Antibody-Dependent Cellular Cytotoxicity-Competent Antibodies against HIV-1-Infected Cells in Plasma from HIV-Infected Subjects

HIV Env-specific nonneutralizing Abs (NnAbs) able to mediate ADCC have been implicated in protection from HIV infection. However, Env-specific NnAbs have the capacity to support ADCC of both HIV-infected and HIV-uninfected bystander cells, potentially leading to misinterpretations when the assay used to measure ADCC does not distinguish between the two target cell types present in HIV cultures. Using a novel ADCC assay, which simultaneously quantifies the killing activity of Env-specific Abs on both infected and uninfected bystander cells, we observed that only a minority of Env-specific Abs in HIV+ plasma mediated ADCC of genuinely HIV-infected cells displaying Env in its native closed conformation. This assay can be used for the development of vaccine strategies aimed at eliciting Env-specific Ab responses capable of controlling HIV infection.

T he RV144 Thai trial was the first and only HIV vaccine trial to date to show moderate (31%) but significant efficacy at protecting against HIV infection (1). Protection was not associated with the presence of broadly neutralizing antibodies (BnAbs) or cytotoxic T cell responses (2). Results from analyses of correlates of protection suggested that protection was associated with the induction of nonneutralizing immunoglobulin G (IgG) Abs directed to the V1/V2 loop of HIV Envelope (Env) gp120 (2)(3)(4). Also reported to be associated with protection from infection were Env-specific IgG nonneutralizing Abs (NnAbs) able to mediate Ab-dependent cellular cytotoxicity (ADCC) provided that no competing IgA Abs were present (2,(5)(6)(7). This has led to heightened interest in describing the determinants of effective anti-HIV directed ADCC activity.
HIV Env glycoprotein is the HIV gene product targeted by ADCC-competent Abs since it is the only viral protein exposed on the surface of infected cells (8). Env is a trimer assembled of heterodimers constituted of gp120 and gp41 glycoproteins. Whereas gp120 forms the outer part of the trimer, gp41 is largely buried at the trimer interface and anchors Env on the plasma membrane (9)(10)(11)(12). Unliganded Env is normally present in a "closed" conformation on the surface of virions and infected cells (13,14). Env interaction with CD4 drives the transition of a closed Env conformation to a CD4-bound "open" conformation (3,4,13). CD4 binding to gp120 occurs mainly during the attachment of viral particles to CD4 ϩ target cells (T) at viral entry, as CD4 is downregulated from the surface of productively infected cells by Nef and Vpu (4,15). However, gp120 is reported to shed from the surface of infected cells and to bind to CD4 on uninfected bystander cells, which then display Env in an open conformation (16,17).
A frequently used ADCC target cell is the CEM.NKR.CCR5 (CEM) cell line coated with monomeric recombinant gp120 (rgp120) (24,25). CEM cells are resistant to direct natural killer (NK) cell cytolysis (26,27). Just as shed gp120 binds bystander cells, rgp120 coats CEM target cells through CD4 interactions, forcing gp120 to assume the CD4bound conformation recognized by anti-cluster A monoclonal antibodies (MAbs) such as prototypical A32 (17,28,29). HIV-infected CEM cells and primary CD4 cells have also been used as ADCC targets (16,(30)(31)(32). HIV infection of CD4 ϩ T cells usually results in only a fraction becoming infected. Uninfected CD4 ϩ bystander cells coated with shed gp120 or virions present in the viral inoculum expose CD4i epitopes, which trigger the recognition and killing of bystander cells through ADCC (16,33). This compromises the measurement of Env-specific ADCC-competent Abs capable of mediating the killing of genuinely infected cells. Other studies evaluating the presence of ADCC-competent Abs in HIV ϩ plasma have used HIV isolates presenting Nef and/or Vpu deletions or having mutations in these gene products that compromise their ability to downmodulate cell surface CD4, making it available to bind Env, which then assumes an open conformation recognized by anti-cluster A MAbs (25,34). The use of rgp120-coated cells, partially HIV-infected cells, or cells infected with a virus unable to downregulate CD4 as ADCC target cells is the basis for the widely held view that Abs to CD4i epitopes dominate the HIV Env-specific ADCC-competent Ab pool in plasma from HIV-infected subjects (25,35).
We describe here the generation of a CEM target cell that is close to 100% HIV infected as determined by intracellular p24 staining. The virus used to infect CEM cells codes for HIV Bal Env and expresses wild-type Nef and Vpu able to downregulate CD4, exposing Env in a closed conformation. A panel of BnAbs, NnAbs, and control Abs was used to probe the epitope structure of cell surface Env on infected CEM cells in comparison to bystander and rgp120-coated CEM (cCEM) cells. The ability of each Ab to support the ADCC activity of the three target cells was assessed. We also describe a novel assay designed to measure ADCC activity using the frequency of annexin V ϩ (AnV ϩ ) target cells as a readout and to compare the performance of this assay with that of the previously described ADCC-GranToxiLux assay (24). We confirm that anti-cluster A MAbs mediate ADCC of CEM cells coated with rgp120 or of bystander cells that have taken up shed gp120 or bound defective HIV virions but not of genuinely HIV-infected CEM cells. Most ADCC-competent Abs in HIV ϩ plasma recognize and kill target cells, exposing open-Env-conformation. Abs to epitopes other than CD4i epitopes that recognize and support ADCC of infected CEM cells expressing closed-conformation Env are also present in HIV ϩ plasma.
or of HIV Ϫ IgG as a negative control (Fig. S2). While the results generated using the two assays were found to correlate, ADCC-AnV was significantly more sensitive at quantifying ADCC than ADCC-GTL in terms of the maximum frequency of apoptotic cells detected (P Ͻ 0.05 for the comparisons between AnV ؉ and GzB ؉ results for all HIV ؉ IgG concentrations lower than 100 g/ml) and of the superior signal/noise ratio achievable in the presence of HIV ؉ IgG versus HIV Ϫ IgG, particularly at low opsonizing Ab concentrations. Furthermore, the frequencies of ADCC-AnV ؉ and ADCC-GzB ؉ target cells peaked at 4 and 40 g/ml of HIV ؉ IgG, respectively (Fig. S2).
Binding characteristics and ADCC competence of a panel of anti-HIV Envelope (Env)-specific MAbs with respect to newly infected CEM cells and bystander CEM cells. The rgp120 commonly used to coat target cells for ADCC assays is monomeric and does not expose the same epitopes as native, trimeric Env on cells infected with replication-competent HIV (37). Quaternary epitopes on trimeric Env are not present on monomeric gp120. In contrast, monomeric gp120 displays the epitopes which are  normally occluded inside the closed conformation of Env on target cells infected with wild-type HIV (16). Moreover, the CD4 binding site (CD4bs) Env epitope is not exposed on gp120-coated cells due to the engagement of the gp120 CD4bs with cell surface CD4 (16). This prompted us to develop a model of ADCC using HIV-infected instead of rgp120-coated cells. The virus used to infect CEM cells expressed all viral proteins from an NL4-3 backbone with the exception of Env, which came from HIV-Bal, and the heat shock protein heat-stable stable antigen (HSA) reporter gene (38). Since only a fraction of these cells were found to be HIV infected at 4 days postinfection, HSA staining was used to distinguish newly infected CEM cells from uninfected bystander CEM cells. This staining allowed us not only to gate on infected cells but also to evaluate binding and ADCC activity mediated by anti-Env Abs on newly infected CEM cells compared to   Targeting HIV-Infected Cells by ADCC ® uninfected bystander CEM cells whose CD4 had bound shed gp120, an interaction that revealed gp120-CD4i epitopes not exposed on newly infected CEM cells (16,17). The binding characteristics of a panel of anti-Env MAbs and polyclonal HIV ϩ IgG (15 g/ml each) to newly infected CEM cells (Live/Dead [LD] Ϫ CFSE ϩ PKH26 ϩ HSA ϩ ), bystander CEM cells (LD Ϫ CFSE ϩ PKH26 ϩ HSA Ϫ ), and uninfected CEM cells (CFSE ϩ PKH26 Ϫ ) are shown in Fig. 3 (see also Fig. S3). As expected, control staining performed with no Ab and with HIV Ϫ IgG generated similar levels of background staining of the three target cells. CD4bs Abs (b12, VRC01, NIH45-46, and 3BNC117) bound newly infected CEM cells with a much higher MFI than was seen with bystander CEM cells. This was expected since the CD4 binding site of gp120 is not available on bystander CEM cells by virtue of its interaction with cell surface CD4 (16). Abs to glycans such as PGT121, 10-1074, and 2G12 also recognized newly infected CEM cells better than bystander CEM cells. In contrast, Abs to CD4i epitopes such as A32, C11, and N12-i2 recognized bystander CEM cells with a higher MFI than newly infected CEM cells. Of note, polyclonal HIV ؉ IgG stained both newly infected CEM cells and bystander CEM cells, with binding to bystander CEM cells resulting in a higher MFI.
Overall, these observations suggested that gp120 was being shed from newly infected CEM cells and taken up by bystander CEM cells. They confirmed that CD4i epitopes are more readily exposed on bystander CEM cells and that CD4bs and V3 loop Abs targeting the closed Env conformation recognized newly infected CEM cells with a higher MFI than was seen with bystander CEM cells.
Results of ADCC activity assays using the same panel of anti-Env MAbs and polyclonal HIV ϩ IgG (15 g/ml each) to opsonize newly infected CEM cells, bystander CEM cells, and uninfected CEM cells are shown in Fig. 4A to C. All MAbs that bound newly infected CEM cells also supported ADCC measured as the frequency of AnV ؉ target cells. The one exception to this was the gp120 outer domain recognizing 2G12 MAb, known to bind both the open and closed Env conformations but also to mediate poor ADCC activity (4,15,30). Indeed, 2G12 bound Env on both newly infected CEM cells and bystander CEM cells with a high MFI without triggering the ADCC of these target cells.
2G12 has an unusual domain-swapped configuration which may support 2G12 dimerization. This might affect the ability of this Ab to interact with Fc receptors to i., infected CEM cells were labeled with CFSE, LD, and PKH26 and combined with uninfected CEM (unCEM) cells labeled with CFSE only as described for Fig. 3. The combined CEM cells were opsonized with 15 g/ml of each anti-Env Abs used as described for Fig. 3 and incubated with effector NK cells for 1 h followed by staining with anti-HSA and AnV. ADCC activity (% ADCC) was measured as the frequency of AnV ϩ cells among niCEM (LD Ϫ CFSE ϩ PKH26 ϩ HSA ϩ , red bars; A), byCEM (LD Ϫ CFSE ϩ PKH26 ϩ HSA Ϫ , blue bars; B) and unCEM (CFSE ϩ PKH26 Ϫ , orange bars; C) cells. Data represent averages Ϯ SD of results from three independent experiments. Each dot represents a single NK cell donor. Significance was determined by comparing the percentages of ADCC between each anti-Env Ab used and HIV Ϫ IgG (*, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001; 2-way ANOVA, Dunnett's multiple-comparison test). (D and E) Spearman correlations between the frequency of anti-Envspecific Ab binding ([Env binding %]; D) and the MFI of anti-Env binding ([Env binding MFI]; E) to niCEM cells, as described for Fig. 3 (red histograms), with the percentage of ADCC generated using the same Abs to opsonize target cells.
Targeting HIV-Infected Cells by ADCC ® support ADCC (39,40). The hypothesis regarding the association between binding and ADCC was supported by the significant correlation between the frequency and MFI of Ab binding and the frequency of AnV ؉ newly infected CEM cells generated in the ADCC-Anv assay (P ϭ 0.0003 and r ϭ 0.6825 for percentages and P ϭ 0.0003 and r ϭ 0.6828 for MFI values, respectively) ( Fig. 4D and E). In contrast to the results determined with newly infected CEM cells, none of the MAbs tested in the panel mediated robust ADCC of bystander CEM cells, with the exception of polyclonal HIV ؉ IgG (Fig. 4B). Gp120 shedding, like viral production, requires Env to be expressed on the surface of infected cells. In a mixture of infected and uninfected cells, uptake by bystander CEM cells of gp120 shed from newly infected CEM cells and infection of bystander CEM cells may occur simultaneously. Thus, to better characterize the infection status of bystander CEM cells in our system, we evaluated levels of intracellular p24 and cell surface CD4 in both HSA ϩ and HSA Ϫ CEM cells 4 days postinfection. As shown in Fig Nonetheless, levels of Env exposed on bystander CEM cells were insufficient to support high frequencies of AnV ϩ bystander CEM cells opsonized with any of the MAbs tested, though opsonization with HIV ؉ IgG did support substantial ADCC as measured by the frequency of AnV ϩ bystander CEM target cells (Fig. 4B).

Binding characteristics and ADCC competence of a panel of anti-Env-specific MAbs with respect to sorted HIV-infected CEM cells and coated CEM cells.
To rule out misinterpretations of these results due to the infection status of bystander CEM cells or the potential bias induced by the viral inoculum used, we enriched and sorted HSA ϩ newly infected CEM cells. These HIV-infected cells grow as an immortalized cell line and are referred to as sorted infected CEM cells here. The proportions of sorted infected CEM cells that expressed cell surface HSA and intracellular p24 were 99% and 94%, respectively (Fig. 5A). The downregulation of CD4, BST-2, and HLA-C from the surface of sorted infected CEM cells was consistent with these cells expressing functional Nef and Vpu (Fig. 5B).
Results of ADCC-AnV assays performed using sorted infected CEM target cells opsonized by increasing doses of HIV ؉ IgG and HIV Ϫ IgG are shown in Fig. 6. Killing of sorted infected CEM cells by effector cells was dose dependent and anti-Env specific since HIV Ϫ IgG did not mediate ADCC activity (Fig. 6A). The ADCC-AnV assay effector cells were NK cells as their depletion from PBMC abrogated the ADCC activity of sorted infected CEM cells, as previously described for coated CEM cells ( Fig. 6B; see also Fig. S1). The results of ADCC-AnV and ADCC-GTL assays using sorted infected CEM cells as targets were correlated (Fig. 6C). As observed for coated CEM target cells, the ADCC-AnV assay was significantly more sensitive at quantifying ADCC activity than the Error bars indicate SD of results from replicates, and significance was determined by comparing the frequencies of AnV ϩ siCEM cells (% ADCC) between HIV ϩ IgG and HIV Ϫ IgG for both concentrations (*, P Ͻ 0.05; ***, P Ͻ 0.001). (C) CFSE ϩ or NFL1 Ϫ TFL4 ϩ labeled siCEM T cells (NFL1 Ϫ marks viable cells and TFL1 marks target cells) were opsonized with increasing doses of HIV ϩ IgG (filled symbols) or HIV Ϫ IgG (empty symbols) and used in an ADCC-AnV assay (black symbols) or in an ADCC-GTL assay (gray symbols). Error bars indicate SD of results from replicates, and significance was determined by comparing the percentages of ADCC as measured by the frequency of AnV ϩ or granzyme B ϩ (GzB ϩ ) siCEM between HIV ϩ IgG and HIV Ϫ IgG for all IgG concentrations (***, P Ͻ 0.001) and by comparing the percentages of ADCC using AnV ϩ or GzB ϩ for all HIV ϩ IgG concentrations (#, P Ͻ 0.05; ###, P Ͻ 0.001). (D) siCEM target cells labeled with CFSE and PKH26 were combined 1:1 with cCEM cells labeled with CFSE before opsonization with HIV ϩ IgG and coculture with NK cells. Error bars indicate the SD of results from replicates, and significance was determined by comparing the percentages of ADCC as measured by the frequencies of AnV ϩ between siCEM cells (black bars) and cCEM cells (gray bars) for each opsonizing HIV ϩ IgG concentration (***, P Ͻ 0.001). Data represent results from one experiment representative of three. Two-way ANOVAs with Tukey's multiple-comparison test were used for all comparisons in panels A, B, C, and D.
Targeting HIV-Infected Cells by ADCC ® ADCC-GTL assay in terms of the maximum frequency of apoptotic cells generated with HIV ؉ IgG-opsonized sorted infected CEM cells (P Ͻ 0.05 for the comparisons between AnV ؉ and GzB ؉ for all HIV ؉ IgG concentrations higher than 1 g/ml) (Fig. 6C). Of note, the concentration of polyclonal HIV ؉ IgG needed to obtain an equivalent frequency of AnV ϩ target cells in the ADCC-AnV assay was at least 10 times lower with coated CEM cells than with sorted infected CEM target cells (compare Fig. 6A, C, and D to Fig. 1B and Fig. S2). Staining of coated CEM cells and sorted infected CEM cells with increasing doses of HIV ؉ IgG demonstrated that the superior killing of coated CEM cells with HIV ؉ IgG in ADCC-AnV assays was consistent with the binding potential of HIV ؉ IgG to these target cells (Fig. S5). Indeed, equivalent concentrations of HIV ϩ IgG bound a higher frequency of coated CEM cells with a higher MFI than sorted infected CEM cells ( Fig. S5A and B). When MAb 2G12 was used instead of HIV ؉ IgG to stain these target cells, the MFI of binding was higher on sorted infected CEM cells than on coated CEM cells (Fig. S5D). Therefore, the superior binding and ADCC activity characteristics of HIV ϩ IgG seen with coated CEM target cells were not due to the amount of Env exposed on coated CEM cells versus sorted infected CEM cells but rather suggested that a majority of the anti-Env Abs in polyclonal HIV ϩ IgG recognized epitopes exposed by monomeric/linear Env rather than the native trimeric, closed Env conformation. We next explored the characteristics of binding of the MAb panel to sorted infected CEM cells and coated CEM cells side by side and the frequency of AnV ϩ cells generated in the ADCC-AnV assay when target cells were opsonized using 15 g/ml of each Ab in the panel (Fig. 7 and Fig. S6). Coated CEM cells were used here as a surrogate for the bystander CEM cells present in CEM cell cultures subjected to 4 days of HIV infection. As expected, the sorted infected CEM cells, like newly infected CEM cells, were preferentially recognized by CD4bs (b12, VRC01, NIH45-46, 3BNC117), 10-1074, and 2G12 MAbs (Fig. 3 and 7; see also Fig. S3A and B and S6). Of note, the PGT121 MAb poorly recognized sorted infected CEM cells compared to newly infected CEM cells. In contrast, coated CEM cells, like bystander CEM cells, were recognized by Abs to CD4i epitopes (A32, C11, and N12-i2) and 2G12, which binds to both "open" and "closed" conformations of Env (16). However, unlike the results seen with bystander CEM cells, we observed no binding of CD4bs MAbs or of PGT121 to coated CEM cells ( Fig. 3 and  7; see also Fig. S3B and S6C and D). The pattern of Ab binding to sorted infected CEM cells and coated CEM cells correlated with their ability to support ADCC activity, and, with the exception of HIV ؉ IgG, the global levels of ADCC using sorted infected CEM cells were higher than those seen using coated CEM cells as targets ( Fig. 7B and C). Furthermore, when sorted infected CEM cells were used as ADCC-AnV target cells, the frequency and MFI of Env binding were positively correlated with the frequency of AnV ϩ target cells generated in this ADCC assay (Fig. S7). This was also the case when coated CEM cells were used as target cells, though the correlation was weaker (data not shown).
ADCC activity of plasma or IgG isolated from HIV ؉ subjects following blocking of CD4i epitopes with Fab fragments. Abs to cluster A-like epitopes, such as the prototypical A32 MAb, have been implicated as dominant ADCC-competent Abs in plasma from HIV-infected individuals. However, most of the work supporting this view used rgp120-coated target cells or target cells infected with HIV bearing mutant Nef and/or Vpu. Such HIV-infected cells retain cell surface CD4, which favors the assumption by Env of an open conformation (4,15,34). To evaluate the contribution of Abs with specificities that overlap those of A32 in HIV ϩ IgG and individual plasma samples from HIV ϩ subjects to the ADCC of coated CEM cells and sorted infected CEM cells, we pretreated these target cells with 10 g/ml of A32 Fab fragment before opsonization with HIV ϩ IgG or HIV ϩ plasma (Fig. 8). As shown in Fig. 8A (see also Fig. S8), preincubation of target cells with 10 g/ml of A32 Fab abolished the binding and ADCC competence of A32 MAb with respect to coated CEM cells. Pretreatment with the A32 Fab significantly reduced the ADCC activity mediated by HIV ؉ IgG with respect to coated CEM cells. On average, there were 11%, 30%, 56%, and 72% decreases in the frequencies of AnV ϩ coated CEM cells generated when decreasing concentrations (i.e., 10, 3.3, 1.1, and 0.37 g/ml) of HIV ϩ IgG were used to opsonize A32 Fab with pretreated coated CEM cells (P Ͻ 0.001 for comparisons between no Fab treatment and A32 Fab pretreatment for all HIV ϩ IgG concentrations; 2-way analysis of variance [ANOVA], Tukey's multiple-comparison test). In contrast, pretreatment with the A32 Fab fragment failed to reduce the ability of HIV ϩ IgG to support ADCC-AnV activity on sorted infected CEM cells at any of the HIV ϩ IgG concentrations tested (Fig. 8B). These results suggest that HIV ϩ IgG contains anti-Env Abs with specificities that overlap those of A32 and support the idea of the ADCC activity of coated CEM cells. Furthermore, HIV ϩ IgG contains anti-Env Abs to epitopes other than A32 that are also capable of supporting ADCC of both coated CEM cells and sorted infected CEM cells (Fig. 8B). We next used individual plasma samples from 10 HIV-infected individuals to opsonize coated CEM cells (Fig. 8C) and sorted infected CEM cells (Fig. 8D). Pretreatment with the A32 Fab fragment alone was able to modestly but significantly reduce the ADCC competence of these HIV ؉ plasma samples for coated CEM target cells. On average, A32 Fab pretreatment of coated CEM cells reduced ADCC activity by 16.2% (P ϭ 0.037, Wilcoxon test) whereas A32 Fab pretreatment of sorted infected CEM cells had no effect on the frequency of AnV ϩ sorted infected CEM cells generated in the ADCC-AnV assay (Fig. 8C  and D). This observation was consistent with sorted infected CEM cells expressing Env in a closed conformation unable to expose CD4i epitopes. In addition, by performing an ADCC-AnV assay with plasma from 10 HIV ϩ individuals on combined coated CEM cells and sorted infected CEM cells, we showed that anti-Env Abs in HIV ϩ plasma preferen-   Targeting HIV-Infected Cells by ADCC ® tially triggered the killing of coated CEM cells, as previously observed with HIV ؉ IgG ( Fig. 6D and Fig. 9). Indeed, the frequency of AnV ϩ target cells was between 3 to 7 times higher for coated CEM cells than for sorted infected CEM cells, depending on the concentration of HIV ؉ IgG used for opsonization (Fig. 6D), and was on average between 4.5 and 8.2 times higher for coated CEM cells than for sorted infected CEM cells when 15 and 1.5 g/ml of total IgG from the 10 HIV ϩ plasma samples were used, respectively ( Fig. 9A and 9B).

DISCUSSION
We describe here a new method for ADCC quantification that measures the frequency of AnV ϩ cells as a readout for dead and dying cells. Using this assay and a panel of anti-Env MAbs to opsonize target coated CEM cells, newly infected CEM cells, bystander CEM cells, and sorted infected CEM cells, we confirmed that Abs to cluster A epitopes such as A32 and C11 predominantly recognized coated CEM cells and bystander CEM cells and supported killing of both types of target cells through ADCC. In contrast, CD4bs and V1/V2/V3 loop-specific Abs known to bind to the closed Env conformation (41) predominantly recognized newly infected CEM cells and sorted infected CEM cells, efficiently supporting their killing through ADCC. Env-specific Abs present in HIV ϩ plasma samples or HIV ϩ IgG supported the killing of all targets tested through ADCC, although a majority of this ADCC activity was directed toward epitopes found on the open Env conformation and exposed on coated/bystander CEM cells. A subset of the Env-specific Abs present in HIV ϩ IgG and HIV ϩ plasma recognized the Env closed conformation present on newly infected CEM cells and sorted infected CEM cells. Blocking experiments confirmed that the Ab specificities that overlapped A32-like epitopes in HIV ϩ plasma or HIV ϩ IgG supported the ADCC of coated CEM cells but not that of sorted infected CEM cells (see model in Fig. S9 in the supplemental material).
The ADCC-AnV assay that we describe here is easy to perform, high throughput, specific for target cells expressing HIV Env, and more sensitive than the ADCC-GTL assay. Unlike the widely used rapid fluorometric ADCC (RFADCC) assay, which quantifies membrane exchange between target and effector cells rather than ADCC activity, the ADCC-AnV assay measures ADCC activity by identifying and quantifying apoptotic target cells (42,43). Staining for AnV to measure ADCC activity is rapid, as AnV ϩ target cells are detected after a 1-h incubation with effector cells. NK cells were confirmed to be the main ADCC-competent effector cell in the assay, and the MFI of CD16 expression on NK cells was directly associated with the ADCC-AnV assay readout. This assay was used to measure ADCC activity of Env-specific BnAbs and NnAbs MAb, HIV ϩ IgG, and plasma from individual HIV ϩ subjects by the use of target CEM cells either coated with rgp120 from HIV-Bal or infected with a virus expressing Env-Bal and HSA. Labeling target cells with CFSE/PKH26 or staining them with anti-HSA allowed us to specifically gate on infected cells (HSA ϩ ) and/or bystander cells (HSA Ϫ ) exposed to shed gp120 by the use of flow cytometry and to include uncoated/uninfected CEM cells as within-well internal negative controls. Some investigators perform ADCC assays using an NK cell line expressing CD16 as a source of effector cells (30). The advantages of using an NK cell line are that the cells are readily available and are consistent from experiment to experiment. Employment of an NK cell line would eliminate the need for large blood draws (i.e., leukapheresis) to obtain enough NK cells from the same source for ADCC assays. It would also reduce the cost associated with isolation of NK cells from PBMCs. The use of an NK effector cell line would be worth exploring as a way to improve the ADCC assay described in this paper.
A panel of BnAbs and NnAbs to various HIV Env epitopes was used to assess their binding and ADCC competence. Anti-Env Abs with specificity to the CD4bs (i.e., b12, VRC01, NIH45-46, and 3BNC117) and glycan-dependent V3 loop (i.e., PGT121, 10-1074), but not to cluster A-specific MAbs, bound newly infected CEM cells. These results were consistent with wild-type Nef and Vpu expression in newly infected CEM downregulating cell surface CD4, which prevented the interaction with gp120 required to open the Env conformation that exposes CD4i epitopes. The Ab panel bound sorted infected CEM cells similarly to newly infected CEM cells with 2 exceptions. PG16, a V1/V2 loop-specific Ab recognizing a quaternary epitope on trimeric Env, bound only sorted infected CEM cells, whereas PGT121, a V3 loop-specific Ab recognizing the Env closed conformation, bound only newly infected CEM cells. In contrast, 10-1074, another V3 loop-specific Ab, bound both newly infected CEM cells and sorted infected CEM cells. For both newly infected CEM cells and sorted infected CEM target cells, the frequency and MFI of anti-Env binding were correlated with the frequency of AnV ϩ target cells generated in the ADCC-AnV assay.
Anti-cluster A-specific NnAbs A32 and C11 bound both bystander CEM cells and coated CEM cells but not newly infected CEM cells or sorted infected CEM cells. This finding is consistent with bystander and coated CEM target cells presenting gp120 in an open conformation, as has been reported previously by others (16,17). Surprisingly, bystander CEM cells, unlike coated CEM cells, were also recognized, though weakly, by the CD4bs-specific BnAbs VRC01, NIH45-46, and 3BNC117. In HIV-infected cell cocultures, gp120 shed by infected cells binds bystander cells by virtue of its interaction with CD4 on these cells, preventing the binding of CD4bs-specific BnAbs (16,33). Thus, the binding of CD4bs-specific BnAbs to bystander CEM is unlikely to be due to the recognition of shed gp120 from newly infected CEM cells. The low level of binding of CD4bs Abs to bystander CEM cells might have been due to recognition of defective viral particles attached to uninfected cells that originated from the surrounding newly infected CEM cells, which continuous produce HIV, or from the viral inoculum used to infect CEM cells. In line with this, Lee et al. suggested that the binding and ADCC activity observed with bystander CEM can be explained by the attachment of viral particles present in the inoculum, which has been shown to generate intermediate levels of p24 (33). Supporting this idea, our bystander CEM cells expressed insufficient HSA levels to be detectable as HIV ϩ by flow cytometry and p24 levels between those of newly infected CEM cells and uninfected CEM cells. Enough defective viral particles may be present to bind bystander CEM cells in a manner that exposes the CD4i epitopes that result from the formation of CD4-gp120 complexes while also maintaining gp120 trimers as previously suggested (33,44). However, we cannot formally exclude the possibility that the bystander CEM cells were present at an early stage of the infection. In fact, we observed that incubation of uninfected CEM cells with supernatant from sorted infected CEM cells, which contains shed gp120 and viral particles, did not block the recognition of surface CD4 by detector OKT4 Ab (data not shown) and therefore cannot account for the partial CD4 downregulation seen on bystander CEM cells. Activity during an early stage of the infection might result in a partial downregulation of CD4 by the early expressed HIV Nef protein, as observed on HSA Ϫ bystander cells, allowing some HIV Env to remain in a closed conformation recognized by CD4bs BnAbs and some HIV Env to expose CD4i epitopes due to interactions with CD4. Taking the data together, the challenges inherent in interpreting the anti-Env-specific BnAb and NnAb binding and ADCC competence results and the infection status of recently infected CEM cells provided the impetus to generate sorted infected CEM cells, which were virtually 100% HIV infected based on HSA, CD4, and p24 expression patterns. The availability of sorted infected CEM cells allowed us to compare these cells with coated CEM cells, which exposed CD4i epitopes in a manner similar to that seen with bystander cells in HIV-infected cocultures. Comparisons of sorted infected CEM cells and coated CEM target cells revealed that CD4bs-specific Abs exclusively recognized sorted infected CEM cells whereas the CD4i-specific NnAbs exclusively bound coated CEM cells.
ADCC-competent Abs in HIV ϩ plasma have been proposed to preferentially target the open Env conformation (15,20). In line with this, Abs to cluster A determinants, such as A32, were described previously as dominant ADCC-competent Abs in HIV ϩ plasma (20,25). However, this observation was made using target cells coated with rgp120 or infected with a virus unable to downregulate CD4, each of which exposes Env in an open conformation. To challenge this, we compared the capacities of polyclonal HIV ϩ IgG and individual HIV ϩ plasma to trigger ADCC of coated CEM cells and sorted infected CEM side by side, assuming that coated CEM cells can be used to measure levels of ADCC-competent Abs targeting an open Env/CD4i conformation and that sorted infected CEM cells can be used to measure levels of ADCC-competent Abs targeting a closed/trimeric Env conformation. We showed that polyclonal HIV ϩ IgG and HIV ϩ plasma elicited ADCC responses to both coated CEM cells and sorted infected CEM cells. The levels of ADCC were on average 6 times higher for coated CEM cells than for sorted infected CEM target cells opsonized with HIV ϩ plasma. This confirmed that the monomeric gp120/CD4i conformation and, by extension, gp120 shed by infected cells are preferentially targeted by anti-Env Abs in HIV ϩ plasma. This is despite the results seen with MAb 2G12, which detects a conformation-independent Env epitope, staining sorted infected CEM cells with a higher MFI than coated CEM cells. Since the open Env conformation preferentially marks bystander cells, these HIV ϩ plasma Abs may contribute to the killing of uninfected cells rather than to controlling HIV infection. Their role in HIV control can now be examined by using sorted infected CEM cells as ADCC target cells.
We investigated the contribution of A32-like Abs to ADCC responses. A32 Fab fragments partially blocked ADCC mediated by HIV ϩ IgG-and HIV ϩ plasma-opsonized coated CEM cells but failed to block ADCC mediated by HIV ϩ Ig-opsonized sorted infected CEM cells. As expected, the reduction in ADCC activity was restricted to coated CEM cells exposing CD4i epitopes. This observation suggested that the level of A32-like Abs mediating ADCC activity in HIV ϩ plasma was relatively modest and that they were strictly directed toward bystander CEM cells. These results highlight the presence of ADCC-competent Abs in HIV ϩ plasma to epitopes other than those recognized by A32 MAbs, capable of supporting the efficient killing of bystander cells displaying an open Env conformation and, to a lesser extent, of infected cells displaying a closed Env conformation. It would be of great interest to evaluate the relevance of bystander cell killing through the ADCC in the global CD4 depletion which occurs in HIV-infected donors since it is well known that only a small fraction of CD4 ϩ cells are actually infected in vivo whereas the majority of apoptotic CD4 ϩ cells in the lymph nodes of HIV ϩ persons consist of bystander CD4 ϩ cells surrounding infected cells (17).
We envision that the ADCC-AnV assay described here using sorted infected CEM cells as target cells may be useful for immune monitoring of HIV vaccine trials and therapeutic approaches that aim to induce anti-Env-specific Abs. The ADCC-AnV assay would aid in distinguishing Abs with specificities directed at bystander cells, which may contribute to CD4 loss versus Abs able to recognize HIV-infected cells that support HIV control. The concept that Abs able to recognize HIV-infected cells can support their lysis through ADCC may have applications in the context of other viral infections. For example, both respiratory syncytial virus (RSV) and Ebola virus (EboV) encode forms of their viral glycoproteins that are secreted or shed from the infected cell surface such as occurs for HIV-infected cells (45)(46)(47)(48)(49). This phenomenon protects virus-infected cells. Anti-virus Abs bind the soluble glycoproteins, making them unavailable to bind infected cells. Strategies aimed at preventing shedding or at identifying epitopes maintained on virus-infected cells have the potential to improve Ab targeting of virally infected cells able to support ADCC. Cells and reagents. PBMCs used as effector cells in ADCC assays were obtained from HIV-uninfected subjects enrolled in the St Luc cohort of injection drug users or from a cohort of couples with discordant HIV characteristics. None of the study subjects met the criteria for consideration as HIV-exposed seronegative (HESN) subjects. PBMCs were isolated from leukapheresis samples by density gradient centrifugation, as previously described (50,51). Cells were frozen in 90% fetal bovine serum (FBS; Wisent BioProducts, St-Jean-Baptiste, QC, Canada)-10% dimethyl sulfoxide (Sigma-Aldrich, St. Louis, MO) and stored in liquid nitrogen until use. Thawed PBMCs were rested overnight in RPMI 1640 medium supplemented with 10% FBS, 2 mM L-glutamine, 50 IU/ml penicillin, and 50 mg/ml streptomycin (R10; all from Wisent) before use.

MATERIALS AND METHODS
CEM cells were obtained from the NIH AIDS Reagent Program, Division of AIDS (DAIDS), NIAID, NIH, as CEM.NKR.CCR5 cells (from Alexandra Trkola) (26,27,52). HIV-1 Bal rgp120 was obtained through the NIH AIDS Reagent Program (DAIDS, NIAID, NIH). Anti-HIV immune globulin (HIVIG; referred to here as HIV ϩ IgG), representing a pool of purified IgG from asymptomatic HIV-positive donors with CD4 ϩ counts above 400/l, was obtained from the National Agri-Food Biotechnology Institute (NABI) and the National Heart, Lung, and Blood Institute (NHLBI) through the NIH AIDS Reagent Program (DAIDS, NIAID, NIH) (53).
Targeting HIV-Infected Cells by ADCC ® Plasma from five healthy donors at low risk for HIV infection (referred to here as HIV Ϫ IgG) was obtained from blood draws and stored in acid citrate dextrose-containing vacutainers. The tubes were centrifuged, the liquid phase was pooled, and total IgG was quantified by enzyme-linked immunosorbent assay (ELISA). A Live/Dead fixable dead cell stain kit (Invitrogen, St Laurent, QC, Canada) was used to quantify dead cells by flow cytometry. For some experiments, plasma was obtained from HIV-infected individuals enrolled in the Montreal Primary Infection Cohort or the Canadian Cohort of HIV-infected Slow Progressors.
HIV infection of CEM target cells. HIV-infected CEM cells were generated by infecting CEM cells with a replication-competent NL4.3-based HIV-1 virus expressing all viral genes from the original NL4.3 backbone except the Env gene, which was replaced by a Bal Env gene and a reporter gene encoding heat-stable antigen (HSA, murine CD24) coexpressed with Nef by the use of an internal ribosome entry site (IRES) sequence. The NL4-3-Bal-IRES-HSA viral construct was a kind gift from Michel Tremblay (Laval University, Quebec, QC, Canada) (38). CEM cells were infected with HIV by adding supernatant from 293T cells cotransfected with NL4-3-Bal-IRES-HSA and vesicular stomatitis virus glycoprotein G (VSV-G) plasmids to 10 6 CEM cells by spinoculation at 2,000 ϫ g for 90 min and incubating these cells for 30 min at 37°C in a 5% CO 2 humidified incubator. After washing, the CEM cells were cultured in R10. Four days after infection, the CEM cells were on average 52% HSA ϩ (range, 45% to 73%).
In order to prepare infected CEM cells exclusively exposing Env in a closed conformation, newly infected CEM cells were stained with PECy7-conjugated anti-mouse CD24 Ab (BD Biosciences, Mississauga, ON, Canada), sorted for HSA expression using a FACSAria instrument (BD Biosciences), and expanded in vitro. The frequencies of CD4 ϩ , HLA-C ϩ , and BST-2 ϩ CEM cells were evaluated by surface staining with MAbs specific for CD4 (clone OKT4; BD Biosciences), HLA-C (clone DT-9; Biolegend, Burlington, ON, Canada), and BST-2 (NIH AIDS Reagent Program, DAIDS, NIAID, NIH), respectively, whereas the frequency of p24 ϩ CEM cells was evaluated by intracellular staining using a phycoerythrin (PE)-conjugated anti-p24 antibody (clone KC57; Beckman-Coulter, Mississauga, ON, Canada). Sorted infected CEM cells expressed levels of ligands for NKG2D that were no higher than those seen with uninfected CEM cells (unpublished results) (54).
IgG ELISA. To detect the total amount of IgG in HIV Ϫ and HIV ϩ plasma samples, we used a human IgG ELISA quantitation set (Bethyl Laboratories, Montgomery, TX) per the manufacturer's instructions.
Target cell labeling. All target cells were stained with the green fluorescent cytosolic cell dye carboxyfluorescein succinimidyl ester (CFSE; Thermo Fisher Scientific, St. Laurent, QC, Canada) to distinguish them from PBMCs or NK effector cells. CFSE staining was performed per the manufacturer's instructions and as previously described (55).
For binding experiments and ADCC assays using two target cells combined (i.e., rgp120-coated CEM cells or infected CEM cells 4 days postinfection (p.i.) with HIV combined with uncoated or uninfected CEM cells), CFSE ϩ cells were also stained with PKH26 red fluorescent membrane cell dye (PKH26 red fluorescent cell linker kit; Sigma-Aldrich) to distinguish them from CFSE ϩ PKH26 Ϫ uncoated or uninfected CEM control cells, used as an internal control for nonspecific binding and killing. PKH26 staining was performed as previously described (55).
Preparation of rgp120-coated CEM target cells. Labeled CEM cells were resuspended to reach a level of 1 ϫ 10 6 cells in 100 l of R10 to which was added 0.5 g of rgp120 (from the NIH AIDS Reagent Program) for 1 h at 37°C in a humidified 5% CO 2 incubator. Excess rgp120 was washed off with R10.
Preparation of effector cells. PBMCs or NK cells were used as ADCC effector cells. Cryopreserved, thawed PBMCs were resuspended to a level of 2 ϫ 10 6 cells per ml of R10 and rested overnight in a 37°C, humidified 5% CO 2 incubator. For the experiments performed with NK cells, the cells were enriched from PBMCs using a negative selection kit (EasySep human NK cell enrichment kit; STEMCELL, Vancouver, BC, Canada) per the manufacturer's instructions. This kit does not include antibodies to CD16 or FcR blocking Abs, either of which could have an impact on ADCC assays. NK cell purity (average, 93% Ϯ 7.2% CD56 ϩ cells) and CD16 expression were evaluated by surface staining with MAbs specific for CD56 (clone HCD56; Biolegend), CD16 (clone 3G8; Biolegend) and CD3 (OKT3; Biolegend). Ninety-nine percent of CD56 dim CD3 Ϫ cells and 64% of CD56 bright CD3 Ϫ cells were CD16 ϩ (see Fig. S10 in the supplemental material).
ADCC-AnV assays. For ADCC assays using coated CEM target cells, cells were stained with CFSE and PKH26 before being coated with rgp120. A total of 10 4 CFSE ϩ PKH26 ϩ coated CEM cells combined with 10 4 CFSE ϩ PKH26 Ϫ uncoated CEM cells were plated into the wells of a 96-well V-bottom plate in 50 l of R10. Target cells were opsonized by adding 50 l of predetermined concentrations of Abs (HIV ϩ IgG, HIV Ϫ IgG, BnAbs, NnAbs, or HIV ϩ plasma) for 20 min at room temperature (RT) in the dark. After Ab incubation, 100 l of PBMC effector (E) cells were added to each well containing opsonized coated CEM target cells (T) at an E:T ratio of 30:1. Plates were centrifuged at 300 ϫ g for 1 min to pellet the cells and incubated at 37°C in a humidified 5% CO 2 incubator for 1 h.
For ADCC assays using infected CEM cells 4 days p.i., cells were stained with PKH26, CFSE, and Live/Dead stain such that HIV-infected cells that died prior to the start of the ADCC assay could be gated out. A total of 10 4 CFSE ϩ PKH26 ϩ infected CEM cells 4 days p.i. combined with 10 4 CFSE ϩ PKH26 Ϫ uninfected CEM cells were plated into the wells of a 96-well V-bottom plate followed by Ab opsonization. At 20 min after incubation with Abs, NK effector cells were added to each well containing target cells at an E:T ratio of between 5:1 and 10:1, unless otherwise specified. Plates were centrifuged and incubated as described above for 1 h. Cells were then stained with anti-HSA Ab to distinguish newly infected CEM cells, which were LD Ϫ CFSE ϩ PKH26 ϩ HSA ϩ , from bystander CEM cells, which were LD Ϫ CFSE ϩ PKH26 ϩ HSA Ϫ .
For ADCC assays using sorted infected CEM target cells, 10 4 CFSE ϩ sorted infected CEM cells were plated into the wells of a 96-well V-bottom plates and opsonized with Ab for 20 min. NK effector cells were then added to these wells at an E:T ratio of between 5:1 and 10:1, unless otherwise mentioned. Plates were centrifuged and incubated as described above for 1 h. In ADCC experiments comparing sorted infected CEM cells and coated CEM target cells in separate wells, both types of target cells were stained with CFSE and opsonized with Abs in parallel. In experiments where sorted infected CEM cells and coated CEM target cells were combined 1:1 and plated in the same well, they were distinguished by staining sorted infected CEM cells with PKH26 before opsonization or by staining with anti-HSA Ab after incubation with effector cells.
We used a new method to quantify ADCC activity in target CEM cells. This method employed AnV as a readout to identify and quantify the frequencies of both early and late apoptotic target cells following incubation of effector and Ab-opsonized target cells. The effector and opsonized target cells in each well were cocultured for 1 h, washed with 1ϫ AnV binding buffer (BD Biosciences), and incubated with 100 l of the same buffer supplemented with 1 l of AnV (BD Biosciences) for 10 min at RT. Cells were washed and resuspended in 1ϫ AnV binding buffer and acquired using an LSR Fortessa or LSR Fortessa X-20 instrument and a high-throughput system (HTS; BD Biosciences). Percentages of ADCC (% ADCC) were obtained by calculating the average frequency of AnV ϩ target cells from duplicate wells after subtracting the frequency of AnV ϩ cells measured under the no-Ab negative-control conditions. Results were analyzed using FlowJo software v10.
For some experiments, Fab fragments of the MAb A32 were used to pretreat coated CEM cells and sorted infected CEM target cells prior to Ab opsonization. Fab (10 g/ml) was added to target cells, and incubation was performed for 20 min at RT before the addition of opsonizing Abs.
Ab panel used for ADCC and to characterize the Env structure on target cells. The Env expressed on coated CEM cells, newly infected CEM cells, bystander CEM cells, and sorted infected CEM cells was probed using a panel of anti-Env-specific BnAbs and NnAbs, HIV ϩ IgG, and HIV Ϫ IgG. These Abs targeted the following Env epitopes: CD4i anti-cluster A (A32 and C11), CD4i-coreceptor binding site (CoRBS) (N12-i2), CD4 binding site (CD4bs) (b12, VRCO1, NIH45-46 G54W, and 3BNC117), V3 glycan (PGT121), V3 loop (10-1074), the V1/V2 glycan (PG16), and the N-linked glycans on the gp120 outer domain (2G12). All these Abs were obtained from the NIH AIDS Reagent Program, with the exception of C11 and N12-i2. Table 1 lists these MAbs, their sources, and relevant references. The binding of unconjugated primary Abs was detected using an allophycocyanin (APC)-conjugated mouse anti-human IgG Fc-specific secondary MAb (BioLegend). Negative controls for staining included HIV Ϫ IgG and staining with secondary Ab alone.
Statistical analysis. Statistical analyses were conducted using GraphPad Prism 7.0 (La Jolla, CA, USA). Comparisons were conducted using nonparametric Wilcoxon tests to assess the significance of the differences between the two conditions and 2-way ANOVA tests with Tukey's, Sidak's, or Dunnett's correction for multiple variables. Correlation analyses were performed using a nonparametric Spearman test. An ␣ level of 5% was used for statistical significance.