Pre- and post-pollination interaction between six co-flowering Pedicularis species via heterospecific pollen transfer

Transcription

1 Research Pre- and post-pollination interaction between six co-flowering Pedicularis species via heterospecific pollen transfer Ze-Yu Tong 1 and Shuang-Quan Huang 2 1 State Key Laboratory of Hybrid Rice, College of Life Sciences, Wuhan University, Wuhan , China; 2 Institute of Evolution and Ecology, School of Life Sciences, Central China Normal University, Wuhan , China Author for correspondence: Shuang-Quan Huang Tel: hsq@mail.ccnu.edu.cn Received: 30 October 2015 Accepted: 6 April 2016 doi: /nph Key words: bumblebee, co-flowering, flower constancy, heterospecific pollen transfer, Pedicularis, pollen tube growth, pollen pistil interaction, stigmatic pollen load. Summary It remains unclear how related co-flowering species with shared pollinators minimize reproductive interference, given that the degree of interspecific pollen flow and its consequences are little known in natural communities. Differences in pollen size in six Pedicularis species with different style lengths permit us to measure heterospecific pollen transfer (HPT) between species pairs in sympatry. The role of pollen pistil interactions in mitigating the effects of HPT was examined. Field observations over 2 yr showed that bumblebee pollinators visiting one species rarely moved to another. Heterospecific pollen (HP) comprised < 10% of total stigmatic pollen loads for each species over 2 yr, and was not related to conspecific pollen deposition. Species with longer styles generally received more HP per stigma. The pollen tube study showed that pollen from short-styled species could not grow the full length of the style of long-styled species. Pollen from long-styled species could grow through the short style of P. densispica, but P. densispica rarely received HP in nature. Flower constancy is a key pre-pollination barrier to HPT between co-flowering Pedicularis species. Post-pollination pollen pistil interactions may further mitigate the effects of HPT because HP transferred to long styles could generally be effectively filtered. Introduction The coexistence of flowering plant species depends largely on various reproductive barriers that minimize interspecific gene flow. Pre-pollination barriers have been recognized in some orchids in which pollinators are very specialized and post-pollination barriers are absent (Xu et al., 2011; Whitehead & Peakall, 2014; but see Scopece et al., 2008). In most plants with granular pollen rather than pollen packed in discrete units as in orchids, prepollination barriers mediated by pollinators are not expected to play a key role in that pollinators are generally not specific and may carry pollen from several species (Armbruster, 2014). Co-flowering plants in sympatry often share similar pollinators and consequently heterospecific pollen transfer (HPT) is not uncommon (Waser et al., 1996; Morales & Traveset, 2008). HPT can be reduced if pollen grains from different species are placed on separate parts of the pollinator body and correspondingly picked up by conspecific stigmas (Armbruster et al., 1994; Muchhala & Potts, 2007; Johnson, 2010). The other means of reducing HPT is flower constancy, in which pollinator species may restrict their foraging to a single plant species (Grant, 1994b; Flanagan et al., 2009). However, the role of flower constancy in minimizing interspecific reproductive interference has been questioned given that an occasional pollinator move to another species may bring a large amount of heterospecific pollen (HP); and even when some animals are specialist floral visitors (e.g. hummingbirds), plants may also attract generalists which may bring HP (Chari & Wilson, 2001; Waser, 2001). Our survey of HPT among 57 co-flowering species in an alpine meadow indicated that 96.5% of species received HP and 89.2% of species donated pollen to other species, and HP accounted for 0% to 66.8% (mean 16.4%) of the total pollen deposition per stigma (Fang & Huang, 2013). A recent review on variation in HP on stigmas showed that in some species over 50% of the pollen on stigmas could be HP, suggesting tolerance of the deleterious effects of HP (Ashman & Arceo-Gomez, 2013), but the mechanism of tolerance to the negative effects of HP has been little studied (Arceo-Gomez et al., 2016b). Given that the fitness costs of HP receipt could play a role in shaping existing plant pollinator interactions, Ashman & Arceo-Gomez (2013) proposed that an understanding of the role of post-pollination interactions in floral trait evolution is much needed. They predicted that style length could be a target trait to lessen the detrimental effects of HP receipt: pollen recipients with short styles will be more susceptible to the negative effects of HP receipt than those with long styles. One may predict that selection against HP receipt would be stronger in species with short styles than in those with long styles because long styles could provide a more effective selective sieve for competing gametophytes (see Nista et al., 2015), minimizing the negative effects of HP receipt. 1452

2 New Phytologist Research 1453 The genus Pedicularis Linnaeus (Orobanchaceae), with over 500 species, is one of the largest genera in the Northern Temperate region (Yang et al., 1998). Its flowers show a wide range of corolla tube length (ranging from < 10 mm to > 100 mm) as well as style length, and vary in floral shape, particularly in the galea (the hood-like upper lip of the corolla, often wrapped around the style) (Yang et al., 1998; Yu et al., 2015). In alpine areas, several Pedicularis species often coexist with overlapping flowering times and shared bumblebee species as the exclusive pollinators, but no natural hybrids are known (Macior, 1982; Eaton et al., 2012). Long-term observation by Macior (1982) on numerous Pedicularis species revealed pollinator-mediated floral isolation; pollen from co-flowering species was assumed to be placed on different sites of the bumblebee s body. It is not surprising that this type of floral isolation was referred to as the Pedicularis type by Grant (1994b). Grant s own study on the modes of floral isolation in two sympatric species, P. groenlandica and P. attolens, in California indicated that both mechanical and ethological isolation were undoubtedly major factors in the floral isolation of these species (Grant, 1994a; see also Yang et al., 2007). He noted that individual bumblebee species were highly flower-constant to one Pedicularis species during a foraging trip. Recent studies using pollen staining and fluorescent powder to examine the segregation of pollen placement and pickup in three Pedicularis species with shared bumblebee pollinators suggested that HPT was reduced, but not completely avoided, by use of different placement sites on the same pollinators (Huang & Shi, 2013; Armbruster et al., 2014). However, floral constancy and HPT among co-flowering Pedicularis species have not been quantified. The contribution of pre- and post-pollination barriers in mitigating the effects of HPT remains unclear. Quantification of the degree of heterospecific pollen transfer between co-flowering species is needed to assess the strength of floral isolation. Previous studies on HPT between congeners have usually lacked a direct measurement of HPT from natural communities (but see Muchhala & Potts, 2007), perhaps due to the difficulty of distinguishing pollen from closely related species. For example, two studies measuring heterospecific pollen deposition used fluorescent dye to mimic HPT (Goulson & Jerrim, 1997; Kay, 2006), and others used emasculated or protogynous flowers among pollen donor species to estimate HP receipt on virgin stigmas (Thomson, 1986). Here, we used six sympatric Pedicularis species which are native to our study site as a model system to quantify HPT and pollen pistil interactions (Supporting Information Figs S1, S2; Table S1). Among them, three species have much longer corolla tubes (associated with long styles and large pollen sizes) than the other three species. Size differences in pollen grains between the species pairs permit us to quantify HPT on stigmas (Fig. 1; Table S2). In particular, we addressed the following questions. (1) Do individual bumblebee pollinators have high constancy when visiting mixed Pedicularis species? Individual bumblebees of different species visiting four naturally occurring Pedicularis species pairs were followed and their constancy was quantified over 2 yr. (2) Do co-flowering Pedicularis species experience HPT? Is HP receipt related to interspecific pollinator switches? What is the Fig. 1 Pollen grain volume and style length of six studied Pedicularis species. Each circle represents one species with its main corolla color and species name marked. Curves link the two species within the four species pairs. relationship between HPT and conspecific pollen deposition (see Arceo-Gomez et al., 2016a)? (3) Can species with long styles provide a selective sieve for pollen from short-styled species (Ashman & Arceo-Gomez, 2013; Nista et al., 2015)? We compared heterospecific pollen tube growth patterns in these Pedicularis species pairs. One might expect that co-flowering species with high levels of interference by HP would evolve post-pollination mechanisms reducing the deleterious effects of HP (Arceo-Gomez et al., 2016b; Nista et al., 2015). Consistent with this prediction, we found that species with short styles, which are more susceptible to the negative effects of HP receipt, received little HP. Materials and Methods Study system This study was conducted at a field station, Shangri-La Alpine Botanical Garden ( N, E, m above sea level (asl)), Yunnan Province, Southwestern China. Six native Pedicularis species distributed around the site bloom from late June to early August. These herbaceous flowers are selfcompatible (Mao, 2010). One species (P. densispica Franch. ex Maxim.) is nectariferous, with pale pink, short-tubed flowers and the other five species are nectarless providing a pollen reward to pollinators or a little nectar (not yet seen to be collected by pollinators) in some individuals. Four species generally flower earlier from late June to mid-august except the short-tubed, pinkflowered P. confertiflora Prain and long-tubed, yellow-flowered P. tricolor Hand.-Mazz. The latter two species usually grow in open dry meadows and co-flowering mixed patches were observed in August. Short-tubed, pink-flowered P. cephalantha Franch., long-tubed, pink-flowered P. siphonantha D. Don and yellow-flowered P. longiflora Rudolph usually co-flowered in relatively wet meadows. Plant height can be up to cm in P. densispica but c cm in the other species. Individual flowers of these six species lasted 2 4 d (Mao, 2010; Huang & Shi, 2013). With distinctive morphological differences (Fig. S1; Table S1), natural patches with two species intermixed were

3 1454 Research New Phytologist studied in summer 2013 and Data on pollen size and other floral traits are presented in the supplemental materials (Tables S1, S2). The pollen of members of a species pair could be distinguished on the stigma by morphology, particularly size and shape, and sometimes also by colour and exine sculpture (Fig. S2). Bumblebee workers were observed to be the only effective pollinators in all of the six species (Huang & Fenster, 2007; Huang & Shi, 2013), in accordance with previous studies conducted in other areas (Macior, 1982; Macior et al., 2001; Yang et al., 2007; Kawai & Kudo, 2009). Assessment of flower constancy In order to investigate pollinator behavior patterns on co-flowering Pedicularis species, and to simplify interactions between these species, four pairs containing two naturally co-flowering Pedicularis species were chosen. The species pairs met the following criteria: (1) there were enough individual plants to fill a 2 9 2m 2 patch; (2) flower numbers of each species in the patch were nearly equal (convenient for adjusting flower number) and the surrounding area was not dominated by either of the species in the pair; (3) more than two replicates of these species pairs were available over 1 yr. The flower number of each paired species in the patch was adjusted to be similar (Table S3) by removing corollas from the calyx tubes in a few inflorescences. Four pairs that could be used in the field station were P. longiflora vs P. cephalantha, P. siphonantha vs P. densispica, P. cephalantha vs P. densispica and P. tricolor vs P. confertiflora. Flowers of other bumblebee-pollinated plant species within a 1 m radius around the patch were removed to exclude the possible effect of other species on bumblebee foraging. Because the selected patches were dominated by Pedicularis, the removal of flowers from other species resulted in little overall destruction. Based on species abundance and weather conditions, for each pair we conducted > 12 h of observations in several patches between 09:00 h and 17:00 h on sunny days (Table S3). A single observation period lasted 15 min. Flower numbers were recorded at the beginning of each observation. In the nectarless species of Pedicularis, a visitor must perform buzzing behavior (Corbet & Huang, 2014) to retrieve pollen. We therefore recorded visits by buzzing bumblebees, but not other behaviors such as hovering or hanging above a flower. Bumblebee specimens were collected for identification at the end of the observation period to avoid influencing the visitation rate. When a bumblebee arrived in a quadrat, the whole foraging bout was recorded including any moves between conspecific flowers and between species. If a bumblebee visited only one flower in the patch and then left, then this record was treated as an occasional visit and excluded from subsequent analyses. Given that style length within the studied species pairs was distinctly different, we designated the species with a long style in the pair as species a, and the short-styled as species b. late anthesis (Fang & Huang, 2013) in the patches containing the four pairs of Pedicularis species. For each species in the pairs in 2013, > 50 stigmas were randomly collected from over 30 individual plants (1 2 stigmas per plant) after the flowers opened on two consecutive fine days. Sixty stigmas were collected from 60 individuals (20 stigmas each from three patches of one species pair) of each species in the pairs in 2015, except species pair P. tricolor vs P. confertiflora, for which only one patch could be found in that year. Stigmas were removed from flowers in the field with clean forceps and stored separately in 0.5-ml microcentrifuge tubes in FAA (formalin : acetic acid : alcohol, 5 : 5 : 90). Each stigma was placed on a slide and then squashed under a coverslip. Then pollen grains were counted, and pollen species were identified under a microscope at 9400 magnification, based on differences in the size and shape (and sometimes color and exine sculpture) of a reference collection of pollen (Fig. 1; Table S2). Given that there was a very small amount pollen from other genera on Pedicularis stigmas (< 1%), we counted only pollen from the species involved in the pairs. Pollen tube growth experiment In order to examine interspecific pollen pistil interactions, we conducted reciprocal interspecific pollination between the two species within four species pairs and cross-pollination within each species. Hand pollination using outcross or interspecific pollen was conducted on first-day flowers which were bagged before treatment. Conspecific pollen (CP) donor samples contained a mixture of pollen from an average of ten individuals with newly dehisced anthers to mimic natural stigmatic pollen loads, which commonly reflect extensive pollen carryover in bee-pollinated systems (Schaal, 1980; Thomson & Plowright, 1980). Each pollination had at least 100 pollen grains (there are fewer than 40 ovules per flower in these Pedicularis species; Table S3) and pollinated flowers were bagged after manipulation. To observe pollen tube growth status, we fixed treated flowers in formaldehyde-acetic acid solution after hand pollination. Our previous investigation showed that in these Pedicularis species, CP can germinate on stigmas within 2 h, and reach the ovary in < 24 h, whereas tube growth of heterospecific pollen had stopped after 24 h (Mao, 2010), so we collected flowers hand-pollinated by heterospecific flowers at 24 h. For each treatment on each species, we pollinated one or two flowers from each of at least 20 individuals. The pistils were stained with 0.01% aniline blue solution. Pollen tubes were examined with an epifluorescence microscope. The number of germinated pollen grains, the length of the longest pollen tube and the number of pollen tubes that arrived at the ovary were recorded; the length of each style also was measured. When pollen tubes did not enter the stylar tissue, we scored their pollen-tube lengths as zero. For each species in the pair, styles were successfully harvested to examine pollen tube growth. Assessment of heterospecific pollen transfer (HPT) In order to quantify HPT between these co-flowering congeners, we collected stigmas randomly from open-pollinated flowers at Data analysis Gegear s constancy index (CI) was used to calculate the constancy of individual bumblebees (Gegear & Thomson, 2004; Natalis &

4 New Phytologist Research 1455 Wesselingh, 2013). CI is calculated as CI = (c e)/(c + e 2ce), where c represents the actual proportion of moves between the same flower types (constant moves) and e is the expected proportion of moves between the same flower types based on the overall frequency of visitation to a particular species in the pair. If p is the proportion of visits to species a, then e = p 2 + (1 p) 2.CI distinguishes between complete constancy (CI = 1), random moves (CI = 0) and complete inconstancy ( 1). CI was calculated for each bumblebee foraging bout. For bee individuals visiting one species without interspecific moves in a foraging bout, we assigned a CI value of 1 (complete constancy). We used a v 2 -test of independence to compare species in the same pair with respect to the proportions of stigmas with pollen or heterospecific pollen (HP). Generalized Linear Models (GLM) with binominal error distribution and a logit link function were used to compare the proportion of HP deposited per stigma between species in the same pair and also between years for the same species. To test if there was any relationship between CP and HP receipt for each species, following the criteria of Arceo-Gomez et al. (2016a), we used regression curve estimation (SPSS) to investigate whether the relationship was linear or nonlinear (exponential). Spearman s rank correlations were performed to examine whether pollen (or HP) receipt in each species was correlated with conspecific (or interspecific) pollinator moves among the species pairs. We used GLM with normal distribution and a linear link function to evaluate the difference between short- and longstyled species in the pair in the proportion of heterospecific pollen tubes that entered the ovary, the proportion of traverses completed by the longest pollen tubes in 24 h and the maximum length of pollen tube growth. Finally, the global effects of style length on the pre-pollination barrier (evaluated as the percentage of HP grains received per stigma) were analyzed using the model with binominal distribution and a logit link function in GLM, fixing the effects of style length, year and their interactions, whereas the effects on post-pollination barriers (evaluated as the completeness of HP growth in the style) were analyzed by using GLM with normal distribution and a linear link function, fixing the effects of style length. All the analyses were performed in SPSS 19.0 (IBM Inc., New York, NY, USA). Results Flower constancy During the 4005 min of quadrat observations in 2013 and 1500 min in 2015, workers of three bumblebee species were recorded as pollinators: Bombus festivus Smith, B. friseanus Skorikov and B. trifasciatus Smith. Bumblebee pollinators rarely visited two flowering species in one bout. Pollinator interspecific moves occurred in 4.0% (18/449) and 6.45% (10/155) of foraging bouts in 2013 and 2015, respectively. For bumblebee individuals, interspecific moves accounted for only 0.7% (50/7195) of the total pollinator moves in 2013 and 1.28% (38/2971) in In one pair, P. tricolor vs P. confertiflora in 2013, interspecific bouts (37.9%) and moves (6.7%) were relatively more frequent (Table 1). The CI calculated per bumblebee individual was high across all Pedicularis species and bumblebee species over 2 yr, ranging from to 1 (Table 1). Most bumblebees visited only one species when foraging in the quadrats. There were nine scores Table 1 Numbers of conspecific and interspecific moves by three bumblebee species in two Pedicularis species co-flowering patches over 2 yr No. of floral visits Pedicularis species pair Bombus Year Interspecific/total bouts a?a b?b a?b b?a CI longiflora vs cephalantha friseanus / / festivus / / siphonantha vs densispica friseanus / / festivus / trifasciatus / cephalantha vs densispica friseanus / / festivus / / trifasciatus / tricolor vs confertiflora friseanus / / Interspecific moves/total foraging bouts and numbers of moves within the patches for each bumblebee species are presented in the upper line for 2013, in the second line for Gegear s constancy index (CI, mean SE) is shown for each species pair (a vs b). Dashes indicate cases where no pollinator visits were observed. In each species pair the long-styled species ( a ) is listed first.

5 1456 Research New Phytologist of complete constancy (CI = 1) and only two scores of CI below 0.9, indicating that bumblebees exhibited constant behavior when visiting the Pedicularis species. Heterospecific pollen transfer The percentage of stigmas with pollen ranged from 5.0% to 95.5% (Fig. 2a), with no significant difference between species in the pair over 2 yr (v 2 -test of independence) except for one pair in 2015, in which only five of 60 stigmas of P. siphonantha had received pollen, much lower than the 41 stigmas of its counterpart P. densispica (v 2 = , df = 1, P < 0.001). Among those pollinated stigmas, overall the percentage of stigmas with heterospecific pollen was 12.4% 3.5% (mean SE) in 2013 and 13.7% 7.9% in In three of the four pairs, species a, with longer styles, had more stigmas with HP (v 2 -test for independence, all P < 0.05), but in the pair P. tricolor and P. confertiflora the species did not differ in the percentage of stigmas with HP in 2013 and no HP grains were deposited on either species in 2015 (Fig. 2b; Table 2). The number of HP grains received per stigma ranged from 0 to in 2013 and from 0 to in The proportion of HP grains in the total pollen load per stigma in each species ranged from 0 to 7.5% 3.7% in 2013 and from 0 to 24.2% 19.1% in 2015 (Fig. 2c). Paired with P. densispica in 2015, P. siphonantha received the highest percentage of HP per stigma (24.2% 19.1%, n = 5), and the highest amount of HP per stigma ( ), but only five stigmas received pollen (Table 2). The percentages of HP received per stigma were significantly higher in species with longer styles than in the shorterstyled species in three pairs over 2 yr (GLM, all P < 0.05), except for the pair P. tricolor and P. confertiflora in which there was no significant difference in 2013 and no HP at all deposited in 2015 (Fig. 2c). In short-styled P. cephalantha, little or no heterospecific pollen was observed when it was paired with long-styled P. longiflora for 2 yr, but HP was relatively more frequent when it was paired with P. densispica, the species with the shortest pistil among the studied species. When P. densispica was paired with P. cephalantha or P. siphonantha, very few or none of its stigmas received HP, and only in small amounts. Fig. 2 Percentages of pollinated stigmas (a), stigmas with heterospecific pollen (HP) (b) and heterospecific pollen number grain divided by total pollen load per stigma (c) in the four pairs in 2013 and 2015 (mean SE). One species in each pair (black bars) has a longer corolla tube and style than the other (white bars). Numbers of observed stigmas from each species in the pair are shown on the bar. *, Significant differences between the species in a pair. Species abbreviations: Lo, P. longiflora; Ce, P. cephalantha; Si, P. siphonantha; De, P. densispica; Tr, P. tricolor; Co, P. confertiflora.

6 New Phytologist Research 1457 Table 2 Conspecific (upper line) and heterospecific (second line) pollen receipt per stigma (mean SE) for 2 yr within each species pair, and percentage of pollen tubes going all the way through the styles from conspecific and heterospecific pollen donors between long-styled ( a, listed first) and short-styled ( b ) Pedicularis species in the four species pairs Species pair (a vs b) No. stigmas No. pollen No. stigmas No. pollen Pollination treatment No. styles Tubes passed through stigma Tubes passed through style % longiflora vs cephalantha Conspecific Heterospecific Conspecific Heterospecific siphonantha vs densispica Conspecific Heterospecific Conspecific Heterospecific cephalantha vs densispica Conspecific Heterospecific Conspecific Heterospecific tricolor vs confertiflora Conspecific Heterospecific Conspecific Heterospecific The numbers of stigmas with pollen deposited naturally and styles after hand pollination treatments are shown. Heterospecific pollen from some longstyled species could go through the style of the short-styled species (data in bold). The relationships of CP and HP receipt in all pairs across 2 yr were neither linear nor nonlinear (curve estimation, all P > 0.05), meaning that HP deposition in our study varied independently of CP deposition. Total pollen load per stigma (as well as CP load) was positively correlated with bumblebee visits for each Pedicularis species in the pair, both in 2013 (n = 8, r = 0.881, P = 0.004) and 2015 (n = 8, r = 0.838, P = 0.009), indicating that stigmatic pollen deposition depended on pollinator visits. However, the number or percentage of interspecific moves was not correlated with the percentage of stigmas with HP (Fig. 2b) either in 2013 (P = or 0.069) or in 2015 (P = or P = 0.642). The number of interspecific moves was not correlated with the number of HP grains received per stigma (Fig. 2c) either in 2013 (P = 0.116) or in 2015 (P = 0.886). These results indicate that the frequency and amount of HP deposition could not be deduced from pollinator observations in these Pedicularis species. The GLM test of temporal variation in CP and HP receipt between 2013 and 2015 showed that conspecific pollination was highly dynamic but HPT was less variable within the four Pedicularis species pairs (Table S4). The GLM test of the effects of year and style length on HPT showed that year did not affect HPT (v 2 = 0.375, df = 1, P = 0.540), but style length did (v 2 = , df = 1, P < 0.001), indicating that species with longer styles in the pair were likely to experience a relatively greater interference by HP. higher percentages of germinated CP could grow through the style and enter the ovary (GLM, all P < 0.05). Consistently, across all pairs, no pollen grain from donor species with shorter styles could grow through the longer styles of species a in 24 h. By contrast, 16.5% 5.8% to 76.0% 3.5% of germinated pollen from species a with longer styles could go through the styles of species b in three of the pairs, but in one pair, pollen from long-styled P. longiflora did not form tubes going through the short styles of P. cephalantha (Table 2). Pollen pistil interactions Both CP and HP could germinate on stigmas of each Pedicularis species, but the degree of further growth of pollen tubes varied (Table 2; Fig. 3). Compared to HP across all pairs, significantly Fig. 3 Maximum pollen tube length (mean SE) in heterospecific styles of the four species-pairs involving six Pedicularis species. White bars, length of pistils; black bars, maximum growth of heterospecific pollen tubes. Six Pedicularis species in four pairs as pollen recipients are listed on the horizontal axis. Species pairs are separated by a dotted-line.

7 1458 Research New Phytologist The maximum length of pollen tubes in heterospecific pistils showed a consistent pattern: pollen from short-styled species could not grow through the longer styles of species in 24 h although it varied between pairs and species. In only two pairs, both involving P. densispica which had the shortest styles, heterospecific pollen from P. siphonantha and P. cephalantha formed tubes through the styles of P. densispica (Fig. 3). By contrast, pollen tubes from P. densispica only grew a very short distance (< 4 mm) in heterospecific pistils. Overall in these pairs, pollen tubes from a short-styled species could grow a maximum distance of < 40% of the style length in a long-styled species (Fig. 3). The GLM test of the effects of style length on HP growth in the style showed that style length could affect HP growth (v 2 = , df = 1, P < 0.001). Discussion With shared bumblebee pollinators, high flower constancy was observed across nearly all studied Pedicularis pairs in the field over 2 yr and consequently interspecific pollen flow was limited. This suggests that pre-pollination isolation by flower constancy contributed more than post-pollination barriers in minimizing reproductive interference in these species. Interestingly, we found that heterospecific pollen transfer (HPT) was generally low and asymmetrical, species with long styles experiencing higher heterospecific pollen (HP) deposition than species with short styles in three of the four pairs. Although HP receipt was relatively higher in long-styled species, the long styles could effectively filter HP during the pollen pistil interaction. Flower constancy Our pollinator observations on the four pairs of Pedicularis species showed that bumblebees showed high flower constancy and that interspecific moves were rare in the field. The bees generally preferred visiting one species over the other in the quadrats. One may speculate that bumblebees prefer nectariferous species over nectarless species. When nectariferous P. densispica was paired with nectarless P. cephalantha or P. siphonantha, the visit frequency by the three bumblebee species was significantly higher to the nectariferous species except for the preference of B. festivus in the pair P. cephalantha vs P. densispica (Table 1). The status of provisions in the nest may determine whether a given bumblebee worker collects pollen or nectar. When Rhinanthus minor and R. angustifolius were present in similar quantities, bumblebees generally preferred the more rewarding and conspicuous species, but B. pascuorum had a preference for less rewarding R. minor when the other species went for R. angustifolius (Natalis & Wesselingh, 2013). Natural hybridization can occasionally be found between the two Rhinanthus species in Europe where HPT was reduced mainly through bumblebee constancy (which was highly sensitive to site characteristics), rather than through mechanical isolation because pollen from the two species was placed on similar sites on the bumblebees bodies (Natalis & Wesselingh, 2012). Great flower constancy in sympatric Pedicularis species may be attributed to the interaction between the complexity of flower form and bumblebee behavior. However, interspecific moves may increase the bumblebees learning time when extracting pollen and manipulating the galea, involving an extra cost of handling time on a new flowering species (Stout et al., 1998; Gegear & Laverty, 2005; Corbet & Huang, 2014). Pedicularis flowers have various distinguishing floral characters between species, such as twisted protruding beak structures, and variation in flower color pattern and in the shape of the labellum (Fig. S1). Different floral forms need different actions by the bumblebee to collect pollen, particularly in nectarless species (Macior, 1982; Macior et al., 2001; Corbet & Huang, 2014). Interspecific differences in flower color and shape can limit reproductive interference, as shown in a large-scale survey of Pedicularis species assembly into natural communities in southwest China (Eaton et al., 2012). Heterospecific pollen transfer Previous studies have used pollen dye to trace HPT (see Goulson & Jerrim, 1997; Kay, 2006; Brys et al., 2014) or set up emasculated flowers of one species surrounded by pollen donors from the other species to estimate HP (Natalis & Wesselingh, 2012). Using emasculated flowers to evaluate HP, Natalis & Wesselingh (2012) found that interspecific stigmatic pollen deposition was similar in both directions between R. minor and R. angustifolius. Although these two approaches aimed to estimate HPT, conspecific pollen (CP) transfer could not be quantified. To examine whether divergence in the position of the reproductive parts of coexisting plants affects HPT, Muchhala & Potts (2007) conducted flight cage experiments involving five bat-pollinated Burmeistera species. Species pairs of Burmeistera were chosen in which pollen was identifiable to the species level (Muchhala & Potts, 2007). Their results demonstrated that HPT was less when paired species had larger difference in exsertion lengths (a measure of the exsertion of both the anthers and the stigma) from the focal species. Differences in pollen size between Pedicularis species pairs permitted us to examine interspecific and conspecific pollen deposition on stigmas among co-flowering species in natural communities, directly quantifying HP and its relationship to pollinator moves. Our examination of HPT showed that the proportions of stigmas with HP and the amounts of HP per stigma were variable but both had the same asymmetrical pattern within species pairs in the field over 2 yr: species with shorter styles generally experienced less HP deposition. One may predict that high HP receipt is caused by a high frequency of pollinator moves. We observed that CP on stigmas and total pollen loads were correlated with pollinator moves in the studied pairs. However, neither the proportion of stigmas with HP nor the number of HP grains per stigma was related to interspecific moves by bumblebees. Actually a negative relationship between the number of HP grains per stigma and interspecific moves was observed in an alpine meadow involving 14 co-flowering species, illustrating an adaptive strategy for minimizing HPT (Fang & Huang, 2016). These results suggested that interspecific moves did not always

8 New Phytologist Research 1459 result in HP deposition. Our observations of temporal variation in the relative proportion of HP and CP per stigma between Pedicularis species (Table S4) were consistent with a 3-yr study of stigmatic pollen loads in over 30 co-flowering species in the same field station, which showed that species generally experienced a similar magnitude of HP interference (Gao, 2015). The low proportion of Pedicularis stigmas with HP may directly correspond to the effect of the observed high constancy; however, HP grains seldom exceeded 10% of the stigmatic pollen load. The predicted positive relationship between pollinator moves and the magnitude of HP was not observed here, which could be attributable to differential pollen placement and stigma contact position among species (see Huang & Shi, 2013). This observation was coincident with no correlation between CP and HP deposition in each species over 2 yr. An analysis of 19 species showed that HP was often positively or negatively related to CP deposition (Arceo-Gomez et al., 2016a), except in one species Sideritis incana in which the number of HP grains was 1 or 2. The very low numbers of HP grains are similar to what we observed in Pedicularis stigmas (Table 2). The low numbers of HP grains and interspecific moves may have weakened the power to detect a potential correlation. Thus, HPT between Pedicularis species was greatly reduced through high constancy along with the interspecific segregation of stigma contact from pollen placement position on the bumblebee body, see Huang & Shi, 2013; Armbruster et al., 2014). Compared to previous investigations on floral isolation, in which the relative contribution of ethological and mechanical isolation was not examined, our direct measurements in natural communities suggested that flower constancy played a key role in reducing the magnitude of HPT. Heterospecific pollen pistil interactions A pollen grain needs to interact with a stigma to germinate and, later, the secretions of the pistil will promote or reduce its rate of tube growth (Skogsmyr & Lankinen, 2002). Although high constancy was shown by pollinators in foraging bouts on Pedicularis, our examination of stigmas still revealed HPT between pairs of Pedicularis species (Fig. 2), suggesting that pollen of co-flowering Pedicularis species could experience a race in the style. Pollen from long-styled species could usually grow through the pistils of short-styled species, but conversely, pollen from short-styled species was less likely to grow through the styles of long-styled species (Table 2; Fig. 3). Therefore, short-styled species were always disadvantaged in the post-pollination stage. Another project of ours examining ecological adaptation of long styles in Pedicularis showed that interference of HP depressed seed set more seriously in short- than in long-styled species (Mao, 2010; X-P. Wang & S-Q. Huang, unpublished). Because long-tongued pollinators were not observed in Himalayan Pedicularis species, selective factors acting on elongate corolla tubes which have evolved multiple times in this genus remain mysterious (Ree, 2005; Huang et al., 2016). If long styles are advantageous in filtering heterospecifc pollen than short styles, then the correlational traits of long tubes would be favored under interspecific pollination competition. Our analyses showed that Pedicularis species with shorter styles experienced lower interspecific pollen loads on the stigmas when they coexisted with long-styled species over 2 yr (see Fig. 2), whereas long styles would benefit more from factors that would reduce the detrimental effects of HPT. Long-styled Pedicularis species usually had a larger stigma surface (Yang et al., 2002), giving a greater probability of receiving both CP and interspecific pollen (Montgomery & Rathcke, 2012; Table 2), suggesting that style length and stigma size might provide targets of selection for tolerance of HPT alternatives to the traits that are involved in avoidance responses to HPT (Ashman & Arceo-Gomez, 2013). A recent study involving two Clarkia species observed that CP tube success in C. xantiana was greater in plants from populations with previous exposure to the HP donor species than in plants from populations without exposure, suggesting that a greater tolerance to HP effects might have evolved in mixed-species communities (Arceo-Gomez et al., 2016b). We found that species with short styles are more vulnerable to pollen pistil interactions, whereas in the three long-styled species, pollen tubes from other species could hardly grow through the styles (Fig. 3). Such species differentiation in style length is mirrored in several species-rich genera including Prunus (Perez & Moore, 1985), Rhododendron (Williams & Rouse, 1988) and Nicotiana (Lee et al., 2008), in which species have disparate style lengths and co-flowering plants may have overlapping pollinators, potentially experiencing HPT. One would expect that selection for reducing the deleterious effects of HPT would be stronger in short-styled species than in the long-styled species (Nista et al., 2015). Indeed, we found that HPT was generally lower in the short- than the long-styled Pedicularis species in natural conditions (Fig. 2b). Such asymmetry of directionality of pollen flow has occasionally been documented in plants (Lowry et al., 2008). Further studies are needed to verify this pattern, particularly in congeners with large variation in style length. Our measurements of HP receipt indicated a low degree of HPT between co-flowering Pedicularis species in natural communities. Although these species shared bumblebee pollinators, high flower constancy greatly contributed to a limited HPT. This would support the idea that quantification of pollen loads on stigmas is more accurate than direct pollinator observation in understanding plant pollinator interactions. We did not observe the predicted correlations between CP and HP deposition, or between interspecific moves and HP receipt, implying that an interspecific mechanical misfit between flower and pollinator morphology may further decrease HPT. This conclusion needs further attention, given the limited samples, notwithstanding that counting pollen on stigmas involves enormous effort. The role of pollen pistil interactions in illuminating the deleterious effects of HP is largely underappreciated (Ashman & Arceo-Gomez, 2013). Here we have shown that Pedicularis species with short styles were highly vulnerable to interference by pollen tubes from HP, but were better able to avoid HP deposition than the species with long styles.

9 Acknowledgements We thank: J. Gao, Y-Y. Mao, X-P. Wang, X-Y. Wang, L-Y. Wu and B-Y. Ye for assistance in the field; Z-D. Fang and staff of the Shangri-La Alpine Botanical Garden for logistical support; S. Armbruster, S. Corbet, Q. Fang, R. A. Wesselingh and two anonymous referees for providing valuable comments; and S. Corbet for correcting the English in the manuscript. This work was supported by the National Science Foundation of China (grants nos , ) to S-Q.H. Author contribution S-Q.H. and Z-Y.T. planned and designed the research. Z-Y.T. performed experiments and conducted fieldwork, Z-Y.T. and S-Q.H. analyzed the data, Z-Y.T. and S-Q.H. wrote the manuscript. References Arceo-Gomez G, Abdala-Roberts L, Jankowiak A, Kohler C, Meindl GA, Navarro-Fernandez CM, Parra-Tabla V, Ashman T-L, Alonso C. 2016a. Patterns of among- and within-species variation in heterospecific pollen receipt: the importance of ecological generalization. American Journal of Botany 103: Arceo-Gomez G, Raguso RA, Geber MA. 2016b. Can plants evolve tolerance mechanisms to heterospecific pollen effects? An experimental test of the adaptive potential in Clarkia species Oikos 125: Armbruster WS Floral specialization and angiosperm diversity: phenotypic divergence, fitness trade-offs and realized pollination accuracy. AoB Plants 6: plu003. Armbruster WS, Edwards ME, Debevec EM Floral character displacement generates assemblage structure of Western Australian triggerplants (Stylidium). Ecology 75: Armbruster WS, Shi X-Q, Huang S-Q Do specialized flowers promote reproductive isolation? Realized pollination accuracy of three sympatric Pedicularis species. Annals of Botany 113: Ashman TL, Arceo-Gomez G Toward a predictive understanding of the fitness cost of heterospecific pollen receipt and its importance in co-flowering communities. American Journal of Botany 100: Brys R, Broeck AV, Mergeay J, Jacquemyn H The contribution of mating system variation to reproductive isolation in two closely related Centaurium species (Gentianaceae) with a generalized flower morphology. Evolution 68: Chari J, Wilson P Factors limiting hybridization between Penstemon spectabilis and Penstemon centranthifolius (Scrophulariaceae). Canadian Journal of Botany 79: Corbet SA, Huang S-Q Buzz pollination in eight bumblebee-pollinated Pedicularis species: does it involve vibration-induced triboelectric charging of pollen grains? Annals of Botany 114: Eaton DAR, Fenster CB, Hereford J, Huang S-Q, Ree RH Floral diversity and community structure in Pedicularis (Orobanchaceae). Ecology 93: S182 S194. Fang Q, Huang S-Q A directed network analysis of heterospecific pollen transfer in a biodiverse community. Ecology 94: Fang Q, Huang S-Q A paradoxical mismatch between interspecific pollinator moves and heterospecific pollen receipt in a natural community. Ecology. doi: /ecy Flanagan RJ, Mitchell JR, Knutowski D, Karron JD Interspecific pollinator movements reduce pollen deposition and seed production in Mimulus ringens (Phrymaceae). American Journal of Botany 96: Gao J On dynamics of heterospecific pollen transfer among sympatric plants in an alpine meadow. PhD thesis, Wuhan University, Wuhan, China. Gegear RJ, Laverty TM Flower constancy in bumblebees: a test of the trait variability hypothesis. Animal Behaviour 69: Gegear RJ, Thomson JD Does the flower constancy of bumble bees reflect foraging economics? Ethology 110: Goulson D, Jerrim K Maintenance of the species boundary between Silene dioica and S. latifolia (red and white Campion). Oikos 79: Grant V. 1994a. Mechanical and ethological isolation between Pedicularis groenlandica and P. attollens (Scrophulariaceae). Biologisches Zentralblatt 113: Grant V. 1994b. Modes and origins of mechanical and ethological isolation in angiosperms. Proceedings of the National Academy of Sciences, USA 91: Huang S-Q, Fenster CB Absence of long-proboscid pollinators for longcorolla-tubed Himalayan Pedicularis species: implications for the evolution of corolla length. International Journal of Plant Sciences 168: Huang S-Q, Shi X-Q Floral isolation in Pedicularis: how do congeners with shared pollinators minimize reproductive interference? New Phytologist 199: Huang S-Q, Wang XP, Sun S-G Are long corolla tubes in Pedicularis driven by pollinator selection? Journal of Integrative Plant Biology. doi: /jipb Johnson SD The pollination niche and its role in the diversification and maintenance of the southern African flora. Philosophical Transactions of the Royal Society B 365: Kawai Y, Kudo G Effectiveness of buzz pollination in Pedicularis chamissonis: significance of multiple visits by bumblebees. Ecological Research 24: Kay KM Reproductive isolation between two closely related hummingbird pollinated neotropical gingers. Evolution 60: Lee CB, Page LE, McClure BA, Holtsford TP Post-pollination hybridization barriers in Nicotiana section Alatae. Sexual Plant Reproduction 21: Lowry DB, Modliszewski JL, Wright KM, Wu CA, Willis JH The strength and genetic basis of reproductive isolating barriers in flowering plants. Philosophical Transactions of the Royal Society B 363: Macior LW Plant community and pollinator dynamics in the evolution of pollination mechanisms in Pedicularis (Scrophulariaceae). In: Armstrong JA, Powell JM, Richards AJ, eds. Pollination and evolution. Sydney, NSW, Australia: Royal Botanic Gardens, Macior LW, Tang Y, Zhang J-C Reproductive biology of Pedicularis (Scrophulariaceae) in the Sichuan Himalaya. Plant Species Biology 16: Mao Y-Y Evolutionary adaptation of long styles in Pedicularis flowers. PhD thesis, Wuhan University, Wuhan, China. Montgomery BR, Rathcke BJ Effects of floral restrictiveness and stigma size on heterospecific pollen receipt in a prairie community. Oecologia 168: Morales C, Traveset A Interspecific pollen transfer: magnitude, prevalence and consequences for plant fitness. Critical Reviews in Plant Sciences 27: Muchhala N, Potts MD Character displacement among bat-pollinated flowers of the genus Burmeistera: analysis of mechanism, process and pattern. Proceedings of the Royal Society B 274: Natalis LC, Wesselingh RA Shared pollinators and pollen transfer dynamics in two hybridizing species Rhinanthus minor and R. angustifolius. Oecologia 170: Natalis LC, Wesselingh RA Parental frequencies and spatial configuration shape bumblebee behavior and floral isolation in hybridizing Rhinanthus. Evolution 67: Nista P, Brothers AN, Delph LF Differences in style length confer prezygotic isolation between two dioecious species of Silene in sympatry. Ecology and Evolution 5: Perez S, Moore JN Prezygotic endogenous barriers to interspecific hybridization in Prunus. Journal of the American Society for Horticultural Science 110: Ree RH Phylogeny and the evolution of floral diversity in Pedicularis (Orobanchaceae). International Journal of Plant Sciences 166: Schaal BA Measurement of gene flow in Lupinus texensis. Nature 284: