Programmed cell death-1 contributes to the establishment and maintenance of HIV-1 latency

Abstract

Objective: 

In HIV-infected individuals on antiretroviral therapy (ART), latent HIV is enriched in CD4⁺ T cells expressing immune checkpoint molecules, in particular programmed cell death-1 (PD-1). We therefore assessed the effect of blocking PD-1 on latency, both in vitro and in vivo.

Methods: 

HIV latency was established in vitro following coculture of resting CD4⁺ T cells with myeloid dendritic cells. Expression of PD-1 was quantified by flow cytometry, and latency assessed in sorted PD-1^high and PD-1^low/− nonproliferating CD4⁺ memory T cells. The role of PD-1 in the establishment of latency was determined by adding anti-PD-1 (pembrolizumab) to cocultures before and after infection. In addition, a single infusion of anti-PD-1 (nivolumab) was administered to an HIV-infected individual on ART with metastatic melanoma, and cell-associated HIV DNA and RNA, and plasma HIV RNA were quantified.

Results: 

HIV latency was significantly enriched in PD-1^high compared with PD-1^low/- nonproliferating, CD4⁺ memory T cells. Sorting for an additional immune checkpoint molecule, T-cell immunoglobulin domain and mucin domain-3, in combination with PD-1, further enriched for latency. Blocking PD-1 prior to HIV infection, in vitro, resulted in a modest but significant decrease in latently infected cells in all donors (n = 6). The administration of anti-PD-1 to an HIV-infected individual on ART resulted in a significant increase in cell-associated HIV RNA in CD4⁺ T cells, without significant changes in HIV DNA or plasma HIV RNA, consistent with reversal of HIV latency.

Conclusion: 

PD-1 contributes to the establishment and maintenance of HIV latency and should be explored as a target, in combination with other immune checkpoint molecules, to reverse latency.

Introduction

Long-lived latently infected resting memory CD4⁺ T cells persist in HIV-infected individuals on antiretroviral therapy (ART) and are a major barrier to HIV cure ^[1]. Latent infection is enriched in CD4⁺ T cells expressing immune checkpoint molecules, including programmed cell death-1 (PD-1) ^[2–4]. Immune checkpoint molecules regulate the immune response via suppression of T-cell activation ^[5,6], following ligation of the T-cell receptor with a specific antigen peptide-major histocompatibility complex molecule displayed on antigen presenting cells ^[7]. We hypothesized that through this process, immune checkpoint molecule signalling may inhibit productive HIV infection, thereby facilitating the establishment of latent HIV infection.

If the expression of immune checkpoint molecules is important in establishing and/or maintaining latency, the administration of a blocking antibody could potentially reverse latency. Here, we investigate the effects of anti-PD-1 on HIV latency both in vitro, using a coculture model of myeloid dendritic cells (mDC) and resting CD4⁺ T cells ^[8], and in vivo, following administration of anti-PD-1 to a single HIV-infected individual on suppressive ART.

Methods

Isolation of cellular subsets

Peripheral blood mononuclear cells (PBMC) were isolated from HIV-uninfected donors (Australian Red Cross Blood Service, Melbourne, Victoria, Australia). Resting (CD69⁻HLA-DR⁻) CD4⁺ T cells and HLA-DR⁺CD11c⁺CD123⁻ mDC were isolated as previously described ^[8].

Plasmids, virus production and infection

CCR5-tropic, replication and Nef competent enhanced green fluorescent protein (EGFP)-reporter virus, NLAD8-E, was kindly provided by Yasuko Tsunetsugu-Yokota (National Institute of Infectious Diseases, Tokyo, Japan) ^[9]. HIV stocks were generated by FuGene (Promega, Madison, Wisconsin, USA) transfection of 293T cells ^[8]. Cells were infected at 37 °C for 2 h (multiplicity of infection = 0.5, determined by limiting dilution on PBMC ^[10]), and then washed to remove unbound virus.

In-vitro HIV latency model

HIV latency was established in vitro as previously described ^[8]. Briefly, eFluor670-labelled resting CD4⁺ T cells were cultured with syngeneic mDC (mDC : T cell ratio 1 : 10) and infected with NLAD8-E. At day 5 postinfection, productive infection (EGFP⁺ cells) was determined. The eFluor670^highEGFP⁻ T cells were sorted and cultured for 72 h (IL-7 + Raltegravir, kind gift from Merck, Kenilworth, New Jersey, USA), with or without stimulation (anti-CD3/CD28). Postintegration latency was defined as the number of EGFP⁺ cells in stimulated minus unstimulated cultures.

Phenotyping and sorting

PD-1 ligand 1 (PD-L1) and 2 (PD-L2) expression was determined using anti-PD-L1-PE or PD-L2-PE, combined with anti-HLA-DR-perCP (BD, Franklin Lakes, New Jersey, USA). PD-1 expression was determined using anti-PD-1-PE/isotype control (BioLegend, San Diego, California, USA), in combination with anti-CD3-PB, CD45RA-V500 and HLA-DR-APC-Cy7 (BD). At day 5 postinfection, eFluor670^highEGFP⁻CD45RA⁻ T cells were sorted into PD-1^high and PD-1^low/− populations. Some donors were labelled with both anti-PD-1-PE and anti-T cell imunoglobulin domain and mucin domain-3 (Tim-3)-PE-Cy7 (BioLegend), and cells positive for both or either immune checkpoint molecule (PD-1/Tim-3^high), or cells negative for both immune checkpoint molecule (PD-1^low/−Tim-3^low/−) were sorted.

Blocking experiments

Resting CD4⁺ T cells were incubated with anti-PD-1 (pembrolizumab) or isotype control (antirespiratory syncytial virus; IgG4; 10 μg/ml; Merck), at 37 °C for 30 min prior to coculture with mDC. Following infection with NLAD8-E, anti-PD-1 (10 μg/ml) was readded to cell cultures. At day 5 postinfection, IFN-γ levels were quantified by ELISA (MAX Deluxe kit; BioLegend), and pembrolizumab binding saturation determined following staining with anti-CD45RA-V500 (BD) and anti-PD-1-PE (BioLegend).

PCR assays

RNA and DNA were extracted (Allprep Isolation Kit, Qiagen, Hilden, Germany) from total blood CD4⁺ T cells (MACS; Miltenyi; purity >95%). Cell-associated unspliced (CA-US) RNA [lower limit of detection (LLOD) = 1 copy/well], and total HIV DNA was quantified as previously described ^[11]. Plasma HIV RNA was quantified using both a standard Roche (LLOD = 20 copies/ml) and an ultrasensitive single copy assay (LLOD = 0.13 copies/ml), as previously described ^[11,12].

Flow cytometry

Cell sorting was performed using a FACSAria (BD) or MoFlo Astrios (Beckman Coulter, Brea, Californiaa, USA). Analysis was performed using a FACSCalibur or LSRII (BD), and Weasel software (Walter and Elisa Hall Institute, Melbourne, Victoria, Australia).

Statistical analysis

Statistical analyses were performed using Prism, version 6 (GraphPad Software Inc., La Jolla, California, USA). A paired t test was used when n was less than 6; a Wilcoxon signed-rank test was used when n was at least 6 ^[13].

Results

Expression of programmed cell death-1 and its ligands within a dendritic cell-T-cell model of latency

Expression of PD-1 and its ligands PD-L1 and PD-L2 were determined within our previously described mDC-T-cell coculture model of HIV latency ^[8].

PD-1 expression was low on resting CD4⁺ T cells isolated from HIV-neg donors (mean 2.4%, n = 4). However, coculture with mDC for 2–5 days significantly increased PD-1 expression on the eFluor670^highEGFP⁻ (nonproliferating, nonproductively infected) memory CD4⁺ T cells, when compared with T cells cultured alone [Fig. 1a, mean fold change (MFC) = 2.9 at day 5 postinfection, P = 0.01]. Similar results were observed in uninfected and HIV-infected cultures (Fig. 1a). Cell viability was comparable in the presence and absence of mDC (data not shown).

Fig. 1: Role of programmed cell death-1 in an in-vitro model of HIV latency.Resting CD4⁺ T cells, labelled with eFluor670, were cultured alone or with syngeneic myeloid dendritic cells (mDC) for 24 h, infected with an enhanced green fluorescent protein (EGFP)-reporter virus, NLAD8-E, and cultured for a further 5 days. (a) Programmed cell death-1 (PD-1) expression was determined on resting CD4⁺ T cells at baseline, and on eFluor670^highEGFP⁻ CD4⁺ T cells during culture. Points represent the mean of four donors and error bars the SD. (b) PD-1 ligand 1 (PD-L1) and 2 (PD-L2) expression was determined on mDC at baseline, and 1 day postinfection of mDC-T cell cocultures with NLAD8-E (day 1) [n = 6; mean (SD) is shown]. (c) At day 5 postinfection, eFluor670^highEGFP⁻ memory CD4⁺ T cells were sorted into PD-1^high and PD-1^low/− populations; (d) inducible latency was determined in the sorted cells, measured by EGFP expression. Each symbol represents a different donor; MFC, mean fold change. (e–h) Anti-PD-1 (10 μg/ml; pembrolizumab, Merck), or isotype control, was added to resting CD4⁺ T cells prior to coculture with mDC and readded following infection. Each donor is represented by the same symbol; columns represent the median. (e) IFN-γ production was determined in supernatants at day 5 postinfection. (f) To confirm a saturating level of neutralizing antibody, cells treated with isotype or anti-PD-1 were labelled with anti-CD45RA-V500 and anti-PD-1-PE at day 5 postinfection. (g) Productive infection was determined at day 5 postinfection and inducible latency was determined in sorted memory eFluor670^highEGFP⁻ CD4⁺ T cells. (h) The ratio of latent to productive infection is shown. For all panels, P values were determined by paired t test (n < 6) or Wilcoxon signed-rank test (n ≥ 6); ^* P < 0.05, ^** P < 0.01.

PD-L1 and PD-L2 expression was low on freshly isolated mDC (mean 0.4 and 0.2%, respectively), as previously reported ^[14]. However, following coculture with resting T cells, and infection with NLAD8-E, mDC significantly upregulated PD-L1 (Fig. 1b, mean 13.5%, P = 0.01) and PD-L2 (mean 2.2%, P = 0.03). Similar expression levels were observed in uninfected cocultures (data not shown).

HIV latency is enriched in programmed cell death-1 expressing CD4⁺ T cells within a dendritic cell-T-cell model of latency

At day 5 postinfection, eFluor670^highEGFP⁻ memory CD4⁺ T cells were sorted into PD-1^high and PD-1^low/− populations (Fig. 1c). Inducible postintegrated latency was significantly enriched in cells expressing PD-1 (MFC = 6.6, P = 0.007, n = 7) compared with cells with little to no PD-1 (Fig. 1d). These findings are consistent with what has been demonstrated ex vivo using CD4⁺ T cells from HIV-infected individuals on ART ^[3,4] and support the use of this in-vitro model to test the effect of immune checkpoint molecule blockade on HIV latency.

Blocking programmed cell death-1 modestly inhibits HIV latency establishment in nonproliferating CD4⁺ T cells

We next determined whether blocking PD-1 could inhibit the establishment of latency within our DC-T-cell model. Anti-PD-1 (pembrolizumab), approved for the treatment of advanced melanoma ^[15] and other malignancies, was added to resting CD4⁺ T cells prior to coculture with mDC, and again following infection with NLAD8-E, to block interactions between PD-1 and its ligands.

At 10 μg/ml, anti-PD-1 was both functional, inducing a significant increase in supernatant IFN-γ (Fig. 1e), and saturating, preventing binding of anti-PD-1-PE (Fig. 1f).

In the presence of anti-PD-1, we observed a modest, but statistically significant, decrease in latent infection (number of EGFP⁺ cells following anti-CD3/28 stimulation of sorted eFluor670^highEGFP⁻ cells) in all donors compared with isotype control (MFC = 1.6, P = 0.01, n = 6; Fig. 1g). There was no significant difference in productive infection (Fig. 1g). This was unexpected and might be explained by death of the productively infected cells, or possibly, a small increase may have been masked by the higher absolute numbers of EGFP⁺ cells in the productively infected population. In addition, we observed a 1.5 mean fold decrease in the ratio of latent to productive infection (n = 6; Fig. 1h), although this did not reach significance (P = 0.1).

Together, these data suggest that although HIV is enriched in PD-1^high cells, blocking PD-1 alone was not sufficient to completely prevent the establishment of latency. Other immune checkpoint molecules may also contribute to the establishment of mDC-induced latency ^[4]. Indeed, sorting for the immune checkpoint molecule, Tim-3, in addition to PD-1 [including both single (majority) and double positive cells], within our DC-T-cell latency model, resulted in a further enrichment of latency (MFC = 15, P = 0.02, n = 6; Supplementary Fig. 1, https://links.lww.com/QAD/B276), as compared with experiments in which cells were sorted for PD-1 alone (MFC = 6.6; Fig. 1d). However, further experiments are required to determine whether blocking other immune checkpoint molecules, in combination with PD-1, will prevent the establishment of HIV latency.

Administration of nivolumab to an HIV-infected individual on suppressive antiretroviral therapy leads to an increase in cell-associated unspliced HIV RNA

In a previous case report of an HIV-infected individual on ART, we demonstrated a 20-fold increase in CA-US HIV RNA in CD4⁺ T cells following multiple infusions of anti-cytotoxic T lymphocyte-associated protein 4 (CTLA-4) (ipilimumab), consistent with reversal of HIV latency ^[16]. This same individual received a single intravenous infusion of anti-PD-1 (nivolumab, Bristol-Myers Squibb, New York, New York, USA; 3 mg/kg; Fig. 2a) for the treatment of metastatic melanoma. Seven days postadministration of nivolumab, we observed a significant increase in CA-US HIV RNA (fold change = 24.3; P = 0.0001; Fig. 2b) and an increase in the ratio of CA-US HIV RNA : HIV DNA (fold change = 7.9; Fig. 2c), with no significant change in plasma HIV RNA (Fig. 2d), as compared with prenivolumab levels. The individual died shortly after, due to complications of metastatic melanoma. Although only a single case report, these data demonstrate a significant change in CA-US HIV RNA, consistent with latency reversal, following the administration of nivolumab in vivo.

Fig. 2: Increase in cell-associated HIV RNA following anti-programmed cell death-1 in an HIV-infected individual on suppressive antiretroviral therapy.(a) A single intravenous administration of anti-programmed cell death-1 (nivolumab, BMS-936558, 3 mg/kg; bold arrow) was given to an HIV-infected individual on suppressive antiretroviral therapy, following multiple infusions of anti-cytotoxic T lymphocyte-associated protein 4 (CTLA-4) (ipilimumab). (b) Prior to, and following, administration of nivolumab, cell-associated unspliced HIV RNA and HIV DNA were quantified in sorted CD4⁺ T cells, and (c) the ratio of cell-associated unspliced (CA-US) HIV RNA to HIV DNA calculated. CA-US HIV RNA and HIV DNA assays were performed in quadruplicate and triplicate, respectively, the mean ± SEM is shown. (d) Plasma HIV RNA levels were determined using a standard commercial assay (Roche; lower limit of detection = 20 copies/ml) and an ultrasensitive single-copy assay (lower limit of detection = 0.13 copies/ml). HIV RNA measured by single-copy assay (SCA) was quantified in triplicate and the mean is shown.

Discussion

We have provided evidence that PD-1 contributes to the establishment and maintenance of HIV latency, and that antibodies to immune checkpoint molecules, including anti-PD-1, should be further explored as a strategy to disrupt or reverse HIV latency.

We have developed a physiologically relevant in-vitro model of DC-induced primary T-cell latency ^[8], that will enable us to now screen multiple immune checkpoint molecule blockers for their ability to disrupt HIV latency, alone and in combination. Within this model, latency was enriched in PD-1^high CD4⁺ T cells. PD-1 blockade prior to infection reduced, but did not eliminate, the establishment of latency. The persistence of other immune checkpoint molecules, which are also enriched for latent HIV, as demonstrated in CD4⁺ T cells from HIV-infected individuals on ART ^[4], may explain why anti-PD-1 alone did not abrogate latency in our model. Coexpression of these immune checkpoint molecules is minimal ^[17,18]; therefore, it is likely that a combination immune checkpoint molecule blockade approach will be necessary. Given the successful clinical trials with concurrent administration of anti-CTLA-4 and anti-PD-1 in individuals with malignancy ^[19,20], this is potentially feasible, although toxicity remains a significant concern with this combination ^[21]. Next-generation engineered immune checkpoint molecule blockers, such as afucosylated anti-CTLA-4 (BMS-986218), may potentially have lower toxicities (ClinicalTrials.gov identifier NCT03110107).

Administration of immune checkpoint molecule blockers in HIV-infected individuals on ART may not only reverse latent infection, but could also have the added benefit of enhanced immune clearance of latently infected cells (recently reviewed in ^[22]). Blocking interactions between PD-1 and PD-L1, as well as CTLA-4 and its ligands (CD80 and CD86), has been shown to increase HIV-specific CD4⁺ and CD8⁺ T cell function both ex vivo^[23] and in vivo in simian immunodeficiency virus-infected macaques ^[24]. In a recent case report, the administration of multiple repeat doses of anti-PD-1 (nivolumab) to an HIV-infected individual on ART was well tolerated and resulted in a dramatic increase in HIV-specific CD8⁺ T cells, a modest and delayed increase in plasma HIV RNA and a 3 log decrease in HIV DNA ^[25]. The mechanism of how HIV DNA decreased in this case was unclear, but could potentially be due to enhanced clearance of infected cells or a consequence of T-cell proliferation, which would dilute unintegrated HIV DNA that is also included in a total HIV DNA assay.

Increases in HIV-specific CD8⁺ T cells have also been observed in two of six otherwise healthy HIV-infected individuals following a single dose of anti-PD-L1 (BMS-936559). In this study, however, there was no effect on plasma HIV RNA or CA HIV RNA or DNA ^[26]. The different virological findings between this clinical trial and our case report may have several explanations. First, a very low dose of 0.3 mg/kg anti-PD-L1 was given, although in our case report, the full dose of 3 mg/kg of nivolumab was administered. Second, following the administration of anti-PD-L1, ligation of PD-1 is still possible through PD-L2. PD-L2 has multiple functions in addition to binding to PD-1 and, in some settings, can enhance rather than suppress T-cell function ^[22,27]. Surprisingly, we were unable to find any publications on the relative expression of PD-L1 and PD-L2 in HIV-infected individuals on ART. Third, the level of PD-1 expression and signalling may potentially differ in the setting of disseminated malignancy compared with otherwise healthy HIV-infected individuals on ART. Fourth, the activity of anti-PD-1 (or anti-PD-L1) on latency reversal could potentially vary significantly in different recipients, as well described in responses to both antibodies for the management of malignancy ^[28]. Finally, it is unclear whether the degree of latency reversal observed here was in any way related to the prior administration of ipilimumab ^[16]. In clinical trials of individuals with metastatic melanoma, a clinical response to ipilimumab may predict a subsequent response to nivolumab ^[29].

Here, we report that anti-PD-1 led to an immediate, marked increase in CA-US HIV RNA, with no observable change in HIV DNA, in an HIV-infected individual on ART followed for 7 days after a single dose of nivolumab. Together, these data suggest that anti-PD-1 can have a direct effect on latently infected CD4⁺ T cells and should be further explored as a clinical intervention in HIV-infected individuals on ART.

Acknowledgements

We thank the patient without whose generosity and commitment to research the study could not have occurred. We thank Yasuko Tsunetsugu-Yokota (National Institute of Infectious Diseases, Tokyo, Japan) for providing us with the EGFP-reporter virus. We acknowledge Merck for the kind gift of pembrolizumab for evaluation in vitro, and Bristol Myers Squibb for provision of nivolumab under their Special Access Scheme. We thank G. Paukovics, J. Le Masurier and P. Donaldson (Alfred Medical Research and Educational Precinct (AMREP) Flow Cytometry Core Facility, Burnet Institute, Melbourne, Australia), and Tina Luke, Catherine Li, Angela Hind and Daniel Blashki (ImmunoID Flow Cytometry Facility, University of Melbourne) for flow cytometric cell sorting. We thank Ashish Nair (University of Melbourne) for his help processing the human buffy coats and Nicholas Urriola (Department of Immunology, Royal Prince Alfred Hospital) for sample collection from the study participant.

ORCID orcid.org/0000-0002-0330-8241.

The current work was presented, in part, at the 20th International AIDS Conference (AIDS2014), Towards an HIV Cure Symposium, Melbourne, Australia, 20–25 July 2014 (abstract OA42) and at the International AIDS Society (IAS) HIV Cure and Cancer Forum, Paris, France, 22–23 July 2017.

Conflicts of interest and funding

Anti-PD-1 antibodies were kindly supplied by Merck (MK-3475/pembrolizumab; in-vitro studies) and Bristol Myers Squibb (nivolumab; supplied under their Special Access Scheme for the in-vivo study). This work was supported by funds from the American Foundation for AIDS Research, amfAR (grant numbers 108237-51-RGRL and 109226-58-RGRL); the National Health and Medical Research Council (NHMRC) of Australia (grant number APP1041795); and the National Institutes of Health Delaney AIDS Research Enterprise (DARE) to find a cure collaboratory (grant numbers U19 AI096109 and UM1AI126611). S.R.L. is an NHMRC Practitioner Fellow. The authors have no conflicting financial interests.