Comparison of Antibody-Dependent Cell-Mediated Cytotoxicity and Virus Neutralization by HIV-1 Env-Specific Monoclonal Antibodies

ABSTRACT Although antibodies to the human immunodeficiency virus type 1 (HIV-1) envelope glycoprotein have been studied extensively for their ability to block viral infectivity, little data are currently available on nonneutralizing functions of these antibodies, such as their ability to eliminate virus-infected cells by antibody-dependent cell-mediated cytotoxicity (ADCC). HIV-1 Env-specific antibodies of diverse specificities, including potent broadly neutralizing and nonneutralizing antibodies, were therefore tested for ADCC against cells infected with a lab-adapted HIV-1 isolate (HIV-1NL4-3), a primary HIV-1 isolate (HIV-1JR-FL), and a simian-human immunodeficiency virus (SHIV) adapted for pathogenic infection of rhesus macaques (SHIVAD8-EO). In accordance with the sensitivity of these viruses to neutralization, HIV-1NL4-3-infected cells were considerably more sensitive to ADCC, both in terms of the number of antibodies and magnitude of responses, than cells infected with HIV-1JR-FL or SHIVAD8-EO. ADCC activity generally correlated with antibody binding to Env on the surfaces of virus-infected cells and with viral neutralization; however, neutralization was not always predictive of ADCC, as instances of ADCC in the absence of detectable neutralization, and vice versa, were observed. These results reveal incomplete overlap in the specificities of antibodies that mediate these antiviral activities and provide insights into the relationship between ADCC and neutralization important for the development of antibody-based vaccines and therapies for combating HIV-1 infection. IMPORTANCE This study provides fundamental insights into the relationship between antibody-dependent cell-mediated cytotoxicity (ADCC) and virus neutralization that may help to guide the development of antibody-based vaccines and immunotherapies for the prevention and treatment of HIV-1 infection.

T he recent isolation of a new generation of monoclonal antibodies with remarkably potent and broad neutralizing activity against diverse human immunodeficiency virus type 1 (HIV-1) isolates has renewed interest in the use of antibodies to treat HIV-1 infection (1,2). Passive transfer experiments in animal models have shown that many of these antibodies can protect against HIV-1 or simian-human immunodeficiency virus (SHIV) challenge (3,4), and in some cases, they are able to suppress virus replication to undetectable levels when administered during chronic infection (5)(6)(7). While the ability to block viral infection is a defining property of neutralizing antibodies, nonneutralizing effector functions may also contribute to antiviral responses. The IgG constant (Fc) domain can recruit cellular mediators of antibody-dependent cell-mediated cytotoxicity (ADCC) and phagocytosis through interactions with Fc␥ receptors (Fc␥Rs) or initiate complement-mediated lysis by binding to soluble factors in plasma.
Studies of nonhuman primates and mice support a role for Fc␥R-dependent functions of antibodies in protection against immunodeficiency virus infection. Passive transfer experiments with Fc variants of an HIV-specific broadly neutralizing antibody (bNAb) revealed that protection of rhesus macaques against pathogenic SHIV challenge is dependent in part on Fc␥R interactions, but not on complement fixation (8,9). The preferential engagement of activating, but not inhibitory, Fc␥Rs was also shown to contribute to the clearance of cell-free virus by antibodies in murine models (10), and Fc␥R-mediated functions of bNAbs interfered with the establishment of persistent HIV-1 reservoirs in humanized mice (11). Thus, the therapeutic potential of HIV-1-specific antibodies may be significantly enhanced by optimizing Fc␥R-dependent antiviral activities.
Emerging evidence suggests that antibodies capable of engaging Fc␥RIIIa on NK cells to direct the lysis of virus-infected cells may be especially important for containing or preventing HIV-1 infection (12,13). ADCC responses are detectable in plasma shortly after the resolution of acute viremia and correlate inversely with disease progression (14)(15)(16)(17)(18)(19)(20). Greater ADCC responses have also been observed in individuals who exhibit elite control of HIV-1 in the absence of antiretroviral therapy (21,22). In the setting of mother-to-child transmission, higher ADCC activity in breast milk is associated with a lower risk of virus transmission by breastfeeding, and passively acquired ADCC correlates with reduced infant mortality (23,24). ADCC may also have contributed to the modest protection observed in the RV144 trial as suggested by exploratory analyses revealing an association between ADCC and reduced risk of infection among vaccinated subjects with low IgA titers (25). Although passive transfer of a nonfucosylated bNAb with increased affinity for Fc␥RIIIa did not enhance the protection of macaques against pathogenic SHIV challenge relative to the fucosylated antibody (26), several studies of nonhuman primates have also revealed correlations between vaccine-induced ADCC and complete protection or reduced postchallenge viral loads (27)(28)(29)(30)(31).
While these studies suggest that ADCC, and possibly other Fc␥R-dependent functions, contribute to the antiviral activity of HIV-1-specific antibodies, the properties of antibodies that mediate ADCC are not well defined. We therefore tested monoclonal antibodies to diverse epitopes of the HIV-1 envelope glycoprotein, including potent bNAbs and nonneutralizing antibodies, for their ability to direct NK cell lysis of cells infected with primary versus lab-adapted HIV-1 and SHIV isolates. These antibodies were also tested for binding to Env on the surfaces of virus-infected cells and for neutralization of viral infectivity. Our results show that although ADCC generally correlates with Env binding and neutralization, there are cases where these functions do not correspond, revealing differences in epitopes exposed on virions versus infected cells that differentiate these antiviral activities.

MATERIALS AND METHODS
Virus production. Virus stocks were produced by transfection of proviral DNA into HEK 293T cells using GenJet transfection reagent (SignaGen). Culture supernatants were collected 48 and 72 h posttransfection, cell debris was removed by centrifugation, and aliquots of virus-containing supernatant were stored at Ϫ80°C. Virus concentrations were determined by anti-p24 or anti-p27 enzyme-linked immunosorbent assay (ELISA). SHIV AD8-EO was provided by Malcolm Martin.
ADCC assay. ADCC activity was measured as previously described (31,41). CEM.NKR-CCR5 -sLTR-Luc target cells, which express luciferase (Luc) under the control of a Tat-inducible promoter, were infected by spinoculation in the presence of 40 g/ml Polybrene. At 4 days postinfection, infected cells were incubated with monoclonal antibodies and an NK cell line expressing human CD16 for 8 h. The dose-dependent loss of Luc activity was measured as an indication of antibody-mediated killing of virus-infected cells. Infected target cells incubated with NK cells in the absence of antibody were used to measure maximal Luc activity, and uninfected target cells cultured with NK cells were used to determine background Luc activity. Percent relative light units (RLU) were used to determine partial area under the ADCC curve (pAUC) values and antibody concentrations required for half-maximal killing (50% ADCC), as previously described (25,41). Differences between log 10 -transformed percent RLU values and 100% RLU, indicating no activity, were calculated. pAUC values were determined by multiplying the sum of these differences at the four highest antibody concentrations tested by the log 10transformed dilution factor of 2, yielding an area. Standard deviations (SD) of individual measurements were propagated to yield the SD of the pAUC.
Neutralization assay. Neutralization of viral infectivity was measured using a TZM-bl reporter cell assay, according to standard methods (42,43). In a flat-bottom 96-well plate, 5,000 cells per well were seeded the day before the neutralization assay. Either 4 ng p24 (HIV-1 NL4-3 ), 10 ng p24 (HIV-1 JR-FL ), or 20 ng p27 (SHIV AD8-EO ) of virus per well was incubated with serial dilutions of monoclonal antibody for 1 h at 37°C before being added to the reporter cells. After 3 days, luciferase activity in cell lysates was measured, and virus neutralization was calculated from reductions in RLU relative to cells incubated with virus but no antibody. Uninfected cells were measured to account for background luciferase activity. Partial area under the neutralization curve (pAUC) values and antibody concentrations for 50% neutralization (IC 50 ) were calculated by using the same methods as for the ADCC assay.
In contrast, cells infected with HIV-1 JR-FL and SHIV AD8-EO were susceptible to lysis only by bNAbs. With the exception of the oligomannose-specific antibody 2G12 and the MPER-specific antibodies 2F5, 4E10, and 10E8, HIV-1 JR-FL -infected cells were sensitive to all of the bNAbs (Table 1 and Fig. 1); however, SHIV AD8-EOinfected cells were resistant to all but a handful of antibodies. ADCC was detected against SHIV AD8-EO -infected cells for PGV04, 3BNC117, PGT126, PGT121, and 10-1074, but only PGV04 and PGT121 mediated potent killing at 50% ADCC concentrations of less than 100 g/ml (58 g/ml and 0.67 g/ml, respectively) ( Table 2). ADCC activity was also measured at lower antibody concentrations for HIV-1 JR-FL -infected cells than SHIV AD8-EO -infected cells (Fig. 1). These observations indicate that SHIV AD8-EOinfected cells are less sensitive to recognition by most HIV-1-specific antibodies, perhaps as a consequence of extensive adaptation of this virus for replication in rhesus macaques, as reflected by changes in the neutralization profile of SHIV AD8-EO relative to the parental HIV-1 AD8 strain (76). In comparison to HIV-1 NL4-3 , the greater resistance of HIV-1 JR-FL -and SHIV AD8-EO -infected cells to  (25,41). Standard deviations were calculated from triplicate measurements. Red indicates potent ADCC or neutralization (Neut) (top tertile), yellow indicates intermediate activity, green indicates weak activity (bottom tertile), and blue indicates a lack of significant antiviral activity (P Ͼ 0.01). The tertiles for ADCC activity and neutralization were calculated separately from the respective values against HIV-1 NL4-3 . n.s., not significant.
ADCC is also consistent with the resistance of these primary isolates to neutralizing antibodies.
ADCC activity correlates with binding to Env on the surfaces of virus-infected cells. Antibody binding to virus-infected cells is a prerequisite for ADCC. Cells infected with HIV-1 NL4-3 , HIV-1 JR-FL , and SHIV AD8-EO were therefore stained with each of the HIV-1 Env-specific monoclonal antibodies and analyzed by flow cytometry to determine the extent to which binding correlates with susceptibility to ADCC (Fig. 2). The geometric mean fluorescence intensities of Env staining (Table 3) were compared by nonparametric Spearman correlation to partial area under the curve values for ADCC (Fig. 3), which capture responses for antibodies that did not achieve 50% ADCC at concentrations less than 100 g/ml.
Antibody binding correlated with ADCC for each of the three viruses tested. Whereas binding was highly predictive of ADCC for HIV-1 JR-FL (P Ͻ 0.0001), less robust, but nevertheless significant, associations were also observed for HIV-1 NL4-3 (P ϭ 0.0104) and SHIV AD8-EO (P ϭ 0.0422) (Fig. 3). The relationship between binding and ADCC for HIV-1 NL4-3 -infected cells reflects greater variability in these measurements and instances of antibody binding in the absence of detectable ADCC. For instance, 2F5 and 4E10 stained cells infected with HIV-1 NL4-3 (Fig. 2), but they did not mediate cell lysis (Fig. 1). The reason for this discrepancy is unclear, but it may be related to the limited accessibility of these antibodies for engagement by Fc␥Rs on NK cells when bound to virus-infected cells due to their specificity for an epitope consist-ing of phospholipids and sequences exposed at the base of gp41 (77)(78)(79)(80). For SHIV AD8-EO -infected cells, there was almost complete correspondence between antibody binding and susceptibility to ADCC ( Fig. 1 and 2); however, the significance of this relationship was limited in comparison to HIV-1 JR-FL by lower responses that were detectable for a smaller subset of antibodies.
Correlation of ADCC activity with virus neutralization. The antiviral activity of HIV-1-specific antibodies is typically defined by their ability to neutralize viral infectivity. Each of the HIV-1 Env-specific monoclonal antibodies was therefore tested for neutralization of HIV-1 NL4-3 , HIV-1 JR-FL , and SHIV AD8-EO to investigate the relationship between their ability to block viral infectivity and to mediate NK cell lysis of virus-infected cells. Antibody concentrations for 50% neutralization (IC 50 ) ( Table 2) were calculated from neutralization curves (Fig. 4), and corresponded well to previously published data (33,50,51,59,63,(81)(82)(83)(84). pAUC values for neutralization were also determined ( Table 1) and compared to pAUC values for ADCC by nonparametric Spearman correlation (Fig. 5).
Neutralization correlated with ADCC for all three viruses. This relationship was strongest for HIV-1 JR-FL (P ϭ 0.0002), followed by SHIV AD8-EO (P ϭ 0.0010) and HIV-1 NL4-3 (P ϭ 0.0302) (Fig.  5). Several instances of neutralization in the absence of detectable ADCC were observed, and in most cases, antibodies that directed ADCC against cells infected with a particular isolate also neutralized that virus (Tables 1 and 2). Indeed, all of the antibodies with TABLE 2 Antibody concentrations for 50% ADCC and 50% neutralization a a Antibody concentrations (g/ml) for half-maximal ADCC (50% ADCC) and virus neutralization (IC50) were calculated as previously described (41). Standard deviations were calculated from triplicate neutralization curves. Red indicates potent ADCC or neutralization (top tertile), yellow indicates intermediate activity, green indicates weak activity (bottom tertile), and blue indicates less than 50% activity at 100 g/ml. The tertiles for ADCC activity and neutralization were calculated separately from the respective values against HIV-1 NL4-3 .
ADCC activity against HIV-1 JR-FL -or SHIV AD8-EO -infected cells also neutralized these viruses (Table 1). Furthermore, antibody concentrations for 50% neutralization were generally lower than for 50% ADCC (Table 2). However, in the case of HIV-1 NL4-3 , a number of instances of ADCC in the absence of detectable neutralization were observed (Table 1), and antibody concentrations for 50% ADCC were sometimes lower than for 50% neutralization (Table 2). Notably, 240D and F240, which recognize epitopes exposed in gp41 that are nonneutralizing, mediated efficient NK cell lysis of HIV-1 NL4-3 -infected cells (50% ADCC of 0.12 and 0.066 g/ml, respectively) ( Table 2). Overall, these results indicate that most antibodies that are able to bind to Env expressed on the surfaces of virus-infected cells to mediate ADCC are also able to bind to Env trimers on the surfaces of virions to neutralize infectivity. However, neutralization is not always predictive of ADCC, and for lab-adapted viruses such as HIV-1 NL4-3 , conformations of Env may be exposed on the surfaces of virus-infected cells that render them susceptible to antibody-dependent killing but are not relevant to blocking viral infectivity.

DISCUSSION
Increasing evidence suggests that Fc␥ receptor-dependent functions of antibodies are important for the optimal antiviral activity of HIV-1-specific antibodies (8)(9)(10)(11)85). Thus, a better understanding of the relationship between neutralizing and nonneutralizing functions of antibodies is needed to guide the development of immunotherapies and antibody-based vaccines for the treat-ment and prevention of HIV-1 infection. In the present study, antibodies targeting diverse epitopes of the HIV-1 Env protein were tested for ADCC against cells infected with HIV-1 or SHIV isolates, and their ADCC activity was compared to their ability to bind to Env expressed on the surfaces of virus-infected cells and to neutralize viral infectivity. Consistent with recent findings (86,87), ADCC activity correlated with Env binding and with neutralization for each of the viruses tested, indicating that these functions are largely overlapping; however, instances of ADCC in the absence of detectable neutralization and neutralization in the absence of detectable ADCC were also observed, revealing differences in Env epitopes exposed on the surfaces of HIV-1-infected cells and virions that confer susceptibility to these antiviral activities.
Sensitivity to ADCC corresponded closely with sensitivity to neutralization for antibody-resistant primary isolates. All of the antibodies that directed ADCC against HIV-1 JR-FL -and SHIV AD8-EOinfected cells also neutralized these viruses. Thus, most antibodies capable of binding to Env on the cell surface and directing the lysis of virus-infected cells are also able to bind functional Env trimers on virions to block viral infectivity; however, this was not always the case for HIV-1 NL4-3 . In accordance with the greater exposure of the Env proteins of lab-adapted viruses to antibodies (73-75), HIV-1 NL4-3 -infected cells were generally more sensitive to ADCC, both in terms of the number of antibodies and the magnitude of responses. Although ADCC corresponded with neutralization for many of the bNAbs, cells infected with HIV-1 NL4-3 were also susceptible to killing by nonneutralizing antibodies. This was particularly evident for F240 and 240D, which mediated potent ADCC against HIV-1 NL4-3 -infected cells despite their inability to block viral infection (Fig. 4) (88). ADCC responses were also detected for 98-6 and 126-7, which recognize epitopes exposed in the postfusion conformation of gp41 (39,70,71), and for A32 and C11, which target CD4i epitopes of the gp120 inner domain (66,67,(89)(90)(91). The ADCC activity of nonneutralizing antibodies against HIV-1 NL4-3 -infected cells suggests that Env epitopes that are not antibodies. Partial area under the curve values (pAUC) for ADCC activity were calculated from percent RLU measurements at the four highest antibody concentrations tested, as previously described (25,41). pAUC values for HIV-1 NL4-3 , HIV-1 JR-FL , and SHIV AD8-EO were compared to the geometric mean fluorescence intensities (gMFIs) of Env staining on the surfaces of virus-infected cells by Spearman correlation. relevant to blocking viral infectivity are exposed on the surfaces of cells infected with lab-adapted viruses that render them susceptible to ADCC. The striking difference in the susceptibility of HIV-1 NL4-3 -versus HIV-1 JR-FL -infected cells to ADCC illustrates the importance of using primary virus isolates for studying antiviral functions of antibodies. To facilitate virus replication in the face of ongoing immune responses, the HIV-1 envelope glycoprotein has evolved structural features that make it inherently resistant to antibodies (92)(93)(94)(95). These features can become attenuated as virus is passaged in T cell lines, accounting for the well-documented increase in the susceptibility of lab-adapted HIV-1 to neutralizing antibodies (73-75, 96, 97). The much greater sensitivity of HIV-1 NL4-3infected cells to opsonization by Env-specific monoclonal antibodies, including antibodies that do not neutralize this virus, suggests that this is also true for ADCC. These observations therefore advocate for the use of primary HIV-1 isolates expressing physiologically relevant conformations of Env on the surfaces of infected cells for studies investigating ADCC or other Fc␥R-dependent functions of antibodies.
Contrary to earlier reports identifying A32 as a potent mediator of ADCC (98,99), we found that HIV-1-and SHIV-infected cells are highly resistant to lysis by this antibody. Indeed, ADCC was detected against HIV-1 NL4-3 -infected cells only at high concentrations of A32, and not at all against HIV-1 JR-FL -or SHIV AD8-EO -infected cells. Recent evidence suggests that this disparity probably reflects differences in the methods used to measure ADCC. A32 is specific for an epitope on the inner domain of gp120 that is normally occluded in the unliganded Env trimer but can be exposed upon CD4 engagement (53,94). The accumulation of Env-CD4 complexes on the surfaces of cells infected with viruses deficient for Nef-and/or Vpu-mediated CD4 downregulation was accordingly found to increase exposure of this epitope (100,101). Shed gp120 released from productively infected cells was also found to sensitize uninfected bystander cells to A32-mediated ADCC (102). These studies help to explain the robust ADCC activity for A32 initially observed using target cells pulsed with soluble gp120 or chronically infected with a Nef-deficient HIV-1 (98). The measurement of NK cell degranulation as a surrogate for the direct lysis of virus-infected cells, which cannot differentiate ADCC responses to virus-infected cells from responses to uninfected cells coated with gp120, may also explain the detection of ADCC activity for A32 against cells infected with HIV-1 isolates that retain CD4 downmodulation (98,99). In accordance with this interpretation, other recent studies using ADCC assays that directly measure the elimination of virus-infected cells have found that viruses that downmodulate CD4 are resistant to A32-mediated lysis (100,101,103). These observations therefore further argue for the use of ADCC assays that directly measure the killing of cells infected with HIV-1 isolates expressing functional accessory proteins and native conformations of Env.
Instances of neutralization in the absence of ADCC include the MPER-specific antibodies 2F5, 4E10, and 10E8, and the glycanspecific antibody 2G12. The lack of ADCC activity for the MPER bNAbs is probably due in part to the lower affinity of these antibodies for Env on virus-infected cells, which is consistent with their specificity for an epitope consisting of gp41 sequences that are transiently exposed during fusion and phospholipids that are preferentially enriched in viral membranes (77)(78)(79)(80). Yet these antibodies still bound to HIV-1 NL4-3 -and HIV-1 JR-FL -infected cells, as indicated by levels of Env staining similar to other antibodies with ADCC against these viruses, such as the V2 apex bNAbs PG9 and PG16. The reason for this discrepancy in binding versus ADCC is unclear at this time but potentially reflects the orientation of MPER-specific antibodies bound to gp41, which may hinder their accessibility for engagement by NK cells. The lack of detected ADCC activity for 2G12 was also surprising considering the ability of this antibody to stain Env on the surfaces of HIV-1 NL4-3 -and HIV-1 JR-FL -infected cells. 2G12 is specific for a cluster of high-mannose glycans on the outer domain of gp120 that should not limit its accessibility (56,104); however, 2G12 has an unusual domain-swapped configuration and propensity for dimerization that may impair Fc␥R interactions (105,106). Although monomeric and dimeric forms of 2G12 were shown to mediate ADCC against a cell line expressing HIV-1 HXB2 gp160 (107), our data are consistent with recent reports that have found negligible ADCC activity for 2G12 against HIV-1-infected cells (86,98,103).
Overall, our results reveal a general correlation between ADCC and neutralization by HIV-1 Env-specific antibodies, which implies, perhaps not surprisingly, that most antibodies that are able to bind to functional Env trimers on virions to block infectivity are also able to bind to Env expressed on the surfaces of virus-infected cells to direct their elimination by ADCC. This correlation was imperfect, however, as several instances where these antiviral activities did not correspond were observed. These exceptions point to underlying differences in Env epitopes on the surfaces of virions and infected cells that differentiate susceptibility to neutralization versus ADCC. Hence, this study provides new insights into the relationship between neutralization and ADCC that may help to guide the development of antibody-based vaccines and immunotherapies for the prevention and treatment of HIV-1 infection.