Polygamous breeding system identified in the distylous genus Psychotria: P. manillensis in the Ryukyu archipelago, Japan

Study plants and sites

P. manillensis (Bartl. ex DC.) is a woody shrub, 2–3 m in height, commonly found in limestone evergreen forests with alkali soil and widely distributed from the Ryukyu Islands to the Lanyu and Ludao Islands of Taiwan and the Philippines (Yamazaki, 1993). Inflorescences are terminal and each possesses 20–150 flowers (Fig. 1). Corollas are white to white-green and have a short funnel shape, 2–3 mm in length, with white hairs inside. Flowers bloom for a single day; they open in the morning and fade in the evening. Normally, one to five flowers in a single inflorescence bloom each day, and one inflorescence blooms for about two weeks. The flowering season is from June to August, and fruits mature from November to January on Okinawa Island.

We examined three populations on Okinawa Island: the Katsuu population at Mt. Katsuu–dake (26.63°N, 127.94°E, 310 m above sea level (a.s.l.)), the Oppa population at Mt. Oppa–dake (26.67°N, 127°96′E, 250 m a.s.l.) and the Sueyoshi population at Sueyoshi Park (26.23°N, 127.72°E, 65 m a.s.l.), and two populations located on Iriomote Island: the Sonai population at the Sonai beach forest (24.37°N, 123.75′E, 5 m a.s.l.) and the Uehara population at the Uehara forest (24.39°N, 123.81°E, 40 m a.s.l.). These three populations on Okianwa Island and two populations on Iriomote Island are representative of limestone forest of those islands, and populations are well separated by other types of forest. All field studies were performed from 2011–2013. Field experiments were approved by Okinawa Prefecture (permit number 23) and Naha city (permit number 272).

Classification of stigmas and anthers of flowers

We collected 294 flowers from 20 individuals (14 or 15 flowers per individual) in the Sueyoshi population in June–August 2011. We observed the sexual organs of all the collected flowers under a light microscope (SMZ800; Nikon, Tokyo, Japan) and the unfixed stigmas under a scanning electron microscope (SEM, S-3000N; Hitachi, Tokyo, Japan) to examine their surface structure.

We classified the stigmas of P. manillensis into three types based on the degree of development of the stigmatic papillae: I, well-developed; II, moderately developed; and III, poorly developed (Fig. 2). We hypothesized the type III stigma does not have female function. We also classified the anthers of P. manillensis into three types based on the size of pollen sacs and the amount of pollen: a, anthers with large pollen sacs (usually more than 10 mm in length) full of pollen grains; b, anthers with small pollen sacs (usually less than five mm in length) with half or less of the total capacity of pollen grains; and c, anthers with no pollen sac nor grains (without male function) (Fig. 3). Thus, there were nine possible combinations of stigma and anther types; Ia, Ib, Ic, IIa, IIb, IIc, IIIa, IIIb, IIIc (Fig. S1).

Function of stigmas

To evaluate the function of each stigma type, we cross-pollinated 103 flowers with type-I stigmas (N: Number of individuals = 21), 111 flowers with type-II stigmas (N = 20), and 96 flowers with type-III stigmas (N = 21) at the Oppa and Katsuu populations by transferring pollen from other individuals. Because the stigmas of some flowers were hidden within the corolla tubes, we partially cut the floral tube for pollination. Approximately 50 d after pollination, we counted the number of immature fruits to evaluate the fruit set percentage.

To determine the ability of pollen tubes to access the styles and ovaries, 22 flowers with type-I stigma (N = 10), 17 flowers with type-II stigma (N = 10), and 18 flowers with type-III stigma (N = 10) were cross-pollinated in the laboratory. After 24 h, the materials were fixed in 5:5:90 formalin: glacial acetic acid: 70% ethanol, and then transferred to 70% ethanol for storage. The fixed styles were softened in eight N NaOH at 4 °C for 24 h, washed with water, transferred to 0.005% aniline blue in Na₂HPO₄ (pH 11), and stored at 4 °C for at least 24 h. The styles were mounted on a slide in glycerol beneath a cover glass and gently pressed to spread the tissue. Pollen grains on the stigmas and pollen tubes in the styles were observed under a fluorescence microscope (Axio Imager Z2; Zeiss, Oberkochen, Germany). Then, the numbers of pollen tubes which reached at the base of the style were counted.

Flower collection and counts of flowers by type

We collected flowers from 20 previously marked individuals from each of the five populations in 2012 and from the three populations on Okinawa Island in 2013. We visited each plant at least five times (days), each covering the flowering period (early, mid and late), and collected all available flowers (50–396 flowers each). We classified them into the nine flower types as defined above (see Fig. S1). Then, we classified the individuals by gender based on the flower sexual type(s) at the time of collection: female individuals (only female flowers), male individuals (only male flowers), monoecious individuals (both male and female flowers and often with some perfect flowers), and hermaphroditic individuals (only perfect flowers). To calculate the relative femaleness (G) of each plant, we used proportions of female, male and perfect flowers in total number of flowers.

Although these proportions cannot represent the “functional femaleness” of the individuals exactly (Lloyd, 1980), they can be the indicator of relative femaleness. We used the following formula to calculate the relative femaleness (G):

G = (F + P/2)/(F + M + P), where F, P and M are the number of female, perfect and male flowers per plant, respectively. Since the perfect flowers have both female and male function, we assumed the perfect flower has the half of the femaleness of the female flower.

Measurements of floral traits

Male and female flowers of dioecious species that were derived from distylous ancestors often still have reciprocal positioning of the anther and stigma height. To understand the relationship between morphological traits (anther and stigma heights) and sexual types of flowers (to check for remnants of distylous traits in this species), we measured the anther and stigma height of 12–15 flowers from six individuals (78 flowers in total) of different genders (one female, one male, one hermaphroditic and three monoecious plants) in the Sueyoshi population.

Self-compatibility and agamospermy test

To test self-compatibility, we self-pollinated 136 flowers (34 Ic, 63 IIc, and 31 IIIa) from 31 individuals in the Oppa population. Inflorescences were bagged 1 d before pollination, and open flowers in bags were pollinated with pollen grains from other flowers in a bagged inflorescence of the same plant. Then, the self-pollinated flowers were bagged again, and the fruit set was observed approximately 30 d after pollination. To test agamospermy (apomixis), we emasculated 10 flowers from nine individuals (90 flowers in total) before anthesis and observed the fruit set approximately 50 d after the treatment.

Fruit set under open pollination and bagging experiments

Fruit set under open pollination was surveyed for three to five inflorescences of 9–31 individuals (average of 19.8 individuals per site per year) from the three populations on Okinawa Island in 2011 and all five populations on Okinawa and Iriomote Islands in 2012–2013. We counted the number of flower buds in inflorescences and the number of immature fruits on the marked inflorescences approximately 50 d after the flowering peak.

A bagging experiment was carried out to test for spontaneous self-pollination/fertilization in 2012. One inflorescence each of 10, 11, and 18 individuals from the Sueyoshi, Katsuu, and Oppa populations, respectively, was bagged using a mesh bag before blooming, and the fruit set ratio (proportion of flowers that produced fruit) was calculated after all flowers faded (approximately 60–70 d after the flowering peak).

Flower phenology

We marked 13 inflorescences on 10 individuals in the Oppa population (130 inflorescences in total) and recorded the number and sexual type of open flowers every 2 d from June 26 to August 3, 2012. We divided the flowering period into four spans and calculated the relative femaleness (G) as described above for each span and plant.

Gender shift over the years

We tagged 21, 18, and 19 plants from Katsuu, Oppa and Sueyoshi populations on Okinawa Island respectively for multiple year tracking of the relative femaleness (G) and the gender of each individual.

Flower visitors

Flower visitors were observed in all five populations over 8 h (from the time flowers opened until they closed; usually between 0700 and 1,800), and some individuals were collected for identification. Observations were performed on July 17 and 18, 2011 and on July 14 and 20, 2012 in the Katsuu population; from July 6 to September 18, 2011 and on July 14, 2012 in the Oppa population; on July 9 and 27, 2011, on June 19 and July 1, 2012 in the Sueyoshi population; and on June 2, 3, 15, 16, and 17, 2011 and on June 13, 2013 in the Sonai and Uehara populations.

Statistical analysis

We used R v. 4.1.0 (R Core Team, 2021) for all statistical analysis.

The effects of stigma types and pollination treatments to fruit set were tested using the generalized linear mixed models (GLMMs, fitting data to a binomial error distribution with log-link function), using the ‘glmer’ function of the R package ‘lem4’. To test the effect of stigma type to fruit set, the GLMMs included stigma type (I, II, and III) as fixed explanatory variables, fruit set as response variable and individual plants as random effect. To test the effects of pollination treatment to fruit set, GLMMs included pollination treatment (cross and self-) as fixed explanatory variables, fruit set as response variable and individual plants and stigma type as random effects. The goodness of fit to models was assessed by the maximum likelihood-ratio chi-square test, using ‘anova’ function of the R package ‘lme4’. To evaluate the effect of each stigma type, we performed a Tukey post hoc test with the ‘glht’ function of the package ‘multcomp’.

We also analyzed the factors that affect fruit set in the field using GLMM (binomial error distribution with log-link function). For the open pollination, the models included the fruit set as the response variable, year and population as the explanatory variables and individual plant as a random factor. We compared four models with different combinations of the explanatory variables: model 1 (population), model 2 (year), model 3 (population and year), and model 0 (null). The model fit was evaluated with the Akaike information criterion (AIC), and the model with the smallest AIC was regarded as the best fit for our data. Then, we tested the effect of the bagging experiment against the open pollination. In this case, the models included the fruit set as the responsible variable, treatment (open/bagging) as the explanatory variables and individual plant and population as nested random factors. Subsequently, we performed the maximum likelihood-ratio chi-square test.

We performed a one-sided Pearson’s correlation test using ‘cor.test’ function of R to evaluate the association between femaleness (G) and fruit set under open pollination in each population. We also performed a two-sided Wilcoxon test (using ‘wilcox.test’ function in R) to test the difference in relative femaleness of individual between years 2012 and 2013.

Function of stigmas and anthers by type

After cross-pollination, fruit set was observed in 64.1% and 65.8% of flowers with type-I and type-II stigmas respectively, but only 2.1% of flowers with type-III stigmas (Table 1). Similarly, pollen tubes reached at the base of styles in 100% (22/22) of type-I stigmas and 82% (14/17) of type-II stigmas, but no pollen tube growth was observed in flowers with type-III stigma (0/18).

Of the nine possible flower types (based on combinations of stigma and anther types, see Fig. S1), eight flower types (namely Ib, Ic, IIa, IIb, IIc, IIIa, IIIb, and IIIc, but not Ia) were found in the field. All flowers had anthers and a stigma, but more than 93% of the flowers differentiated to either male (IIIa and IIIb, 34%) or female (Ic and IIc, 59%). Only 6.5% of the flowers were perfect (Ib, IIa, and IIb) and only 0.5% were sterile (IIIc); no Ia type flowers were observed (Fig. 4). The ratio of type-I and type-II stigmas was 92:8 for the female flowers and 8:92 for the perfect flowers. The proportion of flower types was relatively consistent in all five populations, although the number of perfect flowers was relatively low in the Uehara population (Fig. 5A).

Monoecious individuals were the most common (54%) in the three populations, followed by female (30%), male (14%), and hermaphroditic (2%) individuals (Fig. 5B). Female individuals usually consisted of only Ic (female) flowers that had a high stigma with low anthers (Fig. S2A), whereas male individuals consisted of IIIa (male) flowers that had a low stigma and high anthers (Fig. S2B). Of the monoecious individuals, most of the flowers were Ic (female) and IIIa (male), but those types were not distinct when mixed with perfect flowers (Figs. S2D–S2F).

Self-pollination and agamospermy test

The fruit set after self-pollination was significantly lower (36.2%) than that after cross-pollination (65.0%; P < 0.001, GLMM followed by likelihood ratio test; Table 1). Of 18 studied individuals, 13 showed self-compatibility. Additionally, none of the emasculated flowers set any fruit, revealing that agamospermy did not occur.

Fruit set under open pollination and bagging experiments

The fruit set under open pollination varied among populations and years (1.8–21.9%; Table S1), and the model 3 which includes both populations and years as explanatory variables showed the smallest AIC (binomial GLMM; Table S2). Some male individuals (only male flowers at the time of flower collection) also produced fruit, although femaleness (G) was significantly correlated with fruit set in all populations (P < 0.05, Pearson’s correlation test; Fig. S3), except for the Katsuu population (P = 0.35). Some flowers set fruit after bagging experiments, although the fruit set was significantly lower than that under open pollination (P < 0.0001, the likelihood ratio test after binomial GLMM; Tables S1, S2).

Flowering phenology and flower types within inflorescence through flowering period

The flowering period of 10 individuals observed in the Oppa population was approximately 42 d (Fig. S4A). Femaleness (G) increased in seven individuals and decreased only in one plant during the flowering period, whereas two individuals produced only female flowers (Fig. S4B). Of 127 inflorescences, 44 (35%) had only female flowers, 44 (35%) had male, female, and perfect flowers, 23 (18%) had male and female flowers, 14 (11%) had male and perfect flowers, and two (1.6%) had only male flowers (Fig. 6).

Gender shift over the years

Of the 58 checked plants, relative femaleness increased in 24 plants (41%) and decreased in three plants (5%, Fig. 7A). Although the majority of female and monoecious plants kept the same gender in 2012–2013 season, 20 plants (35%) changed their sex (Fig. 7B). Five female plants changed to monoecious, six monoecious plants to female, one hermaphroditic plant to monoecious, seven male plants to monoecious, and one male plant to female.

Flower visitors

A total of 39 insect species of 20 families belonging to six orders visited the flowers of P. manillensis in the five populations (Table S3, Fig. 8). Of these insects, wasps of the family Vespidae (Vespa and Polistes) and flies of the families Calliphoridae, Syrphidae, and others were the most frequent visitors, and usually pollen grains were attached to their heads. Wasps visited the flowers, sucked nectar, and left quickly for other flowers, whereas flies stayed relatively longer after licking the nectar. Bees (Apis mellifera, Amegilla spp., Apidae and Lasioglossum, Halictidae) also visited the flowers, but much less frequently than wasps and flies. Bees touched both the stigmas and anthers (usually pollen attached to their heads or bodies) and moved quickly from one flower to another, Butterflies (Papilio, Lampides, and Athymaperius) and moths (Crambidae) also visited flowers on Okinawa Island, but only occasionally touched the anthers or stigmas with their proboscises. Hawkmoths (Macrogrossum), beetles, stink bugs, and crickets were occasional visitors. Only syrphid flies (Allobaccha nubilipennis and Sphaerophoria sp.) fed on pollen, whereas all others fed on nectar.