Get 20M+ Full-Text Papers For Less Than $1.50/day. Subscribe now for You or Your Team.

Learn More →

Genome-wide Determinants of Proviral Targeting, Clonal Abundance and Expression in Natural HTLV-1 Infection

Genome-wide Determinants of Proviral Targeting, Clonal Abundance and Expression in Natural HTLV-1... The regulation of proviral latency is a central problem in retrovirology. We postulate that the genomic integration site of human T lymphotropic virus type 1 (HTLV-1) determines the pattern of expression of the provirus, which in turn determines the abundance and pathogenic potential of infected T cell clones in vivo. We recently developed a high-throughput method for the genome-wide amplification, identification and quantification of proviral integration sites. Here, we used this protocol to test two hypotheses. First, that binding sites for transcription factors and chromatin remodelling factors in the genome flanking the proviral integration site of HTLV-1 are associated with integration targeting, spontaneous proviral expression, and in vivo clonal abundance. Second, that the transcriptional orientation of the HTLV-1 provirus relative to that of the nearest host gene determines spontaneous proviral expression and in vivo clonal abundance. Integration targeting was strongly associated with the presence of a binding site for specific host transcription factors, especially STAT1 and p53. The presence of the chromatin remodelling factors BRG1 and INI1 and certain host transcription factors either upstream or downstream of the provirus was associated respectively with silencing or spontaneous expression of the provirus. Cells expressing HTLV-1 Tax protein were significantly more frequent in clones of low abundance in vivo. We conclude that transcriptional interference and chromatin remodelling are critical determinants of proviral latency in natural HTLV-1 infection. Citation: Melamed A, Laydon DJ, Gillet NA, Tanaka Y, Taylor GP, et al. (2013) Genome-wide Determinants of Proviral Targeting, Clonal Abundance and Expression in Natural HTLV-1 Infection. PLoS Pathog 9(3): e1003271. doi:10.1371/journal.ppat.1003271 Editor: Michael Emerman, Fred Hutchinson Cancer Research Center, United States of America Received November 6, 2012; Accepted February 10, 2013; Published March 21, 2013 Copyright:  2013 Melamed et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: This project was supported by the Wellcome Trust (www.wellcome.ac.uk/), grant number P08165. We are grateful for support from the Imperial NIHR Biomedical Research Centre funding scheme. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have declared that no competing interests exist. * E-mail: c.bangham@imperial.ac.uk Mitotic spread of HTLV-1 results in expanded clones of cells Introduction that carry the provirus in the same genomic integration site [7]. It is poorly understood how the flanking host genome influences Infectious spread results in integration of the provirus at a new transcription of an integrated provirus. Experiments on artificially genomic position. We have recently shown that the majority of modified proviral reporter constructs have yielded contradictory naturally infected T-cell clones carry a single proviral copy [8]. evidence on the role of flanking host promoters in either driving Integration of HTLV-1 does not favour specific hotspots, but is proviral transcription, or suppressing it by transcriptional inter- more frequent in transcriptionally active areas of the genome ference [1,2]. Conclusions from experiments on single artificial [9,10,11]. However, the factors that determine integration clones therefore cannot be reliably generalized: evidence is targeting and the abundance and expression of the HTLV-1 required from genome-wide studies of integrated proviruses in provirus in vivo are unknown. Two HTLV-1 gene products are natural infection. thought to play a crucial role in viral persistence in vivo. Tax, the Human T lymphotropic virus Type 1 (HTLV-1) persists in vivo by transcriptional transactivator of the virus, elicits abundant, two routes: by driving selective clonal proliferation of infected T chronically activated CTLs [12,13,14], indicating continuous or lymphocytes (‘mitotic spread’) and by de novo infection (‘infectious repeated expression of Tax in vivo. Ex vivo, Tax protein is spread’) via the virological synapse [3]. HTLV-1 replication is spontaneously expressed in a fraction of infected peripheral blood counterbalanced by a strong, chronically activated cytotoxic T mononuclear cells (PBMCs) after overnight culture [15]. HBZ is lymphocyte (CTL) immune response [4]. The HTLV-1 proviral load the only gene expressed from the minus strand of the provirus. (number of proviral copies per 100 PBMCs) varies between infected HBZ also promotes infected cell proliferation [16] and the CTL individuals by over 1000-fold. The proviral load is the strongest response to HBZ protein is a key determinant of proviral load and correlate of HTLV-1 associated diseases, in particular Adult T-cell the risk of the inflammatory disease HAM/TSP [17,18]. Tax Leukemia-Lymphoma (ATLL, [5]) and HTLV-1 Associated Mye- enhances HBZ expression; HBZ protein exerts negative feedback lopathy/Tropical Spastic Paraparesis (HAM/TSP, [6]). on Tax expression [19,20]. PLOS Pathogens | www.plospathogens.org 1 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression Similarly, we observed a bias (up to 2-fold greater than random) Author Summary towards integration in proximity to CpG islands; again, the bias HTLV-1 is a human retrovirus, estimated to infect over 10 reached a peak at 1 kb from the nearest CpG island (supplemen- million individuals worldwide, which causes the inflamma- tary Figure S4). tory disease HTLV-1-associated Myelopathy/Tropical Spas- We showed previously [11] that HTLV-1 provirus preferentially tic Paraparesis and an aggressive malignancy known as integrates in transcriptionally active regions of the host genome. Adult T-cell Leukemia/Lymphoma. The mechanisms that To test the hypothesis that specific transcription factor binding allow the virus to maintain a life-long infection are not fully sites (TFBS) influence HTLV-1 proviral targeting, expression and understood. Here we identified attributes of the host clonal abundance, we used data on genome-wide TFBS ChIP-seq: genome flanking the integrated HTLV-1 provirus associat- + where available, from primary CD4 T cells; otherwise, from T ed with integration targeting and spontaneous expression cells or other human cell types; see Table S3 for complete listing of of the provirus in vitro, and clonal expansion in vivo. the datasets used. Spontaneous expression (after short-term culture) of the In vitro integration sites showed a remarkably strong bias viral protein Tax, which is known to drive proliferation of (compared with random sites) towards integration in proximity to the infected cell, was significantly more frequent among specific TFBS, in particular STAT1, p53, HDACs (e.g. HDAC3, less expanded clones, suggesting that Tax-expressing HDAC6) and HATs (e.g. p300, CBP) (Table S3). In most cases the clones are more efficiently controlled by the immune effect was localized to within 100–1000 bases of the integration site response. Certain transcription start sites immediately (Figure 2A) and declined sharply at greater distances. Two upstream of the viral integration site were associated with patterns were observed in this biased integration. First, the virus latency, which in turn was associated with clonal preference towards integration in proximity to TFBS was typically expansion in vivo. symmetrical (e.g. p300), i.e. equally strong upstream and downstream of the integration site but in some cases was We hypothesize that the genomic integration site of HTLV-1 asymmetrical (e.g. STAT1), with a bias towards one side (often determines the pattern and intensity of expression of the plus and downstream). Second, in many cases we observed a sharp decrease minus proviral strands, which in turn determine the equilibrium in the preferential integration at 10 bases from the TFBS, such as abundance and the pathogenic potential of an infected T cell clone STAT1 Figure 2A). This pattern was consistently observed across in vivo. To test this hypothesis, we used our recently described several in vitro and in vivo datasets (supplementary Figures S1, protocol [11] of high-throughput mapping and quantification of S2). proviral integration sites in fresh primary PBMCs from HTLV-1- Because certain TFBS are frequently co-located in the human infected individuals. genome [22], we wished to test which TFBS were independently associated with targeting of the integration site. First, a likelihood ratio test was used to test whether the TFBS was selectively Results associated with integration either upstream or downstream of the HTLV-1 preferentially integrates within 1 kb of a host integration site, and each TFBS was then tested individually using transcription start site and is strongly biased to specific a univariate model. We then combined all significant factors using a step-down multivariate logistic regression analysis until only transcription factor binding sites independently significant (p,0.05) factors remained. Most factors To identify genomic factors associated with the targeting of that were independently associated with integration site targeting HTLV-1 integration, we infected Jurkat T cells by short co-culture occurred with equal frequency upstream or downstream of the with the HTLV-1-producing cell line MT2. The integration sites integration site (Figure 2B, see also supplementary Table S7). The were then analysed using our high-throughput protocol and factors with the highest odds ratios were the transcription factor compared to a control list of random sites in the human genome. p53 and the histone deacetylase HDAC6. Figure 1A illustrates the possible orientations (same or opposite) of the nearby genomic features, such as transcription start sites, either upstream or downstream of the integrated provirus. Effect of HTLV-1 integration sites on clonal expansion We previously showed [11] that 47% of de novo HTLV-1 We previously reported [11] a significant association between proviral integration events lie within a RefSeq gene. This certain features in the flanking genome and in vivo expansion of frequency is slightly higher than expected by chance, but is much the infected T-cell clone. Here, we found that proviruses lower than that observed for HIV (,70%), which uses the host integrated within a gene were more frequent in larger (more protein LEDGF to target proviral integration to genes [21]. As abundant) clones than in smaller clones in vivo, but only when the expected by chance, ,50% of proviruses integrated within host provirus was integrated in the same transcriptional orientation as genes were in the same transcriptional orientation as the host gene the host gene (Figure 1D); the frequency of integration in the (Figure 1D, in vitro). opposite orientation was not positively correlated with clonal abundance. Gillet et al [11] reported a significantly higher than expected proportion of in vitro integration sites within 10 kb of a RefSeq High clone abundance (Figure 1C, top two bins) was associated gene. We extended this analysis to identify the optimal (most with the presence of a host TSS within 1 kb downstream of the frequent) distance between the integration site and the nearest host provirus; here, the transcriptional orientation of the provirus had transcription start site (TSS). The results (Figure 1B) show a peak less effect on abundance than in the case of proviruses integrated preference (measured by the odds ratio, OR, observed/expected) within a host gene. The excess frequency of TSS downstream (but towards integration in proximity to TSS at ,1 kb of the integrated not upstream) was much higher in integration sites in vivo than in provirus (upstream or downstream); the OR gradually diminished vitro, in particular when the provirus was integrated in the same until it reached 1 (same as random expectation) at ,1 Mb from orientation as the nearby host gene (p(same) ,10 ; p(opposite) the integration site (Figure 1B). There was a small bias (non- ,0.05, x test). The presence of a host CpG island within 1 kb significant for in vitro integration) towards integration with a TSS downstream was also selectively associated with clone high downstream of the integration site (Figure 1C, in vitro). abundance (Figure 1E). PLOS Pathogens | www.plospathogens.org 2 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression Figure 1. Genomic environment at HTLV-1 proviral integration site determines integration in vitro and abundance in vivo. (A) Blue blocks denote a genomic feature such as a transcription start site. The distance to the nearest genomic feature is calculated (unless otherwise stated) separately for features upstream (closer to 59 LTR) and downstream of the provirus. Unless otherwise stated, distance is calculated to the nearest end of the genomic feature. Where the genomic feature has an orientation (i.e. transcription units) its orientation relative to the transcriptional orientation of the provirus is indicated as ‘‘same’’ or ‘‘opposite’’. (B) to (E): proportion of observed integration sites compared to random expectation. (B) Frequency of integration in proximity to transcriptional units (RefSeq). In vitro denotes a combined dataset from two independent experiments (see Table 1). (C) Frequency of integration within 1 kb of a TSS according to clonal abundance (cells in a given clone per 10 000 PBMCs). (D) The excess frequency (compared with random) of observing a provirus within a transcription unit was greater among abundant clones in vivo integrated in the same transcriptional orientation (blue) but not in opposite orientation (orange). (E) The excess frequency (compared with random) of observing a provirus within 1 kb of a host CpG island increased with increasing clonal abundance, in particular where the CpG island lay downstream of the integration site. doi:10.1371/journal.ppat.1003271.g001 PLOS Pathogens | www.plospathogens.org 3 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression Figure 2. Influence of host TFBS on integration site targeting. (A) Bias in integration in proximity to TFBS (based on ChIP-seq experiments), measured by the odds ratio compared to random expectation. Four representative plots are shown; see also supplementary information. The excess frequency of integration in proximity to TFBS was frequently greater in in vitro infection than in clones isolated from PBMCs in vivo, and greater in low abundance clones in vivo than high abundance clones in vivo (see bottom right panel and supplementary information). Arrows indicate a PLOS Pathogens | www.plospathogens.org 4 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression symmetrical (p300) or asymmetrical (STAT1) bias towards integration in proximity to TFBS, as well as a lower bias in close proximity to IS (STAT1). See also supplementary Table S4 for underlying data. (B) TFBS independently associated with integration frequency in vitro were identified by multivariate analysis. OR – odds ratio. TFBS shown above the line were associated with an excess frequency of integration compared with random (OR.1); TFBS below the line were significantly less likely to lie near the provirus (OR,1). Model 1 and Model 2 (carried out independently) test for TFBS within 1 kb and 100 bp of IS, respectively. doi:10.1371/journal.ppat.1003271.g002 Integration sites observed in vivo showed a similar bias towards PBMCs of 10 infected HAM/TSP patients (to preclude CTL- proximity to TFBS, with two important differences. First, the OR mediated lysis), and the CD8 population was incubated in vitro was in each case lower than that observed in in vitro integration. overnight to allow spontaneous expression of the Tax protein [15]. Second, the magnitude of the bias (OR) declined as clonal We then sorted the cells by flow cytometry to isolate Tax and abundance increased (Figure 3; supplementary Figure S3). Tax cells and analysed the integration sites in the two cell fractions. We measured the proportion of each clone that spontaneously Effect of HTLV-1 integration site on Tax expression expressed Tax by quantifying individual integration sites in the We wished to identify features of the genomic integration site + 2 Tax and Tax cells, (Figure 4E, and supplementary Figure S7). that favour expression of the HTLV-1 provirus. We hypothesized The observed proportion of Tax cells per clone varied between that the genomic environment flanking the proviral integration site 0% and 100%. The majority of clones, regardless of clonal determines the rate of spontaneous expression of the HTLV-1 + 2 abundance, were either .90% Tax or .90% Tax . This transactivator protein Tax by a given infected T-cell clone: that is, observation is consistent with the hypothesis that the rate of the proportion of cells in that clone that express Tax within a given spontaneous expression of Tax is an intrinsic property of each time interval. CD8 T-cells were depleted from fresh unstimulated clone and is determined by the proviral integration site. When the provirus was integrated within a host gene, we observed a slight but significant excess frequency of Tax cells 2 23 2 compared with Tax cells (46% vs 43% respectively, p,10 , x test). However, while the proviruses in the Tax cells were found with equal frequency in the same or the opposite transcriptional orientation to the host gene in which they were integrated, the Tax cells were significantly more frequently present in the same + 2 orientation as the host gene (52% of Tax vs 59% of Tax cells, 215 2 2 p,10 , x test). Thus, T cell clones that were 100% Tax were significantly more likely to carry a provirus in the same orientation as the host gene (Figure 4B). The relative position (upstream or downstream of the integra- tion site) and the transcriptional orientation of the nearest host gene influenced not only the clonal abundance (Figure 1) but also spontaneous Tax expression. Where the nearest host gene lay in the same transcriptional orientation as the HTLV-1 provirus, the presence of a host TSS (Figure 4A) or CpG island (Figure 4C) within 1 kb upstream of the provirus was associated with silencing of Tax, whereas a TSS or CpG island within 1 kb downstream was associated with Tax expression. The closer the upstream gene was to the integration site, the lower was the proportion of Tax cells if the gene was in the same orientation (Figure 4D). In contrast, where the nearest host gene was in opposite transcrip- tional orientation, this asymmetrical effect of the nearby host gene was not observed (Figure 4A, right-hand panel; Figure 4D). The mean proportion of Tax cells in one clone (across all clone abundance classes) was 60%. We wished to test whether proximity to TFBS would alter this proportion. We found that the presence of certain TFBS (including STAT1, cJun, NRSF) within 1 kb upstream of the integration site was associated with a higher proportion of Tax cells in the respective T-cell clone (Figure 5A). A notable exception was BRG-1, which showed a strong opposite asymmetric effect: cells containing a BRG-1 site just upstream of the provirus were more likely to be Tax , whereas cells with a BRG-1 site just downstream of the provirus were more likely to be Figure 3. Influence of host TFBS on clonal abundance. Bias in Tax (Figure 5A, top left panel). integration in proximity to TFBS (based on ChIP-seq experiments), To identify the TFBS that were independently and significantly measured by the odds ratio compared to random expectation. Two associated with spontaneous Tax expression, a logistic regression representative plots are shown; see also supplementary Figure S3. The analysis was carried out as described above (Figure 2B) for excess frequency of integration in proximity to TFBS was greater in low integration site targeting. The results (Figure 5B, see also abundance clones in vivo than high abundance clones in vivo. See also supplementary Table S7) confirmed the asymmetric effects of supplementary Table S5 for underlying data. doi:10.1371/journal.ppat.1003271.g003 the BRG-1 binding site, and in addition revealed significant PLOS Pathogens | www.plospathogens.org 5 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression Figure 4. Genomic environment at HTLV-1 proviral integration site associated with proviral expression after 18 h in culture. CD8- + 2 depleted PBMCs were placed in culture overnight and sorted by flow cytometry to isolate Tax and Tax cells, followed by integration site analysis of sorted cells. (A)–(C): proportion of observed integration sites compared to random expectation. (A) Frequency of integration in proximity to + 2 transcriptional units (RefSeq) in clones that were 100% Tax or 100% Tax , according to the relative transcriptional orientation of the provirus and the host gene. The peak of integration at 1 kb mirrors that observed in vivo in unsorted cells (Figure 1B). However, the integration site in Tax clones was more likely than in Tax clones to possess a nearby upstream TSS in the same orientation, and less likely to lie nearby a downstream TSS in the same orientation (or any relative position in the opposite orientation). (B) The provirus in Tax clones (blue) was oriented in the same transcriptional sense as the host gene in which it was integrated more frequently than random. The orientation of Tax clones (pink) did not differ from random. (C) + 2 Frequency of integration in proximity to CpG islands in clones that were 100% Tax or 100% Tax . The peak of integration at 1 kb mirrors that observed in vivo in unsorted cells and in vitro (Figure S4). (D) Mean fraction of Tax cells in clones with a TSS at a given distance (log scale) from the integration site, according to the relative transcriptional orientation of the provirus and the host TSS. The dotted line denotes the mean fraction of + + Tax cells across all clones. (E) Frequency distribution of clones according to the frequency of Tax cells in the respective clones. See supplementary Figure S7 for detailed frequency distribution separated according to clone abundance. doi:10.1371/journal.ppat.1003271.g004 asymmetric associations between Tax expression and several other Tax cells are more frequent in low-abundance clones TFBS, notably STAT1, NRSF, and HDAC1. Thus, a STAT1 To test the hypothesis that the level of Tax expression is binding site 100 bp upstream of the provirus strongly favoured Tax correlated with the in vivo abundance of the infected T cell clone, expression, but the presence of a downstream STAT1 binding site we divided all detected clones into four abundance bins based on was not an independent predictor of Tax expression after the total number of cells observed in each clone. There was a multivariate analysis. Conversely, an NRSF binding site 100 bp significant negative correlation between clone abundance and the downstream was a significant predictor of Tax negativity, but the proportion of Tax cells in the respective abundance bin (Figure 5). closest upstream NRSF binding site was not independently That is, small clones were more likely to be Tax , and this associated with Tax expression. The asymmetry of these associa- likelihood decreased as clone abundance increased. We conclude tions contrasts with the predominantly symmetrical associations that, at least in cells from HAM/TSP patients, the majority of observed between TFBS and integration site targeting (Figure 2B), spontaneous Tax expression observed is due to the large number and suggests a mechanistic interaction between transcription of the of low-abundance clones, rather than a small number of high- provirus and transcription of the flanking host genome. abundance clones. PLOS Pathogens | www.plospathogens.org 6 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression Figure 5. Influence of proximity to TFBS on Tax expression. (A) Mean fraction of Tax cells in clones with a TFBS (based on ChIP-seq experiments) at a given distance from the IS. Four representative plots are shown. (B) TFBS that were independently associated with Tax expression were identified by multivariate analysis, outcome measure . TFBS shown above the line were associated with Tax expression (OR.1); TFBS below the line were associated with Tax silencing (OR,1). Model 1 and Model 2 (carried out independently) test for TFBS within 1 kb and 100 bp of IS, respectively. doi:10.1371/journal.ppat.1003271.g005 PLOS Pathogens | www.plospathogens.org 7 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression thoroughly studied [27]; the most important is the lens epithelium- Discussion derived growth factor (LEDGF/p75, [28]), which determines An understanding of the regulation of proviral latency is integrase localization [29] and targeting of HIV integrase to required for attempts to eradicate latent retroviruses and to transcription units [21]. A study of host factors associated with optimize retroviral vectors for in vitro and in vivo use. In HIV-1 HTLV-1 integrase is currently underway. infection, a reservoir of latently infected cells persists indefinitely in The observed bias towards integration near certain TFBS was the face of antiretroviral drug therapy and precludes eradication of predominantly symmetrical and short-range, reaching a maximum the infection (reviewed in [23]). In HTLV-1 infection, proviral at 100b from the integration site and falling to random expectation expression is difficult to detect in fresh PBMCs: however, the at ,10 kb (Figure 2A). In many instances the bias dropped sharply strong, chronically activated host immune response and the at less than 100b from the integration site: we suggest that this selective oligoclonal proliferation of HTLV-1-infected T cells drop is due to steric hindrance between the pre-integration argue that the virus is continuously or intermittently expressed in complex and the DNA-bound transcription factor. vivo [4,24]. In contrast to the symmetry observed in the association between The abundance of an HTLV-1-infected T cell clone in vivo will genomic features (such as TFBS) and the frequency of initial be determined by the net effect of two main selection forces: its integration, we found significant asymmetric interactions between ability to proliferate and its susceptibility to killing by the strong the flanking host genome and the integrated provirus in CTL response [4]. If these forces acted upon all clones equally, the determining clonal abundance and spontaneous proviral expres- clones would have the same relative abundance in the host. sion. Both the relative position of the nearest host gene (upstream However, Gillet et al [11] showed a wide variation in clone or downstream of the provirus) and its relative transcriptional abundance both within and between infected individuals and over orientation showed significant associations with clone abundance time. We hypothesized that this variation between clones is caused and expression. Previous studies [1,2] reported contradictory by the genomic environment of the integrated provirus, by evidence on the role of an upstream same-sense host promoter in determining the frequency and intensity of expression of proviral either promoting or suppressing proviral transcription. More genes, in particular Tax and HBZ, which in turn promote cell recently, Shan et al [30] have shown in Bcl-2-transduced CD4 T proliferation and thereby confer a selective advantage on the cells, infected in vitro with GFP expressing modified HIV, that infected T cell clone. persistent expression of GFP was associated with opposite sense To identify the host genomic factors that determine integration orientation, while inducible expression was associated with same site targeting, we mapped and quantified proviral integration sites sense orientation. The evidence obtained here demonstrates that, isolated from two independent in vitro infection experiments. We in natural HTLV-1 infection, the presence of a same-sense host assume that the pattern of integration observed in short-term in gene promoter upstream of the integrated provirus is associated vitro infection reflects the initial pattern of integration in vivo, with inhibition of spontaneous proviral expression, suggesting the before the selection exerted during chronic infection. The results operation of transcriptional interference. We conclude that the confirmed our previous observations [10,11] that the virus is transcriptional interaction between host and HTLV-1 operates at targeted to transcriptionally active regions of the genome, within two levels. First, at a regional level – within 10 kb of the provirus – or near to a host gene. There was no bias in the orientation of the transcriptional activity of the flanking host genome favours provirus in the initial infection, indicating that the bias observed in proviral gene expression [10,11], presumably because of accessi- integration sites isolated from PBMCs is a result of the long-term bility of the euchromatin to transcription complexes. Second, at a selection forces acting on the infected clones in vivo. local level – within 100b to 1000b – transcriptional interference by We observed a bias towards integration in proximity to a same-sense host promoter within 1 kb upstream can override the particular transcription factor binding sites. This bias was regional effect and inhibit proviral transcription. remarkably strong in certain cases (STAT1, NRSF) in single- Two observations reported here demonstrate that proviral factor analysis. Because clusters of different TFBS are frequent in integration and expression are not determined simply by the the genome [22], we carried out a multivariate (logistic regression) accessibility of chromatin. First, whereas some TFBS were analysis to identify the TFBS that were independently and associated with HTLV-1 proviral integration more frequently significantly associated with an excess frequency of integration. than expected, the frequency of other TFBS showed no such bias. The results (Figure 2B) confirmed the identification of p53, Second, the asymmetric associations observed between proviral HDAC6 and STAT1 as significant independent correlates of orientation and position with respect to flanking host genes and the integration. Further independent predictors of integration includ- abundance and expression of the HTLV-1 provirus argue for a ed Ini1 (see below), cMyc, cJun and NF-kB (Figure 2B). p53 and mechanistic interaction between transcription of the HTLV-1 STAT1 both play important roles in HTLV-1 infection. HTLV-1 provirus and transcription of the flanking host genome. dysregulates p53 signalling pathways in vivo [25]; it is not known An observation of particular interest is the opposing effect of a whether insertional mutagenesis contributes to this dysregulation. BRG-1 binding site upstream and downstream of the provirus HTLV-1 also causes widespread activation of interferon-stimulat- (Figure 5A). BRG-1, one of the two ATPase components required ed genes in vivo, including the key transcriptional regulator for the activity of the SWI/SNF complex [31], controls gene STAT1 [25]. A strong association was reported between STAT1 expression by remodelling chromatin, i.e. by repositioning and MLV integration [26]; the authors attributed this to an nucleosomes to control the access of transcriptional complexes to association between MLV integration and particular epigenetic the DNA. BRG-1 can cause both gene repression [32] and gene marks (H3K4me3, H3K4me1 and H3K9ac) at the integration site. activation [33]; the balance appears to depend on which other The proportion of all integration sites near any one TFBS was subunits are recruited to the SWI/SNF complex [34]. Easley et al in the minority. This observation indicates that proximity to the [35] found that BRG-1 is required for Tax expression and HTLV- transcription factor binding site itself is not sufficient for 1 replication in vitro, and Rafati et al [36] found that the BAF integration, but suggests that these transcription factors (or an subclass of the SWI/SNF complex repressed HIV-1 transcription associated host factor) increase the efficiency of proviral integra- whereas the PBAF subclass promoted transcription. Our observa- tion. Host factors associated with HIV integrase have been tion (Figure 5A) that a BRG-1 site upstream of the provirus is PLOS Pathogens | www.plospathogens.org 8 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression associated with silencing of Tax, while a BRG-1 site downstream is associated with Tax expression, is consistent with our conclusion (above) that transcriptional interference dominates the transcrip- tional interaction between the provirus and the flanking host genome. The Ini1 subunit of SWI/SNF interacts directly with HIV-1 integrase [37]; a fraction of Ini1 moves transiently to the cytoplasm to associate with the HIV-1 preintegration complex [38]. However, the function of Ini1 in HIV-1 proviral integration and expression in vivo is not understood. We found that genomic sites for Ini1 binding were significantly associated with HTLV-1 proviral integration (Figure 2B). The presence of an Ini1 site 1 kb downstream of the provirus was associated with spontaneous Tax expression, similar to the effect of the downstream BRG-1 site. Finally, the SWI/SNF subunit BAF155 was overrepresented 1 kb upstream of the integration site (Figure 2B), and was associated with Tax silencing when present either upstream or downstream of the provirus (Figure 5B). The HTLV-1 Tax protein acts in concert with host cell Figure 6. Tax cells are more frequent in smaller clones. Mean fraction of Tax cells in bins of increasing clonal abundance (total transcription factors (notably CBP/p300) on the promoter/ number of cells in each respective clone). The fraction of Tax cells was enhancer in the viral 59 LTR, driving plus-strand transcription 216 2 negatively correlated with clonal abundance (p,10 , x test for in a strong positive feedback loop. Tax also acts on response trend). elements for NF-kB, CREB and the serum response factor (SRE) doi:10.1371/journal.ppat.1003271.g006 to upregulate expression of a wide range of host genes [39]. Finally, Tax promotes cell cycle progression by accelerating difference in amino acid sequence found between one clone and passage through G1 and inhibiting the G1/S and G2/M the others (data not shown). checkpoints [40]. The net effect of Tax expression is therefore to Interestingly, while Tax cells were more frequent in low- drive activation and proliferation of the infected T cell. We abundance clones, certain features favouring proviral expression previously reported that spontaneous Tax expression in fresh (e.g. a downstream host TSS) also favoured clonal expansion. The unstimulated PBMCs was associated with proliferation of the association between clonal abundance and proviral integration respective cell in vivo [41]. We therefore expected to observe a within 1 kb of a downstream host TSS was maintained even positive correlation between the frequency of spontaneous Tax within Tax clones, consistent with the idea that the selective expression by a given clone ex vivo and the abundance of that expansion of these clones is driven by other proviral genes. clone in vivo. However, the results obtained here (Figure 6) These observations raise the possibility that the equilibrium demonstrate the opposite, i.e. a highly significant negative abundance of an HTLV-1-infected T cell clone in vivo is correlation. This correlation is likely to be caused by the strong determined not by Tax but by the HBZ gene, encoded on the host immune response to the virus. The Tax protein is highly negative strand of the provirus. Satou et al showed that HBZ immunodominant in the class 1 MHC-restricted cytotoxic T mRNA promoted proliferation of the infected cell, and whereas lymphocyte (CTL) response to HTLV-1 [14,42], and the tax gene Tax expression is frequently undetectable, HBZ appears to be is frequently silenced in vivo by mutation or epigenetic changes persistently expressed in fresh cells isolated from both non- such as DNA methylation in both untransformed and malignantly malignant cases of HTLV-1 infection and cases of adult T-cell transformed (leukemic) cells [43]. Cells that express a high level of leukemia/lymphoma [16]. Further, Macnamara et al recently Tax are killed by CTLs faster than low Tax-expressing cells [44]. showed that the CTL response to HBZ is a critical determinant of Therefore, suppression or loss of Tax expression may confer a the equilibrium proviral load in vivo [17] survival advantage on the infected clone in vivo. We conclude that In this study we examined Tax expression only among CD4 T the small (low-abundance) HTLV-1-infected clones express Tax at cells. A small percentage of infected cells are CD8 T cells [48,49]; a higher rate and turn over faster in vivo than the high-abundance it is possible that the genomic factors that determine targeting and clones. It is possible that the critical role of Tax in the HTLV-1 expression of HTLV-1 differ in CD8 cells. Also, the propensity of lifecycle is not to maintain clone abundance but rather to promote a cell to express Tax was measured by quantifying the frequency of virion production and infection of new cells by cell contact via the spontaneous Tax expression after 18 hours incubation in vitro. virological synapse [45,46]. Two lines of evidence suggest that this measure is relevant to The negative correlation observed between clone abundance HTLV-1 infection and pathogenesis in vivo. First, cells which and the percentage of Tax cells, although it was highly significant express Tax ex vivo turn over faster in vivo [41]. Second, the in all patients combined, was not uniform in every patient. In a proportion of CD4 cells that express Tax after overnight culture small number of patients (in particular those with a high is significantly associated with the HTLV-1 inflammatory disease oligoclonality index, Supplementary Figure S8), the most abun- HAM/TSP [50]. The individuals studied here were all patients dant clones (bin 4, clones with greater than 10 cells) contained a with HAM/TSP: the mean level of spontaneous Tax expression is high proportion of Tax cells, suggesting that certain antigen- lower in asymptomatic HTLV-1 carriers, but it is unlikely that the expressing clones escaped control by the immune response, for molecular mechanisms that govern proviral latency differ quali- example by CTL escape mutations in the tax gene [47]. However, tatively between asymptomatic carriers and patients with HAM/ clone-specific sequencing of exon 3 of the tax gene of 38 clones TSP. (.10 cells) from 8 patients did not reveal significant differences in It will be important to compare the present results with the the occurrence of Tax mutations between clones with a high or genomic factors associated with HBZ expression or silencing. At low frequency of Tax cells; and only in one patient was a PLOS Pathogens | www.plospathogens.org 9 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression present this cannot be done by flow-sorting because existing HBZ- Table 1. IS datasets used. specific antibodies are insufficiently sensitive to detect the low expression levels of HBZ protein in primary cells. We are currently testing the hypothesis that Tax-specific and HBZ-specific CTL total infected clones selectively lyse different clonal populations in vitro. Dataset total IS individuals reference We have identified host genomic factors that determine the in vitro (1) 4521 N/A [11] integration site, the proviral expression and selective clonal in vitro (2) 1805 N/A This publication expansion of HTLV-1 in natural infection in vivo: these factors In vivo (1) 78563 63 [11] are summarized in Figure 7. These results open the way to test the molecular mechanisms involved. In vivo (2) 20202 10 This publication Tax Negative 6700 10 (pooled) This publication Materials and Methods Tax Positive 13054 10 (pooled) This publication Random UIS 176505 N/A This publication Ethics statement Blood samples were donated by HTLV-1-infected individuals 1 For the purpose of this work, only one time point was used for each patient attending the HTLV-1 clinic at the National Centre for Human (most recent available if multiple time points were originally analysed). Retrovirology (Imperial College Healthcare NHS trust) at St doi:10.1371/journal.ppat.1003271.t001 Mary’s Hospital, London UK, with fully informed written consent. This study was approved by the UK National Research Ethics In vitro infection Service (NRES reference 09/H0606/106). In vitro infection was carried out in two independent assays as previously described [11]. Jurkat (JKT) cells were co-cultured for DNA samples (Table 1) 3 h with c-irradiated ( Cs, 40,000 cGy) MT2 cells [51], labelled PBMCs were isolated using Histopaque-1077 (Sigma-Aldrich) with anti-CD4 MicroBeads (Miltenyi). MT2 cells were then and cryopreserved in FBS (Gibco) containing 10% DMSO depleted from the co-culture using magnetic separation (Miltenyi), (Sigma-Aldrich). DNA extraction was carried out using the and infected JKT cells were maintained in culture for 14 days in DNeasy Blood & Tissue kit (Qiagen) according to the manufac- RPMI (supplemented by L-glutamine, penicillin, streptomycin) turer’s protocol. containing 10% FBS for 18 hours at 37C with 5% CO . Genomic Figure 7. Genomic correlates of HTLV-1 proviral targeting, clonal expansion and proviral expression. (A) Factors associated with the presence or absence of spontaneous Tax expression by a given cell after short-term (18 h) in vitro incubation. (B) Features of the genomic environment of the provirus associated either with initial integration (left panel), or clonal expansion in vivo (right panel). Findings were made in the present study unless otherwise stated. TSS – transcription start site. TFBS – transcription factor binding site. doi:10.1371/journal.ppat.1003271.g007 PLOS Pathogens | www.plospathogens.org 10 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression DNA was extracted and the proviral integration sites (IS) analysed Bioinformatic analysis of genomic environment as previously described [11]. IS from MT2 were also analysed to Transcription units and CpG island data were retrieved from exclude possible contamination of the JKT IS. No contaminating the NCBI (ftp.ncbi.nih.gov/gene/) and UCSC tables [56], MT2 IS were found after 14 days. respectively. Annotations to the human genome were obtained from published datasets (Table S3) including ChIP-seq experi- ments on primary CD4 T cells where available; otherwise, data Tax sorting on human CD4 T cell lines or other human cell lines were used. See also supplementary Figure S5, supplementary Table S6. We used the SISSRs algorithm [57] to identify the position of a PBMCs from 10 patients with HAM/TSP with a high proviral putative transcription factor binding site in published ChIP-seq load (range 12.2–50.6 copies per 100 PBMC) were depleted of data where raw ChIP-seq data were available. CD8 cells using magnetic depletion (Miltenyi) and incubated in Annotations positions were compared to the IS using the R RPMI (supplemented by L-glutamine, penicillin, streptomycin) package hiAnnotator (http://malnirav.github.com/hiAnnotator), containing 10% FBS for 18 hours at 37C with 5% CO . After kindly provided by N. Malani and F. Bushman (University of 18 h culture, the cells were stained for intracellular expression of Pennsylvania, USA). Tax (anti-Tax mAb LT4) and sorted using FACS (FACSAria IIIU, BD Biosciences) to isolate two populations of live CD4 cells based on Tax expression. Gates were set (FACSDiva, BD Biosciences) to Statistical analysis + 2 ensure a clear demarcation between the Tax and Tax Statistical analysis was carried out using R version 2.13.0 populations (Figure S6). DNA was extracted from whole unsorted (http://www.R-project.org/). Two separate logistic regression analyses were carried out, respectively, to identify independent PBMCs from each patient and analysed separately to identify the predictors of HTLV-1 integration targeting and independent patient of origin of each clone; 46% of the clones were attributed predictors of Tax positivity. Genomic annotations used to derive in this way. To calculate the fraction of Tax cells in a given clone, + 2 input variable were published ChIP-seq datasets (see Bioinformatic the frequency of Tax and Tax cells were normalized to the mass analysis above; Table S3). For integration targeting, the binary of genomic DNA per cell from each respective cell population, to outcome measure was a ‘‘true’’ integration site (from 4521 correct for experimental variation in efficiency of genomic DNA identified in vitro integration sites) or a ‘‘false’’ integration site isolation (Table S6), (45210 random genomic locations). For spontaneous Tax expression, the binary outcome was Tax positivity (20813 Tax Analysis of IS cells) or Tax negativity (10326 Tax cells). Each TFBS was tested Identification and quantification of proviral integration sites was (presence or absence of the TFBS within a given distance of the done as previously described [11]. HTLV-1 infected DNA was integration site) as an independent predictor in each analysis. For randomly sheared by sonication (Covaris S2) and then blunt- each outcome variable, two separate analyses were carried out, ended (Klenow polymerase) and ligated to a partly double- respectively at two distances of the integration site - 100 bases and stranded DNA linker. Following a nested PCR step, the resulting 1 kb. DNA libraries were deep sequenced using the Illumina GA-II First, for each TFBS and at each distance, we tested whether the platform. DNA sequence was aligned to the human genome relative position (upstream/downstream) of the integration site reference (UCSC hg18, excluding haplotype and ‘‘random’’ and the TFBS determined the outcome by using a likelihood ratio sequences) using the ELAND algorithm. Distinct IS were grouped test to compare two competing models: 1) presence or absence of based on integration site and quantified based on number of TFBS upstream or (separately) downstream; 2) presence or distinct shear sites isolated and the respective patient’s proviral absence of TFBS, regardless of relative position. Next, we carried load. out univariate analysis of each individual TFBS, based on the DNA sequences from ,190000 random sites in the human model chosen by the likelihood ratio test. Only TFBS that were genome (hg18) were generated using Galaxy [52,53,54] and back- significant (p-value,0.05) after correction for multiple compari- aligned to the human genome using the same pipeline to eliminate sons (Benjamini-Hochberg) were used in the multivariate analysis. any potential bias due to alignment limitations. Multivariate analysis was carried out using a step-down logistic regression method. Calculation of clonal abundance The absolute abundance of a given clone was defined as the Supporting Information number of proviral copies of that clone per 10 PBMCs. Given n - th the number of proviral copies for the i clone, and S – the total Figure S1 Influence of host TFBS on integration site number of clones identified in the sample, the absolute abundance targeting – in vitro. Bias in integration in proximity to TFBS was calculated for PBMC samples according to the following (based on ChIP-seq experiments), measured by the odds ratio formula: compared to random expectation. The bias was maintained across separate datasets, generated by independent in vitro experiments. Dotted line denotes random expectation (OR = 1). absolute abundance~PVL| (TIF) Figure S2 Influence of host TFBS on integration site i~1 targeting – in vivo. Bias in integration in proximity to TFBS Clone abundance bins were defined on a logarithmic scale since (based on ChIP-seq experiments), measured by the odds ratio proviral load (used in calculation of abundance) follows a compared to random expectation. The pattern of bias was logarithmic distribution [55]. The number of clones in each clone maintained between different patient clinical groups. Dotted line abundance bin is given in Table S1. For samples sorted for Tax denotes random expectation (OR = 1). ATLL = Adult T-cell protein expression, where proviral load data were not available, leukaemia/lymphoma . HAM/TSP = HTLV-1 associated mye- the clonal abundance bins were set according to proviral copy lopathy/Tropical spastic paraparesis. AC = Asymptomatic carrier. count. (TIF) PLOS Pathogens | www.plospathogens.org 11 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression Figure S3 Influence of host TFBS on clonal abundance combined (P,10 ). However in certain patients (particularly in vivo. Bias in frequency of integration in proximity to TFBS those with a high oligoclonality Index) the most abundant clones (based on ChIP-seq experiments), measured by the odds ratio contained a high proportion of Tax cells. Clone abundance bins compared to random expectation. TFBS that were associated with were defined as in Figure S7. integration targeting showed a stronger bias (higher OR) in the (TIF) clones least expanded in vivo. Clonal abundance is expressed as 4 Table S1 In vivo integration sites – sample data by the number of cells in given clone per 10 PBMCs. Dotted line clone abundance. denotes random expectation (OR = 1). (DOC) (TIF) Table S2 In vivo integration sites – sample data by Figure S4 The genomic environment at the HTLV-1 patient code. proviral integration site determines integration target- (DOC) ing in vitro and clonal abundance in vivo. Frequency of integration in proximity to CpG islands in clones for in vitro (in Table S3 List of annotations datasets used. blue) and in vivo (purple) integration. (DOC) (TIF) Table S4 Odds ratio and clone counts data for a Figure S5 Protocol for flow-sorting of Tax-expressing selection of TFBS – in vitro, in vivo vs. random sites. cells. (A) CD8 cell-depleted PBMCs were studied from 10 (XLS) patients with HAM/TSP with a high proviral load. The cells were Table S5 Odds ratio and clone counts data for a incubated overnight, fixed and stained for Tax and surface CD4 selection of TFBS – clonal abundance bins vs. random expression, and sorted on a high-speed flow cytometer (see Figure sites. S6 for details). (B) Recovered cells from all 10 patients were + + (XLS) combined in two pools, respectively CD4 Tax cells and + 2 CD4 Tax cells. (C) Genomic DNA was extracted from each Table S6 Tax sorting experiment – sample data. pool of cells and integration site analysis carried out as described. (DOC) (TIF) Table S7 Multivariate analysis results – detailed odds Figure S6 Flow cytometry sorting by Tax expression. (A) ratios and confidence intervals. Representative FACS plots of the gating procedure used (from 1 of (XLS) 10 samples studied). Lower middle panel shows gating of + + + 2 CD4 Tax (‘Tax pos’) and CD4 Tax (‘Tax neg’) populations; Acknowledgments these gates were set to distinguish unequivocally between Tax and 2 2 + Tax populations. (B) Purity testing of Tax sorted cells: Tax We thank Nirav Malani and Frederic D. Bushman at the department of cells not detected. C) Purity testing of Tax sorted cells: 0.2% were Microbiology, University of Pennsylvania, Philadelphia, PA, USA for the Tax . list of random integration sites and for developing software packages; and Laurence Game, Nathalie Lambie and Adam Giess of the Genomics (TIF) Laboratory at the MRC Clinical Sciences Centre, Hammersmith Hospital, Figure S7 Majority of HTLV-1-infected clones were London UK and Robert Sampson at the flow cytometry facility at St + + either 100% Tax or 0% Tax . Frequency distribution of Mary’s Campus, Imperial College, London UK. Finally we thank Lucy clones according to the frequency of Tax cells in each respective Cook, Heather Niederer and Becca Asquith and her team at the Section of clone, binned according to number of sister cells detected in Immunology (Imperial College) for discussions and comments, and the staff and donors at the National Centre for Human Retrovirology (Imperial sample: bin 1: 1 cell detected; bin 2: 2 or 3 cells detected; bin 3: 4 College Healthcare NHS trust) at St Mary’s Hospital, London UK. to 10 cells detected; bin 4: over 10 cells detected. (TIF) Author Contributions Figure S8 Tax cells were more frequent in smaller Conceived and designed the experiments: AM GPT CRMB. Performed clones. Mean fraction of Tax cells within each bin, in bins of the experiments: AM NAG. Analyzed the data: AM. Contributed increasing clonal abundance (total number of cells in each reagents/materials/analysis tools: YT GPT NAG DJL. Wrote the paper: respective clone). In the majority of patients there was an inverse AM CRMB. Clinical diagnosis and assessment: GPT. Designed the correlation between clone abundance bin and fraction of Tax statistical analysis: AM DJL CRMB. cells: this correlation was highly significant in all patients References 1. Han Y, Lin YB, An W, Xu J, Yang HC, et al. (2008) Orientation-dependent analysis of 239 HAM/TSP patients including 64 patients followed up for 10 regulation of integrated HIV-1 expression by host gene transcriptional years. J Neurovirol 7: 228–234. readthrough. Cell Host Microbe 4: 134–146. 7. Wattel E, Vartanian JP, Pannetier C, Wain-Hobson S (1995) Clonal expansion 2. Lenasi T, Contreras X, Peterlin BM (2008) Transcriptional interference antagonizes of human T-cell leukemia virus type I-infected cells in asymptomatic and proviral gene expression to promote HIV latency. Cell Host Microbe 4: 123–133. symptomatic carriers without malignancy. J Virol 69: 2863–2868. 3. Igakura T, Stinchcombe JC, Goon PK, Taylor GP, Weber JN, et al. (2003) 8. Cook LB, Rowan AG, Melamed A, Taylor GP, Bangham CR (2012) HTLV-1-infected Spread of HTLV-I between lymphocytes by virus-induced polarization of the T cells contain a single integrated provirus in natural infection. Blood 120: 3488–3490. cytoskeleton. Science 299: 1713–1716. 9. Derse D, Crise B, Li Y, Princler G, Lum N, et al. (2007) Human T-cell leukemia 4. Bangham CR (2009) CTL quality and the control of human retroviral infections. virus type 1 integration target sites in the human genome: comparison with those Eur J Immunol 39: 1700–1712. of other retroviruses. J Virol 81: 6731–6741. 5. Iwanaga M, Watanabe T, Utsunomiya A, Okayama A, Uchimaru K, et al. 10. Meekings KN, Leipzig J, Bushman FD, Taylor GP, Bangham CR (2008) (2010) Human T-cell leukemia virus type I (HTLV-1) proviral load and disease HTLV-1 integration into transcriptionally active genomic regions is associated progression in asymptomatic HTLV-1 carriers: a nationwide prospective study with proviral expression and with HAM/TSP. PLoS Pathog 4: e1000027. in Japan. Blood 116: 1211–1219. 11. Gillet NA, Malani N, Melamed A, Gormley N, Carter R, et al. (2011) The host 6. Matsuzaki T, Nakagawa M, Nagai M, Usuku K, Higuchi I, et al. (2001) HTLV-I genomic environment of the provirus determines the abundance of HTLV-1- proviral load correlates with progression of motor disability in HAM/TSP: infected T-cell clones. Blood 117: 3113–3122. PLOS Pathogens | www.plospathogens.org 12 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression 12. Jacobson S, Shida H, McFarlin DE, Fauci AS, Koenig S (1990) Circulating 35. Easley R, Carpio L, Guendel I, Klase Z, Choi S, et al. (2010) Human T- CD8+ cytotoxic T lymphocytes specific for HTLV-I pX in patients with HTLV- lymphotropic virus type 1 transcription and chromatin-remodeling complexes. I associated neurological disease. Nature 348: 245–248. J Virol 84: 4755–4768. 13. Parker CE, Daenke S, Nightingale S, Bangham CR (1992) Activated, HTLV-1- 36. Rafati H, Parra M, Hakre S, Moshkin Y, Verdin E, et al. (2011) Repressive LTR specific cytotoxic T-lymphocytes are found in healthy seropositives as well as in nucleosome positioning by the BAF complex is required for HIV latency. PLoS patients with tropical spastic paraparesis. Virology 188: 628–636. Biol 9: e1001206. 14. Goon PK, Biancardi A, Fast N, Igakura T, Hanon E, et al. (2004) Human T cell 37. Kalpana GV, Marmon S, Wang W, Crabtree GR, Goff SP (1994) Binding and lymphotropic virus (HTLV) type-1-specific CD8+ T cells: frequency and stimulation of HIV-1 integrase by a human homolog of yeast transcription factor immunodominance hierarchy. J Infect Dis 189: 2294–2298. SNF5. Science 266: 2002–2006. 15. Hanon E, Hall S, Taylor GP, Saito M, Davis R, et al. (2000) Abundant tax 38. Turelli P, Doucas V, Craig E, Mangeat B, Klages N, et al. (2001) Cytoplasmic protein expression in CD4+ T cells infected with human T-cell lymphotropic recruitment of INI1 and PML on incoming HIV preintegration complexes: virus type I (HTLV-I) is prevented by cytotoxic T lymphocytes. Blood 95: 1386– interference with early steps of viral replication. Mol Cell 7: 1245–1254. 1392. 39. Matsuoka M, Jeang KT (2011) Human T-cell leukemia virus type 1 (HTLV-1) 16. Satou Y, Yasunaga J, Yoshida M, Matsuoka M (2006) HTLV-I basic leucine and leukemic transformation: viral infectivity, Tax, HBZ and therapy. zipper factor gene mRNA supports proliferation of adult T cell leukemia cells. Oncogene 30: 1379–1389. Proc Natl Acad Sci U S A 103: 720–725. 40. Marriott SJ, Semmes OJ (2005) Impact of HTLV-I Tax on cell cycle progression 17. Macnamara A, Rowan A, Hilburn S, Kadolsky U, Fujiwara H, et al. (2010) and the cellular DNA damage repair response. Oncogene 24: 5986–5995. HLA class I binding of HBZ determines outcome in HTLV-1 infection. PLoS 41. Asquith B, Zhang Y, Mosley AJ, de Lara CM, Wallace DL, et al. (2007) In vivo Pathog 6: e1001117. T lymphocyte dynamics in humans and the impact of human T-lymphotropic 18. Hilburn S, Rowan A, Demontis MA, MacNamara A, Asquith B, et al. (2011) In virus 1 infection. Proc Natl Acad Sci U S A 104: 8035–8040. vivo expression of human T-lymphotropic virus type 1 basic leucine-zipper 42. Kannagi M, Harada S, Maruyama I, Inoko H, Igarashi H, et al. (1991) protein generates specific CD8+ and CD4+ T-lymphocyte responses that Predominant recognition of human T cell leukemia virus type I (HTLV-I) pX correlate with clinical outcome. J Infect Dis 203: 529–536. gene products by human CD8+ cytotoxic T cells directed against HTLV-I- 19. Landry S, Halin M, Vargas A, Lemasson I, Mesnard JM, et al. (2009) infected cells. Int Immunol 3: 761–767. Upregulation of human T-cell leukemia virus type 1 antisense transcription by 43. Taniguchi Y, Nosaka K, Yasunaga J, Maeda M, Mueller N, et al. (2005) the viral tax protein. J Virol 83: 2048–2054. Silencing of human T-cell leukemia virus type I gene transcription by epigenetic 20. Gaudray G, Gachon F, Basbous J, Biard-Piechaczyk M, Devaux C, et al. (2002) mechanisms. Retrovirology 2: 64. The complementary strand of the human T-cell leukemia virus type 1 RNA 44. Kattan T, MacNamara A, Rowan AG, Nose H, Mosley AJ, et al. (2009) The genome encodes a bZIP transcription factor that down-regulates viral avidity and lytic efficiency of the CTL response to HTLV-1. J Immunol 182: transcription. J Virol 76: 12813–12822. 5723–5729. 21. Ciuffi A, Llano M, Poeschla E, Hoffmann C, Leipzig J, et al. (2005) A role for 45. Nejmeddine M, Barnard AL, Tanaka Y, Taylor GP, Bangham CR (2005) LEDGF/p75 in targeting HIV DNA integration. Nat Med 11: 1287–1289. Human T-lymphotropic virus, type 1, tax protein triggers microtubule 22. Bernstein BE, Birney E, Dunham I, Green ED, Gunter C, et al. (2012) An reorientation in the virological synapse. J Biol Chem 280: 29653–29660. integrated encyclopedia of DNA elements in the human genome. Nature 489: 46. Nejmeddine M, Negi VS, Mukherjee S, Tanaka Y, Orth K, et al. (2009) HTLV- 57–74. 1-Tax and ICAM-1 act on T-cell signal pathways to polarize the microtubule- 23. Durand CM, Blankson JN, Siliciano RF (2012) Developing strategies for HIV-1 organizing center at the virological synapse. Blood 114: 1016–1025. eradication. Trends Immunol 33: 554–562. 47. Niewiesk S, Daenke S, Parker CE, Taylor G, Weber J, et al. (1995) Naturally 24. Bangham CR, Meekings K, Toulza F, Nejmeddine M, Majorovits E, et al. occurring variants of human T-cell leukemia virus type I Tax protein impair its (2009) The immune control of HTLV-1 infection: selection forces and dynamics. recognition by cytotoxic T lymphocytes and the transactivation function of Tax. Front Biosci 14: 2889–2903. J Virol 69: 2649–2653. 25. Tattermusch S, Skinner JA, Chaussabel D, Banchereau J, Berry MP, et al. 48. Richardson JH, Edwards AJ, Cruickshank JK, Rudge P, Dalgleish AG (1990) In (2012) Systems biology approaches reveal a specific interferon-inducible vivo cellular tropism of human T-cell leukemia virus type 1. J Virol 64: 5682– signature in HTLV-1 associated myelopathy. PLoS Pathog 8: e1002480. 5687. 26. Santoni FA, Hartley O, Luban J (2010) Deciphering the code for retroviral 49. Hanon E, Stinchcombe JC, Saito M, Asquith BE, Taylor GP, et al. (2000) integration target site selection. PLoS Comput Biol 6: e1001008. Fratricide among CD8(+) T lymphocytes naturally infected with human T cell 27. Rain JC, Cribier A, Gerard A, Emiliani S, Benarous R (2009) Yeast two-hybrid lymphotropic virus type I. Immunity 13: 657–664. detection of integrase-host factor interactions. Methods 47: 291–297. 50. Asquith B, Mosley AJ, Heaps A, Tanaka Y, Taylor GP, et al. (2005) 28. Cherepanov P, Maertens G, Proost P, Devreese B, Van Beeumen J, et al. (2003) Quantification of the virus-host interaction in human T lymphotropic virus I HIV-1 integrase forms stable tetramers and associates with LEDGF/p75 protein infection. Retrovirology 2: 75. in human cells. J Biol Chem 278: 372–381. 51. Miyoshi I, Kubonishi I, Sumida M, Hiraki S, Tsubota T, et al. (1980) A novel T- 29. Maertens G, Cherepanov P, Pluymers W, Busschots K, De Clercq E, et al. cell line derived from adult T-cell leukemia. Gann 71: 155–156. (2003) LEDGF/p75 is essential for nuclear and chromosomal targeting of HIV-1 52. Giardine B, Riemer C, Hardison RC, Burhans R, Elnitski L, et al. (2005) integrase in human cells. J Biol Chem 278: 33528–33539. Galaxy: a platform for interactive large-scale genome analysis. Genome Res 15: 30. Shan L, Yang HC, Rabi SA, Bravo HC, Shroff NS, et al. (2011) Influence of 1451–1455. host gene transcription level and orientation on HIV-1 latency in a primary-cell 53. Goecks J, Nekrutenko A, Taylor J (2010) Galaxy: a comprehensive approach for model. J Virol 85: 5384–5393. supporting accessible, reproducible, and transparent computational research in 31. Euskirchen G, Auerbach RK, Snyder M (2012) SWI/SNF Chromatin- the life sciences. Genome Biol 11: R86. remodeling Factors: Multiscale Analyses and Diverse Functions. J Biol Chem 54. Blankenberg D, Von Kuster G, Coraor N, Ananda G, Lazarus R, et al. (2010) 287: 30897–30905. Galaxy: a web-based genome analysis tool for experimentalists. Curr Protoc Mol 32. Ho L, Jothi R, Ronan JL, Cui K, Zhao K, et al. (2009) An embryonic stem cell Biol Chapter 19: Unit 19 10 11–21. chromatin remodeling complex, esBAF, is an essential component of the core 55. Nagai M, Usuku K, Matsumoto W, Kodama D, Takenouchi N, et al. (1998) pluripotency transcriptional network. Proc Natl Acad Sci U S A 106: 5187– Analysis of HTLV-I proviral load in 202 HAM/TSP patients and 243 5191. asymptomatic HTLV-I carriers: high proviral load strongly predisposes to 33. De S, Wurster AL, Precht P, Wood WH, 3rd, Becker KG, et al. (2011) Dynamic HAM/TSP. J Neurovirol 4: 586–593. BRG1 recruitment during T helper differentiation and activation reveals distal 56. Karolchik D, Hinrichs AS, Furey TS, Roskin KM, Sugnet CW, et al. (2004) The regulatory elements. Mol Cell Biol 31: 1512–1527. UCSC Table Browser data retrieval tool. Nucleic Acids Res 32: D493–496. 34. Euskirchen GM, Auerbach RK, Davidov E, Gianoulis TA, Zhong G, et al. 57. Jothi R, Cuddapah S, Barski A, Cui K, Zhao K (2008) Genome-wide (2011) Diverse roles and interactions of the SWI/SNF chromatin remodeling identification of in vivo protein-DNA binding sites from ChIP-Seq data. Nucleic complex revealed using global approaches. PLoS Genet 7: e1002008. Acids Res 36: 5221–5231. PLOS Pathogens | www.plospathogens.org 13 March 2013 | Volume 9 | Issue 3 | e1003271 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png PLoS Pathogens Public Library of Science (PLoS) Journal

Genome-wide Determinants of Proviral Targeting, Clonal Abundance and Expression in Natural HTLV-1 Infection

PLoS Pathogens , Volume 9 (3): e1003271 – Mar 21, 2013

Loading next page...
 
/lp/public-library-of-science-plos-journal/genome-wide-determinants-of-proviral-targeting-clonal-abundance-and-gDERSFiPUd

References (119)

Publisher
Public Library of Science (PLoS) Journal
Copyright
Copyright: © 2013 Melamed et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: This project was supported by the Wellcome Trust (www.wellcome.ac.uk/), grant number P08165. We are grateful for support from the Imperial NIHR Biomedical Research Centre funding scheme. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist.
Subject
Research Article; Biology; Microbiology; Virology; Viral classification; Retroviruses; Virology
ISSN
1553-7366
eISSN
1553-7374
DOI
10.1371/journal.ppat.1003271
Publisher site
See Article on Publisher Site

Abstract

The regulation of proviral latency is a central problem in retrovirology. We postulate that the genomic integration site of human T lymphotropic virus type 1 (HTLV-1) determines the pattern of expression of the provirus, which in turn determines the abundance and pathogenic potential of infected T cell clones in vivo. We recently developed a high-throughput method for the genome-wide amplification, identification and quantification of proviral integration sites. Here, we used this protocol to test two hypotheses. First, that binding sites for transcription factors and chromatin remodelling factors in the genome flanking the proviral integration site of HTLV-1 are associated with integration targeting, spontaneous proviral expression, and in vivo clonal abundance. Second, that the transcriptional orientation of the HTLV-1 provirus relative to that of the nearest host gene determines spontaneous proviral expression and in vivo clonal abundance. Integration targeting was strongly associated with the presence of a binding site for specific host transcription factors, especially STAT1 and p53. The presence of the chromatin remodelling factors BRG1 and INI1 and certain host transcription factors either upstream or downstream of the provirus was associated respectively with silencing or spontaneous expression of the provirus. Cells expressing HTLV-1 Tax protein were significantly more frequent in clones of low abundance in vivo. We conclude that transcriptional interference and chromatin remodelling are critical determinants of proviral latency in natural HTLV-1 infection. Citation: Melamed A, Laydon DJ, Gillet NA, Tanaka Y, Taylor GP, et al. (2013) Genome-wide Determinants of Proviral Targeting, Clonal Abundance and Expression in Natural HTLV-1 Infection. PLoS Pathog 9(3): e1003271. doi:10.1371/journal.ppat.1003271 Editor: Michael Emerman, Fred Hutchinson Cancer Research Center, United States of America Received November 6, 2012; Accepted February 10, 2013; Published March 21, 2013 Copyright:  2013 Melamed et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: This project was supported by the Wellcome Trust (www.wellcome.ac.uk/), grant number P08165. We are grateful for support from the Imperial NIHR Biomedical Research Centre funding scheme. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have declared that no competing interests exist. * E-mail: c.bangham@imperial.ac.uk Mitotic spread of HTLV-1 results in expanded clones of cells Introduction that carry the provirus in the same genomic integration site [7]. It is poorly understood how the flanking host genome influences Infectious spread results in integration of the provirus at a new transcription of an integrated provirus. Experiments on artificially genomic position. We have recently shown that the majority of modified proviral reporter constructs have yielded contradictory naturally infected T-cell clones carry a single proviral copy [8]. evidence on the role of flanking host promoters in either driving Integration of HTLV-1 does not favour specific hotspots, but is proviral transcription, or suppressing it by transcriptional inter- more frequent in transcriptionally active areas of the genome ference [1,2]. Conclusions from experiments on single artificial [9,10,11]. However, the factors that determine integration clones therefore cannot be reliably generalized: evidence is targeting and the abundance and expression of the HTLV-1 required from genome-wide studies of integrated proviruses in provirus in vivo are unknown. Two HTLV-1 gene products are natural infection. thought to play a crucial role in viral persistence in vivo. Tax, the Human T lymphotropic virus Type 1 (HTLV-1) persists in vivo by transcriptional transactivator of the virus, elicits abundant, two routes: by driving selective clonal proliferation of infected T chronically activated CTLs [12,13,14], indicating continuous or lymphocytes (‘mitotic spread’) and by de novo infection (‘infectious repeated expression of Tax in vivo. Ex vivo, Tax protein is spread’) via the virological synapse [3]. HTLV-1 replication is spontaneously expressed in a fraction of infected peripheral blood counterbalanced by a strong, chronically activated cytotoxic T mononuclear cells (PBMCs) after overnight culture [15]. HBZ is lymphocyte (CTL) immune response [4]. The HTLV-1 proviral load the only gene expressed from the minus strand of the provirus. (number of proviral copies per 100 PBMCs) varies between infected HBZ also promotes infected cell proliferation [16] and the CTL individuals by over 1000-fold. The proviral load is the strongest response to HBZ protein is a key determinant of proviral load and correlate of HTLV-1 associated diseases, in particular Adult T-cell the risk of the inflammatory disease HAM/TSP [17,18]. Tax Leukemia-Lymphoma (ATLL, [5]) and HTLV-1 Associated Mye- enhances HBZ expression; HBZ protein exerts negative feedback lopathy/Tropical Spastic Paraparesis (HAM/TSP, [6]). on Tax expression [19,20]. PLOS Pathogens | www.plospathogens.org 1 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression Similarly, we observed a bias (up to 2-fold greater than random) Author Summary towards integration in proximity to CpG islands; again, the bias HTLV-1 is a human retrovirus, estimated to infect over 10 reached a peak at 1 kb from the nearest CpG island (supplemen- million individuals worldwide, which causes the inflamma- tary Figure S4). tory disease HTLV-1-associated Myelopathy/Tropical Spas- We showed previously [11] that HTLV-1 provirus preferentially tic Paraparesis and an aggressive malignancy known as integrates in transcriptionally active regions of the host genome. Adult T-cell Leukemia/Lymphoma. The mechanisms that To test the hypothesis that specific transcription factor binding allow the virus to maintain a life-long infection are not fully sites (TFBS) influence HTLV-1 proviral targeting, expression and understood. Here we identified attributes of the host clonal abundance, we used data on genome-wide TFBS ChIP-seq: genome flanking the integrated HTLV-1 provirus associat- + where available, from primary CD4 T cells; otherwise, from T ed with integration targeting and spontaneous expression cells or other human cell types; see Table S3 for complete listing of of the provirus in vitro, and clonal expansion in vivo. the datasets used. Spontaneous expression (after short-term culture) of the In vitro integration sites showed a remarkably strong bias viral protein Tax, which is known to drive proliferation of (compared with random sites) towards integration in proximity to the infected cell, was significantly more frequent among specific TFBS, in particular STAT1, p53, HDACs (e.g. HDAC3, less expanded clones, suggesting that Tax-expressing HDAC6) and HATs (e.g. p300, CBP) (Table S3). In most cases the clones are more efficiently controlled by the immune effect was localized to within 100–1000 bases of the integration site response. Certain transcription start sites immediately (Figure 2A) and declined sharply at greater distances. Two upstream of the viral integration site were associated with patterns were observed in this biased integration. First, the virus latency, which in turn was associated with clonal preference towards integration in proximity to TFBS was typically expansion in vivo. symmetrical (e.g. p300), i.e. equally strong upstream and downstream of the integration site but in some cases was We hypothesize that the genomic integration site of HTLV-1 asymmetrical (e.g. STAT1), with a bias towards one side (often determines the pattern and intensity of expression of the plus and downstream). Second, in many cases we observed a sharp decrease minus proviral strands, which in turn determine the equilibrium in the preferential integration at 10 bases from the TFBS, such as abundance and the pathogenic potential of an infected T cell clone STAT1 Figure 2A). This pattern was consistently observed across in vivo. To test this hypothesis, we used our recently described several in vitro and in vivo datasets (supplementary Figures S1, protocol [11] of high-throughput mapping and quantification of S2). proviral integration sites in fresh primary PBMCs from HTLV-1- Because certain TFBS are frequently co-located in the human infected individuals. genome [22], we wished to test which TFBS were independently associated with targeting of the integration site. First, a likelihood ratio test was used to test whether the TFBS was selectively Results associated with integration either upstream or downstream of the HTLV-1 preferentially integrates within 1 kb of a host integration site, and each TFBS was then tested individually using transcription start site and is strongly biased to specific a univariate model. We then combined all significant factors using a step-down multivariate logistic regression analysis until only transcription factor binding sites independently significant (p,0.05) factors remained. Most factors To identify genomic factors associated with the targeting of that were independently associated with integration site targeting HTLV-1 integration, we infected Jurkat T cells by short co-culture occurred with equal frequency upstream or downstream of the with the HTLV-1-producing cell line MT2. The integration sites integration site (Figure 2B, see also supplementary Table S7). The were then analysed using our high-throughput protocol and factors with the highest odds ratios were the transcription factor compared to a control list of random sites in the human genome. p53 and the histone deacetylase HDAC6. Figure 1A illustrates the possible orientations (same or opposite) of the nearby genomic features, such as transcription start sites, either upstream or downstream of the integrated provirus. Effect of HTLV-1 integration sites on clonal expansion We previously showed [11] that 47% of de novo HTLV-1 We previously reported [11] a significant association between proviral integration events lie within a RefSeq gene. This certain features in the flanking genome and in vivo expansion of frequency is slightly higher than expected by chance, but is much the infected T-cell clone. Here, we found that proviruses lower than that observed for HIV (,70%), which uses the host integrated within a gene were more frequent in larger (more protein LEDGF to target proviral integration to genes [21]. As abundant) clones than in smaller clones in vivo, but only when the expected by chance, ,50% of proviruses integrated within host provirus was integrated in the same transcriptional orientation as genes were in the same transcriptional orientation as the host gene the host gene (Figure 1D); the frequency of integration in the (Figure 1D, in vitro). opposite orientation was not positively correlated with clonal abundance. Gillet et al [11] reported a significantly higher than expected proportion of in vitro integration sites within 10 kb of a RefSeq High clone abundance (Figure 1C, top two bins) was associated gene. We extended this analysis to identify the optimal (most with the presence of a host TSS within 1 kb downstream of the frequent) distance between the integration site and the nearest host provirus; here, the transcriptional orientation of the provirus had transcription start site (TSS). The results (Figure 1B) show a peak less effect on abundance than in the case of proviruses integrated preference (measured by the odds ratio, OR, observed/expected) within a host gene. The excess frequency of TSS downstream (but towards integration in proximity to TSS at ,1 kb of the integrated not upstream) was much higher in integration sites in vivo than in provirus (upstream or downstream); the OR gradually diminished vitro, in particular when the provirus was integrated in the same until it reached 1 (same as random expectation) at ,1 Mb from orientation as the nearby host gene (p(same) ,10 ; p(opposite) the integration site (Figure 1B). There was a small bias (non- ,0.05, x test). The presence of a host CpG island within 1 kb significant for in vitro integration) towards integration with a TSS downstream was also selectively associated with clone high downstream of the integration site (Figure 1C, in vitro). abundance (Figure 1E). PLOS Pathogens | www.plospathogens.org 2 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression Figure 1. Genomic environment at HTLV-1 proviral integration site determines integration in vitro and abundance in vivo. (A) Blue blocks denote a genomic feature such as a transcription start site. The distance to the nearest genomic feature is calculated (unless otherwise stated) separately for features upstream (closer to 59 LTR) and downstream of the provirus. Unless otherwise stated, distance is calculated to the nearest end of the genomic feature. Where the genomic feature has an orientation (i.e. transcription units) its orientation relative to the transcriptional orientation of the provirus is indicated as ‘‘same’’ or ‘‘opposite’’. (B) to (E): proportion of observed integration sites compared to random expectation. (B) Frequency of integration in proximity to transcriptional units (RefSeq). In vitro denotes a combined dataset from two independent experiments (see Table 1). (C) Frequency of integration within 1 kb of a TSS according to clonal abundance (cells in a given clone per 10 000 PBMCs). (D) The excess frequency (compared with random) of observing a provirus within a transcription unit was greater among abundant clones in vivo integrated in the same transcriptional orientation (blue) but not in opposite orientation (orange). (E) The excess frequency (compared with random) of observing a provirus within 1 kb of a host CpG island increased with increasing clonal abundance, in particular where the CpG island lay downstream of the integration site. doi:10.1371/journal.ppat.1003271.g001 PLOS Pathogens | www.plospathogens.org 3 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression Figure 2. Influence of host TFBS on integration site targeting. (A) Bias in integration in proximity to TFBS (based on ChIP-seq experiments), measured by the odds ratio compared to random expectation. Four representative plots are shown; see also supplementary information. The excess frequency of integration in proximity to TFBS was frequently greater in in vitro infection than in clones isolated from PBMCs in vivo, and greater in low abundance clones in vivo than high abundance clones in vivo (see bottom right panel and supplementary information). Arrows indicate a PLOS Pathogens | www.plospathogens.org 4 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression symmetrical (p300) or asymmetrical (STAT1) bias towards integration in proximity to TFBS, as well as a lower bias in close proximity to IS (STAT1). See also supplementary Table S4 for underlying data. (B) TFBS independently associated with integration frequency in vitro were identified by multivariate analysis. OR – odds ratio. TFBS shown above the line were associated with an excess frequency of integration compared with random (OR.1); TFBS below the line were significantly less likely to lie near the provirus (OR,1). Model 1 and Model 2 (carried out independently) test for TFBS within 1 kb and 100 bp of IS, respectively. doi:10.1371/journal.ppat.1003271.g002 Integration sites observed in vivo showed a similar bias towards PBMCs of 10 infected HAM/TSP patients (to preclude CTL- proximity to TFBS, with two important differences. First, the OR mediated lysis), and the CD8 population was incubated in vitro was in each case lower than that observed in in vitro integration. overnight to allow spontaneous expression of the Tax protein [15]. Second, the magnitude of the bias (OR) declined as clonal We then sorted the cells by flow cytometry to isolate Tax and abundance increased (Figure 3; supplementary Figure S3). Tax cells and analysed the integration sites in the two cell fractions. We measured the proportion of each clone that spontaneously Effect of HTLV-1 integration site on Tax expression expressed Tax by quantifying individual integration sites in the We wished to identify features of the genomic integration site + 2 Tax and Tax cells, (Figure 4E, and supplementary Figure S7). that favour expression of the HTLV-1 provirus. We hypothesized The observed proportion of Tax cells per clone varied between that the genomic environment flanking the proviral integration site 0% and 100%. The majority of clones, regardless of clonal determines the rate of spontaneous expression of the HTLV-1 + 2 abundance, were either .90% Tax or .90% Tax . This transactivator protein Tax by a given infected T-cell clone: that is, observation is consistent with the hypothesis that the rate of the proportion of cells in that clone that express Tax within a given spontaneous expression of Tax is an intrinsic property of each time interval. CD8 T-cells were depleted from fresh unstimulated clone and is determined by the proviral integration site. When the provirus was integrated within a host gene, we observed a slight but significant excess frequency of Tax cells 2 23 2 compared with Tax cells (46% vs 43% respectively, p,10 , x test). However, while the proviruses in the Tax cells were found with equal frequency in the same or the opposite transcriptional orientation to the host gene in which they were integrated, the Tax cells were significantly more frequently present in the same + 2 orientation as the host gene (52% of Tax vs 59% of Tax cells, 215 2 2 p,10 , x test). Thus, T cell clones that were 100% Tax were significantly more likely to carry a provirus in the same orientation as the host gene (Figure 4B). The relative position (upstream or downstream of the integra- tion site) and the transcriptional orientation of the nearest host gene influenced not only the clonal abundance (Figure 1) but also spontaneous Tax expression. Where the nearest host gene lay in the same transcriptional orientation as the HTLV-1 provirus, the presence of a host TSS (Figure 4A) or CpG island (Figure 4C) within 1 kb upstream of the provirus was associated with silencing of Tax, whereas a TSS or CpG island within 1 kb downstream was associated with Tax expression. The closer the upstream gene was to the integration site, the lower was the proportion of Tax cells if the gene was in the same orientation (Figure 4D). In contrast, where the nearest host gene was in opposite transcrip- tional orientation, this asymmetrical effect of the nearby host gene was not observed (Figure 4A, right-hand panel; Figure 4D). The mean proportion of Tax cells in one clone (across all clone abundance classes) was 60%. We wished to test whether proximity to TFBS would alter this proportion. We found that the presence of certain TFBS (including STAT1, cJun, NRSF) within 1 kb upstream of the integration site was associated with a higher proportion of Tax cells in the respective T-cell clone (Figure 5A). A notable exception was BRG-1, which showed a strong opposite asymmetric effect: cells containing a BRG-1 site just upstream of the provirus were more likely to be Tax , whereas cells with a BRG-1 site just downstream of the provirus were more likely to be Figure 3. Influence of host TFBS on clonal abundance. Bias in Tax (Figure 5A, top left panel). integration in proximity to TFBS (based on ChIP-seq experiments), To identify the TFBS that were independently and significantly measured by the odds ratio compared to random expectation. Two associated with spontaneous Tax expression, a logistic regression representative plots are shown; see also supplementary Figure S3. The analysis was carried out as described above (Figure 2B) for excess frequency of integration in proximity to TFBS was greater in low integration site targeting. The results (Figure 5B, see also abundance clones in vivo than high abundance clones in vivo. See also supplementary Table S7) confirmed the asymmetric effects of supplementary Table S5 for underlying data. doi:10.1371/journal.ppat.1003271.g003 the BRG-1 binding site, and in addition revealed significant PLOS Pathogens | www.plospathogens.org 5 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression Figure 4. Genomic environment at HTLV-1 proviral integration site associated with proviral expression after 18 h in culture. CD8- + 2 depleted PBMCs were placed in culture overnight and sorted by flow cytometry to isolate Tax and Tax cells, followed by integration site analysis of sorted cells. (A)–(C): proportion of observed integration sites compared to random expectation. (A) Frequency of integration in proximity to + 2 transcriptional units (RefSeq) in clones that were 100% Tax or 100% Tax , according to the relative transcriptional orientation of the provirus and the host gene. The peak of integration at 1 kb mirrors that observed in vivo in unsorted cells (Figure 1B). However, the integration site in Tax clones was more likely than in Tax clones to possess a nearby upstream TSS in the same orientation, and less likely to lie nearby a downstream TSS in the same orientation (or any relative position in the opposite orientation). (B) The provirus in Tax clones (blue) was oriented in the same transcriptional sense as the host gene in which it was integrated more frequently than random. The orientation of Tax clones (pink) did not differ from random. (C) + 2 Frequency of integration in proximity to CpG islands in clones that were 100% Tax or 100% Tax . The peak of integration at 1 kb mirrors that observed in vivo in unsorted cells and in vitro (Figure S4). (D) Mean fraction of Tax cells in clones with a TSS at a given distance (log scale) from the integration site, according to the relative transcriptional orientation of the provirus and the host TSS. The dotted line denotes the mean fraction of + + Tax cells across all clones. (E) Frequency distribution of clones according to the frequency of Tax cells in the respective clones. See supplementary Figure S7 for detailed frequency distribution separated according to clone abundance. doi:10.1371/journal.ppat.1003271.g004 asymmetric associations between Tax expression and several other Tax cells are more frequent in low-abundance clones TFBS, notably STAT1, NRSF, and HDAC1. Thus, a STAT1 To test the hypothesis that the level of Tax expression is binding site 100 bp upstream of the provirus strongly favoured Tax correlated with the in vivo abundance of the infected T cell clone, expression, but the presence of a downstream STAT1 binding site we divided all detected clones into four abundance bins based on was not an independent predictor of Tax expression after the total number of cells observed in each clone. There was a multivariate analysis. Conversely, an NRSF binding site 100 bp significant negative correlation between clone abundance and the downstream was a significant predictor of Tax negativity, but the proportion of Tax cells in the respective abundance bin (Figure 5). closest upstream NRSF binding site was not independently That is, small clones were more likely to be Tax , and this associated with Tax expression. The asymmetry of these associa- likelihood decreased as clone abundance increased. We conclude tions contrasts with the predominantly symmetrical associations that, at least in cells from HAM/TSP patients, the majority of observed between TFBS and integration site targeting (Figure 2B), spontaneous Tax expression observed is due to the large number and suggests a mechanistic interaction between transcription of the of low-abundance clones, rather than a small number of high- provirus and transcription of the flanking host genome. abundance clones. PLOS Pathogens | www.plospathogens.org 6 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression Figure 5. Influence of proximity to TFBS on Tax expression. (A) Mean fraction of Tax cells in clones with a TFBS (based on ChIP-seq experiments) at a given distance from the IS. Four representative plots are shown. (B) TFBS that were independently associated with Tax expression were identified by multivariate analysis, outcome measure . TFBS shown above the line were associated with Tax expression (OR.1); TFBS below the line were associated with Tax silencing (OR,1). Model 1 and Model 2 (carried out independently) test for TFBS within 1 kb and 100 bp of IS, respectively. doi:10.1371/journal.ppat.1003271.g005 PLOS Pathogens | www.plospathogens.org 7 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression thoroughly studied [27]; the most important is the lens epithelium- Discussion derived growth factor (LEDGF/p75, [28]), which determines An understanding of the regulation of proviral latency is integrase localization [29] and targeting of HIV integrase to required for attempts to eradicate latent retroviruses and to transcription units [21]. A study of host factors associated with optimize retroviral vectors for in vitro and in vivo use. In HIV-1 HTLV-1 integrase is currently underway. infection, a reservoir of latently infected cells persists indefinitely in The observed bias towards integration near certain TFBS was the face of antiretroviral drug therapy and precludes eradication of predominantly symmetrical and short-range, reaching a maximum the infection (reviewed in [23]). In HTLV-1 infection, proviral at 100b from the integration site and falling to random expectation expression is difficult to detect in fresh PBMCs: however, the at ,10 kb (Figure 2A). In many instances the bias dropped sharply strong, chronically activated host immune response and the at less than 100b from the integration site: we suggest that this selective oligoclonal proliferation of HTLV-1-infected T cells drop is due to steric hindrance between the pre-integration argue that the virus is continuously or intermittently expressed in complex and the DNA-bound transcription factor. vivo [4,24]. In contrast to the symmetry observed in the association between The abundance of an HTLV-1-infected T cell clone in vivo will genomic features (such as TFBS) and the frequency of initial be determined by the net effect of two main selection forces: its integration, we found significant asymmetric interactions between ability to proliferate and its susceptibility to killing by the strong the flanking host genome and the integrated provirus in CTL response [4]. If these forces acted upon all clones equally, the determining clonal abundance and spontaneous proviral expres- clones would have the same relative abundance in the host. sion. Both the relative position of the nearest host gene (upstream However, Gillet et al [11] showed a wide variation in clone or downstream of the provirus) and its relative transcriptional abundance both within and between infected individuals and over orientation showed significant associations with clone abundance time. We hypothesized that this variation between clones is caused and expression. Previous studies [1,2] reported contradictory by the genomic environment of the integrated provirus, by evidence on the role of an upstream same-sense host promoter in determining the frequency and intensity of expression of proviral either promoting or suppressing proviral transcription. More genes, in particular Tax and HBZ, which in turn promote cell recently, Shan et al [30] have shown in Bcl-2-transduced CD4 T proliferation and thereby confer a selective advantage on the cells, infected in vitro with GFP expressing modified HIV, that infected T cell clone. persistent expression of GFP was associated with opposite sense To identify the host genomic factors that determine integration orientation, while inducible expression was associated with same site targeting, we mapped and quantified proviral integration sites sense orientation. The evidence obtained here demonstrates that, isolated from two independent in vitro infection experiments. We in natural HTLV-1 infection, the presence of a same-sense host assume that the pattern of integration observed in short-term in gene promoter upstream of the integrated provirus is associated vitro infection reflects the initial pattern of integration in vivo, with inhibition of spontaneous proviral expression, suggesting the before the selection exerted during chronic infection. The results operation of transcriptional interference. We conclude that the confirmed our previous observations [10,11] that the virus is transcriptional interaction between host and HTLV-1 operates at targeted to transcriptionally active regions of the genome, within two levels. First, at a regional level – within 10 kb of the provirus – or near to a host gene. There was no bias in the orientation of the transcriptional activity of the flanking host genome favours provirus in the initial infection, indicating that the bias observed in proviral gene expression [10,11], presumably because of accessi- integration sites isolated from PBMCs is a result of the long-term bility of the euchromatin to transcription complexes. Second, at a selection forces acting on the infected clones in vivo. local level – within 100b to 1000b – transcriptional interference by We observed a bias towards integration in proximity to a same-sense host promoter within 1 kb upstream can override the particular transcription factor binding sites. This bias was regional effect and inhibit proviral transcription. remarkably strong in certain cases (STAT1, NRSF) in single- Two observations reported here demonstrate that proviral factor analysis. Because clusters of different TFBS are frequent in integration and expression are not determined simply by the the genome [22], we carried out a multivariate (logistic regression) accessibility of chromatin. First, whereas some TFBS were analysis to identify the TFBS that were independently and associated with HTLV-1 proviral integration more frequently significantly associated with an excess frequency of integration. than expected, the frequency of other TFBS showed no such bias. The results (Figure 2B) confirmed the identification of p53, Second, the asymmetric associations observed between proviral HDAC6 and STAT1 as significant independent correlates of orientation and position with respect to flanking host genes and the integration. Further independent predictors of integration includ- abundance and expression of the HTLV-1 provirus argue for a ed Ini1 (see below), cMyc, cJun and NF-kB (Figure 2B). p53 and mechanistic interaction between transcription of the HTLV-1 STAT1 both play important roles in HTLV-1 infection. HTLV-1 provirus and transcription of the flanking host genome. dysregulates p53 signalling pathways in vivo [25]; it is not known An observation of particular interest is the opposing effect of a whether insertional mutagenesis contributes to this dysregulation. BRG-1 binding site upstream and downstream of the provirus HTLV-1 also causes widespread activation of interferon-stimulat- (Figure 5A). BRG-1, one of the two ATPase components required ed genes in vivo, including the key transcriptional regulator for the activity of the SWI/SNF complex [31], controls gene STAT1 [25]. A strong association was reported between STAT1 expression by remodelling chromatin, i.e. by repositioning and MLV integration [26]; the authors attributed this to an nucleosomes to control the access of transcriptional complexes to association between MLV integration and particular epigenetic the DNA. BRG-1 can cause both gene repression [32] and gene marks (H3K4me3, H3K4me1 and H3K9ac) at the integration site. activation [33]; the balance appears to depend on which other The proportion of all integration sites near any one TFBS was subunits are recruited to the SWI/SNF complex [34]. Easley et al in the minority. This observation indicates that proximity to the [35] found that BRG-1 is required for Tax expression and HTLV- transcription factor binding site itself is not sufficient for 1 replication in vitro, and Rafati et al [36] found that the BAF integration, but suggests that these transcription factors (or an subclass of the SWI/SNF complex repressed HIV-1 transcription associated host factor) increase the efficiency of proviral integra- whereas the PBAF subclass promoted transcription. Our observa- tion. Host factors associated with HIV integrase have been tion (Figure 5A) that a BRG-1 site upstream of the provirus is PLOS Pathogens | www.plospathogens.org 8 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression associated with silencing of Tax, while a BRG-1 site downstream is associated with Tax expression, is consistent with our conclusion (above) that transcriptional interference dominates the transcrip- tional interaction between the provirus and the flanking host genome. The Ini1 subunit of SWI/SNF interacts directly with HIV-1 integrase [37]; a fraction of Ini1 moves transiently to the cytoplasm to associate with the HIV-1 preintegration complex [38]. However, the function of Ini1 in HIV-1 proviral integration and expression in vivo is not understood. We found that genomic sites for Ini1 binding were significantly associated with HTLV-1 proviral integration (Figure 2B). The presence of an Ini1 site 1 kb downstream of the provirus was associated with spontaneous Tax expression, similar to the effect of the downstream BRG-1 site. Finally, the SWI/SNF subunit BAF155 was overrepresented 1 kb upstream of the integration site (Figure 2B), and was associated with Tax silencing when present either upstream or downstream of the provirus (Figure 5B). The HTLV-1 Tax protein acts in concert with host cell Figure 6. Tax cells are more frequent in smaller clones. Mean fraction of Tax cells in bins of increasing clonal abundance (total transcription factors (notably CBP/p300) on the promoter/ number of cells in each respective clone). The fraction of Tax cells was enhancer in the viral 59 LTR, driving plus-strand transcription 216 2 negatively correlated with clonal abundance (p,10 , x test for in a strong positive feedback loop. Tax also acts on response trend). elements for NF-kB, CREB and the serum response factor (SRE) doi:10.1371/journal.ppat.1003271.g006 to upregulate expression of a wide range of host genes [39]. Finally, Tax promotes cell cycle progression by accelerating difference in amino acid sequence found between one clone and passage through G1 and inhibiting the G1/S and G2/M the others (data not shown). checkpoints [40]. The net effect of Tax expression is therefore to Interestingly, while Tax cells were more frequent in low- drive activation and proliferation of the infected T cell. We abundance clones, certain features favouring proviral expression previously reported that spontaneous Tax expression in fresh (e.g. a downstream host TSS) also favoured clonal expansion. The unstimulated PBMCs was associated with proliferation of the association between clonal abundance and proviral integration respective cell in vivo [41]. We therefore expected to observe a within 1 kb of a downstream host TSS was maintained even positive correlation between the frequency of spontaneous Tax within Tax clones, consistent with the idea that the selective expression by a given clone ex vivo and the abundance of that expansion of these clones is driven by other proviral genes. clone in vivo. However, the results obtained here (Figure 6) These observations raise the possibility that the equilibrium demonstrate the opposite, i.e. a highly significant negative abundance of an HTLV-1-infected T cell clone in vivo is correlation. This correlation is likely to be caused by the strong determined not by Tax but by the HBZ gene, encoded on the host immune response to the virus. The Tax protein is highly negative strand of the provirus. Satou et al showed that HBZ immunodominant in the class 1 MHC-restricted cytotoxic T mRNA promoted proliferation of the infected cell, and whereas lymphocyte (CTL) response to HTLV-1 [14,42], and the tax gene Tax expression is frequently undetectable, HBZ appears to be is frequently silenced in vivo by mutation or epigenetic changes persistently expressed in fresh cells isolated from both non- such as DNA methylation in both untransformed and malignantly malignant cases of HTLV-1 infection and cases of adult T-cell transformed (leukemic) cells [43]. Cells that express a high level of leukemia/lymphoma [16]. Further, Macnamara et al recently Tax are killed by CTLs faster than low Tax-expressing cells [44]. showed that the CTL response to HBZ is a critical determinant of Therefore, suppression or loss of Tax expression may confer a the equilibrium proviral load in vivo [17] survival advantage on the infected clone in vivo. We conclude that In this study we examined Tax expression only among CD4 T the small (low-abundance) HTLV-1-infected clones express Tax at cells. A small percentage of infected cells are CD8 T cells [48,49]; a higher rate and turn over faster in vivo than the high-abundance it is possible that the genomic factors that determine targeting and clones. It is possible that the critical role of Tax in the HTLV-1 expression of HTLV-1 differ in CD8 cells. Also, the propensity of lifecycle is not to maintain clone abundance but rather to promote a cell to express Tax was measured by quantifying the frequency of virion production and infection of new cells by cell contact via the spontaneous Tax expression after 18 hours incubation in vitro. virological synapse [45,46]. Two lines of evidence suggest that this measure is relevant to The negative correlation observed between clone abundance HTLV-1 infection and pathogenesis in vivo. First, cells which and the percentage of Tax cells, although it was highly significant express Tax ex vivo turn over faster in vivo [41]. Second, the in all patients combined, was not uniform in every patient. In a proportion of CD4 cells that express Tax after overnight culture small number of patients (in particular those with a high is significantly associated with the HTLV-1 inflammatory disease oligoclonality index, Supplementary Figure S8), the most abun- HAM/TSP [50]. The individuals studied here were all patients dant clones (bin 4, clones with greater than 10 cells) contained a with HAM/TSP: the mean level of spontaneous Tax expression is high proportion of Tax cells, suggesting that certain antigen- lower in asymptomatic HTLV-1 carriers, but it is unlikely that the expressing clones escaped control by the immune response, for molecular mechanisms that govern proviral latency differ quali- example by CTL escape mutations in the tax gene [47]. However, tatively between asymptomatic carriers and patients with HAM/ clone-specific sequencing of exon 3 of the tax gene of 38 clones TSP. (.10 cells) from 8 patients did not reveal significant differences in It will be important to compare the present results with the the occurrence of Tax mutations between clones with a high or genomic factors associated with HBZ expression or silencing. At low frequency of Tax cells; and only in one patient was a PLOS Pathogens | www.plospathogens.org 9 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression present this cannot be done by flow-sorting because existing HBZ- Table 1. IS datasets used. specific antibodies are insufficiently sensitive to detect the low expression levels of HBZ protein in primary cells. We are currently testing the hypothesis that Tax-specific and HBZ-specific CTL total infected clones selectively lyse different clonal populations in vitro. Dataset total IS individuals reference We have identified host genomic factors that determine the in vitro (1) 4521 N/A [11] integration site, the proviral expression and selective clonal in vitro (2) 1805 N/A This publication expansion of HTLV-1 in natural infection in vivo: these factors In vivo (1) 78563 63 [11] are summarized in Figure 7. These results open the way to test the molecular mechanisms involved. In vivo (2) 20202 10 This publication Tax Negative 6700 10 (pooled) This publication Materials and Methods Tax Positive 13054 10 (pooled) This publication Random UIS 176505 N/A This publication Ethics statement Blood samples were donated by HTLV-1-infected individuals 1 For the purpose of this work, only one time point was used for each patient attending the HTLV-1 clinic at the National Centre for Human (most recent available if multiple time points were originally analysed). Retrovirology (Imperial College Healthcare NHS trust) at St doi:10.1371/journal.ppat.1003271.t001 Mary’s Hospital, London UK, with fully informed written consent. This study was approved by the UK National Research Ethics In vitro infection Service (NRES reference 09/H0606/106). In vitro infection was carried out in two independent assays as previously described [11]. Jurkat (JKT) cells were co-cultured for DNA samples (Table 1) 3 h with c-irradiated ( Cs, 40,000 cGy) MT2 cells [51], labelled PBMCs were isolated using Histopaque-1077 (Sigma-Aldrich) with anti-CD4 MicroBeads (Miltenyi). MT2 cells were then and cryopreserved in FBS (Gibco) containing 10% DMSO depleted from the co-culture using magnetic separation (Miltenyi), (Sigma-Aldrich). DNA extraction was carried out using the and infected JKT cells were maintained in culture for 14 days in DNeasy Blood & Tissue kit (Qiagen) according to the manufac- RPMI (supplemented by L-glutamine, penicillin, streptomycin) turer’s protocol. containing 10% FBS for 18 hours at 37C with 5% CO . Genomic Figure 7. Genomic correlates of HTLV-1 proviral targeting, clonal expansion and proviral expression. (A) Factors associated with the presence or absence of spontaneous Tax expression by a given cell after short-term (18 h) in vitro incubation. (B) Features of the genomic environment of the provirus associated either with initial integration (left panel), or clonal expansion in vivo (right panel). Findings were made in the present study unless otherwise stated. TSS – transcription start site. TFBS – transcription factor binding site. doi:10.1371/journal.ppat.1003271.g007 PLOS Pathogens | www.plospathogens.org 10 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression DNA was extracted and the proviral integration sites (IS) analysed Bioinformatic analysis of genomic environment as previously described [11]. IS from MT2 were also analysed to Transcription units and CpG island data were retrieved from exclude possible contamination of the JKT IS. No contaminating the NCBI (ftp.ncbi.nih.gov/gene/) and UCSC tables [56], MT2 IS were found after 14 days. respectively. Annotations to the human genome were obtained from published datasets (Table S3) including ChIP-seq experi- ments on primary CD4 T cells where available; otherwise, data Tax sorting on human CD4 T cell lines or other human cell lines were used. See also supplementary Figure S5, supplementary Table S6. We used the SISSRs algorithm [57] to identify the position of a PBMCs from 10 patients with HAM/TSP with a high proviral putative transcription factor binding site in published ChIP-seq load (range 12.2–50.6 copies per 100 PBMC) were depleted of data where raw ChIP-seq data were available. CD8 cells using magnetic depletion (Miltenyi) and incubated in Annotations positions were compared to the IS using the R RPMI (supplemented by L-glutamine, penicillin, streptomycin) package hiAnnotator (http://malnirav.github.com/hiAnnotator), containing 10% FBS for 18 hours at 37C with 5% CO . After kindly provided by N. Malani and F. Bushman (University of 18 h culture, the cells were stained for intracellular expression of Pennsylvania, USA). Tax (anti-Tax mAb LT4) and sorted using FACS (FACSAria IIIU, BD Biosciences) to isolate two populations of live CD4 cells based on Tax expression. Gates were set (FACSDiva, BD Biosciences) to Statistical analysis + 2 ensure a clear demarcation between the Tax and Tax Statistical analysis was carried out using R version 2.13.0 populations (Figure S6). DNA was extracted from whole unsorted (http://www.R-project.org/). Two separate logistic regression analyses were carried out, respectively, to identify independent PBMCs from each patient and analysed separately to identify the predictors of HTLV-1 integration targeting and independent patient of origin of each clone; 46% of the clones were attributed predictors of Tax positivity. Genomic annotations used to derive in this way. To calculate the fraction of Tax cells in a given clone, + 2 input variable were published ChIP-seq datasets (see Bioinformatic the frequency of Tax and Tax cells were normalized to the mass analysis above; Table S3). For integration targeting, the binary of genomic DNA per cell from each respective cell population, to outcome measure was a ‘‘true’’ integration site (from 4521 correct for experimental variation in efficiency of genomic DNA identified in vitro integration sites) or a ‘‘false’’ integration site isolation (Table S6), (45210 random genomic locations). For spontaneous Tax expression, the binary outcome was Tax positivity (20813 Tax Analysis of IS cells) or Tax negativity (10326 Tax cells). Each TFBS was tested Identification and quantification of proviral integration sites was (presence or absence of the TFBS within a given distance of the done as previously described [11]. HTLV-1 infected DNA was integration site) as an independent predictor in each analysis. For randomly sheared by sonication (Covaris S2) and then blunt- each outcome variable, two separate analyses were carried out, ended (Klenow polymerase) and ligated to a partly double- respectively at two distances of the integration site - 100 bases and stranded DNA linker. Following a nested PCR step, the resulting 1 kb. DNA libraries were deep sequenced using the Illumina GA-II First, for each TFBS and at each distance, we tested whether the platform. DNA sequence was aligned to the human genome relative position (upstream/downstream) of the integration site reference (UCSC hg18, excluding haplotype and ‘‘random’’ and the TFBS determined the outcome by using a likelihood ratio sequences) using the ELAND algorithm. Distinct IS were grouped test to compare two competing models: 1) presence or absence of based on integration site and quantified based on number of TFBS upstream or (separately) downstream; 2) presence or distinct shear sites isolated and the respective patient’s proviral absence of TFBS, regardless of relative position. Next, we carried load. out univariate analysis of each individual TFBS, based on the DNA sequences from ,190000 random sites in the human model chosen by the likelihood ratio test. Only TFBS that were genome (hg18) were generated using Galaxy [52,53,54] and back- significant (p-value,0.05) after correction for multiple compari- aligned to the human genome using the same pipeline to eliminate sons (Benjamini-Hochberg) were used in the multivariate analysis. any potential bias due to alignment limitations. Multivariate analysis was carried out using a step-down logistic regression method. Calculation of clonal abundance The absolute abundance of a given clone was defined as the Supporting Information number of proviral copies of that clone per 10 PBMCs. Given n - th the number of proviral copies for the i clone, and S – the total Figure S1 Influence of host TFBS on integration site number of clones identified in the sample, the absolute abundance targeting – in vitro. Bias in integration in proximity to TFBS was calculated for PBMC samples according to the following (based on ChIP-seq experiments), measured by the odds ratio formula: compared to random expectation. The bias was maintained across separate datasets, generated by independent in vitro experiments. Dotted line denotes random expectation (OR = 1). absolute abundance~PVL| (TIF) Figure S2 Influence of host TFBS on integration site i~1 targeting – in vivo. Bias in integration in proximity to TFBS Clone abundance bins were defined on a logarithmic scale since (based on ChIP-seq experiments), measured by the odds ratio proviral load (used in calculation of abundance) follows a compared to random expectation. The pattern of bias was logarithmic distribution [55]. The number of clones in each clone maintained between different patient clinical groups. Dotted line abundance bin is given in Table S1. For samples sorted for Tax denotes random expectation (OR = 1). ATLL = Adult T-cell protein expression, where proviral load data were not available, leukaemia/lymphoma . HAM/TSP = HTLV-1 associated mye- the clonal abundance bins were set according to proviral copy lopathy/Tropical spastic paraparesis. AC = Asymptomatic carrier. count. (TIF) PLOS Pathogens | www.plospathogens.org 11 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression Figure S3 Influence of host TFBS on clonal abundance combined (P,10 ). However in certain patients (particularly in vivo. Bias in frequency of integration in proximity to TFBS those with a high oligoclonality Index) the most abundant clones (based on ChIP-seq experiments), measured by the odds ratio contained a high proportion of Tax cells. Clone abundance bins compared to random expectation. TFBS that were associated with were defined as in Figure S7. integration targeting showed a stronger bias (higher OR) in the (TIF) clones least expanded in vivo. Clonal abundance is expressed as 4 Table S1 In vivo integration sites – sample data by the number of cells in given clone per 10 PBMCs. Dotted line clone abundance. denotes random expectation (OR = 1). (DOC) (TIF) Table S2 In vivo integration sites – sample data by Figure S4 The genomic environment at the HTLV-1 patient code. proviral integration site determines integration target- (DOC) ing in vitro and clonal abundance in vivo. Frequency of integration in proximity to CpG islands in clones for in vitro (in Table S3 List of annotations datasets used. blue) and in vivo (purple) integration. (DOC) (TIF) Table S4 Odds ratio and clone counts data for a Figure S5 Protocol for flow-sorting of Tax-expressing selection of TFBS – in vitro, in vivo vs. random sites. cells. (A) CD8 cell-depleted PBMCs were studied from 10 (XLS) patients with HAM/TSP with a high proviral load. The cells were Table S5 Odds ratio and clone counts data for a incubated overnight, fixed and stained for Tax and surface CD4 selection of TFBS – clonal abundance bins vs. random expression, and sorted on a high-speed flow cytometer (see Figure sites. S6 for details). (B) Recovered cells from all 10 patients were + + (XLS) combined in two pools, respectively CD4 Tax cells and + 2 CD4 Tax cells. (C) Genomic DNA was extracted from each Table S6 Tax sorting experiment – sample data. pool of cells and integration site analysis carried out as described. (DOC) (TIF) Table S7 Multivariate analysis results – detailed odds Figure S6 Flow cytometry sorting by Tax expression. (A) ratios and confidence intervals. Representative FACS plots of the gating procedure used (from 1 of (XLS) 10 samples studied). Lower middle panel shows gating of + + + 2 CD4 Tax (‘Tax pos’) and CD4 Tax (‘Tax neg’) populations; Acknowledgments these gates were set to distinguish unequivocally between Tax and 2 2 + Tax populations. (B) Purity testing of Tax sorted cells: Tax We thank Nirav Malani and Frederic D. Bushman at the department of cells not detected. C) Purity testing of Tax sorted cells: 0.2% were Microbiology, University of Pennsylvania, Philadelphia, PA, USA for the Tax . list of random integration sites and for developing software packages; and Laurence Game, Nathalie Lambie and Adam Giess of the Genomics (TIF) Laboratory at the MRC Clinical Sciences Centre, Hammersmith Hospital, Figure S7 Majority of HTLV-1-infected clones were London UK and Robert Sampson at the flow cytometry facility at St + + either 100% Tax or 0% Tax . Frequency distribution of Mary’s Campus, Imperial College, London UK. Finally we thank Lucy clones according to the frequency of Tax cells in each respective Cook, Heather Niederer and Becca Asquith and her team at the Section of clone, binned according to number of sister cells detected in Immunology (Imperial College) for discussions and comments, and the staff and donors at the National Centre for Human Retrovirology (Imperial sample: bin 1: 1 cell detected; bin 2: 2 or 3 cells detected; bin 3: 4 College Healthcare NHS trust) at St Mary’s Hospital, London UK. to 10 cells detected; bin 4: over 10 cells detected. (TIF) Author Contributions Figure S8 Tax cells were more frequent in smaller Conceived and designed the experiments: AM GPT CRMB. Performed clones. Mean fraction of Tax cells within each bin, in bins of the experiments: AM NAG. Analyzed the data: AM. Contributed increasing clonal abundance (total number of cells in each reagents/materials/analysis tools: YT GPT NAG DJL. Wrote the paper: respective clone). In the majority of patients there was an inverse AM CRMB. Clinical diagnosis and assessment: GPT. Designed the correlation between clone abundance bin and fraction of Tax statistical analysis: AM DJL CRMB. cells: this correlation was highly significant in all patients References 1. Han Y, Lin YB, An W, Xu J, Yang HC, et al. (2008) Orientation-dependent analysis of 239 HAM/TSP patients including 64 patients followed up for 10 regulation of integrated HIV-1 expression by host gene transcriptional years. J Neurovirol 7: 228–234. readthrough. Cell Host Microbe 4: 134–146. 7. Wattel E, Vartanian JP, Pannetier C, Wain-Hobson S (1995) Clonal expansion 2. Lenasi T, Contreras X, Peterlin BM (2008) Transcriptional interference antagonizes of human T-cell leukemia virus type I-infected cells in asymptomatic and proviral gene expression to promote HIV latency. Cell Host Microbe 4: 123–133. symptomatic carriers without malignancy. J Virol 69: 2863–2868. 3. Igakura T, Stinchcombe JC, Goon PK, Taylor GP, Weber JN, et al. (2003) 8. Cook LB, Rowan AG, Melamed A, Taylor GP, Bangham CR (2012) HTLV-1-infected Spread of HTLV-I between lymphocytes by virus-induced polarization of the T cells contain a single integrated provirus in natural infection. Blood 120: 3488–3490. cytoskeleton. Science 299: 1713–1716. 9. Derse D, Crise B, Li Y, Princler G, Lum N, et al. (2007) Human T-cell leukemia 4. Bangham CR (2009) CTL quality and the control of human retroviral infections. virus type 1 integration target sites in the human genome: comparison with those Eur J Immunol 39: 1700–1712. of other retroviruses. J Virol 81: 6731–6741. 5. Iwanaga M, Watanabe T, Utsunomiya A, Okayama A, Uchimaru K, et al. 10. Meekings KN, Leipzig J, Bushman FD, Taylor GP, Bangham CR (2008) (2010) Human T-cell leukemia virus type I (HTLV-1) proviral load and disease HTLV-1 integration into transcriptionally active genomic regions is associated progression in asymptomatic HTLV-1 carriers: a nationwide prospective study with proviral expression and with HAM/TSP. PLoS Pathog 4: e1000027. in Japan. Blood 116: 1211–1219. 11. Gillet NA, Malani N, Melamed A, Gormley N, Carter R, et al. (2011) The host 6. Matsuzaki T, Nakagawa M, Nagai M, Usuku K, Higuchi I, et al. (2001) HTLV-I genomic environment of the provirus determines the abundance of HTLV-1- proviral load correlates with progression of motor disability in HAM/TSP: infected T-cell clones. Blood 117: 3113–3122. PLOS Pathogens | www.plospathogens.org 12 March 2013 | Volume 9 | Issue 3 | e1003271 Proviral Integration, Abundance and Expression 12. Jacobson S, Shida H, McFarlin DE, Fauci AS, Koenig S (1990) Circulating 35. Easley R, Carpio L, Guendel I, Klase Z, Choi S, et al. (2010) Human T- CD8+ cytotoxic T lymphocytes specific for HTLV-I pX in patients with HTLV- lymphotropic virus type 1 transcription and chromatin-remodeling complexes. I associated neurological disease. Nature 348: 245–248. J Virol 84: 4755–4768. 13. Parker CE, Daenke S, Nightingale S, Bangham CR (1992) Activated, HTLV-1- 36. Rafati H, Parra M, Hakre S, Moshkin Y, Verdin E, et al. (2011) Repressive LTR specific cytotoxic T-lymphocytes are found in healthy seropositives as well as in nucleosome positioning by the BAF complex is required for HIV latency. PLoS patients with tropical spastic paraparesis. Virology 188: 628–636. Biol 9: e1001206. 14. Goon PK, Biancardi A, Fast N, Igakura T, Hanon E, et al. (2004) Human T cell 37. Kalpana GV, Marmon S, Wang W, Crabtree GR, Goff SP (1994) Binding and lymphotropic virus (HTLV) type-1-specific CD8+ T cells: frequency and stimulation of HIV-1 integrase by a human homolog of yeast transcription factor immunodominance hierarchy. J Infect Dis 189: 2294–2298. SNF5. Science 266: 2002–2006. 15. Hanon E, Hall S, Taylor GP, Saito M, Davis R, et al. (2000) Abundant tax 38. Turelli P, Doucas V, Craig E, Mangeat B, Klages N, et al. (2001) Cytoplasmic protein expression in CD4+ T cells infected with human T-cell lymphotropic recruitment of INI1 and PML on incoming HIV preintegration complexes: virus type I (HTLV-I) is prevented by cytotoxic T lymphocytes. Blood 95: 1386– interference with early steps of viral replication. Mol Cell 7: 1245–1254. 1392. 39. Matsuoka M, Jeang KT (2011) Human T-cell leukemia virus type 1 (HTLV-1) 16. Satou Y, Yasunaga J, Yoshida M, Matsuoka M (2006) HTLV-I basic leucine and leukemic transformation: viral infectivity, Tax, HBZ and therapy. zipper factor gene mRNA supports proliferation of adult T cell leukemia cells. Oncogene 30: 1379–1389. Proc Natl Acad Sci U S A 103: 720–725. 40. Marriott SJ, Semmes OJ (2005) Impact of HTLV-I Tax on cell cycle progression 17. Macnamara A, Rowan A, Hilburn S, Kadolsky U, Fujiwara H, et al. (2010) and the cellular DNA damage repair response. Oncogene 24: 5986–5995. HLA class I binding of HBZ determines outcome in HTLV-1 infection. PLoS 41. Asquith B, Zhang Y, Mosley AJ, de Lara CM, Wallace DL, et al. (2007) In vivo Pathog 6: e1001117. T lymphocyte dynamics in humans and the impact of human T-lymphotropic 18. Hilburn S, Rowan A, Demontis MA, MacNamara A, Asquith B, et al. (2011) In virus 1 infection. Proc Natl Acad Sci U S A 104: 8035–8040. vivo expression of human T-lymphotropic virus type 1 basic leucine-zipper 42. Kannagi M, Harada S, Maruyama I, Inoko H, Igarashi H, et al. (1991) protein generates specific CD8+ and CD4+ T-lymphocyte responses that Predominant recognition of human T cell leukemia virus type I (HTLV-I) pX correlate with clinical outcome. J Infect Dis 203: 529–536. gene products by human CD8+ cytotoxic T cells directed against HTLV-I- 19. Landry S, Halin M, Vargas A, Lemasson I, Mesnard JM, et al. (2009) infected cells. Int Immunol 3: 761–767. Upregulation of human T-cell leukemia virus type 1 antisense transcription by 43. Taniguchi Y, Nosaka K, Yasunaga J, Maeda M, Mueller N, et al. (2005) the viral tax protein. J Virol 83: 2048–2054. Silencing of human T-cell leukemia virus type I gene transcription by epigenetic 20. Gaudray G, Gachon F, Basbous J, Biard-Piechaczyk M, Devaux C, et al. (2002) mechanisms. Retrovirology 2: 64. The complementary strand of the human T-cell leukemia virus type 1 RNA 44. Kattan T, MacNamara A, Rowan AG, Nose H, Mosley AJ, et al. (2009) The genome encodes a bZIP transcription factor that down-regulates viral avidity and lytic efficiency of the CTL response to HTLV-1. J Immunol 182: transcription. J Virol 76: 12813–12822. 5723–5729. 21. Ciuffi A, Llano M, Poeschla E, Hoffmann C, Leipzig J, et al. (2005) A role for 45. Nejmeddine M, Barnard AL, Tanaka Y, Taylor GP, Bangham CR (2005) LEDGF/p75 in targeting HIV DNA integration. Nat Med 11: 1287–1289. Human T-lymphotropic virus, type 1, tax protein triggers microtubule 22. Bernstein BE, Birney E, Dunham I, Green ED, Gunter C, et al. (2012) An reorientation in the virological synapse. J Biol Chem 280: 29653–29660. integrated encyclopedia of DNA elements in the human genome. Nature 489: 46. Nejmeddine M, Negi VS, Mukherjee S, Tanaka Y, Orth K, et al. (2009) HTLV- 57–74. 1-Tax and ICAM-1 act on T-cell signal pathways to polarize the microtubule- 23. Durand CM, Blankson JN, Siliciano RF (2012) Developing strategies for HIV-1 organizing center at the virological synapse. Blood 114: 1016–1025. eradication. Trends Immunol 33: 554–562. 47. Niewiesk S, Daenke S, Parker CE, Taylor G, Weber J, et al. (1995) Naturally 24. Bangham CR, Meekings K, Toulza F, Nejmeddine M, Majorovits E, et al. occurring variants of human T-cell leukemia virus type I Tax protein impair its (2009) The immune control of HTLV-1 infection: selection forces and dynamics. recognition by cytotoxic T lymphocytes and the transactivation function of Tax. Front Biosci 14: 2889–2903. J Virol 69: 2649–2653. 25. Tattermusch S, Skinner JA, Chaussabel D, Banchereau J, Berry MP, et al. 48. Richardson JH, Edwards AJ, Cruickshank JK, Rudge P, Dalgleish AG (1990) In (2012) Systems biology approaches reveal a specific interferon-inducible vivo cellular tropism of human T-cell leukemia virus type 1. J Virol 64: 5682– signature in HTLV-1 associated myelopathy. PLoS Pathog 8: e1002480. 5687. 26. Santoni FA, Hartley O, Luban J (2010) Deciphering the code for retroviral 49. Hanon E, Stinchcombe JC, Saito M, Asquith BE, Taylor GP, et al. (2000) integration target site selection. PLoS Comput Biol 6: e1001008. Fratricide among CD8(+) T lymphocytes naturally infected with human T cell 27. Rain JC, Cribier A, Gerard A, Emiliani S, Benarous R (2009) Yeast two-hybrid lymphotropic virus type I. Immunity 13: 657–664. detection of integrase-host factor interactions. Methods 47: 291–297. 50. Asquith B, Mosley AJ, Heaps A, Tanaka Y, Taylor GP, et al. (2005) 28. Cherepanov P, Maertens G, Proost P, Devreese B, Van Beeumen J, et al. (2003) Quantification of the virus-host interaction in human T lymphotropic virus I HIV-1 integrase forms stable tetramers and associates with LEDGF/p75 protein infection. Retrovirology 2: 75. in human cells. J Biol Chem 278: 372–381. 51. Miyoshi I, Kubonishi I, Sumida M, Hiraki S, Tsubota T, et al. (1980) A novel T- 29. Maertens G, Cherepanov P, Pluymers W, Busschots K, De Clercq E, et al. cell line derived from adult T-cell leukemia. Gann 71: 155–156. (2003) LEDGF/p75 is essential for nuclear and chromosomal targeting of HIV-1 52. Giardine B, Riemer C, Hardison RC, Burhans R, Elnitski L, et al. (2005) integrase in human cells. J Biol Chem 278: 33528–33539. Galaxy: a platform for interactive large-scale genome analysis. Genome Res 15: 30. Shan L, Yang HC, Rabi SA, Bravo HC, Shroff NS, et al. (2011) Influence of 1451–1455. host gene transcription level and orientation on HIV-1 latency in a primary-cell 53. Goecks J, Nekrutenko A, Taylor J (2010) Galaxy: a comprehensive approach for model. J Virol 85: 5384–5393. supporting accessible, reproducible, and transparent computational research in 31. Euskirchen G, Auerbach RK, Snyder M (2012) SWI/SNF Chromatin- the life sciences. Genome Biol 11: R86. remodeling Factors: Multiscale Analyses and Diverse Functions. J Biol Chem 54. Blankenberg D, Von Kuster G, Coraor N, Ananda G, Lazarus R, et al. (2010) 287: 30897–30905. Galaxy: a web-based genome analysis tool for experimentalists. Curr Protoc Mol 32. Ho L, Jothi R, Ronan JL, Cui K, Zhao K, et al. (2009) An embryonic stem cell Biol Chapter 19: Unit 19 10 11–21. chromatin remodeling complex, esBAF, is an essential component of the core 55. Nagai M, Usuku K, Matsumoto W, Kodama D, Takenouchi N, et al. (1998) pluripotency transcriptional network. Proc Natl Acad Sci U S A 106: 5187– Analysis of HTLV-I proviral load in 202 HAM/TSP patients and 243 5191. asymptomatic HTLV-I carriers: high proviral load strongly predisposes to 33. De S, Wurster AL, Precht P, Wood WH, 3rd, Becker KG, et al. (2011) Dynamic HAM/TSP. J Neurovirol 4: 586–593. BRG1 recruitment during T helper differentiation and activation reveals distal 56. Karolchik D, Hinrichs AS, Furey TS, Roskin KM, Sugnet CW, et al. (2004) The regulatory elements. Mol Cell Biol 31: 1512–1527. UCSC Table Browser data retrieval tool. Nucleic Acids Res 32: D493–496. 34. Euskirchen GM, Auerbach RK, Davidov E, Gianoulis TA, Zhong G, et al. 57. Jothi R, Cuddapah S, Barski A, Cui K, Zhao K (2008) Genome-wide (2011) Diverse roles and interactions of the SWI/SNF chromatin remodeling identification of in vivo protein-DNA binding sites from ChIP-Seq data. Nucleic complex revealed using global approaches. PLoS Genet 7: e1002008. Acids Res 36: 5221–5231. PLOS Pathogens | www.plospathogens.org 13 March 2013 | Volume 9 | Issue 3 | e1003271

Journal

PLoS PathogensPublic Library of Science (PLoS) Journal

Published: Mar 21, 2013

There are no references for this article.