An Experimental Evaluation of Traits That Influence the Sexual Behaviour of Pollinators in Sexually Deceptive Orchids

An experimental evaluation of traits that influence the sexual behaviour of pollinators in sexually deceptive orchids

Ryan D. Phillips1,2,3* & Rod Peakall1 5 1 Ecology and Evolution, Research School of Biology, The Australian National University, Canberra, ACT, 2601, Australia 2 Kings Park and Botanic Garden, The Botanic Garden and Parks Authority, West Perth, 6005, Western Australia, Australia 10 3 Department of Ecology, Environment & Evolution, La Trobe University, Victoria, 3086, Australia *Corresponding author: email: [email protected] Ph: 61 2 61252866 Fax: 61 2 61255573 15 Running head: floral traits and sexual behaviour

Abstract

20 Pollination by sexual deception of male insects is perhaps one of the most remarkable cases of mimicry in the plant kingdom. However, understanding the influence of floral traits on pollinator behaviour in sexually deceptive orchids is challenging, due to the risk of confounding changes in floral odour when manipulating morphology. Here, we investigated the floral traits influencing the sexual response of male Zaspilothynnus nigripes (Tiphiidae) 25 wasps, a pollinator of two distantly related sexually deceptive orchids with contrasting floral architecture, Caladenia pectinata and Drakaea livida. In D. livida the chemical sexual attractant is emitted from the labellum, while in C. pectinata it is produced from the distal sepal tips, allowing manipulative experiments. When controlling for visual cues there was no difference in long distance attraction, though the floral odour of D. livida induced copulation 30 more frequently than that of C. pectinata. The role of colour in pollinator sexual attraction was equivocal, indicating that colour may not be a strong constraint on the initial evolution of sexual deception. The frequency of wasp visitors landing on C. pectinata decreased when the amount of floral odour was reduced, but attempted copulation rates were enhanced when the source of floral odour was associated with the labellum. These latter variables may represent

35 axes of selection that operate across many sexually deceptive species. Nonetheless, the observed variation in floral traits suggests flexibility in how sexual deception can be achieved.

Keywords: sexual deception, mimicry, floral traits, pollinator behaviour 40

Introduction

Pollination by sexual deception, where male insects are attracted to the flower by chemical 45 and/or physical mimicry of a female insect (Ellis & Johnson, 2010; Schiestl et al., 1999, Schiestl et al., 2003; de Jager & Peakall, 2016), is perhaps one of the most remarkable of all pollination strategies. The investigation of plants with unconventional floral forms is revealing an increasing diversity of plants pollinated via this strategy (Vereecken et al., 2012; Phillips et al., 2014b; Arakaki et al., 2016). Consequently, sexual deception is now known to 50 be geographically widespread, and to involve at least several hundred species, with examples from Europe, Australia, South and central America, southern Africa and Asia (Singer, 2002; Blanco & Barboza, 2005; Ellis & Johnson, 2010; Vereecken et al., 2012; Phillips et al., 2014b; Arakaki et al., 2016; Bohman et al., 2016a). Despite recently discoveries in the Asteraceae (Ellis & Johnson, 2010) and Iridaceae (Vereecken et al., 2012), the vast majority 55 of species using this strategy are in the Orchidaceae, where at least 20 genera are known to contain sexually deceptive species (Johnson & Schiestl, 2016; Bohman et al., 2016a). While there is a strong trend towards reduced floral size, dull colouration and insectiform structure of the labellum (Johnson & Schiestl, 2016), there are also sexually deceptive species with large, sometimes colourful flowers that lack insectiform structures (Phillips et al., 2009, 60 2017; Phillips & Peakall 2018) raising the important question of which traits are needed to achieve the sexual attraction of pollinators.

In sexually deceptive pollination systems, one intuitively expects selection to favour traits that increase sexual behavior at the flower, since more time on the flower, or more vigorous 65 sexual behavior, may increase the frequency of pollen removal and pollination (de Jager & Peakall, 2018). However, teasing apart the effect of various floral traits on sexual behavior raises some important methodological complexities. For example, in many sexually deceptive orchids, pollinator attraction is achieved via chemical cues (Kullenberg, 1961; Stoutamire, 2

1983; Schiestl et al., 2003; Bohman et al., 2016a). Therefore, when experimentally 70 disentangling the role of the chemical, visual and morphological cues, any manipulation must avoid confounding floral odour effects with other cues (de Jager & Ellis 2012; de Jager & Peakall, 2016, 2018). This is not so easy to achieve in most genera of sexually deceptive orchids where odour release occurs from the labellum (Kullenberg, 1961; Phillips et al., 2013; Phillips et al., 2014b; de Jager & Peakall, 2016), meaning that labellum shape or size 75 cannot be manipulated without the need to control for odour. Furthermore, interspecific comparison of the effect of floral traits on pollinator behaviour is also confounded when plants use different pollinator species. This is a particularly acute challenge for sexually deceptive orchids where pollinator specificity is extreme, and pollinator sharing rare or absent in most genera (but see Gaskett & Herberstein, 2010; Gogler et al., 2009; Phillips et 80 al., 2017 for some exceptions).

In order to understand the floral traits that affect the attraction of sexually deceived pollinators, in this study we take advantage of an exceptional case of pollinator sharing where the thynnine wasp Zaspilothynnus nigripes (Guérin, 1842) pollinates two distantly related 85 orchids, Caladenia pectinata R.S. Rogers and Drakaea livida J.Drumm., by sexual deception (Phillips et al., 2013; Fig. 1). Unlike D. livida, C. pectinata produces the sexual attractant entirely from glands on the distal sepal tips rather than the labellum (hereafter called clubs), meaning that other floral traits can be manipulated without altering the amount of sexual attractant produced. Interestingly, the copulation rate is over three times higher in D. livida 90 than C. pectinata (Phillips et al., 2013), providing the opportunity to explore traits that affect levels of sexual attraction.

Research on other sexually deceptive systems has provided some clues about the factors that are likely to be important for enhancing sexual attraction. For example, experiments using 95 synthesized semio-chemicals have demonstrated that both changes in the blend of chemicals produced, and the quantity of compounds released, affect the number of pollinators responding and the frequency of attempted copulation (Ayasse et al., 2003; Schiestl, 2004; Vereecken & Schiestl, 2008; Peakall et al., 2010; Bohman et al., 2014; Bohman et al., 2017; Xu et al., 2017). In some sexually deceptive orchids, colour also appears to be important for 100 locating the flower, either through visual mimicry or colour contrast to maximize detectability (Streinzer et al., 2009; Gaskett & Herberstein, 2010; Gaskett et al., 2016), though the effect of colour on copulation rate is unknown in any orchid. Once pollinators

have landed on the flower, morphological traits are known to enhance copulation rates and duration at least in some systems (de Jager & Ellis, 2012; Phillips et al., 2014b; de Jager & 105 Peakall, 2016, 2018). For example, experimental reduction in labellum size of Chiloglottis, while avoiding removal of the odour producing callus, showed that a labellum size corresponding to the length of the mimicked female elicited a longer duration of attempted copulation than smaller labella (de Jager & Peakall, 2016).

110 Here, we apply a series of innovative floral manipulations to test if three floral traits (amount of floral odour, site of odour release, and colour), which are predicted to affect reproductive success in sexually deceptive orchids, influence the extent of pollinator response and sexual behaviour with the flower. Further, we seek to understand the basis of pronounced variation in pollinator sexual behaviour with different species of sexually deceptive orchid (see Fig. 1). 115 Due to the large size of Z. nigipres (approximately 20 mm body length) and its vigorous copulatory behaviour with the flower, in this system pollinator responses can be readily classified and into three discrete steps, long-distance attraction, landing on the flower, and attempted copulation. Such a classification is not always possible in other sexually deceptive species where the pollinator and flower is small and/or the sexual behavior is more subtle 120 (e.g. Phillips et al., 2014b). Further, we can subdivide ‘sexual attraction’ into two quantifiable components. We define ‘enhanced sexual response’ as a significant increase in the total number of wasps responding to the flower. We define ‘enhanced sexual behavior’ as a significant increase in the proportion of wasps landing on the flower, and/or the proportion of landing wasps that attempt copulation with the flower. In this context, we address four 125 specific questions for Z. nigripes: (i) Is there interspecific variation in the attractiveness of the floral odour? (ii) Does a decrease in the amount of floral odour reduce sexual attraction? (iii) Does sexual attraction increase when the attractant is associated with the labellum? (iv) Does a dull-coloured labellum enhance sexual attraction?

130 Materials and methods

Study species

In many of the well-known sexually deceptive orchid genera such as Drakaea, Chiloglottis 135 and Ophrys, virtually all species are sexually deceptive with the possible exception of a minority of self-pollinating species (Kullenberg, 1961; Paulus & Gack, 1990; Peakall et al.,

2010; Gaskett, 2011; Phillips et al., 2014a). However, Caladenia is exceptional among orchids in that it contains both species pollinated by sexual deception, and species pollinated by food-foraging insects, both through reward and deception (Stoutamire, 1983; Faast et al., 140 2009; Phillips et al., 2009; Reiter et al., 2018). Although many sexually deceptive Caladenia exhibit the reduced floral morphology often seen in other genera of sexually deceptive orchids, some species, such as C. pectinata, have large flowers that are of similar dimensions and shape to food-deceptive species. However, the food-deceptive species are generally more brightly coloured (most commonly white, pink or yellow; (Phillips et al., 2009; Phillips et al., 145 2017) than sexually deceptive species, including C. pectinata, which tend to be dull red and green (Phillips et al., 2009) (Fig. 1).

Thus far all known cases of sexual deception in Caladenia involve male thynnine wasps. Among the aculeate hymenoptera, thynnine wasps exhibit a unique mating system where the 150 wingless females crawl to a perch and release a sex pheromone to attract the volitant males (Alcock & Gwynne, 1987). The males often compete vigorously for females, with the successful male flying off with her in copula to a food source (Alcock & Gwynne, 1987). In all Drakaea, the combination of a hinge part way along the labellum, and the momentum of the male wasp as he attempts to pick up the pseudo-female, tips the wasp upside down and 155 positions him against the column where pollination occurs (Peakall, 1990, Phillips et al., 2014a). While both C. pectinata and D. livida are dull-coloured, the colouration is not a precise match to the female (Gaskett et al., 2016) (Fig. 1). The chemicals used in pollinator attraction remain unknown, although novel pyrazines may be involved (Bohman et al., 2012) as is the case for Drakaea glyptodon and its pollinator Zaspilothynnus trilobatus (Bohman et 160 al., 2014).

Study site and collection details

Experiments were undertaken during October in 2014-2016 at a single site south of Yallingup 165 in Leeuwin-Naturaliste National Park, south-western Australia (33° 39' 32"S 115° 02' 09"E). Experimental work was conducted on sunny days with a maximum temperature of 20°C or more. Each experiment was conducted within a single season. The three orchid species used in this study, Drakaea livida, Caladenia pectinata and the food-deceptive Caladenia longicauda Lindl. (used as the source of white labellums in experiment 5), do not grow at this 170 study site, but Zaspilothynnus nigripes occurs here in abundance (see Phillips et al., 2013). 5

The source populations of experimental flowers are represented by vouchers in the Western Australian herbarium (C. pectinata: RDP0218 PERTH08643407, RDP0247 PERTH08643261); D. livida: RDP0009 PERTH08603456, RDP0019 PERTH08603537; C. longicauda: RDP0301 PERTH08645825, RDP0329 PERTH08739854). 175

Experimental methods

180 Experiments were undertaken using the pollinator baiting method, where picked flowers that are moved to a new part of the landscape can rapidly attract sexually deceived pollinators (Stoutamire, 1983; Peakall, 1990). In between periods of baiting the orchids were kept in a sealed box. For all experiments flowers were kept fresh for use in a portable refrigerator at 4°C. While Caladenia flowers stored like this can remain attractive to pollinators for 2-3 185 weeks (Phillips et al., 2017), all experiments were conducted within four days of picking. Further, in each experiment, flowers from both treatments were picked at the same time to avoid any confounding age affects. Each individual experiment (1 to 5) was undertaken in a single year, but typically over three consecutive days of suitable climatic conditions.

190 As part of our previous study of this system (Phillips et al., 2013), we demonstrated experimentally that dissections do not affect the attractiveness of the flower to the pollinators. As such, our experiments, which involved modifying floral traits of C. pectinata, were based upon the assumption that the sexual response of the pollinator is not affected by floral dissection. 195 Experiment 1: Long-distance attractiveness of D. livida vs C. pectinata

To test for any species-specific difference in the long-distance attraction of pollinators, while controlling for any confounding effects of appearance or site of odour release, baiting was 200 undertaken with either the clubs of C. pectinata or the labellum of D. livida concealed below the labellum of C. pectinata flowers that had had their sepal clubs removed. The dissected flower parts were secured in this concealed position by placing them on a pin inserted through the ovary. Baiting was conducted for five trials of two minutes, alternating between species. This experiment was replicated six times, alternating between starting species and 6

205 using a different set of new flowers for each replicate experiment. For both species, the number of wasps attracted and the time to first pollinator response was recorded in the field, and the average calculated across trials to give a value for each experiment. We then used this value to test for a difference in the average response between species using a student’s t-test in JMP v12.0.1 (SAS Institute Inc., 2015). 210

General methods for Experiments 2-5

In experiments 2-5 we baited with each treatment for 10 minutes, and repositioned the flower 215 mid-way through the trial after 5 minutes to renew the pollinator response as per (Phillips et al., 2013). The experiment was replicated six times, with a new set of flowers for each experiment, alternating between starting treatment. For each responding wasp we scored whether it landed on the flower and if it attempted copulation. We did not discriminate between copulation with different parts of the flower. However, except for the intact 220 treatment in experiment 4, the attractive clubs were concealed under the labellum such that copulation was almost exclusively with the labellum. For experiments 2-5, we compared the total number of responses per experimental replicate using student’s t-tests in JMP v12.0.1 (SAS Inc., 2015). To test for differences in the proportion of pollinators alighting on the flower, and the proportion of landing wasps that copulated with the flower, we used 225 Generalised Linear Mixed Models (GLMM). These analyses were run in the R package lme4 (Bates et al., 2015) using the function glmer with a binomial distribution. We treated experimental replicate (1 – 6) as a random factor in recognition of the different environmental conditions that wasps were exposed to across experimental replicates, and that a different set of experimental flowers was used for each experimental replicate. 230 Experiment 2: Does the floral odour of D. livida elicit greater sexual attraction than C. pectinata?

As already noted, Phillips et al., 2013 showed that copulation attempts by Z. nigripes are 235 significantly more frequent with D. livida than C. pectinata flowers (D. livida = 34.5%; C. pectinata = 9.3%; Fig. 1). To test if this is attributable to differences in floral odour alone, the floral parts responsible for odour release (labellum in D. livida, three sepal clubs in C. pectinata; Fig. 1) were concealed below the labellum of C. pectinata flowers that had had 7

their sepal clubs removed. As above, dissected flower parts were secured in this concealed 240 position by placing them on a pin inserted through the ovary.

Experiment 3: Does a decrease in floral odour reduce sexual attraction of the pollinator?

To investigate the effect of a reduction in the amount of floral odour, here we evaluated the 245 pollinator response when just one, compared to three, of the odour-producing clubs was concealed under the labellum of C. pectinata. Due to size constraints imposed by the labellum, and the need to keep the clubs concealed from the pollinator, it was not possible to use more than three clubs in any given treatment. Each treatment for each trial used a new set of randomly selected clubs that had been freshly excised from whole flowers taken from the 250 source population and stored as described above. The flowers used for each trial had their clubs removed such that the clubs concealed under the labellum were the only source of sexual attractant.

Experiment 4: Does the sexual attraction of the pollinator increase when the attractant is 255 associated with the labellum?

To test if sexual attraction to the flower increases when the attractant is associated with the labellum, we conducted an experiment comparing intact C. pectinata flowers with C. pectinata flowers where the three clubs were removed and pinned under the labellum (as 260 above). As a control for the incisions made to remove the clubs from the manipulated flower (as per de Jager & Peakall, 2016), on the intact flowers a 2 mm incision was made on the base of the dorsal sepal (which is obscured from pollinator view).

Experiment 5: Does a dull-coloured labellum enhance the sexual attraction of the pollinator? 265 The food-deceptive C. longicauda has a labellum that is similar in size and shape to that of C. pectinata (Hopper & Brown, 2001), but is white rather than red in colour (for spectral reflectance see supplement 1). To test whether the C. pecintata labellum could be substituted for a C. longicauda labellum without chemical inhibition, a pilot experiment was conducted 270 (supplement 2). This confirmed that the odour of the labellum of C. longicauda does not have an inhibitory effect on the sexual behaviour of Z. nigripes. Therefore, we proceeded to compare pollinator behaviour on the red labellum of C. pectinata with the white labellum of 8

C. longicauda. Using C. pectinata without clubs or a labellum as the base structure, we then used a pin to attach three clubs of C. pectinata under either, a red labellum of C. pectinata, or 275 a white labellum of C. longicauda. For each new experimental replicate, new reconstructed flowers were used, although within a trial the same C. pectinata flower was used for both treatments.

280 Results

Experiment 1: Long distance attractiveness of D. livida vs C. pectinata

285 When we controlled for floral odour tissue position by concealing the respective parts of C. pectinata (clubs) and D. livida (labellum) under the labellum of C. pectinata, there was no significant difference in the total number of wasps responding to the manipulated C. pectinata flower (Fig. 2). Similarly, the average length of time for the first wasp to land on the flower did not differ between species (Fig. 2), demonstrating no difference in long 290 distance attraction.

Experiment 2: Does the odour of D. livida elicit greater sexual attraction than C. pectinata?

Overall there was no significant difference in the total number of wasps responding to the 295 manipulated C. pectinata flower when there was three C. pectinata clubs or a single D. livida labellum concealed under the labellum (Fig. 2). Further, the proportion of the wasps landing on the flower was not significantly different between species (Fig. 2). However, the proportion of wasps attempting copulation with the flower was significantly higher with the D. livida labellum compared with the C. pectinata clubs (Fig. 2). As such, it is evident that 300 the odour of D. livida leads to more frequent sexual behaviour than the odour of C. pectinata.

Experiment 3: Does a decrease in floral odour reduce the sexual attraction of the pollinator?

There was no significant difference in the total number of wasps responding to the 305 manipulated C. pectinata flower with three clubs versus flowers with one club concealed under the labellum (Fig. 3). However, a significantly lower proportion of wasps landed on the 9

flowers (Fig. 3) with the solitary club, providing evidence that a decrease in the amount of floral odour reduces sexual behaviour. Nonetheless, of those landing on the flower, there was no significant difference in the proportion of copulation between the treatments (Fig. 3). 310 Experiment 4: Does the sexual attraction of the pollinator increase when the attractant is associated with the labellum?

There was no significant difference in the total number of wasps responding to intact C. 315 pectinata, where the three clubs are presented on the tips of the widely splayed sepals (Fig. 1), compared with manipulated flowers with three clubs concealed under the labellum (Fig. 3). Similarly, there was no significant difference in the proportion of wasps that then landed on the flower (Fig. 3). However, significantly more wasps attempted copulation with the flower when the clubs were concealed under the labellum (Fig. 3), demonstrating increased 320 sexual behaviour when the sexual attractant is associated with the labellum.

Experiment 5: Does a dull-coloured labellum enhance the sexual attraction of the pollinator?

There was no difference in the number of wasps responding to manipulated C. pectinata 325 flowers when their red labellum was switched to the white labellum of C. longicauda (but with C. pectinata clubs concealed under the respective labella) (Fig. 3). Further, there was no significant difference in the proportion of wasps landing on or copulating with the flower depending on whether the labellum was white or red (Fig. 3).

330 Discussion

Floral odour – species and quantity effects on sexual attraction

As is the case for the pollinators of all other sexually deceptive orchids investigated to date 335 (reviewed in Bohman et al., 2016a), floral odour is crucial for the attraction of Z. nigripes by D. livida and C. pectinata (Phillips et al., 2013). We showed that reducing the amount of floral odour leads to a decrease in the number of pollinators landing on the flower. This is a similar result to Schiestl (2004) who used synthesised sex pheromone of Neozeleboria cryptoides wasps, whose males pollinate the sexually deceptive orchid Chiloglottis 340 trapeziformis, to show that more pheromone leads to more frequent attempted copulation. 10

However, for the first time we report experimental evidence for a possible difference in the active chemical constituents of the floral odour of D. livida and C. pectinata. While long- distance attraction as measured by response time and the number of individuals attracted did not differ between the species, the odour of D. livida triggered higher rates of copulation than 345 the odour of C. pectinata. This difference in copulation rate may indicate that there are differences in the semiochemical blend between the two species. It is already known in several other species of thynnine wasps where the semiochemicals involved in orchid pollinator attraction have been elucidated that considerable variation in the copulation rate can be achieved by altering ratios of the different constituents of the pheromone blend 350 (Bohman et al., 2014; Peakall et al., 2010; Bohman et al., 2017). In one case, it has even been shown that analogues to the female pheromone blend can elicit both frequent attraction and attempted copulation (Bohman et al., 2016b). As such, there is clearly a capacity for sexual attraction of the same pollinator to be achieved via different semiochemical blends.

355 Effect of site of odour release on attraction via sexual deception

Intuitively, producing the sexual attractant from the female-like labellum is expected to lead to greater attraction via sexual deception and more frequent pollination than the production of semiochemicals from other floral parts. We tested this prediction by comparing intact C. 360 pectinata with their three clubs in their natural position on the sepal tips with manipulated C. pectinata where the three clubs were concealed under the labellum. This change in position led to an approximately three-fold higher rate of attempted copulation, despite identical chemistry. The increase in attempted copulation may reflect more compelling mimicry, as the olfactory and visual cues now coincide, or wasps being more easily able to firmly grasp 365 the fake female rather than struggling to grasp the relatively fine sepal tips. These results suggest that in many sexually deceptive systems, where strong sexual attraction to the labellum may increase the likelihood of pollination (Rakosy et al., 2017; de Jager & Peakall, 2018), producing the odour from the labellum is likely to be the most effective strategy.

370 The potential benefits of producing the sexual attractant from the labellum raises the question of why C. pectinata does not use this strategy? One possibility is that there are strong developmental and/or morphological constraints on the ability to produce sexual attractants from the labellum, which is a modified petal. However, evidence from other groups of Caladenia suggest that this is unlikely to be the case. Firstly, the production of sexual 11

375 attractant from the labellum occurs in some other groups of Caladenia (Peakall & Beattie, 1996). Secondly, while C. pectinata does not have odour-producing clubs on the tips of the petals, numerous other Caladenia do (at least in some individuals e.g. Hopper & Brown, 2001). A possible alternative explanation is that due to the floral architecture of C. pectinata, attempted copulation with the labellum does not increase reproductive success. Indeed, it 380 seems that for pollination to occur in C. pectinata, the wasp moves head first to the base of the column with its head facing down. However, when they try to copulate with the labellum of C. pectinata in our experiments, they generally attempt to copulate with the labellum tip while facing forwards and upwards, the typical position for initiating copulation. This position fails to bring them into contact with the column, suggesting that in C. pectinata 385 selection may favour traits that lead to long-range attraction of pollinators with a sex pheromone, but not wasps attempting copulation with the labellum.

Effects of labellum colour on attraction via sexual deception

390 The experiment switching the dull red labellum of C. pectinata with the bright white labellum of C. longicauda showed that labellum colour had no significant effect on the frequency of landing or copulation rates. On one hand, this finding is consistent with the observation that some sexually deceptive Caladenia do have a brightly coloured labellum (Phillips et al., 2017), and that in some dull-coloured species the labellum is not a close match to the colour 395 of the female of the pollinator species (Gaskett et al., 2016). However, this finding is at odds with the correlation between the use of a sexually deceptive pollination strategy and dull- coloured, often insectiform labella. For both the frequency of landing and copulation, our results were approaching significance (P = 0.087 and P = 0.058) suggesting that further experimental work would be of value to resolve this issue, perhaps manipulating both 400 labellum colour and colour contrast with the other tepals (e.g. Streinzer et al., 2009). Nonetheless, the frequent copulation attempts elicited by the white labellum of C. longicauda suggests that colour is not a strong constraint on the initial evolution of sexual deception.

Sexual attraction and axes of selection in sexually deceptive orchids 405 Even across distantly related orchids, the evolution of sexual deception is typically accompanied by a similar suite of floral traits (e.g. reduced flower size, dull colouration, insectiform structure), suggesting that common selection pressures along multiple axes of 12

floral variation is associated with the evolution of this strategy (Table 1 and references 410 therein). Our experimental study has shown that the type and amount of floral odour, and the site of odour release, all influence the degree of sexual attraction in terms of either number of responding males, the proportion of responding males that land on the flower, or the proportion of these that attempt copulation. While our data provide mechanisms that may explain the predominant trends in the floral traits of sexually deceptive orchids, the 415 remarkable variability in floral traits between orchids that exploit the same or closely related pollinators (Fig. 1; Phillips et al., 2017) demonstrates that there is flexibility in how pollination via sexual deception can be achieved. An important area of future research will be to understand which axes of floral variation experience stronger selection and therefore impose a greater constraint on transitions in pollination strategy to and from sexual deception 420 (e.g. Ellis et al., 2014; Rakosy et al., 2017). For example, Rakosy et al. (2017) showed that in Ophrys parts of the labellum used as gripping points affect the duration of copulation, and thus may be under stronger selection for shape than other parts of the labellum. Similarly, de Jager and Peakall (2018) show strong evidence for stabilizing selection on the distance from the callus to the labellum tip of Chiloglottis trapeziformis, but not other morphological traits 425 of the labellum. Given the wide diversity of insect families exploited by the various genera of sexually deceptive plants, the floral traits involved in sexual deception, and the strength of selection that they are under, is likely to vary considerably between orchid lineages.

References 430 Alcock, J. & Gwynne, D.T. 1987. Courtship feeding and mate choice in thynnine wasps (Hymenoptera, Tiphiidae). Aust. J. Zool. 35: 451-458. Arakaki, N., Yasuda, K., Kanayama, S., Jitsuno, S., Oike, M. & Wakamura, S. 2016. Attraction of males of the cupreous polished chafer Protaetia pryeri pryeri 435 (Coleoptera: Scarabaeidae) for pollination by an epiphytic orchid Luisia teres (Asparagales: Orchidaceae). Appl. Entomol. Zool. 51: 241-246. Ayasse, M., Schiestl, F.P., Paulus, H.F., Ibarra, F. & Francke, W. 2003. Pollinator attraction in a sexually deceptive orchid by means of unconventional chemicals. Proc. Roy. Soc. B 270: 517-522. 440 Bates, D., Maechler, M., Bolker, B. & Walker, S. 2015. Fitting Linear Mixed-Effects Models Using lme4. J. Stat. Softw. 67: 1-48.

Blanco, M.A. & Barboza, G. 2005. Pseudocopulatory pollination in Lepanthes (Orchidaceae : Pleurothallidinae) by fungus gnats. Ann. Bot. 95: 763-772. Bohman, B., Flematti, G.R., Barrow, R.A., Pichersky, E. & Peakall, R. 2016a. Pollination by 445 sexual deception - it takes chemistry to work. Curr. Opin. Plant Biol. 32: 37-46. Bohman, B., Jeffares, L., Flematti, G., Phillips, R.D., Dixon, K.W., Peakall, R. et al. 2012. The discovery of 2-Hydroxymethyl-3-(3methylbutyl)-5-methylpyrazine: a semiochemical in orchid pollination. Org. Lett. 14: 2576-2578. Bohman, B., Karton, A., Dixon, R.C.M., Barrow, R.A. & Peakall, R. 2016b. Parapheromones 450 for Thynnine Wasps. J. Chem. Ecol. 42: 17-23. Bohman, B., Phillips, R.D., Flematti, G.R., Barrow, R.A. & Peakall, R. 2017. The spider orchid Caladenia crebra produces sulfurous pheromone mimics to attract its male wasp pollinator. Angew. Chem. 129: 1-5. Bohman, B., Phillips, R.D., Menz, M.H.M., Berntsson, B.W., Flematti, G.R., Barrow, R.A. et 455 al. 2014. The discovery of alkyl- and hydroxyl pyrazines as insect sex pheromones and orchid semiochemicals. New Phytol. 203: 939-952. de Jager, M.L. & Ellis A.G. 2012. Gender-specific pollinator preference for floral traits. Funct. Ecol. 26: 1197–1204. de Jager, M.L. & Peakall, R. 2016. Does morphology matter? An explicit assessment of floral 460 morphology in sexual deception. Funct. Ecol. 30: 537-546. de Jager, M.L. & Peakall, R. 2018. Experimental examination of pollinator mediated selection in a sexually deceptive orchid. Ann. Bot., in press. Ellis, A.G., Brockington, S.F., de Jager, M.L., Mellers, G., Walker, R.H. & Glover, B.J. 2014. Floral trait variation and integration as a function of sexual deception in 465 Gorteria diffusa. Phil. Trans. R. Soc. B 369: 20130563. Ellis, A.G. & Johnson, S.D. 2010. Floral mimicry enhances pollen export: the evolution of pollination by sexual deceit outside of the Orchidaceae. Am. Nat. 176: E143-E151. Faast, R., Farrington, L., Facelli, J.M. & Austin, A.D. 2009. Bees and white spiders: unravelling the pollination syndrome of Caladenia rigida (Orchidaceae). Aust. J. Bot. 470 57: 315-325. Gaskett, A.C. 2011. Orchid pollination by sexual deception: pollinator perspectives. Biol. Rev. 86: 33-75. Gaskett, A.C., Endler, J.A. & Phillips, R.D. 2016. Convergent evolution of sexual deception via chromatic and achromatic contrast rather than colour mimicry. Evol. Ecol. 31: 475 205-227. 14

Gaskett, A.C. & Herberstein, M.E. 2010. Colour mimicry and sexual deception by Tongue orchids (Cryptostylis). Naturwissenschaften 97: 97-102. Gögler, J., Stokl, J., Sramkova, A., Twele, R., Francke, W., Cozzolino, S. et al. 2009. Menage trois - two endemic species of deceptive orchids and one pollinator species. 480 Evolution 63: 2222-2234. Hopper, S.D. & Brown, A.P. 2001. Contributions to Western Australian orchidology: 2. New taxa and circumsciptions in Caladenia. Nuytsia 14: 27-307. Johnson, S.D. & Schiestl, F.P. 2016. Floral Mimicry. Oxford University Press, Oxford, United Kingdom. 485 Kullenberg, B. 1961. Studies in Ophrys pollination. Zoologiska Bidrag fr°an Uppsala 34: 1- 340. Paulus, H.F. & Gack, C. 1990. Pollination of Ophrys (Orchidaceae) in Cyprus. Plant Syst. Evol. 169: 177-207. Peakall, R. 1990. Responses of male Zaspilothynnus trilobatus Turner wasps to females and 490 the sexually deceptive orchid it pollinates. Funct. Ecol. 4: 159-167. Peakall, R. & Beattie, A.J. 1996. Ecological and genetic consequences of sexual deception in the orchid Caladenia tentaculata. Evolution 50: 2207-2220. Peakall, R., Ebert, D., Poldy, J., Barrow, R.A., Francke, W., Bower, C.C. et al. 2010. Pollinator specificity, floral odour chemistry and the phylogeny of Australian sexually 495 deceptive Chiloglottis orchid: implications for pollinator-driven speciation. New Phytol. 188: 437-450. Peakall, R. & Handel, S.N. 1993. Pollinators discriminate among floral hieghts of a sexually deceptive orchid - implications for selection. Evolution 47: 1681-1687. Phillips, R.D., Brown, G.R., Dixon, K.W., Hayes, C., Linde, C.C. & Peakall, R. 2017. 500 Evolutionary relationships among pollinators and repeated pollinator sharing in sexually deceptive orchids. Journal of Evol. Biol. 30: 1674-1691 Phillips, R.D., Faast, R., Bower, C.C., Brown, G.R. & Peakall, R. 2009. Implications of pollination by food and sexual deception on pollinator specificity, fruit set, population genetics and conservation of Caladenia. Aust. J. Bot. 57: 287-306. 505 Phillips, R.R. & Peakall, R. 2018. Breaking the rules: discovery of sexual deception in Caladenia abbreviata (Orchidaceae), a species with brightly coloured flowers and a non-insectiform labellum. Aust. J. Bot. in press.

Phillips, R.D., Peakall, R., Hutchinson, H.F., Linde, C.C., Xu, T., Dixon, K.W. et al. 2014a. Specialized ecological interactions and plant species rarity: The role of pollinators and 510 mycorrhizal fungi across multiple spatial scales. Biol. Conserv. 169: 285-295. Phillips, R.D., Scaccabarozzi, D., Retter, B.A., Hayes, C., Brown, G.R., Dixon, K.W. et al. 2014b. Caught in the act: pollination of sexually deceptive trap-flowers by fungus gnats in Pterostylis (Orchidaceae). Ann. Bot. 113: 629-641. Phillips, R.D., Xu, T., Hutchinson, M.F., Dixon, K.W. & Peakall, R. 2013. Convergent 515 specialization - the sharing of pollinators by sympatric genera of sexually deceptive orchids. J. Ecol. 101: 826-835. Rakosy, D., Cuervo, M., Paulus, H.F. & Ayasse, M. 2017. Looks matter: changes in flower form affect pollination effectiveness in a sexually deceptive orchid. Journal of Evolutionary Biology 30: 1978-1993. 520 Reiter, N., Bohman, B., Flemmati, G.R & Phillips, R.D. 2018. Pollination by nectar foraging thynnine wasps: evidence of a new specialised pollination strategy for Australian orchids. Botanical Journal of the Linnean Society, in press. SAS Institute Inc. (2015) JMP. SAS Institute Inc., Cary, North Carolina. Schiestl, F.P. 2004. Floral evolution and pollinator mate choice in a sexually deceptive 525 orchid. J. Evol. Biol. 17: 67-75. Schiestl, F.P., Ayasse, M., Paulus, H.F., Lofstedt, C., Hansson, B.S., Ibarra, F. et al. 1999. Orchid pollination by sexual swindle. Nature 399: 421-422. Schiestl, F. P., Peakall, R., Mant, J. G., Ibarra, F., Schulz, C., Franke, S. et al. 2003. The chemistry of sexual deception in an orchid-wasp pollination system. Science 302: 530 437-438. Singer, R. B. 2002. The pollination mechanism in Trigonidium obtusum Lindl (Orchidaceae : Maxillariinae): Sexual mimicry and trap-flowers. Ann. Bot. 89: 157-163. Stoutamire, W.P. 1983. Wasp-pollinated species of Caladenia (Orchidaceae) in southwestern Australia. Aust. J. Bot. 31: 383-394. 535 Streinzer, M., Paulus, H. & Spaethe, J. 2009. Floral colour signal increases short-range detectability of a sexually deceptive orchid to its bee pollinator. J. Exp. Biol. 212: 1365-1370. Vereecken, N.J. & Schiestl, F.P. 2008. The evolution of imperfect floral mimicry. P. Natl. Acad. Sci. USA 105: 7484-7488.

540 Vereecken, N. J., Wilson, C.A., Hotling, S., Schulz, S., Banketov, S.A. & Mardulyn, P. 2012. Pre-adaptations and the evolution of pollination by sexual deception: Cope's rule of specialization revisited. Proc. Roy. Soc. B 279: 4786-4794. Xu, H., Bohman, B., Wong, D.C.J., Rodriguez-Delgado, C., Scaffidi, A., Flematti, G.R. et al. 2017. Complex sexual deception in an orchid is achieved by co-opting two 545 independent biosynthetic pathways for pollinator attraction. Curr. Biol. 27: 1867- 1877.

Table 1: Axes of variation in floral traits expected to be under selection in sexually deceptive orchids based on experimental behavioural evidence. Axis of variation Expectation Observed behavioural response Species Reference Amount of sexual More odour, greater sexual More pollinators land if more odour; Caladenia pectinata; Present study; attractant released attraction More pollinators attempt copulation if Chiloglottis trapeziformis (Schiestl, 2004) more odour Composition of sexual Blend closest to female sex Varying blend affects rates of attraction Chiloglottis spp.; Drakaea (Ayasse et al., attractant pheromone leads to greatest and copulation glyptodon; Ophrys speculum 2003; Bohman et attraction al., 2014; Peakall et al., 2010; Schiestl et al., 2003)

attraction greatest from close mimicry Ophrys exaltata (Vereecken & or similar blend Schiestl, 2008) Site of release of sexual Greater attraction when odour More copulation when odour comes Caladenia pectinata Present study attractant from labellum from labellum Scape height More pollinators locate and Optimal height for landing on the Chiloglottis trilabra (Peakall & Handel, respond to flower when height flower 1993) matches mate searching behaviour Floral display colour Either maximises detectability of Increase bee green receptor contrast Ophrys heldreichii (Streinzer et al., labellum OR does not detract between labellum and perianth 2009) from mimicry by labellum increases search time Floral display size Large perianth either increases Perianth removal increases pollinator Ophrys heldreichii (Streinzer et al., detectability OR has no benefit search time 2009) no effect Caladenia pectinata (Phillips et al.,

2013) Labellum colour Increased detection and attraction No significant difference, but more Caladenia pectinata Present study when matches female work needed Labellum size Labellum size that is near the Close match to size of female increases Chiloglottis trapeziformis (Schiestl, 2004; de upper size for females increases copulation duration; males prefer large Jager & Peakall, attraction dummy females 2016, 2018) Presence of labellum Presence will increase landing and Not tested N/A N/A callus attempted copulation Labellum texture/shape More copulation when matches Reverse orientation of labellum leads Pterostylis sanguinea (Phillips et al., morphology/texture of female to reduced copulation 2014b) Shape of gripping points affect Ophrys leochroma (Rakosy et al., duration of copulation 2017)

Fig. 1: Summary of the behaviour of Zaspilothynnus nigripes (iii) in response to Drakaea livida (i, DL) and Caladenia pectinata (ii, CP). The percentage landing (A) is the percentage of all responding individuals. Percentage copulation (B) and percentage column contact (C) is calculated based only on those individuals landing on the flower. Behavioural data for both orchid species was collected in the same populations of wasps on the same days. Percentage pollination rate (D) is the average across multiple populations of each orchid species. Data from (Phillips et al., 2013). Illustrated odour plumes represent the site of release of the sexual attractant. Photos by Rod Peakall (i, ii) and Keith Smith (iii).

Fig. 2: Comparison of the sexual response of Zaspilothynnus nigripes to the floral odour of Drakaea livida (DL) and Caladenia pectinata (CP). Panels (A) and (B) refer to an experiment where the number of wasps responding, and the time taken for the first individual to respond, were compared between orchid species. Panels (C), (D) and (E) refer to a separate experiment designed to test if there was a difference in the proportion of wasps landing or attempting copulation between the two orchid species. To control for any confounding effects of appearance or site of odour release, baiting was undertaken with either the odour- producing clubs of C. pectinata or the labellum of D. livida concealed below the labellum of C. pectinata flowers that had had their sepal clubs removed. The percentage landing is the percentage of all responding individuals. Percentage copulation is calculated using only those individuals landing on the flower. In panels (A) and (C) the x-axis refers to the number of wasps attracted per experimental replicate. Error bars are standard errors.

Fig. 3: Floral traits affecting sexual response and sexual attraction in Zaspilothynnus nigripes. All experiments were undertaken using modified flowers of Caladenia pectinata. The percentage landing is the percentage of all responding individuals. Percentage copulation is calculated using only those individuals landing on the flower. Panels (A)-(C) refer to an experiment where the number of odour producing glands (known as clubs) was manipulated. Panels (D)-(F) refer to an experiment where the site of odour release was either from clubs on the tips of the sepals or the labellum. Panels (G)-(I) refer to an experiment where the red labellum of C. pectinata was switched to the white labellum of Caladenia longicauda. In panels (A), (D) and (G) the x-axis refers to the number of wasps attracted per experimental replicate. Error bars are standard errors.