|Home | About | Journals | Submit | Contact Us | Français|
Reactivation of latent HIV-1 infection is considered our best therapeutic means to eliminate the latent HIV-1 reservoir. Past therapeutic attempts to systemically trigger HIV-1 reactivation using single drugs were unsuccessful. We thus sought to identify drug combinations consisting of one component that would lower the HIV-1 reactivation threshold and a synergistic activator. With aclacinomycin and dactinomycin, we initially identified two FDA-approved drugs that primed latent HIV-1 infection in T cell lines and in primary T cells for reactivation and facilitated complete reactivation at the population level. This effect was correlated not with the reported primary drug effects but with the cell-differentiating capacity of the drugs. We thus tested other cell-differentiating drugs/compounds such as cytarabine and aphidicolin and found that they also primed latent HIV-1 infection for reactivation. This finding extends the therapeutic promise of N′-N′-hexamethylene-bisacetamide (HMBA), another cell-differentiating agent that has been reported to trigger HIV-1 reactivation, into the group of FDA-approved drugs. To this end, it is also noteworthy that suberoylanilide hydroxamic acid (SAHA), a polar compound that was initially developed as a second-generation cell-differentiating agent using HMBA as a structural template and which is now marketed as the histone deacetylase (HDAC) inhibitor vorinostat, also has been reported to trigger HIV-1 reactivation. Our findings suggest that drugs with primary or secondary cell-differentiating capacity should be revisited as HIV-1-reactivating agents as some could potentially be repositioned as candidate drugs to be included in an induction therapy to trigger HIV-1 reactivation.
The costs of providing lifetime HIV treatment are estimated at $355,000 per patient (61). In the United States, about 1,000,000 people are currently infected with HIV-1. In addition to ending the personal hardship of the affected patients, a cure for HIV-1 would save more than 300 billion dollars in health care costs, in the United States alone. With >30 million people being HIV-1 infected worldwide, the global impact would be much larger, but no data are available to calculate the actual cost savings. In either case, accelerated efforts to identify drugs that could cure HIV-1 infection seem to be readily justifiable for humanitarian and economic reasons.
Previous drug screens for HIV-1-reactivating compounds or previous attempts to therapeutically reactivate latent HIV-1 infection in patients were developed under the “one target-one drug” hypothesis, which is de facto based on the premise that the “perfect” chemical probe is sufficient to act on a single target (a “magic bullet”). This approach is likely successful in scenarios where a single molecular target has been identified. For HIV-1 latency, research is mostly focused on the utilization of histone deacetylase (HDAC) inhibitors used as single drugs in intensification therapies, which to date have not been clinically successful (5, 6, 43, 67–69). This and the seeming inability of HDAC inhibitors such as sodium butyrate, trichostatin A (TSA), or valproic acid to trigger HIV-1 reactivation in some of the advanced in vitro models of HIV-1 latency in primary T cells (see the work of the Siliciano or Planelles laboratory [14, 82]) and some more recently established latently infected T cell lines (23) cast additional doubts on the therapeutic promise of this class of agents as HIV-1-reactivating drugs, when used as single-drug treatments. Eventually, alternative approaches will have to be considered.
At this time, it remains unclear how latent HIV-1 infection is controlled at the molecular level. Current research suggests that establishment and control of latent HIV-1 infection may be a relatively complex phenomenon in which control is achieved at multiple layers and by multiple mechanisms. In such a scenario, a different approach, which would simultaneously target latent infection at several molecular levels, may be indicated.
For many years, the molecular mechanisms that epigenetically regulate cellular genes were thought to control HIV-1 latency. Following integration of the virus, a restrictive histone code is established at the HIV-1 promoter (long terminal repeat [LTR]) and suppresses HIV-1 gene expression (54, 56, 74). DNA methylation of the viral LTR could stabilize latent infection (9, 10). Then, in 2004, Han et al. (29) published that HIV-1 infection events in vivo in CD4+ memory T cells of patients on successful antiretroviral therapy (ART) were mostly found in actively expressed host genes, which should not provide a DNA environment that lends itself to the formation of a restrictive histone code. This finding was confirmed in several in vitro models of latent infection (16, 23, 29, 63, 66). Some latently HIV-1-infected cell lines from the Verdin laboratory also have been shown to hold the integrated, transcriptionally silent virus in actively expressed genes (44). The idea that HIV-1 latency is governed by transcriptional interference was brought forward to explain latent HIV-1 infection in actively transcribed host genes (23, 30, 44). Host gene RNA polymerase II (RNAP II) would read through the integrated viral genome and thereby prevent transcription initiated at the viral promoter by physical exclusion.
On the other hand, the idea that transcriptional interference would govern latent HIV-1 infection is somewhat in conflict with studies that report that paused RNA polymerase II (RNAP II) is found at the latent HIV-1 promoter (37, 38, 86). In addition, the involvement of upstream transcriptional control mechanisms in HIV-1 latency is suggested by a series of other studies. These studies describe recruitment of HDACs to the HIV-1 LTR and the involvement of higher-level regulators of chromatin structures in HIV-1 latency (22, 36, 48, 54, 72, 73, 79).
While some of the results from these publications may be conflicting, the studies suggest that changes in the chromatin structure at the latent HIV-1 LTR and a favorable transcription factor composition are likely contributing to govern HIV-1 latency at several layers of molecular control. As such, it seems unlikely that a single chemical compound will be able to induce a sufficiently complex cellular response that would trigger the required system-wide, complete reactivation of all latent infection events. The sole mechanism for a single molecular stimulus that should trigger efficient systemic HIV-1 reactivation may be the induction of high levels of NF-κB activity. Maximum NF-κB stimulation has been attempted using an anti-CD3 monoclonal antibody (MAb) (OKT3) and interleukin-2 (IL-2), which represent signal 1 and signal 3 of the T cell activation pathway and should activate NF-κB and NFAT. While these attempts used more than one drug, they were not combination treatments in the sense that they would target different level of molecular control at the latent HIV-1 promoter. Neither of these clinical studies resulted in a meaningful decrease in the viral reservoirs (40, 75). The biggest hurdle to overcome for therapeutic implementation of such a strategy is that NF-κB activation is usually associated with the induction of a large array of cellular genes, and NF-κB-stimulating agents are known to harbor a great risk of inducing a cytokine storm, with potentially detrimental consequences for the patient. Therefore, treatment intensity is limited and may simply be insufficient (1, 25). Combining several drugs that tackle latent infection at several regulatory checkpoints to reduce the requirement of an NF-κB activation component may be a more promising way forward. An example for this strategy was reported in an earlier study by Quivy et al. for latent infection in a cell line, where the authors reported synergistic effects of HDAC inhibitors and the phorbol ester 13-phorbol-12-myristate acetate (PMA), which would target the protein kinase C (PKC)/NF-κB pathway (57).
Here, we describe a drug-screening assay designed to identify HIV-1-reactivating drug combinations and report on a class of selected FDA-approved anticancer drugs that prime latent HIV-1 infection for reactivation or directly trigger HIV-1 reactivation as a result of their cell-differentiating capacity (dactinomycin, aclacinomycin, and cytarabine). Importantly, in T cell populations that would not respond with full reactivation to high activator concentrations, pretreatment with differentiating drugs allowed for full reactivation at the population level.
Complete system-wide reactivation will be a prerequisite for viral eradication. We discuss how our findings relate to previous reports that the cell-differentiating polar compound N′-N′-hexamethylene-bisacetamide (HMBA) and the second-generation cell-differentiating drug suberoylanilide hydroxamic acid (SAHA) both trigger HIV-1 reactivation (7, 20, 24, 60). In summary, our data suggest that differentiating drugs or compounds, as a class of therapeutic agents, could become part of a future induction therapy to eradicate the reservoir of latently HIV-1-infected T cells.
All T cell lines and primary T cells were maintained in RPMI 1640 supplemented with 2 mM l-glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin, and 10% heat-inactivated fetal bovine serum (FBS). Primary T cell cultures were further supplemented with 10 U/ml IL-2. The latently infected J89GFP cells, CA5 T cells, and EF7 T cells and the viruses used to generate these cell lines have been described earlier (23, 41). All cell lines were generated using green fluorescent protein (GFP) reporter viruses that express all viral gene products. FBS was obtained from HyClone (Logan, UT) and was tested on a panel of latently infected cells to ensure that the utilized FBS batch did not spontaneously trigger HIV-1 reactivation (32, 41). The phorbol ester 13-phorbol-12-myristate acetate (PMA), N′-N′-hexamethylene-bisacetamide (HMBA), cytarabine, 5-azacytidine, rebeccamycin, aphidicolin, and oxaliplatin were purchased from Sigma, whereas recombinant human tumor necrosis factor alpha (TNF-α) was obtained from R&D. SAHA was purchased from Cayman Chemicals. Daunorubicin, doxorubicin, α-amanitin, ICRF-193, and camptothecin were purchased from Calbiochem. 5,6-Dichloro-beta-d-ribofuranosylbenzimidazole (DRB) was purchased from Alexis Biochemicals. A retroviral murine stem cell virus (MSCV)-DsRedExpress plasmid was used for generation of the red fluorescent protein (RFP)-barcoded J89GFP cell populations.
Peripheral blood mononuclear cells (PBMCs) from buffy coats of healthy donors obtained through the American Red Cross or another commercial vendor were CD8 T cell depleted using anti-CD8 DynaBeads. The remaining cells were stimulated with an anti-CD3/CD28 MAb combination (OKT3/CD28.2). At 7 to 10 days poststimulation, the cells were infected with HIV-1 NL4-3. The cultures were kept in the presence of a reverse transcriptase (RT) inhibitor for 10 days prior to evaluating drug efficacy as the increase in the percentage of p24-positive T cells over background infection. Intracellular stains for HIV-1 Gag p24 were performed using the anti-p24 MAb KC57-RD1 (Beckman Coulter). This model of latent HIV-1 infection in primary T cells had been first described in reference 23.
J2574 reporter T cells were generated by retrovirally transducing Jurkat T cells with an HIV-1 reporter construct (p2574) in which the HIV-1 LTR controls the expression of GFP. The HIV-1 LTR and the GFP gene are separated by a 2,500-bp spacer element. Lentiviral particles were produced by transfecting 293T cells with p2574 and supplying Gag-Pol-Rev-Tat in trans. Vesicular stomatitis virus G protein (VSV-G) was used as a viral envelope protein. Following lentiviral transduction of Jurkat cells, all cells that spontaneously expressed GFP were removed by cell sorting. The GFP-negative population was then activated with PMA to identify all cells that would harbor an inducible LTR-GFP-LTR integration event. Cells that turned GFP positive following stimulation were again selected by cell sorting. GFP expression in this population ceased after a few days, leaving a population of GFP-negative reporter cells. The amount of founder cells for this population is calculated to represent >50,000 individual integration events.
J89GFP or CA5 T cells (10 × 106) were left untreated or treated with 0.004 μg/ml, 0.01 μg/ml, or 1 μg/ml dactinomycin for 18 h or 1 h, respectively. Cells were washed twice with cold PBS and then lysed for 30 min on ice in lysis buffer (0.5% Triton X-100, 20 mM HEPES [pH 7.9], 150 mM NaCl, 20 mM KCl, 2 mM MgCl2, 1 mM dithiothreitol [DTT], 0.2 mM EDTA, and protease inhibitor cocktail [P8340; Sigma]), followed by centrifugation at 14,000 rpm for 10 min. The same amount of protein lysate was fractionated on 5 ml of a 10 to 45% glycerol gradient in lysis buffer in a SW-Ti55 rotor for 16 h at 245,000 × g. Fractions were resolved on 10% SDS-PAGE gels and transferred to a polyvinylidene fluoride (PVDF) membrane. The primary antibodies used for Western blotting were rabbit anti-Cdk9 (sc-484; Santa Cruz Biotechnology) and rabbit anti-HEXIM-1 (ab25388; Abcam), respectively.
Infection levels in the cell cultures were monitored by flow cytometric (FCM) analysis of GFP expression. FCM analysis was performed on a Guava EasyCyte (Guava Technologies, Inc.) or a FACSCalibur or an LSRII (Becton & Dickinson) cell sorter. Cell sorting experiments were performed using a FACSAria flow cytometer (Becton & Dickinson). Data analysis was performed using either CellQuest (Becton & Dickinson) or Guava Express (Guava Technologies, Inc.).
High-throughput screening (HTS) data acquisition was performed (data not shown) using a HyperCyt autosampler combined with a FACSCalibur flow cytometer. The system was adjusted to acquire ~2,000 counts/population in the life gate to ensure sufficiently high cell counts to perform statistically meaningful data analysis. The assay is characterized by a Z′ factor of 0.83 using PMA as an activating agent (data not shown). Maximum achievable HIV-1 reactivation levels for the three populations using 10 ng/ml PMA were 90% ± 3%. Data analysis was performed using the HyperView data analysis software. Determination of hits can be visually performed using a heat map that is programmed to indicate changes in HIV-1 expression levels by a self-defined color code. HyperView-generated data were transferred to Spotfire (Tibco) or Excel (Microsoft) for statistical analysis. Compound plates for drug screening purposes were generated from a parental 80,000-compound library (Chembridge) using a BioTek Precision platform. In addition, an in-house collection of drugs/compounds with known molecular function was utilized.
As approaches to trigger latent HIV-1 infection with single drugs have shown limited promise, we developed a drug screening assay that would allow for the direct identification of drug combinations that act synergistically to trigger HIV-1 reactivation (data not shown). The underlying idea was to identify drugs/compounds that would lower a putative HIV-1 reactivation threshold to allow for otherwise-weak activating agents to trigger efficient HIV-1 reactivation. During a pilot screen, we identified several drugs/compounds that would fulfill these criteria.
As the basis of the drug screening assay, we used the previously reported latently HIV-1-infected J89GFP T cells (41). J89GFP cells are latently infected with a GFP reporter virus. In a latent state, the cells express no GFP; however, following reactivation by stimuli such as anti-CD3/CD28 MAb combinations, TNF-α, or PMA, the cells start to express high levels of GFP as a direct and quantitative marker of HIV-1 expression. With GFP being used as the specific signal for on-target drug effects, we transduced J89GFP cells with a retroviral DsRedExpress (RFP) vector to produce three distinctive J89GFP populations (J89GFP, J89GFP-R, and J89GFP-R+), distinguishable by an RFP-based fluorescent barcode (Fig. 1A). Retroviral transduction was performed using an MSCV LTR-based retroviral vector to express RFP, as MSCV LTR-driven gene expression in Jurkat T cells remains stable in long-term cell culture and does not respond to activation with changes in fluorescence intensity. The latter characteristic is essential to maintain the integrity of the fluorescent barcode following compound addition.
For the compound-screening/drug-repositioning effort described here, we used a compound library holding an extensive selection of drugs/compounds with defined activities. The drug screen was designed to identify modulator compounds that were able to prime latent HIV-1 infection for reactivation by subthreshold concentrations of three predetermined activators (PMA, OKT3, and HIV reactivating factor [HRF] ) in a single 96-well plate. Final compound concentrations were chosen at 5 μM for compounds derived from our small chemical molecule library (Chembridge). Concentrations of drugs varied according to the known effective concentrations of the respective compounds.
For the modulator compound screen, three individual 96-well plates holding 1 × 105 J89GFP, J89GFP-R, and J89GFP-R+ cells/well were prepared. Compounds were loaded into the individual wells, and after 6 h, the individual plates were stimulated with any suboptimal concentration of the respective activators. At 24 h after addition of the compounds, the 3 corresponding individual 96-well plates were combined using a robotic platform and immediately subjected to high-throughput flow cytometric analysis using a HyperCyt high-throughput autosampler which allows for time-resolved data acquisition (Fig. 1B). In this setup, the data for individual samples are collected not as single files but as time-resolved data. Separation of the individual data sets is achieved using specialized analysis software (HyperView data analysis software). In our experimental setup, by combining HTS flow cytometry with a fluorescent barcode (Fig. 1A), a 96-well plate, or 288 drug combinations, can be analyzed in 8 min. The following parameters can be determined in each plate during the primary drug screen: HIV-1 transcription activity for each cell population, compound-induced changes in the baseline GFP expression, cell density of each cell population as characterized by individual fluorescence signatures (antiproliferative compound effects), and overall cell viability as determined by life gate analysis in the forward scatter (FSC)/side scatter (SSC) plot (compound toxicity). The results of a 96-well sample plate are shown in Fig. 1C. One major advantage of flow cytometry-based drug screening in this system is that we can sensitively detect on-target effects despite significant compound toxicities at the utilized compound concentration. As hits are identified by determining the ratio of cells harboring latent (GFP-negative) to active (GFP-positive) infection within one population determined by the RFP barcode, the assay becomes largely independent of the amount of analyzed cells.
As HIV-1 latency does not offer a defined molecular target and the drug screen is based on a change in phenotype, the mechanism of action for each identified hit has to be determined individually. Toward this goal, we prioritized the analysis of FDA-approved drugs over the analysis of compounds with reported or unknown mechanisms of action. In here, we report that aclacinomycin (Fig. 2A, AcM) and dactinomycin (Fig. 2A, DM) were identified as drugs that prime latent HIV-1 infection for efficient reactivation. The two drugs have no structural similarities and act through entirely different molecular mechanisms. Dactinomycin is a polypeptide antibiotic that acts as a DNA intercalator and transcription inhibitor (35, 77, 78). Aclacinomycin is an oligosaccharide anthracycline that, in contrast with its structural analogs daunomycin and doxorubicin, lacks cardiotoxicity (49, 53).
Dactinomycin exerted its optimal priming activity for latent HIV-1 infection at a concentration of 4 ng/ml (3.18 nM), with a priming effect being observed at concentrations as low as 0.5 ng/ml. Dactinomycin had only a small direct HIV-1-reactivating effect (Fig. 2A). Also, to act as a priming drug for HIV-1 reactivation, dactinomycin needed to be applied prior to stimulation. The optimal pretreatment period for dactinomycin prior to addition of the activating stimulus was 18 h (Fig. 2C). The optimal drug concentration for aclacinomycin to act as a priming drug for HIV-1 reactivation was found to be 0.1 μM. Aclacinomycin and the activating stimulus could be added simultaneously to obtain a synergistic reactivating effect. Pretreatment for up to 24 h was still effective. We did not test longer pretreatment periods.
The primary mechanism by which dactinomycin and aclacinomycin exert their on-target drug effect in cancer cells is DNA intercalation, with a second reported function of aclacinomycin and dactinomycin being their ability to block transcription. To this end, we tested whether other DNA intercalators or transcription inhibitors would prime latent HIV-1 infection for reactivation. The DNA intercalators daunorubicin, doxorubicin, rebeccamycin, oxaliplatin, and amsacrine and the transcription inhibitors α-amanitin, ICRF-193, camptothecin, and 5,6-dichloro-beta-d-ribofuranosyl-benzimidazole (DRB) were titrated on CA5 T cells and incubated for various periods of time prior to triggering reactivation by a suboptimal dose of an activator. None of the tested drugs had any priming effect on latent HIV-1 infection, suggesting that DNA intercalation or inhibition of transcription is not the primary mode of action by which dactinomycin or aclacinomycin primed latent HIV-1 for reactivation. The data for aclacinomycin, dactinomycin, daunorubicin, DRB, and α-amanitin are shown in Fig. 2A. The failure of daunorubicin to produce a priming effect on latent HIV-1 infection is particularly interesting as daunorubicin is a structural analog of aclacinomycin. Both daunorubicin and aclacinomycin are anthracycline antibiotics that differ mostly in the length of the oligosaccharide moiety that extends from the planar anthraquinone nucleus. While daunorubicin or doxorubicin has a monosaccharide side chain, aclacinomycin has a trisaccharide side chain.
Recently, it was demonstrated that latent HIV-1 infection events, similarly to active infection events, are found almost exclusively integrated in actively expressed host genes (29). Transcriptional interference was suggested as an additional mechanism that would control latent infection (30, 44). To this end, we investigated whether the sense of orientation of the integrated latent virus relative to the transcriptional direction of the host gene would influence the ability of dactinomycin or aclacinomycin to prime latent infection for reactivation. In the latently HIV-1-infected CA5 T cells, virus and host gene are oriented in the same transcriptional orientation, whereas in EF7 cells, the virus is integrated into the host gene in the converse transcriptional orientation. As shown in Fig. 3, dactinomycin exerted its priming effect in either of the possible integration scenarios, suggesting that the priming effect is unlikely to be caused by possibly altering transcriptional interference effects that may add to the control of latent infection. The data rather suggest that dactinomycin treatment must favor direct LTR activation or promote elongation efficacy. Most importantly, in both cell lines, TNF-α even at high concentrations would not trigger full reactivation at the population level. However, in the presence of optimal concentrations of dactinomycin, full reactivation of latent HIV-1 infection at the population level was achieved for either T cell population. Similar data were obtained using aclacinomycin (data not shown).
We next tested the ability of dactinomycin and aclacinomycin to trigger reactivation of latent HIV-1 infection in primary T cells, using an in vitro model of latent HIV-1 infection in primary T cells that we have previously used (23). Briefly, CD8-depleted T cells were activated by antibody-mediated CD3/CD28 stimulation. On day 7 poststimulation, the T cells were infected with HIV-1 NL43. Initial infection levels were determined by intracellular HIV-1 p24 staining for flow cytometric analysis. Initial infection levels usually ranged between 8 and 20%. Over the next 7 to 14 days, active infection levels subsided in parallel with cellular activation markers, such as CD25. The cells could then be kept in extended culture for up to 4 weeks in the presence of low levels of IL-2 (10 U/ml). Reactivation following stimulation can be measured as an increase of the percentage of p24-positive cells over background. The size of the latent reservoir in different experiments was donor dependent and varied between 1 and 5%. Using this experimental system, we tested the ability of aclacinomycin or dactinomycin to directly trigger reactivation of latent HIV-1 infection. Representative results generated using T cells from different donors are depicted in Fig. 4. As in T cell lines, aclacinomycin and dactinomycin directly triggered some level of reactivation of latent HIV-1 infection. Reactivation efficacy varied between 20 and 50% of the maximum reactivation levels achievable using anti-CD3/CD28 antibodies. As we observed donor variation in our ability to establish latently infected T cell population, we also observed variation in the ability of dactinomycin or aclacinomycin to induce HIV-1 reactivation. The three depicted experiments represent possible experimental outcomes in regard to the ability of dactinomycin and aclacinomycin to trigger HIV-1 reactivation. The data suggest that, as in latently HIV-1-infected T cell lines, dactinomycin and aclacinomycin exert some direct effect on latent HIV-1 infection but mostly act to prime infection for reactivation that would then have to be fully triggered by a second low-level stimulus. Drug screens to identify low-level activating agents that would synergize with aclacinomycin or dactinomycin have been initiated.
Having confirmed that the priming effect of dactinomycin and aclacinomycin on latent HIV-1 infection is also seen in latently HIV-1-infected primary T cells, we next tested whether aclacinomycin or dactinomycin would eventually boost active infection (Fig. 5). Optimal HIV-1-reactivating agents should not boost active HIV-1 infection to minimize the risk of de novo infections. For this purpose, we titrated both drugs and a panel of structural or functional analogs (DNA intercalators [WP631, amsacrine, daunorubicin, and rebeccamycin], antitumor agents with different mechanisms [oxaliplatin and docetaxel], and transcription inhibitors [ICRF-193, DRB, α-amanitin, and camptothecin]) on two chronically actively HIV-1-infected reporter T cell lines (JNLG#35 and JNLG#44 ). These cell lines are infected with a GFP reporter virus, and compound effects on HIV-1 transcription can be directly determined by flow cytometric analysis quantifying GFP mean channel fluorescence (MCF) intensity. Dactinomycin and aclacinomycin at concentrations relevant for HIV-1 reactivation did not boost active infection but rather inhibited infection. Also, α-amanitin did not have any effect on active infection, which is noteworthy, as a previous report using plasmid-based experimental systems suggested that α-amanitin and dactinomycin would boost LTR activity (17, 84). The differences may be explained by the fact that these previous studies used HeLa cells with stably integrated LTR-luc constructs and did not study drug effects on integrated, replication-competent full-length viruses.
Some agents that did not trigger HIV-1 reactivation boosted active HIV-1 transcription, such as the DNA intercalators daunorubicin and rebeccamycin. This is likely the result of their reported ability to stimulate NF-κB activity (11). In neither case were the effects of the tested transcription inhibitors pronounced in the absence of cytotoxic effects.
We also tested whether aclacinomycin or dactinomycin would boost HIV-1 infection in primary T cells. For this purpose, we infected primary T cells with a GFP reporter virus and determined the level of infection and the GFP mean channel fluorescence intensity as a marker of LTR activity every 24 h for a total of 72 h following drug addition. Again, no boosting effect of aclacinomycin or dactinomycin on active HIV-1 infection was observed, neither at the level of infection (% GFP-positive cells) nor at the level of promoter activity (GFP mean channel fluorescence). The data for dactinomycin are shown in Fig. 5B.
In summary, these data demonstrate that aclacinomycin and dactinomycin achieve their priming effect for HIV-1 reactivation without boosting active HIV-1 infection. As there is no correlation of the proposed primary effects of aclacinomycin and dactinomycin to act as DNA intercalators or as transcription inhibitors with the observed effect on active HIV-1 expression, these data further suggest that the priming effect of the two drugs is not related to their primary mode of action as anticancer agents.
Dactinomycin and aclacinomycin have no structural similarities and reportedly exert their primary drug effect through different mechanisms. However, both drugs were investigated as cell-differentiating agents, an effect that can be observed at subtoxic concentrations (2, 50, 52, 64, 65, 70). Until the early 1990s, they were part of a group of structurally unrelated drugs or compounds that was investigated because of their ability to differentiate cells and therefore act as anticancer compounds. In addition to dactinomycin or aclacinomycin, these included drugs such as vorinostat/SAHA, cytarabine (39, 46), and 5′-azacytidine (21, 55) or compounds such as HMBA (58) and aphidicolin (28, 46, 51). Of these, HMBA and SAHA have been reported to reactivate latent HIV-1 infection (19, 76). In our experimental systems, both HMBA and SAHA triggered some HIV-1 reactivation and at subtoxic concentrations were relatively potent at priming latent HIV-1 infection for full reactivation by a suboptimal activating TNF-α concentration (Fig. 6).
Prior to the discovery of its HDAC-inhibitory capacity, SAHA had been developed as a second-generation, hybrid-polar cell-differentiating anticancer agent, a further development/enhancement of HMBA (60). SAHA was reported to exert its cell-differentiating effect at 100-fold-lower concentrations than those of HMBA (59). In our system, SAHA exerted some HIV-1-reactivating ability by itself at 0.1 μM, which correlated with the onset of drug toxicity. Higher concentrations of SAHA immediately abrogated cell viability. Like other cell-differentiating agents, but unlike HDAC inhibitors (e.g., NaBu or trichostatin A [data not shown]), at low concentrations (40 and 120 nM) SAHA exerted a synergistic HIV-1-reactivating effect with TNF-α on latent infection in CA5 T cells (Fig. 6C). It is difficult to detail the full extent of possible synergistic effects of SAHA with low-level stimuli in this system, as other than for dactinomycin or aclacinomycin, the HIV-1-priming or -reactivating effect for SAHA was observed at the immediate onset of drug toxicity. Nevertheless, these findings suggest that the cell-differentiating capacity of some drugs could be a key component of their ability to trigger HIV-1 reactivation.
To test whether cell-differentiating drugs as a group are likely candidates for HIV-1 reactivation therapy, we decided to probe the ability of cytarabine, another anticancer drug reported to have secondary cell-differentiating capacity, to prime latent HIV-1 infection for reactivation. Cytarabine or cytosine arabinoside is a deoxycytidine analog in which deoxyribose is replaced by an arabinose sugar and has no structural similarities to dactinomycin or aclacinomycin. Due to its extremely short serum half-life, cytarabine is not a first-choice anticancer drug but, for example, had been used during the induction therapy of the “Berlin patient,” an HIV-1 patient who was cured in the wake of a bone marrow transplant operation (3). In our experiments, cytarabine did not act to trigger HIV-1 reactivation by itself but acted as a priming agent for latent HIV-1 infection, albeit to a somewhat lesser extent than dactinomycin or aclacinomycin. The priming effect of cytarabine on reactivation of latent HIV-1 infection in CA5 T cells is shown in Fig. 6D.
We further tested the ability of aphidicolin, another agent that has been reported to have cell-differentiating capacity, to prime latent HIV-1 infection for reactivation (51). Aphidicolin is a tetracyclic diterpene antibiotic with reported antiviral properties. Aphidicolin primarily acts as a reversible inhibitor of eukaryotic nuclear DNA replication. Again, aphidicolin by itself had no HIV-1-reactivating capacity but acted synergistically with low-level TNF-α stimulation to induce potent HIV-1 reactivation (Fig. 7). Taken together, these data raise the possibility that selected cell-differentiating drugs as a class may be suitable to trigger HIV-1 reactivation in an induction therapy setting comprising several drugs.
Our data indicated that the priming effect of aclacinomycin and dactinomycin on latent HIV-1 infection was likely triggered at the level of transcriptional elongation. We thus investigated the possibility that the drugs would release positive transcription elongation factor (P-TEFb) from its inactive complex with HEXIM-1. P-TEFb association with RNAP II is essential to trigger efficient elongation, and the presence of P-TEFb (a complex of cyclin T1 [CycT1] and CDK9) at the RNAP II complex associated with the HIV-1 LTR has been demonstrated as essential for efficient transcriptional elongation (47, 87). Conversely, restriction of P-TEFb has been associated with HIV-1 latency (73). HMBA-mediated release of P-TEFb from its complex with HEXIM-1 has previously been reported to trigger HIV-1 reactivation (18, 19).
For this purpose, we treated the latently HIV-1-infected J89GFP or CA5 T cells either with 1.0 μg/ml dactinomycin for 1 h or with the physiologically optimal concentration of 0.004 μg/ml for 18 h. Cell lysates were then separated on a glycerol gradient (10 to 45%) to reveal possible changes in the composition of the P-TEFb/HEXIM-1 complex. Release of P-TEFb from the inactive complex with HEXIM-1 (large complex), which is found in the glycerol gradient fractions with higher glycerol content, is indicated by a shift to a smaller complex (CDK9/CycT1) found in the gradient fractions with lower glycerol content. Each gradient fraction was separated on an SDS-PAGE gel and subjected to Western blotting and antibody staining. The results of these experiments using J89GFP cells are presented in Fig. 8. Staining with an anti-CDK9 antibody revealed that treatment of J89GFP with 1 μg/ml dactinomycin for 1 h quantitatively released P-TEFb from its complex with HEXIM-1. A shift of CDK9 presence from the large complex to the small complex could also be detected under treatment conditions that represented the optimal conditions for HIV-1 reactivation (0.004 μg/ml dactinomycin for 18 h). Similar results were obtained using anti-HEXIM-1 antibody. However, for HEXIM-1 no shift toward small complex was observed at the optimal condition of 0.004 μg/ml dactinomycin for 18 h. The minimal dactinomycin concentration to induce a shift toward small complex was 0.01 μg/ml. Other than CDK9, HEXIM-1, even in control cells, was found in the small-complex fractions, suggesting that free HEXIM-1 is present in abundance, which would be in line with the idea that it should serve as a regulator of transcription by inactivating P-TEFb. Similar results were obtained for aclacinomycin. Figure 8B shows how aclacinomycin at 5 μM provokes a complete shift of CDK9 into the small-complex fractions 1 h poststimulation, indicating that aclacinomycin also releases P-TEFb from its inactive complex with HEXIM-1.
While it has been previously reported that dactinomycin would act by releasing P-TEFb from its complex with 7SK-RNA in HeLa cells and thereby would directly promote LTR activity of transfected LTR-luciferase constructs (84), we did not observe such an LTR-stimulatory effect of dactinomycin when we used T cell lines (NOMI cells) in which an HIV-1 LTR-GFP was integrated into the cellular genome (Fig. 8C) (33). While PMA stimulation in a concentration-dependent manner induced GFP expression, dactinomycin failed to induce LTR-driven GFP expression by itself. Taken together, these data suggest that while P-TEFb release could act to prime latent HIV-1 infection for reactivation, P-TEFb release by itself is insufficient to trigger reactivation. The release of P-TEFb from its inactive complex with HEXIM-1 would lower a putative reactivation threshold and favor elongation of transcription by the paused RNAP II complex found at the latent HIV-1 LTR triggered by additional factors (37, 54, 86).
The challenge to eradicate the latent HIV-1 reservoir, a prerequisite for a curative therapy, is probably best compared to leukemia treatment. A subset of cells needs to be systemically and completely, but selectively, eradicated. A single cell that escapes therapy will likely result in the reoccurrence of the tumor. Combination therapies where several drugs that target different mechanisms are combined to maximize the therapeutic effect of therapy have proven to be the most successful way forward to treat various forms of leukemia.
Previous unsuccessful HIV-1 eradication attempts used single drugs. On the basis of the success of combination therapy in the treatment of several forms of leukemia, drug combinations that target latent HIV-1 infection at several levels of molecular control will likely also be the most promising way forward to HIV-1 eradication. One of the major hurdles to overcome is that other than in the case of leukemia cells, latently HIV-1-infected T cells are phenotypically not distinguishable from uninfected CD4+ memory T cells that serve as the cellular main reservoir of latent HIV-1 infection. Latently HIV-1-infected cells can thus not be specifically targeted. Thus, a therapy form based on a systemic stimulus that ideally is selective for T cells and most importantly will trigger HIV-1 reactivation without causing detrimental side effects such as a cytokine storm or immune hyperactivation will have to be developed. Dissociation of HIV-1 reactivation from high-level cell activation will likely be essential. We hypothesized that this may be achievable if we can identify combinations of drugs that first lower the activation threshold for latent HIV-1 infection and then trigger HIV-1 reactivation with a low-level activating stimulus.
In a drug screen designed to directly identify drug combinations that would reactivate latent HIV-1 infection, we initially identified two FDA-approved drugs, aclacinomycin and dactinomycin, as compounds that lower the activation threshold required to achieve full reactivation at the population level and that directly trigger HIV-1 reactivation in primary T cells. Our studies revealed that the drugs do not act by their primary mechanism as topoisomerase inhibitors or as DNA intercalators but rather target latent HIV-1 infection by their ability to trigger cell differentiation at subtherapeutic concentrations. To this end, we also demonstrate that cytarabine, a third FDA-approved anticancer drug with cell-differentiating capacity, and aphidicolin prime latent infection for reactivation. Our findings suggest that repositioning of a subgroup of FDA-approved anticancer drugs that exert cell-differentiating effects could be a promising way forward to a novel therapeutic approach to eradicate latent HIV-1 infection.
A major consideration for the development of a curative therapy is the current success of antiretroviral therapy (ART). For most patients who have access to care, ART provides the ability to treat HIV-1 infection like a chronic disease. Based on the data available at this time, ART may provide patients with the possibility to live a relatively normal life with a close-to-normal life expectancy. As beneficial as a curative treatment for HIV-1 would be for many reasons, other than for the treatment of leukemia, given the success of ART, the acceptable side effects need to be minor and should be limited during the treatment.
In the light of these requirements, it needs to be emphasized that aclacinomycin, dactinomycin, and cytarabine in vitro exert their priming or reactivating effects on latent HIV-1 infection at concentrations that are not cytotoxic to T cell lines or primary T cells. This raises the possibility of using these drugs at subtherapeutic doses relative to cancer therapy.
Other than being used at lower concentrations, the three drugs are clinically extremely well defined. Aclarubicin was first described in 1975 as a product of Streptomyces galilaeus (53). Relative to other anthracycline antibiotics, such as daunorubicin or doxorubicin, aclacinomycin has been described to have reduced cardiotoxicity (49). While its primary anticancer activity is linked to its ability to act as a DNA intercalator, the mechanisms by which it induces cell differentiation are ill defined. It has been suggested that the differentiating ability of aclacinomycin would be linked to its ability to block the synthesis of asparagine-linked glycoproteins, which is not affected by anthracyclines that have monosaccharide side chains, but overall, it remains unclear how aclacinomycin treatment results in cell differentiation.
The primary dactinomycin effect during cancer treatment is also based on its ability to intercalate into DNA and to inhibit RNA polymerase (45). Dactinomycin is being used in the treatment of childhood leukemia. How it actually triggers cell differentiation remains largely unclear.
Cytarabine is classified as an antimetabolite and seems to primarily act by inhibiting DNA polymerase. Incorporation of the cytosine analog into the nascent DNA and nascent RNA has been reported. Cytarabine used as part of an HIV-1 eradication treatment may be particularly interesting. In a study comparing low-dose cytarabine with other treatment regimens for acute myeloid leukemia and high-risk myelodysplastic syndrome in patients not considered fit for intensive treatment, cytarabine proved to be most efficient (15). This indicates that cytarabine can exert effects at low doses and is tolerated well at low doses. Cytarabine was actually used in the induction therapy for the “Berlin patient,” an HIV-1-infected individual who was cured of HIV-1 infection in the course of a bone marrow transplant required as the result of reoccurring leukemia (3).
A fourth cell-differentiating agent, which is the first to be reported to trigger HIV-1 expression, is N′-N′-hexamethylene-bisacetamide (76). More recently, Contreras et al. demonstrated that this effect is linked to the ability of HMBA to release P-TEFb from its inactive complex with HEXIM-1 (19). However, transfer of HMBA as an HIV-1-reactivating agent into the clinical situation is unlikely to be successful. HMBA caused severe thrombocytopenia, which limits the amount of drug that can be administered. Furthermore, continuous drug exposure is required to cause the cell-differentiating effect. As the biological half-life of HMBA in the patient is very short (about 1.5 h), HMBA must be administered by continuous infusion to maintain a clinical effect (4).
We demonstrated that just like HMBA, dactinomycin and aclacinomycin release P-TEFb from its inactive complex with HEXIM-1. Given previous findings that P-TEFb restriction is important for the transition into latency (73) and given the overall importance of P-TEFb for HIV-1 transcriptional elongation (47, 87), our results suggest that induced P-TEFb release is an essential part in the effect of these drugs to lower the reactivation threshold. However, it is unlikely that P-TEFb release is the only target of these drugs in the context of HIV-1 reactivation. Rather, it is likely that the drugs alter the cellular transcription factor profile and provide a more permissive cellular environment for viral transcription, also at other levels. Unfortunately, to date the molecular mechanisms underlying the cell-differentiating effects of the various drugs/compounds are ill defined. There are, however, some candidate genes that are reported to be regulated by these cell-differentiating drugs, which also have been implied in the regulation of HIV-1 expression. Aclacinomycin has been reported to induce GATA expression (26), and GATA has been described to induce LTR activity (71, 83). Aclacinomycin also was noted to trigger a rapid but transient decrease in the levels of c-myc and c-myb transcripts (62). c-myc downregulation has also been reported for dactinomycin (12) and for cytarabine. For cytarabine, the first peak of c-myc downregulation in K562 human erythroleukemia cells was correlated with the onset of cell differentiation (8). In the context of latent HIV-1 infection, this is interesting, as valproic acid, an HDAC inhibitor reported to trigger HIV-1 reactivation (85), has also been reported to downregulate c-Myc. Inhibition of c-Myc was shown to reduce HDAC1 occupancy of the HIV-1 LTR, to relieve c-Myc-imposed repression of Tat activation, and to increase LTR expression (31). Interestingly, valproic acid is not only an HDAC inhibitor but also a cell-differentiating agent (27). We are currently investigating whether there is a correlation between the ability of cell-differentiating drugs to trigger HIV-1 reactivation and their ability to downregulate c-myc.
In the context that HMBA, dactinomycin, and aclacinomycin have all been reported to act as cell-differentiating agents, it is further interesting that the HDAC inhibitor SAHA (vorinostat), reported to reactivate latent HIV-1 infection (7, 20, 24), was initially also developed as a very potent cell-differentiating polar agent, a second-generation HMBA (60). Its ability to act as a potent HDAC inhibitor was described only later (59). To generate SAHA, the structures of HMBA and the HDAC inhibitor trichostatin A (TSA) were used as the templates (60). However, while SAHA acts to directly trigger reactivation of latent HIV-1 infection and primes latent HIV-1 infection for reactivation, the HDAC inhibitor TSA exhibits no HIV-1-reactivating capacity in our experimental systems. This raises the question whether the HIV-1-reactivating capacity of SAHA is actually a function of its ability to drive cell differentiation or of its ability to inhibit HDACs.
Of note, while many dispute the value of T cell lines for HIV-1 latency research, a previous study by the Karn laboratory could identify only a single difference. Latent infection in T cell lines, other than in primary T cells, could be reactivated by TNF-α. No other differences at the molecular level were revealed (54, 73). Planelles, Bosque, and their collaborators have recently demonstrated that latently HIV-1-infected primary T cells can even be induced to proliferate by IL-7 without triggering comprehensive HIV-1 reactivation, voiding the argument that proliferating T cell lines must control latent HIV-1 infection by a fundamentally different mechanism than that of resting memory T cells (13). We recently reported that AS601245 would prevent stimulation-induced HIV-1 reactivation despite the induction of high levels of NF-κB activity in either cell type, T cell lines and primary T cells, another example that latently infected T cell lines in many ways are a good experimental substitute for primary T cells (80). Our finding that the priming capacity of aclacinomycin or dactinomycin acting on latent HIV-1 infection that we identified using latently HIV-1-infected T cell lines can also be observed in latently infected primary T cells adds further support to the notion that the more recently developed models of latent HIV-1 infection in T cell lines (23, 34, 41) are representative of the molecular mechanisms that control latent infection in primary T cells. Combining research in the two systems is likely the most promising way forward to understanding latent HIV-1 infection.
From a drug-screening point of view, the identification of these drugs during a screening effort is somewhat remarkable. Drugs such as cytarabine that act only to reactivate latent HIV-1 infection in conjunction with a low-level activating stimulus validate the approach of directly screening for drug combinations rather than screening for single magic-bullet compounds. Obviously, it could be argued that drug screening should exclusively be performed using in vitro latently HIV-1-infected primary T cells. This is likely possible for drug-repositioning efforts with very limited amounts of compounds to be tested. However, screening larger compound libraries that are comprised of >1 × 105 compounds is unlikely to become a reality due to the scarcity of the cell material that can be reliably generated.
Screening other anticancer drug/compound collections, in particular nucleoside analog libraries, now seems a promising approach to identify more priming agents for the treatment of latent HIV-1 infection. It is conceivable that some of these compounds lack potent anticancer effects but remain potent cell-differentiating agents. Screens for low-level activating agents that act in conjunction with identified priming agents would be the logical next step forward. Our data provide concept validation that this approach could be a promising move toward the development of a curative therapy for HIV-1 infection.
This work was funded in part by NIH grant R01AI064012 to O.K. Parts of the work were performed in the UAB CFAR biosafety level 3 (BSL-3) facilities and by the UAB CFAR Flow Cytometry Core/Joint UAB Flow Cytometry Core, which are funded in part by NIH/NIAID P30 AI027767 and by NIH 5P30 AR048311.
Takao Shishido performed this research at the University of Alabama at Birmingham as a visiting scientist from Shionogi & Co., Ltd., Japan.
Published ahead of print 13 June 2012