Combination Immune Checkpoint Blockade to Reverse HIV Latency

Antiretroviral therapy (ART) has revolutionized the treatment of HIV infection and has dramatically reduced mortality and morbidity. However, ART is lifelong, expensive, and often has side effects, so there is an urgent need to identify strategies to cure HIV or induce “remission” to avoid lifelong treatment (1). The major barrier to a cure for HIV infection is the persistence of latent infection in long-lived resting and proliferating CD4⁺ T cells (2–5) that are more frequently detected in lymphoid tissue and the gastrointestinal tract (6–8). It is highly likely that the mechanisms maintaining HIV latency differ in nonproliferating and proliferating T cells, suggesting that multiple interventions may be needed to eliminate latency.

One approach being tested to eliminate latently infected cells in people living with HIV (PLWH) on ART is to activate latent virus and thereby induce death of the infected cell through immune clearance or virus-induced cytolysis (9, 10). To date, clinical trials that have examined latency-reversing agents (LRA), such as histone deacetylase inhibitors (HDACi), disulfiram, or TLR agonists, have shown modest latency reversal but without clearance of infected cells (11–15). Furthermore, HDACi have been shown to induce adverse effects on adaptive immune function in vitro and have multiple off target effects (11, 16–21). Therefore, alternative LRAs that are more potent and have a beneficial effect on adaptive immune function to enhance immune-mediated clearance of infected cells are needed.

Latent infection is enriched in CD4⁺ T cells expressing immune checkpoint (IC) molecules, first described for programmed death 1 (PD-1) in circulating CD4⁺ T cells in blood (22, 23), and more recently in T follicular helper cells in lymphoid tissue (7). In SIV-infected macaques on ART, there is also enrichment of virus in CD4⁺ T cells expressing CTL-associated protein 4 (CTLA-4) and PD-1 in the extrafollicular and follicular lymphoid compartments, respectively (24). We previously demonstrated that coexpression of multiple IC molecules, including PD-1, lymphocyte activation gene 3 (LAG-3), and T cell immunoreceptor with Ig and ITIM domains (TIGIT), on CD4⁺ T cells from blood from PLWH on ART were highly enriched for HIV infection compared with cells that expressed fewer than three IC markers (23).

HIV transcription and virus production is largely dependent on host transcription factors that increase and localize to the nucleus following T cell activation (reviewed in Ref. 1). Ligation of IC molecules can actively suppress these pathways (25). We recently demonstrated that engagement of PD-1 in vitro inhibits the establishment of HIV latency in resting CD4⁺ T cells (26), and in latently infected cells isolated from PLWH on ART, programmed death ligand (PD-L) 1 can block viral production at the transcriptional level by abrogating TCR-induced HIV reactivation (27). Conversely, PD-1 blockade with the mAb pembrolizumab in combination with the LRA bryostatin enhances HIV production ex vivo without increasing T cell activation (27). Together, these data suggest that IC molecules alone or in combination can define latently infected cells, and PD-1 has a functional role in the establishment, maintenance, and reversal of HIV latency in CD4⁺ T cells (reviewed in Ref. 28), in addition to the more global effects PD-1 might have on Ag-specific T cell function.

The administration of Abs to PD-1, PD-L1, or CTLA-4, either alone or in combination, are now licensed in many countries for the treatment of malignancy (29), and multiple small case series have demonstrated that the safety of anti–PD-1 and anti–PD-L1 in PLWH is similar to individuals without HIV infection (30–37). Abs to other IC molecules, such LAG-3 and TIGIT, are also now in clinical development and early phase studies (38).

In a single case report of a person living with HIV on ART with metastatic melanoma, we demonstrated that the administration of anti–CTLA-4 (ipilimumab) led to a significant increase in cell-associated unspliced HIV RNA, potentially consistent with latency reversal (39). Additionally, in the same person, a single dose of anti–PD-1 (nivolumab) increased cell-associated HIV RNA in CD4⁺ T cells (26). Other case reports of anti–PD-1 (pembrolizumab) have not shown a change in cell-associated unspliced HIV RNA, although the timing of sampling blood differed in each of these studies (40). Finally, repeated dosing of anti–PD-1 has been associated with a decline in cell-associated HIV DNA, potentially consistent with clearance of infected cells (27, 41). We therefore hypothesized that IC molecules provide an attractive target to eliminate latent infection and that administration of Abs to multiple IC molecules in combination could synergistically reverse inhibitory signals required to maintain latency, thereby leading to latency reversal.

The HEK 293T and TZM-bl cell lines cells were grown as a monolayer DMEM (Life Technologies) supplemented with 10% (v/v) heat-inactivated FCS (Bovogen Biologicals, Keilor East, Australia), 100 U/ml penicillin, 100 μg/ml streptomycin, and 292 μg/ml glutamine (Life Technologies) at 37°C and 5% CO₂.

PBMC were isolated by Ficoll-Paque density gradient centrifugation (GE Healthcare, Chalfont St. Giles, U.K.) from buffy coats obtained from the Australian Red Cross Blood Service (Melbourne, Australia). Resting CD4⁺ T cells were negatively selected using magnetic cell sorting with an in house mixture of Abs to CD8 (clone OKT8), CD11b (OKM1), CD16 (3G8), HLA-DR (2-06), CD19 (FMC63), CD14 (FMC17) and CD69 (L78), goat anti-mouse IgG magnetic beads (Miltenyi Biotec), and the AutoMACS Pro (Miltenyi Biotec), as previously described (42–44). Purity (≥96%) was based on CD3⁺ and CD4⁺ expression with mouse anti-human(m-a-h)CD3-PE (HIT3a) and m-a-hCD4-FITC (M5E2; BD Biosciences).

Monocytes were positively selected from PBMC using magnetic cell sorting using CD14 Abs (FMC17), as previously described (45). Purity (≥98%) was based on CD14⁺ expression with m-a-hCD14-FITC (M5E2; BD Biosciences). Primary cells were cultured in RPMI 1640 (Life Technologies) supplemented with 10% (v/v) heat-inactivated FCS, 100 U/ml penicillin, 100 μg/ml streptomycin, and 292 μg/ml glutamine (RF10) at 37°C and 5% CO₂.

Plasmid DNA encoding a NL4.3 backbone with an AD8 envelope and EGFP inserted one base downstream of the env open reading frame (46) followed by an IRES-nef coding sequence, kindly provided by Y. Tsunetsugu-Yokota (National Institute of Infectious Diseases, Tokyo, Japan), was transiently transfected into HEK 293T cells with calcium phosphate precipitation (47). Virus supernatant was harvested after 2 d, filtered by passage through a 0.2-μm filter, concentrated using 20% sucrose density ultracentrifugation, and stored in aliquots at −80°C. Cells were infected for 2–3 h at a multiplicity of infection of 0.5 as determined by limiting dilution on TZM-bl using the Reed and Muench method (48), followed by a wash step to remove unbound virus.

Resting CD4⁺ T cells were labeled with proliferation dye eFluor 670 (eBiosciences) per the manufacturer’s instructions. Labeled T cells were cultured with or without syngeneic-sorted monocytes (monocyte/T cell ratio of 1:10) with 20 ng/ml staphylococcal enterotoxin B (SEB; Sigma-Aldrich) and 10 U/ml recombinant human (r-h) IL-2 (Roche Diagnostics) for 24 h, infected with full length nef-competent EGFP reporter virus for 2–3 h, after which excess virus was washed away and cells were cultured for 5 d in media supplemented with SEB and 10 U/ml r-h IL-2. Productive infection was determined day 5 postinfection by detecting EGFP⁺ cells using flow cytometry. Subsequently, the nonproductively infected (EGFP⁻), CD3⁺, HLA-DR⁻ (m-a-hCD3-PB, UCHT1, m-a-hHLA-DR-BV650, G46-6; from BD Biosciences) nonproliferating (eFluor 670^HI) and proliferating (eFluor 670^LO) CD4⁺ T cells were sorted using a FACS-Aria, MoFlo Astrios, or FACSAria Fusion cell sorters (BD Biosciences). In some experiments, the CD3⁺HLA-DR⁻EGFP⁻ nonproliferating and proliferating cells were further sorted based on IC expression using m-a-hCD279/PD-1-PE (EH12.1; BD Biosciences), m-a-hTIM-3-PE (F38-2E2; BioLegend), m-a-hCD152/CTLA-4-PE (BNI3; BD Biosciences), m-a-hCD272/BTLA-PE (MIH26; BioLegend), m-a-hTIGIT–eFluor 710 (MBSA43; eBioscience). In other experiments, nonproliferating and proliferating T cells were sorted based on the expression of any of the following molecules: PD-1, T cell Ig and mucin domain-containing protein 3 (TIM-3), CTLA-4, and B and T lymphocyte attenuator (BTLA). For these experiments, all four PE-conjugated Abs were added to the cells in combination with m-a-hCD45RA-V500 (Hl100; BD Biosciences) to exclude CD45RA⁺ from the nonproliferating cell population prior to IC gating.

To determine latent infection, 100,000–200,000 sorted T cells were cultured in a flat-bottom 96-well plate with 200 μl RF10 supplemented with integrase inhibitor L870812 (Merck, Whitehouse Station, NJ) or raltegravir (AIDS Research and Reference Reagent Program, Division of AIDS, National Institute of Allergy and Infectious Diseases, National Institutes of Health or SelleckChem), both used at a final concentration of 1 μM in the presence (positive control) or absence (negative control) of activation stimuli composed of 3 μg/ml plate-bound m-a-hCD3 (UCHT1; BD Biosciences), 2 μg/ml soluble m-a-hCD28 (L293; BD Biosciences), 50 ng/ml r-hIL-7 (Sigma-Aldrich) and 10 U/ml r-hIL-2. Cells were harvested 72 h after stimulation, and EGFP expression was quantified using the FACSCalibur or FACSCanto II (BD Biosciences). Results were analyzed using Weasel Software (Walter and Elisa Hall Institute of Medical Research, Melbourne, Australia), and to quantify latent infection, the number of EGFP⁺ cells in the unstimulated culture (background) was subtracted from the number of EGFP⁺ cells in the stimulated culture. In some experiments, Abs blocking IC molecules; humanized anti–PD-1 (nivolumab; a kind gift from Bristol-Myers Squibb, New York, NY), humanized anti–CTLA-4 (ipilimumab; a kind gift from Bristol-Myers Squibb), m-a-hTIM-3 (clone F38-2E2; BioLegend), m-a-hTIGIT (clone MBSA43; eBioscience), or isotype controls (hIgG4 [clone ET904; BioLegend], hIgG1 [clone ET901; BioLegend], mIgG1 [clone MOPC-21; BioLegend]), were added to activate latent infection. For these experiments 10 μg/ml IC blocker (ICB) or isotype control was added to sorted cells with back-added thawed syngeneic monocytes (monocyte/T cell ratio of 1:10) in the presence or absence of SEB and with the standard activation stimuli in a parallel culture (positive control) for 72 h. When multiple ICB were added, equal amounts of the appropriate isotype control Ab were added to the control culture.

To calculate the ICB-induced latency reversal expressed as a percentage of maximal stimulation with anti-CD3/CD28 plus IL-7 plus IL-2, the number of EGFP⁺ cells in the unstimulated culture was subtracted from the number of EGFP⁺ cells in the anti-CD3/CD28 stimulated culture, and this was set at 100% (maximal stimulation). To calculate the effect of ICB-induced reversing latency, the number of EGFP⁺ cells in the unstimulated control was subtracted from the number of EGFP⁺ cells in the ICB-treated culture and expressed as a percentage of maximal stimulation.

In other experiments, LRAs romidepsin (FK288, depsipeptide, 1 or 0.1 nM; SelleckChem), JQ1 (1 or 0.1 μM; SelleckChem), bryostatin (10 or 1 nM; Sigma-Aldrich), PMA plus ionomycin (2 nM and 0.5 μM, respectively; both from Sigma-Aldrich), or DMSO (as reconstitution solvent control; Sigma-Aldrich) were added to sorted cells with the standard activation stimuli in a parallel culture (positive control) for 72 h.

In all experiments, live and single cells were gated using forward and side scatter plots. IC ligand expression on myeloid DC and monocytes was determined on CD3⁻HLA-DR⁺ cells using m-a-hCD3-PB (UCHT1; BD Biosciences), m-a-hHLA-DR-PerCP (L243; BD Biosciences), m-a-hCD274/PD-L1-PE (MIHI; BD Biosciences), m-a-hCD273/PD-L2-PE (MIH18; BD Biosciences), m-a-hCD80-PE (L307.4; BD Biosciences), m-a-hCD86-PE (2331; BD Biosciences), m-a-hCD275/ICOSLigand-PE (2D3/B7-H2; BD Biosciences), m-a-hGal9-PE (9M1-3; BioLegend), m-a-hCD112/Nectin-2/PVRL-2-PE (TX31; BioLegend), m-a-hCD155/PVR-PE (TX24; BioLegend), m-a-hCD270/HVEM-PE (eBioHVEM-122; eBioscience), or mIgG2a-PE isotype control (MOPC-173; BioLegend).

Single IC expression on T cells was determined by gating on CD3⁺HLA-DR⁻EGFP⁻ cells, which were subsequently separated into eFluor 670^HI (nonproliferating) and eFluor 670^LO (proliferating) populations using m-a-hCD3-PB (UCHT1; BD Biosciences), m-a-hHLA-DR-PE-TexR (TU36; BD Biosciences), m-a-hCD279/PD-1–PE (EH12.1; BD Biosciences), m-a-hTIM-3-PE (F38-2E2; BioLegend), m-a-hCD152/CTLA-4–PE (BNI3; BD Biosciences), m-a-hCD272/BTLA-PE (MIH26; BioLegend), m-a-hTIGIT–eFluor 710 (MBSA43; eBioscience), mIgG2a-PE isotype control (MOPC-173; BioLegend) or mIgG1–eFluor 710 isotype control (P36281; eBioscience).

Coexpression of PD-1, TIM-3, and TIGIT on CD3⁺HLA-DR⁻EGFP⁻ nonproliferating and proliferating cells was done using m-a-hCD3-PB, m-a-hHLA-DR-PE-TexR, or m-a-hHLA-DR-BV650 (G46-6; BD Biosciences), m-a-hCD279/PD-1–PE, m-a-hTIM-3-PE-Cy7 (F38-2E2; BioLegend), m-a-hTIGIT–eFluor 710, mIgG1–PE isotype control (MOPC-21; BD Biosciences), mIgG1-PE-Cy7 isotype control (MOPC-21; BioLegend) and mIgG1–eFluor 710 isotype control.

Coexpression of PD-1, CTLA-4, TIM-3, and TIGIT on CD3⁺HLA-DR⁻EGFP⁻ nonproliferating and proliferating cells was done using m-a-hCD3-BV711 (UCHT1; BD Biosciences), m-a-hHLA-DR-BV650, m-a-hCD279/PD-1–BV421 (EH12.1; BD Biosciences), m-a-hCD152/CTLA-4–PE, m-a-hTIM-3-PE-Cy7 and m-a-hTIGIT–eFluor 710.

The percentage of positive cells was calculated by subtracting fluorescence minus one (FMO) or isotype control values.

The use of blood samples from HIV-negative donors for this study was approved by the Human Research and Ethics Committees from the Alfred Hospital (HREC156/11), Monash University (CF11/1888) and the University of Melbourne (1443071). Donors were recruited by the Red Cross Blood Transfusion Service, and all provided written informed consent for the use of their blood products for the research.

Differences between experimental conditions were analyzed using Wilcoxon matched-pairs signed-rank test (n > 5) or paired Student t test (n ≤ 5) on GraphPad Prism (version 6). All p values ≤ 0.05 were considered significant.

To study the role of IC molecules in HIV latency, we first assessed the expression of multiple IC molecules: PD-1, CTLA-4, TIGIT, TIM-3, BTLA, and their ligands in an in vitro model of latency (44, 45). The model involves coculture of resting CD4⁺ T cells with monocytes in the presence of SEB, which are known to express high levels of IC ligands, followed by HIV infection with an EGFP-expressing virus (49) and separation of proliferating and nonproliferating T cells based on expression of the proliferation dye eFluor 670 (50) (Fig. 1A). IC molecule expression was determined on T cells before infection and day 1–4 postinfection and coculture. Monocytes were excluded from the analysis by gating for CD3⁺HLA-DR⁻ cells, and IC molecule expression was measured by flow cytometry on the nonproductively infected (EGFP⁻), nonproliferating (eFLuor^HI), and proliferating (eFLuor^LO) T cells (Fig. 1B–G); the percentage of cells expressing IC markers was calculated over the 4 d (Fig. 1H). Postinfection data from the 3 d time point are summarized in Fig. 1I. In the nonproliferating T cell population, the mean percentage of T cells expressing PD-1, TIM-3, BTLA, and TIGIT all significantly increased, but there was no change in expression of CTLA-4 (Fig. 1I). In the proliferating T cells, the increase in expression of all IC molecules was substantial, with up to 30–40% of cells expressing at least one IC molecule (Fig. 1I).

Next, we determined whether the cultured monocytes expressed ligands for IC molecules, including PD-L1 and PD-L2 (ligands for PD-1), galectin 9 (Gal9; ligand for TIM-3), herpes virus entry mediator (HVEM, ligand for BTLA), CD80 and CD86 (ligands for CTLA-4), poliovirus receptor (PVR) and PVR-like 2 (PVRL2, ligands for TIGIT). The same approach was used as described in Fig. 1A, except this time we measured the expression of the IC ligands 1 d following infection, and T cells were excluded from the analysis by gating for CD3⁻HLA-DR⁺ cells (Fig. 2A). IC ligand expression was compared with FMO or isotype control (Fig. 2B), and the percentage of cells expressing the IC ligands was calculated (Fig. 2C). Monocytes expressed all IC ligands tested, but there was a lower frequency of cells expressing PD-L1, PD-L2, CD80, and PVR compared with the other ligands (Fig. 2C). There were no differences in IC ligand expression in the presence and absence of HIV (Fig. 2C).

CD4⁺ T cells from PLWH on ART that express PD-1 (22) or multiple IC molecules are enriched for HIV DNA (23). We first confirmed whether these findings could be reproduced in our in vitro model of HIV latency. T cells were cultured with syngeneic monocytes in the presence of SEB and infected with the EGFP reporter virus. On day 5 postinfection, EGFP⁻ nonproductively infected cells (i.e., potentially latently infected) in nonproliferating and proliferating T cell populations were sorted based on high and low eFluor 670 expression, respectively. These two populations were then sorted on the basis of IC molecule expression (Supplemental Fig. 1), and inducible latent virus was quantified, as previously described (44, 45) (Fig. 3A). Briefly, sorted cells were activated with anti-CD3/anti-CD28 plus IL-7 plus IL-2 in the presence of raltegravir (an HIV integrase inhibitor) to prevent subsequent rounds of infection. An increase in EGFP expression following stimulation was consistent with postintegration-inducible latent infection (44). Postintegration latency was significantly enriched in nonproliferating T cells expressing high levels of PD-1, TIM-3, CTLA-4, or BTLA but not TIGIT when compared with cells that did not express these IC molecules (p = 0.01, 0.03, 0.01, 0.03, and 0.24, respectively) (Fig. 3B, Table I). In proliferating T cells, postintegration latency was significantly enriched in PD-1^HI compared with PD-1^LO cells (p = 0.04, Fig. 3B). No enrichment for HIV-infected cells in IC-expressing cells was detected when assessed for all other IC molecules (Fig. 3B).

To determine whether coexpression of IC molecules resulted in a greater enrichment of HIV within our model, we performed the same experiments as described above, but we sorted cells that expressed any combination of PD-1, TIM-3, CTLA-4, BTLA, or none of these. Postintegration latency was significantly enriched in both nonproliferating (mean fold change [MFC] = 42; p = 0.03) and proliferating (MFC = 48; p = 0.03) T cells expressing one or more of the IC molecules as compared with cells that expressed none of the IC molecules (Fig. 3B, Table I). However, for both nonproliferating and proliferating T cells, latency was still detected in the cells expressing low/none of the IC molecules. Finally, we observed no statistically significant correlations when comparing the frequency of IC-expressing cells and the frequency of latent infection in either the nonproliferating or proliferating T cells (Supplemental Fig. 2). BTLA was excluded from these last experiments, as there is currently no blocking Ab available, and we aimed to investigate IC that have blocking Abs available or in development for clinical use. Taken together, these data demonstrate enrichment of latent HIV infection in proliferating and nonproliferating cells expressing multiple IC molecules alone or in combination.

Given that different cells can express one or more IC molecules, we next defined coexpression of IC molecules on T cells in this in vitro model for HIV latency. We first assessed the coexpression of PD-1, TIM-3, and TIGIT, as some of these IC molecules have been shown to be expressed on cells enriched for HIV using CD4⁺ T cells from PLWH on ART (23), and we had previously shown that PD-1 and TIM-3 were important in the establishment of latency (26). Resting CD4⁺ T cells were cocultured with monocytes and analyzed for IC expression by flow cytometry. The percentage of cells positive for one IC (i.e., PD-1 or TIM-3 or TIGIT), two IC, or three IC in combination was calculated (Fig. 4A). Three days postinfection, nonproliferating cells more frequently expressed a single IC than any combination of two IC (n = 4; p < 0.001, compared with single IC) or three IC (n = 4; p < 0.001, compared with single IC) (Fig. 4B). Concomitant expression of two or three IC was far more common on proliferating than on nonproliferating cells. No statistically significant difference was observed in the frequency of cells expressing a single IC between nonproliferating and proliferating T cells. On day 3 postinfection, at the peak of IC expression compared with nonproliferating T cells, proliferating T cells expressed two and three IC more frequently (p = 0.0017, 0.002, respectively, for both) (Fig. 4B). Overall, these results show that nonproliferating T cells more often express a single IC molecule, whereas proliferating T cells more commonly coexpress two or more IC molecules. These results were similar in the presence and absence of HIV infection (data not shown).

Given the availability of anti–CTLA-4 in the clinic for management of melanoma (51), we also examined the coexpression of the same three IC together with CTLA-4. Proportions of cells expressing CTLA-4 was less frequent compared with the other IC molecules (i.e., PD-1 and TIGIT on nonproliferating and PD-1 and TIM-3 on proliferating T cells) (Fig. 4C). Again, we found that there was more frequent expression of IC molecules in proliferating cells compared with nonproliferating cells, (expression of more than one IC was found in 46 and 7% and a single IC molecule in 33 and 15% of cells, respectively) (Fig. 4C–E). Coexpression of all four IC molecules was the rarest population within either proliferating or nonproliferating cells (Fig. 4E). Collectively, this data shows that the distribution of PD-1, CTLA-4, TIM-3, and TIGIT, expressed individually and in combination, differs between nonproliferating and proliferating T cells, and that expression of more than one IC molecule was more frequent in proliferating T cells.

To examine whether blocking IC, either alone or in combination, can reverse HIV latency in vitro, resting CD4⁺ T cells were cocultured with monocytes in the presence of SEB and sorted into EGFP⁻ nonproliferating and proliferating populations, as described above. ICB were added to the sorted cells in the presence of raltegravir and T20 (an integrase and fusion inhibitor, respectively). As we exclude HLA-DR⁺ T cells during the sort on day 5 postinfection and select for CD3⁺ cells, the monocytes were excluded from the sorted population. Therefore, for the IC ligand to be present in the blocking experiment, syngeneic monocytes were added back to the cultures postsort (Fig. 5A). Additionally, sorted cells were cultured in the absence of monocytes with anti-CD3/CD28 plus IL-7 plus IL-2 to induce maximal HIV production and to which all conditions were compared.

The effect of anti–PD-1 (nivolumab) and anti–CTLA-4 (ipilimumab) were tested alone or in combination, as well as in combination with commercially available Abs to TIM-3 and TIGIT. BTLA was not further examined, as blocking agents are currently unavailable. Effects of ICB were assessed in the presence or absence of SEB. SEB is a super-Ag that will drive proliferation in a subset of cells that express a specific Vb repertoire (52). We assumed that SEB will drive proliferation of both uninfected and latently infected cells in this model. Latency reversal was quantified by measuring the increase in absolute number of cells that express EGFP following stimulation (Supplemental Fig. 3A) or as a proportion of maximal stimulation with anti-CD3/CD28 plus IL-7 plus IL-2 (Fig. 5B). The mean (range) of EGFP⁺ cells per 10⁴ cells in the untreated control was 10.8 (8.4–17.1)/10⁴ cells, and this increased to 25.9 (12.0–37.2)/10⁴ cells following stimulation of the sorted nonproliferating T cells using a maximal stimulus (Supplemental Fig. 3A). For the sorted proliferating T cells, the number of EGFP⁺ cells increased from 22.3 (3.5–47.5)/10⁴ to 46.2 (25.6–73.3)/10⁴ using anti-CD3/CD28 plus IL-7 plus IL-2 (Supplemental Fig. 3B).

In nonproliferating CD4⁺ T cells, there was an increase in the number of EGFP⁺ cells following addition of anti-CD3/CD28 plus IL-7 plus IL-2, as expected (MFC 2.5; n = 6; p = 0.031) (Supplemental Fig. 3A), although there was quite considerable variation in inducible EGFP expression between donors. In the absence of SEB, a significant increase in EGFP⁺ cells was only observed when all four ICB were added (p = 0.03; Fig. 5B, Supplemental Fig. 3A). In the presence of SEB, the number of EGFP⁺ cells increased with anti–PD-1 alone (p = 0.031) and in combination with anti–CTLA-4 (p = 0.031) but not with anti–CTLA-4 alone (p = 0.22), consistent with only anti–PD-1 inducing latency reversal. The addition of Abs to TIM-3 (p = 0.16) and TIGIT (p = 0.22) had no further effect in enhancing EGFP expression (Fig. 5B).

In the proliferating T cells, there was an increase in EGFP⁺ cells following the addition of anti-CD3/CD28 plus IL-7 plus IL-2 addition as expected (n = 6; MFC 3.38; p = 0.031) (Supplemental Fig. 3B). In the absence of SEB, the results were similar to the nonproliferating cells, with an increase in EGFP only observed following the addition of all four ICB. In the presence of SEB, we observed that only anti–CTLA-4 induced a modest increase in the number of EGFP⁺ cells (p = 0.031) (Fig. 5C, Supplemental Fig. 3B). Interestingly, there was a decrease in the number of EGFP⁺ cells in the presence of anti–PD-1 (MFC 0.83; p = 0.031) which was not related to a decrease in cell viability (data not shown).

To understand what factors might be associated with efficient latency reversal, we assessed whether there was a correlation between the induction of EGFP expression and the degree of T cell proliferation, as measured by an increase in the proportion of eFluor 670^LO cells for the nonproliferating cell cultures or a decrease in eFluor 670 mean fluorescent intensity in the proliferating cells, given the dye is diluted with each cell division. Following the addition of the ICB, no relationship was observed between cell proliferation and induction of EGFP expression (data not shown).

In summary, these data demonstrate that latency reversal could be achieved in both nonproliferating and proliferating latently infected cells when multiple ICB were used in combination without SEB. The degree of latency reversal was high; approaching 45% of what was achieved with maximal stimulation. In the presence of SEB, we found that latency reversal was possible using a single ICB in nonproliferating latently infected cells (using anti–PD-1) and in proliferating cells (using anti–CTLA-4) (Table II). These data suggest that the use of a single ICB is unlikely to sufficiently activate latent infection in either proliferating or nonproliferating cells, and an additional stimulus is required.

To investigate the relative potency of IC blockade for latency reversal compared with other LRAs that have been shown to reverse latent HIV, we next assessed commonly used LRAs in the same in vitro model in separate experiments (Fig. 6A). We tested three different classes of LRAs: the HDACi romidepsin, the bromodomain inhibitor JQ1, and an activator of protein kinase C (PKC), bryostatin. Of note, lower doses of romidepsin were used for these cell cultures that were maintained for 3 d. We did this to avoid the toxic effects of 40 nM romidepsin observed in continuous T cell cultures elsewhere (data not shown). As positive controls, we included anti-CD3/CD28 plus IL-7 plus IL-2 and the combination of PMA and ionomycin that mimic T cell activation, activating the transcription factors NF-κB and NFAT to initiate HIV transcription (53). Of the LRAs tested, only bryostatin led to a statistically significant increase in the frequency of EGFP⁺ cells in both nonproliferating and proliferating latently infected cells, reaching 65 and 43% of maximal stimulation induced by anti-CD3/CD28 plus IL-7 plus IL-2 (Fig. 6, Supplemental Fig. 3, Table II). In this model of latency, we demonstrate that IC blockade in combination or using SEB with either anti–PD-1 or anti–CTLA-4 was more potent than commonly used LRAs such as romidepsin (tested at continuous low doses) or JQ1.

In PLWH on ART, HIV persists preferentially in CD4⁺ T cells expressing multiple IC molecules, in particular PD-1 (7, 22, 23). To determine the role of multiple IC molecules and blockers in the maintenance and reversal of HIV latency, we used an in vitro model that can establish latency in both nonproliferating and proliferating T cells (44, 45, 50). These two populations of latently infected cells cannot be distinguished when evaluating CD4⁺ T cells ex vivo from PLWH on ART. We found HIV latency to be enriched in nonproliferating and proliferating T cells expressing IC molecules, and the enrichment was greater in nonproliferating cells that expressed multiple IC molecules. Using multiple ICB that are licensed for the treatment of malignancy (anti–PD-1 and anti–CTLA-4) and in preclinical development (anti–TIM-3 and anti-TIGIT), we demonstrated that reversal of HIV latency was possible with a single IC Ab in combination with a T cell activation stimulus (SEB in this study) or with a combination of multiple Abs to IC molecules without another stimulus. In this latency model, the potency of IC blockade in reversing latency was comparable to reversing HIV latency with the PKC activator bryostatin and greater than commonly used LRAs, such as the HDACi romidepsin (used at low dose) or the bromodomain inhibitor JQ1. Given the additional potential beneficial effects of IC blockade in enhancing the function of HIV-specific CD4⁺ and CD8⁺ T cells (54, 55), IC blockade is an attractive option to move forward into clinical evaluation as both a potential LRA and as a strategy to boost HIV-specific T cell function.

Using an in vitro model of HIV latency, latent infection was found to be enriched in nonproliferating CD4⁺ T cells expressing PD-1, TIM-3, CTLA-4, or BTLA, but in proliferating T cells, this enrichment was only found in T cells expressing PD-1. IC molecules were expressed at high levels but were often coexpressed on proliferating T cells, whereas nonproliferating T cells more commonly expressed a single IC molecule. The mechanisms behind the need for multiple ICB to reverse latent infection in nonproliferating and proliferating cells may be different. For the nonproliferating cells, multiple ICB may be required to target different cells that express a single IC molecule, whereas for the proliferating cells, multiple ICB are needed because of the high frequency of coexpression of different types of ICs on the same cell.

Blocking IC molecules inhibits the negative signaling in the T cell activation process, effectively releasing the breaks on T cell activation. TCR and superantigen signaling induce multiple transcription factors to become active and/or translocate to the nucleus leading to induction of HIV transcription and virus production (56, 57), among many other changes in the host cell. These host transcription factors predominantly include NF-κB, NFAT, specificity protein 1 (SP-1) and AP-1 (reviewed in Ref. 58). We recently showed that PD-1 signaling prevents activation of the positive transcription elongation factor b (P-TEFb), a master regulator of HIV transcription, through inhibition of TCR-induced CDK9 phosphorylation and cyclin T1 expression (27). Whether these or other pathways are important for latency reversal following combination IC blockade remains to be determined.

In our in vitro model, we found that a single ICB could reverse latency but only in the presence of SEB (see Fig. 5). Furthermore, anti–PD-1 was the only Ab that could reverse latency in nonproliferating T cells, whereas only anti–CTLA-4 reversed latency in proliferating T cells. The lack of activity of anti–CTLA-4 in nonproliferating cells may be a result of the low expression of CTLA-4 on these cells (see Fig. 1H). This is expected because CTLA-4 is upregulated on cells following TCR priming (59) and these cells had not undergone proliferation in response to the superantigen SEB. In contrast, the proliferating T cells expressed high levels of CTLA-4 (see Fig. 1H), as expected, following stimulation with a superantigen or TCR-mediated stimulation. Similarly, the failure of anti–PD-1 to reverse HIV latency in proliferating cells when used on its own would be a major limitation as an LRA. Although administration of the combination of anti–PD-1 and anti–CTLA-4 is now recommended in the management of some malignancies, toxicity was significantly increased (29). An increase in toxicity would significantly limit the use of combination IC blockade as an HIV cure intervention, given that PLWH now have a near normal life expectancy on long term ART.

Blocking IC molecules can also induce T cell proliferation. It is therefore feasible that driving T cell proliferation could increase the number of latently infected cells in vivo, which would not be desirable. In our in vitro model, we observed enhanced T cell proliferation upon IC blockade when combined with superantigen SEB, as expected, but no correlation was observed between the degree of proliferation and reversal of latency (data not shown). In our model, we did not directly measure the number of latently infected cells but, rather, the number of cells with inducible virus. It is certainly possible that the cultures also contain cells infected with intact virus that do not respond to stimulation. This has been well described in CD4⁺ T cells from PLWH on ART and referred to as noninduced intact proviruses (60). Further work is needed to also measure the total number of infected cells prior to and following IC blockade to exclude the possibility of proliferation of infected cells that contain noninduced intact provirus.

There were several limitations to this study. First, we used an in vitro model to probe the effects of different IC molecules. These results need confirmation using CD4⁺ T cells from PLWH on ART ex vivo and/or in vivo administration of an ICB Ab. We are currently studying changes in cell-associated HIV RNA and plasma HIV RNA in the context of clinical trials of anti–PD-1 and anti–CTLA-4 in PLWH with cancer (https//clinicaltrials.gov NCT02408861). Despite this limitation, our model has several advantages, including the following: 1) the ability to separate proliferating and nonproliferating latently infected cells, which is currently not possible with cells from PLWH; 2) the measurement of inducible virus rather than cell-associated RNA, which allows quantification of the number of infected cells that undergo latency reversal; and 3) the incorporation of a superantigen to drive proliferation. A second limitation of this study is that we did not assess every possible combination of two or three ICB. Instead, we focused on the clinically relevant combination of anti–PD-1 and anti–CTLA-4 and added other ICB (TIM-3, TIGIT) to this well-established, clinically tested combination. A third limitation is that the use of SEB is not a physiological stimulus but can mimic an Ag-specific response. Other costimuli should be further evaluated. A fourth limitation of our study is that we did not evaluate the role of LAG-3 because we were unable to detect an increase in LAG-3 expression upon coculturing the CD4⁺ T cells with monocytes. Furthermore, we did not have access to a clinically relevant anti-LAG3 Ab to test in vitro. Finally, we only quantified inducible EGFP expression as a marker of latency reversal. Other proximal markers, such as cell-associated unspliced and multiply spliced RNA, were not quantified; however, given there are blocks in transcriptional elongation and splicing described in resting CD4⁺ T cells (61) and that protein expression will be required for either immune-mediated clearance or virus-induced cytolysis, we felt that quantification of protein expression would be most informative.

In conclusion, using an in vitro model that can establish latency in both nonproliferating and proliferating T cells, we show that Abs blocking signaling of specific IC molecules could reverse HIV latency in both proliferating and nonproliferating T cells, but this was only achieved in the presence of an additional T cell–activating stimulus or when four Abs were used in combination. Given that SEB is not a physiological stimulus and not feasible to deliver in vivo (62), further studies are needed to identify alternative secondary stimulus to replace SEB. One approach could be therapeutic vaccination for HIV (63), which could drive T cell proliferation but also enhance the generation of Ag-specific T cells, as recently demonstrated in the setting of cancer (64). Alternatively, the PKC agonist bryostatin or a related compound (65) might be a complimentary stimulus, given that bryostatin appears to be safe in vivo (66), induces T cell activation, and acts as an LRA alone (53) or in combination with anti–PD-1 (27). The administration of a combination of ICB should be further explored as a path to HIV remission following cessation of ART.

We thank Damian Purcell (University of Melbourne, Parkville, Australia) and Yasuko Tsunetsugu-Yokota (National Institute of Infectious Diseases, Tokyo, Japan) for providing the EGFP reporter virus. We thank G. Paukovics, J. Le Masurier, and P. Donaldson (Alfred Medical Research and Educational Precinct Flow Cytometry Core Facility, Melbourne, Australia) and T. Luka, C. Li, A. Hind, and D. Blashki (University of Melbourne Flow Cytometry Facility, Melbourne, Australia) for flow cytometric cell sorting. We thank Ashish Nair (University of Melbourne) for help processing the human buffy coats. We thank Alan Korman from Bristol Myers Squibb for providing nivolumab (anti–PD-1) and ipilimumab (anti–CTLA-4).

S.R.L. has received honoraria to her institution for participation in advisory boards and or educational seminars from Gilead Sciences, Merck, Viiv Healthcare and Abbvie. S.R.L. has received investigator initiated research grant support from Gilead Sciences, Merck and Viiv Healthcare. The other authors have no financial conflicts of interest.