4 DISCUSSION
We integrated whole genome resequencing data from a comprehensive sampling of long-monitored populations of P. vulgaris with knowledge of the recently assembled genomes of P. vulgaris andP. veris to explain the causes and consequences of the transition from distyly to homostyly. We identified a novel loss-of-function structural mutation in CYPT associated with the transition to homostyly that had remained undetected using exonic Sanger sequencing (Mora-Carrera et al., 2021). Importantly, we found no evidence for a potential single origin of homostyly in P. vulgaris via mutations in CYPT promoter region or structural mutations involving CYPT exons, thus the previously supported hypothesis of multiple transitions to homostyly via independent loss-of-function mutations inCYPT exons stands (Mora-Carrera et al.,2021). Furthermore, population genetic analyses validated theoretical expectations for the evolutionary consequences of hemizygosity on S-locus genes and revealed differences of selective constraints among S-locus genes. Finally, the genomic resources newly available in thePrimula system enabled, for the first time, the testing of long-standing predictions on changing frequencies of S-locus genotypes during intraspecific transitions from distyly to homostyly, partially supporting the possible role of viability differences between homostyles with haploid vs. diploid S-locus genotypes in preventing the fixation of homostyly. Jointly, our study provides a detailed overview of the early molecular- and population genetic causes and consequences of mating-system transitions.
Genetic basis of transitions from distyly to homostyly in Primula vulgaris
Shifts from outcrossing to selfing are common in flowering plants and can be caused by loss-of-function mutations in the genes of interest, structural rearrangements of their exons, or mutations in their promoters (Shimizu & Tsuchimatsu, 2015). One question concerns whether disruptive mutations in the alleles that determine outcrossing act dominantly or recessively. In Brassicaceae, loss of self-incompatibility often stems from mutations in dominant alleles of genes controlling this trait (Busch et al ., 2011; Nasrallah et al. , 2017; Tsuchimatsu et al., 2012; Bachmann et al., 2019), although in Arabidopsis lyrata loss of self-incompatibility caused by mutations in recessive alleles has also been discovered (Mableet al ., 2017). In Primula , the S-locus controlling distyly is hemizygous in both P. vulgaris and P. veris (Huu et al., 2016; 2020), but previous models assumed S-locus heterozygosity, with thrum phenotype associated with the dominant S-locus allele (Bateson & Gregory, 1905). Based on this model and greenhouse crossing experiments, Crosby (1949) assumed that S-locus alleles associated with homostyly should be recessive. Previously, we documented sevenCYPT haplotypes (CYPT -2 to CYPT -7) with putative loss-of-function mutations that are occurring exclusively in P. vulgaris homostyles (Mora-Carrera et al., 2021). Two of these haplotypes (CYPT -2 andCYPT -6) have an early stop codon causing premature termination of translation. Our results indicate that, when hemizygous or homozygous, these disrupted CYPT alleles lead to homostyly. However, CYPT -2 behaves recessively in the single heterozygous individual carrying one functional and one disrupted copy of CYPT(i.e., CYPT-1/CYPT-2 , represented as S/S* in Figure 4), determining a thrum phenotype. This finding aligns with results of crossing experiments between thrums and homostyles of Primula oreodoxa showing that S* is recessive to S when the two alleles co-occur (Yuan et al ., 2018), corroborating Crosby’s prediction (1949). Therefore, our results indicate that the presence of a disrupted CYPT allele does not alter the thrum morph when paired with a functionalCYPT allele, hence the disrupted allele acts recessively.
In addition to loss-of-function mutations in coding regions, mating-system shifts can also stem from transcription silencing or exonic rearrangements in pertinent genes (Chakraborty et al. , 2023). For instance, down-regulation caused by transposon-like insertions in the promoter regions of the male-determining self-incompatibility genes MGST and BnSP11-1 trigger the shift from self-incompatibility to self-compatibility in Prunus avium and Brassica napus , respectively (Gao et al ., 2016; Ono et al ., 2020). In Primula vulgaris , previous Sanger sequencing of individual CYPT exons identified homostyles with an apparently functional CYPTallele (CYPT-1 ; Figure 1B), suggesting that homostyles might also arise through CYPTsilencing caused by a disruptive mutation in its promoter region or exonic rearrangements in CYPT not detectable via Sanger sequencing (Mora-Carrera et al ., 2021). However, 20 of the 31 homostyles analyzed in the present study had a promoter region identical to that of 37 thrums, implying that the shift to homostyly in these plants was likely caused by loss-of-function mutations inCYPT exons rather than in its promoter. The remaining 11 homostyles were characterized by a large 2150 bp deletion that eliminated both CYPT exon 1 and its promoter region (Figure 3). Thus, our current results do not support the conclusion that mutations in the promoter region or exonic rearrangements in CYPT can alone cause the shift to homostyly in P. vulgaris .
The evidence above also has implications for determining whether homostyly arose once or multiple times in P. vulgaris . The single origin of homostyly, followed by independent mutations inCYPT , would have been supported if all studied homostyles had shared the same promoter mutation or rearrangement inCYPT . However, this is not the case, favoring the hypothesis of multiple origins of homostyly via independent mutations in CYPT exons, as previously proposed (Mora Carrera et al ., 2021). Nevertheless, a study of a single homostyle from Chiltern Hills, England (population not included in our analyses), found reduced expression of CYPTwhen compared to a thrum, suggesting epigenetic silencing might play a role in the shift to homostyly (Huu et al., 2016). The mentioned study however did not provide sequences of CYPTexons, thus it remains unknown whether they contained any potentially disruptive mutations in the coding region. Therefore, transcriptome analyses of homostylous flowers are necessary to conclusively discard the possibility that disruptive promoter mutations causing reducedCYPT expression might also cause the shift to homostyly.
Finally, it remains to be explained why the three homostyles previously thought to have the functional CYPT -1 allele based on Sanger sequencing of the five CYPTexons (Mora-Carrera et al ., 2021) were here found to contain the 2150 bp deletion including exon 1 (i.e., CYPT-8 haplotype: see Figures 1B, 2, and 3). A possible explanation is that exon 1 was deleted from the S-locus (causingCYPT loss of function, hence homostyly) and translocated to a highly repetitive genomic region. The translocation could have allowed targeted amplification and subsequent Sanger sequencing using exon-1-specific PCR primers, while preventing exon-1 detection via next generation sequencing due to biases arising, for example, during genomic DNA sonication used to produce short DNA fragments prior to short-read library preparation (Poptsova et al. , 2014; Garafutdinov et al. , 2016; Jennings et al. , 2017)⁠. Notably, a few low-quality sequencing reads did map toCYPT exon 1, suggesting this exon is indeed present in the genome of these homostyles but was not successfully sequenced using short-read sequencing methodology. Long-read sequencing technologies capable of sequencing through repetitive regions would be necessary to definitively resolve whether CYPTexon 1 was translocated to a highly repetitive genomic region in these homostyles. To summarize, our findings indicate that the homostyles previously identified as having a functionalCYPT allele in fact possess a disruptedCYPT allele due to exon 1 deletion (designatedCYPT -8 allele: Figure 3). Overall, these results emphasize that not only non-synonymous mutations or small deletions, but also structural rearrangements such as large deletions and translocations can cause mating-system transitions.