Single gene mutation in a plant MYB transcription factor causes a major shift in pollinator preference

Understanding the molecular basis of reproductive isolation and speciation is a key goal of evolutionary genetics. In the South American genus Petunia , the R2R3-MYB transcription factor MYB-FL regulates the biosynthesis of UV-absorbing ﬂavonol pigments, a major determinant of pollinator preference. MYB-FL is highly expressed in the hawkmoth-pollinated P. axillaris , but independent losses of its activity in sister taxa P. secreta and P. exserta led to UV-reﬂective ﬂowers and associated pollinator shifts in each lineage (bees and hummingbirds, respectively). We created a myb-ﬂ CRISPR mutant in P. axillaris and studied the effect of this single gene on innate pollinator preference. The mutation strongly reduced the expression of the two key ﬂavonol-related biosynthetic genes but only affected the expression of few other genes. The mutant ﬂowers were UV reﬂective as expected but additionally contained low levels of visible anthocyanin pigments. Hawkmoths strongly preferred the wild-type P. axillaris over the myb-ﬂ mutant, whereas both social and solitary bee preference depended on the level of visible color of the mutants. MYB-FL , with


INTRODUCTION
Most plants are rooted in the soil and are prevented from actively seeking a mating partner in the way animals do.Male gametes, packed in resilient pollen, can be transferred by wind or water, but such abiotic vectors are rather inefficient. 1 Depending on the geographical region, 78%-94% of flowering plants are pollinated by animals, 2 demonstrating their importance for plant reproduction.In this mutually beneficial interaction, the pollinator transfers pollen between individual plants and is rewarded for its services with a high-energy food source, such as nectar.Floral color, scent, and morphology enhance the specificity of the interaction and further promote the preferential transfer of pollen to conspecific stigmas. 3However, outsourcing pollen transfer to animal vectors carries the risk of failure when pollinators become unavailable or the local pollinator assemblage changes.Under such circumstances, adaptation to new pollinators can be beneficial as has happened frequently in many taxa, and it is widely accepted that animal-mediated pollination has promoted speciation and driven the rapid diversification of the angiosperms. 4 major theme in contemporary pollination research is the elucidation of the molecular-genetic basis of shifts in pollinator preference.Classical theory has long held that phenotypic change proceeds through many individual mutations, each of small phenotypic effect, as small mutations allow descendants to slowly diverge from the parental population without removing them far from their fitness optimum. 5However, experimental data from genetic model organisms provided convincing evidence for mutations in single genes causing major phenotypic effects, an impressive early example being the establishment of apical dominance during the domestication of maize. 68][9] Moreover, it is difficult to reconcile gradual change with sympatric speciation, as ongoing gene flow and recombination will have homogenizing effects on populations. 10,11The plausibility of mutations of major effect when selection is strong has a solid theoretical underpinning. 12,13Loss of an established pollinator is likely to displace a plant far from its fitness optimum, a scenario that may favor mutations of large effect to help a pollinator adapt toward a new optimum.][16][17][18] Plants attract specific pollinators by combinations of floral traits such as color, scent, and morphology, so-called pollination syndromes.0][21][22] While hawkmoths simply prefer UV-absorbing flowers, UV preference for bees is more complex. 18,23,24In fact, for white-colored flowers, UV reflectance can significantly prolong a bee's search time, 25 whereas for red-colored flowers, bees prefer UV-reflecting flowers over UV-absorbing ones. 26Both the visible and UV color displays of a flower are therefore vital in determining pollinator visitation.The major floral UV and visible color pigments responsible for floral color are synthesized by the flavonoid biosynthetic pathway.Two main branches of this pathway produce the visible color pigments: the anthocyanins and UV-absorbing flavonols (Figure 1C). 27Anthocyanins can range in color from red to purple to blue, while flavonols are responsible for UV-absorbent flowers or UV patterns not visible to the human eye.
The South American genus Petunia (Solanaceae) offers a compelling system to study color transitions and associated pollinator shifts.The genus comprises species with distinct pollination syndromes, and the biochemistry and genetics of the flavonoid biosynthetic pathway have been extremely well  S3).characterized. 30,31Species in the long-tube clade display the most discrete phenotypic differences.P. axillaris displays white, UV-absorbing flowers emitting a rich blend of volatiles to attract nocturnal hawkmoths. 18,32,33P. exserta has red flowers that are UV reflective, attracting hummingbirds as their primary pollinator. 34][40] The combination of molecular-genetic data with phylogenetic analyses provides strong evidence that the common ancestor of the long-tube clade was likely hawkmoth pollinated with subsequent shifts to bees and hummingbirds. 39,40P. axillaris and P. secreta in particular have similar floral morphologies, provide a nectar reward, and grow in sympatry.The regain of visible color in P. secreta was simple: a 2-bp deletion in the gene encoding the MYB transcription factor AN2 restored the reading frame and anthocyanin biosynthesis. 39Loss of UV color was achieved by the inactivation of MYB-FL. 18,39The close phylogenetic relationship and the knowledge of causative genes make the P. axillaris/P.secreta species pair particularly interesting to study the effect of individual genetic changes on pollinator preference.
Introgression lines have been widely used to study the effect of single loci on pollinator behavior.A classic example is the behavioral assays in Mimulus where introgressions at the YUP flower color locus demonstrated differential pollinator preference. 41owever, even after generations of backcrossing, an introgressed region will include many linked genes with potentially confounding effects.This is particularly relevant in interspecific crosses, where recombination may be suppressed. 42To rigorously test the effect of a single gene, behavioral assays must be performed comparing isogenic lines that only differ in the gene under study.Rare examples of such an approach are the transgenic overexpression of AN2 in P. axillaris 14 and the reduction of ocimene synthase expression in Mimulus by RNAi. 43Mixta and nivea mutants in Antirrhinum, where single genes have been altered that affect flower color appearance, are rare examples where pollinator behavioral assays have been performed on single gene mutants, elegantly demonstrating their effects on bumblebee behavior. 44,45he Petunia MYB-FL gene offers an attractive system to test the precise impact of a single gene mutation on the flavonoid biosynthetic pathway, floral UV pigmentation phenotype, pollinator preference, and ultimately on reproductive isolation of species.To assess each of these aspects, we generated a knockout of the MYB-FL gene in two accessions of the hawkmoth-pollinated P. axillaris and tested the consequences of the mutation with respect to flavonoid biosynthesis, global gene expression, and innate pollinator preference.MYB-FL expression is highly restricted, and its knockout affects only a few genes in the genome.Moreover, we show that the loss of MYB-FL function has dramatic effects on the preference of both hawkmoths and bees.

RESULTS
CRISPR-Cas9-mediated knockout of MYB-FL gene causes a shift in the balance between UV and visible color In order to inactivate the MYB-FL gene, we used a CRISPR-Cas9 approach with two gRNAs in P. axillaris accessions N and P (Figure 1A).Two independent mutations were generated in each background (Figure 1A).A 1-and 2-bp insertions were detected in the P. axillaris N background, while a 5-and 43-bp deletion were found in the P. axillaris P background.The mutations led to early stop codons through frameshifts, resulting in truncated MYB-FL proteins.All mutant lines showed a shift in floral UV phenotype; flowers were no longer UV absorbent but UV reflective (Figure 1B).Additionally, we observed a shift in the visible color phenotype, going from little to no anthocyanins (white) to shades of light pink.Interestingly, this shift in color differed between the P. axillaris accessions.In the P. axillaris N background, the corolla was light pink and easily observable by the human eye.By contrast, P. axillaris P mutants had a very faint pink corolla color (Figure 1B).Mutating MYB-FL in P. axillaris therefore not only induces a major reduction in UV absorption but also minor background-dependent increases in the visible range of the color spectrum.
To quantify the effects of the MYB-FL knockout on the flavonoid pigments that underlie visible color (anthocyanins) and UV color (flavonols) in Petunia (Figure 1C), we measured total flavonol and anthocyanin absorbance (relative concentration) in all lines using UV-Vis spectrophotometry.Total flavonol absorbance was significantly reduced in the myb-fl mutants compared with the wild-type lines, while anthocyanin absorbance was significantly higher in the P. axillaris N myb-fl mutants (Figure 1D; detailed statistics in Table S3).The differences in flavonol absorbance (and therefore their relative ratios) are marked between the mutant and wildtype lines, with much lower flavonol absorbance in the mutants than wild types.We did detect slightly higher anthocyanin absorbance in P. axillaris N myb-fl mutants compared with P. axillaris P myb-fl mutants, corresponding to the difference in the intensity of visible pink color (Figures 1B and 1D).Overall, myb-fl lines showed a reduction in flavonols responsible for UV color as well as an elevation in anthocyanin pigments responsible for visible color.

MYB-FL specifically affects floral UV-Vis absorption
To determine possible pleiotropic effects of the MYB-FL mutation, we measured several key traits that could potentially influence pollinator preference.Corolla size, pistil and stamen length, nectar volume and sugar concentration, and pollen germination rate showed no differences between wild-type and mutant lines (Figures 2A-2E).Epidermal cell shape has been reported to affect the efficiency of bee foraging 44,45 but is often overlooked.While the tested wild species (P.secreta and P. axillaris N and P) showed subtle differences in shapes of conical cells (Figure 2I), no differences were detected in cell shape between the myb-fl mutants and their corresponding wild-type lines.Similarly, an analysis of scent production revealed no differences between the myb-fl mutants and the respective wild-type lines for the major volatiles methylbenzoate, benzaldehyde, and isoeugenol/ eugenol (Figures 2F-2H).In summary, the myb-fl mutation alters UV-Vis color without confounding effects on other major floral traits with potential relevance in pollinator preference.
MYB-FL is specifically expressed in the floral limb and has a limited number of target genes in the flavonoid pathway Tissue specificity of MYB-FL expression was investigated by qRT-PCR in floral organs and leaves.Levels of MYB-FL  S3).Two independent mutant lines are shown for each P. axillaris line.(A) Corolla areas for mutant and wild-type lines.
(legend continued on next page) expression were negligible in all tested organ types except in floral limbs for both P. axillaris N and P (Figure 3A).In order to study changes in the expression of genes potentially regulated by MYB-FL, an RNA-seq experiment was conducted with two independent myb-fl mutants as well as the isogenic P. axillaris N. Given the observed changes in both UV and visible floral phenotypes, we investigated whether the mutation in MYB-FL affected the expression of downstream flavonoid biosynthetic genes as well as anthocyanin biosynthetic genes.The expression of MYB-FL itself is significantly reduced, possibly due to nonsense-mediated decay. 46Only two genes, FLS and HT1/F3'H, are strongly downregulated in the mutant (Figure 3B; see Figure S1 for qRT-PCR data).FLS catalyzes the synthesis of flavonols from their dihydroflavonol precursors and is the only enzyme exclusively functioning in flavonol synthesis. 29HT1/F3 0 H catalyzes the hydroxylation of the precursor dihydrokaempferol (DHK) into dihydroquercetin (DHQ), the immediate precursor of quercetin if modified by FLS and cyanidin if modified by dihydroflavonol-4reductase (DFR). 29Most of the anthocyanin biosynthetic genes were slightly higher expressed in the mutant, although in most cases not significantly so, and furthermore not confirmed by qRT-PCR (Figures 3B and S1).We therefore consider it likely that the gain in anthocyanins in the mutant is due to release from competition for their common substrates, the dihydroflavonols (Figure 1C).In summary, inactivation of MYB-FL caused a strong decrease in the expression of the two genes involved in flavonol synthesis (FLS, HT1/F3 0 H) and at most a marginal increase in the expression of anthocyanin biosynthetic genes.This points toward MYB-FL being specific in its targets.
8][49][50] To assess effects of the myb-fl mutation beyond the flavonoid biosynthetic pathway, global changes in gene expression between the myb-fl and wild-type lines were studied in petal limb tissue.Only 55 genes were significantly differentially expressed (DE), using a filter of a log2-fold change value greater than 1 or less than À1 and a p value < 0.05 (Table S2).Of these, 43 were downregulated and 12 were upregulated in the mutant.Several of the DE genes are potentially involved in high-light-induced stress responses (e.g., DNA photolyase Peaxi162Scf00034g01311 or cryptochrome/DNA photolyase Peaxi162Scf00213g01024; Table S2).P. axillaris typically grows in high-light environments, 51,52 and therefore flavonols might play a protective role in petals.We conclude that MYB-FL has a highly restricted expression pattern and affects only a small set of target genes.

Distinct innate pollinator preference of wild Petunia species
The evolution from hawkmoth to bee pollination in Petunia involved changes in a combination of floral traits, one of them being the loss of UV color. 18,39The myb-fl mutants make it possible to disentangle the effect of changes in UV absorption from other floral traits on pollinator behavior.To assess innate pollinator preference, we set up pairwise comparisons between P. axillaris N and P and their isogenic myb-fl mutants.Three different pollinators were used (n = 20 for each comparison): the hawkmoth Manduca sexta, a pollinator of P. axillaris in the wild; 53 the bumblebee B. terrestris, a popular model organism; and the solitary bee Osmia cornuta (O.cornuta), which is closer in size and life cycle to the natural pollinator of P. secreta (solitary bee of the genus Pseudagapostemon) than to that of B. terrestris. 37Choice experiments with M. sexta, a nocturnally active animal, were performed in a dark room supplemented with low UV light, similarly to UV reflectance from the moon during nighttime pollinator visitations.Experiments with B. terrestris and O. cornuta took place under natural light on sunny days (see STAR Methods for details).Animal behavior was recorded as first choice (the first plant visited) and total choices (all types of choices recorded during the 10-min duration of the experiments; bee choices are categorized as touch, visit, and forage, whereas hawkmoth choices are categorized as hover, probing, and forage; Figure S2).
To establish a baseline behavior, we first analyzed the preference of the three pollinators for the wild Petunia species.Hawkmoths significantly preferred P. axillaris for first and total choices, whereas bumblebees had a preference for P. secreta for first choice (significant only for total choices).This comparison evoked a more selective bee response from solitary bees than bumblebees: O. cornuta preferred P. secreta for both first and total choices and displayed very few visits to P. axillaris overall (Figure 4A; for separated data on types of choices, see Table S5).This demonstrates differences in innate preference between hawkmoths and bees as well as differences in the strength of the selectiveness between bee species.
Hawkmoths strongly prefer wild-type P. axillaris over isogenic myb-fl mutants In the comparison between P. axillaris N and its isogenic myb-fl mutant, hawkmoths significantly preferred the wild-type flowers over the myb-fl mutant flowers for their first choice as well as the total number of choices (Figure 4B).The same was observed for hovering, probing, and foraging (Table S5).In the absence of UV light, the moths did not discriminate between wild type and mutant, clearly indicating that discrimination depends on the presence of UV light (Figure S3).Note that even under these conditions, the moths do not wander aimlessly around but efficiently approach the flowers, presumably guided by olfactory cues.Bees preferred the mutant over the wild type (not significant for O. cornuta, first choice).In the comparison between P. axillaris P and the isogenic myb-fl (A) Organ specificity of MYB-FL expression was quantified in wild-type lines of both P. axillaris N and P through qRT-PCR.Expression is shown relative to the expression of the three housekeeping genes SAND, RAN1, and ACTIN11.Statistics were calculated using a Mann-Whitney U test; letters depict significant differences (full statistics in Table S4).
(legend continued on next page) mutant, hawkmoths strongly preferred the wild type (Figure 4C), almost identical to what was observed in Figure 4B.By contrast, both bumblebees and solitary bees displayed no significant preference for either genotype.Therefore, the loss of MYB-FL activity compromises the attraction of hawkmoths in both P. axillaris N and P, whereas bees have a preference  S3).Plant species labels on the left side of the graphs show the different pairwise comparisons (light gray, P. axillaris N; light pink, P. axillaris N myb-fl; dark gray, P. axillaris P; faint pink, P. axillaris P myb-fl).Statistical differences between groups were calculated using an exact binomial test and are depicted by letters (Table S5; see also Table S6).(A) Comparison of wild-type lines P. axillaris N versus P. secreta for all three pollinator types.(B) Comparison of P. axillaris N versus P. axillaris N myb-fl mutant for all three pollinator types.(C) Comparison of P. axillaris P versus P. axillaris P myb-fl mutant for all three pollinator types.
for the mutant in the P. axillaris N background but not in the P. axillaris P background.
Intraspecific differences in visible color between myb-fl mutants affect bee preference P. axillaris N and P are almost identical in floral traits relevant for pollinator behavior (Figure 2), but an obvious difference between their myb-fl mutants is the intensity of the visible color: light pink, easily visible by the human eye, in myb-fl mutants of P. axillaris N and barely visible in P. axillaris P mutants (Figure 1B).To analyze the effect of these differences in visible color, we tested pollinator behavior with an array containing the two mutants from which to choose (Figure 5A).In control experiments, hawkmoths and bees equally visited each of the wild-type lines (Figure 5B).In the mutant comparisons, hawkmoths displayed a weak preference for the pale P. axillaris P mutant, whereas bumblebees strongly preferred the darker colored P. axillaris N mutant for both first choice and total choices.We conclude that the loss of MYB-FL activity causes a stronger shift toward bee pollination in P. axillaris N due to its stronger visible color.
Bees strongly prefer P. secreta over the myb-fl mutants If the myb-fl mutants represent an intermediate stage of adaptation toward a different pollinator group, they should be less attractive to a hawkmoth pollinator and more attractive to a bee pollinator.Therefore, we compared the myb-fl mutants of either P. axillaris N or P backgrounds with P. secreta (Figure 6).In these comparisons, all flowers are UV reflective, but P. secreta flowers have a stronger magenta anthocyanin pigmentation than the mutants.Hawkmoths preferred the P. axillaris N myb-fl mutants over P. secreta (Figure 6A; not significant for first choice for P. axillaris N myb-fl).In this same comparison, bumblebees preferred P. secreta (not significant for total choices), and solitary bees preferred P. secreta as well (not significant for first choice).Therefore, in the absence of UV-absorbing flavonols, hawkmoths have a distinct tendency to visit the P. axillaris N mutant and bees to visit P. secreta.In the case of the P. axillaris Bars display the number of first choices and total number of choices for each 10-min experiment (light gray, P. axillaris N; light pink, P. axillaris N myb-fl; dark gray, P. axillaris P; faint pink, P. axillaris P myb-fl).Plant species labels on the left side of the graphs show the different pairwise comparisons.Statistical differences between groups were calculated using an exact binomial test and are depicted by letters (Table S5; see also Table S6).(A) Comparison of both myb-fl mutants, P. axillaris N myb-fl mutant versus P. axillaris P myb-fl mutant.(B) Comparison of both P. axillaris wild types, P. axillaris N versus P. axillaris P. P mutant, the same preferences were observed but were statistically significant in all comparisons: hawkmoths preferred the mutant over P. secreta, whereas bees preferred P. secreta (Figure 6B).The clearer preferences in the comparisons with the P. axillaris P mutants compared with the assays with P. axillaris N mutants are most likely due to the differences in visible color observed between the mutants.In summary, the loss of MYB-FL activity seems to be enough to completely shift hawkmoth preference away from the mutants when comparing them with the isogenic P. axillaris wild types, but it does not suffice to make the mutants equivalent to P. secreta in the eyes of a hawkmoth pollinator.This supports the hypothesis that the P. axillaris myb-fl mutants represent intermediate stages, are no longer adapted to their original pollinator, but are not yet completely adapted to a novel pollinator.

DISCUSSION
Shifts in pollinators are widely recognized as a driving force of floral diversification and speciation in the angiosperms. 4,54,55esting the evolutionary role and specific effects of single gene mutations in potential speciation genes remains challenging in many species, and there are few instances in the literature identifying single mutations while also demonstrating a direct influence of the mutation on pollinator behavior. 14,44,45We constructed a CRISPR mutation in the floral UV-controlling transcription factor MYB-FL that mimics the loss-of-function alleles found in nature 18,39 and studied its specific impact on pollinator preference and thereby on reproductive isolation and speciation in Petunia.
MYB-FL expression in P. axillaris is tightly restricted in time and developmental space, and it has only a few target genes (Figure 3A; Table S2). 18Such properties may well be a prerequisite for genes of large phenotypic effect to induce a favorable evolutionary change with a minimum of undesirable side effects.7][58][59][60][61] It has even been suggested that pollination syndromes are preferably controlled by MYB factors, reminiscent of the prominent role of MADS-box factors in the determination of floral organ whorl identity. 62A classic example from the non-plant literature is the role of Hox genes in Drosophila segmentation, 63 illustrating the idea that complex traits could have evolved through the diversification and specialization of particular classes of transcription factors.Flavonoid synthesis is a beautiful example of a conserved pathway with variations that adapt it to the needs of individual species, moth pollination in Petunia, or for example, the exquisite color patterning that guides bumblebees in Anthirrhinum majus. 64B Figure 6.Bees strongly prefer P. secreta over myb-fl mutants Graphs show comparisons of myb-fl mutants with P. secreta, using three pollinator species: M. sexta, B. terrestris, and O. cornuta (n = 20 pollinators per comparison).Number of first choices as well as total number of choices are shown in each panel.Plant species labels on the left side of the graphs show the different pairwise comparisons (magenta, P. secreta; light pink, P. axillaris N myb-fl; faint pink, P. axillaris P myb-fl).Statistical differences between groups were calculated using an exact binomial test and are depicted by letters (Table S5; see also Table S6).

Major shifts in pollinator preference due to a single gene with dual phenotypic effects
The pollinator choice assays between isogenic genotypes demonstrate that mutation of MYB-FL has a major impact on reproductive isolation in Petunia.The primary pollinator of P. axillaris (hawkmoth) significantly preferred the wild-type plants over the myb-fl mutants (Figures 4B and 4C), supporting the suggested innate preference for UV-absorbing Petunia flowers. 18lthough inactivation of MYB-FL shifts pollination away from hawkmoths, the mutation alone is not enough to fully shift to bee pollination; secondary pollinators of P. axillaris (bumblebees and solitary bees) were still able to discriminate between the myb-fl mutant and P. secreta flowers, especially in the P. axillaris P background (Figures 6A and 6B).This retained discrimination by bee species between these UV-reflective choices (but different in visible anthocyanin pigments) suggests that reproductive isolation still occurs between the myb-fl mutant and P. secreta.
Additionally, our phenotypic analyses showed a dual phenotypic effect due to mutation to MYB-FL with a strong reduction in flavonol absorbance as well as a simultaneous increase in anthocyanin absorbance in the myb-fl mutants, compared with wild-type plants (Figure 1D).The UV-absorbing flowers were important cues for nighttime hawkmoth pollinators but did not have as strong of an effect on bee pollinators (Figure 4).MYB-FL activity is therefore not a prerequisite for daytime pollination to occur, and its loss of function is likely to free up dihydroflavonol precursors for anthocyanin production.Thus, this single mutation results in the loss of flavonols and gain of anthocyanins; both phenotypic traits that are favorable to shift pollinator preference from hawkmoth to bees.Such combined traits have the potential to increase a plant's reproductive success. 65We also note that the innate preferences detected here could be easily transformed by insect learning in an ecological context in which these mutants arise (and are likely in very low frequency).
With its intermediate phenotype, the myb-fl mutant represents a transitional form of the evolutionary process that occurred during the shift from hawkmoth to bee pollination syndrome.Major phenotypic leaps were long thought to be deleterious, but theory now accepts the possibility of genes of large phenotypic effect, 66 and this case demonstrates experimentally that a leap large enough to attract a new pollinator can allow reproductive isolation to occur through changes in a low number of genes of large effect.However, in nature, Petunia visitation by a sole pollinator type is not always the case, with several documented secondary visitors. 14,37For reproductive isolation to be maintained, a strong but not perfect pollinator preference may therefore suffice, especially if hybrids fail. 16,67,68 color as an initial cue for hawkmoth preference Comparing the two myb-fl mutants with different wild types allowed us to study the effect of UV-vis color on pollinator preference in more detail.In all the experimental setups where the plant lines differed in UV color (absorbent versus reflective), hawkmoths strongly preferred UV-absorbent flowers for first choice, regardless of the intensity of visible color, whereas preferences for UV-reflective flowers as a first choice were not equally strong for bee pollinators (Figure 4).Based on these findings, UV color seems to act as an initial and important cue during a hawkmoth's decision-making process to approach a flower but not during a bee's.While UV-absorbent flowers are attractive to hawkmoth pollinators, 18,23,69,70 flowers pollinated by diurnal pollinators often display UV patterning, such as nectar guides, that affects pollinator attraction.[71][72][73][74] The binary UV phenotype (absorbent/reflective) we used in our experiments may therefore be a less informative cue to bees than hawkmoths when deciding to approach a flower.
In our experiments, bees also seemed to rely more strongly on differences in the visible spectrum than hawkmoths.When presented with the two myb-fl mutants, both UV reflective but differing in intensity of visible color (Figure 5A), bumblebees significantly preferred the more visibly colored P. axillaris N mutant, whereas hawkmoths showed no significant preference for either mutant for first choice.6][77][78][79] Although the difference in visible color between the two mutants is rather subtle and not as strong as between wild-type P. axillaris and P. secreta (Figures 1B and 1D), it is detectable for bumblebees (Figure 5).When comparing these findings with the comparisons with P. secreta, bees clearly prefer P. secreta over the P. axillaris P myb-fl mutant (Figure 6B), but the preference is not as clear for the P. axillaris N myb-fl mutant (Figure 6A).In the absence of UVabsorbing flowers, even a minor addition of visible color already has a remarkably strong effect on bee preference (Figure 5A).

Reproductive isolation through multiple major effect genes
If the synthesis of high levels of flavonols limits anthocyanin production due to metabolic competition for common precursors, a decrease in flavonols is required before anthocyanins can be produced.This implies that the loss of MYB-FL activity is most likely an early mutational event in the evolution from hawkmoth to bee pollination.Through its dual phenotypic effect, mutation of MYB-FL may have made a strong initial contribution toward reproductive isolation during the divergence of the two visibly colored species (P.secreta and P. exserta) from the common ancestor.
Pollinators can integrate multiple cues to choose between plant species.Just like color, scent differences between species also affect pollinator preference. 43,70,80,81Previous analyses of hawkmoth preferences for Petunia scent volatiles discovered that hawkmoths preferred scented to non-scented plants, but when presented with conflicting cues (scent and color), both cues were equivalent. 80This demonstrates that a combination of cues (generally referred to as pollination syndromes) is necessary to make efficient choices.Additional genes of major phenotypic effect(s) have been described in Petunia and other systems, and it will be interesting to study their effects on pollinator preference, separately and in combination.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following: then transformed into A. tumefaciens strains GV3101 and LBA4404 and grown on solid selective medium at 28 C for 2 days before being used for subsequent stable plant transformation.Multiple independent lines were carefully phenotyped.
Stable transformation of P. axillaris P. axillaris N and P. axillaris P leaf discs were transformed essentially as described in Horsch et al. 93 using A. tumefaciens (strains GV3101 or LBA4404) carrying the respective CRISPR construct.Leaves from both P. axillaris species (from 4-6 week-old plants) were sterilized in a 10% hypochlorite solution for 10 min and rinsed five times with autoclaved MilliQ water.The sterilized leaves were then cut into 1 cm 2 pieces and inoculated in the A. tumefaciens suspension for 30 min (in the dark with shaking).After drying, the leaf fragments were transferred to Murashige-Skoog (MS) growth medium without antibiotic selection.Leaf pieces were grown in the dark for 7 days at a constant temperature of 24 C.After this first week, leaf pieces were then transferred to fresh selective growth medium containing kanamycin to select for the presence of the CRISPR construct.All plates were sealed with medical tape and incubated at C and a light:dark regime of 16:8 h.Once shoots started to appear, each shoot was excised from the calli and transferred to MS rooting medium containing kanamycin for further selection of the construct.

DNA extractions and genotyping
Genomic DNA extractions were performed with fresh leaf samples using a modified CTAB 94 or modified SDS protocol. 95Samples were analyzed with a Nanodrop ND-1000 (Thermo Fisher) prior to further analysis.Target sequences were amplified using the GoTaq (Promega) or Q5 polymerase (NEB) with primers depicted in Table S7.

RNA extractions, cDNA preparation and RT-qPCR
For each plant line collected, three individual buds (stage 4, 22-30 mm length) from four different plants grown under controlled conditions, were collected (n=12 for each sample type), representing biological replicates.The floral limb tissue of the buds was dissected and frozen immediately in liquid nitrogen.Other floral and plant organs were sampled under the same controlled conditions.For leaf samples, a single young fresh leaf was sampled, whereas all five sepals of an open flower were pooled for the sepal sample types.All stamens of an individual flower were also pooled for analysis while the pistil of the flower was sampled as a separate sample type.Tissue samples were stored at -80 C until further processing.RNA extractions were performed using the innuPREP RNA Mini Kit 2.0 and cDNA was synthesized using the qScriber cDNA synthesis kit (Axonlab) and then used for RT-qPCR with the ORA SEE qPCR Green ROX L mix (Axonlab) according to the manufacturer's recommendations.
The expression levels of flavonoid pathway biosynthetic genes 30 were investigated by RT-qPCR.qPCR primers were published previously in Esfeld et al. 39 The reference genes used for all samples were SAND, RAN1 and ACTIN11. 96Primer efficiencies for each primer pair were determined using standard curves.RT-qPCR experiments were run on a QuantStudio 5, 384 well Real-Time PCR Machine (Applied Biosystems).All reactions were run in triplicate to calculate the means of the individual biological replicates using three data points.Using the QuantStudio Design and Analysis (v1.4.3) software the Cq values were determined for calculations.Values of the mutant and wildtype plants of each line were normalized to the three reference genes.For all analyzed biosynthetic genes pairwise comparisons with a Mann-Whitney U test were performed for each sample type of the mean relative expression levels in R v3.6.1 85 with results being significantly different if p < 0.05.The ggplot2 package was used to visualize the results. 91A sequencing Quality of RNA was assessed using a Qubit 4.0 fluorometer (Thermo Fisher Scientific) and an Advanced Analytical Fragment Analyzer (Agilent).cDNA libraries were generated using a TruSeq Stranded mRNA Library Kit (Illumina) in combination with RNA UD Indexes (Illumina).The quality of the generated NGS libraries was evaluated using the same tools as described for RNA quality analysis.The pooled libraries were 100 bp single end sequenced using a NovaSeq 6000 SP Reagent Kit, v1, 100 cycles (Illumina) on an NovaSeq 6000 instrument (Illumina).All base call files were demultiplexed and converted into FASTQ files using Illumina bcl2fastq conversion software v2.20.The average number of reads/library was 41 million.The RNA quality-control assessments, generation of libraries and sequencing run were performed at the Next Generation Sequencing Platform, University of Bern, Switzerland.

Read and differential expression
The quality of the Illumina read data was processed using Trimmomatic v.0.36. 86Processed reads were then mapped to the reference genome of P. axillaris N v.3.03 using STAR v.2.6.0c 87in two-pass mode, with splice junctions -sjbdOverhang 100.Reads that mapped > 20 times were discarded.Reads were counted per gene using featureCounts v1.5.2. 88Differential expression analysis was performed between mutant P. axillaris N myb-fl and P. axillaris N lines, and was performed with DESeq2 v.1.26.0 89 in R v3.6.1. 85Read counts were normalized with rlog-transformation and sample replicates were used to compute mean counts per gene.Genes were considered significantly differentially expressed between the two lines if expression difference was at least 2-fold and p-value < 0.05.Functional annotations of differentially expressed genes were determined using the functional annotation scripts in Maker (v2.31.9) to call blastp (uniprot database accessed on 09/03/2020) and interproscan (v5.33.72.0).Interproscan additionally adds PFAM hits and assigns GO (gene ontology) terms.

Figure 1 .
Figure 1.Targeting the MYB-FL gene via CRISPR-Cas9 (A) Two independent mutant lines of P. axillaris accessions N and P were generated, all leading to truncated proteins.Red asterisks indicate early stop codons.(B) The myb-fl mutant flowers of both P. axillaris accessions are UV reflective.P. axillaris N mutants showed a clear pink color, while the P. axillaris P mutants were only faintly pink.(C) Flavonols and anthocyanins are produced through individual branches of the flavonoid biosynthetic pathway.Enzymes needed for individual steps are depicted next to the respective arrow.Substrates are in bold.Transcription factors AN2 and MYB-FL (highlighted in blue) activate DFR/anthocyanidin synthase (ANS) and FLS, respectively.Petunia DFR does not accept DHK as a substrate, 28 and FLS has low activity on DHM 29 (light gray color).Abbreviations: for genes that have different names in Petunia, we state the Petunia name in parentheses.DHK, dihydrokaempferol; DHQ, dihydroquercetin; DHM, dihydromyricetin; F3H, flavonone 3-hydroxylase; F3 0 H (HT1), flavonoid-3 0 -hydroxylase, encoded by HT1; F3 0 5 0 H (HF1/2), flavonoid-3 0 5 0 -hydroxylase, encoded by HF1 and HF2; FLS, flavonol synthase; DFR, dihydroflavonol-4-reductase; ANS, anthocyanidin synthase.(D) Total flavonol and anthocyanin absorbance values of the wild-type and mutant lines as determined through spectrophotometric analysis.Letters indicate statistically significant groups calculated using a one-way ANOVA with Tukey's post hoc comparisons (full statistics in TableS3).

Figure 2 .
Figure 2. Phenotypic measurements of myb-fl mutant lines compared to wild-type P. axillaris All bars depict mean values ± SD.Statistics were calculated using a one-way ANOVA with Tukey's post hoc comparisons (TableS3).Two independent mutant lines are shown for each P. axillaris line.(A) Corolla areas for mutant and wild-type lines.

BFigure 3 .
Figure 3. MYB-FL is specifically expressed in the floral limb and strongly induces expression of key genes involved in flavonol synthesis

Figure 4 .
Figure 4. Hawkmoths strongly prefer wild-type P. axillaris over isogenic myb-fl mutants Graphs show comparisons with three pollinator species: M. sexta, B. terrestris, and O. cornuta (n = 20 pollinators per comparison).Number of first choices as well as total number of choices are shown in each panel (see also FigureS3).Plant species labels on the left side of the graphs show the different pairwise comparisons (light gray, P. axillaris N; light pink, P. axillaris N myb-fl; dark gray, P. axillaris P; faint pink, P. axillaris P myb-fl).Statistical differences between groups were calculated using an exact binomial test and are depicted by letters (TableS5; see also TableS6).(A) Comparison of wild-type lines P. axillaris N versus P. secreta for all three pollinator types.(B) Comparison of P. axillaris N versus P. axillaris N myb-fl mutant for all three pollinator types.(C) Comparison of P. axillaris P versus P. axillaris P myb-fl mutant for all three pollinator types.

Figure 5 .
Figure 5. Intraspecific differences in visible color between myb-fl mutants affect pollinator choicePairwise comparisons were run for two pollinator species: M. sexta and B. terrestris (n = 20 pollinators per comparison).Bars display the number of first choices and total number of choices for each 10-min experiment (light gray, P. axillaris N; light pink, P. axillaris N myb-fl; dark gray, P. axillaris P; faint pink, P. axillaris P myb-fl).Plant species labels on the left side of the graphs show the different pairwise comparisons.Statistical differences between groups were calculated using an exact binomial test and are depicted by letters (TableS5; see also TableS6).(A) Comparison of both myb-fl mutants, P. axillaris N myb-fl mutant versus P. axillaris P myb-fl mutant.(B) Comparison of both P. axillaris wild types, P. axillaris N versus P. axillaris P.