Prepollination barriers prevent gene flow between co-occurring bat-pollinated bromeliads in a montane forest

Study site

The bromeliads studied were located at a montane forest ecosystem in the Central Valley of Costa Rica (9°52′–9°54′N and 83°57′–84°00′W). La Carpintera Protective Zone is a small mountain formation between 1,500 and 1,850 m asl with nearly 2,400 hectares in size (35% primary forest and 57% old secondary forest) (Sánchez-González, Duran & Vega, 2008). The study was conducted on the property of Iztarú Field School from the Association of Guides and Scouts of Costa Rica and under permission granted by the former station administrator, Mr. Minor Serrano. The zone receives an average annual rainfall of 1,839.2 mm, with a mean temperature of 16.1 °C, and a mild drier season runs from December to April (Ríos & Cascante-Marín, 2017); however, the presence of fog is frequent during the night and early morning. The Life Zones System of Holdridge (1978) classifies the vegetation as both humid and very humid lower montane forest. The local diversity of vascular epiphytic plants represents nearly one-third of the local flora, and bromeliads contribute with 29 species; represented by the genera Aechmea (1 spp.), Catopsis (3 spp.), Guzmania (3 spp.), Pitcairnia (1 spp.), Racinaea (2 spp.), Tillandsia (11 spp.), Vriesea (1 spp.), and Werauhia (7 spp.) (Sánchez-González, Duran & Vega, 2008).

Study species

We selected the four most abundant Werauhia species based on previous research on the flowering phenology of epiphytic plants at the study site (Cascante-Marín, Trejos & Alvarado, 2017): W. ampla (L. B. Sm.) J. R. Grant, W. nephrolepis (L. B. Sm. & Pittendr.) J. R. Grant, W. pedicellata (Mez & Wercklé) J. R. Grant, and W. subsecunda (Wittm.) J. R. Grant. These epiphytic species develop small to medium-sized rosettes, simple spiked (W. ampla and W. subsecunda), or compound inflorescences (W. nephrolepis and W. pedicellata) (Fig. 1). These bromeliads share the same pollinator at the study site, the nectar-feeding bat Hylonycteris underwoodi (Phyllostomidae) and exhibit a highly self-compatible mating system with an autonomous delayed mechanism of selfing (Núñez-Hidalgo & Cascante-Marín, 2024). Their geographic distribution mostly encompasses the very humid and cloudy forests between 1,000 and 2,750 m asl on the Talamanca Mountain range in Southern Central America between Costa Rica and Panama (Morales, 2003).

Prepollination mechanisms

Postpollination mechanisms

To estimate the strength of postpollination barriers, we followed the formula proposed by Sobel & Chen (2014): RI_4C = 1 – 2(H/H + C), where H = hetero-specific events (percentage of fruits, seed number or seed germination from interspecific manual crosses), and C = conspecific events (percentage of fruits, number of seeds or seed germination from intraspecific manual crosses). Using the previous formula, we calculated the indices corresponding to: (i) inter-specific incompatibility or RI_I (pre-zygotic barrier), (ii) hybrid seed production or RI_S (post-zygotic barrier) and (iii) seed viability or RI_V (post-zygotic barrier). The RI value indicates the amount of interspecific gene flow at each stage, where −1 indicates the presence of inter-specific pollen flow (absence of barriers), 0 indicates random pollen flow, and 1 indicates complete isolation of gene flow between species. Data on intraspecific events (manual cross-pollinations) were obtained from a related work on the breeding systems of the study species at the same site (Núñez-Hidalgo & Cascante-Marín, 2024).

Contribution of each barrier type and total reproductive isolation

Each reproductive barrier contributes to isolation in proportion to the order in which they occur in the plant’s life. As a result, the first barrier to action will reduce gene flow, implying a greater contribution to reproductive isolation (Coyne & Orr, 1989; Ramsey, Bradshaw & Schemske, 2003). We used the formulas proposed by Sobel & Chen (2014) and provided in their supplementary material (evo12362-sup-0003) to estimate the relative and absolute contribution of each type of barrier and total reproductive isolation between each pair of species. We also identified which category of isolation barrier (pre- or postpollination) contributes more to reducing gene flow between species. The calculations were performed using an Excel spreadsheet provided by Sobel & Chen (2014) and included here as a Supplemental File.

Isolation by floral phenology

The species exhibited an annual pattern of flowering (sensu Newstrom, Frankie & Baker, 1994) but the intensity and distribution of flowering peaks varied among species and between years (Fig. 2). The blooming periods were seasonally divided: W. ampla, W. pedicellata, and W. subsecunda mostly bloomed in the dry season and W. nephrolepis flowered separately during the rainy season (Fig. 2). Werauhia ampla and W. subsecunda showed the longest reproductive periods (5–6 months), both with a pattern of constant intensity or “steady-state” (sensu Janzen, 1967) and flowering peaks of relatively low intensity (usually <40% of the observations) (Fig. 2). W. pedicellata showed a defined bimodal pattern and a low to moderate overlap with the two previous species. The flowering of W. nephrolepis was of short duration with a very marked peak resembling the “cornucopia type” pattern (sensu Janzen, 1967) and temporally isolated from the rest (Fig. 2).

Estimations of temporal isolation between species-pair combinations using the four-year average value of the index were very variable, ranging from 0.128 to 0.991 (Table 1). It was the lowest between W. ampla and W. subsecunda (RI_F = 0.128 and 0.258), and for all paired comparisons involving W. nephrolepis, it indicated strong temporal isolation in both directions (RI_F ≥ 0.97) (Table 1). There were some variations between years in the strength of this barrier for some species pairs (Table S1).

Isolation by floral morphology

Differences in floral morphometry among species were statistically significant (PERMANOVA test: r² = 0.932; DF = 3, 118; F-value = 544.67; p-value = 0.001). Similarly, all paired comparisons among species were significant (p = 0.006) which indicates substantial variation in flower size. For instance, the length of the corolla, pistil and stamens in W. ampla and W. nephrolepis were 2‒3 times longer compared to W. subsecunda and W. pedicellata (Fig. 3A). The PCA biplot suggested two distinctive groups based on dimensions of floral parts, large (W. ampla and W. nephrolepis) vs. small-flowered (W. subsecunda and W. pedicellata) species (Fig. 3B). The first component explained most (92.48%) of the variation in the data, with stamens and pistil length showing the highest scores (Table S2). These size differences in reproductive structures were deemed to represent complete isolation (RI_MS = 1; Table 1), since such morphological dissimilarity precludes effective interspecific pollen transfer between those two species groups.

The species W. ampla, W. nephrolepis, and W. pedicellata share the same position of stamens and stigma on the dorsal part of the corolla aperture (Fig. 4), suggesting the probability of gene flow (RI_MP = 0). In W. subsecunda the pistil and stamens separate into two triplets that locate in lateral position at the corolla aperture (Fig. 4C). This conformation of the reproductive organs represents a strong barrier to gene flow with respect to the other species, thus paired comparisons involving W. subsecunda were assigned a complete reproductive isolation for this aspect of floral morphology (RI_MP = 1) (Table 1).

Isolation by interspecific incompatibility

Reciprocal crosses between W. subsecunda and W. pedicellata showed partial interspecific incompatibility and resulted in fruit percentages of 37% and 47.8% (RI_I = 0.097 and 0.348, respectively; Table 2). Crosses involving W. ampla only produced fruits with W. subsecunda, when the latter species acted as pollen recipient (54.5%, RI_I = 0.168), and it represented a case of asymmetric incongruity. Because of full incompatibility, the reproductive barriers between W. ampla and W. pedicellata were complete (RI_I = 1).

Isolation by hybrid progeny unviability

When compared to the respective intraspecific crossings (Table 2), the number of hybrid seeds per fruit from reciprocal crosses between W. subsecunda and W. pedicellata resulted in a reduction of 48% and 66.4%, respectively. This represented a relatively low to moderate isolation barrier between the two species (RIs = 0.316 and 0.414, respectively). When W. subsecunda (the pollen recipient) crossed with W. ampla, fruits produced about a third as many hybrid seeds as fruits from intraspecific crosses of the same species (214 vs. 636 seeds, respectively; Table 2). This represented a moderate barrier to reproduction (RI_S = 0.497).

The germination capacity of hybrid seeds from W. subsecunda sired with pollen from W. pedicellata was high (92.3%; Table 2). This resulted in a nearly absent isolation barrier (RIv = 0.034). On the contrary, hybrid seeds from W. pedicellata sired with pollen from W. subsecunda did not germinate as well (31.9% vs. 92.5% for intraspecific crosses) (Table 2) and represented a moderate isolation barrier (RIv = 0.487). Hybrid seeds from W. subsecunda and W. ampla (as pollen donor) had lower viability compared to seeds from intraspecific crosses of the latter species (73.1 vs. 98.8%; Table 2). This loss of viability represented a relatively weak isolation barrier (RIv = 0.149).

The role of temporal barriers

Research has shown that flowering time can have varying effects on plant reproductive isolation, ranging from minimal to significant (mean RI = 0.375, Christie, Fraser & Lowry, 2022). Our study found that differences in flowering time were an important barrier to gene flow among the Werauhia species studied, with a mean RI_F of 0.677. The overall strength of this barrier varied between years from 0.624 (2020–2021 season) to 0.712 (2012–2013 season), though indicating rather modest variation. Between species pairs, however, variation in isolation strength of flowering time ranged from as low as 0.128 to nearly complete isolation at 0.991, reflecting the diversity of flowering patterns at the study site. Similarly, the strength of this barrier varied among years between some species pairs (Table S1). This is the result of interannual variation in flowering patterns, which is primarily attributed to alterations in local climate or larger meteorological events that affect plant flowering (McNeilly & Antonovics, 1968; Frankie, Baker & Opler, 1974; Marquis, 1988; Elzinga et al., 2007). Thus, the importance of considering several flowering episodes to obtain a better RI estimation since year-to-year fluctuations may alter the magnitude of this reproductive barrier.

Our results showed that over half of the species-pair comparisons exhibited nearly complete reproductive isolation attributed to non-overlapping flowering phenology. Conversely, the remaining comparisons demonstrated either weak or moderate isolation (RI_F = 0.13–0.56). However, in those instances of incomplete isolation due to temporal reproductive overlap, isolation was subsequently enhanced by a more effective premating barrier linked to floral morphology (discussed further). In two species pairs (W. nephrolepis-W. pedicellata and W. nephrolepis-W. subsecunda), we found almost complete reproductive isolation by non-overlapping phenology, while the floral morphological barriers were also significant yet redundant.

In a guild of sympatric bat-pollinated bromeliads from the genera Pitcairnia, Pseudalcantarea, and Werauhia in southern Mexico, Aguilar-Rodríguez et al. (2019) documented non-overlapping phenologies which apparently prevented interspecific pollen transfer. However, for closely related species growing in sympatry, the flowering time may be constraint by shared ancestry (Rathcke & Lacey, 1985). In our case, three species of Werauhia bloomed during the dry season and one in the rainy period, while other sympatric species from the study site, W. notata and W. haberi (Cascante-Marín, Trejos & Alvarado, 2017; Cascante-Marín, Trejos & Morales, 2019), also flowered during the rainy season. The observed variation in flowering patterns suggests an absence of phylogenetic constraint regarding reproductive timing and suggests phenotypic plasticity that may contribute to mitigate reproductive interference.

Mechanical barriers related to flower morphology

Differences in flower architecture can also prevent gene flow by influencing how pollen is deposited on the pollinator’s body (Grant, 1994). Reproductive isolation through morphological differences in flower size plays a major role among some sympatric neotropical groups (Kay, 2006; Ramírez-Aguirre et al., 2019; Albuquerque-Lima, Lopes & Machado, 2024). In our study, species conformed into two groups: large (W. ampla and W. nephrolepis) and small-flowered species (W. subsecunda and W. pedicellata). The size disparity was nearly 2- to 3-fold between both groups, suggesting that flower-visiting bats would contact the anthers and stigma on different areas of their bodies, thereby preventing interspecific pollen flow. The pollen of the large-flowered Werauhias is likely deposited on the bat’s head, while for small-flowered species, it is carried on the face (forehead and cheeks) (Fig. 4).

For species exhibiting comparable floral dimensions and overlapping phenology, such as W. pedicellata and W. subsecunda, differences in the positioning of anthers and stigma in relation to the corolla aperture are key in preventing interspecific pollen transfer. The lateral positioning of the anthers in W. subsecunda flowers likely causes pollen to be deposited on the bat’s cheeks, while in W. pedicellata, pollen is probably deposited on the top region of the snout and forehead. In a previous work on the reproductive systems of the studied species (Núñez-Hidalgo & Cascante-Marín, 2024), we recovered pollen from Werauhia found on the head and snout of captured bats, but we could not identify the species it originated from due to the morphological similarities of the pollen grains.

Research has shown that in taxonomically unrelated plants pollinated by bats, the differential placement of pollen on the pollinator’s body serves as an effective reproductive barrier to interspecific pollination (Tschapka, Dressler & von Helversen, 2006; Muchhala & Potts, 2007; Muchhala, 2008; Stewart & Dudash, 2016). Recently, Pontes, Machado & Domingos-Melo (2024) illustrated how various strategies of pollen deposition on the pollinator’s body contribute to facilitating the reproductive coexistence of a guild of chiropterophilous plants in a Neotropical dry forest. Muchhala (2008) proposed that the evolution of this isolation mechanism is particularly facilitated in bat-pollinated plants, attributed to the larger size of bats relative to other pollinator groups, enabling more accurate pollen deposition on specific areas of their bodies. Similar studies on hummingbird-pollinated bromeliads could determine whether reproductive isolation involving pollen-placement strategies are also significant in this pollination guild or are exclusive to bat flowers.

Our estimations of the strength of mechanical barriers due to floral morphology (RI_MS and RI_MP) were based on clear-cut and statistically significant differences in size of corolla and reproductive organs (Fig. 3). Even though the differences in the position of stamens and stigma were sufficiently intuitive to assume the existence of a full impediment to interspecific pollen transfer, both measurements should be interpreted as conservative estimates. The high similarity in pollen grain morphology and size overlap among the studied species (unpublished data) precluded any analysis based on detecting differential pollen deposition on stigmas because of unreliable identification. Further experimental studies of specific pollen deposition on the pollinator body using pollen dyes or novel techniques (Minnaar & Anderson, 2019) may corroborate our interpretation.

Interspecific incompatibility and hybrid progeny

The examined postpollination barriers showed a lower strength compared to prepollination barriers, with an average RI value of 0.225 vs. 0.614, respectively. The results of the reciprocal crosses indicated that interspecific incompatibility plays an inconsistent role in preventing gene flow among species pairs. In two of the three possible reciprocal crosses, the absence of fruit production reflected a marked incompatibility or incongruity (Vervaeke et al., 2001), although it was not symmetrical in all cases. For example, unilateral incompatibility (Lewis & Crowe, 1958) was observed in crosses between W. subsecunda and W. ampla, when the former acted as a pollen recipient. However, reciprocal crosses between W. subsecunda and W. pedicellata revealed partial incompatibility, which allowed the production of hybrid progeny. These permeable postpollination barriers have also been documented among bromeliad species of the genera Aechmea (Bromelioideae), Pitcairnia (Pitcairnioideae) and Vriesea (Tillandsioideae) (Parton et al., 2001; Wendt et al., 2002; Souza et al., 2017). The precise location and mechanisms of operation of this barrier are unknown.

Covas & Schnack (1945) explained the phenomenon of unilateral incompatibility by proposing a positive relationship between pollen size and pistil length in the two species. Stroo (2000), in a compilation of studies on bat-pollinated plants, found a positive correlation between pollen size and stigma length. They suggested that the pollen grain needs to accumulate sufficient resources for tube growth as it traverses the stigma to reach the egg cell (Levin, 1971; Cruden, 2009). Moreover, the size and depth of the stigma also play a role, as the pollen grain can draw resources for tube growth from the stylar liquid (Cruden, 2009; Wang et al., 2016). In this context, the larger pollen of W. ampla (62–75 μm) successfully reached the ovary of W. subsecunda, which has a shorter pistil. Conversely, the smaller pollen of W. subsecunda (50–64 μm) was apparently unable to traverse the longer pistil of W. ampla. Although this needs confirmation through an analysis of pollen tube growth, this pattern of unilateral incompatibility has also been observed in crosses between congeneric bromeliads of the genera Aechmea, Alcantarea, and Vriesea (Vervaeke et al., 2001; Matallana et al., 2016; Souza et al., 2017).

Further isolation barriers related to seed vigor also showed varying levels of effectiveness in preventing gene flow. These levels ranged from low to moderate; however, similar to interspecific incompatibility, they were also redundant.

Are there other potential barriers to gene flow among the studied Werauhia?

Although we did not directly address the following parameters in this experiment, we recognize that they may contribute to the reproductive isolation of the studied species, and we discuss them below. Variations in microhabitat specialization could serve as a spatial reproductive barrier among sympatric species (Schluter, 2001); however, we lack quantitative data regarding the spatial distribution or microhabitat preferences of the Werauhia species in question. Based on our field observations, the examined species tend to occupy similar microhabitats, typically colonizing the lower sections of tree trunks and branches within the inner areas of tree crowns, thus the role of this barrier is likely minor or null.

A temporal barrier related to the time of flower anthesis may constitute an additional mechanism of isolation (Levin, 1971). In the studied species, flowers open during the same period in the late afternoon (between 16–17 h) and before the nocturnal pollinator is active (Núñez-Hidalgo & Cascante-Marín, 2024), thus it does not constitute an isolation mechanism. Autogamy or the ability to spontaneously self-fertilize (i.e., selfing) has been proposed by Levin (1971) as a reproductive barrier. For the Bromeliaceae family, Matallana et al. (2016) suggested that selfing was a mechanism to avoid hybridization, due to the high frequency of self-compatibility and autogamy among bromeliads. In a previous study, we found that our studied species showed high levels of autonomous self-fertilization (Núñez-Hidalgo & Cascante-Marín, 2024). This study demonstrated that selfing occurs at the end of the flower’s life (i.e., delayed selfing) after the opportunities for cross-pollination have diminished, primarily serving as a reproductive assurance mechanism. Additionally, differences in the number of chromosomes or ploidy levels between species may represent a postpollination barrier to prevent the formation of hybrid progeny (Stebbins, 1950; Grant, 1981). Polyploidy has been reported in several groups of bromeliads (McWilliams, 1974; Brown & Gilmartin, 1983, 1986; Gitaí, Horres & Benko-Iseppon, 2005) but basic information on chromosome numbers is lacking for Werauhia species in general.