Genetic Signatures of HIV-1 Envelope-mediated Bystander Apoptosis

Background: Determinants of HIV-1 Env-mediated apoptosis remain poorly understood. Results: We studied the bystander apoptosis-inducing activity of a panel of primary HIV Envs. Conclusion: Residues Arg-476 and Asn-425 are associated with differences in HIV-1 Env-mediated bystander apoptosis induction. Significance: We identified specific genetic signatures within the HIV-1 Env that are associated with the bystander apoptosis-inducing phenotype. The envelope (Env) glycoprotein of HIV is an important determinant of viral pathogenesis. Several lines of evidence support the role of HIV-1 Env in inducing bystander apoptosis that may be a contributing factor in CD4+ T cell loss. However, most of the studies testing this phenomenon have been conducted with laboratory-adapted HIV-1 isolates. This raises the question of whether primary Envs derived from HIV-infected patients are capable of inducing bystander apoptosis and whether specific Env signatures are associated with this phenomenon. We developed a high throughput assay to determine the bystander apoptosis inducing activity of a panel of primary Envs. We tested 38 different Envs for bystander apoptosis, virion infectivity, neutralizing antibody sensitivity, and putative N-linked glycosylation sites along with a comprehensive sequence analysis to determine if specific sequence signatures within the viral Env are associated with bystander apoptosis. Our studies show that primary Envs vary considerably in their bystander apoptosis-inducing potential, a phenomenon that correlates inversely with putative N-linked glycosylation sites and positively with virion infectivity. By use of a novel phylogenetic analysis that avoids subtype bias coupled with structural considerations, we found specific residues like Arg-476 and Asn-425 that were associated with differences in bystander apoptosis induction. A specific role of these residues was also confirmed experimentally. These data demonstrate for the first time the potential of primary R5 Envs to mediate bystander apoptosis in CD4+ T cells. Furthermore, we identify specific genetic signatures within the Env that may be associated with the bystander apoptosis-inducing phenotype.

this HIV-induced CD4 ϩ T cell depletion like direct cell killing due to infection (1), bystander apoptosis (2), immune activation (3), immune exhaustion (4,5), etc. Several lines of evidence suggest that CD4 ϩ T cells undergo apoptosis in HIV infections (6 -8). A majority of these cells are shown to be uninfected, lying in close proximity to HIV-infected cells (9), suggesting a role for bystander apoptosis in this pathology. The phenomenon of bystander apoptosis has been demonstrated in both SIV 2 (4,10) and the humanized mouse model (11) of HIV infection, and a role of viral proteins from infected cells mediating apoptosis in bystander cells has been proposed. Among the viral proteins, the Env glycoprotein remains as the frontrunner for mediating bystander apoptosis because it is expressed on the surface of infected cells and can interact selectively with bystander CD4 ϩ T cells (2,12,13).
HIV Env glycoprotein is the most dynamic viral protein that constantly evolves throughout the course of the disease (5, 14 -16). The high rate of Env evolution and variability within the HIV-infected population (17) has stimulated a number of studies aimed at understanding the polymorphic nature of the Env glycoprotein (5, 18 -20). One of the best studied phenomena associated with HIV Env is the evolution of virus from a CCR5-utilizing phenotype (non-syncytia-inducing) early during the infection to a CXCR4-utilizing phenotype later during the disease (15,21,22). This change in co-receptor use has been associated with evolution of the virus to a syncytia-inducing phenotype, leading to rapid decline in CD4 counts and accelerated disease progression (23). However, co-receptor switch is not an absolute requirement for disease progression, and in ϳ50% of the patients evolution of virus to X4 usage does not occur prior to AIDS development or terminal disease (24,25). Interestingly, these AIDS-associated R5 viruses have been shown to be more fusogenic than early asymptomatic phase viruses (26,27). Furthermore, the evolution of virus toward X4 usage has also been variable between different subtypes of HIV with lower rates of X4 tropism seen in subtype C viruses (28). These findings emphasize the importance of the viral Env in regulating the course of disease with more fusogenic Envs being potentially more deleterious to CD4 ϩ T cells versus the less fusogenic ones.
HIV disease progression is a complex interplay between several viral and host factors. The host immune system in most cases can suppress the virus to undetectable set point/levels after an acute phase. The CD8 cytotoxic T cell response (29,30) and the neutralizing antibody production are largely responsible for the viral control leading into the chronic phase. During the chronic phase, the highest selection pressure due to neutralizing antibodies is on the Env glycoprotein of the virus (31), which undergoes rapid evolution via changes in the pattern of N-glycosylation and charge on the Env glycoprotein (32). Recently, several genetic signatures have been identified that can distinguish viruses from different stages of the disease that include but are not limited to changes in putative N-glycosylation sites (PNGS) (18,(32)(33)(34)(35). Moreover, during the course of intrapatient HIV-1 evolution, it is also documented that the viral Env acquires mutations that alter CD4 binding and/or Env fusogenicity (36 -39). However, how this variability in HIV Env affects bystander apoptosis and virus pathogenesis remains undetermined.
We and others have extensively studied the phenomenon of bystander apoptosis mediated by HIV Env both in vitro (40 -43) and in vivo (11) in a humanized mouse model of HIV infection. Our studies demonstrate that Env fusogenic activity determines bystander apoptosis and CD4 decline but not virus replication (11). Moreover, with respect to R5 viruses, using cell lines that express different levels of CCR5, we have shown that cell surface CCR5 expression levels determine Env-mediated bystander apoptosis (44). Although experimental data from us and others demonstrate that bystander apoptosis can be induced by HIV Env in vitro, the clinical implication of these findings is limited due to a lack of data from primary HIV isolates and patient-derived Env clones.
In this study, we developed a high throughput assay to study a panel of primary patient-derived HIV Envs for their bystander apoptosis-inducing activity and determined its correlation with both phenotypic and genotypic characteristics of the Envs. These included virus infectivity, sensitivity to neutralizing antibodies, putative N-glycosylation sites, Env charge, and association of specific residues in the Env with bystander apoptosis. Our study demonstrates that although primary R5 Envs vary considerably in their bystander apoptosis-inducing potential, the phenomenon correlates inversely with PNGS and to some extent neutralizing antibody sensitivity. Moreover, using computational sequence and structure analysis, we identified specific Env signatures associated with differential bystander apoptosis-inducing phenotypes. Specific residues associated with these signatures were confirmed by site-directed mutagenesis experiments. These findings not only corroborate the role of viral Env in bystander apoptosis but take our previous studies a step further in attempting to identify specific Env signatures associated with bystander apoptosis. To the best of our knowledge, this is the first study aimed at studying the bystander apoptosis-inducing activity of a large panel of primary Envs utilizing a high throughput assay and correlating this phenomenon with specific Env signatures.

EXPERIMENTAL PROCEDURES
Cell Lines and Transfections-SupT1 cells were maintained in RPMI medium supplemented with 10% FBS and penicillin streptomycin (5000 units/ml). SupT1 cells expressing CCR5 were generated by transduction of wild type SupT1 cells with a lentiviral vector expressing the CCR5 gene and have been described previously (44). These cells were maintained in RPMI medium supplemented with 10% FBS and penicillin streptomycin (5000 units/ml) and blasticidin at a concentration of 3 g/ml. 293T, HeLa, and TZM-bl cells (National Institutes of Health AIDS Research and Reference Reagent Program) were maintained in Dulbecco's modified Eagle's medium supplemented with 10% FBS and penicillin/streptomycin (5000 units/ ml). TZM-bl are HeLa-derived cells that express the HIV receptor CD4 and the co-receptors CXCR4 and CCR5 along with the luciferase and ␤-galactosidase genes under the control of HIV LTR (45). These cells thus readily support HIV infection/replication. All transfections were conducted using the Exgen 500 transfection reagent (Fermentas) following the manufacturer's instructions.
Env Constructs-Reference panels for different subtype Env clones were obtained from the NIH AIDS research and reference reagent program (supplemental Table 1). These include the HIV-1 subtype A/G Env clones (catalog no. 11673), (46) the reference panel for subtype B HIV-1 Env clones (catalog no. 11227) (35,47,48), and clade C HIV-1 reference panel of Env clones (catalog no. 11326) (49,50). These constructs include different patient-derived primary Envs as insert in the pcDNA3.1 expression vector (Invitrogen). For the purpose of comparison and to be used as normalizing control, we used a laboratory-adapted R5-utilizing HIV-1 clone, YU-2. The YU-2 Env was thus similarly cloned into the pcDNA3.1 directional Topo expression vector (Invitrogen) and contains the open reading frames for the env and rev genes. Point mutations were introduced into different Env constructs using specific mutagenic primers and the QuikChange site-directed mutagenesis kit (Stratagene). The introduced mutations were confirmed by sequencing. For detection of Env expression, HeLa cells transfected with the various Envs were cultured in growth medium devoid of methionine and cysteine and supplemented with [ 35 S]Met/Cys. Thereafter, cell lysates were immunoprecipitated with HIV-Ig (kindly provided by the NIH AIDS Research and Reference Reagent Program) resolved by SDS-PAGE followed by PhosphorImager analysis.
High Throughput Assay for Bystander Apoptosis-HeLa cells were seeded in 96-well plates at 7500 cells/well and transfected with various Env constructs using the Exgen 500 transfection reagent. The next day, SupT-R5-H6 cells were added at 40,000 cells/well. The cells were co-cultured for 24 h, following which caspase-3 substrate was added directly to the wells (Caspase-Glo 3/7 assay, Promega). The plates were incubated for 30 min at 37°C and read on a luminescence plate reader (Fluostar Omega, BMG LabScience). The Caspase-Glo 3/7 assay is a luminescence-based assay that directly measures caspase-3/7 activity in cultures. Apoptosis for each Env was calculated as the percentage of YU-2 Env-mediated apoptosis after subtraction of the background derived from pcDNA3.1 empty vectortransfected cells.
To determine if bystander apoptosis induction by primary Envs requires CCR5 binding or gp41-mediated membrane fusion, the gp41 inhibitor enfuvirtide (used at a concentration of 2 M) or CCR5 inhibitor maraviroc (used at a concentration of 1 M) was added at the time when SupT-R5-H6 were added to each well. After an overnight incubation, the cultures were assayed for caspase-3 activity as indicated above.
Apoptosis Assay-For the standard flow cytometry-based apoptosis assay, Env-transfected HeLa cells were co-cultured with CCR5-expressing SupT-R5-H6 cells followed by the classical method of apoptosis detection via annexin V staining. HeLa cells transfected with different Env constructs were seeded in 24-well plates at a concentration of 10 5 cells/well. The cells were allowed to adhere for 4 -6 h. Subsequently, the medium was removed, and SupT-R5-H6 cells were added at a concentration of 0.5-1 ϫ 10 6 cells/well. The cells were co-cultured for 24 h, following which the suspension cells were carefully collected, stained with annexin V (BD Biosciences), and analyzed by flow cytometry using a Gallios flow cytometer (Beckman Coulter). At least 10,000 events were collected and analyzed using Flow Jo software (Tree Star).
Virus Infectivity-293T cells were transfected with the pNLLuc-RϪ/EϪ HIV backbone along with different Env constructs. Virus stocks were harvested 48 h post-transfection and used to infect the indicator TZM-bl cell line in the presence of 20 g/ml DEAE-dextran (Sigma). Luciferase activity was determined 72 h postinfection using the BriteLite plus luciferase assay substrate (PerkinElmer Life Sciences). Infectivity for each Env was calculated as a percentage of YU-2 Env control after subtraction of the background derived from pcDNA3.1 empty vector-transfected cells.
Neutralizing Antibody Sensitivity-pNLLuc-RϪ/EϪ virus stocks pseudotyped with different envelopes were prepared as described above. Virus stocks were then added to different concentrations of the TriMab antibody mix (IgGb12, 2F5, and 2G12) starting with a concentration of 25 g/ml. The virus antibody mix was incubated at 37°C 1 h and subsequently added to TZM-bl cells. Infection was determined 48 h postinfection by measuring luciferase activity in the cultures. Infectivity curves of each Env were fitted using Sigma Plot software, and IC 50 values were calculated from the curves.
Identification of Bystander Apoptosis Determinants in gp120-Nine pairs of genetically similar high-low apoptosis strains were chosen based on a phylogenetic tree generated using the maximum likelihood method with ␥-distributed evolutionary rates (shape parameter, 0.4327) using MEGA5 (51). Mutations between these nine pairs were compared for distinctive changes in their physiochemical properties (positive charge, negative charge, aromatic groups, aliphatic groups, prolination, glycosylation, size) between the two groups. Sites that were found in at least three pairs of high-low apoptosis strains to possess distinctive differences between the two groups were considered as candidates for determinants of bystander apoptosis. We examined each of these sites in crystal structures (Protein Data Bank entries 3U2S, 2QAD, and 2B4C) (52-54) for possible structural effects that they may have in altering the function of gp120. Protein Data Bank structure 3U2S is the crystal structure of an antibody (PG9 Fab) in complex with the V1V2 region from HIV-1 strain ZM109 (54). 2QAD is the structure of tyrosinesulfated 412d antibody in complex with HIV-1 YU2 strain GP120 and CD4 (52). The structure of 2B4C is that of HIV-1 JR-FL strain gp120 containing the V3 loop region in complex with CD4 and the X5 antibody (53). Sites where the mutation was predicted to affect CD4 binding were selected for mutagenesis experiments. All crystal structures were rendered using Yasara (55).
Analysis of Charge, PNGS, and Variable Loop Lengths-The variable loop regions of gp120 are V1(6615-6692), V2(6693-6812), V3(7110 -7217), V4(7377-7478), and V5(7596 -7637), numbered according to the HxB2 nucleotide sequence. V1V2 length is the combined peptide length of the V1 and V2 loop regions. A site with the sequence motif NX(S/T)X, where X represents a non-proline amino acid, is used to predict PNGS. Molecular models of gp120 were generated with MODELLER (56), using Protein Data Bank entry 2B4C (53) as a template. These models were used as inputs for the PROPKA server (version 3.1) (57) to estimate the pK a values of the amino acids for the calculation of charges at pH 7.0. Sequences were aligned using MAFFT, and the resulting alignment was used for the PNGS prediction, charge, and loop length calculations and the search for molecular determinants of bystander apoptosis (58).

Development of a High Throughput
Assay for HIV-1 Env-mediated Bystander Apoptosis-Several groups, including ours, have demonstrated unequivocally that the Env glycoprotein expressed on the surface of infected cells can induce apoptosis in uninfected bystander CD4 ϩ T cells (41,59,60). Classical methods of apoptosis detection like annexin V staining, TUNEL assay, etc. have been used for studying this phenomenon. Previously, we have shown that determination of caspase-3 activity in cells correlates well with bystander apoptosis induction by HIV Env (41,44). Building on this information, we asked whether this phenomenon could be developed as a high throughput assay for determination of bystander apoptosis-inducing activity of a panel of primary HIV Envs derived from patients. For this, HeLa cells were transfected with various primary HIV Envs in a 96-well format. The next day, SupT-R5-H6 cells expressing CD4 and CCR5 (44) were added to the transfected wells. After an overnight incubation, the cells were analyzed for caspase-3 activity using a luminescence substratebased assay (Caspase-Glo 3/7 assay, Promega). A schematic representation of the assay is shown in Fig. 1A. Because all manipulations are done in a single 96-well plate, the assay provides a rapid and easy method for analysis of bystander apoptosis-inducing activity of different Envs. We optimized the assay for maximum signal/noise ratio by using different ratios of effector (HeLa cells transfected with HIV-1 YU-2 Env) and target (SupT-R5-H6) cells. As seen in Fig. 1B, the highest signal/ noise ratio was seen for 7500 effector and 40,000 target cells/ well. We next asked whether the assay can accurately reproduce variability in bystander apoptosis-inducing potential of various HIV-1 Envs from laboratory-adapted strains. We cloned the Env and Rev region of a number of laboratory strains of HIV like YU-2, AD-8, JRCSF, Bal, and 89.6. We found that the assay showed differences in bystander apoptosis-inducing activity of these Envs with high level of reproducibility (Fig. 1C). Furthermore, the high throughput assay correlated well with the classical method of apoptosis detection via annexin V or 7-aminoactinomycin D staining followed by flow cytometry (Fig. 1C) used by us in the past. Interestingly, the caspase assay was more sensitive at detecting bystander apoptosis of Env variants that are low inducers of apoptosis. This is most likely a difference due to measurement of the percentage of apoptotic cells in annexin V and 7-aminoactinomycin D staining compared with total caspase-3 activity in the high throughput assay. Moreover, H6 cells but not peripheral blood mononuclear cells (PBMCs) are best suited for this high throughput assay due to a limited availability of CD4 ϩ CCR5 ϩ cells in the peripheral blood, which makes apoptosis detection via R5 isolates like YU-2 difficult (Fig. 2). Hence, our optimized assay could be used for studying a large panel of primary HIV-1 Envs with a high degree of reproducibility.
Primary Envs Vary Considerably in Their Bystander Apoptosis-inducing Activity-Using the assay developed above, we next asked whether there were differences in the bystander apoptosis-inducing activity of a panel of primary patient-derived HIV-1 Envs. We obtained the reference panel of HIV Envs from the National Institutes of Health AIDS Reference Reagent Program, representing subtypes B, C, and AG. A total of 37 primary Envs were tested with YU-2 Env, a laboratory-adapted R5 strain, as a normalizing control and the parent vector pcDNA3.1 as the negative control. Apoptosis was referenced as the percentage of WT or YU-2 after subtracting vector control from each sample. As seen in Fig. 3A, there was a wide variation in the bystander apoptosis-inducing activity of the primary HIV-1 Envs. Based on this variability, we classified the Envs into three groups: high (H) apoptosis-inducing, with the percentage of apoptosis more than 40; low (L) apoptosis-inducing, with the percentage of apoptosis less than 20; and medium (M) apopto-FIGURE 1. Development and validation of high throughput assay for HIV-1 Env-mediated bystander apoptosis. A, flow chart depicting the procedure for detection of apoptosis mediated by different HIV Envs via the caspase-3-based apoptosis detection assay. B, optimization of the high throughput assay for bystander apoptosis. Different numbers of HeLa cells were transfected with the YU2-Env clone expressing the Env and Rev proteins in a 96-well plate. The next day, the indicated numbers of SupT-R5-H6 were added to the transfected wells. After an overnight incubation, the cells were analyzed for caspase-3 activity using a luminescence substrate-based assay (Caspase-Glo 3/7 assay, Promega). Bars represent means of duplicate wells. One representative of three independent experiments is shown. C, comparison of the high throughput caspase-3 based assay for bystander apoptosis detection with the classical method of apoptosis detection via annexin V staining or by 7-aminoactinomycin D staining. 7500 HeLa cells were transfected with HIV-1 Envs from different laboratoryadapted strains of HIV-1 like YU-2, AD-8, JRCSF, Bal, and 89.6. The next day, 40,000 SupT-R5-H6 cells were added to each well. After an overnight incubation, the cells were analyzed for caspase-3 activity using the Caspase-Glo 3/7 assay. In parallel, HeLa cells were transfected with the above Envs in a 6-well plate. After an overnight incubation, the target cells were harvested and stained for surface phosphatidylserine expression using the annexin V staining kit (BD Biosciences) or by use of cell viability stain 7-aminoactinomycin D (7AAD) (BD Biosciences) followed by flow cytometry. Data are mean Ϯ S.D. (error bars) from triplicate observations. sis-inducing, between 20 and 40% apoptosis (Fig. 3B). When classified based on subtypes, there was no statistically significant difference between the bystander apoptosis-inducing activity of different Envs (Fig. 3C). All of the Envs used in the study were expressed as detected after immunoprecipitation using pooled sera from HIV-infected individuals (Fig. 3D). This suggests that although primary Envs vary in their bystander apoptosis-inducing activity, this phenomenon may be universal for the subtypes tested. The lack of differences between the subtypes may be due to the selection of the reference panel to represent the most diverse Envs from each subtype. Hence, overrepresentation of certain phenotypes in the population cannot be ruled out.
Bystander Apoptosis-inducing Activity of Primary Envs Correlates with Virion Infectivity-Because both virion infectivity and apoptosis induction are dependent on Env function, we next asked whether there were differences in the virion infectivity of the Envs studied. We used a pseudotyped virus particle approach to determine the single round infectivity of the Envs tested. Here again we found that there was a huge variation in the infectivity of the Envs tested (Fig. 4A). We found that high apoptosis-inducing Envs were more infectious compared with low apoptosis (p Ͻ 0.05) (Fig. 4B), although subtype-based classification did not show significant differences between the Envs (Fig. 4C). This is not surprising because both virion infectivity and bystander apoptosis are dependent on the Env fusion activity. Although as a group, virus infectivity correlated with bystander apoptosis, this was not necessarily true for each Env. For instance, we found a number of Envs that showed relatively low apoptosis but very high infectivity and vice versa (Fig. 4D). Hence, the spectrum of Envs studied here also includes a number of Envs that may be highly infectious without causing significant apoptosis. This is consistent with several mutagenesis studies whereby point mutations in HIV Env significantly alter bystander apoptosis while having a limited effect on virion infectivity (40,61).
Bystander Apoptosis-inducing Activity of Primary Envs Correlates Negatively with PNGS-One of the characteristics of HIV evolution in patients over the course of the infection is acquisition of variations in PNGS (35). It has been demonstrated that PNGS tend to increase during the course of infection, with the highest PNGS seen during chronic infection (33) associated with a minimal decline in CD4 T cell counts. The PNGS have been shown to decrease at the late stages of the disease, where there is an accelerated CD4 cell decline (32). Because we hypothesize that bystander apoptosis mediated by Env may be one of the contributing factors in CD4 T cell decline, we asked whether PNGS in Env correlate with bystander apoptosis. PNGS were estimated using the sequence information for each Env and the N-Glycosite software available at the LANL database (Fig. 5A). When the Envs were classified based on high, medium, or low bystander apoptosis activity and PNGS were determined for the Env region (gp160) in each group, there was a significant difference in PNGS between FIGURE 2. Suitability of H6 cells for the high throughput assay. A, PBMCs were obtained from Stem Cell Technologies, and CD4 ϩ T cells were isolated using the Human CD4 ϩ T Cell Enrichment Kit (Stem Cell Technologies). HeLa cells were transfected with the indicated Env clones or pcDNA3.1 control vector. The next day, phytohemagglutinin-activated purified CD4 ϩ PBMCs were added to the transfected wells, incubated overnight, and analyzed for caspase-3 activity using the Caspase-Glo 3/7 assay as described in the legend to Fig. 1. B, in parallel, the assay described in A was also conducted with the H6 cell line. C, bright field analysis of syncytia formation in cultures in A and its inhibition via use of AMD-3100 or maraviroc. D, purified CD4 ϩ T cells isolated from PBMCs were stained using anti-CXCR4 and anti-CCR5 antibodies followed by flow cytometry. RLU, relative light units. Error bars, S.D. the high and low apoptosis groups (Fig. 5B). The high apoptosis group shows significantly lower PNGS compared with the low apoptosis group (p Ͻ 0.05). We also determined whether there were differences between PNGS in the gp120 and gp41 region. Interestingly, most of the differences in PNGS were seen in the gp120 region (Fig. 5C) and not gp41 (Fig. 5D). This suggests that adaptation of the virus via an increase in PNGS may be counterproductive to bystander apoptosis induction. However, when the Envs were grouped on the basis of subtype, no significant differences were seen (Fig. 5, E-G). Once again, this may be due to the diverse set of Envs selected for the reference panel.
Neutralizing Antibody Sensitivity of Primary Envs Correlates with Bystander Apoptosis Induction-Primary Envs derived from HIV-infected patients are known to vary considerably in their sensitivity to neutralizing antibodies (62). Interestingly, the sensitivity to broadly neutralizing antibodies also varies with the stage of the disease as well as PNGS, with end stage viruses being more sensitive and having lower PNGS (32). Because we saw a strong correlation between PNGS and apoptosis induction, we next asked whether there was a correlation between bystander apoptosis activity of primary Envs and sensitivity to broadly neutralizing antibodies. For this purpose, we  Fig. 1. The YU-2 Env (100%) was used as a normalizing control, whereas cells transfected with empty pcDNA3.1 vector were used as background, which was subtracted from all wells for data analysis. A, the graph represents pooled data from three independent experiments with each well assayed in triplicate in each experiment. B, classification of the apoptotic potential of each primary Env into high, medium, or low apoptosis inducers based on the caspase assay data from A. C, classification of the apoptotic potential of each primary Env based on the virus subtype. D, detection of Env expression via HIV Ig-mediated immunoprecipitation. Error bars, S.D.
used a mixture of the three well characterized broadly neutralizing antibodies 2F5, 2G12, and IgGb12 commonly referred to as TriMab mix. We utilized HIV particles pseudotyped with the panel of HIV Envs to infect the TZM-bl target cell line in the presence of serial dilutions of the TriMab antibody mix (Table  1). Interestingly, the low apoptosis-inducing Envs showed a trend toward being more resistant to inhibition via neutralizing antibodies (Table 1 and Fig. 6A). In this analysis, the Envs that showed no inhibition to the TriMab mix were excluded because an IC 50 could not be calculated. Interestingly, when classified based on virus subtype, the non-BC (comprising largely of AG subtype) type isolates showed higher resistance to neutralization when compared with the B subtype (Fig. 6B). It is worth noting that subtype B has been shown to be more susceptible to broadly neutralizing antibodies like b12 (62) used in our study, and hence it is not surprising that the IC 50 values for this subtype were low. Several previous studies have reported a correlation between PNGS and neutralizing sensitivity, whereby higher PNGS in Env correlated with increased resistance to neutralizing antibodies (32)(33)(34)63). Although the most significant association of bystander apoptosis was found to be with PNGS, the indirect correlation of PNGS and neutralizing anti-body sensitivity reported by others (32)(33)(34)63) is indicative of possible selective pressure on the virus altering its bystander apoptosis-inducing phenotype. However, this is largely speculative because data are limited in this regard in our study.
Sequence Analysis of Primary Envs Reveals Several Residues Associated with Bystander Apoptosis-inducing Activity-Using genotypic information to predict the co-receptor usage of primary HIV Envs is a routine practice (64). The co-receptor prediction software programs utilize sequence information gathered from numerous laboratory experiments using pseudotyped virus-based infection assays as a standard. Based on early findings, a set of rules has been established for prediction of CXCR4/CCR5 co-receptor usage by HIV isolates. These rules include the V3 loop sequence and charge rules (65). However, there is very limited information on the Env sequence changes that can predict or may be associated with pathogenicity of primary Envs. Recently, Sterjovski et al. (66) demonstrated that the presence of Asn-362 in the HIV gp120 region was associated with increased cell-to-cell fusion activity. We hence asked whether there were specific amino acids in the HIV Env that could be associated with bystander apoptosis induction. A search for distinctive physicochemical properties between pairs of phylogenetically close high-low apoptosis strains ( Fig. 7 and Supplemental Fig. 1) revealed several important residues that may be relevant to the bystander apoptosisinducing phenotype. Twelve sites (134, 139, and 189 (in the V1 loop); 328, 330, 332, and 334 (V3 loop and vicinity); and 362, 363, 425, 462, and 476 (adjacent to the CD4 binding site) (HXB2 numbering)) were found (Table 2) in at least three pairs of highlow apoptosis strains with distinctive differences in amino acid residues in the above positions. Notably, we found that Asn-362, which has previously been shown to be associated with  To further our analysis and to avoid complications from founder effects through amino acid differences that may arise as a result of separate evolution of the different subtypes, we considered the relative phylogenetic positions of all analyzed strains. Thereafter, we selected genetically closely related pairs with differing apoptosis-inducing results to increase the specificity of identifying mutations directly related to this phenotype (Fig. 8A). In essence, this approach identifies Envs that are most closely related in sequence but varied in phenotypic characteristics, thus classifying them as nearest pair neighbors. For instance, using this approach, Envs 11034 and 11035 were found to be closely related in sequence but widely varied in apoptosis induction. This novel analysis revealed a number of sites that may be associated with bystander apoptosis induction with two sites that stood out in particular. These include amino acid positions 425 and 476. Interestingly, four of the nine high apoptosis inducers selectively showed the presence of an asparagine at position 425 by our pairwise analysis approach (Fig.  8B). Additionally, position 476 was of particular interest (Fig.  8C) because five of nine high-low apoptosis pairs possessed a lysine in the low apoptosis inducers that was not found in any of the high apoptosis strains examined. In all nine pairs, the high apoptosis group had an arginine at this position. In addition to our semimanual phylogeny and structure-aware candidate selection procedure, we also compared the results with the related sequence harmony approach (67), which confirms our two studied candidates as being highly significant ( Fig. 8D and supplemental Fig. 2). Using this comprehensive analysis, we were able to identify several potential residues important for bystander apoptosis, of which residues 425 and 476 were most promising due to the reasons described below and hence chosen for further in depth analysis.
Structural Analysis of Candidates for Determinants of HIV Env-mediated Bystander Apoptosis-Having identified specific amino acid residues that may be involved in bystander apoptosis induction, we next conducted structural analysis to determine the location of these residues with respect to the viral Env. Structural mapping of the above identified 12 candidate sites onto Protein Data Bank structure 2QAD (52) revealed that these sites could be found in the V1 loop (positions 134, 139, and 189), V3 loop and its vicinity (positions 328, 330, 332, and 334), and the CD4 binding site region (positions 362, 363, 425, 462, and 476). For this study, we chose candidate sites that could play a role in directly affecting CD4 binding (positions 362, 363, 425, and 476) for further experimental analysis. Because position 362 has already been shown to affect HIV cell fusion (66) and position 363 may possibly play a similar role because it lies next to residue 362, we tested the other promising positions, 425 and 476, for association with bystander apoptosis. Because both of these positions are within 5 Å of the CD4 binding site, residue changes R476K and N425R in the high apoptosis-inducing strains, converting them to low apoptosis inducers, were also expected to affect CD4 binding via steric hindrance and also lead to charge alterations in the Env (Fig. 9).
Mutational Analysis Shows That R476K and N425R May Be Relevant to Bystander Apoptosis-inducing Activity of Env Glycoprotein-Our sequence analysis data and molecular modeling identified specific amino acid residues that may be linked to apoptosis induction by the primary HIV Env used in this study. Of those, the most noticeable ones were Asn-425 and Arg-476 because these sites are probably involved in CD4 binding based on our structure analysis. To confirm the involvement of these residues in apoptosis induction, we used a sitedirected mutagenesis approach to induce R476K and N425R mutations in the high apoptosis-inducing Envs. As seen in Fig.  9, A and B, the N425R and R476K mutations resulted in a loss of function in some of the Envs tested. Whereas N425R showed the most significant loss in two of the four Envs tested (Fig.  10A), R476K was found to result in loss of function (apoptosis) in three of five Envs tested (Fig. 10B). These findings corroborate the potential role of these two sites in determining bystander apoptosis potential of Envs. To further corroborate our findings, we constructed a gain of function mutation in the corresponding low apoptosis-inducing variant. Because Env 11035 R476K showed the most significant loss of function in the experiments above, we selected its low apoptosis-inducing counterpart (11034) to induce a K476R mutation. Interestingly, we saw a gain in apoptosis-inducing activity for 11034 K476R compared with the wild type counterpart 11034 (Fig. 10C). Although both

Concentration of TriMab antibody required for 50% infectivity inhibition (IC 50 ) of pseudotyped virus stocks
Data are the mean of duplicate observations. Name, classification of Envs as high medium or low apoptosis inducers; Env. no, NIH-ARP designated catalog number for the Envs used in the study; Clade, virus subtype; NA, not available. N425R and R476K are associated with bystander apoptosis in our study, R476K was also confirmed to be a bystander apoptosis-regulating site via gain of function mutations. Env expression between each pair was found to be identical (Fig. 10H), confirming that changes in Env bystander apoptosis are not likely to be due to differences in expression levels.  Because our approach involves a pairwise analysis of the nearest neighbors that are genetically similar but with differing apoptosis induction capability, we tested the various characteristics of 11034 and 11035 (R476K pair). A side by side comparison of this pair revealed maximum differences with respect to all of the characteristics studied like apoptosis (Fig. 10D), virus infectivity (Fig. 10E), antibody neutralization (Fig. 10F), and PNGS (Fig. 10G). Thus, this pair represents the perfect scenario for identification of characteristics associated with low versus high apoptosis-inducing Envs. Also this pair corroborates many of the associations revealed in our study, including virion infectivity, PNGS, neutralizing antibody sensitivity, and specific amino acids, in determining bystander apoptosis activity. This approach also reveals that the Env phenotype in terms of bystander apoptosis is probably quite complex, and multiple Env regions and characteristics may collectively determine the apoptosis outcome. Thus, studying genotypically closely related Envs that show a diverse apoptosis phenotype is probably the best approach for identifying the residues/genetic signatures important for this phenomenon.

Bystander Apoptosis Induction by Primary Envs Requires CCR5 Binding as Well as gp41-mediated Membrane Fusion-
The membrane fusing activity of HIV Env has long been suspected to have a role in HIV pathogenesis (68). Specifically, the role of the gp41 subunit in mediating bystander apoptosis has only recently been demonstrated (41,59). Studies by our group and others have shown that gp41-mediated fusion/hemifusion is critical for apoptosis induction, and binding of gp120, although required, cannot alone induce apoptotic signaling by HIV Env (40,41,59). Although studies with laboratory viruses have provided proof of principle for this hypothesis (69), there is a lack of data from primary Envs derived directly from patients to corroborate the physiological relevance of these findings. We therefore asked whether bystander apoptosis induction by primary Envs is also gp41-dependent. To answer this question, we selected the highest apoptosis-inducing Envs from our panel and tested bystander apoptosis induction in the presence or absence of CCR5 antagonist maraviroc and gp41 inhibitor enfuvirtide. As seen in Fig. 11, both maraviroc and enfuvirtide inhibited bystander apoptosis by all of the primary Envs. Inhibition by CCR5 inhibitor confirms a role of gp120 binding to the co-receptor as a requirement for apoptosis induction, whereas inhibition by enfuvirtide is indicative of a critical role played by the gp41 unit in this phenomenon. These findings are consistent with previous studies and further support the role of HIV gp41-mediated fusion in apoptosis induction not only with laboratory isolates but also primary viruses. Furthermore, this also suggests that the mechanism of bystander apoptosis by primary and laboratory variants of HIV is probably the same.

DISCUSSION
HIV infection leads to a selective depletion of CD4 ϩ T helper cells, resulting in immunodeficiency. However, the mechanism by which HIV mediates this process remains highly debated (2, 6, 12, 70 -73) and forms the basis of our study. It is proposed that the loss of CD4 cells during HIV infection is due to the process of bystander apoptosis (2). This is supported by early studies conducted by Finkel et al. (9), who demonstrated that the majority of cells undergoing apoptosis during HIV infection are not infected but are in close proximity to infected cells. A role of direct infection in the loss of CD4 cells is also refuted by natural infections in sooty mangabeys with the simian SIVsm, where high levels of infection and viremia do not result in CD4 loss or AIDS development (74). Hence, bystander apoptosis is believed to be one of the major causes of CD4 loss leading to AIDS (6,70,75). Bystander apoptosis is a phenomenon that has been demonstrated in various animal models of HIV, including the SIV model in rhesus macaques (4, 10) and more recently in the humanized mouse model (11). Interestingly, bystander apoptosis induction in these models correlates with CD4 loss, further supporting the role of this phenomenon in disease progression.
The mechanism of bystander apoptosis of CD4 T cells in HIV infection remains unresolved. Interestingly, the role of the Env glycoprotein in this process is becoming increasingly evident (12,13). This is largely supported by the arguments that 1) as cell death in HIV infection outnumbers the infected cell population, a role of bystander cell death in the progression to AIDS is likely; 2) as the depletion of immune cells is restricted to CD4 ϩ helper phenotype and as the Env glycoprotein of HIV binds to CD4, it probably plays a role either directly or indirectly in CD4 T cell death; and 3) the Env glycoprotein is the only viral protein expressed on the surface of infected cells and Amino acid residues that may be associated with bystander apoptosis induction using the approach delineated in Figs. 7 and 8 The region of those residues identified, their amino acid phenotype, and their structural attributes are listed. Position, amino acid position based on HXB2 protein sequence starting from the first methionine residue of GP120; 3U2S, crystal structure of PG9 Fab in complex with the V1V2 region from HIV-1 strain ZM109 (52); 2QAD, structure of tyrosine-sulfated 412d antibody complexed with HIV-1 YU2 gp120 and CD4 (53); 2B4C, crystal structure of HIV-1 JR-FL gp120 core protein containing the third variable region (V3) complexed with CD4 and the X5 antibody (54); NAG, N-acetylglucosamine; AI, apoptosis inducers.
has been shown to interact with and mediate apoptosis in bystander cells, albeit only in in vitro studies (41,59,60,76).
With support from in vitro studies, understanding the genetic signatures associated with the bystander apoptosis-inducing phenotype of the Env glycoprotein is of increasing importance. Although the phenotypic and genotypic characterization of HIV Env has been well documented for co-receptor usage and neutralizing antibody sensitivity, there is virtually no information on the phenotypic characterization of HIV Env for the phenomenon of bystander apoptosis induction. We have previously developed cell lines that express varying levels of the HIV co-receptor CCR5 (44). Using these cell lines and a laboratory-adapted YU-2 clone of HIV, we studied the bystander apoptosis-inducing activity of HIV Env and the role of CCR5 co-receptor levels on this phenomenon. Our findings suggest that the bystander apoptosis induction is dependent on the Env fusogenic activity as well as the surface CCR5 expression levels on cells (44). Although this study provided valuable insights into the mechanism of R5 Env-mediated apoptosis, the physiological relevance of these findings remains untested. Hence, we undertook this study with the objectives to understand whether and how primary HIV Env variants induce apoptosis in bystander cells, whether there is variability in bystander apoptosis induction between different HIV subtypes, and whether specific genetic signatures in the Env glycoprotein are associated with increased bystander apoptosis phenotype.
To achieve these objectives, we developed a high throughput assay that can determine the relative bystander apoptosis-inducing activity of primary HIV-1 Envs. Our caspase-based assay was highly reproducible and correlated well with other methods of apoptosis determination. More importantly, the assay could distinguish small changes in bystander apoptosis-inducing activity between various HIV Env variants tested. With the help of this assay, an important characteristic of HIV Env can be added to the profile of patient viruses.
We found that the bystander apoptosis-inducing activity of primary Envs of R5 tropism varies considerably. Although we did not find subtype-specific differences in our panel, it should be noted that the assay was done using a reference panel of Envs that represents a highly diverse set of Envs. Whether certain subtypes contain high apoptosis-inducing Envs that are overrepresented in the population can only be determined by studying larger populations of Envs directly from patients. We also found that virion infectivity correlated with bystander apoptosis-inducing activity in our panel. This does not come as a surprise because both virion infectivity and bystander apoptosis are dependent on the function of the Env glycoprotein. Although as a group, the Envs showed these differences, the bystander apoptosis and virion infectivity of individual Envs did not necessarily correlate in every case. We did find some Envs that showed high infectivity but low bystander apoptosis and vice versa. These findings underscore both the genotypic and phenotypic variability of Envs seen in patients. This also suggests that in some cases, the viruses may be capable of infection/ replication in the absence of inducing bystander apoptosis. Such non-pathogenic viruses probably contain changes in the Env glycoprotein that alter their apoptosis phenotype while maintaining virion infectivity. This is consistent with our previous findings, whereby a single amino acid change in the Env glycoprotein altered bystander apoptosis and CD4 loss in a humanized mouse model of HIV infection (11) while maintaining virus infectivity. However, from a viral evolution point of view, the acquisition of bystander apoptosis-inducing activity by a virus must come with some selective advantage, and in this case, it might be increased fitness/infectivity. This increased fitness may help in virus dissemination via infection with cellfree virus in different anatomical sites.
Putative N-glycosylation sites have been shown to vary considerably in Env glycoproteins from HIV-infected patients based on the stage of the disease (18,33). During the chronic phase, HIV Env has been shown to contain more PNGS compared with the acute phase of infection (33). This same characteristic is reflected in viruses isolated from late stages of the disease, whereby loss of PNGS is observed during late versus chronic viruses (32). Furthermore, the presence of higher glycosylation or PNGS has also been associated with reduced viral fitness (32). The question that arises is what mediates the increase in PNGS during the chronic phase of HIV infections. The leading hypothesis is that the neutralizing antibody pressure is the primary driving force behind PNGS evolution in HIV. This is supported by the fact that the presence of PNGS on FIGURE 8. Phylogenetic analysis of primary Envs to identify sequences that share genetic signatures but vary in apoptosis induction. A, phylogeny of primary Envs by a maximum likelihood method based on the HKY model conducted in MEGA5 (51). ␥-Distribution (five categories) with shape parameter (0.4327) was used to model evolutionary rate differences among the sites. Pairs of high-low bystander apoptosis are colored with the same color. Shown are sequence signatures of high and low apoptosis inducers at amino acid positions 425 (B) and 476 (C). The high apoptosis-inducing Envs predominantly bear an Asn at position 425 and an Arg at position 476, whereas low apoptosis-inducing Envs bear an Arg at position 425 and a Lys at position 476. D, sequence harmony score coloring of GP120 in complex with CD4. The red colored residue is sequence harmony's best estimate of an important residue distinguishing high/low apoptosis, but based on its structural position away from CD4, such an effect is not intuitively clear. The further ranking can be seen by color transition from red to white to blue. . Structure (Protein Data Bank entry 2B4C) of GP120 (gray ribbon) binding to CD4 (yellow ribbon) rendered using Yasara (55). Arg-476 (red residue) from gp120 is in close proximity with Gln-25 and Ser-23 (pink residues) from CD4. Asn-425 (blue residue) from GP120 is also within 5 Å of Phe-43 and Arg-59 (light blue residues) from CD4.
HIV Env is responsible for masking the epitopes and increasing neutralization resistance (33)(34)(35). However, this phenomenon may also come at the price of reduced viral fitness (77), and hence, when the opportunity arises, as in the case of late stages where the immunocompromised patients are no longer able to mount an effective neutralizing antibody response, the virus reverts to low PNGS, increased fitness, and possibly increased bystander apoptosis phenotype.
Our study demonstrates that many of the characteristics of Env glycoprotein, including PNGS, neutralizing antibody sen-sitivity, and viral fitness, that have previously been shown to correlate with CD4 loss and disease progression are also incidentally associated with bystander apoptosis induction. The preliminary hypothesis that emerges from our study is that during the chronic phase of the disease, immune pressure selects for Envs with higher PNGS. This potentially results in a loss/ reduction in bystander apoptosis, which may be reflected in the slower depletion of CD4 ϩ T cells seen during the chronic phase. Subsequently, the gradual loss of CD4 cells results in an immunocompromised state that paves the way for the viruses with FIGURE 10. Mutation of residues Arg-476 and Asn-425 to R476K and N425R, respectively, leads to a reduction in Env-mediated bystander apoptosisinducing activity. HeLa cells transfected with WT Env clones or their corresponding N425R (A) or R476K (B) counterparts were assayed for bystander apoptosis induction as described in Fig. 1. C, gain of function analysis for the 11034 and 11035 Env pairs. HeLa cells transfected with the indicated WT DNAs or their gain of function counterpart mutants were analyzed for bystander apoptosis induction. Data represent mean Ϯ S.D. (error bars) from triplicate observations. Shown is pairwise analysis of the nearest neighbors 11034 and 11035 for various Env characteristics like apoptosis (D), virus infectivity (E), antibody neutralization (F), and PNGs (G). This pair represents the perfect case scenario for all of the parameters studied. H, detection of Env expression for various WT and mutant Env pairs used in the study by HIV Ig-mediated immunoprecipitation.
lower PNGS, increased fitness, and possibly higher bystander apoptosis-inducing phenotype to emerge and precipitate a rapid CD4 loss. Although our study only provides circumstantial evidence of this correlation, it does open the door to further studies focused on this aspect.
Interestingly, most of the PNGS differences as well as the 12 most relevant sites found to be associated with the apoptosis phenotype were in the gp120 region, with many mapping around the CD4 binding site. The fusion process mediated by Env is a concerted effort between both the gp120 and the gp41 subunits of HIV Env. Because the gp120 subunit is exposed to host defense mechanisms, it is not surprising that the largest variability is seen in this subunit, possibly via antibody pressure. It is also not surprising that most of the changes associated with bystander apoptosis induction also map to gp120. Supporting these findings, Holm et al. (78) demonstrated that changes in the gp120 co-receptor binding domain can affect Env function and bystander apoptosis.
As mentioned above, one of our major objectives was also to determine whether certain HIV phenotypic and genotypic signatures correlate with bystander apoptosis and, more specifically, whether sequence analysis alone may be able to predict the bystander apoptosis-inducing phenotype of HIV Envs. To this end, we adopted a novel pairwise analysis approach that focused on phylogenetically closely related Env pairs with differing apoptosis activity in an attempt to identify candidate amino acid signatures related to this unique phenotype. The Env fusogenic activity has been associated with a specific stage of disease and specific residues like Asn-362 (66). Recently, Wade et al. (27) demonstrated that Env glycoproteins derived from late stage patients are more fusogenic and induce enhanced CD4 ϩ T cell apoptosis compared with early stage Envs. In our study, we identified a total of 12 amino acid positions that were potentially associated with high bystander apoptosis-inducing Envs, including the previously mapped Asn-362 linked to increased fusogenicity. Among these, R476K and N425R were identified as novel sites through our pairwise analysis and molecular modeling. We also confirmed the potential role of these amino acid signatures experimentally using sitedirected mutagenesis.
The primary function of HIV Env is to mediate fusion of the viral and cellular membranes to facilitate entry of the virus core into cells. However, the same phenomenon can also mediate fusion between an infected and uninfected bystander cell. This fusion process and the intermediate of this phenomenon referred to as hemifusion (kiss of death) have been shown to be the primary cause of bystander apoptosis in in vitro culture systems (40,41,59,79). Most of the data supporting the role of Env-mediated fusion/hemifusion in bystander apoptosis comes from laboratory-adapted viruses. Hence, in this study, we also asked whether bystander apoptosis mediated by primary HIV-1 Envs is dependent on the Env fusogenic activity or gp41 function. We indeed demonstrate that bystander apoptosis induction by primary Envs is dependent on gp41 function (Env fusogenic activity). Interestingly, fusogenic activity of Env glycoprotein has been indirectly associated with HIV pathogenesis in clinical studies where the presence of a highly fusogenic syncytia-inducing phenotype has been associated with poor prognosis in patients (80,81), increased pathogenesis (23), and CXCR4 (82) tropism of viruses. The syncytia-inducing phenotype has also been associated with increased pathogenesis in the SCID-hu mouse model. In SCID-hu mice, the lack of CD4 T cell loss by certain CCR5 tropic laboratory strains underscores this difference (83). However, it has also been demonstrated that CCR5 tropic viruses from late stages of disease are more fusogenic, which correlates with pathogenicity (26). This is further supported by recent studies by Wade et al. (27), demonstrating the apoptotic potential of late stage R5 viruses. Probably the strongest evidence for a role of Env membrane-fusing activity and HIV pathogenesis and CD4 decline comes from SHIV studies in rhesus macaques, where the pathogenicity of the SHIV-KB9 variant can be linked directly to the membrane fusion activity of the virus (84 -88). The fact that primary Env-mediated bystander apoptosis in our study could be inhibited by inhibiting Env-mediated fusion using a gp41 inhibitor adds further support for the role of membrane fusion in this phenomenon. Recently, Hunt et al. (89) have shown the clinical benefits of maraviroc and ENF therapy that target Env, supporting the role of Env in HIV pathogenesis.
Our findings suggest that bystander apoptosis-inducing activity of HIV Env may be an important phenotypic characteristic that warrants studies in a larger panel of Envs for two reasons. 1) It will help strengthen the hypothesis that bystander apoptosis is an important phenomenon for CD4 loss in HIV infection. 2) With a larger panel of Envs, we will be able to obtain enough preliminary data to determine whether certain genetic signatures are associated with bystander apoptosis, much in line with the current methods for determining co-receptor tropism using sequence information alone.
Although the Env glycoprotein has previously been characterized for numerous phenotypic properties like infectivity, coreceptor use, neutralizing antibody sensitivity, and PNGS, bystander apoptosis induction is a unique phenotype of HIV that has previously not been studied, probably due to a lack of an appropriate high throughput assay. Because a number of these characteristics of the Env glycoprotein are associated with FIGURE 11. Bystander apoptosis induction by primary Envs requires CCR5 binding as well as gp41 function. HeLa cells were transfected with the indicated Envs. The next day, SupT-R5-H6 cells were added to each well in the presence of medium alone, enfuvirtide (2 M), or maraviroc (1 M). After an overnight incubation, the cells were analyzed for caspase-3 activity using the Caspase-Glo 3/7 assay as described in the legend to disease progression and CD4 loss, our findings provide further support for the phenomenon of bystander apoptosis in CD4 loss and the role of Env glycoprotein in this process. The actual apoptotic potential of Env variants may best be calculated using our assay and pairwise computational analysis. We believe that using this strategy, we will be able to generate a database that can be used for developing a methodology for predicting bystander apoptosis. Recently, a machine-learning method has been developed for identifying genetic signatures in HIV Env predictive of HIV-associated dementia (90) based on a database of HIV Env sequences from the brain (91). Using similar strategies, our study should lead to development of an HIV Env sequence database characterizing the bystander apoptosis potential of Envs. Our study establishes a first set of preliminary rules like PNGS and specific amino acid signatures that can be further developed into an algorithm. Ultimately, this should lead to determination of bystander apoptosis phenotypes of Envs by sequence analysis alone. Whether these findings would have a diagnostic value remains to be seen.