Spatial variation of pollen receipt and effects of heterospecific pollen on seed set in Salvia przewalskii

Abstract Generalized pollinators visit multiple co‐flowering plant species and may transfer heterospecific pollen grains. Recent studies have indicated that the effect of heterospecific pollen (HP) on reproduction success is variable and depends on the identity of donor and recipient species. However, few studies have documented variation in HP receipt and evaluated the reproductive effects of HP receipt across geographic locations under natural conditions. We investigated the spatial variation of pollen deposition across eight sites and how the pollen receipt related to the seed set of Salvia przewalskii, a subalpine perennial herb in Hengduan Mountain in southwest China. We found that stigmatic pollen loads substantially varied among sites for several metrics, including quantities of conspecific and heterospecific pollen, the proportion of HP, and species composition of HP donors. Five different plant families were the most common HP source at one or two sites, and the proportion of HP ranged from 3.4% to 51.3% across sites. The association of conspecific pollen with seed set was positive and variable among sites, whereas the association of HP receipt and seed set was negative and not significantly different among sites. Our results demonstrate variation in the quantity and fitness effect of pollen receipt across sites, which is a precondition for evolution of local adaptation. Further study of variation in patterns and effects of HP receipt for the same recipient species across natural communities would allow better understanding of the ecological and evolutionary consequences of HP receipt.

occluding the stigmatic surface (Caruso & Alfaro, 2000), or for closely related species by usurping ovules that subsequently abort (Burgess et al., 2008). Effects of HP on plant fitness have been found to vary from being negative to neutral or even positive, with effects depending on the identity of donor species, and with larger heterospecific pollen loads generally causing greater impacts (Ashman & Arceo-Gómez, 2013;Morales & Traveset, 2008). Moreover, the magnitude of HP's effect on seed production is influenced by the interactions of donor and recipient plant species (Arceo-Gómez et al., 2019;Lanuza et al., 2021). Negative fitness effects of HP deposition could result in selection of a variety of floral and reproductive traits to reduce pollen sharing (Moreira-Hernández & Muchhala, 2019). For example, there could be selection to modify flowering phenology to reduce overlap with HP donors (Waser, 1978), to modify where floral parts contact pollinators (Armbruster et al., 2014;Minnaar et al., 2019), or to modify traits to favor reliance on other pollinators (Hopkins & Rausher, 2012). Alternatively, negative fitness effects could select for traits that allow tolerance of HP deposition.
For example, larger stigmatic surfaces may provide tolerance to HP because they are less likely to be clogged with pollen (Ashman & Arceo-Gómez, 2013). However, the effect of selection on traits to avoid HP deposition depends on temporal and geographical variation in patterns and reproductive consequences of HP receipt within a species (Arceo-Gómez, 2021;Fang et al., 2019). Thus, better understanding of these patterns could be important for understanding HP-mediated adaptation.
Most studies evaluating the effects of HP have been carried out with potted plants or in greenhouses and have involved experimental manipulation of HP loads, which may lead to an incomplete understanding of natural environmental effects (Ashman & Arceo-Gómez, 2013;Lanuza et al., 2021). For example, HP receipt could decrease conspecific pollen-tube growth under resource-limited but not high-resource conditions (Celaya et al., 2015). Moreover, in typical HP studies, a certain proportion of HP, ranging from 5% to 20% or higher, from a single-donor species is applied in an attempt to achieve consistent quantities on all treatment flowers (Caruso & Alfaro, 2000;Da Silva & Sargent, 2011;Moragues & Traveset, 2005; Thomson et al., 1982). In contrast to manipulative experiments, observations of HP incidence under natural pollination regimes have revealed that quantities of HP received tend to be variable and from diverse donor species (Emer et al., 2015;Fang & Huang, 2013;Johnson & Ashman, 2019;Montgomery & Rathcke, 2012;Wei et al., 2021;Zhang et al., 2021). Consequently, effects may differ between manipulative studies and those investigating natural variation (e.g., Montgomery, 2009). Studies that evaluated the impact of natural HP receipt on plant reproduction in the field have observed no detrimental effect of HP (Montgomery, 2009), negative effects on conspecific pollen-tube success (Parra-Tabla et al., 2020), or negative interaction between conspecific pollen (CP) and HP quantities, with the effect of CP on seed production becoming weaker with larger quantities of HP deposition (Briggs et al., 2016). Moreover, because the composition of co-flowering species and pollinators can differ greatly across communities, different populations of the same species may receive HP from different donor species (Herrera, 1988;Johnson & Ashman, 2019), which suggests that the effects of HP receipt may vary among populations across a species' range. Documentation of patterns of HP receipt across the species' range would help to understand the importance of pollen transfer in floral evolution and species coexistence (Arceo-Gómez, 2021; Arceo-Gómez, Abdala-Roberts, et al., 2016;Mitchell et al., 2009). A recent study showed consistent patterns of pollen load size and HP proportion in 34 co-flowering species over three consecutive years, which suggests that propensity for HP receipt is likely a species property rather than the result of occasional random events (Fang et al., 2019 Waites and Ågren (2004) and Arceo-Gómez, Abdala-Roberts, et al. (2016) both documented the intensity of HP receipt at multiple sites and found that qualities and compositions of pollen receipt varied significantly among populations. However, the pollen grains were classified into general categories (compatible, incompatible, and heterospecific in Waites & Ågren, 2004;conspecific and heterospecific in Arceo-Gómez, Abdala-Roberts, et al., 2016), but not identified to the species level, which limited the interpretation of interspecific pollen transfer.
Investigations of variation in HP receipt and its effects on reproductive success across populations of a species could improve understanding of the spatial variation in HP-mediated selection pressures.
In this study, we collected naturally pollinated stigmas of Salvia przewalskii (Lamiaceae) while leaving remaining floral structures undamaged in order to investigate how the pattern of pollen receipt correlated with fruit and seed set. This approach allows linking pollen receipt to seed set within a species across individuals and communities (see Briggs et al., 2016). We identified the HP at the species level and then aggregated counts to the family level, which allowed us to assess interactive effects between pollen from different HP donor families and conspecific pollen on fruit and seed sets. S. przewalskii is a common subalpine herb in the southwest of China and is known to be pollinated by several bumblebee species (Ye et al., 2017). Frequent interspecific pollinator movements and heterospecific pollen transfer were observed between S. przewalskii and co-flowering species (Fang & Huang, 2013. Our study addresses the following questions: (1) How does pollen composition on S. przewalskii stigmas vary among sites? (2) How do the quantities of HP from different donors and conspecific pollen affect seed set? (3) Does HP have larger effects at some sites due to being more abundant? (4) Does the relationship between HP receipt and fruit or seed set vary among sites? By addressing these questions, we assess whether there is geographic variation in fitness effects of HP.

| Study system and sites
Salvia przewalskii is a native perennial herb of sub-alpine meadows, hillsides, or forest margins in Southwest China. Individuals usually produce multiple stems. The branched inflorescence has widely spaced whorls of flowers opening a few a time, with blooming from mid-July to late-August. Including side inflorescences, there are up to 60 flowers per individual (Ye et al., 2017). Flowers are zygomorphic and have purple-red or red-brown corollas. They are epipetalous with a slightly exserted style and four ovules, which mature into fruits with 0-4 seeds. Flowers typically remain open for 2-3 days.
This study was conducted at eight sites in the southeast of the Hengduan Mountains, Yunnan Province, Southwest China. Study sites were sub-alpine meadows separated by 2-50 km ( Figure 1a, Table S1), and S. przewalskii was one of the major flowering species at each site. At each site, one 30 × 30 m plot was established containing about 100-150 flowering S. przewalskii individuals. Sampling from only one plot per site minimized within-site variation in the pollination environment. We identified the co-flowering species in each plot. On average, 15.9 ± 4.2 (mean ± SD) co-flowering species were in bloom in study plots during our fieldwork.

| Stigma and seed collection and pollen identification
Field collections occurred from July 25 to August 19, 2021, during peak flowering of S. przewalskii. We observed abundant pollinators, suggesting saturated pollination. We labeled 1065 fully blooming flowers (133.1 ± 10.2 per site) from 369 individuals selected haphazardly ( Table 1). Density of S. przewalskii was similar among sites, and it was the most or second-most abundant flowering species at all sites. We collected 2.9 (median, 3) stigmas per plant, with a range from 1 to 10 stigmas due to variations in plant size. Because the wilted corolla and pistil usually abscise, we secured the corolla to the corona with a single horizontal pin at the corona's midpoint to keep it from falling off. For each site, we performed a given procedure (labeling, stigma collection, or seed counts) for all flowers on the same day (Table S1 for labeling dates). After 3 or 4 days, when the corolla wilted, we removed the stigma with clean forceps and stored it in a microcentrifuge tube containing 70% ethanol. In most cases, the style abscised naturally, which indicated that the fertilization had already occurred, and germinated pollen exines were still affixed to the stigma. Ten days after stigma collection, we counted the developed and undeveloped seeds of each flower, which were readily distinguishable based on size. This approach allowed us to link pollen F I G U R E 1 (a) Location of the eight study sites (the coordinates of each location are included in Table S1). (b) Quantities of heterospecific and conspecific pollen deposited on stigmas of each flower at eight sites. (c) The proportion of heterospecific pollen (HP/total grains) at eight sites. Bars represent 95% confidence intervals. Different letters indicate significant differences at p < .05. receipt and seed number in the same flower (Briggs et al., 2016;Waser & Price, 1991). In some cases, stigmas were damaged or flowers (plants) were missing or showed evidence of herbivore damage.
We then observed the stigmatic pollen grains in the laboratory at 400× magnification and captured a series of images using a digital camera (Nikon D90) linked to a microscope (Nikon E100), with one image per species of HP pollen on each stigma, the pollen from each HP species being sufficiently aggregated to be captured in a single image. Pollen grains were identified to the species level on the basis of morphological features including size, shape, and exine ornamentation, with use of a pollen reference library of co-flowering species within sites (Fang & Huang, 2013. We excluded the observations for which there was no pollen deposition (10 flowers) or conspecific pollen quantities were smaller than the number of seeds (19 flowers). We assume that these observations were due to loss of stigmatic pollen in the field during fertilization and stigma collection.

| Data analyses
To test whether S. przewalskii populations experienced differences in pollen deposition, we compared the quantities of CP and HP received, as well as the proportion of HP across sites. For comparisons of CP and HP among sites, we used GLM (Poisson) with site as a fixed factor, with individual flowers treated as the unit of replication. For comparisons of the proportion of HP among sites, we used a binomial GLM (logit-link) with the dependent variable consisting of HP (event variable) and total pollen grains (trial variable) and site as a fixed factor (GLM, SPSS 19.0). To test whether S. przewalskii received HP from similar donor species across sites, we calculated Bray-Curtis dissimilarity of HP compositions between all site pairs. The Bray-Curtis similarity ranges from 0 (no difference) to 1 (completely different HP composition).
The major HP donor species varied substantially across sites (Table 1) and no single species or genus contributed over 30% of total HP. Consequently, we did not test the effect of each HP species or genus individually. Instead, we identified HP to the species level and then summed counts to the family level. For each HP family, one or two species contributed most of the pollen (e.g., Picris hieracioides for Asteraceae, Astragalus pullus for Fabaceae, Pedicularis densispica for Orobanchaceae, and Potentilla lancinata for Rosaceae). Even at the family level, most families were too rarely observed to allow for meaningful comparisons of their effects on reproductive success across sites. Consequently, we categorized HP loads into three HP categories based on the observed prevalence of pollen from each HP family: the most common family (Orobanchaceae, Oro hereafter, including 8 species, all in Pedicularis, 23.6% observed prevalence and 22.7% contribution of total HP); the second-most common family (Asteraceae, Ast hereafter, including 10 species, 22.1% prevalence and 11.5% contribution); and "Other," a category including all 47 TA B L E 1 The sampling effort, total quantities of conspecific pollen (CP), and heterospecific pollen (HP), as well as species richness of HP donors across all stigmas, richness of co-flowering community in site's vicinity, and major heterospecific pollen categories for eight study sites in Shangri-La, Southwest China.
There was significant among-site variation in the quantities of CP  Note: Models include data from all sites and were analyzed for the indicated response variables and independent variables of CP (all models) and HP or HP category, as listed parenthetically.

| The associations of HP with fruit set and seed set among sites
The associations of CP and HP quantities on fruit set did not significantly differ among sites (Table 2). In all candidate models, the CP quantity was positively associated with fruit set (p < .001, Table 3), while the HP quantity had no significant effect (p > .18 in all models).
Thus, CP quantity but not HP quantity is the primary influence on fruit set.
The association of CP quantity with seed set differed among sites ( Table 2). The model that best explained the influence of pollen quantity on seed set included the interaction between CP quantity and site, but not HP quantity and site (ΔAIC = 13.99,  (Figure 2). When the HP was separated into three categories, the best model included the interaction between CP quantity and site, but not interactions between any of the HP categories and site or between CP quantity and any HP categories (ΔAIC = 13.53, Table 3). In the best model, CP quantity was positively associated with seed set (z = 5.19, p < .001); the quantity of Other HP was negatively associated (z = −1.98, p = .048,  to the quantity of any pollen category ( Figure S1).

| DISCUSS ION
In this study, we evaluated variation in HP receipt for S. przewalskii flowers across eight sites, and the effects of pollen deposition on fruit and seed set under natural conditions. Conspecific pollen receipt was positively associated with fruit and seed set while HP receipt was negatively associated with seed set. The positive association of CP with seed set varied among sites, whereas the negative association of HP with seed set was consistent across sites. However, the quantities of HP and CP, as well as the HP proportions and HP compositions varied substantially across sites, such that the impact of HP on seed set varied among sites. Pollen from Orobanchaceae and Asteraceae, the two most common HP sources, had no measurable detrimental effect on seed set, while the quantity of all other rare HP species was associated with reduced seed set.
Heterospecific pollen was prevalent on stigmas, as has been found in other studies (e.g., Arceo-Gómez, Abdala-Roberts, et al., 2016;Fang & Huang, 2013;Wei et al., 2021). In S. przewalskii, HP grains were deposited on more than two-thirds of stigmas and represented about one fifth of the total pollen load. So far, there is little empirical indication of the extent of within-species variation in pollen receipt (see Arceo-Gómez, 2021). A recent study that investigated pollen receipt in 34 species, including S. przewalskii, over three consecutive years found temporal stability in patterns of pollen receipt with greater variation in the proportion of HP among species than among years (Fang et al., 2019). Those results suggest that propensity for HP receipt is affected by species traits, not a result of occasional random events. In that study, S. przewalskii had a comparatively low HP proportion across 3 years, less than 20% (Fang et al., 2019). In the present study, despite substantial variation in CP and HP loads, the proportion of HP was typically low, less than 30% of total pollen receipt at 7 of 8 sites. However, at site SC, the HP and CP quantities were similar, suggesting greater potential for pollen interference to affect reproduction. Further studies of the within-species variation of pollen receipt could provide better understanding of the cause and consequence of HP deposition (Arceo-Gómez, 2021;Mitchell et al., 2009).
Under natural conditions, species often receive heterospecific pollen from multiple species at different ratios (Fang & Huang, 2013;Johnson & Ashman, 2019;Wei et al., 2021), but few studies have investigated the effect of heterospecific pollen deposition in the field (Celaya et al., 2015). Similar to Briggs et al. (2016), we found that CP quantity exerts a primary influence on seed set; however, we detected only a direct negative effect of HP on seed set rather than a negative interaction with CP. The effect of HP may depend on interactions between donor and recipient species, with HP having strong detrimental effects on seed set in some donor and recipient combinations, but neutral or positive effects in others (Arceo-Gómez et al., 2019;Lanuza et al., 2021). In this study, we observed only neutral or negative fitness effects of different HP categories.
Our results demonstrate that the quantity of CP and HP as well as the relationship between CP and seed set vary among sites.
S. przewalskii is self-compatible but outcrossing increases seed set (Ye et al., 2017). The variable benefit of CP across sites may result from smaller benefits at sites with higher selfing or inbreeding rates. There may be selection for increased floral attractiveness at sites with low CP receipt and traits that promote outcrossing or F I G U R E 2 Scatterplots showing relationship between natural log-transformed conspecific (a) and heterospecific pollen quantities (b) and predicted seed set per flower based on GLMM from Table 3 (b) that best explains developed seeds based on HP quantity. Linear regressions ±95% confidence intervals are depicted.
reduce inbreeding depression at sites with reduced benefits of CP receipt. The variable benefit of CP across sites may also relate to variation in resource limitation-benefits of CP may be reduced at sites where provisions rather than CP receipt limits seed production (Herrera, 1988). The random effect of site was significant in seed set models, which indicated that other factors, such as resource limitation, may cause among-site variation in seed set (Totland & Birks, 1996). The site habitats were different from meadow to hillside (Table S1) We found consistent negative effects of HP but variable HP loads across sites, which suggests that populations may experience different levels of HP interference. Populations that have been continually exposed to a variety of HP donors may experience selection for adaptations that increase tolerance of HP (Kay & Schemske, 2008).
For example, smaller HP effects were observed in populations of Clarkia xantiana that previously experienced continual exposure to HP compared to populations without such exposure (Arceo-Gómez, Raguso, & Geber, 2016). However, the potential role of HP receipt as a selection force is largely unknown (Hopkins & Rausher, 2012).
Longer-term study is required to elucidate whether the spatial variation of HP receipt in S. przewalskii could lead to spatial heterogeneity in selection. Alternatively, it is possible that pollen from these families diminishes F I G U R E 3 Correlations among HP proportion (a), conspecific pollen quantity (b), heterospecific pollen quantity (c) and seed set per flower across eight sites. Each point represents one site. Bars represent ±SE.
seed set less than pollen from less common HP sources. In previous studies in the same region, several Orobanchaceae (Pedicularis) and Asteraceae (Picris and Aster) species were prevalent HP donors to a diverse group of recipient species (Fang & Huang, 2013.
There may be more consistent selection to tolerate receipt of HP from frequent donors.
Our results demonstrate variation in the quantity and fitness effect of HP receipt across sites, which is a precondition for local adaptation. Future work should evaluate the spatial-temporal variation in HP receipt and its consequences for seed set. If withinpopulation HP receipt or the detrimental effect of HP deposition is highly stochastic over years, it would strongly limit the opportunity for HP to influence selection on flower traits. Conversely, if spatial variation in HP receipt and detrimental effects of HP deposition were consistent across years, these conditions would increase the potential for selective effects of heterospecific pollen to influence floral traits. Thus, understanding spatial and temporal variation in patterns of HP receipt and effects would provide insight into selective pressures related to reproductive success of co-existing plant species. Similarly, observational and experimental approaches to assessing factors that influence HP receipt, such as the co-flowering species, species arrangement, and pollinator composition, would also help uncover the underlying drivers of selections (Thomson et al., 2019). In conclusion, our study indicated that plant species were exposed to different HP transfer environments across sites. The results of our study highlight the importance of within-species variation in HP receipt across natural communities. Such studies would help to understand the way that co-flowering species interact through pollinator sharing and heterospecific pollen transfer, as well as the potential ecological and evolutionary consequences.

ACK N OWLED G M ENTS
We thank Kexing Li for the help in field and Chenzhou Liu for help creating the coordinate map and two anonymous reviewers for constructive comments to help improve this work. The Shangri-la Alpine Botanical Garden provided the research facilities and intellectual community that helped make this study possible.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no confict of interest.