Fig 1.
Diversity of the fine-scale recombination landscapes in eukaryotes.
Ancestral eukaryotes probably had evolutionarily stable hotspots associated with regions of open chromatin/promoter-like features. In animals, the gene PRDM9 emerged which redirected hotspots outside of promoter-like features. The mode of hotspot specification by PRDM9 induces the erosion of the hotspots. Associated with a high mutation rate of PRDM9’s zinc finger domain, this erosion leads to a rapid turnover of recombination hotspots. Finally, in some animals in which PRDM9 has been lost, promoter-associated hotspots have been lost and recombination does not vary much at a fine-scale. Silhouette images are taken from www.phylopic.org.
Fig 2.
Relative hotspot density as a function of meiotic expression in mice.
The density of PRDM9 hotspots in mice was measured with the maps of PRDM9 hotspots constructed by Chip-seq experiment on male mice in [59]. The SPO11 hotspot density was obtained from SPO11 oligo-seq performed by [139]. The ZCWPW1 hotspots was obtained from the CUT&RUN experiments of [140]. DMC1 hotspots in B6 mice were retrieved from the study of [85]. The CO and NCO events come from the study of [61] and [62]. They were obtained by sperm sequencing of B6/CAST hybrid mice. We considered six categories of transcription activity: five quantiles of gene expression in leptotene retrieved from the study of [141], and intergenic regions. Hotspot density was computed as the number of peaks/events that overlapped each category, divided by the sum of the length of the transcripts of this category. This density was normalized by the total number of peaks/events called in the whole genome. For all the overlaps, we excluded blacklisted regions in which peaks cannot be called. Error bars correspond to binomial error in the sampling of hotspots. Note that NCO density corresponds to observed NCO event whose detection power depends on heterozygocity which can vary between gene expression categories.
Fig 3.
Deficit of CO rate in the body of highly expressed genes in dogs and humans.
We cut by bins of distance to the nearest TSS or TES. All dot between the TSS and TES are inside a protein-coding gene. We distinguished 3 equal sized categories of genes regarding their expression in testis for dog [150] and in ovary for humans [151,152]. We kept only genes that have detectable expression in dog testis or human ovary (Dog: N = 23,116 genes, Humans: N = 20,050 genes). The sex-averaged recombination rates in humans were estimated from large pedigrees in [153]. The dog recombination map was estimated from linkage disequilibrium in [38]. Silhouette images are taken from www.phylopic.org.
Fig 4.
Distribution of mice DSB hotspots relative to repeated elements.
We retrieved hotspots of DMC1-ChiP-seq signal from the study of [85]. Random expectation was computed as the mean of the overlap of repeated elements and 40 sets of random DMC1 ChiP-seq peaks (see methods). If the value of the true overlap lies between those of the 40 sets of random DMC1 ChiP-seq peaks, the enrichment is not significant (NS). For a given allele, numbers correspond to the percentage of all DMC1 ChiP-seq peaks whose center is at less than 200 bp from a given repeat family. Numbers in parentheses correspond to the random expectation of the percentage of overlap.
Fig 5.
Model of the role of the default DSB-directing system (Spp1-like) and PRDM9 in the formation of crossovers.
In this schematic example, the binding probability of the DSB inducer is 0.5 for epigenetic marks, 0.25 for weak PRDM9 binding sites and 1 for strong PRDM9 binding sites. In this arbitrary setting, before the erosion of strong binding sites, PRDM9 provides an advantage compared to the default system to efficiently repair DSBs. After the erosion of strong binding sites, the default system becomes more advantageous. One can hint from this schematic representation that the distribution of binding probability of potential DSB targets, as well as the erosion level of high-affinity PRDM9 targets will be key to determine which system is more efficient.
Fig 6.
Summary of the different hypotheses for the evolutionary origins of the two types of hotspots presented throughout this review, along with the sections of the manuscript discussing them.