HIV transcription depends on host cellular transcription factors and consequently on the activation state of the infected cell, on the viral Tat protein, and on the chromatin state of the viral promoter at the integration site. When HIV infects a homogeneous cell population, viral expression is heterogeneous among individual integration events (clones) (29
). This heterogeneity can be attributed to the integration site and, consequently, to the chromatin state of the HIV promoter that influences the accessibility of the transcription factors and the loading of the preinitiation complex. It cannot be ruled out that chromatin compaction may also influence progression of the elongating polymerase, termination, or reinitiation. HIV is a highly effective virus that is expressed in the majority of infected cells, allowing viral replication, presumably by integration into open chromatin represented by active genes (51
). Occasionally, the HIV provirus is not basally expressed but can be reactivated by disturbing the equilibrium between activation and repression with agents that interfere with chromatin, such as histone deacetylase inhibitors, or inducers of cell signaling that activate transcription factors. This behavior resembles viral latency, which in vivo
certainly depends on additional components (11
). Analysis of the integration sites that are associated with a latent phenotype has shown that it is produced by integration into regions of obviously compacted chromatin or, more importantly, into introns of highly expressed genes (24
). It is not obvious how the chromatin environment could be a determinant for repression when a provirus is embedded in a transcribing gene. Instead, it has been proposed that read-through transcription from an upstream promoter may interfere with HIV transcription by disturbing assembly of the preinitiation complex, irrespective of the relative orientation between the host gene and the provirus (23
). Nonetheless, there have also been described situations in which upstream transcription could indeed enhance transcription from HIV-1 proviruses that are in the same orientation as the host gene (25
). Position effects may explain the discrepancy between the different models used. The mechanism by which transcriptional interference is exerted has not previously been unraveled. Here, we propose that the chromatin reassembly machinery and associated factors traveling with the elongating form of RNA polymerase II (Pol II) through the HIV promoter actively maintain a chromatin environment refractory to HIV promoter activation.
In this paper, we have addressed the question of how HIV latency is established upon integration into a highly expressed gene, in particular, what transcriptional consequences the provirus integration has for the host genes and for the virus itself and what host factors are involved in maintaining HIV repression in this context. For this, we have used a minimal model system proven to be useful for studying the genomic determinants of postintegration latency. Two J-Lat clones harboring HIV integrations into introns of highly active genes, in the same direction as or opposite to the transcription of the host gene, have been analyzed and compared to clones where the HIV provirus is in other genomic environments.
We observed that HIV integration into an intron of a gene does not abolish expression of its normally spliced transcript, although expression is diminished to some extent. Analyzing the J-Lat clone E27, we were not able to determine the proportion in which normally spliced UBXD8 transcripts derived from the intact or the HIV-containing allele, although we have measured expression of the unintegrated allele (intron 8) and it was normal, proportional to the number of copies in the genome for different clones. Host gene expression was further decreased when HIV was reactivated, presumably at the provirus-containing allele, as the unintegrated allele was not affected by treatments that induce HIV (Fig. ). Moreover, in clone A2, as HIV is in the single-allele UTX gene (chromosome X), we were able to ascertain that normally spliced UTX transcripts are being produced despite the provirus integration but that this expression is reduced upon HIV reactivation. The reason for this is not clear, and possible mechanisms have not yet been investigated, but it could be due to transcription factor or polymerase trapping by the strong initiating activity of the HIV promoter, a putative form of transcriptional interference. In this line, we have measured by chromatin immunoprecipitation (ChIP) a strong recruitment of Sp1 to the HIV LTR upon TNF-α treatment (see Fig. S10 in the supplemental material). It has recently been proposed that enhancer-blockers and insulators could resemble specialized promoters that sequester the transcriptional machinery (49
). The activated HIV promoter could be acting as one of these enhancer-blockers.
A relative comparison of transcripts representing host gene exons or the stimulated initiating HIV indicated that HIV transcription is comparable to or higher than host gene transcription (Fig. ). Notably, host gene downregulation is much lower than the activation of HIV observed upon stimulation. A limited decrease in host gene expression upon latent HIV reactivation has been also observed in a study (38
) that analyzed J-Lat clones 9.2 and 15.4 generated with a full-length, GFP-expressing HIV genome (28
). We also observed this with the related J-Lat clone 6.3 (see Fig. S2 in the supplemental material). In a different report, downregulation of the host gene upon latent HIV reactivation by Tat expression was not attributed to transcriptional interference, but rather to a Tat effect, as it affected both alleles (15
Lenasi et al. (38
) reported that host-viral chimeric transcripts at the infected allele in J-Lat 9.2 and 15.4 cells terminate at the poly(A) contained in the 5′ LTR (R-U5 junction). In that case, chimeric transcripts were due to the appearance of cryptic splicing acceptor sites at the intron preceding the 5′ LTR. We have also characterized the existence of several species of chimeric transcripts in J-Lat E27 cells extending into the 5′ LTR and transcripts that cross the 5′ LTR without terminating at the poly(A), with different splicing variants (Fig. ). Similarly, we have identified transcription through the 3′ LTR. This indicates that poly(A) function is leaky; otherwise, all transcription of the infected allele would terminate at the HIV LTRs and no correctly spliced gene transcripts would be detected. The existence of mature, chimeric transcripts containing the U3 region of the 5′ LTR has been used to argue that transcriptional interference may occur. Moreover, transcription through the 5′ LTR, presumably initiated at an upstream host gene and terminating at the 5′ LTR poly(A), can be detected in heterogeneous populations of HIV-infected primary T cells (38
). Nonetheless, transcription interference would still take place even if chimeric transcripts were not detected, because the provirus-containing intron could be efficiently removed from mature transcripts and be present only in short-lived pre-mRNAs. Relative quantification of transcripts shows that chimeric transcripts are significantly less abundant than are transcripts containing upstream exons (chimeric plus wild type).
Interestingly, the alternative splicing varies as a function of PMA treatment: the intron 8/5′ LTR containing the A1
band is decreased compared to spliced A2
, while the unspliced C1
product was substituted for the splicing variants C2
. Apparently, PMA stimulation of the host gene promoter favors splicing events, although in all three cases splicing was accepted at cryptic sites, at either the 5′ LTR or the Tat sequences. The reciprocal relationship between transcription elongation and splicing has been extensively reported: elongation rates control alternative splicing and splicing factors can, in turn, modulate Pol II elongation (34
). It has to be noted that upon TNF-α treatment, HIV transcripts that initiated shortly upstream of the canonical transcription start site are detected.
Upon stimulation, transcripts initiated at the HIV 5′ LTR are produced, with several splicing variants. Unexpectedly, a transcript that uses the HIV splicing donor and UBXD8 exon 9 acceptor is highly induced. This could explain in part why the relative quantification of HIV transcripts covering different regions shows that Tat, GFP, and 3′ LTR are underrepresented compared to the 5′ LTR region (Fig. ). Nonetheless, an elongation block or incomplete read-through elongation after HIV transcription initiation also seems to exist, as has been reported previously (1
). In agreement with this, exogenously added Tat can reactivate latent HIV (40
). Another consequence of our study is that simple measurement of a reporter gene such as GFP is an oversimplification of the transcriptional events taking place around the provirus.
Transcription from the 3′ LTR is also induced upon cell stimulation, being responsible for the expression of downstream UBXD8 exons, most strongly if elongation is stimulated with HMBA in addition to TNF-α treatment (Fig. ). Despite the 3′ LTR being identical to the 5′ LTR, its promoter activity is always lower, basally and after induction, due to interference of the upstream HIV promoter in the downstream LTR (14
). Expression of downstream host gene exons from either or both LTRs could lead to the synthesis of truncated proteins, this being one of the bases of the mutagenic capacity of retroviruses. However, this was not the case in clone E27 according to the predictions made from the sequencing data on amplified cDNAs.
Transcriptional interference has been proposed to explain how a highly transcribing gene may maintain as silent a downstream HIV promoter leading to HIV latency. We report that components of the machinery which reassemble chromatin concomitantly with transit of the transcribing RNA polymerase are necessary for the repression of the cryptic HIV promoter in this context. Upon depletion of these factors, HIV reactivation occurs when integrated into introns of active genes, but not for other genome sites, concomitantly with increased chromatin accessibility at the HIV promoter and host gene. This effect produces synergies with other treatments that induce HIV promoter activation and could be explored as therapeutic interventions against the latent reservoir of the virus, although the general effects of knocking down these factors on gene expression would in principle discourage this.
The situation described here resembles the mechanism reported in Saccharomyces cerevisiae
to suppress cryptic transcription from within coding regions. In a first report, the transcription elongation factor Spt6 was identified as a repressor of transcription initiation from cryptic promoters by maintaining the normal chromatin structure during transcription elongation (31
). A yeast spt6
mutant permitted aberrant initiation from within coding regions, and transcribed chromatin became hypersensitive to micrococcal nuclease. Concomitantly with activation of the cryptic promoter, the level of transcription of the wild-type host gene decreased. More recently, a comprehensive analysis of cryptic transcription identified at least 50 factors, many involved in chromatin structure and transcription, required to repress cryptic promoters throughout the yeast genome (7
). These factors would maintain the global integrity of gene expression during normal growth but may also allow the expression of alternative genetic information under altered genetic or physiological conditions. These factors include histone-encoding genes; histone regulators, such as Hir1 to -3 or Hpc2; chromatin assembly and remodeling factors, such as Spt6, Chd1, Asf1, and Spt16; histone deacetylases and accompanying factors; components of the mediator complex; and transcription elongation factors, such as Spt4 and Spt5 (orthologs of subunits of the DSIF complex, a known inhibitor of HIV elongation) (22
), among others.
By using a chimeric yeast-HIV promoter system, we initially found that spt6
, and spt16
mutants derepressed the viral promoter (57
). We have now added Hir1 to -3, Hpc2, and Asf1A to the list of factors that mediate HIV repression in this model system. RNA interference-mediated depletion of all these factors in the human HIV latency J-Lat system shows their involvement in the repression of HIV when integrated at certain coding regions. Further, all these factors can be related to the process of chromatin reassembly associated with transcription elongation, either as histone chaperones or as chromatin remodeling or disassembly/reassembly factors. Accordingly, we propose a model in which CRFs are involved in the repression of the HIV promoter by transcriptional interference when integrated at highly active transcription units (Fig. ). The transcriptional machinery elongating through a highly expressed cellular gene containing an integrated HIV provirus includes CRFs that maintain a repressive chromatin configuration. In the absence of any chromatin reassembly factor, nucleosomes are poorly rebuilt and the chromatin configuration allows transcription factors and the transcriptional machinery to access the HIV promoter. Concomitantly, host gene expression decreases due to the global effect of CRF depletion, contributing to a reduction in transcriptional interference, and HIV activation further downregulates the upstream promoter.
A model for the involvement of chromatin reassembly factors in the repression of integrated HIV promoter by transcriptional interference.
The histone chaperone Spt6, in addition to depositing histones during reassembly, is known to interact with the RNA Pol II C-terminal domain (CTD) and Set2, a histone methyltransferase that methylates H3K36 through transcription elongation (5
). H3K36 methylation is a mark for histone deacetylation (37
). Set2, histone deacetylases, and the histone modifications that they determine provide restoration of normal chromatin in the wake of elongating RNA Pol II and prevent inappropriate initiation within coding regions masked by chromatin. This would include initiation at the HIV promoter in certain chromatin environments. In agreement with this model, histone deacetylase (HDAC) inhibitors are known inducers of HIV transcription (56
), and this is also true in the context of the latent integrations at active coding regions analyzed here.
CRFs participate in the mechanism that controls the equilibrium between activation and repression of HIV when integrated in the human genome, which depends greatly on the chromatin environment at the integration site. Disturbance of this equilibrium, by depleting CRFs, for example, makes some of the latent integrations become activated without the need for further activating stimuli. Similarly, it has been reported that transcriptional activators are dispensable for transcription in the absence of Spt6-mediated chromatin reassembly of particular yeast promoter regions (2
). Many of the factors involved in preventing internal cryptic promoter initiation in yeast, including Spt2, Spt4, Spt5, Spt6, and Spt16 (7
), were originally identified in genetic screening based on the yeast Ty1 retroelement (10
). It is, therefore, plausible that the repressive function of CRFs to maintain the global integrity of gene expression is conserved from yeast to humans and may have been evolutionarily shaped by the selective pressure of endogenous retroelements. It would be interesting to investigate what the global consequences of Spt6 inhibition on the expression and mobility of these elements are and whether the observed downregulation of host genes is due to Spt6 being essential for transcription or to side effects produced by the induction of overlapping cryptic promoters.
Infection of T cells with an HIV vector leads to viral expression in the vast majority of integration events. Latency is rare, as in the in vivo
situation. Integration into heterochromatin or gene deserts favors the establishment of latency, but it can also occur with the integration into introns of active genes (3
). We have found that depletion of Spt6 reactivates latent HIV when inserted into an active gene (J-Lat clones E27 and A2) but does not further activate HIV in an active clone (J-Act C9) where the host gene (TYK2) is expressed. Since we have observed TYK2 downregulation upon Spt6 depletion (data not shown), the ability of CRFs to repress HIV must be context dependent. The easiest explanation is that chromatin reassembly exerts its repressive effect when the host gene is highly expressed and polymerases are thoroughly reading the host gene. Accordingly, in active host genes, latency seems to be favored when they are highly expressed (3
). Nonetheless, local determinants such as position, orientation, and distance to other regulatory elements (enhancers, insulators, and splice sites) may also be of consideration. After repression is set, the situation is maintained by CRFs and histone deacetylation. Perturbing the equilibrium of promoter chromatin leads to HIV reactivation, but once the promoter becomes active, these perturbations have a minor effect.
Further work is required to clarify whether, in addition to CRFs, other cis or trans determinants may participate in transcriptional interference of the viral promoter.