HIV DNA integration sites and genomic annotation mapped on the human genome. The human chromosomes are shown numbered. Unsequenced regions are shown in gray. The uppermost track on each chromosome indicates the locations of (1) ENCODE regions and (2) loci showing either greater or less HIV integration than predicted by the multiple regression model. ENCODE regions hosting more integration events than predicted (using a false discovery rate of 5%) are shown in red, those showing less are colored blue, and those ENCODE regions that contain sites not significantly different from that predicted by the model are colored black. Genome-wide loci showing greater integration than predicted are shown in red (hot) or orange (warm), indicating regions with 10 (red) or 50 (orange) sites where fewer were expected. Gray (cool) indicates regions with at least 10 integration events, where more were expected. Proceeding downward, integration site densities for the HIV-Mse (green; densities color-coded ranging from zero to 103 sites per 500-kb genomic segments) or HIV-Avr (red; ranging from one to 122 sites per 500-kb window) are shown in the next two (wider) tracks. In the next track down, the blue gradient indicates the number of Refseq transcription start sites, ranging from zero to 66 sites, located in a given 500-kb window. The next track (purple) shows Refseq gene coverage density, which quantifies the fraction of a given 500-kb window (0%–100%) that contains Refseq transcription units. The bottom track (brown) shows the G/C content (33%–61%) averaged over 500-kb intervals.

Favored DNA sequences for HIV integration, analyzed using WebLogo. The point of HIV integration in the target DNA sequence occurs between positions 0 and −1 (for the sequenced HIV DNA ends). For the complementary strand, the point of integration occurs between positions 4 and 5. Y-axis represents the information content at each target base position (perfect conservation would have a score of 2 bits). The height of the letter reflects the degree of base conservation. (A) Analysis 10,000 HIV-Avr sites. (B) Analysis of 10,000 HIV-Mse sites. (C) Forty-one integration sites that host two independent integration events, compared with 10,000 HIV-Avr sites (D). Essentially identical results were obtained for comparison to HIV-Mse sites (data not shown). Note the difference in scale of the Y-axis.

The effect of epigenetic modifications on HIV integration frequency in the ENCODE region. HIV integration sites and their matched random control sites were annotated using the indicated ENCODE tracks and then distributed into 10 bins based on the ENCODE annotation values. The expectation for random distribution is indicated by the horizontal line at one. For each annotation in A–D, both HIV-Avr (dark gray) and HIV-Mse (light gray) data sets achieved P < 10⁻⁶ for comparison to random distribution. ENCODE annotation: (A) H3 acetylation, (B) H3 K4 mono-methylation, (C) H4 acetylation, (D) H3 K27 tri-methylation, and (E) DNA CpG methylation (P < 0.0002).

HIV favors integration in the major grooves of DNA on nucleosomes. Positions of HIV integration sites on nucleosomes were predicted based on the chicken nucleosome prediction model developed by Segal et al. (2006). The percentage of integration events occurring on nucleosomes is shown for each DNA position relative to the symmetric dyad axis (position 0): (A) 19,962 HIV-Avr sites (gray) and 59,763 matched random control sites for Avr (black line); (B) 20,607 HIV-Mse sites (gray) and 61,692 matched random control sites for Mse (black line). (C, D) Fourier transformation of the data from A and B, respectively, showing periodicity for both HIV-Avr and HIV-Mse.

Integration sites in the ENCODE regions, emphasizing regions that diverge from the multiple regression model. (A) An unexpectedly favorable ENCODE region—ENr332. (B) An unexpectedly disfavorable ENCODE region—ENm009 (beta-globin region). (Top panel) The observed (black line) cumulative histogram of HIV integration events is plotted against the prediction (green line) based on the fitted model as described in Supplemental Data S4. The bottom panel shows the corresponding ENCODE region captured from the UCSC genome browser. The top five tracks are custom annotations: (from top to bottom) the ENCODE-wide matched random controls for the two Mse and Avr integration data sets (MRC-E-Mse and MRC-E-Avr), the experimental HIV-Mse and HIV-Avr integration sites, and transcriptional profiling data from Jurkat cells (shades of gray; highly expressed genes are dark). The genome-wide annotations are shown (from top to bottom): G/C percentage, RefSeq genes, CpG islands, and DNase I hypersensitivity sites. Below are selected ENCODE annotations (from top to bottom): ENCODE tracks showing intensities of various histone post-translational modification in lymphoid cells (GM06990) and DNA CpG methylation. The P-values and observed versus expected numbers of sites are shown on the figure panels.