Table 1.
Datasets used in this study
Figure 1.
Integration of HTLV-1 in transcriptionally active genomic regions.
Panel A: Near gene transcriptional start sites and CpG islands. HTLV-1 showed significantly greater than expected frequency of integration in the vicinity of RefSeq transcriptional start sites (TxStart sites) and CpG islands in cell culture in vitro and in persistent infection in vivo. The vertical axes for the in vitro and in vivo datasets show the proportion of observed HTLV-1 integration sites or expected MRC sites in the vicinity of the respective genomic feature. * indicates a significant difference between HTLV-1 sites and MRCs, by χ-squared analysis (p<0.05). There was a significantly greater association with integration in the proximity of TxStart sites and CpG islands comparing the distribution of observed HTLV-1 integration sites to expected MRC sites in vivo and in vitro (vertical axis, in vitro vs in vivo). $ indicates a significant difference between the frequency of HTLV-1 in vivo sites and in vitro sites, by logistic regression (p<0.005). Panel B: In regions of high gene density. HTLV-1 showed an increased frequency of integration in gene dense regions. Gene density in regions from 25 kb to 1 Mb around the integration site was analysed. In all region sizes, there was a greater association of HTLV-1 integration in gene dense regions both in vivo and in vitro. However, there was a significantly greater association between proviral integration frequency and gene density in persistent infection in vivo than was seen in vitro. A graphical illustration of the 25 kb region and data on the logistic regression results comparing the in vivo and in vitro HTLV-1 datasets for the first three region sizes is given (Panel B).
Table 2.
Percentage of integration sites residing within gene coding regions/CpG islands in vitro and in vivo
Figure 2.
Chromosomal distribution of HTLV-1 integration sites in vitro and in vivo.
The distribution of HTLV-1 integration sites within the genome in vivo during persistent infection was non-random across all chromosomes compared to matched random controls (logistic regression analysis, p = 0.00095). In particular, chromosome 13 had a significantly increased frequency of integration sites compared to MRC sites (χ-squared test; p = 0.014 after correction for multiple comparisons). The vertical axis indicates the proportion of integration sites lying in each chromosome. Integration of HTLV-1 in vitro (combined data from the co-culture sites obtained in this report and sites generated in a previous report [22]) showed a weak overall chromosome bias (logistic regression analysis, p = 0.016) but no particular chromosome was favoured. Logistic regression analysis confirmed that the chromosomal distribution differed significantly in vivo compared to in vitro. * indicates a significant difference between HTLV-1 dataset and MRCs (logistic regression analysis, p = 0.014).
Figure 3.
The genomic distribution of HTLV-1 integration sites was associated with disease status.
Integration sites isolated from the PBMCs of individuals with HAM/TSP were compared to integration sites from ACs using logistic regression analysis. The integration sites from the HAM/TSP group were associated with RefSeq genes (logistic regression model, p = 0.049). The numbers of integration sites compared in each group are indicated.
Figure 4.
The genomic distribution of HTLV-1 integration sites within individuals was associated with provirus expression.
CD8+ cell-depleted PBMCs from an HTLV-1-infected individual were separated into Tax+ and Tax− fractions after 18 hrs incubation in vitro. Integration sites were cloned from each fraction. The integration sites from three independent experiments were combined and the distribution of sites between the Tax+ and Tax− fractions compared using logistic regression analysis. The Tax+ fraction had significantly more integrations in RefSeq genes than did the Tax− fraction (Panel A, p = 0.04). The Tax+ fraction also had a higher proportion of integrations lying within 10 kb of a CpG island (Panel B, p = 0.034) and a higher gene density around the integration site (Panel C, p = 0.019 using 1 Mb around each integration site). The numbers of integration sites compared in each group are indicated.