Interspecific Competition for Pollination Lowers Seed Production and Outcrossing in Mimulus Ringens

The University of Akron IdeaExchange@UAkron

Biology Faculty Research Biology Department

2005 Interspecific ompC etition for Pollination Lowers Seed Production and Outcrossing in Mimulus Ringens John M. Bell University of Wisconsin - Milwaukee

Jeffrey D. Karron University of Wisconsin - Milwaukee

Randall J. Mitchell University of Akron Main Campus, [email protected]

Please take a moment to share how this work helps you through this survey. Your feedback will be important as we plan further development of our repository. Follow this and additional works at: http://ideaexchange.uakron.edu/biology_ideas Part of the Biology Commons

Recommended Citation Bell, John M.; Karron, Jeffrey D.; and Mitchell, Randall J., "Interspecific ompeC tition for Pollination Lowers Seed Production and Outcrossing in Mimulus Ringens" (2005). Biology Faculty Research. 24. http://ideaexchange.uakron.edu/biology_ideas/24

This Article is brought to you for free and open access by Biology Department at IdeaExchange@UAkron, the institutional repository of The nivU ersity of Akron in Akron, Ohio, USA. It has been accepted for inclusion in Biology Faculty Research by an authorized administrator of IdeaExchange@UAkron. For more information, please contact [email protected], [email protected]. Ecology, 86(3), 2005, pp. 762±771 ᭧ 2005 by the Ecological Society of America

INTERSPECIFIC COMPETITION FOR POLLINATION LOWERS SEED PRODUCTION AND OUTCROSSING IN MIMULUS RINGENS

JOHN M. BELL,1,3 JEFFREY D. KARRON,1,4 AND RANDALL J. MITCHELL2 1Department of Biological Sciences, P.O. Box 413, University of Wisconsin-Milwaukee, Milwaukee, Wisconsin 53201 USA 2Department of Biology, University of Akron, Akron, Ohio 44325-3908 USA

Abstract. Sympatric plant species with similar ¯owering phenologies and ¯oral mor- phologies may compete for pollination, and as a consequence potentially in¯uence each other's reproductive success and mating system. Two likely competitors are Mimulus ringens and Lobelia siphilitica, which co-occur in wet meadows of central and eastern North Amer- ica, produce blue zygomorphic ¯owers, and share several species of bumble bee pollinators. To test for effects of competition for pollination, we planted experimental arrays of Mimulus ringens, each consisting of genets with unique combinations of homozygous marker genotypes. In two arrays we planted mixtures of Mimulus and Lobelia, and in two additional arrays we planted Mimulus without a competitor for pollination. Bumble bee pollinators frequently moved between Mimulus and Lobelia ¯owers in the mixed-species arrays, with 42% of plant-to-plant movements being interspeci®c transitions. Pollinator movements between species were associated with a reduction in the amount of conspeci®c pollen arriving on Mimulus stigmas. The presence of Lobelia led to a signi®cant 37% reduction in the mean number of Mimulus seeds per fruit. In addition, Mimulus had a signi®cantly lower rate of outcrossing in the mixed-species arrays (0.43) than in the ``pure'' arrays (0.63). This is the ®rst study to demonstrate that competition for pollination directly in- ¯uences outcrossing rates. Our work suggests that in self-compatible populations with genetic load, competition for pollination may not only reduce seed quantity, but may also lower seed quality. Key words: Bombus fervidus; competition for pollination; ®eld experiment; improper pollen transfer; Lobelia siphilitica; mating system; Mimulus ringens; outcrossing rate; seeds per fruit; seed set; pollen loss; visitation rate.

INTRODUCTION ospeci®c pollen is deposited on stigmas of one or both competitors (Rathcke 1983). Accumulation of foreign When two or more sympatric plant species have pollen on a recipient's stigma may interfere with fer- overlapping ¯owering phenologies, they may compete tilization of ovules by conspeci®c pollen (Waser 1978, for pollination (Levin and Anderson 1970, Schemske Rathcke 1983, Brown and Mitchell 2001). Heteros- et al. 1978, Waser 1978, Pleasants 1980, 1983, Camp- peci®c pollen deposition may also lead to pollen loss bell 1985a, Feinsinger et al. 1986, Caruso 1999, 2000). (pollen wastage), a reduction in the amount of pollen This competition may lower the quantity and quality deposited on conspeci®c stigmas (Waser 1983). Both of pollen deposited on conspeci®c stigmas (Harder and of these mechanisms of competition may affect plant Barrett 1996, Caruso 1999, Brown et al. 2002), and reproductive success and outcrossing rates by altering may reduce reproductive success and outcrossing rates the amount and type of pollen arriving on stigmas. (Campbell 1985a, b). The extent of competition for pollination largely de- Two potential mechanisms of competition for pol- pends upon patterns of pollinator foraging within a lination are pollinator preference and improper pollen population (Levin 1978, Campbell 1985a, Campbell transfer. Competition through pollinator preference oc- and Motten 1985, Feinsinger et al. 1986, Brown and curs when plant species B attracts pollinators away Mitchell 2001, Brown et al. 2002). Improper pollen from species A, reducing the reproductive success of transfer is much more likely when pollinators move species A (Levin and Anderson 1970, Waser 1978, 1983, Campbell 1985a, Campbell and Motten 1985, frequently between co-occurring species on a single Sih and Baltus 1987, Brown et al. 2002). Competition foraging bout (Grant 1950, Waser 1978, 1986). Camp- through improper pollen transfer occurs when heter- bell (1985a, b) demonstrated that frequent pollinator movements between co-occurring species may lead to substantial pollen loss. She hypothesized that in self- Manuscript received 22 April 2004; revised 23 August 2004; compatible species, this pollen loss would reduce the accepted 24 August 2004. Corresponding Editor: L. M. Wolfe. amount of pollen deposited on stigmas of other con- 3 Present address: Applied Ecological Services, Inc., 17921 Smith Road, Brodhead, Wisconsin 53520 USA. speci®c individuals, leading to a lower rate of out- 4 Corresponding author. E-mail: [email protected] crossing (Campbell 1985b). However, the effect of 762 March 2005 COMPETITION FOR POLLINATION IN MIMULUS 763 competition for pollination on outcrossing rates has not meadows in central and eastern North America. The yet been demonstrated empirically. azure-blue zygomorphic ¯owers are similar in size to We tested Campbell's outcrossing-rate hypothesis by Mimulus ¯owers, with typical Lobelia corolla tube quantifying the effects of Lobelia siphilitica, a com- length ϳ17 mm and corolla tube width ϳ6 mm (Caruso petitor for pollination, on outcrossing rates and seed et al. 2003b). Plants ¯ower from early August to early production in Mimulus ringens. These two blue-¯ow- October in southeastern Wisconsin. Flowers last for 5± ered perennials broadly overlap in ¯owering phenol- 6 days, and are bumble bee pollinated (Johnston 1991). ogy, and bumble bee pollinators move freely between Lobelia siphilitica is gynodioecious and the fre- them (Bell 2003), so there is strong potential for com- quency of females varies considerably among popu- petition. A rigorous test of Campbell's hypothesis re- lations (see references cited in Caruso et al. [2003a]). quires accounting for many other variables that can Some L. siphilitica populations lack females (Caruso potentially in¯uence plant mating systems, including et al. 2003a), and in our experimental arrays only her- ¯oral morphology (Karron et al. 1997), population den- maphroditic plants were present. The hermaphroditic sity (Karron et al. 1995a), and ¯oral display size (Kar- ¯owers are strongly protandrous. On each day of our ron et al. 2004). To control these variables, we planted experimental study, every Lobelia plant had both male experimental arrays of Mimulus ringens with constant phase and female phase ¯owers present. spacing and ¯oral display size. In two arrays we planted To assess whether these species co-occur and share Mimulus ringens without a competitor for pollination. pollinators in the wild, in the summer of 2002 we sur- In two other arrays we added a competitor for polli- veyed 23 natural populations of Mimulus ringens in nation by planting mixtures of Mimulus and Lobelia in southeastern Wisconsin. Five of these populations close sympatry. Our study addressed the following (21.7%) co-occurred with Lobelia siphilitica, and bum- questions: (1) How does the presence of a competitor ble bees were observed moving between the two spe- for pollination in¯uence patterns of pollinator move- cies (Bell 2003). The Mimulus populations, ranging in ment? (2) How does competition for pollination in¯u- size from 12 to more than 50 individuals, were found ence patterns of pollen receipt and numbers of seeds along the edges of wet meadows and deciduous mesic per fruit? (3) How does competition for pollination forests. in¯uence outcrossing rates and pollen-mediated gene When worker bumble bees probe Mimulus ringens dispersal? ¯owers, pollen is transported on the face and proboscis of the bee (Mitchell et al. 2004: cover photo). Lobelia METHODS siphilitica pollen is also transported on the face and proboscis, and pollen of the two species mixes at these Study species locations on the bee (J. M. Bell, personal observation). Mimulus ringens L. (Scrophulariaceae) is an her- baceous perennial native to wet meadows of central Propagation of genets with unique marker genotypes and eastern North America. Populations tend to be To facilitate mating-system analysis, we utilized 16 small (Ͻ50 genets), with distances between conspe- Mimulus ringens genets with unique multilocus com- ci®cs ranging from 0.5 m to 3 m (J. M. Bell, J. D. binations of homozygous genotypes at four unlinked Karron, and R. J. Mitchell, personal observations). allozyme loci. This design enabled us to unambigu- Plants produce showy blue zygomorphic ¯owers with ously assign paternity to all sampled seeds, and to corolla tube length ϳ19 mm and corolla tube width ϳ5 quantify the outcrossing rate of individual plants. We mm. Flowering occurs from mid-July through early used this same approach in our earlier work on sel®ng September in southeastern Wisconsin, USA. Flowers rates and gene dispersal in Mimulus ringens (Karron last for half a day, and are visited primarily by bumble et al. 1995a, b, 1997, 2004). bees foraging for nectar and pollen (Mitchell et al. Mimulus ringens occasionally produces vegetative 2004). offshoots in natural populations and is readily cloned Mimulus ringens ¯owers are self-compatible and in the greenhouse. In October 2000 we divided ramets have a mixed mating system (Karron et al. 1995a). from each of the 16 genets with unique combinations Nearly all ¯owers develop into fruits, even in the ab- of genetic markers and transplanted them into 10-cm sence of pollinator visitation, due to delayed self-fer- (four-inch) pots. We then stored these ramets in the tilization at the time of corolla abscission (Dole 1990, dark at 4ЊC for seven months. In early May 2001 we Karron et al. 2004). Outcrossing rates vary widely moved the ramets to a cool greenhouse, where they among fruits, both within and among genets (Karron were hardened off prior to planting in the ®eld. et al. 2004). This variation may re¯ect the composition Hand-outcrossed Lobelia siphilitica seeds collected of the pollen load deposited on each stigma (Karron et from 40 plants in a natural population in Cook County, al. 2004, Mitchell et al. 2005). Illinois, USA, were strati®ed and germinated in January Lobelia siphilitica L. (Campanulaceae) is an her- 2001. In May 2001 we transplanted Lobelia seedlings baceous perennial that grows in moist soils along the to 10-cm (four-inch) pots and hardened them off in a edges of lakes, streams, mesic woodlands, and wet cool greenhouse prior to planting in the ®eld. 764 JOHN M. BELL ET AL. Ecology, Vol. 86, No. 3

FIG. 1. Arrangement of Mimulus ringens genets in two of four experimental arrays: one with no competitor (left), and one with a mixture of Mimulus and Lobelia siphilitica (right). Each Mimulus genet is identi®ed with a capital letter on a shaded background. Single ramets of 15 genets are surrounded by 13 ramets of border genet ``D.'' Lobelia plants are shown on a white background and identi®ed with the word Lobelia. In both gardens there is 1.0-m spacing between Mimulus plants along the diagonal.

Experimental arrays and used a different random order in ``pure'' array 3. On 6 June, 2001 we planted four experimental arrays We used the same spatial arrangement of Mimulus gen- at the University of Wisconsin-Milwaukee Field Sta- ets in array 1 to plant array 2 in competition with Lo- tion (Saukville, Wisconsin, USA; 43Њ23Ј29.4Љ N, belia. Similarly, we used the same spatial arrangement 88Њ01Ј25.0Љ W). To minimize pollen dispersal between of Mimulus genets in array 3 to plant Mimulus genets experimental populations, the arrays were planted in in competition with Lobelia in array 4. isolated gardens separated by at least 75 m of vege- Manipulation of ¯oral display tation containing several bumble bee-pollinated plant species. We fenced the gardens to exclude mammalian Because natural variation in Mimulus ¯oral display herbivores, then tilled and mulched with hay before strongly affects outcrossing rate (Karron et al. 2004), planting to control weeds and maintain similar moisture we trimmed all Mimulus in each array to eight ¯owers levels. In two of the four experimental arrays (arrays per plant during our experimental observations. We 1 and 3) we planted Mimulus without a competitor (Fig. trimmed these ¯owers in the pre-dawn morning, before 1). Neighboring Mimulus plants were separated by 1.0 pollinators were foraging in our populations. Since m along the diagonal. In the remaining two arrays (ar- Mimulus plants in our gardens often grew to be slightly rays 2 and 4) we planted a mixture of Mimulus and taller than in the wild, Mimulus ¯oral displays were Lobelia in a checkerboard pattern (Fig. 1). The density trimmed so that displays of the two species were similar of Mimulus in the mixed-species arrays was the same in height, as they are in natural populations. as the density in the ``pure'' arrays. We watered the Pollinator observations well-drained gardens 2±3 times each week to ensure that these wetland plants did not experience drought On three consecutive fair-weather days, 12±14 Au- stress. Because plants were watered well and post-ex- gust 2001, a single two-person team observed and re- periment excavations revealed that root systems of corded a total of 12 hours of pollinator visitation pat- neighboring Mimulus and Lobelia plants were sepa- terns in the four arrays. We initiated 20-min observa- rated by at least 25 cm, we concluded that any differ- tion periods at sunrise, immediately following com- ence among arrays in reproductive success of Mimulus pletion of ¯oral-display manipulations. Observations individuals was unlikely to be due to belowground continued until 11:00 hours, when pollinator visitation competition. noticeably declined and Mimulus stigmas had begun to To facilitate mating-system analysis, we planted one close. We rotated observation periods randomly among ramet of each of 15 Mimulus genets with unique mul- arrays. Each array was observed three times each day tilocus genotypes in the center of each array. To min- (60 min total), which was ϳ20% of the total time be- imize edge effects on pollinator visitation to these 15 tween anthesis and stigma closure. ``central genets,'' we surrounded them with a buffer In each observation period we followed the ®rst row consisting of 13 ramets of genet D (Fig. 1). We bumble bee to arrive in the array for as long as possible, randomly planted the central genets in ``pure'' array 1 recording the full ¯oral-visitation sequence. We then March 2005 COMPETITION FOR POLLINATION IN MIMULUS 765 quanti®ed the sequence of ¯oral probes for each sub- harvested prior to delayed sel®ng, which typically oc- sequent bumble bee visitor until the end of the 20-min curs after 14:00 hours. observation period. When two bumble bees were foraging simultaneously in an array, we did not begin Seeds per fruit following the second visitor until the ®rst bee had de- To quantify seed production in the two competition parted. For each ¯oral-visitation sequence we recorded treatments, we tied labeled plastic tags to pedicels of the pollinator species and noted the spatial position and open ¯owers on each Mimulus central genet in each species of each plant visited. Pollinator movements be- array. Pedicels were tagged immediately following pol- tween ¯owers were classi®ed as follows: (a) transitions linator observations on 12±14 August. We then air within displays of individual Mimulus plants; (b) tran- dried fruits for 14 d and stored them in a low-humidity sitions between Mimulus plants; (c) transitions from chamber at 4ЊC. We used a dissecting microscope to Mimulus to Lobelia; (d) transitions within displays of count seeds in each of two randomly selected fruits individual Lobelia plants; (e) transitions between Lo- from each ramet for each of the three days of pollinator belia plants; (f) transitions from Lobelia to Mimulus. observation (N ϭ 355 fruits). From these data we calculated visitation rate per ¯ower per hour of observation by tallying the total number of Genotyping progeny ¯oral probes to each Mimulus plant during each ob- To quantify sel®ng rates we germinated seeds from servation period (transitions (a) ϩ (b) ϩ (f) from the each of four fruits on each central genet (N ϭ 235 list above), and dividing by the number of open ¯owers fruits). Two fruits from each plant were derived from per Mimulus plant (eight), and by the amount of time ¯owers open on 12 August, and two fruits were derived observed (0.333 h). from ¯owers open on 14 August. We germinated progeny arrays in separate pots, and transplanted two-week- Analysis of pollen loads on stigmas old seedlings into individual cells in plastic ¯ats. After To quantify patterns of improper pollen transfer, we three additional weeks of growth, the seedlings were measured pollen deposition on Mimulus and Lobelia large enough for genotyping. We genotyped 10 ran- stigmas. Pollen grains of the two species can readily domly selected seedlings from each progeny array us- be distinguished with a dissecting microscope. Mim- ing the tissue extraction and electrophoretic methods ulus ringens pollen is small (12 ␮ diameter), granular, of Karron et al. (2004). Seed germination rates were and white, whereas Lobelia siphilitica pollen is larger high (Ͼ85%) and seedling mortality was near zero. (26 ␮ diameter), yellow-orange, and usually found in Also an earlier study (J. D. Karron, R. J. Mitchell, K. clumps. Mimulus ringens stigmas are papillose and G. Holmquist, and J. M. Bell, unpublished manuscript) Mimulus pollen is often deposited in several layers, found no evidence for inbreeding depression at early each with thousands of pollen grains (J. M. Bell, per- stages of the life cycle. Therefore it is unlikely that our sonal observation). Therefore, we could not accurately estimates of outcrossing rate are biased due to early count numbers of conspeci®c and heterospeci®c pollen mortality of inbred zygotes (Farris and Mitton 1984). grains on stigmas. Instead, we noted the presence or absence of conspeci®c and heterospeci®c pollen, and Data analysis visually estimated the proportion of stigmatic area oc- All analyses were conducted using SAS 8.02 (SAS cluded by conspeci®c or heterospeci®c pollen. Institute 2000) statistical software. We compared pro- Sampling of stigmas to characterize pollen loads is portion data on pollinator species composition in the destructive. Therefore, we did not sample stigmas on ``pure'' and mixed arrays using log-likelihood ratio ␹2 12±14 August because these ¯owers were needed for analyses. quantifying seed number and outcrossing rate of each In order to understand how patterns of pollinator fruit. Instead, we sampled stigmas on 15 August, the movement within and among Mimulus genets might ®rst day following the three consecutive days of pol- in¯uence outcrossing rates in our ``pure'' and mixed- linator observation. Like the preceding three days, 15 species arrays, we determined the ratio [b/(a ϩ b)], August was a fair-weather day and we manipulated where a ϭ transitions within displays of individual ¯oral displays in the early morning. We collected eight Mimulus genets and b ϭ transitions between Mimulus stigmas from each of the 15 central Mimulus genets in genets. These proportions were calculated for the both ``pure'' array 3 and mixed-species array 4. We ``pure'' and mixed arrays and tested with a log-like- also sampled eight stigmas from each of the 15 central lihood ratio ␹2 analysis. Lobelia plants in array 4. Sampling of stigmas was To quantify the effects of competition treatment on performed at 11:00 hours, when most stigma lobes were pollinator visitation rate or behavior, we used a split- still open and therefore could readily be spread apart plot analysis (Steel and Torrie 1980), with competition on a wet-mount microscope slide. Although this sam- treatment and day of measurement as whole-plot fac- pling period facilitated analysis of pollen loads, it may tors, and observation period as a sub-plot factor. Be- have led to a slight underestimate of the proportion of cause the unit of sampling for whole-plot factors is the stigmas receiving Mimulus pollen, since stigmas were individual array, we tested for their signi®cance using 766 JOHN M. BELL ET AL. Ecology, Vol. 86, No. 3 a denominator mean square of array within treatment Mimulus did not differ signi®cantly between pure and Ͼ ␹2 ϭ using the ``test'' statement in SAS Procedure GLM mixed-species arrays (P 0.7; likelihood ratio 1 (SAS Institute 2000). 0.08; all non-B. fervidus visitors were pooled to avoid To determine whether the proportion of Mimulus expected cells Ͻ5). ¯owers receiving conspeci®c pollen was in¯uenced by Bombus fervidus also was the predominant pollinator the presence of Lobelia, we used a t test to compare to Lobelia in the mixed-species arrays, accounting for means for the two competition treatments. It was nec- 75.5% of pollinator visits to Lobelia plants. Lobelia essary to perform a t test rather than a split-plot analysis received a slightly higher proportion of visits by B. because quanti®cation of pollen loads on stigmas was impatiens (16.6%) and B. vagans Smith (8.0%) than very labor intensive, and we were only able to sample did Mimulus in the mixed-species arrays (B. impatiens from two arrays on a single day. To determine whether ϭ 10.4%; B. vagans ϭ 2.8%). These differences in competition treatment in¯uenced the amount of Mim- composition of bee species visiting Mimulus and Lo- ulus pollen received per ¯ower we used one-way AN- belia in the mixed-species arrays were signi®cant (P Ͻ ␹2 ϭ OVA. We arcsine transformed pollen variables to ap- 0.003, likelihood ratio 1 9.0, again pooling non- proximate a normal distribution of residuals (Zar B. fervidus visitors) (Appendix A). 1999). Following arcsine transformation, these vari- Bee activity in the pure arrays was about half that ables satis®ed the assumptions of normality and ho- in the mixed-species arrays. In the pure arrays we ob- mogeneity of variance. served 25 bumble bee ¯oral-visitation sequences with To test the hypothesis that competition treatment af- 783 ¯oral probes, while in the mixed-species arrays we fected number of seeds per fruit and outcrossing rate, observed a total of 42 visitation sequences with 1377 we used a split-plot analysis, with competition treat- ¯oral probes to the two species. The higher total visits ment and day of measurement as whole-plot factors, in the mixed-species arrays was largely due to the pres- and genet as a sub-plot factor. Because the unit of sam- ence of Lobelia, which received 814 of the 1377 ¯oral pling for whole-plot factors is the individual array, we probes. The combined number of Mimulus and Lobelia tested for their signi®cance using a denominator mean ¯owers probed by a bee during a ¯oral-visitation se- square of array within treatment. For sub-plot factors, quence was similar in the two competition treatments we took the conservative approach of testing for sig- (31.3 Ϯ 6.4 ¯owers [mean Ϯ 1 SE] in the pure arrays, ni®cance using a denominator MS of Genet ϫ Day ϫ 32.8 Ϯ 5.2 ¯owers in the mixed-species arrays). Array(Treatment), which re¯ects variation among ra- The visitation rate to Mimulus ¯owers tended to be met±day combinations. higher in the pure arrays (0.65 Ϯ 0.04 probes per ¯ower To analyze patterns of pollen-mediated gene dis- per hour) than in the mixed-species arrays (0.47 Ϯ 0.04 persal, we excluded seeds sired by self-pollen and cal- probes per ¯ower per hour). However, because of sub- culated the distance between the pollen donor and the stantial variation among arrays, statistical power for maternal parent for each outcrossed seedling in our among-plot factors was low (power ϭ 0.05), and the paternity data set. We used ␹2 analysis to test for the split-plot ANOVA (Appendix B) showed no signi®- effect of competition treatment on pollen dispersal dis- cance for this 28% difference between competition tance, assigning each outcrossed seedling to 2-m dis- treatments (P Ͼ 0.3). Probes to Mimulus ¯owers that tance categories based on distance to the paternal plant. were immediately preceded by probes to Lobelia ¯ow- We combined the two longest distance categories to ers may be ineffective in transferring Mimulus pollen. ensure that expected values in each cell were Ͼ5 prog- Therefore, we also calculated a rate of ``effective'' vis- eny. itation to Mimulus ¯owers in the mixed-species arrays, which excludes Mimulus probes that were immediately RESULTS preceded by a probe to a Lobelia ¯ower. The rate of effective visitation to Mimulus ¯owers in the mixed- Pollinator observations species arrays was 0.40 Ϯ 0.04 probes per ¯ower per During 12 h of observation we recorded 2160 ¯oral hour. This rate is slightly lower than the overall visi- probes, 626 plant visits, and 67 separate bumble bee tation rate to Mimulus in the mixed-species arrays (0.47 ¯oral-visitation sequences. The predominant pollinator Ϯ 0.04 probes, reported above), but patterns of sig- visiting Mimulus ¯owers in both ``pure'' and mixed- ni®cance were unchanged (unpublished analysis). species arrays was Bombus fervidus Fabricius, which In the mixed-species arrays, bees frequently moved was responsible for 86.3% of visits to Mimulus plants between species during individual visitation sequences. in pure arrays, and 86.8% of visits to Mimulus plants Nearly half (42%) of the 321 plant-to-plant movements in mixed-species arrays (Appendix A). The next most were interspeci®c transitions (either Mimulus to Lo- common visitor to Mimulus plants in the two types of belia or Lobelia to Mimulus), while 10.1% of 1335 arrays was Bombus impatiens Cresson, which contrib- movements between ¯owers were interspeci®c. uted 13.3% of visits to Mimulus plants in pure arrays, The ratio of transitions between Mimulus genets di- and 10.4% of visits to Mimulus plants in mixed-species vided by all intraspeci®c Mimulus transitions was high- arrays. The species composition of pollinators visiting er in the pure arrays (0.314) than in the mixed-species March 2005 COMPETITION FOR POLLINATION IN MIMULUS 767

␹2 ϭ ϭ arrays (0.227) (likelihood ratio 2 10.7, P 0.005). Such among-plant movements may promote outcross pollen transfer. Bees did not change the number of ¯owers they probed consecutively on Mimulus plants in response to competition treatment (in the pure arrays, 2.80 Ϯ 0.19 ¯owers [mean Ϯ 1 SE], in the mixed- species arrays, 2.68 Ϯ 0.16 ¯owers; P Ͼ 0.6 [from split-plot ANOVA]; Appendix C). This suggests that geitonogamy on Mimulus plants is similar in the two competition treatments. Pollen load analysis The proportion of Mimulus ¯owers receiving conspeci®c pollen prior to 11:00 hours was signi®cantly lower in the mixed-species arrays (85.8%) than in the pure arrays (95.9%) (t ϭ 2.75, P ϭ 0.007). In addition, the proportion of Mimulus stigmatic area occluded with conspeci®c pollen was signi®cantly lower in the mixed- species arrays (0.51 Ϯ 0.01) than in the pure arrays Ϯ ϭ ϭ (0.62 0.02) (F1,1 6.43, P 0.012). The presence of Lobelia did not lead to much heterospeci®c pollen deposition on Mimulus stigmas. Only 12% of Mimulus stigmas in the mixed-species arrays had detectable Lobelia pollen present (Fig. 2a). By con- trast, 44% of Lobelia stigmas received Mimulus pollen (Fig. 2a). In addition, the proportion of Mimulus stigmatic area occluded with Lobelia pollen was lower FIG. 2. Conspeci®c and heterospeci®c pollen loads on (0.05 Ϯ 0.01) than the proportion of Lobelia stigmatic Mimulus ringens and Lobelia siphilitica stigmas in the mixed- area (0.15 Ϯ 0.019) occluded with Mimulus pollen (Fig. species arrays: (a) the proportion of Mimulus and Lobelia stigmas with conspeci®c or heterospeci®c pollen present; (b) 2b). the proportion of stigmatic surface occluded by conspeci®c Seeds per fruit and heterospeci®c pollen. Data are means and 1 SE; N ϭ 120± 127 stigmas for each bar. The presence of Lobelia led to a signi®cant 37% reduction in the mean number of Mimulus seeds per ϭ Ͻ fruit (ANOVA, F1,2 114.76, P 0.009, Table 1, Fig. 3). Genets also differed signi®cantly in seeds per fruit, and these differences were reasonably consistent across

TABLE 1. Results of split-plot ANOVA of the effects of competition treatment (abbreviated ``Trt''), day, and genet on the number of seeds per fruit.

Source of variation df MS FP Whole-plot Treatment 1 4.61 ϫ 107 114.76 0.009 Day 2 3.48 ϫ 106 8.68 0.10 Trt ϫ Day 2 5.95 ϫ 105 1.48 0.4 Array(Trt) 2 4.02 ϫ 105 Sub-plot Genet 14 1.43 ϫ 106 2.78 0.004 Trt ϫ Genet 14 8.10 ϫ 105 1.57 0.12 Day ϫ Array(Trt) 4 1.95 ϫ 106 3.78 0.009 Genet ϫ Array(Trt) 28 5.07 ϫ 105 0.98 0.5 Genet ϫ Day 28 3.28 ϫ 105 0.64 0.9 Trt ϫ Genet ϫ Day 28 4.07 ϫ 105 0.79 0.7 Genet ϫ Day ϫ Array(Trt) 54 5.16 ϫ 105 1.64 0.008 Residual error 177 3.14 ϫ 105 Notes: Competition treatment and day were applied to whole plots (arrays), so their effects were tested over Array(Trt), which is among-array error. The subplot effect was Genet, which was conservatively tested over Genet ϫ Day ϫ Array(Trt). The variation among ramet±day combinations [Genet ϫ Day ϫ Array(Trt)] was tested over the variation among individual fruits. N ϭ 355 fruits. Model R2 ϭ 0.739. Signi®cant values are in boldface type. 768 JOHN M. BELL ET AL. Ecology, Vol. 86, No. 3

FIG.4. Mimulus ringens outcrossing rate (proportion of FIG.3. Mimulus ringens seeds per fruit in the presence seeds that are outcrossed) in the presence and absence of (mixed species) and absence (Mimulus only) of Lobelia si- Lobelia siphilitica, a competitor for pollination. Outcrossing philitica, a competitor for pollination. Seed number was de- rates for Mimulus were determined by genotyping 10 off- termined by counting seeds from two randomly selected fruits spring from each of two randomly selected fruits from each from each of the 15 central genets in each array (Fig. 1). of the 15 central genets in each array. Data are means and 1 Separate means are reported for each day of pollinator ob- SE; N ϭ 58±59 fruits for each bar. The total number of seed- servation. Data are means and 1 SE; N ϭ 56±71 fruits for lings genotyped was 2344 seedlings from 238 fruits. each bar.

teraction indicates that individual ramets responded competition treatments (interaction P Ͼ 0.12). The oth- differently to competition treatment on different days. er signi®cant interactions concern sub-plot factors, and Outcrossing rates were not correlated with seeds per do not alter conclusions about the whole-plot effects. fruit in either treatment (P Ͼ 0.8 in both cases, N Ͼ 116). Additionally, patterns of pollen-mediated gene Outcrossing rates dispersal did not vary signi®cantly across competition ␹2 ϭ ϭ The rate of outcrossing (proportion of seeds that are treatments ( 4 6.9, P 0.14); both had similar, outcrossed) in Mimulus ringens was signi®cantly lower strongly leptokurtic distributions. in the mixed-species arrays (0.43 Ϯ 0.02) than in the DISCUSSION pure arrays (0.63 Ϯ 0.02) (mean Ϯ 1SE) (Table 2, Fig. 4). This effect was consistent across days and across Competition for pollination with Lobelia siphilitica genets. No other main effects were signi®cant, and dramatically lowered both the number of seeds per fruit most interactions were weak. The only signi®cant in- and the rate of outcrossing in Mimulus ringens. To our

TABLE 2. Results of split-plot ANOVA of the effects of competition treatment, day, and genet on outcrossing rate.

Source df MS FP Whole-plot Treatment 1 2.1432 148.36 0.0067 Day 2 0.0772 5.35 0.15 Trt ϫ Day 2 0.1228 8.50 0.10 Array(Trt) 2 0.0144 Sub-plot Genet 14 0.0636 0.64 0.8 Trt ϫ Genet 14 0.0314 0.32 0.9 Day ϫ Array(Trt) 4 0.0619 0.63 0.5 Genet ϫ Array(Trt) 28 0.0783 0.79 0.7 Genet ϫ Day 28 0.0458 0.46 0.9 Trt ϫ Genet ϫ Day 28 0.0901 0.91 0.6 Genet ϫ Day ϫ Array(Trt) 54 0.0988 2.40 0.0008 Residual error 177 0.0412

Notes: Denominator MS are as in Table 1. N ϭ 2344 seeds, 235 fruits. Model R2 ϭ 0.690. Signi®cant values are in boldface type. March 2005 COMPETITION FOR POLLINATION IN MIMULUS 769 knowledge, this is the ®rst study to empirically dem- times attributed to a reduced rate of pollinator visitation onstrate that pollinator sharing directly in¯uences out- (Waser 1978, 1983, Brown and Mitchell 2002). In our crossing rates, as predicted by Campbell (1985b). This study, the frequency of ``effective'' pollinator visitation result adds competition for pollination to a growing list to Mimulus ¯owers was 38% lower in the mixed- spe- of ecological and demographic variables known to in- cies arrays than in the ``pure'' arrays. This result was ¯uence outcrossing rates, including population density, not signi®cant, but we had low statistical power for population size, ¯oral-display size, weather conditions, detecting differences in visitation rate. If such differ- and availability of pollinators (Barrett and Eckert 1990, ences in visitation indeed exist, they may partially ex- de Jong et al. 1992, Jarne and Charlesworth 1993, Wil- plain the signi®cant reduction in amount of conspeci®c liams et al. 2001, Karron et al. 2004). In our study, the pollen deposited on Mimulus stigmas in the mixed- effect of competition for pollination was striking. The species arrays. 20-percentage-point reduction in outcrossing rate is Reduced seed set in species mixtures may also result comparable to the effects of other ecological factors from improper pollen transfer and pollen loss (Waser known to in¯uence outcrossing in Mimulus. For ex- 1978, Campbell and Motten 1985, Galen and Gregory ample, Karron et al. (1995a) noted that a 16-fold de- 1989). Frequent pollinator moves from Mimulus to Lo- crease in population density led to a 16-percentage- belia led to considerable deposition of Mimulus pollen point reduction in outcrossing rate. Also, Karron et al. on Lobelia stigmas, and probably also to loss on other (2004) found that outcrossing rates of plants with large ¯oral surfaces. These factors may have contributed to daily ¯oral displays (16 open ¯owers) were 14 per- the smaller loads of conspeci®c pollen on Mimulus stig- centage points lower than outcrossing rates of plants mas in the mixed-species arrays. with small daily ¯oral displays (two open ¯owers). Pollinator moves from Lobelia to Mimulus may also Our ®ndings have important implications for ®tness cause improper pollen transfer through deposition of in mixed-species populations because reproductive Lobelia pollen on Mimulus stigmas. However, only success depends upon both the quality and quantity of 12% of Mimulus stigmas received any Lobelia pollen offspring produced. The Mimulus genets used in this (Fig. 2b). Therefore, heterospeci®c pollen deposition study are derived from a population that exhibits a on Mimulus stigmas was probably not a major factor moderate level of inbreeding depression (mean ®tness in¯uencing the number of ovules fertilized per fruit. of self progeny is 21% lower than mean ®tness of out- Multiple factors may have limited the amount of Lo- cross progeny (J. D. Karron, unpublished data)). To belia pollen deposited on Mimulus stigmas. For ex- quantify the overall effects of competition for polli- ample, Lobelia anthers usually do not dehisce until late nation on ®tness through seed function, we used these morning, after many Mimulus stigmas have already data on inbreeding depression to weight the relative closed. In addition, Lobelia pollen grains are much ®tness of self and outcross progeny in our two com- larger than Mimulus grains, and do not adhere well to petition treatments. These calculations indicate that, on the small papillae on Mimulus stigmas. average, competition with Lobelia siphilitica resulted in a 39.4% reduction in maternal ®tness for Mimulus Effect of pollen loss on outcrossing ringens. The effects of competition on both seed set The lower rate of outcrossing in the mixed-species and outcrossing rate may be different in natural com- arrays (Fig. 4, Table 2) is most likely caused by the munities, where competitors may be clumped or may loss of Mimulus pollen to Lobelia. Such heterospeci®c occur at a range of densities (Caruso 1999, Fishman pollen loss reduces the amount of bee-transported Mim- and Wyatt 1999). Separate clusters or patches of con- ulus pollen that could potentially be deposited on ¯ow- speci®cs may reduce the proportion of interspeci®c ers of other conspeci®cs (outcrossing). The presence transitions by pollinators, and may lower the amount of Lobelia is unlikely to have had much in¯uence on of improper pollen transfer (Campbell 1985b, Camp- the extent of Mimulus self-pollination. In particular, bell and Motten 1985). Alternatively, the presence of geitonogamous self-pollination is likely to have been several co-occurring species competing for pollination similar in the two competition treatments, since there could lead to a greater reduction in ®tness than we was no signi®cant difference in the distribution of num- observed in this study. Additional research is needed bers of consecutively probed ¯owers on Mimulus plants to quantify the effects of competition for pollination in the ``pure'' and mixed-species arrays. Also, the pres- on offspring quality in natural populations. ence of a competitor for pollination may have little effect on autogamous (within-¯ower) self-pollination. Competition mechanisms in¯uencing seed set Three of the modes of autogamy (prior, competing, and Our study was designed primarily to examine the delayed self-pollination) occur without involvement of consequences of competition for pollination on seed a pollinator (Lloyd and Schoen 1992). The fourth set and outcrossing rates in Mimulus ringens. However, mode, facilitated sel®ng, may be modest in Mimulus our results also provide some indications of the mech- due to the position of the stigma above the anthers anisms responsible for the signi®cant responses we (LeClerc-Potvin and Ritland 1994). If the number of found. Lower seed set in mixed-species arrays is some- outcross pollen grains on Mimulus stigmas declines, 770 JOHN M. BELL ET AL. Ecology, Vol. 86, No. 3 but the number of self pollen grains on stigmas remains to R. J. Mitchell supported this research. This paper is derived unchanged, the outcrossing rate would be lower in the from a Ph.D. dissertation submitted by J. M. Bell to Uni- versity of WisconsinÐMilwaukee. mixed-species arrays. Note that there is some similarity between the effects of heterospeci®c pollen loss on LITERATURE CITED outcrossing rates and the effects of pollen discounting Barrett, S. C. H., and C. G. Eckert. 1990. Variation and on outcrossing rates (Holsinger 1996). In both pro- evolution of mating systems in seed plants. Pages 229±554 cesses a reduced amount of pollen transfer to conspe- in S. Kawano, editor. Biological approaches and evolu- ci®cs leads to a lower rate of outcrossing. tionary trends in plants. Academic Press, London, UK. Bell, J. M. 2003. The effect of interspeci®c competition for Pollen-mediated gene dispersal pollinator service on seed production and outcrossing rates in Mimulus ringens (Scrophulariaceae). Dissertation. Uni- Campbell (1985b) suggested that competition for versity of Wisconsin-Milwaukee, Milwaukee, Wisconsin, pollination would not only in¯uence outcrossing rates, USA. but might also in¯uence the distance pollen disperses Brown, B. J., and R. J. Mitchell. 2001. Competition for pollination: effects of pollen of an invasive plant on seed set between conspeci®c plants. She empirically con®rmed of a native congener. Oecologia 129:43±49. this prediction by ®nding that the presence of Claytonia Brown, B. J., R. J. Mitchell, and S. A. Graham. 2002. Com- virginica decreased the distance of dye dispersal in petition for pollination between an invasive species (purple Stellaria pubera (Campbell 1985b). We found no ev- loosestrife) and a native congener. Ecology 83:2328±2336. idence for an effect of competition treatment on real- Campbell, D. R. 1985a. Pollinator sharing and seed set of Stellaria pubera: competition for pollination. Ecology 66: ized pollen-mediated gene dispersal. It is possible that 544±553. a different result would be obtained in larger arrays or Campbell, D. R. 1985b. Pollen and gene dispersal: the in- large natural populations, which would be less subject ¯uences of competition for pollination. Evolution 39:418± to edge effects, and might permit detection of longer 431. Campbell, D. R., and A. F. Motten. 1985. The mechanism of distance gene dispersal. competition for pollination between two forest herbs. Ecol- ogy 66:554±563. Conclusion Caruso, C. M. 1999. Pollination of Ipomopsis aggregata (Po- When Mimulus ringens and Lobelia siphilitica grow lemoniaceae): effects of intra- vs. interspeci®c competition. in close sympatry, they share bumble bee pollinators American Journal of Botany 86:663±668. Caruso, C. M. 2000. Competition for pollination in¯uences and strongly compete for pollination. The outcrossing selection on ¯oral traits of Ipomopsis aggregata. Evolution rate of Mimulus in mixed-species arrays was 20 per- 54:1546±1557. centage points lower than the outcrossing rate in pure Caruso, C. M., H. Maherali, and R. B. Jackson. 2003a. Gen- Mimulus. arrays. Also, seed set in Mimulus fruits was der-speci®c ¯oral and physiological traits: implications for the maintenance of females in gynodioecious Lobelia si- 37% lower in the presence of Lobelia than in Mimulus- philitica. Oecologia 135:524±531. only arrays. Therefore, competition in¯uences both Caruso, C. M., S. B. Peterson, and C. E. Ridley. 2003b. Nat- seed quality and seed quantity, resulting in a dramatic ural selection on ¯oral traits of Lobelia (Lobeliaceae). reduction in reproductive success. In the present study American Journal of Botany 90:1333±1340. the effect of competition for pollination on seed quan- deJong, T., N. M. Waser, M. V. Price, and R. M. Ring. 1992. Plant size, geitonogamy, and seed set in Ipomopsis aggre- tity was stronger than the effect on seed quality, largely gata. Oecologia 89:310±315. because the experimental population had a fairly low Dole, J. A. 1990. Role of corolla abscission in delayed self- level of genetic load. However, our results suggest that pollination of Mimulus guttatus (Scrophulariaceae). Amer- studies of competition for pollination in self-compat- ican Journal of Botany 77:1505±1507. ible species should consider effects on seed quality as Farris, M. A., and J. B. Mitton. 1984. Population density, outcrossing rate, and heterozygote superiority in ponderosa well as seed quantity. It is possible that in a species pine. Evolution 38:1151±1154. with substantial delayed self-fertilization and fairly Feinsinger, P., K. G. Murray, S. Kinsman, and W. H. Busby. high levels of genetic load, competition for pollination 1986. Floral neighborhood and pollination success in four might have little effect on seed quantity, but might hummingbird-pollinated cloud forest plant species. Ecol- dramatically affect seed quality. Further research is ogy 67:449±464. Fishman, L., and R. Wyatt. 1999. Pollinator-mediated com- needed to determine whether competition for pollina- petition, reproductive character displacement, and the evo- tion in¯uences outcrossing rates of other ¯owering lution of sel®ng in Arenaria uni¯ora (Caryophyllaceae). plant species. Evolution 53:1723±1733. Galen, C., and T. Gregory. 1989. Interspeci®c pollen transfer ACKNOWLEDGMENTS as a mechanism of competition: consequences of foreign We thank T. Schuck for assistance in the greenhouse and pollen contamination for seed set in the alpine Wild¯ower, K. Havens for providing seeds of Lobelia siphilitica. We are Polemonium viscosum. Oecologia 81:120±123. grateful to T. Bell, B. Funk, R. Claire, J. Reinartz, G. Meyer, Grant, V. 1950. The ¯ower constancy of bees. Botanical Re- and L. Nelson for help and encouragement in the ®eld. We view 16:379±398. thank N. Waser, S. Schnitzer, J. Reinartz, G. Meyer, P. Dunn, Harder, L. D., and S. C. H. Barrett. 1996. Pollen dispersal and T. Ehlinger for helpful comments on an earlier version and mating patterns in animal-pollinated plants. Pages 140± of the manuscript. A Botanical Society of America Karling 190 in D. G. Lloyd and S. C. H. Barrett, editors. Floral Research Award to J. M. Bell and National Science Foun- biology: studies on ¯oral evolution in animal-pollinated dation grants DEB 9816712 to J. D. Karron and DEB 9903308 plants. Chapman and Hall, New York, New York, USA. March 2005 COMPETITION FOR POLLINATION IN MIMULUS 771

Holsinger, K. E. 1996. Pollination biology and the evolution Mitchell, R. J., J. D. Karron, K. G. Holmquist, and J. M. of mating systems in ¯owering plants. Evolutionary Bi- Bell. 2005. Patterns of multiple paternity in fruits of Mim- ology 29:107±149. ulus ringens (Phrymaceae). American Journal of Botany Jarne, P., and D. Charlesworth. 1993. The evolution of the 92(5), in press. sel®ng rate in functionally hermaphrodite plants and ani- Pleasants, J. M. 1980. Competition for bumblebee pollinators mals. Annual Review of Ecology and Systematics 24:441± in Rocky Mountain plant communities. Ecology 61:1446± 466. 1459. Johnston, M. O. 1991. Natural selection on ¯oral traits in Pleasants, J. M. 1983. Structure of plant and pollinator com- two species of Lobelia with different pollinators. Evolution munities. Pages 375±393 in C. E. Jones and R. J. Little, 45:1468±1479. editors. Handbook of experimental pollination ecology. Karron, J. D., R. T. Jackson, N. N. Thumser, and S. L. Van Nostrand Reinhold, New York, New York, USA. Schlicht. 1997. Outcrossing rates of individual Mimulus Rathcke, B. 1983. Competition and facilitation among plants ringens genets are correlated with anther±stigma separa- for pollination. Pages 305±329 in L. Real, editor. Polli- tion. Heredity 79:365±370. nation biology. Academic Press, Orlando, Florida, USA. Karron, J. D., R. J. Mitchell, K. G. Holmquist, J. M. Bell, SAS Institute. 2000. SAS/STAT User's Guide, Version 8. and B. Funk. 2004. The in¯uence of ¯oral display size on SAS Institute, Cary, North Carolina, USA. sel®ng rates in Mimulus ringens. Heredity 92:242±248. Schemske, D. W., M. F. Willson, M. N. Melampy, L. J. Miller, Karron, J. D., N. N. Thumser, R. Tucker, and A. J. Hessenauer. L. Verner, K. M. Schemske, and L. B. Best. 1978. Flow- 1995a. The in¯uence of population density on outcrossing ering ecology of some spring woodland herbs. Ecology 59: rates in Mimulus ringens. Heredity 75:175±180. 351±366. Karron, J. D., R. Tucker, N. N. Thumser, and J. A. Reinartz. Sih, A., and M. Baltus. 1987. Patch size, pollinator behavior, and pollinator limitation in catnip. Ecology 68:1679±1690. 1995b. Comparison of pollinator ¯ight movements and Steel, R. G. D., and J. H. Torrie. 1980. Principles and pro- gene dispersal patterns in Mimulus ringens. Heredity 75: cedures of statistics: a biometrical approach. Second edi- 612±617. tion. McGraw-Hill Book Company, New York, New York, Leclerc-Potvin, C., and K. Ritland. 1994. Modes of self- USA. fertilization in Mimulus guttatus (Scrophulariaceae): a ®eld Waser, N. M. 1978. Competition for hummingbird pollination experiment. American Journal of Botany 81:199±205. and sequential ¯owering in two Colorado wild¯owers. Levin, D. A. 1978. Pollinator behaviour and the breeding Ecology 59:934±944. structure of plant populations. Pages 133±150 in A. J. Rich- Waser, N. M. 1983. Competition for pollination and ¯oral ards, editor. The pollination of ¯owers by insects. Aca- character differences among sympatric plant species: a re- demic Press, London, UK. view of evidence. Pages 277±293 in C. E. Jones and R. J. Levin, D. A., and W. W. Anderson. 1970. Competition for Little, editors. Handbook of experimental pollination bi- pollinators between simultaneously ¯owering species. ology. Van Nostrand Reinhold New York, New York, USA. American Naturalist 104:455±467. Waser, N. M. 1986. Flower constancy: de®nition, cause, and Lloyd, D. G., and D. J. Schoen. 1992. Self- and cross-fer- measurement. American Naturalist 127:593±603. tilization in plants I. Functional dimensions. International Williams, C. F., J. Ruvinsky, P. E. Scott, and D. K. Hews. Journal of Plant Science 153:358±369. 2001. Pollination, breeding system, and genetic structure Mitchell, R. J., J. D. Karron, K. G. Holmquist, and J. M. in two sympatric Delphinium (Ranunculaceae) species. Bell. 2004. The in¯uence of Mimulus ringens ¯oral display American Journal of Botany 88:1623±1633. size on pollinator visitation patterns. Functional Ecology Zar, J. H. 1999. Biostatistical analysis. Fourth edition. Pren- 18:116±124. tice-Hall, Upper Saddle River, New Jersey, USA.

APPENDIX A A ®gure showing species composition of bumble bee visitors to Mimulus ringens and Lobelia siphilitica in the experimental arrays is available in ESA's Electronic Data Archive: Ecological Archives E086-039-A1.

APPENDIX B A table showing results of a split-plot ANOVA for pollinator visitation as a function of Treatment, Day, and Array(Trt) is available in ESA's Electronic Data Archive: Ecological Archives E086-039-A2.

APPENDIX C A ®gure showing the number of consecutively probed ¯owers on individual Mimulus plants in the pure-species and mixed- species arrays is available in ESA's Electronic Data Archive: Ecological Archives E086-039-A3.