Specialization for Tachinid Fly Pollination in the Phenologically Divergent Varieties of the Orchid Neotinea ustulata

Despite increased focus on elucidating the various reproductive strategies employed by orchids, we still have only a rather limited understanding of deceptive pollination systems that are not bee- or wasp-mediated. In Europe, the orchid Neotinea ustulata has been known to consist of two phenologically divergent varieties, neither of which provide rewards to its pollinators. However, detailed studies of their reproductive biology have been lacking. Our study aimed to characterize and understand the floral traits (i.e., morphology, color, and scent chemistry) and reproductive biology of N. ustulata. We found that the two varieties differ in all their floral traits; furthermore, while Neotinea ustulata var. ustulata appears to be pollinated by both bees (e.g., Anthophora, Bombus) and flies (e.g., Dilophus, Tachina), var. aestivalis is pollinated almost entirely by flies (i.e., Nowickia, Tachina). Tachinids were also found to be much more effective than bees in removing pollinaria, and we show experimentally that they use the characteristic dark inflorescence top as a cue for approaching inflorescences. Our results thus suggest that while both N. ustulata varieties rely on tachinids for pollination, they differ in their degree of specialization. Further studies are, however, needed to fully understand the reproductive strategy of N. ustulata varieties.


INTRODUCTION
The orchid family is one of the largest plant families in the world, accounting for around 28,000 described species (Willis, 2017;Fay, 2018). At least one third of these species are deceptive, not providing their pollinators with a reward (Renner, 2005; but see Shrestha et al., 2020). Deceptive orchids employ various strategies to lure in their pollinators, which range from the advertisement of false nectar or pollen sources to the imitation of brood sites or mating partners (Jersáková et al., 2006;Johnson and Schiestl, 2016). Each type of deception is thought to be associated with a particular combination of floral traits, often influenced by their reliance on particular groups of pollinators (Johnson and Schiestl, 2016;Valenta et al., 2017). Conspicuous visual displays coupled with reduced floral scent emission and the attraction of male and female pollinators are traditionally considered an indication of food-deceptive pollination (Galizia et al., 2004;Jersáková et al., 2012Jersáková et al., , 2016Johnson and Schiestl, 2016). Less conspicuous visual displays, the exclusive attraction of either male or female insects and the emission of particular scent bouquets are, in contrast, associated with pollination by sexual deception or brood-site mimicry (Vereecken and Schiestl, 2009;Ayasse et al., 2011;Bohman et al., 2016;Martel et al., 2019; but see Streinzer et al., 2010). For many deceptive orchids the investigation of the floral traits employed can provide an accurate estimate of their pollination strategy; however, some orchid species may have complex strategies that are much more difficult to categorize (e.g., Vogel, 1972;Bino et al., 1982;Vöth, 1989;Valterová et al., 2007;Scopece et al., 2009). This seems also to be the case for the European orchid species, Neotinea ustulata (L.) R. M. Bateman, Pridgeon and M. W. Chase.
Neotinea ustulata, like up to 81% of orchid taxa occurring in Europe, Asia Minor and North Africa (Paulus, 2005;Delforge, 2006), is deceptive, i.e., does not offer any readily identifiable reward to potential floral visitors (Claessens and Kleynen, 2011). While most of these deceptive orchids rely on either bumblebees or solitary bees for reproduction, N. ustulata has been considered to be potentially fly-and beetle-pollinated (Vöth, 1984;van der Cingel, 1995). The evidence for N. ustulata's pollination system has, however, been mostly anecdotal. The identification of its reproductive strategy has been further complicated by the occurrence of two phenologically distinct varieties, the early flowering (from April to June, or even until July) Neotinea ustulata var. ustulata and the late flowering (from June to August) N. ustulata var. aestivalis (Kümpel) Tali, M. F. Fay and R. M. Bateman. The first recorded pollination event in N. ustulata var. ustulata probably dates back to Godfery (1933), who observed two Tachina (Echinomyia) aff. magnicornis (Zett.) flies, one of them with pollinaria attached to its proboscis, visiting flowers in a population from Eastern France. Further records stem from Vöth (1984), who observed several individuals of T. aff. magnicornis carrying pollinaria in Austria. While tachinid flies seem to be common visitors, other records include bees (Borsos, 1962), bumblebees (Paulus, 2005;Claessens and Kleynen, 2016), and empidid flies (Paulus, 2005;Claessens and Kleynen, 2016). Pollinator records are even more scarce for N. ustulata var. aestivalis, for which there have been only observations of beetles, namely Chrysanthia sp. (Oedemeridae) (Danesch and Danesch, 1962) and Vadonia (Leptura) livida (F.) (Cerambycidae) (Mrkvicka, 1991), carrying pollinaria on various parts of their head. These records are, however, insufficient to determine the reproductive system of N. ustulata as they only account for sporadic and isolated visitation events.
Throughout their range of distribution, the two N. ustulata varieties do not usually occur in sympatry ; var. ustulata being mostly found in dry meadows, whereas var. aestivalis in both dry and wet meadows (e.g., Molinetum vegetation) (Paulus, pers. obs.). Plants of var. aestivalis are also usually taller than those of var. ustulata (Haraštová-Sobotková et al., 2005). Despite these differences, the two varieties are genetically undifferentiated (Tali et al., 2006; but see Haraštová-Sobotková et al., 2005) and have the same chromosome number (Tali, 2004), which indicates no association of autopolyploidy with the phenological differences as reported in others plants (e.g., Simón-Porcar et al., 2017). The mechanisms driving their divergence, thus remains unknown. It is possible that both pollinator-and non-pollinator-mediated selection could have played a role in driving the phenological divergence among populations, leading to the evolution of these two varieties. However, testing the role of pollinator as selection mediators in N. ustulata requires firstly a better assessment of its pollinators and of the traits involved in pollinator attraction. To this date there are only very few assessments of floral traits of N. ustulata varieties (i.e., Haraštová-Sobotková et al., 2005;Schiestl and Cozzolino, 2008;D'auria et al., 2021). Inflorescences of both varieties have a similar color pattern (Paulus, 2005), both having a very dark brown-purple top (hereafter referred as "dark top"). This provides a strong contrast with the remaining part of the inflorescence (Godfery, 1933;Stace, 2010), and may serve as a signal for attracting pollinators; however, this hypothesis so far has not been tested. Perhaps the most unusual trait of N. ustulata is its floral scent. Fooddeceptive orchids rarely rely upon flower scent for pollinator attraction, this floral trait is, however, essential in sexually deceptive species (Ayasse et al., 2011;Bohman et al., 2016). Schiestl and Cozzolino (2008) were the first to show that the floral scent of N. ustulata contains alkenes, which, although uncommon in most flowers, are frequently associated with sexually deceptive ones (Ayasse et al., 2011;Bohman et al., 2016); in particular, N. ustulata flowers produce high relative quantities of 11-tricosene. The role of this specific compound in pollinator attraction is so far unknown, but other tricosene isomers are known to act as sex pheromones in the house fly Musca domestica (Carlson et al., 1971;Rogoff et al., 1973) or aggregation pheromones in Drosophila (Hedlund et al., 1996). At the moment, we do not know if alkenes could be found in both varieties or these are restricted to just one of them. The particular combination of visual signals associated with the chemical composition of the flower scent raises the interesting possibility that N. ustulata may rely on intricate signals to attract its pollinators.
In order to understand the reproductive strategy and, eventually, the evolution of N. ustulata, we need to accurately quantify the differences in floral traits between its two varieties as well as identify their reproductive biology. Therefore, our study aims are to compare N. ustulata varieties in terms of (i) morphological, visual and chemical floral traits, (ii) pollination success and patterns of fruit set, and (iii) pollinator spectra. Furthermore, we aimed to (iv) assess efficiency in pollinaria removal among the principal pollinator guilds, and (v) identify the role of the inflorescence dark top in the attraction of predominant pollinators.

Study Species
Neotinea ustulata is a perennial tuberous herb, which is typically found on extensively used meadows . Each plant of this species bears a single inflorescence, which has a burnt appearance when its flowers are in bud; the species is therefore known as the burnt-tip orchid. Neotinea ustulata flowers have three semi-fused sepals, which form a hood; the outer side of the sepals (especially in buds) displays a dark reddish to blackish color (Figure 1). The lip is white with dark red spots and is deeply divided into three parts (Figure 1). The flowers have a nectarless spur, which is blunt. Neotinea ustulata occurs in central and southern Europe with populations reaching Spain and Greece in the south, England and southern Sweden in the north, and as far east as the Caucasus and Ural Mountains Delforge, 2006). Despite its broad geographical range, the species is vulnerable to extinction and the number of populations have been greatly reduced during the last century (Neiland, 2001;Tali et al., 2004;Baumann et al., 2005;Jacquemyn et al., 2005).
The species consists of two varieties, which, while phenologically distinct, do not differ strongly in their vegetative traits (Kümpel and Mrkvicka, 1990;Wucherpfennig, 1992;Haraštová-Sobotková et al., 2005): the early flowering N. ustulata var. ustulata and the late flowering var. aestivalis. Neotinea ustulata var. ustulata is 10-35 cm tall, rarely larger, and its flowering period ranges from April till June in most parts of Europe Baumann et al., 2005), but from June to July at high altitudes in the Alps (Paulus, pers. obs.). The dense, narrow, up to 8 cm long inflorescence is composed on an average of 30 flowers Baumann et al., 2005;Haraštová-Sobotková et al., 2005; Figure 1). Neotinea ustulata var. aestivalis has in turn a somewhat taller stem, of 33 cm long in average, (rarely up to 80 cm;Haraštová-Sobotková et al., 2005) with longer and narrower leaves, and sharply pointed inflorescences (especially in young stage) (Figure 1). The inflorescences have on average around 40 flowers (Jensen and Pedersen, 1999). Plants of this variety flower from the end of June or July till mid-August Baumann et al., 2005;Paulus, pers. obs.). In order to confirm the identity of the two varieties, detailed digital photographs of individuals from our sampling localities were compared against material deposited in diverse herbaria (e.g., CB, BRNM, OHN, PR, PRC, UPS, VN, and W).

Study Sites
Field observations and experiments were carried out in Austria, the Czech Republic, the Slovak Republic and Sweden (Figure 2), during May-June (for details see Supplementary Table 1). In total, thirteen populations of var. ustulata and ten populations of var. aestivalis were sampled to analyze different aspects of their reproductive biology (Figure 2; for population details see Supplementary Table 1).

Morphology
We sampled inflorescences of N. ustulata var. ustulata and var. aestivalis from two populations each (i.e., 30, 30, 23, and 15 inflorescences from U8, U9, A2, and A4, respectively, see Supplementary Table 1 for population codes). From the middle of each inflorescence, we picked a single, fully opened and yet unpollinated flower. Flowers were placed in 70% ethanol (Sigma Aldrich, Munich, Germany) and stored until further usage. Individual flowers were mounted on a styrofoam base and then photographed from three different angles with the aid of a Dino-Lite digital microscope (AnMo Electronics Corporation, Taiwan). Size calibration and distance measurements were performed with the in-house software DinoCapture (AnMo Electronics Corporation, Taiwan). Eight floral traits (i.e., hood opening width, hood opening height, spur opening width, spur width at 3/4 spur length, width at spur bottom, spur length, distance from column to the spur opening, and lip length) were measured.  To identify which morphological traits, if any, are responsible for differences among varieties of N. ustulata, we performed a principal component analysis (PCA) using the factoextra package (Kassambara and Mundt, 2020) of the R software (R version 3.6.1; R Core Team., 2017). Morphological data were scaled and normalized prior to conducting the PCA. To explore differences in morphology between the two varieties, we performed a permutational multivariate analysis of variance (PERMANOVA) with the raw morphological data. We constructed a matrix based on Euclidean distances to run the PERMANOVA with 10,000 permutations, considering "variety" as fixed factor and "population" as nested factor in "variety." PERMANOVA was performed using the vegan package (Oksanen et al., 2019) of R.

Color
The spectral reflectance was measured for 44 flowers (one flower per plant) from two populations per N. ustulata variety (i.e., 14, 10, 10, and 10 flowers from U8, U9, A2, and A4, respectively). We specifically measured the reflectance of the dark top, outer sepals and lips. All floral reflectance measurements were performed using an Ocean optics UV-VIS Spectrometer JAZ and an Ocean Optics DT-mini light source (200-1,100 nm; Dunedin, FL, United States), calibrated with a white reflection standard (WS-1-SL, Ocean Optics). Readings were taken through a fiber-optic reflection probe (UV/VIS 400 µm) held at 45 • ca. 5 mm from the surface of the object. The visual similarity between varieties, as perceived by their visitors, was determined by plotting their reflectance spectra as loci in bee-and fly-specific color spaces, assuming adaptation to foliage background (Chittka and Kevan, 2005). Selection of these two insect models was based on our pollinator list (see section "Results"). To model bee color vision, we used the color hexagon (Chittka, 1992), based on the spectral sensitivities of Apis mellifera (L.). To model fly color vision, we used Vorobyev and Osorio, 1998) receptor noise model. This method uses spectral sensitivities and the relative numbers of each fly photoreceptor type to calculate the perceived color contrast between objects. We did not use Troje's (1993) fly categorical color vision model because it was already shown to be unreliable when contrasted with experimental data (e.g., Jersáková et al., 2012;Renoult et al., 2014;An et al., 2018). Based on color loci, we also run a PERMANOVA, considering "variety" and "inflorescence part" as fixed factors and "population" as a nested factor in variety, to test for differences in spectral reflectance between varieties while considering population origin. The visual modeling and PERMANOVA were performed using the pavo (Maia et al., 2021) and vegan packages of R, respectively.

Floral Scent
Highly volatile compounds of the floral scents of both N. ustulata varieties were collected using dynamic headspace techniques (i.e., 10, 9, and 5 flowers from U8, U9, and A2, respectively), whereas semi-and low-volatile compounds were collected through solvent extracts (i.e., 10, 10, 11, and 10 flowers from U8, U9, A2, and A4, respectively). Samples were analyzed by a gas-chromatograph and gas chromatograph/mass spectrometer using a DB-5 non-polar column (for detailed description of the analytical techniques see Supplementary Methods 1). Chemical identification was carried out by comparing the mass spectra of pure compounds with the ones of commercial libraries (NIST, ADAMS) or of our own libraries, and the retention indexes of detected compounds with those from the literature (e.g., Adams, 2017). Mass spectra were processed using the Automated Mass Spectral Deconvolution and Identification System-AMDIS-software (National Institute of Standards and Technology, Gaithersburg, United States). Differences in floral scent bouquets of both varieties were analyzed by comparing the relative proportions of compounds identified in our two type of samples. We calculated the Bray-Curtis similarity index to determine semi-quantitative differences in odor bouquets. Then, we performed a PERMANOVA with 10,000 permutations to test for differences in scent bouquets between the two varieties while considering population origin. A nonmetric multidimensional scaling (NMDS) was performed to graphically display the (dis)similarities in floral scent patterns among samples. PERMANOVA and NMDS were performed using the vegan package and displayed graphically using the ggplot2 package (Wickham et al., 2020) of R.

Reproductive Success Patterns
During 6 years (i.e., 1978, 1981, 1983, 1984, 2000, and 2001), we examined seven populations of N. ustulata var. ustulata and four populations of var. aestivalis for fruit production (see Supplementary Table 1 for population details). We recorded the total number of flowers and position of fruits within inflorescences for each inflorescence sampled. The differences in the number of flowers, the number of fruits produced and the proportion of flowers that set fruits between the N. ustulata varieties were tested by means of generalized linear mixedmodels (GLMMs) with Poisson model distribution with log link function for number of flowers and fruits, and binomial model distribution with logit link function for the proportion of flowers that set fruits; "population" was set as a random factor.
In deceptive systems, it is expected to find position effects on fruit production, as first open flowers are more likely to be visited by yet unexperienced insects than late flowers; however, it is unknown if this pattern generalized and also happens in N. ustulata. Therefore, we checked whether the development of fruits is affected by the relative position of the flower in the inflorescence by dividing each inflorescence into three sections (i.e., bottom, middle, and top), calculated mean reproductive success for each section, and then performed a GLMM with a binomial model distribution and log link functions, and conducted Tukey post hoc tests to compare pairs of means; "plant individual" was set as a random factor. GLMMs and post hoc tests were carried out using the lme4 (Bates et al., 2019) and emmeans packages (Lenth et al., 2020), respectively, in R.

Flower Visitors, Pollinators, and Their Overall Efficiency
Insect visitors on N. ustulata flowers were recorded by video footages (73 and 37 h of footage on N. ustulata var. ustulata and var. aestivalis, respectively) and complemented by direct observations in our study sites (see Supplementary Table 1 for population details). Floral visitor identity and identification of visitors as pollinators were assessed based on both direct observations and video recordings; whereas frequency of visits and floral visitor behavior on the inflorescence were assessed based on the video footage only. Visitors were identified to the order and family level, and when possible up to genus/species level. To test for differences in the flower visitor composition, a G-test of independence comparing the frequency of species within orders between varieties was performed using the DescTools package (Signorell, 2020) of R. Representatives of each floral visiting species were collected using entomological nets in parallel to the video recordings in search of attached pollinaria and for insect identification. We also checked for presence and placement of pollinaria in insects feeding on rewarding flowers of plants co-occurring with N. ustulata, such as species of Aegopodium L. (Apiaceae), Crataegus L. (Rosaceae), Dorycnium Mill. (Fabaceae) and Euonymus L. (Celastraceae). To corroborate the pollinaria identity, they were compared with pollinaria directly collected from N. ustulata individuals. Insects were considered pollinators based on both the presence of attached pollinaria and evidence of pollinaria removal (assessed from the video footage or direct observations).
Furthermore, we also calculate male (i.e., pollinaria removed/total flowers × 100) and female (i.e., pollinia deposited/total flowers × 100) reproductive success, and pollen transfer efficiency (i.e., ratio of pollinia deposited to pollinaria removed × 100; Nunes et al., 2016), which can be linked to specialization. We performed a GLMM with Poisson model distribution and log link function to estimate the effects of variety on pollen transfer efficiency and with "population" set as a random factor. GLMM was carried out using the lme4 package of R.

Efficiency in Pollinaria Removal Between Pollinator Guilds
We compared the pollinaria removal efficiency between the tachinid fly Tachina fera (L.) and the bee Anthophora plumipes (Pallas), in order to identify whether pollinators of different guilds are equally efficient. Our set up was as follows: one inflorescence of N. ustulata var. ustulata with pollinaria present in all its open flowers was placed in a plastic container (100 × 50 mm i.d., Greiner Bio-One GmbH., Frickenhausen, Germany) in which a single individual of either T. fera (n = 6) or A. plumipes (n = 12) was introduced. We recorded the removal of pollinaria in up to three attempts (i.e., insertion of the proboscis in the flower) of each T. fera and A. plumipes. We performed a GLMM with binomial model distribution and logit link function to test for removal efficiency between pollinators. We also conducted Tukey post hoc tests to compare pairs of means. Our GLMM for repeated measures, compound symmetry was used as covariance pattern model, included "pollinator guild" and "attempt number" as fixed factors and "insect individual" as a random factor. GLMM and post hoc tests were performed using the lme4 and emmeans packages in R.

Role of Visual Traits in Pollination Success
In 1999 we sampled inflorescences from eight populations from Sweden and Czech Republic belonging to both varieties (see Supplementary Table 1 for population details). They were examined for inflorescence display and dark top display. Neotinea ustulata inflorescences have a cylindrical arrangement and, therefore, the inflorescence display was calculated as its two-dimensional projection (i.e., the area of a rectangle: length × width of inflorescence); while the dark top has a circular shape and, therefore, dark top display was approximated as the area of a circle (i.e., π × {width of dark top/2} 2 ). We also assessed the number of flowers with removed pollinaria, deposited pollinia and flowers where pollination events occur; they were then transformed to proportions by dividing them by the number of open flowers. We performed GLMMs with binomial model distribution and logit link function to estimate the effects of visual display on the probability of a flower to be pollinated. Our models included "inflorescence display, " "dark top display" and "variety" as fixed effects, and "population" as random factor. GLMMs were performed using the lme4 package in R.

Role of Dark Top in Fly Pollinator Attraction
We tested the importance of the dark top in the attraction of tachinids since it might be an attractant to flies (Godfery, 1933;van der Cingel, 1995). Tachina flies have been previously considered the main pollinators of N. ustulata (Vöth, 1984;Trunschke et al., 2017), and their importance as pollinators is supported by our own observations (see section "Results"). Experiments were carried out with both varieties. For var. ustulata, four inflorescence pairs were selected for experiments. For each trial, two inflorescences of equal height and similar phenological state were placed less than 10 cm apart. Each inflorescence pair was first observed for about 30 min, during which the number of tachinids attracted by each inflorescence was recorded. After the initial observations the dark top of one of the two inflorescences was arbitrarily cut off. The inflorescence pairs (i.e., one intact, one without the dark top) were then observed again for the same amount of time and the number of tachinids which landed and/or probed the flowers was recorded. In case of var. aestivalis, only two inflorescence pairs were selected, but the experimental setup was similar to that described for var. ustulata, with two exceptions: (i) For the second trial, the cut off dark top of var. aestivalis was placed below the selected plant, in such a way that it was hidden from the view to approaching flies. This setup ensured that olfactory cues released by the dark top were still present when the dark top was cut off. (ii) An additional trial was carried out, in which the dark top was reattached to the plant. Then, the inflorescence pairs were offered again to tachinids. This trial controlled for any effect of the damage (e.g., green leaf volatiles after plant tissue damage) to the inflorescence on their attractiveness to Tachina flies. We performed GLMMs with negative binomial model distribution and log link function to test for differences in the attractiveness of manipulated and non-manipulated inflorescences within each variety. Our GLMMs for repeated measures incorporated a compound symmetry covariance matrix included "treatment stage" and "manipulation condition" as fixed factors, "plant individual" as the subject for repeated measures before and after the manipulation, and "number of visits" as the response variable.
GLMMs were performed using the MASS package (Ripley et al., 2020) in R.

Morphology
The flowers of N. ustulata var. ustulata are larger in all measured traits in comparison to var. aestivalis (Table 1). In our PCA, the first two PCs accounted for 62.1% of the total variance (PC1: 50.5%, PC2: 11.6%). All the morphological floral traits loaded negatively on the PC1 (Supplementary Table 2), which separates var. ustulata (negative scores) from var. aestivalis (positive scores) (Figure 3). Indeed, the PERMANOVA analysis revealed that the floral morphology differed significantly between var. ustulata and var. aestivalis

Color
The dark top and the lightly colored lip plotted apart in both the fly and bee color models, indicating that these two parts of the flower appear highly contrasting, and are therefore easily discriminated by bees as well as flies. Loci of the individual color measurements of the dark top (n = 44) plotted, except by one case, in the UV-blue color category within the bee hexagon and also grouped together within the fly tetrahedron (Figure 4). Similarly, color loci of the lip (n = 44) clustered together within both the bee hexagon and the fly tetragon (Figure 4). The color loci of the hood (n = 44) were more broadly distributed, spanning the space from the dark top loci to the lip loci (Figure 4). In terms of color information, the dark top and hood may often be perceived as uncolored against the leaf background by bees based on color hexagon distances. The color loci of different inflorescence parts were different, independently of variety, as our PERMANOVA analysis confirmed [Pseudo-F (7, 131) = 52.64, p < 0.0001]. Furthermore, PERMANOVA analyses also indicated that there are statistical differences in the spectral reflectance between N. ustulata varieties [Pseudo-F (1, 131) = 9.28, p < 0.01] and within varieties among populations [Pseudo-F (3, 131) = 2.81, p < 0.05].

Flower Visitors, Pollinators, and Their Overall Efficiency
Inflorescences of both varieties were visited by a broad range of insect taxa within the Coleoptera, Diptera, Hymenoptera and Lepidoptera. Diptera and Hymenoptera were more frequently observed in our video recordings and together accounted for around 90% of the total insect visits (i.e., 415 insect visitors) to both N. ustulata varieties (Figure 7 and see Supplementary  Table 5 for details of insect identity and frequency). Based on the same video recordings, the frequencies of the attracted insect orders significantly differed between var. aestivalis and var. ustulata (G = 71.68, df = 3, p < 0.0001; Figure 7). Anthophora plumipes and Lasioglossum sp. bees were particularly frequent on N. ustulata var. ustulata, and T. aff. magnicornis flies were most abundant on var. aestivalis (Supplementary Table 5). Despite the high number of insect visits, only three Anthophora bees carrying pollinaria of N. ustulata var. ustulata and one T. aff. magnicornis removing pollinaria of var. aestivalis were observed in our videos. By direct observations, we recorded some Tachina individuals visiting the flowers and carrying pollinaria (Supplementary Table 6), thereby acting as pollinators of N. ustulata var. ustulata and var. aestivalis. While bees almost exclusively landed on the white part of the inflorescence moving upwards in their search for a reward, Tachina flies always settled briefly (for 1-4 s.) on the dark top of the inflorescence; then they searched for potential nectar in the uppermost freshly open flowers and introduced their head to reach the base of the flower spur with their body in downright position (Figure 8). The pollinaria then became attached on the ventral side of the lower part of the proboscis (Figure 8). Identification of tachinid specimens revealed that they belong to three morphologically similar species, Tachina fera, T. magnicornis and Nowickia ferox. Both taxa pollinate flowers of N. ustulata var. ustulata, while var. aestivalis was only observed to be pollinated by T. magnicornis and N. ferox (Supplementary Table 6). Beside tachinids, solitary bees and bumblebees acted as regular pollinators, whereas beetles and other flies were highly occasional (see Supplementary Table 6 and Supplementary Figure 2).

Efficiency in Pollinaria Removal Between Pollinator Guilds
While all tachinids but one, which probed the flower with the body in an upright position, removed pollinaria at their first attempt, some bees did not remove pollinaria even after three attempts. When testing the efficiency of pollinaria removal, we found that pollinator type had significant effects on pollinaria removal (χ 2 = 7.01, df = 1, p < 0.01; Figure 9) but no effects of number of attempts on pollinaria removal success (χ 2 = 2.34, df = 2, p = 0.31; Figure 9). Post hoc tests indicated that the efficiency of pollinaria removal was significantly higher within each attempt for T. fera compared to A. plumipes (z = −2.850, p ≤ 0.05; Figure 9).

Role of Dark Top in Fly Pollinator Attraction
In our experiment with manipulation of the inflorescence dark top, there were significant effects of the treatment stage (χ 2 = 5.68, df = 1, p < 0.05), the inflorescence type (i.e., selected for dark top excision or not; χ 2 = 10.29, df = 1, p < 0.01) and the interaction treatment stage × inflorescence type (χ 2 = 15.97, df = 1, p < 0.001) on the number of Tachina aff. fera visits to var. ustulata (Figure 10). In the case of var. aestivalis, we did not find significant effects of the treatment stage on the number of Nowickia ferox visits (χ 2 = 0.017, df = 2, p = 0.99), however the inflorescence type (χ 2 = 10.27, df = 1, p < 0.01) and the interaction treatment stage × inflorescence type (χ 2 = 6.98, df = 1, p < 0.05) had significant effects on pollinator visits (Figure 10). Post hoc tests indicated that inflorescences with the dark top had significantly more visits than inflorescences without it (z = 3.507, p < 0.01; Figure 9), but no other pair-wise differences were found.

DISCUSSION
Despite being widely distributed, Neotinea ustulata is one of the most enigmatic European orchids in terms of its reproductive biology. Our study reveals that morphological (e.g., larger flowers in var. ustulata), visual (e.g., spectral reflectance pattern) and chemical (e.g., pattern of semi-and low-volatile compounds) floral traits distinguish N. ustulata var. ustulata from var. aestivalis. Our observations across different locations in Europe also support some early records that N. ustulata is mainly pollinated by tachinid flies (Godfery, 1933;Vöth, 1984). However, the relative importance of tachinids differs between the two varieties, playing a predominant role as pollinator for the late flowering var. aestivalis. The complexity of floral traits and pollinator spectra of N. ustulata may indicate a complex, and not yet fully understood, food-deceptive mechanism.
Morphological, Visual, and Chemical Differences Between N. ustulata Varieties

Morphological Cues
Flower morphology can play a key role in pollination by promoting reproductive isolation, particularly in highly specialized pollination systems (Schiestl and Schlüter, 2009). For example, in the sexually deceptive orchid genus Chiloglottis R.
Br., differences in floral morphology have been shown to be enough to achieve pollinator specificity in species relying on the same chemical attractant (de Jager and Peakall, 2016). Variations in floral size can also affect morphological fit and, therefore, pollen/pollinia transfer by their respective pollinators (Anderson and Johnson, 2008;Solís-Montero and Vallejo-Marín, 2017;de Jager and Peakall, 2019). Even relatively subtle differences in flower structures can have an impact on floral isolation such as the position of pollinaria in the moth pollinated Platanthera bifolia (L.) Rich. and P. chlorantha (Custer) Rchb., which are then attached on different parts of the pollinators (Nilsson, 1983;Schiestl and Schlüter, 2009). Although the differences in size among floral structures of N. ustulata var. ustulata and var. aestivalis were usually quite small, they are consistent even at the population level, and suggest that the two varieties might be morphologically adapted to different pollinator spectra, although FIGURE 9 | Pollinaria removal success in up to three attempts by the two main pollinator guilds of Neotinea ustulata, bees (Anthophora plumipes; n = 12) and flies (Tachina fera; n = 6). Back-transformed means and standard errors are shown.
phenotypic plasticity and non-pollinator selection should not be ruled out (Caruso et al., 2019;Sletvold, 2019). For instance, the width of the hood, the length of the lip and the length of the spur could all act as filters for potential pollinators. In fact, a previous study has already shown that the spur length of N. ustulata could be under pollinator-mediated selection (Trunschke et al., 2017). The role of flower traits for ensuring mechanical fit has been documented in diverse plant groups (e.g., Steiner and Whitehead, 1990;Muchhala, 2007;Anderson and Johnson, 2008;Pauw et al., 2009), including orchids (e.g., Chapurlat et al., 2015;Gögler et al., 2015;Rakosy et al., 2017;de Jager and Peakall, 2019). The smaller dimensions of N. ustulata var. aestivalis flowers could thus favor tachinids as pollinators and, conversely, larger flowers with deeper spur could favor a wider range of pollinators and improve pollinaria transfer. Assessment of mechanical fitting and behavioral experiments are necessary to test the role of morphology in N. ustulata pollinator filtering.

Visual Cues
Floral color is one of the main traits used by pollinators to identify both rewarding flowering plants and their deceptive mimics (Reverté et al., 2016). While all pollinators are able to respond to flower color signals, their preference for particular color hues and their ability to discriminate between them varies among groups (Reverté et al., 2016). Color reflectance patterns were structured at the population within and between varieties of N. ustulata. Although these differences can be related to phenotypic plasticity and non-pollinator-mediated selection (Ellis and Johnson, 2009;Caruso et al., 2019;Sletvold, 2019), pollinator preferences could be also involved. For instance, pollinator preferences have shaped daisy communities, which function as fly pollination specialist (Ellis et al., 2021). Variation in color hue among N. ustulata populations can be related to pollinator detectability and environmental conditions such as background context (e.g., coloration of surrounding plants and soil). Independently of the population or variety origin, the stark chromatic contrast between the dark top and the white lips can be discriminated by both bees and flies, and its presence may be result of pollinator selection (see below). This is supported by the occurrence of bright top forms (e.g., light pink or white top) in N. ustulata inflorescences, which are not stable and appear from time to time in different populations (Foley, 1990). Color pattern and color contrast have been shown to be important in pollinator attraction (Zhang et al., 2017), and the dark top coloration would play a similar role (see below). Experiments are needed to test the role of color pattern in the pollination success of N. ustulata varieties.

Chemical Cues
Both N. ustulata varieties produced a rather complex floral scent and the scent bouquets of N. ustulata individuals were highly variable. This pattern is not restricted to N. ustulata populations from Czech Republic since the floral scent of a var. ustulata plant from Sweden was similarly dominated by α-pinene, β-myrcene, limonene, eucalyptol, and (E)-β-ocimene (Kaiser, unpublished data). A study based on a plant from Italy also reported a complex bouquet including eucalyptol and D-limonene (i.e., D'auria et al., 2021); although we are cautious with the identity of some compounds also reported by the authors as they seem to be environmental contaminants. Differences between N. ustulata varieties were only detected when comparing compounds with low-and semi-volatility; however, larger scale sampling would be required to confirm that there are not differences between varieties when comparing highly volatile compounds. Differences in hydrocarbon compounds, which were predominant among compounds with lower volatility, could induce differential responses of pollinators once they approach or land on the flowers. Indeed, the function of these compounds, especially alkenes, has been intensively investigated in sexually deceptive orchids (Ayasse et al., 2011;Bohman et al., 2016). In fact, it is extremely unusual for food-deceptive species to produce such a high proportion of alkenes (Schiestl and Cozzolino, 2008), and in particular the production of (Z)-11-tricosene, which is extremely rare even among sexually deceptive plants but has been recorded in some tachinids (Martel et al., unpublished data). Therefore, alkenes and other lowvolatile compounds could play a major role at short-range attraction in N. ustulata by eliciting certain behaviors in tachinid pollinators, while highly volatile compounds and visual cues remain probably the most important stimuli for long-range attraction. Further physiological and behavioral analyses are needed to elucidate the differential importance of N. ustulata floral scent in pollination.

Reproductive Success in N. ustulata Varieties
As typical for deceptive orchids, fruit set was rather low in most populations of N. ustulata. Reproductive success of N. ustulata was found to be higher than previously reported (i.e., Neiland, 2001;Harder and Johnson, 2008), but matches with those of some food-deceptive orchids (Gill, 1989;Fritz and Nilsson, 1994;Scopece et al., 2009Scopece et al., , 2010Trunschke et al., 2017). Contrary to previous reports (i.e., Tali et al., 2004;Haraštová-Sobotková et al., 2005), we did find differences in the characters associated with reproductive success between varieties. Furthermore, N. ustulata var. aestivalis was found to be slightly more successful than var. ustulata. Nevertheless, independently of the variety, the bottom and middle section of the inflorescence developed more fruits than the top section. This is consistent with gradual acropetalous opening of N. ustulata inflorescences and with tachinid behavior, as they always visit the freshest flowers on the top, but as flies habituate and more buds open, later opening flowers get less visits. This pattern is similar to those recorded in food-deceptive plants pollinated by bumblebees, which contrary to tachinids usually pursue the inflorescence from the bottom toward the top (Nilsson, 1980(Nilsson, , 1983(Nilsson, , 1984Fritz, 1990;Jersáková and Kindlmann, 1998) and in sexually deceptive plants pollinated by bees and wasps, which learn to avoid deceitful plants (Peakall, 1990;Peakall and Handel, 1993;Ayasse et al., 2000). Thus, reproductive success pattern of N. ustulata indicates that pollinators are initially attracted to visit the plants, but then learn to avoid them.
Pollinator Spectra in N. ustulata Varieties Diverse insect taxa visited flowers of both N. ustulata varieties, but only a few acted as true pollinators and the two varieties differ in their main pollinators. Both Anthophora and Tachina are able to remove pollinaria of N. ustulata var. ustulata with their proboscis, but Tachina flies were more efficient in doing so. Furthermore, while the tachinids were unable to detach the pollinaria from their proboscis, we frequently observed Anthophora bees grooming them off. Our observations of N. ustulata var. ustulata are in line with previous records in which bees and flies have been recorded (Paulus, 2005;Claessens and Kleynen, 2016). In contrast, our observations do not support the assumption that N. ustulata var. aestivalis is adapted to beetle pollination (see Danesch and Danesch, 1962;Mrkvicka, 1991;van der Cingel, 1995), though beetles can also act as occasional pollinators. While we observed a broad range of beetles, bees and even butterflies visiting the flowers of var. aestivalis, only Nowickia and Tachina were was consistently observed removing pollinaria and occasionally Bombus spp. Thus, N. ustulata var. aestivalis may have become more specialized for fly pollination than var. ustulata and this adaptation could be triggered by differences in availability of insects through space and time of flowering, and the higher efficiency of tachinids in transferring pollinia compared to bees. Despite the difference in pollinator spectra, pollen transfer efficiency was similar between both varieties, which would also be linked to their similar degrees of specialization or adaptation to fly pollinators. Our tachinid collection from Austrian and Czech populations suggests that the two N. ustulata varieties have one species of Nowickia and two of Tachina as main pollinators (i.e., N. ustulata var. ustulata is pollinated by T. fera and T. magnicornis, and var. aestivalis by T. magnicornis and N. ferox); however, differences in tachinid attraction can also be due to differences in species availability throughout the year and not specific adaption to one of them. Tachina fera and T. magnicornis are morphologically very similar, can co-occur and can be easily misidentified without a close inspection by a trained person. Although previous reports indicated that T. magnicornis flies were visiting N. ustulata var. ustulata flowers (e.g., Vöth, 1984;Paulus, 2005), some of those flies could have actually been T. fera (see Supplementary Table 6), which is more abundant and common than T. magnicornis (Paulus, pers. obs.). Fly pollination is rather rare among European orchids (Jersáková et al., 2016) and, to the best of our knowledge, no tachinid species has been recorded as the main pollinator of any European orchid. Tachinids have been previously reported in only a couple of specialized pollination systems such as those of Schizochilus (van der Niet et al., 2010) and Telipogon (Martel et al., 2016). Independently of the identity of their pollinators, both varieties have a similar male success and pollen transfer efficiency, but different female success. Hence, based on pollen transfer efficiency, pollinators of both varieties seem to be very efficient in delivering pollinia (i.e., a ratio of pollinia deposited to pollinia removed above 0.50). The pollinator efficiency is linked to the pollinator constancy as well as pollinator identity (Peter and Johnson, 2009). We therefore suggest that the reported efficiency may be related to tachinids, which seem to be more constant and are more efficient in removing pollinaria.

Role of the Dark Top in N. ustulata
We have reliably demonstrated that the dark top is an attractant signal of N. ustulata as inflorescences bearing the dark top received significantly more visits of tachinids than inflorescences without it. This increasing attractiveness can be result of a combined effect of color, shape and size of the dark top (i.e., visual stimuli). These findings appear to support the theory put forward almost 90 years ago by Godfery (1933), which considered that the dark top may be an adaptation to attract fly pollinators. The dark top of N. ustulata is quite unique among European orchids and although other species can also present a dark top (e.g., Orchis purpurea Huds.), they seem to have a far weaker contrast against the lips. Dark spots have also arisen in diverse plant groups such as Apiaceae, Asteraceae, Geraniaceae, Iridaceae, Liliaceae, Papaveraceae and Ranunculaceae (Dafni et al., 1990;Westmoreland and Muntan, 1996;Johnson and Midgley, 1997;van Kleunen et al., 2007;Goulson et al., 2009;Ellis and Johnson, 2010). It was suggested that dark, contrasting spots in flowers and inflorescences might even act as an insect mimic as demonstrated in some beetle-and fly-pollinated plants (e.g., Johnson and Midgley, 1997;van Kleunen et al., 2007;Goulson et al., 2009;Ellis and Johnson, 2010); however, at this point, it is unknown how tachinids are perceiving the contrasting dark top of N. ustulata. Although dark spots in flowers promotes pollinator attraction in some plants (e.g., Johnson and Midgley, 1997;van Kleunen et al., 2007;Goulson et al., 2009;Ellis and Johnson, 2010), it does not necessarily involve enhancement of fruit set (e.g., Westmoreland and Muntan, 1996;Johnson and Midgley, 1997;Ellis and Johnson, 2010). However, as dark spots promote visits of the main pollinators, their presence is likely to impact the fruit set of N. ustulata. In situ manipulation experiments would be needed to know how tachinids are perceiving the contrasting dark and its impact on pollination success.

Pollination Strategy of N. ustulata
Neotinea ustulata varieties are visited by generalist insects that seek for nectar and pollen (e.g., bees and flies), and the lip is bright white as in other food-deceptive plants (Johnson and Schiestl, 2016). However, other floral traits such as a strong and complex scent including alkenes, the evolutionary innovation of a dark top on the inflorescence and the predominance of tachinid male visitors (Martel et al., unpublished data) suggest that N. ustulata may have evolved a complex food-deceptive pollination system. Based on previous studies (i.e., Vogel, 1972;Bino et al., 1982;Valterová et al., 2007;Scopece et al., 2009), it seems that some orchid species potentially classified as fooddeceptive have also evolved signals not usually associated with this strategy (e.g., some kind of chemical mimicry of their pollinators) and their pollinators are gender biased (i.e., attract predominantly female or male pollinators). This may also occur in N. ustulata as most abundant alkenes have been also detected in Tachina flies (Martel et al., unpublished data). Most floral visitors of N. ustulata are only seeking food, but tachinids would in addition perceive alkenes as a signal of other yet unknown resources. For instance, alkenes are key in the chemical mimicry and pollinator attraction of the sexually deceptive Telipogon peruvianus T.Hashim. to its tachinid specialist pollinator (Martel et al., 2019). In other complex pollination systems, the flypollinated Gorteria diffusa Thunb. (Asteraceae) attracts flies seeking food, but male flies are also sexually attracted and try to copulate with the black petal ornamentation (Ellis and Johnson, 2010); similarly, some populations of the beetlepollinated Luisia teres Gaudich attracts nectar-feeding females and male beetles, whereas others sexually attract males (Sugiura et al., 2021). Hence, the combination of traits present in N. ustulata indicates that this orchid would exploit not only food-based signals but also sexual ones (e.g., gender-specific pheromones). To address this issue further studies should identify physiologically active scent compounds and carry out behavioral bioassays.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ACKNOWLEDGMENTS
We thank Jaromír Vaňhara, Peter Sehnal, and Hans-Peter Tschorsnig for tachinid determination; Ivana Jongepierová, Karel Fajmon, Vlastimil Vlačiha, Dano Dítě for locations of N. ustulata populations. We also thank Casper van der Kooi and Ewan Bodenham for the invitation and support to publish our study in this special issue. Roxana Rojas kindly prepared the map shown in Figure 2. The reviewers make very helpful comments that improved an earlier version of the manuscript. DAAD and MŠMT supported this study by a shared Grant to JJ (7AMB16DE003) and MA (DAAD-16-13/2016(DAAD-16-13/ -2017. CM thanks the Max Planck Partner Group for providing financial supports at PUCP.