Late in 2009, several research groups independently alighted on a histone 3 lysine 4 (H3K4) trimethylase, PRDM9, as the first known determinant of meiotic recombination hotspots in metazoa (Cheung et al., 2010, and references therein). Earlier, hotspot activity in budding yeast was found to be regulated by its only H3K4 trimethylase, encoded by the Set1 gene (Sollier et al., 2004). This apparent conservation of mechanism across diverse eukaryotes led Wahls and Davidson (2010), in a recent opinion article, to propose a ‘unifying model’ of recombination. They suggest that the positioning of recombination in eukaryotic species as different as fungi and humans is determined by the transcription factor recognition of specific DNA sites followed by histone modification and chromatin remodelling (Figure 1).

Figure 1
figure 1

A simplified representation of Wahls and Davidson's pan-eukaryotic model for recombination positioning. (a) Consensus DNA recognition sites (dark segment) within recombination hotspots are recognized by a transcription factor (TF), such as Prdm9 in mammals or Atf1-Pcr1 in yeast. Identifiable sequence motifs near hotspots fall into many classes, each presumably recognized by a separate TF. (b) In yeast, TFs recruit histone modification enzymes that add, or subtract, methyl, acetyl or ubiquitin marks (circles or stars), for example, to adjacent nucleosomes. In mammals, the hotspot-recognizing Prdm9 is itself a H3K4 trimethylase. (c) These histone modifications then recruit further elements of the recombination machinery, including Spo11, presumably to more open chromatin. This catalyses DSBs in the vicinity of the hotspot.

Full size image

In mammals and in yeast, meiotic recombination is essential for the regular segregation of homologous chromosomes during meiotic cell divisions (Coop and Przeworski, 2007). For example, if meiotic recombination is reduced or positioned suboptimally, this may result in congenital trisomatic birth defects. Recombination is nonrandomly concentrated at ‘hotspots’ that together occupy only a small fraction of the genome. This has profound implications for gene and genome evolution. For example, meiotic recombination is correlated with increases in mutation rate and in GC content, and it disrupts the co-inheritance of physically linked alleles, thereby generating diversity and promoting efficient response to natural selection.

In mammalian meiosis, double-stranded breaks (DSBs), which precede recombination, cluster at these hotspots. The zinc-finger domains of human PRDM9 bind preferentially to a minority (∼40%) of such hotspots through a degenerate 13-bp consensus motif (Myers et al., 2008). Hotspots are also evident in yeast, with one or more of five classes of degenerate DNA sequence motifs being present in 60% of randomly generated DSB hotspots for Schizosaccharomyces pombe (Steiner et al., 2009). Across the eukaryotic lineage, meiotic DNA cleavage activity is catalysed by orthologous Spo11 enzymes, thereby placing a conserved enzymatic activity at the heart of this fundamental cellular process.

Are molecular processes controlling recombination, other than Spo11-mediated DNA cleavage, also broadly conserved over eukaryotic evolution? Wahls and Davidson (2010) present evidence that this is, indeed, the case. They point to associations between histone H3K4 trimethylation and recombination at hotspots that are seen in both mouse and yeast (Borde et al., 2009; Buard et al., 2009). Furthermore, most yeast hotspots might be bound by zinc-finger or other transcription factors, as they are in mammals (by PRDM9). Could the participation of histone H3K4 trimethylases and transcription factors in recombination for organisms as different as yeasts and mammals be mere coincidence? Not at all, argue Wahls and Davidson. They instead propose that these molecules contribute to a phyletically conserved process, which requires first the binding of transcription factors to hotspot motifs, which then induces histone modifications, and is followed by the recruitment of the recombination molecular machinery to the hotspot (Figure 1).

Their model is attractive both in its simplicity and its relevance to a broad swathe of cellular life. It is also consistent with the conservation of many other fundamental cellular processes across single- and multicellular eukaryotic life. Nevertheless, they acknowledge that problems remain, two of which we discuss here.

First, although histone modification has been associated with recombination, it is yet unclear which particular histone marks (for example, acetylation, ubiquitylation, dimethylation or trimethylation) facilitate Spo11 recruitment and DSB formation, and whether these histone signals for hotspots are conserved across Eukarya. For example, the apparent pre-eminent role of methylation in determining budding yeast recombination hotspots has yet to be supported in fission yeast. Even in Saccharomyces cerevisiae, H3K4me3 cannot be a sufficient code for DSB hotspot initiation because this histone mark also reflects an active transcriptional state of genes (Berger, 2007). Moreover, DSBs still occur in the absence of H3K4me3 (Borde et al., 2009).

Second, the currently identified DNA-binding proteins that determine DSB positioning show no signs of being conserved between fission and budding yeasts, or between these yeasts and metazoans. The extremely rapid evolution of Prdm9's DNA-binding determinants is strikingly different to the situation in yeast. Moreover, although there is little evidence for turnover either among yeast hotspots or the consensus sequences that define them (Tsai et al., 2010), primate recombination hotspots are extremely short lived, with the Prdm9 zinc-fingers that bind them changing at a correspondingly rapid pace (Oliver et al., 2009; Myers et al., 2010).

The rapidity by which mammalian hotspots are turned over might resolve the ‘recombination hotspot paradox’ (Boulton et al., 1997) and may explain the extreme positive selection driving Prdm9 sequence change. As DSBs are repaired using the homologous chromosome as a template, ‘hot’ alleles with high recombination-initiation activity are constantly being replaced by ‘cold’ low-activity homologues. Once hotspots become substantially depleted, new Prdm9 alleles with altered DNA-binding specificity will be favoured, redirecting recombination to newly ‘hot’ sites, and thus ensuring a necessary number of crossover events to support disjunction. These rapid changes, however, come with an evolutionary cost, as combinations of incompatible Prdm9 alleles apparently result in male infertility (Vyskocilova et al., 2009). This, presumably, is the basis for why Prdm9 is a species-incompatibility gene (Mihola et al., 2009).

By contrast to this situation in mammals, recombination hotspots appear to be highly conserved between Saccharomyces paradoxus and S. cerevisiae (Tsai et al., 2010), species that are considerably more divergent than are chimpanzees and humans that do not share hotspots. The low frequency of sex and outcrossing (estimated as 1 in 10 000 generations in S. paradoxus), and hence meiotic recombination, in some wild-type yeasts (Tsai et al., 2008) may explain why a similarly rapid attrition of yeast recombination sites is not seen. There is accordingly little selective pressure for sequence change among yeast transcription factors governing the positioning of DSBs. This suggests that when seeking additional regulators of recombination hotspots in animals, we should be considering rapidly changing DNA-binding proteins rather than sequence-conserved yeast orthologues.

Several meiotic proteins, and their functions, appear to be conserved across the eukaryotes. This has allowed the use of tractable model organisms, including yeast, to reveal mechanisms whose disruptions underlie rare developmental abnormalities in humans. However, it is also becoming clear that other meiotic proteins show little sequence conservation and do not have corresponding orthologues across diverse taxa. Even Prdm9, which appears to have such a central role in mice and humans, has either lost its DNA-binding specificity or protein-coding function multiple times in different animal lineages (Oliver et al., 2009). It is not difficult to think of reasons why meiotic pathways might have co-opted different genes and mechanisms in different lineages: for those species in which crossovers are essential for chromosomal segregation, the positioning of hotspots could vary with the great differences among eukaryotic genome architectures. To what extent can an overarching model explain the positioning of recombination across eukaryotic species? On balance, it appears to us that the numerous differences outweigh the few similarities: the control of recombination position appears not to be substantially conserved across diverse Eukarya.