Background: Retroviral integration favors weakly conserved palindrome sequences at the sites of viral DNA joining
and generates a short (4–6 bp) duplication of host DNA flanking the provirus. We previously determined two key
parameters that underlie the target DNA preference for prototype foamy virus (PFV) and human immunodeficiency
virus type 1 (HIV-1) integration: flexible pyrimidine (Y)/purine (R) dinucleotide steps at the centers of the integration
sites, and base contacts with specific integrase residues, such as Ala188 in PFV integrase and Ser119 in HIV-1 integrase.
Here we examined the dinucleotide preference profiles of a range of retroviruses and correlated these findings with
respect to length of target site duplication (TSD).
Results: Integration datasets covering six viral genera and the three lengths of TSD were accessed from the literature
or generated in this work. All viruses exhibited significant enrichments of flexible YR and/or selection against rigid RY
dinucleotide steps at the centers of integration sites, and the magnitude of this enrichment inversely correlated with
TSD length. The DNA sequence environments of in vivo-generated HIV-1 and PFV sites were consistent with integration
into nucleosomes, however, the local sequence preferences were largely independent of target DNA chromatinization.
Integration sites derived from cells infected with the gammaretrovirus reticuloendotheliosis virus strain A (Rev-A),
which yields a 5 bp TSD, revealed the targeting of global chromatin features most similar to those of Moloney
murine leukemia virus, which yields a 4 bp duplication. In vitro assays revealed that Rev-A integrase interacts with
and is catalytically stimulated by cellular bromodomain containing 4 protein.
Conclusions: Retroviral integrases have likely evolved to bend target DNA to fit scissile phosphodiester bonds
into two active sites for integration, and viruses that cut target DNA with a 6 bp stagger may not need to bend
DNA as sharply as viruses that cleave with 4 bp or 5 bp staggers. For PFV and HIV-1, the selection of signature
bases and central flexibility at sites of integration is largely independent of chromatin structure. Furthermore,
global Rev-A integration is likely directed to chromatin features by bromodomain and extraterminal domain proteins.