Molecular evolution and variation in genomic regions with low recombination

The overall goal of the research is to address the classic problem of the evolutionary significance of sex and genetic recombination, using large-scale comparisons of the properties of variation and evolutionary change among genomic regions with different levels of genetic recombination. We are exploiting the genomic resources available in Drosophila melanogaster and its close relatives. This is a particularly useful system for our purpose, since there is wide variation in the rate of recombination across the Drosophila genome. We are using these differences between genomic regions to see if recombination enhances the efficiency of natural selection, as predicted by several evolutionary models. For our comparisons with freely recombining genes, We are using a set of 200 genes that have recently been discovered in the heterochromatin, which has a very low rate of genetic recombination. This enhances our ability to detect the predicted effects of reduced recombination, especially reduced genetic diversity and levels of adaptation, and to discriminate among different specific causes of such patterns.

One of the classical problems of biology is the evolutionary role of sexual reproduction. This involves the bringing together of the genomes of the two parents, and reshuffling them by the process known as genetic recombination, so that an offspring individual receive a mixtures of contributions from each parent. This allows the evolutionary fates of genetic variants at different places in the genome to behave more or less independently of each other. One consequence of this is that natural selection can act at one site in the genome without interfering with what happens at other sites. Many specific models of evolutionary processes that can cause interference between different sites when sex is absent have been proposed: they all predict that selection is less effective in the absence of sex.
It is, however, hard to test these predictions in nature, since asexual species or populations are very rare, and in most cases have arisen only recently from sexual ancestors. It is important to do this, both for the intellectual interest in understanding why sexual reproduction is so common in nature, and because there are plans to develop asexual strains of plants for the purpose of breeding crops.
A way around this difficulty is to compare different regions of the same genome. Genetic recombination, the process that creates the reshuffling, is rare or absent in some parts of the genome, especially the part of the genome known as the heterochromatin. Such parts of the genome are expected to behave like asexual species, in terms of their evolutionary patterns. Until very recently, however, the heterochromatin has been impossible to study at the level of DNA sequences, since it contains large amounts of DNA that are repeated over and over again, making it hard to study at the sequence level. Recent breakthroughs in research on the fruitfly Drosophila, the best-studied model animal species, have led to the characterisation of several hundred genes in heterochromatin. This means we are now in a position to study evolution and variation of genes in the heterochromatin almost as easily as genes in the rest of the genome, and can therefore see whether or not they show the patterns expected from their lack of recombination. Sophisticated statistical methods are available for this purpose, but require large datasets to be used effectively.
We plan to exploit new technologies for sequencing DNA, that allow large quantities of information to be generated rapidly and cheaply. We will use these to generate data on variability in a large number of genes in the heterochromatin and in other part of the genome, within populations of two closely related species of Drosophila. By combining the results of these studies with computer-based analyses of the published sequences of other species of Drosophila, we will be able to determine whether or not the patterns that we seen in region of the genome with low levels of recombination agree with our theoretical models. If they do, we will have much more convincing evidence that sexual reproduction has an evolutionary advantage than currently exists.

1. Comparisons between related species of Drosophila of sequences obtained from publicly available databases.
We divided the D. melanogaster genome into four recombination categories: High (H), Intermediate (I), low (L), and no crossing over (N). We found 401 coding genes in the heterochromatic regions in release 5.34 of the D. melanogaster genome. We selected D. yakuba as an outgroup for estimating sequence divergence, because its divergence from D. melanogaster is sufficiently large to remove any major influence of ancestral polymorphisms, and its genome is well annotated with a high coverage. The final dataset contained 10,642 genes. The main results were as follows. Autosomal non-recombining regions (strictly speaking, non-crossing over regions, since gene conversion may occur) showed much higher levels of non-synonymous divergence (KA), slightly higher synonymous divergence (KS), and higher KA/KS than the recombining autosomal regions. Codon usage bias, as measured by the frequency of optimal codons (Fop), was significantly lower for the non-recombining genes for both the A and X datasets, although the reduction in Fop was smaller for X than A. The GC content at third position coding sites (GC3) showed the same patterns as for Fop. Levels of gene expression are similar for genes in regions with and without crossing over, which rules out the possibility that the reduced level of adaptation that we detect is caused by relaxed selection due to lower levels of gene expression in the heterochromatin. All the patterns observed are consistent with a reduction in the efficacy of selection in all regions of the genome of D. melanogaster that lack crossing over, as a result of the effects of enhanced Hill-Robertson interference. However, we also detected differences among different non-recombining locations: the X chromosome seems to exhibit the weakest effects, whereas the fourth chromosome and the heterochromatic genes on the autosomes most proximal to the centromere showed the largest effects. In addition, signatures of selection on both nonsynonymous mutations and on codon usage persist in all heterochromatic regions. The results of these analyses have been published in a leading journal. It represents the most comprehensive study to date of differences in evolutionary patterns related to the effects of low levels of genetic recombination on levels of adaptation at the molecular level.
We also investigated the following problem. Codon usage bias in Drosophila is higher for X-linked genes than for autosomal genes. In addition, silent site diversity is about the same for the X and A in East African populations of D. melanogaster, contrary to the expectation that it should be three-quarters of the autosomal value when there is a 1:1 sex ratio and equal variances in fitness for males and females. One possible explanation is that the higher effective recombination rate for genes on X compared with A (due to the absence of crossing over in males) reduces their susceptibility to the effects of selection at linked sites (Hill-Robertson effects), and thus increases the effective population size of X relative to A. The genome sequence of D. melanogaster and the Rwandan resequencing data from the DPGP were used to test this hypothesis.
After correcting for the effective recombination rates on X versus A, the ratio of X to A diversity is close to ¾, confirming a previous analysis of much more limited data. It is therefore extremely likely that selection at linked sites is indeed the cause of the equality of diversities for the chromosomes as a whole, consistent with our theoretical study of the effects of selection on linked deleterious mutations 2.
In addition, and contrary to expectation, it was found that, after correcting for the effective recombination rate, codon usage bias remained higher for X than A. In addition, analyses of the frequency distributions of synonymous polymorphisms showed that selection for preferred versus unpreferred synonymous variants is stronger on the X than the autosomes. Thus, there is inherently stronger selection on codon usage on the X than the A. Examination of possible factors such as dominances and X/A differences in gene expression levels failed to shed light on the cause of this difference in selection pressure.
These results shed new light on the factors determining chromosome-wide levels of natural variability and codon usage bias, suggesting that the former is strongly affected by the effects of selection acting on sites linked to neutral or nearly variants, and that the latter is under different selective pressures on the X chromosome relative to the autosomes. This work has also been published in a major journal.