Population genetics of self-incompatibility in a clade of relict cliff-dwelling plant species

OPEN ACCESS – RESEARCH ARTICLE Population genetics of self-incompatibility in a clade of relict cliff-dwelling plant species Jose L. Silva 1, *, Adrian C. Brennan 2 and José A. Mejías 1 1 Departamento de Biología Vegetal y Ecología, Universidad de Sevilla, CP 41012, Sevilla, España. 2 School of Biological and Biomedical Sciences, University of Durham, Durham, UK. *Corresponding author Corresponding author’s e-mail address: [email protected]. Running head: Self-incompatibility in relict Sonchus species. 1 Published by Oxford University Press on behalf of the Annals of Botany Company. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. AoB PLANTS Advance Access published May 6, 2016 at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from RESEARCH ARTICLE Population genetics of self-incompatibility in a clade of relict cliff-dwelling plant species ABSTRACT The mating systems of species in small or fragmented populations impact upon their persistence. Small self-incompatible (SI) populations risk losing S allele diversity, responsible for the SI response, by drift thereby limiting mate availability and leading to population decline or SI system breakdown. But populations of relict and/or endemic species have resisted these demographic conditions over long periods suggesting their mating systems have adapted. To address a lack of empirical data on this topic, we studied the SI systems of three relict cliff-dwelling species of Sonchus section Pustulati (Asteraceae): S. masguindalii, S. fragilis and S. pustulatus in the western Mediterranean region. We performed controlled pollinations within and between individuals to measure index of SI (ISI) expression and identify S alleles in multiple population samples. Sonchus masguindalii and S. pustulatus showed strong SI (ISI = 0.6–1.0) compared to S. fragilis (ISI = 0.1–0.7). Just five S alleles were estimated for Spanish S. pustulatus and a moderate 11-15 S alleles for Moroccan S. pustulatus and S. fragilis, respectively. The fact that autonomous fruit set was generally improved by active self-pollination in self-compatible S. fragilis suggests that individuals with weak SI can show a wide range of outcrossing levels dependent on the degree of self or outcross pollen that pollinators bear. We conclude that frequent S allele dominance interactions that mask the at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 2 incompatibility interactions of recessive S alleles leading to higher mate availability and partial breakdown of SI leading to mixed mating, both contribute to reproductive resilience in this group. Keywords: Breakdown of SI; diallel crosses; index of self-incompatibility; S allele diversity; Sonchus section Pustulati; sporophytic self-incompatibility; resilience. at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 3 INTRODUCTION Self-incompatibility (SI) is a genetic barrier to inbreeding that is broadly distributed among hermaphroditic angiosperms (Busch and Schoen 2008). Two major types of self-incompatibility are recognized based on the genetic control of the incompatibility reaction: gametophytic (GSI) and sporophytic (SSI) systems. In both SI types, the incompatibility reactions are controlled by a linked cluster of genes collectively known as the ‘S locus’ (‘S’ denotes self-sterility), and individual plants that share alleles at this locus do not produce offspring in cross-pollinations (Richards 1997). In GSI, the more common type of SI, the incompatibility S allele phenotype is governed by the genotype of each single haploid pollen grain. In SSI, which has so far been detected in seven dicotyledonous families (Asteraceae, Betulaceae, Brassicaceae, Caryophyllaceae, Convolvulaceae, Polemoniaceae and Sterculiaceae; Hiscock and Kües 1999), the pollen incompatibility phenotype is genetically controlled by the plant that produces the pollen grains (i.e. the sporophyte) through the diploid pollen grain coat. In both cases, if the pollen of the donor plant is recognized by the receptor plant as own pollen (i.e. they share at least one allele, codominantly or dominantly expressed in the case of SSI), the receptor triggers the SI reaction. The molecular basis of SSI is best understood in Brassicaceae, where the maternal receptor and paternal ligand are known, and while the molecular basis of SSI in other families including Asteraceae seems to be distinct, a similar receptor-ligand “lock-and-key” model of SSI is generally assumed (Allen et al. 2011). at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 4 Regardless of the type of SI, reproductive success in populations of SI species depends on the number and frequency of alleles at the S locus (Busch and Schoen 2008) and, in turn, the number of S alleles depends on the longterm population size and the mutation rate of the S locus (Wright 1939; Busch et al. 2014). S allele diversity may therefore have important consequences for population biology, and in large populations many S alleles are maintained by negative-frequency-dependent selection (Wright 1939; Schierup et al. 1997; Lawrence 2000). Meanwhile, in small or highly fragmented populations S allele diversity can be lost due to genetic drift (Wagenius et al. 2007). In this case, SI species can suffer from limited reproduction through scarcity of compatible mates (S Allee effect) or strong inbreeding depression following breakdown of SI (e.g. Byers and Meagher 1992; Willi and Fisher 2005; Glémin et al. 2005; Wagenius et al. 2007; Caujapé-Castell et al. 2008; Young and Pickup 2010; Leducq et al. 2010), which can lead to the extinction of rare or endemic taxa (Demauro 1993; Reinartz and Les 1994; Caujapé-Castells et al. 2008). The demographic consequences of the S Allee effect in SSI systems may be especially important in populations with fewer than 10 S alleles (Busch et al. 2014), whose spatial distribution within the population can also influence local reproductive success. These reproductive problems may be aggravated for insect pollinated species due to associated declines in specialist pollinators or limited pollinator attractiveness when rare (Rymer et al. 2005). Recovery of reproductive assurance in SI plants subjected to strong mate limitation in declining or new populations is usually expected to involve selection for self-compatibility (SC) (Reinartz and Les 1994; Ortiz et al. 2006; Wagenius et al. 2007) that can be detected from low index of incompatibility (ISI) at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 5 measures (Lloyd 1965). Nevertheless, selection for SC does not always occur as has been shown in cases of maintenance of SI through recent colonization events (Carr and Powel 1986; Kim et al. 1999; Brennan et al. 2002, 2003; Miller et al. 2008). In some cases, alternative solutions to increasing mate availability have been found such as increasing dominance interactions among remaining S alleles to increase cross-compatibility in SSI systems (Brennan et al. 2002). Increased S allele dominance leads to increased cross-compatibility because more recessively expressed S alleles are masked, resulting in more compatible crosses between individuals that share these recessive S alleles. Alternatively, SC alone might not be sufficient to increase self-fertilization, if the plant species still requires pollinators for effective pollen transfer from anthers to stigmas (Gandhi et al. 2005). Moreover, between the two ends of the mating system spectrum (SI and SC), sexual systems of intermediate nature that seem to combine the advantages of both selfing and outcrossing, i.e. the partial selfincompatibility (pSI) systems, have frequently been reported (Levin 1996, Igic and Busch 2013). Studies of populations undergoing stressful conditions due to major demographic changes (fragmentation, decline, founder events or colonization) have been invaluable to our understanding of the factors shaping mating system evolution (e.g. Willi and Fisher 2005; Nielsen et al. 2007; Wagenius et al. 2007; Busch et al. 2010; Young and Pickup 2010; Leducq et al. 2010). However, not all species with small and/or fragmented populations are in the process of major recent demographic upheaval (Mable and Adam 2007; Hoebe et al. 2009). Many endemic species with narrow environmental requirements are relicts, i.e. have stably persisted for long periods maintaining small and at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 6 fragmented distribution ranges which often reflect relatively old vicariance events. This makes them potentially very interesting models to understand mating system evolution in small isolated populations over long sustained periods. In addition, these species also make important contributions to the biodiversity of many regions and are increasingly of conservation concern due to habitat loss (Thompson 2005). Mating system evolution and particularly the maintenance of SI systems have hardly been studied in these species to date. Sonchus pustulatus, S. fragilis and S. masguindalii constitute the wellsupported clade Sonchus section Pustulati (Asteraceae, Cichorieae; Boulos 1973; Kim et al. 2007; Silva 2015a). These species are narrow endemics restricted to both sides of the western Mediterranean Basin (Spain and Morocco) occurring on localized cliffs at low altitude (Fig. 1, Silva et al. 2015b). All three species are found in North Africa, where they are considered to be very rare (Fennane and Ibn Tatoo 1998). S. pustulatus also occurs in the SE Iberian Peninsula, where it is categorized as “critically endangered” (Cueto et al. 2003; Silva et al. 2015b). Phylogenetic and phylogeographic data suggest that these taxa are relicts of the late Tertiary (Silva et al. 2015a). Their restricted and disjunct distribution seems to be related to old geological events of large biogeographic impact: the Messinian Salinity Crisis and the subsequent Zanclean reflooding of the Mediterranean Basin 5.96 to 3.60 million years ago (Krijgsman et al. 1999; Rosenbaum et al. 2002; Fauquette et al. 2006), and the subsequent establishment of the Mediterranean climate (3.2–2.8 Ma; Suc, 1984). These relict and endemic taxa seem to have suffered from erosion of genetic diversity by genetic drift during long periods of small population size that is reflected in previous observations of relatively low genetic diversity at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 7 (HE=0.088–0.177; see Silva et al. 2015a). Therefore, it is also of interest to investigate their SI mating systems. These species are pollinated by generalist insects (Silva 2014) and preliminary hand pollinations suggested that S. pustulatus in Spain is SI (Mejías 1992). We predict that strong long-term genetic drift could have led to limited S allele diversity, capacity of selfing or frequent dominance interactions among remaining S alleles. Here we report a detailed population genetic study of the mating system in the species of Sonchus section Pustulati based on an intensive program of hand pollinations. Though in recent decades the development of biochemical and molecular techniques for SSI in Brassica (e.g. Gaude et al. 1991; Brace et al. 1993) has allowed a faster and less tedious S allele identification than using classical diallel crosses (Glémin et al. 2005), the molecular mechanism of SSI in the Brassicaceae is not shared by the Asteraceae (Allen et al. 2011; Gounthier et al. 2013). Fortunately, a recent study with Senecio species showed how extensive controlled crossing surveys can be applied to investigate the SSI systems in species of particular ecological or evolutionary interest (Brennan et al. 2013). Based on this experimental approach the main goals of this study were to investigate across multiple relict Sonchus populations and species: (1) variation in the strength of SSI within and among populations of these relict species (2) dependency on insects for successful pollination (3) S allele diversity and (4) S allele dominance interactions. We interpret and discuss our results in terms of mating system responses to the past and present demographic conditions faced by these species. METHODS at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 8 Plant species The species Sonchus pustulatus, S. fragilis and S. masguindalii constitute the section Pustulati of the subgenus Sonchus and are all diploid (n=9, 2n=18; Mejías and Andrés 2004; Vogt and Oberprieler 2008). They are perennial suffrutescent chamaephytic cliff plants, usually procumbent, which develop moderate-long branches, particularly lignified in S. masguindalii (Silva et al. 2015b). Depending on the species, individuals range approximately from only 5 cm2 to 200 cm2 in surface and can reach up to 45 cm in height, with intricate branching. They commonly attain sexual maturity during the first year of life (Silva et al. 2015b) and mainly flower in spring. Flower heads are terminal, solitary or in groups from two to four, with bright yellow ligule florets. All florets are hermaphrodites and protandrous, which develop centripetally over a period of 3–6 days. Flower heads comprise 30–120 florets in the case of S. pustulatus and S. fragilis, and 60–250 florets in S. masguindalii. Fruits are achenes with a short-lasting pappus, released during late spring and early summer. Currently, there are only 19 known populations of these species, which are located in four small non-overlapped distribution areas restricted to the Baetic-Rifan geological complex in the western Mediterranean Basin (Silva et al. 2015b). Plant sampling and pollinations In May and June 2008, we collected and georeferenced 281 plants (128 as cuttings and 153 as flower heads with mature seeds) from 12 populations (Fig. 1; Table 1). We collected at a minimum distance of 20 m between plants to at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 15 higher in S. fragilis than in the remaining taxa (Χ2 = 21.77, DF = 3, P <0.001; U >321.00, P <0.022). Surprisingly, some of these plants (10%), particularly of S. fragilis (6%), showed even a higher fructification after ASP than after FSP. S allele diversity estimation Our three complete diallels for populations ANT, TAL and GHO1 of Spanish and Moroccan ranges of S. pustulatus and S. fragilis allowed the identification of five, eight and eight incompatibility groups, respectively (fig. 5a- c). These were interpreted as the numbers of different expressed S alleles in each population sample (fig. 5a-c and [see SUPPORTING INFORMATION Tables S3 and S4]). It is worth noting that as the number of individuals used in the crosses was 14, 14 and 11, respectively, the number of potential alleles were 28, 28 and 22. We numbered the S alleles found from S1 to S21, although individuals from different populations may share some of these alleles among them. From our diallel results we interpreted S alleles to be expressed dominantly in those individuals belonging to a single incompatibility group. A SSI model of complete dominance among alleles satisfactorily explained the majority of the crossing results in each population (94.3–99.1% of results supported; [see SUPPORTING INFORMATION Table S3]). Furthermore, assuming co-dominance of alleles S1 and S2 in the pollen of plant 8 from ANT, and co-dominance of alleles S6 and S9 in the stigma of plant 1 from TAL [see SUPPORTING INFORMATION Figure S5], the SSI model explained a further 2.8% and 1.1% of the crossing results for each of these populations, respectively [see SUPPORTING INFORMATION Table S3]. Very few crossing anomalies, i.e. those not fitted into the SSI model, were detected across the at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 16 diallels (0.9–4.6%; [see SUPPORTING INFORMATION Table S3]). The S allele number estimator of Brennan et al. (2002) predicted approximately 5, 11, and 15 S distinct alleles for the entire populations ANT, TAL and GHO1, respectively (table 2). Repeatability values were 0.88, 0.77 and 0.70 for the three populations, respectively (table 2), suggesting that sampling has been sufficient to capture most of the S allele diversity present in these populations. The relatively high numbers of identified S alleles and small sample sizes preclude a formal statistical analysis of S allele spatial structure. Visual inspection suggested that the S alleles identified did not appear to show a strong spatial distribution structure within the populations apart from the population ANT, where the individuals of each of the incompatibility groups most represented in the diallels (those assigned with the S1 and S3 alleles; fig. 5a and SUPPORTING INFORMATION Table S4) were mainly distributed in each one of the two clusters of individuals forming the population [see SUPPORTING INFORMATION Figure S6]. DISCUSSION Strength of expression of SI across Sonchus section Pustulati The generally low fruit set observed after self-pollination (particularly autonomous pollination), relative to the large amounts of fruits for outcrossed treatments across individuals, indicate that Sonchus pustulatus and S. masguindalii have a relatively strong SI mechanism. This is not the case for S. fragilis, which according to our results has a weak SI system. Nevertheless, virtually all populations of S. pustulatus and S. masguindalii showed low at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 17 proportions of SC and pSI individuals, which indicates that some flexibility of expression of SI is always present in the section. According to the ISI criteria used in the present study, the three Sonchus species of the section Pustulati would be therefore included in the long list of Asteraceae taxa that have been considered to bear a not-always-100% efficient SSI system (i.e. SI-pSI species; e.g. Brauner and Gottlieb 1987; Reinartz and Less 1994; Young et al. 2000; Nielsen et al. 2003; Brennan et al. 2005; Ortiz et al. 2006; Lafuma and Maurice 2007; Scheffknecht et al. 2007; Ferrer et al. 2009). Historical events involving major demographical constraints could have favoured the partial breakdown of SI (Mable and Adam 2007; Hoebe et al. 2009). A series of demographic, ecological and genetic features of these species described in Silva (2014) and Silva et al. (2015a, b) suggest an ancient origin and diversification, a relict condition, and high ecological resilience despite a narrow ecological amplitude within a very restricted geographical distribution of the section Pustulati. The persistence of these populations could have been reinforced through the partial or complete breakdown of SI, decreasing therefore the S Allee threshold, that is, the number of S alleles under which a population shows a decline through scarcity of potential mates (Wagenius et al. 2007). However, strongly expressed SSI was still the rule for the majority of individuals examined (Fig. 2, table 1), indicating that it is largely still an advantageous mating strategy. Selection to maintain strongly expressed SSI might be important for these long-lived perennial species where recessive deleterious mutations arising from mitotic mutation probably maintain high at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 18 inbreeding depression (Morgan 2001). Studies of other cliff-dwelling species have found that they are frequently long-lived and persistent and have relatively stable population size (Larson et al. 2000; Picó and Riba 2002; García et al. 2003; Lavergne et al. 2004; Thompson 2005; Silva et al. 2015b). This relative demographic stability might buffer against periods when the S Allee threshold of limited S allele numbers is passed and breakdown of SI would otherwise be selected for. Implications of highly pollinator-dependent self-fertilization for mixed mating systems Differences in seed production between the ASP and FSP treatments were especially high in plants of S. fragilis (fig. 4). Such a difference indicates that, despite of the capacity of self-fertilization, this species shows limited autonomous self-pollination. Therefore, SC and pSI plants likely need pollinators both to be either outcrossed or self-pollinated, and hence the mating system (from inbreeding to outcrossing) in the SC and pSI populations will primarily depend on the pollen load of pollinators and, secondly, on which type of pollen (self or non-self) has higher fertilization success [see SUPPORTING INFORMATION Figure S7]. Cross pollen appears to have a fertilization advantage in these mixed mating populations as evidenced by the higher values of fruit set detected after the crossed pollinations than in the forced selfpollinations. In the Asteraceae, SC taxa tend to constitute mixed mating systems rather than fully inbreeding populations (e.g. Ellstrand et al. 1978; Sun and Ganders 1988). For example, additional adaptations to promote crosspollination such as attractive blooms (as for these Sonchus species) can lead to at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 19 mixed mating outcomes for otherwise SC taxa. The use of molecular techniques would be necessary to determine the actual rates of self- and cross-fertilization in the SC and pSI populations of Sonchus section Pustulati. Another possible explanation for low ASP fruit set, also discussed later in terms of S allele diversity, is that the emergence of SC and pSI might be relatively recent in S. fragilis and complementary adaptations to increase autonomous pollination have not yet had time to evolve. SC alleles might nonetheless persist in these populations due to their automatic transmission advantage in any selfed progeny that are produced (Stone et al. 2014). S allele diversity The results and interpretations from our three diallels support the presence of a sporophytic genetic control of self-incompatibility in the species of Sonchus section Pustulati. The high frequency of reciprocally compatible or incompatible inter-individual crossing results could be explained by S allele dominance interactions occurring both in pollen and stigma. Similarly, nonreciprocal compatibility or single incompatibility group crossing results could be explained by frequent S allele dominance interactions. The crossing results could thus be fitted to a sporophytic incompatibility model, and five to eight incompatibility groups were identified in each of the three diallels supporting the multiallelic nature of the S locus (e.g. Brennan 2002, 2006 and 2013; Young and Pickup 2010). It is of interest that the S. fragilis population was estimated to have the highest S allele number (15 versus 5 to 11 for S. pustulatus), despite showing higher levels of pSI. However, it is worth noting that other features of this at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 20 population support the observation of higher S allele diversity. The GHO1 population is far larger than the S. pustulatus populations based on direct counts of flowering individuals and it has also maintained similar genetic diversity to these populations (table 2). Therefore, the high S allele diversity suggest that the breakdown of obligate SI might be recent so that selfing has not gone on long enough to cause reductions in S allele number or heterozygosity. The lower S allele diversity identified for the Spanish ANT population of S. pustulatus (5) compared to Moroccan populations (11-15) probably reflects greater long-term isolation on the Spanish side of the Alboran Sea compared to larger Moroccan populations of S. pustulatus that occur in relative proximity to S. fragilis and S. masguindalii. These related species might contribute to each other’s S allele diversity because negative frequency dependent selection favours the introduction of new S alleles between species even when migration and hybridization might be rare (Castric et al. 2008; Brennan et al. 2013). For example, Spanish and Moroccan individuals of S. pustulatus and S. fragilis seem to show high levels of interfertility (Silva, Mejías and Mendoza unpublished data) but levels of interspecific sharing of S alleles have not yet been studied for these species. The number of S alleles estimated by means of diallel crosses in three populations of two species from this group was relatively low (5-15) compared to the number of S alleles estimated in other species with SSI (range 2.1–54; mean 16.8; Busch et al. 2014). This number of S alleles was similar to another Asteraceae species so far studied, Senecio squalidus (7–11; Brennan et al. at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 21 2006), which also retained SSI despite a strong colonization bottleneck. A range of 4 to 22 S alleles were observed in samples of a second Asteraceae species, Rutidosis leptorrhynchoides, that were correlated with their population sizes spanning from 5 to 100,000 individuals (Young and Pickup 2010). Similarly to S. squalidus, dominance interactions among S alleles were common in the three Sonchus populations sampled by means of diallels. Dominance interactions are thought to lead to a higher number of compatible crosses within a population than in the case of common co-dominant interactions among S alleles, so increase in dominance is predicted to be an evolutionary response to reduced S allele frequency (Byers and Meagher 1992; Brennan et al. 2003; Hiscock and Tabah 2003). The high levels of S allele dominance interactions observed for these species would further contribute to the resilience of their SI systems because more recessive S alleles are masked, thereby increasing population mate availability (Brennan et al. 2003). The evidence of stigma dominance interactions detected in the present study is consistent with the hypothesis that fecundity selection (i.e. reduced mate availability affecting female fitness) has been important in these populations (Busch and Schoen 2009; Llaurens et al. 2009). It would be interesting to see how widely this SSI feature of frequent S allele dominance interactions might also apply to other narrow endemic taxa within the large Asteraceae family and other SSI families. Implications of SSI for the current negative demographic trends of Spanish S. pustulatus at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 22 The presence of pSI and SC plants and the dominance interactions among S alleles increasing the S Allee threshold may have been crucial for the relative demographically stability of the Spanish population ANT of S. pustulatus (Silva et al. 2015b), for which just five S alleles were estimated. However, in the case of the other Spanish populations CAB and TEL, which have been found to be in decline based on population viability analysis, these compensatory strategies do not now seem to be sufficient (see wild fruit set values table 2, Silva et al. 2015b). The two populations (CAB and TEL) could have surpassed the S Allee threshold of adequate S allele numbers for reproductive success (Wagenius et al. 2007), although other factors such as low recruitment, increasing human disturbance or drought have also been implicated to be contributing to the population decline (Silva et al. 2015b). Any further slight decreases in the fruit set, due for instance to mate limitation, would further negatively affect the stability of these small populations. All three Spanish populations of S. pustulatus showed signals of possible inbreeding depression in comparison with the Moroccan populations of the group for multiple demographical, reproductive and genetic features (table 2; Silva 2014; Silva et al. 2015a, b). More empirical data about the level of inbreeding depression is required to better understand how SSI systems affect populations under threat of extinction as well as in the first stages of recolonization after strong bottleneck events (Winn et al. 2011). As habitat areas decrease and fragmentation increases, it becomes increasingly important to consider breeding systems when designing conservation plans (Wagenius et al. 2007). Here, we suggest that it would be at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 23 beneficial to carry out a genetic rescue among the genetically impoverished Spanish populations of S. pustulatus (Silva et al. 2015a); at least from the most demographically stable and genetically isolated population ANT to the declining ones, TEL and CAB (Silva et al. 2015a, b). In a metapopulation where individual populations have lost S allele diversity, interpopulation crosses should more often alleviate mate limitation compared to crosses between plants from the same population (Busch and Schoen 2008; Young and Pickup 2010). As gene flow does not appear to now occur among the three Spanish populations (Silva et al. 2015a), it would therefore be useful for their conservation to perform interpopulation crosses in the field (e.g. Paschke et al. 2002; Willi and Fischer 2005; Pickup and Young 2008) or, due to low observed recruitment rates (Silva et al. 2015b), attempt the use of interpopulation cuttings transplants. CONCLUSIONS This study highlights the value of performing detailed mating system studies in plant species of high conservation value, such as the rare endemic and relict species of Sonchus section Pustulati described here. This study adds to the evidence that outcrossing mating systems based on SSI are highly resilient even under long-term conditions of small, fragmented, and isolated populations, possibly due to mating system flexibility with the presence of some selfing and the fact that high cross-compatibility is achieved for relatively modest dominantly expressed S allele polymorphism. We also find evidence for a mating system shift from SSI to facultative SC in some populations of S. fragilis suggesting the value of relict species as a system for studying mating system evolution. Finally, as additional anthropogenic and climate changes are at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 24 placing additional pressure on vulnerable endemic and relict species, such as the Spanish populations of S. pustulatus, it will become increasingly important to take mating system factors into account as part of conservation efforts. SUPPORTING INFORMATION Figure S1. Fertilized and unfertilized achenes. Figure S2. Fruit (achene) set variation across the species of Sonchus section Pustulati according to the different types of pollinations applied over sampled individuals. Table S3. Summary statistics for cross-classifications and the explanatory power of the sporophytic self-incompatibility model of near-complete S allele dominance interactions for the analysed diallels of cross results in the sampled populations. Table S4. S allele designations for the individuals in the population samples. Figure S5. Different dominance and co-dominance relationships among S alleles in the stigmas and the pollen grains of some individuals inferred from results of diallel crosses. Figure S6. Spatial distribution of sampled incompatibility groups in the populations. Figure S7. Pollinator dependency for outcrossing levels in SC and pSI plants. SOURCES OF FUNDING at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 31 Mejías JA, Andrés C. 2004. Karyological studies in Iberian Sonchus 712 (Asteraceae: Lactuceae): S. oleraceus, S. microcephalus and S. asper 713 and a general discussion. Folia Geobotanica 39:275–291. 714 Miller JS, Levin RA, Feliciano NM. 2008. A tale of two continents: Baker's rule 715 and the maintenance of self‐incompatibility in Lycium (Solanaceae). 716 Evolution 62:1052–1065. 717 Morgan MT. 2001. Consequences of life history for inbreeding depression and 718 mating system evolution in plants. Proceedings of the Royal Society of 719 London B: Biological Sciences 268:1817–1824.Nielsen LR, Siegismund 720 HR, Hansen T. 2007. Inbreeding depression in the partially self-721 incompatible endemic plant species Scalesia affinis (Asteraceae) from 722 Galápagos islands. Evolutionary Ecology 21:1–12. 723 Nielsen LR, Siegismund HR, Philipp M. 2003. Partial self-incompatibility in the 724 polyploid endemic species Scalesia affinis (Asteraceae) from the 725 Galápagos: remnants of a self-incompatibility system? Botanical Journal 726 of Linnaean Society 142:93–101. 727 Ortiz MA, Talavera S, Castaño JL, Tremetsberger K, Stuessy T, Balao F, 728 Casimiro-Soriguer R (2006) Self-incompatibility and floral parameters in 729 Hypochaeris sect. Hypochaeris (Asteraceae). American Journal of 730 Botany 93:234–244. 731 Paschke M, Abs C, Schmid B. 2002. Effects of population size and pollen 732 diversity on reproductive success and offspring size in the narrow 733 endemic Cochlearia bavarica (Brassicaceae). American Journal of 734 Botany 89:1250–1259. 735 Paxman GJ. 1963. The maximum likelihood estimation of the number of self-736 sterility alleles in a population. Genetics 48:1029–1032. 737 Pickup M, Young AG. 2008. Population size, self-incompatibility and genetic 738 rescue in diploid and tetraploid races of Rutidosis leptorrhynchoides 739 (Asteraceae). Heredity 100:268–274. 740 at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 32 Picó FX, Riba M. 2002. Regional-scale demography of Ramonda myconi: 741 Remnant population dynamics in a preglacial relict species. Plant 742 Ecology 161:1–13. 743 Raduski AR, Haney EB, Igic B. 2011. The expression of self-incompatibility in 744 angiosperms is bimodal. Evolution 66:1275–1283. doi:10.1111/j.1558-745 5646.2011.01505.x. 746 Reinartz JA, Les DH. 1994. Bottleneck-induced dissolution of self-747 incompatibility and breeding system consequences in Aster furcatus 748 (Asteraceae). American Journal of Botany 84:446–455. 749 Richards AJ. 1997. Plant Breeding Systems. Chapman and Hall. Garland 750 Science. 751 Rosenbaum G, Lister GS, Duboz C. 2002. Relative motions of Africa, Iberia and 752 Europe during Alpine orogeny. Tectonophysics 359:117–129. 753 Rymer PD, Whelan RJ, Ayre DJ, Weston PH, Russell KG. 2005. Reproductive 754 success and pollinator effectiveness differ in common and rare Persoonia 755 species (Proteaceae). Biological Conservation 123:521–532. 756 Scheffknecht S, Dullinger S, Grabherr G, Hülber K. 2007. Mating 757 systems of snowbed plant species of the northeastern Calcareous Alps 758 of Austria. Acta Oecologica 31:203–209. 759 Schierup MH, Vekemans X, Christiansen FB. 1997. Evolutionary dynamics of 760 sporophytic self-incompatibility alleles in plants. Genetics 147:835–846. 761 Schoen DJ, Busch JW. 2009. The evolution of dominance in sporophytic self‐762 incompatibility systems. II. Mate availability and recombination. Evolution 763 63:2099–2113.Silva JL. 2014. Biology and conservation of Sonchus sect. 764 Pustulati (Asteraceae): cliff-dwelling species endemic to the Baetic-Rifan 765 complex (Western Mediterranean). PhD Thesis, University of Seville, 766 Spain. 767 Silva JL, Lim S-Y, Kim S-C, JA Mejías. 2015a. Phylogeography of a cliff-768 dwelling relicts with a highly narrow and disjunct distribution in the 769 western Mediterranean. American Journal of Botany 102:1538–1551. 770 at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 33 Silva JL, Mejías JA, García MB. 2015b. Demographic vulnerability in cliff-771 dwelling Sonchus species endemic to the western Mediterranean. Basic 772 and Applied Ecology 16:316–324. doi.org/10.1016/j.baae.2015.02.009. 773 Stevens JP, Kay QON. 1989. The number, dominance relationships and 774 frequencies of self-incompatibility alleles in a natural population of 775 Sinapis arvensis L. in South Wales. Heredity 62:199–205. 776 Stone JL, VanWyk EJ, Hale JR. 2014. Transmission advantage favours selfing 777 allele in experimental populations of self-incompatible Witheringia 778 solanacea (Solanaceae). Evolution 68:1845-1855. 779 Suc JP. 1984. Origin and evolution of the Mediterranean vegetation and climate 780 in Europe. Nature 307:429–432. 781 Sun MY, Ganders FR. 1988. Mixed mating systems in Hawaiian Bidens 782 (Asteraceae). Evolution 42:516–527. 783 Thompson JD. 2005. Plant Evolution in the Mediterranean. Oxford University 784 Press, USA, 306 pp. 785 Vogt R, Oberprieler C. 2008. Chromosome numbers of North African 786 phanerogams. VIII. More counts in Compositae. Willdenowia 38:497–787 519. 788 Wagenius S, Lonsdorf E, Neuhauser C. 2007. Patch aging and the S‐allee 789 effect: breeding system effects on the demographic response of plants to 790 habitat fragmentation. The American Naturalist 169:383–397. 791 Willi Y, Fischer M. 2005. Genetic rescue in interconnected populations of small 792 and large size of the self-incompatible Ranunculus reptans. Heredity 793 95:437–443. 794 Winn AA, Elle E, Kalisz S, Cheptou P-O, Eckert ChG, Goodwillie C, Johnston 795 MO, Moeller DA, Ree RH, Sargent RD, Vallejo-Marín M. 2011. Analysis 796 of inbreeding depression in mixed-mating plants provides evidence for 797 selecting interference and stable mixed mating. Evolution 65:3339–3359. 798 at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 34 Wright S. 1939. The distribution of self-sterility alleles in natural populations. 799 Genetics 24:538–552. 800 Young AG, Pickup M. 2010. Low S‐allele numbers limit mate availability, reduce 801 seed set and skew fitness in small populations of a self‐incompatible 802 plant. Journal of Applied Ecology 47:541–548. 803 Young AG, Miller C, Gregory E, Langston A. 2000. Sporophytic self-804 incompatibility in diploid and tetraploid races of Rutidosis 805 leptorrhynchoides (Asteraceae). Australian Journal of Botany 48:667–806 672. 807 808 at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 35 FIGURE LEGENDS 809 Figure 1. Distribution range of the species of Sonchus section Pustulati, 810 locations of all known populations, and area of occupancy (Silva et al. 2015b). 811 Sampled populations are in bold. 812 Figure 2. Strength and variation of the self-incompatibility across populations of 813 the species of Sonchus section Pustulati assessed by the index of self-814 incompatibility (ISI; Lloyd 1965; Raduski et al. 2011). Individuals were classified 815 according to the ISI states: self-compatible (SC, ISI < 0.2), partial self-816 compatible (pSI, 0.2 ≥ ISI < 0.8) and self-incompatible (SI, ISI ≥ 0.8). Numbers 817 above bars are sampled individuals used to calculate frequencies of ISI states. 818 Figure 3. Relationships between the mean index of self-incompatibility (Mean 819 ISI; Lloyd 1965; Raduski et al. 2011) across populations of the species of 820 Sonchus section Pustulati and a) effective population size, and b) population 821 plant density. 822 Figure 4. Variation of the individual difference on fruit set obtained after forced 823 (FSP) and autonomous (ASP) self-pollinations across the Sonchus section 824 Pustulati taxa. Small circles and asterisks indicate atypical and extreme values, 825 respectively; i.e. outliers that are more than 1.5 and 3 box lengths from the 826 upper hinge (75th percentile), respectively. Only plants that produced fruits after 827 ASP and/or FSP pollinations were included (sample size = N). 828 Figure 5. Analysed diallel of cross results between pairs of individuals from the 829 Spanish population ANT of Sonchus pustulatus (a), the Moroccan population 830 TAL of S. pustulatus (b) and the population GHO1 of S. fragilis (c). The first row 831 at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from 36 and column refer to the dominantly or co-dominantly-expressed S alleles that 832 have been identified. Codes of individuals acting as maternal or paternal are in 833 the second column and second row, respectively. Symbols indicate 834 compatibility reaction of crosses involving each pair of individuals according to 835 the fruit set (see Methods): compatible (+ and white shading), incompatible (- 836 and coloured or dark grey shading) and indeterminate (+/- and pale grey 837 shading). No symbol indicates missing data. The last column refers to the 838 compatibility reaction obtained after autonomous self-pollinations. 839 840 at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from at UNIVERSIDAD DE SEVILLA on December 14, 2016http://aobpla.oxfordjournals.org/Downloaded from