The latent viral reservoir is the critical barrier for the development of a cure for HIV-1 infection. Previous studies have shown direct antiviral activity of potent HIV-1 Env-specific broadly neutralizing antibodies (bNAbs) administered when antiretroviral therapy (ART) was discontinued, but it remains unclear whether bNAbs can target the viral reservoir during ART. Here we show that administration of the V3 glycan-dependent bNAb PGT121 together with the Toll-like receptor 7 (TLR7) agonist vesatolimod (GS-9620) during ART delayed viral rebound following discontinuation of ART in simian–human immunodeficiency virus (SHIV)-SF162P3-infected rhesus monkeys in which ART was initiated during early acute infection. Moreover, in the subset of monkeys that were treated with both PGT121 and GS-9620 and that did not show viral rebound after discontinuation of ART, adoptive transfer studies and CD8-depletion studies also did not reveal virus. These data demonstrate the potential of bNAb administration together with innate immune stimulation as a possible strategy for targeting the viral reservoir.
The viral reservoir in latently infected CD4+ T lymphocytes1,2,3,4 is responsible for viral rebound in the vast majority of HIV-1-infected individuals who stop taking ART and represents the key challenge to a cure for HIV-1 infection5,6. Multiple strategies for an HIV-1 cure are currently being pursued. One hypothesis is that activation of reservoir cells may render them more susceptible to immune-mediated destruction7,8,9, but, to our knowledge, direct evidence that this strategy can reduce the viral reservoir in vivo has not previously been reported.
Potent HIV-1-specific bNAbs have been shown to reduce viraemia in untreated, chronically SHIV-infected rhesus monkeys10,11,12 and in HIV-1-infected humans13,14. Moreover, bNAbs have been reported to delay viral rebound in HIV-1-infected humans when the antibodies were administered at the time of discontinuation of ART15,16. These studies demonstrate that bNAbs can exert direct antiviral activity, but whether they can target the viral reservoir during ART suppression remains to be determined. Such an experiment would require that bNAbs no longer be present at therapeutic levels when ART is discontinued. To explore this concept, we evaluated the capacity of the potent neutralizing antibody PGT12117,18 and the TLR7 agonist vesatolimod (GS-9620)19,20 to target the viral reservoir in ART-suppressed, SHIV-SF162P3-infected rhesus monkeys.
We infected 44 Indian-origin rhesus monkeys (Macaca mulatta) by the intrarectal route with our rhesus peripheral blood mononuclear cell (PBMC)-derived stock of SHIV-SF162P321 and initiated daily subcutaneous administration of a pre-formulated ART cocktail (tenofovir disoproxil fumarate, emtricitabine, dolutegravir)20 on day 7 of infection (Extended Data Fig. 1). Monkeys had median plasma viral loads of 4.2–4.3 log RNA copies per ml (range 2.8–5.8 log RNA copies per ml) on day 7 before initiation of ART (Fig. 1a). Viral loads were balanced among the different groups. Following initiation of ART, plasma viraemia declined to undetectable levels in most monkeys by week 3 and in all monkeys by week 6.
All monkeys were treated with continuous daily ART for 96 weeks and then received one of the following interventions (n = 11 monkeys per group; Fig. 1b, Extended Data Fig. 1): sham (group 1), GS-9620 alone (group 2), PGT121 alone (group 3), or both PGT121 and GS-9620 (group 4). In groups 2 and 4, monkeys received ten oral administrations of 0.15 mg kg−1 GS-9620 (Gilead Sciences) every two weeks from weeks 96 to 114. In groups 3 and 4, monkeys received five intravenous infusions of 10 mg kg−1 PGT121 (Catalent Biopharma) every two weeks from weeks 106 to 114. Thus, in the combination group, monkeys first received five doses of GS-9620 alone and then received five doses of both PGT121 and GS-9620. ART was continued for 16 weeks after the final administration of PGT121 and GS-9620 to allow antibody washout before discontinuation of ART at week 130. Viral loads remained undetectable in all monkeys throughout the period of ART suppression with no evidence of viral ‘blips’ to week 130 (Fig. 1b).
Pharmacodynamics and pharmacokinetics
Triggering of TLR7 leads to innate immune activation22,23, and we previously reported that treatment with the TLR7 agonist GS-986 resulted in robust CD4+ T cell activation in rhesus monkeys20. In the present study, the related TLR7 agonist GS-9620 similarly led to activation of CD4+ T cells, as evidenced by increased CD69 and CD38 expression on CD4+ T cells on the day after each GS-9620 administration (Fig. 2a, Extended Data Fig. 2a; P = 0.001–0.03, Mann–Whitney tests). Moreover, GS-9620 activated natural killer (NK) cells (Fig. 2b; P = 0.001–0.01, Mann–Whitney tests) and monocytes (data not shown), and led to increased plasma levels of IFNα, IL-1RA, I-TAC, eotaxin, MIG, MCP-1, IL-1β, IL-6 and IP-10 (Extended Data Fig. 2b).
PGT121 was detected in serum after each of the five infusions from weeks 106 to 114 without evidence of induction of anti-drug antibodies (Extended Data Figs. 3, 4a). After the final PGT121 infusion, PGT121 levels fell rapidly, reflecting the short half-life of this human antibody in monkeys10. Serum PGT121 declined to undetectable levels (<0.5 μg ml−1) in most monkeys by week 120 and in all monkeys by week 122 (Extended Data Fig. 4a). PGT121 was also undetectable in cell lysates and supernatants from lymph nodes and colorectal biopsies from week 120 (Extended Data Fig. 4b). We previously defined a PGT121 level of 1 μg ml−1 as the threshold for viral rebound in SHIV-SF162P3-infected rhesus monkeys for PGT121-mediated virologic suppression10. Thus, PGT121 levels were below this rebound threshold in peripheral blood and lymphoid and gastrointestinal tissues for 8–10 weeks before discontinuation of ART at week 130.
Viral DNA was largely undetectable in PBMCs in all groups of monkeys at week 96 and week 120 using quantitative PCR with reverse transcription (RT–PCR) with a sensitivity of 3 DNA copies per 106 PBMCs20 (Fig. 3a, b), presumably as a result of the early initiation and the extended duration of suppressive ART. We still detected viral DNA in lymph nodes in the majority of monkeys at week 96 before the interventions, consistent with previous findings in rhesus monkeys treated with ART during acute SIVmac251 infection24. Following the interventions at week 120, we observed similar levels of viral DNA in the sham and GS-9620-alone groups. By contrast, monkeys treated with PGT121 and GS-9620 had lower levels of viral DNA in lymph nodes than did sham controls at week 120 (Fig. 3b; P = 0.004, Mann–Whitney test). We were unable to detect cellular RNA in all groups, both at week 96 and at week 120 (data not shown). These data suggest that bNAb administration with innate immune stimulation may have reduced the viral reservoir in these monkeys.
Cellular immune responses
Monkeys showed low levels of Gag-, Env-, and Pol-specific cellular immune responses at week 4, and these responses waned by week 96 and week 120 (Extended Data Fig. 5), presumably reflecting a lack of antigen stimulation during the extended period of ART suppression. Multi-parameter intracellular cytokine staining at week 120 corroborated these findings and showed minimal to no detectable CD8+ T cell responses in PBMCs in all groups of monkeys (Extended Data Fig. 6a). Lymph node CD8+ T cell responses and follicular CXCR5+CD8+ T cell responses were also marginal in all groups (Extended Data Fig. 6b). In particular, Gag-, Env- and Pol-specific CD8+ T cell responses in PBMCs and lymph nodes were not higher in monkeys that received PGT121 than in those that did not, suggesting that there was no ‘vaccinal effect’ following antibody administration in this study.
Viral rebound following ART discontinuation
At week 130, we discontinued ART in all monkeys and monitored plasma viral loads for 196 days to assess viral rebound (Fig. 4a). In the sham control group, 11 of 11 monkeys (100%) rebounded with high peak viral loads and moderate setpoint viral loads typical of SHIV-SF162P3 infection20,21. In the GS-9620-alone group, 10 of 11 monkeys (91%) rebounded, and in the PGT121-alone group, 9 of 11 monkeys (82%) rebounded. In the group treated with both PGT121 and GS-9620, however, only 6 of 11 monkeys (55%) rebounded by day 196 following ART discontinuation. The monkeys treated with PGT121 and GS-9620 that rebounded also showed a 2.6 log reduction in peak viral loads and a 1.5 log reduction in setpoint viral loads as compared with sham controls (Fig. 4b; P < 0.001, χ2 test comparing total viral load AUC). All monkeys that rebounded generated robust Gag-, Env- and Pol-specific cellular immune responses following viral rebound, which probably contributed to post-rebound virologic control (Extended Data Fig. 7a).
The median time to viral rebound in the sham group and in the GS-9620-alone group was 21 days, whereas the median time to viral rebound in the group that received PGT121 and GS-9620 was 5.3-fold longer at 112 days (Fig. 5a). Overall, the group that received PGT121 and GS-9620 demonstrated significantly delayed viral rebound compared with the sham group, and the PGT121-alone group showed a detectable but more modest effect (Fig. 5a, b; P = 0.0001, Kruskall–Wallis test comparing all groups; P < 0.001, censored Poisson regression model comparing PGT121+GS-9620 versus sham, PGT121 alone versus sham, and PGT121+GS-9620 versus PGT121 alone). These findings demonstrate that treatment with PGT121 and GS-9620 was superior to PGT121 alone in delaying viral rebound following ART discontinuation.
Monkeys treated with PGT121 and GS-9620 that did not rebound by day 196 following ART discontinuation had lower peak pre-ART plasma viral loads at week 1 of infection than those that did rebound (Fig. 5c; P = 0.03, Mann–Whitney test). Moreover, pre-ART viral loads correlated inversely with time to rebound in this group (Fig. 5c; P = 0.03, R = −0.62, Spearman rank-correlation test). Similar trends were observed in the monkeys that did not rebound in the other groups (Extended Data Fig. 7b). These data suggest that the extent of viral exposure before ART initiation was a key determinant of the therapeutic efficacy of PGT121 and GS-9620, probably by limiting initial seeding of the viral reservoir.
Adoptive transfer and CD8 depletion studies
Current viral DNA assays are insufficiently sensitive to predict viral rebound following ART discontinuation, as shown by the rapid rebound in 100% of sham controls (Fig. 4a), despite the absence of detectable viral DNA in PBMCs and lymph nodes in a subset of these monkeys at week 120 (Fig. 3b). These findings are also consistent with clinical observations that have shown viral rebound even in patients with undetectable levels of viral DNA25,26. To assess residual replication-competent virus, we first performed an adoptive transfer study using PBMCs and lymph node mononuclear cells (LNMCs) from five of the seven monkeys that did not rebound and for which we had sufficient LNMCs available from day 140 following ART discontinuation, including four of the five monkeys treated with PGT121 and GS-9620 and one of the two monkeys treated with PGT121 alone. We infused 30 million PBMCs and LNMCs from these five monkeys by intravenous infusion into naive rhesus monkeys, and we used a similar number of PBMCs and LNMCs from two monkeys treated with PGT121 and GS-9620 that showed transient rebound followed by virologic control as positive controls. All monkeys had undetectable plasma viral RNA on day 140 when PBMCs and LNMCs were collected. Adoptive transfer of cells from the two monkeys that rebounded led to efficient infection of the naive hosts, as shown by high plasma viral loads of 6.1–6.8 log RNA copies per ml in recipients by day 14–21 following adoptive transfer (Fig. 6a, red lines). These data show that replication-competent virus persisted in these two monkeys despite post-rebound virologic control. By contrast, adoptive transfer of cells from the five monkeys that did not rebound failed to transfer infection to naive hosts (Fig. 6a, black lines).
We next depleted CD8+ T cells and NK cells in the monkeys treated with PGT121 and GS-9620 that had undetectable viral loads on day 196 following ART discontinuation. Monkeys received a single intravenous infusion of 50 mg kg−1 of the anti-CD8α CDR-grafted rhesus IgG1 antibody MT807R1 (provided by K. Reimann, MassBiologics). Following infusion, all monkeys showed marked CD8 depletion in PBMCs (Extended Data Fig. 8) as well as in lymph nodes and colorectal biopsies (data not shown). In all the monkeys treated with PGT121 and GS-9620 that rebounded following ART discontinuation, plasma viral loads spiked to 6.0–6.8 log RNA copies per ml by day 7–14 after CD8 depletion (Fig. 6b, red lines). By contrast, in the five monkeys that were treated with PGT121 and GS-9620 and did not rebound following ART discontinuation, viral loads remained undetectable following CD8 depletion (Fig. 6b, black lines).
We developed a computational model that included cellular activation, innate plasma cytokines, and adaptive immune responses to define the correlates of delayed viral rebound in this study. A model that included NK cell activation correlated best with delayed viral rebound (Extended Data Fig. 9a; P = 5.31 × 10−4, R = 0.50, Spearman rank-correlation test). A similar model involving NK cell and monocyte activation correlated with reduced viral loads following ART discontinuation (Extended Data Fig. 9b; P = 3.52 × 10−6, R = 0.64, Spearman rank-correlation test). These data suggest that activated NK cells and monocytes may have a key role in PGT121-mediated elimination of infected CD4+ T cells, consistent with the observed NK cell activation following GS-9620 activation (Fig. 2b).
Finally, we assessed the ability of PGT121, with or without GS-9620, to kill HIV-1-infected CD4+ T cells in vitro. GS-9620 led to activation of NK cells and T cells in vitro (Extended Data Fig. 10a), consistent with the in vivo data (Fig. 2, Extended Data Fig. 2). Moreover, the combination of PGT121 and GS-9620 led to optimal killing of HIV-1-infected CD4+ T cells in vitro (Extended Data Fig. 10b), consistent with previous studies19. Together, these data suggest a mechanism by which the TLR7 agonist GS-9620 stimulated innate immunity and activated multiple immune cell subsets in vivo including infected CD4+ T cells (Fig. 2, Extended Data Fig. 2), rendering them more susceptible to PGT121-mediated recognition, as well as effector cells such as NK cells and monocytes (Fig. 2b, Extended Data Fig. 9), which may have facilitated PGT121-mediated elimination of these infected CD4+ T cells.
Our data demonstrate that PGT121 together with the TLR7 agonist vesatolimod (GS-9620) delayed viral rebound following discontinuation of ART in acutely treated, SHIV-SF162P3-infected rhesus monkeys. Moreover, five of eleven monkeys that were treated with PGT121 and GS-9620 showed no viral rebound for more than 6 months after cessation of ART and also did not reveal virus by highly sensitive adoptive transfer and CD8 depletion studies. This proof-of-concept study suggests that bNAb administration combined with innate immune stimulation may represent a potential strategy to target the viral reservoir.
Our findings extend previous observations that HIV-1-specific bNAbs have direct antiviral activity in SHIV-infected rhesus monkeys10,11,12 and in HIV-1-infected humans13,14,15,16. However, these previous studies did not assess the potential of bNAbs to target the viral reservoir, which would require bNAbs to be administered during ART suppression and no longer be present at therapeutic levels following discontinuation of ART. We previously defined a PGT121 level of 1 μg ml−1 as the threshold for viral rebound in SHIV-SF162P3-infected rhesus monkeys for PGT121-mediated virologic control10. In the present study, PGT121 levels were undetectable (below 0.5 μg ml−1) in peripheral blood, lymph nodes, and colorectal tissue for 8–10 weeks before discontinuation of ART (Extended Data Fig. 4). These findings suggest that the delayed viral rebound in monkeys treated with PGT121 and GS-9620 may reflect reservoir targeting rather than just direct antiviral activity.
The five monkeys in the PGT121- and GS-9620-treated group that showed sustained remission for more than 6 months following discontinuation of ART also did not show evidence of virus in adoptive transfer and CD8 depletion studies (Fig. 6), which are sensitive tests for residual replication-competent virus. Viral rebound can occur in humans following extended periods of ART-free remission25,26. We therefore cannot exclude the possibility that exceedingly low levels of replication-competent virus may still exist in these monkeys. Nevertheless, there was a clear difference between the monkeys that rebounded and those that did not rebound in the adoptive transfer and CD8 depletion studies (Fig. 6).
We hypothesize that the mechanism of the observed effects of PGT121 and GS-9620 involves activation of multiple cell types by GS-9620 followed by efficient binding and elimination of virally infected CD4+ T cells by PGT121. GS-9620 activated CD4+ T cells and NK cells, both in vivo (Fig. 2, Extended Data Fig. 2) and in vitro (Extended Data Fig. 10), and activated NK cells and monocytes correlated with delayed viral rebound following ART discontinuation (Extended Data Fig. 9). These data suggest that GS-9620 may have activated latently infected CD4+ T cells, possibly rendering them more susceptible to PGT121 binding, and effector cells such as NK cells and monocytes, which may have facilitated antibody-mediated elimination of the infected CD4+ T cells. This proposed mechanism predicts that a longer duration of combined PGT121 and GS-9620 therapy might improve therapeutic efficacy. In monkeys treated with PGT121 and GS-9620, we did not detect evidence of a ‘vaccinal effect’ of increased autologous antigen-specific CD8+ T cell responses following bNAb administration (Extended Data Figs. 5, 6), in contrast to a previous study that involved bNAb administration during acute SHIV infection27.
In summary, our data show that bNAb administration combined with innate immune stimulation can delay viral rebound following discontinuation of ART. This study used rhesus monkeys in which ART was initiated on day 7 of acute infection and then received ART continuously for 2.5 years. Moreover, the maximum therapeutic effect was observed in monkeys with the lowest pre-ART viral loads (Fig. 5c). It will therefore probably be far more difficult to achieve similar results in monkeys in which ART is initiated during chronic infection, and thus implications for typical HIV-1-infected humans remain unclear. Nevertheless, our data provide an initial proof-of-concept in primates showing the potential of innate immune activation with immune-based targeting of the viral reservoir.
Monkeys and study design
Forty-four outbred Indian-origin, young adult male and female rhesus monkeys (M. mulatta) were genotyped, and animals that expressed protective MHC class I alleles and susceptible and resistant TRIM5α alleles were distributed among the groups. Each group had 0 Mamu-A*01, 1 or 2 Mamu-B*17, and 0 or 1 Mamu-B*08 animals. Animals were otherwise randomly allocated to groups. All monkeys were housed at Bioqual, Rockville, MD. Animals were infected with a single 500 TCID50 dose of our rhesus PBMC-derived SHIV-SF162P3 challenge stock21 by the intrarectal route. ART was initiated on day 7. Monkeys were bled up to twice per week for viral load determinations. Monkeys received the following interventions starting at week 96 (n = 11/group): (1) sham controls, (2) vesatolimod (GS-9620) alone, (3) PGT121 alone, or (4) both (Extended Data Fig. 1). In groups 2 and 4, animals received 10 oral administrations of 0.15 mg/kg GS-9620 (Gilead Sciences, Foster City, CA) every 2 weeks from weeks 96 to 114. In Groups 3 and 4, animals received 5 intravenous infusions of 10 mg/kg PGT121 (Catalent Biopharma, Madison, WI) every 2 weeks from weeks 106 to 114. Cytokine levels were determined by Luminex assays, and T cell and NK cell activation was assessed by multiparameter flow cytometry. Immunologic and virologic assays were performed blinded. ART was discontinued at week 130. All animal studies complied with all relevant ethical regulations and were approved by the Bioqual Institutional Animal Care and Use Committee (IACUC).
The formulated antiretroviral therapy (ART) cocktail contained 5.1 mg/ml tenofovir disoproxil fumarate (TDF), 40 mg/ml emtricitabine (FTC), and 2.5 mg/ml dolutegravir (DTG) in water containing 15% (v/v) kleptose adjusted to pH 4.2. This ART cocktail was administered once daily at 1 ml/kg body weight via the subcutaneous route.
Adoptive transfer studies
For adoptive transfer studies, 30 million LNMCs and PBMCs from PGT121+GS-9620- or PGT121-treated animals, collected on day 140 following ART discontinuation, were infused intravenously into healthy, SHIV-uninfected rhesus monkeys. Viral loads were assessed in recipient animals weekly following adoptive transfer.
CD8 depletion studies
For CD8 depletion studies, animals received a single intravenous infusion of 50 mg/kg of the anti-CD8α CDR-grafted rhesus IgG1 antibody MT807R1 (Keith Reimann, MassBiologics, Mattapan, MA). CD8 T cell counts and viral loads were assessed weekly following anti-CD8 infusion.
ImmunoClear ELISA plates (Thermo Scientific) were coated with 100 ng/well clade C (C97ZA.012) gp140 Env protein capture reagent overnight at 4 °C. Plates were then washed with PBS–0.05% Tween 20 and blocked for 2 h at room temperature with Blocker Casein in PBS (Pierce). Standard curve calibrators and diluted serum samples were incubated on the plates for 1 h before further washing and subsequent incubation with a mouse PGT121 anti-idiotype (1 μg/ml) monoclonal antibody. Plates were washed again and then incubated with 1:1,000 dilution of rabbit anti-mouse IgG-horseradish peroxidase (Thermo Scientific). Finally, plates were washed and developed for 5 min using SureBlue (KPL Laboratories) followed by addition of TMB Stop solution (KPL Laboratories). Plates were read at 450 nm on a VersaMax microplate reader (Molecular Devices) using Softmax Pro version 6.5.1 software. The SoftMax Pro software calculated 4-Parameter Logistic (4-PL) curve fits for the standard calibrators and the test sample concentrations were determined by interpolation into the calibration curves.
PGT121 anti-drug antibody (ADA) assays
ADA assays were performed using the MesoScale Discovery (MSD) electrochemiluminescence (ECL) platform. In brief, serum samples were diluted in assay buffer and incubated with a master mix containing sulfo-tagged and biotinylated-tagged PGT121 (at equimolar concentrations of 0.5 µg/ml) overnight at 4 °C on an orbital plate shaker. The mixture was then incubated on a streptavidin-functionalized plate and washed, and tripropylamine (TPA) was added and read using an MSD SQ-120 ECL imager. Luminescence was proportional to the amount of ADA. The assay positivity cut-point was defined as the 95th percentile of ECL signal response observed using naive NHP serum samples.
Cellular immune assays
SIV-specific cellular immune responses were assessed using IFN-γ ELISPOT assays and multiparameter intracellular cytokine staining (ICS) assays essentially as described20. ICS assays were performed with 106 PBMCs or LNMCs that were incubated for 6 h at 37 °C with medium, 10 pg/ml phorbol myristate acetate (PMA) and 1 µg/ml ionomycin (Sigma-Aldrich), or 1 µg/ml HIV-1 Env, SIV Gag, or SIV Pol peptide pools. Cultures contained monensin (GolgiStop; BD Biosciences), brefeldin A (GolgiPlug; BD Biosciences), and 1 µg/ml of a mAb against human CD49d (clone 9F10). Cells were then stained with predetermined titres of mAbs against CD3 (clone SP34.2;, Alexa 700), CD4 (clone L200; BV786), CD8 (clone SK1; APC H7), CD28 (clone L293, PerCP-Cy5.5), CD95 (clone DX2, BV711), CD20 (clone 2H10, BV570), CCR5 (clone 3A9, PE), BCL6 (clone K112-91, PE-CF594), CXCR5 (clone MU5UBEE, PE-cy7), and PD-1 (clone EH12.2H7, Pacific Blue); and stained intracellularly with IFN-γ (clone B27; BUV395), IL-2 (clone MQ1-17H12; BUV737), TNF-α (clone Mab11; BV650), CD69 (clone FN50, BV510), and Ki67 (clone B56, FITC). IFN-γ backgrounds were <0.05% in PBMCs.
Viral RNA assays
Viral RNA was isolated from cell-free plasma using a viral RNA extraction kit (Qiagen) and was quantified essentially as described24.
Viral DNA assays
Levels of proviral DNA were quantified as previously described24. Total cellular DNA was isolated from 5 × 106 cells using a QIAamp DNA Blood Mini kit (Qiagen). The absolute quantification of viral DNA in each sample was determined by qPCR using primers specific to a conserved region SIVmac239. All samples were directly compared to a linear virus standard and the simultaneous amplification of a fragment of human GAPDH gene. PCR assays were performed with 100–200 ng sample DNA.
The frequencies of cell populations at week 106 (corresponding to the sixth GS-9620 administration) before and after the intervention were used for the computational model. Z-score standardization was performed to have all features mean centred and unit variance scaled. The minimal correlates that best predicted viral rebound (or total viral loads) were identified by using a two-step model: least absolute shrinkage and selection operator (LASSO) regularization followed by partial least squares regression analysis (PLSR). The LASSO method was used to remove irrelevant features in order to improve the robustness of high-dimensional modelling. Next, PLSR was used to model the covariance relationship between the feature variables (X) and the outcome variables (Y), in the way that PLSR decomposes both X and Y into a hyperplane and maximizes covariance between the hyperplanes. To estimate the minimal correlates that best explain the outcome without overfitting, 5,000 repeated fivefold nested cross-validation was designed. In each repetition, the data set was randomly divided into five folds, where 80% of the data set was used for building the model and the remaining holdout set was used to test the model prediction, where the goodness-of-fitness of the model was measured by mean squared error (MSE) between prediction and the outcome. This approach resulted in the generation of a model with the minimal set of the features that generates the best outcome prediction in cross-validation test. In addition, variable importance in projection (VIP), a weighted sum of squares of the PLSR weights that summarized the importance of the features in a PLSR model with multiple components, was computed. To estimate the statistical significance of the optimized model with the defined correlates, we employed two types of permutation tests (shuffling the outcome label and selecting the randomized correlates) to test the likelihood of obtaining a model prediction accuracy by chance. Each permutation test performed 1,000 times to generate an empirical null distribution, and an exact P value of the model was computed.
PGT121-mediated CD4+ T cell killing assay
Human PBMCs were isolated from fresh blood from healthy donors using SepMate PBMC Isolation tubes (StemCell Technologies) and histopaque (Sigma Aldrich) according to the manufacturer’s protocol. Using an aliquot of PBMCs, CD4+ T cells were isolated for HIV-1 infection using the EasySep Human CD4+ T Cell Isolation Kit (StemCell Technologies). Remaining PBMCs were treated with 1,000 nM GS-9620 or DMSO and cultured at 3.0 × 106 cells/ml for 5 days in RPMI 1640 (Sigma Aldrich) medium supplemented with l-glutamine, HEPES, and IL-15 (1 ng/ml) at 37 °C. For HIV-1 infection, freshly isolated CD4+ T cells were spinfected for 45 min at 1,500 RPM with NL4-3 virus (or medium for mock infection) and then treated with CD3/CD28 Human T Activator Dynabeads (ThermoFisher Scientific). Infected CD4+ T cells were subsequently cultured for 5 days in R10 and IL-2 (30 units/ml) at 37 °C. Cells were counted daily and supplemented with fresh medium with IL-2 as needed. On day 5, both sets of cells were counted, washed twice and resuspended in fresh medium. HIV-infected CD4+ T cells were incubated with 10 μg/ml of PGT121 for 30 min. Co-cultures of infected CD4+ T cells and autologous donor PBMCs were then set up at a ratio of 1:10 (5.0 × 104 CD4+ T cells; 5.0 × 105 PBMCs) and incubated overnight in R10 and IL-2 at 37 °C. Following the overnight incubation, cells were stained with Fixable Aqua viability dye (Invitrogen) followed by surface staining with CD4-APC (Invitrogen), CD56-PE-Cy7 (BD), CD3-AF700 (BD), and CD69-BV605 (BD). Cells were fixed and permeabilized with Cytofix/Cytoperm (BD Biosciences) and stained intracellularly with anti-KC57-PE-labelled p24 antibody (Beckman Coulter) in 1× Perm/Wash Buffer (BD Biosciences, 554723). Gating on viable CD3+CD4+p24+ T cells was used to evaluate viral killing with each treatment condition normalized to an uninfected control. Per cent killing was normalized to the untreated antibody condition.
Analysis of virologic and immunologic data was performed using GraphPad Prism v6.03 (GraphPad Software). Comparisons of groups was performed using two-sided Mann–Whitney tests without Bonferroni adjustments. Correlations were assessed by two-sided Spearman rank-correlation tests. Analysis of viral rebound was performed using Kruskall–Wallis tests to compare all groups. For group pairwise comparisons, area under the curve (AUC) for total viral RNA following ART discontinuation was compared with chi-square tests, and viral rebound was compared with a censored Poisson regression model. In vitro killing data were analysed with paired Student’s t-tests.
No statistical methods were used to predetermine sample size. Except as stated, the experiments were not randomized, and investigators were not blinded to allocation during experiments and outcome assessment.
Further information on experimental design is available in the Nature Research Reporting Summary linked to this paper.
All data generated and analysed in this study are available from the corresponding author upon reasonable request. Source data for figures from individual animals are available online.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank K. Reimann and A. Hill for advice, assistance, and reagents. We acknowledge support from the Bill & Melinda Gates Foundation (OPP1107669), the American Foundation for AIDS Research (109219-58-RGRL), the National Institutes of Health (AI096040, AI124377, AI126603, AI129797, AI128751, OD024917), and the Ragon Institute of MGH, MIT, and Harvard.