Porcine endogenous retrovirus integration sites in the human genome: features in common with those of murine leukemia virus

Abstract

Porcine endogenous retroviruses (PERV) are a major concern when porcine tissues and organs are used for xenotransplantation. PERV has been shown to infect human cells in vitro, highlighting a potential zoonotic risk. No pathology is associated with PERV in its natural host, but the pathogenic potential might differ in the case of cross-species transmission and can only be inferred from knowledge of related gammaretroviruses. We therefore investigated the integration features of the PERV DNA in the human genome in vitro in order to further characterize the risk associated with PERV transmission. In this study, we characterized 189 PERV integration site sequences from human HEK-293 cells. Data showed that PERV integration was strongly enhanced at transcriptional start sites and CpG islands and that the frequencies of integration events increased with the expression levels of the genes, except for the genes with the highest levels of expression, which were disfavored for integration. Finally, we extracted genomic sequences directly flanking the integration sites and found an original 8-base statistical palindromic consensus sequence [TG(int)GTACCAGC]. All these results show similarities between PERV and murine leukemia virus integration site selection, suggesting that gammaretroviruses have a common pattern of integration and that the mechanisms of target site selection within a retrovirus genus might be similar.

FIG. 1.

Map of PERV integration sites on human chromosomes.

FIG. 2.

Distribution of PERV integration sites. (A) Within a 60-kb window centered on transcription start sites. The percentage of PERV integration events was calculated for each 5-kb gap and compared with those of MLV and HIV integration events. (B) With respect to target transcription unit length and flanking region. Transcription units were divided into eight equal parts, and the percentage of integration events in each part was determined. As with MLV (Wu et al.) (53-55) and avian sarcoma virus, fixed 5-kb windows were used for the flanking regions. ^a, data set from Wu et al. (54); ^b, data set from Schroder et al. (43)

FIG. 3.

Effects of gene activity on integration frequency and localization. Expression levels were assayed using Affymetrix HU133A microarray or Affymetrix HU95A microarray data sets. All genes on the chips were divided into eight bins according to their relative levels of expression. (A) The genes targeted by PERV, MLV, and 10,000 in silico random target genes were assigned to the corresponding expression bins and summed, and then the data were plotted as the percentages of all the integration sites in genes on the array. P values were determined using the chi-square test. (B) Relationship between distribution of the PERV and MLV integration sites within genes and gene activity. In each bin, the integration events were separated into two categories according to their localization within the gene: near transcriptional start sites (±5 kb) or in TUs without the start site (+5 kb).

FIG. 4.

Base frequency distribution at PERV integration sites. Integration occurs between positions −1 and 0 on the top strand. (A) The base composition of the top DNA strand at each position flanking the integration site was calculated. Bold positions indicate base frequencies that were statistically higher than that of the global position (left column). P values in italics were obtained by χ² analysis comparing observed base frequencies with the expected frequencies. (B) Comparison of the observed integration preferences with the expected preferences. The x axis shows the offset for each base from the integration site. The vertical line, between offsets 1 and 2, indicates the expected axis of symmetry based on the characteristic 4-base spacing between the sites of PERV DNA integration. The y axis represents the percentage of expected frequency observed for each base (27.3% for A, 22.8% for C, 22.3% for G, and 27.5% for T). The horizontal line is drawn at 100% of the expected frequency.