What causes mating system shifts in plants? Arabidopsis lyrata as a case study

Abstract

The genetic breakdown of self-incompatibility (SI) and subsequent mating system shifts to inbreeding has intrigued evolutionary geneticists for decades. Most of our knowledge is derived from interspecific comparisons between inbreeding species and their outcrossing relatives, where inferences may be confounded by secondary mutations that arose after the initial loss of SI. Here, we study an intraspecific breakdown of SI and its consequences in North American Arabidopsis lyrata to test whether: (1) particular S-locus haplotypes are associated with the loss of SI and/or the shift to inbreeding; (2) a population bottleneck may have played a role in driving the transition to inbreeding; and (3) the mutation(s) underlying the loss of SI are likely to have occurred at the S-locus. Combining multiple approaches for genotyping, we found that outcrossing populations on average harbour 5 to 9 S-locus receptor kinase (SRK) alleles, but only two, S1 and S19, are shared by most inbreeding populations. Self-compatibility (SC) behaved genetically as a recessive trait, as expected from a loss-of-function mutation. Bulked segregant analysis in SC × SI F2 individuals using deep sequencing confirmed that all SC plants were S1 homozygotes but not all S1 homozygotes were SC. This was also revealed in population surveys, where only a few S1 homozygotes were SC. Together with crossing data, this suggests that there is a recessive factor that causes SC that is physically unlinked to the S-locus. Overall, our results emphasise the value of combining classical genetics with advanced sequencing approaches to resolve long outstanding questions in evolutionary biology.

Figure 1

Minimum evolution genealogy of B80 alleles, indicating associations with SRK alleles and geographic distribution. The frequency of each allele is indicated in parentheses after its name. The tree was reconstructed using MEGA 6.0, under a Kimura 2 parameter model of evolution, with rate heterogeneity modelled under a gamma distribution using a rate parameter of 0.45. Numbers on the nodes indicate bootstrap support based on 1000 pseudoreplicates. As low phylogenetic resolution is expected for genes evolving under balancing selection, the main purpose of the tree is for visualisation of relatedness among B80 alleles in relation to their association with SRK alleles. Associated SRK alleles are indicated by name and using coloured branches. Occurrences of each B80 allele in inbreeding and outcrossing populations and in each of the six genetic clusters predicted by STRUCTURE are indicated in the table to the right.

Figure 2

SHOREmap output for chromosomes 5, 6 and 7. The trace in red shows comparison of the SC pool with the reference sequence AL4 (from an SC individual from RON) and the trace in blue shows that for the SI pool. The scale at the bottom shows the position along the chromosome. The plots were produced using a step size of 10 000 and a window size of 200 000 bp. For each chromosome plot, the y axis indicates the proportion of reads either matching or showing an alternative to the reference sequence: 0 indicates fixation of variants that match AL4 and 1 indicates fixation for a different variant; the red line in the middle shows 50% heterozygosity. Note that for most regions, there is no difference between the SI and SC pools, whereas on the short arm of chromosome 5 (and near the centromere) and the long arm of chromosome 7 there are extended regions where the SC pool is more homozygous than the SI pool and skewed towards values near 0 (indicating that it is the same as the AL4 sequence); several examples are shown with arrows on the two chromosomes. The most concentrated region showing this pattern is between 9 and 10 Mb on chromosome 7, the location of the S-locus.