In Orchid Pollinator Deception, Size May Not Matter But Shape Certainly Does

Pollinator-mediated selection on floral morphology is predicted to be driven by two non-exclusive mechanisms: selection exerted via the sensory preferences of pollinators on attractive traits and selection driven by the mechanical fit between pollinators and flowers. Several studies have revealed patterns consistent with selection for mechanical fit to primary pollinators (Steiner & Whitehead 1990; Muchhala 2007; Anderson & Johnson 2008). For example, floral tube length of plants pollinated by long tongue flies typically matches the tongue lengths of their local pollinators (Pauw, Stofberg & Waterman 2009). Pollinator preference may also influence floral morphology with contrasting pollinator preferences potentially driving divergence in floral colour (Bradshaw & Schemske 2003; De Jager & Ellis 2014a) and odour (Shuttleworth & Johnson 2010). Strong adaptive responses to pollinator preference are illustrated by the matching flower colours between the rewardless food-deceptive orchids and the rewarding plants they mimic (Peter & Johnson 2008; Newman, Anderson & Johnson 2012).

Pollinator sensory preferences are also predicted to be a key driver in sexually deceptive orchids. These plants attract male insects as pollinators by mimicking female sex pheromones (Schiestl et al. 1999, 2003; Ayasse, Stökl & Francke 2011) and achieve pollination when mate-seeking males attempt copulation with their flowers. Variation in floral odour has been linked to speciation in sexually deceptive orchids (Peakall et al. 2010; Xu et al. 2011; Peakall & Whitehead 2014), indicating a key role for chemistry in their evolution. Beyond sex pheromone mimicry, these orchids exhibit unusual, often morphologically complex insectiform labella that have long been interpreted as morphological mimicry of the females of the pollinator species (Kullenberg 1961; Paulus & Gack 1990; Peakall 1990). Alternatively, complex floral morphology in sexually deceptive orchids may be under selection via mechanical fit with male pollinator morphology to enhance copulation behaviour with the labellum and/or contact with the floral reproductive structures.

To date, no studies of sexually deceptive orchids have comprehensively investigated the role of flower morphology in either female mimicry or mechanical fit. Some support for female mimicry is found in Cryptostylis, where the labella exhibit similar reflectance spectra as the abdomen of its pollinator’s female (Gaskett & Herberstein 2010). Floral colour in Ophrys has also been found to enhance pollinator detection over short distances (Streinzer, Paulus & Spaethe 2009). Ophrys may also provide some support for mechanical fit as a significant correlation between labellum length and pollinator length has been reported (Paulus 2006). However, due to a general size overlap between male and female sexes in pollinators of Ophrys (Amiet et al. 2010), it may not be possible to disentangle mimicry of female form from mechanical fit to male pollinators within this system, as no clear predictions can be made for each hypothesis. Much needed experimental manipulations are also lacking, and as a consequence, the function of these intriguing and morphologically complex labellum structures remains poorly understood.

The parasitoid thynnine wasps that pollinate the vast majority of Australian sexually deceptive orchids exhibit unique mating behaviour and unusual sexual dimorphism. The smaller wingless females spend most of their adult lives underground, while the larger winged males search above-ground for emerging females (Ridsdill Smith 1970). This sexual dimorphism offers a unique opportunity to tease apart the role of female mimicry from mechanical fit. Here, we conduct an explicit assessment of the role of floral morphology in the thynnine-pollinated sexually deceptive orchid genus Chiloglottis. The highly specific orchid–pollinator interactions (average of 1·1 pollinators/orchid species) in the genus involve 1–3 floral volatile components from a pool of six related 2,5-dialkylcyclohexane-1,3-diones or ‘chiloglottones’ (Schiestl et al. 2003; Franke et al. 2009; Peakall et al. 2010).

Pollinators can be readily attracted to and may even attempt copulation with plastic beads (as simplistic female dummies), when spiked with the appropriate synthetic chiloglottone(s) (Schiestl et al. 2003; Peakall et al. 2010). Chiloglottis species, however, exhibit consistent differences in the size and shape of their labella, including the placement of insectiform calli structures (Bower 1996; Peakall et al. 2002; Peakall & Whitehead 2014). Pollinators have also been reported to exhibit higher tendencies to attempt copulation with flowers than with spiked beads (Whitehead & Peakall 2014). Thus, despite the critical importance of semiochemicals, morphology may also be important for pollination.

Our study takes advantage of two strategically targeted species, Chiloglottis trapeziformis Fitzg. and C. valida D.L. Jones. These orchids are drawn from distinct clades (Peakall et al. 2010) and represent the two divergent floral morphologies that characterize the genus (Peakall et al. 1997). Yet, both orchids employ chiloglottone 1 for pollinator attraction (Schiestl & Peakall 2005). Each species has a distinct primary pollinator, male Neozeleboria cryptoides Smith and male N. monticola Turner wasps, respectively (Peakall et al. 2010). While essentially allopatric, both orchids may share each other’s pollinator when distributions overlap and rare F1 hybrids are known (Peakall et al. 1997, 2002).

Each wasp species represents a distinct clade within Neozeleboria, and they exhibit some behavioural differences that likely reflect their phylogenetic divergence (Schiestl & Peakall 2005; Griffiths et al. 2010). For example, orientations during attempted copulation at the two species are consistently different (Figs 1 and S1, Video S1, Supporting Information). Thus, working with this pair of orchids, and their respective pollinators, provides a powerful opportunity to evaluate the role and importance of morphology while minimizing the confounding effects of chemistry.

Our study asks two main questions. First, we ask, does morphology matter in sexual deception? We address this question through a series of experiments where we compare attempted copulation frequency between two morphologically distinct Chiloglottis species and between morphologically manipulated conspecific flowers. Secondly, we ask, is there morphological evidence for mechanical fit with male pollinators in labellum or floral reproductive structures and is there evidence for female mimicry? We answer this by comparing orchid morphological traits of size and colour with male and female wasp morphology. We then evaluate our findings against a priori predictions that take into account the distinct orientations of attempted copulation at the flowers, allowing us to predict separate morphological responses for each orchid species under both the mechanical fit and mimicry hypotheses.

Notwithstanding the well-established primary role of semiochemicals for securing long-range pollinator attraction in sexually deceptive orchids (Schiestl et al. 1999, 2003; Ayasse, Stökl & Francke 2011; Bohman et al. 2014), we present here compelling new experimental evidence that morphology also matters. Both C. trapeziformis and C. valida elicited more and longer copulation attempts from their primary pollinator despite using the same chiloglottone for pollinator attraction. This finding suggests that the two orchids are likely morphologically adapted to their primary pollinators. Although previous studies have also reported higher rates of attempted copulation for primary pollinators in congeneric sexually deceptive orchids (Peakall et al. 2010; Whitehead & Peakall 2014), these studies always focussed on taxa with divergent odour chemistry.

While it is potentially possible that C. trapeziformis and C. valida differ in their rate of chiloglottone 1 emission, chiloglottones are emitted in such low amounts that unfortunately neither standard headspace nor SPME (solid-phase microextraction) techniques can detect them (Falara et al. 2013; Amarasinghe et al. 2015). However, even if systematic variation in chiloglottone emission rates did exist between the two orchids, it cannot explain the consistent pattern of behavioural responses we uncovered. These include reciprocally stronger responses to the orchid that each wasp pollinates, and differences in landing place, orientation and probing point. The relative percentages of wasps that landed on the two species were also broadly overlapping, suggesting that these fundamental differences in sexual responses to C. trapeziformis and C. valida are unlikely to be chemical.

This is further supported by results from our labellum manipulation experiments, which avoided any systematic emission rate differences by manipulating non-chiloglottone-producing tissues within the same species. In these experiments, we thus retained common chemistry while specifically manipulating labellum size. Consistent with our a priori expectation that C. trapeziformis labellum size overlaps with pollinator size (and the correlated Callus-Tip overlaps with MaleAbdomenL; see Table 2), manipulations in this orchid yielded a significant reduction in the duration of attempted copulation (Fig. 4b). Conversely, there was no significant change in C. valida (Fig. 4b), consistent with the a priori expectation that labellum size is less important given the reverse orientation of the pollinator.

Seed set is strongly pollinator-limited in both C. valida (Peakall & Whitehead 2014) and C. trapeziformis (Schiestl 2004). Under this norm of pollinator limitation, Chiloglottis flowers with morphological traits that elicit more frequent and prolonged copulation attempts are predicted to exhibit higher reproductive success in the field, as found in other sexually deceptive species (Gaskett, Winnick & Herberstein 2008; Ellis & Johnson 2010). Conducting field experiments to directly test this prediction in Chiloglottis, while not impossible, would be difficult as extensive replications across multiple populations would be required. With these constraints in mind, we suggest that the frequency and duration of attempted copulation measured in this study likely provides a reasonable proxy for plant reproductive fitness.

Furthermore, in the case of diamond-shaped Chiloglottis, our experimental morphological manipulations suggest that flowers with maladapted morphology will be at a significant fitness disadvantage under pollinator limitation, indicating the likelihood of pollinator-mediated selection in the wild for morphological traits such as labellum size. Thus, we conclude that morphology does indeed matter in sexual deception.

Our strategic focal orchid species were chosen both because they represented contrasting floral morphologies and because they were broadly representative of the two floral shapes across the genus. C. valida is larger and showed little overlap in size with C. trapeziformis in any traits (Fig. 2). Similarly, there was no overlap in multivariate morphometric space between the two orchids (Fig. S2b). By contrast, the differences in size between N. monticola and N. cryptoides were smaller (Fig. 2), with considerable overlap in multivariate space (Fig. S3b), although trait means differed significantly for most measured traits (Table S2).

From the outset of the study, we knew that pollinator orientation differed between the two focal species and that these orientation differences hold true across the genus for the two flower morphologies (Fig. S1). In the light of these behavioural differences, we predicted there would be two labellum traits for mechanical fit to copulating males in C. trapeziformis, length (LabellumL) and callus to tip length (Callus-Tip), which would not be relevant in C. valida (Table 2). As expected, these two traits showed overlap in size with male pollinator total length (MaleTotalL) and abdomen length (MaleAbdomenL) in C. trapeziformis (Fig. 2). The labellum-shortening experiments further revealed that the duration of copulation attempts on C. trapeziformis was severely disrupted (Fig. 4b), likely due to the breakup of mechanical fit between Callus-Tip and male abdomen length.

In the light of this interesting finding, we investigated the correlation between mean Callus-Tip length and mean male pollinator abdomen length for all Chiloglottis species and pollinators measured in this study (see Table S3). A significantly positive relationship was found for diamond-shaped species (Pearson’s product–moment correlation: r = 0·839, P = 0·037, Fig S6), indicating that mechanical fit between these traits may be common in all diamond-shaped Chiloglottis.

By contrast, in the cordate-shaped C. valida whose pollinator attempts copulation in the reverse orientation, we found no match between pollinator and labellum size. Nor did experimentally shortening the labellum length alter pollinator behaviour. We also found no relationship between the mean Calli-Tip length and mean male pollinator abdomen length for cordate-shaped species (r = −0·241, P = 0·603, Fig. S6). Other mechanical fit traits may also be associated with the fit between the male pollinator and the floral reproductive structures where pollination occurs. One such floral trait was the distance between the calli and the pollinia. As predicted, we found overlapping sizes between pollinator thorax height and Calli-Pollinia distance (Fig. 2) in both C. trapeziformis and C. valida.

We defined female mimicry as floral traits such as size and colour that match the female of the pollinator. Here, we explicitly excluded sex pheromone mimicry, which is already clearly established in our study system (Schiestl et al. 2003; Schiestl & Peakall 2005). The role of such morphological traits is to enhance both attraction and the intensity of sexual responses of the male pollinator beyond that achieved by chemical mimicry alone. The implicit assumption is that such traits contribute to the false perception by the male that he has found a female. Directly testing this assumption is challenging. Furthermore, when specific floral structures, such as the calli, are both the known source of the chemical mimicry and putative morphological mimics, experimental floral manipulations may not be feasible.

Mimicry traits must enhance the reproductive fitness of the plant; thus, our fitness proxy of attempted copulation holds true for both mechanical fit and mimicry traits. We postulated in the introduction that the sexual dimorphism in our wasp study system of large winged males vs. small wingless females might provide a unique opportunity to help tease apart traits for mechanical fit from traits for female mimicry. In C. trapeziformis, all of our a priori predictions for trait matches with female wasp morphology were confirmed. In C. valida, we also found overlapping sizes between orchid structures and female components consistent with the mimicry hypothesis.

Additional evidence for mimicry was found in the colour similarity of calli and female abdomen. The spectral reflectance of both fell within the very centre of the colour hexagon representing hymenopteran vision (Chittka 1992), effectively making them both achromatic (Chittka et al. 1994). Interestingly, a recent experimental study revealed that achromatic flowers might exclude hymenopteran visitors because they are difficult to detect (Lunau et al. 2011). In this case, both labellum calli and female wasps may thus be similarly cryptic in colour.

Other systems also provide evidence for morphological adaptation in sexual deception. For example, in the Australian fungus gnat pollinated sexually deceptive orchid, Pterostylis sanguinea, experimental rotation of the labellum generated a significant reduction in the copulation attempts of the pollinator (Phillips et al. 2014). In the sexually deceptive African daisy, Gorteria diffusa, floral morphology, rather than semiochemicals, is the main form of attraction (De Jager & Ellis 2012). The male bee fly pollinators of this species are attracted to complex spots on G. diffusa‘s ray florets and grip these spots when exhibiting copulation behaviour (Ellis & Johnson 2010), much like male wasps grip calli on Chiloglottis labella. Floral forms of G. diffusa differing in deceptive spot morphology also elicit different behaviours from their shared pollinator, including where they land and the likelihood and duration of attempted copulation (De Jager & Ellis 2013), confirming the potential for deceptive morphology to strongly influence pollinator behaviour (De Jager & Ellis 2014b).

In conclusion, our results indicate that there is considerable scope for pollinator-mediated selection to operate on floral morphology in sexually deceptive plants. We found evidence for both mechanical fit to male pollinators and mimicry of female wasps. We also discovered that pollinator behavioural differences, such as orientation during copulations attempts, are correlated with contrasting floral types, suggesting a key role for pollinator behaviour in shaping floral morphology. Thus, despite the critical and long-distance role of chemical cues in securing pollinator behaviour, we offer a new and compelling case for the importance of floral morphology in sexual deception. We recommend that future studies experimentally investigate the role of pollinator behaviour as a driver of floral morphological trait variation that is likely under pollinator-mediated selection.