly increases neutralization of pre-triggered Envs by an anti-coreceptor binding site antibody 17b.

Mutant I309L and wildtype Envs pseudoviruses were incubated with their cognate IC50 nM of sCD4 (100nM of sCD4 if 50% inhibition was never reached), and then neutralization sensitivity was assessed by monoclonal antibody 17b (A). The data is shown as the percent viral infectivity (luciferase units in the presence of antibody divided by in the presence of control anti-parvovirus antibody) on the y-axis. The concentration of 17b is shown on the x-axis on a log-10 scale. Mutant I309L Envs are shown by green lines and wildtype Envs are shown by blue lines. Each line on the graph represents an individual Env pseudovirus. Error bars represent the standard error of the mean of at least 2 independent experiments. A vertical point plot of the mean percent viral infectivity at 2μg/ml of 17b was generated using Graphpad Prism (B). Wildtype Envs in blue were compared to I309L Envs in green while the mean infectivity for each group is represented by the black horizontal line. A paired Wilcoxon test was performed to generate the p-value shown.

Discussion

In sequences representative of established infection, the subtype C V3 domain is more conserved than other subtypes. Consistent with this observation, we recently demonstrated that V3 remained remarkably conserved during the first two years of infection despite increasing sequence variation in V1V2, the gp120 outer domain, and gp41 (Rong et al., 2009). In the present study, we investigated the underlying biological basis for the conservation of V3 by focusing specifically on a residue that can exhibit high entropy in subtypes other than C. While this and other subtype-specific mutational patterns in V3 have been well established (Gaschen et al., 2002; Gilbert, Novitsky, and Essex, 2005; Korber et al., 1994; Patel, Hoffman, and Swanstrom, 2008; Rong et al., 2007b; Stanfield et al., 2006), this study is among the first to delve into the biology that drives these observations.

Replication is not dependent on conservation of I309

As Ile is the highly conserved consensus residue for position 309 in subtype C, we evaluated whether a conservative substitution of Leu would lead to any defect reflected in the capacity for in vitro replication. Replication and spread in two primary target cells, CD4 T cells and MDM, was not overtly reduced by the I309L substitution. Although, it should be noted that we could not rule out subtle differences in replication kinetics between the wildtype and I309L mutants that might have been evident in a more sensitive growth competition assay. Overall, the subtype C Envs replicated to higher levels in CD4 T cells compared to MDM, suggesting that none of the wildtype Envs were inherently macrophage-tropic. This finding is consistent with our previous results in which subtype C Envs from recently and chronically infected subjects mediated single round infection of MDM much less efficiently than CD4 T cells (Isaacman-Beck et al., 2009). Low efficiency replication in MDM compared to CD4 T cells has been attributed to an inability of patient-derived Envs to utilize the low CD4 levels expressed by MDM (Bannert et al., 2000; Collman et al., 1990; Duenas-Decamp et al., 2009; Thomas et al., 2007). The moderate enhancement of replication in MDM when I309L was present suggests that this mutation could have altered the viral interaction with CD4. Indeed, a majority of the I309L mutant Envs more efficiently infected a cell line expressing a low level of CD4, focusing the effects of the mutation on utilization of CD4. Taken together, these results suggest that subtype-specific conservation of I309 is not driven by a detrimental effect on replication in CD4 T cells or MDM. Rather, this finding is consistent with the idea that Env residues that are critical for viral entry, like the GPG motif in V3, are conserved across all subtypes (Freed, Myers, and Risser, 1991). Thus, subtype-specific conservation of I309 is more likely driven by properties that differ between subtypes, such as the antigenic nature of V3 and other Env regions.

Conservation of I309 prevents exposure of neutralization targets

sCD4 sensitivity was used in this study as a surrogate for exposure of the CD4bs, and all but two Envs (both from subject 185F) became more sensitive to sCD4 when I309L was introduced. This finding supports the idea that V3, specifically I309, modulates exposure of the CD4bs, and is consistent with data from other studies (Agrawal-Gamse et al., 2009; Hwang et al., 1992; Lynch et al., 2009; Wyatt et al., 1992). Increased exposure of the CD4bs does not necessarily indicate that these Envs would be more sensitive to CD4bs-directed antibodies, however, a monoclonal targeting this region and capable of neutralizing these subtype C Envs was not available to test this hypothesis. However, increased sCD4 sensitivity was consistently observed, and this was accompanied by increased neutralization via V3 and CD4-induced epitopes to varying degrees. It has been shown that HIV-1 infected patients develop high titers of anti-CD4 induced and anti-V3 antibodies, and more variable titers of anti-CD4bs antibodies (Binley et al., 2008; Davis et al., 2009; Decker et al., 2005; Gray et al., 2009; Li et al., 2009; Moore et al., 2008; Scheid et al., 2009). It is plausible that I309L Envs could be more susceptible to these types of antibodies in vivo, and would therefore be selected against by immune pressure. It is important to note that we were unable to detect an increase in sensitivity to autologous patient plasma with I309L (data not shown). However, only the dominant circulating Nab specificities are measured in assays that utilize patient plasma, while other lower titer specificities may also exert selective pressure (Blish et al., 2009; Moore et al., 2009; Rong et al., 2009; Scheid et al., 2009). Alternatively, I309 could protect immunogenic structures so effectively that antibodies against these hidden epitopes are never induced. Since the number and potency of monoclonal antibodies available to probe differences between wildtype and I309L Envs was limited, the effect of I309L on neutralization sensitivity could be underestimated. Thus, even a subtle increase in susceptibility to neutralization by sCD4, anti-V3, and anti-CD4i monoclonal antibodies in vitro could reflect a substantial growth disadvantage in vivo. In fact, the differences in neutralization sensitivity between parent and mutant Envs were subtle, trending toward significance, but really driven by the Env from subject 55F. The impact of the L309 mutation was dramatic in this Env, a fact that may result from the unusual M307 residue (see Fig. 1) and highlights how many changes in Env sequence are context dependent and rarely result in a globally reproducible phenotype. We also acknowledge, that over-all neutralization sensitivity and immune evasion is clearly influenced by multiple Env domains, as shown by our group and others (Blish, Nguyen, and Overbaugh, 2008; Moore et al., 2008; Moore et al., 2009; Pinter et al., 2004; Rong et al., 2007a Rong et al., 2009; Wei et al., 2003).

V3-mediated masking of the CD4 binding site

We previously highlighted a cluster of hydrophobic residues flanking the V3 tip (anchored by residues I307, I309, and F317) that are highly conserved in subtype C sequences in the database and are largely restricted to hydrophobic amino acids across subtypes (Gilbert, Novitsky, and Essex, 2005; Lynch et al., 2009; Rong et al., 2007b; Stanfield et al., 2006). Coarse-grained calculations suggested that this hydrophobic cluster has the potential to interact with multiple gp120 core residues, several of which are proximal to the CD4 binding site and may impact CD4 binding (Lynch et al., 2009). Based on these previous observations, and our current experimental results, we propose that the I309L substitution perturbs the hydrophobic cluster and, by altering the position of the V3 loop, impacts the CD4bs. Thus, for subtype C viruses, maintaining this residue could represent an intrinsic strategy of immune evasion, particularly concerning Nab recognition of the CD4bs. Consistent with this idea, I309 was not critical for viral replication or usage of CCR5, but its major effect was manifested in exposure of the CD4bs prior to receptor ligation.

Thus, V3 in subtype C Env could be placed in an optimal position to contribute to immune evasion prior to CD4 binding and additionally aid viral entry after CD4 binding. This theory provides a biologically plausible explanation for the I309L CD4 enhanced phenotype described here, as well as a compelling reason for the unique conservation pattern of specific residues within the subtype C V3 domain that may not be critical for viral entry functions. The idea of conformational masking of neutralization targets is not new; and others have demonstrated concealment of epitopes on the native trimer (Binley et al., 2008; Davis et al., 2009; Moore et al., 2008; Wu et al., 2008; Wu et al., 2009). It is important to note that this tactic of V3-masking could also occur in other subtypes, but the specific mechanism may subtly differ. Thus, this specific feature of subtype C Envs may limit their inherent neutralization susceptibility to certain types of antibodies and could also influence their immunogenicity, which is of interest for vaccine design.