HIV integration site distributions in resting and activated CD4+ T cells infected in culture

Abstract

Objective: The goal of this study was to investigate whether the location of HIV integration differs in resting versus activated T cells, a feature that could contribute to the formation of latent viral reservoirs via effects on integration targeting.

Design: Primary resting or activated CD4 T cells were infected with purified X4-tropic HIV in the presence and absence of nucleoside triphosphates and genomic locations of integrated provirus determined.

Methods: We sequenced and analyzed a total of 2661 HIV integration sites using linker-mediated PCR and 454 sequencing. Integration site data sets were then compared to each other and to computationally generated random distributions.

Results: HIV integration was favored in active transcription units in both cell types, but integration sites from activated cells were found more often in genomic regions that were dense in genes, dense in CpG islands, and enriched in G/C bases. Integration sites from activated cells were also more strongly correlated with histone methylation patterns associated with active genes.

Conclusion: These data indicate that integration site distributions show modest but significant differences between resting and activated CD4 T cells, and that integration in resting cells occurs more often in regions that may be suboptimal for proviral gene expression.

Fig. 1. Diagram of the experiment and purification of CD4⁺ T cells

(a) Experimental design to investigate HIV integration distributions in resting and activated CD4⁺ T cells. (b) Resting CD4⁺ T cells were purified by negative selection using antibodies against CD25, CD69, and human leukocyte antigen (HLA-DR) to deplete cells expressing these activation markers. The purity of the resting CD4⁺ T cells was monitored by flow cytometry before and after purification. A fluorescence-minus-one control (FMO; cells labeled for CD4 but not for activation markers) was used as negative control to set the gate that measures activation in purified resting CD4⁺ T cells. The quadrants were set such that 0.5% of cells were present in the upper right quadrant when the cells were not stained with activation markers. (The activation markers used for staining were CD25, CD69, and HLA-DR). (c) Kinetics of viral infection measured using quantitative Alu-PCR. ‘+ dN’ indicates addition of deoxynucleosides to the cell culture. Arrows indicate the time point at which cells were harvested for genomic DNA extraction. PBMC, peripheral blood mononuclear cell.

Fig. 2. Primary human DNA sequences at sites of integration

(a) Examples of sequences recovered from resting and activated cells. HIV long terminal repeat (LTR) sequence is shown in green with 25 bp of human genomic sequence in black. The viral-host DNA junction is marked by an arrow. (b) Information content at aligned integration target sites from activated and resting integration site data sets. The Y-axis indicates base conservation, with perfect conservation equaling two bits and no conservation equal to zero. The X-axis shows nucleotide positions relative to the integration site (10 bases on each side with the point of integration between positions –1 and 0).

Fig. 3. Construction of heat maps using the receiver operating characteristic area method

(a) Conventional histogram of data from the activated cell data set (without added deoxynucleosides) showing favored integration of HIV in gene-dense regions. All integration sites and matched random controls were annotated for gene density in the 1 Mb region surrounding the integration site. The pool of sites was then separated into 10 equal bins by relative gene density, with lower gene density to the left and higher to the right. The proportions of the experimental sites and matched random controls were then plotted for each bin, allowing visualization of the higher proportions in the experimental set at higher gene densities compared with the control set. (b) Annotating and ranking integration sites and matched random controls for construction of receiver operating characteristic (ROC) areas. Genes are represented by short vertical lines, integration sites by arrows. The gene count is shown beside each genomic interval; the rank over the pooled data set is shown colored blue for experimental integration sites or green for matched random controls. (c) Use of rank information to generate the ROC areas. For each set of one experimental site and three matched random controls, the experimental site is ranked relative to the matched random control. Positive correlation is indicated by values greater than 0.5, negative correlation as values less than 0.5. (d) Color-coding of heatmap tiles. Increasingly intense shades of red = enriched compared to random; increasing shades of blue depleted compared to random. The heatmap tile in (d) represents the same data shown in histogram form in (a). The P value, determined by a logistic regression method that respects the pairing in the data (clogit), is overlaid on the heatmap tile. (*P < 0.05; **P < 0.01; ***P < 0.001).

Fig. 4. Genomic heat map of integration frequency relative to genomic features

Integration site data set names are shown above the columns. Genomic features analyzed are shown to the left of the corresponding row of heat map. A colored receiver operating characteristic (ROC) area scale is shown along the bottom of the panel with increasing shades of blue indicating negative correlation relative to the genomic feature and increasing shades of red indicating positive correlation relative to the comparison set. P values showing significance of the departure from the comparison set are shown with asterisks (*P < 0.05; **P < 0.01; ***P < 0.001). (a) Comparisons of each experimental data set to the matched random controls relative to frequency of the indicated genomic feature. Asterisks summarize the statistical significance of departures from random. (b) Heat map similar to that in (a), but here the statistical test compares the activated cell data set to each of the other experimental sets. The naming of genomic features is described in the text. A version of these heat maps with user-configurable statistical tests can be found in the interactive supplementary information. (c) Heat map of integration frequency relative to epigenetic marks and chromatin-bound proteins. Associations of integration with histone methylation and chromatin-bound proteins were quantified using ROC areas [22], comparing the association of integration site data sets with the frequency in corresponding matched random controls. A colored ROC area scale is shown along the bottom of the panel with increasing shades of yellow indicating negative correlation of the experimental data set relative to matched random control and increasing shades of blue indicating positive correlation relative to the matched random control. Data sets and window sizes analyzed are shown above each column. CCCTC-binding factor (CTCF) is a DNA-binding protein proposed to be associated with chromatin boundaries. P values showing significance of the departure from the comparison set are shown with asterisks (*P < 0.05; **P < 0.01; ***P < 0.001) Pair-wise comparisons of all sets to matched random controls relative to correlations with histone modifications or chromatin-binding proteins. Colored tiles and asterisks represent significance of departure from random. (d) Pair-wise comparisons of the activated set to each of the other T cell data sets. Asterisks summarize the significance of the departure of each experimental data set from the activated set. Colored tiles represent comparisons to matched random controls as in (c).