Journal of Research https://doi.org/10.1007/s10265-020-01226-8

REGULAR PAPER – ECOLOGY/ECOPHYSIOLOGY/ENVIRONMENTAL BIOLOGY

Post‑pollination barriers enable coexistence of pollinator‑sharing ornithophilous

Anina Coetzee1 · Claire N. Spottiswoode1,2 · Colleen L. Seymour1,3

Received: 6 June 2020 / Accepted: 23 September 2020 © The Botanical Society of Japan 2020

Abstract Some evolutionary radiations produce a number of closely-related species that continue to coexist. In such plant systems, when pre-pollination barriers are weak, relatively strong post-pollination reproductive barriers are required to maintain spe- cies boundaries. Even when post-pollination barriers are in place, however, reproductive interference and pollinator depend- ence may strengthen selection for pre-pollination barriers. We assessed whether coexistence of species from the unusually speciose Erica in the fynbos biome, South Africa, is enabled through pre-pollination or post-pollination barriers. We also tested for reproductive interference and pollinator dependence. We investigated this in natural populations of three bird-pollinated Erica species (Erica plukenetii, E. curvifora and E. coccinea), which form part of a large guild of congeneric species that co-fower and share a single pollinator species (Orange-breasted Sunbird Anthobaphes violacea). At least two of the three pre-pollination barriers assessed (distribution ranges, fowering phenology and fower morphology) were weak in each species pair. Hand-pollination experiments revealed that seed set from heterospecifc pollination (average 8%) was signifcantly lower than seed set from outcross pollination (average 50%), supporting the hypothesis that species boundaries are maintained through post-pollination barriers. Reproductive interference, assessed in one population by applying outcross pollen three hours after applying heterospecifc pollen, signifcantly reduced seed set compared to outcross pollen alone. This may drive selection for traits that enhance pre-pollination barriers, particularly given that two of the three species were self-sterile, and therefore pollinator dependent. This study suggests that post-pollination reproductive barriers could facilitate the coexistence of congeneric species, in a recent radiation with weak pre-pollination reproductive barriers.

Keywords Breeding system · Cape foristic region · · Hybrid · Nectariniidae · Pollen interference

Introduction Relatively strong reproductive barriers are necessary to maintain species boundaries among such coexisting and Although evolutionary radiations often result in geographi- recently radiated species (Mallet 2005). Reproductive bar- cally separated species, some produce a diversity of coex- riers include premating and postmating isolation barriers isting, closely-related species (Weber and Strauss 2016). (Arceo-Gómez et al. 2016; Nosil et al. 2007). The latter can impose prezygotic isolation (via sperm competition and Electronic supplementary material The online version of this gametic incompatibility) and/or postzygotic isolation (e.g. article (https​://doi.org/10.1007/s1026​5-020-01226​-8) contains via reduced ftness in hybrids) (Rieseberg and Willis 2007). supplementary material, which is available to authorized users. In , most species are preserved by multiple reproduc- tive isolation barriers, and prezygotic barriers are thought * Anina Coetzee [email protected] to be about twice as strong as postzygotic barriers (Lowry et al. 2008; Rieseberg and Willis 2007). Thus, understand- 1 FitzPatrick Institute of African Ornithology, DSI‑NRF ably, studies on the coexistence of closely-related species Centre of Excellence, University of Cape Town, Private Bag have predominantly focussed on premating barriers (Weber X3, Rondebosch 7701, South Africa and Strauss 2016). However, studies from hummingbird- 2 Department of Zoology, University of Cambridge, Downing pollination systems and other plant communities with high Street, Cambridge CB2 3EJ, UK species diversity suggest that the frequency of heterospecifc 3 South African National Biodiversity Institute, Kirstenbosch pollen transfer can be high (Da Fonseca et al. 2016; Fang Research Centre, Claremont 7753, South Africa

Vol.:(0123456789)1 3 Journal of Plant Research and Huang 2013), in which case postmating barriers become predominantly pollinated by a single species, the endemic crucial. Orange-breasted Sunbird Anthobaphes violacea (Nectari- In plants, reproductive barriers are most easily distin- niidae) (Rebelo et al. 1985). In some communities, South- guished as pre- and post-pollination barriers. Pre-pollina- ern Double-collared Sunbird Cinnyris chalybea and Mala- tion barriers include geographical separation, segregation chite Sunbird Nectarinia famosa pollinate these , but of fowering periods or pollinator isolation (Rieseberg and again, pollination is carried out almost exclusively by a Willis 2007). The latter requires diferent pollinator guilds single bird species within these sites (Malan 2013; Rebelo (e.g. birds, insects or wind), pollinator species or pollen et al. 1984, 1985; Turner et al. 2012; Van der Niet et al. placement sites (through diferential fower morphology), 2014). This guild of bird-pollinated ericas often co-fower or can arise when pollinator individuals show fower con- (Rebelo et al. 1984). Almost all have straight or curved stancy (foraging selectively on one plant species) (Grant tubular corollas > 15 mm in length (Rebelo et al. 1985), 1994; Waser 1986). When pre-pollination barriers fail, but vary in the length of their reproductive parts (Anina plants can receive heterospecifc pollen and hybridisation Coetzee, unpublished data). The breeding systems of only could occur (Rieseberg and Willis 2007). This could lead to nine bird-pollinated Erica species have been tested, but of genetic swamping (Rieseberg and Willis 2007), particularly those, three displayed self-compatibility and only one was in closely-related species with little genetic divergence (Lar- partially capable of autonomous self pollination (Angoh combe et al. 2015; Mallet 2005). Coexisting closely-related et al. 2017; Arendse 2015). Self-sterility in this genus species are thus likely to experience reduced plant ftness seems to be a function of a late-acting mechanism within when interspecifc mating occurs, favouring the evolution of the ovule (Arendse 2015): embryo development is arrested post-pollination reproductive barriers (Arceo-Gómez et al. after fertilisation, which prevents wasting resources on 2016), which include premating barriers at the stigma, and selfed seeds but carries the cost of reducing the number pre- and postzygotic barriers downstream of the stigma. of ovules available for cross fertilisation (Gibbs 2014). Even when post-pollination barriers are efcient, the risk Regular natural hybridisations in sunbird-pollinated of reproductive interference remains when pre-pollination orchid and protea species have been reported in the CFR barriers are weak. Post-pollination reproductive interference (Johnson 2018; Mitchell and Holsinger 2018). The success- results in reduced seed set through interactions at the stigma, ful interspecifc crossing and cultivating of many Erica spe- style and/or ovule (Brown and Mitchell 2001; Nishida et al. cies by eighteenth century horticulturists (Nelson and Oliver 2014), which may drive selection for stronger pre-pollina- 2004) reveals that hybridisation is indeed possible between tion barriers (Morales and Traveset 2008). The deposition of certain Erica species, too. Currently, the rate of interspecifc incompatible pollen onto stigmas before, or together with, pollen transfer in the natural system is unknown, but aviary compatible pollen is a common form of post-pollination experiments show that Orange-breasted Sunbirds forage ran- reproductive interference in co-occurring, closely-related domly between diferent Erica fower morphotypes (Heystek species (Weber and Strauss 2016). Furthermore, condi- et al. 2014). If such random foraging occurs between species tions that drive selection for traits that enhance pollinator as well, these ericas would experience at least low levels of efectiveness (e.g. a fower morphology that increases pol- heterospecifc pollen transfer. len placement accuracy or fower constancy) may also, as a Here, we test whether coexistence of bird-pollinated by-product, strengthen pre-pollination barriers (Hopkins and Erica species is enabled through pre-pollination or post- Rausher 2012). Such conditions include pollinator depend- pollination barriers, and whether they experience processes ence. Pollinator dependence results from one or more of an that could drive selection for stronger pre-pollination barri- inability to reproduce (1) asexually, (2) through self pol- ers. We estimate the reduction in potential gene fow through lination (using external vectors) or (3) through autonomous pre-pollination barriers (separation of geographic range, selfng (without an external pollen vector; Lloyd et al. 1992). fowering phenology and reproductive fower parts) and Therefore, an assessment of the conditions that strengthen post-pollination barriers (by assessing seed set from hetero- pre-pollination barriers is also important for understanding specifc pollination relative to outcross pollination). Next, how closely-related species coexist. we use hand-pollination experiments to detect two condi- In the relatively small Cape Floristic Region (CFR, tions (reproductive interference and self-incompatibility) 90 000 ­km2), we frequently fnd coexisting congeneric that would promote evolution of pre-pollination barriers. species due to the extensive major radiation of several First, for E. coccinea L., we test for reproductive interference clades (Linder 2003), of which Erica (Ericaceae) is by by quantifying seed loss when applying outcross pollen three far the largest (Pirie et al. 2011). Some Erica species are hours after heterospecifc pollen has been applied to fowers. almost certainly reproductively isolated by their use of Second, we test for self-compatibility by comparing seed diferent pollinators (Heystek and Pauw 2014; Rebelo set from self and autonomous pollination to that produced et al. 1985). However, the 67 bird-pollinated species are by outcross pollination. Taken together, these tests aim to

1 3 Journal of Plant Research provide insights into the mechanisms that enable coexistence study sites (Cape Point and Brodie Link Nature Reserve). of congeneric species that share a single pollinator. Since phenological barriers may vary in space (Johnson 1993), a phenological isolation index was calculated for each site where -1 indicates complete disassortative mating, 0 Materials and methods indicates random mating and 1 indicates complete reproduc- tive isolation due to a phenological barrier (Sobel and Chen Study species 2014; equation 4S2). Flower morphological isolation was quantifed as the We conducted experiments in 2016 and 2019 at three natural linear distance between the maternal species’ stigma and sites, on three bird-pollinated Erica species (E. plukenetii the pollen donor species’ anther. This distance represents subsp. plukenetii Oliver (2002), E. curvifolia L. and E. coc- the likelihood of heterospecifc pollen deposition since the cinea subsp. coccinea L.), each of which co-fowers and stigma and anther length may indicate the areas of pollen grows intermixed with another bird-pollinated congeneric at deposition and reception (Da Fonseca et al. 2016; Much- each site (Tables 1, S1, Fig. 1). We applied hand-pollination hala 2008; Turner et al. 2012) on sunbirds as they probe experiments on two populations of E. coccinea at diferent the narrow, tubular fowers which do not restrict access to sites that co-occur with diferent Erica species. The two E. sunbirds. Pistil lengths and anther lengths (from corolla base coccinea populations could resprout (multiple stems at the to stigma and anther pore, respectively) were measured with base of the plant), while E curvifolia and E. plukenetii are digital callipers from two fowers per plant and averaged non-resprouting and rely on seed for regeneration after fre. (27–38 plants per species). Although the extent to which stigma-anther distances reduce pollen transfer has not yet Pre‑pollination barriers been quantifed, we compared heterospecifc distances to conspecifc distances. For each maternal species plant, we We assessed three pre-pollination barriers: geographical, calculated the mean distance (mm) between the maternal phenological and fower morphological isolation. Large- species’ stigma and the anthers of heterospecifc individuals, scale geographical isolation was estimated from distribu- as well as the mean distance between the maternal species’ tion range data at quarter degree square resolution, with an stigma and the anthers of conspecifc individuals. Paired area of 23 × 23 km (Oliver and Forshaw 2012; Oliver et al. Wilcoxon signed rank tests were applied. 1983). For each maternal species (the pollen-receiving spe- cies, Table 1), we calculated the percentage of its quarter Hand‑pollination experiments degree cells that is not shared with the pollen donor spe- cies. Flower phenological isolation between species pairs For each species, 20–25 plants were used for hand-pollina- was determined by recording fower abundance of 26–37 tion experiments. Five pollination treatments were applied: individual plants once a month for 6 months at two of the outcross pollination (pollen transferred from other plants

Table 1 Study species and sites of coexisting bird-pollinated Erica species Study site Maternal species Heterospecifc pollen Geographic Phenological Post-polli- IAS ISI donor species isolation (%) isolation nation RI

Brodie Link Nature Reserve E. plukenetii plukenetii E. coccinea coccinea 38 0.91 0.76 0.0 0.05 Brodie Link NR E. coccinea coccinea E. plukenetii plukenetii 27 0.91 0.80 0.01 0.03 Cape Point E. coccinea coccinea E. abietina atrorosea 95 0.13 0.85 0.03 0.14 Somerset West E. curvifora E. plukenetii plukenetii 51 – 0.49 0.12 0.42

Pre-pollination barriers for maternal species were estimated as the quantity of unshared geographical space and fowering time. Indices of post- pollination reproductive isolation (RI), autonomous selfng (IAS) and self-incompatibility (ISI) are provided for the maternal species (formulas provided in the footnotes) 100 U Geographic isolation = T  , U = number of quarter degree cells not shared with the pollen donor species, T = total number of quarter degree cells Phenological isolation = equatiion 4S2 in Sobel and Chen (2014); 0 indicates random mating and 1 indicates complete reproductive isolation 1 − 2 H Post-pollination RI = C+H  , H = average heterospecifc pollination seed set, C = average cross pollination seed set IAS = A C  , A = autonomous self pollination seed set ISI = S C  , S = self pollination seed set

1 3 Journal of Plant Research

Fig. 1 Bird-pollinated Erica species and the sites at which pollination Brodie Link NR; e E. curvifora and f E. plukenetii at Somerset West experiments were conducted: a E. coccinea at Brodie Link NR; b E. (scale bar = 20 mm) coccinea and c E. abietina atrorosea at Cape Point; d E. plukenetii at of the same species), heterospecific pollination (pollen randomly allocated branch. This accounts for diferences transferred from another species), self pollination (pollen between individual plants but assumes that the efects of transferred from the same plant), autonomous pollination resource allocation are limited by diferent branches with (no pollen manually added, but vector-less self pollination leaves functioning as integrated physiological units (Wes- is possible) and natural pollination. We treated a subset of selingh 2007). For each treatment, three to four fowers with fowers per plant so that all treatments could be applied receptive stigmas were treated on an inforescence (recep- to each plant, with each treatment applied to a separate, tive stigmas identified by stigmatic glossiness or slight

1 3 Journal of Plant Research colour change) and marked on the calyx with a permanent The treated fowers’ mature fruit capsules were collected marker pen. The number of fowers per branch difered little before opening (4–8 weeks later), dissected in the lab, and between branches. No fowers were removed, to avoid creat- seeds, aborted seeds and ovules counted with a Leica Zoom ing a bias in seed set due to disturbance in natural resource 2000 dissecting microscope under 20 × magnification. allocation. Inforescences with fower buds were bagged Seeds were distinguished from ovules by their larger size in advance with an organza material bag that allows sun- (E. plukenetii and E. curvifora), glossy texture (E. coccinea) light and air fow but prevents access by birds and insects, or “honey-combed” seed testa (E. plukenetii). Aborted seeds and limits wind pollination. After a treatment was applied, were intermediate in size, and the testa texture difered from the inforescence was again covered by the exclusion bags seeds. We excluded fruits that were partly or heavily dam- (except for the natural pollination treatment) to prevent her- aged by insects or fungi (< 18% for all treatments except bivory. Flowers lasted multiple days and their exact age and for the reproductive interference treatment with 41%). The herkogamy at the time of the experiment was not measured. average (± standard error) fruit content (ovules + viable and Outcross and heterospecifc pollination treatments were non-viable seed) were as follows: E. coccinea 303 ± 7, E. applied to test for post-pollination barriers. Pollen was col- plukenetii 211 ± 4 and E. curvifora 86 ± 4. lected by rupturing the anther ring of donor fowers directly We tested the diference in seed set (number of seeds per into a clean petri dish. A clean, fne paintbrush was used to fruit as a proportion of the total seeds, aborted seeds and mix the pollen and brush onto the stigma of experimental ovules) between treatments with a generalised linear mixed- fowers. Separate petri dishes and paint brushes were used efect model with a binomial error structure (glmer function for each treatment. For the cross pollination treatment, pol- from the lme4 package (Bates et al. 2015); R version 3.6.0). len was collected from at least three surrounding conspecifc Plant identity was added as a random factor, and we used an plants (2–10 m away from experimental plants) that were not observational level factor to account for overdispersion. This used as experimental plants (a maximum of two fowers per was followed by multiple comparisons of means with Tukey plant). Individual plants are only recruited after fre from contrasts (glht function from the multcomp package; seed or resprouting (thus almost all plants in a population Hothorn et al. 2008). Post-pollination reproductive isolation 1 − 2 H are the same age), and since these species do not reproduce was calculated as C+H  , where H is average hetero- clonally, separate plants are unlikely to be genetically identi- specifc pollination seed set and C is average cross pollina- cal individuals. Pollen was collected in the same way from tion seed set (Sobel and Chen 2014). This estimates the at least two heterospecifc plants for the heterospecifc pol- proportional reduction in gene fow compared to what might lination treatment. Seed set from heterospecifc pollination be expected if there were no reproductive isolation. We also was also compared to seed set from natural pollination to S calculated indices of self-incompatibility C  (Lloyd and estimate the ftness reduction relevant to the natural pollina- A tion conditions experienced. We marked open inforescences Schoen, 1992) and autonomous selfng C  for each species, without exclusion bags and allowed natural pollination. where S is seed set from self pollination and A is autono- Orange-breasted sunbirds were observed probing fowers mous selfng seed set. on experimental plants with exclusion bags on them, and therefore their pollination rate on open fowers should be similar to natural conditions. Three ripe and undamaged Results fruits per plant were collected to estimate natural seed set, which biased the counts towards higher than average seed Pre‑pollination barriers set, since unpollinated fowers were not considered. Never- theless, comparisons of seed set between treatments were Geographical isolation was low for three of the four popula- informative. tions (Table 1). The exception was Erica coccinea at Cape In order to test for reproductive interference, an additional Point, although this is simply because the pollen donor spe- treatment was applied to E. coccinea at the Cape Point site. cies, E. abietina atrorosea, has a small distribution com- Heterospecifc pollen was applied to fowers and then con- pletely overlapping that of E. coccinea. Phenological isola- specifc cross pollen was applied three hours later. tion was strong at one of the two sites where it was measured The ability to self pollinate was assessed in two ways. (Table 1). In the two E. coccinea populations, heterospecifc First, in the self pollination treatment, pollen was collected stigma-anther distances between co-occurring species were from fowers on the same inforescence or plant and applied signifcantly greater than conspecifc distances, suggest- directly onto the stigma. Second, to test for autonomous self ing relatively strong morphological foral isolation (Brodie pollination, fowers were left unmanipulated in exclusion Link population p < 0.001, V = 560, median heterospecifc bags. distance 3.3 mm, median conspecifc distance 1.8 mm, Cape Point population p < 0.001, V = 372, median heterospecifc

1 3 Journal of Plant Research distance 4.5 mm, median conspecific distance 2.8 mm, Reproductive interference Fig. S1). Weak foral isolation was found for E. plukenetii (p < 0.001, V = 43, median heterospecifc distance 2.2 mm, When heterospecifc pollen was added before outcross pol- median conspecific distance 3 mm) and E. curviflora len, seed set was 32% lower compared to outcross pollina- (p < 0.001, V = 43, median heterospecifc distance 2.7 mm, tion only (Fig. 2b), indicating that reproductive interference median conspecifc distance 3 mm) since their heterospecifc exists. The reproductive interference treatment produced stigma-anther distances were smaller than their conspecifc signifcantly greater seed set than the heterospecifc treat- distances (Fig. S1). ment (Table S2).

Post‑pollination barriers Self‑compatibility

All species produced signifcantly lower seed set from het- Seed set was low for all three species in the self erospecifc pollination compared to outcross pollination (mean = 0.08 ± 0.1 per fruit) and autonomous selfing (26–53% lower, Fig. 2), indicating that post-pollination bar- (mean = 0.02 ± 0.06 per fruit) treatments, and these were riers exist. Compared to natural pollination seed set, hetero- both signifcantly lower than the outcross pollination seed specifc pollination produced 2–50% lower seed set. Seed set set (mean = 0.5 ± 0.3 per fruit, Fig. 2, Table S1). Only E. from heterospecifc pollination was similar to that produced curvifora can be considered self-compatible (ISI > 0.3; by self pollination, except for E. coccinea at Brodie Link Table 1). The low seed set from autonomous selfng also NR, in which it was higher. Post-pollination reproductive suggests that wind pollination makes only a small contribu- isolation was relatively strong between all species pairs but tion to the total seed set, if any. weakest in E. curvifora (Table 1). Complete results of the multiple comparisons of seed set are provided in Table S2. Total fruit set per treatment (the

Fig. 2 Median seed set per fruit (black horizontal lines) resulting to cross pollination. Medians that share the same letter are not sig- from hand-pollination treatments: outcross (cross), heterospecifc nifcantly diferent (p > 0.05, Table S1). Flower images in each panel outcross (hetero), self pollination through an external vector (self), show the pair of maternal species (top, names indicated in the panel) autonomous self pollination (autogamy) and natural pollination (natu- and heterospecifc pollen donor species (bottom: a E. plukenetii, b E. ral). Reproductive interference was tested at Cape Point, where treat- abietina, c E. coccinea and d E. plukenetii) ment h + c involved the application of heterospecifc pollen 3 h prior

1 3 Journal of Plant Research percentage of treated fowers that set fruit across all plants required in this system to maintain species boundaries. from a population) supported the fndings from compari- At least one study has shown that strong post-pollination sons of seed set (Fig. S2), but we have not expanded on this barriers can enable coexistence of congeneric species in because of the low sample of fowers per treatment within the face of very weak pre-pollination barriers (Schemske plants. 1981).

Discussion Post‑pollination reproductive barriers

This study tested for reproductive barriers in co-fowering, This study reports on the frst formal heterospecifc polli- bird-pollinated Erica species, to explain their coexist- nation experiments conducted on the speciose Erica genus ence. Three of the four populations had only one rela- (c. 680 species) from the CFR. Since an estimated 9% of tively strong pre-pollination barrier. The low seed set from plant species (Mallet 2005) are involved in hybridisation at heterospecifc pollination treatments suggests that post- some point (Whitney et al. 2010), the production of a small pollination reproductive barriers are efective. Despite amount of hybrid seed is perhaps unsurprising. However, this, reproductive interference (through heterospecifc all species pairs showed strong reproductive isolation, pollen deposition prior to outcross pollen) reduced seed indicating that the post-pollination reproductive barriers set, indicating that there should be selection for traits that are well developed. This is interesting considering that enhance pre-pollination barriers. Such selection may be many hybrid Erica cultivars have been created by horti- strengthened by self-sterility, which was present in three culturists (Nelson and Oliver 2004), although the number of the four study populations. of successful crosses between bird-pollinated species is unknown. By contrast, three Satyrium species in the CFR showed greater interspecifc compatibility than the Erica Pre‑pollination reproductive barriers species in this study, with seed sets of 20–78% from heter- ospecifc pollination treatments (Ellis and Johnson 1999). Only one of the four populations had relatively strong Despite all species in this study showing some of degree pre-pollination barriers. This, taken together with previ- of reproductive isolation, it varied in extent. Previous studies ous work, indicates that this sunbird-Erica system may have shown that this variation may be explained by phyloge- comprise a relatively rare case of a radiation in which a netic distance (Larcombe et al. 2015) or phenotypic diver- diversity of congeneric species coexist despite weak pre- gence (Suni and Hopkins 2018), but here, the variation was pollination barriers. Bee-pollinated Erica species expe- not associated with phylogenetic distance, stigma-anther dis- rience similarly weak pre-pollination barriers (Bouman tance or corolla colour distance (Table S1, Fig S1), although et al. 2017), but fower constancy by honeybees appears more cases should be examined. Any of these factors could to be the mechanism that limits interspecifc pollen fow infuence the success of diferent combinations of pollen (van der Niet et al. 2020). Diferences in foral traits and/ donor and recipient species (Arceo-Gómez et al. 2019): for or the dominance of certain sexes or individuals may example, both E. coccinea and E. curvifora received pollen induce fower constancy by sunbirds, as has been seen in from E. plukenetii, but seed set was much more reduced in hummingbirds (Aldridge and Campbell 2007; Gowda and E. coccinea. We expected the greatest reproductive isolation Kress 2013), but this remains to be tested in a multispecies between the non-resprouting species, since natural selection context. Two study populations may avoid interspecifc should act faster in species with a more rapid turnover of pollen transfer via diferences in fower sizes among spe- generations (Ojeda et al. 2016). However, we found the low- cies in the same community, which should infuence where est reproductive isolation in E. curvifora and E. plukenetii, pollen is positioned on sunbirds’ bills. Experimental tests which are obligate reseeders, compared to the resprouting are still needed to test the extent to which this imposes E. coccinea. Nonetheless, all species pairs sufered ftness reproductive isolation, especially given that Erica species costs of heterospecifc pollination that may drive selection have slightly explosive pollen release (Geerts and Pauw for pre-pollination reproductive barriers such as divergence 2011). Phenological isolation was strong for one of the in fower colour and size (Table S1, Fig S1), which appear to study species pairs. Flowering phenology may be the most be labile traits in the Cape Erica (Pirie et al. 2011). common pre-pollination barrier in this guild, considering the low bird pollinator diversity across the CFR (Rebelo Reproductive interference 1987) and the rather seasonally unconstrained fowering of Cape Erica (Johnson 1993). Overall, these ecological Even though post-pollination barriers were strong, het- factors indicate that strong post-pollination barriers are erospecifc pollen deposition prior to outcross pollen did

1 3 Journal of Plant Research reduce seed set signifcantly in the one species we tested. well-developed post-pollination isolation barriers can This shows that this type of reproductive interference results facilitate their coexistence. However, reproductive inter- in ftness losses, and so may consequently drive selection ference could still cause signifcant female ftness costs for pre-pollination barriers. Selection against interspecifc in such systems, which may drive selection for pre-pol- pollen transfer may also be further driven by the male com- lination barriers. An investigation of the historical and ponent of ftness loss, through pollen lost to heterospecifc contemporary gene fow between coexisting species would stigmas (Morales and Traveset 2008). Some fower traits, confrm the reproductive barriers found in this study. The such as style length, may determine the degree of repro- fndings indicate that we should expect pollinators to have ductive interference, since long-styled species may receive a strong infuence on plant coexistence and fower evolu- more heterospecifc pollen (Mazer et al. 2016; Tong and tion in this, and similar, systems. Huang 2016) perhaps due to increased contact with pollina- tors’ bodies, although they could also prevent heterospecifc Acknowledgements We thank Stuart Hall, Kylie Pires, Geraldine Jacobs, Caroli Diener, Genevieve Theron, Mvana Sibiya and Gina pollen tube growth more efectively (Tong and Huang 2016). Arena for assistance with the pollination experiments, Sally Reece for In this guild of Erica species, pistil lengths vary at least allowing us to work at Lourensford Estate, and two anonymous refer- between 15–31 mm (n = 10 species, Anina Coetzee unpub- ees for their comments which greatly improved the manuscript. This lished data, Turner et al. 2012). This variation may prevent study was conducted under South African National Parks (CRC/2016- 2017/020--2016/V1) and CapeNature permits (AAA043-00008-0056). reproductive interference and interspecifc mating. AC was funded by the Department of Science and Innovation/National Research Foundation Centre of Excellent at the FitzPatrick Institute of Self‑compatibility African Ornithology, the South African National Biodiversity Insti- tute, the Botanical Education Trust and the Claude Leon Foundation. CNS was partly funded by a BBSRC David Phillips Fellowship (BB/ If there is selection for pre-pollination barriers in this sys- J014109/1). tem, it may be strengthened by pollinator dependence. Two Author contributions of the three species had low self-compatibility, suggesting All authors designed the study, AC collected and analysed data, AC led the writing, to which all authors contributed that their reproduction is highly pollinator dependent. This critically. concurs with the general pattern found for this guild of Erica species (Angoh et al. 2017; Arendse 2015), and contrasts Compliance with ethical standards with the high prevalence of self-compatibility the compara- ble system of a large diversity of coexisting Bromeliaceae Conflict of interest The authors declare that they have no confict of species that share hummingbird pollinators (Matallana et al. interest. 2010). Our results confrm that E. plukenetii is self-ster- ile, although the index values of self-incompatibility and autonomous selfng were lower than the previously found References values of 0.13 and 0.27, respectively (Arendse 2015), hinting at a possibility of interpopulation variation in self-sterility. Aldridge G, Campbell DR (2007) Variation in pollinator preference Ipomopsis There was no linear association between the degree of self- between two contact sites that difer in hybridization rate. Evolution (N Y) 61:99–110 compatibility and herkogamy (distance between the stigma Angoh SYJ, Brown M, Midgley J (2017) Selfng versus out-crossing in and anther within fowers, Table S1) of the study species. six Erica species of the southern Cape. S Afr J Bot 111:387–391 Interestingly, E. curvifora had the highest self-compatibility Arceo-Gómez G, Raguso RA, Geber MA (2016) Can plants evolve and interspecifc compatibility, perhaps supporting studies tolerance mechanisms to heterospecifc pollen efects? An experi- mental test of the adaptive potential in Clarkia species. Oikos that have proposed a link between these two types of com- 125:718–725 patibilities (Rieseberg and Willis 2007). Overall, a conclu- Arceo-Gómez G, Kaczorowski RL, Patel C, Ashman TL (2019) Inter- sion from our fndings is that any selection on pollination active efects between donor and recipient species mediate ftness enhancement traits should be strengthened by their high costs of heterospecifc pollen receipt in a co-fowering community. Oecologia 189:1041–1047 dependence on pollinators. Arendse B (2015) Variation in breeding systems and consequences for reproductive traits in Erica. Dissertation, University of Cape Town. https​://hdl.handl​e.net/11427​/15494​. Accessed 2015 Bates D, Maechler M, Bolker B, Walker S (2015) Fitting Linear mixed- Conclusion efects models using lme4. J Stat Softw 67:1–48 Bouman RW, Steenhuisen SL, Van Der Niet T (2017) The role of the Erica This study suggests that segregation of fowering phe- pollination niche in community assembly of species in a biodiversity hotspot. J Plant Ecol 10:634–648 nology and fower reproductive parts potentially serve Brown BJ, Mitchell RJ (2001) Competition for pollination: Efects as pre-pollination barriers in closely-related, pollinator- of pollen of an invasive plant on seed set of a native congener. sharing plant species. Where these barriers are weak, Oecologia 129:43–49

1 3 Journal of Plant Research

Da Fonseca LCN, Rech AR, Bergamo PJ et al (2016) Heterospecifc Nishida S, Kanaoka MM, Hashimoto K et al (2014) Pollen-pistil inter- pollen deposition among plants sharing hummingbird pollinators actions in reproductive interference: comparisons of heterospecifc in the Brazilian Atlantic Forest. Rodriguesia 67:335–345 pollen tube growth from alien species between two native Taraxa- Ellis AG, Johnson SD (1999) Do pollinators determine hybridization cum species. Funct Ecol 28:450–457 patterns in sympatric Satyrium (Orchidaceae) species? Plant Syst Nosil P, Crespi BJ, Gries R, Gries G (2007) Natural selection and Evol 219:137–150 divergence in mate preference during speciation. Genetica Fang Q, Huang SQ (2013) A directed network analysis of heterospecifc 129:309–327 pollen transfer in a biodiverse community. Ecology 94:1176–1185 Ojeda F, van der Niet T, Malan MC et al (2016) Strong signature of Geerts S, Pauw A (2011) Easy technique for assessing pollination rates selection in seeder populations but not in resprouters of the fynbos in the genus Erica reveals road impact on bird pollination in the heath Erica coccinea (Ericaceae). Bot J Linn Soc 181:115–126 Cape fynbos, South Africa. Austral Ecol 36:656–662 Oliver EGH, Oliver IM (2002) The genus Erica (Ericaceae) in southern Gibbs PE (2014) Late-acting self-incompatibility-the pariah breeding Africa: taxonomic notes 1. Bothalia 32:37-61 system in fowering plants. New Phytol 203:717–734 Oliver EGH, Forshaw N (2012) Genus Erica: an identifcation aid soft- Gowda V, Kress WJ (2013) A geographic mosaic of plant-pollina- ware version 3.00. Published by Bolus Herbarium, University of tor interactions in the Eastern Caribbean Islands. Biotropica Cape Town, Cape Town, South Africa 45:224–235 Oliver EGH, Linder HP, Rourke JP (1983) Geographical distribution of Grant V (1994) Modes and origins of mechanical and ethological isola- present-day Cape taxa and their phytogeographical signifcance. tion in angiosperms. Proc Natl Acad Sci USA 91:3–10 Bothalia 14:427–440 Heystek A, Pauw A (2014) Does competition for pollinators contribute Pirie MD, Oliver EGH, Bellstedt DU (2011) A densely sampled ITS to structuring Erica communities? J Veg Sci 25:648–656 phylogeny of the Cape fagship genus Erica L. suggests numer- Heystek A, Geerts S, Barnard P, Pauw A (2014) Pink fower preference ous shifts in foral macro-morphology. Mol Phylogenet Evol in sunbirds does not translate into plant ftness diferences in a 61:593–601 polymorphic Erica species. Evol Ecol 28:457–470 Rebelo AG (1987) Bird pollination in the Cape fora. In: Rebelo AG Hopkins R, Rausher MD (2012) Pollinator-mediated selection on (ed) A preliminary synthesis of pollination biology in the Cape fower color allele drives reinforcement. Science 335:1090–1092 fora. CSIR, Pretoria, pp 83–108 Hothorn T, Bretz F, Wesfall P (2008) Simultaneous inference in general Rebelo AG, Siegfried WR, Crowe AA (1984) Avian pollinators and parametric models. Biometrical J 50:346–363 the pollination syndromes of selected mountain fynbos plants. S Johnson SD (1993) Climatic and phylogenetic determinants of fower- Afr J Bot 3:285–296 ing seasonality in the Cape fora. J Ecol 81:567 Rebelo AG, Siegfried WR, Oliver EGH (1985) Pollination syndromes Johnson SD (2018) Natural hybridization in the orchid fora of South of Erica species in the southwestern Cape. S Afr J Bot 51:270–280 Africa: Comparisons among genera and foristic regions. S Afr J Rieseberg LH, Willis JH (2007) Plant speciation. Science 317:910–914 Bot 118:290–298 Schemske DW (1981) Floral convergence and pollinator sharing in two Larcombe MJ, Holland B, Steane DA et al (2015) Patterns of reproduc- bee-pollinated tropical herbs. Ecology 62:946–954 tive isolation in Eucalyptus. A phylogenetic perspective. Mol Biol Sobel JM, Chen GF (2014) Unifcation of methods for estimating the Evol 32:1833–1846 strength of reproductive isolation. Evolution (N Y) 68:1511–1522 Linder HP (2003) The radiation of the Cape fora, Southern Africa. Suni SS, Hopkins R (2018) The relationship between postmating repro- Biol Rev 78:597–638 ductive isolation and reinforcement in Phlox. Evolution (N Y) Lloyd DG, Schoen DJ (1992) Self- and cross-fertilization in plants. l. 72:1387–1398 Funct Dimens Int J Plant Sci 153:358–369 Tong ZY, Huang SQ (2016) Pre- and post-pollination interaction Lowry DB, Modliszewski JL, Wright KM et al (2008) The strength and between six co-fowering Pedicularis species via heterospecifc genetic basis of reproductive isolating barriers in fowering plants. pollen transfer. New Phytol 211:1452–1461 Philos Trans R Soc B Biol Sci 363:3009–3021 Turner RC, Midgley JJ, Barnard P et al (2012) Experimental evidence Malan MC (2013) Intraspecifc variation in Erica coccinea. Disserta- for bird pollination and corolla damage by ants in the short-tubed tion, University of Cape Town. https://hdl.handl​ e.net/11427​ /6672​ . fowers of Erica halicacaba (Ericaceae). S Afr J Bot 79:25–31 Accessed 2015 Van der Niet T, Pirie MD, Shuttleworth A et al (2014) Do pollinator Mallet J (2005) Hybridization as an invasion of the genome. Trends distributions underlie the evolution of pollination ecotypes in the Ecol Evol 20:229–237 Cape shrub Erica plukenetii? Ann Bot 113:301–316 Matallana G, Godinho MAS, Guilherme FAG et al (2010) Breeding van der Niet T, Pires K, Steenhuisen SL (2020) Flower constancy of systems of Bromeliaceae species: evolution of selfng in the con- the Cape honey bee pollinator of two co-fowering Erica spe- text of sympatric occurrence. Plant Syst Evol 289:57–65 cies from the Cape Floristic Region (South Africa). S Afr J Bot Mazer SJ, Moghaddasi A, Bello AK, Hove AA (2016) Winning in 132:371–377 style: Longer styles receive more pollen, but style length does Waser NM (1986) Flower constancy: defnition, cause and measure- not afect pollen attrition in wild Clarkia populations. Am J Bot ment. Am Nat 127:593–603 103:408–422 Weber MG, Strauss SY (2016) Coexistence in close relatives: beyond Mitchell N, Holsinger KE (2018) Cryptic natural hybridization between competition and reproductive isolation in sister taxa. Annu Rev two species of Protea. S Afr J Bot 118:306–314 Ecol Evol Syst 47:359–381 Morales CL, Traveset A (2008) Interspecifc pollen transfer: Magni- Wesselingh RA (2007) Pollen limitation meets resource allocation: tude, prevalence and consequences for plant ftness. CRC Crit Rev towards a comprehensive methodology. New Phytol 174:26–37 Plant Sci 27:221–238 Whitney KD, Ahern JR, Campbell LG et al (2010) Patterns of hybridi- Muchhala N (2008) Functional signifcance of interspecifc variation zation in plants. Perspect Plant Ecol Evol Syst 12:175–182 in Burmeistera fower morphology: evidence from nectar bat cap- tures in Ecuador. Biotropica 40:332–337 Publisher’s Note Springer Nature remains neutral with regard to Nelson EC, Oliver EGH (2004) Cape heaths in European gardens: the jurisdictional claims in published maps and institutional afliations. early history of South African Erica species in cultivation, their deliberate hybridization and the orthographic bedlam. Bothalia 34:127–140

1 3