Genetic basis and dual adaptive role of floral pigmentation in sunflowers

  1. Marco Todesco  Is a corresponding author
  2. Natalia Bercovich
  3. Amy Kim
  4. Ivana Imerovski
  5. Gregory L Owens
  6. Óscar Dorado Ruiz
  7. Srinidhi V Holalu
  8. Lufiani L Madilao
  9. Mojtaba Jahani
  10. Jean-Sébastien Légaré
  11. Benjamin K Blackman
  12. Loren H Rieseberg  Is a corresponding author
  1. Department of Botany and Biodiversity Research Centre, University of British Columbia, Canada
  2. Department of Biology, University of Victoria, Canada
  3. Department of Plant and Microbial Biology, United States
  4. Michael Smith Laboratory and Wine Research Centre, University of British Columbia, United States

Abstract

Variation in floral displays, both between and within species, has been long known to be shaped by the mutualistic interactions that plants establish with their pollinators. However, increasing evidence suggests that abiotic selection pressures influence floral diversity as well. Here, we analyse the genetic and environmental factors that underlie patterns of floral pigmentation in wild sunflowers. While sunflower inflorescences appear invariably yellow to the human eye, they display extreme diversity for patterns of ultraviolet pigmentation, which are visible to most pollinators. We show that this diversity is largely controlled by cis-regulatory variation affecting a single MYB transcription factor, HaMYB111, through accumulation of ultraviolet (UV)-absorbing flavonol glycosides in ligules (the ‘petals’ of sunflower inflorescences). Different patterns of ultraviolet pigments in flowers are strongly correlated with pollinator preferences. Furthermore, variation for floral ultraviolet patterns is associated with environmental variables, especially relative humidity, across populations of wild sunflowers. Ligules with larger ultraviolet patterns, which are found in drier environments, show increased resistance to desiccation, suggesting a role in reducing water loss. The dual role of floral UV patterns in pollinator attraction and abiotic response reveals the complex adaptive balance underlying the evolution of floral traits.

Editor's evaluation

The enlarged petals of sunflowers contain pigments that absorb ultraviolet light and are perceived by pollinators as dark ‘bullseyes’ that function as nectar guides. Todesco et al. identify the primary genetic mechanism underlying variation in the size of this bullseye pattern and provide evidence suggesting that abiotic variables, rather than pollinators, may maintain this phenotypic and genotypic variation.

https://doi.org/10.7554/eLife.72072.sa0

eLife digest

Flowers are an important part of how many plants reproduce. Their distinctive colours, shapes and patterns attract specific pollinators, but they can also help to protect the plant from predators and environmental stresses.

Many flowers contain pigments that absorb ultraviolet (UV) light to display distinct UV patterns – although invisible to the human eye, most pollinators are able to see them. For example, when seen in UV, sunflowers feature a ‘bullseye’ with a dark centre surrounded by a reflective outer ring. The sizes and thicknesses of these rings vary a lot within and between flower species, and so far, it has been unclear what causes this variation and how it affects the plants.

To find out more, Todesco et al. studied the UV patterns in various wild sunflowers across North America by considering the ecology and molecular biology of different plants. This revealed great variation between the UV patterns of the different sunflower populations. Moreover, Todesco et al. found that a gene called HaMYB111 is responsible for the diverse UV patterns in the sunflowers. This gene controls how plants make chemicals called flavonols that absorb UV light.

Flavonols also help to protect plants from damage caused by droughts and extreme temperatures. Todesco et al. showed that plants with larger bullseyes had more flavonols, attracted more pollinators, and were better at conserving water. Accordingly, these plants were found in drier locations.

This study suggests that, at least in sunflowers, UV patterns help both to attract pollinators and to control water loss. These insights could help to improve pollination – and consequently yield – in cultivated plants, and to develop plants with better resistance to extreme weather. This work also highlights the importance of combining biology on small and large scales to understand complex processes, such as adaptation and evolution.

Introduction

The diversity in colour and colour patterns found in flowers is one of the most extraordinary examples of adaptive variation in the plant world. As remarkable as the variation that we can observe is, even more of it lays just outside our perception. Many species accumulate pigments that absorb ultraviolet (UV) radiation in their flowers; while these patterns are invisible to the human eye, they can be perceived by pollinators, most of which can be seen in the near UV (Chittka et al., 1994; Tovée, 1995). UV patterns have been shown to increase floral visibility and to have a major influence on pollinator visitation and preference (Brock et al., 2016; Horth et al., 2014; Rae and Vamosi, 2013; Sheehan et al., 2016). Besides their importance for pollinator attraction, patterns of UV-absorbing pigments in flowers have increasingly been recognized to have a role in responses to other biotic and abiotic factors, including defence against insect herbivory (Gronquist et al., 2001), protection against UV radiation (Koski and Ashman, 2015; Koski et al., 2020), and adaptation to different temperatures (Koski and Ashman, 2016; Koski et al., 2020).

Sunflowers are one of the most recognizable members of the Asteraceae family, which comprises circa 10% of all flowering plants (Mandel et al., 2019). Besides cultivated sunflower, about 50 species of wild sunflowers are found across North America. Wild sunflowers are adapted to a variety of different habitats and display a remarkable amount of phenotypic and genetic diversity, which makes them a model system for studies of adaptation, speciation, and domestication (Bock et al., 2020; Heiser et al., 1969; Todesco et al., 2020). In addition to being a major crop, sunflowers are also ubiquitous in popular culture, largely due to their iconic yellow inflorescences. Indeed, like many Asteraceae species, wild sunflowers have ligules (the enlarged modified petals of the outermost whorl of florets in the sunflower inflorescence) that appear of the same bright yellow colour to the human eye. However, ligules also accumulate UV-absorbing pigments at their base, while their tip reflects UV radiation (Harborne and Smith, 1978; Wojtaszek and Maier, 2014). Across the whole inflorescence, this results in a bullseye pattern, with an external UV-reflecting ring and an internal UV-absorbing ring. Considerable variation in the size of UV bullseye patterns has been observed between and within plant species (Koski and Ashman, 2013; Koski and Ashman, 2016); however, few studies have investigated the ecological factors that drive this variation or the genetic determinants that control it (Brock et al., 2016; Koski and Ashman, 2015; Moyers et al., 2017; Sheehan et al., 2016). Here, we explore the diversity of floral UV pigmentation in wild sunflowers and the genetic mechanisms and environmental factors that shape this variation.

Results and discussion

Floral UV patterns in wild sunflowers

A preliminary screening of 19 species of wild sunflowers, as well as cultivated sunflower, suggested that UV bullseye patterns are common across sunflower species (Figure 1—figure supplement 1). In several cases, we also observed substantial within-species variation for the size of UV floral patterns. Patterns of floral UV pigmentation have been previously investigated in the silverleaf sunflower Helianthus argophyllus, which is endemic to Southern Texas (Figure 1—figure supplement 1). Limited diversity was found between individuals, but transgressive segregation was observed in mapping populations; while several QTL affecting this trait were detected, genetic mapping resolution was insufficient to identify individual causal genes (Moyers et al., 2017).

To better understand the function and genetic regulation of variation for floral UV pigmentation, we focused on two widespread species of annual sunflowers, Helianthus annuus and Helianthus petiolaris. H. annuus, the common sunflower, grows across most of North America; it is probably the most diverse of the sunflower species and is the progenitor of domesticated sunflower (H. annuus var. macrocarpus). H. petiolaris also has a broad distribution across North America, but prefers sandier soils. It includes two subspecies: subsp. petiolaris, which is common in the central plains of the United States, and subsp. fallax, which is found in the Southwestern USA and has repeatedly adapted to growing on sand dunes (Heiser et al., 1969; Todesco et al., 2020). Over two growing seasons, we measured floral UV patterns (as the proportion of the ligule that absorbs UV radiation, henceforth ‘ligule ultraviolet proportion’ [LUVp]) in 1589 H. annuus individuals derived from 110 distinct natural populations and 351 H. petiolaris individuals from 40 populations, grown in common garden experiments in Vancouver, Canada (Todesco et al., 2020). The populations of origin of these plants were selected to represent the whole range of H. annuus, and most of the range of H. petiolaris (Figure 1a and b, Figure 1—source data 1). While extensive variation was observed within both species, it was particularly striking for H. annuus, which displayed a phenotypic continuum from ligules with almost no UV pigmentation to ligules that were entirely UV-absorbing (Figure 1c–e, Figure 1—figure supplement 2, Figure 1—source data 2). Floral UV patterns have been proposed to act as nectar guides, helping pollinators orient towards nectar rewards once they land on the petal (Daumer, 1958), although recent experiments have challenged this hypothesis (Koski et al., 2014). A relatively high proportion of H. annuus individuals in our survey (~13%) had completely UV-absorbing ligules and therefore lacked UV nectar guides, suggesting that pollinator orientation is not a necessary function of floral UV pigmentation in sunflower.

Figure 1 with 3 supplements see all
Diversity for floral ultraviolet (UV) pigmentation patterns in wild sunflowers.

(a) Geographical distribution of sampled populations for H. annuus and (b) H. petiolaris. Yellow/orange dots represent different populations, overlaid grey dot size is proportional to the population mean ligule ultraviolet proportion (LUVp). (c) Range of variation for floral UV pigmentation patterns in the two species. Scale bar = 2 cm. (d) LUVp values distribution for H. annuus and (e) H. petiolaris subspecies.

Figure 1—source data 1

Populations used in this study, average ligule ultraviolet proportion (LUVp) values, environmental variables, and inflorescence traits.

https://cdn.elifesciences.org/articles/72072/elife-72072-fig1-data1-v1.xlsx
Figure 1—source data 2

Individuals used in this study, ligule ultraviolet proportion (LUVp) values, Chr15_LUVp SNP genotypes, and inflorescence traits.

https://cdn.elifesciences.org/articles/72072/elife-72072-fig1-data2-v1.xlsx

Genetic control of floral UV patterning

To identify the loci controlling variation for floral UV patterning, we performed a genome-wide association study (GWAS). We used a subset of the phenotyped plants (563 of the H. annuus and all 351 H. petiolaris individuals) for which we previously generated genotypic data at >4.6M high-quality single-nucleotide polymorphisms (SNPs) (Todesco et al., 2020). Given their relatively high level of genetic differentiation, analyses were performed separately for the petiolaris and fallax subspecies of H. petiolaris (Todesco et al., 2020). While no significant association was identified for H. petiolaris fallax (Figure 2—figure supplement 1), we detected several genomic regions significantly associated with floral UV patterning in H. petiolaris petiolaris, and a particularly strong association (p=5.81e–25) on chromosome 15 in H. annuus (Figure 2a and b). The chromosome 15 SNP with the strongest association with ligule UV pigmentation patterns in H. annuus (henceforth ‘Chr15_LUVp SNP’) explained 62% of the observed phenotypic and additive variation (narrow-sense heritability for LUVp in the H. annuus dataset is ~1). Additionally, allelic distributions at this SNP closely matched that of floral UV patterns (Figure 2c, compare to Figure 1a; Figure 1—source data 2).

Figure 2 with 3 supplements see all
A single locus explains most of the variation in floral ultraviolet (UV) patterning in H. annuus.

(a) Ligule ultraviolet proportion (LUVp) genome-wide association studies (GWAS). (b) Zoomed-in Manhattan plot for the chromosome 15 LUVp peak in H. annuus. Red lines represent 5% Bonferroni-corrected significance. GWAS were calculated using two-sided mixed models. Number of individuals: n = 563 individuals (H. annuus); n = 159 individuals (H. petiolaris petiolaris). Only positions with -log10 p-value >2 are plotted. HaMYB111 is the only annotated feature in the genomic interval shown in Figure 1b; the single-nucleotide polymorphism (SNP) with the strongest association to LUVp (Chr15_LUVp SNP) is highlighted in red. Linkage disequilibrium (LD) decays rapidly in wild H. annuus (average R2 at 10 kbp is ~0.035; Todesco et al., 2020), and all SNPs significantly associated with LUVp in H. annuus are included in the depicted region. The chromosome 15 association in H. petiolaris petiolaris is distinct from the one in H. annuus as it is located ~20 Mbp downstream of HaMYB111. (c) Geographical distribution of Chr15_LUVp SNP allele frequencies in H. annuus. L, large; S, small allele. (d) LUVp associated with different genotypes at Chr15_LUVp SNP in natural populations of H. annuus grown in a common garden. All pairwise comparisons are significant for p<10–16 (one-way ANOVA with post-hoc Tukey HSD test, F = 438, df = 2; n = 563 individuals). LUVp values for the individuals in the GWAS populations and genotype data for Chr15_LUVp SNP are reported in Figure 1—source data 2. (e) LUVp associated with different genotypes at Chr15_LUVp SNP in H. annuus F2 populations grown in the field or in a greenhouse (GH). Measurements for the parental generations are shown: squares, grandparents (field-grown); empty circles, F1 parents (GH-grown; Figure 2—figure supplement 2). Boxplots show the median, box edges represent the 25th and 75th percentiles, whiskers represent the maximum/minimum data points within 1.5× interquartile range outside box edges. Differences between genotypic groups are significant for p=0.0057 (Pop. 1 Field, one-way ANOVA, F = 5.73, df = 2; n = 54 individuals); p=0.0021 (Pop. 2 Field, one-way ANOVA, F = 7.02, df = 2; n = 50 individuals); p=0.00015 (Pop. 1 GH, one-way ANOVA, F = 11.13, df = 2; n = 42 individuals); p=0.054 (Pop. 2 GH, one-way ANOVA, F = 3.17, df = 2; n = 38 individuals). p-Values for pairwise comparisons for panels (d) and (e) are reported in Figure 2—source data 1.

Figure 2—source data 1

Ligule ultraviolet proportion (LUVp) values and Chr15_LUVp SNP genotypes for F2.

https://cdn.elifesciences.org/articles/72072/elife-72072-fig2-data1-v1.xlsx

Genotype at the Chr15_LUVp SNP had a remarkably strong effect on the size of UV bullseyes in inflorescences. Individuals homozygous for the ‘large’ (L) allele had a mean LUVp of 0.78 (SD ±0.16), meaning that ~3/4 of the ligule was UV-absorbing, while individuals homozygous for the ‘small’ (S) allele had a mean LUVp of 0.33 (SD ±0.15), meaning that only the basal ~1/3 of the ligule absorbed UV radiation. Consistent with the trimodal LUVp distribution observed for H. annuus (Figure 1d), alleles at this locus showed additive effects, with heterozygous individuals having intermediate phenotypes (LUVp = 0.59 ± 0.18; Figure 2d). The association between floral UV patterns and the Chr15_LUVp SNP was confirmed in the F2 progeny of crosses between plants homozygous for the L allele (with completely UV-absorbing ligules; LUVp = 1) and for the S allele (with a small UV-absorbing patch at the ligule base; LUVp < 0.18; Figure 2e, Figure 2—figure supplement 2). Average LUVp values were lower, and their range narrower, when these populations were grown in a greenhouse rather than in a field. Plants in the greenhouse experienced relatively uniform temperatures and humidity, and were shielded from most UV radiation. These results suggest that although floral UV patterns have a strong genetic basis (consistent with previous observations; Koski and Ashman, 2013), their expression is also affected by the environment.

HaMYB111 regulates UV pigment production

While no obvious candidate genes were found for the GWAS peaks for floral UV pigmentation in H. petiolaris petiolaris, the H. annuus chromosome 15 peak is ~5 kbp upstream of HaMYB111, a sunflower homolog of the Arabidopsis thaliana AtMYB111 gene (Figure 2b). Together with AtMYB11 and AtMYB12, AtMYB111 is part of a small family of transcription factors (also called PRODUCTION OF FLAVONOL GLYCOSIDES [PFG]) that controls the expression of genes involved in the production of flavonol glycosides in Arabidopsis (Stracke et al., 2007). Flavonol glycosides are a subgroup of flavonoids known to fulfil a variety of functions in plants, including protection against abiotic and biotic stresses (e.g., UV radiation, cold, drought, herbivory) (Pollastri and Tattini, 2011). Crucially, they absorb strongly in the near UV range (300–400 nm) and are the pigments responsible for floral UV patterns in several plant species (Rieseberg and Schilling, 1985; Sheehan et al., 2016; Thompson et al., 1972). For instance, alleles of a homolog of AtMYB111 are responsible for the evolutionary gain and subsequent loss of flavonol accumulation and UV absorption in flowers of Petunia species, associated with two successive switches in pollinator preferences (from bees, to hawkmoths, to hummingbirds; Sheehan et al., 2016). A homolog of AtMYB12 has also been associated with variation in floral UV patterns in Brassica rapa (Brock et al., 2016). Analysis of sunflower ligules found two main groups of UV-absorbing compounds: glycoside conjugates of quercetin (a flavonol) and di-O-caffeoyl quinic acid (CQA, a member of a family of antioxidant compounds that includes chlorogenic acid and that accumulates at high levels in many sunflower tissues; Koeppe et al., 1970). Both quercetin glycosides and CQA were more abundant at the base of sunflower ligules, and in ligules of plants with larger LUVp. However, this pattern was much more dramatic for flavonols, and they represented a much larger fraction of the total UV absorbance in UV-absorbing (parts of) ligules, suggesting that flavonols are the main pigments responsible for UV patterning in sunflower ligules (Figure 3a and b).

Figure 3 with 2 supplements see all
MYB111 is associated with floral ultraviolet (UV) pigmentation patterns and flavonol accumulation in sunflower and Arabidopsis.

(a) UV chromatograms (350 nm) for methanolic extracts of the upper and lower third of ligules with intermediate UV patterns, and (b) of ligules with large and small floral UV patterns. Peak areas are proportional to the total amount of absorbance at 350 nm explained by the corresponding compounds in the extracts. Relevant peaks are labelled: M, myricetin; QG, quercetin glucoside; QdG, quercetin diglucoside; Q3OG, quercetin-3-O-glucoside (co-elutes with quercetin-glucoronide); QMG, quercetin malonyl-glucoside; CQA, di-O-caffeoyl quinic acid (Figure 3—source data 1). (c) Expression levels of AtMYB111 in Arabidopsis. RNAseq data were obtained from Klepikova et al., 2016. RPKM, reads per kilobase of transcript per million mapped reads. (d) Arabidopsis petals. HaMYB111 from H. annuus plants with small or large ligule ultraviolet proportion (LUVp) was introduced into the Arabidopsis myb111 mutant under the control of a constitutive promoter (p35ScaMV) or of the promoter of the Arabidopsis homolog (pAtMYB111). Scale bar = 1 mm. (e) UV chromatograms (350 nm) for methanolic extracts of petals of Arabidopsis lines. Upper panel: wild-type Col-0 and mutants. Bottom panel: p35ScaMV::HaMYB111 lines in myb111 background. Relevant peaks are labelled: KRGRR, kaempferol-rhamnoside-glucoside-rhamnoside-rhamnoside; QGR, quercetin-rhamnoside-glucoside; KRG, kaempferol-rhamnoside-glucoside; IRG, isorhamnetin-rhamnoside-glucoside; QX, quercetin-xyloside. (Figure 3—source data 1). (f) Expression levels of HaMYB111 in the XRQ line of cultivated sunflower. RNAseq data were obtained from Badouin et al., 2017. (g) Pigmentation patterns in ligules of wild H. annuus at different developmental stages: R3, closed inflorescence bud; R4, inflorescence bud opening; R5, inflorescence fully opened. (h) Expression levels in the UV-absorbing base (grey) and UV-reflecting tip (yellow) of mature (mat; collected from inflorescences at R5 stage) and developing (dev; collected from inflorescences at R4 stage) ligules for HaMYB111 and (i) HaFLS1, one of its putative targets. One representative individual with intermediate LUVp values was chosen for each species. Each bar represents average expression over three technical replicates for a biological replicate (different inflorescence from the same individual). For each gene, expression data are normalized to the average expression levels in the base of developing ligules of H. annuus. (j) HaMYB111 expression levels in ligules of field-grown wild H. annuus with contrasting floral UV pigmentation patterns. Expression data are normalized to the average expression levels across all the samples. The difference between the two groups is significant for p=0.009 (Welch t-test, t = 2.81, df = 27.32, two-sided; n = 24 individuals for the large LUVp group; n = 22 individuals for the small LUVp group). Similar correlations are observed when HaMYB111 expression levels are compared to individuals’ LUVp values or their genotype at the Chr15_LUVp SNP (see Figure 3—source data 2). Boxplots show the median, box edges represent the 25th and 75th percentiles, whiskers represent the maximum/minimum data points within 1.5× interquartile range outside box edges.

Figure 3—source data 1

Flavonols in methanolic extractions of sunflower ligules and Arabidopsis petals.

https://cdn.elifesciences.org/articles/72072/elife-72072-fig3-data1-v1.xlsx
Figure 3—source data 2

Expression analyses in sunflower and Arabidopsis.

https://cdn.elifesciences.org/articles/72072/elife-72072-fig3-data2-v1.xlsx

In Arabidopsis, AtMYB12 and AtMYB111 are known to have the strongest effect on flavonol glycoside accumulation (Stracke et al., 2007; Stracke et al., 2010). We noticed, from existing RNAseq data, that AtMYB111 expression levels are particularly high in petals (Klepikova et al., 2016; Figure 3c) and found that Arabidopsis petals, while uniformly white in the visible spectrum, absorb strongly in the UV (Figure 3d). To our knowledge, this is the first report of floral UV pigmentation in Arabidopsis, a highly selfing species that is seldom insect-pollinated (Hoffmann et al., 2003). Accumulation of flavonol glycosides is strongly reduced, and UV pigmentation is almost completely absent, in petals of mutants for AtMYB111 (myb111). UV absorbance is further reduced in petals of double mutants for AtMYB12 and AtMYB111 (myb12/111). However, petals of the single mutant for AtMYB12 (myb12), which is expressed at low levels throughout the plant (Klepikova et al., 2016), are indistinguishable from wild-type plants (Figure 3d and e). This shows that flavonol glycosides are responsible for floral UV pigmentation also in Arabidopsis, and that AtMYB111 plays a fundamental role in controlling their accumulation in petals.

To confirm that sunflower HaMYB111 is functionally equivalent to its Arabidopsis homolog, we introduced it into myb111 plants. Expression of HaMYB111, either under the control of a constitutive promoter or of the endogenous AtMYB111 promoter, restored petal UV pigmentation and induced accumulation of flavonol glycosides (Figure 3d and e). HaMYB111 coding sequences obtained from wild sunflowers with large or small LUVp were equally effective at complementing the myb111 mutant. Together with the observation that the strongest GWAS association with LUVp fell in the promoter region of HaMYB111, these results suggest that differences in the effect of the ‘small’ and ‘large’ alleles of this gene on floral UV pigmentation are not due to differences in protein function, but rather to differences in gene expression.

Analysis of HaMYB111 expression patterns in cultivated sunflower revealed that, consistent with a role in floral UV pigmentation and similar to its Arabidopsis counterpart, it is expressed specifically in ligules, and it is almost undetectable in other tissues (Badouin et al., 2017; Figure 3f). Similar to observations in Rudbeckia hirta, another member of the Heliantheae tribe (Schlangen et al., 2009), UV pigmentation is established early in ligule development in both H. annuus and H. petiolaris as their visible colour turns from green to yellow before the inflorescence opens (R4 developmental stage; Schneiter and Miller, 1981; Figure 3g, Figure 3—figure supplement 1). HaMYB111 is highly expressed in the part of the ligule that accumulates UV-absorbing pigments, and especially in developing ligules, consistent with a role in establishing pigmentation patterns (Figure 3h). We also observed a matching expression pattern for HaFLS1, the sunflower homolog of a gene encoding one of the main enzymes controlling flavonol biosynthesis in Arabidopsis (FLAVONOL SYNTHASE 1, AtFLS1), whose expression is regulated directly by AtMYB111 (Stracke et al., 2007; Figure 3i). Finally, we compared HaMYB111 expression levels in a set of 46 field-grown individuals with contrasting LUVp values, representing 21 different wild populations. HaMYB111 expression levels differed significantly between the two groups (p=0.009; Figure 3j). Variation in expression levels within phenotypic classes was quite large; this is likely due at least in part to the strong dependence of HaMYB111 expression on developmental stage (Figure 3g) and the difficulty of accurately establishing matching ligule developmental stages across diverse wild sunflowers.

These expression analyses further point to cis-regulatory rather than coding sequence differences between HaMYB111 alleles being responsible for LUVp variation. Accordingly, direct sequencing of the HaMYB111 locus from multiple wild H. annuus individuals, using a combination of Sanger sequencing and long PacBio HiFi reads, identified no coding sequence variants associated with differences in floral UV patterns, or with alleles at the Chr15_LUVp SNP (Figure 3—figure supplement 2, Supplementary files 1 and 2). However, we observed extensive variation in the promoter region of HaMYB111, differentiating wild H. annuus alleles from each other and from the reference assembly for cultivated sunflower (Supplementary files 3 and 4). Relaxing quality filters to include less well-supported SNPs in our LUVp GWAS did not identify additional variants with stronger associations than Chr15_LUVp SNP (Figure 2—figure supplement 2). However, many of the polymorphisms we identified by direct sequencing were either larger insertions/deletions (indels) or fell in regions that were too repetitive to allow accurate mapping of short reads, and would not be included even in this expanded SNP dataset. While several of these variants in the promoter region of HaMYB111 appeared to be associated with the Chr15_LUVp SNP, further studies will be required to confirm this, and to identify their eventual effects on HaMYB111 activity (see discussion in the legend of Figure 3—figure supplement 2).

Interestingly, when we sequenced the promoter region of HaMYB111 in several H. argophyllus and H. petiolaris individuals, we found that they all carried the S allele at the Chr15_LUVp SNP, and that their promoter regions were generally more similar in sequence to those of H. annuus individuals carrying the S allele at the Chr15_LUVp SNP (Supplementary files 3 and 4). Similarly, in a set of previously re-sequenced wild sunflowers, we found the S allele to be fixed in several perennial (Helianthus decapetalus, Helianthus divaricatus, and Helianthus grosseserratus) and annual sunflower species (H. argophyllus, Helianthus niveus, Helianthus debilis), and to be at >0.98 frequency in H. petiolaris (Figure 1—source data 2). Conversely, the L allele at Chr15_LUVp SNP was almost fixed (>0.98 frequency) in a set of 285 cultivated sunflower lines (Mandel et al., 2013). Consistent with these patterns, UV bullseyes are considerably smaller in H. argophyllus (mean LUVp ± SD = 0.27 ± 0.09), H. niveus (0.15 ± 0.09), and H. petiolaris (0.27 ± 0.12; Figure 1e) than in cultivated sunflower lines (0.62 ± 0.23). Additionally, while 50 of the cultivated sunflower lines had completely or almost completely UV-absorbing ligules (LUVp > 0.8), no such case was observed in the other three species (Figure 1—figure supplement 3).

A dual role for floral UV pigmentation

Although our results show that HaMYB111 explains most of the variation in floral UV pigmentation patterns in wild H. annuus, why such variation exists in the first place is less clear. Several hypotheses have been advanced to explain the presence of floral UV patterns and their variability. Like their visible counterparts, UV pigments play a fundamental role in pollinator attraction (Horth et al., 2014; Koski et al., 2014; Rae and Vamosi, 2013; Sheehan et al., 2016). For example, in Rudbeckia species, artificially increasing the size of bullseye patterns to up to 90% of the petal surface resulted in rates of pollinator visitation equal to or higher than wild-type flowers (which have on average 40–60% of the petal being UV-absorbing). Conversely, reducing the size of the UV bullseye had a strong negative effect on pollinator visitation (Horth et al., 2014). To test whether the relative size of UV bullseye patterns affected pollination, we assessed insect visitation rates for wild H. annuus lines with contrasting UV bullseye patterns. An initial experiment compared inflorescences from pairs of plants from two populations (ANN_03 from California and ANN_55 from Texas), which were selected to have large or small floral UV patterns. In this setup, inflorescences with large UV patterns received significantly more visits (Figure 4a). While this experiment revealed a clear pattern of pollinator preferences, it involved plants from only two different populations, and effects of other unmeasured factors unrelated to UV pigmentation on visitation patterns cannot be excluded. Therefore, we monitored pollinator visitation in plants grown in a common garden experiment including 1484 individuals from 106 H. annuus populations, spanning the entire range of the species. Assaying a much more diverse population of H. annuus individuals should reduce effects on pollinator preferences of traits unrelated to floral UV pigmentation. Within this field, we selected 82 plants, from 49 populations, which flowered at roughly the same time and had comparable numbers of flowers. We selected plants falling into three categories of LUVp values, representatives of the more abundant phenotypic classes across the range of wild H. annuus (Figure 1d): small (LUVp = 0–0.3), intermediate (LUVp = 0.5–0.8), and large (LUVp >0.95). Plants with intermediate UV patterns had the highest visitation rates (Figure 4b, Figure 4—figure supplement 1). Visitation to plants with small or large UV patterns was less frequent, and particularly low for plants with very small LUVp values (<0.15). Pollination rates are known to be yield-limiting in sunflower (Greenleaf and Kremen, 2006), and a strong reduction in pollination could therefore have a negative effect on fitness; this would be consistent with the observation that plants with very small LUVp values were rare (~1.5% of individuals) in our common garden experiment, which was designed to provide a balanced representation of the natural range of H. annuus. Although pollinator preferences in this experiment could still be affected by other unmeasured factors (nectar content, floral volatiles), these results are consistent with previous results showing that floral UV patterns play a major role in pollinator attraction (Horth et al., 2014; Koski et al., 2014; Rae and Vamosi, 2013; Sheehan et al., 2016). They also agree with earlier findings in other plant species, suggesting that intermediate-to-large UV bullseyes are preferred by pollinators (Horth et al., 2014; Koski et al., 2014). While we cannot exclude that smaller UV bullseyes would be preferred by pollinators in some parts of the H. annuus range, this does not seem likely; the most common pollinators of sunflower are ubiquitous across the range of H. annuus, and many bee species known to pollinate sunflower are found in both regions where H. annuus populations have large LUVp and regions where they have small LUVp (Hurd et al., 1980). Therefore, while acting as visual cues for pollinators is clearly a major function of floral UV bullseyes, it is unlikely to (fully) explain the patterns of variation that we observe for this trait.

Figure 4 with 5 supplements see all
Accumulation of ultraviolet (UV) pigments in flowers affects pollinator visits and transpiration rates.

(a) Rates of pollinator visitation measured in Vancouver in 2017 (p=0.017; Mann–Whitney U-tests, W = 150, two-sided; n = 143 pollinator visits) and (b) 2019 (differences between ligule ultraviolet proportion [LUVp] categories are significant for p=0.0058, Kruskal–Wallis test, χ2 = 14.54, df = 4; n = 1390 pollinator visits). Letters identify groups that are significantly different for p<0.05 in pairwise comparisons, Wilcoxon rank sum test. Exact p-values are reported in Figure 4—source data 14. Boxplots show the median, box edges represent the 25th and 75th percentiles, whiskers represent the maximum/minimum data points within 1.5× interquartile range outside box edges. (c) Correlation between average LUVp for different populations of H. annuus and summer UV radiation (R2 = 0.01, p=0.12, n = 110 populations) or (d) summer average temperature (R2 = 0.44, p=2.4 × 10–15, n = 110 populations). Grey areas represent 95% confidence intervals. (e) Sunflower inflorescences pictured in the visible, UV, and infrared (IR) range. In the IR picture, a bumblebee is visible in the inflorescence with large LUVp (right; the warmer abdomen of the bee is visible as a bright yellow spot under the asterisk). The higher temperature in the centre (disc) of the inflorescence with small LUVp does not depend on ligule UV patterns (Figure 4—figure supplement 3). These inflorescences belong to two of the plants that were used for the pollination preference experiments reported in (a) and are representative of the differences in floral UV patterns between LUVp categories in that experiment. (f) H. annuus ligules after having been exposed to sunlight for 15 min. (g) Five pairs of ligules from different sunflower lines were exposed to sunlight for 15 min, and their average temperature was measured from IR pictures. (h) Correlation between average LUVp in H. annuus populations and summer relative humidity (RH) (R2 = 0.51, p=1.4 × 10–18, n = 110 populations). The grey area represents the 95% confidence interval. (i) Rate of water loss from ligules and (j) leaves of wild H. annuus plants with large or small LUVp. Values reported are means ± standard error of the mean. n = 16 inflorescences (ligules) or 15 plants (leaves). Three detached ligules and one or two leaves for each individual were left to air-dry and weighed every hour for 5 hr, after they were left to air-dry overnight (o.n.), and after they were incubated in an oven to remove any residual humidity (oven-dry). Asterisks denote significant differences (p<0.05, two-sided Welch t-test; exact p-values are reported in Figure 4—source data 14). (k) Genotype-environment association (GEA) for summer average temperature (Av. T) and summer RH in the HaMYB111 region. The dashed orange line represents Bayes factor (BFis) = 10 deciban (dB). GEAs were calculated using two-sided XtX statistics. n = 71 populations.

Figure 4—source data 1

Pollinator experiment data.

https://cdn.elifesciences.org/articles/72072/elife-72072-fig4-data1-v1.xlsx
Figure 4—source data 2

Temperature measurements from infrared pictures for individual detached ligules.

https://cdn.elifesciences.org/articles/72072/elife-72072-fig4-data2-v1.xlsx
Figure 4—source data 3

Ligules and leaves desiccation experiment data.

https://cdn.elifesciences.org/articles/72072/elife-72072-fig4-data3-v1.xlsx
Figure 4—source data 4

Genotype-environment association (GEA) results for the HaMYB111 region.

https://cdn.elifesciences.org/articles/72072/elife-72072-fig4-data4-v1.xlsx

In recent years, the importance of non-pollinator factors in driving selection for floral traits has been increasingly recognized (Strauss and Whittall, 2006). Additionally, flavonol glycosides, the pigments responsible for floral UV patterns in sunflower, are known to be involved in responses to several abiotic stressors (Korn et al., 2008; Nakabayashi et al., 2014b; Pollastri and Tattini, 2011; Schulz et al., 2015). Therefore, we explored whether some of these stressors could drive diversification in floral UV pigmentation. An intuitively strong candidate is UV radiation, which can be harmful to plant cells (Stapleton, 1992). Variation in the size of UV bullseye patterns across the range of Argentina anserina (a member of the Rosaceae family) has been shown to correlate positively with intensity of UV radiation. Flowers of this species are bowl-shaped, and larger UV-absorbing regions have been proposed to protect pollen from UV damage by absorbing UV radiation that would otherwise be reflected toward the anthers (Koski and Ashman, 2015). However, sunflower inflorescences are much flatter than A. anserina flowers, making it unlikely that any significant amount of UV radiation would be reflected from the ligules towards the disc flowers. Studies in another plant with non-bowl-shaped flowers (Clarkia unguiculata) have found no evidence of an effect of floral UV patterns in protecting pollen from UV damage (Peach et al., 2020). Consistent with this, the associations between the intensity of UV radiation at our collection sites and floral UV patterns in H. annuus was weak (H. annuus: R2 = 0.01, p=0.12; Figure 4c, Figure 4—figure supplement 2).

Across the Potentillae tribe (Rosaceae), floral UV bullseye size is also weakly associated with UV radiation, but is more strongly correlated with temperature, with lower temperatures being associated with larger UV bullseyes (Koski and Ashman, 2016). We found a similar, strong correlation with temperature in our dataset, with lower average summer temperatures being associated with larger LUVp values in H. annuus (R2 = 0.44, p=2.4 × 10–15; Figure 4d, Figure 4—figure supplement 2). It has been suggested that the radiation absorbed by floral UV pigments could contribute to increasing the temperature of the flower, similar to what has been observed for visible pigments (Koski et al., 2020). This possibility is particularly intriguing for sunflower, in which flower temperature plays an important role in pollinator attraction; inflorescences of cultivated sunflowers consistently face east so that they warm up faster in the morning, making them more attractive to pollinators (Atamian et al., 2016; Creux et al., 2021). By absorbing more radiation, larger UV bullseyes could therefore contribute to increasing temperature of the sunflower inflorescences, and their attractiveness to pollinators, in cold climates. However, UV wavelengths represents only a small fraction (3–7%) of the solar radiation reaching the Earth’s surface (compared to >50% for visible wavelengths), and might therefore not provide sufficient energy to significantly warm up the ligules (Nunez et al., 1994). In line with this observation, different levels of UV pigmentation had no effect on the temperature of inflorescences or individual ligules exposed to sunlight (Figure 4e–g, Figure 4—figure supplement 3).

While several geoclimatic variables are correlated across the range of wild H. annuus, the single variable explaining the largest proportion of the variation in floral UV patterns in this species was summer relative humidity (RH; R2 = 0.51, p=1.4 × 10–18; Figure 4h, Figure 4—figure supplement 2), with lower humidity being associated with larger LUVp values (i.e., higher concentrations of flavonol glycosides in ligules). Lower RH is generally associated with higher transpiration rates in plants, leading to increased water loss, and flavonol glycosides are known to play an important role in responses to drought stress (Nakabayashi et al., 2014a); in particular, Arabidopsis lines that accumulate higher concentrations of flavonol glycosides due to overexpression of AtMYB12 lose water and desiccate at slower rates than wild-type plants (Nakabayashi et al., 2014b). Similarly, in a set of plants representing seven independent natural populations of H. annuus, we found that completely UV-absorbing ligules desiccate at a significantly slower rate than largely UV-reflecting ligules (Figure 4i). This is not due to general differences in transpiration rates between genotypes since we observed no comparable trend for rates of leaf desiccation in the same set of sunflower lines (Figure 4j). Transpiration from flowers can be a major source of water loss for plants, and this is known to drive, within species, the evolution of smaller flowers in populations living in dry locations (Galen, 2000; Herrera, 2005; Lambrecht, 2013; Lambrecht and Dawson, 2007; see Figure 4—figure supplement 4). While desiccation rates are only a proxy for transpiration in field conditions (Duursma et al., 2019; Hygen, 1951), and other factors might affect ligule transpiration in this set of lines, this evidence (strong correlation between LUVp and summer RH; known role of flavonol glycosides in regulating transpiration; and correlation between extent of ligule UV pigmentation and desiccation rates) suggests that variation in floral UV pigmentation in sunflowers is driven by the role of flavonol glycosides in reducing water loss from ligules, with larger floral UV patterns helping prevent drought stress in drier environments.

One of the main roles of transpiration in plants is facilitating heat dispersion at higher temperatures through evaporative cooling (Burke and Upchurch, 1989; Drake et al., 2018), which could explain the strong correlation between LUVp and temperature across the range of H. annuus (Figure 4d). Consistent with this, summer RH and summer temperatures together explain a considerably larger fraction of the variation for LUVp in H. annuus than either variable alone (R2 = 0.63, p=0.0017; Figure 1—source data 1), with smaller floral UV patterns being associated with higher RH and higher temperatures (Figure 4—figure supplement 2). Consistent with a role of floral UV pigmentation in the plant’s response to variation in both humidity and temperature, we found strong associations (dB > 10) between SNPs in the HaMYB111 region and these variables in genotype-environment association (GEA) analyses (Figure 4k, Figure 4—source data 4). Despite a more limited range of variation for LUVp, a similar trend (larger UV patterns in drier, colder environments) is present also in H. petiolaris (Figure 4—figure supplement 5). Interestingly, while the L allele at Chr_15 LUVp SNP is present in H. petiolaris (Figure 1—figure supplement 3), it is found only at a very low frequency and does not seem to significantly affect floral UV patterns in this species (Figure 2a). This could represent a recent introgression since H. annuus and H. petiolaris are known to hybridize in nature (Heiser, 1947; Yatabe et al., 2007). Alternatively, the Chr_15 LUVp SNP might not be associated with functional differences in HaMYB111 in H. petiolaris, or differences in genetic networks or physiology between H. annuus and H. petiolaris could mask the effect of this allele, or limit its adaptive advantage, in the latter species.

Conclusions

Connecting adaptive variation to its genetic basis is one of the main goals of evolutionary biology. Here, we show that regulatory variation at a single major gene, the transcription factor HaMYB111, underlies most of the diversity for floral UV patterns in the common sunflower, wild H. annuus. Variation for these floral UV patterns correlates strongly with pollinator preferences, but also with geoclimatic variables (especially RH and temperature) and desiccation rates in sunflower ligules. While the effects of floral UV patterns on pollinator attraction are well-known, these associations suggest a role of environmental factors in shaping diversity for this trait. Larger floral UV patterns, due to accumulation of flavonol glycoside pigments in ligules, could help reduce the amount of transpiration in environments with lower RH, preventing excessive water loss and maintaining ligule turgidity. In humid, hot environments (e.g., Southern Texas), lower accumulation of flavonol glycosides would instead promote transpiration from ligules, keeping them cool and avoiding overheating. The presence of UV pigmentation in the petals of Arabidopsis (also controlled by the Arabidopsis homolog of MYB111) further points to a more general protective role of these pigments in flowers since pollinator attraction is likely not critical for fertilization in this largely selfing species. It should be noted that, while we have examined some of the most likely factors explaining the distribution of variation for floral UV patterns in wild H. annuus across North America, other abiotic factors could play a role, as well as biotic ones (e.g., the aforementioned differences in pollinator assemblages, or a role of UV pigments in protection from herbivory; Gronquist et al., 2001). However, a role of floral UV patterns in reducing water loss from petals is consistent with the overall trend in increased size of floral UV patterns over the past 80 years that has been observed in herbarium specimens (Koski et al., 2020); due to changing climates, RH over land has been decreasing in recent decades, which could result in higher transpiration rates (Byrne and O’Gorman, 2018). Further studies will be required to confirm the existence of this trend and assess its strength.

More generally, our study highlights the complex nature of adaptive variation, with selection pressures from both biotic and abiotic factors shaping the patterns of diversity that we observe across natural populations. Floral diversity in particular has long been attributed to the actions of animal pollinators. Our work adds to a growing literature demonstrating the contributions of abiotic factors to this diversity.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Gene (Helianthus annuus)HaMYB111INRA Sunflower Bioinformatics ResourcesHanXRQChr15g0465131
Gene (H. annuus)HaFLS1INRA Sunflower Bioinformatics ResourcesHanXRQChr09g0258321
Gene (H. annuus)HaEF1αINRA Sunflower Bioinformatics ResourcesHanXRQChr11g0334971
Gene (Arabidopsis thaliana)AtMYB111; PFG3The Arabidopsis Information ResourceAt5G49330
Gene (A. thaliana)AtMYB12; PFG1The Arabidopsis Information ResourceAt2G47460
Strain, strain background (Helianthus spp.)Various Helianthus species and individualsUSDA, North Central Regional Plant Introduction StationSee Figure 1—source data 1 for full list
Strain, strain background (A. thaliana)Col-0Arabidopsis Biological Resource CenterCS28167
Genetic reagent (A. thaliana)myb111Arabidopsis Biological Resource CenterCS9813
Genetic reagent (A. thaliana)myb12Arabidopsis Biological Resource CenterCS9602
Genetic reagent (A. thaliana)myb12/myb111Arabidopsis Biological Resource CenterCS9980
Recombinant DNA reagentp35SCaMV::HaMYB111largeThis paperHaMYB111 CDS from sunflower lines with large LUVp, constitutive promoter
Recombinant DNA reagentp35SCaMV::HaMYB111smallThis paperHaMYB111 CDS from sunflower lines with small LUVp, constitutive promoter
Recombinant DNA reagentpAtMYB111::HaMYB111largeThis paperHaMYB111 CDS from sunflower lines with large LUVp, endogenous Arabidopsis promoter
Recombinant DNA reagentpAtMYB111::HaMYB111smallThis paperHaMYB111 CDS from sunflower lines with small LUVp, endogenous Arabidopsis promoter
Sequence-based reagentHaMYB111 CDS FThis paperPCR primerATGGGAAGGACCCCGTGTT
Sequence-based reagentHaMYB111 CDS RThis paperPCR primerTTAAGACTGAAACCATGCATCTACC
Sequence-based reagentAtMYB111 promoter FThis paperPCR primerCCTGTGCTTTAAGGCTCGAC
Sequence-based reagentAtMYB111 promoter RThis paperPCR primerTGCTTCTCGGTCTCTTCTGT
Sequence-based reagentHaMYB111 qPCR FThis paperPCR primerATGGGAAGGACCCCGTGTT
Sequence-based reagentHaMYB111 qPCR RThis paperPCR primerGCAACTCTTTCCGCATCTCA
Sequence-based reagentHaFLS1 qPCR FThis paperPCR primerAAACTACTACCCACCATGCC
Sequence-based reagentHaFLS1 qPCR RThis paperPCR primerTCCTTGTTCACTGTTGTTCTGT
Sequence-based reagentEF1α qPCR FThis paperPCR primerGTGTGTGATGTCGTTCTCCA
Sequence-based reagentEF1α qPCR RThis paperPCR primerATTCCACCCAAAGCTTGCTC
Commercial assay or kitCloneJET PCR cloning kitThermo Fisher ScientificCat. #: K1231
Commercial assay or kitCustom TaqMan SNP Genotyping AssayThermo Fisher ScientificAssay ID: ANKCD29
Commercial assay or kitTaqMan Genotyping Master MixThermo Fisher ScientificCat. #: 4371355
Commercial assay or kitRevertAid RT Reverse Transcription KitThermo Fisher ScientificCat. #: K1691
Commercial assay or kitSsoFast EvaGreen SupermixBio-RadCat. #: 1725201
Peptide, recombinant proteinPhusion High-Fidelity DNA PolymeraseThermo Fisher ScientificCat. #: F530L
Chemical compound, drugPPM (Plant Preservative Mixture)Plant Cell Technologies
Chemical compound, drugTRIzol ReagentThermo Fisher ScientificCat. #: 15596026
Software, algorithmImageJImageJ (https://imagej.nih.gov/ij/)RRID:SCR_003070v2.0.0-rc-43/1.51o
Software, algorithmTrimmomaticUsadel lab (http://www.usadellab.org/cms/?page=trimmomatic)RRID:SCR_011848v0.36
Software, algorithmNextGenMapNextGenMap (https://cibiv.github.io/NextGenMap/)RRID:SCR_005488v0.5.3
Software, algorithmGATKBroad Institute (https://gatk.broadinstitute.org/hc/en-us)RRID:SCR_001876v4.0.1.2
Software, algorithmBeagleUniversity of Washington (https://faculty.washington.edu/browning/beagle/beagle.html)RRID:SCR_00178910 Jun 18.811
Software, algorithmBWABWA (https://github.com/lh3/bwa)RRID:SCR_010910v0.7.17
Software, algorithmEMMAXUniversity of Michigan – Center for Statistical Genetics (http://csg.sph.umich.edu//kang/emmax/download/index.html)v07 Mar 2010
Software, algorithmGEMMAGEMMA (https://github.com/genetics-statistics/GEMMA)v0.98.3
Software, algorithmGCTA_GREMLGCTA (https://cnsgenomics.com/software/gcta)v1.93.2beta
Software, algorithmBOLT-REMLBOLT-LMM (https://alkesgroup.broadinstitute.org/BOLT-LMM/BOLT-LMM_manual.html)v.2.3.5
Software, algorithmAgilent MassHunterAgilent (https://www.agilent.com/en/promotions/masshunter-mass-spec)RRID:SCR_015040
Software, algorithmRThe R Project for Statistical Computing(https://www.r-project.org/)RRID:SCR_001905v3.6.2
Software, algorithm‘raster’ packageSpatial Data Science (https://rspatial.org/raster/index.html)
Software, algorithm‘interactions’ packageCRAN (https://cran.r-project.org/web/packages/interactions/)v1.1.5
Software, algorithmBayPassBayPass (http://www1.montpellier.inra.fr/CBGP/software/baypass/)v2.1
Software, algorithmFluke ConnectFluke (https://www.fluke.com/en-ca/products/fluke-software/connect)v.1.1.536.0

Plant material and growth conditions

Request a detailed protocol

Sunflower lines used in this paper were grown from seeds collected from wild populations (Todesco et al., 2020) or obtained from the North Central Regional Plant Introduction Station in Ames, IA. For all experiments except the plants used for Figure 1—figure supplement 1b, sunflower seeds were surface sterilized by immersion for 10 min in a 1.5% sodium hypochlorite solution. Seeds were then rinsed twice in distilled water and treated for at least 1 hr in a solution of 1% PPM (Plant Preservative Mixture; Plant Cell Technologies, Washington, DC), a broad-spectrum biocide/fungicide, to minimize contamination, and 0.05 mM gibberellic acid (Sigma-Aldrich, St. Louis, MO). They were then scarified, dehulled, and kept for 2 weeks at 4°C in the dark on filter paper moistened with a 1% PPM solution. Following this, seeds were kept in the dark at room temperature until they germinated. For common garden experiments, the seedlings were then transplanted in peat pots, grown in a greenhouse for 2 weeks, then moved to an open-sided greenhouse for a week for acclimation, and finally transplanted in the field at the Totem Plant Science Field Station of the University of British Columbia (Vancouver, Canada). For all other experiments, seedlings were transplanted in 2-gallon pots filled with Sunshine #1 growing mix (Sun Gro Horticulture Canada, Abbotsford, BC, Canada). Plants grown in greenhouses at the Vancouver campus of the University of British Columbia were kept at 26°C during the day and 20°C during the night, supplemented with LED light on a cycle of 16 hr days and 8 hr nights. For the wild sunflower species shown in Figure 1—figure supplement 1b, following sterilization, seeds were scarified and then dipped in fusicoccin solution (1.45 µM) for 15 min, dehulled, germinated in the dark for at least 8–10 days, and then grown in pots for 3 weeks before transplantation. One group of species was transplanted into 2-gallon pots filled with a blend of sandy loam, organic compost and mulch, and grown at the UC Davis Plant Sciences Field Station (Davis, CA) from July to October 2017. Several additional species were grown in single rows covered with mulch and spaced 0.75 m apart at the Oxford Tract Facility field (Berkeley, CA) from June to October 2021, or in a greenhouse facility at Berkeley, CA. A complete list of sunflower accessions and their populations of origin is reported in Figure 1—source data 1 and Figure 1—source data 2.

Seeds from the following Arabidopsis lines were obtained from the Arabidopsis Biological Resource Center: Col-0 (CS28167), myb111 (CS9813), myb12 (CS9602), and myb12/myb111 (CS9980). Seeds were stratified in 0.1% agar at 4°C in the dark for 4 days, and then sown in pots containing Sunshine #1 growing mix. Plants were grown in growth chambers at 23°C in long-day conditions (16 hr light, 8 hr dark).

Common garden

Request a detailed protocol

Two common garden experiments were performed, in 2016 and 2019. After germination and acclimation, plants were transplanted at the Totem Plant Science Field Station of the University of British Columbia (Vancouver, Canada). In the 2016 common garden experiment, each sunflower species was grown in a separate field. Pairs of plants from the same population were randomly distributed within each field. In the 2019 common garden experiment, plants were sown using a completely randomized design.

In the summer of 2016, 10 plants from each of the 151 selected populations of wild H. annuus, H. petiolaris, H. argophyllus, and H. niveus were grown. Plants were transplanted in the field on 25 May (H. argophyllus), 2 June (H. petiolaris and H. niveus), and 7 June 2016 (H. annuus). Up to four inflorescences from each plant were collected for visible and UV photography.

In the summer of 2019, 14 plants from each of the 106 populations of wild H. annuus were transplanted in the field on 6 June. These included 65 of the populations grown in the previous common garden experiment, and 41 additional populations that were selected to complement their geographical distribution. At least three ligules from at least two different inflorescences for each plant were collected for visible and UV photography. Ligules were selected to be as far apart from each other as possible across the inflorescence, taking care to avoid damaged or otherwise unrepresentative ligules.

Sample size for the common garden experiments was determined by the available growing space and resources. 10–14 individuals were grown for each population because this would provide a good representation of the variation present in each population, while maximizing the number of populations that could be surveyed. Researchers were not blinded as to the identity of individual samples. However, information about their populations of origin and/or LUVp phenotypes was not attached to the samples during data acquisition.

Ultraviolet and infrared photography

Request a detailed protocol

Ultraviolet patterns were imaged in whole inflorescences or detached ligules (see ‘Common garden’ section) using a Nikon D70s digital camera, fitted with a Noflexar 35 mm lens and a reverse-mounted 2-inch Baader U-Filter (Baader Planetarium, Mammendorf, Germany), which only allows the transmission of light between 320 and 380 nm. Wild sunflower species shown in Figure 1—figure supplement 1b were imaged using a Canon DSLR camera in which the internal hot mirror filter had been replaced with a UV bandpass filter (LifePixel, Mukilteo, WA). Floral UV patterns were scored as LUVp, rather than total area or diameter of the UV bullseye, because LUVp is less influenced by genetic or environmental factors affecting inflorescence size (Moyers et al., 2017). The length of the whole ligule (LL) and the length of the UV-absorbing part at the base of the ligule (LUV-abs) were measured using ImageJ (Schindelin et al., 2012; Schneider et al., 2012). LUVp was measured as the ratio between the two (LUVp = LUV-abs/LL). In some H. annuus individuals, the upper, ‘UV-reflecting’ portion of the ligules (LUV-ref) also displayed a lower level of UV absorption; in those cases, these regions were weighted at 50% of fully UV-absorbing regions using the formula LUVp = (LUV-abs/LL) + ½(LUV-ref/LL). Partial UV absorbance in the tip of ligules was more common in plants with larger floral UV patterns (Figure 1—figure supplement 2). To avoid possible confounding effects, for all experiments plants in the ‘small’ and ‘intermediate’ LUVp classes were selected to have no noticeable UV absorbance in the tips of ligules. For UV pictures of whole inflorescences, LUVp values were measured for three representative ligules chosen to be as far apart from each other as possible, and the average of those three values was used as the LUVp for the inflorescence. LUVp values for all the inflorescences or detached ligules available for each plant were averaged to obtain the LUVp value for that individual.

Infrared pictures for the experiments shown in Figure 4e–g and Figure 4—figure supplement 3 were taken using a Fluke TiX560 thermal imager (Fluke Corporation, Everett, WA) and analysed using the Fluke Connect software (v1.1.536.0). For time-series experiments on whole inflorescences, plants from populations ANN_03 (from CA, USA, with large LUVp) and ANN_55 (from TX, USA, with small LUVp) were germinated as above (see ‘Common garden’), grown in 2-gallon pots in a greenhouse until they produced four true leaves, and then moved to the field. On three separate days in August 2017, pairs of inflorescences with opposite floral UV patterns at similar developmental stages were selected and made to face east. Infrared images were taken just before sunrise, ~5 min after sunrise, and then at 0.5, 1, 2, 3, and 4 hr after sunrise.

For infrared pictures of detached ligules, plants were grown in a greenhouse. Plants with large LUVp came from populations ANN_03 (CA, USA), ANN_16 (NM, USA), and ANN_19 (NM, USA); plants with small LUVp came from populations ANN_55 and ANN_58 (both from TX, USA). Flowerheads were collected and kept overnight in a room with constant temperature of 21°C, with their stems immersed in a beaker containing distilled water. The following day, pairs of inflorescences were randomly selected from the two LUVp categories, and representative, undamaged ligules were removed and arranged on a sheet of white paper. Infrared pictures were taken immediately before exposing the ligules to the sun, and again 5, 10, and 15 min after that, at around 1 pm on 5 October 2020 (Figure 4—source data 2).

Library preparation, sequencing, and SNP calling

Request a detailed protocol

Whole-genome shotgun (WGS) sequencing library preparation and sequencing, as well as SNP calling and variant filtering, for the H. annuus and H. petiolaris individuals used for GWAS analyses in this paper were previously described (Todesco et al., 2020). Briefly, DNA was extracted from leaf tissue using a modified CTAB protocol (Murray and Thompson, 1980; Zeng et al., 2002), the DNeasy Plant Mini Kit, or a DNeasy 96 Plant Kit (QIAGEN, Hilden, Germany). Genomic DNA was sheared to an average fragment size of 400 bp using a Covaris M220 ultrasonicator (Covaris, Woburn, MA). Libraries were prepared using a protocol largely based on Rowan et al., 2015, the TruSeq DNA Sample Preparation Guide from Illumina (Illumina, San Diego, CA), and Rohland and Reich, 2012, with the addition of an enzymatic repeats depletion step using a Duplex-Specific Nuclease (DSN; Evrogen, Moscow, Russia) (Matvienko et al., 2013; Shagina et al., 2010; Todesco et al., 2020). All libraries were sequenced at the McGill University and Génome Québec Innovation Center on HiSeq2500, HiSeq4000, and HiSeqX instruments (Illumina) to produce paired end, 150 bp reads.

Sequences were trimmed for low quality using Trimmomatic (v0.36) (Bolger et al., 2014) and aligned to the H. annuus XRQv1 genome (Badouin et al., 2017) using NextGenMap (v0.5.3) (Sedlazeck et al., 2013). We followed the best practice recommendations of the Genome Analysis ToolKit (GATK) (Poplin et al., 2017) and executed steps documented in GATK’s germline short variant discovery pipeline (for GATK 4.0.1.2). During genotyping, to reduce computational time and improve variant quality, genomic regions containing transposable elements were excluded (Badouin et al., 2017). We then used GATK’s VariantRecalibrator (v4.0.1.2) to select high-quality variants. SNP data were then filtered for minor allele frequency (MAF) ≥ 0.01, genotype rate ≥ 90%, and to keep only biallelic SNPs.

Filtered SNPs were then remapped to the improved reference assembly HA412-HOv2 (Staton and Lázaro-Guevara, 2020) using BWA (v0.7.17) (Li, 2013). These remapped SNPs were used for all analyses, excluding the GWAS for the region surrounding the HaMYB111 locus that used unfiltered variants based on the XRQv1 assembly (Figure 2—figure supplement 3).

The SNP dataset used to determine the genotype at the Chr15_LUVp SNP in other species (H. argophyllus, H. niveus, H. debilis, H. decapetalus, H. divaricatus, and H. grosseserratus) was based on WGS data generated for Todesco et al., 2020 and is described in Owens et al., 2021. Sequence data for the Sunflower Association Mapping population are reported in Hübner et al., 2019.

Genome-wide association mapping

Request a detailed protocol

Genome-wide association analyses for LUVp were performed for H. annuus, H. petiolaris petiolaris, and H. petiolaris fallax using two-sided mixed models implemented in EMMAX (v07Mar2010) (Kang et al., 2010) or in the EMMAX module in EasyGWAS (Grimm et al., 2017). For all runs, the first three principal components (PCs) were included as covariates, as well as a kinship matrix. Only SNPs with MAF > 5% were included in the analyses, and variants were imputed and phased using Beagle (version 10Jun18.811) (Browning et al., 2018 #497). A GWAS with MAF > 1% in H. petiolaris petiolaris failed to find any additional association between LUVp and variation at the Chr15_LUVp SNP (the L allele is found at a frequency of ~2% in H. petiolaris petiolaris). Sample size was estimated to be sufficient to provide an 85% probability of detecting loci explaining 5% or more of the phenotypic variance in H. annuus. An 85% probability of detecting loci explaining 8% of variance in H. petiolaris was estimated for the whole species set (488 individuals); upon analysis of resequencing data for this species, three distinct clusters of individuals were detected (H. petiolaris petiolaris, H. petiolaris fallax, H. niveus canescens), and GWAS were performed independently on H. petiolaris petiolaris and H. petiolaris fallax (the H. niveus canescens cluster included only 86 individuals). Subspecies dataset were found to provide sufficient power to detect strong associations with adaptive traits (Todesco et al., 2020). Narrow-sense heritability (h2) in the H. annuus dataset was estimated using EMMAX (Kang et al., 2010), GEMMA (Zhou and Stephens, 2012), GCTA-GREML (Yang et al., 2011), and BOLT_REML (Loh et al., 2015). All software produced h2 values of ~1: while it is possible that the presence of a single locus of very large effect would lead to inflation of these estimates, all individuals in the GWAS populations were grown at the same time under uniform conditions, and limited environmental effects are therefore expected.

F2 populations and genotyping

Request a detailed protocol

Individuals from population ANN_03 from CA, USA (large LUVp), and ANN_55 from TX, USA (small LUVp), were grown in 2-gallon pots in a field. When the plants reached maturity, they were moved to a greenhouse, where several inflorescences were bagged and crossed. The resulting F1 seeds were germinated and grown in a greenhouse, and pairs of siblings were crossed (wild sunflowers are self-incompatible). The resulting F2 populations were grown both in a greenhouse in the winter of 2019 (n = 42 individuals for population 1, 38 individuals for population 2) and in a field as part of the 2019 common garden experiments (n = 54 individuals for population 1, 50 individuals for population 2). DNA was extracted from young leaf tissue as described above. All F2 plants were genotyped for the Chr15_LUVp SNPs using a custom TaqMan SNP genotyping assay (Thermo Fisher Scientific, Waltham, MA) on a Viia 7 Real-Time PCR system (Thermo Fisher Scientific).

Metabolite analyses

Request a detailed protocol

Methanolic extractions were performed following Stracke et al., 2007. Sunflower ligules (or portions of them) and Arabidopsis petals were collected and flash-frozen in liquid nitrogen. For sunflower, all ligules, or part of ligules, were collected from the selected inflorescence (avoiding damaged ligules). At least two ligules (or parts of ligules) were then randomly chosen, pooled, and weighed for methanolic extraction from each inflorescence. For Arabidopsis, hundreds of petals from several plants for each genotype were collected, pooled, and weighed to obtain a sufficient amount of tissue. The frozen tissue was ground to a fine powder by adding 10–15 zirconia beads (1 mm diameter) and using a TissueLyser (QIAGEN) for sunflower ligules, or using a plastic pestle in a 1.5 ml tube for Arabidopsis petals. 0.5 ml of 80% methanol were added, and the samples were further homogenized and incubated at 70°C for 15 min. They were then centrifuged at 15,000 × g for 10 min, and the supernatant was dried in a SpeedVac (Thermo Fisher Scientifics) at 60°C. Samples were then resuspended in 1 µl (sunflower) or 2.5 µl (Arabidopsis) of 80% methanol for every milligram of starting tissue.

The extracts were analysed by LC/MS/MS using an Agilent 1290 UHPLC system (Agilent Technologies, Santa Clara, CA) coupled with an Agilent 6530 Quadrupole Time of Flight mass spectrometer. The chromatographic separation was performed on Atlantis T3- C18 reversed-phase (50 mm × 2.1 mm, 3 µm) analytical columns (Waters Corp, Milford, MA). The column temperature was set at 40°C. The elution gradient consisted of mobile phase A (water and 0.2% formic acid) and mobile phase B (acetonitrile and 0.2% formic acid). The gradient program was started with 3% B, increased to 25% B in 10 min, then increased to 40% B in 13 min, increased to 90% B in 17 min, held for 1 min, and equilibrated back to 3% B in 20 min. The flow rate was set at 0.4 ml/min and injection volume was 1 µl. A photo diode array (PDA) detector was used for detection of UV absorption in the range of 190–600 nm.

MS and MS/MS detection were performed using an Agilent 6530 accurate mass Quadrupole Time of Flight mass spectrometer equipped with an ESI (electrospray) source operating in both positive and negative ionization modes. Accurate positive ESI LC/MS and LC/MS/MS data were processed using the Agilent MassHunter software to identify the analytes. The ESI conditions were as follows: nebulizing gas (nitrogen) pressure and temperature were 30 psi and 325°C; sheath gas (nitrogen) flow and temperature were 12 l/min, 325°C; dry gas (nitrogen) was 7 l/min. Full scan mass range was 50–1700 m/z. Stepwise fragmentation analysis (MS/MS) was carried out with different collision energies depending on the compound class.

Transgenes and expression assays

Request a detailed protocol

Total RNA was isolated from mature and developing ligules, or part of ligules, using TRIzol (Thermo Fisher Scientific), and cDNA was synthesized using the RevertAid First Strand cDNA Synthesis kit (Thermo Fisher Scientific). All ligules, or part of ligules, were collected from the selected inflorescence in a single tube (avoiding damaged ligules) and flash-frozen in liquid nitrogen. At least two full ligules (or parts of ligules) were then randomly chosen and pooled for RNA extraction from each inflorescence. Genomic DNA was extracted from leaves of Arabidopsis using CTAB (Murray and Thompson, 1980). A 1959-bp-long fragment (pAtMYB111) from the promoter region of AtMYB111 (At5g49330), including the 5′-UTR of the gene, was amplified using Phusion High-Fidelity DNA polymerase (New England Biolabs, Ipswich, MA) and introduced in pFK206 derived from pGREEN (Hellens et al., 2000). Alleles of HaMYB111 (HanXRQChr15g0465131) were amplified from cDNA from ligules of individuals from populations ANN_03 (large LUVp, from CA) and ANN_55 (small LUVp, from TX). These are the same populations from which the parental plants of the F2 populations shown in Figure 2e were derived. A comparison of the patterns of polymorphisms between these two alleles (HaMYB111_large and HaMYB111_small), other HaMYB111 CDS alleles from wild H. annuus, and the cultivated reference XRQ sequence is shown in Figure 3—figure supplement 2. These alleles were placed under the control of pAtMYB111 (in the plasmid described above) or of the constitutive CaMV 35S promoter (in pFK210, derived as well from pGREEN; Hellens et al., 2000). Constructs were introduced into Arabidopsis plants by Agrobacterium tumefaciens -mediated transformation (strain GV3101) (Weigel and Glazebrook, 2002). At least five independent transgenic lines with levels of UV pigmentation comparable to the ones shown in Figure 3d were recovered for each construct. For expression analyses, qRT-PCRs were performed on cDNA from ligules using the SsoFast EvaGreen Supermix (Bio-Rad, Hercules, CA) on a CFX96 Real-Time PCR Detection System (Bio-Rad). Expression levels were normalized against HaEF1α. HaEF1α (HanXRQChr11g0334971) was selected as a reference gene because, out of a set of genes that showed constitutively elevated expression across different tissues and treatments in cultivated sunflower (Badouin et al., 2017), it displayed the most robust expression patterns across ligules of different H. annuus and H. petiolaris individuals, and across ligule tips and bases in the two species. For the expression analyses shown in Figure 3h and i, portions of ligules were collected at different developmental stages from three separate inflorescences from one individual for each species (biological replicates). Three qRT-PCRs were run for each sample (technical replicates). For the expression analysis shown in Figure 3j, samples were collected from wild H. annuus individuals grown as part of the 2019 common garden experiment. Ligules were collected on the same day from developing inflorescences of 24 individuals with large LUVp (from 10 populations) and 22 individuals with small LUVp (from 11 populations). qPCRs for three technical replicates were performed for each individual. These plants were genotyped for the Chr15_LUVp SNP using a custom TaqMan assay (see ‘F2 populations and genotyping’) on a CFX96 Real-Time PCR Detection System (Bio-Rad). Sample size for this experiment was determined by the number of available plants with opposite LUVp phenotypes and at the appropriate developmental stage on the day in which samples were collected. Primers used for cloning and qRT-PCR are given in the Key resources table.

Sanger and PacBio sequencing

Request a detailed protocol

Fragments ranging in size from 1.5 to 5.5 kbp were amplified using Phusion High-Fidelity DNA polymerase (New England Biolabs) from genomic DNA of 20 individuals that had been previously resequenced (Todesco et al., 2020) and whose genotype at the Chr15_LUVp SNP was therefore known. Fragments were then cloned in either pBluescript or pJET (Thermo Fisher Scientific) and sequenced on a 3730S DNA analyzer using BigDye Terminator v3.1 sequencing chemistry (Applied Biosystems, Foster City, CA).

For long read sequencing, seed from wild H. annuus populations known to be homozygous for different alleles at the Chr15_LUVp SNP were germinated and grown in a greenhouse. After confirming that they had the expected LUVp phenotype, branches from each plant were covered with dark cloth for several days, and young, etiolated leaves were collected and immediately frozen in liquid nitrogen. High molecular weight (HMW) DNA was extracted from six plants using a modified CTAB protocol (Stoffel et al., 2012). All individuals were genotyped for the Chr15_LUVp SNP using a custom TaqMan SNP genotyping assay (Thermo Fisher Scientific, see above) on a CFX96 Real-Time PCR Detection System (Bio-Rad). Two individuals, one with large and one with small LUVp, were selected. HiFi library preparation and sequencing on a Sequel II instrument (PacBio, Menlo Park, CA) were performed at the McGill University and Génome Québec Innovation Center. Each individual was sequenced on an individual SMRT cell 8M, resulting in average genome-wide sequencing coverage of 6–8×.

Pollinator preferences assays

Request a detailed protocol

In September 2017, pollinator visits were recorded in individual inflorescences of pairs of plants with large (from population ANN_03, from CA) and small LUVp (from population ANN_55, from TX) grown in pots in a field adjacent the Nursery South Campus greenhouses of the University of British Columbia. Populations ANN_55 and ANN_03 were chosen because they flowered at about the same time in our 2016 common garden experiment and had inflorescences of similar size and appearance. Pairs of size-matched inflorescences, made to face towards the same direction, were filmed using a Bushnell Trophy Cam HD (Bushnell, Overland Park, KS) in 12 min intervals. Visitation rates were averaged over 14 such movies (Figure 4—source data 1). The only other sunflowers present in the field were H. anomalus individuals, grown in a separate field about 15 m away. H. anomalus has uniformly small floral UV patterns (Figure 1—figure supplement 1), and is therefore unlikely to have affected pollinator preferences.

In summer 2019, pollinator visits were scored in a common garden experiment consisting of 1484 H. annuus plants at the Totem Plant Science Field Station of the University of British Columbia (see the ‘Common gardens’ section for details on field design). Over 5 days, between 29 July and 7 August, pollinator visits on individual plants were directly observed and counted over 5 min intervals for a total of 435 series of measurements on 111 plants from 51 different populations (Figure 4—source data 1). Observers were careful to be at least 2 m away from the plant, and not to overshadow it. Visits to all inflorescences for each plant were recorded; pollinators visiting more than one inflorescence per plant were recorded only once. To generate a more homogenous and comparable dataset, measurements for plants with too few (1) or too many (>10) inflorescences were excluded from the final analysis (Figure 4—source data 1).

Correlations with environmental variables and GEA analyses

Request a detailed protocol

Twenty topo-climatic factors were extracted from climate data collected over a 30-year period (1961–1990) for the geographical coordinates of the population collection sites using the software package Climate NA (Wang et al., 2016; Figure 1—source data 1). Additionally, UV radiation data were extracted from the glUV dataset (Beckmann et al., 2014) using the R package ‘raster’ (Hijmans, 2020; R Development Core Team, 2020). Correlations between individual environmental variables and LUVp was calculated using the ‘lm’ function implemented in R. A correlation matrix between all environmental variables, and LUVp was calculated using the ‘cor’ function in R and plotted using the ‘heatmap.2’ function in the ‘gplots’ package (Warnes et al., 2009). Plots of the interactions between relative humidity and average temperature in relation to LUVp were generated using the ‘interact_plot’ function implemented in the ‘interactions’ R package (Long, 2020). It should be noted that the values for climate variables used in these analyses are extrapolated from weather stations across North America, and not measured in situ, meaning that they might not account for microclimatic variation. For example, two populations in Southern Arizona do not fit the pattern we proposed – they have small floral UV patterns and high frequency of S alleles at the Chr15_LUVp SNP, despite being associated with relatively low RH values in our datasets. However, one of them (ANN_13) was collected along the Verde River, near Deadhorse lake, and the description of the collection site is ‘riparian forest and wetland,’ suggesting that humidity might be locally higher than in the surrounding region. Similarly, from satellite pictures, the collection site for the other population (ANN_47) appears considerably more verdant than other collection sites in Arizona.

GEAs were analysed using BayPass (Gautier, 2015) version 2.1. Population structure was estimated by choosing 10,000 putatively neutral random SNPs under the BayPass core model. The Bayes factor (denoted BFis as in Gautier, 2015) was then calculated under the standard covariate mode. For each SNP, BFis was expressed in deciban units [dB, 10 log10 (BFis)]. Significance was determined following Gautier, 2015 and employing Jeffreys’ rule (Jeffreys, 1961), quantifying the strength of associations between SNPs and variables as ‘strong’ (10 dB ≤ BFis < 15 dB), ‘very strong’ (15 dB ≤ BFis < 20 dB), and ‘decisive’ (BFis ≥ 20 dB; Figure 4—source data 4).

Desiccation assays

Request a detailed protocol

Water loss was determined by measuring changes in the weight of detached ligules and leaves over time (Duursma et al., 2019; Hygen, 1951). In the summer of 2020, fully developed inflorescences and the one or two youngest fully developed leaves from each individual were collected from well-watered, greenhouse-grown plants that had large (LUVp = 1) or small (LUVp ≤ 0.4) floral UV patterns. They were brought immediately to an environment kept at 21°C and were left overnight with their stems or petioles immersed in a beaker containing distilled water. The following morning leaves from each plant, and three ligules removed from each inflorescence (selected to be as far apart from each other as possible across the inflorescence and taking care to avoid damaged or otherwise unrepresentative ligules), were individually weighed and hanged to air dry at room temperature (21°C). Their weight was measured at 1 hr intervals for 5 hr, and then again the following morning. Leaves and ligules were then incubated for 48 hr at 65°C in an oven to determine their dry weight. Total water content was measured as the difference between the initial fresh weight (W0) and dry weight (Wd). Water loss was expressed as a fraction of the total water content of each organ using the formula [(Wi-Wd)/(W0-Wd)] × 100, where Wi is the weight of a sample at a time i. The assay was performed on ligules from 16 inflorescences from 12 individuals belonging to seven different populations of H. annuus, and on leaves from 15 individuals from eight different populations. Of the individuals used for assays on leaves, 10 were also used for assays on ligules, 4 were half-siblings of individuals used for ligule assays, and 1 belonged to a different population (Figure 4—source data 3).

Data availability

All raw sequenced data are stored in the Sequence Read Archive (SRA) under BioProjects PRJNA532579, PRJNA398560 and PRJNA736734. Filtered SNP datasets are available at https://rieseberglab.github.io/ubc-sunflower-genome/. Raw short read sequencing data and SNP datasets have been previously described in (Todesco et al., 2020). The sequences of individual alleles at the HaMYB111 locus and of HaMYB111 coding sequences have been deposited at GenBank under accession numbers MZ597473-MZ597536 and MZ410295-MZ410296, respectively. Full details and links are provided in Supplementary file 5. All other data are available in the main text or in the source data provided with the manuscript.

The following data sets were generated
    1. Todesco M
    2. Rieseberg LH
    3. Bercovich N
    4. Owens GL
    (2021) NCBI Sequence Read Archive
    ID PRJNA736734. Floral UV patterns in sunflowers: HiFI sequences.
The following previously published data sets were used
    1. Todesco M
    2. Rieseberg LH
    3. Owens GL
    4. Drummond EBM
    (2017) NCBI Sequence Read Archive
    ID PRJNA398560. Wild and Weedy Helianthus annuus whole genome resequencing.

References

    1. Daumer K
    (1958) Blumenfarben, wie sie die Bienen sehen
    Zeitschrift Für Vergleichende Physiologie 41:49–110.
    https://doi.org/10.1007/BF00340242
  1. Book
    1. Heiser CB
    2. Smith DM
    3. Clevenger SB
    4. Martin WC
    (1969)
    The North American Sunflowers
    Durham, N.C., USA: Published for the Club by the Seeman Printery.
  2. Book
    1. Jeffreys H
    (1961)
    Theory of Probability
    Clarendon Press.
  3. Software
    1. R Development Core Team
    (2020) R: a language and environment for statistical computing
    R Foundation for Statistical Computing, Vienna, Austria.
    1. Strauss SY
    2. Whittall J
    (2006)
    Non-pollinator agents of selection on floral traits
    Ecology and Evolution of Flowers 10:120–138.
  4. Book
    1. Weigel D
    2. Glazebrook J
    (2002)
    Arabidopsis: A Laboratory Manual
    CSHL Press.
    1. Wojtaszek J
    2. Maier C
    (2014)
    A microscopic review of the sunflower and honeybee mutualistic relationship
    International Journal of AgriScience 4:272–282.
    1. Zeng J
    2. Zou Y
    3. Bai J
    4. Zheng H
    (2002)
    Preparation of total DNA from recalcitrant plant taxa
    Acta Botanica Sinica 44:694–697.

Decision letter

  1. Jeffrey Ross-Ibarra
    Reviewing Editor; University of California, Davis, United States
  2. Jürgen Kleine-Vehn
    Senior Editor; University of Freiburg, Germany
  3. Ian T Baldwin
    Reviewer; Max Planck Institute for Chemical Ecology, Germany

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Genetic basis and dual adaptive role of floral pigmentation in sunflowers" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Jeff Ross-Ibarra as Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Jürgen Kleine-Vehn as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Ian T Baldwin (Reviewer #2).

All three reviewers are quite enthusiastic about the manuscript and I believe that with some relatively straightforward revisions it would be suitable for publication in eLife. The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

The primary concern echoed by the reviewers, and which came up in our discussion, was the feeling that several of the claims of the paper could do with a bit more caveats and consideration of alternative explanations. For example, reviewer #2 points out previous results on additional metabolites that could be involved. We all agreed the claims about fitness advantages could be tempered some, given that e.g. the water loss experiment is not the same as the actual measure of transpiration, and the link between bullseye size and fitness under different humidity, while likely given other evidence in the paper, is also untested. Although we felt that additional experiments could be done to solidify these claims, we also all felt the current manuscript stands as is with the conclusions softened somewhat.

Aside from this major concern, each of the reviewers had a number of additional points to consider, and I ask you kindly respond to these in turn as well.

Thank you again for submitting such a wonderful paper to eLife, and I look forward to seeing the revision – I doubt very much it would need to go back out to review.

Reviewer #1 (Recommendations for the authors):

I'm not a fan of the last line of the conclusion. While I'm not against having some fun with papers, this felt a bit flippant perhaps.

Any speculation as to what causes the odd non-linear patterns (low UV radiation and temperature) at intermediate LUVP in figure 4c and d?

Any speculation as to why is the association with temperature stronger than RH in 4k?

I couldn't quite understand figure 3h. The bars show expression relative to what? Are these stacked bars or overlapping?

Reviewer #2 (Recommendations for the authors):

The putative fitness benefits of variable bullseye size under different humidity regimes, proposed to explain the observed geographical clines in bullseye size remain untested. These functional hypotheses could have been tested in accessions that naturally vary in bullseye size, or with genome-edited lines, which are subject to alternative explanations and onerous, respectively.

Alternatively, these hypotheses could have been tested by comparing the reproductive outputs under different watering regimes of myb12 Arabidopsis mutants and those complimented with the pAtMyb111 transgene; both species are largely selfing, so that seed set would be a reasonable proxy of functional significance. Water-loss associated fitness effects of floral flavonol glycoside expression in Arabidopsis would argue strongly for a generalized effect of these glycosides that does not depend on the particular molecular composition of the glycosides.

While bar graphs of pollinator visitation data from 2019 are presented in a figure, the data from 2017 are only mentioned in the figure caption of Figure 4 supplemental.

Reviewer #3 (Recommendations for the authors):

First, thanks for a great read! I enjoyed this manuscript.

My line notes elaborate on (and identify relevant lines for) the weaknesses I identified in my public review, and I offer possible solutions for some of them. I have only two additional comments here. (1) The authors go a bit too light on methodological details for my liking (e.g. sampling design for UV absorbance, transpiration), or leave out results that could be important and interesting (e.g. would like to Sanger sequencing variation in supplement or data). I think these should be available in the methods or a supplementary text. (2) I think there are a few areas where more could be learned (optionally!), including further exploration of GWAS, and using linear models for proportion data instead of categorizing

There are many line notes, but these are solely meant to be helpful to the authors. There are NO further critiques. Line notes are only examples, typos, or compliments.

Line 26: UV patterns are not manipulated in pollinator experiments, and they are observed across populations. Other phenotypes could vary across populations that matter to pollinators. Suggest "strongly correlate with" instead of "have a strong effect on" Same is true at line 29 about UV and transpiration.

Line 43: This is just my opinion, but the phrasing of the emoji reference feels a little informal.

Figure 1 is very easy to understand without even reading the caption. Well done. My only suggestion to emphasize the scale bar, it is a bit hard to pick out in panel c.

Line 95: This portion is a bit light on details. Here, it is not clear in R and D text what happened to ssp. fallax. Also, nowhere in results (neither Figure 1b nor text) do the authors clarify if other features (e.g. genes) are in the region that might be linked to Chr15_LUVp SNP. What is the scale of LD decay in this GWAS panel? If bigger than 30kpb, I might suggest expanding Figure 2b axis to the scale of LD breakdown. I would suggest highlighting the Chr15_LUVp SNP. Lastly, we read that from Sanger sequencing results that there are non-SNP promoter-region features and coding sequence variation, I would love to see these in the supplement.

Line 99: 62% is a lot of the variation for one SNP. I'm curious how much of the heritable variation is this? Does that increase in the greenhouse for the F2 populations?

I also wonder if there might be more to learn from the GWAS, especially for H. petiolaris? GWAS works best with a normally distributed phenotype and intermediate frequency SNPs, and these are somewhat small GWASes. Computation methods exist to account for MAF and phenotype distribution effects on spurious SNP associations (or under-associations). I doubt this would change the primary result, but might be worthwhile if the authors wish to explain more variation. See, Stanton-Geddes, John, et al., "Candidate genes and genetic architecture of symbiotic and agronomic traits revealed by whole-genome, sequence-based association genetics in Medicago truncatula." PloS one 8.5 (2013): e65688.

Line 104: how do these proportions translate for individuals where UV abs was simply reduced distally? In how many of the plants was absorbance reduced distally versus absent distally? And was this evenly distributed across populations? Did this correlate with pollinator visitation? Expression?

From line 138 onwards, individuals are referred to in phenotype bins. I imagine that the binning addresses the non-normal distribution of proportion phenotypes, but it could introduce statistical issues into the visitation and transpiration design/analyses. Maybe see: Douma, Jacob C., and James T. Weedon. "Analysing continuous proportions in ecology and evolution: A practical introduction to β and Dirichlet regression." Methods in Ecology and Evolution 10.9 (2019): 1412-1430.

Line 155: "we found flavonol glycosides to be the main UV-absorbing pigments". I see that this evidence is probably in figure 3 panels a and b, but I cannot figure out how UV-absorbance is on these graphs, nor do I understand from the methods how this was evaluated. I suggest adding detail.

Line 178: "equally effective" The figures show some differences between these alleles when paired with the local promoter. It's unclear to me what we should learn from this, please interpret. Also, which plants were selected for the large and small allele coding sequences? I can't find it in the methods. I see that variation in CDS obtained from Sanger sequencing was not associated with phenotypes, but still good to report which plants were used. Similarly, I don't understand why a myb112 mutant is explored only in the background of a myb111 mutant? I buy that MYB111 is the gene that's important regardless, just letting the authors know where they lost me if I missed something critical.

Line 196: Did HaMYB111 expression correlate with genotype at the promoter region? Did it correlate quantitatively with phenotype (beyond bins)?

Line 206: How extensive is the variation in the promoter region? Obviously across the range the GEA analysis is ideal, but within populations that have fixed/nearly fixed / maintained one allele vs the other we might expect reductions in diversity. What about other population genetic tests for selection?

Line 216: in sequencing these individuals, were the promoter regions also invariant?

Line 221: "haplotype" confuses me. I thought the S allele was a single SNP?

Line 223: Given figures in the supplement showing UV patterns across Helianthus species, I am unsure whether smaller LUVp phenotype is ancestral. The authors might run a phylogenetic analysis, or soften that claim.

Line 231: panels a, b, e: see early comment. Does peak area tells you something about amount? I'm not sure from caption/main text. Was the data for panel c generated in this study or is it from the previous study mentioned in text? For panels f, h, i, and j, the number of individuals, which genotypes/populations are not given in the caption, main text, or methods – please point to source data if they are there, and ideally also summarize. Also, for qPCR, why was HaEF1alpha chosen as the comparison?

Line 264: I disagree. Pollinators learn on short timescales. I would expect learning to play a role. For example, pollinators often display "constancy" behavior, by re-visiting a phenotype they have just visited. So if big bullseyes are common (which I expect based on figure 1), there could be a self-reinforcing bias towards visits to big bullseyes. What BC won't have is genetic variation or species filtering patterns in pollinators that might change the relative fitness of UV phenotypes across the sunflower range (though I otherwise agree with the authors' reasoning in lines 283-288). Also, can the authors comment on whether it is known if pollinators care more about proportion of UV absorbing area or the total size? Here, the authors have considered proportion only.

Line 269: This design confounds any phenotypes that are co-correlated to the UV phenotype across sunflower populations (e.g. due to local adaptation or n\eutral population structure). Therefore, pollinator preferences could arguably be due to co-varying phenotypes. Possible solution: Given that F2s within two different populations were included in the field study, was any pollinator preference data taken for these? Or are any other within population large versus small LUVp comparisons?

Line 279: I have no idea from the evidence presented how visitation links to fitness. Even the lowest rates of visitation to inflorescences seem high (>20 visits per hour, to, I assume, a limited number of florets where one visit might be sufficient).

Line 312: "explaining" is a strong word choice for a correlation.

Line 336: I have the same concern here that I do for line 269 the pollination experiment. The authors do consider and eliminate one possible co-correlated phenotype (leaf transpiration, well done) – but there are others I could think of: ligule total size, stomata density on ligules… The authors could consider the F2 individuals, or even just within population individuals, and test if this pattern is independent of population structure.

Line 339: Is there any evidence that ligules are the main source of inflorescence transpiration in sunflowers? It occurs to me that water loss might also drive pollinator preferences. Presumably pollinators avoid inflorescences with wilt – e.g. because nectar production would be low, or flowers might be old with no pollen.

Line 348-350: I am struggling to understand. The ANOVA in the source data doesn't include the fitted slopes. How does the expected effect of RH change with temperature?

Line 356: This is one result the authors interpret with the ideal amount of caution (this is a compliment). I'd be curious to know if other regions of the genome also appear in GEA, and whether any of those co-locate to other SNPs with larger estimated associations with phenotype (even if those are n.s.).

Figure 4. Why is the x-axis in 4k different from 2b? Figure 4 is otherwise functionally excellent and very aesthetically pleasing. I'm very impressed with these great figures overall, despite my critiques of Figure 3.

Conclusions: The statements on effects of pollinators and transpiration should be softened to fit the evidence. On line 390: Makes sense for Texas, but Arizona also has high S allele prevalence in Figure 2, and to my knowledge is less humid. Line 406: How is the transpiration cost of sex not about sex? I suggest deleting this sentence.

https://doi.org/10.7554/eLife.72072.sa1

Author response

Reviewer #1 (Recommendations for the authors):

Minor comments:

I'm not a fan of the last line of the conclusion. While I'm not against having some fun with papers, this felt a bit flippant perhaps.

We have removed the last sentence.

Any speculation as to what causes the odd non-linear patterns (low UV radiation and temperature) at intermediate LUVP in figure 4c and d?

We believe that simply reflects the fact that those environmental factors (even average summer temperature, for which there is an overall good correlation with LUVp) are not the best predictors of floral UV patterns (or not as good as relative humidity). In both cases, the visible dips in the intermediate LUVp are at least partially driven from several populations with intermediate LUVp found at high latitudes (Canada, North Dakota), where UV radiation is weaker and temperatures are lower.

Any speculation as to why is the association with temperature stronger than RH in 4k?

The stronger association with temperature could hint at an effect of HaMYB111 on other traits that are important for responses to temperature, but not, or not as much, for responses to changes in relative humidity. However, in our experience the strength of associations in GEA analyses is quite sensitive to non-biological factors (using different sets of putatively neutral SNPs can affect the relative strengths of associations), and it is therefore hard to tell how much of that difference is biologically relevant. Additionally, the resulting Bayes Factors also depend on the environmental trait distribution across the landscape, so we’re cautious in overinterpreting one factor as being more important purely based on this.

I couldn't quite understand figure 3h. The bars show expression relative to what? Are these stacked bars or overlapping?

The bars in that figure are stacked bars – that was meant to recall the organization of the sunflower ligule, with the UV-absorbing part at the bottom, and the UV-reflecting part at the top. However, in hindsight that might not have been as intuitive as we had hoped. We have re-designed figures 3h and 3i to hopefully make the interpretation of the experiment more straightforward, by separating expression data for ligule tips and bases.Apologies for omitting to specify that the expression levels are normalized against the average expression levels in the UV-absorbing part of developing ligules in H. annuus. We have included that information in the figure legend, also for Figure 3j.

Reviewer #2 (Recommendations for the authors):

The putative fitness benefits of variable bullseye size under different humidity regimes, proposed to explain the observed geographical clines in bullseye size remain untested. These functional hypotheses could have been tested in accessions that naturally vary in bullseye size, or with genome-edited lines, which are subject to alternative explanations and onerous, respectively.

We would have loved to be able to use transgenic or GE lines to test directly the effects of different alleles of HaMYB111, but unfortunately sunflower is exceptionally resistant to transformations (despite the fact that several sunflower transformation protocols have been proposed, none worked in our hand, and no functional work using those protocol has ever been published). Throughout the manuscript, we have tried to compensate for that by comparing groups of accessions with different floral UV pattern, to try to average out other factors that could differentiate individual lines. However, this would be particularly complicated for measurements of fitness, which are affected by virtually all other plant traits, meaning that all but the strongest effects would be impossible to detect (and we don’t think that variation in levels of transpiration in ligules would have an oversized effect on classic measures of plant fitness like seed set). An acceptable compromise would be to use isogenic lines in which different alleles of HaMYB111 have been introduced into cultivated (self-compatible) sunflower. While we are working on that, obtaining suitable lines is likely to take several years.

Alternatively, these hypotheses could have been tested by comparing the reproductive outputs under different watering regimes of myb12 Arabidopsis mutants and those complimented with the pAtMyb111 transgene; both species are largely selfing, so that seed set would be a reasonable proxy of functional significance. Water-loss associated fitness effects of floral flavonol glycoside expression in Arabidopsis would argue strongly for a generalized effect of these glycosides that does not depend on the particular molecular composition of the glycosides.

While this would be a much more feasible experiment than using sunflowers, it would be complicated by the fact that AtMYB111, while strongly expressed in petals, is expressed also throughout the plants, and the myb111 mutant has altered flavonol profiles in different tissues, including seedlings and rosette and cauline leaves (Stracke et al., Plant J. 2007; Stracke et al., New Phytol. 2010). Since accumulation of flavonol glycosides affects water loss rates in leaves/rosettes (Nakabayashi et al., Plant J. 2014), as well as possibly other traits, it would be difficult to isolate the effect of floral UV pigmentation of plant fitness. Additionally, since sunflower ligules are much larger than Arabidopsis petals (even in proportion to the whole plant), the eventual effect on transpiration might be much less relevant. As mentioned above, sunflower isogenic lines will hopefully be a more suitable system to answer these questions.

While bar graphs of pollinator visitation data from 2019 are presented in a figure, the data from 2017 are only mentioned in the figure caption of Figure 4 supplemental.

We did not provide a plot with pollinator-specific information for the 2017 experiment because almost all visits were from bumblebee, with only nine visits from sylphid flies, and it would have been difficult to visualize a difference. We have now added more detailed information on the number of visits from sylphid flies in the 2017 experiment to the legend of Figure 4 —figure supplement 1, which now reads:

“In the 2017 field experiment, pollinators were overwhelmingly bumblebees. The only other pollinators recorded were syrphid flies, which visited inflorescences with large LUVp seven times (7.9% of total visits on these inflorescences) and inflorescences with small LUVp two times (3.7% of total visits on these inflorescences).” (new lines 1297-1300)

Reviewer #3 (Recommendations for the authors):

First, thanks for a great read! I enjoyed this manuscript.

Thank you!

My line notes elaborate on (and identify relevant lines for) the weaknesses I identified in my public review, and I offer possible solutions for some of them. I have only two additional comments here. (1) The authors go a bit too light on methodological details for my liking (e.g. sampling design for UV absorbance, transpiration), or leave out results that could be important and interesting (e.g. would like to Sanger sequencing variation in supplement or data). I think these should be available in the methods or a supplementary text. (2) I think there are a few areas where more could be learned (optionally!), including further exploration of GWAS, and using linear models for proportion data instead of categorizing

There are many line notes, but these are solely meant to be helpful to the authors. There are NO further critiques. Line notes are only examples, typos, or compliments.

Line 26: UV patterns are not manipulated in pollinator experiments, and they are observed across populations. Other phenotypes could vary across populations that matter to pollinators. Suggest “strongly correlate with” instead of “have a strong effect on” Same is true at line 29 about UV and transpiration.

We have modified the abstract as suggested on line 26. We have re-phrased line 29 to better reflect the results of our experiments (that is, that ligules with larger LUVp patterns show reduced water loss, but not directly that larger UV patterns cause reduced water loss). The two sentences now read:

“Different patterns of ultraviolet pigments in flowers are strongly correlated with pollinator preferences.” (new lines 26-27)

“Ligules with larger ultraviolet patterns, which are found in drier environments, show increased resistance to desiccation, suggesting a role in reducing water loss.” (new lines 29-31)

We have also modified slightly line 26 to include a definition of “ligules”.

Line 43: This is just my opinion, but the phrasing of the emoji reference feels a little informal.

That was our attempt to lighten the tone of the introduction, and was received with mixed review even among authors. The author that wrote the initial draft of the manuscript still finds it moderately funny, so instead of removing it we have contextualized the reference a bit better. The sentence now reads:

“Much of the popularity of sunflowers (as testified by countless references in the visual arts and, more recently, by the arguably dubious honour of being one of the only five flower species with a dedicated emoji (Unicode.org, 2020)) is due to their iconic yellow inflorescences.” (new lines 58-61)

Figure 1 is very easy to understand without even reading the caption. Well done. My only suggestion to emphasize the scale bar, it is a bit hard to pick out in panel c.

Thank you. We have increased the size bar to 2 cm, and used a thicker line.

Line 95: This portion is a bit light on details. Here, it is not clear in RandD text what happened to ssp. Fallax.

We have included some more information about the (sub-)species of wild sunflowers that are the focus of the manuscript in the previous section of RandD (“Floral UV patterns in wild sunflowers”), and we have added a mention of the fact that no significant association was found in GWAS for LUVp in H. petiolaris fallax (a reference to the fallax Manhattan plot in Figure 2 —figure supplement 1 was already present). The sentence now reads:

“While no significant association was identified for H petiolaris fallax (Figure 2 —figure supplement 1), we detected several genomic regions significantly associated with UV patterning in H. petiolaris petiolaris…” (new lines 150-152)

Also, nowhere in results (neither Figure 1b nor text) do the authors clarify if other features (e.g. genes) are in the region that might be linked to Chr15_LUVp SNP. What is the scale of LD decay in this GWAS panel? If bigger than 30kpb, I might suggest expanding Figure 2b axis to the scale of LD breakdown. I would suggest highlighting the Chr15_LUVp SNP.

We have highlighted the Chr15_LUVp SNP in Figure 2b and have include the following sentence to the legend of Figure 1b, to clarify that no other gene or annotated feature is found in the depicted interval, and that LD decays rapidly in our wild H. annuus panel:

HaMYB111 is the only annotated feature in the genomic interval shown in Figure 1b; the SNP with the strongest association to LUVp (Chr15_LUVp SNP) is highlighted in yellow. Linkage disequilibrium (LD) decays rapidly in wild H. annuus (average R2 at 10 kbp is ~0.035 (Todesco et al., 2020)), and all SNPs significantly associated with LUVp in H. annuus are included in the depicted region.” (new lines 184-187)

Lastly, we read that from Sanger sequencing results that there are non-SNP promoter-region features and coding sequence variation, I would love to see these in the supplement.

We have added new Figure 3 —figure supplement 2 showing the polymorphisms in the coding sequence of HaMYB111 across a set of 18 wild H. annuus alleles, compared to the cultivated sunflower reference XRQ. Multiple sequence alignments for genomic regions were too large to be included as figure supplements, and contained too many polymorphisms to be meaningfully summarized in a figure similar to Figure 3 —figure supplement 2. We have provided graphical representations of these alignments as Supplementary files 1-4 (CDS, genomic HaMYB111 region, proximal promoter region, distal promoter region), and included additional details about patterns of variation in the promoter region of HaMYB111 in the legend of Figure 3 —figure supplement 2.

“Extensive sequence and structural variation across H. annuus individuals was found in intron regions as well as in the putative promoter region of HaMYB111. While the vast majority of the polymorphisms in introns or in the proximal promoter region (from the Chr15_LUVp SNP to the transcription start) did not appear to correlate with LUVp values or with genotypes at the Chr15_LUVp SNP, several large polymorphisms associated with both were found in the distal promoter region, upstream of the Chr15_LUVp SN. However, the distal promoter region could be amplified and sequenced in its entirety only from a subset of sunflower lines carrying the L allele at Chr15_LUVp SNP, precluding a conclusive determination of a link between this sequence variation and functional diversity between alleles of HaMYB111. Additionally, given that the promoter region had to be split in two large regions (proximal and distal), with limited overlap, to be amplified before Sanger sequencing, we cannot exclude the presence of more complex rearrangements in the region. Alignments for HaMYB111 coding and genomic sequences, and for the proximal and distal promoter regions, are provided as Supplementary files 1-4.” (new lines 1275—1288)

Line 99: 62% is a lot of the variation for one SNP. I’m curious how much of the heritable variation is this? Does that increase in the greenhouse for the F2 populations?

I also wonder if there might be more to learn from the GWAS, especially for H. petiolaris? GWAS works best with a normally distributed phenotype and intermediate frequency SNPs, and these are somewhat small GWASes. Computation methods exist to account for MAF and phenotype distribution effects on spurious SNP associations (or under-associations). I doubt this would change the primary result, but might be worthwhile if the authors wish to explain more variation. See, Stanton-Geddes, John, et al., "Candidate genes and genetic architecture of symbiotic and agronomic traits revealed by whole-genome, sequence-based association genetics in Medicago truncatula." PloS one 8.5 (2013): e65688.

As mentioned in one of our replies to Reviewer 1, heritability is extremely high for LUVp in our H. annuus GWAS population (~1, as calculated by four separate software) – meaning that the percentage of phenotypic variation explained by the Chr15_LUVp SNP is the same as the percentage of additive variation explained by that SNP.

We ran GWA analyses for H. annuus with a different software (GEMMA, Zhou and Stephens, Nat Genet. 2012), which has sometimes produced clearer associations in our hands, but results were virtually indistinguishable. Although other approaches might identify better associations for the H. petiolaris datasets, we suspect that their limited sample size (due to the separation between subspecies) would allow only limited power to detect most associations. While exploring more in details the genetics of floral UV patterns in other sunflower species is of course of interest to us, we preferred therefore to focus on H. annuus in the present manuscript.

Line 104: how do these proportions translate for individuals where UV abs was simply reduced distally?

We found that accounting for partial UV-absorbance in the distal part of the ligule improved the strength of the association between LUVp and HaMYB111, but did not affect the overall GWA pattern. Ignoring that partial pigmentation (“unmodified LUVp”) resulted in only minor reductions in average LUVp values for different genotype classes for the Chr15_LUVp SNP (L/L = 0.78 -> 0.73; L/S 0.59 -> 0.55; S/S 0.33 ->0.32). We have added this information to the methods section, and included the unmodified LUVp measurements for both individuals and genotypic classed in Figure 1 – source data 2.

In how many of the plants was absorbance reduced distally versus absent distally? And was this evenly distributed across populations?

Excluding plants with completely UV-absorbing ligules, some degree of UV-absorbance in the distal part of ligules was found on ~43% of individuals. Distal UV absorbance was generally more common in ligules with larger unmodified LUVp, and rarer in plants with small unmodified LUVp values (see Author response image 1); however, it was segregating in most populations with intermediate unmodified LUVp values.

Author response image 1

Did this correlate with pollinator visitation? Expression?

In all experiments, plants in the “small” or “intermediate” LUVp classes were selected to be without noticeable distal UV absorption, to simplify data interpretation.

The corresponding methods section was amended to provide additional information on this:

“Partial UV absorbance in the tip of ligules was more common in plants with larger floral UV patterns; while accounting for this in LUVp measurements (as outline above) improved the strength of the association with the Chr15_LUVp SNP in GWAS (from P = 8.52e-19 to P = 5.81e-25), it did not change the overall pattern. Similarly, ignoring UV absorbance in the tip of ligules had only a minor effect on the average LUVp values for genotypic classes at the Chr15_LUVp SNP (Figure 1 – source data 2). To avoid possible confounding effects, for all experiments plants in the “small” and “intermediate” LUVp classes were selected to have no noticeable UV absorbance in the tips of ligules.” (new lines 644-651 Partial UV absorbance in the tip of ligules)

From line 138 onwards, individuals are referred to in phenotype bins. I imagine that the binning addresses the non-normal distribution of proportion phenotypes, but it could introduce statistical issues into the visitation and transpiration design/analyses. Maybe see: Douma, Jacob C., and James T. Weedon. "Analysing continuous proportions in ecology and evolution: A practical introduction to β and Dirichlet regression." Methods in Ecology and Evolution 10.9 (2019): 1412-1430.

As mentioned in our response to Reviewer 1, experiments were designed to compare defined phenotypic classes to reduce experimental noise and simplify interpretation. As a consequence, plants with LUVp values falling outside of these categories (e.g. 0.3 < LUVp < 0.5 and 0.8 < LUVp < 0.95 in the 2019 pollinator visitation experiment) are not represented. Therefore, we believe that analyzing these experiments in a different way than they were designed would be more problematic. However, in the revised manuscript we have provided a modified Figure 4 —figure supplement 1 in which individual data points are show (colour-coded by pollinator type), as well as a fitted lines showing the general trend across the data. We have also modified the main text to clarify that plants were purposely selected to belong to those three phenotypic classes:

“We selected plants falling into three categories of LUVp values, representatives of the more abundant phenotypic classes across the range of wild H. annuus (Figure 1d): small (LUVp = 0-0.3); intermediate (LUVp = 0.5-0.8) and large (LUVp > 0.95).” (new lines 361-370)

Line 155: "we found flavonol glycosides to be the main UV-absorbing pigments". I see that this evidence is probably in figure 3 panels a and b, but I cannot figure out how UV-absorbance is on these graphs, nor do I understand from the methods how this was evaluated. I suggest adding detail.

Apologies for the lack of details. We have now specified in the legend of panels 3a,b that the areas of the peaks in the chromatograms are proportional to the total UV absorbance at 350 nm for the corresponding compound in each extract. We have also included the full names of the compounds in the legend of those panels, as well as in that of Figure 3e, and specified in the text that quercetin glycosides are the main flavonols in sunflower ligules (as well as addressing the presence of caffeoyl quinic acid, CQA, in ligules).

The statement that flavonol glycosides are the main UV-absorbing pigment was based on the fact that the total area under the corresponding peaks is considerably larger than that of CQA in UV-absorbing (parts of) ligules. We have modified the sentence in line 155 to better qualify this point:

“Analysis of sunflower ligules found two main UV-absorbing compounds: glycoside conjugates of quercetin (a flavonol) and di-O-caffeoyl quinic acid (CQA, a member of a family of antioxidant compounds that includes chlorogenic acid and that accumulates at high levels in many sunflower tissues (Koeppe et al., 1970)). Both quercetin glycosides and CQA were more abundant at the base of sunflower ligules, and in ligules of plants with larger LUVp. However, this pattern was much more dramatic for flavonols, and they explained a much larger fraction of the total UV absorbance in UV-absorbing (parts of) ligules, suggesting that flavonols are the main pigments responsible for UV patterning in sunflower ligules (Figure 3a,b)” (new lines 219-227)

Line 178: "equally effective" The figures show some differences between these alleles when paired with the local promoter. It's unclear to me what we should learn from this, please interpret.

We modified Figure 3d to show, for each transgenic line, a petal from three independent primary transformation events representing the range of variation observed for that line, rather than three petals from a single line. Hopefully this better represents that lack of obvious differences in the ability of different HaMYB111 alleles in complementing the myb111 mutant.

Also, which plants were selected for the large and small allele coding sequences? I can't find it in the methods. I see that variation in CDS obtained from Sanger sequencing was not associated with phenotypes, but still good to report which plants were used.

The large and small allele used in those complementation experiments were cloned from individuals from the same wild populations as the parental lines of the F2s shown in Figure 2e: ANN_03, from California (HaMYB111_large) and ANN_55, from Texas (HaMYB111_small). We have added these sequences to the new Figure 3 —figure supplement 2, where they are compared to other HaMYB111 alleles from wild H. annuus individuals, as well as to the cultivated reference XRQ sequence. We have included this information in the Methods section:

“Alleles of HaMYB111 (HanXRQChr15g0465131) were amplified from cDNA from ligules of individuals from populations ANN_03 (large LUVp, from California) and ANN_55 (small LUVp, from Texas). These are the same populations from which the parental plants of the F2 populations shown in Figure 2e were derived. A comparison between the patterns of polymorphisms between these two alleles (HaMYB111_large and HaMYB111_small), other HaMYB111 CDS alleles from wild H. annuus, and the cultivated reference XRQ sequence is shown in Figure 3 —figure supplement 2.” (new lines 791-797)

Similarly, I don't understand why a myb112 mutant is explored only in the background of a myb111 mutant? I buy that MYB111 is the gene that's important regardless, just letting the authors know where they lost me if I missed something critical.

We originally provided visible and UV pictures for several lines (including myb12) in Figure 3 —figure supplement 1, but we have now incorporated all the content of that figure supplement in Figure 3d. That figure now shows that the myb12 mutant by itself has no visible effect on petal UV absorbance, and its effect is only visible in conjunction with the myb111 mutant. Similarly, the myb111/myb12 mutant was added to Figure 3e only to show that, while AtMYB12 does contribute to flavonol accumulation in petals, its contribution is negligible compared to that of AtMYB111.

Line 196: Did HaMYB111 expression correlate with genotype at the promoter region? Did it correlate quantitatively with phenotype (beyond bins)?

The strength of the correlation is largely unchanged when HaMYB111 expression levels are analyzed based on LUVp categories (R2 = 0.123, P = 0.0097), LUVp values (R2 = 0.121, P = 0.0102), or genotypes at the Chr15_LUVp SNP (R2 = 0.088, P = 0.0282). Like for the pollinator experiments, plants in the two LUVp categories were chosen to have very divergent LUVp values (small: LUVp <0.35; large: LUVp > 0.8), which explains why the correlation is almost unchanged when using LUVp values or LUVp categories.

As mentioned in the manuscript, collecting at exactly the same ligule developmental stage across a set of 46 wild lines is nearly impossible. Together with the strong dependency of HaMYB111 expression on developmental stage, this resulted in rather noisy data, which could explain the relatively weak correlation coefficients.

We have added to the manuscript a new source data file (Figure 3 – source data 2), which includes expression data for Figures 3c,f,h,i,j, as well as new Chr15_LUVp SNP genotype data for the plants in Figure 3j (determined using a custom TaqMan assay, see Methods section). We have added information about the strength of correlation with LUVp values and Chr15_LUVp SNP genotype to the legend of Figures 3j and in Figure 3 – source data 2.

Line 206: How extensive is the variation in the promoter region? Obviously across the range the GEA analysis is ideal, but within populations that have fixed/nearly fixed / maintained one allele vs the other we might expect reductions in diversity. What about other population genetic tests for selection?

The presence of numerous large indels in the regions makes it difficult to provide a more precise quantification of the level of divergence in the region; however, to provide that information, we added graphic representations of alignments for the promoter regions of up to 15 H. annuus alleles (based on Sanger sequencing data) as Supplementary files 3 and 4.

SNP data in the HaMYB111 promoter region is sparse, since several portions of it were deemed too repetitive to be used for short read mapping (see large regions with no SNPs in Figure 2b and Figure 4k), and Sanger sequencing made it clear that those SNPs provide a very limited representation of the diversity present in the region. As a consequence, we believe selection tests done using those SNP data would not be informative.

Line 216: in sequencing these individuals, were the promoter regions also invariant?

The promoter regions that we sequenced for H. argophyllus and H. petiolaris had several polymorphisms that differentiated them from H. annuus alleles. However, they all were overall more similar to promoter regions of individuals carrying the S allele at Chr15_LUVp than to individuals carrying the L allele.

We have added that information to the main text, and provided alignments comparing the promoter region in H. annuus, H. argophyllus and H. petiolaris individuals as Supplementary files 3 and 4.

The modified sentence now reads:

“Interestingly, when we sequenced the promoter region of HaMYB111 in several H. argophyllus and H. petiolaris individuals, we found that they all carried the S allele at the Chr15_LUVp SNP, and that their promoter regions were generally more similar in sequence to those of H. annuus individuals carrying the S allele at the Chr15_LUVp SNP (Supplementary files 3,4).” (new lines 288-291)

Line 221: "haplotype" confuses me. I thought the S allele was a single SNP?

We have removed that part of the sentence, ad have clarified that the similarities between HaMYB111 alleles from H. annuus individuals with small LUVp, and H. argophyllus and H. petiolaris extends to the whole promoter region (see above).

Line 223: Given figures in the supplement showing UV patterns across Helianthus species, I am unsure whether smaller LUVp phenotype is ancestral. The authors might run a phylogenetic analysis, or soften that claim.

In that sentence we were referring specifically to the S allele at the Chr15_LUVp SNP, and by extension to the associated HaMYB111 allele, rather than to the small LUVp phenotype (other alleles or genes might be responsible for large UV patterns in other sunflower species). However, we agree that this was not very clear and not a particularly well-supported claim, so we have removed it.

Line 231: panels a, b, e: see early comment. Does peak area tells you something about amount? I'm not sure from caption/main text. Was the data for panel c generated in this study or is it from the previous study mentioned in text? For panels f, h, i, and j, the number of individuals, which genotypes/populations are not given in the caption, main text, or methods – please point to source data if they are there, and ideally also summarize. Also, for qPCR, why was HaEF1alpha chosen as the comparison?

We have clarified in the legend of the figure that the peaks area in those UV chromatograms is proportional to the total absorbance for that compound in the extract. We have also specified that the data used for Figure 3c and 3f were obtained from previous publications.

We have added a new Figure 3 – source data 2 file containing the expression data that Figure 3c,f,h,i,j are based on, as well as the IDs, populations of origin, LUVp values and genotype at the Chr15_LUVp SNP of the individuals used for the experiment in Figure 3j. We have expanded the legend and Methods section to provide more details about the qPCR experiments, and to clarify how HaEF1alpha was chosen as reference gene.

HaEF1α (HanXRQChr11g0334971) was selected as a reference gene because, out of a set of genes that showed constitutively elevated expression across different tissues and treatments in cultivated sunflower (Badouin et al., 2017), it displayed the most robust expression patterns across ligules of different H. annuus and H. petiolaris individuals, and across ligules tips and bases in the two species.” (new lines 805-809)

Line 264: I disagree. Pollinators learn on short timescales. I would expect learning to play a role. For example, pollinators often display "constancy" behavior, by re-visiting a phenotype they have just visited. So if big bullseyes are common (which I expect based on figure 1), there could be a self-reinforcing bias towards visits to big bullseyes. What BC won't have is genetic variation or species filtering patterns in pollinators that might change the relative fitness of UV phenotypes across the sunflower range (though I otherwise agree with the authors' reasoning in lines 283-288). Also, can the authors comment on whether it is known if pollinators care more about proportion of UV absorbing area or the total size? Here, the authors have considered proportion only.

The experiment shown in Figure 4a compared individual plants with either very large or very small UV patterns (the inflorescences shown in Figure 4e are actually from that experiment). Only four plants from each of these two classes were present in the field. The only other sunflowers in the vicinity (~10-20 meters away) were a couple dozen H. anomalus plants, which have uniformly small LUVp values (see Figure 1 —figure supplement 1). While a learned bias toward large UV patterns in sunflowers would therefore be unlikely in this case, we agree that sentence might over-simplify the situation, and have removed it.

Throughout the manuscript, we have focused on UV proportions because it has been shown to be a more robust way to quantify UV patterns than the total size of UV bullseyes, at least in sunflowers (Moyers et al., Ann Bot 2017), and because it is the standard in the field. While differences in inflorescence size clearly have a genetic basis, there can be large variation in size between inflorescences of a same plant, depending on developmental stage or environmental factors (and this will affect total UV size). While it is possible that total UV size would affect pollinator preferences, we do not therefore have sufficient data to test that; it should be noted, however, that the plants used for the experiment in Figure 4a were selected to have matching flowering time and inflorescence size, and the individual inflorescences that were recorded were size-matched.

We have added the information reported above in the legend of Figure 4 and in the relevant parts of the Methods section, where we have also added additional information about the experimental setting for pollinator preference assays.

Line 269: This design confounds any phenotypes that are co-correlated to the UV phenotype across sunflower populations (e.g. due to local adaptation or neutral population structure). Therefore, pollinator preferences could arguably be due to co-varying phenotypes. Possible solution: Given that F2s within two different populations were included in the field study, was any pollinator preference data taken for these? Or are any other within population large versus small LUVp comparisons?

We did not record pollinator preferences in F2 populations, and most wild populations have relatively uniform LUVp values. We have added a sentence acknowledging that presence of other, unmeasured traits that could also affect the pollinator preference patterns we observed. However, we believe that the fact that our results are consistent with what is reported in literature for the effect of UV bullseyes on pollinator preferences in other species supports our interpretation.

The section now reads:

“Therefore, we monitored pollinator visitation in plants grown in a common garden experiment including 1484 individuals from 106 H. annuus populations, spanning the entire range of the species. Assaying a much more diverse population of H. annuus individuals should reduce effects of traits unrelated to floral UV pigmentation on pollinator preferences” (new lines 356-359).

Line 279: I have no idea from the evidence presented how visitation links to fitness. Even the lowest rates of visitation to inflorescences seem high (>20 visits per hour, to, I assume, a limited number of florets where one visit might be sufficient).

Unfortunately, accurately measuring the effect of pollinator rates on fitness (i.e. seed set) would have required that we constantly measure pollinator visits to individual inflorescences, and to have hand pollinated inflorescences as a comparison, which was not feasible in our experimental setup. However, it has been shown before that pollination rates are yield-limiting in hybrid sunflower production (Greenleaf et al., PNAS 2006). We have included this information and softened our claims about selection and fitness in this section, which now reads:

“Pollination rates are known to be yield-limiting in sunflower (Greenleaf et al., 2006), and a strong reduction in pollination could therefore have a negative effect on fitness; this would be consistent with the observation that plants with very small LUVp values were rare (~1.5% of individuals) in our common garden experiment, which was designed to provide a balanced representation of the natural range of H. annuus” (new lines 373-379)

More in generally, there is ample literature showing that pollinator visits increase seed yield even in self-compatible cultivated sunflower, albeit with considerable differences in effect size (e.g. Dag et al., Am. Bee J. 2002; Nderitu et al., Span. J. Agric. Res. 2008; Mallinger and Prasifka, Crop Sci. 2017; Said et al., Pak. J. Zool. 2017; but see also Bartual et al., 2018 PLoS one and Astiz et al., Helia 2011); however, it is of course harder to translate this information to wild sunflowers.

Line 312: "explaining" is a strong word choice for a correlation.

Changed to “…with lower average summer temperatures being associated with larger LUVp values in H. annuus.” (new lines 422-423)

Line 336: I have the same concern here that I do for line 269 the pollination experiment. The authors do consider and eliminate one possible co-correlated phenotype (leaf transpiration, well done) – but there are others I could think of: ligule total size, stomata density on ligules… The authors could consider the F2 individuals, or even just within population individuals, and test if this pattern is independent of population structure.

As for the pollination experiments, we chose to focus on a diverse collection of wild genotypes because they are more representative of phenotypic variation found in the wild, but we recognize that some confounding factors could persist, and have acknowledged that in the revised manuscript. In this case as well, however, we think that the presence of independent evidence in the literature on the role of flavonol glycosides in limiting desiccation/transpiration lends support to our interpretation.

The last sentence of this section now reads:

“While desiccation rates are only a proxy for transpiration in field conditions (Duursma et al., 2019, Hygen et al., 1951), and other factors might affect ligule transpiration in this set of lines, this evidence (strong correlation between LUVp and summer relative humidity; known role of flavonol glycosides in regulating transpiration; and correlation between extent of ligule UV pigmentation and desiccation rates) suggests that variation in floral UV pigmentation in sunflowers is driven by the role of flavonol glycosides in reducing water loss from ligules, with larger floral UV patterns helping prevent drought stress in drier environments.” (new lines 462-469)

Line 339: Is there any evidence that ligules are the main source of inflorescence transpiration in sunflowers? It occurs to me that water loss might also drive pollinator preferences. Presumably pollinators avoid inflorescences with wilt – e.g. because nectar production would be low, or flowers might be old with no pollen.

To our knowledge, the amount of transpirations from different parts of the sunflower inflorescence has not been measured – however, given that they are the largest exposed surface on the inflorescence, and have a high surface-to-volume ration, it seems plausible that they represent a sizable part of the total transpiration from inflorescences.

It also seems plausible that pollinators might not like inflorescences with wilted ligules, as the reviewer suggests, since they might be old or sickly, and in an earlier version of the manuscript we had a sentence suggesting as much. However, we decided to remove it, since we had no direct evidence of that in sunflowers (although we do know that pollinators do not like inflorescences without ligules, see legend of Figure 4 —figure supplement 3).

Line 348-350: I am struggling to understand. The ANOVA in the source data doesn't include the fitted slopes. How does the expected effect of RH change with temperature?

We have included new panels (Figure 4 —figure supplement 2b; and Figure 4 —figure supplement 4e) that show how the correlation between temperature and LUVp changes at different levels of relative humidity in H. annuus and H. petiolaris. For H. annuus in particular, there is a strong interaction between the two variables, with a stronger negative correlation between temperature and LUVp at higher values of relative humidity.

Line 356: This is one result the authors interpret with the ideal amount of caution (this is a compliment). I'd be curious to know if other regions of the genome also appear in GEA, and whether any of those co-locate to other SNPs with larger estimated associations with phenotype (even if those are n.s.).

GEAs generally have many significant associations throughout the genome (see Todesco et al., Nature 2020 for some examples with the same datasets used in this manuscript); this is because climate variables like temperature and relative humidity have major effects on many aspects of plant development and plant fitness, and elicit complex adaptations. We have not found any major GEA signal linked to other significant or suggestive GWAS associations; however, it is likely that a larger population size would be required to reliably detect weaker signals in GWA experiments.

Why is the x-axis in 4k different from 2b?

The choice of the region shown on the x-axis was arbitrary, but in both cases we wanted to show the “shape” of the association (that is, that the strongest association in the region was on top of HaMYB111); since the peak is broader for GEA, we included a larger region.

Figure 4 is otherwise functionally excellent and very aesthetically pleasing. I'm very impressed with these great figures overall, despite my critiques of Figure 3.

Thank you! Hopefully the changes we made to Figure 3 have addressed the Reviewer’s concerns.

Conclusions: The statements on effects of pollinators and transpiration should be softened to fit the evidence. On line 390: Makes sense for Texas, but Arizona also has high S allele prevalence in Figure 2, and to my knowledge is less humid.

It is correct that RH values for Arizona are generally lower than for Texas, and we do not mean to claim a perfect correlation between temperature/RH and floral UV patterns. It should be noted, however, that the RH values that we used for our analyses are extrapolated from weather stations across North America, and not measured in situ, meaning that they do not account for microclimatic variation. Of the two populations in Arizona that have smaller floral UV patterns and high frequency of S alleles, one (ANN_13) was collected along the Verde River, near Deadhorse lake, and the description of the collection site is “riparian forest and wetland”, suggesting that humidity might be locally higher than in the surrounding region. We have added this observation to the Methods section:

“It should be noted that the values for climate variables used in these analyses are extrapolated from weather stations across North America, and not measured in situ, meaning that they might not account for microclimatic variation. For example, two populations in Southern Arizona do not fit the pattern we proposed – they have small floral UV patterns and high frequency of S alleles at the Chr15_LUVp SNP, despite being associated with relatively low RH values in our datasets. However, one of them (ANN_13) was collected along the Verde River, near Deadhorse lake, and the description of the collection site is “riparian forest and wetland”, suggesting that humidity might be locally higher than in the surrounding region. Similarly, from satellite pictures, the collection site for the other population (ANN_47) appears considerably more verdant than other collection sites in Arizona.“ (new lines 881-892)

As suggested, we have softened the conclusions in regard to the connection between floral UV patterns and transpiration, and added a sentence to explore possible additional factors affecting pollinator preferences and/or geographic distribution of floral UV patterns.

The relevant section now reads:

“Here, we show that regulatory variation at a single major gene, the transcription factor HaMYB111, underlies most of the diversity for floral UV patterns in the common sunflower, wild H. annuus. Variation for these floral UV patterns correlates strongly with pollinator preferences, but also with geoclimatic variables (especially relative humidity and temperature) and desiccation rates in sunflower ligules. While the effects of floral UV patterns on pollinator attraction are well-known, these associations suggest a role of environmental factors in shaping diversity for this trait. Larger floral UV patterns, due to accumulation of flavonol glycoside pigments in ligules, could help reduce the amount of transpiration in environments with lower relative humidity, preventing excessive water loss and maintaining ligule turgidity. In humid, hot environments (e.g. Southern Texas), lower accumulation of flavonol glycosides would instead promote transpiration from ligules, keeping them cool and avoiding overheating.” (new lines 527-537)

Line 406: How is the transpiration cost of sex not about sex? I suggest deleting this sentence.

We have removed the sentence.

https://doi.org/10.7554/eLife.72072.sa2

Article and author information

Author details

  1. Marco Todesco

    Department of Botany and Biodiversity Research Centre, University of British Columbia, Vancouver, Canada
    Contribution
    Conceptualization, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Visualization, Writing - original draft, Writing – review and editing
    For correspondence
    mtodesco@biodiversity.ubc.ca
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-6227-4096
  2. Natalia Bercovich

    Department of Botany and Biodiversity Research Centre, University of British Columbia, Vancouver, Canada
    Contribution
    Conceptualization, Data curation, Investigation, Methodology, Project administration, Supervision, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7703-2858
  3. Amy Kim

    Department of Botany and Biodiversity Research Centre, University of British Columbia, Vancouver, Canada
    Contribution
    Investigation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2623-4118
  4. Ivana Imerovski

    Department of Botany and Biodiversity Research Centre, University of British Columbia, Vancouver, Canada
    Contribution
    Conceptualization, Data curation, Investigation, Methodology, Project administration, Supervision, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1164-3664
  5. Gregory L Owens

    1. Department of Botany and Biodiversity Research Centre, University of British Columbia, Vancouver, Canada
    2. Department of Biology, University of Victoria, Victoria, Canada
    Contribution
    Data curation, Formal analysis, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4019-5215
  6. Óscar Dorado Ruiz

    Department of Botany and Biodiversity Research Centre, University of British Columbia, Vancouver, Canada
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  7. Srinidhi V Holalu

    Department of Plant and Microbial Biology, University of California, Berkeley, United States
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-1948-8216
  8. Lufiani L Madilao

    Michael Smith Laboratory and Wine Research Centre, University of British Columbia, Vancouver, United States
    Contribution
    Investigation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4161-2540
  9. Mojtaba Jahani

    Department of Botany and Biodiversity Research Centre, University of British Columbia, Vancouver, Canada
    Contribution
    Investigation, Visualization
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1844-1464
  10. Jean-Sébastien Légaré

    Department of Botany and Biodiversity Research Centre, University of British Columbia, Vancouver, Canada
    Contribution
    Data curation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1483-9643
  11. Benjamin K Blackman

    Department of Plant and Microbial Biology, University of California, Berkeley, United States
    Contribution
    Conceptualization, Funding acquisition, Supervision, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4936-6153
  12. Loren H Rieseberg

    Department of Botany and Biodiversity Research Centre, University of British Columbia, Vancouver, Canada
    Contribution
    Conceptualization, Funding acquisition, Supervision, Writing – review and editing
    For correspondence
    lriesebe@mail.ubc.ca
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2712-2417

Funding

Genome Canada (LSARP2014-223SUN)

  • Marco Todesco
  • Loren H Rieseberg

Genome British Columbia (LSARP2014-223SUN)

  • Marco Todesco
  • Loren H Rieseberg

National Science Foundation (IOS-1444522)

  • Loren H Rieseberg

National Science Foundation (IOS-1759942)

  • Benjamin K Blackman

Human Frontier Science Program (LT000780/2013)

  • Marco Todesco

University of California Berkeley

  • Benjamin K Blackman

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This research was conducted in the ancestral and unceded territory of the xʷməθkʷəy̓əm (Musqueam) People. We thank Andrea Todesco, Daniela Rodeghiero, Emma Borger, Quinn Anderson, Jennifer Lipka, Jasmine Lai, Hafsa Ahmed, Dominique Skonieczny, Ana Parra, Cassandra Konecny, Chris Zan, Juan Chavez, Victor Canta-Gallo, Anna Dmitrieva, Patrick Jacobsen, Kelsie Morioka, and Daniel Yang for assistance with field work and data acquisition, Melina Byron and Glen Healy at UBC, Christina Wistrom at the UC Berkeley Oxford Tract Facility, and the UC Davis Plant Sciences Field Station personnel for assistance with greenhouse and field experiments, Elizabeth Elle and Tyler Kelly for help planning the pollinator preference experiments, Laura Marek and the USDA-ARS in Ames, IA, USA, for providing sunflower seeds, and Chase Mason for providing cuttings for Phoebantus tenuifolius. Maps were realized using tiles from Stamen Design (https://stamen.com), under CC BY 3.0, from data by OpenStreetMaps contributors (https://openstreetmap.org), under ODbL. Funding was provided by Genome Canada and Genome BC (LSARP2014-223SUN), the NSF Plant Genome Program (IOS-1444522, IOS-1759442), the University of California, Berkeley, and an HFSP long-term postdoctoral fellowship to MT (LT000780/2013).

Senior Editor

  1. Jürgen Kleine-Vehn, University of Freiburg, Germany

Reviewing Editor

  1. Jeffrey Ross-Ibarra, University of California, Davis, United States

Reviewer

  1. Ian T Baldwin, Max Planck Institute for Chemical Ecology, Germany

Publication history

  1. Preprint posted: June 27, 2021 (view preprint)
  2. Received: July 9, 2021
  3. Accepted: November 28, 2021
  4. Version of Record published: January 18, 2022 (version 1)

Copyright

© 2022, Todesco et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,999
    Page views
  • 313
    Downloads
  • 2
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Marco Todesco
  2. Natalia Bercovich
  3. Amy Kim
  4. Ivana Imerovski
  5. Gregory L Owens
  6. Óscar Dorado Ruiz
  7. Srinidhi V Holalu
  8. Lufiani L Madilao
  9. Mojtaba Jahani
  10. Jean-Sébastien Légaré
  11. Benjamin K Blackman
  12. Loren H Rieseberg
(2022)
Genetic basis and dual adaptive role of floral pigmentation in sunflowers
eLife 11:e72072.
https://doi.org/10.7554/eLife.72072
  1. Further reading

Further reading

  1. Patterns of ultraviolet pigments in sunflowers can attract pollinators and reduce water loss.

    1. Evolutionary Biology
    2. Genetics and Genomics
    Dawei Cai et al.
    Research Article Updated

    The exceptionally rich fossil record available for the equid family has provided textbook examples of macroevolutionary changes. Horses, asses, and zebras represent three extant subgenera of Equus lineage, while the Sussemionus subgenus is another remarkable Equus lineage ranging from North America to Ethiopia in the Pleistocene. We sequenced 26 archaeological specimens from Northern China in the Holocene that could be assigned morphologically and genetically to Equus ovodovi, a species representative of Sussemionus. We present the first high-quality complete genome of the Sussemionus lineage, which was sequenced to 13.4× depth of coverage. Radiocarbon dating demonstrates that this lineage survived until ~3500 years ago, despite continued demographic collapse during the Last Glacial Maximum and the great human expansion in East Asia. We also confirmed the Equus phylogenetic tree and found that Sussemionus diverged from the ancestor of non-caballine equids ~2.3–2.7 million years ago and possibly remained affected by secondary gene flow post-divergence. We found that the small genetic diversity, rather than enhanced inbreeding, limited the species’ chances of survival. Our work adds to the growing literature illustrating how ancient DNA can inform on extinction dynamics and the long-term resilience of species surviving in cryptic population pockets.