RESUMEN
Latent Kaposi sarcoma-associated herpesvirus (KSHV) genomes rapidly acquire distinct patterns of the activating histone modification H3K4-me3 as well as repressive H3K27-me3 marks, a modification linked to transcriptional silencing by polycomb repressive complexes (PRC). Interestingly, PRCs have recently been reported to restrict viral gene expression in a number of other viral systems, suggesting they may play a broader role in controlling viral chromatin. If so, it is an intriguing possibility that latency establishment may result from viral subversion of polycomb-mediated host responses to exogenous DNA. To investigate such scenarios we sought to establish whether rapid repression by PRC constitutes a general hallmark of herpesvirus latency. For this purpose, we performed a comparative epigenome analysis of KSHV and the related murine gammaherpesvirus 68 (MHV-68). We demonstrate that, while latently replicating MHV-68 genomes readily acquire distinct patterns of activation-associated histone modifications upon de novo infection, they fundamentally differ in their ability to efficiently attract H3K27-me3 marks. Statistical analyses of ChIP-seq data from in vitro infected cells as well as in vivo latency reservoirs furthermore suggest that, whereas KSHV rapidly attracts PRCs in a genome-wide manner, H3K27-me3 acquisition by MHV-68 genomes may require spreading from initial seed sites to which PRC are recruited as the result of an inefficient or stochastic recruitment, and that immune pressure may be needed to select for latency pools harboring PRC-silenced episomes in vivo. Using co-infection experiments and recombinant viruses, we also show that KSHV's ability to rapidly and efficiently acquire H3K27-me3 marks does not depend on the host cell environment or unique properties of the KSHV-encoded LANA protein, but rather requires specific cis-acting sequence features. We show that the non-canonical PRC1.1 component KDM2B, a factor which binds to unmethylated CpG motifs, is efficiently recruited to KSHV genomes, indicating that CpG island characteristics may constitute these features. In accord with the fact that, compared to MHV-68, KSHV genomes exhibit a fundamentally higher density of CpG motifs, we furthermore demonstrate efficient acquisition of H2AK119-ub by KSHV and H3K36-me2 by MHV-68 (but not vice versa), furthermore supporting the notion that KSHV genomes rapidly attract PRC1.1 complexes in a genome-wide fashion. Collectively, our results suggest that rapid PRC silencing is not a universal feature of viral latency, but that some viruses may rather have adopted distinct genomic features to specifically exploit default host pathways that repress epigenetically naive, CpG-rich DNA.