Integration targeting by avian sarcoma-leukosis virus and human immunodeficiency virus in the chicken genome

Abstract

We have analyzed the placement of sites of integration of avian sarcoma-leukosis virus (ASLV) and human immunodeficiency virus (HIV) DNA in the draft chicken genome sequence, with the goals of assessing species-specific effects on integration and allowing comparison to the distribution of chicken endogenous retroviruses (ERVs). We infected chicken embryo fibroblasts (CEF) with ASLV or HIV and sequenced 863 junctions between host and viral DNA. The relationship with cellular gene activity was analyzed by transcriptional profiling of uninfected or ASLV-infected CEF cells. ASLV weakly favored integration in active transcription units (TUs), and HIV strongly favored active TUs, trends seen previously for integration in human cells. The ERVs, in contrast, accumulated mostly outside TUs, including ERVs related to ASLV. The minority of ERVs present within TUs were mainly in the antisense orientation; consequently, the viral splicing and polyadenylation signals would not disrupt cellular mRNA synthesis. In contrast, de novo ASLV integration sites within TUs showed no orientation bias. Comparing the distribution of de novo ASLV integration sites to ERVs indicated that purifying selection against gene disruption, and not initial integration targeting, probably determined the ERV distribution. Further analysis indicated that ERVs in humans, mice, and rats showed similar distributions, suggesting purifying selection dictated their distributions as well.

FIG. 1.

De novo integration targeting in the chicken genome and its relationship to ERVs. The chicken chromosomes are shown numbered (macrochromosomes 1 to 5, intermediate chromosomes 6 to 10, microchromosomes 11 to 32, and sex chromosomes W and Z). Note that, due to the incomplete status of the draft chicken genome sequence, some of the microchromosomes are not represented. Each de novo integration site is shown as a “lollipop” (ASLV, red; HIV, blue). Endogenous ERVL sequences are shown by the purple shading in the upper half of each chromosome, and ERVK is shown below by green. The most intense purple shading represents 85 ERVL integrations per 250 kbp; the most intense green shading represents 12 ERVK integrations per 250 kbp. Centromere locations are denoted by chromosomal indentations. The centromere positions for chromosomes 6, 9, 13 to 16, 18 to 22, 24, 27, and 32 are currently unavailable and have been arbitrarily placed at the chromosomal midpoints. The software used to draw the ideogram was obtained and adapted from http://www.uni-essen.de/∼bt0756/cc/.

FIG. 2.

Frequency of retroviral integration in and around TUs and CpG islands: influence of cell type and retrovirus studied. The percentages of total integrations into TUs (using the “mRNA” gene catalog) and regions of DNA 5 kb upstream of the transcription start site and 5 kb downstream of the transcription end sites were plotted separately for each virus. The viruses and cell type studied are as marked above each graph. (A) ASLV in CEF cells; (B) HIV in CEF cells; (C) ASLV in human 293T-Tva cells; (D) HIV data from several human cell types (see Table 1); (E) MLV in human HeLa cells. The diagram below each graph shows the regions in and around TUs that were scored for integration events. The arrow represents the transcription start site, and the black box represents the TU. (F) The percentage of total integrations for each virus within 2.5 kb upstream and 2.5 kb downstream of CpG island midpoints compared to matched random controls. Comparison of the data on matched random control sites for human and chicken shows that a much larger fraction of the chicken genome is annotated as CpG island, perhaps an artifact of the higher G/C content of the chicken genome. If CpG islands are in fact “overcalled” in the chicken genome, then this will reduce our ability to detect biases in integration in these sites due to increased noise. P values were determined using the chi-square test and comparison to matched random controls. *, 0.01 < P < 0.05; **, 0.001 < P < 0.01; ***, P < 0.001.

FIG. 3.

Relationship between gene activity and integration frequency. (A) ASLV and HIV show a bias towards integration into active TUs. The relative expression levels of genes in CEF cells were assayed on six Affymetrix chicken genome arrays, and the relative expression levels were averaged over the six arrays. The median expression signals for ASLV and HIV were plotted and compared to all the genes queried on the chip (P value on figure). In addition, comparison of genes targeted for integration to those in the matched random control also showed significance (P = 0.021 for ASLV and P = 0.0066 for HIV; Mann-Whitney test). In another analytical approach, the signals for genes hosting integration events were compared to the signals for genes not hosting integration events, and this similarly showed a significant difference (data not shown). (B and C) Analysis of integration frequency as a function of gene expression intensity for ASLV (B) and HIV (C). All genes assayed on the Affymetrix microarrays were divided into eight “bins” according to their relative levels of expression (x axis). The leftmost bin contains genes with the lowest expression levels, and the rightmost bin contains the highest. Genes hosting integration events were distributed into the corresponding expression bins and summed, and then the data wereplotted as the percentages of all integrations in that bin (y axis). P values were determined using the chi-square test for trend by comparison to the null hypothesis of no bias due to expression level.

FIG. 4.

Relative integration frequency in macrochromosomes, intermediate chromosomes, and microchromosomes. Chicken chromosomes were grouped into gene-sparse macrochromosomes (chromosomes 1 to 5; (A), intermediate chromosomes (chromosomes 6 to 10; (B), and gene-dense microchromosomes (chromosomes 11 to 32; (C). Integrations into each chromosomal group were summed and expressed as the percentage of total integrations for each virus. The distribution of ASLV (black bars) integrations shows an association with macrochromosomes, whereas HIV (dark grey bars) integration shows an association with microchromosomes. P values were determined by chi-square test (comparison to the matched random control [light grey bars]).