Determinants of the Establishment of Human Immunodeficiency Virus Type 1 Latency

All T-cell lines (Jurkat, Molt, SupT1, PM1, C8166, and CEM-GFP) and the CD4-positive B-cell line (AA2) used in infectivity assays were obtained from the NIH AIDS Research and Reference Reagent Program. These cells, as well as the latently HIV-1 infected J89GFP cells (41), 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. All latently infected cell lines generated in the described experiments were maintained in supplemented RPMI 1640. Fetal bovine serum was obtained from HyClone (Logan, UT) and was tested on a panel of latently infected cells to ensure that it did not spontaneously trigger HIV-1 reactivation (35, 41).

Anti-CD3 MAb (clone UCHT1) and anti-CD28 MAb (clone CD28.2) were purchased from Pharmingen. The phorbol ester 13-phorbol 12-myristate acetate (PMA) or prostratin and the HDAC inhibitors (sodium butyrate [NaBu], trichostatin A [TSA], and valproic acid), as well as the DNMT inhibitor 5-azacytidine, were obtained from Sigma. The enhanced green fluorescent protein (EGFP) reporter virus that was used, HIV-1 NLENG1-IRES, has been described elsewhere (41, 44).

The HIV-1 reverse transcriptase (RT) inhibitor (RTI) lamivudine, the HIV-1 protease inhibitor (PI) indinavir, and the integrase inhibitor 118-D-24 were obtained from the NIH AIDS Research and Reference Reagent Program. The HIV-1 transcription inhibitor Ro24-7429 was a kind gift from Roche Scientific.

Infection levels in the cell cultures were monitored by flow cytometric (FCM) analysis of EGFP expression. FCM analysis was performed on a Guava EasyCyte (Guava Technologies, Inc.) or an LSRII (Becton Dickinson) cytometer. 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.).

To determine the level of cell proliferation and cell death in the cell cultures, we added a defined number of Sphero blank calibration particles (Pharmingen) to each individual culture, in which cell numbers had been adjusted by manual counting using a Neubauer chamber. Due to differences in size and granularity, Sphero blank calibration particles (diameter, 6 to 6.4 μm) can be easily distinguished from T cells in a forward scatter/side scatter dot plot analysis using flow cytometry. At the time of analysis, we set up the flow cytometer to acquire a defined amount of beads while acquiring all cells in the live gate. The ratio of beads to cells in each culture allows for the determination of absolute cell numbers for each time point analyzed.

The population of chronically actively infected cells used in the reported HIV-1 transcription inhibitor experiments was generated by infecting a recently described HIV-1 reporter cell line (NOMI; Jurkat T cell based) with a primary patient HIV-1 isolate (CUCY) (36). In the NOMI reporter cells, EGFP expression can be used as a direct and quantitative correlate of HIV-1 expression. Two weeks postinfection (p.i.), EGFP-positive cells were enriched by FCM cell sorting and the resulting EGFP-positive cell population was expanded. We observed that in the next 12 weeks EGFP expression was gradually shut down in >30% of the cells. After that time, the percentage of EGFP-positive cells remained constant. At this time point, the cell culture was subjected to a second round of cell sorting to enrich for EGFP-positive cells and the resulting cell population was employed in the described experiments (see Fig. 3).

Peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll Paque centrifugation and cultured in supplemented RPMI 1640 to which phytohemagglutinin L (PHA-L) (2 μg/ml) and Giant cell tumor conditioned medium (Igen International, Inc., Gaithersburg, MD)-conditioned media (10%) were added. At 2 or 7 days following stimulation, fractions of the parental cell culture were infected with HIV-1 NLENG1-IRES (37). On day 1 p.i. a lamivudine-indinavir combination was added to the infection cultures to inhibit de novo infection and to prevent the formation of preintegration latency. In the absence of IL-2, each culture was maintained until day 14 poststimulation and then subjected to fluorescence-activated cell sorting to remove EGFP-positive and therefore actively HIV-1-expressing cells. The EGFP-negative cell fraction, which would hold noninfected and silently HIV-1-infected cells, was then activated with the indicated stimuli, and reactivation of silent HIV-1 infection was measured 24 h poststimulation by determining EGFP fluorescence using fluorescence-activated cell sorting analysis.

Nuclear extracts were generated using the NE-PER nuclear and cytoplasmic extraction reagent according to the manufacturer's instructions (Pierce; ThermoScientific). Quantification of NF-κB activity in the respective nuclear extracts was performed using a TransAM NF-κB family transcription factor assay kit according to the manufacturer's instructions (Active Motif, Carlsbad, CA).

The underlying idea of this nested PCR strategy is that, if the virus is latently integrated into an actively expressed host gene, viral sequences should be detectable in host gene mRNA (30). To exclude the possibility that low-level viral gene transcription could be detected, the 5′ primers are selected to bind prior to the HIV-1 LTR transcription start site. By this means, viral integration into an actively expressed host gene can be determined without any knowledge of the host gene sequence. cDNA from latently infected cell clones was generated using the μMACS mRNA isolation kit with an added digestion step with RNase-free DNase and the μMACS cDNA synthesis module from Miltenyi according to the manufacturer's instructions. PCR amplification was then performed as nested PCR. Primers used for the first amplification are 5′-outLTR, 5′-AGGTGTGGCCTGGGCGGGACT-3′, and 3′-out-gag, 5′-CAGCAAGCCGAGTCCTGCGTCG-3′. The primers 5′-inLTR, 5-GGCGAGCCCTCAGATCCTGC-3, and 3′-in-gag, 5′-GTCCCTGTTCGGGCGCCACTGCTA-3′, were then used for second-round amplification using 5 μl of the first-round PCR product as template. Conditions for both PCRs were as follows: 94°C for 5 min; 30 cycles of 94°C for 30 s, 65°C for 30 s, and 72°C for 30 s; and a final extension at 72°C for 7 min. For visualization, the PCR products were then loaded on a 2% agarose gel and separated by gel electrophoresis. Control reactions in which RT was omitted during cDNA generation were performed to ensure that no genomic DNA was amplified.

Understanding the mechanisms governing HIV-1 latency establishment may provide new insights into how to therapeutically interfere with HIV-1 latency in vivo. Currently, HIV-1 latency at the molecular level is studied mostly using latently HIV-1-infected clonal T-cell lines in which HIV-1 latency has already been established at the time point of analysis (e.g., U1, ACH-2, J89GFP, and J-LAT) (28, 37, 41). As such, these cells do not permit the study of mechanisms underlying HIV-1 latency establishment but are limited in use to the study of factors that govern latency maintenance or HIV-1 reactivation.

To study HIV-1 latency establishment at the population level, we initially infected Jurkat T cells with an EGFP-expressing reporter virus (41, 44). Routinely, RTIs and PIs were added to the infection cultures on day 1 p.i. to prevent the establishment of preintegration latency (47) and to restrict continuous de novo infection that would perturb the stability of a possibly established latently infected cell population (reactivation by superinfection). RTIs and PIs were then continuously added throughout the entire experimental period. To follow the level of inactive or silent viral integration events over time and to monitor the establishment of a stable latently HIV-1-infected cell population, we removed samples of the infected parental culture at various time points and stimulated these samples with PMA or tumor necrosis factor alpha (TNF-α), two potent HIV-1-(re)activating agents (29). Levels of active infection in the unstimulated parental culture, as determined by FCM quantification of EGFP expression, were then compared with the levels of total infection in the activated samples 24 h poststimulation. The difference between total infection (% EGFP-positive cells in activated sample) and active infection (% EGFP-positive cells in untreated sample) represented the number of silent integration events.

In the representative experiment depicted in Fig. 1A, the mean active infection level in the parental infection cultures on day 3 p.i. was 28% ± 2%, with 43% ± 3% mean total infection (TNF-α activated), indicating that at this time point, 15% of the cell population contained integrated infection events that were not transcriptionally active but could be activated (silent infection). Active infection levels initially rapidly declined, mostly due to the cytopathic effect of the virus, since no difference in the cell proliferation rates of the uninfected and the infected population could be detected (PKH26 stain; data not shown). After day 10, the decay of active infection levels significantly slowed and complete self-eradication of the active infection took >50 days (Fig. 1B). The pool of silently infected cells also initially quickly decreased, stabilized between days 10 and 15, and remained at the 5% level until the end of the experiment on day 64, when it was evident that this cell population consisted of stably latently HIV-1-infected cells. In control experiments, addition of the integrase inhibitor 118-D-24 on day 1 p.i. produced identical results (data not shown), validating the previously described use of RTIs to prevent preintegration latency (47).

It has been suggested that in vivo cells that are close to a hypothetical minimum intracellular activation threshold become latently HIV-1 infected (30). These cells are sufficiently activated to allow for viral integration and initial viral gene expression. If the cellular activation levels then drop below this arbitrary threshold prior to the onset of any virally induced cytopathicity or immune recognition, viral gene expression is shut off, leaving the virus latently integrated. A second possibility is that the virus integrates but never actively expresses its genes. Obviously, the latter possibility would deprive even a functional immune system of any chance to attack and eradicate these infected cells prior to latency formation. To investigate whether either of these two forms of latency establishment would be predominant, we performed experiments in which we separated actively HIV-1-expressing and thus EGFP-positive cells from the EGFP-negative cell population through fluorescence-activated cell sorting on day 2 p.i. (Fig. 2A). The EGFP-negative population contained noninfected cells, as well as silently infected cells. RTIs and PIs were added to the cultures on day 1 p.i. and continuously replenished throughout the entire duration of the experiment. A part of the parental infection culture was not subjected to cell sorting to serve as a reference culture. The parental population and, after the cell sorting, the EGFP-positive and the EGFP-negative fractions were placed into continuous cell culture, and levels of active and silent infection were quantified throughout a 41-day period, while cell proliferation and cell death were monitored for the first 10 days (Fig. 2B).

If latent HIV-1 infection were to develop exclusively from silencing of active infection, then, after a prolonged culture period, the entire pool of latently infected cells identified in the parental infection culture should be contained in the actively infected, EGFP-positive cell population. Conversely, if HIV-1 latency were to develop following silent integration, then, after an extended experimental period, the sorted EGFP-negative and the unsorted parental population should hold the same level of latently HIV-1-infected cells. Obviously, if both pathways contribute to latency establishment, the total number of latent events observed in the parental culture would be equal to the sum of the latent events observed in the two sorted cell populations following normalization for cell death/proliferation.

The results depicted in Fig. 2 demonstrate that HIV-1 latency is almost exclusively a result of silent integration and originates from the EGFP-negative cell population. In four independent experiments, using Jurkat or Molt-4 T-cell-based HIV-1 infection cultures, we physically separated T cells that were EGFP positive from the EGFP-negative cell fraction (Fig. 2A). Purity of these sorts in all cases was greater than 98%. We then followed the establishment of a latently infected cell population in these sorted cell populations as well as in the parental, unsorted population until day 41 p.i. The kinetics of absolute cell numbers were determined in all cultures for the first 10 days (Fig. 2B). These experiments revealed that the initially EGFP-negative cell populations on day 41 p.i. contained the same level of latent infection events as did the unsorted parental cell cultures (Jurkat cells [Fig. 2C] and Molt-4 cells [Fig. 2D]), arguing that the vast majority of the latently infected cells are derived from the EGFP-negative population and are established by silent integration.

Latently HIV-1-infected cells can also be detected in the initially EGFP-positive cell cultures. However, by tracking cell proliferation and HIV-1-induced cell death in this population, we found that the contribution of these cells to the total population (unsorted) on day 10 p.i. is already less than 1% (Fig. 2B). As latency levels in these populations did not exceed 10%, it can be calculated that the total contribution of latent infection events derived from initially active infection events in the nonsorted population is <0.1% (<1% × 10%). As the separation of the EGFP-positive and EGFP-negative populations had been performed on day 2 p.i., the results further show that the decision whether integration occurs in an active or latent state is made early after infection.

While our results indicate that HIV-1 latency establishment is a function of silent integration, we also sought to investigate the possibility that active infection events could be rendered latent by long-term suppression of active HIV-1 transcription. This idea assumes that pharmacological suppression of transcriptional activity at the integrated viral LTR would assist the establishment of a restrictive histone code at the LTR. This would in turn, as previously suggested, stabilize the promoter in a transcriptionally inactive/latent state. For the experiments shown in Fig. 3, we utilized a NOMI reporter cell population in which chronically actively integrated viruses drive EGFP expression of a stably integrated HIV-1 LTR-EGFP reporter plasmid (36). The cells were treated with an optimal dose of the HIV-1 transcription inhibitor Ro24-7429 (10 μg/ml), and EGFP expression, as a direct correlate of LTR activity, was measured at the indicated time points using FCM analysis. Ro24-7429 was readded to the culture every 5 days, prior to any loss of its fully suppressive activity. Five days after compound addition, maximum suppression of HIV-1 activity was obtained, with EGFP expression being at background level in >40% of the cells. While Ro24-7429 was continuously added to the parental culture, we removed cell culture samples every 5 days and monitored whether virus expression would rebound in all cells (Fig. 3A). Incomplete rebound of EGFP expression in some cells would be indicative that continuous suppression of active HIV-1 infection had resulted either in the establishment of a latent infection or eventually in permanent transcriptional silencing. However, when Ro24-7429 suppression was removed after 5, 10, 15, or 20 days, virus expression rebounded. Even following 30 days of continuous suppression of HIV-1 expression, HIV-1 activity completely recovered in the entire population. The single-cell analysis of one representative experiment is depicted in Fig. 3B. The analysis regions were adjusted according to baseline EGFP fluorescence in uninfected NOMI reporter cells (upper left quadrant). Maximum achievable suppression of HIV-1 expression as indicated by EGFP expression on day 30 was 73% (lower left quadrant). Following cessation of Ro24-7429 addition to the culture, HIV-1 expression gradually rebounded in all cells and on day 48 (lower right quadrant) was found identical at the population level and for mean channel intensity to the levels seen in untreated chronically actively infected NOMI cells (upper right quadrant). These data suggest that prolonged suppression of HIV-1 transcription is insufficient to silence an initially active HIV-1 promoter. Consequently, a restrictive histone code or an inhibitory DNA methylation pattern that could have potentially formed during this time period of suppression was insufficient to stabilize the inactive state of these viral promoters.

Prerequisite for silent infection, viral integration in the absence of viral gene expression is likely a relatively low level of intracellular activity of the host cell at the time point of infection. While several transcription factors have been reported to be involved in HIV-1 transcription (e.g., Sp-1, AP-1, and NFAT), the nuclear presence of active NF-κB, in particular the p50/p65 heterodimer, is necessary for efficient transcription driven by the HIV-1 LTR (5, 24, 25). If latency establishment is controlled by the presence, or rather the absence, of NF-κB, then increasing the levels of active NF-κB by external stimulation prior to infection should abrogate latency formation. In exchange, stimulated cells should exhibit higher levels of active infection, with total infection remaining at the same level, provided that stimulation does not alter expression of the viral receptor/coreceptor profile. To test this hypothesis, we initially infected Jurkat T cells and simultaneously stimulated the cells with PMA or prostratin, two phorbol esters, which are known to (re)activate HIV-1 expression by stimulating NF-κB activity, the key transcription factor in activating HIV-1 expression (Fig. 4A). We then subjected the cells to the standard culture protocol to determine the size of the silent infection on day 3 and the size of latent infection on day 37. As predicted, while silent infection could be observed in the control infection, stimulation either with PMA or with prostratin abrogated the establishment of a pool of silently infected cells at the day 3 time point. This translated into a greatly diminished pool of latently HIV-1-infected cells at the day 37 time point, confirming that the size of the silently infected cell population on day 3 is predictive for the size of the latently infected cell pool at a later time point. This also provides additional evidence that silent integration is a prerequisite for latent infection. Similar results, although slightly less pronounced, could be obtained with other stimuli that target the NF-κB pathway, such as TNF-α or CD3/CD28 MAb stimulation (Fig. 4B). Again stimulation decreased the pool of cells holding silently integrated infection events on day 3 p.i., which translated to a decreased pool of latently infected cells on day 37 p.i. Thus, increasing the intracellular activation level with three different stimuli that trigger cell activation by different signaling pathways but converge in the NF-κB pathway diminished the likelihood of the establishment of latent infection events.

Previous reports have shown that various T-cell lines exhibit different levels of basal NF-κB activation in the absence of stimulation (12). We thus hypothesized that T-cell lines that have been reported to exhibit low NF-κB baseline activity, such as Jurkat or SupT1 cells, would be permissive for the establishment of latent HIV-1 infection, while cell lines with high levels of baseline NF-κB expression, such as PM1 or C8166 cells, would be nonpermissive for the establishment of latent infection (Fig. 5A). As predicted by the NF-κB profiles, infection of Jurkat or SupT1 cells resulted in the establishment of a silently infected cell reservoir, whereas PM1 and C8166 cells did not allow for silent integration to occur (Fig. 5B). Other cell lines frequently used in HIV-1 infection experiments that were found nonpermissive for HIV-1 latency establishment at the population level were the CD4-positive AA2 B-cell line and CEM-GFP cells. It is also noteworthy that the NF-κB activation profile of three latently infected Jurkat T-cell lines (CA5, 3F12, and 11B10) (Fig. 5A) did not differ from the NF-κB activation profile of the parental Jurkat cells, suggesting that once latency is established, it can be maintained in the presence of an NF-κB activation status that in the setting of a de novo infection is generally supportive of active infection.

To explain the finding that latent HIV-1 infection, in a cell line that is permissive for latency establishment at the population level (e.g., Jurkat T cells), is established only in a small fraction of the cells, we must postulate that NF-κB activity at the single-cell level is stochastically distributed around a median value. Only a small portion of the cells exhibit NF-κB activity below the threshold that is required for active HIV-1 transcription and thus, in turn, is permissive for silent integration. If this hypothesis is correct, and the percentage of cells exhibiting subthreshold intracellular activation levels does not change over time, the percentage of latently infected cells relative to the corresponding initial infection level should remain constant and should be independent of the initial infection level. As we can demonstrate that HIV-1 infection is not biased within a Jurkat T-cell population and all cells can be infected (we have achieved infection levels of >95% using an NL43-based EGFP reporter virus), we can test this idea, by infecting Jurkat T cells with various amounts of virus and calculating the ratio of established latent infection (>day 35) over active infection on day 3 for the different infection levels. Analysis on day 3 p.i. revealed that relative silent infection levels increase with decreasing infection levels. This is to be expected as superinfection levels, which would camouflage latent infection events, increase with increasing total infection levels (Fig. 6A). However, as proposed, we found that the ratio of latent infection as determined on day 41 to active or total infection as determined on day 3 was indeed independent of the level of the initial infection and found to be around 7% (6.9% ± 0.6% to 7.5% ± 0.4%) of initial infection levels (Fig. 6B). We repeated this experiment several times throughout a 2-year period and found that the relative level of latency established varied very little between experiments (7 to 10% of total infection). These small alterations can be best explained by minor changes in the level of baseline NF-κB activity throughout the cell population caused by differences in cell culture conditions (e.g., cell density directly prior to the experiments). Thus, these data provide additional evidence that the transcriptional fate of a virus to integrate in a latent state is dependent on the individual host cell activation state at the time of infection.

To test whether these findings could be translated into a primary T-cell setting, we hypothesized that HIV-1 infection of primary T cells immediately following stimulation would not allow for the establishment of silent/latent HIV-1 infection, while infection at a later time point, after peak cellular activation, would permit the establishment of silent infection at the population level. We stimulated PBMCs with PHA-L (2 ng/ml), and half of the cell population was infected with an EGFP-reporter virus on day 2 poststimulation (6 to 12% infected cells on day 2 p.i.). From day 1 p.i. the infected cell population was then continuously treated with an RTI-PI combination to prevent the formation of preintegration latency and de novo infection. The second half of the cell population was kept in culture and was infected on day 7 poststimulation, after the peak of PHA-L-induced cell activation. On average, infection levels at this time point were ∼50% lower than peak infection levels (3 to 8%). Again, an RTI-PI combination was added 24 h p.i., and all infection cultures were continued until day 14 poststimulation in the continuous presence of an RTI-PI combination. At this time, all remaining actively infected, and therefore EGFP-positive, cells were removed by flow cytometry-based cell sorting. Immediately following the cell sorting, cells were transferred into fresh cell culture medium and stimulated with an anti-CD3/CD28 MAb combination, IL-2, or TNF-α. A fraction of each cell population was left untreated to serve as a negative control. Again, during this 24-h period, all cultures were continuously supplemented with the RTI-PI combination. Figure 7 depicts the results of experiments using PBMCs from four different donors. In the sorted T-cell populations that were infected on day 2 following activation, only minimal levels of activatable HIV-1 infection were detected (squares). In contrast, silent/latent infection events could be detected in T-cell cultures in which the infection was initiated after the peak of cellular activation. The level of reactivatable infection was dependent on the stimulus, with TNF-α activation providing no increase in HIV-1 levels. In contrast, the anti-CD3/CD28 MAb combination and IL-2 activation produced active HIV-1 infection as indicated by an increase in the expression of EGFP cells relative to untreated control cells. Despite high sorting purity (>99.9% EGFP negative), the controls exhibited increased levels of infection events in which spontaneous HIV-1 reactivation was observed, which is likely to be a result of the physical stress imposed upon the cells during the cell sorting procedure.

Although different stimuli were required, these experimental results are in line with the experimental results obtained using T-cell lines and suggest that high cellular activation levels prohibit the formation of silent HIV-1 infection. While useful to address questions concerning the establishment of HIV-1 latency, the system is likely to have some limitations. The finding that IL-2 (100 U/ml) as a single treatment efficiently activates HIV-1 infection suggests that the cells, while not having an activated phenotype (95% of the cultures were CD25 negative; small forward scatter/side scatter pattern), do not have a true memory T-cell phenotype either. Memory T cells have been reported to express only the low-affinity IL-2 receptor, and we are uncertain whether in vivo latent HIV-1 infection in memory T cells would be reactivated by the applied IL-2 concentrations. In addition, while we can state that infection after the peak of intracellular activation allowed for silent integration of HIV-1, we cannot extend the culture past the day 14 time point in the absence of IL-2 supplementation, in order to test whether the observed silent viral integration state would reflect truly latent infection events found in memory T cells in an in vivo setting.

Histone modifications, such as the generation of a repressive histone code, are key mechanisms for the control of cellular gene expression. Histone occupancy is also thought to protect silent cellular promoters from the nonspecific activity of DNMTs, and the absence of the appropriate histone composition could result in permanent transcriptional silencing of the unprotected promoter. Since we have demonstrated that silent integration is key to the establishment of latent HIV-1 infection (Fig. 2), one might assume that, similarly to cellular genes, a restrictive histone composition is formed at a silently/latently integrated HIV-1 promoter to (i) stabilize the silent promoter state and (ii) protect the promoter from nonspecific DNA methylation and subsequent permanent transcriptional silencing. Indeed, the formation of a repressive histone code has been associated with HIV-1 latency (23, 60, 61), and HDAC inhibitors have been reported to reactivate latent HIV-1 infection (49, 52, 60). These findings imply that latency establishment would be regulated by HDAC activity, and we postulated that a lack of HDAC activity would abrogate latency formation. We thus treated Jurkat T cells prior to infection with a single dose of the HDAC inhibitor NaBu (100 μM), valproic acid (300 μg/ml), or TSA (100 nM). These compounds were used at the maximum tolerated concentration without biasing the experimental outcome by affecting cell viability. We found that pretreatment of the cells with a single dose of an HDAC inhibitor (data for NaBu shown in Fig. 8) prior to infection increased the level of infection (28% active infection in control cells versus 37% infection in NaBu-treated cells), demonstrating that the compounds exerted activity at the concentration utilized. In the experiments, NaBu not only increased the level of active infection on day 3 but also increased the size of the generated latently infected cell population on day 37. If HDAC activity were essential for the establishment of latent infection events, then, in the presence of NaBu, we would have expected to see a reduction in the size of the latent reservoir that was found established on day 37 p.i. These data thus suggest that the formation of a restrictive histone code is not essential to establish latent HIV-1 infection. The observed increase in infectivity could be attributed to the ability of HDAC inhibitors to open heterochromatin, thereby allowing for integration of retroviruses into DNA areas that usually are not accessible for integration.

This unexpected finding led us to revisit previous findings that HDAC inhibitors reactivate latent HIV-1 infection in the absence of cellular activation in the cell clones generated under our experimental conditions. For this purpose, we generated ∼800 single-cell clones from the infection culture described for Fig. 1 and stimulated each clone with PMA, the HDAC inhibitor NaBu, the DNMT inhibitor 5-azacytidine, or combinations thereof. Of the ∼800 generated clones, 36 responded to PMA treatment with reactivation as indicated by an increase in the percentage of EGFP-positive cells on day 2 poststimulation (see Fig. 10B and C). None of the clones responded with HIV-1 reactivation to treatment with either the HDAC inhibitor NaBu (100 μM) or the DNMT inhibitor 5-azacytidine (1 μM) alone (data not shown).

For a more detailed analysis, we randomly selected seven of the latent clones that would respond to PMA stimulation with efficient HIV-1 reactivation (>80% EGFP-positive cells). For each of the selected cell clones, a dose-response curve for the HIV-1-activating agents TNF-α, PMA, and prostratin and for the HDAC inhibitors NaBu, valproic acid, and TSA was generated. Levels of HIV-1 reactivation as indicated by EGFP expression and cell viability were determined after 48 h using FCM analysis. Reactivation efficiency relative to drug toxicities for the cell-activating agents in comparison to the HDAC inhibitors is shown in Fig. 9. As expected, TNF-α activation efficiently reactivated latent HIV-1 infection at the population level and the onset of cytotoxicity directly correlated with the increase in active HIV-1 infection levels. Similar results were obtained for the HIV-1-activating phorbol esters PMA and prostratin. Throughout all latently infected cell clones, we found that the ability of HDAC inhibitors to reactivate latent HIV-1 infection was clearly inferior to that of cell-stimulating agents. No HIV-1-reactivating effect of HDAC inhibitors on latent HIV-1 infection was observed until the onset of massive compound-caused cell death. In all cell lines, the observed HDAC inhibitor-caused increase in cytotoxicity was disproportionate to the level of achievable reactivation, suggesting that a cell stress response, rather than changes in a restrictive LTR histone composition, may drive or at least be required for HIV-1 reactivation. The absence of an HDAC inhibitor effect on established latent infection in these cell lines suggests that a restrictive histone code is also not essential to maintain latent HIV-1 infection in these cells.

We next tested the influence of the DNMT inhibitor 5-azacytidine on HIV-1 latency establishment. Using the same experimental setup as used above for the HDAC inhibitors, we found no evidence that inhibition of DNMT activity would influence HIV-1 latency establishment (Fig. 10A). However, of the 36 latently infected T-cell clones generated, we identified 18 clones in which the addition of 5-azacytidine to the culture prior to stimulation with PMA would increase the level of HIV-1 reactivation, suggesting that a restrictive DNA methylation pattern had formed on the integrated latent viruses (Fig. 10C). We believe that this theoretical conflict, the lack of influence of DNMT inhibitors during silent integration and early latency establishment (up to day 41) and the formation of an inhibitory DNA methylation pattern at a later time point, is an artifact of the utilization of immortalized T-cell lines and needs to be carefully evaluated for its significance. Evidence for the idea that methylation patterns in our system could be an artifact comes from long-term culture experiments on our latently infected cell populations. We repeatedly observed that continuous growth of latently infected cell populations would decrease the amount of latently infected cells in these cultures without any signs that this decrease could have been the result of spontaneous reactivation and subsequent depletion of these cells from the culture. For example, the cell population depicted in Fig. 1 was initially characterized as holding 5% latently infected cells on day 64 p.i. (Fig. 1). In continuous culture, levels of latently infected cells in the population gradually declined to 3% over the next 6 months (data not shown). While we do not provide direct evidence, this is likely attributable to low-level de novo methylation, enabled by continuous cell division. However, cell replication, which allows for de novo methylation, would not be found in memory T cells, which are considered the in vivo reservoir of latent HIV-1 infection.

The inability to provide sufficient amounts of HIV-1 Tat and thus active transcription has been associated with the establishment of HIV-1 latency. We thus tested the question of whether Tat-mediated HIV-1 transcription activity was relevant for latency formation. For this purpose, we infected Jurkat T cells in the presence of an optimal dose of the HIV-1 transcription inhibitor Ro24-7429 (10 μg/ml) (26, 33). As expected, Ro24-7429 delayed the course of active infection but surprisingly did not alter peak infection levels (Fig. 11). Ro24-7429-mediated inhibition of HIV-1 expression also did not alter the size of the pool of latently HIV-1-infected cells in the infection culture, suggesting that lack of immediate active HIV-1 transcription does not favor latency establishment (Fig. 11). These results are consistent with the inability to generate stable latent infection events from previously active infection events (Fig. 3) and again argue against the establishment of a restrictive histone code as a prerequisite for HIV-1 latency.

In our experiments, we could not find any evidence for a role of a restrictive histone code for either latency establishment or maintenance. This is in line with previous findings by Han et al., who reported that >90% of all viral integration events in memory T cells of infected patients are found in actively expressed genes (30). Given these findings, we investigated whether the latent integration events in the clonal cell lines generated from our experimental system would be found integrated into actively transcribed host genes and thus reflect the in vivo situation. For this purpose, we used a modification of a PCR strategy described by Han et al. that would exclusively amplify viral sequences integrated into actively transcribed host genes but not viral mRNA products generated under the control of the viral LTR. This was achieved by utilizing 5′ primers that bind the LTR sequence prior to the transcription initiation site and 3′ primers that bind to the gag sequence prior to the splice donor site. Using this experimental strategy, we analyzed cellular mRNA from 13 of the latently infected cell clones described for Fig. 10B that had been archived for the presence of integrated viral genomes. All clones responded to PMA stimulation with reactivation levels of >85%, confirming that the cells were latently infected. Figure 12A shows the reactivation profile of five representative clones. In all 13 analyzed cell clones, the HIV-1 LTR sequence was found integrated into cellular mRNA, demonstrating that HIV-1 in our latently infected cell clones, similar to the in vivo situation, is integrated into actively expressed host genes (Fig. 12B). Combined with the inability of these clones to respond to HDAC inhibitor treatment with HIV-1 reactivation, these data provide quantitative support for the idea that transcriptional host gene interference may be the major mechanism to control HIV-1 latency.