Population genetic consequences of hemizygosity and the transition to homosyly on S-locus genes
One of the most notable, recent discoveries on the S-locus is that it is hemizygous and present only in thrums in all systems where its genetic architecture has been investigated, including in Primula (Huuet al., 2016; Li et al., 2016), Turnera (Shoreet al. , 2019)⁠, possibly Fagopyrum (Matsui & Yasui, 2020), Linum (Gutiérrez-Valencia et al., 2022), andGelsenium (Zhao et al. , 2023), representing different families and orders of flowering plants. The hemizygosity of the S-locus should affect patterns of molecular diversity. Specifically, tight genetic linkage provided by recombination suppression and thrum-specific occurrence of S-locus genes are expected to cause a reduction of genetic diversity inside the S-locus compared to other genomic regions (Gutiérrez-Valencia et al. , 2021). Our results demonstrate that the mean πS of S-locus genes (CCMT , CYPT ,GLOT , and KFBT ; πS: 0.0012 ± 0.0006) is lower than that of their paralogs (CCM1 , CYP734A51 , GLO1 , and KFB1 ; πS: 0.0034 ± 0.0007) located elsewhere in the genome (Table 3). This result corroborates previous studies that found an overall decrease in genetic diversity between the S-locus and its flanking regions in Primula (Potente et al., 2022) andLinum (Gutiérrez-Valencia et al., 2022). Thus, our work confirms predictions that S-locus genomic architecture influences patterns of molecular evolution in S-locus genes.
The non-recombining nature of the S-locus also affects its response to natural selection. Specifically, absence of recombination is expected to reduce S-locus Ne , decreasing the efficacy of purifying selection (Gossmann et al. , 2011) on S-locus genes compared to genes outside the S-locus. Thus, increased degeneration due to accumulation of deleterious mutations is expected in these genes (Charlesworth & Charlesworth, 2000; Huu et al. , 2016). Conversely, if selection to maintain function were strong, purifying selection should be more efficient on S-locus genes than on their paralogs due to the dominant nature of the hemizygous S-locus (Gutiérrez-Valencia et al. , 2021; Potente et al., 2022). Regarding the former hypothesis, a greater accumulation of transposable elements in S-locus non-coding regions compared to the rest of the genome was detected, supporting the conclusion that purifying selection on the S-locus might be relaxed (Potente et al ., 2022). However, whether the efficacy of purifying selection differs between coding regions of S-locus genes and their paralogs remains poorly understood (Potente et al ., 2022). Our results indicate that, on average, S-locus genes exhibit higher accumulation of non-synonymous mutations than their paralogs, implying purifying selection is less effective on the former (πNS = 1.01 ± 0.37 and 0.53 ± 0.25, respectively; Table 3), conformant with predicted effects of reduced S-locus Ne . However, patterns of selective constraints within and outside the S-locus vary among gene duplicates. For example, the strength of purifying selection is similar betweenCYPT and CYP734A51, albeit slightly stronger in the former (πNS = 0.28 and 0.38, respectively). Conversely, purifying selection is less efficient in the S locus for KFBNS = 1.83 [KFBT ] and 0.23 [KFB1 ]), whereas CCM shows the opposite pattern (πNS = 0.91 [CCMT ] and 1.50 [CCM1 ]; Table 3). Taken together, the results imply that the effects of hemizygosity on purifying selection vary among P. vulgaris S-locus genes, corroborating previous results in P. veris (Potente et al ., 2022).
A key question for the genetics of distyly concerns whether the strength and nature of selection differ between S-locus genes with and without a demonstrated function in distyly. Among the three, nine, and five protein-coding genes identified in the S-locus of Gelsemium ,Linum , and Primula, respectively, (Li et al., 2016; Gutiérrez-Valencia et al. , 2022; Potente et al., 2022; Zhao et al., 2023) only two, namely CYPTand GLOT of Primula , have been functionally characterized, showing that CYPTdetermines short styles and female self-incompatibility (Huu et al., 2016, 2022), while GLOT determines high anthers in thrums (Huu et al., 2020). However, it remains unclear whether CCMT, PUMT , and KFBT play a role in Primula distyly. The markedly reduced and non-floral specific expression ofCCMT, PUMT , andKFBT compared to CYPTand GLOT in both P. vulgaris andP. veris (Cocker et al., 2018; Potente et al.,2022) cast doubt on whether the former three genes are essential for distyly. In the distylous Gelsemium elegans (Gentianales), the homolog of Primula CCMT was absent from the genome, while homologs of PUMT andKFBT were present but did not localize to the putative S-locus and were expressed in both pin and thrum flowers (Zhaoet al., 2023). Taken together, previous evidence suggests thatCCMT, PUMT , andKFBT may not be essential for the core traits of distyly (i.e., reciprocal placement of sexual organs and self-incompatibility), hence they might be under relaxed purifying selection. If this is true, one might expect thrums to exhibit higher accumulation of non-synonymous mutations inCCMT , KFBT , andPUMT than in CYPT andGLOT . Indeed, our results support this prediction, for we found weaker purifying selection onCCMT , KFBT , andPUMT NS = 0.91, 1.83, and 10.36, respectively) compared toCYPTNS = 0.28; Table 3A). It is unlikely that the results are explained by positive directional selection on advantageous non-synonymous mutations of the three genes above in thrums, because positive selection should cause rapid fixation of advantageous mutations, hence absence of polymorphism at non-synonymous sites (Hahn, 2020), which is not what we found (Table 3A). To summarize, in P. vulgaris purifying selection seems stronger on the only two S-locus genes for which a key function in distyly has been demonstrated (namely,CYPT and GLOT ) than onCCMT , KFBT , andPUMT , which were not found in the S-locus of other species and for which no differential expression between pin and thrum flowers was detected. Discovering whether the three genes above may play a role in controlling ancillary traits of distyly (e.g., pollen size and number, male incompatibility) requires additional functional studies in Primula and other distylous taxa.
Comprehensive population genetic analyses of variability in S-locus genes and their paralogs had never been performed until now, due to missing knowledge of relevant genes, unavailability of sequences from said genes, and inadequate population sampling. Here, we expanded on previous Sanger sequencing analyses of CYPT in Somerset (England) populations (Mora-Carrera et al ., 2021) by analyzing also sequences of S-locus genes and their paralogs extracted from WGR data of Slovakian, Swiss, and Turkish populations of P. vulgaris . First, homostyles, found exclusively in three Somerset populations, exhibited lower genetic diversity than thrums for both S-locus genes and their paralogs (Table 3), corroborating previous reports of reduced genetic diversity in homostyles (Husband & Barrett, 1993; Ness et al. , 2010; Yuan et al. , 2017; Zhou et al. , 2017; Zhong et al. , 2019). Second, both S-locus genes and their paralogs have markedly lower genetic variation in English populations than in other Eurasian populations of P. vulgaris(Table S2). This finding suggests a recent genetic bottleneck in English populations. This bottleneck could be associated with colonization of England following glacial retreat during the Last Glacial Maximum (ca. 10,000-12,000 years ago), as suggested for other plant species (Birks, 1989). Future genomic and demographic investigations will determine whether the signatures of genetic bottlenecks detected in S-locus genes and their paralogs apply to the entire genome, thus helping to infer the timing and mode of P. vulgaris colonization of the British Isles.
Does lower viability of S*/S*-homostyles prevent the fixation of homostyly in P. vulgaris?
Theoretical and experimental work suggests that, all else being equal, once selfing originates, the selfing phenotype should increase in frequency and eventually become fixed over time (Fisher, 1941; Lande & Schemske, 1985; Charlesworth et al. , 1990)⁠. In the transition from distyly to homostyly, Crosby’s model (1949) predicted that the rate of increase and ultimate fixation of homostyles in a population depends on whether homostyles with diploid S-locus have lower or equal viability as the other genotypes in the population (Figure 1C and D). The assumption of lower viability for S*/S*-homostyles of P. vulgarisexpanded upon evidence from crossing experiments in P.sinensis suggesting that homozygous dominant thrums had 30% lower viability than heterozygous thrums (de Winston & Mather, 1941). More recently, results of crossing experiments in a Primulahybrid (Primula x tommasinii) were interpreted as evidence of inviability for S/S-thrums (Kurian & Richards, 1997). Furthermore, population surveys of pin-to-thrum ratios in P. oreodoxaindicated that thrums were overrepresented at the seed (~1:3) but not adult stage (~1:1), implying that differences in viability could occur during the life cycle (Yuan et al. , 2018). However, genotyping of thrums was not carried out, thus preventing the determination of whether the decrease of thrums from seed to adult stage was caused by lower viability ofS /S -thrums. Our observed frequencies ofS* /0- and S*/S*- homostyles from the two trimorphic, English populations EN4-T and EN5-T of P. vulgaris are consistent with Crosby’s prediction of a recent transition to homostyly (20-30 generations) under 30-40% lower viability ofS* /S* -homostyles (Table 4), supporting the model that assumes lower fitness for S*/S*- homostyles than S*/0- homostyles (Figure 1C).
Conversely, the occurrence of a monomorphic, homostylous population ofP. vulgaris in England, first reported by Curtis and Curtis (1985) 38 years ago and recently sampled by Mora Carrera et al . (2021 and present study) is congruent with the assumption of equal viability for S*/S* homostyles. All 11 genotyped homostyles in this population (here named EN6-M) carry the S* /S*- genotype (Table 1 and Figure 3), thus EN6-M could represent a case in which homostyly increased in frequency over time and became fixed in the population by displacing pins and thrums, as predicted under the assumption of equal viability for S*/S* homostyles (Figure 1D). Alternatively, EN6-M could have been established by anS* /S*- homostyle stemming from a nearby population, thus it might have been monomorphic homostylous from the beginning. Indeed, Curtis and Curtis (1985) reported that this monomorphic population was located only about 240 m away from a trimorphic population which might have served as a source for the initial homostyle that gave origin to EN6-M. Finally, EN6-M had a very low population size (n = 19; Mora-Carrera et al., 2021) suggesting that stochasticity could have played a role in the fixation of S*/S*-homostyles in this population and that homozygosity of an S-locus with disruptedCYPT might have detrimental effects on population growth.
To summarize, our results suggest that a diploid S-locus with inactivated CYPT* may not per se be incompatible with homostyle viability. However, the occurrence of two copies of the remaining S-locus genes [i.e.,CCMT , GLOT ,KFBT , and PUMT ] in the genome of a homostyle could have detrimental effects on viability at different stages of the life cycle, possibly stemming from gene-dosage effects (Rice & McLysaght, 2017; Ascencio et al ., 2021; Liet al ., 2015; Tasdighian et al ., 2017). Future research combining S-locus genotyping and characterization of function and dosage effects of S-locus genes at different life-cycle stages with fitness measurements in the field and in greenhouse experiments is essential to address whether differences in viability prevent the widespread fixation of homostyly in P. vulgaris .