Plants can evolve rapidly after pollinator changes, but the response of different floral traits to novel selection can vary. Floral morphology is often expected to show high integration to maintain pollination accuracy, while nectar traits can be more environmentally sensitive. The relative role of genetic correlations and phenotypic plasticity (PP) in floral evolution remains unclear, particularly for nectar traits, and can be studied in the context of recent pollinator changes.
Digitalis purpurea shows rapid recent evolution of corolla morphology but not nectar traits following a range expansion with hummingbirds added as pollinators. We use this species to compare PP, heritability, evolvability and integration of floral morphology and nectar in a common garden.
Morphological traits showed higher heritability than nectar traits, and the proximal section of the corolla, which regulates access to nectar and underwent rapid change in introduced populations, presented lower integration than the rest of the floral phenotype. Nectar was more plastic than morphology, driven by highly plastic sugar concentration. Nectar production rate showed high potential to respond to selection.
These results explain the differential rapid evolution of floral traits previously observed in this species and show how intrafloral modularity determines variable evolutionary potential in morphological and nectar traits.

Introduction

Flower local adaptation plays a leading role in the diversification and speciation of angiosperms (Stebbins, 1950, 1970; Harder & Barrett, 2006; Friis et al., 2011). The potential for floral traits to respond to selection and undergo adaptive change depends on how much of their heritable phenotypic variation (i.e. additive genetic variance, V_A) is exposed to natural selection. This is in turn influenced by their environmental sensitivity. In addition, in complex organs such as flowers, traits tend to be highly integrated into functional clusters and do not evolve in isolation from each other (Ashman & Majetic, 2006). Genetic correlations among clusters of floral traits, caused by linkage disequilibrium and pleiotropic effects among other factors, determine the direction and magnitude of the response to selection of each trait (Hansen et al., 2003). All these factors might vary across different clusters of floral traits, such as morphological and nectar traits. Exploring to what extent additive genetic variance, environmental sensitivity and genetic correlations interact and determine the potential for adaptive change in different floral traits under novel pollination environments remains a key issue to explain the patterns that shape flower evolution and to comprehend the potential of flowering plants to adapt to future environmental changes (Pascoal et al., 2012; Payne & Wagner, 2019).

One of the mechanisms determining phenotypic variation via environmental sensitivity is phenotypic plasticity (PP), understood as the ability of a genotype to express different phenotypic values under different environments (Valladares et al., 2006). Plant survival to environmental changes is greatly determined by PP, which can play a major role in adaptive change and evolution from both the microevolutionary and macroevolutionary perspectives (Bradshaw, 1965; Davidson et al., 2011; Schneider, 2022). At the population level, PP allows the expression of environmentally induced novel phenotypes better adjusted to changes in ecological time scales, thus determining plant survival in response to new environmental conditions (van Kleunen, 2007; Nicotra et al., 2010; Valladares et al., 2014; Corl et al., 2018; Schneider, 2022). At macroevolutionary scales, PP can be key in the evolution of species and lineages by exposing novel environmentally induced phenotypic variation that can be the base for genetic change to take place (West-Eberhard, 2003; Pigliucci et al., 2006; Crispo, 2007; Levis & Pfennig, 2016).

In pollinator-dependent flowering plants, the floral phenotype is directly related to reproductive success and is therefore expected to show reduced variability. Floral traits tend to be more phenotypically integrated and canalised than vegetative traits to maintain functional accuracy for optimal pollination, showing limited environmental sensitivity and lower phenotypic variation than vegetative traits (Berg, 1960; Pélabon et al., 2011). This canalisation becomes more evident in specialised pollination systems due to strong selection for accuracy acting on traits involved in pollen transfer (Armbruster et al., 2009). Yet, different types of floral traits (for instance, morphological and nectar traits) can show variation in their environmental sensitivity and integration, depending on their developmental and physiological characteristics. Still, both micro- and macroevolutionary patterns across plant lineages show that all clusters of floral traits have the ability to adapt to different pollination environments (i.e. to be evolvable; Armbruster et al., 1999; Opedal, 2019). In fact, rapid floral adaptation to novel pollination environments is key for plant reproductive success under a changing environment, for example after range expansion events such as introductions to new habitats (Ollerton et al., 2012; Pfennig, 2021).

Introduction events pose a perfect opportunity to study how PP and the differential evolvability of clusters of floral traits determine flower adaptability under novel selection pressures. Populations of Digitalis purpurea (common foxglove) introduced to the American continent around the 1850s (c. 85 generations ago for this facultative biennial species) have undergone important changes in their pollinator assemblage. In its native European range, D. purpurea is pollinated mostly by long-tongued Bombus species (most frequently Bombus hortorum) that can reach the nectar, produced at the base of the ovary in the deepest, proximal section of the tubular corolla (Broadbent & Bourke, 2012). However, American populations are visited and efficiently pollinated by other Bombus spp. and several species of hummingbirds. The birds perform 20–27% of the visitation and deposit more pollen grains per visit than bumblebees (Mackin et al., 2021b). We have also shown that floral phenotypes in these populations are under novel directional selection (Mackin et al., 2021b).

Bees and hummingbirds are known to exert conflicting selection on floral traits. Melittophilous (bee-pollinated) flowers generally show lower production of more concentrated nectar than ornithophilous (bird-pollinated) flowers (reviewed in Fenster et al., 2004). Additionally, flowers pollinated by hummingbirds and other long-beaked birds tend to have long tubular corollas, limiting access to short-tongued nectar-feeding visitors and optimising the morphological fit for pollen transfer by birds (reviewed Herrera & Pellmyr, 2002). We previously found positive selection for larger proximal corollas in the non-native populations of D. purpurea but not in the native ones. Consistently, the proximal corolla of non-native populations shows rapid adaptive evolution and is now from 13% to 26% larger than in native populations. However, none of the studied non-native populations showed an adaptive change in nectar volume or concentration despite the theoretically expected more abundant and diluted nectar linked to ornithophilia, and even though some populations are still under positive selection for higher nectar volumes (Mackin et al., 2021b).

Different sections of the corolla also differ in their response to selection in this study system, regardless of the expected evolutionary constraints derived from integration between morphological traits (Lande, 1979; Opedal et al., 2022). In plants with tubular corollas, corolla dimensions, directly related to pollination efficiency are expected to be under stabilising selection and to show reduced variation (Berg, 1959). However, in D. purpurea, different portions of the tubular corolla play different roles in the interaction with pollinators. The distal part of the corolla serves as a landing platform and provides visual cues, regulating attraction. The size of the medial section of the corolla regulates pollen transfer and determines the mechanical fit between flower and pollinator. Its dimensions must be so that legitimate flower visitors contact the reproductive organs, located in the inner upper part of the corolla tube. Finally, the narrow proximal section of the corolla regulates access to the nectar reward, produced and presented at the base of the ovary, and accessible only to pollinators with long mouthparts. This system allows us to study the potential genetic constraints governing the evolvability of floral phenotypes after pollinator change and the ability of differential selection pressures to decouple developmentally integrated morphological traits.

Conversely to corolla morphology, nectar traits are known to show high phenotypic variability from the intraindividual to the intraspecies level (Parachnowitsch et al., 2019). Despite the important role of nectar for plant–animal interactions, estimates of additive genetic variance (V_A) for nectar-related traits have been scarcely reported (Mitchell, 2004). Most of the studies estimating narrow-sense heritability (V_A/V_P, where V_P = total phenotypic variance) of nectar traits have done so under artificial environments to control for the enormous environmental variation (Mitchell, 2004). Little research on this area has been conducted in natural or semi-natural conditions, likely due to the difficulties of measuring nectar traits under those conditions (but see Campbell, 1996; Leiss et al., 2004; Kaczorowski et al., 2008).

In this study, we use D. purpurea as a model to analyse trait variation and co-variation as potential reasons behind the differential evolutionary response in nectar and morphological traits after a change in pollination environments. We quantify PP, narrow-sense heritability, and potential and conditional evolvabilities of both functional types of traits as well as their degree of autonomy and integration within the floral phenotype. For this, we use controlled crosses and a multi-population common garden in the native range to compare the sources of phenotypic variation determining the evolutionary potential between flower morphology and nectar traits. We discuss the implications of floral variability and phenotypic integration for rapid adaptive floral change under novel pollinator environments.

Materials and Methods

Study system

Digitalis purpurea L. (Plantaginaceae) is a biennial monocarpic herbaceous plant, native to Western and Southwestern Europe, where it thrives in forest openings, forest edges, and exposed and disturbed sites. In early summer, it produces a tall spike-like inflorescence with numerous protandrous purple tubular flowers. Though self-compatible, it relies on pollinators for effective fertilisation and full seed set (Nazir et al., 2008; Mackin et al., 2021b). Digitalis purpurea has been introduced and naturalised in all continents except Antarctica, but none of its native pollinators (except for B. hortorum, also introduced in New Zealand) coexists with it in its non-native range. In American non-native populations, other Bombus spp. and several hummingbird species act as effective pollinators (Mackin et al., 2021b). Nectar robbing is rare in native populations (but see Rojas-Nossa et al., 2016), but flowers in the American populations are frequently robbed by several species of flower piercers (Diglossa spp.) with hummingbirds and bumblebees often acting as secondary robbers (Riveros et al., 2006; Mackin et al., 2021a).

Common garden and crossing design

In order to quantify PP, heritability and evolvability of different floral traits, as well as genetic correlations among them, we used a paternal half-sib full-sib nested crossing design in a common garden setup (Fig. 1). Crossings were done between plants of the same original D. purpurea population. Note that our experimental design was optimised to quantify plasticity and quantitative genetic parameters, and not to compare trait values across native and non-native populations. This was done for multiple populations in Mackin et al. (2021b).

Details are in the caption following the image — **Fig. 1**
Open in figure viewer PowerPoint

Crossing design and common garden setup of *Digitalis purpurea* plants. Crossings were done within the population of origin, using F1 plants originating from three populations. N♀ = 38–40, N♂ = 14–15 (icons from www.flaticon.com).

Seeds were collected from native (Holly Cross and Loder Valley, UK) and non-native (La Floresta, Colombia) populations, and an F1 generation was grown in a common garden at the University of Sussex (Falmer, UK) in summer 2018. Seeds were sown in pots with a peat-free compost and vermiculite and kept in the glasshouse until the seedlings were established, and then transferred outdoors. In summer 2020, we carried out hand pollination crosses. Inflorescences were bagged before flower anthesis to avoid pollinator visitation. Within each population, 14–15 plants were used as pollen donors (sires) to pollinate three different plants (dams) each, when possible (F1 generation in Fig. 1). The crossing design was nested so that dams were only pollinated by a unique sire plant. A total of 118 dams (38–40 per population) produced fruits. From each of the 118 families, four full-sib seeds were sown in the glasshouse in early September 2020 (n = 472) and 448 seeds germinated. After 45 d, seedlings were supplemented with a NPK 7 : 7 : 7 liquid fertiliser (Doff Liquid Grow More). To encourage flowering in the following summer, plants experienced a prolonged growing season in the glasshouse with a light cycle of 16–8 h and summer-like temperatures. In December 2020, plants were transferred to 4-l pots with a ratio of peat-free compost to horticultural sharp sand of 4 : 1 and kept under the same conditions until mid-January, when the temperature and photoperiod were gradually decreased. Potted plants were transferred in February to a nearby field (50.864563°, 0.078840°) for vernalisation.

To promote the expression of PP, we exposed full sibs to one of two different environments from before the onset of flowering. We did not intend to test the effects of the contrasting environments on phenotypes, but to increase the opportunity for plastic variation. In April 2021, two of each four full-sibs within each seed family were placed in a plot under a 50% shading net and the other two in an adjacent plot under full sunlight (Fig. 1). Light was monitored weekly at the centre of each plot at 80 cm from the ground with a light meter (RS PRO RS-3809), and temperature and humidity recorded every 10 min using two dataloggers (RS PRO RS-172TK) installed at the centre of each plot and protected from rain. After testing for normality (Shapiro–Wilk and Kolmogorov–Smirnov) and homoscedasticity (Levene's test) of the data, we performed t-tests (Welch's t-test for nonhomoscedastic distributions) to compare the abiotic environment between plots. The shading treatment significantly conditioned multiple environmental factors. Light intensity was significantly lower in the shaded plot (32.01 ± 16.39 klx) than in the full sunlight plot (61.06 ± 30.94 klx; Welch's t_21.28 = −3.21, P = 0.004). Daily average temperature was lower (t₂₈₆ = −2.60, P = 0.01) in the shaded plot (15.80 ± 2.76°C) than in the sunlight (16.73 ± 3.30°C), while relative humidity followed the inverse pattern (t₂₈₆ = 6.43, P < 0.001; shaded: 88.35 ± 6.49%, sunlight: 83.13 ± 7.27%). Daily average values for temperature and relative humidity were negatively correlated (r = −0.53, P < 0.001).

At the end of the experiment, we counted the number of fruits per plant. The number of fully developed seeds was quantified with ImageJ 1.53t (http://rsb.info.nih.gov/ij/) in one fruit per plant, and total seed production was estimated as a measure of lifetime reproductive success.

Measurement of floral traits

During the flowering season in June 2021, three flowers per plant were collected for phenotypic characterisation. Inflorescences were bagged with white bridal veil bags (mesh size < 1 mm) 24 h before sampling to avoid pollinator visits. Flowers live for several days (> 5) and nectar production can vary between stages of flower development (Percival & Morgan, 1965; Best & Bierzychudek, 1982), so we only sampled those in the male stage before nectar production wanes. We considered a flower to be in its male stage if at least its two front anthers were dehisced, and its stigmatic lobes were completely closed. Nectar was collected with microcapillary tubes (BLAUBRAND®), and the total volume of nectar in μl was used as the nectar production rate (NPR) in 24 h for each flower. Sugar concentration was measured with a Brix refractometer (Bellingham + Stanley™). Because Brix° equals grams of sucrose per 100 g of solution, we rescaled the concentration (CON) by calculating the molarity and converting to mg μl⁻¹ with the molecular weight of sucrose provided in the CRC Handbook of Chemistry and Physics (Haynes, 2014; pp. 5–123). Total sugar production in 24 h (SUG) was obtained from the product of NPR and CON.

After nectar collection, the corolla and fused anthers were detached from the rest of the flower, pressed in a lateral position using a botanical press with cardboard and filter paper, dried in silica gel for at least a week and digitalised at 300ppi using a scanner. Width and length of the whole (WCW and WCL) and proximal (PCW and PCL) sections of the corolla were quantified with ImageJ 1.53t (http://rsb.info.nih.gov/ij/; Fig. 2). Dimensions of the whole corolla and the proximal corolla were analysed as independent traits due to differences in their functionality (see the Introduction section) and their different evolutionary response in the introduced populations reported in our previous study (see Mackin et al., 2021b). Proximal and whole corolla size, calculated as the geometric mean between their length and width components, were included in the analysis by Mackin et al. (2021b). We excluded these derived measures here since they did not show different patterns to those observed when analysing the dimensions individually, and because this allowed us to analyse relationships among the different dimensions.

All statistical analyses were conducted using RStudio (Posit Team, 2022; R Core Team, 2022). A phenotypic covariance matrix, or P matrix, was computed for all nectar and morphological traits. Additionally, a phenotypic correlation matrix was calculated from pairwise Pearson correlation tests of all traits.

Phenotypic plasticity

We used the trait phenotypic variation among full sibs in the common garden as a measure of PP expressed under the environmental heterogeneity generated between the plot treatments (see ‘Common garden and crossing design’ in the Materials and Methods section). For this, we calculated the Phenotypic Dissimilarity (PhD) index (Puglielli et al., 2022) to estimate phenotypic variation within each full-sib family while accounting for intraindividual trait variation between different flowers of a same plant. Using the PhD()function in Puglielli et al. (2022), we computed one PhD index value for each trait per family. PhD indices were then compared among traits and functional groups of traits (morphological and nectar traits) using linear models. Because the PhD values range from 0 to 1, a logistic transformation (log[P/(1−P)]) was applied to improve their fit to a normal distribution before ANOVA analyses (Warton & Hui, 2011). A pairwise post hoc test of the PhD indices of each floral trait was carried out with the glht() function of the multcomp package in R (Hothorn et al., 2008) using the Tukey multiple comparisons of means as the linear contrast function. Natural selection on each trait and its PP in the common garden was assessed using the total seed production of each plant as a proxy of lifetime fitness. We used R packages gsg and mgcv to estimate linear and quadratic selection gradients by fitting general additive models (Morrissey & Sakrejda, 2014; Wood, 2023). We calculated the statistical significance of selection gradients with a bootstrap method set to 500 iterations following the approach by Morrissey & Sakrejda (2013).

Estimation of quantitative genetic parameters

Single trait heritability and evolvability

We estimated the additive genetic variance (V_A) of all traits using Bayesian inference on a multivariate mixed ‘animal model’ run with v.2.34 of the MCMCglmm R package (Hadfield, 2010). The model partitioned the phenotypic variance (V_P) of each trait into its additive genetic and residual components by accounting for the relatedness across individuals and allowed for the inclusion of fixed factors. Information on relatedness was included as the relatedness matrix calculated from the pedigree of the controlled crosses. We extracted estimates of V_A from the variance across individual plants, included in the model as the random effect ‘animal’ with its variance–covariance structure defined by the inverse relatedness matrix. Note that in nested paternal full-sib, half-sib crossing designs, the sire variance component is assumed to represent the covariance between half-sibs and to be unbiased by dominance, maternal and environmental effects (Falconer & Mackay, 1996; Lynch & Walsh, 1998).

Trait values included in the multivariate MCMCglmm model were ln-transformed to homogenise the variances and correct for different trait units before fitting the model. Because trait dimensionality affects trait variability and to avoid overestimation of V_A, NPR was introduced in all models as millimetres of nectar measured in the microcapillary tubes instead of using cubic volumetric units. Population of origin and average relative humidity of the day before measurement were included as fixed factors to account for potential effects of environmental variation. The multivariate model was run for 1.1 million iterations, with a burn-in period of 100 000 iterations and a thinning interval of 1000, which produced a sample size of 1000 for the posterior distributions of the parameters. We used the least informative inverse Wishart prior that allowed for an autocorrelation below 0.1 after the first thinning interval (lag 1000) on all traits. We defined the prior with V = 1 and nu = 0.02 for the variance–covariance matrix of all traits. This is equivalent to an inverse-gamma distribution of shape and scale equal to 0.01 (Hadfield, 2017). More details on the multivariate model are provided in Supporting Information Methods S1.

We obtained evolvabilities (e) for each trait, understood as the expected proportional change in its average value in one generation under a selection gradient of Strength 1, from the posterior distribution of the genetic component (i.e. variance explained by ‘animal’ in the model) of each trait. Because the measurements were ln-transformed for the model, the posterior variances and covariances obtained are expressed on a proportional scale and thus equivalent to mean-scaled evolvabilities (Hansen et al., 2011). Because the posterior distributions of the genetic and the residual variance (‘units’ in the model) are both in a proportional scale due to the ln-transformation of the trait measurements, we estimated narrow-sense heritability (h²) as the ratio of the total variance of the multivariate model explained by the genetic component (i.e. $h^{2} = \frac{V_{A}}{V_{P}} \equiv \frac{V_{animal}}{V_{animal} + V_{units}}$ ). We report the posterior medians and the 95% credible intervals for both e and h² extracted from all posterior distributions with the HPDinterval()function.

Genetic correlations

We used the posterior distributions of the additive genetic variances (V_A) and covariances (COV_A) between traits yielded by the multivariate model to calculate the additive genetic correlation ( $r_{G}$ ) between pairs of traits as $r_{G (j, k)} = \frac{{COV}_{A} (j, k)}{\sqrt{V_{A} (j) V_{A} (k)}}$ (for traits j and k) to construct 1000 genetic correlation matrices. Note that the posterior variances and covariances were both obtained from the ln-transformed data so that calculating $r_{G}$ using the proportional or absolute values of V_A and COV_A yields the same result. We calculated the 95% credible intervals of the genetic correlation matrices obtained from the posterior distributions of the multivariate model and considered the genetic correlations to be significant when their credible intervals did not overlap with zero.

Conditional evolvability, autonomy and integration

We used the proportional V_A and COV_A of the six directly measured traits (i.e. excluding total sugar production) to study the evolvability parameters of the floral phenotype. We analysed the 1000 G-matrices resulting from the posterior distributions of the animal model using the package evolvability (Bolstad, 2021) and following the methodology developed by Hansen & Houle (2008). We estimated the expected mean evolvability (e, equivalent to the mean-standardised additive genetic variance V_A), respondability (r, the length of the response of the phenotype to a vector of selection gradients), conditional evolvability (c, the ability to respond to selection when all other traits are not allowed to change), autonomy (a, the fraction of V_A that is independent of constraining traits) and integration (i, the relative reduction in evolvability due to correlated traits) of the floral phenotype, morphological and nectar traits, along 1000 randomly generated unit-length vectors of selection gradients. Note that, being $β$ a vector of selection gradients acting on a set of traits, $a (β) \equiv 1 - i (β) = \frac{c (β)}{e (β)}$ .

To construct an integration matrix (I), we used the median G-matrix of the posterior distribution from the model to calculate c for each trait under scenarios of stabilising selection acting on one trait at a time. We first estimated autonomy of each floral trait with respect to each of the other traits and constructed an autonomy matrix (A). To achieve this, we assumed stabilising selection to be acting on one trait at a time and calculated the conditional sub-G-matrix with all other traits using conditionalG(). We then estimated the autonomy of all other traits in relation to the stabilised trait. We repeated this process stabilising one trait at a time and constructed A with the relative pairwise a of each pair of traits. We then calculated I as $I = 1 - A$ (details in Methods S2).

Results

Phenotypic correlations of floral traits

We measured six floral traits in over 1000 flowers from 362 individual plants (Table 1) and constructed a phenotypic correlation matrix with all morphological and nectar traits (Fig. 3, upper diagonal elements). All pairs of morphological traits showed significant positive phenotypic correlations between them, while only some of the dimensions of the whole corolla showed positive correlations with nectar traits. Whole corolla length (WCL) was positively correlated to NPR and total sugar production (SUG), while whole corolla width (WCW) was positively correlated with nectar concentration (CON). Conversely, none of the proximal corolla traits showed a significant phenotypic correlation to any nectar trait. Among morphological traits, proximal corolla length (PCL) and WCW showed the weakest correlation (r = 0.11). The length and width of the whole corolla (WCL and WCW, r = 0.4) showed a stronger positive linear relationship than the dimensions of the proximal corolla (PCL and PCW, r = 0.21). As expected, NPR and CON were strongly negatively correlated. Interestingly, SUG was strongly correlated to NPR but showed a weak and nonsignificant correlation with CON.

Table 1. Summary of measured traits in Digitalis purpurea.

	PCL (cm)	PCW (cm)	WCL (cm)	WCW (cm)	NPR (μl per 24 h)	CON (mg μl⁻¹)
Plants	352	352	352	352	360	360
Flowers	1041	1041	1018	1013	1053	1020
Mean	0.52	0.62	4.81	2.48	6.63	0.35
SD	0.09	0.09	0.43	0.29	6.01	0.10
Variance	0.01	0.01	0.19	0.08	36.17	0.01
Max	0.91	0.93	6.02	3.45	47.07	0.73
Min	0.31	0.35	3.47	1.38	0.17	0.14

CON, nectar sugar concentration; flowers, total number of flowers measured (sample sizes vary across traits in cases with broken/deformed flowers or not enough nectar volume to measure concentration); NPR, nectar production rate; PCL, proximal corolla length; PCW, proximal corolla width; Plants, number of individuals measured; SD, standard deviation; WCL, whole corolla length; WCW, whole corolla width.

Phenotypic plasticity

We aimed to quantify differences in PP between and within functional clusters of floral traits. For this, we computed a total of 917 within-family PhD indices and compared them across the six directly measured traits (Fig. 4; Table S1). Plasticity was marginally higher in nectar than in morphological traits (F_1,610 = 2.17, P = 0.141; Fig. 4a). When compared individually, PP varied significantly among traits (F_5,606 = 6.18, P < 0.001, n = 6 traits; Fig. 4b). PhD indices of the proximal corolla were consistently but not significantly lower than those of the whole corolla. Sugar concentration was the most plastic of all traits and significantly more plastic than NPR. However, it did not differ significantly from WCL (Fig. 4b). We did not detect significant selection acting on PP for any trait, suggesting that PP did not affect reproductive output in the common garden.

Narrow-sense heritability

We calculated the narrow-sense heritability h² for all traits in the common garden. On average, h² values were higher for corolla morphology and lower for nectar traits (Fig. 5a). Within the proximal corolla dimensions, PCL showed higher heritability than PCW. Conversely, dimensions of the whole corolla showed the opposite pattern, with WCW being the most heritable of all analysed traits.

Single trait evolvability

We compared the potential evolvability of each trait calculated as a proxy for the proportional change expected in response to a unit strength of selection (Fig. 5b). Within floral morphology, PCL was the most evolvable trait (e = 0.98%) in contrast to WCL (e = 0.27%). Overall, SUG and NPR showed not only the highest evolvability of all traits (e = 5.12% and 4.66% respectively) but also the highest uncertainty. CON showed the lowest evolvability of nectar traits (e = 0.69%).

Genetic correlations

Genetically correlated traits tend to be more integrated, and this can affect their response to selection. To quantify the genetic correlations between individual traits, we rescaled the estimates of additive genetic covariances of all trait pairs (Table S2) to construct a genetic correlation matrix (Fig. 3 lower diagonal elements; Table S3). Morphological traits showed positive genetic correlations, except for PCL and PCW, which presented a slightly negative genetic correlation. Conversely, those same dimensions of the whole corolla showed a significant positive genetic correlation. In contrast to the phenotypic correlations, nectar concentration and production rate showed a nonsignificant negative genetic correlation. Nectar and morphological traits were positively but nonsignificantly genetically correlated. See Table S4 for complete P and G-matrices with variances and covariances between traits.

Conditional evolvability, autonomy and integration

Since genetic correlations may affect the evolvability of single traits and their level of integration and autonomy within the flower, we analysed the genetic variances and covariances from the posterior distribution (n = 1000) of the G-matrix to obtain estimates of evolvability, conditional evolvability, integration and autonomy. A summary of the mean evolvability parameters extracted from the posterior distribution of G-matrices over 1000 randomly generated unit-length selection gradients is shown in Fig. 6 and Table S5. The floral phenotype as a whole showed a relatively high median evolvability of e = 1.32%. Its conditional evolvability was very low (median c = 0.37%) probably due to the relatively high integration between floral traits (i = 64.58%). Conversely, analysing the G-matrix of the corolla dimensions in isolation from nectar traits, the corolla showed lower integration (i = 38.59%) than nectar traits (i = 43.34%; Fig. 6; Table S5).

We constructed an integration matrix by assuming stabilising selection on one trait at a time (Table 2). The two dimensions of the proximal corolla were much less integrated among them (i = 0.02%) than those of the whole corolla (i = 24.78%). Whole corolla length was the least autonomous trait since it showed the highest integration percentages with the rest of traits. Nectar traits were on average the most autonomous traits, with concentration being the least integrated trait of the whole floral phenotype.

Table 2. Integration matrix of Digitalis purpurea floral traits.

Trait	PCL	PCW	WCL	WCW	NPR	CON
PCL		0.02	18.13	3.62	2.59	5.46
PCW	0.02		9.18	9.30	4.23	0.04
WCL	18.13	9.18		24.78	7.42	2.56
WCW	3.62	9.30	24.78		8.74	0.48
NPR	2.59	4.23	7.42	8.74		1.26
CON	5.46	0.04	2.56	0.48	1.26

Values indicate the percentage of integration (i) between traits. Headers indicate which trait was assumed to be under stabilising selection (i.e. conditioning the others). CON, nectar sugar concentration; NPR, nectar production rate; PCL, proximal corolla length; PCW, proximal corolla width; WCL, whole corolla length; WCW, whole corolla width.

Discussion

Flowers tend to be highly integrated organs, but different types of traits show different levels of evolutionary potential in response to selection. We studied the factors underlying the variation of nectar and morphological traits in pollinator-dependent D. purpurea, and their importance in the rapid adaptive change observed in the morphology of the proximal corolla, but not in other morphological or nectar traits, after the addition of hummingbirds as pollinators. Morphological traits showed overall higher levels of heritable variation than nectar traits, and the proximal corolla presented higher evolvability in its length dimension and lower integration than the rest of the corolla. As expected, nectar traits showed higher, albeit marginally, levels of plastic variation than morphological traits due to the high plasticity of nectar concentration. Nectar production rate showed high evolvability and could experience rapid adaptive change under novel selection. We discuss our findings in the context of previous measurements of selection and rapid evolution in this system, as well as the wider implications for understanding floral diversification in response to changes in pollination.

Environmental sensitivity and correlations vary throughout the floral phenotype

Patterns of genetic and phenotypic correlations differed for nectar and morphological traits. While phenotypic and genetic correlations of morphological traits showed similar trends, phenotypic correlations of nectar traits, both among them and with corolla traits, were weaker and did not reflect the genetic correlations. This is likely due to differences in environmental sensitivity derived from ontogenetic and physiological differences among groups of traits. While corolla morphology is stable throughout anthesis once flower tissues complete development, nectar traits remain flexible throughout the life of the flower, varying with conditions such as humidity and temperature, which determine evaporation rates (Mitchell, 2004; Parachnowitsch et al., 2019). Therefore, any potential response to selection depending on the interaction of NPR and concentration (CON) with other traits may be constrained by their high plastic variation.

Consistent with that, the nectar phenotype was marginally more plastic than corolla morphology. However, no clear patterns were detected in a trait-by-trait comparison, with some of the corolla traits showing higher plasticity than NPR. Nectar concentration was the most plastic trait overall and significantly more plastic than NPR. The high environmental variability of CON could make its response to selection slower, reducing its evolvability and buffering any potential long-lasting effect of selection. This is consistent with the lack of adaptive change in nectar traits reported by Mackin et al. (2021b), even when under positive selection as is the case of NPR.

The significant, positive but inconsistent phenotypic correlations found between the dimensions of the whole corolla and nectar traits match those reported in the closely related genus Penstemon (Wessinger et al., 2014) and are likely due to allometric relationships between flower size and nectar traits. This is reported in other species, with larger flowers producing more nectar (Tavares et al., 2016). If correlations between reward and flower size originate via pollinator selection on signal accuracy they may turn corolla size into an honest signal (Benitez-Vieyra et al., 2014). Notoriously, we found that dimensions of the proximal corolla are not correlated with nectar traits in D. purpurea.

Total sugar production (SUG) and production rate (NPR) were strongly and positively correlated to each other genetically and phenotypically. While this correlation might seem to be an artefact of SUG being derived from NPR, it was absent between SUG and concentration (CON). Postsecretion regulation mechanisms such as sugar and water reabsorption contribute to decoupling nectar concentration and sugar content (Nepi & Stpiczynska, 2008). Previous research reports similar patterns. Klinkhamer & Wijk (1999) found a strong correlation between nectar volume and sugar production but not between sugar production and concentration in Echium vulgare. Similar genetic correlational patterns were reported in Nicotiana alata (Kaczorowski et al., 2008).

Conversely to nectar traits, genetic correlations of corolla morphology were largely preserved in the floral phenotype. Length components of the two portions of the corolla were phenotypically and genetically correlated. This is likely a consequence of linkage disequilibrium caused by shared developmental regulation and the expected canalisation of floral morphology. Since D. purpurea relies on pollinators for reproduction, genetic correlation and integration between different sections of the corolla can be favoured by pollinator-mediated selection, to maintain adaptive accuracy (Armbruster et al., 2009). Our results are consistent with the pollinator-mediated stabilising selection hypothesis, stating that zygomorphic flowers, which tend to rely on relatively specialised pollinators, show low environmental variation in flower size to minimise the costs of pollen loss (Wolfe & Krstolic, 1999). Berg (1960) discussed that canalisation of corolla dimensions in entomophilous outcrossing species, such as D. purpurea, responds to evolutionary constraints imposed by the body size of pollinators and the average corolla dimensions in each population. This may lead to any response to novel selection acting on one dimension of the corolla to be constrained by stabilising selection acting on other dimensions (Hansen et al., 2003).

Heritability and evolvability differ among types of floral traits

Our analysis of single trait narrow-sense heritability (h²) and evolvability (e) showed a lack of correlation between the two measures and among the two functional groups. Morphological traits showed higher h² than nectar traits, most likely due to the higher environmental phenotypic variation shown by nectar traits increasing their overall phenotypic variation. Differences between h² of morphological and nectar traits have been reported, for example for N. alata (Kaczorowski et al., 2008) and Penstemon centranthifolius (Mitchell & Shaw, 1993). The dimensions of the whole corolla showed higher heritability than those of the proximal corolla in D. purpurea. This is consistent with expectations that high variability of corolla sections involved in pollen transport that deviate from the population mean would imply costs for individual fitness (Berg, 1960).

Conversely, our evolvability estimates, which express V_A in a proportional scale, showed very different trends to h². This is likely due to correlations between V_A and other variance components (Hansen et al., 2011). Within corolla morphology, the PCL showed the highest evolvability, consistent with the observed rapid adaptive change in the introduced populations under positive selection pressure for larger PCL. The high uncertainty of the estimates of evolvability for NPR and total sugar content (derived from NPR) obscured reliable comparisons. We attribute the broad credible intervals in these traits to high environmental variation under natural conditions.

We are currently working on comparisons in controlled environments with large sample sizes for a more accurate estimation of the variance of nectar traits in D. purpurea (including relative sugar composition). However, the variation observed in the common garden is likely a good reflection of what would be expected to occur in natural conditions and of the limitations of selection to act on environmentally sensitive traits. This is interesting given that nectar production has been repeatedly found to be determined by large-effect loci in QTL studies of linages with shifts to hummingbird pollination (reviewed by Wessinger, 2024). From a genetic architecture perspective, the evolution of nectar traits could happen rapidly under strong selection. Our results show that environmentally induced variability could slow the process.

Integration and evolvability vary within corolla morphology

We estimated several evolvability parameters of different traits and of the floral phenotype through the response of their G-matrix to selection. The floral phenotype showed a high mean evolvability ( $\bar{e}$ ), which forecasts a change of 1.32% in the average values of floral traits per generation under a selection gradient of Strength 1. However, the mean conditional evolvability ( $\bar{c}$ ), or the expected change in average values of floral traits assuming the rest of them to be under stabilising selection, dropped to 0.37%. Note that e and c represent the upper and lower limit of the evolutionary potential of the target trait and that c could be a better predictor for adaptive change over longer timescales (Opedal et al., 2023).

We also analysed the G-submatrices of morphological and nectar traits separately. Corolla morphology was less integrated than nectar traits, with a higher portion of the evolvability of morphological traits being independent from the selection on the rest of the corolla. This partially contradicts reports of zygomorphic flowers showing high integration of flower morphological traits (Ashman & Majetic, 2006). While this is true for the dimensions of the whole corolla of D. purpurea, our results suggest that both sections of the corolla behave as independent modules with different autonomy to respond to selection, with the proximal section (not directly involved in pollen transfer) being more evolutionarily flexible than the rest of the corolla. The genetic architecture of these traits in D. purpurea is unknown, but QTL studies in several lineages with shifts to hummingbird pollination show multiple loci of small or medium effect underlying the variation in corolla morphology (Stuurman et al., 2004; Wessinger et al., 2014; Alexandre et al., 2015; Kostyun et al., 2019), suggesting potential for complex responses. Regardless of the genetic basis, this decoupling between sections of the corolla might be derived from them being shaped by differential patterns of natural selection (to be described later).

The higher autonomy of the proximal corolla is illustrated by the integration matrix, where the median integration percentage between dimensions of the whole corolla (24.78%) is three orders of magnitude higher than for the proximal corolla (0.02%). Thus, the proximal corolla is less constrained in its response to the novel directional selection observed in the introduced populations, consistent with the rapid adaptive change observed in the non-native range (Mackin et al., 2021b). This also means that stabilising selection on the length or width of the proximal corolla would exert a very small effect on the evolvability of the other dimension.

The lack of integration of the proximal corolla, together with the lower plasticity of the PCL, makes this trait more evolutionarily flexible than the rest of the floral phenotype. This allows the length and width of the proximal corolla to respond differently to novel selection, unlike the whole corolla, where dimensions are more integrated. In D. purpurea, medial and distal corolla sections are directly involved in pollen transfer due to the location of the anthers and stigma, so higher integration of their dimensions may reduce pollen loss. Different selection pressures acting on different clusters of traits can promote intrafloral over floral integration and intrafloral modularity (Ordano et al., 2008; Bissell & Diggle, 2010). We hypothesise that the functionally different roles of the proximal and distal sections of the corolla of D. purpurea led to evolutionarily autonomous intrafloral modules. While the dimensions of the rest of the corolla are canalised and integrated to assure the accuracy of pollen transfer, the proximal corolla may respond separately to selection on access to nectar. This, combined with directional selection for longer corollas in the non-native range, has allowed the proximal corolla to undergo rapid adaptive change.

Concluding remarks

Our multivariate approach, combined with previous work in wild populations, provided novel information on the quantitative genetic basis of D. purpurea flower evolution. A caveat is that we studied populations introduced to novel environments c. 85 generations ago and even within this short period, selection regimes might have fluctuated. The G-matrix is itself evolvable and subject to change over different timescales (Arnold et al., 2008). Further research in natural populations remains important to understand patterns of selection and population-level variation in evolutionary potential. Ideally, a reciprocal transplant would capture the environmental differences between native and introduced ranges, which may be wider than in our common garden. Other factors like nectar robbing, practically absent in native but widespread in introduced populations, might exert selection on flower traits affecting their adaptive potential (Mackin et al., 2021a).

Nevertheless, we show how phenotypic variation of floral traits, determined by PP, additive genetic variation and integration, affects their differential ability to respond to selection. Nectar and morphology differ in evolutionary potential, and within the corolla of D. purpurea, the proximal section is more evolutionarily flexible than the rest. This is an example of how floral modularity plays a fundamental role in maintaining flower evolvability and enables quick adaptive responses (Dellinger et al., 2019). The fact that this can happen in very short time scales suggests an important role of rapid adaptation in floral evolution of angiosperms.

Acknowledgements

We thank Chris Mackin and Julián Felipe Peña for their help with seed collection in the introduced populations, C. Mackin for sharing the F1 generation of plants used for the crossings in this study, Crispin Holloway for his support in the glasshouse and field facilities, Noah Bourne, Ethan Millward and Lucy Johnson for their help with the data collection in the common garden, and Øystein Opedal for his valuable comments on data analysis. ARB was supported by a PhD scholarship from the School of Life Sciences (University of Sussex).

Competing interests

None declared.

Author contributions

ARB and MCC designed and performed the research. ARB conducted the data collection and performed the data analysis. ARB and MCC interpreted the results and wrote the manuscript.

Open Research

Data availability

All data and analysis scripts are openly available in figshare doi: 10.25377/sussex.25590582.

Supporting Information

References

Alexandre H, Vrignaud J, Mangin B, Joly S. 2015. Genetic architecture of pollination syndrome transition between hummingbird-specialist and generalist species in the genus Rhytidophyllum (Gesneriaceae). PeerJ 3: e1028.
10.7717/peerj.1028
PubMedWeb of Science®Google Scholar
Armbruster WS, Di Stilio VS, Tuxill JD, Flores TC, Velásquez Runk JL. 1999. Covariance and decoupling of floral and vegetative traits in nine Neotropical plants: a re-evaluation of Berg's correlation-pleiades concept. American Journal of Botany 86: 39–55.
10.2307/2656953
CASPubMedWeb of Science®Google Scholar
Armbruster WS, Hansen TF, Pélabon C, Pérez-Barrales R, Maad J. 2009. The adaptive accuracy of flowers: measurement and microevolutionary patterns. Annals of Botany 103: 1529–1545.
10.1093/aob/mcp095
PubMedWeb of Science®Google Scholar
Arnold SJ, Bürger R, Hohenlohe PA, Ajie BC, Jones AG. 2008. Understanding the evolution and stability of the G-matrix. Evolution 62: 2451–2461.
10.1111/j.1558-5646.2008.00472.x
PubMedWeb of Science®Google Scholar
Ashman T-L, Majetic CJ. 2006. Genetic constraints on floral evolution: a review and evaluation of patterns. Heredity 96: 343–352.
10.1038/sj.hdy.6800815
PubMedWeb of Science®Google Scholar
Benitez-Vieyra S, Fornoni J, Pérez-Alquicira J, Boege K, Domínguez CA. 2014. The evolution of signal–reward correlations in bee- and hummingbird-pollinated species of Salvia. Proceedings of the Royal Society B: Biological Sciences 281: 20132934.
10.1098/rspb.2013.2934
PubMedWeb of Science®Google Scholar
Berg RL. 1959. A general evolutionary principle underlying the origin of developmental homeostasis. The American Naturalist 93: 103–105.
10.1086/282061
Web of Science®Google Scholar
Berg RL. 1960. The ecological significance of correlation pleiades. Evolution 14: 171–180.
10.1111/j.1558-5646.1960.tb03076.x
Web of Science®Google Scholar
Best LS, Bierzychudek P. 1982. Pollinator foraging on foxglove (Digitalis purpurea): a test of a new model. Evolution 36: 70–79.
10.1111/j.1558-5646.1982.tb05011.x
PubMedWeb of Science®Google Scholar
Bissell EK, Diggle PK. 2010. Modular genetic architecture of floral morphology in Nicotiana: quantitative genetic and comparative phenotypic approaches to floral integration. Journal of Evolutionary Biology 23: 1744–1758.
10.1111/j.1420-9101.2010.02040.x
CASPubMedWeb of Science®Google Scholar
Bolstad GH. 2021. evolvability: calculation of evolvability parameters. [WWW document] URL https://CRAN.R-project.org/package=evolvability [accessed 29 January 2023].
Google Scholar
Bradshaw AD. 1965. Evolutionary significance of phenotypic plasticity in plants. Advances in Genetics 13: 115–155.
10.1016/S0065-2660(08)60048-6
PubMedWeb of Science®Google Scholar
Broadbent A, Bourke A. 2012. The bumblebees Bombus hortorum is the main pollinating visitor to Digitalis purpurea (Common Foxglove) in a U.K. population. Journal of Pollination Ecology 8: 48–51.
10.26786/1920-7603(2012)10
Google Scholar
Campbell DR. 1996. Evolution of floral traits in a hermaphroditic plant: field measurements of heritabilities and genetic correlations. Evolution 50: 1442–1453.
10.1111/j.1558-5646.1996.tb03918.x
PubMedWeb of Science®Google Scholar
Corl A, Bi K, Luke C, Challa AS, Stern AJ, Sinervo B, Nielsen R. 2018. The genetic basis of adaptation following plastic changes in coloration in a novel environment. Current Biology 28: 2970–2977.
10.1016/j.cub.2018.06.075
CASPubMedWeb of Science®Google Scholar
Crispo E. 2007. The Baldwin effect and genetic assimilation: revisiting two mechanisms of evolutionary change mediated by phenotypic plasticity. Evolution 61: 2469–2479.
10.1111/j.1558-5646.2007.00203.x
CASPubMedWeb of Science®Google Scholar
Davidson AM, Jennions M, Nicotra AB. 2011. Do invasive species show higher phenotypic plasticity than native species and, if so, is it adaptive? A meta-analysis. Ecology Letters 14: 419–431.
10.1111/j.1461-0248.2011.01596.x
PubMedWeb of Science®Google Scholar
Dellinger AS, Artuso S, Pamperl S, Michelangeli FA, Penneys DS, Fernández-Fernández DM, Alvear M, Almeda F, Scott Armbruster W, Staedler Y et al. 2019. Modularity increases rate of floral evolution and adaptive success for functionally specialized pollination systems. Communications Biology 2: 1–11.
PubMedWeb of Science®Google Scholar
Falconer DS, Mackay TFC. 1996. Introduction to quantitative genetics, 4^th edn. Longman, Harlow, Essex, UK: Pearson Education.
Google Scholar
Fenster CB, Armbruster WS, Wilson P, Dudash MR, Thomson JD. 2004. Pollination syndromes and floral specialization. Annual Review of Ecology, Evolution, and Systematics 35: 375–403.
10.1146/annurev.ecolsys.34.011802.132347
Web of Science®Google Scholar
Friis EM, Crane PR, Pedersen KR. 2011. Early flowers and angiosperm evolution. Cambridge, UK: Cambridge University Press.
10.1017/CBO9780511980206
Google Scholar
Hadfield JD. 2010. MCMC methods for multi-response generalized linear mixed models: the MCMCglmm R package. Journal of Statistical Software 33: 1–22.
10.18637/jss.v033.i02
PubMedWeb of Science®Google Scholar
Hadfield JD. 2017. MCMCglmm course notes. [WWW document] URL http://cran.nexr.com/web/packages/MCMCglmm/vignettes/CourseNotes.pdf [accessed 4 June 2022].
Google Scholar
Hansen TF, Armbruster WS, Carlson ML, Pélabon C. 2003. Evolvability and genetic constraint in dalechampia blossoms: genetic correlations and conditional evolvability. Journal of Experimental Zoology Part B: Molecular and Developmental Evolution 296B: 23–39.
10.1002/jez.b.14
Web of Science®Google Scholar
Hansen TF, Houle D. 2008. Measuring and comparing evolvability and constraint in multivariate characters. Journal of Evolutionary Biology 21: 1201–1219.
10.1111/j.1420-9101.2008.01573.x
CASPubMedWeb of Science®Google Scholar
Hansen TF, Pélabon C, Houle D. 2011. Heritability is not evolvability. Evolutionary Biology 38: 258–277.
10.1007/s11692-011-9127-6
Web of Science®Google Scholar
Harder LD, Barrett SC, eds. 2006. Ecology and evolution of flowers. New York, NY, USA: Oxford University Press.
10.1093/oso/9780198570851.001.0001
Google Scholar
Haynes WM. 2014. CRC handbook of chemistry and physics, 95^th edn. Boca Raton, FL, USA: CRC Press.
10.1201/b17118
Google Scholar
Herrera CM, Pellmyr O, eds. 2002. Plant animal interactions: an evolutionary approach. Oxford, UK: Wiley-Blackwell.
Google Scholar
Hothorn T, Bretz F, Westfall P. 2008. Simultaneous inference in general parametric models. Biometrical Journal 50: 346–363.
10.1002/bimj.200810425
CASPubMedWeb of Science®Google Scholar
Kaczorowski RL, Juenger TE, Holtsford TP. 2008. Heritability and correlation structure of nectar and floral morphology traits in Nicotiana alata. Evolution 62: 1738–1750.
10.1111/j.1558-5646.2008.00400.x
PubMedWeb of Science®Google Scholar
van Kleunen M. 2007. Adaptive genetic differentiation in life-history traits between populations of Mimulus guttatus with annual and perennial life-cycles. Evolutionary Ecology 21: 185–199.
10.1007/s10682-006-0019-7
Web of Science®Google Scholar
Klinkhamer PGL, Wijk CAMVDV. 1999. Genetic variation in floral traits of Echium vulgare. Oikos 85: 515–522.
10.2307/3546700
Web of Science®Google Scholar
Kostyun JL, Gibson MJS, King CM, Moyle LC. 2019. A simple genetic architecture and low constraint allow rapid floral evolution in a diverse and recently radiating plant genus. New Phytologist 223: 1009–1022.
10.1111/nph.15844
CASPubMedWeb of Science®Google Scholar
Lande R. 1979. Quantitative genetic analysis of multivariate evolution, applied to brain:body bize allometry. Evolution 33: 402–416.
10.1111/j.1558-5646.1979.tb04694.x
PubMedWeb of Science®Google Scholar
Leiss KA, Vrieling K, Klinkhamer PGL. 2004. Heritability of nectar production in Echium vulgare. Heredity 92: 446–451.
10.1038/sj.hdy.6800439
CASPubMedWeb of Science®Google Scholar
Levis NA, Pfennig DW. 2016. Evaluating “plasticity-first” evolution in nature: key criteria and empirical approaches. Trends in Ecology & Evolution 31: 563–574.
10.1016/j.tree.2016.03.012
PubMedWeb of Science®Google Scholar
Lynch M, Walsh B. 1998. Genetics and analysis of quantitative traits. Sunderland, MA, USA: Sinauer.
CASGoogle Scholar
Mackin CR, Goulson D, Castellanos MC. 2021a. Novel nectar robbing negatively affects reproduction in Digitalis purpurea. Ecology and Evolution 11: 13455–13463.
10.1002/ece3.8068
PubMedWeb of Science®Google Scholar
Mackin CR, Peña Polaina JF, Blanco MA, Balfour NJ, Castellanos MC. 2021b. Rapid evolution of a floral trait following acquisition of novel pollinators. Journal of Ecology 109: 2234–2246.
10.1111/1365-2745.13636
Web of Science®Google Scholar
Mitchell RJ. 2004. Heritability of nectar traits: why do we know so little? Ecology 85: 1527–1533.
10.1890/03-0388
Web of Science®Google Scholar
Mitchell RJ, Shaw RG. 1993. Heritability of floral traits for the perennial wild flower Penstemon centranthifolius (Scrophulariaceae): clones and crosses. Heredity 71: 185–192.
10.2307/1940769
Web of Science®Google Scholar
Morrissey MB, Sakrejda K. 2013. Unification of regression-based methods for the analysis of natural selection. Evolution 67: 2094–2100.
10.1111/evo.12077
PubMedWeb of Science®Google Scholar
Morrissey M, Sakrejda K. 2014. gsg: calculation of selection coefficients. [WWW document] URL https://cran.r-project.org/package=gsg [accessed 24 November 2022].
Google Scholar
Nazir R, Reshi Z, Wafai BA. 2008. Reproductive ecology of medicinally important Kashmir Himalayan species of Digitalis L. Plant Species Biology 23: 59–70.
10.1111/j.1442-1984.2008.00214.x
Web of Science®Google Scholar
Nepi M, Stpiczynska M. 2008. Do plants dynamically regulate nectar features through sugar sensing? Plant Signaling & Behavior 3: 874–876.
10.4161/psb.3.10.6228
PubMedGoogle Scholar
Nicotra AB, Atkin OK, Bonser SP, Davidson AM, Finnegan EJ, Mathesius U, Poot P, Purugganan MD, Richards CL, Valladares F et al. 2010. Plant phenotypic plasticity in a changing climate. Trends in Plant Science 15: 684–692.
10.1016/j.tplants.2010.09.008
CASPubMedWeb of Science®Google Scholar
Ollerton J, Watts S, Connerty S, Lock J, Parker L, Wilson I, Schueller S, Nattero J, Cocucci AA, Izhaki I et al. 2012. Pollination ecology of the invasive tree tobacco Nicotiana glauca: comparisons across native and non-native ranges. Journal of Pollination Ecology 9: 85–95.
10.26786/1920-7603(2012)12
Google Scholar
Opedal ØH. 2019. The evolvability of animal-pollinated flowers: towards predicting adaptation to novel pollinator communities. New Phytologist 221: 1128–1135.
10.1111/nph.15403
PubMedWeb of Science®Google Scholar
Opedal ØH, Armbruster WS, Hansen TF, Holstad A, Pélabon C, Andersson S, Campbell DR, Caruso CM, Delph LF, Eckert CG et al. 2023. Evolvability and trait function predict phenotypic divergence of plant populations. Proceedings of the National Academy of Sciences, USA 120: e2203228120.
10.1073/pnas.2203228120
CASPubMedWeb of Science®Google Scholar
Opedal ØH, Hildesheim LS, Armbruster WS. 2022. Evolvability and constraint in the evolution of three-dimensional flower morphology. American Journal of Botany 109: 1906–1917.
10.1002/ajb2.16092
PubMedWeb of Science®Google Scholar
Ordano M, Fornoni J, Boege K, Domínguez CA. 2008. The adaptive value of phenotypic floral integration. New Phytologist 179: 1183–1192.
10.1111/j.1469-8137.2008.02523.x
CASPubMedWeb of Science®Google Scholar
Parachnowitsch AL, Manson JS, Sletvold N. 2019. Evolutionary ecology of nectar. Annals of Botany 123: 247–261.
10.1093/aob/mcy132
CASPubMedWeb of Science®Google Scholar
Pascoal S, Carvalho G, Creer S, Rock J, Kawaii K, Mendo S, Hughes R. 2012. Plastic and heritable components of phenotypic variation in Nucella lapillus: an assessment using reciprocal transplant and common garden experiments. PLoS ONE 7: e30289.
10.1371/journal.pone.0030289
CASPubMedWeb of Science®Google Scholar
Payne JL, Wagner A. 2019. The causes of evolvability and their evolution. Nature Reviews Genetics 20: 24–38.
10.1038/s41576-018-0069-z
CASPubMedWeb of Science®Google Scholar
Pélabon C, Armbruster WS, Hansen TF. 2011. Experimental evidence for the Berg hypothesis: vegetative traits are more sensitive than pollination traits to environmental variation. Functional Ecology 25: 247–257.
10.1111/j.1365-2435.2010.01770.x
Web of Science®Google Scholar
Percival M, Morgan P. 1965. Observations on the floral biology of Digitalis species. New Phytologist 64: 1–22.
10.1111/j.1469-8137.1965.tb05370.x
Web of Science®Google Scholar
DW Pfennig, ed. 2021. Phenotypic plasticity & evolution: causes, consequences, controversies. Boca Raton, FL, USA: CRC Press.
10.1201/9780429343001
Google Scholar
Pigliucci M, Murren CJ, Schlichting CD. 2006. Phenotypic plasticity and evolution by genetic assimilation. Journal of Experimental Biology 209: 2362–2367.
10.1242/jeb.02070
PubMedWeb of Science®Google Scholar
Posit Team. 2022. RStudio: integrated development environment for R. Boston, MA, USA: Posit Software, PBC. [WWW document] URL http://www.posit.co/ [accessed 15 May 2024].
Google Scholar
Puglielli G, Carmona CP, Varone L, Laanisto L, Ricotta C. 2022. Phenotypic dissimilarity index: correcting for intra and interindividual variability when quantifying phenotypic variation. Ecology 103: e3806.
10.1002/ecy.3806
PubMedWeb of Science®Google Scholar
R Core Team. 2022. R: a language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. [WWW document] URL https://www.R-project.org/ [accessed 15 May 2024].
Google Scholar
Riveros AJ, Hernández EJ, Nates-Parra G, Building SS. 2006. Morphological constraints and nectar robbing in three andean bumble bee species (Hymenoptera, Apidae, Bombini). Caldasia 28: 111–114.
Google Scholar
Rojas-Nossa SV, Sánchez JM, Navarro L. 2016. Nectar robbing: a common phenomenon mainly determined by accessibility constraints, nectar volume and density of energy rewards. Oikos 125: 1044–1055.
10.1111/oik.02685
Web of Science®Google Scholar
Schneider HM. 2022. Characterization, costs, cues and future perspectives of phenotypic plasticity. Annals of Botany 130: 131–148.
10.1093/aob/mcac087
CASPubMedWeb of Science®Google Scholar
Stebbins GL. 1950. Variation and evolution in plants, variation and evolution in plants. New York, NY, USA: Columbia University Press.
10.7312/steb94536
Google Scholar
Stebbins GL. 1970. Adaptive radiation of reproductive characteristics in angiosperms, I: pollination mechanisms. Annual Review of Ecology and Systematics 1: 307–326.
10.1146/annurev.es.01.110170.001515
Google Scholar
Stuurman J, Hoballah ME, Broger L, Moore J, Basten C, Kuhlemeier C. 2004. Dissection of floral pollination syndromes in petunia. Genetics 168: 1585–1599.
10.1534/genetics.104.031138
CASPubMedWeb of Science®Google Scholar
Tavares DC, Freitas L, Gaglianone MC. 2016. Nectar volume is positively correlated with flower size in hummingbird-visited flowers in the Brazilian Atlantic Forest. Journal of Tropical Ecology 32: 335–339.
10.1017/S0266467416000250
Web of Science®Google Scholar
Valladares F, Matesanz S, Guilhaumon F, Araújo MB, Balaguer L, Benito-Garzón M, Cornwell W, Gianoli E, van Kleunen M, Naya DE et al. 2014. The effects of phenotypic plasticity and local adaptation on forecasts of species range shifts under climate change. Ecology Letters 17: 1351–1364.
10.1111/ele.12348
CASPubMedWeb of Science®Google Scholar
Valladares F, Sanchez-Gomez D, Zavala MA. 2006. Quantitative estimation of phenotypic plasticity: bridging the gap between the evolutionary concept and its ecological applications. Journal of Ecology 94: 1103–1116.
10.1111/j.1365-2745.2006.01176.x
Web of Science®Google Scholar
Warton DI, Hui FKC. 2011. The arcsine is asinine: the analysis of proportions in ecology. Ecology 92: 3–10.
10.1890/10-0340.1
PubMedWeb of Science®Google Scholar
Wessinger CA. 2024. How the switch to hummingbird pollination has greatly contributed to our understanding of evolutionary processes. New Phytologist 241: 59–64.
10.1111/nph.19335
PubMedWeb of Science®Google Scholar
Wessinger CA, Hileman LC, Rausher MD. 2014. Identification of major quantitative trait loci underlying floral pollination syndrome divergence in Penstemon. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences 369: 20130349.
10.1098/rstb.2013.0349
PubMedWeb of Science®Google Scholar
West-Eberhard MJ. 2003. Developmental plasticity and evolution. Oxford, UK: Oxford University Press.
10.1111/j.1365-2435.2007.01283.x
CASGoogle Scholar
Wolfe LM, Krstolic JL. 1999. Floral symmetry and its influence on variance in flower size. The American Naturalist 154: 484–488.
10.1086/303249
PubMedWeb of Science®Google Scholar
Wood S. 2023. mgcv: mixed GAM computation vehicle with automatic smoothness estimation. [WWW document] URL https://CRAN.R-project.org/package=mgcv [accessed 24 November 2022].
Google Scholar

Citing Literature

Volume243, Issue2

July 2024

Pages 753-764

Nectar and floral morphology differ in evolutionary potential in novel pollination environments

Summary

Introduction