Ecological implications of reduced pollen deposition in alpine plants: a case study using a dominant cushion plant species

The reproductive assurance hypothesis states that self-incompatible female plants must produce twice the number of seeds relative to their self-compatible hermaphroditic counterparts to persist in gynodioecious populations. This is a viable life-history strategy, provided that pollination rates are sufficiently high. However, reduced pollination rates in alpine plants are likely due to climate induced plant-pollinator mismatches and general declines in pollinators. Using a gynodioecious population of the dominant plant Silene acaulis (Caryophyllaceae), we tested the reproductive assurance hypothesis and also the stress gradient hypothesis with a series of pollinator exclusion trials and extensive measurements of subsequent reproductive output (gender ratio, plant size, percent fruit-set, fruit weight, seeds per fruit, total seeds, seed weight, and seed germination). The reproductive assurance hypothesis was supported with female plants being more sensitive to and less likely to be viable under reductions in pollination rates. These findings are the first to show that the stress gradient hypothesis is also supported under a gradient of pollen supply instead of environmental limitations. Beneficiary abundance was negatively correlated to percent fruit-set under current pollen supply, but became positive under reduced pollen supply suggesting that there are important plant-plant-pollinator interactions related to reproduction in these alpine plant species.

At least two future climate change scenarios predict that pollination rates will be directly reduced, and these are due to either pollinator declines (Potts et al., 2010) and/or plant-pollinator mismatches (Hegland et al., 2009). Recently there has been concern over general global trends of reduced pollinator species abundance and diversity that are both predicted to reduce pollination rates to plants (Memmott et al., 2007;Potts et al., 2010). Climate induced plantpollinator mismatch can reduce pollination rates by creating a temporal mismatch in pollinator emergence and plant flowering times (Hegland et al., 2009). Pollinator emergence is regulated by temperature, whereas plant bloom time is regulated by photoperiod (Hegland et al., 2009). If climate warming shifts pollinator emergence but not plant bloom time, then a temporal mismatch between plants and pollinators occurs (Hegland et al., 2009). This scenario is likely more pronounced in alpine and polar environments that are experiencing a more rapid increase in annual temperature than the global average (very high confidence; Intergovernmental Panel on Climate Change [IPCC], 2013). Conceivably, both of these reductions in pollination rates occur simultaneously and thus adaptability of different sexual morphs in alpine plants can be an important consideration in predicting responsiveness and variation in reproductive output.
The reproductive assurance hypothesis (RAH) and the stress gradient hypothesis (SGH) are thus highly relevant hypotheses to explore in better understanding climate change effects on alpine communities. The RAH proposes that when pollen supply is low, self-compatible plants are favored over self-incompatible plants (Lloyd, 1992;Lloyd & Schoen, 1992). This is because self-compatible plants create their own pollen thereby being more adapted to low or variable pollination rates (Larson & Barrett, 2000;Muñoz & Arroyo, 2006;García-Camacho & Totland, 2009). Further, it has been proposed that self-compatible plants are less likely to become extinct if pollinators drastically decrease or disappear from a given system (Richards, 1997;Morgan et al., 2005). Therefore, selfcompatible plants may become favored in the future if pollination supply declines. The SGH is also an important ecological theory to consider with respect to potential climate impacts on pollinators. The SGH states that facilitation between plant species is more common when resources are limited (Bertness & Callaway, 1994).
Typically, the SGH is tested using environmental limitations such as temperature or moisture in the alpine (He et al., 2013;Liczner & Lortie, 2014;McIntire & Fajardo, 2014), but has not been applied to the concept of pollen supply as an important limitation for plants in stressful environments. Taken together, these ecological theories provide a solid platform to build pollen limitation studies upon and also provide a set of potential ecological drivers that can help better predict pollination rate changes in the alpine.
Here, we use a gynodioecious population of Silene acaulis to assess the sensitivity of different genders to pollen limitation. We test the following predictions associated with the reproductive assurance hypothesis: (1) that the reproduction of self-incompatible plants is more sensitive to reduced pollen deposition than self-compatible plants and (2) that self-incompatible plants will be less viable under experimentally reduced pollen loads compared to self-compatible plants. In doing so, we also explore whether the SGH applies to the plant-pollinator system in the alpine. Specifically, we predict that facilitation between plants is more common under reduced pollen, i.e. that less pollen can be a novel stressor for alpine plants and that this can in turn relate to plant-plant interactions.

Study site
The experiments were conducted on the Whistler Mountain in British Columbia 50°03′31.68″N, 122°57′22.53″W, 2168m elevation), Canada, during the snow-free season of July and August 2010. This area is classified as alpine tundra with ten months of snow cover per year (Pojar et al., 1987). A total of 273 S. acaulis plants were measured. Three S. acaulis plants were excluded from the study because they were infected with the pollinator-transmitted anther smut-fungus Microbotryum violaceum that renders the flowers of both genders sterile (Baker, 1947;Alexander & Antonovics, 1988;Hermanutz & Innes, 1994;Marr, 1997).

Treatments
Before bud-burst, S. acaulis plants were covered with cloth mesh to prevent pollinators contacting the flowers (Donnelly et al., 1998). As the plant gender was unknown when initially covering, 60 plants were covered to ensure that there would be sufficient replicate plants of each gender. Plant gender was established after bud-burst, and at that time, plants were randomly assigned a treatment such as covered with mesh or open to insect pollinators and marked with a unique identification code. Reduced pollination treatments were the ones covered with mesh and were applied to 20 hermaphrodite and 20 female plants. The first 20 female and male plants found were used as replicates with the additional 20 plants being uncovered.
The 40 plants (20 of each gender) selected for the reduced pollination treatments remained covered with mesh for the entire flowering season to exclude all insect pollination.
All flowers of the reduced-pollination treatment plants were handpollinated with pollen collected from S. acaulis plants within 10 meters from the treatment plants. All hermaphrodite flowers with mature anthers were collected in the morning of the hand pollination days. Pollen was then applied using small paintbrushes or by directly touching the anthers to the stigmas of all the treatmentplant flowers. We found that direct contact of the anthers to the stigmas was the most effective method of hand pollination. The exact amount of pollen applied to each flower at each hand-pollination event was not quantified. Hand pollination was repeated on three different days between July 20 th and August 1 st , 2010.

Reproductive output measures
Reproductive output measures were collected from the 40 hand-pollinated treatment plants as well as 231 naturally pollinated S. acaulis plants. These measures included total number of flowers, total number of fruits, percent fruit-set, seeds per fruit, fruit weight, and seed weight. The percentage of germination was calculated on a subset of 60 plants, including the 40 treatment plants and 20 naturally pollinated plants.
The total number of flowers was counted during fruit collection including both successfully and unsuccessfully (i.e. no fruit) pollinated flowers. Fruits were collected when mature but not yet dehiscing, so that the seeds remained in the fruit capsule. This occurred between August 11 th and 25 th . All fruits were placed in small labeled paper envelopes and were allowed to dry at room temperature to avoid decomposition. The percentage of fruit-set was calculated using the measures of total number of fruits and total number of flowers. The mean fruit weight (g) was calculated by averaging the weight of five randomly selected fruits per plant. These five fruits were dissected and the seed counted. The mean seed number per fruit was calculated from the seed counts. The total seed number per plant was estimated by multiplying total number of fruits with mean seed number per fruit. Mean seed weight (mg) was calculated by averaging the weight of ten randomly selected seeds per plant.
All weighing was done to four significant digits. When a plant produced less than five fruits or ten seeds, the average was based on the maximum number of fruit or seed produced. Weighed seed was stored separately and cold-stratified at 4°C for two months. A test germination trial was conducted with limited success likely because the cold stratification was not sufficient. Therefore, seeds were then stored at 0°C for two additional months in preparation for germination trials.
Germination trials were conducted on the weighed and cold stratified seeds from the 40 S. acaulis plants in reduced pollination treatments and the remaining 20 labeled plants that were left open to natural pollination. Growth chambers were set to standard optimum growing conditions of 20°C and light for 12 hours, then 10°C and dark for the remaining 12 hours of the day (Baskin & Baskin, 1998). Relative humidity was set to 90%. The ten seeds from an individual plant were placed on a labeled filter paper in a Petri dish. Seeds were checked weekly for three months, after which germination is rare (Milbau et al., 2009). Germination was considered to have occurred when the radical broke open the seed (Milbau et al., 2009). Germinated seeds were removed to speed-up counting during the subsequent weeks and reduce counting errors (Milbau et al., 2009). Percent germination was expressed as the fraction of total number of germinated seeds with respect to the total number of seeds per Petri dish.
Cushion area and floral density were measured because of their possible effect on reproductive output. Cushion area was defined by the external boundary of vegetation and calculated as an ellipse with the formula, where a is the longest diameter of the plant and b the diameter perpendicular to a. We calculated the floral density by dividing the total flower number by the cushion area.

Beneficiary abundance
To test if facilitation became more common under reduced pollen loads (stress gradient hypothesis), we measured beneficiary abundance on all cushion plants. Beneficiary abundance is the total number of individual plants living on the cushions.

Statistical analysis
To assess the sensitivity of female and hermaphrodite reproductive output under reduced pollen deposition, we calculated the percent change of reproductive success measures within each gender between current and reduced pollen deposition. Percent change was calculated using the following equation: where T is the reproductive output measures under the reduced pollination treatments and C is the reproductive output measures under the current pollination rates (Ayres, 1993). Negative numbers indicate that reduced pollination treatments decrease reproductive success and positive numbers indicate that reduced pollination treatments increase reproductive success. This method facilitates comparisons of the direction and magnitude of change.
To statistically test if reduced pollination, gender, and their interaction effects significantly explained the variation in the reproductive output measures, we used a generalized linear model (GLM) with Poisson distribution and a log link function. Covariate measures of S. acaulis included surface area, S. acaulis floral density and beneficiary abundance.
Hermaphrodites provide one-half of the genetic material to the population through pollen production. Therefore, female plants must produce at least twice the number of seeds as hermaphrodite plants to persist in the population (Charnov, 1982) or have offspring that are more fit (Lewis, 1941). To statistically analyze the viability of females under current and reduced pollen deposition levels, we compared female reproductive output to twice that of hermaphrodite reproductive output. In this way, if female reproduction (F) is greater than two times hermaphrodite reproduction (2H), then females are viable in the population. For females to be viable, not all measures of reproductive success need to be twice that of hermaphrodites, but all measures are shown to be comprehensive. GLMs were also used to test if gender significantly affected reproductive output measures under the current ambient and experimentally reduced pollination regimes.
Instead of testing the effect of beneficiaries on measures of reproductive success over a range of environmental gradients (Bertness & Callaway, 1994) we tested whether variation in the current and experimentally reduced pollen supply can be viewed as potential stressor. To test the stress gradient hypothesis in this plant-pollinator system, we conducted correlation analysis and tested for significance in the interaction term between the effect pollination and beneficiary abundance on percent fruit set in a GLM. A significant p-value (p<0.05) indicates that the response of percent fruit set to beneficiary abundance significantly differs between plants in the current and reduce pollination regimes. All analyses are appropriate for dealing with the unbalanced number of replicates between the current and reduced pollination regimes and were done in JMP 10 (SAS, 2012).

Results
Pollination regime significantly influenced percent fruit-set, seeds per fruit, fruit weight and percent germination (Table 1). Percent fruitset, seeds per fruit, fruit weight, and percent germination decreased in both genders with reduced pollen deposition ( Figure 1). There was a significant interaction effect between gender and pollination regime for percent fruit set, and percent germination (Table 1) indicating that these measures differed in their response. The direction of these differences was significant and is illustrated in the percent change calculations. Percent fruit-set was more significantly reduced in females relative to hermaphrodites (Figure 1), whereas percent germination reduced to a greater in hermaphrodites compared to females (Figure 1).
Under current pollination rates, female plants had more than twice (2.98 times) the percent fruit-set compared to hermaphrodites ( Figure 2). Female plants had less than twice the seeds/fruit, fruit weight, seed weight and percent germination compared to hermaphrodites ( Figure 2). Under current pollination rates, all reproductive output measures were significantly different between females and two times hermaphrodite reproductive measures (Table 2).
Under reduced pollination rates, none of the female reproductive measures were greater than two times that of hermaphrodites ( Figure 2). Under reduced pollination rates, all reproductive output measures,

Figure 2. Reproductive output measures for female plants (white) and two times that of hermaphrodite plants (grey) under current
and reduced pollination regimes (p < 0.05). ± 1 standard error bars shown. except seed weight and percent germination, were significantly different between females and two times hermaphrodite reproductive measures (Table 3).
Beneficiary abundance had a significant effect on percent fruit-set (Table 1). Under the current pollination rates, percent fruit-set and beneficiary abundance are negatively related (slope = -0.48, R 2 = 0.04, Figure 3). In contrast, under reduced pollination rates, percent fruit-set and beneficiary abundance are positively related (slope = 0.24, R 2 = 0.06, Figure 3). The slopes of these lines significantly differ (Chi 2 161.25, p-value <0.0001).

The reproductive effects of reduced pollen deposition via exclosures and hand pollination on the cushion plant S. acaulis
S. acaulis is an important alpine plant species commonly demonstrated to have positive impacts on other plants and insects.
In this experiment, we covered sets of plants (gynodioecious species with hermaphrodite and female flowers) with mesh with hand pollination only to examine the relative importance of pollen load as a potential stressor in alpine ecosystems.

Discussion
Pollinator declines and climate mismatches are important in understanding the capacity for alpine plants to respond to possible future scenarios with reduced pollen loads. Using the cushion plant S. acaulis, we tested two predictions associated with the reproductive assurance hypothesis and more broadly we investigated whether pollination stress influences plant-plant interactions. All predictions were supported. Females were more sensitive than hermaphrodites to reduced pollen loads resulting in reproductive output dropping to below twice that of hermaphrodites. As the pollen supply conditions became more stressful (i.e. reduced), beneficiary plant species on these cushions positively related to the percent fruit-set of the cushions. Hence, the reproductive assurance hypothesis and use of pollen reduction experiments can be important tools for ecological experiments on the responsiveness of alpine plant-pollinator systems to future changes in pollen availability. Importantly, loss of keystone alpine plant species such as cushions may in turn have significant and reciprocal negative impacts on the pollinator communities.
Pollinator declines and plant-pollinator mismatches are important potential drivers of broad plant-community dynamics in the alpine if dominant cushion plant species are impacted because they often function as keystone plant species (Arroyo et al., 2003;Bertness & Callaway, 1994;Callaway & Walker, 1997;Cavieres et al., 2006;Molenda et al., 2012;Butterfield et al., 2013;McIntire & Fajardo, 2014). Current trends of decreasing native pollinator populations are a pressing concern globally (Memmott et al., 2007;Potts et al., 2010;Bartomeus & Winfree, 2013). In these alpine environments, bumblebees in particular are suggested to be critical because they are the most effective alpine pollinator in these ecosystems (Bingham & Orthner, 1998;Chittka et al., 1999;Gegear & Laverty, 1998), and for instance, because some alpine bumblebee populations are in decline (Colla & Ratti, 2010). The future scenario that pollinators may emerge before flowers are in bloom due to a warming climate has also been proposed (Hegland et al., 2009) and shown in the alpine environments of Japan (Kudo, 2013). Although it may not mimic the exact future pollen deposition rates, the experimental design tested herein begins to explore how alpine plants may respond to reduced-pollen loads. Differences in reproduction between alpine plant genders are thus a critical avenue of research and are important because cushion plants are common facilitators (for instance, see meta-analysis in Liczner & Lortie, 2014).
Two findings were particularly useful from an ecological perspective. Female cushion plants became less viable under reduced pollen loads and the stress gradient hypothesis was supported as means to model a gradient of pollen deposition rates because it is also clearly a potential limitation in some stressful ecosystems. Percent fruit-set was the only measure that indicated a drop in female viability. Indeed, this more than compensated for female plants' reproductive disadvantage over hermaphrodite plants. Hence, a reasonable proxy or single measure to consider in similar future studies using dominant cushion plants is percent fruit-set only. Interestingly, our results also supported the application of the stress gradient hypothesis to pollen limitation in addition to its original formulation for environmental stress or consumer pressure. The relationship between beneficiary abundance and percent fruit-set shifted from negative to positive as the pollination rates were reduced. This supports previous findings that under current pollination to the plant community, beneficiary plants living on cushions generally have a cost associated with the cushion plants' reproductive fitness (Cranston et al., 2012;Schöb et al., 2013). The findings here however also further suggest that under reduced pollen loads this cost of facilitation can be diminished likely because competition between cushions and the other species is significantly reduced. Clearly, additional research is needed to identify the causal relationships between plant-plant interactions and plant-pollinator interactions with dominant plant species that host other species in stressful environments such as the alpine.
Author contributions AR and CJL wrote the paper and analyzed the data. AR, RH and OM performed the experiments and collected the data. All authors read and approved the final content of the manuscript.

Competing interests
No competing interests were disclosed. This is an interesting study dealing with the effects that facilitated plants may have on the performance of the reproductive success of the nurse species. As far as I am aware, this is the first study reporting such effects. The article is well written and is easy to read. The arguments provided in the introduction section support the hypotheses raised by the authors, and the methods used to test these hypotheses are sound. I believe that this paper makes an important contribution to the conceptual framework of positive interactions among plants along environmental gradients, mainly because it helps one to understand the outcome of these interactions in future climate change scenarios. I just have a few comments and queries about the methods used by the authors. The other sections of the article are clear and would not require major improvement.

Materials and Methods
In the subsection "Study species", several sentences begin with the abbreviated scientific name of species ( ). Nevertheless, it is not appropriate to use the abbreviated name of the species S. acaulis for beginning a sentence. In this case, the full name of the species should be provided (Silene ) while its abbreviated form ( ) must be used when the species is mentioned in the acaulis S. acaulis middle of a sentence.
In the second paragraph of the subsection "Study species" the authors indicate that, besides S.
, several other plant species are relatively abundant. My question on this issue is: Are all acaulis these other species beneficiary of or they are just sharing the habitat with this cushion S. acaulis plant? This should be clarified to understand the aim of testing the Stress Gradient Hypothesis.
In the first paragraph of the subsection "Treatments" the authors indicate that they choose 20 female and male plants for the experiment. This caught my attention because in the previous sections of the article the authors stated that they used a gynodioecious population of S. acaulis and, therefore, male plants should not exist. I think that there is a mistake in this sentence, where instead of " " the authors should say "20 female plants and plants with 20 female and male plants hermaphroditic flowers".
I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.
No competing interests were disclosed. This is a well conducted experiment and well written manuscript. It is the first experimental test of the stress-gradient hypothesis under pollen limitation stress, showing how we can further our understanding of plant-plant interactions with a new perspective. In my view, this is a neat paper, though I have some concerns that I hope the author can address in a revision.
The main limitation of this paper is that plant-plant interactions are inferred by association, not experimental removal of neighbours. Although a number of studies from alpine habitats suggest that facilitation can be indicated by positive association relationships, this limitation I feel should be clarified in the paper.
It is also a bit confusing to me why the authors did hand pollination for reduced-pollination treatment plants -was this because reduced-pollination treatment plants did not receive pollination at all without hand pollination? If so, the question would be how the authors controlled the magnitude of hand pollination as compared to natural pollination.
A third issue is that the pollen reduction treatments lacked procedural control. For example, one cannot tell the effect of mesh on the data. Also surely, the authors need to provide some details about the cloth mesh they used, such as mesh size etc.
It is unclear what can be the potential mechanisms for facilitation under pollen limitation. Facilitation in physically stressful habitats is produced by neighbours' relief of abiotic stress, and in biologically stressful habitats by associational defence. It is not a must for the authors to experimentally demonstrate this in the paper but I think they should at least discuss some of the possibilities. The authors may also want to include descriptions to encourage future research that experimentally tests for these possibilities.
I have several more specific comments: " " Clarify the start and end dates of your experimental treatments. What is the Before bud-burst ... total duration? " " Does the male here mean "hermaphrodite"? The first 20 female and male plants found were ... Use terms consistently.
For treatments, how many plants in current pollination treatments are female and male,