Evolution of self-fertilization may be initiated by a historical population bottleneck, which should diagnostically reduce lineage-wide genetic variation. However, selfing can also strongly reduce genetic variation after it evolves. Distinguishing process from pattern is less problematic if mating system divergence is recent and geographically simple. Dramatically reduced diversity is associated with the transition from outcrossing to selfing in the Pacific coastal endemic Abronia umbellata that includes large-flowered, self-incompatible populations (var. umbellata) south of San Francisco Bay and small-flowered, autogamous populations (var. breviflora) to the north. Compared to umbellata, synonymous nucleotide diversity across 10 single-copy nuclear genes was reduced by 94% within individual populations and 90% across the whole selfing breviflora lineage, which contained no unique polymorphisms. The geographic pattern of genetic variation is consistent with a single origin of selfing that occurred recently (7–28 kya). These results are best explained by a historical bottleneck, but the two most northerly umbellata populations also contained little variation and clustered with selfing populations, suggesting that substantial diversity loss preceded the origin of selfing. A bottleneck may have set the stage for the eventual evolution of selfing by purging genetic load that prevents the spread of selfing.

We randomly sampled two individuals from each of 24 populations across the species geographic range, including 15 putatively outcrossing umbellata and nine putatively selfing breviflora populations. We defined a population as a group of plants separated from other such groups by 1 km, usually much more. We extracted DNA from each silica-dried leaf sample using a modified CTAB protocol and amplified the coding regions of 10 single-copy genes using primers derived from a transcriptome for A. umbellata. GenomeQuebec (Montreal, QC) sequenced the forward and reverse strands of these amplicons, using the same primers. We visualized chromatograms with FinchTV (v 1.4.0) and manually edited to identify heterozygous positions (double peaks) and remove low quality ends (peaks of poor integrity). Forward and reverse strands were aligned using BioEdit (v 7.2.5) and then collapsed to one consensus sequence per individual (mean = 530 bp/locus) from which haplotypes were estimated using PHASE (v 2.1.1). One breviflora genotype (population CGB) was removed from further analysis because it was unusually heterozygous, possessed unique single nucleotide polymorphisms (SNPs) and was heterozygous indels at several loci  suggesting that it was a hybrid with sympatric A. latifolia.

The data are open-source .csv or .fasta files and can be opened with any text reader. Most key analyses were done using open-source R. Details for useage are included in a readme file.