Genetic structure, UV‐vision, wing coloration and size coincide with colour polymorphism in Fabriciana adippe butterflies

Colour polymorphisms have long served as model systems in evolutionary studies and continue to inform about processes involved in the origin and dynamics of biodiversity. Modern sequencing tools allow for evaluating whether phenotypic differences between morphs reflect genetic differentiation rather than developmental plasticity, and for investigating whether polymorphisms represent intermediate stages of diversification towards speciation. We investigated phenotypic and genetic differentiation between two colour morphs of the butterfly Fabriciana adippe using a combination of ddRAD‐sequencing and comparisons of body size, colour patterns and optical properties of bright wing spots. The silvery‐spotted adippe form had larger and darker wings and reflected UV light, while the yellow cleodoxa form displayed more green scales and reflected very little UV, showcasing that they constitute distinct and alternative integrated phenotypes. Genomic analyses revealed genetic structuring according to source population, and to colour morph, suggesting that the phenotypic differentiation reflects evolutionary modifications. We report 17 outlier loci associated with colour morph, including ultraviolet‐sensitive visual pigment (UVRh1), which is associated with intraspecific communication and mate choice in butterflies. Together with the demonstration that the wings of the adippe (but essentially not the cleodoxa) morph reflect UV light, that UV reflectance is higher in females than males and that morphs differ in wing size, this suggests that these colour morphs might represent genetically integrated phenotypes, possibly adapted to different microhabitats. We propose that non‐random mating might contribute to the differentiation and maintenance of the polymorphism.

There is ample evidence that differences in animal colour patterns can directly contribute to variation in performance and lifetime reproductive success (i.e.relative fitness) of individuals, for example, by influencing susceptibility to predators, intraspecific signalling for mate choice, capacity for temperature regulation, protection from UV-radiation and resistance to desiccation and abrasion (Huey & Kingsolver, 1989;Ruxton et al., 2004;Wellenreuther et al., 2014).Because animal colour patterns have multifarious functions, a change in one component may induce selection that modifies the combination of size, shape, colour and position of other pattern components, and this may contribute to the integration and differentiation of multiple colour pattern elements (Fisher, 1930;Nijhout, 1991;Polic et al., 2023;Tsai et al., 2020).In many species, colour patterns are associated with body size, behaviours or other phenotypic traits (Ahnesjö & Forsman, 2003;Forsman et al., 2008;McKinnon & Pierotti, 2010;True, 2003).Such associations between traits, also known as phenotypic integration, may represent evolutionary responses to correlational selection that favours certain combinations of trait values over other combinations, that is, when the selective benefit of a given trait value depends on the value of another trait (Brodie, 1992;Dingemanse et al., 2020;Fisher, 1939;Forsman & Appelqvist, 1998;Lande & Arnold, 1983;Svensson et al., 2021).Their underpinnings include (but are not limited to) linkage disequilibrium, pleiotropy, close linkage and shared developmental pathways (Andersson, 2001;Falconer & Mackay, 1996;Fruciano et al., 2016;Svensson et al., 2021;True, 2003); however, they can also be a result of phenotypic plasticity (Schlichting & Wund, 2014).
Once established, alternative morphs and trait-value combinations may be maintained by the evolution of assortative mating, which can prevent the disruption and break-up of co-adapted gene complexes and favourable trait-value combinations (Jiang et al., 2013;Lancaster et al., 2014;Tamin & Doligez, 2022).As colour patterns play an important role for visual signalling and mate choice in many species, assortative mating according to colour morph or selection against offspring from mixed parents might contribute to continued differentiation between colour morphs (Gray & McKinnon, 2007).This can ultimately lead to reproductive isolation with respect to colour morph, as previously reported in visually oriented animals such as butterflies, birds and fishes (e.g.Kronforst et al., 2006;Roulin, 2004;Whitney et al., 2018).
Despite the considerable and growing scientific interest that colour polymorphisms attract (see figure 1 in Forsman, 2016), there remain open questions awaiting to be answered.For example, our understanding of the drivers, evolutionary dynamics and consequences of colour polymorphisms is potentially complicated by that discrete variations in colour patterns do not always have an underlying genetic basis; they can also result from developmental plasticity (polyphenism) when a single genotype produces alternative phenotypes in response to some environmental cue (Pfennig et al., 2010;Schwander & Leimar, 2011;Whitman & Agrawal, 2009).Similarly, associations between different colour pattern elements, and associations of colour pattern with other traits need not reflect underlying genetic correlations, they can also result from plasticity integration (Pfennig et al., 2010;Pigliucci & Preston, 2004;Whitman & Agrawal, 2009).
The effects of selection and reproductive isolation, the evolutionary modifications and divergence of integrated phenotypes, and the consequences for population performance all depend on whether colour morphs have a genetic underpinning (Garner et al., 2020;Hughes et al., 2008;Lande, 1998;Wennersten & Forsman, 2012).
The recent developments in DNA sequencing tools have opened new possibilities to further the understanding of the above issues, and more.For instance, high-throughput technologies such as double-digested restriction-site-associated DNA sequencing (ddRADseq) can generate thousands of single nucleotide polymorphism (SNP) loci randomly distributed across both coding (functional) and non-coding regions of the genome (Andrews et al., 2016).This makes ddRADseq a powerful tool to analyse genetic diversity and structure, genetic associations with phenotypic traits such as colour morphs, and to identify genomic regions potentially under selection (Andrews et al., 2016;Peterson et al., 2012).
The butterfly species Fabriciana adippe is well-suited for examining whether sympatric colour morphs are genetically manifested, whether colour morphs are phenotypically and genetically correlated with other traits, and whether the dynamics and maintenance of the polymorphism is influenced by selective processes such as assortative mating (Lank, 2002;Pérez i de Lanuza et al., 2013).Throughout the distribution area, the High Brown Fritillary occurs in several discrete colour morphs, two of which are present in our study area in Sweden (Figure 1), the nominate form F. adippe f. adippe with silvery, iridescent mother-of-pearl like spots on the ventral hindwings, and F. adippe f. cleodoxa, which instead has light yellow spots (Eliasson et al., 2005;Tolman & Lewington, 2012).In butterflies, reflective silvery patterns may enhance camouflage and thereby protect against predators, and may be involved in intraspecific signalling during mate choice, particularly in habitats with complex light conditions such as forest habitats (Wilts et al., 2013), but this has not yet been confirmed for F. adippe.To our knowledge, the colour polymorphism in F. adippe has not been well characterized, potential multivariate phenotypic integration within and differentiation between the adippe and the cleodoxa morph has not been evaluated, and the genetic background of the colour polymorphism and associated traits has not been previously investigated in this species.
In the present study of colour polymorphism in F. adippe, we integrate phenotypic and genetic sequencing data for butterflies collected from one mainland and two insular natural populations in Sweden.Specifically, we (i) characterize the wing colour pattern polymorphism and quantify phenotypic integration and differentiation by examining whether the adippe and cleodoxa colour morphs are phenotypically and genetically correlated with other traits (viz. forewing length, degree of melanism, the amount of green and orange wing scales, and optical properties (i.e.reflectance of UV and longer wavelengths (390-700 nm)) of the bright spots).As colour polymorphisms are often associated with other phenotypic (e.g.morphological) traits in many organisms (see above), we expect some covariation between morphological characteristics in the respective colour morphs.If phenotypic traits covary with genetic structuring, this might suggest that these trait combinations have a genetic underpinning, rather than stemming from plasticity responses.We also (ii) assess whether the colour polymorphism is

| Study species
Fabriciana adippe occurs in large parts of the Palearctic and is widely distributed in Scandinavia, ranging as far north as central Sweden (Eliasson et al., 2005, Tolman & Lewington, 2012).While both colour morphs can occur in the same area, the nominate form F. adippe f. adippe is progressively replaced by the yellow form F. adippe f. cleodoxa southwards through central Europe to the Balkans and Greece, where the latter morph is dominating (Tolman & Lewington, 2012).In their northern distribution area, F. adippe butterflies breed in bracken (Pteridium aquilinum) dominated habitats within woodland landscapes, grass and P. aquilinum mosaics, as well as in glades, clear-cuttings and meadows in forest dominated landscapes (Eliasson et al., 2005;Warren, 1995).The univoltine species usually lays its eggs, which constitute the overwintering life stage, on or in close proximity to its larval food plants, viz.Viola species (Eliasson et al., 2005).

| Study area and data collection
We collected butterflies belonging to both colour morphs F. adippe f. adippe and F. adippe f. cleodoxa from three Swedish populations, viz.Småland, and the islands Öland and Gotland in summer 2017 (see map in Figure 1).From this collection, we selected 18 (14 adippe, 4 cleodoxa), 39 (23 adippe, 16 cleodoxa) and 29 (25 adippe, 4 cleodoxa) individuals from Småland, Öland, and Gotland, respectively, such that both sexes were represented in each population and colour morph (Table 1).Butterflies were either directly transferred to 95-99% ethanol or dried and kept at −20°C in the laboratory at the Department of Biology and Environmental Science, Linnaeus University in Kalmar until DNA extraction.

| Quantifying body size and wing colour patterns
To obtain data on body size and wing colour patterns, each butterfly was photographed in front of standard white paper with a Nikon D3000 camera using an 18-55 mm lens without flash, with the left ventral fore-and hindwing and the right dorsal fore-and hindwing (i.e.four parts) facing the camera, together with a ruler (Polic et al., 2023).The camera was placed in a fixed position, that is, 34 cm from the wings.Each picture was taken with the same settings (focal length 35 mm, ISO 400, shutter speed 1/200 s, aperture f/4.5) and light conditions, in the butterfly laboratory at Institut de Biologia Evolutiva in Barcelona, Spain.We used the software GIMP 2 to measure the wing size and colours.Using the ruler tool, we measured forewing (and hindwing) length as the straight linear distance from the apex of the forewing (and hindwing) to its base.In each of the 4 wing parts (left ventral fore-and hindwing, right dorsal fore-and hindwing), we used the "select by colour" tool to select and measure the number of pixels of a certain colour, viz."light brown" (#92580f), "dark brown" (#4a3822), "black" (#0a0907), "light orange" (#c88d00), "dark orange" (#a55400), "white" (#f2f4e7), "grey" (#b5b498), "yellow" (#c5a035), "light green" (#7b7f20), and "dark green" (#565d17) with a threshold of 15 (Polic et al., 2023).Then, we calculated the percentage of each colour in each wing part.Dark brown and black pixels were added up and defined as the degree of melanism in each wing, which is also referred to as the degree of darkness in the text.The added percentages of dark and light green in the ventral hindwing (only present in this wing part) were defined as the variable "green".
Hereafter, "orange" refers to the percentage of orange in all 4 wing parts.For the statistical analyses of phenotypic integration and composite phenotypic differentiation between colour morphs, we used data on wing size (forewing length, hindwing length) and colour patterns of the wings (forewing melanism, hindwing melanism, green and orange).
For the statistical analyses used to evaluate phenotypic associations with the genetic structure (described below), we used dorsal melanism as a measure of darkness and forewing length as a measure of wing size.Dorsal melanism was used since butterflies expose their dorsal wings when sun basking to increase their body temperature (Clench, 1966;Kingsolver, 1985).Forewing length is commonly used as a proxy for body size in butterflies and is highly correlated with wing area, which is important for temperature regulation (Kingsolver & Koehl, 1985;Stuhldreher et al., 2014) and flight capacity (Berwaerts et al., 2002;Kalarus et al., 2013;Skórka et al., 2013).

Note:
The right side of the table shows the mean wing size (measured as forewing length) in mm, the mean summed percentage of dark brown and black scales in the dorsal wings ("Melanism"), the mean percentage of green scales in the ventral hindwings ("Green"), and the mean percentage of orange scales in all four parts of the wings ("Orange").Standard deviations are shown in brackets.
Individuals were classified according to colour morph based on the presence of the silvery or yellow spots, respectively, on the ventral hindwings (Figure 1, Table S1).

| Quantifying optical properties of the bright wing spots
For a straightforward approach of comparing the optical properties of the bright wing spots between individuals, we chose two of the spots (silvery or yellow respectively) on the ventral side of the right hind wing (Figure 1), and used a JAZ spectrophotometer with a JAZ-PX light source (both Ocean Optics, Dunedin, FL) and a small-tip reflection probe (Avantes, the Netherlands) for the measurement of spectral reflectance of the spots.All measurements were conducted relative to a white, diffuse reflectance standard.
Because the measured spots of the adippe morph were iridescent, their reflectance spectrum and intensity vary with the angle of light and viewing.We therefore used five individuals for an initial approximation of the angle of the probe, yielding highest maximal reflectance of the measured spots.Based on these measurements, we used a fixed 75° angle to the plane of the wing for the final measurements.For each measurement, we adjusted the rotational (horizontal) angle by starting at a perpendicular angle to the wing base and the longitudinal axis of the body, and then slightly rotated the measured wing clockwise and anti-clockwise, to find the angle that yielded the highest reflectance.This was repeated three times for each measured spot.In the analysis of reflectance, we used the highest value of the three repeated measurements for each spot (Table S1).
Specifically, we aimed at assessing the relative intensity of the UV component in the bright spots.To our knowledge, neither the spectral sensitivity of UVRh1 nor the maximum wavelength sensitivity for F. adippe has been directly tested.Studies of other butterflies report that the UV-sensitive photoreceptor (dark purple) encoded by the UVRh1 gene covers the band from around 325 to 385 nm, with a peak in sensitivity around 355 nm, and that the visual sensitivity interval extends to around 650 nm, albeit with some variation between species and sexes (Briscoe et al., 2010;Bybee et al., 2012;Finkbeiner & Briscoe, 2021;Ogawa et al., 2013).Although F. adippe may differ from other species, we set the range of UV sensitivity of UVRh1 from 325 to 385, and the maximum wavelength sensitivity to 700 nm.For each individual and spot, we then estimated the summed reflectance within the UV (325-385 nm) and within longer wavelengths (hereafter LW, 390-700 nm), and then calculated the ratio of UV/LW reflectance to estimate the salience of the UV component.

| Statistical evaluation of phenotypic integration and composite phenotypic differentiation between colour morphs
To evaluate and compare phenotypic integration between morphs, we first analysed pairwise associations between the two wing size measures and the four wing colour pattern variables (see above) using Spearman correlation analysis.This was done separately for the two morphs.Next, to evaluate whether the integration between different phenotypic traits is parallel or independent in the two colour morphs, we tested whether pairs of phenotypic traits that were highly positively or weakly negatively correlated in one morph were also highly positively or weakly negatively correlated in the other morph, using a correlation analysis.Finally, to evaluate whether the strength of phenotypic integration was equal or different in the two morphs, we tested for a difference in the absolute value of the 15 pairwise trait correlations using a paired t-test.
To evaluate whether overall composite phenotypic variation differed between colour morphs, after controlling for variation due to source population and sex, we first used a principal component analysis (PCA, a dimension reducing approach) to describe the total phenotypic variation in the six original phenotypic traits in two dimensions (principal component axes, PCA1 and PCA2).This was done for all three populations pooled, with procedure PRINCOMP in SAS 9.4 (Windows version 1.0.19041), and the results visualized in a biplot, to evaluate whether composite phenotypic variation overlapped or differed between morphs.Next, to formally evaluate whether the composite phenotypic variation differed markedly according to colour morph after statistically accounting for variation due to population and sex, we performed a multivariate analysis of variance (MANOVA, implemented with procedure GLM in SAS 9.4) using PC1 and PC2 as dependent variables, and population, sex and colour morph as explanatory factors.To enable visualization of the average composite differentiation, we estimated least squares means and associated 95% confidence limits of PC1 and PC2 for the two morphs, from the MANOVA.
In addition to the multivariate analyses, we tested for differentiation between colour morphs by performing separate 3-way analyses of variance for the different traits.These analyses were implemented with procedure GLM in SAS 9.4.Colour morph, source population and sex were treated as discrete explanatory variables, and forewing length, hindwing length, dorsal melanism, green, and orange were treated as response variables (the two latter variables were square root transformed to achieve normal distribution).First, we tested for both main and interaction effects of colour morph, source population and sex on the phenotypic traits.Non-significant interaction effects were removed stepwise from the models.
To identify sources of variation in the optical properties of wing spots, differences in summed reflectance between colour morphs, sexes and populations were evaluated using mixed model ANOVAs, implemented with procedure MIXED in SAS 9.4 (Windows version 1.0.19041), with individual identity included as a random factor to account for repeated measures within individuals.2018) with some modifications.For details, see Library preparation and ddRAD-sequencing in Appendix S1.We used the process_radtags unit in Stacks 2.2 Catchen et al., 2011Catchen et al., , 2013) ) to de-multiplex the raw data and for quality control, and the integrated approach (Catchen et al., 2011(Catchen et al., , 2013;;Paris et al., 2017) for assembly of ddRADtag loci and SNP calling.The recently published Fabriciana adippe genome (GenBank assembly GCA_905404265.1) was used for aligning the catalogue.For details, see Quality control and data trimming in Appendix S1.The final catalogue used for downstream analyses contained 6401 loci.

| Wolbachia testing to evaluate the potential role of parasite genetic contamination
The sequences obtained after populations in Stacks were mapped to the Wolbachia pipientis genome (GenBank: NZ_JQAM01000001) using Geneious 11.1.05(Biomatters) to evaluate the risk of contamination by the bacterial parasite Wolbachia, which could influence the outcome of the population genetic analyses (Werren et al., 2008).As none of the catalogue loci could be mapped to Wolbachia, we concluded that Wolbachia contamination would not affect the outcome of our analyses to a significant extent.

| Analysing genetic structure
We used multiple approaches with different algorithms to investigate the genetic structure among colour morphs and populations for the purpose of obtaining a more comprehensive picture.
We assessed phylogenetic relationships between individuals using the maximum likelihood (ML) tree inference in IQ-TREE 2.1.2(Nguyen et al., 2015) and used Interactive Tree Of Life (iTOL) 6 for visualization (Letunic & Bork, 2019).
2.9 | Evaluating associations of genetic structure with population, sex, colour morph and phenotypic traits Since we found that discrete colour morphs also differed significantly in other traits (see Section 3.1), we tested for associations of the genetic structure with phenotypic variables (forewing length, dorsal melanism, the percentage of green on the ventral hindwing and the total percentage of orange in the wings), colour morph (i.e. adippe with silvery-white spots or cleodoxa with yellow spots), sex and source population by utilizing a distance-based redundancy analysis (db-RDA), which constitutes a constrained version of PCA (Legendre & Anderson, 1999).The same distance matrix as for PCA was converted into a Manhattan dissimilarity matrix and used for principal coordinates analysis (PCoA).The capscale function in the R package VEGAN 2.5 (Oksanen et al., 2019) was used to build db-RDA models, which assessed the variables contributing the most to the PCoA ordination.The function anova determined the statistical significance of the explanatory variables using 999 permutations.
We also performed separate db-RDA for each population (Småland, Öland, Gotland) to assess associations of the genetic structure with the above-mentioned variables on a higher geographic resolution, and to explore whether the pattern of overall genetic differentiation between colour morphs was consistent across populations.When building the db-RDA models, we calculated the variance inflation factor (VIF) for each predictor variable, and excluded variables with a VIF > 10 from the model, as such high values might indicate a problematic level of collinearity with other variables (James et al., 2013).
Since our phenotypic variables are dimensionally heterogeneous, we standardized them using z-scoring; for each variable, we subtracted the mean value from the observed value and then divided this by the standard deviation.

| Outlier analysis to identify genes associated with colour morph
To identify genes putatively under selection associated with colour morph identity, we searched for outliers correlated with colour morph using two different approaches: BayeScEnv 1.1 (de Villemereuil & Gaggiotti, 2015), which is an F ST -based genome-scan method, and a coalescent approach, Fdist in Arlequin (Excoffier & Lischer, 2010).For details, see Outlier analysis in Appendix S1.
Manhattan plots were generated in R 3.6.1 (R Core Team, 2020) using the p values from these analyses.To assess the potential functional roles of the outlier loci, we interrogated the outlier sequences against coding sequences from all available Lepidopteran reference genomes in Lepbase 4 (Challi et al., 2016) using the BLAST interface (Priyam et al., 2019).We calculated allele frequencies in each population for the outlier loci that mapped to coding regions.
We also checked the protein-coding genes residing between the two SNPs flanking each outlier locus in Ensembl genome browser 110 (genome release GCA_905404265.1), and annotation was conducted with Ensembl Genebuild method by Lohse et al. (2022).
Inbreeding coefficient (F IS ) and violations of Hardy-Weinberg equilibrium (HWE) at these loci were calculated in the populations unit in Stacks.
We also searched for outlier loci associated with sex using BayeScEnv by assigning −1 and 1 as environmental variables to females and males respectively.

| Phenotypic integration and differentiation between colour morphs
The analyses of phenotypic integration based on pairwise correlations between different traits showed that forewing size and hindwing size were positively correlated in both colour morphs, that the amount of green was negatively correlated with hindwing melanism in the adippe morph, and that forewing melanism tended to decrease with forewing size in the adippe morph (Table S2).The analyses of pairwise correlations further showed that phenotypic integration between the different wing size and colour pattern traits was not parallel in the two colour morphs (r = 0.47, n = 15, p = .08),and that trait correlations were stronger overall in cleodoxa than in adippe (paired t-test based on absolute correlation coefficients, t = 2.43, n = 15, p = .029,Table S2).
The quantification of composite phenotypic variation as described by the first two principal component scores (Prin1 and Prin2) accounted for 51% of the total variance in the two wing size traits and the four colour pattern traits (Table S3).Composite phenotypic variation, as quantified by Prin1 and Prin2, differed significantly between colour morphs, sexes and populations (MANOVA, colour morph: Wilks' lambda = 0.84, F 2,71 = 6.61, p = .0023;sex: Wilks' lambda = 0.51, F 2,71 = 18.10, p < .0001;population: Wilks' lambda = 0.84, F 4142 = 4.06, p = .0038,Figure 2).On average, individuals belonging to the adippe morph had larger wings, a higher degree of melanism (darker coloration) and a lower proportion of green scales on the underside of the hindwing compared with cleodoxa, whereas individuals belonging to the cleodoxa morph had paler (less melanic) hindwings, a higher proportion of green scales on the underside of the hindwings and a higher percentage of orange summed across all four wing parts compared with adippe (Figure S1).The results from separate univariate analyses of variation in the different traits according to colour morph, sex and population (reported in Table S4) support that colour morphs differed in darkness (degree of melanism) and the proportion of green scales on the underside of the hindwings.

| Optical properties of the bright wing spots
The optical properties (spectral reflectance) of the two measured wing spots varied according to colour morph, sex and population (Figure 3).Overall, reflectance in the UV interval was about 15 times higher in adippe than in cleodoxa individuals when controlling for the variation owing to sex and population reported in   S2 and  S3).Differences between colour morphs are statistically significant (Prin1: F 1,72 = 4.56, p = .036;Prin2: F 1,72 = 7.72, p = .007,see text for further details).
reflectance in the UV, but the magnitude of the UV reflectance and the relative intensity of the UV component was two to seven times higher in the adippe morph (Figure 3).The above variation among populations was paralleled in the adippe morph, with individuals from Småland reflecting more in the UV band compared with individuals from Öland and Gotland, in both males and females (Figure 3).
According to results from the outlier analyses (for details, see Section 3.6 below, and Table 3), the differences between populations in reflectance in the UV band coincided with a molecular difference between populations in the frequency of the UVRh1 re-  Note: Reflectance was quantified for two silvery iridescent spots on the ventral hind wings in the adippe morph and the corresponding yellow spots in cleodoxa morph in males and females from three populations.Results from mixed model ANOVAs, implemented with procedure MIXED in SAS, treating individual identity as a random factor.The covariance parameter estimate associated with the random effect of individual identity pointed to individual variation in UV reflectance also within populations, sexes, and colour morphs (UV: Z = 1.47, p = .070;ratio UV/LW: Z = 1.42, p = .078).

TA B L E 2 Sources of variation in
spectral reflectance in the UV (summed over the 325-385 nm interval) and in the ratio of summed reflectance in the UV and longer wavelengths (LW, 390-700 nm) in Fabriciana adippe butterflies.

TA B L E 3
The position of outlier loci, the adjacent SNPs at both 5′ and 3′ ends, and the protein-coding genes residing at these loci.Lepbase 4 (Challi et al., 2016).The protein-coding genes were searched in Ensembl website (genome release GCA_905404265.1).

TA B L E 3 (Continued)
Öland belonged to the A/A genotype (Tables 4 and 5).In the sample (n = 22) used to quantify reflectance, only the four male adippe individuals from Öland and Gotland, with relatively low UV reflectance (Figure 3), belonged to the A/A genotype, whereas all others, including the six individuals from Småland (both males and females and regardless of colour morph) belonged to the T/T genotype.

| Genetic structure
All populations were significantly differentiated from each other according to F ST values (Gotland -Öland: 0.023, Gotland -Småland: 0.018, Öland -Småland: 0.013, p < .001for all comparisons).The F ST values for all pairwise comparisons between colour morphs within and among populations ranged up to 0.037.Despite the low F ST values, the majority of comparisons showed significant differentiations (p < .05, Figure 4).The genetic differentiation between colour morphs within populations was highly significant in Öland (F ST = 0.012, p < .001)and Småland (F ST = 0.023, p = .012),but was not statistically significant in Gotland (F ST = 0.000, p = .99).Notice that the sample size for Gotland-cleodoxa was the lowest (n = 4), which likely decreased the power of the permutation tests.However, population sample size can be significantly reduced, as long as a large enough number of bi-allelic genetic markers is used (>1000), which is the case in the present study (Willing et al., 2012).
According to the results generated by PCA, individuals belonging to the same colour morph largely clustered together (Figure 4b), but the structuring according to source population was manifested  cross-validation with leave-one-out method showed that the assignment success of the different morphs was 83.9% (73/87) for the full dataset, and 84.6% (33/39) for Öland (Table S5).We used only the Öland population separately for CAP analyses as it had a more balanced sampling for the respective colour morphs.For detailed results of PERMANOVA, PERMDISP and CAP, please see Results from PERMANOVA, PERMDISP and CAP in Appendix S1.

| Phylogenetic analysis
The phylogenetic (ML) tree revealed similarities in the genetic structure with results from F ST and PERMANOVA analyses.Clades were primarily associated with sampling population, and an additional level of clustering was largely in line with colour morph identity within the populations, and this was least apparent on Gotland (Figure 5).

| Associations of phenotypic variables and the genetic structure
Performing db-RDA, the final selected models included colour morph, sex, source population (when analysing all three populations together), dorsal melanism, forewing length, the percentage of green in the ventral hindwing and the percentage of orange in all wings.

| Associations of colour morph with loci putatively under selection
Outlier analyses with Fdist and BayeScEnv suggested 116 (p < .01, Figure 7a) and 19 loci (p < .05, Figure 7b), respectively, associated with colour morph.Seventeen of these loci were identified using both methods.The overlapping loci were lying within the coding regions of one uncharacterized gene, one hypothetical gene, the peptidyl-tRNA hydrolase 2 (PTRH2, a role in a functional protein synthesis in regulating cell survival and apoptosis (Sharkia et al., 2023)), the peptidyl-prolyl cis-trans isomerase-like 2 (PPIL2, a role in spliceosome activity (Bai et al., 2021)), the ribosome-releasing factor 2 (RRF2, involved in the ribosome's inherent catalytic activity in specific circumstances (Rodnina, 2013), and the ultraviolet-sensitive visual pigment gene; UVRh1, responsible for UV-vision in butterflies; Briscoe et al., 2010;Bybee et al., 2012) respectively.The remaining 11 loci could not be matched to any gene.We examined the protein-coding genes residing between the two SNPs flanking each outlier locus in Ensembl genome browser 110.There were 229 protein-coding genes in total at these candidate loci.The locus information for the outlier loci and the flanking SNPs, as well as the ID of the proteincoding genes can be found in Table 3. Please note that the F. adippe genome is not characterized, and protein-coding regions were annotated based on models.Moreover, the number of candidate genes is most probably higher than the actual number because none of the SNPs flanking the candidate outlier loci were themselves outliers, and the actual candidate loci should be much shorter in length.A denser genotyping could help identify the boundaries of the outlier loci.
We calculated the allele frequencies and assessed the genotype of individuals in each population at the outlier locus in UVRh1 due to its potential role in colour perception.The allele distribution for UVRh1 was different among the populations; the major allele frequency (MAF) was 0.52, 0.61 and 0.94 in Gotland, Öland and Småland, respectively (Table 4).Inbreeding coefficient (F IS ) values showed that all populations were significantly heterozygote deficient at the locus (Gotland: F IS = 1, p < .001;Öland: F IS = 0.81, p < .001;Småland F IS = 1, p = .028).We determined the genotype of the individuals for which the data was available.
While the adippe morph had all three genotypes (T/T, A/A, T/A) at UVRh1, all cleodoxa individuals belonged to genotype T/T.In other words, MAF for individuals representing the cleodoxa morph was 1 regardless of their source population whereas the MAF for the adippe morph varied according to the sampling location (0.40 ≤ MAF ≤0.93, Table 5).
BayeScEnv analysis did not suggest any association of outlier loci with sex.

| DISCUSS ION
The morphological and genomic analyses of three Swedish F. adippe populations revealed that the adippe morph was significantly larger, darker and had fewer green scales than the cleodoxa morph.The comparison of optical properties of the bright spots also provides evidence that UV reflectance of wing spots is about 15 times higher overall in adippe than cleodoxa, and about two times higher in female than in male adippe.Together, these findings suggest that the differ- although we cannot rule out the possible role of plasticity integration.We also identified 17 outlier loci putatively under selection, including UVRh1, which were correlated with colour morph identity.
Collectively, these results indicate that the phenotypic differentiation between colour morphs is likely adaptive and coincides with an underlying genomic differentiation that appears to have been driven by selection, although we cannot rule out the potential role of neutral processes, such as genetic drift.

| Colour morphs represent alternative integrated phenotypes
That the adippe and cleodoxa colour morphs appear to represent alternative phenotypes with distinct trait value combinations raises the question which functions and selective factors may have contributed to their evolutionary divergence.The evolution of phenotypic integration and trait associations can be driven by correlational selection, such as when susceptibility to visually oriented predators is reduced by specific combinations of for example prey colour pattern and avoidance behaviour (Brodie, 1992;Dingemanse et al., 2020;Forsman & Appelqvist, 1998), prey colour pattern and body size (Hagman & Forsman, 2003;Karpestam et al., 2014), or when the utilization of a certain habitat or niche favours specific associations of two or more phenotypic characteristics (Forsman et al., 2002;McKinnon & Pierotti, 2010).In F.
adippe, the silvery adippe morph is likely better protected against predators in habitats with complex light conditions and visual backgrounds, such as forest glades, or clear-cut areas in woodland landscapes (Wilts et al., 2013).When at rest, the ventral wings are exposed, such that the iridescent silver scales can act as a mirror to the foliage, thus possibly enhancing camouflage and making the butterfly harder to detect for visually oriented predators (Kjernsmo et al., 2020).The specular reflection of light by the mother-of-pearl-like spots while flapping the wings might also complicate pursuit and capture of the butterfly in the interplay of light and shade.Conversely, such specular reflection might attract attention in open and visually more homogeneous grasslands, where the yellow cleodoxa morph with more green scales on the ventral hindwings might be better protected against detection and predation, as in the Green Hairstreak butterfly Callophrys rubi (Michielsen et al., 2010).However, since we did not specifically test for differences in predation avoidance between colour morphs in distinct microhabitats, our interpretations remain speculative.
One way to evaluate the roles of predation would be to perform systematic detection and predation experiments, and to compare the escape behaviour and capture rates of the two morphs in contrasting environments.
That the silvery adippe form had both larger and darker wings might be another adaptation to shaded forest habitats, as this trait combination allows for more efficient absorption of sun light where the opportunities for sun-basking are limited (Bishop et al., 2016;Schweiger & Beierkuhnlein, 2016).A difference in wing size between colour morphs in F. adippe was also apparent in a study based on analyses of phenotypic data collected across a large spatial scale (>32° latitude and >47° longitude), with individuals belonging to the adippe morph having significantly larger wings than cleodoxa (see

| Genetic differentiation between colour morphs
In principle, the type of alternative integrated phenotypes indicated by our results may represent the outcome of plasticity integration, which stems from developmental responses to environmental conditions experienced during development and growth (Pfennig et al., 2010;Whitman & Agrawal, 2009).However, that our results show signs of genetic differentiation associated with morphological dimensions instead points to a role of pleiotropic effects, supergenes, or linkage disequilibrium (McKinnon & Pierotti, 2010).
Although a structuring according to colour morph was apparent, the structuring according to source population was more obvious.
Please note that since the studied populations are geographically relatively isolated from each other, the genetic signature of colour morph might be masked by a stronger background genomic signal associated with source population.However, the genetic differentiation between colour morphs within each population remained apparent (Figures 4-6, Figure S3).In parallel with the results from our analyses of phenotypic variation, db-RDA demonstrated that the genetic structure was associated not only with the two distinct morphs but also with forewing length (Öland only).In the mimetic butterfly Hypolimnas misippus, from the same family as F. adippe (Nymphalidae), distinct colour pattern variation is correlated with differences in body size, presumably owing to a supergene linked to several loci (Gordon & Smith, 1998).A recent study on the polymorphic H. numata found a supergene responsible for several independent loci associated with distinct wing pattern characteristics (Jay et al., 2022).An association in the grasshopper species Tetrix undulata of maternal colour morph with offspring body size variations, regardless of differences in environmental conditions during rearing, also suggests a genetic link between these traits (Ahnesjö Forsman, 2003).
Although we were unable to test for genetic correlations, as the genome for F. adippe has not been characterized, it is possible that the genomic regions associated with colour morph identity and wing size are linked.Both pleiotropy and linkage disequilibrium could be responsible for patterns where traits do not vary independently of one another (Falconer & Mackay, 1996;Wright, 1984).A stable maintenance of genetic correlations might indicate pleiotropic effects, while physically linked loci might be expected to break up or reduce the associations between characteristics within a few generations due to recombination and segregation, although linkage equilibrium is expected to be reached slowly for linked genes (Conner, 2002).However, strong selection or tight linkage due to physical proximity can prevent allele separation via recombination (Li & Nei, 1974).Linkage disequilibrium can also be maintained in deep evolutionary time through supergene formation (Fisher, 1939;Ford, 1965;Küpper et al., 2016;Nabours, 1929;Thompson & Jiggins, 2014).That the associations of colour morph with melanism, wing size, the amount of green scales and the presence of iridescent structures that reflect UVlight were evident across all studied populations, despite that populations were genetically distinct from each other, might indicate that trait correlations are rather stable and likely have existed since before populations spread from the Swedish mainland to the islands.That the majority of outlier loci, including UVRh1, reside in separate chromosomes might suggest some pleiotropic involvement in these integrated phenotypes.Please note that some outliers lie on the same chromosomes, which suggests that there is some physical linkage between these SNPs, at least at the chromosomal level.However, since we are unable to identify the main loci responsible for the colour polymorphism (silvery adippe vs. yellow cleodoxa) and many more genes might contribute to the aforementioned phenotypes, these genes could be in linkage disequilibrium with the detected outlier loci.Although we can only speculate about the exact genetic architecture, we suggest that linkage disequilibrium likely underlies the observed trait value combinations, as our results point to that many genes are involved in the distinction of colour morphs (e.g.overall genetic differentiation between morphs).

| Prevalence and mediators of genetic differentiation between colour morphs
Our study provides rare empirical evidence for an overall genetic divergence that is correlated with phenotypic (wing size, colour pattern, and UV reflectance) variation in the High Brown Fritillary.
While several other studies discovered genetic divergences that coincided with colour pattern variation in other species, our study stands out as it shows a general genetic difference between colour morphs occurring in the same population.For example, the wood tiger moth Parasemia plantaginis exhibits geographical differences in hindwing warning coloration, with high levels of differentiation between populations at contrasting margins of the species' distribution range, and a weaker genetic divergence coinciding with hindwing coloration (Hegna et al., 2015).However, the tiger moth study covered a much larger spatial scale, and colour variation was highly correlated with geographic location.Similarly, a study on the polymorphic ballan wrasse Labrus bergylta revealed an overall genetic structuring along a latitudinal gradient that coincides with morph frequencies (Casas et al., 2021).In the highly polymorphic strawberry poison-dart frog Dendrobates pumilio, the genetic distance between populations is associated with phenotypic differences rather than with geographic distance, however, divergent phenotypes are allopatric in this species (Wang & Summers, 2010).Similar to the findings in our present study of butterflies, studies of genetic structure of marine taxa with sympatric distinct phenotypes, for example, in the arc-eye hawkfish Paracirrhites arcatus, the sea urchin Paracentrotus gaimardi or the sea anemone Parazoanthus axinellae, report on genomic associations with colour morphs, however, they used fewer nuclear and mitochondrial genomic regions (Calderon et al., 2010;Villamor et al., 2020;Whitney et al., 2018).
The existence of distinct and (partially) genetically diverged colour morphs in a species has previously been interpreted as a sign of reproductive isolation (e.g.Boratynski et al., 2014;Casas et al., 2021), which can be mediated through assortative mating, or as a result of spatial sorting (Van Belleghem et al., 2016).If morphs utilize different habitats, assortative mating may occur as a byproduct even in the absence of specific mate preferences or avoidance behaviours (Berggren et al., 2012;Calderon et al., 2010;Shine et al., 2011).The silvery patterns in the adippe morph might be important for intraspecific signalling, where the flashing of the reflective spots in sunlight might facilitate conspecific perception and attract the attention of potential mating partners, a phenomenon observed in various taxa (Mouchet & Vukusic, 2018;Vukusic & Sambles, 2003).Our finding of sex-based differences in UV reflectance suggests a role in intraspecific communication, such as mate choice or intra-sexual signalling.Female adippe butterflies had roughly double the UV reflectance of males, potentially indicating male mate choice, such as observed in Bicyclus anynana, where females have more hindwing eyespots than males, and males, but not females, learn to choose mates based on the number of these spots (Westerman et al., 2014).Although many butterfly species exhibit female (Kemp, 2008;Papke et al., 2007;Silberglied, 1984) or mutual (Westerman et al., 2014) mate choice, there are many documented cases of male mate choice.Such examples include Pieris napi and P. rapae where UV reflectance is higher in females, and males actively search for females, with UV reflection playing an important role in mate localization, recognition and choice (Fukano et al., 2012;Stella et al., 2018).Another case is Colias philodice eriphyle, where males choose females based on wing melanisation (Ellers & Boggs, 2003).We propose that F.
adippe males might find and choose females based on the bright spots on the ventral hindwings, however, further studies on mating behaviour are needed to evaluate this proposition.
As F. adippe is a non-model organism, our data cannot identify with certainty the specific mechanisms shaping the phenotypic and genetic differentiation between distinct colour morphs documented in this species.However, we propose that both genetic and environmental variables influence the genomic architecture of the colour polymorphism.Because each distinct morph might be better suited for different microhabitats within the same population (see Section 4.1), diverging mating signals and preferences might have evolved, thereby potentially enabling the occupancy of a broader niche (Nosil & Feder, 2012).This is supported by the detection of an outlier locus appearing to be under divergent selection associated with colour morph identity, that is, ultraviolet-sensitive visual pigment (UVRh1).This gene plays an important role in intraspecific signalling, especially for mate choice correlated with wing colouration that reflects UV light, as reported for polymorphic Heliconius butterflies, a tropical genus in the family Nymphalidae, the same as F. adippe (Bybee et al., 2012;Chamberlain et al., 2009).In Heliconius butterflies, another UVRh opsin, which has formed as a duplication of UVRh1, is associated with perceiving the colour yellow in wings and there is evidence that UV receptors in their eyes have co-evolved with wing pigmentation (Briscoe et al., 2010).
Our demonstrations that the iridescent spots of the adippe morph (but barely the corresponding yellow spots on the cleodoxa morph) reflect UV light, that morphs differ in UVRh1 genotypes, and that there is an overall genetic divergence between colour morphs, suggest that selective mating is involved in the evolution and maintenance of the colour polymorphism in the High Brown Fritillary.
That all sampled populations showed a significant heterozygote deficiency at the outlier locus coding for UVRh1 (Table 4) also points towards non-random mating, as inbreeding decreases the frequency of heterozygotes (Gaffney et al., 1990;Rousset & Raymond, 1995).
It should be mentioned that all possible genotypes at UVRh1 were present in adippe, while only one (T/T) was found in cleodoxa (Tables 4 and 5), possibly indicating that the in Sweden rarer cleodoxa morph might be more selective.This is similar to what has been documented for Heliconius cydno butterflies, which show different gene expression levels between yellow and white butterflies at a locus that also plays a dominant role in mediating assortative mating (Chamberlain et al., 2009;Westerman, VanKuren, et al., 2018).In H.
cydno, the distinct colour patterns have a mimetic function across geographic races and assortative mating seems to contribute to ongoing ecological speciation in this system (Chamberlain et al., 2009).
Due to the very similar sympatric wing pattern polymorphism and assortative mating system, we carefully point towards the possibility of ecological speciation in F. adippe.Still, it remains an open question whether the here documented phenotypic and genomic differentiation between these colour morphs represents an early stage of ongoing evolutionary diversification.That F. adippe's sister species, F. niobe, shares a similar colour polymorphism with the nominate form F. niobe f. niobe exhibiting iridescent spots on the ventral hindwings, and F. niobe f. eris displaying yellow spots, seems to suggest that this colour polymorphism might be ancient (Eliasson et al., 2005;Tolman & Lewington, 2012).Such homologous polymorphisms highlight that multiple species may react similarly to analogous selective pressures (convergent evolution), and emphasize the benefits and the phylogenetic antiquity of phenotypic intraspecific trait variation as a response to shared prevailing selective regimes (Forsman, 2016;Mayr, 1963).

| CON CLUS I ON S AND FUTURE DIREC TIONS
We provide phenotypic and molecular evidence that the F. adippe colour morphs adippe and cleodoxa represent two genetically distinct and alternative integrated phenotypes.The overall genetic differentiation between morphs is in contrast to many other examples of colour polymorphism, where alternative colour pattern variants have been linked to differences in one or a few genes as in the Soay sheep (Gratten et al., 2007) and the black-headed bulbul (Shakya et al., 2021), or with a structural genomic change in the form of a super gene (chromosome inversion), as in, for example, Heliconius butterflies (Joron et al., 2011), the common ruff (Lamichhaney et al., 2016), and redpoll finches (Funk et al., 2021).
Our discovery of the genetic diversification of the UV-sensitive visual pigment together with the demonstration of higher reflectance in the UV range of the iridescent spots on the wings of the adippe morph, particularly in females, points to a possible role of intraspecific communication and sexual selection as a contributing driver of the evolution of genetic and phenotypic integration within and differentiation between colour morphs, similar to what we see in H. cydno butterflies (Chamberlain et al., 2009, Westerman, VanKuren, et al., 2018).In addition, we propose that divergent selection for thermoregulation and predator avoidance associated with differences in microhabitat use may also have contributed to the separation in size, darkness and iridescence between morphs.
Our study spurs several questions about the evolutionary dynamics of this polymorphism.For example, additional sampling from other parts of the distributional range may help identify environmental drivers of diversification and inform whether the differences accompanied by an overall genetic differentiation between sympatric morphs across populations.A genetic differentiation between morphs that is accompanied by other distinct trait values might suggest evolutionary modifications, possibly owing to selection.Lastly, we (iii) performed outlier analyses to identify loci putatively under selection and discuss environmental factors or processes contributing to evolutionary differentiation and maintenance of the colour polymorphism.F I G U R E 1 Map of study area.The red spots indicate the three studied Swedish populations in Småland, and the islands Öland and Gotland (from left to right).The pictures on the right side show the dorsal side of Fabriciana adippe, the ventral side of F. adippe f. adippe with silvery spots on the ventral hindwings, and F. adippe f. cleodoxa with yellow spots (from top to bottom).The red lines and numbers indicate the location of the two spots on the ventral side of the hind wing, where the reflection probe was used to quantify optical properties (spectral reflectance) of the silvery spots of the adippe morph and the corresponding yellow spots in the cleodoxa morph.
Genomic DNA (gDNA) was extracted by grinding thorax tissue with stainless steel beads (Next Advance, USA) in a bullet blender, and E.Z.N.A.® Genomic DNA Isolation Kit (OMEGA Bio-Tek, USA) was used for further processing.We used agarose gel electrophoresis to estimate the quality of gDNA.The DNA samples with relatively low quality were subjected to whole genome amplification using REPLI-g Mini Kit (Qiagen) to obtain sufficient gDNA in quantity and quality prior to double-digested restriction site-associated DNA sequencing (ddRADseq).Double-digested restriction site-associated DNA sequencing (ddRADseq) libraries were prepared following the protocol inPeterson et al. (2012) andLee et al. ( None of the cleodoxa individuals from Öland and Gotland reflected in the UV-spectrum.By contrast, all six (four females and two males) cleodoxa individuals from Småland showed weak F I G U R E 2 Biplot illustrating the composite phenotypic differentiation between Fabriciana adippe butterflies resulting from PCA, according to colour morph (adippe/cleodoxa) indicated by symbol colours and source population (Småland, Öland and Gotland) indicated by symbol shapes.Large positive values on Prin1 indicate large wings, a higher degree of melanism (darker coloration), and a low proportion of green scales on the underside of the hindwing.Large positive values on Prin2 indicate that hindwings are pale (less melanic), a high proportion of green scales on the underside of the hindwings, and a high percentage of orange summed across all four wing parts.Large diamonds show least squares means (± 95% confidence limits) of PC1 and PC2 scores for each colour morph obtained from a MANOVA (see Tables ceptor genotypes (alleles T and A).In the entire sample, nearly all (17 of 18) individuals from Småland belonged to the T/T genotype, whereas nearly half (25 of 69) of the individuals from Gotland and F I G U R E 3 Comparisons of reflectance in Fabriciana adippe f. adippe and f. cleodoxa (a) females and (b) males from three different source populations.Reflectance was quantified for two of the silvery iridescent spots in the adippe morph and two corresponding yellow spots in the cleodoxa morph (Figure 1).Bottom panels show (c) sum of reflectance in the UV (325-385 nm) and (d) ratio of the summed reflectance in the UV and longer wavelength (LW, 390-700 nm) spectra for adippe and cleodoxa females (open squares) and males (black diamonds).Values represent least-squares means ± SE estimated from the mixed model ANOVAs presented in

a
The outlier sequences were interrogated against coding sequences from all available Lepidopteran reference genomes in Lepbase 4(Challi et al., 2016) using the BLAST interface.b U R E 4 (a) Matrix of F ST values for pairwise comparisons of individuals from colour morphs within and across populations.Darker colours indicate higher genetic differentiation between groups, p-values for each comparison are in the respective square.(b) Bi-plot according to PCA of ddRADseq data.Colours represent colour morphs, shapes represent populations.more clearly.PERMANOVA results were in overall agreement with the outcome of the other analyses, suggesting a significant effect of population and colour morph on genetic structure (population, F 2,86 = 1.91, p < .001;colour morph, F 1,85 = 1.39, p < .001;interaction, F 2,85 = 1.23, p < .001).Overall, CAP analyses supported the genetic distinctiveness of the colour morphs for both the full dataset (tr = 0.70, p < .001)and Öland (tr = 0.82, P < .001, Figure S2).The morphs might represent alternative integrated phenotypes with distinct trait value combinations, possibly constituting different fitness peaks in distinct microhabitats (discussed below).Importantly, the reported phenotypic associations with colour morph were largely consistent across all study populations, despite the clear population genetic structure indicated by our results.The results further show that colour morph identity was linked with the overall genetic structure, suggesting that the phenotypic differentiation between morphs reflects genetically based evolutionary modifications, F I G U R E 5 Maximum Likelihood inference tree according to ddRADseq data.Numbers on branches display bootstrap values.The top left corner shows a manipulated picture of Fabriciana adippe illustrating the ventral sides of both colour morphs, with the adippe morph on the left, and the cleodoxa morph on the right.Photo: Daniela Polic.

F
I G U R E 6 Bi-plots according to db-RDA of ddRADseq data showing associations of the genetic structure with colour morph identity, source population, sex and other phenotypic variables.Colours are according to colour morph, shapes represent source populations, and arrows show the direction of association of analysed variables with the genetic structure.

F I G U R E 7
Outliers suggested by BayeScenv and Fdist.Blue lines represent the cut-off values (p < .05for BayeScEnv and p < .01 for Fdist).Outliers identified by both methods are highlighted with encircled markers.Additionally, loci within a gene locus with known names are specifically indicated.
between morphs are general or context specific.Studies of habitat use, thermoregulation and mate choice, together with comparisons of viability of offspring from pure and mixed parental combinations in staged mating experiments, and comparisons of detectability and predation risk of the morphs in different habitats can help identify the underlying selection pressures.Finally, whole-genome sequencing may identify genomic regions and possible chromosomal re-arrangements responsible for differences in colour patterns and other traits.AUTH O R CO NTR I B UTI O N S DP, AF, YY and MF designed the study.Funding was secured by AF, MF, DP and YY.DP, MF and YY collected the samples.DP and YY performed the laboratory work.SM conducted the reflectance measurements.DP, YY and AF analysed the data.DP and AF wrote the first draft of the manuscript.All authors contributed to the final version and approved the submitted manuscript.
Number of individuals from each colour morph, sex and population used in this study.

Table 2 . Source of variation Num DF Den DF UV (325-385 nm) Ratio UV/LW
Allele frequencies of the gene ultraviolet-sensitive visual pigment (left) in F. adippe populations.
TA B L E 5Genotype information for the gene ultraviolet-sensitive visual pigment (UVRh1) in F. adippe populations.

table S1 in
Polic et al. (2023)).Still, firm evaluation of the hypothesis that these colour morphs represent alternative adaptive peaks associated with different microhabitats (i.e.forest vs. open habitats) would require to systematically study and compare habitat utilization where the two morphs occur in sympatry.