Flowers of Berberis and Mahonia have six petals and six stamens inserted between paired nectar glands (organ sizes are given in Supplementary file 1). Each pollen sac opens by a separate valve and contains about 610 ± 6 sticky yellow pollen grains that remain attached to the pollen sac (Figure 1, Figure 1—figure supplement 1A, Figure 1—figure supplement 2). When a flower visitor (or a pointed object) contacts the adaxial surface of a filament base, the respective stamen rapidly moves toward the flower center, placing pollen grains on the visitor’s tongue (Figure 1A and B, Figure 1—figure supplement 2E and F). The stamen movement takes 0.44 ± 0.02 s in Berberis julianae, 0.17 ± 0.02 s in B. jamesiana, and 0.23 ± 0.04 s in B. forrestii (Supplementary file 1). Within 1 min, the stamen moves back from the flower center, where the single style with its large stigma is located, to the petal, taking 227.70 ± 10.06 s, 110.37 ± 6.64 s, and 155.31 ± 14.07 s, respectively, to return its original position (Supplementary file 1). In Mahonia bealei, the stamen movement takes 0.09 ± 0.01 s (N = 10 flowers), and 3.46 ± 0.71 s later, the stamen starts moving back, taking 7.74 ± 1.96 s to return to its original position (Video 2).

In B. julianae, fruit set in open-pollinated flowers (80.0 ± 6.4%) was significantly higher than in bagged and self-fertilized flowers (27.5 ± 7.1%; Figure 2). However, seed set per fruit did not differ between manually self-pollinated and cross-pollinated flowers. Self-pollen receipt by stigmas of B. julianae after a single stamen movement (14 ± 3 grains, N = 16) was only 6% of the pollen receipt resulting from a single visit by the most common visitors, Apis cerana, which deposited between 230 and 260 grains (section ‘Effects of forward-snapping stamens on visitor behavior and pollination in the 22 field and under-enclosed conditions’) and roughly 1% of the pollen grains of a single anther with its two pollen sacs (ca. 1220 grains), indicating that intra-flower self-pollination mediated by the stamen movements plays a minor role in total pollen receipt.

Figure 2

Download asset Open asset

At our study site, B. julianae was visited mainly by five insect species (Supplementary file 2), the bee A. cerana Fab., 1793 (Figure 1A), two anthophorid bees Anthophora waltoni Cockerell, 1910, and Habropoda sichuanensis Wu, 1986 (Figure 1B), and the syrphid flies Rhingia campestris Meigen, 1822, and Meliscaeva spec. In the field, it was not always possible to securely distinguish the anthophorid bees, although H. sichuanensis was clearly more frequent, and some of our results, for example, on visitation rates, therefore pool these species. These bees foraged for nectar, but not pollen, while the flies fed on both nectar and pollen (Figure 1C). Consistent with these feeding habits, pollen transfer efficiency of the bees was significantly higher than that of the flies (Supplementary file 2).

We found no effect of any lingering alcohol scents (in stamen-immobilized flowers) on visitor behavior: Visitation rates of A. cerana to B. julianae (under enclosed conditions) did not differ between untreated flowers with mobile stamens (SM), flowers with immobilized stamens (SI) via alcohol immersion, and untreated flowers in a fixed position above alcohol, called SMA flowers (Wald χ² = 0.194, df = 2, p=0.908; Figure 3A). However, A. cerana stayed longer in SI compared to SM and SMA flowers (Wald χ² = 64.599, df = 2, p<0.001; Figure 3B), showing that it was the forward-snapping stamen that caused these bees to leave. Visitation rates of A. cerana to B. julianae (again under enclosed conditions) also did not differ among SM, SI, and filament-damaged (FD) flowers (Wald χ² = 0.44, df = 2, p=0.802; Figure 3C) because all three types of flowers offered the nectar sought by these bees. However, bees stayed less time in SM flowers than in flowers without stamen bending (SI and FD flowers) and equally long in SI and FD flowers (Figure 3D), showing that it was the stamen forward-snapping per se that caused visitors to move on.

Figure 3

Download asset Open asset

In the field, when three or four SI flowers of B. julianae were inserted on racemes with the same number of SM flowers (Figure 1H), A. cerana stayed much longer on SI flowers than on SM flowers (15.46 ± 1.54 s vs. 3.63 ± 0.33 s, Wald χ² = 56.055, p<0.001 in 2020; 16.076 ± 1.515 s vs. 5.675 ± 0.382 s, Wald χ² = 68.421, p<0.001 in 2021). Despite the longer visits, fewer pollen grains were loaded onto A. cerana after a single visit to SI flowers than to SM flowers (716 ± 85 vs. 1223 ± 100 grains in 2020; 890 ± 73 vs. 1401 ± 134 grains in 2021; Wald χ² = 12.873, p<0.001 in 2020; Wald χ² = 10.511, p=0.001 in 2021), and the numbers of pollen grains deposited on stigmas by a single A. cerana visit also were much lower in SI flowers than in SM flowers (87 ± 10 vs. 230 ± 25 grains in 2020; 63 ± 8 vs. 260 ± 35 grains in 2021; Wald χ² = 26.847, p<0.001 in 2020; Wald χ² = 40.042, p<0.001 in 2021; Figure 5, Figure 5—figure supplement 1). Pollen transfer efficiency by A. cerana was therefore reduced in SI compared to SM flowers, and this was significant in 2021 but not 2020 (0.084 ± 0.016 vs. 0.303 ± 0.102; Wald χ² = 4.505, p=0.034 in 2021; 0.205 ± 0.024 vs. 0.246 ± 0.091; Wald χ² = 0.188, p=0.665 in 2020; Figure 5—figure supplement 1I and J).

When the experiment was repeated under enclosed condition (using five SI and five SM flowers), the bees and syrphid flies all stayed longer on SI than on SM flowers (33.7 ± 4.2 s vs. 16.0 ± 1.7 s; Wald χ² = 30.106, p<0.001; Figure 4E′–T). All visitor species exploited more nectar (Figure 4M′–P) and touched more stamens in SI than in SM flowers (Figure 4I′–L). The pollen transfer efficiency of the bees was higher than that of the flies (Wald χ² = 13.319, df = 3, p=0.004, N = 80). In SI flowers, A. cerana was loaded with fewer grains than in SM flowers (just as under outdoor conditions; above), while the anthophorid bees and the flies under enclosed conditions removed more grains from SI flowers than from SM flowers (Figure 4Q′–T).

Figure 4

Download asset Open asset

After experimental single visits, both A. cerana (Figure 5A, Wald χ² = 142.565, p<0.001) and the anthophorids (Figure 5B, Wald χ² = 14.236, p<0.001) carried fewer pollen grains on their tongues after visiting SI flowers compared to SM flowers.

Figure 5 with 1 supplement see all

Download asset Open asset

In three trials in which we held anesthetized bees between forceps and simulated visits to SM and SI flowers in different sequences (‘Materials and methods’), the number of pollen grains deposited on stigmas by A. cerana (mean ± SE, Figure 5C) and the anthophorids (mean ± SE, Figure 5D) in trial 1 (SM + SM flower) was significantly higher than in trials 2 (SM flower + SI flower) and 3 (SI + SI flowers; Wald χ² = 118.887, p<0.001 vs. Wald χ² = 69.274, p<0.001, respectively). Moreover, pollen receipt by A. cerana and the anthophorids was significantly higher (p<0.001) in trial 2 (SM + SI) than in trial 3 (SI +SI), indicating that the stamens precisely place pollen grains on tongues, which then deposit them on stigmas during the insect’s next flower visit.

Pollen export of SM and SI flowers was quantified with pollen-tracking experiments in which pollen of flowers of B. julianae and B. jamesiana was stained in situ with either eosin or aniline blue, and stigmas of all flowers in the vicinity (about 25–100 cm from the source plant) were checked for stained grains.

Of over 700 flowers of B. julianae whose stigmas we checked, 44 of 772 flowers received pollen from SM flowers and 14 of 733 flowers from SI flowers (Supplementary file 3), indicating that flowers with mobile stamens donated pollen to about 3.1× more flowers (44/772 = 0.057 vs. 14/733 = 0.019, G = 15.341, p<0.001). The mean number of pollen grains deposited per stigma in these trials was also higher from SM than SI flowers (0.16 ± 0.03 vs. 0.03 ± 0.01; Wald χ² = 76.536, p<0.001). All four runs of this experiment showed a consistent pattern: more pollen grains from SM flowers were delivered to more flowers (Figure 6, Figure 6—figure supplement 1).

Figure 6 with 1 supplement see all

Download asset Open asset

Within <50 cm from the dyed pollen source, 25 of 338 flowers received pollen from SM flowers, while only 10 of 395 flowers received pollen from SI flowers (G = 9.645, p=0.0019, Supplementary file 3). At distances of 50–100 cm, 19 of 434 flowers received pollen from SM flowers, while only 4 of 338 flowers received pollen from SI flowers (G = 7.439, p=0.0064, Supplementary file 3), and at distances >100 cm, only pollen from SM flowers was detected on stigmas (on 6 of 222 inspected flowers).

When we conducted the same experiment in B. jamesiana (Figure 6—figure supplement 1), 75 of 400 nearby flowers received pollen from SM flowers and 40 of 400 flowers from SI flowers, again indicating that mobile stamens were likely to donate pollen to more flowers (75/400 vs. 40/400, G = 12.612, p<0.001). The mean number of pollen grains deposited per stigma was also higher in SM than SI flowers (0.4 ± 0.05 vs. 0.2 ± 0.04; Wald χ² = 22.95, p<0.001).

Within <25 cm from the dyed pollen source, 13 of 60 flowers received pollen from SM flowers, while 9 of 46 flowers received pollen from SI flowers (G = 0.07, p=0.791). At distances >25 cm, 7 of 60 flowers received pollen from SM donors, while only 1 of 56 flowers received pollen from SI flowers (G = 4.961, p=0.026, Supplementary file 4).

High-performance liquid chromatography (HPLC) of the berberine concentrations in B. julianae leaves, petals, pollen, and nectar indicated high berberine concentrations in leaves, petals, and pollen, while no berberine was detectable in the nectar (Supplementary file 5). That bees can taste the berberine in the pollen is suggested from the observation that individuals of all bee species used their front legs to remove pollen grains that stuck to their tongues.

Even though botanists have speculated about the adaptive value of the visitor-triggered forcefully forward-snapping stamens of Berberis since 1755 (Linnaeus, 1755), this is the first experimental investigation of how this trait impacts the flowers’ pollen export and receipt. Our results demonstrate surprisingly large effects of stamen bending on pollen export (involving both quantity and distance) and nectaring times (involving lower nectar costs per pollen grain transfer) and reveal another mechanism by which plants meter out their pollen.

During each March between 2019 and 2022, we studied a natural population of B. julianae C.K.Schneider in a field located at 29°52′26″N, 105°30′32″E, 427 m above sea level, about 50 km southeast of Anyue County, Sichuan Province, China. Experiments to evaluate the effect of stamen movements on pollen dispersal were also carried out in a natural population of B. jamesiana Forrest & W.W.Smith at Shangri-La Alpine Botanical Garden (27°54′05″N, 99°38′17″E, 3300–3350 m above sea level), Yunnan Province, Southwestern China. We also studied planted populations of B. forrestii Ahrendt at the Shangri-La field station and of M. bealei (Fortune) Carrière in the Wuhan Botanical Garden (30°33′2″N, 114°25′48″E, 23 m above sea level) in Hubei Province to test the effects of the alcohol treatment on stamen mobility. Our Berberis and Mahonia taxonomy follows the Flora of China (Ying et al., 2011). Herbarium vouchers of each species have been deposited in the herbarium of Central China Normal University (CCNU). All species are hermaphroditic perennial shrubs with clusters of 9–25 yellow flowers produced between March and May, depending on species. Individual flowers last for 3–5 days, and each anther dehisces upward by two valves exposing the pollen grains (Figure 1). The bottle-shaped pistils have one ovary containing 2–4 ovules with a discoid stigma with a peripheral ring of papillae (Figure 1E).

Insect visitors were observed, in some cases filmed, captured with insect nets in the field, and preserved for later identification by insect taxonomists.

When we immersed floral pedicels in a solution of 75% alcohol for 40 min, all stamen movement was blocked (Videos 1 and 3 of B. julianae and Video 2 of M. bealei). To test whether any lingering alcohol scent affected visitor behavior in alcohol-treated flowers, we set up arrays with different types of flowers as follows. We first bagged >20 flowers on different individuals of B. julianae before they opened. Once open, 18 flowers were gently cut off and subjected to one of the three treatments: (1) stamen-mobile (SM) flowers: six flowers without any treatment; (2) stamen-immobilized (SI) flowers: six flowers whose pedicels (ca. 10 mm long) were immersed in 75% alcohol for about 40 min; and (3) six natural flowers in a fixed position above alcohol (SMA flowers). For this, the pedicels of freshly opened flowers were inserted into 30-mm-long Eppendorf microcentrifuge tubes that contained 10 μL alcohol, such that they did not touch the alcohol (Figure 1—figure supplement 1A). All 18 flowers were then fixed in position by inserting them into small holes on the surface of a paper box covered by a clean glass cup (as shown in Figure 1—figure supplement 1). We then placed freshly caught Berberis visitors inside the glass cup so they could interact with the enclosed floral arrays and recorded visitation rates (visits per flower per 10 min) and handling times.

Video 1

Download asset

Video 2

Download asset

Video 3

Download asset

To further examine the possible effects of the alcohol treatment versus the loss of stamen movements on the duration of insect visits, we set up additional floral arrays consisting of three types of flowers: six SM flowers, six alcohol-treated SI flowers, and six FD flowers, the latter being flowers in which all six filaments were damaged with clean forceps so that the stamens could not move but petal nectaries and the anther sacs were still there (Figure 1—figure supplement 1C). The experimental procedure for these arrays was the same as above (Figure 1—figure supplement 1B).

To test whether B. julianae is self-compatible and whether fruit or seed production are pollinator-limited, in 2020 and 2021, 80 flowers from 12 individuals were subjected to the following four pollination treatments. (1) Control: 20 randomly chosen flowers from 12 individuals were open pollinated, while the remaining 60 flowers were bagged and subjected to one of three treatments: (a) automatic self-pollination: flowers not manipulated; (b) self-pollination: flowers hand-pollinated with pollen mixtures collected from flowers of the same individual; (c) cross-pollination: flowers hand-pollinated with pollen mixtures from multiple flowers of individuals at least 20 m away. Flowers were then bagged with mesh until the petals dropped 1 week later. Fruits were harvested 3 months later and seeds and undeveloped ovules per fruit counted. Fruit set was calculated as fruit number divided by flower number in each treatment. Seed set was calculated as seed number per fruit divided by total number of seeds and undeveloped ovules. Aborted fruits were not included.

To examine intra-flower selfing induced by the stamen movement, we counted pollen grains deposited per stigma under a light microscope in 16 bagged flowers in which we had triggered one stamen by a needle.

Flowers of B. julianae were visited by several species of bees and two species of syrphid flies (‘Results’). To examine the effects of stamen bending on foraging behavior and pollen transfer, floral visitors were allowed to interact with the above-described flower arrays (section ‘Alcohol as an inhibitor of the stamen movement, and tests for confounding the effects of the alcohol treatment on visitor behavior’) in the field (Figure 1) and under enclosed conditions (Figure 1—figure supplement 1D). Under enclosed conditions, we compared visitation rates, handling times, numbers of stamens touched, nectar volume remaining, pollen removal, and pollen receipt on the stigma after single visits of the four species of visitors. We also quantified pollen transfer efficiency after single visits by harvesting flowers that had been visited a single time in the field and then counting pollen grains on their stigmas as well as pollen grains remaining in their stamens. To quantify pollen loads placed on insect bodies and stigmas of next-visited conspecific flowers, we held captured and anesthetized bees between forceps and made their tongues contact the filament bases in SI or SM flowers. Using this method, we carried out three trials with SM and SI flowers: trial 1, SM + SM; trial 2, SM + SI; and trial 3, SI + SI flowers.

To quantify pollen export from SM and SI flowers, we conducted four pollen-tracking experiments each in B. julianae in March 2022 and B. jamesiana in May 2021. Each trial was conducted on a sunny day and involved 60 flowers from 3 to 5 individuals whose pollen was stained as the pollen donor: 30 flowers were alcohol-treated (SI flowers) and the remaining 30 flowers were SM flowers. The manipulated flowers were carefully inserted among natural flowers on densely blooming racemes. The stigmas of >100 flowers (126–260) on the same and on different individuals at varying distances from the donor were then examined for dyed pollen grains, these were counted, and the straight-line distance from donor to recipient recorded with a meter rule.

Visitation rates of visitor species in the field (not normally distributed) were examined with nonparametric Kruskal–Wallis tests. To compare visitor behavior on SM, SMA, FD, and SI flowers, we performed a generalized linear model (GLM) analysis with normal distributions and identity-link function to test for differences in visitation rates, handling time, and pollen transfer efficiency (log10-transformed) of the four main visitor species. Nectar volumes remaining after different visitors had visited were examined with a nonparametric Kruskal–Wallis test. To compare the number of stamens touched in different flowers, pollen removal, and pollen deposition by each visitor species, and pollen export and import, we performed GLM analyses with a Poisson distributions and loglinear-link function. Floral traits among species were compared under a GLM with a normal distribution and identity-link function, while for numbers of pollen ovules per flower we used a Poisson distribution and loglinear-link function (Supplementary file 1). Data of pistil height and stamen length were log10-transformed to achieve normal distribution. We used the G-test of independence to test whether stained pollen from SM or SI donors differs in the distances to which it is dispersed to (McDonald, 2014). A GLM with binomial distribution and logistic-link function was used to detect the effects of the selfing and outcrossing on fruit set and seed set (with fruit/seed number as event variable, total treated flower/ovule number as trial variable, and different treatments as factors). The GLM analyses were performed in SPSS 19.0 (IBM, Armonk, NY).

Flower size, length of the pedicel, and other flower traits were measured with a digital caliper to 0.01 mm (Supplementary file 1). To estimate nectar volume per flower of B. julianae, clean 10 µL microliter syringes (Agilent Technologies Inc, USA) were used to extract nectar drops from bagged flowers. To count pollen grains and ovules per flower, we randomly collected one of six anthers from virgin flowers that had been bagged as buds with fine-mesh cotton bags to exclude visitors. Each selected anther was placed on a microscope slide and then squashed under a coverslip. All pollen grains were counted under a microscope, and the number in one anther was multiplied by six to obtain the pollen grain number per flower. Meanwhile, ovules of sampled flowers were also counted. Sample sizes are shown in Supplementary file 1.

To find the minimum time required to completely block stamen movements in each of the three species of Berberis, we checked the response of stamen movements in a time series of alcohol-treated flowers. Pedicels were gently cut off and the bases (about 10 mm long) immersed in 75% alcohol. Every 5 min, 8–10 flowers were checked by touching the filaments of each flower with a dissecting needle to identify whether the stamens remain mobile. For B. julianae, we tested 64 flowers, for B. jamesiana 100, and for B. forrestii 80.

Visitation rates (visits/flower/hour) of the different visitors to B. julianae were obtained on sunny days in March 2019, 2020, and 2021. Visitor were observed on fully flowering shrubs with over 100 open flowers, and each observation period lasted 1 hr, during 9:00–12:00 and 12:30–16:30. The sex and foraging behavior of each visitor were recorded, such a feeding on nectar or pollen or groomed pollen grains into pollen loads. Total observation times in the 3 years are shown in Supplementary file 2.

To see whether floral visitors discriminated between SI and SM flowers, we compared visitation rates of the Asian honeybee A. cerana under open pollination and under enclosed conditions. The other insect species were only studied under enclosed conditions (illustrated in Figure 1—figure supplement 1).

In March 2020, 72 flowers on different individuals of B. julianae were randomly bagged with mesh nets, before they opened. Also, 12 SI and 12 SM flowers were observed each day. When flowers were beginning to open, we gently cut off 36 flowers (12 flowers × 3 days) at the base of the pedicels and immersed the pedicels in 75% alcohol. All stamens in each alcohol-treated flower became touch-insensitive in 40 min. These SI flowers were carefully inserted into racemes of flowering individuals with the other 36 SM flowers on these racemes as controls, allowing the bees to visit in open pollination conditions (Figure 1D and H). The number of flowers visited per hour was recorded during 09:00–12:00 and 13:00–17:00 for 3 days. Visitation rates were calculated as the number of flowers visited per hour divided by the number of observed flowers, that is, visits per flower per hour.

In March 2021, a total of 160 floral buds from five individuals were randomly bagged. As above, 80 newly opened flowers were alcohol-treated and the remaining 80 flowers were free of alcohol as SM flowers. On each observation day, we set up four arrays each having four SM and four SI flowers. The visitation rates were recorded during 09:30–12:30 and 13:00–17:00 for 5 days each with 16 SI and 16 SM flowers (Figure 5—figure supplement 1A-B).

To compare the handling time, pollen removal, and pollen transfer efficiency of the pollinator A. cerana in SM and SI flowers under open conditions, 100 flowers from different individuals of B. julianae were randomly bagged with mesh before the flowers opened in March 2020 and 2021. When the flowers were newly opened, 12 flowers per day were alcohol-treated for 5 days per year with fine weather. Four of the SI flowers were carefully inserted into each of three racemes with buds, but no open flowers (4 flowers × 3 racemes × 5 days = 60), and four bagged SM flowers from two racemes were uncovered per day (4 flowers × 2 racemes × 5 days = 40) for A. cerana visits. In the five arrays, each of four flowers per day, we recorded handling times of A. cerana during each visit to one flower. To estimate pollen transfer efficiency, 20 SI and 20 SM flowers visited once by A. cerana were harvested immediately in the field. Six anthers and the stigma of each flower were collected. We counted pollen grains remaining within anthers per flower and deposited per stigma under a light microscope. Pollen removal per flower was calculated as the mean number of pollen grains (see ‘Results’) minus the remaining grains. Pollen transfer efficiency was calculated as pollen deposition divided by pollen removal. In both years, we obtained 20 pairs of pollen removal and deposition data for SM flowers, and 20 pairs for SI flowers. In March 2021, we also compared the number of stamens touched and the remaining nectar volume left by A. cerana in SM and SI flowers that it had it visited just once. As visitor switched to collect nectar from another nectary, their bodies turned. Therefore, counting the visitor’s body turns allowed us to record the number of stamens touched by a visitor in SI flowers (Video 3).

To estimate nectar collection by different visitors, we measured the remaining nectar volume within each SM and SI flower after one visit using a clean 10 µL microliter syringe (Agilent Technologies Inc).

To compare the foraging behavior of bees and flies on SM and SI flowers under enclosed conditions (Figure 1—figure supplement 1), we set up floral arrays each consisting of five SM and five SI flowers. The flowers were fixed by inserting the pedicels into small holes on the surface of a paper box covered by a clear glass cover. We then allowed individual of the different insect species to freely interact with the floral array. We recorded the visitation rates (visits per flower per 10 min) and handling time of the visitor and the number of stamens touched by the visitor in each flower, measured the nectar volume left by the visitor, and assessed pollen removal and pollen deposition after one visit to calculate pollen transfer efficiency. It took four trials of ‘interviews’ (cafeteria experiments) over 2 days to yield 20 once-visited flowers for each insect species. Four individuals of each insect species were captured in the field and released within 2 hr after their visits to the SM and SI flowers.

To estimate the effects of the stamen mobility on pollen placement on the pollinator and on stigmas, we simulated bee visits by to SI and SM flowers. In March 2021, we bagged 320+ flower buds from different individuals of B. julianae. When the flowers opened, we held anesthetized bees between forceps so that they probed the flower, pollen grains placed on the bee’s tongue (and occasionally on the head) were pasted onto a piece of tape. The tape was attached to a slide and pollen grains placed on the bee’s tongue in 20 SI and 20 SM flowers were then counted under a light microscope (Figure 5A and B). Using this method, we also carried out three trials with simulated visits to flowers in specific sequence: Trial 1, SM + SM flowers; trial 2, SM + SI; and trial 3, SI + SI flowers. Each trial was repeated 20 times. After each trial, the stigma of the second-visited flower (the pollen recipient) was collected to count the pollen grains deposited on its stigma (Figure 5C and D).

To examine the effects of stamen movements on the fate of pollen, we conducted four trials of pollen-tracking experiments in a transplanted population of B. jamesiana at Shangri-La Alpine Botanical Garden, where the interference of stained pollen with the sexual reproduction of wild plants could be minimal. Trials were conducted on sunny days and involved 60 flowers from 3 to 5 individuals whose pollen was stained as pollen donors: 30 flowers were alcohol-treated (SI flowers) and the remaining 30 flowers were used as SM flowers. Pollen grains in SM or SI flowers were stained with eosin (stained red) or aniline blue (blue). The two dyes were alternately used between SM and SI flowers in four trials. The dyes dried within 5–10 min, and all pollen-stained flowers were taken to the field and carefully inserted into racemes of two flowering individuals along the roadside for pollinators to visit. To track pollen flow, the two clusters containing either 30 SM or SI flowers were arranged separately (about 100 cm apart) on the flowering branches. Flowers of each cluster were within a 40 cm × 40 cm square on four erect racemes. Previous studies indicated that pollen flow mediated by generalist insects usually occurs within meters, with a highly leptokurtic distribution (Williams, 2001). One day (24 hr) later, given that A. cerana visits were infrequent to B. jamesiana in May 2021, stigmas of 100 flowers from nearby racemes were collected. Dyed pollen grains deposited on each stigma were counted under a stereomicroscope.

To compare the distance of pollen dispersal from SM and SI flowers, we conducted two trials of pollen-tracking experiments to compare distance of pollen dispersal between SM and SI flowers of B. jamesiana. Each trial was the same as above. At 17:30 on the day of the experiment, stigmas of all flowers from nearby racemes were collected and we noted the straight-line distance within 25 cm and over 25 cm from the racemes with the sampled flowers to the racemes with pollen-stained SM or SI flowers. Dyed pollen grains deposited on each stigma were counted under a stereomicroscope.

To examine whether berberine is present in Berberis pollen and nectar, we collected leaves, petals, pollen, and nectar from 10 plants of B. julianae. Leaves, petals, and pollen grains were dried using an oven at 65°C, while the nectar was stored at –20°C before chromatographic analysis. The berberine content of the different samples was analyzed using HPLC. To extract berberine from leaves, a 0.1 g leaf sample was weighed with a balance (Sartorius BAS124S) and transferred to a 2 mL microcentrifuge tube. A steel bead was added to the tube. The leaves were ground using a tissue homogenizer (Tissuelyser, QIAGEN) at 30 Hz for 10 min, and then the leaf sample was transferred to a vial. Enzymolysis of the leaf sample was conducted by adding 2 mL pure water, 1 μL dilute sulfuric acid, and 6 mg cellulose at 50°C in a magnetic stirrer for 3 hr. The leaf sample was extracted for 2.5 hr in 5 mL pure water, and then transferred to a 5 mL microcentrifuge tube. Following centrifugation at 8000 rpm for 10 min (Centrifuge 5430R, Eppendorf), the supernatant was transferred to a vial. The volume of the sample was recorded. Berberine in petals, nectar, and pollen grains was measured similarly.

Components were separated using an Acquity HPLC BEH C18 column (50 × 2.1 mm, 1.7 μm) (Waters, Milford, MA) set at 30°C, and the injection volume was 20 μL. All aspects of system operation and data acquisition were controlled by software (Agligent 1100 series) at the Center of Analysis and Test of Wuhan University, Wuhan, China. The mobile phase was acetonitrile–0.3% phosphoric acid–pure water (35:5:60); flow rate: 1.0 mL/min; determination wavelength: 346 nm; the detection signal was diode array detector (DAD). The berberine concentrations in the different samples were compared using a GLM with normal distribution and identity-link function.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We thank lab members Q-M Quan, X-W Lv, and Z-X Tian for field assistance and Z-Y Tong, and Y-Z Xiong for methodological and statistical advice; Dr. Huan-Li Xu, Department of Entomology, China Agricultural University, Beijing, for the identification of flies and bees; Dr. Antti Haarto, Zoological Museum, University of Turku, Finland, for the identification of Meliscaeva; Dr. Michael Orr, Institute of Zoology, Chinese Academy of Sciences, Beijing, for the identification of Habropoda cf. sichuanensis; Dr. Chih-Chieh Yu for confirming our plant identifications; Z-D Fang and staff of Shangri-La Botanical Garden for logistical support; Sarah Corbet, Steven Johnson, and Nathan Muchhala for providing helpful suggestions on early versions of this manuscript; and the reviewers Bernhard Schmid, Zong-Xin Ren, and Felipe Yon for their constructive comments. This work was supported by the National Natural Science Foundation of China (grant nos. 31730012 and 32030071) and Fundamental Research Funds for the Central Universities (no. CCNU22LJ003) to S-QH.

© 2022, Li et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.