Premise: A central goal of pollination biology is to connect plants with the identity of their pollinator(s). While predictions based on floral syndrome traits are extremely useful, direct observation can reveal further details of a species’ pollination biology. The wildflower, Phlox drummondii, has a floral syndrome consistent with pollination by a variety of lepidoptera. We describe pollination in P. drummondii and use empirical data to test this prediction.

Methods: We directly observe each step of the pollination process in P. drummondii. First, we observe 55.5 hours of floral visitation throughout the day/evening (7:00-20:30) at sites across the species range. We use a temporal pollinator exclusion experiment to determine the contribution of diurnal and nocturnal pollination to reproductive output. We then quantify P. drummondii pollen pickup and deposition by the dominant floral visitor, Battus philenor. Finally, we test the effect of B. philenor visitation on P. drummondii reproductive output by quantifying fruit set following visitation to greenhouse-grown flowers.

Results: B. philenor is the primary pollinator of P. drummondii. Pollination is largely diurnal, and we observe a variety of lepidopteran visitors during this period. However, B. philenor is by far the most frequent visitor, representing 88.5% of all observed floral visits. We show that B. philenor is not only an extremely common visitor but also an effective pollinator by demonstrating that individuals transfer pollen between flowers and that a single visit can elicit fruit set.

Conclusion: Our data are consistent with the syndrome prediction of lepidopteran pollination and further reveal a single butterfly species, B. philenor, as the primary pollinator. Collectively, our study demonstrates the importance of empirical pollinator observations, adds to our understanding of pollination mechanics, and offers a specific case study of butterfly pollination.

Study species: P. drummondii (Polemoniaceae) is an annual herb native to central Texas that occurs along roadsides, fields, and pastures. Seeds germinate in early spring and individuals flower and fruit from April through June. Flowers are radially symmetric with a diameter of 18 to 24mm (Figure 1A,B) (Wherry 1955). Floral buds open in the morning (07:00-9:00) in both field and greenhouse conditions. If unpollinated, open flowers last approximately 2-4 days before wilting. Fruits begin developing 1-2 days after pollination and each flower can produce a single fruit. P. drummondii has an active self-incompatibility system and displays low rates of autonomous selfing (~1%) (Roda and Hopkins 2019).

Previous studies within the Phlox genus demonstrate visitation by a broad group of lepidopteran pollinators (Strakosh and Ferguson 2005; Wiggam and Ferguson 2005). Like other members of its genus, P. drummondii exhibits floral traits consistent with a lepidopteran pollination syndrome (Faegri and Van Der Pijl 1979; Grant and Grant 1965; Stebbins 1970). Flowers produce nectar at the base of a narrow, fused, tubular corolla ranging in length from 13 to 17mm (Wherry 1955). Sexual organs (stigma and anthers) are fully inserted within the corolla tube. Brightly colored petals are light-blue/purple throughout most of the species range and emit a faint fragrance (Hopkins and Rausher 2012). 

Observing floral visitation: To characterize the diversity and frequency of floral visitors, we conducted 55.5 hours of direct pollinator observations. We observed visitation throughout the day/evening (07:00 to 20:30). Coinciding with peak bloom, observations were conducted from April 23, 2022 to May 4, 2022. We identified three focal populations for pollinator observations throughout Central Texas (Appendix S1). These focal study sites were located on rural roadsides or in a protected meadow at the Brackenridge Field Station. At each site, we delimited a patch area ranging from 30 to 140 square meters. Observers stood just outside the patch boundary for 1 hour observation periods and noted every animal interacting with flowers inside the study patch. Visitors were recorded if any part of the animal’s body inserted into the corolla tube opening. Species identity was assessed visually. We conducted a handful of additional observation hours near range edges to assess potential variation in pollinator assemblage across space (Appendix S1).

Temporal pollinator exclusion: We used temporal pollinator exclusion to assess the contribution of diurnal pollination, nocturnal pollination and autonomous selfing towards overall reproductive output. In April 2022, a population of P. drummondii was identified at the Brackenridge Field Laboratory in Austin, TX (30.2832 N, -97.7801 W) near the geographic center of the species range (Appendix S1). Throughout an approximately 100m² area, 86 experimental plants were randomly chosen and assigned to one of four pollinator treatments: total exclusion (n=15), diurnal exclusion (n=23), nocturnal exclusion (n=23), and no exclusion (n=25). All open flowers and fruits were removed before covering plants in fine-mesh cages to exclude pollinators of all sizes. At 07:00 each morning, fine-mesh cages were removed from the nocturnal exclusion group. At 19:00 each evening, fine-mesh cages were removed from the diurnal exclusion group. The total exclusion group remained covered while the no exclusion group remained uncovered. After five days, new but unopened buds were removed and all flowers that opened during the experiment were counted (as measured by remaining calyx number). All plants were then covered with exclusion cages and monitored for fruit set. We counted the number of fruits on days five and eight after beginning treatment. On day eight, we observed minor plant death (n=84 live plants remaining) and caterpillar herbivory resulting in the removal of entire inflorescences leading to some uncertainty in final reproductive success. We implemented a binomial mixed model using the lme4 R package (v1.1-30) to model fruit set in our exclusion conditions (Bates et al. 2015). Each individual flower that opened during the experiment was represented in the data set as either fruit present or fruit absent. 124 flowers were recorded in the total exclusion treatment, 309 in the diurnal exclusion treatment, 235 in the diurnal treatment and 250 in the no exclusion treatment. Because our dataset includes multiple flowers from the same individual, we included plant individual as a random effect in the model. To compare fruit set across treatment groups, we conducted pairwise post-hoc Tukey-Kramer tests using the glht() function in the multcomp R package (v1.4-20) (Hothorn, Bretz, and Westfall 2008). 

Quantifying pollen transfer: To determine if visitors transfer pollen between flowers, we assessed both pollen pickup and deposition by insect visitors. We characterized pollen pickup by capturing insects foraging on roadside populations of P. drummondii and quantifying pollen grains on proboscises. These insects were captured outside of patches used for observation and represented a small fraction of all insects observed in the area. Because Battus philenor was by far the most frequent visitor, most insects captured and included in our analysis were B. philenor (n=24). We collected at least one representative individual of each pollinator type observed visiting P. drummondii (n=1 Hyles lineata; n=4 Skipper spp.; n=2 ) except for the Common Buckeye (Junonia coenia) which was only seen making a single floral visit throughout all observation periods.

After capture, insects were immediately immobilized in a glassine envelope and placed on ice. Specimens were stored at -20°C. Following twenty-four hours in a humidifying chamber to increase pliability, we pinned each specimen through the thorax onto glassine covered Styrofoam. The proboscis was carefully unrolled with a clean pin and length was measured. To remove pollen from the proboscis for counting, the dorsal, ventral and lateral sides of the unrolled proboscis were rubbed with a gelatin-glycerin cube (Kearns and Inouye 1993). We melted the gelatin-glycerin cube onto a microscope slide and imaged the entirety of the melted gelatin in brightfield at 2.5x magnification using the Zeiss Axioimager (Jena, Germany). Pollen grain number per image was quantified automatically using ImageJ and confirmed by manual count (Schindelin et al. 2012).

Based on the literature and extensive experience observing Phlox pollen using a light microscope, we identified pollen grains matching Phlox pollen’s distinctive morphology in size and shape (Wherry 1955; Roda and Hopkins 2019; Suni and Hopkins 2018; Buthod and Skvarla 2014). Pollen grains consistent with Phlox pollen morphology were included in our quantifications. Another Phlox, P. cuspidata, is found geographically near, although not at, some of our study sties. Because of shared pollen morphology, distinguishing between pollen of these species is challenging. However, we do not expect this to present an issue in our analyses as P. cuspidata is highly selfing with low pollinator visitation (Levin 1989).

We assessed B. philenor pollen deposition per visit by allowing wild-caught individuals to visit flowers in succession on greenhouse grown plants maintained indoors. After each single visit, corollas were removed and stigmas were collected, fixed in 80% isopropanol, and stored at 4°C (n=20 unvisited; n=28 visited by B. philenor). We removed corollas and collected stigmas from unvisited flowers as a control. Stigmas were then mounted in glycerol media on a slide and imaged in brightfield at 10x magnification using the Zeiss Axioscope 7 (Jena, Germany). Pollen grain number per image was quantified using ImageJ (Schindelin et al. 2012). We used a linear model implemented in the lme4 R package (v1.1-30) to test for an effect of B. philenor visitation on the number of pollen grains on stigmas.

Although quantifying pollen transfer by each visitor type would be informative, the low visitation rates of non-Battus insects made capturing other visitors extremely challenging. We were unable to gain adequate sample sizes of non-Battus visitors to assess pollen transfer despite extensive time in the field.

Quantifying pollination efficiency: To assess whether B. philenor visitation affects reproductive output, we performed a controlled foraging experiment. Plants were grown from seed in the greenhouse and maintained indoors to exclude pollinators. We captured B. philenor foraging on natural populations of P. drummondii and allowed them to forage on a ‘pollen donor plant’ while inside of a mesh cage. Because B. philenor is a generalist, wild-foraging insects likely carry some heterospecific pollen which may affect overall pollination efficiency. To capture this effect in our measurement of efficiency, insects were not cleaned prior to controlled foraging.

Next, we presented the butterfly with a single, new plant and noted the first three flowers visited in succession by the butterfly (geitonogamous visitation). Corollas were removed following B. philenor visitation and calyxes were labeled with tape. On the same plant, we removed corollas from three unvisited flowers as a control for autonomous selfing. Additionally, we manually applied pollen from the ‘pollen donor plant’ to stigmas of three open flowers. Pollen donor and experimentally visited plants were from independent genetic lines grown in our lab, to reduce the likelihood of sharing incompatibility genotypes. We included 18 plants in our experiment. Following pollinator exposure, plants were maintained in the greenhouse and monitored for fruit set for one week. We used a binomial mixed model using the lme4 R package (v1.1-30) to model fruit set across the five pollination treatments (manual outcross, autonomous selfing, first, second, and third flower visited by B. philenor) (Bates et al. 2015). We included plant individual as a random effect in the model. To compare fruit set across treatment groups, we conducted pairwise post-hoc Tukey-Kramer tests using the glht() function in the multcomp R package (v1.4-20) (Hothorn, Bretz, and Westfall 2008).