Convergent specialization – the sharing of pollinators by sympatric genera of sexually deceptive orchids

Authors

  • Ryan D. Phillips,

    Corresponding author
    1. Kings Park and Botanic Garden, The Botanic Garden and Parks Authority, West Perth, WA, Australia
    2. School of Plant Biology, The University of Western Australia, Nedlands, WA, Australia
    • Evolution, Ecology and Genetics, Research School of Biology, The Australian National University, Canberra, ACT, Australia
    Search for more papers by this author
  • Tingbao Xu,

    1. Fenner School of Environment and Society, The Australian National University, Canberra, ACT, Australia
    Search for more papers by this author
  • Michael F. Hutchinson,

    1. Fenner School of Environment and Society, The Australian National University, Canberra, ACT, Australia
    Search for more papers by this author
  • Kingsley W. Dixon,

    1. Kings Park and Botanic Garden, The Botanic Garden and Parks Authority, West Perth, WA, Australia
    2. School of Plant Biology, The University of Western Australia, Nedlands, WA, Australia
    Search for more papers by this author
  • Rod Peakall

    1. Evolution, Ecology and Genetics, Research School of Biology, The Australian National University, Canberra, ACT, Australia
    Search for more papers by this author

Correspondence author. E-mail: ryan.phillips@bgpa.wa.gov.au

Summary

  1. Pollinator sharing can offer powerful insights into the floral traits associated with the evolution of a pollination system and the consequences of floral differences for pollinator behaviour. Here, we investigate the first known case of pollinator sharing between two sexually deceptive plant genera. Floral manipulations were used to test the importance of floral traits for pollinator behaviour and pollination efficiency. We also explored the ecological differences enabling species co-occurrence.
  2. Drakaea livida and Caladenia pectinata (Orchidaceae) exhibit dramatic differences in floral display and the insectiform appearance of the labellum, yet both are pollinated by sexually attracted males of the thynnine wasp Zaspilothynnis nigripes. Because of the prevalence of cryptic species in some genera of thynnine wasps, we confirmed pollinator sharing by a mark–recapture study and sequencing of the mtDNA CO1 region.
  3. Floral dissections revealed that semiochemicals used to attract the pollinator are released from the labellum in D. livida and sepaline clubs in C. pectinata. Drakaea livida was more efficient at converting pollinator attraction into potential pollen deposition leading to higher fruit set. Floral manipulations showed that pollinator contact with the labellum increases when it is the point of semiochemical release. However, sexual attraction to the labellum remained infrequent in C. pectinata in all experimental treatments.
  4. While their distribution and climatic range show extensive overlap, the differences in edaphic requirements of the two orchid species suggest that they rarely co-occur. Therefore, the potential cost of sharing the same pollinator species is not realized.
  5. Synthesis. This case of pollinator sharing confirms that morphological traits do not place a strong constraint on the evolution of sexual deception. However, interspecific differences in floral traits have important consequences for converting attraction into pollination, suggesting that selection can act to increase efficiency at multiple steps of the pollination process. This system provides a novel opportunity to elucidate the chemical, visual and morphological adaptations underpinning the evolution of sexual mimicry.

Introduction

Cases of convergent evolution for the same pollinator(s) can offer powerful insights into the floral traits necessary for the evolution and success of a particular pollination strategy (e.g. Wilson et al. 2004; Johnson 2010). Species with more specialized pollination strategies may evolve exaggerated floral traits as natural selection favours refinement on the preferences, behaviours and morphology of one or few favourable pollinator species (e.g. Castellnos, Wilson & Thomson 2003; Whittall & Hodges 2007). One of the most remarkable of all pollination strategies is that of sexual deception, where plants chemically and/or physically mimic a female insect leading to pollination via the sexually excited males (Schiestl et al. 1999, 2003; Ellis & Johnson 2010). In this study, we confirm the first known case of the sharing of a sexually deceived pollinator by members of two plant genera. This provides a unique opportunity to determine which chemical and morphological adaptations are critical to the evolution of this pollination strategy.

With one documented exception in the Asteraceae (Ellis & Johnson 2010), the strategy of pollination via sexual deception is only known from the Orchidaceae, where it has evolved on multiple continents in multiple lineages (Gaskett 2011). In those cases most well studied to date (e.g. Chiloglottis in Australia and Ophyrs in Europe), chemical rather than morphological mimicry is of most importance for pollinator attraction. For example, pollinators of Chiloglottis are readily attracted to and attempt to copulate with plastic beads spiked with semiochemicals that match the sex pheromone (e.g. Schiestl et al. 2003; Peakall et al. 2010). Notwithstanding the importance of chemicals in long-range pollinator attraction (Schiestl et al. 2003; Streinzer, Paulus & Spaethe 2009), many sexually deceptive species show remarkable morphological specialization with a trend towards insectiform labella and reduction in the remaining tepals (Gaskett 2011). This suggests that morphology may be critical for assisting in the location of the flower, sustaining pollinator attraction at close range (e.g. Streinzer et al. 2010) and correct positioning of the pollinator to achieve pollination (Ascensão et al. 2005).

One consequence of convergent specialization resulting in pollinator sharing is the risk of reduced seed set through competition for visitation by pollinators, pollen wastage and allocation of resources to unfit hybrid seed (Waser 1978; Mitchell et al. 2009; Muchhala & Thomson 2012). Given the strong attraction of sexually deceived pollinators to the flowers they pollinate (e.g. Stoutamire 1983; Peakall 1990), in the absence of differential pollen placement, there is likely to be interspecific pollen transfer if two sexually deceptive orchids share the same pollinator in sympatry. An additional possible outcome is extensive introgression, particularly given the tendency for weak post-zygotic barriers in orchid genera characterized by strong pre-mating barriers (Cozzolino & Scopece 2008). Accordingly, there appear to be very few cases of pollinator sharing by sympatric sexually deceptive orchids. In Ophrys, most known examples exhibit morphological isolation (Paulus & Gack 1990; Cortis et al. 2009; though see Gögler et al. 2009), while in Cryptostylis, large differences in chromosome number between sympatric species probably enforce strong post-zygotic isolation (Peakall & James 1989; Dawson, Molloy & Beuzenberg 2007; Gaskett 2011).

An unusual case of pollinator sharing is believed to occur in south-western Australia, where the distantly related Caladenia pectinata R.S. Rogers and Drakaea livida J. Drumm. (subtribes Caladeniinae and Drakaeinae, Orchidaceae) were reported to be pollinated via sexual deception of Zaspilothynnus nigripes Guérin (Thynnidae) (Hopper & Brown 2007; Phillips et al. 2009). The courtship behaviour of Z. nigripes is typical of thynnine wasps, with the flightless female releasing a pheromone to attract mates. The successful male will grasp the calling female and fly off in copula (Ridsdill-Smith 1970).

Although both orchid genera place pollen on the wasp in the same position (Hopper & Brown 2007; Phillips et al. 2009), there are pronounced differences in size and morphology between the two orchid genera with important consequences for the behaviour of the pollinator on the flower. In Drakaea, when a wasp attempts to fly off with the hinged insectiform labellum, momentum swings the pollinator upside down and onto the column (Stoutamire 1975; Peakall 1990; Fig. 1). Attempted copulation with the semiochemical-producing labellum is frequent, but not essential for pollination (Peakall 1990). Many sexually deceptive Caladenia possess prominent linear lanceolate tepals terminating in clubs that are a source of the chemical attractant, along with the labellum in some species (Stoutamire 1983; Peakall & Beattie 1996; Phillips et al. 2009). For pollination to occur, the wasp must crawl to the base of the labellum, where it is brought into contact with the column (Stoutamire 1983). Caladenia appears to be a less refined system than Drakaea because pollinators are known to attempt to copulate or fly off with the odour-producing clubs (Phillips et al. 2009), thereby missing the column where pollination occurs.

Figure 1.

Sexual deception of Zaspilothynnus nigripes by Drakaea livida and Caladenia pectinata. In D. livida, semiochemicals are released from the labellum, while in C. pectinata, they are released from the sepaline clubs (L = labellum, Co = column, Cl = clubs). (a): Drakaea livida; (b): Caladenia pectinata, (c): D. livida being pollinated by Znigripes (photograph by Suzi Bond) (d): Z. nigripes in copula (photograph by Keith Smith), where the flightless female is carried to a food source; (e): C. pectinata with clubs pinned to back of the column; (f): C. pectinata with tepals pinned behind the labellum. Labellum width in C. pectinata is approximately 15 mm, and width in D. livida is approximately 7 mm.

Pollinator sharing in D. livida and C. pectinata provides the opportunity to investigate the consequences of differences in floral display and site of chemical attractant for pollinator behaviour. Given the prevalence of cryptic species in other genera of related thynnine wasps (Peakall et al. 2010; Griffith et al. 2011), we combined field experiments and molecular genetic analysis to confirm pollinator sharing. Using floral manipulations and field pollination data, we tested the prediction that in C. pectinata, the production of semiochemicals from sepaline clubs in concert with prominent floral display would reduce pollinator contact with the column leading to lower reproductive success than D. livida. Finally, we quantify the extent of spatial overlap between the two species to determine the ecological differences enabling species co-occurrence.

Materials and methods

Confirmation of pollinator sharing

Caladenia pectinata and D. livida show extensive overlap in flowering time, with both species reaching peak flowering in late September/October. Baiting for pollinators using picked flowers was undertaken across the distribution of D. livida and C. pectinata in suitable habitat for at least one of the orchid species. Two orchid flowers of both species were presented simultaneously at each site. Pollinator response to bait flowers is highest in the first minutes of presentation (Peakall 1990), so baiting was undertaken for six periods of 2 min at each site. All pollination studies took place between 10:00 am and 4:00 pm on sunny days with maximum temperatures exceeding 20 °C. Observations of pollinator behaviour were made to confirm that Z. nigripes is capable of removing and depositing pollinia in both species of orchid.

To confirm that the same individual wasp will respond to both D. livida and C. pectinata, a mark–recapture study was undertaken in remnant bushland adjacent to Ruabon Nature Reserve (33° 38′ S 115° 29′ E) in October 2010. A total of 254 wasps were marked using paint pens over 4 days, following their capture at artificially presented D. livida flowers. Subsequently, baiting was undertaken with C. pectinata on a single day, and captured wasps were examined for the presence of paint marks.

DNA sequencing of the CO1 region was used to test whether all pollinators represent a single species, with other species of Zaspilothynnus included as outgroups (Appendix S1). The CO1 region has previously been shown to resolve species level variation in thynnine wasps (Mant et al. 2002; Griffith et al. 2011). Laboratory and statistical procedures followed Griffith et al. (2011) with minor modifications. Briefly, statistical parsimony analysis was undertaken in the software TCS (Clement, Posada & Crandall 2000) to define groups of similar haplotypes. This method of analysis has been shown to yield haplotype groups that closely correspond with species defined by non-DNA evidence (Hart & Sunday 2007). This analysis was supported by a phylogeny generated using maximum-likelihood analysis in RaxML with 100 bootstrap replicates (Stamatakis, Hoover & Rougemont 2008).

Establishing the source of the semiochemicals

A pilot study showed that floral dissection did not reduce the attractiveness of orchids (Appendix S2). Floral dissections were undertaken to determine the site of release of the semiochemicals that attract pollinators in C. pectinata and D. livida. In C. pectinata, four treatments were used: labellum only, column only, all three clubs of the sepals and ‘declubbed’ perianth parts (the petals and the sepals with the clubs removed). For each treatment, the floral parts were pinned to the top of a wooden skewer using a pin that terminated in a black bead. Initial trials suggested that only the clubs were attractive, so sequential choice trials were undertaken. The labellum, column and floral display were offered simultaneously but arranged randomly in a row for 2 min, while the clubs remained sealed in a box. After 2 min, the clubs were introduced and placed in one of the four positions. For D. livida, a similar approach was used where the column and the tepals were presented for 2 min, before the labellum was also introduced for a further 2 min. In both experiments, all Z. nigripes alighting on either the floral parts or the bead were recorded. For six specimens of each species, five trials were conducted giving a total of 30 trials per species.

Pollinator behaviour in response to floral morphology

To compare the effectiveness of converting pollinator attraction into contact with the column, observations of pollinator behaviour on intact flowers of D. livida and C. pectinata were undertaken. For C. pectinata, two manipulations were performed to test whether scent production by the clubs and/or the presence of floral display reduced pollinator contact with the labellum and column. In the first manipulation, the clubs were cut from the sepals and pinned flush against the back of the column (Fig. 1). In this manipulation, the visual display was the same, but the scent was now associated with the labellum. In the second manipulation, the petals and sepals were folded back behind the labellum and the clubs pinned close to the back of the column (Fig. 1). Here, the visual display of the petals and sepals was removed, reducing overall size of the floral presentation, and the scent was again associated with the labellum. As such, observations were made for a total of four treatments: intact D. livida, intact C. pectinata, C. pectinata with the clubs removed and pinned behind the labellum and C. pectinata with the petals and sepals pinned behind the labellum. Statistical comparisons of wasp behaviour were undertaken between: (i) intact C. pectinata and intact D. livida, (ii) intact C. pectinata and C. pectinata with the clubs removed and pinned behind the labellum and (iii) intact C. pectinata and C. pectinata with the petals and sepals pinned behind the labellum.

Pollinator observations associated with the treatments described above were conducted over 4 days with morning trials run between 10:00 am and 12:00 pm and afternoon trials run between 12:30 pm and 2:30 pm. Each treatment was observed for 20 min each morning and afternoon, giving a total of eight observation periods per treatment. Observations were made of a single fresh flower for each observation period. The order of treatments was rotated so that each treatment was observed first, second, third and fourth both in the morning and afternoon. The orchid was placed at a random position within the bushland and switched to another random position after 10 min. Trials were split over four locations due to the constraints of limited suitable weather for wasp activity over the flowering period (Appendix S3).

Wasps that are attracted to experimentally presented flowers display either repeated circling in flight or a characteristic zigzag flight as they track the pheromone plume (Stoutamire 1983; Peakall 1990). For every wasp exhibiting one of these behaviours, it was recorded if they alighted on the flower, where they first landed, if they contacted the labellum, if they flipped the labellum hinge (Drakaea only), if they contacted the column and if they attempted copulation and where. The distinction between contact with the labellum and contact with the column was made because contact with the labellum means the mimicry has been sufficient to induce landing, but contact with the column is needed for pollinia transfer. For these trials, pollinia were manually removed from the orchid to avoid pollination of the experimental flowers, which may lead to a change in odour production and the attractiveness of the flower. Therefore, we could not quantify the proportion of contacts with the column that lead to pollinia removal. Differences in wasp behaviour between treatments were tested for via G-tests in the software GenAlEx 6.5 (Peakall & Smouse 2006, 2012). To test whether differences in pollinator behaviour have an impact on levels of reproduction, pollination rate was observed at eight populations of C. pectinata and 19 populations of D. livida. Pollination rate was expressed as pollination events per flower, noting that D. livida is exclusively single flowered and C. pectinata primarily single flowered (Hopper & Brown 2001, 2007). The variance in fruit set was significantly different between the two species (Levene's test, < 0.001), so a nonparametric Mann–Whitney U-test in JMP 9.0.0 (Dicke & Baldwin 2010; SAS Institute Inc 2010) was used to test for a difference in pollination rate.

The potential cost of pollinator sharing

To test whether there is a cost to orchids through wasted allocation of resources to intergeneric pollen transfer, artificial crosses between D. livida and C. pectinata were performed. Pollen was taken from six C. pectinata and deposited onto the stigma of six D. livida and vice versa. As a control, 10 C. pectinata and 10 D. livida were pollinated by hand using conspecific pollen from the same site. After 4 weeks, flowering stems with a capsule were collected, matured in water and the dehisced seed collected. Viability was determined by the presence or absence of embryos. Due to the similarity of the pollinia of D. livida and C. pectinata (Hopper & Brown 2001, 2007), they cannot be reliably distinguished on pollinators or stigmas. This prevented a direct assessment of intergeneric pollen transfer and whether or not the same individual wasps were visiting both species under natural conditions.

A survey of D. livida populations and a surrounding radius of 50 m was undertaken to quantify the proportion of populations at which C. pectinata co-occurred. To test whether D. livida and C. pectinata tended to occupy sites with different edaphic conditions, soil type and topography were recorded for each site. Geographic location was also recorded for subsequent analysis of the bioclimatic range occupied by both of these species. This data were supplemented using herbarium records and their associated habitat information from the Western Australian Herbarium (PERTH Perth, WA, Australia).

Bioclimatic ranges were calculated by ANUCLIM V6.1 (Xu & Hutchinson 2011) from the values of selected bioclimatic variables at the known locations of each species. Following McKenny et al. (2007) and Manning et al. (2010), six key bioclimatic variables were chosen to quantify heat and moisture, two important climatic gradients for plants (e.g. Shao & Halpin 1995; Stephenson 1998). The selected bioclimatic variables were annual mean temperature, minimum temperature of the coldest month, maximum temperature of the warmest month, annual precipitation, precipitation in the coldest quarter and precipitation in the warmest quarter. These were calculated from elevation-dependent climate surfaces fitted by the ANUSPLIN package (Hutchinson 1995, 2004) to monthly mean climate data for the period 1976–2005. For both species, we calculated the predicted range and increasingly restricted core climatic regions based on the 2.5–97.5 percentile, the 5–95 percentile and the 10–90 percentile areas. For each class, we calculated the range area, the area of the overlap between species and the percentage of the ranges that overlap between species. Distribution areas were calculated based on the Lambert Conformal Conic projection with a central meridian of 120° longitude to minimize the geographic distortion of the study area.

Results

Confirmation of pollinator sharing

Based on morphological identification, Zaspilothynnus nigripes was the only pollinator observed for both D. livida and C. pectinata. This species was observed to deposit and remove pollinia of both species by carrying pollinia on the same central region of the thorax. The only other floral visitor was one individual of the butterfly Heteronympha merope Fabricius (Nymphalidae) observed perching on and probing at the surface of the labellum of C. pectinata, but without contacting the column.

Six wasps marked at D. livida flowers were recaptured at C. pectinata bait flowers. The recapture rate of marked wasps at C. pectinata (2.8%) was similar to that of D. livida (3.9%) during the marking phase. Genetic analysis also confirmed the sharing of pollinators between the two orchid species. Sequencing of the CO1 region revealed 29 different DNA haplotyes within Z. nigripes, five of which were carried by wasps responding to both orchid species. Statistical parsimony analyses revealed two major haplotype groupings within Z. nigripes (Fig. 2), which was supported by the maximum-likelihood analysis. While Z. nigripes was monophyletic, there was 100% bootstrap support for one of the major haplotype groups (Fig. 2). Both major clades of Z. nigripes contained multiple wasps that have been collected from both C. pectinata and D. livida.

Figure 2.

A phylogeny of Zaspilothynnus nigripes and related species using a maximum-likelihood analysis of the CO1 region. Haplotype networks from an analysis in TCS are demarked for Znigripes (HN = haplotype network), showing support for the clades within this taxon. CP = wasp collected from Caladenia pectinata; DL = wasp collected from Drakaea livida. Numbers at the nodes are bootstrap values.

Establishing the source of the semiochemicals

In the floral dissection experiments, D. livida wasps were exclusively attracted to the labellum, with an average of 28 ± 10.6 (SE) individuals alighting on the labellum per trial. By contrast, in C. pectinata, the sepal clubs were the most attractive floral part, with an average of 19.7 ± 3.1 (SE) wasps alighting per trial. On six occasions, Z. nigripes approached within 30 cm of the labellum, three of which were approached when the clubs were being presented.

Pollinator behaviour in response to morphology

The total number of wasps responding in each experimental treatment was 313 to intact D. livida (252 alighting), 233 to intact C. pectinata (140 alighting), 213 to C. pectinata with the clubs removed and pinned behind the labellum (118 alighting) and 234 in C. pectinata with the petals and sepals pinned behind the labellum (124 alighting).

Pronounced differences were observed in the behaviour of Z. nigripes attracted to D. livida and C. pectinata (Table 1). A higher proportion of wasps alighted on D. livida than C. pectinata (80.5% vs. 60.1%, < 0.001). Furthermore, those wasps alighting on D. livida were much more likely to initially land on the labellum (98.8% vs. 66.4%, < 0.001), contact the labellum (99.6% vs. 86.4%, < 0.001), contact the column (42.5% vs. 3.6%, < 0.001) and attempt copulation (34.5% vs. 9.3%, < 0.001) than wasps alighting on C. pectinata. As such, wasps visiting D. livida displayed stronger attraction to the flower and greater efficiency at converting initial contact into potential pollination.

Table 1. Behaviour of sexually deceived Zaspilothynnus nigripes at Drakaea livida and Caladenia pectinata
  D. livida C. pectinata C. pectinata C. pectinata
  1. Percentage values in the lower section of the table are for those animals that alighted on the flower. Labellum landing and club landing refers to the point that the wasp landed upon initial arrival at the flower. Labellum contact refers to a touch at some point during the visit. For D. livida, significant differences are given for G-tests with intact C. pectinata. For C. pectinata (clubs removed), significant differences are given for G-tests with intact C. pectinata. For C. pectinata (tepals pinned behind labellum), significant differences are given for G-tests with intact C. pectinata. ***< 0.001, **< 0.01, *< 0.05.

ManipulationIntactIntactClubs removed, pinned to back of columnPetals and sepals pinned behind labellum
N 313233213234
Alight (%)80.5***60.155.4 n.s.53.0 n.s.
Labellum landing (%)98.8***66.478.0*64.5 n.s.
Club landing (%)29.39.3***34.7 n.s.
Labellum contact (%)99.6***86.489.0 n.s.74.2*
Hinge flip (%)81.3
Column contact (%)42.5***3.69.3 n.s.4 n.s.
Copulation (%)34.5***9.39.3 n.s.3.2*
Where copulatedLabellum onlyClubs only8 clubs only, 1 labellum and clubs, 1 labellum and 1 stemClubs only

When the clubs were removed from C. pectinata and pinned to the back of the column, the number of wasps initially alighting on the clubs dropped from 29.3% to 9.3% (≤ 0.001). When the entire sepals and petals were pinned behind the labellum, the values changed very little when compared with the intact flower (Table 1). The percentage of wasps alighting on the flower, contacting the column and attempting copulation changed very little between treatments. Across all C. pectinata treatments and a total of 680 observations, there were only two instances of wasps attempting to copulate with the labellum, compared with 27 instances with the clubs. Of the 21 visitors that made contact with the column, only two of these attempted copulation with the flower, demonstrating that attempted copulation is not necessary for pollination. For D. livida, 87 of 313 wasps attracted to the flower attempted copulation, always with the labellum. The pollination rate of D. livida (55.5 ± 7.6 SE, = 19) was significantly higher than that of C. pectinata (12.3 ± 5.6, = 8, = 0.01) suggesting that differences in pollinator behaviour may have consequences for plant reproduction.

Potential cost of pollinator sharing

Capsule formation was observed following artificial cross-pollination in both directions [five of six (D. livida maternal); four of six (C. pectinata maternal)]. However, on all occasions, seeds lacked a pro-embryo and were deemed non-viable. When conspecific pollen was deposited on stigmas of both D. livida and C. pectinata, all plants produced capsules with an average of 98.0 ± 0.6% and 81.2 ± 7.0% of embryos being filled, respectively.

A survey of 50 sites where the distribution of D. livida and C. pectinata overlap revealed no locations where the two species co-occurred at the scale of 50 m. Edaphic variables recorded from survey sites and herbarium records showed that D. livida favours grey sandy sites and rarely occurred in seasonally (winter) damp sites (Table 2). Alternatively, C. pectinata occurred in a range of soil types and prefers seasonally damp habitats. Overall, 90% of D. livida populations occurred in well-drained, grey sand, while for C. pectinata, this value was only 8%.

Table 2. Edaphic and climatic overlap in the habitat preferences of Caladenia pectinata with Drakaea livida
  C. pectinata D. livida
  1. Percentages for the edaphic variables are the percentage of surveyed sites with that characteristic. Climatic categories include the full climate range and percentiles of the full range. C. pectinata,= 52; D. livida,= 72.

Edaphic niche
% seasonally damp727
% grey sand2392
% well-drained grey sand890
Climatic niche
Distribution size (km2)101,04485,960
Range overlap (%)66.756.8
Overlap 2.5–97.5%56.252.8
Overlap 5–95%32.735.5
Overlap 10–90%18.716.4

A total of 49 populations of C. pectinata and 63 populations of D. livida were used in the climate analysis. Modelling of the climatic niches of C. pectinata and D. livida showed that while there is broad overlap between the two species, their core climatic ranges show relatively little overlap (Table 2, Appendix S4). The predicted overlap between the climatic range of C. pectinata and D. livida was 66.7% and 56.8%, respectively, while the overlap of their core range was just 18.7% and 16.4%, respectively.

Discussion

Floral morphology and pollinator behaviour

We have confirmed the first known case of sharing of a single sexually deceived pollinator species by members of two distantly related plant genera. While C. pectinata demonstrates that the point of semiochemical release does not have to match the morphological mimic of the female to achieve pollination, the prediction that production of floral odour by the clubs would reduce pollination success was supported. Further, C. pectinata has a low pollination rate [12.3 ± 5.6 (SE)] when compared with closely related food-deceptive species (Caladenia longicauda; 42.6 ± 9.2, C. speciosa; 40.0 ± 6.9, C. serotina; 50.2 ± 15.7; Phillips et al. 2009). Given the comparatively low fruit set in C. pectinata and reliance on a single pollinator species, there may be strong ongoing selection favouring floral traits that increase the rate at which attraction of Z. nigripes leads to pollination.

The contrasting floral form of the two species creates the opportunity to address the question of how changes in morphology might increase the efficiency of pollinator behaviour. The prediction that production of the scent from the labellum leads to greater attraction to the labellum, and more frequent contact with the column was evaluated by pinning the clubs to the back of the column. This manipulation increased the percentage of pollinators alighting directly on the labellum, suggesting that production of the scent from the labellum may increase pollinator efficiency on the flower creating a pathway by which selection could favour morphological reduction. This hypothesis could be further tested using comparative behavioural studies with Caladenia that produce chemical attractant from the labellum as well as the clubs, such as C. tentaculata (Peakall & Beattie 1996).

For all treatments of C. pectinata, pollinators frequently contacted the labellum at some point during the visit. However, the key inefficiency is the small percentage of wasps contacting the labellum that go on to make contact with the column. While pinning clubs behind the labellum increased contact with the labellum, it did not increase copulation or contact with the column. Visual inspection of the two orchid species reveals pronounced textural differences between the labella, potentially contributing to different levels of sexual attraction to the flower. As seen in other species, correct pollinator positioning may be strongly influenced by tactile flower features (Goyret & Raguso 2006; Whitney et al. 2009). Alternatively, quantitative or qualitative differences in pheromone production between the two orchid species could affect the level of sexual attraction to the labellum. This hypothesis could be evaluated by concealing a labellum of D. livida under that of C. pectinata and comparing the response of the pollinators.

One of the key morphological differences between D. livida and C. pectinata is the presence of the hinged labellum in D. livida. While hinged labella are present in multiple Australian genera including Arthrochilus, Caleana, Paracaleana, Spiculaea and some Caladenia (Brown et al. 2008), the mechanisms favouring the evolution of this feature remain unresolved. The hinged labellum in D. livida coupled with the wasp attempting to fly off with the female (or pseudo female) in copula places the wasp consistently in close proximity to the column compared with C. pectinata. This suggests that the hinged labellum could have evolved under selection to increase pollinator efficiency on the flower.

Floral display

While floral odour is important in attracting sexually deceived pollinators over longer distances, floral morphology and colouration may be critical for short-range location of the flower. Despite differences in floral size, shape and colour, pollinators could be responding to the same visual signals in both orchid species. For example, Australian members of the genus Cryptostylis are strikingly different to the human eye. However, all are pollinated by sexual deception of the ichneumonid Lissopimpla excelsa (Coleman 1928; Gaskett 2011) and use a similar colour scheme when floral spectral reflectance is plotted based on hymenopteran vision models (Gaskett & Herbenstein 2010).

The visitation of pollinators to two plant species with very different floral morphology suggests that some floral traits can vary extensively without inhibiting pollinator attraction. The prominent floral display of C. pectinata is of particular interest in this regard through the potential to detract from the insectiform appearance of the mimic. However, artificial reduction in this floral display did not lead to an increase in pollinators alighting on the labellum. Given that for hymenopterans, red colouration without UV is relatively difficult to detect against green backgrounds compared with other colours (Chittka et al. 1994), the red/green colouration of the tepals of C. pectinata may minimize the distraction of the wasp away from the labellum and the adjacent column. Alternatively, the perianth parts could provide contrast with the labellum to aid location of the flower at short range as seen in some Ophrys (Streinzer et al. 2010). Drakaea and Caladenia occupy different microhabitats in the environment suggesting that the optimal colouration to facilitate the landing of pollinators may differ between the genera.

Despite strong selection for increased efficiency, developmental constraints may have prevented C. pectinata from evolving the same level of floral specialization as that seen in D. livida. For example, the production of semiochemicals in the sepal clubs may prevent the evolution of the highly reduced and efficient flower seen in D. livida, where only the scent-producing labellum is needed for pollinator attraction. Further, food deception is the ancestral state in Caladenia (Kores et al. 2001) suggesting that species such as C. pectinata are secondarily adapted to sexual deception and may still retain traits associated with their food-deceptive ancestry.

While the flower of C. pectinata is much less efficient at converting wasp attraction into pollination, there may also be advantages to the less insectiform appearance. Caladenia pectinata may maintain sufficient floral display to also lure nectar-foraging insects. A mixed food and sexual deception strategy could prove advantageous in habitats/regions where the sexually deceived pollinator is absent or scarce. Food-foraging insects are occasionally observed visiting and pollinating other sexually deceptive orchids (Steiner, Whitehead & Johnson 1994; Dickson & Petit 2006), although they have not been observed to achieve pollination in C. pectinata. A common feature of food-deceptive Caladenia is production of a floral scent detectable to humans (Hopper & Brown 2001). Interestingly, floral scent can be detected from some but not all individuals of several sexually deceptive Caladenia including C. pectinata (R. Phillips & R. Peakall unpublished data) providing further support that a level of food deception may also be operating.

Implications for the study of floral odour

Studies of the pheromone chemistry of D. livida and C. pectinata are now needed to better understand the chemical basis of this convergence. The higher rate of attempted copulation with the flower in D. livida and the greater total number of wasps attracted suggest that there may be qualitative or quantitative chemical differences between these species. Recent studies of the floral odour of D. livida have shown that a blend of pyrazines appears to be involved in pollinator attraction (Bohman et al. 2012a,b). Therefore, as in Ophrys, it may not be essential to completely replicate the odour produced by the female (e.g. Vereecken & Schiestl 2008). If so, this may facilitate pollinator sharing between Drakaea and Caladenia. This scenario contrasts with a well-studied case of pollinator sharing in the closely related genus Chiloglottis, where the same single compound (chiloglottone 1) facilitates sharing of the same species of sexually deceived thynnine wasp among two largely allopatric orchids (Schiestl et al. 2003; Schiestl & Peakall 2005; Franke et al. 2009; Peakall et al. 2010). Studies of the pheromone chemistry of Z. nigripes are also needed to resolve whether the chemical attractant is identical to the sex pheromone as in Chilogottis trapeziformis (Schiestl et al. 2003) or if analogues are used to lure pollinators.

Ecological mechanism of pollinator sharing

While their distribution and climatic range show extensive overlap, the differences in edaphic requirements suggest that the potential cost of sharing the same pollinator species is not realized. Evidence from mark–recapture studies of thynnines and surveys of D. livida and C. pectinata suggest that the two species will very rarely share pollinator individuals. Studies of the movements of thynnines have all recorded mean recapture distances of < 50 m, with occasional movementots up to several hundred metres (Ridsdill-Smith 1970; Peakall 1990; Peakall & Beattie 1996; Whitehead & Peakall 2012). Surveys of 50 orchid populations failed to detect any co-occurrence at the scale of 50 m, with the nearest populations being 3 km apart. While orchid populations may occur within the flight range of pollinators when suitable habitats are adjacent, current evidence suggests that this rarely occurs. Staining of pollinia (Peakall 1989) would prove a useful technique to test for interspecific pollen movement between orchid species when they closely occur.

The regional sympatry but small-scale allopatry exhibited by D. livida and C. pectinata could have arisen from natural selection against co-occurrence. Alternatively, the low frequency of co-occurrence could have arisen from phylogenetic conservatism of habitat preferences. Of the species closely related to C. pectinata, all of them have different habitat preferences to Drakaea and rarely co-occur with them (Hopper & Brown 2001, 2007; Hoffman & Brown 2011). In this scenario, C. pectinata may be physiologically unable to co-occur in the same habitat as Drakaea, and consequently, the present-day patterns may merely reflect evolutionary history not selection favouring different habitats.

The climatic and habitat data support the conclusion of Pearson & Dawson (2003) that while climatic requirements may determine distribution at regional scales, topographic and edaphic conditions will be critical at finer scales. While D. livida is confined to well-drained grey sandy soils, C. pectinata only occurs in grey sands when they are in relatively low parts of the landscape. Further, C. pectinata shows a preference for soil types with a greater ability to retain water (Bettenay 1984), suggesting that soil moisture is an important determinant of the difference in habitat preferences. At the habitat–patch scale, differences in microhabitat preferences place a further limit on co-occurrence between D. livida and C. pectinata. Drakaea prefer low organic matter soils and, based on seed baiting studies, germinate more frequently in these soils (Phillips et al. 2011). On the other hand, species closely related to C. pectinata have been shown to prefer high organic carbon microsites (Batty et al. 2001). As such, even within habitats, D. livida will be limited to more open, sparsely vegetated areas, while C. pectinata will tend to occur in areas with more dense vegetation and leaf litter.

Conclusions

Here, we report a remarkable case of convergence, with the diminutive and warty D. livida attracting the same sexually deceived pollinator as the robust and colourful C. pectinata. However, comparative pollinator observations have shown that the characteristics of these flowers lead to very different pollinator behaviours with different consequences for plant reproductive success. Floral manipulations demonstrated the role of both floral morphology and site of scent release for determining pollinator efficiency. Using a multidisciplinary approach of chemical, visual and morphological investigations, this system will provide a unique opportunity to establish the traits required for the evolution of sexual mimicry and their consequences for reproductive success. Further, it may also yield insight into the function of less morphologically specialized sexual deception systems and potentially the evolution of the strategy.

Acknowledgements

Initial funding was provided by grants to R.D.P. by the Australian Orchid Foundation and the Holsworth Wildlife Research Endowment. The project was completed with the support of an ARC linkage project to R.P., K.W.D. and M.F.H. (LP098338) and an ARC linkage project to R.P. and K.W.D. (LP110100408). During the initial stages of this project, R.D.P. was supported by an Australian Postgraduate Award. We thank Keith Smith, Myles Menz and Suzi Bond for assistance with field work, Graham Brown for providing initial identifications of pollinators, Chris Hayes for undertaking the DNA sequencing and the researchers who contributed vouchers to the Western Australian Herbarium, in particular Andrew Brown, Stephen Hopper and Garry Brockman.

Ancillary