Evolution of Recombination Hotspots in Vertebrates

Table of contents

1. The Recombination Hotspot Paradox
2. Death and Resuscitation of Recombination Hotspots
3. Rapidly evolving Recombination Landscapes
4. Evolutionary origin of PRDM9
5. Published Articles

The Recombination Hotspot Paradox

In many species, recombination does not occur uniformly throughout the genome; rather, it usually concentrates in areas of high crossing-over activity called Recombination Hotspots. There are two kinds of such hotspots. In some taxa like birds, recombination concentrates near Transcription Start Sites. In these cases, the mechanism of recombination localization is thought to be passive: crossing-over happens in these areas as the protein machinery takes advantage of open chromatin state associated with the initiation of transcription. In other taxa such as the vast majority of mammals, recombination localization is actively determined by a protein called PRDM9. This protein, whose structure has now been determined, uses its highly variable zinc-finger subdomains to recognize and bind specific target sequences in the genome. Once bound, PRDM9 facilitates the opening of chromatin and recruits the crossing-over machinery.

Interestingly, this recombination-determination mechanism means that recombination hotspots are self-destructive in nature. Indeed, crossing-over can result in the biased gene conversion of the target allele bound by PRDM9 into the homologous allele. As PRDM9 is sequence-specific, heterozygosity means that individual hotspots are expected to be progressively eroded, and eventually disappear. Yet, PRDM9-directed hotspots are still alive and going strong in a wide range of vertebrate taxa. This discrepancy between prediction and empirical data is what has been referred to as the Recombination Hotspot Paradox . Current resolution of the paradox relies on selection for crossing-overs. Indeed, it has been shown that a reduction in the number of crossing-overs can lead to aneuploidy: crossing-overs seem to be involved in the stabilization of the pairing of homologous chromosomes in Prophase I, allowing for their correct segregation. If selection for crossing-over is very strong, models predict that it should act against biaised gene conversion and maintain hotspots. However, this seems to require unrealistic selection intensity, and is in contradiction with empirical data that show that PRDM9, hotspot localization and overall recombination landscapes are all rapidly evolving. If selection is mild, the "Red Queen Theory of Recombination Hotspots" predicts that while individual hotspots indeed "die" from biased gene conversion, PRDM9 sequence evolves to "give birth" to new recombination hotspots at alternative target sites. It is important to note that current resolution of the paradox critically relies on the availability of alternative PRDM9 sequences able to bind new sets of targets. This availability has often been supposed, though it may be not that clear. Indeed, suitable target sequences must: (i) be present in sufficient numbers in the genome, (ii) exist all throughout the genome, in every chromosome, and (iii) not be present in unsuitable locations for crossing-over, such as exons or regulatory sequences. In finite genomes, chances are that the number of alternative target sequence is limited. Arguably, the evolutionary process suggested by the Red Queen model would rapidly exhaust them. In some way, it seems that the Red Queen does not resolve the paradox, but merely push back the necessary resolution.

Death and Resuscitation of Recombination Hotspots

To better understand the evolutionary dynamics of PRDM9-directed recombination hotspots, we build the first model formally taking into account:

• allele matches between PRDM9 and target sequences

• gene conversion at target loci

• selection for crossing-over

• mutation at PRDM9 and target sequences

In an extension of the model, we look at the case where there are two target loci. Interestingly, we see on Fig. 4 that crossing-over levels at the two targets tend to fluctuate in asynchrony: recombination landscapes vary rapidly while overall crossing-over levels remain stable. This is also a pattern that fits empirical knowledge and is difficult to explain via the Red Queen theory.

Rapidly evolving Recombination Landscapes

We ran numerical simulations of this model with a varying number of target loci. An example of results is shown on Fig. XX. We see that in stochastic simulations, each target locus fluctuates between hot and cold states, while overall mean crossing-over rates are much more stable. Interestingly, we see that two simulations with the same sets of loci and with the same initial conditions rapidly diverge, showing at any given time different combinations of hot targets. As such, our model is compatible with the observation that closely-related species exhibit very different recombination landscape: even though they use overall the same set of targets, the actual combinations of hotspots can be very different.

PRDM9 is known from empirical data to exhibit a particularly high mutation rate, 10-fold higher than for other genes. Thought to be in part responsible for PRDM9's fast evolutionary changes, the cause of this feature is still unknown. To explore whether our model can account for that too, we build a 3-locus extension with: one target locus, one PRDM9 locus and one mutation modifier locus.

Mathematical analysis shows that a modifier allele that increases mutation rates of PRDM9 relative to the target is selectively favored. Interestingly, this is not necessarily because such an allele increases individuals' fitness; rather, it is due to linkage effects.

Evolutionary origin of PRDM9

Ongoing top-secret work, more in the upcoming months.