Highly active antiretroviral therapy can efficiently suppress human immunodeficiency virus type 1 (HIV-1) replication to below the detection limit. However, even after years of effective viral suppression, cessation of therapy results in the immediate rebound of viremia. During treatment, viral infection is thought to be primarily sustained by a long-lived reservoir of latently infected CD4+
memory T cells (13
). As a result of the long life span of memory T cells that serve as cellular hosts, the latent HIV-1 reservoir is extremely stable. Details of its half-life (t1/2
) are still discussed controversially, with measured t1/2
being up to ∼40 months (20
). At this rate, the natural decay of a reservoir consisting of only 1 × 106
cells would take as long as ∼70 years. Thus, as natural depletion of the latent reservoir is unlikely to occur during the lifetime of an infected patient, HIV-1 latency is believed to represent the principal obstacle to curative AIDS therapy (13
To understand the molecular basis of HIV-1 latency, early studies were performed in latently HIV-1-infected transformed clonal cell lines such as ACH-2, J1.1, U1, and OM-10.1 (3
). Studies of these cells and other systems proposed a role for the site of viral integration (62
), for cellular proteins (9
), for viral proteins (1
), and for histone acetylation or DNA methylation in regulating HIV-1 latency (3
The insights gained from these and other studies produced a model of HIV-1 latency that suggests that the host cell, which initially exhibits a minimum level of activation sufficient to promote infection, returns to a quiescent state (30
). Since the virus is dependent on the availability of certain cellular key transcription factors for active gene expression, these key transcription factors are no longer available to the virus as the cells return to a quiescent state, and viral gene expression is shut down prior to the onset of viral cytopathicity or immune clearance. Integration into genome sites that are not supportive of viral transcription could favor this process (51
). The latent state of the integrated provirus is then stabilized by the establishment of a suppressive histone code, in particular at the viral promoter (23
Based on this molecular understanding of HIV-1 latency, several attempts to therapeutically deplete the latent HIV-1 reservoir have been previously made. The underlying concept of these strategies has been to activate the integrated but transcriptionally silent viral promoter. This was attempted either by stimulation of the infected cells (interleukin 2 [IL-2] or anti-CD3 monoclonal antibody [MAb] OKT3) (17
) or by triggering changes in the histone composition at the viral promoter using histone deacetylase (HDAC) inhibitors (e.g., valproic acid) to favor viral transcription in the absence of cellular activation (17
). These protocols have not resulted in a reduction of the size of the latent reservoir, or the clinical significance of the reported reduction has been disputed (54
). While some evidence for the role of histone modifications in HIV-1 latency has been presented in vitro, a more recent comprehensive investigation of HIV-1 integration events in patients, by J. Siliciano and coworkers, is somewhat in conflict with the idea that latency is governed by a restrictive histone code. In this study, Han et al. demonstrate that >95% of all infection events found in memory T cells of infected patients are located in actively expressed genes (30
) and are thus integrated in a DNA environment that is unlikely to allow for the formation of a stable suppressive histone code at the latent HIV-1 promoter. Indeed, in a subsequent study, in which the orientation-dependent regulation of HIV-1 gene expression by transcriptional interference was analyzed, no evidence for the establishment of a particular restrictive histone code was provided (31
). At the same time, a second study demonstrated the importance of transcriptional interference of the host gene with the integrated virus as a mechanism to stabilize the latent viral expression state (43
We add to these most recent findings on the role of host gene transcriptional interference as a governing factor for HIV-1 latency by demonstrating that histone deacetylation or DNA methylation is not important for the establishment of latent infection events. Our results reveal that the decision whether a latent infection event is established is a function of the availability of NF-κB at the time of infection. Transcriptional silent integration is a prerequisite for the establishment of a latent infection event, which is then maintained by transcriptional interference. Latent infection events that are governed by transcriptional interference are resistant to reactivation by HDAC or DNA methyltransferase (DNMT) inhibitors. We here discuss how these findings are complementary to current ideas on HIV-1 latency and the possible consequences of these findings for therapeutic depletion of the latent HIV-1 reservoir.
Based on our results, we propose a model for HIV-1 latency in which we differentiate between factors governing latency establishment and mechanisms controlling latency maintenance. According to our data, whether infection occurs in a transcriptionally active or latent state is decided upon viral integration and is a strict function of the availability of NF-κB for binding to the viral promoter at the time of infection. Below a certain threshold of NF-κB availability, the virus will integrate in a transcriptionally silent state (silent integration), which is a prerequisite for latency establishment. Once established, latent infection is maintained by transcriptional interference, as latent viruses are generally found integrated into actively expressed genes and we find no evidence for an involvement of histone modifications or DNA methylation in latency establishment or maintenance.
Until recently, HIV-1 latency was regarded as a gene regulation phenomenon that would be governed by the same mechanisms that control cellular gene expression, histone modifications and DNA methylation. However, findings by Han et al., published in 2004 (30
), suggested that viral integration events in CD4-positive memory T cells of highly active antiretroviral therapy-suppressed patients were in >90% of all analyzed events found integrated into the exons or introns of genes that are, in general, actively expressed in memory T cells. This finding is not supportive of the idea that latent integration would be governed by a suppressive histone code, which is unlikely to form in the exon/intron region of actively expressed genes. A likely explanation would be that transcriptional interference in which the transcriptional machinery initiates at the promoter of the gene into which the virus is integrated (host gene) reads through the viral genome. The constant presence of the transcriptional machinery initiating at the host gene promoter then prevents transcription factors from binding to the viral promoter and thus stabilizes latent infection. As the initial ex vivo studies were performed using primary CD4-positive memory T cells, the authors could determine only individual viral integration sites and correlate these integration sites with a general gene expression profile for memory T cells. However, in follow-up studies using the latently infected clonal J-LAT cell line or a clonal cell line in which an HIV-1 reporter construct had been integrated in an orientation-dependent manner relative to the host gene, two groups independently demonstrated that latent HIV-1 infection could indeed be a result of transcriptional interference (31
We here demonstrate that these results on the governing role of transcriptional interference for HIV-1 latency are not limited to selected clonal cell lines but can be transferred to our population-based experimental system, strongly supporting the idea that transcriptional interference is the primary mechanism controlling HIV-1 latency in vivo. We found that all tested integration events that had given rise to latent HIV-1 infections in our population-based assay had occurred in actively expressed host genes.
The idea that transcriptional interference is the key governing mechanism for the maintenance of HIV-1 latency is supported by several of our other experimental results. First, pretreatment of T cells with the HDAC inhibitor valproic acid, NaBu, or TSA did not prevent latency formation. As HDAC inhibitors have been reported to reactivate latent HIV-1 infection by removing a restrictive histone code, the inhibitors should have prevented latency formation if a restrictive histone code were involved, which was not the case in our experiments. As we have demonstrated that the decision whether a virus integrates in a latent state is made early within the first 48 h after infection and that bolus application of the HDAC inhibitors maintains full activity during this time period, we conclude that a histone code is not required to either establish or maintain latent infection.
Second, this is also suggested by the results obtained using the HIV-1 transcription inhibitor Ro24-7429 during the infection phase. As the data demonstrate, the presence of the inhibitor did not alter the number of total viral integration events but relatively efficiently suppressed active viral transcription in the majority of the infected cells. This should have increased the likelihood of the formation of a suppressive histone code at the viral LTR. However, as our data show, it is unlikely that a restrictive histone code was established, as the application of Ro24-7429 did not alter the level of latency formation.
Third, while we were able to fully suppress HIV-1 expression in a population of otherwise chronically actively HIV-1-infected cells over a prolonged period (>20 days), this extended suppression of viral gene expression was insufficient to generate any latent infection events, which would have been suggestive of the formation of a repressive histone code that would exert control over viral gene expression.
Similarly, we could not find evidence that DNA methylation events are important for the establishment of latent infection. The literature is inconclusive on the possible importance of DNA methylation for the maintenance of HIV-1 latency, and no particular methylation pattern of the viral LTR has been associated with HIV-1 latency (3
). We also saw evidence for the formation of DNA methylation patterns that influence reactivation (Fig. 10C
), as in a substantial selection of latently infected cell clones the addition of a DNMT inhibitor increased the ability to reactivate latent infection following provision of a cellular stimulus. However, when we continuously passaged cell populations that held latently infected cells, we observed a slow but gradual decrease in the size of the latently infected cell population. As there is no evidence that this decline of the level of latently infected cells is associated with spontaneous HIV-1 reactivation and deletion of these cells from the culture by the ensuing cytopathic effect of the then-active virus, we conclude that the decrease in reactivatable infection events over time is caused by DNA methylation resulting in permanent transcriptional silencing of the integrated viral LTR. This is not particularly surprising, as silencing of retroviral vectors in dividing cells is a problem that has been extensively described (59
). As it seems that the occurrence of DNA methylation patterns that are reversible by treatment with 5-azacytidine is associated with continuous cell division, it seems unlikely that a similar restrictive DNA methylation pattern would be established in latently HIV-1-infected, nondividing memory T cells in vivo.
The second finding central to our experiments is that we report silent integration to be a prerequisite to the establishment of latent HIV-1 infection events. As we have demonstrated that low intracellular activation levels are a prerequisite for silent integration, this would require that in vivo or ex vivo low-level-activated or even resting CD4+
T cells could be infected with HIV-1. While HIV-1 preferentially infects activated CD4-positive T cells, it has been experimentally demonstrated that resting T cells can be infected and that infection in these cells is mostly latent (2
). Our data generated in immortalized T-cell lines and primary T cells would thus be consistent with these earlier findings in primary T cells. Obviously, our data do not conclusively rule out the possibility that latent infection in primary T cells can be generated by a decrease in the availability of crucial transcription factors that would be associated with the transition of the infected cell to a resting state. However, we demonstrate that the generation of latent infection events is independent of the requirement for a reduction of the cellular activation state, which would be a statistically highly unlikely coincidence. Silent integration of the virus into actively expressed genes is thus a model concept that can explain both latency establishment and maintenance.
We here add to recently presented evidence that HIV-1 latency is not governed by the same mechanisms as is cellular gene expression but, following silent integration, is maintained by transcriptional interference. Even if only a minor portion of the latent infection events, and not the majority of these integration events, as suggested by the work of Han et al. (30
) and our results, are governed by transcriptional interference, current therapeutic strategies will have to be revisited. Taken together, these findings have wide-ranging consequences for the future design of therapeutic strategies. Targeted activation of latently infected cells still remains a therapeutic option, but the devastating results from the TGN1412 clinical phase I trial, in which the application of an agonistic anti-CD28 antibody had a near-fatal outcome for all six volunteers, would suggest extremely careful evaluation of this path (32
). Fundamentally different ideas will be needed to therapeutically target HIV-1 latency.