A universally observed feature of meiotic recombination is a highly non-uniform distribution of meiotic crossovers across the genome. Although it has become evident that hotspots of recombination are relatively unstable in evolution, the magnitude of the variation of recombination profiles on a genome-wide scale is unknown. To study the underlying mechanisms responsible for the birth, life and death of recombination hotspots we decided to estimate the genome-wide variation in profiles of recombination. This has become possible due to the recent confluence of two major developments: improvements in the computational approaches to calculate genome-wide recombination profiles from linkage disequilibrium data and the availability of a very dense catalog of genetic markers (single nucleotide polymorphisms, SNPs) in human beings (phase I of the HapMap, 106 SNPs). We have shown that although for a given population we find only about 60% of present-day recombination (crossovers) at recombination hotspots (roughly equivalent to the edges of haplotype blocks) in the historical computed map, nearly all crossovers in present-day individuals are predicted by including computed hotspots from several populations (Khil and Camerini-Otero (2010) PLoS Genetics 6, 1). We have hypothesized that this reflects the existence of a universal human recombinome. In the last few years we have been able to generate a genome-wide map for hotspots for double-strand breaks in mouse meiosis by usingSolexa/Illumina ChIP-Seq for Dmc1 and Rad51 foci, both of which mark these sites (all in collaboration with the laboratory of Galina Petukhova at the Department of Biochemistry and Molecular Biology at the Uniformed Services University of Health Sciences). Depending on the level of statistical significance as many as 40,000 of these hotspots can be enumerated and a large majority of both Rad51 and Dmc1 sites are identical. About 10,000 of these hotspots have a high degree of statistical significance. Compared to the LD recombination that has an accuracy of about 5 KB, our physical recombination map, has an accuracy of about 200 bp. This is the first, and to date only, high-resolution genome-wide map of recombination hotspots in a multicellular organism. Using such a map has allowed to identify novel structural features for recombination hotspots. For example, we determined that recombination hotspots share a centrally distributed consensus motif (in the vast majority of hotspots), possess a nucleotide skew that changes polarity at the center of the hotspots, and have both a calculated and experimental preference to be occupied by a nucleosome. Finally, we find that the vast majority of recombination hotspots in mice are associated with testis-specific H3K4 trimethylation that do not overlap transcription start sites even though these sites are well-known to be marked by H3K4 trimethylation. Thus, H3K4 trimethylation per se is not a sufficient mark for directing the meiotic double-strand break machinery. Recently, we developed a novel method that is a variant of chromatin immunoprecipitation followed by sequencing (ChIP-seq)single-stranded DNA sequencing (SSDS)- that specifically detects protein-bound single-stranded DNA. SSDS consists of a new sequencing library preparation procedure for the enrichment of fragments originating from ssDNA that creates a signature sequence that is computationally identified after high-throughput sequencing. We have used this novel method to show that the product of the highly polymorphic and rapidly evolving gene Prdm9 not only determines the positions ofpractically all hotspots but also actively sequesters recombination away from functional genomic elements, such as promoters and enhancers, in mice. Previously we (see above) and others (Myers et al. Science (2005)) have identified meiotic recombination hotspots in the human genome using computational analysis of patterns of linkage disequilibrium (LD) in populations; however, this method only permits the evaluation of sex-averaged and population-averaged recombination rates. Furthermore, the resolution of this method is limited to approximately 2 Kb by the availability of informative single-nucleotide polymorphisms and the hotspots identified may not reflect present-day crossovers. In this work, we exploit a sequencing based method recently developed by us (Khil et al. Genome Res (2012), Brick et al. Nature (2012), Smagulova et al. Nature (2011)) to obtain the first direct high-resolution genome-wide map of meiotic recombination initiation hotspots in human males. The meiosis specific methyltransferase PRMD9 has been shown to define the location of the vast majority of meiotic DSB hotspots (Brick et al. Nature (2012)). We mapped DSBs in several individuals: homozygous for the most common Prdm9 allele (A), heterozygous for the A allele and a closely related variant, the B allele, and heterozygous for the A allele and the C allele (a variant commonly found in African populations). We found that the A and B alleles of Prdm9 defined similar recombination initiation hotspots while we confirmed that the C allele defines a distinct set of hotspots. Approximately 60% of population LD hotspots are explained by A-allele hotspots, while C-allele hotspots explain an additional 10%. This demonstrates that relatively minor alleles significantly contribute to the LD map. We also found that the DSB distribution exhibits a strong telomeric bias which closely resembles that of male, but not female crossovers. This indicates that the recombination landscape is largely shaped at the level of initiation. Examination of two A/A individuals revealed inter-individual variation at about 5% of hotspots. Whole genome sequencing of these individuals determined that at least 50% of A/A polymorphic hotspots could be explained by a single nucleotide variant at a putative PRDM9 binding site. Interestingly, we also found that polymorphic hotspots were frequently found in clusters. Also, we used the H3K4me3 ChIP-Seq signal strength as a proxy for Prdm9 binding affinity and found that unlike in mouse, the hotspot strength is not well correlated with the strength of the H3K4me3 signal. In aggregate, these data indicate that Prdm9 binding is not the only factor modulating hotspot strength in humans. Also, we detect distinct signatures of GC-biased gene conversion and of recombination-coupled mutagenesis at DSB hotspots. Finally, we explored the role of DSB hotspots in genomic rearrangements. We found that DSB hotspots were enriched at structural variants that arise via homology-mediated mechanisms and that meiotic DSBs occur at well known disease-associated chromosomal breakpoints. We recently showed that, unlike what is seen in mouse males, meiotic DSB hotspots in human males are more frequent and stronger at the subtelomeric regions regardless of the PRDM9 alleles examined (Pratto et al., 2014). To determine whether this bias is determined by DNA sequence or some other species-specific factors we exploited a model system where the human PRDM9 and a human chromosome 21 were placed into the mouse. To do that we generated a cross between the TC1 mouse strain that carries a human chromosome 21 and a knock-in mouse strain where the mouse PRDM9 wasreplaced by its human B allele counterpart (produced in a collaboration with Peter Donnelly's group at Oxford). In this scenario both mouse and human chromosomes. We found that DSBs defined by the human allele of PRDM9 were not enriched near the telomeres of either human or mouse chromosomes indicating that a mechanism other than DNA sequence is driving the enhanced DSB formation at the subtelomeric region in human males. This model system uncovered the mechanism of another interesting phenomenon, pseudo-dominance of PRDM9 alleles (Brick et al., 2012).
Showing the most recent 10 out of 15 publications
