1 INTRODUCTION
A central topic in evolutionary biology is the repeated transition from outcrossing to selfing (Cutter, 2019)⁠. A main advantage of selfers vs. outcrossers is the increase in the number of gene copies transmitted to the next generation (i.e., automatic selection advantage; Busch & Delph, 2012). Additionally, selfers can reproduce when mates or pollinators, in the case of plants, are scarce (i.e., reproductive assurance; Darwin, 1876). These advantages are counteracted by fitness reduction in selfed progeny due to inbreeding depression (Schemske & Lande, 1985). In plants, transitions to selfing are enabled by the loss of self-incompatibility (SI). The genetic basis of shifts from SI to self-compatibility (SC) has been investigated in systems where the genes responsible for self-pollen recognition have been characterized (Shimizu & Tsuchimatsu, 2015). Repeated transitions to SC via the inactivation of SI genes are often studied by comparing closely related species with contrasting mating systems (Vekemans, Poux, Goubet, & Castric, 2014). Although independent losses of SI have also been reported within species (Busch, Joly, & Schoen, 2011; Foxe et al., 2010; Goodwillie 1999), only a few studies have investigated the molecular basis of these losses (Chantha, Herman, Platts, Vekemans, & Schoen, 2013; Shimizu, Shimizu-Inatsugi, Tsuchimatsu, & Purugganan, 2008).
A classic model for the study of mating system transitions is the shift from heterostyly to homostyly (Barrett, 2019)⁠. Heterostyly, occurring in at least 28 angiosperm families, is a floral polymorphism that promotes outcrossing, whereby two (distyly) or three (tristyly) floral morphs differing in the position of reproductive organs co-occur⁠ in a population (Darwin, 1877). In distylous species, flowers of the long-styled morph, henceforth L-morph (also known as pin), exhibit the stigma above the anthers, whereas flowers of the short-styled morph, henceforth S-morph (also known as thrum), show the reciprocal arrangement (Ganders, 1979; Figure 1). The reciprocal position of male and female sexual organs in flowers of heterostylous taxa simultaneously promotes pollen export to individuals of the opposite floral morph and reduces self-pollination (Keller, Thomson, & Conti, 2014; Lloyd & Webb, 1992)⁠. Moreover, heterostyly is often, but not always, associated with a heteromorphic self-incompatibility system that reduces self- and intra-morph fertilization (Dulberger, 1992)⁠. Heterostyly is controlled by a single Mendelian locus known as the heterostyly supergene or S-locus (Lewis & Jones, 1992). Recent genomic and functional studies in Primula indicated that the S-locus is entirely absent from the L-morph and hemizygous in the S-morph, where it consists of five tightly linked genes (Huu et al., 2016; Li et al., 2016). Two genes in the S-locus, GLOBOSA2 (GLOᵀ) and CYP734A50 (CYPᵀ), have been experimentally demonstrated to control anther and stigma positions, respectively (Huu et al., 2016; 2020)⁠.
As common for complex traits, heterostyly has been repeatedly lost. Conforming to predictions about mating system transitions, losses of heterostyly are usually associated with the decline or absence of specialized pollinators (Pérez-Barrales & Arroyo, 2010; Yuan et al., 2017) or the absence of plant mates, as in colonization after long-distance dispersal (Ness, Wright, & Barrett, 2010) and recolonization after glacial retreat (Guggisberg, Mansion, Kelso, & Conti, 2006). The breakdown of heterostyly can occur either through the loss of one or more of the heterostylous floral morphs (e.g., transitions from tristyly and distyly to monomorphism; Ness et al., 2010; Pérez-Barrales, Pino, Albaladejo, & Arroyo, 2009) or the invasion of homostylous plants (Barrett, 2019). Homostylous flowers are characterized by reduced distance between male and female sexual organs (i.e., reduced herkogamy) and, typically, loss of SI (de Vos et al., 2014)⁠. Both long- and short-homostyles are known, the latter being rare (Lewis & Jones, 1992). Long- and short-homostyles, respectively, bear both anthers and stigmas either higher (H-morph in Figure 1) or lower in the corolla tube. Homostyly enables reproductive assurance through autonomous self-pollination, as demonstrated in Primula (Carlson, Gisler, & Kelso, 2008; de Vos, Keller, Isham, Kelso, & Conti, 2012; Yuan et al., 2017)⁠, and increases selfing rates (Husband & Barrett, 1993; Schoen, Johnston, L’Heureux, & Marsolais, 1997; Zhong et al., 2019)⁠. Repeated shifts from heterostyly to homostyly have been reported both between (de Vos, Hughes, Schneeweiss, Moore, & Conti, 2014; Kissling & Barrett, 2013; Kohn, Graham, Morton, Doyle, & Barrett, 1996; Ruiz-Martín et al., 2018) and within species (Ness et al., 2010; Shao et al., 2019; Zhou et al., 2017). However, none of these studies identified the molecular basis of such transitions, likely because the necessary genomic resources have only recently become available.
Recent studies in Primula revealed that the transition from heterostyly to homostyly is achieved via disruption of either GLOᵀ or CYPᵀ. Silencing GLOᵀ caused a reduction of anther height in S-individuals, leading to the expression of short-homostyly in Primula forbesii (Huu, Keller, Conti, Kappel, & Lenhard, 2020). However, heteromorphic SI was retained, preventing self-fertilization⁠ and possibly explaining, together with the lower mutation rate of GLOᵀ than CYPᵀ, the low frequency of short-homostyles observed in nature (Huu et al., 2020). Similarly, a transposable element insertion likely compromising the function of GLOᵀ was detected in one short-homostyle of a cultivated P. vulgaris variety (Li et al., 2016)⁠. Conversely, silencing CYPᵀ in S-individuals of P. veris increased stigma height, causing the shift to long-homostyly (Huu et al., 2016)⁠⁠. Correspondingly, the expression of CYPᵀ was reduced in long-homostyles compared to S-individuals of both P. vulgaris and P. oreodoxa (Huu et al., 2016; Zhao, Luo, Yuan, Mei, & Zhang, 2019). Furthermore, all five exons of CYPᵀ were likely lost in long-homostyles of the ancestrally heterostylous P. forbesii, whereas exon three of CYPᵀ was not detected by PCR assays in the long-homostylous P. grandis, P. halleri, and P. scotica (Huu et al., 2016). ⁠Moreover, two loss-of-function mutations in CYPᵀ were reported in self-compatible long-homostyles of P. vulgaris, suggesting that CYPᵀ might additionally be involved in the control of SI (Li et al., 2016), but this has not been experimentally demonstrated (Kappel, Huu, & Lenhard, 2017). The findings above demonstrate that any genetic changes disrupting the function or expression of CYPᵀ, a gene present only in S-individuals, cause long-homostyly. However, comprehensive analyses of the different types of mutations associated with shifts to homostyly based on extensive sampling of heterostyles and homostyles from multiple natural populations have never been performed.
After homostyles originate, their reproductive advantages over heterostyles should prompt changes in floral morph composition and genetic diversity of populations. Previous theoretical models predicted that homostyles should increase in frequency by replacing S-individuals (Crosby, 1949; 1960). Concomitantly, occasional crosses between homostyles and L-individuals and reduced fitness of homostyles, presumably caused by homozygosity of genes at the S-locus, were expected to prevent the fixation of homostyly (Crosby, 1949; 1960). Both mechanisms should ultimately result in the co-occurrence of homostyles with L-individuals, the latter remaining at low frequency. The predictions above were supported by two surveys of the frequency of L-, S- and homostylous individuals from natural populations of P. vulgaris performed at a 36-yr. interval (Crosby, 1949; Curtis & Curtis, 1985). Subsequent studies testing whether homostyles were affected by inbreeding depression found no evidence of seed-mass differences between heterostyles and homostyles (Piper, Charlesworth, & Charlesworth, 1986). However, inbreeding depression at subsequent life-cycle stages has never been investigated. At the population genetic level, higher selfing in homostyles should increase homozygosity, thus decreasing diversity within and increasing differentiation among populations (Hamrick & Godt, 1996)⁠. These microevolutionary consequences of mating system transitions have been quantified only in a few homostylous species (Ness et al., 2010; Yuan et al., 2017; Zhou et al., 2017; Zhong et al., 2019).
Primula vulgaris Huds. (the common primrose) represents a classical model for the study of heterostyly and intraspecific transitions to homostyly (Bodmer, 1958; Cocker et al., 2018; Crosby, 1959; Piper et al., 1986). While distylous populations consisting exclusively of L- and S-individuals are common throughout Eurasia, mixed populations with varying frequencies of long-homostylous, S-, and L-individuals have been reported and widely documented for decades only in Somerset, England (Crosby, 1940; Curtis & Curtis, 1985). The occurrence of homostyles in multiple populations and the wide variation in their frequency in this region prompted Crosby (1949) to suggest that homostyles migrated from an initial center of origin to neighboring populations. However, two different loss-of-function mutations in CYPᵀ were recently detected in two long-homostyles from different natural populations of P. vulgaris in England (Li et al., 2016)⁠, suggesting that homostyly might arise independently rather than via migration. Therefore, population genetic analyses aimed at clarifying the role of migration in the spread of homostyly are needed.
This study investigates the origins of homostyly, its effects on the mating system, and its population genetic consequences using the well-characterized P. vulgaris model. We thus combined surveys of floral morph frequencies from multiple natural populations of the ancestrally heterostylous P. vulgaris with both DNA sequence analyses of CYPᵀ and microsatellite data from multiple individuals to ask the following questions: 1) Is an increase in the frequency of homostyles associated with a reduction of S-individuals and the maintenance of L-individuals at low frequencies, as predicted by Crosby’s (1949) model and found in earlier studies (Curtis & Curtis, 1985)? 2) Is the transition to homostyly associated with a single or multiple molecular changes in CYPᵀ and which ones? 3) Is the loss of heterostyly accompanied by a shift from outcrossing to selfing? And, if so, is there evidence of inbreeding depression as well as increased inbreeding within and genetic differentiation among populations? Our study generates new knowledge on the molecular basis of transitions from outcrossing to selfing, its evolutionary consequences, and the putative mechanisms precluding the fixation of homostyly within species.