Angiosperms present enormous variation in flower traits, such as colour, shape, size, content and quality of the floral reward. This variation is also encountered among closely related species (e.g., Johnson et al., 1998; Wilson et al., 2004; Smith et al., 2007; Thomson and Wilson, 2008; Guzman et al., 2017; Roguz et al., 2018). Since Darwin (1862), flower divergence has largely been attributed to the selection of pollinators. The phenomenon of adaptation to the preferences of the most common and efficient pollinator has been formalised as pollination syndromes—recurring suites of flower traits associated with different groups of pollinators (Faegri and van der Pijl, 1966). Despite the ongoing debate about the assumptions of this concept (Ollerton, 1996; Waser et al., 1996; Rosas-Guerrero et al., 2014; Dellinger, 2020) it serves as a framework for studies of floral diversity (Kay and Schemske, 2003; Smith et al., 2007).

Several empirical studies have supported the relation between flower traits and pollinators groups (Dellinger, 2020). This relation is even more visible, when pollinators shifts are considered. These new pollinators may be more efficient (Ashworth et al., 2015), or the only present, if e.g., the occurrence of pollinator shift is enforced by the loss of ancestral pollinators (Cox and Elmqvist, 2000). Indeed, significant changes in floral traits, such as corolla symmetry, petal colour, and type of reward, are generally associated with pollinator shifts, strongly supporting the idea that flower traits reflect new pollinators preferences (Bruneau, 1997; Ramírez-Aguirre et al., 2019; Barrionuevo et al., 2021).

Among visual flower traits, colour is one of the key features involved in signalling to pollinators (Willmer, 2007). Colour constitutes to be one of the main traits used in floral syndromes (Faegri and van der Pijl, 1966), and although its attractiveness for specific pollinators groups remains a source of contention, the assumption that e.g., bird pollinated flowers are red remains common (Smith et al., 2007). In Penstemon Schmidel, for example, red petal colour was one of the characters that best predicted hummingbird visitation (Wilson et al., 2004). Flower colour change was also an important step involved in the transition from bee to nocturnal hawk moth pollination in the genus Petunia Juss (Hoballah et al., 2007). Flower colour, however, does not always relate well to pollinator shifts (Smith et al., 2007; Opedal, 2019; Roguz et al., 2020a). For example, red colouration may be a good predictor of bird-pollination in Iris L. species, but not vice-versa (Roguz et al., 2020a). The explanation for this lack of correlation is that ultimate flower choice is based on a combination of stimuli. Moreover, most flower species are pollinated by generalists (Reverté et al., 2016).

Other signal-transmitting flower traits, such as flower size and the arrangement of its reproductive parts, can also be subjected to strong selection pressure (Lavi and Sapir, 2015; Souto-Vilarós et al., 2018). For more efficient pollen transfer, flower size and arrangement of reproductive parts should fit the body and behaviour of the pollinator (Fenster et al., 2006). Studies conducted on Ipomopsis Michx. have shown that the display size was positively correlated with hummingbird importance and negatively correlated with dipteran importance (Smith et al., 2007). However, like in the case of colour, shift to the new pollinator does not always result in a consistent flower size or the arrangement of flower reproductive parts (Dupont et al., 2004).

While visual flower traits are an important signal for potential pollinators, floral rewards are also crucial in shaping the interaction. In most cases plants provide a food reward of either pollen or nectar (Parachnowitsch et al., 2018). In the latter case, animals visiting flowers may also exert selection on floral reward properties, such as volume, sugar, or amino acid (AA) concentration (Dupont et al., 2004). Flower reward chemistry was proposed as a key factor to transition from moth to bee pollination in Satyrium Sw. orchid (Castañeda-Zárate et al., 2021). Also in the case of Ipomopsis reward was positively correlated with hummingbirds importance. However, in Clivia Lindl. the shift from bird- to butterfly pollination was accompanied by the change of flower shape, smaller nectar volume, and emission of scent, while flower colour and nectar chemistry were not substantially modified (Kiepiel and Johnson, 2014).

The set of new flower traits depends on the identity of the new pollinator. The most often recorded pollinator shifts include the transition from insect (bee)-pollinated to bird-pollinated plants (Wilson et al., 2006; Thomson and Wilson, 2008), which is usually caused by the higher efficiency of birds in transferring pollen as well as lower pollen discounting (Thomson and Wilson, 2008). For example, in Penstemon and related genera, there have been at least 10 separate transitions toward hummingbird pollination (Wilson et al., 2007). In Mimulus sect. Erythranthe hummingbird pollination has arisen from melittophily twice (Beardsley et al., 2003). The following genera also include hummingbird-pollinated plants and species that are pollinated by other bird lineages: Agave, which is predominately bat-pollinated (Agavaceae, Good-Avila et al., 2006); Erythrina, which is entirely ornithogamous (Fabaceae; Bruneau, 1997; Etcheverry et al., 2012), Mucuna (Fabaceae; Kalin Arroyo and Penaloza, 1990), and Puya (Bromeliaceae, Hornung-Leoni and Sosa, 2005).

The above examples and other pollinator shift studies show that floral evolution has been highly labile and also directional. Moreover, even specialised pollination systems are not a dead-end (Tripp and Manos, 2008). There have been few reverse transitions from bird to insect pollination (Mast et al., 2012); for example, in the tribe Sinningieae (Gesneriaceae; two transitions from hummingbird to bee pollination, Perret et al., 2003) and in the genus Aquilegia (Ranunculaceae; five transitions from hummingbird to hawk moth pollination, Whittall and Hodges, 2007).

Such shifts to new pollinators, or reverse transitions to an ancestral pollinator, usually result in a different suite of traits, attracting new, more efficient pollinators and repelling less effective pollinators or flower robbers (Castellanos et al., 2004; Salas-Arcos et al., 2017). While flower colour-related traits, which often occur with pollinator shift, are well understood, little is known about the other flower traits, like nectar properties, that are the potential outcome of this shift. To understand the influence of the new pollinators on flower traits, it is necessary to explore plant genera exhibiting pollinator shift.

In our study, we focussed on Fritillaria L. (Liliaceae), which is a genus that includes approximately 150 species of bulbous plants, predominantly found in temperate Holarctic regions of both the Old and New World (Tamura, 1998; Day et al., 2014; Zych et al., 2014). The highest diversity of Fritillaria is observed in the Mediterranean region (Beetle, 1944; Rix, 1984; Zaharof, 1986; Rønsted et al., 2005; Tekñen and Aytaç, 2011; Hill, 2016; Kiani et al., 2017). Fritillaria flowers are generally actinomorphic and have a nodding tulip-like trimerous perianth. Despite the similarity, flowers of fritillaries exist in various sizes and colours, and can be white, pink, greenish, yellow, or purplish/reddish. We still do not fully understand the diversity of fritillaries, but it is possibly related to pollination systems, at least to some extent. Most fritillaries are described as or presumed to be pollinated by insects; however, there have been at least two pollination shifts from insect to bird pollination in the genus. This shift involves distinct bird groups, namely passerines and hummingbirds (Búrquez, 1989; Peters et al., 1995; Pendergrass and Robinson, 2005) and results in new flower traits in fritillaries. For example the only red and orange fritillaries are pollinated by birds. Moreover, properties of nectar reward reflect preferences of their bird pollinators. Hummingbird-pollinated fritillaries produce nectar of medium concentration and with medium AA concentration, while passerine bird pollinated species produce huge amounts of AA rich nectar, dominated by sucrose preferred by passerines. Still, most of the members of this genus, which are presumably insect pollinated, produce smaller amounts of balanced nectar (Roguz et al., 2018, 2019). Also within insect pollinated fritillaries, there may have occurred shifts of the main pollinators. Fritillaria camtschatcensis is a species with distinct flower traits and reward properties (Roguz et al., 2018, 2019), pollinated by flies (Zox and Gold, 2008). Nectaries of this species are covered with protrusions. Traces of hardly accessible, viscous and almost solid nectar are available as a thin film overlying the protrusions (Roguz et al., 2018, 2019). This way of nectar presentation likely excluded insects other than flies with a cushion-like labium.

Fritillaria is also a genus in which reversals from bird to insect pollination may have occurred. Fritillaria raddeana is a species closely related to passerine bird pollinated F. imperialis (Wietsma et al., 2015). However, flowers of F. raddeana are greenish, smaller and this species produces small amounts of highly concentrated, balanced nectar (Roguz et al., 2018, 2019). The question remains, what was the characteristic of the common ancestor of this species.

Given the unique pollination system (at least two transitions from insect to bird pollination, possible reversals), our research goals were to build a phylogenetic framework in which flower traits and reward properties within Fritillaria can be evaluated. We asked the following questions: to what extent do flower traits fulfil criteria of ornithophilous/enthomogamus flowers? Can quantitative traits, for example flower size, influence the diversification dynamic? Do shifts in phenotypic optima of flower traits overlap with the pollinator shift? We tested the assumption that nectar (as an important flower reward) and colour (as an important visual sign) play crucial roles in pollinator shits. We also anticipated reversals from hummingbird-bird to insect-pollination because of the low degree of specialisation of bird-pollinated fritillaries.

Most of the previously described large-scale phylogenetic relationships in Fritillaria were also found in our study (Rønsted et al., 2005; Day et al., 2014). The phylogenetic tree of fritillaries, based on five sequences, resolved the genus as monophyletic, which was also obtained in previous studies. However, subtrees inferred from plastid sequence data, also as in previous studies (Rønsted et al., 2005; Day et al., 2014), do not resolve Fritillaria as monophyletic, with the Lilium species nested within the studied genus.

The diversity of flower traits and pollination systems observed in fritillaries is likely due to several changes in the ecology and evolution of the studied genus. The most recent common ancestors of Fritillaria probably had purple or pink flowers of medium size, with the reproductive elements arranged in a similar manner to those of modern species. The ancestor likely had flowers with medium amounts of nectar rich in sugars, but with lower AA amount compared to modern species. We did not find phylogenetic signal for most of the studied traits, which indicate lack of tendency for closely related Fritillaria species to resemble one another. This result may suggest that several important shifts in floral characteristics e.g., flower colour in Fritillaria may be related to plant–pollinator interactions or other habitat-related factors. We were especially curious to study the new flower traits, e.g., reward properties in the context of new pollinators preferences. Our study revealed, for example, that flower colour is particularly good at separating bird- from insect-pollinated flowers. Colour shift may have been one of the triggers for bird pollination in Fritillaria. Red colour appeared among the modern, hummingbird-pollinated members of the subgenus Rhinopetalum. Similarly, in Asian species pollinated by passerine birds, orange flowers appeared. Red and orange flowers in Fritillaria are only found among bird-pollinated species, and in contrast to studies on Iochroma (Smith et al., 2007), Iris (Roguz et al., 2020a), and Polemonium (Landis et al., 2018), there seems to be a clear correlation between pollinator type and flower colour in the case of ornithophilous Fritillaria.

The two most common colour shifts were purple/green and purple/yellow, in both directions. We assume that these transitions types may have had little or no association with the influence of interacting animals. Green-coloured floral parts, common among Fritillaria species and most probably not visually attractive for pollinators (Roguz et al., 2020b), may be related to plant physiology. The photosynthetic activity of green flowers may be sufficient to maintain their own structures (Bazzaz and Carlson, 1979). Additionally, transition to a purple, anthocyanin-based colour, can result from environmental stress. Increased anthocyanin production may correlate with the level of sun damage or protection against herbivory (Chalker-Scott, 1999; Winkel-Shirley, 2002; Arista et al., 2013; Irwin et al., 2013). However, purple colouration may reduce visual attractiveness for insect pollinators, since several bee species have preference toward yellow flowers (Lunau and Maier, 1995; Spaethe et al., 2001). In some Fritillaria species, such as F. persica, both colour morphs (green and purple) are maintained, which is not typical. In most cases, one of the colour morphs will eventually be lost and the species will again be monomorphic (Ellis and Field, 2016). While gene flow usually leads to the fixation of one colour morph over the long term (Gray and McKinnon, 2007), some factors, including environmental heterogeneity (Tucić et al., 1998) and vegetative reproduction by rhizomes, may also maintain colour polymorphism in Fritillaria.

Tracking flower colour shifts in fritillaries suggests that colour loss in this genus may be reversible, in contrast to studies of other plant families (Smith and Goldberg, 2015; Landis et al., 2018). For example, the common ancestor of F. eduardii and F. sewerzowii most likely had orange flowers, while F. sewerzowii regains the purple colour present in the deeper nodes. The exception to this rule seems to be white flower colouration, which, similarly to Iris, is probably irreversible (Roguz et al., 2020a).

Other flower traits may also be related to plant–pollinator interactions in Fritillaria. Strong corolla constriction found in American hummingbird-pollinated fritillaries, and anthers and stigmas exerted beyond the corolla may promote contact between the birds’ bodies and the flower reproductive parts (Temeles et al., 2006; Martén-Rodríguez et al., 2009, 2010). A shift to bird pollination may also account for the increased size of the corolla entrance in the subgenus Petilium. These passerine bird-pollinated species have large flowers, growing close to the stem on a relatively short scape, with long stigma and anthers, distant from the tepals. These new traits may enable both to exploit the reward and make contact with anthers and stigmas. Nevertheless, similar to Dupont et al. (2004), we found no unique set of differences in the arrangement of flower reproductive parts between ornithophilous species and their entomophilous relatives. For example, long, exerted styles and stamens often associated with bird pollination (Beardsley et al., 2003), were found in both insect- and bird-pollinated Fritillaria species. These results may also suggest that some of the studied flower traits are just by change present in the bird-pollinated Fritillaria and do not result from pollinator driven trait evolution.

Flower size or arrangement of the reproductive elements often enable legitimate pollinators to access the reward. It’s properties seem to play an important role in the evolution of new pollination systems in fritillaries (Roguz et al., 2018, 2019). Our study revealed that the ability to produce AA-rich nectar with low sugar concentration evolved only once, in the case of the subgenus Petilium. These tendencies fulfil many of the criteria that are characteristic of passerine bird-pollinated flowers. On the other hand, the ability to produce both highly concentrated and/or copious nectar evolved several times, for example, in F. raddeana and F. recurva. The ability to produce highly concentrated nectar has never been lost, which may be related to the huge proportion of bee-pollinated fritillaries. The scarlet flowers of hummingbird pollinated species also produce reward attractive for its bird pollinators—copious amounts of relatively diluted nectar, with low concentrations of AA.

It is important to note that flower size and entrance, the arrangement of reproductive elements, or reward properties do not exclude insect pollinators in bird-pollinated fritillaries. Moreover, balanced proportions of different sugar types found in hummingbird-pollinated flowers may attract both insects and birds (Roguz et al., 2018). This scenario is also supported by the fact that F. eastwoodiae, which is a cross between F. micrantha (insect-pollinated) and F. recurva (bird-pollinated), must have arisen as a hybrid between two insect-visited species (Dewoody et al., 2013), since the flowers of F. micrantha are too small for other pollen vectors.

We also assume a pollinator shift within insect-pollinated species. In the case of fly-pollinated, typically purple-flowered F. camtschatcensis the reward properties (i.e., small amounts of highly concentrated nectar) and the petal structure (uneven surface, covered with numerous protrusions) may act as a trigger for pollinator shift. Flowers of this species, looking like and emitting the smell of rotting flesh, may attract new pollinators. On the other hand, the way of nectar presentation (thin film, almost solid) excludes most of the insect groups. Observed pollinator shifts in Fritillaria follow the predominant directionality in angiosperms, from insect to bird pollination (Wilson et al., 2007; Rausher, 2008; Van der Niet and Johnson, 2012). This transition can be explained by the more efficient transfer of pollen in bird-pollinated plants (Thomson and Wilson, 2008). However, we recorded potential for several reversals, from hummingbird- to insect-pollination, and from passerine bird- to insect-pollination. The most striking example is F. raddeana, a member of the subgenus Petilium. Flowers of this species resemble those of F. eduardii and F. imperialis but are half the size and pale-yellow-green in colour. The reward is also typical for insect-pollinated species, with a small volume of highly concentrated nectar, rich in sucrose and with low AA content (Roguz et al., 2018, 2019).

In addition to F. raddeana, it is likely that most of the floral diversity in Fritillaria is the result of adaptation to insect pollinators. Several species, even distantly related, for example F. pudica and F. carica, have similar flower traits, including colour, flower size, nectary shape (Rix and Strange, 2014), and reward properties. These similarities may reflect adaptation to pollinator preferences, especially taking into consideration species similarity and lack of phylogenetic signal. Insect-pollinated Fritillaria flowers are also of medium size, with anthers and stigmas that are often shorter than tepals, enabling small insects to touch the reproductive parts while foraging. Insect-pollinators often seek flowers with hexose-dominated nectar of medium volume and AA concentration, which are common in Fritillaria flowers (Roguz et al., 2018, 2019).

Finally, it is important to note the potential influence of other environmental factors on the evolution of Fritillaria flowers. Our study suggests that changes in Fritillaria flowers occurred rather recently, with more than 10 changes in the phenotypic optima of studied flower traits. Many flower traits were similar at the level of the deep internal nodes, whereas most changes in floral display and nectar properties appeared in shallow internal nodes. This evolutionary pattern and lack of phylogenetic signal for most traits suggest forces that act on an ecological time scale, rather than changes associated with deep phylogenetic relationships (Gómez et al., 2016). Fritillaries are found in a variety of climatic regions and in different habitats, including coasts, riparian zones, meadows, woodland, steppe, deserts, mountain screes, and alpine zones (Tomović et al., 2007; Tekñen and Aytaç, 2008; Zox and Gold, 2008; Hill, 2011, 2016; Tekñen and Aytaç, 2011; Rix and Strange, 2014; Zych et al., 2014; Gao et al., 2019). Consequently, some aspects of the floral display may have arisen by selection pressure exerted by abiotic factors related to the habitat type, such as temperature, altitudinal gradients, or water stress (Zhao and Wang, 2015; Landis et al., 2018; Gao et al., 2019). In addition, some of the observed variability should be attributed to the potential for natural hybridisation (Naruhashi et al., 2006; Kawano et al., 2008; Hill, 2011).

Further studies of Fritillaria flower traits and pollination systems in natural habitats, as well as molecular analyses of flower traits, would be of great importance. The results obtained in such studies may be crucial for understanding the influence of pollinators on the flower traits evolution.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ncbi.nlm.nih.gov/genbank/, MW081399 and https://www.ncbi.nlm.nih.gov/genbank/, MW081400.

KR and MZ conceived the study and wrote the draft version of the manuscript. KR and LH assembled field data. KR performed the nectar analysis. KR performed phylogenetic analyses. KR, MZ, AR, and LH analysed the data. All authors contributed to the final version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.