First, we developed the “resting-like” or RELI-model to evaluate HIV-1 reactivation by LRA (shock) simultaneously with the elimination of reactivated cells by HIV-1-specific cytotoxic CD8+ T-cell lines (CTL) (kill), as schematized in Figure 1A and detailed in the Materials and Methods section. Briefly, we infected U937-HLA-B^*27:05 cells with the HIV-1_NL43GFP reporter virus as previously reported (31, 32). After 72 h of infection, we sorted the GFP negative cells and cultured them for 4 days in the presence of the protease inhibitor ritonavir (RTV). At this point, we obtained a RELI population enriched in uninfected and non-productively HIV-1-infected cells. RELI cells were then washed, treated with the HIV-1 integrase inhibitor raltegravir (RAL) to avoid multiple rounds of infection, and exposed to LRA shock for 48 h. We also treated RELI cells with azidothymidine (AZT) to avoid the formation of unintegrated episomal forms that may contribute to the expression of viral proteins (data not shown) (33). LRA-treated cells were then extensively washed to uncouple the LRA effect from CD8+ T-cell function against reactivated cells (24), and co-cultured with HLA-class I–matched HIV-1-specific CTLs for 20 h (kill) (Figure 1A).

We measured HIV-1 reactivation in RELI cells in response to LRA for 48 h including the HDACi panobinostat (PNBN), romidepsin (RMD), trichostatin A (Tricho), and suberoylanilide hydroxamic acid (SAHA). We also included combinations of HDACi and bryostatin A (Bryo) to evaluate potential additive effects between HDACi and protein kinase C agonists (PKCag) as previously reported (34). Drug concentrations were selected based on previous clinical or in vitro studies (7, 35–37) and in the absence of direct cellular toxicity (Table S1). The use of HIV-1_NL43GFP enabled us to monitor viral reactivation by flow cytometry using two markers the expression of p24 protein (Figure 1B) and the expression of GFP encoded in frame with the HIV-1 Nef protein, as previously reported (38, 39) (Figure 1C). We observed a direct correlation between HIV-1 fold induction measured according to p24 or GFP expression (p < 0.0001; ρ = 0.94, Figure 1D). In this context, the expression of p24 was consistently higher than that of GFP (p24 vs. GFP p < 0.005, Wilcoxon matched-pairs signed-rank test) (Figure 1E), thus indicating its greater sensitivity as a marker for viral antigen expression upon reactivation and supporting its application in subsequent experiments. However, the use of GFP expression as a marker of HIV-1 induction is particularly relevant given the correlation between p24 and GFP expression for the screening of large compound libraries and, thus, the increased cost-effectiveness of the method.

As shown in Figure 1F, we compared HIV-1 reactivation after LRA treatment with Tricho, PNBN, and SAHA alone or combined with Bryo with untreated or DMSO controls. The levels of HIV-1 reactivation ranged between a 2- and 4-fold viral induction, with a significant increase for PNBN and SAHA alone (p < 0.05) or combined with Bryo (p < 0.005). The additive effects of combining HDACi/Bryo were modest or absent.

Next, we assessed whether LRA treatment enabled HIV-1 antigen presentation in the context of HLA-class I molecules for their recognition and killing by HIV-1 cytotoxic CD8+ T-cell lines (CTL). For this purpose, we used two CTLs, CTL1, and CTL2, which recognized HLA-B^*27:05 restricted epitopes (Gag KK10 and Pol KY9, respectively) as previously described (31). Briefly, we exposed RELI cells to PNBN and SAHA alone or combined with Bryo and co-cultured them with CTLs restricted by HLA-B^*27:05. After 20 h of co-culture, we assessed killing by measuring the frequency of p24-expressing cells in the absence or presence of CTL1 (Figure 2A and Figure S1). As shown in Figure 2B, we observed significantly increased killing by CTL1 and CTL2 after LRA treatment (p < 0.0005, one-way ANOVA). In fact, CTL1 and CTL2 were highly functional and had similar 50% effective concentrations (EC₅₀, −4.8, and −5.5, respectively), despite differences in epitope specificity and a higher response for CTL1 at low HIV-1 antigen concentrations (Figure S2).

In addition, we added an anti-HLA-A/B/C (W6/32) antibody or an HLA-mismatched (MM) CTL3 to demonstrate that, upon viral reactivation, CTL killing was mediated through contacts with class I HLA HIV-1 antigen complexes. The elimination of reactivated cells was significantly abolished both by means of the W6/32 antibody and by impeding the recognition of the HLA-class I peptide complexes with an HLA-MM co-culture (p < 0.05, one-way ANOVA) (Figure 2C). Thus, the compiled analysis of all LRA tested showed a significant increase in the magnitude of killing of cells latently infected by HIV-1 after LRA treatment in both CTL (p < 0.05, Figure 2D). However, we did not observe any additional benefit in the frequency of reactivated cell killing through the combination of HDACi with Bryo (Figure 2D).

These data confirm that inducible reactivation of HIV-1 by LRA enables antigen presentation in the context of HLA-class I molecules to increase killing of latently infected cells by CTL.

It is essential to investigate the kinetics of recognition and killing of LRA-reactivated cells and to establish a window of opportunity between shock and kill by CD8+ T cells for therapeutic optimization. By using the RELI model, we monitored the kinetics of recognition and elimination of HIV-1 reactivated cells by CTLs at 3, 6, and 20 h post co-culture. We measured changes in the frequency of p24-positive cells from time 0 by flow cytometry at 3, 6, and 20 h after co-culture of untreated and LRA-treated RELI cells with CTLs (Figure 3A). We observed a reduction in the frequency of p24-expressing cells over time in culture with CTLs (Figure 3B). HIV-1 reactivated cells were eliminated rapidly after 3 h of co-culture in SAHA, PNBN, SAHA/Bryo, and PNBN/Bryo conditions compared with untreated cells for both CTLs (untreated vs. HDAC ± Bryo; p < 0.0001) (Figure 3C). These differences were sustained at 6 h after co-culture (untreated vs. HDAC ± Bryo; p < 0.0005) and 20 h after co-culture (untreated vs. HDAC ± Bryo; p < 0.005) (Figures 3D,E). These data indicate a general increase in the speed of recognition and killing of reactivated cells by CTLs starting at 3 h in the presence of PNBN, PNBN/Bryo, SAHA, and SAHA/Bryo, although our results do not reveal a hierarchy for LRAs. The data suggest a prolonged effect over the first 6 h by PNBN/Bryo, SAHA, and SAHA/Bryo compared with untreated cells (Figures 3D,E) (untreated vs. PNBN/Bryo; p < 0.05, untreated vs. SAHA, p < 0.005; and untreated vs. SAHA/Bryo; p < 0.005). By contrast, the use of Bryo alone did not significantly increase the speed or magnitude of killing compared with untreated condition at 3, 6, or 20 h of co-culture (Figures 3C–E). Although we observed a marginal augment in the killing at early time points, suggesting an indirect effect of Bryo in CTL activation despite extensive washout of the drug as previously reported (24).

Interestingly, as shown in Figure 3F, we found a direct correlation between the magnitude of CTL killing and the levels of HIV-1 inducible reactivation by LRA at early time points (3 h p < 0.0005, ρ = 0.74; 6 h p < 0.005, ρ = 0.68), thus indicating that the potency of the LRA for HIV-1 antigen expression dictates the speed of CTL recognition. Based on these data, we propose a threshold of >2-fold HIV-1 induction upon LRA treatment to allow antigen presentation and engage rapid recognition and killing of HIV-1 reactivated cells by CTL.

Besides, we investigated whether HIV-1 killing by CTL can be monitored by changes in cytokine expression and degranulation upon antigen recognition using CTL as biosensors (40), we measured CD107a/MIP1β and IFNγ expression (Figure 3G, dot plots). Despite the increase of HIV-1 expression by LRAs, the use of SAHA or PNBN did not alter the kinetics of CD107a/MIP1β or IFNγ expression in the tested CTLs, as compared with untreated conditions. By contrast, Bryo indirect effects increased both CD107a/MIP1β and IFNγ secretion by CTLs at early time points (Figure 3H). We also found different profiles for CD107a/MIP1β and IFNγ expression over time, with increasing frequencies for CD107a/MIP1β and decreasing frequencies for IFNγ secretion. Thus, the kinetics of cytokine expression by CTLs did not mirror changes associated with HIV-1 inducible reactivation or killing.

Overall, these data indicate a direct relationship between the potency of inducible reactivation of HIV-1 by LRA and the speed of recognition and killing of reactivated cells by CD8+ T cells. Therefore, the potency of LRA to cross the threshold of HIV-1 antigen presentation in reactivated cells is a crucial event to shortening the window of opportunity for antigen recognition, thus maximizing CD8+ T-cell sensing for elimination of HIV-1 reactivated cells.

Although the use of CTL supports rapid recognition of HIV-1 reactivated cells based on LRA potency, these CTL are highly functional and sensitive to HIV-1 antigen presentation and may not fully recapitulate the diversity of HIV-1–specific CD8+ T cells in infected individuals.

To delineate the kinetics of recognition and killing of LRA-reactivated cells by CD8+ T cells, we initially co-cultured SAHA-treated cells with ex vivo CD8+ T cells obtained from three of the eight HIV-1-infected individuals tested (participants characteristics in Table 1). As shown in Figure 4A, we observed variation in the kinetics of recognition and killing of reactivated cells by ex vivo CD8+ T cells. While we observed killing in response to SAHA at 3 h for PT1, we did not observe elimination of reactivated cells for PT2; in addition, elimination of reactivated cells was only observed after 20 h of co-culture for PT3. These divergences suggest functional differences between CD8+ T cells from HIV-1-infected individual associated a particular profile of recognition and killing of LRA-reactivated cells (41).

Next, we extended our initial analyses to an additional group of HIV-1-infected individuals (Table 1). As in the previous experiments, we found high inter-individual variability in the killing capacity of HIV-1-reactivated cells by CD8+ T cells ranging from 2% (PT2 and PT7) to 70% (PT4 and PT5) (Figure 4B). The killing observed was specific for CD8+ T-cell recognition in the context of HLA-class I presentation, as supported by the lack of killing observed by mismatch CD8+ T cells (MM). Despite this biological divergence, we observed a significant increase in the killing of SAHA-treated cells by CD8+ T cells (p < 0.05, Figure 4C). Moreover, we were able to enhance the magnitude of CD8+ T-cell killing ex vivo, either by increasing the concentration of SAHA 2-fold or the effector-to-target ratio 3 times (Figure 4D). These results demonstrate the possibility of modulating and improving the efficacy of CD8+ T-cell killing in response to HIV-1 inducible expression by higher LRA dosage or higher frequency of effector cells. We also carried out a detailed study of cell-to-cell interactions in our co-culture system. We observed an increase in the number of CD8+ T-cell contacts with target cells upon reactivation of SAHA (p < 0.005). These contacts were specifically blocked by the addition of an anti-HLA-A/B/C antibody (p < 0.05) (Figure 4E), thus demonstrating the dependency of cell-to-cell contact on the interactions between TCR and HLA-class I/peptide complexes after LRA treatment. In addition, CD8+ T cells that established contact were highly functional, as measured by the levels of CD107a/MIP1β and IFNγ when compared with CD8+ T cells lacking contacts with target cells (p < 0.0001) (Figure 4F).

These findings support an overall increase in the killing of HIV-1-infected cells by CD8+ T cells upon LRA induction of viral reactivation despite inter-individual divergences. Moreover, our data point to the opportunity to optimize the immune clearance of HIV-1 reactivated target cells by CD8+ T cells by increasing the LRA dosage and the frequency of effector CD8+ T cells.

Inter-individual differences in functional ability to recognize and eliminate HIV-1-infected cells ex vivo by CD8+ T cells could be the result of differences in time to initiation of ART (42), in the persistence of antigen exposure and in the chronic pro-inflammatory environment despite ART (30). The pro-inflammatory environment and the long-term exposure to antigen lead to dysfunctional HIV-1–specific CD8+ T-cell responses and eventually to immune exhaustion.

To delineate the determinants of CD8+ T-cell dysfunction in response to LRA-reactivated cells, we cultured CTLs with HIV-1 cognate peptide for 9 weeks with weekly rounds of peptide re-stimulation to simulate in vitro long-term antigen exposure (Figure 5A). At weeks 0, 5, 7, and 9 we performed an immune phenotype of inhibitory receptors (PD-1, LAG3, and TIM-3) and the immune-metabolic marker CD39. All these markers have been previously associated with the exhaustion of HIV-1–specific CD8+ T-cell responses (28, 43). As shown in Figure 5B, we found a significant reduction in the killing of LRA-reactivated cells by CTL after 9 weeks of continuous antigen exposure (p < 0.005). Despite the overall reduction in the killing by CTL after 9 weeks, we did observe an increase in killing between untreated and SAHA conditions (35–58%) (Figure 5B).

We also monitored phenotypic changes in CTL and found a contraction in the frequency of subpopulations without expression or expressing a single inhibitory receptor concomitant with the expansion of cells co-expressing three or four inhibitory receptors over time (Figure 5C, pie charts). This phenotype of expansion/contraction was coincident between CTL1 and CTL2, which vary in HIV-1 antigen specificity. However, the specific combinations of inhibitory receptors diverge between CTL lines and time, thus highlighting the diversity of T-cell inhibitory pathways. Also, CD39+ subpopulations appeared after 5 weeks in cells already expressing two or more inhibitory receptors (Figure 5C, arcs). These data suggest the presence of terminally exhausted CD8+ T cells identified by CD39 expression, as previously described (43).

Based on these data and the biological diversity found in the recognition and killing of LRA-reactivated cells by ex vivo CD8+ T cells, we measured HIV-1–specific CD8+ T-cell function in response to a pool of Gag overlapping peptides. We monitored expression of CD107a, IFNγ, and MIP1β cytokines and of PD-1, LAG3, TIM-3, and CD39 by multiparametric flow cytometry. As shown in Figure 6A, we observed variation in the profile of cytokine production among individuals in response to HIV-1. Meanwhile, the functional response to SEB control stimuli was very homogeneous, as expected. Furthermore, we did not find a correlation between the frequency of polyfunctional cells among the HIV-1–specific CD8+ T cells and the killing in response to LRA-reactivated cells, as it has previously been associated with spontaneous control of HIV-1 replication (44, 45). Importantly, individuals harboring HIV-1–specific CD8+ T cells lacking expression of inhibitory receptors retained the highest suppressive capacity against HIV-1 LRA-reactivated cells (Figure 6B). Thus, PT5 responded to LRA treatment with a 68% in vitro suppressive capacity and 63% of HIV-specific CD8+ T cells lacking inhibitory receptors. Participants PT3, PT1, and PT7 with 36, 35, and 24% of HIV-1–specific CD8+ T cells lacking inhibitory receptors had suppressive capacities achieving 19, 10, and 2% of killing of SAHA reactivated cells. Although limited by sample size, our data delineate a correlation between the percentages of HIV-1–specific CD8+ T cells lacking inhibitory receptors and their immune responsiveness to killing LRA treatment (p < 0.005, Figure 6C).

According to these data, changes in the profile of inhibitory receptors in HIV-1-specific CTL and primary CD8+ T cells modulate their antiviral potency. Therefore, the levels of inhibitory receptor co-expression in CD8+ T cells can limit the magnitude of the killing of HIV-1-reactivated cells.

The virus NL43_GFP was generated by co-transfection of p83–2 and p83–10_eGFP plasmids in MT4 cells (NIH AIDS Reagent program, USA) based on electroporation, as previously described (38, 39). The cell-free supernatant was concentrated by spinoculation at 24,000 rpm at 4°C for 90 min, and stored at −80°C. The 50% tissue culture infective dose (TCID₅₀) was determined on TZM-bl cells (NIH AIDS Reagent program, USA) as described in Kloverpris et al. (31). The HIV-1-permissive U937 cell line transfected with the HLA-B^*27:05 gene (kindly provided by Paul Bowness, UK) (53) was cultured in RPMI media supplemented with 10% fetal calf serum (FCS) (HyClone), 2 mM L-glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin, and 0.5 mg/ml geneticin (Invitrogen, Life Technologies). The HIV-1 cytotoxic CD8+ T-cell lines (CTL) used were restricted by HLA-B^*27:05. The CTL1 was specific for the KK10 epitope in Gag and CTL2 for the KY9 epitope in Pol. In addition, CTL3 was restricted by HLA-B^*57:01 and specific for the KF11 epitope in Gag (31, 53).

The U937 HLA-B^*27:05 permissive cell line was infected with NL43_GFP by magnetofection at a dose equivalent to a nominal multiplicity of infection (MOI) of 0.1, as previously reported (31, 32). After magnetofection, cells were washed twice with 1X PBS and cultured in R10 + geneticin (0.5 mg/ml). At day three post-infection, the GFP-negative population, which included the non-productively infected cells enriched in HIV-1 integrated provirus, was sorted with a FACS Aria II Cell Sorter (BD Biosciences) and cultured in the presence of 1 μM of RTV (Sigma, Spain). After 4 days, we obtained a culture enriched in “resting-like” cells (RELI). The mean levels of cell-associated total HIV-1 DNA in RELI-lysates was 5.1–log ± 0.19 log₁₀ copies/10⁶ RELI as measured by droplet digital PCR previously described in Martínez-Bonet et al. (54). RELI cells were treated with LRA in the presence of RAL at 100 nM (Sigma, Spain) to monitor HIV-1 reactivation or “shock.” The LRA panel included the HDACi: panobinostat (PNBN) at 30 nM (SelleckChem, USA), trichostatin A (Tricho) at 250 nM (Sigma, Spain), SAHA (suberoyl aniline hydroxamic acid) at 1 μM (SelleckChem, Spain), and romidepsin (RMD) at 40 nM (SelleckChem, Spain), and the PKC agonist bryostatin A (Bryo) at 10 nM (Sigma, Spain). RELI cells were treated for 48 h with LRA except for RMD-treated cultures, where the drug was washed-off after 4 h to avoid cellular toxicity.

To monitor the “shock” or HIV-1-fold induction at 48 h after treatment with LRA, the RELI cells were harvested and stained with a Live/Dead probe (APC-Cy7, Invitrogen). After washing, cells were fixed and permeabilized with the Fix&Perm kit (A and B solutions, Thermo Fisher Scientific) for intracellular HIV-1 p24 (PE, clone KC57-RD1, Beckmann Coulter). Cells were acquired on an LSR II flow cytometer (Beckmann Coulter), and data were analyzed using FlowJoV (Tree Star Inc.). The HIV-1-fold induction was calculated based on the following equation: % of p24+ or GFP+ in CD4+ under LRA conditions/% of p24+ or GFP+ under untreated conditions. To proceed with the co-cultures to monitor the “killing” of reactivated cells by CD8+ T cells the RELI cells were extensively washed with 1X PBS and left in R10 media plus RAL. Thus, we cultured LRA-reactivated RELI cells in the absence or presence of HIV-1-specific CTLs or B^*27-HLA-matched ex vivo CD8+ T cells obtained from HIV-1-infected individuals. Cells were co-cultured to a 1:1 effector: target ratio for 20 h. After 20 h of co-culture, we simultaneously measured by flow cytometry the “shock” by intracellular p24, and the “kill” by the reduction of p24-positive RELI cells in the absence or presence of effector cells. We also assessed LRA-related toxicity both in uninfected U937 cells and in the RELI model by measuring cellular viability using flow cytometry and staining with a Live/Dead probe (APC-Cy7, Invitrogen) at the end of the co-culture with RELI and CTL.

We performed CD8+ T-cell killing experiments of LRA-reactivated cells with CTL or ex vivo CD8+ T cells at 20 h after culture. We also monitored the kinetics of CD8+ T-cell killing of LRA-reactivated cells with CTL or ex vivo CD8+ T cells after 3, 6, and 20 h after co-culture. We monitored CTL and CD8+ T-cell activation by intracellular expression of CD107a, MIP-1β, and IFNγ. For analysis of HLA-class I restriction, the W6/32 antibody (BioLegend, Spain) or the isotype control LEAF purified IgG2aÎ (BioLegend) was added to the co-culture at 10 μg/ml. For analysis of CD8+ T-cells in contact with RELI cells, we analyzed the percentage of doublets gated cells that expressed the CD8+ surface marker in the target cells gate. For analysis of early CD107a secretion, the antibody CD107a (PerCP-Cy5.5, BD Biosciences) was added to the co-culture. For the immune-phenotype, samples were collected and washed twice with 1X PBS and incubated for 3 h in the presence of the protein transport inhibitor Golgiplug (Brefeldin A solution, BD Biosciences, Spain) and Golgistop (Monensin solution, BD Biosciences, Spain). Cells were then harvested, stained with a Live/Dead probe (APC-Cy7, Invitrogen). Next, cells were washed and stained to identify the surface marker CD8 (Pacific blue, clone RPA-T8, BD Biosciences). After washing, cells were fixed and permeabilized with the Fix&Perm kit (A and B solutions, Thermo Fisher Scientific) for intracellular cytokine staining of MIP1β (FITC, clone 24,006, R&D Systems), IFNγ (PE-Cy7, clone 4S.B3, BD Biosciences), and HIV-1 p24 (PE, clone KC57-RD1, Beckmann Coulter). Cells were acquired on an LSR II flow cytometer (Beckmann Coulter). Data were analyzed with FlowJoV (Tree Star Inc.). The killing of reactivated HIV-1 cells by CTL/CD8+ was calculated based on the following equation: 100—[(% p24+ in CD4+ co-cultured with CD8+/% p24+ in CD4+ in the absence of CD8+) × 100].

To characterize potential markers associated with CTL dysfunction in response to LRA-reactivated cells, we exposed CTL1 and CTL2 to cognate HIV-1 antigen over nine weeks. To do so, we stimulated CTL with a 1:1 mixture of irradiated autologous B-cell lines (BCLs) pulsed with cognate HIV peptide (10 μg/ml) and irradiated PBMCs from three healthy HIV-seronegative donors weekly. We used flow cytometry to evaluate variations in the expression of the inhibitory receptors PD-1, TIM-3, LAG-3, and the immune-metabolic marker CD39. Briefly, cells were taken from the culture at weeks 0, 5, 7, and 9 and stained with a Live/Dead probe (APC-Cy7, Invitrogen) and surface markers CD3 (Alexa700, BD Biosciences), CD4 (APC-Cy7, BD Biosciences), CD8 (V500, BD Biosciences), PD-1 (BV421, BD Biosciences), TIM-3 (Alexa 647, BD Biosciences), LAG-3 (PE, BD Biosciences), and CD39 (FITC, BD Biosciences) and incubated at room temperature for 25 min. Samples were washed twice with 1X PBS, fixed in 1% formaldehyde, and acquired on an LSR Fortessa. Data were analyzed with FlowJoV (Tree Star Inc.). Patterns of co-expression of PD-1, LAG-3, TIM-3, and CD39 were analyzed using Pestle and SPICE v5 software (55).

The PBMCs from HIV-1-infected individuals were stimulated with HIV-1 Gag peptide pool (2 μg/peptide/ml, EzBiolab), or the positive control SEB (enterotoxin B from Staphylococcus aureus, Sigma, Ref. S4881) or no stimuli in the presence of CD28/49d co-stimulatory molecules (1 μl/ml, BD Biosciences), Golgistop solution (Monensin A, BD Biosciences, Spain), and anti-human CD107a (PE-Cy5, BD Biosciences) for 6 h at 37°C in a 5% CO₂ incubator and overnight at 4°C. Stimulation cells were then washed with 1X PBS and stained for 25 min with the Live/Dead probe (APC-Cy7, Invitrogen). Next, cells were washed with 1X PBS and surface stained for 25 min with CD3 (A700, BD Biosciences), CD4 (APC-Cy7, BD Biosciences), CD8 (V500, BD Biosciences), PD-1 (BV421, BD Biosciences), LAG3 (PE, BD Biosciences), TIM-3 (A647, BD Biosciences), and CD39 (FITC, BD Biosciences). Subsequently, cells were washed twice in 1X PBS and fixed and permeabilized with the Fix&Perm kit (A and B solutions, Thermo Fisher Scientific) for intracellular cytokine staining of IFNγ (BV711, BD Biosciences) and IL-2 (BV650, BD Biosciences). After 20 min at room temperature in the dark, stained samples were washed twice with 1X PBS, resuspended in 1% formaldehyde and acquired on an LSR Fortessa. Data were analyzed with FlowJoV (Tree Star Inc.). Patterns of simultaneous expression of IFNγ, CD107a, and IL-2, or combinations of PD-1, LAG-3, TIM-3, and CD39 were analyzed using Pestle and SPICE v5 software (55).

Statistical analyses were performed using Prism v4 (GraphPad Software, Inc.). Reported p-values were calculated using the Wilcoxon single-rank test in paired comparisons and the t-test and Mann-Whitney test for independent median comparisons. Also we used one-way ordinary analysis of variants (ANOVA), and one-way ANOVA for multiple group comparisons adjusting the pairwise analyses by Tukey. The relationship between HIV-1 induction measured according to the expression of p24 and GFP, and the relationship between the lack of inhibitory receptors in CD8+ and percentage of killing were assessed using the Spearman correlation coefficient. All statistical tests were under a significance level of 0.05.

RB and BH were employed by the company Merck. JGP and JM-P report research funding from Merck. JGP and AR have a patent on the methods for screening of HIV-1 latency reversing agents issue (62590081). The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.