Batesian mimicry in the nonrewarding saprophytic orchid Danxiaorchis yangii

Abstract Batesian mimicry, a type of deceptive pollination, is a complicated strategy used by nonrewarding plants to attract pollinators, but some hypotheses concerning this have not been systematically verified. In order to show in detail a case of Batesian mimicry on saprophytic orchid Danxiaorchis yangii, the ecological relationship between Danxiaorchis yangii, Lysimachia alfredi and Dufourea spp. was explored. Lysimachia alfredi could provide a reward to Dufourea sp., whereas Danxiaorchis yangii not. The floral morphology and geographical distribution of these two plants were highly overlapping, and the fruit set rate of Danxiaorchis yangii was significantly positively correlated with the number of nearby L. alfredi individuals. In a glass cylinder experiment, Danxiaorchis yangii and L. alfredi attracted Dufourea spp. through visual signals, but the insect could not distinguish between flowers of the two plants before landing on flowers. The ultraviolet reflection spectra of flowers between the two plant species were highly similar. In the hexagonal color models constructed according to the visual characteristics of bees, the flower color signals of these two plant species highly overlap, indicating that the visual signals of the flowers of the two plants to the pollinator were greatly similar. All of these results provided evidence that Danxiaorchis yangii simulated the visual signals of L. alfredi through Batesian mimicry, thereby deceptively attracting Dufourea spp.


LUO et aL.
Batesian mimicry among animals, as both systems involve a signal receiver (pollinator or predator) that mistakes the imitator for a model. Simulation of rewarding plants can therefore be termed Batesian mimicry as well (Schiestl & Cozzolino, 2008). This phenomenon differs, however, between the two types of organisms. In animals, Batesian mimicry functions to repel signal receivers, whereas it functions in plants to attract (Newman et al., 2012;Whitehead et al., 2019).
Batesian mimicry in plant pollination is usually derived from generalized food deception, which mainly attracts insects, such as bees, beetles, small flies, and other generalist pollinators (Bae et al., 2019;Pellissier et al., 2010). Generalized food deception occurs when plants provide false food signals, such as spurs without nectar or fake pollen, to trick insects into foraging on flowers, thereby achieving pollination according to insect foraging behaviors (Nepi et al., 2018, Walsh & Michaels, 2017, Internicola et al., 2008. Most of the species pollinated by bees in Cypripedium belong to this type. Among them, there were as many as 14 species of flower-visiting insects in C. plectrochilum, including bumblebees, small bees, hoverflies, ants and butterflies, but only Lasioglossum spp. matches the functional structure of the orchids and is an effective pollinator (Li et al., 2008). Most cases of generalized food deception have no specific mimicry objects, that is, mimicry models. The targets of deception are "naive" insects with no learning experience (Gould & Gould, 1982;Lunau & Wester, 2017). If insects have developed a higher discriminating ability to reduce cheating, however, food-based mimics change accordingly to match the signals of the models. In such systems of food-deceptive mimicry, the visual signal is the most important trait for attracting pollinators. Studies to date have shown that the most important feature is color, followed by inflorescence architecture and flower shape, with the least important feature being scent chemistry (Fantinato et al., 2017;Galizia et al., 2005).
Batesian mimicry in plant pollination generally has the following properties: (a) the mimic and model share the same pollinator (Johnson, 2000); (b) geographical distribution and floral morphology overlap among mimic and model (Wüster et al., 2004) (c) the number of mimicked individuals is lower than that of the model (Schiestl, 2005); (d) the mimic and model release similar signals indistinguishable to the pollinator (Jersáková et al., 2016); and (e) the model can act as a "magnetic species," thereby increasing the mimic's pollinator visitation frequency and fruit set rate. Among these criteria, the fourth and fifth ones have rarely been met in published studies of Batesian mimicry in pollination (Cheney, 2010;de Jager et al., 2016).
Batesian mimicry is the least documented deceptive pollination strategy in orchids and is even controversial in the zoology literature. Few studies have tested most of the above-mentioned Bates mimicry features, let alone tested all features (Schaefer & Ruxton, 2009, O'Hanlon et al., 2014, Schiestl, 2005. Some early studies only looked for an increase in the reproductive success of a species when the species was growing with its assumed model (Johnson, 1994, Johnson, 2000, Gigord et al., 2002. Danxiaorchis yangii, is a recently described species, and also a fully mycoheterotrophic, extremely endangered orchid with a narrow distribution (Yang et al., 2017). In our early field observations, we found that Danxiaorchis yangii might deceptively attract Dufourea spp. to pollinate, while Dufourea spp. can also pollinate L.
alfredi and get rewarding. In addition, the two plant species bloomed at the same time and were distributed together. Therefore, in this case study, we analyze the potential Batesian mimicry from multiple perspectives such as pollination behavior, geographic distribution of plant species, effect of L. alfredi on fruit setting rate of Danxiaorchis yangii, and insect visual models. The research results could show a detailed case of Batesian mimicry and provide scientific guidance for resource conservation of Danxiaorchis yangii.

| Floral morphology investigation
We recorded the floral morphology of Danxiaorchis yangii and L. alfredi, including flowering duration of a flower, individual, and patch, according to the standard of Dafni (Dafni, 1992). Floral organ morphological data were obtained using a Vernier caliper. On a sunny day, at 8, 10, 12, 14, and 16 o'clock, 30 flowers of Danxiaorchis yangii were selected. The flowers were dissected to determine whether there were nectar, oil, related secretion and storage tissues inside.
Glucose test paper was used to detect whether there were traces of nectar in the flower. period, the inflorescences were bagged in the daytime and exposed at night. The selected flowers were examined for pollen deposition and removal each day at 8:00.

| Breeding system
To prevent insects or alien pollinia from entering flowers prior to study, 280 flowers of Danxiaorchis yangii were selected and bagged into seven separate groups over 2 consecutive years (2018-2019).
Three to four days before flowering, one of seven unique treatments was applied to each group: (a) continual bagging (bag kept on until flowers faded), (b) emasculation + bagging (pollinium removal followed by bagging); (c) artificial self-pollination; (d) artificial geitonogamy; (e) artificial xenogamy; (f) gynostemium removal + bagging; and (g) natural (no bagging). The fruit set rate was counted after flowering, and SPSS software was used to analyze differences among pollination methods.

| Spatial distribution of Danxiaorchis yangii and L. alfredi
Ten patches in the distribution range of Danxiaorchis yangii were selected, and each circular patch had a radius of 500 m. To analyze the relationship between the natural distribution of Danxiaorchis yangii and L. alfredi, the geographical coordinates of all individuals of these two species in each patch were recorded and used to generate a scatter plot.

| Glass cylinder experiment
To investigate whether plants attract pollinators through olfactory or visual signals, an experiment was performed with three types of glass cylinders (Milet-Pinheiro et al., 2015): (a) a black cylinder with holes, so that odor was emitted without any visual signal (O cylinder); (b) a sealed, transparent cylinder, so that odor was not emitted but visual signal was present (V cylinder); and (c) a transparent cylinder with holes, so that odor was emitted and a visual signal was present

| Analysis of flower color
10 flowers that just opened from the two plant species were selected for reflectance spectrum detection by following the method used in some classic papers (Dalrymple et al., 2015;Dyer et al., 2012;Shrestha et al., 2019), while the leaf of L. alfredi was served as control (Danxiaorchis yangii has no leaf). During operation, reflectance spectra for wavelengths from 300 to 700 nm were recorded using Ocean Optics spectrophotometer (Dunedin, FL, USA) using a PX-2 pulsed xenon light sources attached with SPECTRASUITE software to PC. We used UV-reflecting white standard and black standard to calibrate the spectrophotometer. The reflectance of the flower organs that could be seen from the front of the flower were measured, including the sepals, petals, labellums and gynostemium of Danxiaorchis yangii, as well as petals, pistils, and stamens of L. alfredi.
A reflectivity curve based on the raw spectral reflectance data was drawn using OriginPro 6.1 SR1 software.
To represent flower color perception by bees, a hexagon color space model of hymenopteran vision was employed. Our current model was based on an integration range of 300-650 nm and trichromatic photoreceptors with spectral sensitivity peaks at 350 nm (UV), 440 nm (blue: B) and 540 nm (green: G), using a vitamin A1 visual template (Stavenga et al., 1993), which closely matches trichromatic photoreceptors in many bee species (Briscoe & Chittka, 2001). Several previous studies have used this hexagonal model of hymenopteran color vision (Bischoff et al., 2013;Dyer et al., 2012;Ohashi et al., 2015;Shrestha et al., 2014). Hymenopteran color vision is phylogenetically conserved (Briscoe & Chittka, 2001), thus in the absence of receptor sensitivity values of studies species, it is also appropriate to use the general hymenopteran color model .

| Data analysis
Routine statistical analyses were performed in IBM SPSS (version 19), while linear regression analysis between the fruit setting rate of Danxiaorchis yangii and the individual number of L. alfredi, as well as analysis of difference in visit frequency, was carried out by GraphPad Prism 8. Pollination by Dufourea spp. was observed primarily from 11:00-15:00. When Dufourea spp. visited L. alfredi flowers, it landed directly on the purple blotch, that is, the intersection of petals and stamens.

| Dufourea spp. pollinated Danxiaorchis yangii and L. alfredi
In this location, the insect first touched the anther with its upper jaw, and its three pairs of feet also moved frequently as it attempted to scrape pollen off the stamen, resulting in a large amount of pollen on its thorax and appendages. During this activity, the insect continuously changed its position, which allowed it to cover most of the stamens. After approximately 30-40 s, the insect was covered with pollen and left. When the pollinator visited Danxiaorchis yangii, it landed on the labellum and climbed onto the purple blotch located at the junction of the labellum and gynostemium. In this location, the appendage of the labellum is raised, forming a narrow channel with the gynostemium, just enough to accommodate the pollinator's head. The pollinator therefore extended its head into the channel and continued to move around the labellum after exiting. At the same time, the insect also continuously moved its upper jaw and foot and left approximately 2-5 s later. While entering or exiting, an insect could bring in or take out the pollinium of Danxiaorchis yangii.
The morphological characteristics of Dufourea spp., Danxiaorchis yangii, and L. alfredi were shown in Table 1 and Table S3.
Anatomical observation and glucose test paper showed that no nectar, oil, and related secretion and storage tissues were found in the flowers of Danxiaorchis yangii. In addition, as an orchid plant, the pollen of Danxiaorchis yangii is clumped and inedible for bees. On the other hand, L. alfredi has been reported to provide edible pollen for bees (including Dufourea spp.) (Müller, 2018;Simpson & Neff, 1983).

| Pollination restrictions caused a low seed set rate
In the breeding system experiment, the fruit set rate under both "emasculation + bagging" and "gynostemium removal + bagging" treatments was zero, which indicated that no apomixis occurs in Danxiaorchis yangii and that the sperm-egg combination was necessary for seed formation. Flowers that were bagged during the entire period did not bear fruit, thus indicating the absence of automatic selfing in Danxiaorchis yangii and hence implying that seed production must depend on pollinators. The fruit set rate following artificial pollination, including self-pollination, geitonogamy, and xenogamy, was above 90%. In contrast, the natural fruit set rate was only 23%, which suggested that Danxiaorchis yangii had severe pollination restrictions (Table 2). to Danxiaorchis yangii individuals increased to 4:1, however, the fruit set rate of Danxiaorchis yangii tended to stabilize or even slightly decrease. A possible reason for this latter result was that Danxiaorchis yangii was more readily ignored by pollinators because of the abundance of rewarding L. alfredi (Table 4). and O-LA. The visitation frequency in the first four groups was significantly higher than that of the last five groups, with no significant differences detected among the first four groups or last five groups. According to these results, Danxiaorchis yangii and L. alfredi attracted pollinators by visual signals, and olfactory signals basically did not work. In addition, no significant difference was observed in the frequency of pollinator visits to these two plant species (Table 5).

| Breeding system and population characteristics
Our breeding system experiment revealed that Danxiaorchis yangii could not produce seed without pollination, and no apomixis or selfcompatibility existed. The results indicated that pollinators played a vital role in the successful reproduction of Danxiaorchis yangii under natural conditions. The fruit set rate under artificial pollination was much higher than the natural fruit set rate, which indicated a pollination limitation was present in Danxiaorchis yangii (Table 2). The widely accepted explanation for the deceptive pollination strategy chosen by certain plants is that deceptive pollination could promotes outcrossing and effective pollen output (Jersáková et al., 2006;Kudo, 2006). Danxiaorchis yangii has rhizomes and exhibits a scattered distribution pattern in the wild. In a patch, most clumps of Danxiaorchis yangii comprise rhizomatous clones. As the genet grows, its flowers generally become gradually surrounded by other flowers from the same individual (Cronberg et al., 2006;Luo et al., 2013). If Danxiaorchis yangii can reward pollinators, pollinators will stay in this small area for a long time and travel between flowers. These flowers are likely to be asexual offspring from the same parent, which indirectly reduces the possibility of genetic recombination and impedes plant evolution (Eckert, 2000;Routley et al., 2004). On the other hand, because pollinators could not obtain a reward from Danxiaorchis yangii, they left the patch after visiting one to five flowers, thereby reducing the proportion of inbreeding and increasing the chance of outcrossing.

| Pollinator specificity
Because approximately 60% of orchid taxa have only one species of pollinator, the family Orchidaceae exhibits a very high degree of pollinator uniqueness (Cozzolino & Widmer, 2005). In some plants with narrow, scattered distributions, inbreeding readily occurs. The evolutionary development of more specialized pollination systems can effectively reduce the risk of inbreeding (Brosi & Briggs, 2013).
In this study, Dufourea spp. was the only pollinator of Danxiaorchis yangii, and the orchid was found in seven intermittently distributed natural patches, but they were at least several kilometers away from each other. This fragmented distribution largely hinders gene flow between patches. Specific pollinators should significantly improve pollen flow between intermittently distributed patches. In addition, perennial plants with strong asexual reproduction are thought to generally prefer specialized pollination systems, as they can use asexual propagation to overcome the loss of seed yield caused by specialized pollination. The adoption of a specialized pollination system can also reduce a plant's investment in sexual reproduction; this is particularly important for fully mycoheterotrophic orchids, as their nutrient sources are already scarce (Nattero et al., 2010;Shuttleworth & Johnson, 2009).

| Mechanism of Danxiaorchis yangii pollination
Lysimachia alfredi and Danxiaorchis yangii shared a pollinator species, the former could provide edible pollen, whereas the latter not. Pollinators therefore stay longer on L. alfredi compared with Danxiaorchis yangii.
The floral morphology and geographical distribution of these two plant species were highly coincident, which raised the question: was there a special relationship between the two plant species?
In a 3-year quadrat survey, a significant positive correlation was found between the fruit set rate of Danxiaorchis yangii and the number of L. alfredi inflorescences (Table 3) alfredi flowers was highly coincident. In addition, the color model analysis revealed that the flower colors of these two plants were highly similar. No significant differences in floral reflectance spectra have been found between Batesian mimics and their model plants (Kraemer et al., 2015;Peter & Johnson, 2008). The importance of visual signals in food-deceptive systems has been proven by experiments. Manipulation of visual signal factors, such as UV reflectance, flower color, and flower and inflorescence shapes, has been shown to increase or decrease the frequency of pollinator visits (Newman et al., 2012;Peter & Johnson, 2008, Jersáková et al., 2012.
Dufourea spp., a pollen-feeding insect, has been reported to collect pollen from members of the genus Lysimachia. According to our observations, this pollinator behaved similarly on flowers of the two studied plant species. During flower visiting, the target was a purple blotch, and the insect used its upper jaw or foot to explore and collect pollen. Because only one of these two plant species can provide pollen, however, the duration of the pollinator's visit to their flowers was quite different.
Several deceptive orchids, such as Viola aethnensis (Ging.) Strobl and Cephalanthera rubra (L.) Rich., are assumed to be examples of Batesian mimics, as their fruit set rate is significantly improved when rewarding plants with similar colors are present (Nilsson, 1983, Dafni & Ivri, 1981b, Pellegrino et al., 2008, Dafni & Ivri, 1981a. These cited studies did not determine, however, the relative preferences of pollinators toward those orchids and their putative models; instead, they only demonstrated that having visual signals similar to those of familiar rewarding plants provides more opportunities for the deceptive orchids to be visited. Contrary to one criterion of Batesian mimicry, however, the deceptive orchids are more widely distributed than their putative models. Those putative Batesian mimicry representatives thus attract general pollinators by means of generalized food deception, not Batesian mimicry. In the present study, we have presented the first detailed example of a pollination strategy based on Batesian mimicry in fully mycoheterotrophic orchids, a finding achieved by observations of pollination behavior, glass cylinder experiments, analysis of the relationship between Danxiaorchis yangii fruit set rate and the number of L. alfredi individuals, and comparative analysis of visual signals from the two plant species.

CO N FLI C T S O F I NTE R E S T
None declared.