Trait divergence, not plasticity, determines the success of a newly invasive plant

Abstract

Background and Aims

Phenotypic plasticity and genetic differentiation both play important roles in the establishment and spread of species after extra-range dispersal; however, the adaptive potential of plasticity and genetic divergence in successful invasions remains unclear.

Methods

We measured six anatomical traits associated with drought tolerance in contrasting water environments for individuals from the invasive and native range of the bunchgrass Brachypodium sylvaticum. To represent sources contributing to admixed genotypes in the invasive range accurately, we used unique alleles to determine probabilities of genetic contribution, and utilized these as weights in our analyses. The adaptive values of plasticity and genetic differentiation were assessed using regression.

Key Results

No plasticity was found in response to water availability for any of the measured traits. Bulliform cell area and three traits related to xylem morphology displayed genetic differentiation between invasive and native ranges, indicating a shift in the invasive range towards drought-tolerant phenotypes. Genetic divergence was not consistently in the direction indicated by selection, suggesting that responses are limited by trade-offs with other traits or physical constraints.

Conclusions

Our results indicate that invasive adaptation is the consequence of post-introduction selection leading to genetic differentiation. Selection, rather than plasticity, is driving B. sylvaticum success in its invaded range.

INTRODUCTION

Phenotypic plasticity and genetic change are thought to play major roles in the establishment and spread of invasive species (Lambrinos, 2004; Dybdahl and Kane, 2005; Geng et al., 2007; Dlugosch and Parker, 2008). After introduction to a novel habitat, plasticity may promote the establishment of colonizing populations and the subsequent spread of invasive plants into a wide range of environments (Droste et al., 2010; Davidson et al., 2011; Martina and von Ende, 2012). Although many previous studies have investigated the role of plasticity in invasion, there is no consensus on whether plasticity is more common in invasive species or if it aids invasion (reviewed in Davidson et al., 2011; Palacio-Lopez and Gianoli, 2011). This is particularly true of functional traits (i.e. traits that determine how species interact with their environment), which generally do not exhibit higher plasticity for invasive species (Godoy et al., 2011; Murphy et al., 2016). Alternatively, genetic change can occur over short time spans once a species is introduced into a new environment and may result in local adaptation that aids invasion (van Kleunen and Fischer, 2008; Alexander et al., 2009; Eriksen et al., 2012; Felker-Quinn et al., 2013). While it has been suggested that bottlenecks during introduction would limit the adaptive potential of invasive populations (Schrieber and Lachmuth, 2017), the large majority of invasions have been initiated by multiple introductions, suggesting that admixture is important for invasion success (e.g. Lavergne and Molofsky, 2007; Rosenthal et al., 2008; Olivieri, 2009; Estoup and Guillemaud, 2010). Further investigation into the contributions of plasticity and genetic adaptation to success in novel environments will aid in our understanding of processes underpinning invasion success.

Phenotypic plasticity is defined as the ability of organisms to modify morphological or physiological traits in response to environmental conditions, and can be adaptive if plasticity increases the ability of a species to survive and reproduce in comparison with non-plastic genotypes (Callaway et al., 2003; Pigliucci and Murren, 2003; Richards et al., 2006; Ghalambor et al., 2007). The ability of invasive plants to colonize a novel environment is commonly attributed to species possessing generalist genotypes with high phenotypic plasticity, as plasticity allows introduced individuals to shift trait expression towards optimal values favoured by the local environment (Matesanz et al., 2012; Zhao et al., 2012). For example, plants that exhibit plasticity in response to drought stress tolerate limited water availability by changing morphological or physiological traits to match the prevailing environment (Picotte et al., 2007; Meier and Leuschner, 2008; Bartlett et al., 2012). There is evidence that phenotypic plasticity is common in invasive species in response to nutrient, light and water availability (Pattison et al., 1998; Niinemets et al., 2003; Burns and Winn, 2006; Davidson et al., 2011). While adaptive plasticity has been found in invasive plant populations in response to climate (Hahn et al., 2012; Zenni et al., 2014; Bock et al., 2015; Turner et al., 2015), no studies are available that examine the plasticity responses of anatomical traits that contribute to drought tolerance in invasive species.

Selection in the invasive range is thought to allow adaptation to novel environments during establishment and range expansion (Willis et al., 2000; Muller-Scharer et al., 2004; Keller and Taylor, 2008; Prentis et al., 2008; Whitney and Gabler, 2008). Selective pressures can lead to genetic differentiation, displayed as shifts in trait values between individuals from invasive and native ranges that are not a result of plasticity. Post-introduction selection can lead to local adaptation to the new abiotic environment, resulting in genetic differentiation that supports population persistence (Piersma and Drent, 2003; Prentis et al., 2008). Evidence of genetic differentiation has been found in invasive plants that have rapidly shifted growth and reproductive rates subsequent to introduction (Sexton et al., 2002; Siemann and Rogers, 2003; Blair and Wolfe, 2004; DeWalt and Hamrick, 2004; Maron et al., 2004; Brown and Idris, 2005; Marchini et al., 2018). Selection has also been found to increase plasticity in invasive populations in response to light, but not to water availability (Zou et al., 2009), and higher plasticity to water availability after hybridization with native subspecies (Lavergne and Molofsky, 2007). However, selection may not be responsible for all genetic changes following introduction, and divergence may be the result of non-adaptive processes such as genetic drift (Bossdorf et al., 2005; Dlugosch and Parker, 2008; Kilkenny and Galloway, 2013). In a previous study, we demonstrated genetic changes in hydraulic traits associated with drought that were due to selection rather than genetic drift during introduction in the invasive bunchgrass Brachypodium sylvaticum (Hudson) Beauv. (Poaceae; Marchini et al., 2018). In the current study, we examine the roles of plasticity and genetic change in anatomical traits associated with drought tolerance for the success of this invasive grass in Western North America.

Here, we utilized tests for adaptive plasticity – specifically, developmental plasticity that results in non-reversible phenotypes – and genetic differentiation of six morphological traits in two water treatments for B. sylvaticum (slender false brome). Brachypodium sylvaticum is an aggressive invader of the mixed-conifer understorey habitats of the Pacific North-west, USA, where it is capable of forming extensive monocultures. The B. sylvaticum invasion was facilitated by multiple introductions from across its native range, with invasive individuals consisting mainly of hybrids from populations in Western and Southern Europe (Rosenthal et al., 2008). Multiple introductions can provide a larger gene pool for selection to facilitate local adaptation and are associated with invasive success (Riis et al., 2010). In B. sylvaticum, the multiple introductions and admixture have contributed to adaptation during introduction (Marchini et al., 2018), and to some degree of purging of genetic load during range expansion (as measured by reduced inbreeding depression; Marchini et al., 2016).

Brachypodium sylvaticum displays broad climatic tolerances in its native source ranges, occurring in arid, warm temperate and boreal regions. Upon introduction to the invasive range, populations of B. sylvaticum had to acclimatize to the temperate Mediterranean climate of the Pacific North-west USA, tolerating dramatic variations in seasonal rainfall and a long summer drought that contrasts with the relatively consistent precipitation patterns characteristic of the species’ native temperate maritime climatic regions in Western Europe. We hypothesized that the long, dry summers of the invasive range would be challenging for successful reproduction of B. sylvaticum because it flowers and produces seed during late summer when the drought is typically most severe.

Our study addresses the hypothesis that plasticity is a mechanism aiding invasive success of B. sylvaticum. We evaluate the plasticity of plants from native and invasive populations of B. sylvaticum by testing for environmentally induced trait changes, and genotype by environment interactions for drought-associated anatomical and morphological traits in water-available and water-limited treatments. Alternatively, if traits are non-plastic (i.e. they are canalized; Waddington, 1942), then genetic change during introduction rather than plasticity may have contributed to invasion success. We test this hypothesis by comparing the direction of non-plastic trait changes in the introduced range with those predicted for under increased drought. The adaptive potential of plasticity and differentiation was assessed with directional (β) and quadratic (ϒ) selection gradients by regressing trait values and plasticity across environments onto fitness-related traits (shoot biomass and final plant size; Lande and Arnold, 1983). Populations in the invasive range of B. sylvaticum consist of recombinant hybrids resulting from admixture among distinct genetic lineages from native source regions (Rosenthal et al., 2008), so we weighted our models by the probability of contribution to the invasive populations, which we determined from multilocus genotypes at microsatellite markers using the methods of Marchini et al., (2018). If selection in the novel environment were driving invasive success, we would expect to find genetic differentiation between invasive and native populations of B. sylvaticum, instead of observing treatment-driven phenotypic changes. We utilize selection gradients to test for selection on levels of plasticity and genetic differentiation to provide insight into the underlying drivers of adaptation in this invasive species.

MATERIALS AND METHODS

Study system

Brachypodium sylvaticum (slender false brome) is a diploid perennial C₃ bunchgrass native to Europe, Asia and Africa, with invasive populations in North America and Australia. The first herbarium specimen of B. sylvaticum was collected in 1939 in Eugene, OR, USA (Rosenthal et al., 2008). Brachypodium sylvaticum maintained small populations close to the points of introduction near Corvallis and Eugene, OR, until the 1980s, when it began to spread throughout central Oregon. The range is now rapidly expanding, and invasive B. sylvaticum can be found in Washington, California, Virginia, New York, and British Columbia and Ontario, Canada (Miller et al., 2011). Invasive populations of B. sylvaticum show a lot of variability in habitat, and can be found in areas ranging from wetland and riparian habitat to dry forests. Invasive populations form a monoculture carpeting the area and preventing establishment of native plants and tree seedlings. The recent spread of B. sylvaticum, the wide native range, varied habitat and substantial genomic resources for this species and its close relative, B. distachyon (Brkljacic et al., 2011; Fox et al., 2013), make it an ideal model system for study of the evolutionary changes accompanying invasion.

Native B. sylvaticum populations encompass a wide latitudinal distribution over several climatic regions. The hybrid genotypes that are currently spreading throughout Oregon are a combination of populations from oceanic (the UK), hot-summer Mediterranean (Greece) and cold semi-arid desert (Spain) climate types (Rosenthal et al., 2008; Ramakrishnan et al., 2010; Marchini et al., 2018). All of these climates represent a significant shift in precipitation from the cool, dry summers that are characteristic of the temperate Mediterranean climate in Oregon. The Oregon climate provides significantly less precipitation than in the UK, and significantly more during the winter months than in Spain or Greece (Marchini et al., 2018).

Brachypodium sylvaticum seeds were collected from 11 invasive populations located in Oregon, USA and 22 native populations located throughout Europe, North Africa and the Middle East (Supplementary Data Appendix S1). Seeds were obtained from the Western Regional Plant Introduction Station (Western Regional Plant Introduction Station, Pullman, WA, USA) and the Millennium Seed Bank Project (Kew, Surrey, UK). Seeds were planted in the greenhouse at Portland State University, Portland, OR, USA, and plants were maintained for a period of 6 months. Two tillers per plant (ramets of the same genet) were collected for use in the current experiment, with one tiller per individual placed in each of the water-limited and water-available treatments.

Greenhouse methods

We performed a complete block common garden study to assess plasticity and phenotypic differentiation of leaf morphological traits in response to water availability of the environment. Plants were grown in a research greenhouse instead of in the field due to the ethical complications associated with growing the European progenitors in natural environments where they might interbreed with invasive populations. Plants from 33 invasive and native populations of B. sylvaticum were grown in a common garden experiment from August 2012 to December 2012 (n = 5 individuals per population; see Supplementary Data Appendix S1 for population sampling locations). Supplemental lighting and thermostats were set to match field growing conditions in August throughout the experiment. Six study plots were established in 1 m² raised-bed containers in the research greenhouse at Portland State University. Boxes were filled with sandy loam soil (5 % clay, 45 % sand and 50 % silt; Woodco Fuel, Beaverton, OR, USA). The six study plots were given 0.5 L of water after transplanting to acclimate the plants to the new growing conditions. Half of the plots were designated for drought treatment and were minimally watered to maintain drought stress while avoiding mortality. Plants in the drought treatment were checked daily and watered minimally at the first signs of wilting, while the other half of the plots were designated for regular watering (approx. 0.5 L per week).

Five tillers per population were each planted in water-limited and well-watered treatments (with the exception of populations ESH, M3, M7, TUK and UK4, which had four tillers, and M16, which had three tillers) representing different individuals from each population of B. sylvaticum. Tillers were randomly placed in each raised bed at a fixed distance of 10 cm from one another. This spacing was sufficient to maintain separation among individuals, but roots and canopies of plants were beginning to intermingle by the end of the experiment. Water treatment was started after the plants had been allowed to grow and establish for 1 month, and was carried out for approx. 5 months. At the end of the 5 month growing period, average plant height and tiller number data were collected to quantify plant growth. Because B. sylvaticum exhibits growth horizontally and through tiller production, we define plant size as the product of height and tiller number prior to harvest. This measure of plant size is reasonable because it reflects both the sexual and asexual reproductive success of individuals (Younginger et al., 2017). Plants were harvested 5 months after planting, and shoot biomass was obtained by placing plants in convection drying ovens (70 °C) until a constant mass was achieved (minimum of 48 h). We were not able to acquire root biomass as the fibrous roots become tangled among neighbouring plants after long growth periods, making them difficult to separate without significant loss.

Morphological traits measured

We chose six leaf and xylem anatomical traits that are important for drought tolerance to test for genetic differentiation and plasticity in plants from native and invasive populations of B. sylvaticum. We also measured two fitness-related traits, final size and shoot biomass, to investigate the possibility that plasticity and genetic differentiation are adaptive. Rapid growth and large size are thought to increase competitive ability and contribute to invasion success (van Kleunen et al., 2010), and if adaptation to novel environments were occurring we would expect to find high values of these fitness-related traits in invasive individuals.

The traits we chose for analysis are related to water uptake in plants: xylem vessel frequency (VF), maximum xylem vessel diameter (MVD), hydraulically weighted mean xylem vessel diameter (HMD), bulliform cell area (BA), stomatal index (SI) and specific leaf area (SLA). In water-limited conditions, VF is expected to increase while vessel diameter (MVD and HMD) decreases, resulting in xylem vessels that are less prone to cavitation while maintaining water uptake (Tyree and Zimmermann, 2002; Tombesi et al., 2010; Smith et al., 2013). Bulliform cells control leaf rolling, a behaviour that reduces surface area, and consequently, transpiration; a smaller BA allows more rapid loss of turgor, and promotes more rapid rolling of leaves in water-limited environments (Kadioglu et al., 2012). The SI is a normalized measure of stomatal density across a leaf’s surface; many, small stomata are expected to reduce water loss in drought compared with larger, less frequent stomata (Beerling and Chaloner, 1992; Xu and Zhou, 2008). The SLA, the ratio of leaf surface area to mass, is expected to decrease in drought as leaves become thicker to reduce water losses (Larcher, 1995; Poorter et al., 2009). See Supplementary Data Appendix S2 for trait details.

Leaf samples were collected after plants in water-limited and well-watered treatments were well established and had produced new crops of leaves under treatment conditions. Due to differences in the growth rates, leaves were collected in two separate batches: half at the end of November and half at the end of December. During each sampling session, the youngest fully expanded leaf was collected from each individual in each of the three replicates in each treatment (153 samples per treatment, and 306 total, from 11 invasive and 21 native populations as described above), and a range of populations was sampled to avoid bias based on sampling date. One of the two leaves collected from each plant was placed in a plant press until it was dried for later use for the SI and SLA measurements. Approximately 2 cm of tissue was collected from the base of the leaf and placed in 2.0 mL vials filled with a 50 % ethanol–glacial acetic acid (200:5, 17.5 m) mixture to preserve tissue for analysis of metaxylem characteristics (VF, MVD and HWMD). Leaves collected in the greenhouse were kept in water and processed on the same day.

Stomatal and metaxylem counts and measurements were made using a Leica MZ16 stereomicroscope (Leica Microsystems, Wetzlar, Germany) linked to a Q-Imaging Retiga 1300 camera (Q-Imaging, Surrey, British Columbia, Canada) and analysed with Image-Pro 6.0 Software (Media Cybernetics, Warrendale, PA, USA). Leaf epidermal impressions were obtained according to methods in Beerling and Chaloner (1992) by counting stomata and epidermal cells over a standard area (Fig. 3; Supplementary Data Appendix S3). SI was calculated as stomatal density/(stomatal density + epidermal cell density) × 100, where stomata consist of the stomatal pore and two flanking guard cells. SI normalizes leaf stomatal density in relation to the confounding influence of epidermal cell expansion that may be initiated by factors other than those influencing stomatal development (Royer, 2001; Xu and Zhou, 2008).

Fig. 3.

Examples of Brachypodium sylvaticum leaves from invasive (Corvallis, OR, USA, left) and native (Larissa, Greece, right) ranges. (A) Cross-sections of leaves. Invasive plants have smaller bulliform cells and fewer metaxylem vessels. (B) Surface imprints of the leaf epidermis. Invasive individuals have lower stomatal density.

Cross-sections of preserved leaf samples were photographed and measured for metaxylem and bulliform cell characteristics on at least three individuals per population (Supplementary Data Appendix S4). Images were analysed with ImageJ (National Institutes of Health, Bethesda, MD, USA). Measurements on each section included cross-sectional area, area of the major veins and proportion of bulliform cells and vascular tissue to cross-sectional area (Fig. 3). Multiple measurements for each leaf were averaged to obtain HMD, VF, MVD and BA for each plant. To assess the hydraulic effects of metaxylem diameter, the principles of the Hagen–Poiseuille relationship, which state that a conduit’s hydraulic conductivity is proportional to its diameter to the fourth power, were taken into account. Thus, to find HMD, raw measurements of diameter were analysed for their frequency in 1 μm size classes and for the relative contribution of each class of diameter to the sum of all the conduits raised to the fourth power calculated as:

$$\text{HMD}=2\frac{\Sigma r^5}{\Sigma r^4},$$

where r is the radius of a xylem conduit. This equation weights the importance of radii in proportion to the estimated hydraulic conductance of the xylem conduits (Sperry and Saliendra, 1994). VF was calculated as the average number of metaxylem vessels per leaf area. MVD is the largest metaxylem vessel per individual. The cross-sectional area of bulliform cells, large leaf surface cells that are the first to lose turgor in times of water stress, resulting in leaves rolling in to prevent transpirational water loss, was found for each individual. BA was calculated as the proportion of bulliform cells present per cross-sectional area.

To obtain SLA, the surface area for each plant was measured by first photographing the top half of the most recently fully expanded leaf using a Kaiser RA-1 copy stand (Buchen, Germany) with a Panasonic WV-BD400 camera (Newark, NJ, USA). The surface area of this portion of the leaf was estimated using the drawing tool in Image J. The top half of each leaf was gathered and dried in an oven for at least 48 h at 60 °C to constant weight. SLA was calculated as leaf surface area in m²/leaf dry weight in kg for each plant.

Statistical analysis

Weights of genetic contribution.

Prior to performing statistical analysis, we weighted traits of native individuals by probability of contribution to invasive populations using the methods described by Marchini et al. (2018). Assignment tests of the probability of the contribution of native to invasive populations were performed using the genotypes of B. sylvaticum plants at eight microsatellite loci with primers developed and optimized at Portland State University (Ramakrishnan et al., 2008). Primers are deposited in GenBank (accession nos: EF450748, EF450751, EF450752, EF450754, EF450756, EF450757, EF450759 and EF450765). Assignment tests were performed in Geneclass2 utilizing the Rannala and Mountain (1997) assignment method of identifying immigrants based on their multilocus genotypes to detect immigrant ancestry (Rannala and Mountain, 1997; Piry et al., 2004). The assignment probabilities were pooled across invasive populations, to create an average probability of genetic contribution of each native population to the invasive range. These weights based on assignment probabilities provide a more precise comparison to test for changes in phenotypic traits after introduction because they exclude the majority of native populations in Europe and Asia that did not contribute to the invasion (Marchini et al., 2018). These weights were used for comparison of traits between native source and invasive populations in the analyses described in the following sections.

Phenotypic plasticity and genetic differentiation.

We used restricted maximum likelihood models with random effects to determine plasticity (mean of the water-available treatment minus mean of the water-limited treatment) and divergence for each trait (trait) across water-limited and water-available treatments (lmer and lmerTest, lme4 package; Bates et al., 2015). Mixed models were applied for testing for differences for each measured trait between the invasive and native ranges (range) and plasticity (treatment). Values for individual plants in the native range were weighted (weight) by the source population’s probability of contribution to the invasion. Measurements on invasive plants were all weighted with a value of 1.0. Water treatment, range (invasive or native), environment (PC1; environment), the interaction environment × range and the interaction treatment × range were treated as fixed factors in the model; raised bed (plot) nested within treatment was a random effect [model: trait × weight ~ treatment × range × environment + (1|plot/treatment)]. Additional interactions explained minimal levels of variation and so were dropped from the model. A significant effect of range in these models would indicate differentiation in the given trait between plants from invasive and native ranges; a significant effect of treatment would indicate plasticity for the given trait. All response variables were logx + 1 transformed to meet assumptions of normality for regression.

To account for phenotypic variation based on climate of the source population, each model included a composite abiotic environmental covariate determined by a principal component analysis (PCA) of latitude and six climatic variables (aridity, potential evapotranspiration, temperature, evaporation, precipitation and cloud cover). Climatic data were obtained from the Prism Group, Oregon State University (PRISM Climate Group, 2015) and United States Geological Climate Survey climate data (Hearn et al., 2003) for the Oregon populations, and from WorldClim (Hijmans et al., 2005) for accessions from the native range. The first axis from the PCA (PC1) explained most variation across regions, and was utilized in all trait models as the environment term.

Plasticity and selection gradients.

For all selection analyses, we utilized standardized explanatory values (mean = 0, s.d. = 1) to allow comparison between regression coefficients of traits measured on different scales (van Kleunen and Fischer, 2001). The relative fitness of shoot biomass and final plant size was calculated for response variables. Response variables were logx + 1 transformed and also standardized to mean = 0, s.d. = 1. All models were weighted by the probability of contribution of native to invasive ranges.

Selection on plasticity for traits across water-available and water-limited treatments was tested using linear regression (Weis and Gorman, 1990). The standardized values of absolute plasticity (differences in trait values for ramets from the same genet between treatments) and standardized mean trait values for each ramet in each treatment were explanatory variables in the model. Response variables were the fitness-related traits of shoot biomass and final plant size. A positive effect of phenotypic plasticity on a fitness-related trait is indicated by a positive regression coefficient for a trait, while a negative selection coefficient suggests selection against plasticity (van Kleunen and Fischer, 2001).

We estimated linear (β) and univariate non-linear (quadratic, ϒ) selection for morphological traits in water-limited and water-available environments. Selection models were weighted by the probability of genetic contribution of native to invasive populations. Directional selection gradients were estimated using linear models regressing standardized trait values onto the standardized relative fitness values for shoot biomass and final size (lmer; Bates et al., 2015). The linear selection coefficient β is found as the partial regression coefficient for each morphological trait. Significance of linear selection coefficients is indicative of directional selection for higher (positive β) or lower (negative β) phenotypic values that effect fitness-related traits. Non-linear models included all linear and quadratic values of standardized traits and their interaction terms. The partial regression coefficient of the quadratic terms was the quadratic selection coefficient. Quadratic selection gradients were doubled (Stinchcombe et al., 2008). Stabilizing selection is revealed by a negative quadratic term, while disruptive selection is indicated by a positive quadratic term.

RESULTS

Principal component analysis of climate

Results from the PCA of population climatic characteristics show that PC1 and PC2 explained 50 % and 19 % of the variance across populations, respectively. Axis 1 was correlated with aridity and latitude (a smaller value for aridity indicates a drier environment); Axis 2 was correlated with surface evaporation and precipitation. Although there is overlap in the ellipses of climatic variation between native and invasive source ranges, regions within the invasive range are more strongly influenced by aridity (PC1; Fig. 1).

Fig. 1.

Principal component analysis biplot of variables associated with climate in populations of Brachypodium sylvaticum. Points represent the range of native (N) and invasive (I) populations.

Plasticity and genetic differentiation

Probabilities of contribution used in analyses are found in Supplementary Data Appendix S5. There was no effect of treatment or a treatment × range interaction for any measured traits, indicating a lack of plasticity. There was significant genetic differentiation between ranges of three traits related to xylem morphology, HMD, MVD and VF, and one trait related to cell size, BA (Fig.1, Table 1). MVD and HMD were significantly lower in invasive populations than in native populations by nearly 20 % for both traits, while VF was higher by 20 % in invasive populations than in native populations, indicating that plants in invasive populations possess smaller, more frequent xylem vessels than plants from native populations. BA was significantly lower by around 10 % in the invasive compared with the native populations. The expression of these four traits, VF, HMD, MVD and BA, in individuals from the invasive range conformed to hypotheses of optimal trait values expected in water-limited environments (Figs 2 and 3).

Table 1.

Probabilities for F-ratios from mixed-model analysis of variance showing the effects of range (invasive vs. native), abiotic environment (principle components axis 1 for climate data) and treatment (well-watered vs. water-limited) on six measured traits in Brachypodium sylvaticum

			VF	HMD	MVD	BA	SLA	SI
Effect	Source	d.f.	P	P	P	P	P	P
Fixed	Treatment	1	0.75	0.92	0.93	0.68	0.83	0.81
	Environment	1	0.05	0.04*	0.04*	<0.01**	0.19	0.26
	Range	1	0.04*	0.02*	0.07*	<0.01**	0.42	0.93
	Environment × range	1	0.50	0.41	0.53	0.20	0.30	<0.01*
	Treatment × range	1	0.48	0.46	0.49	0.30	0.26	0.48
			Variance	Variance	Variance	Variance	Variance	Variance
Random	Raised bed (treatment)		0.32	0.003	0.001	0	27.41	2.57

Asterisks and bold face indicate statistical significance of regression coefficients: *P < 0.05.

Fig. 2.

Trait means for (A) VF, (B) MVD, (C) HMD, (D) SI, (E) SLA and (F) BA in individuals from the invasive and native range of Brachypodium sylvaticum. Arrows indicate the direction of trait shifts under drought stress. Data are weighted by the probability of genetic contribution found by multilocus comparisons. Asterisks indicate a statistical difference of trait values between ranges: *P < 0.05. Traits are defined in Supplementary Data Appendix S2.

Environment (based on PC1) was associated with values of HMD, MVD and BA (Table 1; Fig. 4A–C); however, the environment × range interaction term was not significant for these variables. The environment × range interaction was significant for SI only (Table 1). Opposite trends for SI and PC1 are seen in invasive and native environments (Fig. 4D).

Fig. 4.

Environmental gradients and phenotypic expression (population means) for four morphological traits of Brachypodium sylvaticum: (A) maximum vessel density (MVD); (B) hydraulically weighted mean diameter (HMD); (C) bulliform cell area (BA); and (D) an interaction between range and abiotic environment for stomatal index. PC1 is strongly correlated with aridity. Error bars are ± s.e.

Plasticity, and linear and quadratic selection gradients

The majority of linear selection coefficients for plasticity were negative (Table 3). There was a significant negative selection coefficient for plasticity in VF on shoot biomass (Table 2), indicating that greater plasticity in VF across water-limited and water-available treatments is associated with a lower mean shoot biomass (Fig. 5). No other traits had significant selection coefficients of plasticity.

Table 3.

Directional (β) and quadratic (ϒ) selection gradients on final size and shoot biomass of invasive and native populations of Brachypodium sylvaticum

	Water-available		Water-limited
Value	β	ϒ	β	ϒ
Final size
VF	–0.2078	0.427	–1.215**	0.396
MVD	–2.8161	–3.798	3.116*	–3.1
HMD	3.0408	3.897	–3.175*	2.614
SI	0.3524	0.389	–0.466	–0.468
SLA	–0.9065	3.743	0.166	0.054
BA	–0.3966	–3.039	–0.436*	–0.076
Shoot biomass
VF	–0.4853	0.794**	–1.076**	0.268
MVD	–2.1455	–2.22	4.031**	1.178
HMD	1.8938	2.884	–4.326**	–1.038
SI	0.1366	–0.052	–0.133	–0.036
SLA	–0.1358	1.564	0.331	0.292
BA	–0.2212	2.782**	–0.331*	–0.252

The table lists standardized regression coefficients of plasticity for final size and shoot biomass.

Asterisks indicate statistical significance of regression coefficients: *P < 0.05, **P < 0.01.

The quadratic gradients are twice the estimated partial regression coefficients.

Traits are defined in Supplementary Data Appendix S2.

Table 2.

Directional (β) selection gradients of plasticity in six anatomical traits associated with drought tolerance on final size and shoot biomass of invasive and native populations of Brachypodium sylvaticum

Trait	Final size	Shoot biomass
VF	–0.359	–0.789*
MVD	–4.078	–2.997
HMD	3.854	2.255
SI	0.017	0.211
SLA	–0.105	0.083
BA	–0.607	–0.594

Asterisks and bold face indicate statistical significance of regression coefficients: *P < 0.05.

Traits are defined in Supplementary Data Appendix S2.

Fig. 5.

Significant negative directional selection (β) for vessel frequency (VF) plasticity in response to water availability in Brachypodium sylvaticum. Plasticity is measured as mean values of the water-available treatment minus the water-limited treatment for each population.

Four out of six traits had significant directional selection gradients for shoot biomass and final size in the water-limited treatments (Table 3). Gradients ran in the same direction for shoot biomass and final size in response to all traits, and were consistent for plants from the native and invasive ranges. A negative selection coefficient for VF describes reduced above-ground size with an increasing frequency of xylem vessels. The selection gradient of MVD is positive, indicating that possession of at least one large xylem vessel results in increased size and biomass, although the negative coefficient of HMD indicates selection towards a smaller size of xylem vessels. Negative selection gradients for BA imply negative selection for increasing bulliform cell area in water-limited environments.

In the water-available treatment, the majority of linear selection coefficients were negative, although none was statistically significant (Table 3). There was a significant positive quadratic term for VF (Fig. 6A) and for BA (Fig. 6B). Inspection of these curves show that the quadratic functions are mostly linear.

Fig. 6.

Quadratic selection (ϒ) for (A) vessel frequency (VF) and (B) bulliform area (BA) in the water-available treatment for Brachypodium sylvaticum. Selection curves are significant at P < 0.05.

Genetic differentiation and direction of selection

Of the four traits found to be significantly different across invasive and native ranges of B. sylvaticum, the direction of selection for greater relative fitness was only consistent with the direction of phenotypic differentiation occurring in the invasive range for two traits in the water-limited treatment. One of these traits is related to xylem morphology (HMD), and one of these traits is related to cell morphology (BA). Both HMD and BA were associated with negative selection gradients, consistent with the smaller values of HMD and BA found for individuals from invasive populations (Table 3; Fig. 2).

Two of the four traits that significantly differed across invasive and native ranges of B. sylvaticum displayed trait values inconsistent with the direction of selection for relative fitness of shoot biomass and final size in the water-limited treatment. These traits, VF and MVD, are both related to xylem vessel morphology. A negative selection coefficient implies greater relative fitness with decreasing frequency of xylem vessels, in contrast to the greater VF found in plants from invasive populations. The positive selection coefficient for MVD suggests greater relative fitness with a larger maximum diameter of the xylem vessels, while plants from invasive populations possess a smaller MVD than their native counterparts.

DISCUSSION

Our results indicate that the success of invasive Brachypodium sylvaticum in Oregon is the result of post-introduction selection leading to genetic differentiation. This conclusion contrasts with the hypothesis that plasticity has allowed invasive individuals the flexibility to establish in multiple habitats. Though populations exhibit no plasticity in the measured traits, invasive phenotypes conform to assumptions of expected divergence that would occur under water limitation.

Plasticity

There were no significant differences in the measured trait responses between the water-available and water-limited treatments, indicating a lack of plasticity in response to contrasting water environments for plants from both invasive and native ranges of B. sylvaticum. These results are in contrast to findings that plasticity is high in populations of invasive species (Davidson et al., 2011), and in line with studies concluding that there are no differences in plasticity between invasive and native populations (Godoy et al., 2011; Palacio-Lopez and Gianoli, 2011; Murphy et al., 2016). Moreover, selection gradients for above-ground biomass were negatively correlated with plasticity for the frequency of xylem vessels in leaves (VF), indicating possible selection against plasticity for this trait. This result suggests that there were some differences in plasticity among genets, which is perhaps due to the recent origin of the invasive populations from admixture, which may have elevated levels of plasticity (Lande, 2015). Overall, it appears that morphological traits associated with drought tolerance in native and invasive populations of B. sylvaticum are not plastic (i.e. they are canalized; Waddington, 1942).

Positive quadratic selection gradients were found for both VF and BA in the water-available environment. The quadratic relationship signifies that increasing values for both traits (greater frequency of xylem vessels; a larger area of bulliform cells within the leaf) is initially detrimental to above-ground size. However, if values of VF and BA rise above a certain threshold, fitness as measured by above-ground biomass increases. It is not clear what is contributing to this apparent pattern of disruptive selection for VF and BA, but it may be related to different strategies associated with either low or high root/shoot ratios under drought conditions. This observed variation may also be due to the recent origin of invasive populations as a result of admixture (Rosenthal et al., 2008).

The lack of plasticity for traits associated with water availability in B. sylvaticum is striking, and perhaps unexpected given the prevalence of plasticity in plants (Sultan, 2000; Pigliucci, 2001). The treatments chosen represent extremes in water availability, and were maintained for the duration of our experiment, so conditions should have been optimal to induce plasticity. Consistent expression across treatments in all of the measured traits, and evidence for selection to decrease plasticity in one trait, suggests that canalization has been favoured in both the native and invaded environments. Canalization of functional traits of plants may be a consequence of phenology and leaf longevity. Plants of B. sylvaticum are dormant over the winter and establish a new crop of leaves during the late winter and early spring. Environmental conditions in the Pacific North-west USA are wet and cool during leaf development, and these same leaves are retained over the entire growing season into the hot, dry summer. This growth pattern is in contrast to other plants that exhibit indeterminate growth and develop new leaves throughout the growing season that respond plastically to seasonal conditions (e.g. Picotte et al., 2007). In B. sylvaticum, selection appears to have favoured trait canalization towards leaf morphologies that are ideal for the hot and dry late summer months when plants are flowering and producing seeds.

A hypothesis proposed by Lande (2015) provides one explanation for the inconsistent findings for the role of plasticity in invasion success. Lande describes how introduction to a novel habitat can spur on a rapid increase in plasticity during the early phase of establishment as a consequence of interactions among genetic elements derived from divergent lineages. This peak in phenotypic response to the environment results in novel phenotypes that are then fixed by genetic assimilation (increased canalization) of the new phenotypes in invasive populations (Lande, 2015). Plasticity is reduced as adjustment to the novel habitat proceeds because selection favours a constant phenotype in response to environmental perturbations. An alternative explanation for the pathway to trait divergence in the invaded range is that the variation necessary for adaptive responses in the novel environment was generated by admixture due to interbreeding among genotypes from multiple sources in the native range. Previous research which found phenotypic differentiation during the establishment phase of B. sylvaticum supports this hypothesis (Marchini et al., 2018). Under this scenario, selection favoured trait combinations in recombinant hybrids that conferred the highest fitness under conditions of extensive summer drought. Whether admixture during invasion generates greater plasticity that facilitates genetic assimilation to new adaptive norms remains an open question.

Phenotypic differentiation and selection

Differences in the direction of selection gradients in the water-available and water-limited environments, combined with evidence that environmental variation at population locations influences trait values, suggests that moisture availability is a selective pressure driving genetic differentiation in the invasive range of B. sylvaticum. The direction of genetic divergence predicted to occur under water limitation was supported by the observed changes in trait values in the invasive range; however, examination of the direction of selection gradients for above-ground biomass and size measured in the water-limited treatment was not in the same direction as observed changes in the invaded range for all traits.

Two traits, HMD and BA, displayed genetic divergence in the invasive range consistent with the direction of selection for increased shoot biomass and final size in the water-limited treatment. Selection for reduced bulliform cell area in the invasive range translates into more rapid leaf rolling in response to moisture stress. Leaf rolling decreases leaf surface area, reducing transpirational losses, and is a common response to low water availability in drought-tolerant species (O’Toole and Cruz, 1980; Turner, 1986; Kadioglu et al., 2012). Reduced xylem diameter decreases susceptibility to cavitation, which reduces the transport of water to leaves in plants under severe drought stress (Hacke and Sperry, 2001; Tyree and Zimmermann, 2002; Tombesi et al., 2010). Genetic differentiation of these traits in the invasive range has probably aided population persistence of B. sylvaticum during summer droughts.

We found seemingly maladaptive selection gradients in the water-limited treatment for two traits related to xylem morphology (VF and MVD) that were opposite to the direction of trait divergence in the invasive range of B. sylvaticum. The apparent inconsistency in selection gradients for these two traits could be due to trade-offs with other traits. Frequency of xylem vessels is related to xylem diameter; as diameters decrease and the threat of cavitation is reduced, the frequency of vessels throughout the stem should increase so that water uptake can be maintained (Tyree and Zimmermann, 2002; Smith et al., 2013). Similar to expectations of HMD in drought, MVD is predicted to decrease, lowering the probability of cavitation (Tyree and Zimmermann, 2002). In contrast to predictions, we found that the direction of selection for increased above-ground size and biomass in our water-limited treatment appears to favour less frequent xylem and larger maximum xylem diameter. Additionally, although not statistically significant, the change in SI (stomatal index) in populations from invasive regions was in contrast to the increase in stomatal density that would be expected under drought conditions. This may be because leaves of B. sylvaticum develop during the spring of the Pacific North-west, and stoma are formed when the weather is cool and precipitation is heavy. It could also be due to the fact that SI has less influence on water loss in drought, but this hypothesis requires further testing.

Previous studies examining trade-offs between growth and resource allocation and tolerance to environmental stress may provide insight into the apparent contradictory patterns of trait divergence. Grime (1977) suggested that plants cannot be both highly competitive and highly tolerant to environmental stress. The existence of these trade-offs has been supported by recent studies finding that increased tolerance to abiotic stress often comes at a cost of reduced competitive ability (Petru et al., 2006; Sambatti and Rice, 2007; Liancourt and Tielborger, 2009; Hodgins and Rieseberg, 2011). Previous experiments have shown that B. sylvaticum can be a strong competitor, but only under high nutrient conditions (Holmes et al., 2010; G. L. Marchini, unpubl. data). A more stressful abiotic environment in the invasive range may be driving selection for increased drought tolerance, while limiting above-ground growth potential and competitive ability.

A second explanation for the observed inconsistency between the direction of selection and divergence is the possibility that fitness and competitive ability of B. sylvaticum in the water-limited treatment is not driven by above-ground size and biomass. Though there are few studies finding that increased above-ground size is not a trait contributing to invasiveness (Willis et al., 2000; Thebaud and Simberloff, 2001), there are an abundance of studies finding that selection in water-limited environments often results in increased below-ground biomass (Chaves et al., 2003; Farooq et al., 2009). For example, a study of invasive and non-invasive woody species grown in a summer drought found that invasive species had much larger root masses than non-invasive species (Grotkopp and Rejmanek, 2007). In our study, plants in the water-limited treatment were closely spaced and their roots became entangled during the course of the growing period. In this treatment, water became available in pulses, so it is possible that the advantage of maintaining several large xylem vessels, permitting rapid water uptake, outweighed cavitation risks and led to greater above-ground size and biomass. Based on the scope of the current study, selection on B. sylvaticum individuals in water-limited environments may result in larger vessels and greater resource allocation below-ground, which would maintain the capacity for rapid uptake when water became available. The same conditions may not occur in natural environments because soil depth is not limited and summer precipitation is light so soils do not become saturated with water until late autumn and early winter.

Effect of environment

An abiotic environment was correlated with two traits related to xylem vessel size (MVD and HMD) and one trait related to cell morphology (BA), and there was a significant environment × range interaction for SLA. The sensitivity of these traits to environmental gradients suggests environmental specialization (Richards et al., 2006). Local adaptation to environmental variability can shape genetic divergence between ranges and lead to rapid evolution of introduced species (Callaway and Maron, 2006; Colautti et al., 2010).

Invasive populations of B. sylvaticum are composed of intraspecific recombinant hybrid genotypes as a result of multiple introductions (Rosenthal et al., 2008). Invasive hybridization may be a factor promoting local adaptation of B. sylvaticum to the abiotic environment of the invaded region. Evidence of hybridization is commonly found in populations of invasive species (Ellstrand and Schierenbeck, 2000; Gaskin and Schaal, 2002; Zalapa et al., 2010; Kolbe et al., 2012). In B. sylvaticum, hybridization and multiple introductions have reduced genetic bottlenecks at sites of primary introduction (Rosenthal et al., 2008). These primary introduction sites now act as sources for gene flow to satellite populations (Ramakrishnan et al., 2010), and probably provide the genetic variation necessary for strong selection to evolve phenotypes that are adapted to the local climate in the invasive range.

Conclusions

We have presented evidence that genetic differentiation and adaptation are drivers of invasive success in Brachypodium sylvaticum. Individuals from the invasive and native ranges of B. sylvaticum were not phenotypically plastic in environments of contrasting water availability, which is consistent with other studies concluding that plasticity is not a trait inherent to invasive species. We found genetic divergence between invasive and native ranges for four out of six traits measured. Differentiation of these traits was in the direction hypothesized to occur in response to drought stress in the introduced range. Divergence of invasive plant traits was not consistently in the direction indicated by selection gradients, implying that there are limitations of selection that may be a consequence of physical constraints and/or trade-offs between growth and abiotic tolerance. Significant relationships between genetic divergence and environment suggest local adaptation of B. sylvaticum. Our results highlight how genetic variation generated through admixture resulted in adaptation of introduced B. sylvaticum populations to new, challenging environments. The high levels of canalization observed in this study highlight the hypothesis that increased plasticity is not always adaptive, and contribute to a growing literature on the roles of plasticity and genetic changes during introduction for the success of invasive species.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Appendix S1: location information for sampled native and invasive populations of Brachypodium sylvaticum. Appendix S2: functional traits measured in Brachypodium sylvaticum and their units; significance and expected behaviour in drought. Appendix S3: stomatal density measures. Appendix S4: leaf metaxylem and bulliform cell measurements. Appendix S5: probabilities of genetic contribution from multilocus genotypes.