Florian P. Schiestl, Geobotanical Institute ETH Zürich, Zollikerstrasse 107, Zürich CH-8008, Switzerland. Tel.: +41 1 6327429; e-mail: firstname.lastname@example.org
Sexually deceptive orchids mimic sex pheromones and appearance of female insects to attract males, which pollinate the flowers in an attempted mating. This study examines the effects of pollinator mate choice on orchid floral evolution using the Thynnine wasp Neozeleboria cryptoides (Smith) (Hymenoptera: Tiphiidae), which pollinates the sexually deceptive orchid Chiloglottis trapeziformis Fitzg. (i) When male wasps were given the choice between two female dummies of different sizes and identical amount of synthetic pheromone, they preferentially attempted to copulate with medium-sized dummies over small dummies. (ii) When given the choice between two dummies of identical size but different amounts of pheromone, males preferred the larger amount of pheromone. Larger amounts of pheromone generally attracted more males than smaller amounts. (iii) Orchid flower labella, which mimic a female body, were significantly longer and broader than female wasp bodies, and the flowers also produced on average 10 times more ‘pheromone’ than females. The evolution and maintenance of these exaggerated mating signals is likely to be mediated by the male pollinator behaviour demonstrated here. (iv) When five dummies were offered simultaneously in a 10 cm circular array, males rarely attempted copulation on more than one dummy during a single visit. This behaviour may foster the evolution or maintenance of clonality in C. trapeziformis, as it will minimize pollen exchange within clones.
In sexually deceptive orchids, pollination is brought about by mating behaviour rather than food seeking behaviour of the pollinators. Mating behaviour and its elicitation are generally ‘hard wired’ and therefore less prone to modification by learning processes (Ayasse et al., 2001; but see Ayasse et al., 2000). Pollinators of most Australian sexually deceptive orchids belong to the Thynnine wasps (Hymenoptera: Tiphiidae). In these parasitoid wasps, females are wingless and ‘call’ for males with sex pheromones. Males then carry them, in copula, to a food source to feed (Given, 1954; Ridsdill Smith, 1970a; Alcock, 1981; Peakall, 1990). Males deposit females after feeding and copulation, presumably at a suitable oviposition site. Thynnine-wasp males thus invest a considerable amount of time and energy into mating and are therefore expected to be choosy with respect to available females (Trivers, 1972; Thornhill & Alcock, 1983; Alcock & Gwynne, 1987). Indeed, this seems to be the case. In choice experiments, Alcock & Gwynne (1987) found that males of the Thynnine wasp Megalothynnus klugii carried off larger females more often than smaller ones. If male Thynnine wasps also choose females according to the quality of specific mating signals, such choice behaviour should impose selection pressures on floral traits in orchids that mimic these visual and chemical signals, to maximize fruit production and pollen export. If floral traits under selection pressure are variable and heritable, one would expect the pollinator behaviour to influence the evolution of the flowers. However, the influence of this form of mate choice on floral evolution has not yet been investigated.
In this study, I use the plant–pollinator relationship between an Australian sexually deceptive orchid Chiloglottis trapeziformis Fitzg. and its Thynnine wasp pollinator Neozeleboria cryptoides (Smith) as a model system to investigate the effect of mate choice on floral evolution. The great advantage of this system is that the female sex pheromone, which consists of only one compound and is produced in identical form in the orchid flowers, has recently been identified (Schiestl et al., 2003). A synthetic analogue was found to be equally attractive to male wasps as the sex pheromone and the orchid floral odour. While this compound is both a pheromone (in the pollinator species) and an allomone (in the orchid), I will call it ‘the pheromone’ throughout the paper to avoid confusion. The availability of synthetic pheromone gives the opportunity to test the importance of different mating signals for mate choice behaviour in the male wasps. I used different amounts of synthetic pheromone applied on dummies of variable size to test the following questions:
1Do males choose wasp dummies according to size and amount of pheromone emitted?
2Do males attempt to copulate with more than one female/flower if many are available in the proximity?
3How do orchid flowers and female wasps differ in size and pheromone production?
Materials and methods
Insects and orchids
Males of N. cryptoides patrol open woodland in search for females in early spring. Females crawl on plant stems and other objects a few centimetres above ground where they call with sex pheromone for males, which pick up females and fly with them in copula to a food source, e.g. honeydew secretions of scale insects on Eucalyptus sp. trees (F. P. Schiestl & R. Peakall, unpublished data).
Chiloglottis trapeziformis is a colony-forming orchid, with clones known to exceed 3 m2 and consist of hundreds of individuals (Oakwood, 1990). While only a small proportion of plants flower in any given year, with a single flower borne per plant, some populations may contain tens to hundreds of flowers in a good season (Peakall et al., 1997).
As in sexually deceptive orchids in Australia more generally, pollination success in Chiloglottis is likely to be highly variable in time and space. For example, in the sexually deceptive Caladenia tentaculata, pollination success varied between years and sites from very low (<15%) to very high (>80%) (Peakall & Beattie, 1996). Peakall & Handel (1993) report pollination success of 41% for C. trilabra. For C. trapeziformis, Oakwood (1990) recorded only 4% pollination success. However, in the Canberra region where this study was performed, F. P. Shackleton & R. Peakall (2002, unpublished data) detected pollination success varying per site from 47 to 70% in 2002 (four sites, n flowers = 210). Besides a likely limitation of reproduction by pollinators in C. trapeziformis, plants, or clones, will, however, also be selected for maximum pollen export and therefore compete for pollinators.
Behavioural experiments with N. cryptoides males were conducted on an approximately 2 ha patch of open woodland in the Black Mountain Nature Reserve, Canberra, Australia. No orchids occurred in the immediate surroundings but wasps were common. Males of N. cryptoides patrolled in large numbers close to the ground. To study the mating behaviour of male wasps, black plastic beads were used as female-wasps models. Artificial beads treated with synthetic pheromone were found to be highly attractive (mean of up to 20 copulation attempts in a 3-min test period, Fig 2) and offer the advantage of controlled visual and chemical stimuli. Beads of elliptical shape were obtained in three sizes [small: 6 × 3 mm (approximately size of a female); medium: 8 × 4 mm (approximately size of orchid flower); and large: 10 × 6 mm], mounted on bent insect pins, and used as dummy ‘females’. Synthetic sex pheromone compound was dissolved in different concentrations in dichloromethane and 10 μL of a dilution was applied on a dummy, yielding total amounts of compound applied on the dummies from 0.2 pg (2 × 10−13 g) to 10 μg (1 × 10−5 g). Appropriate time was allowed for the solvent to evaporate, and thereafter the dummies were offered to males in the field. In the dual choice experiments, two dummies were spaced 3 cm apart and in the multiple choice experiments five dummies were grouped in a 10 cm circular array. Copulation attempts of males on the dummies during a 3 min period were recorded on a voice recorder. Three experiments were conducted.
Experiment 1 – Size preference
Dual choice tests comparing different sizes of dummies (small vs. medium and small vs. large) using the same amount of pheromone (20 ng) were conducted.
Experiment 2 – Pheromone-amount preference
Dual choice tests comparing different amounts of pheromone (0.1 vs. 1; 1 vs. 10; 10 vs. 100; 100 vs. 1000; 1000 vs. 10 000 ng) using small dummies were conducted. In each of these five test groups, the total number of copulation attempts was calculated. Additionally, small amounts of pheromone (0.2 pg–2 ng) were tested on a single dummy, to find the response threshold of the insects to the compound.
Experiment 3 – Switch from one dummy to the next
Tests with five small dummies, all with the same amount of pheromone (100 ng), offered simultaneously in a circular array, were carried out. Copulation attempts of males during 3 min were recorded and it was observed how often males attempted to copulate with multiple dummies (‘switched’) within the array during one visit of the experimental spot.
Sizes of females and orchids
Pairs of N. cryptoides were caught at food sources at the margin of open woodland in the Black Mountain Nature Reserve, Canberra, Australia in September 2001 and 2002 using hand nets. Pairs were gently separated, transported to the laboratory in a chilled box and killed by freezing. Females and males were weighed. Headwidth of males and females, and length and width of females were measured using a dissection microscope with an ocular grid.
Unpollinated C. trapeziformis plants were collected from a large population on Black Mountain, Canberra. Length and width of flower labella were measured in the same way as for female wasps.
As the amount of pheromone emitted by females or orchid flowers is too small to be detected with headspace sorption techniques (F. P. Schiestl, unpublished data), the amounts contained in female mandibular glands and flower labella were measured. The female mandibular gland has been found to be the source of the sex pheromone in N. cryptoides, and the flower labellum is the major source of the ‘pheromone’ in C. trapeziformis (Schiestl et al., 2003).
Heads of females were cut off, the mandibles spread, and the whole head extracted in dichloromethane for 24 h. Orchid labella were also extracted in dichloromethane for 24 h. After extractions, 100 ng of octadecane was added to each sample as internal standard, and 5 μL of each sample were injected ‘on-column’ into a gas chromatograph (GC; Varian 3400, Palo Alto, CA, USA). The GC was equipped with a deactivated guard column (5 m, 0.53 mm) connected to an EC-1000 analytical column (Alltech Inc., Deerfield, IL, USA; 30 m, 0.32 mm × 0.25 μm). Oven parameters were 40 °C (1 min) to 230 °C at 10°/min. Compounds were detected by an flame ionisation detector. Absolute amounts of pheromone were calculated from the relation of the pheromone peak area to the internal standard peak area.
To compare the sizes and amounts of pheromone of females and orchids, the Mann–Whitney U-test was used (Norušis, 1993). Correlation of male and female sizes was estimated by calculating Pearson's correlation coefficient. To compare behavioural responses in the dual choice tests (Experiments 1 and 2), the Wilcoxon sign rank test for matched pairs was used. To compare the total number of responses in each test group a Kruskal–Wallis test was used. U-tests were used for a posteriori multiple comparison, with the level of significance set to 0.005 through a Bonferroni correction (0.05 divided by the number of comparisons).
Experiment 1 – Size preference
When small and medium sized dummies were offered together to the males, medium sized dummies were significantly more attractive than small ones, as they elicited more than double the number of copulation attempts by males (Wilcoxon sign rank test for matched pairs: Z15 = −2.8, P < 0.01; Fig 1a). Therefore, males seem to assess size of potential females, and prefer to mate with larger individuals. However, when males had the choice between small and large dummies, no difference in attractiveness was found (Wilcoxon sign rank test for matched pairs: Z7 = −0.37, P = 0.71; Fig 1b). Objects excessively larger than females therefore seem to lose their superior attractiveness.
Experiment 2 – Pheromone-amount preference
When males were given the choice between two dummies with different amounts of pheromone, the larger amount was significantly more attractive. This preference for larger amounts was apparent through almost the whole range of amounts tested (Wilcoxon sign rank test for matched pairs: 1 ng vs. 10 ng: Z14 = −3.4, P < 0.01; 10 ng vs. 100 ng: Z24 = −3.87, P < 0.001; 100 ng vs. 1000 ng: Z14 = −3.3, P < 0.01; 1000 ng vs. 10 000 ng: Z11 = −2.63, P < 0.01; Fig 2). Only differences between 0.1 and 1 ng were not significant, suggesting a threshold effect (Z10 = −1.86, P = 0.063; Fig. 2). In another series of tests with very small amounts of pheromone, the threshold of pheromone needed to attract any males was 0.2 ng, as lower amounts did not elicit any copulation attempts or approaches (Table 1).
Table 1. Responses of Neozeleboria cryptoides males to small amounts of pheromone (n = number of trials).
0.2 pg (n = 1)
2 pg (n = 1)
0.02 ng (n = 3)
0.2 ng (n = 2)
2 ng (n = 2)
6.5 ± 0.5
4.5 ± 1.5
2.5 ± 0.5
3.0 ± 2.0
When pooled responses of each choice group were compared, it was evident that larger amounts of pheromone attracted a greater total number of males (Fig 3). Whereas two dummies with 1 and 0.1 ng attracted a mean of 1.1 males within the 3-min test period, two dummies with 100 and 1000 ng attracted a mean of 29.2 males (Mann–Whitney U-test: U10,14 = 1, P < 0.001). The following groups were not significantly different from each other and hence pooled together: 1 + 10 ng vs. 10 + 100 ng; 100 + 1000 ng vs. 1000 + 10 000 ng. However, there was a significant difference between the remaining three groups [group 1: 0.1 + 1 ng; group 2: (1 + 10) + (10 + 100) ng; group 3: (100 + 1000) +(1000 + 10 000) ng; Kruskal–Wallis test: = 42.65, P < 0.001]. In a posteriori multiple comparison, all groups showed significant differences (Mann–Whitney U-tests: group 1 vs. 2: U10,39 = 35, P < 0.001; group 1 vs. 3: U10,26 = 1, P < 0.001; group 2 vs. 3: U39,26 = 149, P < 0.001).
Experiment 3 – Switch from one dummy to the next
During one visit to the spot with five dummies males rarely switched from one dummy to another after attempting copulation. During the 14 trials using five dummies at one spot with 100 ng pheromone each, 119 males attempted copulation on these dummies (mean: 8.5 ± 2.8 males per trial). Of these males, 114 flew off after one copulation attempt, and only five individuals (4.2%; mean: 0.36 ± 0.17 per trial) attempted copulation on a second dummy immediately after the first attempt. None of the males attempted to copulate with a third dummy.
Sizes of female wasps and orchids
Neozeleboria cryptoides showed a remarkable sex dimorphism in terms of size with males being four times heavier than females (Table 2). There was a significant positive correlation between headwidth and body mass in males and females (males: r9 = 0.97, P < 0.001; females: r9 = 0.96, P < 0.001), demonstrating that headwidth is a good indicator of body mass in both sexes. However, there was no significant correlation between body mass or headwidth of males and females caught as a pair (weight: r9 = 0.1, P = 0.779; headwidth: r9 = 0.24, P = 0.504).
Table 2. Mean (±SE) size of males and females of Neozeleboria cryptoides (n = 10).
Body mass (mg)
16.03 ± 1.22
3.92 ± 0.53
1.98 ± 0.05
0.85 ± 0.04
Orchid labella were significantly longer and broader than wasp females (Table 3). Furthermore, ratios of length/breadth were significantly different in female wasps and orchids (Table 3) with the orchid labella being relatively broader than the females.
Table 3. Mean (±SE) sizes (length of body, breadth of abdomen) and amounts of pheromone in females of Neozeleboria cryptoides and orchid labella (n = 10, if not otherwise indicated).
‡n = 4.
*P < 0.001.
**P < 0.01.
6.0 ± 0.3
8.1 ± 0.2
1.2 ± 0.1
7.0 ± 0.1
4.04 ± 0.3
1.16 ± 0.01
22.05 ± 4.38‡ (min: 16.7; max: 35.1)
218.64 ± 53.17 (min: 21; max: 464.9)
Amounts of sex pheromone
The sex pheromone compound was detected in four of the 10 examined female wasps; the individuals that did not contain detectable amounts were excluded from analysis. There was no significant correlation between pheromone amount and body length of females (r2 = −0.99, P = 0.091). All examined orchid labella contained the pheromone in detectable amounts, and the quantities of compound were significantly larger than in female wasps (Table 3).
Male choice behaviour in wasps
My dual choice experiments with dummies and synthetic sex pheromone clearly show the existence of male choice behaviour in N. cryptoides. Specifically, both the amount of pheromone and the size of a potential mate appear to be important. Males preferred larger dummies to a certain extent. Dummies approximately the size of an orchid labellum (medium) were clearly more attractive than dummies of female size (small). Larger dummies, which exceed the size of an orchid labellum, were, however, not preferred over small ones. These larger sized stimuli may no longer be recognized as female bodies by males and hence lose their attractiveness. Size is generally believed to be an important factor in mate choice, as it may be correlated with female fecundity (Thornhill & Alcock, 1983; Hedrick & Temeles, 1989; van Dongen et al., 1998). In another Thynnine species, larvae have also been found to grow more slowly and to remain smaller when affected by diseases (Ridsdill Smith, 1970b). This suggests that if larval and adult size of females are correlated, a small female may be avoided as a potentially parasitized mate.
Mate choice may, however, be dependent on additional factors or limited by natural conditions, as suggested by the lack of size correlation within copulating pairs of N. cryptoides caught at food sources. A similar result was obtained by Alcock & Gwynne (1987) in another species of Thynnine wasp. Possibly, females are not choosy at all and/or males are choosy only when they encounter two or more females simultaneously, which may happen rarely, as the operational sex ratio in N. cryptoides and other Thynnine species is strongly biased towards males (Alcock & Gwynne, 1987; Peakall, 1990; F. P. Schiestl, unpublished data). However, the orchids usually flower in colonies with many of them open at a time (Oakwood, 1990; F. P. Schiestl, unpublished data). Therefore, males visiting orchids will often be in a situation where they have to choose between two or more flowers.
In addition to size, the amount of pheromone emitted by a female is an important factor affecting mate choice in N. cryptoides males. In the behavioural tests, dummies with larger amounts of pheromone were always more attractive, except for the choice between 0.1 and 1 ng, where only few individual males responded. This preference for larger quantities of pheromone was obvious even with amounts that were over 100 times greater than the typical content of a female gland. In a choice situation, males therefore seem to prefer females that emit larger amounts of pheromone. Sex pheromones have often been assumed to be an important factor in mate choice in insects (Andersson, 1994; Svensson, 1996; Ayasse et al., 2003), but quantitative or qualitative traits that influence choice have rarely been identified. In some Hymenopteran insects, individual odour bouquets are thought to be important signatures that may be used in mate choice to avoid inbreeding (summarized in Ayasse et al., 2001). In a sandfly species, Jones & Hamilton (1998) suggested a link between the quantities of a sex pheromone compound in males and their mating success. In the arctiid moth Utetheisa ornatrix, females choose males according to the amount of a pheromone compound, which is correlated with male size (Iyengar et al., 2001). In my study, no significant correlation was found between amount of pheromone and body size in N. cryptoides females, which may be due to small sample size and/or varying age of individuals, as pheromone reservoirs may deplete with age. However, the demonstrated influence of pheromone amount and body size on choice behaviour by males may translate to choice for females and their mimics, the flowers of sexually deceptive orchids.
Floral evolution in the orchid
The data presented here suggest that both the amount of pheromone produced by a flower and the size of the flower labellum have evolved under selection imposed by the preferences of the pollinators. The flower labella, which mimic the body of a female, are significantly longer and broader than the actual female insects. As pollinators prefer large females, orchids with a larger flower labellum may enjoy an increased likelihood of pollination, as pollinators will often have to choose among flowers. The optimal labellum size appears to be determined mostly by the size perception mechanism of the pollinators. In my experiments, only the dummies approximating labellum size (medium-sized dummies) were more attractive than small dummies, whereas dummies exceeding the size of an orchid labellum (large dummies) were no more attractive than small dummies. This suggests that floral size has reached an optimum maintained by stabilizing selection mediated by the pollinator. In another species of Chiloglottis, C. trilabra, it was shown that the pollinators also select for floral height, but the natural mean height of plants was, however, lower than the experimentally determined optimum for visitation (Peakall & Handel, 1993). Apart from visual cues, pollinators of sexually deceptive orchids may also, and especially so, select for olfactory cues as the odour of a flower is the most important cue to attract the pollinator (Kullenberg, 1961; Peakall, 1990).
As is the case for size, the amount of pheromone in flowers is significantly greater than in females. The threshold amount necessary to attract males ranges around 0.2 ng, which is about 1000 times less than the mean amount produced by flowers. Larger amounts of pheromone may, however, translate into a reproductive advantage for two reasons: (i) pollinators prefer larger amounts of pheromone and thus will, in a choice situation, choose the stronger smelling flower; (ii) larger amounts also attract more pollinators, probably because the odour plume is larger and/or travels a longer distance and may hence reach more males. If orchids bloom in an area with few or no males, males will be attracted only to flowers that emit amounts of pheromone that travel far enough to pull in the pollinators. It is, however, unknown if pollination success in Chiloglottis is linked to the amount of odour produced by a flower. My experiments show that the amounts produced by flowers are not at the upper limit of attractiveness: greater quantities attract even more males, and males still discriminate between larger and smaller amounts. Costs of synthesizing the ‘pheromone’ may, however, at some point outweigh the benefits of increasing the likelihood of pollination.
Floral odour may influence pollinator behaviours in multiple ways. In the European sexually deceptive orchid Ophrys sphegodes, a high degree of odour variability in flowers, both within and between plants, was interpreted as promoting visits of pollinators to more than one flower of a plant (Ayasse et al., 2000). Chiloglottis trapeziformis is a clonal, self-compatible plant (Peakall et al., 1997) and its pollinia do not have stipites, which in Ophrys prevent self-pollination within the first 2 min after withdrawal (Ayasse et al., 2000). Therefore, pollinators visiting more than one flower within a clone will lead to inbreeding. My experiments suggest that pollinators of C. trapeziformis will rarely visit more than one flower per patch, at least if the variability in amount of odour is low. Only about 4% of males attempted copulation with more than one of the five dummies close together during one visit. Other experiments have indicated that Thynnine males rarely pollinate multiple flowers within a patch (Peakall, 1990; Peakall & Beattie, 1996; Wong & Schiestl, 2002), suggesting that the apparent avoidance of multiple copulations within a patch of flower will promote outcrossing and longer distance pollen flow. This may be particularly important in clonal Chiloglottis where the risk of pollination within clones (self-fertilization) is high.
Body size and the amount of pheromone emitted are probably constrained to a certain extent in females of the pollinator species. In the Hymenoptera, size is often correlated with the amount of food intake of the larvae, and calling females generally emit only minute quantities of sex pheromones (Ayasse et al., 2001), which may be explained by biosynthetic costs and/or the possibility of exploitation of mating signals by parasites and predators (Zuk & Kolluru, 1998). Therefore, although males prefer larger amounts of pheromone, Fisher's runaway scenario for increasing the production of sex pheromone in females is unlikely. In plants, these constraints may not be as severe or may even be nonexistent, e.g. parasites of the pollinator species will certainly be of no threat for the orchid. Therefore, mimicry of mating signals, i.e. flower size and amount of pheromone, can evolve to an enhanced intensity in the plant (Ayasse et al., 2003). The data presented in my study suggest that male pollinator choice has likely imposed strong selection to maximize pollination success and pollen export by the evolution of exaggerated mating stimuli.
I sincerely thank C. Schulz and W. Francke (Hamburg) for synthesizing the pheromone compound used in this study. J. Mant (Sydney) and R. Peakall (Canberra) helped with field work and S. Armbruster (Trondheim), G. Bernasconi (Zürich), H. Dobson (Walla Walla), A. Leuchtmann (Zürich), J. Mant, R. Nyffeler (Zürich), R. Peakall and B. Wong (Canberra) commented on the manuscript. This study was financially supported by the Swiss Federal Institute of Technology, The Fonds zur Förderung der wissenschaftlichen Forschung Austria (J2016-BOT) and The American Orchid Society.