(A) DNA breaking and joining steps during integration. Viral and target DNA strands are represented by thick black and parallel lines, respectively, and the viral long terminal repeats (LTRs) are depicted as grey boxes. Nucleotides at the top and bottom strands are denoted by uppercase and lowercase letters, respectively. During 3′-end processing, IN removes two nucleotides from the 3′ end of each strand of linear viral DNA so that the viral 3′ ends terminate with a conserved CA dinucleotide. Closed arrowheads denote the positions of strand transfer, a concerted cleavage-ligation reaction during which IN makes a staggered break in the target DNA. Host DNA repair enzymes fill in the resulting single-stranded gaps, denoted by D1 to D4 in the upper strand and d1 to d4 in the lower strand of target DNA, and remove the two unpaired nucleotides at the 5′ ends of the viral DNA (open arrowheads), thereby generating the short direct repeats flanking the provirus. (B) A potential pathway for generating a base transversion in the short direct repeat during XMRV integration. A coordinated integration of the two viral ends occurred at the 4-bp staggered positions as depicted by the closed arrowheads. During repair of the single-stranded gap adjacent to the upstream LTR, an adenine nucleotide was introduced at the D4 position either by misincorporation or aberrant processing of the unpaired AA-dinucleotide at the viral 5′ end. Subsequent repair of the mismatch resulted in the observed transversion (denoted by bold types).

Base compositions of the 4-bp target site duplication (positions D1 to D4; demarcated by the thick vertical lines) and 10 bp upstream (positions −1 to −10) and downstream (positions +1 to +10) of the direct repeat were calculated. The datasets include the 13 integration sites with correct 4-bp direct repeat (Table 1), 472 integration sites from acutely infected DU145 cells (GenBank accession numbers EU981292 to EU981799) and 14 integration sites from human prostate cancer tissues (GenBank accession numbers EU981800 to EU981813) [14]. Integration occurs between positions −1 and D1 on the top strand, and between positions D4 and +1 on the bottom strand (blue arrows). Any base in a position that is significantly overrepresented than the random dataset (P<0.0001) is highlighted in green, while any base in a position that is significantly underrepresented than the random dataset (P<0.0001) is highlighted in red.