Self‐ and intra‐morph incompatibility and selection analysis of an inconspicuous distylous herb growing on the Tibetan plateau (Primula tibetica)

Abstract There is discussion over whether pollen limitation exerts selection on floral traits to increase floral display or selects for traits that promote autonomous self‐fertilization. Some studies have indicated that pollen limitation does not mediate selection on traits associated with either pollinator attraction or self‐fertilization. Primula tibetica is an inconspicuous cross‐fertilized plant that may suffer from pollen limitation. We conducted a selection analysis on P. tibetica to investigate whether pollen limitation results in selection for an increased floral display in case the evolution of autonomous self‐fertilization has been difficult for this plant. The self‐ and intra‐morph incompatibility features, the capacity for autonomous self‐fertilization, and the magnitude of pollen limitation were examined through hand‐pollination experiments. In 2016, we applied selection analysis on the flowering time, corolla width, stalk height, flower tube length, and flower number in P. tibetica by tagging 76 open‐pollinated plants and 37 hand‐pollinated plants in the field. Our results demonstrated that P. tibetica was strictly self‐ and intra‐morph incompatible. Moreover, the study population underwent severe pollen limitation during the 2016 flowering season. The selection gradients were found to be significantly positive for flowering time, flower number, and corolla width, and marginally significant for the stalk height. Pollinator‐mediated selection was found to be significant on the flower number and corolla width, and marginally significant on stalk height. Our results indicate that the increased floral display may be a vital strategy for small distylous species that have faced difficulty in evolving autonomous self‐fertilization.

Moth-pollinated plants produce white flowers that bloom at night and generate floral scents, two features that are consistent with the foraging behaviors of moths (Boberg et al., 2014;Knudsen & Tollsten, 1993). An evolutionary transition between self-and cross-fertilization may also be accompanied by the modification of floral morphological characters. Self-fertilized plants usually produce small flower and fewer rewards, whereas cross-fertilized plants preferentially produce big, showy flowers to attract pollinators. For instance, Vos, Wüest, and Conti (2014) demonstrated that the evolution of self-fertilization from cross-fertilization in Primula is accompanied by smaller and less conspicuous flowers.
The directions of the selection mediated by pollen limitation on the increase in floral displays are strikingly different from those that promote autonomous self-fertilization. This difference is because increased autonomous self-fertilization tends to weaken self-incompatibility, reduce herkogamy, and produce small flowers, whereas an increase in the floral display requires a big, showy corolla and a conspicuous inflorescence (Vos et al., 2014). However, a study by Fishman and Willis (2007) on selection in Mimulus guttatus, suggests that pollen limitation does not mediate selection on characters associated with either pollinator attraction (corolla width) or self-fertilization (herkogamy). Thus, more case studies are required to improve our understanding of how pollen limitation mediates selection on the floral traits of plants.
Angiosperm cross-fertilization relies on various agents, such as bees, birds, or bats, for the delivery of male gametes (pollen grains) from the anther to the stigmas of other individuals (Harder & Barrett, 2006), whereas autonomous self-fertilization provides reproductive assurance for the flowering plants in habitats that lack pollinators (Carlson, Gisler, & Kelso, 2008;Zhang & Li, 2008). Thus, it is widely acknowledged that pollen limitation is more commonly found in flowering plants with obligate out-crossing (Larson & Barrett, 2000).
Distyly is one of the most common floral features that promote crossfertilization in angiosperms (Barrett & Shore, 2008). The distylous syndrome is characterized by flowers in which male and female organs are reciprocally placed between long-and short-styled morphs within a population (reciprocal herkogamy); this syndrome is usually accompanied by heteromorphic self-and intra-morph incompatibility (heteromorphic SI system) (Barrett & Shore, 2008;Charlesworth & Charlesworth, 1979). Primula is a typical distylous genus that is highly cross-fertilized (Richards, 2003) and is evolving from crossfertilized distyly toward self-fertilized homostyly (Mast, Kelso, & Conti, 2006;Vos et al., 2014). However, fewer than 10% of Primula species have been found to be self-fertilized and homostylous (Richards, 2003). Primula is an ideal species to investigate the question of how pollinator-mediated selection occurs on the obligate out-crossing plant. To date, only two studies have addressed selection pressures on Primula (Ågren, Fortunel, & Ehrlén, 2006;Ågren et al., 2013;Li et al., 2017).
The evolution of self-fertilization in distylous species requires the generation of a self-fertilized homostylous variant via genetic recombination (Barrett & Shore, 2008); before this event can occur, the investment in characters that improve reproductive assurance by, for example, reducing herkogamy and weakening the SI system, would be redundant if the plant species is strictly distylous. Meanwhile, small, inconspicuous plants would probably not be attractive enough for pollinators (Grindeland & Sletvold, 2005;Peakall & Handel, 1993). Thus, it would be a good approach for the small distylous herb to increase its female fitness by increasing its floral display. We therefore raised the question of whether pollen limitation selects on the traits that increase the floral display in a case in which a small, inconspicuous plant undergoes obligate out-crossing and faces difficulty in evolving selffertilization. Primula tibetica is an inconspicuous herb distributed in the Himalayas where lacks pollinators (Duan, Zhang, & Liu, 2007;Zhao, Du, Zhou, Wang, & Ren, 2006). Thus, this species may undergo severe pollen limitation in the field, making it an ideal model system to resolve this question. Two specific questions are addressed: (1)

| Study species and site
Primula tibetica is a small perennial herb belonging to sect. Alurita of Primula (Figure 1a; Richards, 2003). The stigma of the long-styled morph rises slightly above the corolla, and the stigma of short-styled morph is located half way down the floral tube. The anther and stigma are reciprocally placed in the long-styled morph and the short-styled morph. Primula tibetica is predominantly found in Tibet, north India, Nepal, and Bhutan along the range of the Himalayas Richards, 2003). Each individual of P. tibetica generally produces 1-6 pink flowers. Each flower lasts for 8-12 days, and an individual plant flowers for 15-20 days. P. tibetica is an inconspicuous perennial herb that grows in moist meadows between 3,000 and 4,000 m.a.s.l. (Richards, 2003). Based on in-field measurements, the average height of the stalk is only 50 ± 1.23 mm. The flowering season of P. tibetica

| SI system and the capacity of autonomous selffertilization
Four treatments were included in our hand-pollination experiment: hand self-fertilization, autonomous self-fertilization, hand intra-morph pollination, and hand inter-morph pollination. Both the long-styled morph and short-styled morph were subjected to the four treatments.
In the early flowering season, 80 plants of each of the long-styled and short-styled morphs were randomly selected to be moved to a pollinator-free house. As the flower tube of P. tibetica was too narrow to be manipulated, especially for the flowers of the long-styled morph, emasculation would inevitably destroy the female organ. Thus, we were not able to emasculate the maternal flowers before handpollination. However, at the end of the experiment, we discovered that neither of the two morphs of P. tibetica could be autonomously self-fertilized.
On the first day that the maternal plant flowered, we picked out the anthers from the paternal flower using forceps and carefully touched the stigma of the maternal flower with the anthers. The stigmas of the long-styled morphs were raised slightly above the mouth of the flower tube, which allowed us to immediately pollinate the flowers of the long-styled morph. To pollinate the flowers of the short-styled morph, we carefully withdrew the corolla of the maternal flower and touched the stigma using the anther of the paternal flower.
Mature fruits were collected 30 days after hand-pollination, and the seeds of each capsule were counted. The results of the hand-pollination experiments were so evident that we did not use complicated statistical methods. The capacity for autonomous selffertilization in each morph of P. tibetica was calculated by dividing the average seed production following inter-morph pollinations by the average seed production following autonomous self-fertilizations. The index of self-incompatibility proposed by Lloyd and Schoen (1992) and further modified by Ferrero et al., (2011) was applied to infer the degree of the heteromorphic SI system of P. tibetica for both the shortstyled and long-styled morphs independently. The self-incompatibility index (SCI) was calculated as follows: The average seed production per capsule following hand inter-morph pollinations was divided by the average seed production per capsule following hand self-pollinations.
The intra-morph incompatibility index (MCI) was calculated as follows: The average seed production per capsule of hand intra-morph pollination was divided by the average seed production of the hand intermorph pollination.

| SI system and the capacity of autonomous selffertilization
The differences among pollination treatments were evident, and two of the four levels for the short-styled morph showed zero values (hand inter-morph mating, 21.13 ± 2.29; hand intra-morph mating, 0; hand selfing, 0; and autonomous self-fertilization, 0).
For the long-styled morph, one treatment produced zero seeds, and the differences among the remaining treatments were sharply contrasted (hand inter-morph mating, 24.85 ± 1.80; hand intramorph mating, 5.67 ± 1.93; hand selfing, 6.16 ± 1.74; and autonomous self-fertilization, 0) ( Figure 3). The SCI and MCI values of the short-styled morph were 0, whereas the SCI and MCI values of the long-styled morph were 0.25 and 0.23, respectively. The capacity of autonomous self-fertilization in both the long-and short-styled morphs of P. tibetica was 0.

| Phenotypic selection on P. tibetica
In the 2016 flowering season, the magnitude of pollen limitation of P. tibetica was 0.32 in the studied population. Stalk height was significantly positively correlated with other traits (Table 1) Table 3). The selection pressure on the stalk height was found to be marginally significant (p = .099) (  Figure 4a,b). The pollinatormediated selection on the stalk height was marginally significant (p = .098, Table 3 and Figure 4c).

| DISCUSSION
Primula tibetica was demonstrated to be highly self-and intra-morph incompatible and was unable to autonomously self-fertilize (Figure 3).
These results indicate that P. tibetica highly relies on pollinators for sexual reproduction. We demonstrated that P. tibetica underwent se-  (Table 3 and Figure 4). Thus, we can reasonably conclude that P. tibetica was under selective pressure to increase its floral display.

| SI system and autonomous self-fertilization
We demonstrated that P. tibetica was highly self-and intra-morph incompatible. We found that the heteromorphic SI system of P. tibetica was extremely strong, and neither the long-styled nor the short-styled morph of P. tibetica was capable of autonomous selffertilization. In many primrose species, the heteromorphic SI system is stronger in the short-styled morph than in the long-styled morph (Ernst, 1955), which may be due to the genetic background of the distylous syndrome (Barrett & Shore, 2008;Charlesworth & Charlesworth, 1979). Although the self-and intra-morph pollinated long-styled morphs could produce a small number of seeds, the capacity for autonomous self-fertilization in the long-styled morph was zero (Figure 3b). This result indicated that the autonomous self-fertilization was not only limited by the heteromorphic SI system but also strongly limited by the reciprocal herkogamy. As the self-compatible homostylous variant has been demonstrated to occur only through genetic recombination instead of being gradually selected by pollinators (Barrett & Shore, 2008), we suggest that P. tibetica will face difficulty in evolving self-fertilization by lessening the degree of herkogamy and weakening the magnitude of the SI system under the pressure of pollen limitation.

| Selection on flowering time and flower tube length
Flowering time undergoes selection due to temporal variations in pollinator and herbivore abundances (Chapurlat et al., 2015). The abundance of pollinators is generally low in early spring; thus, the flowers that open in early spring may undergo greater pollen limitation than those that open during summer (Dupont, Padrón, Olesen, & Petanidou, 2009). Our results suggest that the selection gradient for the flowering time of P. tibetica, although significant, was not significantly mediated by pollinators (Table 3). Indeed, few pollinators visited P. tibetica throughout the flowering season (Pers. Obs. 2016).
This is likely to be because P. tibetica was less attractive compared with other plants flowering synchronously in the same habitat (including Primula florinda, Iris tibetica, and several species of Pedicularis).
Furthermore, our unpublished data demonstrate that P. florinda was less pollen-limited than P. tibetica over the same period in 2016. The study population of P. tibetica was also severely damaged by the trampling or grazing of domesticated mammals (horses, yaks, and pigs) in 2016; thus, we assumed that the negative selection gradient for the flowering time of P. tibetica may be mediated by animal grazing or trampling.  The length of the flower tube is highly correlated with the proboscis length of pollinators and thus is related to pollination efficiency. We did not find significant selection pressure on the flower tube length in P. tibetica. According to our observations during the 2016 flowering season, P. tibetica was visited by both long-tongued pollinators, such as butterflies and bumble bees, and short-tongued pollinators, such as flies and syrphid flies. The most frequently found pollinator of P. tibetica was a pollen-feeding Tachinidae fly (Figure 1b). Thus, the variation in the length of the flower tube may not greatly limit the visits of pollinators.

| Selection on traits involved in the floral display
Maximizing the floral display increases the attractiveness of flowering plants to pollinators and is especially important for obligate outcrossing species. A variety of morphological traits, including the flower number, corolla size, and plant height, are related to the floral display (Chapurlat et al., 2015;. We found significantly positive selection gradients for the corolla width and flower number of P. tibetica as well as marginally significant selection pressure on the stalk height (Table 3 and Figure 4). The hand-pollinated treatment did not change the morphological traits of P. tibetica significantly ( to the net selection gradient for the flower number and explain the entire natural selection gradient for the corolla width (Table 3); these results indicate that the selection pressures on the flower number and corolla width of P. tibetica were mediated by the pollinators (Table 3, Figure 4a,b). The pollinator-meditated selection on the stalk height was marginally significant, and pollinators contributed to 99.12% of the net selection gradient for the stalk height of P. tibetica (Table 3 and Figure 4c). Thus, we assumed that the selection pressure on the stalk height of P. tibetica was probably pollinator mediated. The selection pressure on the stalk height was marginally significant, which likely resulted from an insufficient sample size. In conclusion, our results suggest that in the investigated population, the traits involved in In that study, there was a population in which the pollinators exerted negative selection on the corolla size. Parachnowitsch & Kessler (2010) suggested that significantly positive pollinator-mediated selection is found on the flower size and flower number of Penstemon digitalis, but negative pollinator-mediated selection is found to act on the plant height (though without significance).
We found that the traits involved in the floral display (corolla width, flower number, and stalk height) were selected positively by the pollinators. Our results indicated that the increase of the floral display would be especially important for P. tibetica, an inconspicuous out-crossing flowering species that grows in the Tibetan plateau.
We assumed it might be because the evolution of self-fertilization for P. tibetica would need to reduce the herkogamy and weaken the dimorphic SI system at the same time, and these two traits may be not significantly selected by the pollinators for the distylous species.
Therefore, there would be redundancy in the investment on the traits for reproductive assurance (such as reducing the herkogamy or weakening the heteromorphic SI system) as P. tibetica is difficult evolving toward self-fertilization. Meanwhile, because P. tibetica is inconspicuous compared with the co-occurring species, an increase in the floral display would greatly promote pollinator attraction. Consequently, for P. tibetica, increase in the floral display to attract pollinators would be a better strategy than selection on the traits correlated with reproductive assurance.

| CONCLUSION
During the 2016 flowering season, we found significantly positive selection gradients for traits involved in the floral display (corolla width, flower number, and stalk height) in the focal population (Table 3 and Figure 4). Although the sample size of hand-pollinated plants was relatively low, we could reasonably conclude that the traits involved in the floral display of P. tibetica were positively selected by pollinators. Our results suggested that enlarging the flower display may be vital for this inconspicuous distylous plant. This study, as one of few to analyze selection on this small, distylous herb that grows in the Tibetan plateau, provides evidence that traits involved in the floral display are especially important for such small cross-fertilized alpine plants.