Emergence of a floral colour polymorphism by pollinator-mediated overdominance

Maintenance of polymorphism by overdominance (heterozygote advantage) is a fundamental concept in evolutionary biology. In most examples known in nature, overdominance is a result of homozygotes suffering from deleterious effects. Here we show that overdominance maintains a non-deleterious polymorphism with black, red and white floral morphs in the Alpine orchid Gymnadenia rhellicani. Phenotypic, metabolomic and transcriptomic analyses reveal that the morphs differ solely in cyanidin pigments, which are linked to differential expression of an anthocyanidin synthase (ANS) gene. This expression difference is caused by a premature stop codon in an ANS-regulating R2R3-MYB transcription factor, which is heterozygous in the red colour morph. Furthermore, field observations show that bee and fly pollinators have opposite colour preferences; this results in higher fitness (seed set) of the heterozygous morph without deleterious effects in either homozygous morph. Together, these findings demonstrate that genuine overdominance exists in nature.

Google Maps, 2018, Digital Globe, European Space Imaging; plot square sizes not to scale). b, Within the fourteen plots, the G. rhellicani plants are non-randomly distributed , D = 0.19312, P = 1.578 × 10 -9 ): Overall plant density increases towards the centre of the population (plots G-J), and plants are outcompeted by other vegetation in certain plot areas. c, A Studentised Permutation Test shows that the spatial distribution pattern does not differ between the colour morphs (SPM-Test (n=281), T = 127.09, P = 0.714, lines show the mean of the summary function for each colour morph).

Supplementary Figure 4 | VOC quantities in the floral scent bouquet of the three G. rhellicani colour morphs
Emission rates in pg per l air of the 15 compounds which each account for ≥ 0.1% of the total floral scent bouquet recorded in the Puflatsch population. Apart from a trend towards a reduced emission of aromatic compounds in red and white morphs (only statistically significant for the minor compound creosol) there was no difference in VOC composition between the three colour morphs (see Supplementary Tables 2 and 5). Centre lines denote medians, bounds of boxes denote first and third quartiles, whiskers denote 1.5 × interquartile ranges.

Supplementary Figure 5 | UHPLC-MS/MS quantifications of anthocyanins and precursors in flowers of the three G. rhellicani colour morphs
The anthocyanin pathway is divided into three branches producing blue delphinidin, orange pelargonidin, and red cyanidin pigments. Quantification of anthocyanins (area under curve (AUC) / (mg tissue × molecular weight (MW)), coloured frames) and colourless precursors (black frames) showed that the metabolite flux is mainly channelled through the cyanidin branch in G. rhellicani, hence the red colour of the plants. Further modifications such as the addition of different sugar residues and chemical groups can slightly change the hue of an anthocyanin compound. The red colour in G. rhellicani flowers is primarily due to cyanidin-3-glucoside, its derivative cyanidin-3-(6-malonylglucoside) and, to a lesser degree another derivative, peonidin-3-glucoside. The amount of the two main cyanidins is reduced more than 7.5 × in red, and more than 30 × in white colour morphs (ANOVA and TukeyHSD post hoc test, see Supplementary Table  6). All compounds were identified and quantified using authentic reference standards, except for cyanidin-3-(6-malonylglucoside), which was identified by an interpretation of the data spectrum obtained and based on previous reports (see Methods section in the main text). Standards for the two delphinidin precursors pentahydroxyflavonone and dihydromyricetin were unavailable, and leucodelphinidin, leucopelargonidin, and leucocyanidin are unstable intermediates. Enzyme abbreviations: CHI: chalcone isomerase, A5GT: anthocyanin 5-O-glucosyltransferase; MaT: anthocyanin malonyl transferase. Centre lines denote medians, bounds of boxes denote first and third quartiles, whiskers denote 1.5 × interquartile ranges.

Supplementary Figure 6 | Quantification and correlation of floral pigments
with the phenotype and genotype of the three G. rhellicani colour morphs a, Concentration of the two main floral carotenoids, β-carotene and lutein, is not different between the colour morphs (β-carotene: ANOVA (n=43), F(2, 40)=0.520, P=0.598, lutein: ANOVA (n=45), F(2, 42)=0.614, P=0.546); centre lines denote medians, bounds of boxes denote first and third quartiles, whiskers denote 1.5 × interquartile ranges. b, Correlation between phenotype (spectral reflectance at 517 nm), metabolomics (combined amount of the two main cyanidins), and transcriptomics data (GrANS1 expression level). Samples are coloured according to genotype (SNP at position 663 in the GrMYB1 transcript, R = A or G). c, GrANS1 genotype is neither correlated with GrANS1 expression level, nor with plant colour (GrMYB1 genotype). Letters a-d denote different GrANS1 alleles identified by a combination of nine SNPs in the coding sequence (SNP states of allele a: GGGGCACTC, allele b: GGGGTACCC, allele c: GAAGCGATG, allele d: AGGTCGACG).

Supplementary Tables
Supplementary Table 1 | Coordinates of the populations assessed in this  study  Table showing the country, latitude and longitude coordinates in WGS 84 format, as well as the elevation in metres above sea level (m.a.s.l.) of the assessed G. rhellicani populations (sorted from west to east).

Supplementary Table 4 | ANOVA tables from the linear mixed-effects models of floral VOC emission in G. rhellicani colour morphs
Prior to analysis, all VOC data was BoxCox transformed to approach normality. Differences in VOC emission between the colour morphs were modelled using Holm-adjusted linear mixed-effects models with sampling year as random factor (n=62), and Tukey HSD post-hoc tests. % bouquet: average proportion of the entire floral volatile bouquet, F (2,58): conditional F-statistic, P adj : Holm-adjusted P-value (P<0.05 in bold), post-hoc: summary of TukeyHSD post-hoc comparisons (B: black, R: red, W: white, n.a. = post-hoc test not applicable).