1 INTRODUCTION
The repeated shift from outcrossing to selfing is a central topic in plant evolution (Stebbins, 1957; Cutter, 2019)⁠. Previous studies used phenotypic traits typically associated with selfing to estimate, for example, the number and tempo of transitions to selfing in phylogenies of ancestrally outcrossing taxa (Goldberg & Igic, 2012; de Vos et al ., 2014)⁠. However, missing knowledge of the genes that control mating systems has hindered the study of molecular processes associated with transitions to selfing until recently, especially in non-model organisms. Current advances in genomics now facilitate the identification of the genes and mutations associated with mating-system shifts.
A prime model to investigate the transition from outcrossing to selfing has been the shift from distyly to homostyly in Primula (Barrett, 2019)⁠. Distyly is characterized by the co-occurrence in populations of two types of self-incompatible individuals, called pins and thrums, distinguished by the reciprocal arrangement of male (anthers) and female (stigma) sexual organs in their flowers (Figure 1A; Ganders, 1979; Lloyd & Webb, 1992; Keller et al. , 2014). This floral heteromorphism represents an adaptation for outcrossing reported in at least 26 angiosperm families (Naiki, 2012). Conversely, homostyly is a floral homomorphism that enables selfing. It is characterized by self-compatible individuals bearing flowers with both stigma and anthers at the same level in the corolla tube (Figure 1A; Barrett, 2019). Evidence supporting higher selfing in homostylous than distylous plants has been reported in diverse taxa (Belaoussoff & Shore, 1995; Schoenet al. , 1997; Mora-Carrera et al. , 2021). Independent shifts from distyly to homostyly have been documented both within and among species (Zhou et al. , 2012; Kissling & Barrett, 2013; de Vos et al. , 2014; Ruiz-Martín et al. , 2018; Costa et al. , 2019).
It has long been known that the S-locus supergene controls distyly and the shift to homostyly (Lewis & Jones, 1992). However, the molecular and functional characterization of the S-locus has been performed only recently. The breakthrough occurred in the Primula system, where the S-locus comprises five genes (CCMT ,CYPT , GLOT ,KFBT , and PUMT ) and is hemizygous in thrums (S/0) but absent in pins (0/0; Figure 1A; Liet al. , 2016; Potente et al. , 2022). Two S-locus genes were recently shown to control key traits in thrum flowers:GLOT determines high anthers, whileCYPT determines short stigma and self-incompatibility (Huu et al ., 2016; 2020; 2022). Specifically, experimental silencing of GLOT inPrimula forbesii thrums lowered anther position, producing flowers with both anthers and stigma in the middle of the corolla tube (i.e., short-homostyly). However, self-incompatibility was retained, preventing self-fertilization in short-homostyles (Huu et al. , 2020). Conversely, silencing of CYPT inPrimula veris was associated with both style elongation and loss of self-incompatibility, thus turning self-incompatible thrum flowers into self-compatible, homostylous flowers with both stigma and anthers at the mouth of the corolla tube (i.e., long-homostyly; Huu et al., 2016; 2022). Although both short- and long-homostyly have been reported, the latter type is most common in Primula (Charlesworth & Charlesworth, 1979; Lewis & Jones, 1992), likely because self-compatibility in long-homostylous flowers enables self-fertilization and reproductive assurance (Mora-Carrera et al., 2021). Therefore, we hereafter refer to long-homostyly simply as homostyly (Figure 1A).
The shift from distyly to homostyly has been intensely studied in populations of P. vulgaris from Somerset, England, that display variation of thrums, pins, and homostyles (Crosby 1940; 1949). Targeted Sanger sequencing of the five individual CYPTexons of homostyles from the mentioned populations revealed that all tested thrums shared the same functional CYPT allele (CYPT -1; Figure 1B). Contrariwise, 21 homostyles harbored six different CYPT alleles, each with a unique, potentially disruptive mutation (CYPT -2 to CYPT -7; Figure 1B; Li et al., 2016; Mora-Carrera et al., 2021). One possible explanation for the lack of sharedCYPT mutations among the homostyles is that homostyly evolved independently multiple times. However, the same study also found that six homostyles from two different populations had the same CYPT allele as that of thrums (i.e.,CYPT -1). This result raised the possibility that homostyly initially arose via CYPT silencing caused by either a structural rearrangement (such as an inversion or a translocation) involving any of theCYPT exons or an inactivating mutation in theCYPT promoter, followed by multiple, unique mutations in CYPT exons, as those found inCYPT-2 to CYPT-7(Mora-Carrera et al. , 2021; Charlesworth, 2022). Both types of mutations (structural rearrangements in CYPT or silencing of CYPT promoter) cannot be detected using Sanger sequencing of individual CYPTexons. Determining whether homostyly in P. vulgaris arose multiple times via independent mutations inCYPT exons or once through a shared structural rearrangement involving CYPT exons or a mutation in the CYPT promoter requires the mapping against a genomic reference of extensive genomic sequences covering both the S-locus and its upstream region. Both types of resources are now available from whole genome resequencing data (WGR) and published genomes for P. vulgaris (Cocker et al ., 2018) and its close relative P. veris (Potente et al ., 2022)
Furthermore, the availability of WGR data and reference genomes in the selected study group facilitates the testing of population genetic predictions concerning the evolution of the entire S-locus and S-locus gene-paralogs in thrums, pins and homostyles. First, the thrum-specific segregation of the hemizygous S-locus should cause a 3/4th reduction of effective population size (Ne ) (Huu et al., 2016), hence a decrease of genetic diversity in S-locus genes compared to the rest of the genome (Gutiérrez-Valencia et al. , 2021). Secondly, hemizygosity could have contrasting effects on the efficacy of purifying selection on S-locus genes. On the one hand, the reduction of Nein the S-locus should make purifying selection less efficient (Huuet al., 2016). On the other hand, similarly to what happens in the Y sex-chromosome (Gossmann et al ., 2011), selection to maintain function of S-locus genes and the exposure of recessive deleterious mutations under hemizygosity should enhance the efficacy of purifying selection. However, the extent to which the efficacy of purifying selection differs between genes within and outside the S-locus remains poorly understood (Potente et al ., 2022). Finally, the transition to homostyly could also reduce genetic diversity in S-locus genes due to increased homozygosity in homostyles (Mora-Carrera et al., 2021). The high-quality annotation of the five S-locus genes and their four paralogs (CCM1 , CYP734A51 , GLO1 , andKFB1 ) in Primula (Li et al., 2016; Potente et al., 2022), combined with sequences of these nine genes extracted from WGR data, provides an ideal opportunity to test the above predictions for the molecular evolution of the S-locus.
Additionally, the ability to assess the ploidy level of specific genomic regions extracted from WGR data now enables the testing of predictions about the changing frequencies of haploid and diploid S-locus genotypes during the transition from distyly to homostyly. In a pioneering study, Crosby (1949) proposed a model for how the frequencies of thrum, pin, and homostylous phenotypes change over time (Figure 1C and D). This model rested on then accepted genetic model for the S-locus, which assumed that thrums were typically heterozygous dominant at the S-locus, pins homozygous recessive, and homostyles stemmed from thrums via recombination at the S-locus (Bateson & Gregory, 1905). Crosby assumed that the viability of homozygous homostyles was either 35% lower than or equal to the viability of pins, thrums, and heterozygous homostyles. The assumption of lower viability for homozygous homostyles rested on previous studies by Mather and Winton (1941) proposing that homozygous dominant thrums (S/S) had lower viability than heterozygous thrums. Crosby’s model is applicable also under the recently demonstrated hemizygosity of the S-locus in Primula , by assuming that homostyles with a diploid S-locus (S*/S*-genotypes; where S* indicates an S locus with disrupted CYPT ) can have either lower or equal viability as that of homostyles with a haploid S-locus (S*/0-genotype), thrums (S/0-genotype), and pins (0/0-genotype; see Figures 1C and D).
In P. vulgaris , repeated phenotypic surveys conducted in Somerset, England, have shown that, when homostyles are present at high frequency, thrums tend to be less frequent and, in some cases, absent, compared to pins (Crosby, 1949; Curtis & Curtis, 1985; Mora-Carreraet al. , 2021). These findings align with the predictions of Crosby’s model under lower viability of S*/S*-homostyles (Figure 1C). However, one Somerset population consisted exclusively of homostyles (Curtis & Curtis, 1985; Mora-Carrera et al., 2021), suggesting that the fixation of homostyly is possible, as expected under the model with equal viability for S*/S*- and S*/0-homostyles. However, previous phenotypic surveys could not discriminate between S*/S*- and S*/0- genotypes for homostyles. Recently developed sequencing technologies enable the estimation of sequencing depth at the S-locus (Gutiérrez-Valencia et al. , 2022), allowing us to determine whether the S-locus is haploid or diploid in both homostylous and thrum individuals. Therefore, it is now possible to estimate whether the observed frequencies of S*/S*- and S*/0- homostyles in natural populations support the model assuming lower or equal viability for S*/S*- genotypes in relation to the other genotypes.
Here, we analyze WGR data from nine populations of P. vulgariswith varying frequencies of pins, thrums, and homostyles, to answer the following questions: 1) Do all homostyles carrying different disruptedCYPT alleles share either a mutation in the promoter region and/or a structural rearrangement involvingCYPT exons that might disruptCYPT function, allowing for the possibility of a single origin of homostyly? 2) Do S-locus genes have lower genetic diversity and efficacy of purifying selection than their paralogs? 3) Do homostyles have lower genetic diversity in S-locus genes than thrums? 4) Do observed frequencies of S* /0- andS*/S*- homostyles in natural populations better match genotypic frequencies predicted under the assumptions of lower or equal viability for S*/S*- homostyles? Our study illustrates how knowledge of the genes controlling mating systems combined with high-quality genomic resources generates novel insights into the genotypic changes and evolutionary consequences associated with phenotypic transitions from outcrossing to selfing.