Article15 April 2003free access HIV reproducibly establishes a latent infection after acute infection of T cells in vitro Albert Jordan Albert Jordan Gladstone Institute of Virology and Immunology, University of California, San Francisco, CA, 94141 USA Present address: Center for Genomic Regulation, Passeig Marítim, 37–49, E-08003 Barcelona, Spain Search for more papers by this author Dwayne Bisgrove Dwayne Bisgrove Gladstone Institute of Virology and Immunology, University of California, San Francisco, CA, 94141 USA Search for more papers by this author Eric Verdin Corresponding Author Eric Verdin Gladstone Institute of Virology and Immunology, University of California, San Francisco, CA, 94141 USA Department of Medicine, University of California, San Francisco, CA, 94141 USA Search for more papers by this author Albert Jordan Albert Jordan Gladstone Institute of Virology and Immunology, University of California, San Francisco, CA, 94141 USA Present address: Center for Genomic Regulation, Passeig Marítim, 37–49, E-08003 Barcelona, Spain Search for more papers by this author Dwayne Bisgrove Dwayne Bisgrove Gladstone Institute of Virology and Immunology, University of California, San Francisco, CA, 94141 USA Search for more papers by this author Eric Verdin Corresponding Author Eric Verdin Gladstone Institute of Virology and Immunology, University of California, San Francisco, CA, 94141 USA Department of Medicine, University of California, San Francisco, CA, 94141 USA Search for more papers by this author Author Information Albert Jordan1,2, Dwayne Bisgrove1 and Eric Verdin 1,3 1Gladstone Institute of Virology and Immunology, University of California, San Francisco, CA, 94141 USA 2Present address: Center for Genomic Regulation, Passeig Marítim, 37–49, E-08003 Barcelona, Spain 3Department of Medicine, University of California, San Francisco, CA, 94141 USA *Corresponding author. E-mail: [email protected] The EMBO Journal (2003)22:1868-1877https://doi.org/10.1093/emboj/cdg188 PDFDownload PDF of article text and main figures. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InMendeleyWechatReddit Figures & Info The presence of latent reservoirs has prevented the eradication of human immunodeficiency virus (HIV) from infected patients successfully treated with anti-retroviral therapy. The mechanism of postintegration latency is poorly understood, partly because of the lack of an in vitro model. We have used an HIV retroviral vector or a full-length HIV genome expressing green fluorescent protein to infect a T lymphocyte cell line in vitro and highly enrich for latently infected cells. HIV latency occurred reproducibly, albeit with low frequency, during an acute infection. Clonal cell lines derived from latent populations showed no detectable basal expression, but could be transcriptionally activated after treatment with phorbol esters or tumor necrosis factor α. Direct sequencing of integration sites demonstrated that latent clones frequently contain HIV integrated in or close to alphoid repeat elements in heterochromatin. This is in contrast to a productive infection where integration in or near heterochromatin is disfavored. These observations demonstrate that HIV can reproducibly establish a latent infection as a consequence of integration in or near heterochromatin. Introduction Combination anti-retroviral therapy can control HIV-1 replication and delay disease progression. However, despite the complete suppression of detectable viremia in many patients, viremia re-emerges rapidly after interruption of treatment, consistent with the existence of a latent viral reservoir (Finzi et al., 1997; Wong et al., 1997). This reservoir is thought to consist mainly of latently infected resting memory CD4+ T cells (recently reviewed by Chun and Fauci, 1999; Butera, 2000; Pierson et al., 2000). Due to the long half-life of this reservoir (44 months), it has been estimated that its total eradication with current treatment would require over 60 years (Finzi et al., 1999). Latently infected cells contain replication-competent integrated HIV-1 genomes that are blocked at the transcriptional level, resulting in the absence of viral protein expression. HIV depends on both cellular and viral factors for efficient transcription of its genome, and the activity of the HIV promoter is tightly linked to the level of activation of its host cell. HIV transcription is characterized by two temporally distinct phases. The early phase occurs immediately after integration and relies solely on cellular transcription factors. Because of a transcriptional elongation defect in the basal HIV promoter, most transcripts cannot elongate efficiently and terminate rapidly after initiation. However, a few transcripts elongate throughout the genome, resulting in transcription of the viral transactivator Tat. The late phase of transcription occurs when enough Tat protein has accumulated. Tat binds to TAR, recruits the pTEFb complex and causes the hyperphosphorylation of RNA polymerase II, dramatically increasing its ability to elongate (Karn, 1999). To understand how postintegration latency is established, an in vitro cell system reflecting the state of HIV-1 latency is required. While several latently infected cell lines have been established after HIV infection, the proviruses integrated in these cell lines harbored mutations in their Tat–TAR transcriptional axis (Emiliani et al., 1996, 1998) or in NF-κB binding sites in the HIV promoter (Antoni et al., 1994). The presence of mutations in HIV-infected latent cell lines has raised questions about the significance of these lines to latency in vivo. However, the very presence of these mutations has also strengthened the concept that transcription inhibition is critical to the establishment and maintenance of HIV latency. We reported previously that the site of integration of the HIV-1 provirus into the cell genome affects its basal Tat-independent transcriptional activity (Jordan et al., 2001). We therefore predicted that a small fraction of integration sites might lead to such low basal promoter activity that no Tat mRNA would accumulate. Such a provirus would be locked in the early phase of transcription, resulting in functional latency. In this manuscript, we have used recombinant viruses that express green fluorescent protein (GFP) to selectively enrich for rare cells blocked at the transcriptional level during an acute infection. Using this approach, we have isolated a large number of clonal cell lines harboring HIV in a latent state. The latent provirus can be reactivated at the transcriptional level in these cell lines by a variety of agents. This study documents that HIV reproducibly establishes a latent infection with low frequency after acute infection of T cells in vitro. Results Establishment of an in vitro model of HIV-1 latency To determine whether unique integration events can lead to latency, we used an HIV-based retroviral vector containing the Tat and GFP open reading frames both under the control of the HIV promoter in the 5′ long terminal repeat (LTR). We infected a culture of the lymphocytic cell line Jurkat with viral particles containing this vector and used differential fluorescence-activated cell sorting (FACS) based on GFP expression (Figure 1A). First, we infected Jurkat cells with the LTR–Tat–IRES–GFP virus at a low m.o.i. and isolated GFP-negative cells by FACS 4 days after infection (Figure 1A). This population presumably harbored both uninfected cells and cells with transcriptionally silenced proviruses. To activate HIV expression, we treated this population with TPA or tumor necrosis factor α (TNF-α) and purified GFP-positive cells by FACS (Figure 1A). These cells, representing <0.06% of the original population, corresponded to the latent phenotype: GFP-negative under basal conditions and GFP-positive after activation with TPA or TNF-α. By comparing the proportions of productively infected cells (4%) and cells exhibiting a latent phenotype (0.06%), we calculate that ∼1.5% of infections (1 in 66) resulted in a latent state in this system. These cells were both grown as a group and individually sorted for further characterization. Re-analysis 17 days after sorting showed that a significant proportion of the cells had no GFP expression, indicating transcriptional silencing in the absence of TPA (Figure 1A). Figure 1.Enrichment and purification of HIV-latently infected cells by FACS. (A) A schematic representation of our enrichment protocol is shown. See text for details. The percentage of GFP-positive cells obtained after infection (4%) or after TNF-α treatment of GFP-negative cells (0.06%) is shown. Similar data were obtained with TPA. (B) Clonal cell lines isolated using the procedure described above were analyzed for GFP expression under basal and stimulated conditions (24 h treatment with TNF-α). (C) mRNA levels were measured for HIV and for GAPDH using TaqMan PCR in untreated and TNF-α-treated cell lines. Results are expressed as a percentage of mRNA levels measured in Jurkat cells acutely infected with HIVNL4-3 (day 6 post-infection). Clone A72 is infected with an LTR–GFP construct (Jordan et al., 2001), clone H2 is infected with the LTR–Tat–IRES–GFP vector while clones F11 and G10 are infected with HIV-R7/E−/GFP. Download figure Download PowerPoint Individual cells exhibiting a latent phenotype were cloned and clonal cell lines were further characterized. Flow cytometry analysis of individual clones 4 weeks after their isolation showed low basal GFP expression (Figure 1B, left panel). After TPA treatment, all clones were activated, and GFP levels reached a maximum that was relatively independent of basal GFP activity (Figure 1B, right panel). Similar results were obtained after stimulation with TNF-α (data not shown). Examination of mRNA levels in these cell lines under repressed (−TNF-α) and activated (+TNF-α) conditions confirmed the presence of very low transcriptional levels under basal conditions and the activation of HIV transcription by TNF-α (Figure 1C, see clone H2). Importantly, HIV-specific transcript levels after TNF-α treatment were of the same order of magnitude as mRNA levels measured in Jurkat cells acutely infected with wild-type HIV (NL4-3; Figure 1C). Mutations in a cellular factor important for HIV transcription, such as CDK9 or cyclin T1 for example, could be responsible for a latent phenotype. To test this possibility, we infected one of our latent cell lines (clone 82) and control Jurkat cells with an LTR–GFP virus or with the LTR–Tat–IRES–GFP virus (Figure 2A). Both cell lines produced similar amounts of GFP after infection with the LTR–GFP–virus (Figure 2A). GFP levels were higher but also comparable in both cell lines after infection with the LTR–Tat–GFP virus (Figure 2A). Similar results were observed in three other latent cell lines (clones H2, A2 and A10; Figure 2C). Figure 2.Characterization of latently infected clones. (A) Clone 82 is competent for basal and Tat-dependent HIV promoter activity. Clone 82 was infected with viral particles containing HIV-derived vectors LTR–GFP or LTR–Tat–IRES–GFP, and GFP expression was measured by flow cytometry. As a control, Jurkat cells were infected in parallel. (B) The integrated latent HIV promoter in clone 82 is unresponsive to Tat stimulation. Clone 82 was infected with a Tat-expression retroviral vector, or treated with TPA, as a positive control. As a control of the infection process and Tat expression, a clone containing a single integration of the HIV-derived LTR–GFP vector was infected in parallel. (C) HIV promoter activity (percentage of GFP-positive cells in R2 gate) was measured in four latent clonal cell lines (82, H2, A2 and A10) 24 h after infection with retroviral particles containing the following vectors: LTR–GFP, LTR–Tat, LTR–Tat–GFP or after treatment with TNF-α. Control Jurkat cells and a cell line containing a stably integrated LTR–GFP retroviral construct (Jordan et al., 2001) were used as controls. Download figure Download PowerPoint This experiment demonstrates that the cellular environment in latent cell lines can support HIV transcription and Tat transactivation with the same efficiency as control Jurkat cells and is therefore not responsible for the latent phenotype. HIV transcription is characterized by an early, Tat-independent phase and a late, Tat-dependent phase. Since no GFP was produced by our latent cell lines, it is likely that Tat was not expressed since both proteins are located on a polycistronic mRNA. To test whether Tat alone was sufficient to reactivate latent HIV gene expression, we infected clone 82 with a HIV-derived retroviral vector expressing the Tat protein (LTR–Tat virus; Figure 2B). Infection of a Jurkat clone containing a single copy of an integrated LTR–GFP retroviral vector (clone A, described in Jordan et al., 2001) with the same LTR–Tat virus stock showed high levels of GFP expression but had no effect in the latently infected cell line (Figure 2B). Similar results were obtained after transfection of a Tat-expression plasmid into these cell lines (data not shown). Similar results were also observed in three other latent cell lines (clones H2, A2 and A10) although weak stimulation of GFP expression was noted in clones H2 and A2 (Figure 1C). We conclude that the latent HIV promoter is relatively unresponsive to Tat stimulation, suggesting that the blockage of transcription in this clone lies primarily at the level of transcription initiation. Preferential HIV integration in or near alphoid DNA in latently infected cells The basal activity of the HIV promoter is determined by a combination of cis- and trans-acting variables. The observations described above indicate that the cellular environment in each latent cell line is fully capable of supporting the transcription of the HIV genome and therefore points to the role of the site of integration of the provirus as a likely cause of low basal transcription. Accordingly, we isolated and sequenced the provirus integration sites in eight distinct clonal cell lines (Figure 3A) Mapping of the integration site on the human genome using BLAST showed that the retroviral vector had integrated in an alphoid repeat element in four out of eight clones on chromosomes 7, 10 and 16 (Figure 3A). The site of integration of three clones was identified in the human genome in non-alphoid repeat DNA while the site of integration in one clone did not yield significant homology to any region of the human genome (Figure 3A). Figure 3.Latency is associated with preferred integration in or near alphoid repeats. (A) The sequence of integration site of the HIV vector is shown for seven clonal cell lines aligned with the corresponding genomic sequence. The match in GenBank for each clone corresponded to DDBJ/EMBL/GenBank accession number M93288 for clone 82, AL591625 for clone A1, AC023948 for clone A5, M16037 for clone A7, Al354920 for clone A10, AC019063 for clone H2 and AC079801 for clone F2. The sequence corresponding to the HIV promoter is indicated by a closed box. The chromosomal location of each integration site is shown and integration sites into alphoid repeat elements are indicated. (B) A nested PCR assay designed to quantify integration in or near alphoid repeats is schematically represented. The position of primers in the two sequential PCRs are shown aligned with the HIV 5′ LTR and a putative alphoid element in the genome. (C) Four PCRs were conducted for each cell line using either primer α1, α2, α3 or α4 in combination with primer A for the first reaction and the primer pair B+C for the second PCR. PCR products amplified according to this scheme are shown for each latent cell line. (D) Real-time PCR analysis of integration in or near alphoid repeats. Primer G represents an internal fluorescent primer used for the quantification of the TaqMan reaction. (E) Quantification of PCR products after TaqMan PCR analysis as indicated in (D). Productively infected cells (GFP-positive cells in Figure 1A, panel 1; productive infection) are compared with latently infected cells (Figure 1A, panel 4; latent infection) using the Alu and alphoid PCR assay. Data is expressed as the relative signal intensity (alphoid/Alu). Numbers on top of bar indicate the ratio between latent and productive infection. Download figure Download PowerPoint To confirm this observation, we developed a novel PCR assay in which degenerate primers matching alphoid DNA consensus sequence (α1, α2, α3, α4) are used in combination with an LTR specific primer (primer A; Figure 3A). Amplified products were detected with two nested primers in the LTR (primers B and C; Figure 3B). Since primers α1 and α3 are orientated in the same way in the alphoid repeat, a positive signal with one primer should be accompanied by a positive signal for the other (Figure 3B). The same holds true for primers α2 and α4 (Figure 3B). Similarly, a positive response with primers α1 and α3 would be accompanied by a negative signal for primers α2 and α4. Examination of the PCRs showed perfect concordance with the sequencing data: clones 82, A7, H2 and F2 showed a positive PCR and had a documented alphoid repeat integration (Figure 3C) while clones A1, A5, A10 (Figure 3C) and A11 (not shown) tested negative in the PCR assay and did not integrate in an alphoid element. For comparison, we subjected DNA obtained from a library of 34 random integrations (not sorted for a latent phenotype) to the same procedure (Jordan et al., 2001). None of these 34 clones showed a pattern consistent with alphoid repeat integration, in agreement with previous reports that HIV integration in or near alphoid repeat elements is disfavored (Carteau et al., 1998). To examine the relative distribution of alphoid integration in latently versus productively infected cells at a population level, we adapted the alphoid PCR assay to TaqMan PCR. Because Alu elements are randomly distributed within the genome, this analysis included a control using primers specific for the Alu element and the HIV LTR. The product amplified by this primer pair is assumed to represent random integration events within the genome and has been used as a reliable marker of HIV integration (Minami et al., 1995; Chun et al., 1997b; Butler et al., 2001). Results are therefore presented as the ratio of the alphoid–LTR product to the Alu–LTR product as an indication of the frequency of integration in or near alphoid DNA (Figure 3E). When cell populations corresponding to productively infected cells (GFP-positive in first panel of Figure 1A) were compared to latently infected cell population (GFP-positive in third panel of Figure 1A), we observed preferential integration (9-fold to >100-fold enrichment) in or near alphoid repeats in latent cells in comparison to productively infected cells (Figure 3E). Similar results were obtained after infection of human peripheral blood mononuclear cells (PBMCs) with the HIV-derived vector (data not shown). Cloning and sequencing of several PCR products obtained after amplification with the alphoid specific primers confirmed that these contained HIV integration events into alphoid DNA (data not shown). Transcriptional activation of the HIV promoter in latently infected cells We have tested a number of biological and chemical agents for their ability to reactivate latent HIV expression. Several NF-κB activators, including TPA, TNF-α and PHA, independently induced HIV expression, as expected, since NF-κB strongly induces the HIV promoter activity. TPA was the strongest inducer tested (Figure 4). All treatments increased the number of GFP-positive cells and the mean fluorescence intensity reflective of GFP levels (data not shown). Incubation with anti-CD3 antibodies, which crosslink the surface T-cell receptors, an alternative pathway leading to NF-κB activation, also induced GFP expression (data not shown). These results suggest that activation of the NF-κB pathway may boost HIV transcription initiation, the production of Tat, and transition to Tat-dependent transcriptional activation. Trichostatin A (TSA), an inhibitor of histone deacetylases also activated expression but to a lesser extent than NF-κB activators and only in some cell lines (see clones A1 as an example, Figure 4). Treatment with 5-axa-2-deoxycytidine (aza-dC), an inhibitor of DNA methylation, had little effect on the fraction of cells induced to transcribe HIV alone or in combination with a histone deacetylase inhibitor (Groudine et al., 1981; Chen et al., 1997; data not shown). Figure 4.Transcriptional activation of the HIV promoter in latently infected cells. Cells from clones 82, A1, A7 and A10 were treated as described in Materials and methods with several indicated agents and LTR expression was measured by flow cytometry. Data are expressed as percentage of cells becoming GFP-positive after a 24 h treatment. Download figure Download PowerPoint Latent cell lines containing a full-length integrated HIV genome To confirm that HIV latency can be established in the context of a full-length provirus, we used a recombinant HIV molecular clone containing the GFP open reading frame in place of the Nef gene (Bieniasz and Cullen, 2000) (HIV-R7/E−/GFP; Figure 5A). To restrict our analysis to a single infection cycle, the env gene was suppressed by introduction of a frameshift mutation. This defect was complemented by coexpression of a VSV-G envelope protein to generate pseudotyped viral particles. We infected a culture of the lymphocytic cell line Jurkat with viral particles containing this HIV genome and used differential FACS based on GFP expression to isolate GFP-negative cells by FACS 4 days after infection (Figure 5B). Based on our previous experiments, we predicted that this population harbored both uninfected cells and cells with transcriptionally silenced proviruses. To activate HIV expression, we treated this population with TNF-α and observed that a small fraction of the cell population (1.9%) became positive. These activated cells were purified by FACS based on GFP expression levels (Figure 5B). These cells were both grown as a group and individually sorted for further characterization. Re-analysis after sorting showed that a small proportion of the cells had no GFP expression, indicating transcriptional silencing has occurred after withdrawal of TNF-α (Figure 5B). Flow cytometry analysis of individual clones showed low basal GFP expression (Figure 5C). After TNF-α treatment, HIV expression was increased in all clones both in terms of the fraction of cells that became GFP-positive (Figure 5C) but also in terms of mean fluorescence intensity (data not shown). Measurement of virus-specific mRNA showed that the mechanism of latency in these clones was controlled at the transcriptional level (Figure 1C, clones F11 and G10). Levels of viral mRNA after activation were similar to those measured after a productive HIV infection (compare clones F11 and G10 with NL4-3, Figure 1C). Similar results were obtained using northern blot analysis (data not shown). Analysis of HIV-specific protein expression in several clones by western blotting using an antiserum from an HIV-infected individual showed no detectable expression of HIV proteins under basal conditions (Figure 5D). Treatment of the same clones with TNF-α led to a dramatic increase in HIV protein expression, particularly the gag p55 precursor (Figure 5D). When the culture supernatants of the same clones were examined for HIV-specific p24 expression, no or low picogram amounts could be detected under basal conditions (Table I). Treatment with TNF-α led to a >1000-fold increase in p24 measurement for several representative clones (Table I). These observations demonstrate that transcriptional latency can also be established in the context of a full-length HIV infection. Figure 5.Establishment of latently infected cell lines with a full-length HIV provirus. (A) Genome organization of a molecular clone of HIV encoding GFP and containing a frameshift mutation in env. (B) Schematic representation of protocol for enrichment of latently infected cells after infection of Jurkat cells with HIV-R7/E−/GFP (see text for details). (C) Clonal cell lines isolated using the procedure described above were analyzed for GFP expression under basal and stimulated conditions (24 h treatment with TNF-α). (D) Western blot analysis of four representative Jurkat clones latently infected with HIV-R7/E−/GFP. Clones were treated for 24 h with TNF-α (10 ng/ml) and cell lysates were analyzed by western blotting using an antiserum from an HIV-infected individual (provided by the NIH AIDS Research and Reagent Reference Program). A predominant band of 55 kDa corresponds to the Gag precursor protein. The same samples were analyzed using an antiserum specific for α-tubulin to ensure equal loading. Download figure Download PowerPoint Table 1. Low to undetectable HIV protein expression in several latently infected clonal cell lines under basal conditions GFP-positive (%) GFP signal (MFI) p24 (pg/ml) TNF-α − TNF-α + TNF-α − TNF-α + TNF-α − TNF-α + Clone 15.4 <1 46 ± 6 6 ± 0.5 188 ± 34 7 ± 12 6066 ± 1960 Clone 6.3 <1 27 ± 9 5 ± 0.2 135 ± 43 0 10 100 ± 4573 Clone 8.4 <1 77 ± 6 5 ± 0.3 488 ± 74 0 32 967 ± 10 537 Clone 9.2 <1 75 ± 7 7 ± 0.5 522 ± 61 23 ± 6 41 067 ± 9100 Clone 10.6 <1 96 ± 1 5 ± 1.4 645 ± 45 14 ± 3 85 500 ± 5981 Five representative Jurkat clones latently infected with HIV-R7/E−/GFP (same clones as shown in Figure 5D) were treated for 24 h with TNF-α (10 ng/ml). HIV expression was quantified using flow cytometry for GFP. Results are shown both as the fraction of cells expressing GFP above background (control Jurkat cells) and as the mean fluorescence intensity (MFI). Cell culture supernatant fluids were also assayed for the Gag-derived p24 HIV protein using an ELISA. Average of three measurements ± SD are shown. Discussion We present evidence that HIV can reproducibly lead to a state of transcriptional silencing and true postintegration latency. We have used a FACS-based protocol to highly enrich for latently infected cells and show that HIV integration leads with low frequency to integration sites that cannot support basal transcription of the HIV promoter. The low frequency of integration events leading to latency (∼1%) is likely to be the reason why this phenomenon has eluded discovery until now. Integration into alphoid repetitive DNA, a component of centromeric heterochromatin, occurred frequently in our latent clone population. A recent comprehensive analysis of the site of integration of HIV during a productive infection showed that HIV integrates preferentially in actively transcribed genes (69% of integrations sites; Schroder et al., 2002). In the same study, α satellite integration represented <1% of all integration events (Schroder et al., 2002). In contrast, in our small sample of sequenced integration sites from latent infections, we observed 50% (four of eight) of integration sites in alphoid repeats. Analysis of a larger sample using the alphoid repeat PCR assay indicated that the true frequency of alphoid repeat integration is probably lower, between 20 and 30% in latent cells (data not shown). Heterochromatin is found in transcriptionally inactive regions of the genome and is associated with specialized chromosome structures, such as centromeres and telomeres. It is characterized by a strongly condensed nucleosomal structure that impairs access to the underlying DNA for transcription factors and the basal transcriptional machinery. The unique properties of hetero chromatin are associated with the incorporation of particular histone variants, unique post-translational modifications of the histone tails and, heterochromatin-specific proteins (Wallrath, 1998; Jenuwein, 2001; Jenuwein and Allis, 2001). The transcriptional silencing effect of heterochromatin is not restricted to the region packaged into heterochromatin itself but extends to several kilobases of adjoining DNA. Heterochromatin-mediated transcriptional silencing has been characterized extensively in Saccharomyces cerevisiae, where the integration of a retrotransposon in or near heterochromatin leads to transcriptional repression and, in Drosophila, where chromosomal rearrangements (inversions, translocation), occurring as a result of X-ray irradiation, placed euchromatic genes close to a heterochromatic breakpoint. This chromosomal rearrangement led to a position-dependent inactivation of the transposed euchromatic gene. Because the expression of the rearranged euchromatic gene or of the transposon is often characterized by a variegated phenotype, transcriptional repression by integration in or near heterochromatin has been referred to as position effect variegation (Tartof, 1994). Interestingly, transcription of the HIV promoter in latent clones also shows a variegated phenotype after activation, i.e. only a fraction of the population becomes reactivated in response to a global signal. Variegated expression of the HIV promoter was also observed during progressive repression after withdrawal of the activation signal. Different clones reverted to the silenced phenotype at different rates, possibly indicating that different chromatin environments lead to transcriptional repression with different efficiencies. We propose as a working hypothesis that a repressive chromatin environment is the common underlying mechanism for transcriptional repression in all of our clones. As discussed above, we estimate that 20–30% of integration events in latent cell