The Evolution of Selfing in Arabidopsis thaliana

See allHide authors and affiliations

Science  24 Aug 2007:
Vol. 317, Issue 5841, pp. 1070-1072
DOI: 10.1126/science.1143153


Unlike most of its close relatives, Arabidopsis thaliana is capable of self-pollination. In other members of the mustard family, outcrossing is ensured by the complex self-incompatibility (S) locus,which harbors multiple diverged specificity haplotypes that effectively prevent selfing. We investigated the role of the S locus in the evolution of and transition to selfing in A. thaliana. We found that the S locus of A. thaliana harbored considerable diversity, which is an apparent remnant of polymorphism in the outcrossing ancestor. Thus, the fixation of a single inactivated S-locus allele cannot have been a key step in the transition to selfing. An analysis of the genome-wide pattern of linkage disequilibrium suggests that selfing most likely evolved roughly a million years ago or more.

The transition from outcrossing to selfing is a major theme in the evolution of flowering plants, having occurred independently in numerous lineages (1). Although it leads to inbreeding depression, the ability to self can be advantageous when colonizing new territory and is therefore associated with weedy and invasive species. A. thaliana, a member of the Brassicaceae, is a highly selfing plant that is separated from its closest relative, the self-incompatible A. lyrata,by about 5 million years (2, 3).

In the Brassicaceae, the main components of the S locus are the tightly linked S-locus cysteine-rich protein (SCR) and S-locus receptor kinase (SRK) genes, which encode male and female specificity determinants of self-incompatibility, respectively (4). The S locus is characterized by highly diverged haplotypes, as predicted by theory (5). Because A. thaliana is self-compatible, the self-incompatibility system must be inactive, and both SCR and SRK are pseudogenes (ΨSCR and ΨSRK) in the reference accession Col-0 (5). Because transformation with S-locus alleles from A. lyrata can restore self-incompatibility in A. thaliana (6), inactivation of the S locus could have been the key step in the transition to selfing. If this occurred only once, it should have sharply reduced variability at this locus [a so-called “selective sweep” (7, 8)]. Even if such an event happened a long time ago, it should be readily detectable because the ancestral locus would have been extremely polymorphic (5, 9).

We investigated variation at the S locus using several approaches. Polymerase chain reaction (PCR) amplification of ΨSCR and ΨSRK failed in many of 96 surveyed accessions (10), indicating that these genes are either very diverged in sequence or missing. These results were supported by whole-genome resequencing data for 20 accessions with oligonucleotide arrays (11). The S-locus region of many accessions did not hybridize well to the arrays, which had been designed based on the reference sequence. Several, highly diverged haplotypes extend throughout the S-locus region (table S1). About 35 accessions are very similar to the Col-0 reference. Cvi-0, from the extreme south of the species' range, carries a very distinctive haplotype that occurs only once in our sample. Another distinctive group includes nine accessions, with origins ranging from Kashmir (12) to Spain. The remaining accessions are characterized by different but consistent patterns of missing data. This latter group includes C24, which carries an extensively rearranged haplotype (see below).

We dideoxy-sequenced bacterial artificial chromosomes (BACs) covering the S locus from two accessions: C24, which was the only accession for which self-incompatibility was restored upon transformation by S-locus alleles from A. lyrata (6), and the highly divergent Cvi-0 (13). In C24, the S locus is extensively rearranged as compared with that of Col-0, and ΨSCR is deleted (14), whereas in Cvi-0, the S locus is similar in structure to Col-0 but is highly diverged in sequence (Fig. 1A). Most of the sequence between the U-box gene and ΨSRK cannot be aligned for the three haplotypes. Cvi-0 SCR carries no obvious null mutation, whereas Cvi-0 ΨSRK has a distinctive splice-site change (fig. S1).

Fig. 1.

(A) Plots illustrating sequence conservation (18) across the S-locus region between C24 and Col-0 (bottom plot) and between Cvi-0 and Col-0 (top plot). Colored regions designate the U-box, ΨSCR, ΨSRK, and ARK3 loci, respectively. Sequence similarity is represented on the y axis. Most of the S locus is too diverse to be aligned for the three haplotypes. (B) Plot of LD [measured as r2, with the darkest and lightest colors indicating complete linkage and no linkage, respectively (19)] across the S-locus region. See table S1 for a summary of investigated polymorphisms. Because the observed polymorphisms are highly clumped, the spacing in the plot is nonlinear: Thin lines connect the actual positions of polymorphism along the chromosome. The highlighted blocks illustrate strong LD not only within the U-box and ARK3 genes but also (i) across the S locus from At 4g21360 (a transposable element) through ΨSRK and (ii) across the entire region from the U-box to ARK3. (C and D) Observed LD decay in genome-wide single-nucleotide polymorphism data (11) as compared with coalescent simulation results (10), assuming a constant population size (C) or assuming population growth (D). The simulations fit the data only if the transition to selfing occurred so recently that no trace would be visible (“no change in selfing rate”) or so long ago that the trace would have been lost [“change at T = 1 million years (My)”]. Ky, thousand years.

Because we found high variation, our data rule out the fixation of a single loss-of-function allele at the S locus as a key step in the transition to selfing. Instead, the ancestral balanced polymorphism at the S locus is gradually being eroded through genetic drift, perhaps in combination with selection for inactivity: a process that may be very slow, especially in highly structured populations. The A. thaliana SRK alleles are shared with closely related species (15). Given that linkage disequilibrium (LD) extends throughout the S-locus region (Fig. 1B), it seems likely that the same will be true for SCR. At the same time, the observation that transformation with S-locus alleles from A. lyrata does not always restore self-incompatibility (6) suggests multiple evolutionary routes to selfing.

Our results contradict a report of low variability at ΨSCR (but not ΨSRK) resulting from a recent selective sweep (12). The disagreement is not due to differences in the sample that was used: We were unable to replicate previously published results (12) using the published PCR primers and accessions (table S1 and fig. S2), and the reported ΨSCR sequence in Cvi-0 disagrees with our BAC sequence, although the highly divergent ΨSRK is identical in our BAC sequence and in the sequence in the previous report. We therefore conclude that the published results are erroneous. It has been noted that a selective sweep at SCR, but not SRK, seemed unlikely (15): Our data resolve this contradiction.

Finally, we considered when selfing might have evolved. It has been estimated that SRK started to become a pseudogene no more than 413,000 years ago (15); however, the transition to selfing could have taken place earlier if loss of S-locus function was not the first step [as suggested by the discovery of a modifier of self-incompatibility in A. thaliana (16)]. To obtain an alternative estimate, we examined the genome-wide pattern of LD (11). LD in a partially selfing organism should decay like that in an outcrosser, albeit more slowly (17). However, a very different pattern is expected with a recent, dramatic change in outcrossing, because recombination events that took place before the transition to selfing would have been more effective in breaking up LD. As a result, long-range LD, which reflects recent events, should be too high relative to short-range LD, which reflects more ancient events. This pattern should be detectable unless selfing evolved either a very long time ago (on the order of the coalescence time, or about 106 years in A. thaliana) or very recently, in which case the LD pattern should still look like that of an outcrosser. The LD decay reveals no indication of a recent change in selfing (Fig. 1, C and D), and because the LD pattern is very different from that in an outcrosser, species-wide selfing most likely evolved on the order of a million years ago or more. This would also have provided ample time for the suite of selfing-associated traits that distinguish A. thaliana from A. lyrata to evolve (3).

Supporting Online Material

Materials and Methods

Figs. S1 and S2

Tables S1 and S2


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article