Enhanced clearance of HIV-1–infected cells by broadly neutralizing antibodies against HIV-1 in vivo

See allHide authors and affiliations

Science  20 May 2016:
Vol. 352, Issue 6288, pp. 1001-1004
DOI: 10.1126/science.aaf1279

Insights into antibody therapy for HIV-1

Despite the success of antiretroviral therapy, HIV-1-infected individuals still harbor latent virus. Thus, other therapeutic strategies are needed. A single injection of a broad and potent monoclonal antibody targeting the HIV-1 envelope protein reduced viral loads in HIV-1-infected individuals, albeit only transiently. Lu et al. now report that antibody treatment not only blocked free virus from infecting new cells, it also accelerated the clearance of infected cells. Furthermore, Schoofs et al. demonstrate that therapeutic antibody treatment enhanced infected individuals' humoral response against the virus. Thus, neutralizing antibodies may be a promising therapy for HIV-1 because of their potential to reduce the viral reservoir.

Science, this issue pp. 1001 and 997


Antiretroviral drugs and antibodies limit HIV-1 infection by interfering with the viral life cycle. In addition, antibodies also have the potential to guide host immune effector cells to kill HIV-1–infected cells. Examination of the kinetics of HIV-1 suppression in infected individuals by passively administered 3BNC117, a broadly neutralizing antibody, suggested that the effects of the antibody are not limited to free viral clearance and blocking new infection but also include acceleration of infected cell clearance. Consistent with these observations, we find that broadly neutralizing antibodies can target CD4+ T cells infected with patient viruses and can decrease their in vivo half-lives by a mechanism that requires Fcγ receptor engagement in a humanized mouse model. The results indicate that passive immunotherapy can accelerate elimination of HIV-1–infected cells.

Broadly neutralizing antibodies (bNAbs) to HIV-1 can block acquisition and suppress viremia in chronically infected humanized mice and macaques (1, 2). In humans, a single infusion of 3BNC117, a bNAb that targets the CD4-binding site on the HIV-1 envelope glycoprotein gp160, led to a rapid but transient reduction in viral loads by an average of 1.48 log10 copies per ml (3).

Antibodies differ from small-molecule drugs that interfere with viral replication in that antibodies have the potential to affect the half-lives of both free virus and infected cells. Indeed, antibodies accelerate the clearance of free virions from the blood of macaques (4) and induce killing of infected cells in vitro by Fcγ receptor (FcγR)–mediated mechanisms (5, 6). However, the majority of infected cells die rapidly by apoptosis or pyroptosis (7, 8), and whether bNAbs can accelerate HIV-1–infected cell clearance in vivo has not been tested directly.

To examine the components that contribute to viral clearance in humans given a single infusion of 3BNC117, we adapted an existing model of HIV-1 viral dynamics (3, 9, 10). The model (11) (fig. S1) includes virus-producing infected cells, as well as transport of free plasma virus to lymphoid tissues (LTs) and vice versa. To this basic model, we added the feature that antibodies bind to virus particles, which leads to virus neutralization and loss of antibody. Measurements of the decline of antibody concentrations in healthy humans were fitted to a two-compartment model (12, 13) to obtain the parameters characterizing the intrinsic antibody decay rates and transport between tissue and plasma over the time scale during which viral loads decay in patients treated with 3BNC117 (fig. S2). The rate of free virus neutralization was fitted to the virus kinetics in 19 patients (fig. S3), but we focused on patients showing an initial monophasic viral load decline (2B3, 2C1, 2C5, 2D3, 2E1, and 2E2), which tended to coincide with those receiving a higher antibody dose (3).

This model is unable to recapitulate the kinetics of viral load decline for any of the 3BNC117-treated viremic patients (Fig. 1, green; and fig. S3). If we fit the overall extent of viral load decrease, the rate of viral load decay is predicted to be too fast. Conversely, matching the initial rate of viral load decline results in insufficient overall reduction of the viral load. Thus, we adjusted our model to incorporate a mechanism that includes antibodies acting to clear infected cells and explored the revised model to see if it provided additional reduction of virus over a longer time scale (11). The rates of free-virus neutralization and infected cell clearance are fit to the measured plasma viral load. Including cell clearance substantially improves the fit to patient data (Fig. 1, purple; fig. S3; and table S3), because reducing the number of infected cells in tissues results in a second-order decay in the plasma viral load over a longer time scale.

Fig. 1 Comparison of viral load measurements with best-fit model predictions.

Viral load measurements (filled circles, solid black lines) with best-fit model predictions (solid colored lines). Each green line shows the predicted viral load over time, normalized by its initial amount, VP(t)/VP(0), in a model in which antibody can only neutralize free virus particles. Each purple line shows a modified model in which antibody can also lead to clearance of infected cells. Only those patients with a day-1 viral load lower than baseline are shown. Open circles and dashed black lines represent data points that were not used for parameter estimation. Within each subfigure, we note the quantity Δlog10VP(tmin) = log10[VP,min/VP(0)], i.e., the viral load at the nadir and the time in days at which this occurs for the data (black letters), as well as the predictions for a model with free virus clearance only (green) and for a model that also includes infected cell clearance (purple). This predicted minimum for each patient and model is denoted with a star.

Our modeling clearly shows that the patient data cannot be explained if 3BNC117 acts only to neutralize free virions and thus makes them incapable of infecting target cells. Including infected cell clearance improves the fit but does not quantitatively recapitulate the data; the reasons for these results are noted in the supplementary materials (11). Further evidence for such a mechanism is indicated by comparison of our modeling and clinical data for patients treated with a constant high level of entry-inhibitor drugs like maraviroc (11, 14) (fig. S4).

To determine whether 3BNC117 can recognize the HIV-1 envelope (Env) trimer expressed on the surface of infected cells, we stained CD4+ T cells infected with HIV-1YU2 or primary isolates obtained from patients 2C1, 2C5, 2D3, and 2E5 before they were infused with 3BNC117 (3). Consistent with their neutralizing activity in TZM-bl assays, 3BNC117, PG16, and 10-1074 specifically stained HIV-1YU2–infected cells (fig. S5A). Also, 3BNC117 and 10-1074 recognized nearly all Gag+ cells infected with primary isolates from patients 2C1, 2C5, 2D3, and 2E5; however, the mean fluorescence intensity of staining for Env was lower than for HIV-1YU2 (fig. S5B), possibly because of lower levels of Env on the surface of cells infected with patient viruses or variable levels of tetherin antagonism by HIV-1 accessory protein Vpu (15). We conclude that 3BNC117 recognizes the HIV-1 Env glycoprotein on the surface of infected CD4+ T cells.

To examine whether bNAbs can accelerate clearance of infected cells in vivo, we performed adoptive transfer experiments using nonobese diabetic (NOD) mice with Rag1–/–Il2rgnull (NRG mice). To prevent spread of infection between human cells, animals and infected cell suspensions were treated with antiretroviral therapy (ART) before adoptive transfer. Anti–HIV-1 bNAbs or isotype control antibodies were administered 12 hours before infected cell transfer (fig. S6).

Either 3BNC117 alone (P = 0.0012) (Fig. 2A) or a combination of 3BNC117 and 10-1074 rapidly reduced the percentage of HIV-1YU2–infected cells among CD3+CD8 cells compared with that in mice treated with an isotype control (P < 0.0001) (Fig. 2B). Concomitant with reduction in the percentage of infected cells, cell-associated HIV-1 RNA levels were lower in bNAb-treated mice than in those treated with isotype controls (P = 0.0054) (Fig. 2B). These data indicate that bNAbs can accelerate clearance of HIV-1YU2–infected cells in vivo.

Fig. 2 bNAbs accelerate clearance of HIV-1–infected cells in vivo.

(A) The percentage of Gag+ cells among CD3+CD8 cells in 3BNC117 (600 μg) or mice treated with isotype control 5 hours after HIV-1YU2–infected cell transfer. (B) (Left) Percentage of Gag+ cells among CD3+CD8 cells in bNAbs (3BNC117 + 10-1074, 300 μg each) or mice treated with isotype controls 5 hours after HIV-1YU2–infected cell transfer. (Right) Cell-associated HIV-1 RNA was measured in enriched human cells extracted from the spleen of mice 5 hours after transfer, plotted as the ratio of HIV-1 RNA to the number of CD3+CD8 cells for each mouse. (C) Transfer experiments with cells infected by HIV2C1, HIV2C5, HIV2D3, or HIV2E5. Each dot represents one mouse. Lines represent median values. Data represent two to four independent experiments with a total of 6 to 13 mice per condition. *P < 0.05; **P < 0.005; ***P < 0.001; two-tailed Mann-Whitney U test.

To determine whether HIV-1–specific antibodies can accelerate clearance of cells infected with primary HIV-1 isolates, we repeated the adoptive transfer experiment described above using human CD4+ T cells infected with HIV2C1, HIV2C5, HIV2D3, or HIV2E5. As with HIV-1YU2, bNAbs accelerated clearance of cells infected with patient isolates (Fig. 2C). We conclude that bNAbs can accelerate clearance of CD4+ T cells infected with primary HIV-1 isolates.

To determine whether enhanced clearance of HIV-infected cells by antibodies depends on their ability to engage FcγR-expressing cells, we repeated the adoptive transfer experiments using bNAbs that carry mutations that specifically abrogate mouse FcγR binding [Arg substituted for Gly236 and for Leu328 (G236R/L328R or GRLR)] (16) (fig. S7). Although GRLR-bNAbs show normal levels of neutralizing activity in TZM-bl assays (17), they do not interact with cytotoxic or phagocytic cells and should therefore fail to accelerate clearance of HIV-1–infected cells in vivo. The frequency of infected cells remaining after treatment with GRLR-bNAbs was comparable to that found in mice receiving the isotype control (Fig. 3A) and significantly higher than that found in mice receiving wild-type bNAbs (P < 0.0001) (Fig. 3A). In addition to testing GRLR-bNAbs, we also blocked wild-type bNAb Fc-FcγR interactions using a combination of antibodies 2.4G2 and 9E9, targeting mouse FcγRs II/III and IV, respectively (fig. S7). Mice receiving FcγR-blocking antibodies failed to accelerate clearance of HIV-1–infected cells in response to bNAbs (Fig. 3B).

Fig. 3 FcγR engagement is required to facilitate clearance of HIV-1–infected cells.

(A) Percentage of Gag+ cells among CD3+CD8 cells in bNAbs, GRLR-bNAbs, or mice treated with isotype controls. (B) NRG mice were injected with mouse antibodies blocking FcγRII, III, and IV or isotype controls 6 hours before injection of bNAbs. The percentage of Gag+ cells among CD3+CD8 cells is shown. (C) Percentage of Gag+ cells among CD3+CD8 cells in mice receiving FcγRIV-blocking antibody or isotype control. (D) Infected cell clearance in chronically HIV-1YU2–infected hu-mice. The ratio of the percentage of Gag+ cells among CD3+CD8 cells before and 5 hours after 3BNC117, GRLR-3BNC117, or isotype control injection. Each dot represents one mouse. Lines represent median values. Data represent two to four independent experiments for each condition with a total of 7 to 12 mice per condition. *P < 0.05; **P < 0.005; ***P < 0.001; two-tailed Mann-Whitney U test.

Human immunoglobulin-G1 (IgG1), the isotype of 3BNC117 and 10-1074, binds with highest affinity to mouse FcγRI and FcγRIV (mFcγRI and mFcγRIV) (18). These receptors are murine orthologs of human FcγRI and FcγRIII (hFcγRI and hFcγRIII) (19), which are expressed on human monocytes and natural killer (NK) cells that can perform antibody-dependent cell-mediated phagocytosis (ADCP) or cytotoxicity (ADCC) in vitro (5). To examine the role of mFcγRIV in clearance of HIV-1–infected cells in the NRG mouse, we blocked this receptor specifically with an mFcγRIV-specific monoclonal antibody (anti-mFcγRIV). Mice treated with bNAbs plus anti-mFcγRIV were comparable to isotype-treated mice (Fig. 3C), which indicated that mFcγRIV is essential for accelerating clearance of HIV-1YU2–infected cells in vivo. We conclude that bNAbs accelerate infected cell clearance in NRG mice by a mechanism that requires mFcγRIV engagement.

We repeated adoptive transfer experiments in FcγR-humanized (hFcγR) mice that express only the human FcγRs (20). Similar to NRG mice, HIV-1YU2–infected cells were reduced in bNAb-treated hFcγR mice (P = 0.0064) (fig. S8). These data suggest that bNAbs can use human FcγRs and not just murine FcγRs to accelerate clearance of HIV-1YU2–infected cells in vivo.

To examine whether 3BNC117 accelerates clearance of HIV-1–infected cells in the context of chronic viral infection, we performed hemisplenectomy experiments in chronically infected humanized mice (fig. S9). The frequency of HIV-1YU2–infected cells before treatment was comparable in all groups of mice (fig. S10). As expected, mice treated with the Fc mutant antibody GRLR-3BNC117, which neutralizes HIV-1 but does not interact with effector cells, had lower frequencies of infected cells compared with isotype-treated controls (P = 0.0219) (Fig. 3D). However, mice treated with wild-type 3BNC117, which can mediate ADCC and ADCP, had substantially fewer infected cells compared with mice treated with GRLR-3BNC117 (P = 0.0167) (Fig. 3D). These data suggest that 3BNC117 can accelerate clearance of HIV-1YU2–infected cells in the context of chronic viral infection.

In contrast to ART, antibodies have the potential to engage host immune cells in defense against the virus. They do so by binding to cell-free virions, which accelerates clearance and prevents their entry into target cells. Antibodies can also bind to HIV-1 Env on the surface of infected cells to induce ADCC or phagocytosis. Finally, immune complexes can activate antigen-presenting dendritic cells to elicit adaptive immune responses (19).

Both neutralizing and non-neutralizing antibodies support anti–HIV-1 ADCC activity in vitro (5, 21), and Fc receptor binding is essential for optimal protection, postexposure prophylaxis, and therapy by bNAbs in animal models (17, 18, 2224). Moreover, ADCC has been indirectly associated with both control of and protection against infection (15, 2528). However, the rapid death of HIV-1–infected cells has made it difficult to establish that antibodies can accelerate the clearance of infected cells in vivo. Our mathematical analysis of patient data, and the antibody-mediated reduction in infected cells seen in adoptive transfer experiments, establish that bNAbs alter the half-life of infected cells. This observation may help explain why postexposure prophylaxis with bNAbs is more effective than ART in hu-mice (23).

Experiments with human cells in mice cannot fully recapitulate the human host; nevertheless, these experiments establish that antibodies can accelerate clearance of infected cells in vivo and do so by an FcγR-dependent mechanism. The finding that antibodies can clear infected cells in vivo has important implications for therapies aimed at HIV prevention and viral reservoir reduction or elimination.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S10

Tables S1 to S3

References (2937)

  • * D.K.M. led the computational studies.

References and Notes

  1. Supplementary analysis is available as supplementary materials on Science Online.
Acknowledgments: We thank all study participants who devoted time to our research. We thank the Rockefeller University Hospital Clinical Research Support Office and nursing staff for patient care and recruitment. We thank T. Eisenreich for help with mouse colony management, K. Yao and S. Hinklein for technical help, P. Smith for FcγR humanized mice, CellDex Therapeutics for providing 3BNC117 and 10-1074, and all members of the Nussenzweig laboratory for helpful discussion and advice. The data reported in this paper are tabulated in the main paper and in the supplementary materials. This work was supported by Collaboration for AIDS Vaccine Discovery grant OPP1033115 (M.C.N. and J.V.R.). This work was also supported, in part, by grant 8 UL1 TR000043 from the National Center for Advancing Translational Sciences (NCATS); NIH Clinical and Translational Science Award (CTSA) program; NIH Center for HIV/AIDS Vaccine Immunology and Immunogen Discovery (CHAVI-ID) 1UM1 AI100663-01 (M.C.N); Bill and Melinda Gates Foundation grants OPP1092074 and OPP1124068 (M.C.N); the Robertson Foundation to M.C.N.; German Research Foundation postdoctoral fellowship SCHO 1612/1-1 (T.S.); NIH grant F31 AI118555-01 (J.A.H.); the American Foundation for AIDS research (amfAR) Mathilde Krim Fellowship in Basic Biomedical Research (108977-57-RKVA) (S.B.); the Ragon Institute of Massachusetts General Hospital, Massachusetts Institute of Technology, and Harvard (A.K.C.); a NSF Graduate Research Fellowship under grant no. 1122374 (D.K.M.); and National Institute of Allergy and Infectious Diseases (NIH) grants AI100148-02 and AI081677-05 (M.C.N. and J.V.R.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. M.C.N. is a Howard Hughes Medical Institute Investigator. There are pending patent applications on the 3BNC117 and 10-1074 antibodies by Rockefeller University on which M.C.N. is an inventor. The patents are not licensed by any companies.
View Abstract

Navigate This Article