Measuring reversal of HIV-1 latency ex vivo using cells from infected individuals

HIV-1 can persist in a latent state in the form of integrated and transcriptionally silent proviruses, which are unaffected by antiretroviral therapy (ART) or host immune responses (1–3). This latent reservoir of HIV-1, found primarily in resting memory CD4⁺ T cells, has an extremely long half-life, making eradication impossible by ART alone (2, 4, 5). To cure HIV-1–infected individuals, it will be necessary to eliminate this reservoir or render it incapable of rekindling viremia after the cessation of ART. The “kick and kill” strategy has been proposed to eliminate the latent reservoir (6, 7). In this approach, latent HIV-1 would be reactivated by small-molecule latency reversing agents (LRAs). Cells harboring reactivated proviruses would then presumably die from viral cytopathic effects or be eliminated by cytolytic T lymphocytes. Global T-cell activation effectively reactivates latent HIV-1, but the associated systemic toxicity makes it unsuitable for clinical use. Candidate LRAs have been identified using in vitro models of latency, and clinical trials were subsequently initiated with some of these candidate compounds (6, 8–10), including the histone deacetylase inhibitor (HDACi) suberoylanilide hydroxamic acid (SAHA, vorinostat). However, to evaluate the extent of latency reversal achieved by candidate LRAs, it is important to have ex vivo assays using cells from HIV-1–infected patients. The development of such ex vivo assays is essential to validate LRAs identified using in vitro models and inform the design of future clinical trials. In PNAS, Cillo et al. describe a quantitative ex vivo approach that provides a high-resolution measurement of latency reversal efficiency (11). The authors use measurements of cell-associated RNA and of virus released into supernatant combined with a maximum likelihood estimate to determine the fraction of latent proviruses affected ex vivo by T-cell activation or vorinostat treatment. The authors demonstrate that a small minority of latent proviruses (1.5%) are induced to produce virions by T-cell activation, although a slightly larger fraction (7.5%) are induced to produce intracellular unspliced HIV-1 RNA. Furthermore, the authors show that the candidate LRA vorinostat induces virion production or intracellular unspliced HIV-1 RNA from an extremely small minority of latent proviruses (0.12% and 0.14%, respectively).

Ex Vivo Measurement of the Latent Reservoir

The standard method for measurement of the latent reservoir is the viral outgrowth assay, which exponentially amplifies replication-competent viruses released from latently infected cells after T-cell activation (2). It provides a definitive minimal estimate of the frequency of latently infected cells. Although recently simplified (12), this assay is labor intensive, requiring the addition of irradiated peripheral blood mononuclear cells (PBMCs) from healthy donors and at least 10 d of tissue culture, which may not be practical in some research or clinical laboratories. PCR-based assays for proviral DNA are technically easier to perform, but interpretation of proviral DNA measurements is complicated by the presence of a large excess of defective proviruses (Fig. 1A). Cillo et al. demonstrate that the vast majority of proviruses found in resting CD4⁺ T cells (98.5%) fail to produce virions after activation for 7 d with anti-CD3/CD28 costimulation (11) (Fig. 1B, blue circles). This finding is consistent with recent work that directly identified lethal defects in 88.3% of proviruses examined (13).

Fig. 1.

Ex vivo measurement of HIV-1 latency reversal. (A) Patient resting CD4⁺ T cells cultured ex vivo with anti-CD3/CD28 costimulation, vorinostat, or media control. Although the majority of the HIV-1 proviruses are not induced by these stimuli (blue), a small fraction of the proviruses can be transcribed on stimulation, as detected by measurement of cell-associated HIV-1 RNAs (magenta). Even fewer proviruses can produce viral particles as detected by supernatant viral RNA (yellow). (B) Comparison of supernatant viral RNA (yellow), cell-associated RNA (magenta), and total proviral DNA (blue) after stimulation with anti-CD3/CD28, vorinostat, or media control. Percentage represents fractional provirus expression (fPVE) as defined by Cillo et al. (11).

Ex vivo measurements of cell-associated HIV-1 RNA (6, 14) and of HIV-1 RNA released into culture supernatants (15) after stimulation of patient cells have been used to assess latency reversal. In studies of this kind, the quantity of HIV-1 RNA transcripts is typically normalized to cell input number or transcripts of a cellular housekeeping gene. Cillo et al. use a different approach. They quantify the induction of latent HIV-1 using limiting dilution conditions and normalizing induction to the number of input cells carrying HIV-1 proviruses as determined by quantitative PCR. Thus, this approach can help to estimate the proportion of the latent reservoir being induced by a particular stimulus.

Although only 1.5% of proviruses were induced by T-cell activation to release virions into the culture supernatant (Fig. 1B, yellow circles), a slightly larger fraction (7.5%) was induced to produce intracellular unspliced HIV-1 RNA (Fig. 1B, magenta circles). It is unclear from this study why some cells in which HIV-1 transcription was induced failed to produce virus particles. Proviruses with intact proviral promoters but defects in components required for virion assembly or release may provide the explanation (13).

Evaluating the “Kick” to Latent Proviruses by Candidate LRAs

Numerous in vitro models of latency have been developed using transformed T-cell lines or primary CD4⁺ T cells infected in vitro with HIV-1 constructs carrying reporter genes. These models have been used to identify or test candidate LRAs (16). The histone deacetylase inhibitors vorinostat, romidepsin, and panobinostat, along with the alcoholism drug disulfiram, showed promise in these in vitro systems. As such, these compounds have been advanced to clinical trials. Although a single dose of vorinostat induced a small increase in intracellular RNA containing HIV-1 gag sequences (6), sustained increases in intracellular HIV-1 RNA were not seen in the follow-up multidose trial (17). Interestingly, vorinostat did not cause increases in residual viremia (6) or reductions in the size of the reservoir (17). In the clinical trial of disulfiram, neither a reduction in the size of the reservoir nor a significant increase in residual viremia was observed (18). Additional clinical trials of multidose vorinostat, panobinostat, romidepsin, and disulfiram are planned or ongoing.

Although in vitro models suggest that these candidate LRAs are active, the results from clinical trials to date argue that the effects seen in these in vitro models are not predictive of substantial in vivo effects on the latent reservoir. Cillo et al. apply their limiting dilution approach to determine the fraction of latent proviruses in patient resting

Cillo et al. describe a quantitative ex vivo approach that provides a high-resolution measurement of latency reversal efficiency.

CD4⁺ T cells that were induced by ex vivo exposure to vorinostat. Compared with latency reversal caused by T-cell activation, vorinostat failed to perturb latency ex vivo in the majority of latently infected cells from HIV-1–infected patients, as measured by the induction of virion production (1.5% for T-cell activation vs. 0.12% for vorinostat) (11). When latency reversal was measured by the induction of intracellular unspliced HIV-1 RNA, vorinostat again failed to induce most latent proviruses (7.5% for T-cell activation vs. 0.16% for vorinostat, even lower than the 0.21% media control) (11). The conclusion that vorinostat does not effectively reverse latency in cells from infected individuals is consistent with a previous study in which vorinostat also failed to induce the release of free virus from patient cells treated ex vivo (5). Moreover, another recent study comparing the efficacy of numerous candidate LRAs using cells from HIV-1–infected individuals on suppressive ART observed that none of the leading candidate LRAs, including vorinostat, perturbed the latent reservoir when used as single agents (14).

The work of Cillo et al. and others in the field illustrates one of the major challenges facing cure efforts and raises important questions. Growing evidence suggests that the current leading candidate LRAs do not significantly perturb the latent reservoir compared with the benchmark achieved with T-cell activation. The data suggest that more effective latency reversal strategies will likely be required for elimination of the latent reservoir, perhaps involving combinations of LRAs administered to patients on combination antiretroviral therapy. Data gathered from early attempts at HIV-1 cure through allogeneic stem cell transplantation or extremely early initiation of ART suggest that very large reductions in the reservoir must be achieved to realize a significant clinical outcome. Understanding the true size of the HIV-1 latent reservoir, the fraction of proviruses that must be targeted by eradication strategies, and the properties of these proviruses is essential in HIV-1 cure efforts. The assays and data presented here by Cillo et al. will no doubt be important in furthering our understanding of the latent reservoir. Since the first reports of the cure of Timothy Ray Brown, the so-called “Berlin patient” (19), the field has been reinvigorated, and a realistic assessment of the challenges that remain will facilitate the eventual development of a general cure for HIV-1 infection.