Visibility and attractiveness of Fritillaria (Liliaceae) flowers to potential pollinators

Visual floral characters play an important role in shaping plant-pollinator interactions. The genus Fritillaria L. (Liliaceae), comprising approximately 140 species, is described as displaying a remarkable variety of flower colours and sizes. Despite this variation in visual floral traits of fritillaries, little is known about the potential role of these features in shaping plant-pollinator interactions. Here, we seek to clarify the role of visual attraction in species offering a robust food reward for pollinators early in the spring, which is the case for Fritillaria. We also searched for potential tendencies in the evolution of floral traits crucial for plant-pollinator communication. The generality of species with green and purple flowers may indicate an influence of environmental factors other than pollinators. The flowers of the studied species seem to be visible but not very visually attractive to potential pollinators. The food rewards are hidden within the nodding perianth, and both traits are conserved among fritillaries. Additionally, visual floral traits are not good predictors of nectar properties. When in the flowers, pollinators are navigated by nectar guides in the form of contrasting nectary area colouration. Flower colour does not serve as a phenotypic filter against illegitimate pollinators—red and orange bird-pollinated fritillaries are visible to bees.


Materials and methods
Spectral analysis. Most Fritillaria species are rare in nature and grow in hardly accessible, remote places.
Therefore, performing studies in natural habitats on a satisfying number of species is often not possible. Nonetheless, to gain better insight into plant-pollinator communication in fritillaries, we decided to use material available at the University of Warsaw Botanic Garden (hereafter BG) and in private collections. For this analysis, we used flowers from plants cultivated in the BG and in private collections of Colin Everett (Somerton, Somerset, UK; hereafter CE), Laurence Hill (Richmond, Surrey, UK; hereafter LH) and Paweł Kalinowski (Szczeglacin, Korczew, Poland; hereafter PK). One species (F. messanensis) was analysed in situ on Crete. Since most of the Fritillaria species are also rare in cultivation, the number of flowers analysed varied due to the availability of plant material (the sources and sample sizes used in this study are listed in Supplementary materials Table 1).
To better understand the visual attractiveness of Fritillaria flowers, we performed spectral reflectance analyses of 35 species (Fig. 1, Supplementary materials Table 1). Since there are few species of fritillaries with red flowers (with the exception of F. eastwoodiae, where the red tepals have a large amount of yellow admixed in) and all of them are bird-pollinated (Supplementary materials Fig. 1), we wanted to test the eventual deterrence of insect visitors by these red flowers, which was previously detected in studies of bird-pollinated species 57 . Flowers from www.nature.com/scientificreports/ the BG collection were sampled directly prior to analysis. Flowers from the collections of CE, LH and PK were collected and stored in a refrigerator (4 °C) for three to five days before analysis. Storing flowers at a low temperature may potentially change the spectral reflectance of collected specimens, yet to the best of our knowledge, this is one of most common ways of preserving flower colours. For the collected flowers, we measured the reflectance of all the dominant-coloured areas of the tepals (the most widespread colours within a flower, with the assessment of colour based on human perception) inside and outside of the flowers. Additionally, reflectance was measured for the nectaries (scheme showing the analysis pattern in Supplementary materials Fig. 2). The reflectance curves represent a mean value of five measurements from one flower (prepared using Python version 3.7).
The reflectance measurements were performed with a JAZ Spectrometer System (Ocean Optics Inc., Florida, USA) at an angle of 45° to the measurement site (connected via a fibre-optic cable; UV-VIS 400 μm; World Precision Instruments, Inc., Florida, USA). To calibrate the spectrometer, we used a black standard (black PTFE powder, Spectralon diffuse reflectance standard SRS-02-010, reflectance factor of 2.00%, Labsphere, Inc., North Sutton, USA) and a white standard (white PTFE powder, Spectralon diffuse reflectance standard SRS99-010, reflectance factor of 99.00%).
To analyse the spectra according to a potential pollinator's visual system, we used the vismodel function in the 'pavo' package. This tool enabled us to show the reflectance spectra in combination with colour-space models of pollinator vision (settings of the vismodel analysis in Supplementary materials S1). We analysed the visual systems of bees (Apis mellifera), flies (Musca domestica and Eristalis tenax), butterflies (Pieris rapae) and birds (Cyanistes caeruleum, a passerine bird, and Sephanoides sephaniodes, a hummingbird). These species were selected as representatives of potential pollinator groups based on the literature 41,42,[53][54][55][56] . Prior to analyses, the reflectance data were smoothed, and the negative values were set to 0 (procspec function in the 'pavo' package).
To find out more about bee vision and test for the presence of nectar guides in fritillaries, we took pictures with several filters (for details, see below). Subsequently, we prepared false-colour pictures of Fritillaria flowers. Many details of fritillary flowers, such as colour patterns (produced by chequered tepals), nectaries, and stamens, are too finely scaled for spectral reflectance measurements. We applied false-colour photography to reveal these colour patterns. These analyses were performed for 34 Fritillaria species (Fig. 1, Supplementary materials Table 1). Photographs of the flower details were taken by using a UV-sensitive Nikon D60 (Tokyo, Japan) camera, which was then combined with a filter transmitting human-visible light (UV/IR cut filter: Baader UV-IR-Cut/L-filter) or a UV-transmittance filter (Baader U-filter 2, transmission peak wavelength 350 nm, half-bandwidth 60 nm; Baader Planetarium, Mammendorf, Germany), in daylight. The false-colour images of Fritillaria flowers were prepared in Photoshop (version CS5) according to the procedure described in Luanau and VerhoeVen 58 .
Phylogenetic tree. To understand the evolutionary tendencies of the floral traits of fritillaries, we constructed a phylogenetic tree for phylogeny-based analysis. We used the nuclear ITS (DNA internal transcribed spacer) and the plastid genes matK and rpl16. We used Lilium L. as an outgroup based on the established relationship 43,49 . All of the gene sequences were acquired from GenBank (accession numbers in Supplementary Materials S2) and downloaded using the MatPhylobi program 59 , which is a command-line tool for constructing taxonomic data sets for phylogenetic inference based on NCBI data. To create the phylogenetic database in MatPhylobi, we seeded F. michailovskyi and L. regale as representatives of the studied genera. The sequences downloaded represent 113 Fritillaria species.
All three types of sequences were independently aligned using the online implementation of the multiple alignment program MAFFT (version 7.64 60 ). Subsequently, all alignments were imported into a modular system for evolutionary analysis, called Mesquite (version 3.6 61 ), for visual inspection. Poorly aligned positions and divergent regions were eliminated using TrimAl (version 1.3, http:// phyle mon2. bioin fo. cipf. es). The trimmed alignments were then concatenated with RAxML, a tool for phylogenetic analysis (version 8.0 62 ).
Using the combined data set, we estimated the divergence time of Fritillaria. To do so, we used a log normal relaxed clock in BEAST 2 version 2.6.3 63 . On the basis of the jModelTest results, we used GTR as the substitution model. We specified the tree prior model as the fossilized birth-death model and calculated the posterior distributions of internal nodes with Markov chain Monte Carlo (MCMC) analyses of 10 million generations with a 10% burn-in. To calibrate the tree, we used four dated fossils for Liliales previously used in Fritillaria dating 64 : Luzuriaga peterbannisteri (age range: 23.03-15.97 Ma; in our study, we set the age to 20 Ma), Smilax labidurommae 65 (age range: 37.2-33.9 Ma; in our study, we set the age to 35 Ma), S. pristina 66 (age range: 55.8-48.6 Ma; in our study, we set the age to 51 Ma), and Ripogonum tasmanicum 67 (age range: 55.8-48.6 Ma; in our study, we set the age to 51 Ma). We set the mean and standard deviation of these distributions to 1 and 1.25, respectively, and the confidence interval to 10 Ma.
Using our refined database of sequences, we also used RAxML to generate our final phylogenetic tree for ancestral state analysis. RAxML uses a series of maximum-likelihood (ML) tests to generate the tree. To find the best phylogenetic tree, we used a bootstrap analysis with 1000 replicates for each of the partitions (with models recommended by PartitionFinder 68 ).

Data collection-nectar properties, floral traits, and species distribution.
To better understand plant-pollinator communication in Fritillaria, we prepared a database describing the diversity of nectar properties and several floral traits that are potentially important for increasing attractiveness in the studied genus. Our sources were scientific literature describing fritillaries, regional floras, and internet sources (http:// www. friti llari aicon es. com, http:// www. eflor as. org, and http:// www. friti llaria. org. uk). Supplementary materials Table 1 presents all the data. We assessed available information for nectar properties and six floral traits for 113 species present on the phylogenetic tree. www.nature.com/scientificreports/ Nectar properties, i.e., volume and sugar concentration, were obtained from Roguz 38,41 . In that study, the authors measured the volume of nectar collected from flowers during anthesis. Flowers for nectar sampling were bagged with mesh during the bud stage to prevent visits by insects. During anthesis but before anther dehiscence, nectar was sampled. All samples assigned to a specific treatment were analysed with the use of the same methods. Sample size varied depending on the species, with a range from four to 70 (Supplementary materials Table 1).
All the studied fritillaries produce nectar, which is easily accessible to pollinators; however, nectaries are usually hidden within the nodding perianth. We assessed the number of species where flowers are not facing down and wide open with visible nectaries. Additionally, we counted species where the nectary presents contrasting colours. Several pollinators visit flowers mainly for pollen 61 ; therefore, their presentation may play a crucial role in luring potential pollinators. Therefore, we calculated the number of Fritillaria species with anthers visible from a distance, i.e., flowers with a wide open perianth (e.g., F. bucharica) or with the anthers exerting beyond the perianth (e.g., F. imperialis). Fritillaria is one of very few genera where a chequered pattern covers the tepals, so we also collected information regarding the presence or absence of this pattern on the tepals. To estimate the visual attractiveness of the flowers, we collected information about the size of the flowers, i.e., entrance diameter (entrance to the flower at the widest point) and the length of the tepals.
To assess the differences among Fritillaria species, we collected information about the distributions of modern representatives of this genus. We defined five distributional areas, namely, North Africa, Asia-temperate, Asiatropical, Europe, and North America, and a combination of these distributions.
Phylogenetic studies. Next, using the obtained tree and trait databases, we determined the ancestral states of reward presentation, the chequered pattern, and the species distribution. The analysis was performed using single rate model reconstruction, which gives empirical Bayesian probabilities ('ape' package, version 5.0 69 ). First, we determined the appropriate transition probability model, choosing among ER-equal rate, SYM-symmetrical rate, and ARD-all-rates different, using a log likelihood ratio analysis. In all cases, the ARD transition ratio model was chosen because it had the highest likelihood value. Then, to compute the total number of character changes, as well as the number of character changes between all states, we used the make.simmap function in the 'phytools' package (1000 simulations across the tree 70 ).
Finally, using the phylogenetic tree of Fritillaria, we tested for correlations between continuous traits and nectar properties (nectar volume, sugar concentration, entrance diameter and tepal length) and discrete traits (anther and nectary presentation, colour of the nectaries, presence of the chequered pattern and species distribution). We conducted phylANOVA on other visual traits (performed in R using the phylANOVA function in the 'phytools' package). This function performs the simulation-based phylogenetic ANOVA of Garland and Jones 71 . The level of correlation between the continuous traits was assessed by calculating the phylogenetic generalized least squares (PGLS). In this study, the structure of the phylogenetic signal was controlled by the lambda parameter. The branch length transformation was optimized between bounds using maximum likelihood ('caper' package, version v1.10.1 72 ). The statistical analyses were performed with the statistical computing software R 3.5.3 73 (R analysis code accessible in Supplementary materials S3).

Results
Spectral analysis. For most of the studied fritillaries, the reflectance was higher above 500 nm and lower in the blue and UV wavelength range. Species with bright flowers (white or pink) were an exception, with reflectances higher above 350 nm. The ultraviolet reflectance properties of Fritillaria species varied within subgenera, with no obvious pattern of UV reflection. For example, the subgenus Fritillaria includes UV-reflecting species (F. montana) and UV-absorbing species (F. graeca). Closely related F. imperialis and F. eduardii, representing the subgenus Petilium, had UV-absorbing tepals and UV-reflecting stigmas ( Fig. 2; Supplementary materials Fig. 3). Flowers of Fritillaria ariana displayed a UV bull's eye pattern with a UV-reflecting centre. Additionally, flowers of F. liliacea had a UV-absorbing centre. The species with chequered tepals had UV-reflecting white dots (Fig. 3).
For almost all Fritillaria species, the reflectance of the nectaries was much lower than that of tepals. Only the nectaries of F. eastwoodiae, F. dasyphylla and F. minima had reflectances similar to or higher than the values recorded for the tepals. The reflectances of the inner and outer sides of the tepals were similar (Figs. 2, 3; Supplementary materials Fig. 3).
False-colour pictures indicated that the nectary area in fritillaries is distinctive. As a result, nectaries may act as visible nectar guides for bees. The anthers and pollen grains, mostly yellow under human vision, proved to be visible to bees in both insect-and bird-pollinated flowers. The chequered pattern present on the tepals of some species was also visible to bees. The nectaries of the subgenus Petilium had an accumulation of starch, producing a white appearance, which was also confirmed in the false-colour pictures (Fig. 3). Bird-pollinated species of fritillaries, i.e., F. imperialis, F. gentneri and F. recurva, are visible to pollinators, with distinctive anthers and nectary areas.
Modelling of fritillary flower colour appearance in the chromaticity diagram of Apis mellifera and Eristalis tenax showed that the studied species densely occupied the portions of long waves: the centre and around the corner with long wavelengths. Bird-pollinated Fritillaria flowers occupied the same sections. In the case of Musca domestica, the model also showed a large accumulation of species presumably pollinated by insects in the corner representing long waves. Species pollinated by passerine birds were located closer to the green photoreceptor type. For Lepidoptera, represented by Pieris rapae, all fritillaries occupied mostly the centre of the diagram. For potential bird pollinators, both representatives of hummingbirds and passerine birds, mostly possessing cones with long-and medium-wavelength sensitivities, were stimulated. The accumulation of most points in the lm range suggested the stimulation of cone oil droplets with carotenoid filters (Fig. 4). www.nature.com/scientificreports/ www.nature.com/scientificreports/ Flower traits, nectar properties, pollinators, and species distribution. All the studied Fritillaria species produced nectar, which is an important reward for pollinators. Its concentration and volume were variable, ranging from 5.3 to 64.9% and from 0.3 to 205 µl per flower, respectively. The flowers were usually bell-shaped and nodding, and as a result, nectaries and anthers were not visible to approaching animals. In general, 13.9% of fritillaries (16 species) had nectaries visible from the outside, and 21.7% (25 species) had anthers that may be perceived from a distance. The nectaries usually had different colours than the background (75.6%, 87 species). The chequered pattern was common among the Fritillaria species, where 54% (61 species) had spotted tepals. The perianth entrance diameter and the length of the tepals were variable and depended largely on the species, with ranges from 9.3 to 67.8 mm and 10.5 to 53.2 mm, respectively.
The relationships among the studied features-phylogeny-based analyses. Fritillaria was estimated to have originated approximately 38 Mya, in the early Oligocene (Fig. 5). The most recent common ancestor of Fritillaria occurred in temperate Asia, where the highest diversity of fritillaries is still found (Supplementary materials Fig. 4). Expansion to North America occurred early in the evolution of this genus. Species pollination by passerine birds evolved earlier (approximately 27.5 Mya) and hummingbird-pollinated species appeared recently in the phylogeny of the studied genus (approximately 24 Mya).
The ancestral state reconstruction in Fritillaria suggests that the most recent common ancestor probably had flowers with nectaries hidden deeply within the perianth but with a distinctive colour (Supplementary materials Fig. 5A,B). Both of these states were also the most common across the phylogeny and among modern species  Table 2). The anthers of most fritillaries are and always have been not visible from a distance (Supplementary materials Fig. 5A,B). Even when visible nectaries and/or anthers evolved, these features were lost several times. The chequered pattern also evolved early during the evolution of this genus. The majority of internal nodes had a pattern on the tepals. Several subgenera, such as Petilium, Theresia and Korolkovia (species representing the subgenera mentioned in Supplementary materials Table 1), appear to have arisen from an ancestor with chequered patterns on the tepals, with recent loss of this character (Supplementary materials Fig. 5A,B). All the studied flower features were labile, with the highest number of changes observed for the presence or absence of the chequered pattern (747 shifts between the states) and the lowest for the visibility of the anthers (56.8, Supplementary materials Table 2). Taking into account the phylogenetic relationships among fritillaries, we did not find associations between floral traits and nectar properties (Supplementary materials Table 3). Testing for correlations between the studied characters with PGLS, we found only two significant negative correlations: between entrance diameter and tepal length (r = − 0.09, p value < 0.05) and between nectar volume and entrance diameter (r = − 0.02, p value < 0.05, Supplementary materials Table 4). www.nature.com/scientificreports/

Discussion
The flowers of fritillaries show a wide diversity of colours and colour patterns. In this study, we assess the attractiveness of Fritillaria flowers to potential pollinators and search for correlations between visual traits and reward properties. We also track the evolutionary tendencies of floral traits related to reward presentation.
Most Fritillaria flowers reflect wavelengths > 500 nm, while hymenopteran pollinators are likely to have the best discrimination for wavelengths close to 400 nm and 500 nm 74 . Therefore, the plant-pollinator relationship in fritillaries seems to be based on robust food rewards; however, these rewards are not advertised. Neither flower entrance diameter nor tepal length is correlated with nectar properties. Most of the modern species also have both nectaries and anthers hidden deeply in a nodding perianth, which is also the most likely state of the recent common ancestor. When pollinators are in the flower, the nectaries are distinctive, and their colours differ from that of the background. Different nectary colour is and has been common among fritillaries. Red and orange flowers are found only among bird-visited fritillaries. The principles of the pollination syndrome theory, however, may be applied to fritillaries only to a limited extent. Flowers of these species are also visited by bees, suggesting ecological generalization, despite some level of phenotypic specialization. One of the reasons for this 75 may be relatively late diversification. Fritillaries originated approximately 38 Mya, with the most recent common ancestor found in temperate Asia, but specific pollination syndromes appeared later. For example, species pollinated by hummingbirds appeared approximately 24 Mya.
False-colour pictures prepared for fritillaries proved that flower colours of bird-pollinated species and the rewards within were visible to bees. This also confirms observations from natural habitats, where North American red-flowered fritillaries are visited by bees 55 . It is important to note that among fritillaries, an evolutionary transition from purple to red flowers is common 51 . This type of shift often accompanies shifts in adaptation from bee to hummingbird pollination or co-occurs with other floral adaptations that facilitate hummingbird pollination 76 . Several studies proved the attractiveness of red-flowered species to insects [77][78][79] ; however, to protect the reward, some hummingbird-pollinated flowers reflect mostly long-wavelength light (median wavelengths > 585 nm), as this decreases their visibility and helps avoid visits by illegitimate pollinators 7,16 . This was not the case for the studied species, which may be related to their relatively recent evolution. Hummingbird-pollinated fritillaries evolved in North America and are absent in Europe and Asia. Due to this relatively recent evolution, hummingbird-pollinated fritillaries probably had no chance to adapt to these birds in the Old World 80,81 .
Similarly, as in the case of hummingbird-pollinated flowers, passerine bird-pollinated fritillaries are also visible to insects. Flowers of these species also have distinctive colouration within the genus: typically, orange-red (rarely yellow) flowers that lack reflectance in the UV range. As with hummingbird-pollinated North American fritillaries, it is still unclear why Asian bird-pollinated F. imperialis and F. eduardii exhibit this colouration or if passerine birds have a preference for these colours. We assume that tepal colour is an important long-distance signal 57 , although the rewards reflecting the preference of these pollinators (hexose-and amino acid-rich nectar, lacking sucrose 40,41 in passerine bird-pollinated flowers and very large amounts of nectar with low amino acid concentrations in hummingbird-pollinated species) are equally important. The observed ecological generalization, despite some level of phenotypical specialization 75 , observed in fritillaries may be a result of relatively shortduration fritillary-bird interactions. Studies testing the evolution of flower colour and rewards in members of this genus suggest that changes in these flower features may have appeared only recently 52 , and Asian F. imperialis and F. eduardii flowers visited and pollinated by passerine birds 53,54 probably evolved approximately 27.5 Mya.
Despite the importance of visual floral traits in shaping plant-pollinator interactions 82 , we did not find visual attractants described as important for insect pollinators [83][84][85] in most of the studied species. For example, our analyses indicated overall UV absorbance and did not indicate additional ultraviolet-reflecting areas. UVreflecting flower parts are common among insect-pollinated species 7,86 , e.g., in yellow bee-pollinated flowers 86,87 , where they form the so-called bull's eye pattern with a UV-reflecting periphery and UV-absorbing centre 88 . In fritillaries, we detected the bull's eye effect in species with bright pink flowers, namely, F. ariana, F gibbosa and F. stenanthera. These closely related species have wide open, erect flowers; therefore, we assume that the angle of the flower may correlate with the presence of the bull's eye pattern-it could not be seen in downward-facing bell-shaped flowers. The lack of a strong visual signal, common among fritillaries, is most likely caused by the thick wax cover present in almost all fritillaries 41 . Additionally, the similarity of the reflectance curves found in our study may be due to the similarity of tepal ultrastructure in Fritillaria, which may determine the reflectance amplitude 89 . Previous studies conducted by Roguz 39 revealed that almost half of the genus has nearly the same tepal and nectary ultrastructure.
We assume that in fritillaries, where the reflectance does not fit the preferences of the most important pollinators, visual signals perceived from a short distance may play an important role. The presence of the chequered pattern, although labile, evolved early in fritillary diversification and is very common among modern species. The chequered pattern, rare among other plant species, can enhance flower attractiveness 90 by increasing the contrast between the object and background through motion parallaxis 91 . Such purple-yellow mottled tepals, common among fritillaries, can also increase the attractiveness of their flowers to flies. These insects are pollinators of at least one fritillaria species (F. camschatcensis 56 ) and have been seen visiting several other species (authors' observations).
Additionally, nectaries and anthers, hidden deeply within a nodding perianth, are common and conserved states. In most of the fritillaries, the placement of the reward is "highlighted" by nectaries in different colours, and contrasting nectaries evolved early and are frequent among modern species. Previous studies on Fritillaria revealed a lack of phylogenetic signal for nectar properties. These traits show high variability at the community level, which might suggest adaptive evolution in response to different environmental conditions 27 . These variabilities in Fritillaria nectar volume and concentration may also be a reason for the lack of correlations between www.nature.com/scientificreports/ floral traits and nectar properties. These results, however, are surprising, since several studies have shown an association between flower size and rewards 27,32,92 . Nevertheless, the most luring advertisement is to display the reward itself, which then confers the dual role of both signal and reward 93,94 . Open flowers with nectar displayed in the form of large droplets and bright yellow pollen contrasting with the dull flower evolved in Fritillaria only a few times. Both of these traits were lost several times across the phylogeny. Although plant pollinators are among the most important factors influencing visual floral traits, their variability may also be maintained by selection related to environmental heterogeneity. Two of the most common colours in the case of Fritillaria flowers are green and purple, which may be related to random or non-adaptive processes maintaining flower colour variation 95 . However, this colour polymorphism may also be related to the ecological role of the pigments responsible for this type of colouration. Studies by Bazzaz and Carlson 96 revealed, for example, that the photosynthetic activity of green flowers may be enough to maintain their own structure. On the other hand, floral pigments such as anthocyanins, related to violet-purple colouration, function in response to plant stress caused, e.g., by drought, cold or nitrogen deficiency.

Conclusions
The attractiveness of Fritillaria flowers appears to be linked to early blooming and copious rewards. Flower signalling in Fritillaria seems to follow a rather simple pattern, with most flowers having almost uniform displays or occasional coloured peripheral and central parts of the flower. For animals already visiting the flowers, fritillaries have nectar guides in the form of contrasting nectary area colouration. This study and recent works on other aspects of pollination biology in Fritillaria indicate that plant-pollinator interactions may be responsible for a significant share of the floral diversity in the genus. The principles of the pollination syndrome theory, however, may be applied to fritillaries with some limitations. Flower colour does not play a role as a phenotypic filterred bird-pollinated fritillaries are also attractive to bees. The widespread purple and green colouration among Fritillaria, however, may potentially indicate some role of the environment in this trait. Clarifying the extent to which flower colour in Fritillaria is maintained by active selection pressures from pollinator preference and/or the environment remains an open question.