Introduction

Evolutionary shifts among pollinator functional groups are common in plant lineages and can have major implications for the evolution of floral phenotypes (van der Niet and Johnson 2012). Identifying these shifts and their implications for trait evolution requires a combination of natural history observations, phylogenetics and studies of floral morphology, advertising and reward traits. Shifts involving moths and butterflies are of interest because these two insect groups are closely related and similar morphologically, yet have differences in sensory modalities and diel activity patterns that have major implications for the evolution of floral advertising traits. As a generalization, moth-pollinated flowers tend to be pale and scented at night, while butterfly-pollinated flowers tend to be brightly coloured and are not usually strongly scented (Faegri and van der Pijl 1979; Goldblatt and Manning 2002; Liu et al. 2022). However, moths and butterflies are both highly diverse groups, and within each group, there is considerable variation in traits such as proboscis length, visual systems and behaviour; for example, whether species settle or flutter while feeding (Mertens et al. 2021). Plants also exploit different body parts of these insects for pollen transport (Hapeman and Inoue 1997; Butler and Johnson 2020). As a consequence, there is marked variation in the floral traits of plants pollinated by Lepidoptera (Mertens et al. 2021).

Habenaria is one of the largest orchid genera worldwide and is known to be paraphyletic with several currently recognized genera being embedded within the broader Habenaria clade (Ngugi et al. 2020). One example is Platycoryne, an African genus of about 20 species, that has been shown to be phylogenetically nested within Habenaria (Ngugi et al. 2020). Earlier morphological studies also supported the inclusion of Platycoryne in Habenaria (Kurzweil and Weber 1992). The majority of Habenaria species have white or green flowers, and many of these species have been shown to be moth-pollinated. It is notable that the flowers of most Platycoryne species are orange. This trait is very rare in Habenaria and has been linked positively to butterfly pollination in the well-studied Asian species Habenaria rhodocheila (Chen et al. 2021; Zhang et al. 2021). Similarly in the orchid genus Platanthera which is not part of the Habenaria clade, moth-pollinated species tend to have white flowers (Nilsson 1978), whereas butterfly-pollinated species tend to have orange flowers (Robertson and Wyatt 1990; Hapeman and Inoue 1997). Orange flower colour is also strongly linked with butterfly pollination in other plant families (Goldblatt and Manning 2002; Kiepiel and Johnson 2014; Daniels et al. 2020; Liu et al. 2022). However, orange flowers also occur in some bird-pollinated orchids (Johnson and Van der Niet 2019) and butterflies have been shown to pollinate white-flowered species in the Habenaria clade (Pedron et al. 2012; Balducci et al. 2019b; Tan et al. 2023). It is therefore essential to conduct field-based pollinator observations and not to rely solely on floral syndromes for reaching conclusions about shifts in pollination systems (van der Niet 2021).

I investigated the pollination biology of the orange-flowered species Platycoryne mediocris in south-central Africa and also searched for anecdotal records of insect visitors to flowers of other Platycoryne species to test my hypothesis that the evolution of orange flowers in this clade of Habenaria is linked with pollination by butterflies. The aims of this study were 1) identify pollinators of P. mediocris, 2) to document floral morphology, nectar rewards and advertising traits such as colour and scent, and 3) to determine levels of pollination success and fruit set.

Methods

Study species

Platycoryne mediocris Summerh. has a wide distribution in south-central Africa that encompasses Zambia, Zimbabwe, northern South Africa, Malawi, the Democratic Republic of Congo, Burundi and Tanzania. All parts of the flower including the ovary are bright orange (Fig. 1). Plants are 15–50 cm tall, and in the population that was studied, plants produced a mean (± SE) of 4.11 ± 0.10 flowers (n = 26).

Fig. 1
figure 1

Morphology of flowers of Platycoryne mediocris

Study site

Observations and trait measurements were carried out at Mutinondo Wilderness (12°27′18"S, 31°17′28"E) in central Zambia in 2022 and 2023. The study population of P. mediocris consisted of about 200 plants growing on a large granite inselbergs. A general account of the flora at the study site is given by Vollesen and Merrett (2020). Voucher specimens (Bytebier and Wightman 3819) are deposited in the Bews Herbarium (NU), the Kitwe Forest Herbarium (NDO) and Meise Botanic Garden Herbarium (BR).

Pollinator observations

Pollinator observations were carried out from 09h00 to 16h00 on 20–23 February 2022 and 7–11 February 2023. Plants were also observed intermittently in the evenings from 18h00 to 23h00, while I was conducting a separate study of moth visitors to flowers of Chamaepentas nobilis (Rubiaceae) which grows intermingled with P. mediocris at the study site. I also used four passive infrared (PIR) motion-activated cameras (Bushnell Nature View 119740) to make indirect observations of flower visitors over the same time period as the direct observations. Insects observed on the flowers were photographed for later identification, and a sample of each species was captured for morphological measurements (proboscis length and distance between inner and outer surfaces of the eyes) and confirmation of the identity of pollinaria they were carrying.

Floral traits

To assess traits that could influence the morphological fit between flowers and pollinators, I measured floral spur length, the length and width of the sepals and petals, the length and spacing of the rostellum arms, the stigma size, and the dimensions of pollinaria for one flower on each of 15 plants. I also measured the volume and concentration of nectar in these flowers using 5-µL pipettes and a 0–50% hand-held refractometer. Spectral reflectance of flowers was measured using an Ocean Optics S2000 spectrometer following the methods described by Johnson and Anderson (2002). Floral scent was analysed using headspace collection and gas chromatography–mass spectrometry (GC–MS) using the apparatus and methods described by Johnson et al (2020). I took four headspace samples from 12 plants (three plants per sample) by enclosing them in polyacetate bags (Kalle, Germany) and sucking air from the bags through an absorbent filter for 60 min at 200 ml per min. Compounds present in similar amounts in a control ambient sample were excluded from analysis. Emission rate of volatiles emitted was calculated from injection of a known amount of methyl benzoate as described by Johnson et al (2020).

Pollination and fruiting success

To assess the level of pollination success in the population, I examined 46 flowers on 24 plants and recorded the number of pollinaria removed and the number of pollen massulae deposited on stigmas. I counted the number of pollen massulae in pollinaria from eight flowers and used these data to calculate pollen transfer efficiency (the proportion of pollen removed from flowers that is deposited on stigmas). When examining flowers, I noted whether pollinia of wilted unvisited flowers remained enclosed within the anther as this would prevent autonomous self-pollination. I also calculated the percentage of flowers that set fruit for 26 plants sampled four weeks after the end of flowering.

Results

Pollinator observations

Flowers of P. mediocris were visited exclusively by medium-sized nymphalid and pierid butterflies (Fig. 2, Table 1, Video S1). I directly observed 28 butterflies feeding on flowers in the population, and a further two butterflies were recorded using camera traps (Video S1). In all cases, the butterflies probed the spur while hanging from the flower, using their feet to grip the ridged ovary or the reflexed lateral sepals. Six butterfly individuals, representing three species, were confirmed to carry pollinaria (Table 1). Pollinaria were mostly attached to the eyes of butterflies, but some were attached either to the palps or between the palps (Fig. 2). In one case, a pollinarium was attached to the legs of a butterfly. The pollinaria on butterflies were confirmed as those of P. mediocris on account of their length and distinctive sickle shape with the massulae segments lined up on one side of the caudicle. The pollinaria attached to butterflies curl forwards over the palps, such that the massulae segments are pressed against the stigma when the tongue is fully inserted. The mean (± SE) proboscis length of butterflies captured on the orchid was 13.1 ± 0.35 mm.

Fig. 2
figure 2

Butterfly visitors to flowers of Platycoryne mediocris (a–e) and Platycoryne sp. (f). a Male Acraea natalica probing a flower. b Acraea caldarena carrying pollinaria. c Female Acraea natalica carrying pollinaria. d Catopsilia florella probing a flower. e Mylothris rueppellii probing a flower. f Eurema hecabe carrying pollinaria on its eyes. Scale bars = 5 mm, except b = 2 mm. Arrows indicate pollinaria attached to the heads of insects. Photographs: Steve Johnson (a–e) and Craig Peter (f)

Table 1 Butterfly individuals observed to visit flowers of Platycoryne mediocris, the number of those confirmed to carry pollinaria of the orchid, and mean (± SE) proboscis lengths. The mean spur length in the orchid population was 15.7 ± 0.25 mm

Floral traits

The mean (± SE) length of the floral spur of P. mediocris was 15.7 ± 0.25 mm which is slightly longer than the mean proboscis length of the butterfly flower visitors (t = 6.15, P < 0.001). The length of the rostellum arms was 1.41 ± 0.06 mm, and the distance between the rostellum arms was 1.8 ± 0.10 mm which is less than the distance between the mean (± SE) outer margins of the butterfly eyes (3.6 ± 0.21 mm), but greater than the distance between the inner margins of their eyes (1.46 ± 0.08 mm). Pollinaria were c. 3.7 ± 0.19 mm in length, with the caudicle measuring 1.7 ± 0.13 mm in length and the pollinium 2.0 ± 0.11 mm in length. Other floral trait dimensions are provided in Table S1.

The volume of the standing crop of nectar in flowers was 1.1 ± 0.19 µL with a sugar concentration of 22.8 ± 1.20%. The mean height of the standing crop nectar column from the tip of the spur was 4.4 ± 0.65 mm which means that it was accessible to butterflies with proboscides of 11.3 mm or longer.

Flowers of P. mediocris absorb ultraviolet light and show an increase in reflectance around 550 nm resulting in the orange appearance (Fig. 3).

Fig. 3
figure 3

Spectral reflectance of flowers of Platycoryne mediocris. Mean spectra are bold lines, and individual spectra are light grey lines

Flowers have no discernible scent to humans. Analysis revealed only trace emissions of three compounds (mean percentage of total peak area ± SE): benzaldehyde (68.2 ± 6.8%), linalool (12.0 ± 2.2%), and phenylethyl alcohol (19.8 ± 8.4%). The mean (± SE) emission rate of volatiles was 0.55 ± 0.1 ng/flower/hour.

Pollination and fruiting success

Of the 46 flowers surveyed, 47.8% were pollinated with an average of 12.2 ± 3.1 massulae (range: 0–70) deposited per stigma, 69.6% had at least one pollinarium removed, and 63% of all pollinaria were removed. Each pollinarium contained 178.6 ± 5.4 massulae. Pollen transfer efficiency in the population was calculated as 4.9%. At the end of the flowering season, 88.1 ± 3.9% of the flowers on 26 plants set fruit.

Discussion

The results of this first detailed study of pollination in Platycoryne are consistent with the hypothesis that orange-flowered Platycoryne species are pollinated by butterflies. In addition to the observations of butterflies visiting P. mediocris that are reported here, I have seen Acraea butterflies visiting the orange flowers of P. buchananii in northern Zambia, and they have also been photographed on the orange flowers of P. guingangae in Gabon (Ramette 2023) and P. pervillei in Zimbabwe (Fibeck and Phiri 1998). Acraea natalica, the most common visitor of the flowers of P. mediocris (Table 1), was also the only recorded species to visit flowers and carry pollinaria of P. pervillei in Zimbabwe (Fibeck and Phiri 1998). A pierid butterfly Eurema hecabe has also been observed by C. Peter (pers. comm.) on the orange flowers of an unidentified Platycoryne species in Angola and also carried pollinaria on its eyes (Fig. 2F). The spur length of this orchid was c. 13 mm, closely matching the mean (± SE) proboscis length of three captured E. hecabe butterflies which was 12.1 ± 0.17 mm (C. Peter, pers. comm.).

The evolution of orange butterfly-pollinated flowers in Platycoryne in Africa parallels that of the independent evolution of orange flowers in the lineage leading to the Asian species Habenaria rhodocheila which is also pollinated by butterflies (Chen et al. 2021; Zhang et al. 2021). The importance of colour for attraction of butterflies has been demonstrated in numerous experiments (Chen et al. 2021; Kiepiel and Johnson 2021). There is a strong overall association between orange-red flowers and butterfly pollination in Africa (Newman et al. 2012; Kiepiel and Johnson 2014; Butler and Johnson 2020; Mertens et al. 2021). The butterfly Catopsilia florella which was one of the species observed to carry pollinaria of P. mediocris was also observed to visit the orange flowers of Gloriosa superba at the study site, and butterflies are known to visit flowers of G. superba across its extensive range in Africa (Daniels et al. 2020).

A notable result was the almost complete lack of scent emission by flowers of P. mediocris. The emission rate of 0.55 ng per flower per hour is at least 100-fold less than values reported for moth-pollinated species in the Habenaria clade—321–2551 ng/fl/hr for H. clavata (Johnson et al. 2020), 120 ng/fl/hr for Bonatea polypodantha (Balducci et al. 2020) and 152.5 ng/fl/hr for Bonatea steudneri (Balducci et al. 2019a). Other authors have reported only emission rates per inflorescence for moth-pollinated Habenaria species, but when divided by the reported average number of flowers, these give similarly high rates of emission—c. 4566 ng/fl/hr for Habenaria limprichtii (Tao et al. 2018) and c. 2083 ng/fl/hr for Habenaria epipactidea (Peter et al. 2009). Conversely, a low rate of 28.8 ng/fl/hr was reported for Bonatea cassidea, which is a butterfly-pollinated member of the Habenaria clade (Balducci et al. 2019b).

Pollinaria of P. mediocris were attached the eyes or palps of butterflies (Fig. 1B, C). This is also a common mode of pollinaria placement in the genus Habenaria and is associated with short diverging rostellum arms that flank the entrance to the spur (Singer and Cocucci 1997; Pedron et al. 2012; Zhang and Gao 2017; Tao et al. 2018). The pollinaria of P. mediocris are sickle-shaped and on withdrawal, project forward and slightly downwards such that the massula segments are pushed against the stigma. The high percentage of flowers of P. mediocris that were pollinated and set fruit is not unusual for orchids that produce nectar rewards (Tremblay et al. 2004). The values are also consistent with pollination rates of “50–80%” reported for a population of P. pervillei in Zimbabwe (Fibeck and Phiri 1998). The pollen transfer efficiency of 4.9% that was recorded in the study population is lower than the average of c. 10% for orchids with massulate (sectile) pollinia (Johnson and Harder 2023). This may reflect that many of the pollinaria attached to butterflies were out of alignment (Fig. 2B) which is likely due to physical interference when large number of pollinaria become attached to the same insect (Harder et al. 2021).

Although it is clearly apparent that a shift took place from moth to butterfly pollination in the Habenaria clade that includes Platycoryne, more detailed phylogenetic studies and pollinator observations are required to infer whether this transition was directly from moth to butterfly pollination or involved an intermediate stage with a different pollination system. There are some Platycoryne species (e.g. P. macroceras, P. isoetifolia) with pale white or greenish flowers which suggest that they may be moth-pollinated, and a more detailed phylogeny is therefore needed to reconstruct the pollination system of the original ancestor of the Platycoryne clade.