Geographic variation of reproductive traits and competition for pollinators in a bird‐pollinated plant

Abstract Geographic variation in the reproductive traits of animal‐pollinated plants can be shaped by spatially variable selection imposed by differences in the local pollination environment. We investigated this process in Babiana ringens (Iridaceae), an enigmatic species from the Western Cape region of South Africa. B. ringens has evolved a specialized perch facilitating cross‐pollination by sunbirds and displays striking geographic variation in perch size and floral traits. Here, we investigate whether this variation can be explained by geographic differences in the pollinator communities. We measured floral and inflorescence traits, and abiotic variables (N, P, C, and rainfall) and made observations of sunbirds in populations spanning the range of B. ringens. In each population, we recorded sunbird species identity and measured visitation rates, interfloral pollen transfer, and whether the seed set of flowers was pollen limited. To evaluate whether competition from co‐occurring sunbird‐pollinated species might reduce visitation, we quantified nectar rewards in B. ringens and of other co‐flowering bird‐pollinated species in local communities in which populations occurred. Variation in abiotic variables was not associated with geographical variation of traits in B. ringens. Malachite sunbirds were the dominant visitor (97% of visits) and populations with larger‐sized traits exhibited higher visitation rates, more between‐flower pollen transfer and set more seed. No sunbirds were observed in four populations, all with smaller‐sized traits. Sunbird visitation to B. ringens was not associated with local sunbird activity in communities, but sunbird visitation was negatively associated with the amount of B. ringens sugar relative to the availability of alternative nectar sources. Our study provides evidence that B. ringens populations with larger floral traits are visited more frequently by sunbirds, and we propose that visitation rates to B. ringens may be influenced, in part, by competition with other sunbird‐pollinated species.


| INTRODUC TI ON
Much of the striking floral variation within and among angiosperms has been attributed to evolutionary responses to variation in the pollinator environment (Johnson, 2010;Kay & Sargent, 2009).
Geographic variation in the pollinator landscape can potentially drive mosaics in selection regimes and floral phenotype (Herrera, Castellanos, & Medrano, 2006;Newman, Manning, & Anderson, 2015;Paudel et al., 2016). For example, differences in the relative abundance of pollinators (Anderson, Alexandersson, & Johnson, 2010;Boberg et al., 2014;van der Niet, Pirie, Shuttleworth, Johnson, & Midgley, 2014), as well as morphology (Anderson & Johnson, 2008) and preference differences in a single pollinator species (Newman, Anderson, & Johnson, 2012) may select for distinct floral ecotypes adapted to different pollinators across the geographical range of a species. Plant responses to mosaics in the quantity and quality of pollinators may include morphological adaptations to the composition of different pollinator communities (e.g., Anderson, Ros, Wiese, & Ellis, 2014;Newman et al., 2015). In addition, evolutionary responses may include changes in mating system associated with variation in the abundance of pollinators (i.e., degree of pollinator limitation; Barrett & Husband, 1990;Eckert, Samis, & Dart, 2006;Lloyd, 1979), although other adaptive responses can also occur (Harder & Aizen, 2010).
This transition is associated with changes to a suite of floral traits, including reduced flower size and stigma-anther separation (herkogamy), greater autonomous self-pollination, less attractive rewards, and lower pollen-to-ovule ratios (e.g., Lloyd, 1965;Morgan & Barrett, 1989;Sicard & Lenhard, 2011). Although shifts to selfing have been documented in numerous angiosperm genera through comparative studies (Stebbins, 1974), empirical studies of the ecological context promoting the evolution of selfing are less well-investigated (Barrett & Harder, 2017;Levin, 2012).
A mechanism proposed for the evolution of increased selfing concerns competition among co-flowering species for shared pollinators (Campbell, 1985;Levin, 1972). In plant communities where some species are more abundant, or offer superior floral rewards, pollinator service to less rewarding species may be compromised resulting in conditions favoring increased selfing. Although array experiments have demonstrated that interspecific competition between co-flowering animal-pollinated species can lower seed production and outcrossing rates (Bell, Karron, & Mitchell, 2005), the extent to which similar processes operate in plant communities with contrasting mating systems is poorly understood (but see Briscoe Runquist, Grossenbacher, Porter, Kay, & Smith, 2016). The relation between pollinator sharing and visitation is further complicated because the co-occurrence of species may also facilitate increased visitation rates (Moeller, 2004;Thomson, 1978). In addition, the influence of the abiotic environment on resource availability may also indirectly affect visitation rates by changing floral rewards and attractiveness (Carroll, Pallardy, & Galen, 2001;Galen, 2000).  Figure 1; Video S1; Anderson, Cole, & Barrett, 2005;. Perch removal results in F I G U R E 1 Malachite sunbirds using the perch of Babiana ringens while foraging for nectar. (a) Female malachite sunbird leans over the reproductive organs of the flower as it reaches toward the tubular flowers of B. ringens situated on the ground. To obtain nectar, the bird must fully insert its bill into the floral tube, at which point the bird's chest will brush against the anthers and stigmas of the inflorescence. (b) Male malachite sunbird uses the perch to gain some elevation from the ground where it is able to call and survey his surroundings, which it does regularly between probing bouts on an inflorescence (also Video S1) reduced visitation rates by sunbirds and reduced seed set and quality (Anderson et al., 2005). This remarkably specialized adaptation for sunbird pollination occurs in only two species of Babiana, the other being the very localized B. avicularis Goldblatt & J.C. Manning (de Waal, Anderson, et al., 2012).
de Waal,  documented striking variation in B. ringens perch size, with population averages ranging from 92.1 to 216.2 mm in the Western Cape ( Figure 2). Variation in perch size was positively correlated with flower size, floral-tube length, and the degree of herkogamy (de Waal, , suggesting the correlated evolution of these traits. Geerts and Pauw (2009) Table S1 populations with robust perches, large flowers, and long floral tubes would be visited primarily by large, long-billed malachite sunbirds, Nectarinia famosa Linnaeus. In contrast, we predicted that populations with smaller floral and inflorescence traits would be visited primarily by small sunbird species with short bills such as the southern double-collared sunbird, Cinnyris chalybeus Linnaeus, and orangebreasted sunbird, Anthobaphes violacea Linnaeus. We

| Study system
Babiana ringens flowers from August to October and has characteristics which fit the sunbird pollination syndrome (Figure 1), including red flowers with long floral tubes, a perch for birds to forage from, large volumes of dilute nectar, and an absence of scent (Anderson et al., 2005;Van der Pijl, 1961). Babiana ringens is one of the four sunbird-pollinated species in this genus of approximately 92 species (Goldblatt & Manning, 2007a, 2007b de Waal, . The striking variation in flower and perch size has led to an east-west division into two subspecies:  , and marker gene studies have established that the species has a mixed mating system with low to moderate frequencies (t = 0.25-0.55) of outcrossing (Anderson et al., 2005;. Outcrossing rates could only be measured in populations with larger floral and perch phenotypes as no polymorphism at allozyme loci was detected in populations with smaller phenotypes. This probably reflects high selfing rates in these populations .

| Are there associations between variation in vegetative and floral traits?
During August to October 2009, 2014 and 2015, we made morphological measurements in situ from 13 flowering populations of B. ringens and obtained additional measurements from another five populations based on herbarium specimens (see Table S1 for localities, measurement, and details of sample sizes). The measurements included perch length, dorsal tepal length, corolla length, length of the longest leaf, and stigma-anther separation. We also recorded the total number of open flowers and buds on each inflorescence (hereafter flowers per inflorescence). In 11 populations, we examined whether perch length in each population covaried with floral trait measurements.

| Are there associations between Babiana ringens trait variation and abiotic factors?
We collected five soil samples (one kg each) from eight B. ringens populations (2, 3, 4, 12, 13, 14, 15, and 16) spanning the phenotypic variation within the species. The soil was removed from a depth of approximately 25 cm (the average depth of B. ringens bulbs). We determined total N and C content of soil through total combustion using a Leco Truspec ® CN N analyser, by means of the Walkley-Black method (Non-Affiliated Soil Analysis Work Committee, 1990), and total phosphate was determined by a method adapted from that described by Sommers and Nelson (1972). We extracted phosphate from soil by acid digestion using a 1:1 mixture of 1 N nitric acid and hydrochloric acid at 80°C for 30 min. We then determined the P concentration in the extract with a Varian ICP-OES optical emission spectrometer. We obtained the historical mean monthly rainfall from the WorldClim data set (Hijmans, Cameron, Parra, Jones, & Jarvis, 2005) with a resolution of 1 km 2 for each population. The monthly means were summed to give a mean annual rainfall for each of the 11 populations (1, 2, 3, 4, 6, 12, 13, 14, 15, 16, and 18).
To determine whether abiotic factors were associated with size variation in B. ringens traits, we conducted regression analyses explaining perch length with % N, % C, phosphate content, and mean annual rainfall as explanatory variables. Because the measurements that we made for inflorescence and floral traits were highly correlated (see Results), we used perch length as a surrogate for corolla length, dorsal tepal length, display size, stigma-anther separation, and the longest leaf. We chose perch length because previous work demonstrated the importance of the perch on visitation and outcrossing rates (see Anderson et al., 2005;).

| Is geographic variation in reproductive traits associated with visitation by different sunbird species?
In 11 populations, we recorded sunbird visitors to B. ringens flowers. We chose populations distributed across the entire range of the species, spanning the phenotypic variation in traits of interest. To determine whether trait variation was associated with particular sunbird species, we conducted a total of 60 hr of pollinator observations over 1-5 days per site (mean = 2.9 days), with a minimum of 30 min of observation each day in fine weather (Table S2)

| Is geographic variation in visitation rates and reproductive traits associated with variation in seed set, pollen limitation, and pollen transfer?
To assess the effect of sunbird visitation on a female fitness com-  Table S1 for sample sizes in each population). We marked unmanipulated buds and these were left undisturbed, whereas receptive stigmas of pollen-supplemented flowers received pollen from two pollen donors by removing anthers from donor flowers and rubbing them on all stigma lobes of recipient flowers. Donors were located more than two meters from recipients. Using regression, we then determined whether the severity of pollen limited seed set decreased with increasing sunbird visitation rate across populations.
Although bird visitation rates may not affect seed set in species where 0 indicates no interfloral pollen transfer and 1 shows complete interfloral pollen transfer. We conducted all three treatments (unmanipulated, emasculated, and pollen-supplemented flowers) on each replicate plant to limit the influence of between-plant habitat and plant condition effects. Although our measure of "interfloral pollen transfer" does not distinguish between outcross and geitonogamous components of pollination, it is probable that outcrossing in a population is positively affected by the amount of pollen dispersed among flowers.

| Is sunbird activity and relative sugar availability in communities positively associated with visitation rate to Babiana ringens?
To determine whether sunbird activity in plant communities was associated with visitation rates to B. ringens, we collected additional data on sunbird activity independent of our visitor observations, described above. This involved an additional 20.7 hr (mean = 1.9 hr, range = 90 -300 min per population) of observations of plant communities in which B. ringens populations occurred (Table S2). We recorded sunbirds when they were within the immediate observation area where B. ringens plants occurred and could be identified accurately (Table S2). If a sunbird perched on a different flowering plant species than B. ringens or on substrate, it was considered to be a new record, as it was not possible to keep track of individual sunbirds. We then determined by regression whether there was an association between overall sunbird activity in the community and visitation rates to B. ringens.
Because some B. ringens populations co-occurred with other species of bird-pollinated plants (Table S2), we assessed whether competition for pollinators may play a role in determining visitation rates.
In each B. ringens population, as described in more detail below, we  Table S2). We also estimated the For a subset of each bird-pollinated species at each site (see Table S2 for a list of bird-pollinated species at each site), we measured nectar volumes using a 5-µl microcapillary tube and concentrations (g sucrose per 100 g solution) using a handheld refractometer (Bellingham and Stanley, UK). We collected flowers for nectar measurements between 06hr00 and 07hr00 to reduce opportunities for birds to consume nectar. We calculated the total amount of sugar per flower by using a combination of nec-

| Is geographic variation in reproductive traits associated with visitation by different sunbird species?
During field observations of the 11 populations, a total of 110 sunbird visits were recorded to inflorescences of B. ringens. Of these,  Table S2).
Two visits from southern double-collared sunbirds were observed in population 14. Thus, there was no strong evidence to support the hypothesis that morphological differentiation in B. ringens is associated with visitation by contrasting sunbird species.

| Is sunbird activity and relative sugar availability in communities positively associated with visitation rate to Babiana ringens?
Observations of sunbirds visiting communities of bird-pollinated plants at B. ringens populations revealed that malachite sunbirds were present at all 11 sites although in four populations they were not observed visiting B. ringens. Site 2 had the lowest sunbird activity while site 6 had the highest, with 1.0 and 16.6 sightings per hour, respectively. Despite this variation in sunbird activity among F I G U R E 4 Relationship between visitation rate (visits/plant/hour) by sunbirds and (a) perch length (mm) (F 1,9 = 22.57, r 2 = .715, p = .001). Population 13 was visited by two groups of sunbirds, and the colors within the symbol represent the percent contribution of each sunbird group (small sunbird species and large malachite sunbirds) to the total visitation rate; (b) interfloral pollen transfer (F 1,9 = 8.40, r 2 = .48, p = .018) in 11 populations of B. ringens. The shaded area around the trend line indicates the 95% confidence interval, and the circle size represents the relative perch length in each population. Numbering follows those in Figure 3 and Table S1 sites, there was no evidence that sunbird activity in the community was significantly associated with sunbird visitation to B. ringens (F 1,9 = .140, r 2 = .154, p = .717). Similarly, the activity in communities with only malachite sunbirds was not associated with their visitation rate to B. ringens (F 1,9 = .004, r 2 < .001, p = .953).
In eight of the 11 B. ringens populations, other bird-pollinated species were co-flowering to varying degrees of abundance, and these species received many sunbird visits during our observation periods (Table S2). Babiana ringens populations in communities with relatively low quantities of alternative sugar sources for sunbirds received significantly higher visitation rates than populations in communities with relatively larger quantities of alternative sugar sources (F 1,9 = 5.534, r 2 = .381, p = .04, W 9 = 26.50; p = .013, Figure 6). The total standing crop of sugar in B. ringens populations with smaller perches did not differ from the total standing crops of sugar in populations with larger perches (W 9 = 3.00, p = .082). However, at the plant level, total sugar per plant tended to be lower in populations with small perches than in plants from populations with large perches (F 1,9 = 7.96, r 2 = .469, p = .020). Populations with smaller perches were associated with communities of other bird-pollinated plants that had significantly larger standing crops of sugar than populations with larger perches (W 9 = 25.00, p = .028, Figure S1).

Geographic variation in floral and inflorescence traits of B. ringens
is associated with variation in the visitation rates of pollinating sunbirds. However, there was no evidence that morphological variation among populations of B. ringens was associated with visitation by contrasting sunbird species (i.e., large vs. small sunbirds), as is typically predicted in models of pollinator-driven divergence through pollinator shifts (e.g., Grant & Grant, 1965;Johnson, 2010). Instead, populations of B. ringens with larger flowers and perches had significantly higher overall visitation rates by sunbirds than populations with smaller-sized traits. Below, we discuss how geographic variation in pollinator visitation may influence the reproductive F I G U R E 5 Relationship between interfloral pollen transfer and mean perch length (mm) (t 9 = 3.75, r = .781, p = .004, red) and stigma-anther separation (t 9 = 2.20, r = .592, p = .055, black) across 11 populations. Numbering follows those in Figure 3 and Table S1 F I G U R E 6 Differences in sunbird visitation rates to Babiana ringens in populations in which most of the sugar available to sunbirds was produced by B. ringens (high) versus populations in which most of the sugar was produced by other bird-pollinated species (low) (W 9 = 26.5; p = .013). The box margins indicate the upper and lower quartiles, the midline indicates the median, and the whiskers indicate the highest and lowest values excluding outliers (dots) success of B. ringens populations, how reproductive-trait variation could have arisen, and why pollinator visitation rates vary so greatly among populations.

| How might variation in pollinator visitation rates affect maternal fitness components?
Variation in visitation rates to animal-pollinated plants can affect maternal fitness components through two main avenues: seed quantity and quality. Insufficient pollinator service can reduce the quantity of seeds produced and may result in selection on floral traits. For example, low visitation can select for reproductive assurance to maintain maternal fertility resulting in reduced flower size and herkogamy (Barrett & Harder, 2017;Eckert et al., 2006;Lloyd, 1965;Opedal, 2018;Sicard & Lenhard, 2011). Although we found no evidence for a relation between pollinator visitation rate and seed production in our study, this is probably because of the capacity for autonomous seed production in B. ringens (de Waal, Anderson, et al., 2012, see Figure 5), which would have maintained some degree of maternal fertility despite limited sunbird visitation. The occurrence of reproductive assurance by autonomous self-pollination in species with specialized pollination systems, such as B. ringens, appears to be more widespread among animal-pollinated species than is often appreciated (Fenster & Marten-Rodriguez, 2007). Babiana ringens relies solely on sunbirds for cross-pollination making populations especially vulnerable to unreliable pollinator service, and this probably explains why populations have evolved the capacity for autonomous self-pollination.
Visitation rates may affect female fitness as a result of the quality of seeds that a plant produces and if plants are capable of autonomous selfing, as in B. ringens, low visitation rates can potentially lead to increased selfing rates. Indirect evidence for this hypothesis was that sunbird visitation rates were positively associated with the amount of pollen transferred between flowers. This should have promoted more outcrossing and higher seed quality. However, our measurements of pollen movement could not distinguish between transfer of pollen between flowers on the same plant (geitonogamous self-pollination) and pollen dispersal between plants (outcrossing) so this inference is tentative. Further work using genetic markers would be required to determine the quantitative relations between visitation rate and outcrossing rate and how much selfing arises from within-flower and between-flower self-pollination (see Eckert, 2002;Hu, Barrett, Zhang, & Liao, 2005;Schoen & Lloyd, 1992  . Low allozyme heterozygosity is generally associated with high rates of selfing (Hamrick & Godt, 1996), and this may have led to relaxed selection on perch characteristics (i.e., dissolution of perches) in these populations. Furthermore, investigations of related outcrossers and selfers commonly report that high selfing rates are associated with reduced vegetative, floral, and inflorescence traits among populations (Lloyd, 1965;Vallejo-Marín & Barrett, 2009), between related species (Armbruster et al., 2002;Goodwillie, 1999;Wyatt, 1984), and in phylogenies (Goodwillie et al., 2010). Reductions in floral traits (e.g., flower size, reward size, and stigma-anther separation) are often interpreted as adaptive responses to the evolution of selfing (Brunet & Eckert, 1998;Charlesworth & Charlesworth, 1981;Charnov, 1982) resulting from chronic pollinator limitation (e.g., Lloyd, 1965;Morgan & Barrett, 1989;Sicard & Lenhard, 2011).  (2018)  Another plausible explanation for geographic variation in B. ringens phenotype is that it is driven by features of the abiotic environment, either as plastic responses or as evolved responses to local conditions. However, we found no relations between the most limiting resources (N, P, rainfall) in the Cape (see Cramer, West, Power, Skelton, & Stock, 2014) and perch size. It is of course still possible that unmeasured abiotic factors may contribute to phenotypic variation. In addition, it is also unclear to what extent developmental constraints associated with allometry might also limit the independent evolution of different morphological traits (e.g., Niklas, 1994;West, Brown, & Enquist, 1999; but see Biddick, Hutton, & Burns, 2018).
Common garden studies would be useful to determine how much of the observed variation has a heritable basis, regardless of the specific ecological and evolutionary processes that may be responsible for the origin and maintenance of geographical differentiation.

| Why do pollinator visitation rates to Babiana ringens populations vary?
Our study demonstrated that pollinator visitation rates in B. ringens Although the idea that competition among plants for pollinators could provide a stimulus for the evolution of selfing was originally proposed by Levin (1972) over 45 years ago, it has received surprisingly limited empirical support from studies of natural populations.
This general lack of evidence is probably because of the challenge of experimentally disentangling the numerous community-level influences on pollinator visitation to animal-pollinated species and the diverse ecological factors affecting plant mating.