Quiescence Promotes Latent HIV Infection and Resistance to Reactivation from Latency with Histone Deacetylase Inhibitors

ABSTRACT

Human immunodeficiency virus type 1 (HIV-1) establishes transcriptionally silent latent infections in resting memory T cells and hematopoietic stem and progenitor cells (HSPCs), which allows the virus to persist in infected individuals despite antiretroviral therapy. Developing in vitro models of HIV-1 latency that recapitulate the characteristics of latently infected cells in vivo is crucial to identifying and developing effective latency-reversing therapies. HSPCs exist in a quiescent state in vivo, and quiescence is correlated with latent infections in T cells. However, current models for culturing HSPCs and for infecting T cells in vitro require that the cells be maintained in an actively proliferating state. Here we describe a novel culture system in which primary human HSPCs cultured under hypothermic conditions are maintained in a quiescent state. We show that these quiescent HSPCs are susceptible to predominantly latent infection with HIV-1, while actively proliferating and differentiating HSPCs obtain predominantly active infections. Furthermore, we demonstrate that the most primitive quiescent HSPCs are more resistant to spontaneous reactivation from latency than more differentiated HSPCs and that quiescent HSPCs are resistant to reactivation by histone deacetylase inhibitors or P-TEFb activation but are susceptible to reactivation by protein kinase C (PKC) agonists. We also demonstrate that inhibition of HSP90, a known regulator of HIV transcription, recapitulates the quiescence and latency phenotypes of hypothermia, suggesting that hypothermia and HSP90 inhibition may regulate these processes by similar mechanisms. In summary, these studies describe a novel model for studying HIV-1 latency in human primary cells maintained in a quiescent state.

IMPORTANCE Human immunodeficiency virus type 1 (HIV-1) establishes a persistent infection for which there remains no feasible cure. Current approaches are unable to clear the virus despite decades of therapy due to the existence of latent reservoirs of integrated HIV-1, which can reactivate and contribute to viral rebound following treatment interruption. Previous clinical attempts to reactivate the latent reservoirs in an individual so that they can be eliminated by the immune response or viral cytopathic effect have failed, indicating the need for a better understanding of the processes regulating HIV-1 latency. Here we characterize a novel in vitro model of HIV-1 latency in primary hematopoietic stem and progenitor cells isolated from human cord blood that may better recapitulate the behavior of latently infected cells in vivo. This model can be used to study mechanisms regulating latency and potential therapeutic approaches to reactivate latent infections in quiescent cells.

INTRODUCTION

Human immunodeficiency virus type 1 (HIV-1) is known to establish a latent infection in some cells, in which the viral genome is integrated into the host cell genome but remains as a transcriptionally silent provirus (1, 2). Such latent infections are believed to contribute to the phenomenon in which HIV-1 infections in individuals treated with antiretroviral therapy that suppresses the plasma viral load to below detectable levels will rebound if therapy is interrupted (1, 3). The primary cellular target of HIV-1 is CD4⁺ T cells, and the most abundant reservoir of latent HIV proviruses is also likely to reside in these cells (1, 3). Notably, latency in CD4⁺ T cells is observed most frequently in quiescent resting memory T cells, and quiescence has been shown to correlate with HIV latency (4, 5). While T cells are the primary target of HIV-1 infection, other cell types have also been shown to be infected, including macrophages and hematopoietic stem and progenitor cells (HSPCs), and non-T cell sources of residual viremia in treated patients have been described (6–13).

Reactivation of transcriptionally silent proviruses and subsequent killing of cells harboring these proviruses by immune-mediated clearance or viral cytopathic effect comprise one approach to eliminating the residual virus present in individuals receiving antiretroviral therapy. Expression of HIV-1 genes is known to be regulated by a number of mechanisms that can be targeted for reactivation in vitro. NF-κB is essential to efficient transcription from the HIV long terminal repeat (LTR) (14), and NF-κB signaling can be induced by tumor necrosis factor alpha (TNF-α) stimulation (12) or protein kinase C (PKC) activation with agonists such as bryostatin (15). In addition to NF-κB, P-TEFb, which is composed of cyclin T1 and CDK9, acts with HIV Tat to promote RNA polymerase elongation during HIV-1 gene transcription (16, 17). P-TEFb is held in an inactive state by the 7SK snRNP but can be released from this complex and activated following treatment with hexamethylene bisacetamide (HMBA) (18). Thus, TNF-α, bryostatin, and HMBA each have the potential to reactivate latent HIV-1 infections, as demonstrated previously (12, 19). Other latency-reversing agents have also been proposed, including histone deacetylase inhibitors (HDACi), such as vorinostat (20, 21) and romidepsin (22–24), which are believed to achieve reactivation through chromatin remodeling.

Heat shock protein 90 (HSP90), a molecular chaperone of the heat shock family of proteins, is a known positive regulator of HIV-1 transcription. A growing body of literature supports the role of HSP90 in regulating HIV-1 gene expression, and the mechanisms of this effect are likely pleiotropic. HSP90 has been implicated in the activation of the NF-κB pathway (25), the formation of stable P-TEFb complexes (26, 27), and the formation of RNA polymerase II (Pol II) complexes in the cytoplasm (28). Hyperthermia has been shown to increase HIV-1 gene expression in persistently infected cell lines, in an HSP90-dependent manner, increasing the colocalization of HSP90 with the HIV-1 promoter (29–32). Inhibition of HSP90 by use of the specific inhibitor 17-N-allylamino-17-demethoxygeldanamycin (17-AAG) can prevent viral rebound in an HIV-1-infected humanized mouse model (30), and HSP90 inhibition is also linked to cell cycle arrest, similar to that which occurs in quiescent cells (33, 34). Thus, emerging evidence suggests that HSP90 may be a useful target in strategies to eliminate the latent reservoir of HIV-1.

To determine which reactivation strategies are most likely to be successful in vivo, it is important to develop in vitro models of HIV-1 latent infection. These in vitro systems must recapitulate the in vivo nature of the latent HIV-1 reservoir, including the diverse cell types that can harbor latent infections and the quiescent state of many cells that contain transcriptionally silent proviruses. While latent HIV-1 infection has been observed in several in vitro systems (19), there is a notable absence of primary cell models in which HIV-1 preferentially establishes a latent infection in quiescent cells. Furthermore, treatments that are effective in many of these models have failed to reduce the viral reservoir in vivo (35–40).

HSPCs reside in the bone marrow and are responsible for generating the hematopoietic cell compartment throughout the life of an individual. While both active and latent HIV-1 infections of HSPCs in vitro and from patient bone marrow have been described by some investigators (10–13), others have been unable to detect HIV-1 provirus in HSPCs from optimally treated HIV-infected donors (41, 42). Helping to resolve this apparent discrepancy, a recent publication by Sebastian et al. demonstrated that the frequency of HIV genomes in HSPCs from people is significantly lower than that in T cells and that prior negative studies lacked the necessary statistical power for reliable detection. Additionally, Sebastian et al. provided clear in vivo examples of infected HSPCs passing clonal defective genomes to differentiated progeny. Because the clonal genomes were defective, this observation could not have been attributed to coincident infection (43). Thus, there is evidence supporting the possibility that HSPCs form a reservoir of HIV in vivo.

The most primitive HSPCs, hematopoietic stem cells (HSCs), exist in a quiescent state and are long-lived and capable of self-renewal (44, 45). HSCs differentiate into more mature progenitor cells, which in turn continue to differentiate and give rise to the entire repertoire of mature hematopoietic cells (46). Differentiation from the most primitive to more mature HSPCs can be determined by expression of cell surface antigens, including CD133 and CD34 (47–49). In HSPCs purified from human cord blood (CB), the most primitive progenitors, including most stem cells, express both CD133 and CD34, whereas intermediate progenitors, such as common myeloid progenitors and megakaryocyte/erythrocyte progenitors, express reduced levels of CD133 but remain CD34⁺, making these surface markers a suitable choice for assessing HSPC differentiation (43).

HSPCs cultured in vitro differ from those in vivo in that the viability of the cells is dependent on the presence of growth factors in the culture medium. For this reason, previous work investigating latent infection of primary human HSPCs in vitro has been performed with cells cultured at 37°C in the presence of growth factors, including thrombopoietin, stem cell factor, insulin-like growth factor 1, and FLT3L (12, 50). Here we demonstrate that cells cultured under these conditions are actively proliferating and differentiating and thus do not recapitulate the quiescent state of HSPCs in vivo. Furthermore, we show that while latent infections are observed, these cells are susceptible to predominantly active infections with HIV-1. In contrast, we show that cells cultured in the presence of the same growth factors at 30°C are maintained in a quiescent state, with no loss of viability, and that these cells harbor predominantly latent HIV-1 infections. Additionally, we demonstrate that latent infections in the most primitive quiescent HSPCs are resistant to spontaneous reactivation and remain in a latent state for extended periods in culture. We further demonstrate that differentiation to more mature progenitors correlates with spontaneous reactivation from latency and that some stimuli that are sufficient for reactivation from latency in proliferating and differentiating HSPCS are not effective in quiescent HSPCs. Latency in this system is dependent on the active maintenance of latently infected cells in a quiescent state and is regulated postintegration. We have also shown that inhibition of HSP90 at standard temperatures recapitulates the quiescence and latency phenotypes observed under hypothermic conditions. In all, we identified and characterized two distinct models of HIV-1 latency by using primary human HSPCs maintained in a quiescent state in vitro. We further implicate HSP90 as an important modulator of HIV latency, and we provide evidence that quiescent cells may be more resistant to reactivation than previously believed.

RESULTS

HSPCs cultured under hypothermic conditions maintain a quiescent state.

To develop conditions that simulate the quiescent state typical of HPSCs in vivo, we examined the effect of hypothermic culture conditions, which we hypothesized would slow cellular proliferation and differentiation. Cellular proliferation was measured using PKH26, a membrane-binding dye that is diluted with each cell division, as depicted in Fig. 1A. We found that HSPCs cultured at 37°C proliferated at a significantly higher rate than those cultured at 30°C (Fig. 1B and C). Cells initially maintained at 37°C and switched to 30°C also proliferated consistently less than those maintained at 37°C throughout the 6 days of culture, suggesting that switching to 30°C after initial culture at 37°C could halt proliferation (Fig. 1B and C). Furthermore, cell counts were also measured after 3 days of culture at 37°C and 30°C, and the cells had expanded significantly more at 37°C (4.6-fold) than at 30°C (1.5-fold), reflecting what was observed with PKH26 staining (Fig. 1D) (P < 0.0001; Wilcoxon signed-rank test). Thus, HSPCs cultured at 30°C are maintained in a more quiescent state in which proliferation and expansion in culture are dramatically reduced, with no loss of viability (Fig. 1E).

FIG 1.

HSPCs cultured in vitro under hypothermic conditions are maintained in a quiescent state. (A) Schematic demonstrating the experimental process for panels B, C, F, and G. (B) Representative histogram for one donor, demonstrating the intensity of PKH26 staining in HSPCs following 6 days of culture as assessed by flow cytometry, where dilution of the PKH26 stain represents proliferation. Day 0, cells harvested immediately following PKH26 staining, prior to any dilution; unstained, cells never stained with PKH26 and harvested at 6 days postisolation. (C) Summary graph of flow cytometric data from 3 independent experiments performed as described for panel B, with the median fluorescence intensity of PKH26 normalized to that for condition A. (D) Summary graph depicting the fold change in cell number at each temperature over a 3-day culture period (****, P < 0.0001; Wilcoxon signed-rank test). (E) Summary graph depicting the viability of HSPCs, calculated as the percentage of events falling within both FSC/SSC and 7-AAD viability gates following 6 days of culture at the respective temperature. (F) Representative flow cytometric analysis of HSPCs cultured as shown in panel A and stained for the indicated surface markers. (G) Summary graph of flow cytometric data from 3 experiments performed as described for panel F, demonstrating the frequency of each of the indicated cellular subsets in the whole population. Mean values are shown, with error bars indicating standard deviations. (H) Schematic demonstrating the experimental process for panels I to K. (I) Summary graph of flow cytometric data from 3 experiments, demonstrating the frequency of each of the indicated cellular subsets in the whole population of HSPCs cultured at 30°C or 37°C postexpansion. Mean values are shown, with error bars indicating standard deviations. (J) Summary graph of flow cytometric data from HSPCs cultured as shown in panel H and harvested on day 4. HSPCs were stained with DAPI, and the frequency of each cell cycle phase was determined using FlowJo software (*, P < 0.05; Mann-Whitney tests). Each symbol represents an independent experiment using cells from a unique donor. (K) Summary graph of normalized flow cytometric data from panel J, displaying the ratio of cells in G₁ phase to those in S phase (*, P < 0.05; Mann-Whitney test).

HSPC proliferation is often linked to differentiation. To determine whether hypothermia also maintains HSPCs in an undifferentiated state, we stained cells for the HSPC markers CD133 and CD34. As HSPCs differentiate and become more mature progenitors, expression of CD133 is lost, followed by loss of CD34 expression in the most mature cells (43). We found that HSPCs maintained at 30°C from the time of isolation remained undifferentiated based on expression of both CD133 and CD34 (Fig. 1F and G). In contrast, HSPCs cultured at 37°C for as little as 2 days began to lose expression of CD133, and by day 6 postisolation, approximately half of the cells no longer expressed CD133 (Fig. 1F and G). Thus, reduced proliferation and expansion of HSPCs at 30°C are also accompanied by reduced differentiation.

Only a small number of CD133⁺ CD34⁺ cells can be isolated from a single donation of human cord blood, requiring an initial expansion of the cells to acquire sufficient cell numbers for downstream experiments. To verify that a quiescent state could be induced after HSPCs had been actively expanding, we performed a temperature shift experiment in which cells initially cultured at 37°C were shifted to the lower temperature (Fig. 1H). We found that switching HSPCs to 30°C after initially expanding them at 37°C effectively halted differentiation at the time of the switch (Fig. 1I). In contrast, HSPCs maintained at 37°C continued to differentiate; the frequency of CD133⁻ CD34⁻ cells was significantly higher than that at 30°C (Fig. 1I) (P < 0.05 for day 4, P < 0.01 for day 6, and P < 0.01 for day 8; Mann-Whitney tests), and the frequency of CD133⁺ CD34⁺ cells was significantly lower than that at 30°C (P < 0.05 for day 4, P < 0.05 for day 6, and P < 0.01 for day 8; Mann-Whitney tests). These data support the conclusion that HSPCs cultured at 37°C will continue to differentiate and proliferate, while HSPCs cultured at 30°C will enter a sustained quiescent state even after 4 days of expansion. Thus, this expansion phase was employed for the remainder of the study, as it allowed us to acquire sufficient cell numbers for downstream analyses.

To further characterize the quiescent state of HSPCs cultured at 30°C in vitro, we asked whether hypothermia led to arrest at a specific stage of the cell cycle. We found that HSPCs cultured at 37°C had significantly fewer cells in G₁ phase and significantly more cells in S phase than HSPCs cultured at 30°C (Fig. 1J) (P < 0.05; Mann-Whitney tests). After normalization, the ratio of G₁/S-phase cells was 2.2-fold higher in HSPCs cultured at 30°C than in those cultured at 37°C (Fig. 1K) (P < 0.05; Mann-Whitney test). Taken together, these data demonstrate that primary human HSPCs cultured at 30°C in vitro are maintained in a quiescent state with minimal proliferation, differentiation, and cell cycle progression.

Quiescence promotes HIV-1 latency in HSPCs.

To determine whether quiescence affected the frequencies of active and latent infections with HIV-1, we used a previously published system that efficiently detects the rate of latent infection in HSPCs established with a single-round green fluorescent protein (GFP) reporter virus (Fig. 2A) (12). In this assay system, infected cells expressing viral proteins are detectable by expression of a GFP-envelope fusion protein, while uninfected or latently infected cells remain GFP negative. At the time of infection, HSPCs were split into 37°C and 30°C groups as shown in Fig. 2B, and at 3 days postinfection (dpi), actively infected, GFP⁺ cells were removed by fluorescence-activated cell sorting (FACS) analysis. The integrase inhibitor raltegravir was added immediately after sorting to prevent de novo integration. At 24 h postsorting, cells were harvested, and the frequency of inducible latent infection was determined by subtracting the frequency of GFP⁺ cells in the raltegravir-only sample, defined as spontaneous reactivation, from the frequency of GFP⁺ cells under TNF-α-stimulated conditions (Fig. 2B and C). To ensure that reactivation was truly the result of integrated genomes that were not producing viral proteins at 3 days postinfection, we verified that raltegravir was indeed capable of completely blocking infection of HSPCs at both 37°C and 30°C (Fig. 2D).

FIG 2.

Assessing postintegration latency in HSPCs. (A) Schematic depicting the NL4-3-ΔGPE-GFP HIV-1 construct expressing GFP from the env open reading frame. (B) Schematic of the experimental setup for Fig. 3. (C) Representative flow cytometric plots of the latency reactivation assay used for Fig. 3, in which HSPCs infected at the indicated temperatures were sorted to remove actively infected cells and treated with TNF-α or a solvent control in the presence of raltegravir to ensure that the assay exclusively measured postintegration latency reactivation. (D) Representative flow cytometric plots for HSPCs infected as shown in panel B, with the addition of raltegravir at the time of infection. Numbers in the lower right corners represent frequencies of GFP⁺ cells at 3 days postinfection.

We found that HSPCs cultured at 30°C had a significantly lower frequency of active infection than inducible latent infection (Fig. 3A) (5.4-fold) (P < 0.0001; Wilcoxon signed-rank test), whereas the reverse was true for cells cultured at 37°C (Fig. 3A) (P < 0.001; Wilcoxon signed-rank test). Correspondingly, the frequency of active infection was significantly higher at 37°C than at 30°C (Fig. 3B) (4.9-fold) (P < 0.0001; Wilcoxon signed-rank test), and the frequency of inducible latent infection was significantly higher at 30°C than at 37°C (Fig. 3B) (1.8-fold) (P < 0.0001; Wilcoxon signed-rank test). Thus, hypothermia promotes latent HIV infection in HSPCs.

FIG 3.

Quiescent HSPCs are susceptible to predominantly latent HIV-1 infections in vitro. (A) Summary graph of flow cytometric data from 15 experiments as in Fig. 2, where active infection is the frequency of GFP⁺ cells at 3 dpi and inducible latent infection is the frequency of GFP⁺ cells at 4 dpi following 24 h of TNF-α and raltegravir treatment. For this analysis, the frequency of spontaneous reactivation occurring in control cells treated with raltegravir and run in parallel was subtracted. (***, P < 0.001; ****, P < 0.0001; Wilcoxon signed-rank test). (B) Summary graph of flow cytometric data as in Fig. 2, comparing total frequencies of active and inducible latent infections in HSPCs cultured at the indicated temperatures (****, P < 0.0001; Wilcoxon signed-rank test). (C) Summary graph depicting the frequency of spontaneous reactivation in raltegravir-only samples normalized to the frequency of inducible infection resulting from 24 h of TNF-α stimulation (****, P < 0.0001; Wilcoxon signed-rank test). (D) Summary graph showing spontaneous reactivation, normalized as described for panel C, in each subset of HSPCs at the indicated temperatures. Numbers above the symbols indicate fold reductions in spontaneous reactivation frequency (****, P < 0.0001; Wilcoxon signed-rank test) (n = 15 for panels D to G).

To further define the latency observed in HSPCs cultured at 30°C, we compared the frequencies of spontaneous reactivation from latency that occurred in the absence of additional stimuli under these two conditions. To facilitate comparison across multiple experiments, these results were normalized to the maximal reactivation induced by TNF-α. At 37°C, 42% of the reactivation that could be induced with TNF-α reactivated spontaneously within 24 h. Notably, at 30°C, only 15% of inducible proviruses reactivated spontaneously within 24 h (Fig. 3C) (P < 0.0001; Wilcoxon signed-rank test).

Differentiation is associated with spontaneous reactivation from latency.

Although increased latency was observed in HSPCs maintained at 30°C, it is important to note that this represents a heterogeneous population of various hematopoietic progenitor cells, which can be identified as primitive or mature based on expression of the surface markers CD133 and CD34 (43). To determine whether the differentiation state of the cell influenced active versus latent infection, we analyzed cells infected and sorted as described for Fig. 2B and compared the frequencies of spontaneous reactivation for each differentiation subset. To facilitate comparison across independent experiments, the frequency of spontaneous reactivation was normalized to the frequency of TNF-α-inducible infection in a sample treated in parallel. This analysis revealed that in each subset, proviruses in HSPCs maintained under hypothermic conditions were significantly less likely to reactivate spontaneously than those in actively proliferating HSPCs maintained at 37°C (Fig. 3D) (P < 0.0001 for each; Wilcoxon signed-rank tests). However, we also observed that among cells cultured at the reduced temperature, not all differentiation subsets were equally susceptible to spontaneous reactivation. The most undifferentiated progenitors (CD133⁺ CD34⁺) were almost 5-fold less likely to reactivate spontaneously at 30°C than at 37°C, while more differentiated progenitors were about 2-fold less likely to reactivate spontaneously. These data show that the most primitive progenitors harbor the latent proviruses that are least likely to reactivate in quiescent cells absent additional stimulation.

Resistance to reactivation is reversible with a temperature shift.

Resistance of latent infection to spontaneous reactivation in quiescent HSPCs could be the result of permanent differences established during integration in quiescent or proliferating cells, such as a preference for different integration sites in the different populations. However, it is also possible that latency is maintained by differences in quiescent and proliferating cells postintegration and thus can easily be reversed when a cell exits a quiescent state. To determine which of these possibilities contributed to the differences in latency observed in quiescent HSPCs compared to proliferating HSPCs, we performed a series of temperature shifts (Fig. 4A). After performing a FACS sort to isolate latently infected GFP⁻ cells, cells were split into 37°C and 30°C groups for reactivation. As demonstrated previously, cells maintained at 37°C had a higher frequency of spontaneous reactivation than cells maintained at 30°C (Fig. 4B) (P < 0.01; Mann-Whitney test). Cells infected at 37°C and switched to 30°C for reactivation, however, had a significantly reduced frequency of spontaneous reactivation compared to that for cells maintained at 37°C throughout, suggesting that a more stable form of latency could be established by allowing the cells to enter a quiescent state (Fig. 4B) (P < 0.01; Mann-Whitney test). Conversely, cells infected in a quiescent state at 30°C and subsequently switched to a proliferating state at 37°C had dramatically higher spontaneous reactivation frequencies, which were not significantly different from those for cells infected and reactivated at 37°C. Thus, under the conditions of our assay, the relative frequency of spontaneous reactivation is determined, at least in part, by postintegration cellular conditions that are readily reversible.

FIG 4.

Postintegration latency in quiescent cells is sustained for extended culture periods but easily reversible with removal from the quiescent state. (A) Schematic representation of the experimental workflow for panel B. (B) Summary graph of the frequencies of spontaneous reactivation 24 h after isolating GFP⁻ cells by FACS, normalized to TNF-α-inducible reactivation run in parallel, as shown in panel A (n = 5 or 6 distinct donors) (**, P < 0.01; Mann-Whitney tests). (C) Schematic representation of the experimental workflow for panel D. NT, no treatment. (D) Summary graph depicting the frequencies of inducible latent infection in HSPCs treated as shown in panel C. For this analysis, the frequency of spontaneous reactivation occurring in control cells treated with raltegravir and run in parallel was subtracted.

Latency in quiescent cells persists for extended culture periods.

To assess whether the increased latent infection observed in HSPCs at 30°C is stable over longer culture periods, latently infected cells were maintained at their respective temperatures in the presence of raltegravir until 7 days postinfection, at which point some were stimulated for 24 h with TNF-α (Fig. 4C). Indeed, we found that hypothermia-induced latency continued to be detectable at a higher frequency than when infected cells were cultured under standard conditions, even at this extended time point, for four of four donors (Fig. 4D). These data suggest that proviruses in quiescent cells will maintain latency for extended periods but will readily produce viral proteins upon stimulation with TNF-α or removal of the quiescent state.

Preferential establishment of latency under hypothermic conditions is not due to deficient expression of NF-κB.

Given that TNF-α is a potent reactivator of latent proviruses in HSPCs and acts by activating NF-κB, we asked whether a deficiency in steady-state NF-κB expression could explain the reduced HIV protein expression in HSPCs at 30°C. However, we found that expression and nuclear localization of NF-κB p65 were identical in HSPCs as assessed by Western blotting following culture at 37°C or 30°C as shown in Fig. 1H and harvesting at day 4 following the temperature split (Fig. 5A; the results are quantified in Fig. 5E). Additionally, HSPCs at 37°C and 30°C showed no differences in steady-state activity or responsiveness to TNF-α by p65 DNA-binding enzyme-linked immunosorbent assay (ELISA) (Fig. 5B). This suggests that baseline levels of NF-κB activity and responsiveness of the NF-κB signaling pathway to activation are identical in HSPCs cultured at 30°C and 37°C. Taken together, these data suggest that while NF-κB activation is sufficient for reactivation from latency, differences in NF-κB expression, DNA-binding capacity, and nuclear localization are not responsible for the increased susceptibility to latent infection of quiescent HSPCs.

FIG 5.

Expression and cellular localization of P-TEFb, NF-κB, and HSP90 fail to explain latency in quiescent HSPCs in vitro. (A, C, and D) Representative Western blots of lysates from HSPCs cultured at the indicated temperatures for 4 days postexpansion. The antibody to pCDK9 recognizes the activating phosphorylation at Thr186 (n = 4 distinct donors in separate experiments). p84 and GAPDH served as controls for separation of the nuclear and cytoplasmic fractions, respectively. Where indicated, fractions were loaded as serial dilutions for enhanced comparison. (B) Summary data from NF-κB p65 DNA-binding ELISA with whole-cell lysates from HSPCs cultured at the indicated temperatures for 4 days postexpansion and receiving no treatment (−) or stimulated with TNF-α (+) for 6 h (n = 2 or 3). Each symbol represents an independent experiment using cells from a unique donor. (E) Summary graphs of Western blot band intensity quantifications for Western blots performed as described for panels A, C, and D. The average pixel density of the band was normalized to the pixel density of the loading control for that sample and subsequently normalized to the relative intensity of the 37°C condition. Each symbol within a condition represents a band from a different gel for a distinct donor (n = 3 or 4 donors).

Preferential establishment of latency under hypothermic conditions is not due to deficient expression of P-TEFb.

P-TEFb, composed of CDK9 and cyclin T1, also plays an important role in expression of HIV genes through regulation of transcriptional elongation by RNA polymerase II (16, 17). However, no differences in the level of expression or cellular localization of cyclin T1 and an activated phosphorylated form of CDK9 (Thr186) were observed in HSPCs cultured at 37°C or 30°C (Fig. 5C; the results are quantified in Fig. 5E). These data suggest that a deficiency in P-TEFb is not responsible for regulating the increased susceptibility of quiescent HSPCs to latent infection with HIV-1.

HSP90 protein expression is unchanged in hypothermic HSPCs.

Heat shock protein 90 (HSP90) is a heat-sensitive chaperone that is known to promote HIV gene transcription. This effect is likely pleiotropic, as several mechanisms for this activity have been described, including via activation of the NF-κB pathway, stabilization of the P-TEFb complex, or formation of the RNA Pol II complex in the cytoplasm (26–31, 33, 34). In response to the growing body of literature supporting the role of HSP90 in promoting HIV gene transcription, we hypothesized that diminished HSP90 activity in response to hypothermia may regulate quiescence and latency in this model. As an initial assessment, we measured HSP90 protein expression in cytoplasmic and nuclear fractions of HSPCs cultured at 37°C or 30°C and observed no significant difference (Fig. 5D; the results are quantified in Fig. 5E). This result, however, did not exclude the possibility that HSP90 activity was reduced at the lower temperature despite a similar abundance of HSP90 protein.

Inhibition of HSP90 recapitulates hypothermia-induced quiescence.

Despite the lack of change in HSP90 protein levels, it remained possible that hypothermia reduced HSP90 functional activity. Thus, we asked whether inhibition of HSP90 with 17-AAG (tanespimycin) recapitulated the effects of hypothermia on the quiescent cell state of HSPCs. Remarkably, when we assessed the effect of HSP90 inhibition on cellular proliferation, differentiation, and viability, we found that 17-AAG treatment achieved a degree of quiescence in HSPCs at 37°C similar to that in HSPCs at 30°C. Inhibition of HSP90 reduced proliferation as assessed by PKH26 dilution after 2 and 4 days of culture (Fig. 6A, upper panels). These results were significant when compiled across three independent experiments (Fig. 6A, lower panels) (P < 0.05 for all comparisons; Mann-Whitney tests). Moreover, inhibition of HSP90 with 17-AAG limited the expansion of cells in culture to a degree similar to that with hypothermia; HSPCs cultured at 37°C in the absence of 17-AAG expanded dramatically, and this expansion was significantly greater than that observed in the presence of 17-AAG or under hypothermic conditions (Fig. 6B) (P < 0.001 for each; Mann-Whitney tests).

FIG 6.

Inhibition of HSP90 by use of 17-AAG recapitulates latency and quiescence phenotypes of hypothermia. (A) Representative flow plots (upper panels) and summary graphs (lower panels) of PKH26 staining following 2 (left) or 4 (right) days of culture at the indicated temperature, with or without 17-AAG, as indicated. For the summary graphs, values were normalized to the 30°C plus 17-AAG condition (P < 0.05 for all comparisons; Mann-Whitney tests). (B) Summary graph depicting the fold change in cell number for each condition over a 3-day culture period (***, P < 0.001; Mann-Whitney tests). (C) Summary plots of flow cytometric data, depicting the frequency of viable cells after 2 or 4 days under each condition, based on inclusion in both FSC/SSC and 7-AAD viability gates. Each symbol represents an independent experiment using cells from a unique donor. (D) Summary graph of flow cytometric data from 3 experiments, demonstrating the frequency of each of the indicated cellular subsets in the whole population after 3 days under the corresponding conditions. Mean values are shown, with error bars indicating standard deviations. (E and F) Summary graphs of frequencies of spontaneous reactivation for cells cultured at the indicated temperatures for 3 days and treated as indicated for 24 h. Values were normalized to the amount of inducible infection by dividing the frequency of GFP⁺ cells in the solvent control by the frequency in TNF-α-stimulated cells run in parallel. (E) Connecting lines indicate samples from the same donor (n = 5 or 6). (F) Comparison of the effects of 24 h of treatment with 17-AAG or hypothermia on spontaneous reactivation. Temperature switching was performed as shown in Fig. 4A (n.s., not significant; **, P < 0.01; Mann-Whitney tests).

Importantly, we demonstrated that treatment with 17-AAG was minimally toxic over the course of 4 days (Fig. 6C). In addition, we found that inhibition of HSP90 activity resembled the effect of hypothermia in that it slowed differentiation of HSPCs at 37°C, leading to a significant increase in the proportion of the most undifferentiated progenitors (CD133⁺ CD34⁺) and a significant decrease in the proportion of the most differentiated progenitors (CD133⁻ CD34⁻) after 3 days (Fig. 6D) (P < 0.01; Wilcoxon signed-rank test). Taken together, these data suggest that inhibition of HSP90 at standard temperatures is sufficient to recapitulate the quiescence observed in HSPCs cultured at hypothermic temperatures.

Inhibition of HSP90 suppresses spontaneous but not TNF-induced reactivation from latency.

Having determined that HSP90 inhibition was sufficient to recapitulate the quiescent cellular state observed in hypothermic HSPCs, we asked whether this correlated with reduced spontaneous reactivation from latency. We found that similar to what was observed under hypothermic conditions, 17-AAG reduced spontaneous reactivation from latency for 4 of 5 donors at 37°C, although this trend was not statistically significant (Fig. 6E) (P = 0.07). Consistent with hypothermia and 17-AAG acting on the same factor, we found that spontaneous reactivation from latency at 30°C remained low and was not significantly affected by inhibiting HSP90 (Fig. 6E). While the frequency of spontaneous reactivation was higher at 37°C with HSP90 inhibition than at 30°C, it was not significantly different from that observed in cells infected at 37°C and shifted to 30°C for the 24-h reactivation period, as in Fig. 4A (Fig. 6F) (P = 0.42), suggesting that quiescence following inhibition of HSP90 is similar to that with hypothermia in suppressing postintegration spontaneous reactivation of latent proviral genomes.

Latently infected quiescent HSPCs are resistant to reactivation from latency by HDACi and HMBA.

As latently infected cells likely exist in a quiescent state in vivo, it was important to assess whether quiescent cells differed from actively proliferating cells in their sensitivity to reactivation by chemical latency reactivators currently being used in therapeutic trials. Indeed, we observed minimal reactivation by the HDACi vorinostat and romidepsin and by the P-TEFb activator HMBA under hypothermic conditions, whereas there was potent reactivation with these compounds at 37°C (Fig. 7A). In a compiled analysis of data from multiple independent experiments, vorinostat, romidepsin, and HMBA were inefficient at reactivating virus from latency at 30°C. Indeed, romidepsin and HMBA did not significantly reactivate virus above the solvent control level at 30°C despite potent activity at 37°C (Fig. 7B). (To facilitate comparison across multiple experiments, the frequency of inducible genomes [GFP⁺ cells] observed under each condition was normalized to the frequency of TNF-α-inducible proviruses observed in paired samples. For this analysis, the frequency of spontaneous reactivation observed in paired solvent control samples was subtracted prior to normalization.) These results suggest that latent genomes in quiescent cells differ dramatically from those in proliferating cells in their ability to respond to HDACi and P-TEFb stimulation.

FIG 7.

Latent HIV-1 infections in quiescent HSPCs are resistant to reactivation by HDAC inhibitors and HMBA but can be reactivated with bryostatin. (A) Representative flow cytometric plots of latently infected HSPCs treated for 24 h with the indicated latency-reversing compounds, as in Fig. 2. The frequency of GFP⁺ HSPCs is depicted in bold at the bottom right of each plot, with the mean fluorescence intensity (MFI) of GFP in the GFP⁺ cells shown in italics at the top right (n = at least 4 experiments). (B) Summary graphs of flow cytometric data as described for panel A. Spontaneous reactivation, as assessed by the frequency of GFP⁺ cells in the solvent control, was subtracted prior to normalization to the frequency of GFP⁺ cells in the TNF-α-stimulated sample (n.s., not significantly different from the solvent control; ****, P < 0.0001; Mann-Whitney tests). (C) Summary graphs of flow cytometric data as described for panel A, depicting the GFP MFI in the indicated GFP⁺ gate for each condition normalized to that of TNF-α-stimulated HSPCs (*, P < 0.05; **, P < 0.01; ****, P < 0.0001; Mann-Whitney tests). (D) Western blot of lysates of HSPCs cultured at the indicated temperatures for 3 days postexpansion and then treated with DMSO (solvent) or vorinostat and lysed at the indicated time posttreatment. (E) Quantification of the bands in panel D, performed by measuring the pixel density of each band for acetylated histone H4 and dividing it by that of total histone H4 for each sample. Data were normalized to the baseline proportion of acetylated histone H4 2 h after treatment with the solvent control for the respective temperature. (F) Schematic of the experimental setup used for panel G. (G) Summary graph of flow cytometric data from cells treated as shown in panel F, stimulated with the indicated reactivation regimen for 24 h, adjusted for spontaneous reactivation, and normalized as described for panel B. Columns indicate means, and error bars indicate standard deviations (n = at least 4) (n.s., not significant; *, P < 0.05; ****, P < 0.0001; Mann-Whitney tests). Solvent, matched DMSO sample.

In contrast, bryostatin, which acts by activating PKC and then NF-κB, was effective under both conditions. Additionally, while HDACi and HMBA were not sufficient for reactivation of quiescent HSPCs from latency, they were still capable of enhancing the reactivation frequency in combination with bryostatin, suggesting that the drugs are functional at 30°C (Fig. 7A and B).

The median fluorescence intensity (MFI) of GFP in GFP⁺ cells was also measured and normalized to that in TNF-α-treated cells to determine the level of HIV protein expression in reactivated cells. This measurement reflects the amount of protein expressed on a per-cell basis. Although most treatments had no appreciable effect on GFP MFI (Fig. 7A) (compiled data not shown), we observed that HMBA increased GFP expression in combination with bryostatin over that observed with bryostatin or HMBA alone, even at 30°C, where HMBA alone had no positive effect on reactivation (Fig. 7C). These data demonstrate that while HMBA was not sufficient for enhancement of protein expression in quiescent HSPCs, it was still capable of this activity in combination with bryostatin, indicating that the drug had activity at this temperature but was ineffective in the absence of additional stimulation.

Similarly, the failure of vorinostat to reactivate latent genomes in quiescent cells was not due to a loss of activity under hypothermic conditions, as treatment of HSPCs with the HDACi vorinostat efficiently induced histone H4 acetylation with kinetics similar to that in HSPCs at both temperatures (Fig. 7D and E). No changes in acetylation were observed following treatment with the solvent control, confirming that the increased acetylation was due to vorinostat HDACi activity and verifying that the drug was capable of inducing chromatin remodeling at 30°C. Taken together, these data suggest that HDACi and HMBA are active at hypothermic temperatures in HSPCs but that their activity is not sufficient to induce reactivation from latency in quiescent cells without additional stimulation.

To determine whether the reduced ability of HDACi and HMBA to reactivate postintegration latency under hypothermic conditions was reversible, we performed temperature shift experiments as depicted in Fig. 7F. We observed that cells reactivated for 24 h at 30°C responded similarly regardless of whether they were infected at 37°C or 30°C, with low responsiveness to HDAC inhibitors or HMBA and high responsiveness to bryostatin. In comparison, cells reactivated at 37°C in a proliferative state were responsive to all reactivation treatments, regardless of whether they were infected at 37°C or 30°C (Fig. 7G). These data suggest that mechanisms preventing reactivation with HDACi or HMBA in quiescent cells respond rapidly to changes in the quiescent state of the cell and are regulated postintegration.

Quiescence mediated by HSP90 inhibition also prevents reactivation with HDACi and HMBA.

We further hypothesized that quiescence induced by inhibition of HSP90 at 37°C would recapitulate the reactivation profile of hypothermia-induced quiescence. Indeed, HSPCs reactivated with these approaches in combination with 17-AAG yielded a reactivation profile similar to that at 30°C (Fig. 8). Addition of 17-AAG significantly reduced the effects of vorinostat, romidepsin, and HMBA at 37°C but had no effect on reactivation with bryostatin (Fig. 8) (P < 0.01; Mann-Whitney tests). In contrast, addition of 17-AAG to HSPCs infected and reactivated at 30°C had no significant effect, maintaining the decreased responsiveness to vorinostat, romidepsin, and HMBA compared to that of HSPCs at 37°C without affecting reactivation with bryostatin (Fig. 8) (P < 0.01; Mann-Whitney tests). Inhibition of HSP90 can therefore recapitulate the reduced frequency of spontaneous reactivation and reduced susceptibility to certain reactivation approaches observed for latent HIV-1 proviruses at 30°C, supporting the hypothesis that quiescence leads to greater resistance to reactivation from latency.

FIG 8.

Quiescence following inhibition of HSP90 restricts reactivation in the same way as that for hypothermia-induced quiescence. The summary graph shows flow cytometric data from cells cultured at the indicated temperature for 3 days and treated with the indicated latency reactivator for 24 h, as shown in Fig. 2. Background reactivation in the DMSO solvent control was subtracted from the value for each sample prior to normalization to the frequency of GFP⁺ cells in the TNF-α-stimulated control. Columns indicate means, and error bars indicate standard deviations (n = at least 3) (n.s., not significant; **, P < 0.01; Mann-Whitney tests).

DISCUSSION

Here we have demonstrated that culturing primary human HSPCs at 30°C maintains these cells in a quiescent state characterized by lower rates of proliferation, differentiation, and progression beyond the G₁ phase of the cell cycle. Furthermore, we showed that these quiescent HSPCs are susceptible to infections with HIV-1 that are predominantly latent, while proliferating and differentiating HSPCs are susceptible to predominantly active infections. Thus, culturing HSPCs at 30°C as a means of inducing quiescence represents a novel in vitro model of HIV-1 latency that uses primary human cells. We have further shown that latency in this model system is maintained for extended periods in culture, is resistant to spontaneous reactivation, and can be reactivated by NF-κB activation in response to TNF-α stimulation. In addition, the most undifferentiated progenitor cells found in this model were the most resistant to spontaneous reactivation, while differentiation to more mature progenitors correlated with viral protein expression. This suggests that primitive HSPCs maintained in a quiescent state without undergoing differentiation may harbor latent HIV proviruses for long periods. Furthermore, the fact that quiescence-induced latency is reversible in this model of postintegration latency means that our results cannot be explained by integration site differences. Instead, our results support the hypothesis that quiescence-induced latency is maintained by an unknown factor with activity that is modulated by the cellular state.

In an effort to determine the mechanisms regulating the latency observed in this model, we demonstrated that differences in NF-κB and P-TEFb expression, activation, and cellular localization were not responsible for the increased frequency of latent infection at 30°C. Interestingly, we found that HMBA treatment, which presumably activates P-TEFb, and chromatin remodeling with HDAC inhibitors enhanced viral protein expression in quiescent HSPCs only in combination with bryostatin, a PKC agonist, although they were functional in the absence of bryostatin in proliferating cells. These data suggest that either a factor downstream of PKC stimulation or a deficiency of some factor that can be overcome by PKC stimulation is responsible for maintaining latency in quiescent HSPCs but is functional at some level in the absence of PKC activation in proliferating HSPCs. Although NF-κB would seem to be a likely candidate, we found no differences in NF-κB expression, localization, or activity in HSPCs at 37°C or 30°C. Future studies will be needed to confirm the existence and identity of such a factor and to clearly define the mechanism promoting latency in this model. The determination of this mechanism may shed light on the factors regulating whether a particular cell will maintain a latent infection in vivo and may guide the development of targeted therapies for inducing HIV gene expression from latent proviruses.

The observation that quiescent HSPCs are resistant to reactivation by previously described latency-reversing agents that are effective in actively proliferating HSPCs calls into question whether these therapies will be effective at reactivating latent reservoirs of virus in HSPCs in vivo, where the cells harboring latent proviruses tend to be maintained in a quiescent state. This model system may allow for more informative testing of latency-reversing agents in vitro by establishing a higher threshold for reactivation. In fact, the reactivation profile observed in hypothermic HSPCs closely resembles that of the Greene and Planelles models of HIV latency, which use T cells maintained in a relatively quiescent state, while other models behave more similarly to proliferating HSPCs (19). Notably, previous clinical trials with the HDAC inhibitors vorinostat and romidepsin failed to reduce the size of the HIV-1 latent reservoir in vivo, consistent with the inability of these compounds to induce reactivation in quiescent HSPCs (35, 36). Further studies will be needed to determine whether the requirements for reactivation in this quiescent model of latency more accurately reflect those observed in vivo than those of other model systems do, including those using actively proliferating HSPCs cultured at 37°C.

We also showed that inhibition of HSP90 is sufficient to induce quiescence and regulates HIV-1 latency similarly to hypothermia in this model. While differences in HSP90 protein expression were not observed, it is still possible that a deficiency in HSP90 activity under hypothermic conditions is responsible for the observations we described. HSP90 has been shown to be a positive regulator of HIV gene expression in numerous studies (26–34), and the mechanism of this activity is likely pleiotropic. Our findings further implicate HSP90 as potentially playing an important role in regulating HIV-1 latency, drawing a novel connection between quiescence-induced latency and HSP90 inhibition. Determining the mechanisms modulating latency downstream of hypothermia and HSP90 inhibition, which may be identical based on their numerous similarities, warrants further investigation. It will be important to determine if these mechanisms are also involved in maintaining latency in quiescent resting memory T cells, which would enhance their utility as targets for reactivation of the latent reservoir in HIV-infected individuals.

We observed that latent proviruses in the least differentiated HSPCs at 30°C were unlikely to reactivate spontaneously and that spontaneous reactivation correlated with differentiation to more mature progenitors. HSCs are long-lived and are capable of self-renewal. This leads to speculation that a quiescent HSC in vivo could be infected with HIV-1 and maintain the provirus in a latent state. This latent provirus could remain stable without reactivating spontaneously for long periods, only reactivating upon differentiation of the cell into a more mature progenitor cell. As primitive HSCs persist without differentiating throughout the life of an individual, it is possible that these cells represent a long-lived reservoir of latent HIV-1 infection that is resistant to spontaneous reactivation but continues to contribute to the pool of replicating virus as HSC progeny differentiate and undergo spontaneous reactivation. The possible existence of HSCs as a reservoir of latent virus in HIV-1-infected individuals would represent an additional barrier to the development of a cure for HIV, as reactivation-based approaches would be required to reactivate latent proviruses in both T cells and quiescent HSCs. While the existence of such a reservoir of latent HIV-1 proviruses in HSPCs has been controversial, a recent publication by Sebastian et al. (43) provides compelling evidence supporting the existence of rare proviruses in bone marrow HSPCs from optimally treated donors that cannot be attributed to T cell contamination. Furthermore, the low rate of infection observed explains prior negative studies, which lacked statistical power to reliably detect the low frequency of latently infected HSPCs.

Several future lines of investigation will be of interest to expand on the findings presented here. While we used HIV virions pseudotyped with vesicular stomatitis virus glycoprotein (VSV-G) to infect HSPCs in this study, characterizing the infection of quiescent HSPCs with full-length, wild-type HIV-1 will be important to better assess how infection of these cells proceeds in vivo. Furthermore, while HSPCs likely contribute to the latent reservoir of HIV-1, the major reservoir is in resting memory T cells. Determining whether hypothermia or the underlying mechanisms regulating latency in quiescent HSPCs can be used to enhance T cell models of HIV-1 latency warrants further investigation. Initial attempts suggest that differences in the cell biology of T cells cause them to respond poorly to hypothermia (data not shown). The model system described here, in combination with improving models of T cell latency, can serve as an important tool to identify effective therapies and gain a deeper understanding of the mechanisms regulating the establishment and maintenance of latency in quiescent cells.

MATERIALS AND METHODS

Ethics statement.

Anonymized whole umbilical cord blood (CB) from uninfected donors was obtained from the New York Blood Center.

Isolation and culture of HSPCs from human cord blood.

HSPCs were isolated from whole umbilical cord blood (New York Blood Center) as previously described (12). Briefly, cord blood mononuclear cells (CBMCs) were isolated from cord blood by Ficoll-Paque (GE Healthcare) density gradient centrifugation and either frozen in bovine serum albumin (BSA) (7.5% in phosphate-buffered saline [PBS]; Gibco) and dimethyl sulfoxide (DMSO) (10%; Sigma-Aldrich) in liquid nitrogen or used immediately. CBMCs were then adherence depleted in Stemspan II medium for 2 h, and nonadherent cells were purified for CD133⁺ cells by magnetic sorting using a CD133 microbead kit (Miltenyi Biotech) according to the manufacturer's protocol, with the modification that 1.5 times the recommended number of beads was added to increase the purity of the sort. Purity following the magnetic sort was >92% CD133⁺ cells, which were also unanimously positive for CD34 expression. Purified HSPCs were then maintained in Stemspan II medium supplemented with 100 ng/ml stem cell factor, 100 ng/ml thrombopoietin, 100 ng/ml Flt3 ligand (all from Stemcell Technologies), and 100 ng/ml insulin-like growth factor binding protein 2 (R&D Systems) (STIF medium) at 37°C or 30°C with 5% CO₂.

Flow cytometry surface staining.

Antibodies with binding specificity for the following surface antigens were used for flow cytometry: CD133 (phycoerythrin [PE] conjugated; Miltenyi Biotec) and CD34 (fluorescein isothiocyanate [FITC] or allophycocyanin [APC] conjugated; Miltenyi Biotec). Cells were stained with 2 μg/ml 7-aminoactinomycin D (7-AAD; Calbiochem) to exclude dead cells. For staining of surface proteins, cells were suspended in FACS buffer (2% fetal bovine serum [FBS], 1% human serum, 2 mM HEPES, 0.025% sodium azide in PBS) with 7-AAD and antibodies (CD34 and CD133), incubated on ice for 15 min, washed, and then fixed in 2% paraformaldehyde in PBS. Flow cytometry data were collected with a BD FACSCanto cytometer or a BD FACScan cytometer with a Cytek 6-color upgrade and analyzed with FlowJo software.

PKH26 proliferation assay.

HSPCs were stained with PKH26 cell membrane-binding dye (Sigma-Aldrich) according to the manufacturer's protocol immediately following isolation of CD133⁺ cells. HSPCs were maintained as described above, and the intensity of PKH26 staining was assessed by flow cytometry using a BD FACSCanto cytometer. Histograms were generated, and the median fluorescence intensity of PKH26 was calculated using FlowJo software.

DAPI cell cycle analysis.

Total cellular DNA content was assessed by staining HSPCs with DAPI (4′,6-diamidino-2-phenylindole; Thermo Scientific) and determining the intensity of DAPI staining by flow cytometry. HSPCs were harvested, washed twice with PBS, fixed on ice in 70% ethanol in PBS, washed with PBS, permeabilized, and stained with 1 μg/ml DAPI in 0.1% Triton X-100 (Fisher Biotech) in PBS for 30 min on ice. Data on the intensity of DAPI staining were collected using a BD FACSCanto cytometer, and cell cycle analysis was performed using FlowJo software. Following doublet exclusion, the DAPI signal intensity was plotted on a linear scale, and the proportion of singlets in G₁, S, or G₂/M phase was determined by averaging the results from the Watson (pragmatic) and Dean-Jett-Fox models, with the G₂ coefficient of variation (CV) set equal to the CV for G₁.

Viral constructs and HIV-1 infection of HSPCs.

HSPCs were infected with the HIV-1 construct NL4-3-ΔGPE-E-GFP (12) after 4 days of expansion at 37°C following isolation of CD133⁺ cells. Infectious viral supernatants were produced by transfection of the proviral NL4-3-ΔGPE-E-GFP plasmid into 293T cells by use of polyethylenimine. The proviral plasmid was cotransfected with a plasmid containing vesicular stomatitis virus glycoprotein (VSV-G) to generate VSV-G-pseudotyped viral particles and with a helper plasmid (p-CMV-HIV) which encodes the necessary structural proteins for virion formation that are absent in NL4-3-ΔGPE-E-GFP. Culture supernatants from these 293T cells were collected at 48 to 72 h posttransfection, filtered through a 0.4-μm syringe filter (GE Healthcare), and frozen at −80°C. 293T cells were cultured in Dulbecco's modified Eagle medium (DMEM; Gibco) supplemented with 10% fetal bovine serum (FBS; Gibco), 1 U/ml penicillin, 1 μg/ml streptomycin, 292 μg/ml glutamine (Gibco), and 56 μg/ml Plasmocin (Invivogen) (D10 medium).

HSPCs were infected by spin inoculation at 1,049 × g for 2 h at room temperature. Mock-infected controls were treated with D10 culture medium. Following spin inoculation, the viral supernatants were removed, and HSPCs were resuspended in STIF medium. Cells were then split and maintained at 37°C or 30°C in STIF medium immediately postinfection.

FACS sorting and reactivation studies.

HSPCs infected with NL4-3-ΔGPE-E-GFP were sorted for GFP⁻ cells by FACS using a FACSAria III (BD Biosciences) cytometer to remove GFP⁺ (actively infected) cells and to obtain a pure population of GFP⁻ (uninfected or latently infected) cells. GFP⁺ and GFP⁻ populations were defined using mock-infected HSPCs.

Following isolation of GFP⁻ HSPCs, 70,000 to 100,000 cells were immediately resuspended in 200 μl Stemspan II medium supplemented with 8 μM raltegravir (Selleck Chemicals) in all experiments, along with one of the following treatment conditions: 3 ng/ml tumor necrosis factor alpha (TNF-α; R&D Chemicals), 1 μM vorinostat (Vor; Cayman Chemical), 25 nM romidepsin (Rom; Selleck Chemicals), 10 mM hexamethylene bisacetamide (HMBA; Sigma-Aldrich), 2.5 nM bryostatin-1 (Bryo; Sigma-Aldrich), or 750 nM 17-N-allylamino-17-demethoxygeldanamycin (17-AAG; Selleck Chemicals). Combination treatments used the same concentrations of the compounds as those in individual treatments. Solvent controls were performed with cells cultured in Stemspan II medium supplemented with raltegravir and either water or DMSO matched to the concentration in the treated samples. Reactivation observed in solvent controls was termed spontaneous reactivation and was subtracted from calculations of the frequency of inducible latent infection or the frequency of reactivated cells with different treatments. Treatments were performed immediately following the FACS sort, and cells were harvested for flow cytometric analysis after 24 h.

HSPCs cultured for extended periods following the FACS sort were cultured in STIF medium supplemented with 8 μM raltegravir until 7 days postinfection (dpi), at which point cells were resuspended in Stemspan II medium supplemented with 8 μM raltegravir and stimulated with TNF-α or the solvent control for 24 h as described above.

Nuclear and cytoplasmic fractionation and Western blotting.

HSPCs were expanded for 4 days at 37°C and then split into 37°C and 30°C groups for 4 additional days. Cells were harvested, and nuclear and cytoplasmic fractions were isolated using a nuclear extraction kit (Active Motif) according to the manufacturer's protocol. Nuclear and cytoplasmic extracts were loaded as serial dilutions into wells of a polyacrylamide gel for Western blotting to enhance comparisons of protein expression levels. Western blotting was performed using antibodies directed against the following proteins: NF-κB p65 (clone 572; Invitrogen), pCDK9 (with activating phosphorylation on Thr186; Cell Signaling Technology), cyclin T1 (Santa Cruz Biotechnology), glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (Abnova), and p84 (Abcam) (the last two as cytoplasmic and nuclear protein controls, respectively). Horseradish peroxidase (HRP)-conjugated secondary antibodies were used (Invitrogen), and the signal was detected using Pierce ECL (Thermo Scientific) or Amersham ECL Prime (GE Healthcare) reagents.

For acetylated histone H4 Western blots as shown in Fig. 7D and E, whole-cell lysates were obtained using blue loading buffer (Cell Signaling Technology) according to the manufacturer's protocol, and membranes were probed with primary antibodies against the following proteins: histone H4 (pantropic) (04-858; Millipore) and acetyl histone H4 (Lys12) (04-119; Millipore). Secondary antibodies and detection reagents were as described above.

NF-κB p65 ELISA.

Whole-cell lysates were obtained from HSPCs after 8 days in culture by use of a TransAM NF-κB kit (Active Motif) according to the manufacturer's protocol. Protein concentrations were determined by Bradford assay, and equal protein amounts were loaded into each ELISA reaction mixture. ELISA was performed with a TransAM NF-κB kit (Active Motif) according to the manufacturer's protocol.

Cell counts.

For cell counting experiments, CD133⁺ HSPCs were isolated as described above and expanded for 4 days in STIF culture medium at 37°C. Cells were then split into 37°C and 30°C groups, with or without 17-AAG (750 nM) (Selleck Chemicals) added to the STIF culture medium. Cells were sampled 2, 3, and 4 days after splitting, and live cells were counted with a hemocytometer following trypan blue (Gibco) staining.

Quantification of Western blot results.

Western blot results were quantified using Photoshop, by determining the average pixel density in a box of equal size overlaid on each band from a single, unedited film displaying a single gel. No quantification comparisons were made for bands on different films or gels at any point.

Statistical analyses.

All statistical analyses were performed using GraphPad Prism software as described in the figure legends for each experiment.