Introduction Combination antiretroviral therapy (cART) has dramatically improved the life expectancy and health of patients infected with HIV. In the setting of controlled clinical trials with optimal cART, up to 90% of treatment-naïve patients can achieve undetectable virus in plasma and normalization of CD4 T-cell levels [1], [2]. However, when cART is interrupted in patients who initiated therapy during the chronic phase of infection, virus replication resumes in virtually all patients [3]–[5], indicating that current cART is not sufficient to cure HIV infection. The failure of cART to cure HIV infection is due, in part, to the ability of HIV to establish latency in a subset of infected CD4 T cells [6]. The state of latency is characterized by the presence of integrated but transcriptionally silent proviral HIV DNA, which makes the infected cells invisible to the immune system and resistant to both innate antiviral defenses and antiretroviral therapy [6], [7]. Although latent proviral DNA has been detected in multiple different immune cell subsets permissive to HIV infection, long-lived resting memory CD4 T cells are believed to represent the predominant reservoir of proviruses that can be activated to produce infectious virions [8], [9]. Initial quantification of latent HIV proviruses in peripheral blood lymphocytes from patients on cART revealed approximately 200 copies per 106 resting CD4 T cells; however, in general, less than 1% of these proviruses was shown to produce infectious HIV after T-cell mitogenic stimulation with substantial inter-patient variation observed in the fraction of total proviruses that could be activated [10]. The pool of latently infected memory CD4 T cells is believed to be maintained throughout a patient's life by homeostatic proliferation of memory T cells and/or intermittent antigen-driven clonal expansion [11]. Alternatively, low levels of HIV replication confined to lymphatic tissues and undetectable in the periphery may also contribute to the maintenance of the latent virus reservoir [11], [12]. The decay rate of latent virus reservoirs in peripheral blood lymphocytes has been estimated to have a half-life of >3 years, indicating that even life-long cART is unlikely to cure HIV infection [7]. Chronic HIV infection, even when suppressed by cART, poses long-term health risks that include accelerated cardiovascular disease, liver and renal disease, non-AIDS-associated cancers, neurocognitive impairment, and accelerated senescence of immune responses [13]–[15]. Thus, there is a clear unmet medical need for novel therapeutic interventions that could lead either to host-mediated control of HIV in the absence of cART or complete clearance of viral reservoirs. Such virus eradication interventions will need to be well tolerated with minimal side effects. Specifically, therapeutic interventions need to be found that do not cause global T cell activation or a chronic state of inflammation. These interventions should also exhibit minimal drug-drug interactions with medications frequently administered to HIV-infected patients. Identifying a safe modality to activate latent HIV in memory CD4 T cells is an important goal and potentially represents the first key step towards a cure for HIV. Establishment and maintenance of HIV latency is a complex process that appears to involve multiple mechanisms restricting productive viral transcription. These mechanisms include promoter occlusion via steric hindrance, insufficient levels of cellular transcription factors, modification of the HIV 5′ long-terminal repeat (LTR) by methylation and alteration of the chromatin environment in the vicinity of the LTR by histone deacetylation and other epigenetic modifications [16], [17]. Histone deacetylases (HDACs) have been implicated in maintaining HIV in a latent state. In this process, HDACs are recruited to the LTR by various transcriptional regulators and deacetylate lysine residues on histones, inducing chromatin condensation, thereby repressing proviral transcription [18], [19]. Consistent with this mechanism, HDAC inhibitors (HDACi) have been reported to activate latent HIV in cell lines and primary cells, including CD4 T cells from HIV-infected patients on cART [20], [21]. Vorinostat (suberoylanilide hydroxamic acid; VOR) is an HDACi approved for clinical use to treat cutaneous T-cell lymphomas and has also been shown to activate HIV transcription in various latency models [22]–[24]. Administration of a single dose of VOR to eight HIV-infected patients on cART increased HIV RNA levels in resting CD4 T-cells by a mean of 4.8-fold [25]. Although it is unclear whether a single dose of VOR diminished the activatable latent HIV reservoir, these results represent an important milestone by demonstrating that HIV expression can be increased pharmacologically in HIV-infected patients on cART. In a recent study exploring a panel of HDACi for HIV activation using an in vitro latency assay, panobinostat (LBH589; PNB) displayed superior potency to multiple other HDACi tested including givinostat, belinostat, and VOR [26]. Notably, a potent HDACi romidepsin (RMD; Istodax) was not tested in this study. RMD is a cyclic peptide naturally synthesized by Chromobacterium violaceum [27] that has received regulatory approval for the treatment of patients with peripheral T-cell lymphomas (PTCL) or cutaneous T-cell lymphoma (CTCL) [28]. The present study explores the ability of RMD, in comparison with VOR and other HDACi currently in clinical development, to reverse HIV latency in vitro in primary T cells infected with a reporter virus as well as ex vivo in resting and memory CD4 T cells from HIV-infected patients on suppressive cART. Results Romidepsin is a potent activator of HIV in an in vitro model of latency To assess clinically tested HDACi for their ability to activate HIV from latency, we first employed a previously described in vitro HIV latency model [29], [30] with several modifications to increase the sensitivity of detection. The assay involves freshly isolated naïve CD4 T cells from healthy donors polarized to a Th0 phenotype, which mimics memory CD4 T cells. These Th0 cells are then infected with HIV expressing a luciferase reporter gene and cultured for an additional 7 to 10 days until a latent infection is established. The HDACi that we tested in this in vitro HIV latency assay included RMD, VOR, PNB, givinostat, mocetinostat, and pracinostat (SB939). All compounds showed dose-dependent activity in the assay, but displayed varying levels of potency in activating HIV expression (Fig. 1A). Based on experiments conducted with cells prepared from three independent donors, RMD was the most potent HDACi with a mean EC50 value of 4.5 nM (Table 1). Global T-cell activators such as anti-CD3/CD28 antibodies and PMA+ ionomycin consistently showed 2- to 4-fold higher maximum induction of the luciferase signal relative to RMD (data not shown). When cytotoxicity was determined under the identical conditions, RMD displayed a CC50 (50% cell viability reduction) value of 100 nM, resulting in an approximately 20-fold selectivity window (Table 1). PNB was the second most potent compound tested with an EC50 value of 10 nM and a relatively high selectivity window of >250-fold. VOR was substantially less potent in this assay with EC50 and CC50 values of 4 µM and >25 µM, respectively. 10.1371/journal.ppat.1004071.g001 Figure 1 In vitro activation of HIV expression by HDAC inhibitors in an in vitro latency model. Primary CD4 T cells latently infected in vitro with reporter HIV were established as previously described [29], [30] with additional minor modifications described in Materials and Methods. The infected cells were incubated in the presence of the indicated HDACi. (A) A dose response of HIV activation by HDACi was determined by the quantification of luciferase reporter activity after a 48-hour treatment. Results are mean ± SD from a representative experiment performed in quadruplicate. (B) Induction of p24 expression by RMD and VOR. Flow cytometry analysis of cells from a representative donor is shown with gating on the live cell population. Anti-CD3/CD28 antibodies conjugated to beads were used as a positive control. (C) Time course of the induction of p24 expression by RMD. Cells isolated from 2 independent donors were treated with 40 nM RMD or anti-CD3/CD28 antibodies for 24 to 72 hours in the presence of antiretrovirals. Percentage of p24-positive cells was determined by flow cytometry with gating on live cell population. 10.1371/journal.ppat.1004071.t001 Table 1 Activity and cytotoxicity of HDAC inhibitors in the in vitro HIV latency model. EC50 (nM) CC50 (nM) RMD 4.5±1.0 107±126 VOR 3,950±1,900 >25,000 PNB 10.1±1.0 >2,500 Givinostat 95.8±20.3 24,000±5,020 Mocetinostat 13,600±4,664 10,100±106 SB939 212±14 >50,000 Latently infected primary CD4 T cells prepared from three independent healthy donors were treated with the indicated compounds continuously for 48 hours. EC50 and CC50 values were determined from each dose-dependent response using a multi-parameter algorithm. Data are mean ± SD from three independent experiments performed in quadruplicate. To confirm that the HDACi were activating HIV expression in this latency model, we used flow cytometry analysis to quantify intracellular p24 antigen levels following treatment of latently infected CD4 T-cell cultures with RMD, VOR, and the positive control anti-CD3/CD28 antibodies (Fig. 1B). Treatment with 5 and 80 nM RMD, representing the minimal and maximal concentrations leading to the activation of HIV expression in this model, resulted in 3.3% and 5.5% of cells expressing p24 antigen, respectively. In comparison, treatment with 3.0 µM VOR induced p24 antigen expression in approximately 4.4% of cells. As expected, anti-CD3/CD28 antibodies induced p24 antigen expression in 2- to 3-fold higher fractions of CD4 T cells than RMD (Fig. 1B). While the p24 expression in anti-CD3/CD28-treated cells reached a plateau at 48 hours, the fraction of p24 positive cells in RMD-treated cultures continued to increase for 72 hours (Fig. 1C). These data indicate that RMD induces the expression of HIV proteins following the activation of latent provirus in vitro. HIV activation by HDACi correlates with the inhibition of HDAC isoenzymes Since the specific mechanism of HIV latency reversal by HDACi remains to be fully understood, we investigated whether the relative potency of selected compounds in the HIV latency assay correlates with their ability to directly inhibit the activity of individual HDAC isoenzymes. RMD, VOR, and PNB were tested against 11 individual HDAC isoenzymes (HDAC-1 to -11) from four distinct classes: 1, 2a, 2b, and 4. Overall, RMD was the most potent inhibitor, especially against class 1 HDACs (HDAC-1 to -3) as well as HDAC-10 and HDAC-11 isoenzymes (Table 2). PNB showed lower potency, particularly against class 1 and 4 enzymes with up to 40-fold higher IC50 values relative to RMD. VOR was a substantially weaker inhibitor of the majority of HDAC enzymes than RMD with the exception of HDAC-6. The greatest difference in potency between RMD and VOR was observed with the class 1 HDAC enzymes that were 60- to 2,500-fold more susceptible to RMD than VOR (Table 2). The greater activity of RMD against a broad range of HDAC enzymes correlated with its higher potency to activate HIV expression relative to VOR (∼1,000-fold difference; compare Tables 1 and 2). 10.1371/journal.ppat.1004071.t002 Table 2 Effect of RMD, VOR, and PNB on the activity of individual human HDAC isoenzymes. IC50 (nM) HDAC Class Enzyme RMD VOR PNB 1 HDAC-1 0.102 148 4.35 HDAC-2 0.376 418 5.28 HDAC-3 0.211 509 6.04 HDAC-8 24.6 1,700 31.9 2a HDAC-4 314 27,000 442 HDAC-5 1,160 20,800 336 HDAC-7 3,405 101,550 10,825 HDAC-9 8,910 31,650 4,805 2b HDAC-6 488 27.1 5.74 HDAC-10 0.133 527 2.54 4 HDAC-11 0.407 504 9.56 Recombinant human HDAC isoenzymes were incubated with the indicated HDACi and fluorogenic peptide substrates as described in Materials and Methods. IC50 values were calculated from 10-point dose-response curves using a multi-parameter curve fit. Data represent mean values from two independent experiments. RMD activates latent HIV ex vivo in CD4 T cells from virally suppressed patients Previous studies revealed that VOR activates HIV transcription ex vivo in resting CD4 T cells isolated from HIV-infected patients on cART [21]. We used a similar approach to compare RMD to VOR in total memory as well as resting CD4 T cells isolated from patients chronically infected with HIV who were treated with cART and maintained their plasma HIV RNA at 80% cell death during the same incubation period, but did not lead to release of detectable levels of HIV RNA into cell culture supernatants (data not shown). 10.1371/journal.ppat.1004071.g003 Figure 3 Induction of extracellular viral RNA release from CD4 T cells treated with RMD and VOR. Memory or resting CD4 T cells isolated from HIV-infected patients on suppressive cART were treated with RMD or VOR, and viral RNA was quantified in cell culture supernatants 6 days after the addition of drugs. Results are depicted as fold increase in viral RNA relative to control cultures. Each symbol represents one HIV subject. Solid circles, p 0.05 compared to vehicle-treated controls from the same donors; solid squares, p value not calculated. Red lines represent the mean fold HIV induction across all analyzed donors. Symbols # and * denote a statistically significant difference (p 12 months), CD4 counts (>350 cells/µL), and absence of co-infection with hepatitis B or C virus (Supplementary Table S1). Clinical laboratory results were reconfirmed 2 weeks before leukapheresis or blood draw. Leukapheresis was conducted for 3–4 hours, and samples were processed within 2 hours after collection. The leukapheresis product was diluted 1∶1 with PBS and layered over Ficoll for isolation of PBMCs. PBMCs were treated with red blood cell lysis buffer (eBioscience) and rested overnight (10 million cells/ml) in tissue culture medium (RPMI 1640 supplemented with 10% FBS and PenStrep) before the isolation of memory CD4 T cells according to the manufacturer's recommendation (EasySep Human Memory CD4 T cell Enrichment Kit). Resting CD4 T cells were isolated by first enriching total CD4 T cells (StemCell Technologies) and subsequently depleting HLA-DR-, CD25-, and CD69-positive cells via negative selection (Miltenyi Biotec, Auburn, CA). Flow cytometry was used to assess the purity of both T-cell subsets (>98%). Resting CD4 T cells for the analysis of continuous RMD exposure and longitudinal responses to RMD in the same donors were purified from fresh blood using the same protocol. To assess HIV activation by HDACi, up to 5 million CD4 T cells were plated in 24-well plates in 2.5 ml of media, supplemented with antiretrovirals (ARVs) (100 nM elvitegravir and 100–300 nM efavirenz) for the entire duration of culture incubation. HDACi were added at specified concentrations on day 0 and cells or culture supernatants were harvested for the analysis of HIV RNA at the indicated time points (6–48 hours or 6–7 days, respectively). For HDACi pulse treatment, cultures were washed twice using ARV-containing media at 4 or 24 hours after addition of HDACi and incubated with ARVs until harvest. To measure HIV RNA levels, 1 ml of culture supernatant was analyzed by a robotic COBAS AmpliPrep/TaqMan system (Roche Diagnostics, Indianapolis, IN), which extracts total nucleic acid and quantifies HIV RNA in copies per milliliter using the HIV-1 Test, v2.0 kit (Roche Diagnostics). For the measurement of cell-associated HIV RNA levels, cells were washed with PBS, counted, and lysed (Qiagen RLT buffer, 400 µl for every 2 million cells; Qiagen, Venlo, Netherlands). Lysates were filtered through a Qiashredder (Qiagen). Total RNA was prepared from the lysates (400 µl) using a robotic system (QIAsymphony, Qiagen) that incorporates a DNase I digestion step to eliminate cellular DNA. The resulting total RNA was eluted with 200 µl of buffer, diluted to 1 ml with nuclease-free water, and analyzed by COBAS using the HIV-1 Test v2.0 kit. Preparations of both cell-associated RNA and supernatant total nucleic acids were tested for potential contamination with HIV DNA and/or host DNA by performing the PCR amplification in the presence and absence of reverse transcriptase and by a detection of GAPDH-encoding host DNA sequence. These methods confirmed that there was no contaminating HIV DNA in either the intracellular RNA or supernatant total nucleic acid preparations (Supplementary Figs. S3 and S4). Cell activation analysis by flow cytometry For cell activation analysis, either complete PBMC cultures or resting CD4+ T cells isolated from HIV-infected cART-suppressed patients were incubated with the tested agents under specified conditions and were stained with relevant antibodies. All antibodies were purchased from BD Biosciences and included Alexa Fluor 700-labeled antibody to CD4, APC-H7-labeled antibody to CD8, PE-Cy7 -labeled antibody to CD19, PE-labeled antibody to CD69, APC-labeled antibody to CD25, and V450-labeled antibody to HLA-DR. Live cells were gated by forward and side scatter and exclusion of the dead cells by Live/Dead Fixable Aqua Stain (Invitrogen). Marker staining was assessed by flow cytometry analysis on a LSR Fortessa with data processing using FlowJo software. Cell-associated HDAC enzymatic assay Resting CD4 T cells obtained from HIV-infected patients on suppressive cART were incubated with serial dilutions of HDACi in 96-well plates at 10,000 cells/well for indicated time and then lysed by repeated freezing and thawing. Total cellular HDAC activity was measured using the HDAC-Glo I/II assay kit (Promega) according to the manufacturer's protocol. Phylogenetic analysis of HIV DNA and RNA during ex vivo latency reversal Single-genome sequencing of a portion of HIV-1 gag-pro-pol amplified from cell culture supernatants or cellular DNA extracts was performed as previously described [34]–[36]. Sequences were aligned using ClustalW and neighbor-joining phylogenetic analysis was performed using MEGA5. Trees were rooted on the subtype B consensus sequence (www.hiv.lanl.gov). Experimental and statistical analyses HIV RNA-induction experiments were conducted using 5 to 6 replicates for every experimental condition, including vehicle-treated controls. Fold induction in HIV RNA levels was calculated as a ratio of mean signals from compound-treated and vehicle (DMSO)-treated control samples from the same donor under identical conditions. Student's t-test (one-tailed distribution, two-sample equal variance) was used to assess statistical significance of the difference between the detected HIV RNA copies in vehicle control- and HDACi-treated samples or between RMD- and VOR-treated samples. Time course experiments testing the latency reversal agents had matching vehicle-treated controls for each time point. For experiments using blood-derived resting CD4 T cells, culture supernatants from 3 replicate wells were combined and analyzed using the COBAS system. Primary results in absolute HIV RNA copy numbers from all experiments presented in this study are provided in Supplementary material (Supplementary Table S4). Supporting Information Figure S1 Effect of RMD and VOR on the viability of primary memory CD4 cells. Memory CD4 cells were isolated from HIV-infected patients and treated with RMD or VOR for 4 or 24 hours, respectively. Cell viability was determined using a Cell TiterGlo reagent 6 days after the initiation of treatment and is expressed as a percentage of cell viability relative to control vehicle-treated memory CD4 cells from the same donors. The data represent mean +/− S.D. from three independent donors. (TIF) Click here for additional data file. Figure S2 Effects of RMD and VOR on the expression of activation markers in primary resting CD4+ T cells, compared to the effect of PMA plus ionomycin stimulation. Cells were treated with a 4-hour pulse of the indicated concentrations of RMD or continuously with either VOR or PMA+ ionomycin (positive control), followed by staining 48 hours later. The surface expression of CD25 and CD69 in viable CD4 cells was analyzed by flow cytometry. Data are mean ± SD of two independent experiments performed with cells isolated from two HIV-infected patients on suppressive cART. (TIF) Click here for additional data file. Figure S3 Lack of HIV DNA contamination in extracted intracellular RNA samples following the treatment with DNase I. (A) Two million memory CD4 T cells isolated from three HIV-infected cART-suppressed patients (Donors A–C) were treated with control (blank, bk) or romidepsin (RMD) for 48 hours, washed, lysed, and filtered through a Qiagen shredder to obtain homogenized cell lysates before additional analyses. Cell lysates were extracted using QIAsymphony, with or without DNase I digestion, before the entire sample was analyzed by COBAS for the quantification of HIV viral sequences. (B) Cells from identical donors were lysed, shredded, and then extracted for total RNA using QIAsymphony with DNase I digestion. Samples aliquots were analyzed by qPCR for HIV Gag and GAPDH sequences, with or without addition of reverse transcriptase (RT+ or RT−). Asterisks (*) indicate none detected. (C) Random lysates of vehicle-treated memory CD4 T cells from virally suppressed HIV patients (#1–8) were divided into identical duplicates and extracted for total RNA using QIAsymphony with DNase I digestion. The total RNA was then treated with additional DNase I digestion or not (yes vs. no) before quantification of HIV viral sequences by COBAS. (TIF) Click here for additional data file. Figure S4 Lack of HIV DNA contamination in total nucleic acid extracts from cell culture supernatants. Memory CD4 cells isolated from four HIV-infected cART-suppressed patients (Donors A–D) were treated with no drug control (blank; bk), 5 nM romidepsin (RMD) or PMA+ ionomycin (P/I) for 6 days. Cell culture supernatants were extracted for total nucleic acid (tNA) using COBAS TNAI kit before additional analyses. (A) HIV Gag DNA and host GAPDH DNA were quantified in tNA by qPCR without reverse transcriptase. Asterisks (*) indicate none detected. (B) The same tNA samples were further incubated with or without DNase I (yes vs. no), re-extracted for tNA, and analyzed for HIV copies by COBAS HIV viral load analyzer. Hash marks (#) indicate the limit of HIV quantification (<20 copies/ml). (TIF) Click here for additional data file. Table S1 Demographic characteristics of HIV-infected patients participating in the study. (XLS) Click here for additional data file. Table S2 HIV RNA released from resting CD4 T cells treated with RMD can be pelleted by high-speed centrifugation. a Percentage of total nucleic acid in the sample. Resting CD4 T cells isolated from an HIV-infected patient on suppressive cART were treated with RMD for 6 days and the collected supernatants were subjected to ultracentrifugation (21,000 g×60 min). HIV DNA and RNA were quantified in pellet and supernatant using Taqman quantitative PCR. (DOCX) Click here for additional data file. Table S3 Systemic clinical exposures of RMD and VOR compared to concentrations used in the ex vivo experiments. a Istodax (romidepsin) prescribing information (www.istodax.com). b Zolinza (vorinostat) prescribing information www.zolinza.com/vorinostat/zolinza).c Determined by an equilibrium dialysis followed by HPLC/mass spectrometry analysis. d Ratio of free drug concentration in cell culture media and free drug concentration in serum of clinically treated patients. (DOCX) Click here for additional data file. Table S4 Summary of datasets from analyses of HIV RNA induction in the ex vivo primary CD4 T cell cultures isolated from virologically suppressed HIV-infected patients. The table shows compiled primary data and statistical analyses from the quantitation of HIV RNA (copies/million cells for intracellular HIV RNA; copies/mL for supernatant HIV RNA) in various types of CD4 T cell cultures isolated from HIV-infected patients and treated with tested HDACi or vehicle control. The datasets represent results displayed in Figures 2, 3, 4, 5, and 7. (XLS) Click here for additional data file.