Environmental contributions to the evolution of trait differences in Geum triflorum: implications for restoration

Abstract

Premise of the Study
Understanding how environment influences the distribution of trait variation across a species’ range has important implications for seed transfer during restoration. Heritable genetic differences associated with environment could impact fitness when transferred into new environments. Here, we test the degree to which the environment shapes the evolution and distribution of genetic effects for traits important to adaptation.

Methods
In a common garden experiment, we quantified trait differentiation for populations of Geum triflorum sourced from three distinct ecoregions and evaluated the ability of climate to predict trait variation. Populations were sourced from alvar ecoregions which experience predictable extremes in seasonal water availability and the prairie ecoregion which exhibits unpredictable changes in water availability.

Key Results
Plants sourced from alvar ecoregions exhibited smaller but more numerous stomata and greater intrinsic water use efficiency relative to prairie plant populations supporting the evolution of ecotypic differences. Estimates of standing genetic variance and heritable genetic variation for quantitative traits suggest alvar populations have greater adaptive potential. However, reduced evolvability suggest all populations of G. triflorum may have limited capacity to evolve in response to environmental change.

Conclusions
These results point towards the importance of understanding the role of environment in shaping the distribution and evolution of genetic differences across seed populations and how these data may inform recommendations for seed transfer across novel environments and our expectations of populations’ adaptive potential.

Full text

1
Title: 1
Environmental contributions to the evolution of trait differences in Geum triflorum: implications 2
for restoration 3
Authors: Kate Volk1, Joseph Braasch1,2, Marissa Ahlering3 & Jill A. Hamilton1,4 4
Affiliations: 5
1 North Dakota State University; Department of Biological Sciences, Fargo, ND 58102, USA 6
2 Rutgers University Camden; Department of Biological Sciences, Camden, NJ 08102, USA 7
3 The Nature Conservancy; Moorhead, MN 56560, USA 8
4 Pennsylvania State University; Department of Ecosystem Science and Management, University 9
Park, PA, USA 10
11
12
Manuscript received _____; revision accepted ____. 13
14
Short title: 15 Environment predicts trait differences in Geum triflorum 16
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

2
ABSTRACT 17
Premise of the Study 18
Understanding how environment influences the distribution of trait variation across a species’ 19
range has important implications for seed transfer during restoration. Heritable genetic 20
differences associated with environment could impact fitness when transferred into new 21
environments. Here, we test the degree to which the environment shapes the evolution and 22
distribution of genetic effects for traits important to adaptation. 23
Methods 24
In a common garden experiment, we quantified trait differentiation for populations of Geum 25
triflorum sourced from three distinct ecoregions and evaluated the ability of climate to predict 26
trait variation. Populations were sourced from alvar ecoregions which experience predictable 27
extremes in seasonal water availability and the prairie ecoregion which exhibits unpredictable 28
changes in water availability. 29
Key Results 30
Plants sourced from alvar ecoregions exhibited smaller but more numerous stomata and greater 31
intrinsic water use efficiency relative to prairie plant populations supporting the evolution of 32
ecotypic differences. Estimates of standing genetic variance and heritable genetic variation for 33
quantitative traits suggest alvar populations have greater adaptive potential. However, reduced 34
evolvability suggest all populations of G. triflorum may have limited capacity to evolve in 35
response to environmental change. 36
Conclusions 37
These results point towards the importance of understanding the role of environment in shaping 38
the distribution and evolution of genetic differences across seed populations and how these data 39
may inform recommendations for seed transfer across novel environments and our expectations 40
of populations’ adaptive potential. 41
42
Key Words 43
Heritability, evolvability, ecotypic variation, water-use efficiency, grassland restoration, 44
common garden, stomata, alvar, prairie 45
46
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

3
INTRODUCTION 47
Understanding how the environment influences trait variation is essential, particularly 48
within the context of restoration (Wang et al., 2010). The evolution of ecotypic differences for 49
vegetative, physiological, or reproductive life history traits can lead to differential success 50
following seed transfer across environments during restoration (McKay et al., 2005; Anderson et 51
al., 2016; Braasch et al., 2021; VanWallendael, Lowry & Hamilton, 2022). In addition, the 52
history of selection may influence the distribution of genetic variance of traits important to 53
adaptation. Variance in the heritability or evolvability of traits is expected to impact the success 54
of ecotypes when planted in novel restored environments (Broadhurst et al., 2008; Crowe & 55
Parker, 2008; Havens et al., 2015). Thus, quantifying how environment contributes to the 56
evolution of trait differences and the distribution of genetic variance provides important insight 57
into contemporary adaptation and future adaptive capacity (Broadhurst et al., 2008; Bucharova et 58
al., 2019; Hamilton et al., 2020; Kulbaba et al., 2021). This is particularly important to 59
restoration, which aims to establish populations resilient to change. 60
Trait differences arise through a combination of deterministic and stochastic processes 61
(Kawecki and Ebert, 2004; Crow et al 2018; Galliart et al., 2018). For example, climatic 62
gradients have contributed to ecotypic differentiation among grass species for morphological 63
(Aspinwall et al., 2013; Olsen et al., 2013), phenological (Lowry et al., 2019), physiological 64
(Aspinwall et al., 2013; Maricle et al., 2017), and fitness traits (McMilan,1959; Galliart et al., 65
2018). To establish seed transfer recommendations during restoration, teasing apart the 66
contributions of environment to the evolution of trait differences may be useful to predicting 67
populations’ response to new environments. In this study, we focus on the evolution of 68
physiological traits, which may evolve in response to varying water availability (Dudley et al., 69
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

4
1996; Picotte et al., 2007; Dittberner et al., 2019). Range wide variation in Arabidopsis thaliana 70
for stomatal characteristics suggests that climatic factors have led to the evolution of changes in 71
stomatal size and density (Dittberner et al., 2019). With smaller, but more numerous stomata, 72
plants have a greater ability to respond rapidly to changing water availability associated with 73
increased temperatures (Drake et al., 2013; Dittberner et al., 2019). Thus, variation in 74
physiological traits may correspond with the evolution of ecotypes associated with environments 75
across a species’ range. 76
The history of selection, particularly the degree to which environmental heterogeneity has 77
been predictable or unpredictable across a species’ range, may impact the distribution of genetic 78
variation underlying traits and consequently their capacity to adapt. Here, we define 79
environmental predictability as repeatable seasonal cues associated with a given climate variable 80
(Reed et al., 2010). Theory suggests that where populations have experienced predictable 81
environmental cues, heritable genetic variance for phenotypic traits will increase as the total 82
phenotypic variance is reduced (Fig. 1; Levins, 1963; Reed et al., 2010; Baythavong, 2011; 83
Kulbaba et al., 2021). In such a scenario, heritable trait differences among ecotypes may lead to 84
increased risk of maladaptation when seed is transferred to new environments (Reed et al., 2010; 85
March-Salas et al., 2019). In contrast, populations sourced from unpredictable environments are 86
expected to exhibit greater plasticity and reduced trait heritability (Fig 1; Chevin et al., 2010; 87
Reed et al., 2010; Ghalambor et al., 2007; Baythavong, 2011; March-Salas et al., 2019). If 88
adaptive, plasticity enables plants to modify their phenotype in response to the changed 89
environment to maintain fitness (Reed et al., 2010; Baythavong, 2011; Becklin et al., 2016; 90
March-Salas et al., 2019). If plasticity is non-adaptive it may come with a fitness cost (Gilbert et 91
al., 2019). Evolvability, which is the expected change in a trait per generation for a given 92
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

5
selection coefficient (Hansen and Houle, 2008; Hansen et al., 2011), quantifies how rapidly 93
adaptation is predicted in a continuously shifting environment (Shaw and Etterson, 2001; 94
Kulbaba et al., 2021). Therefore, quantifying trait heritability and evolvability for seeds sourced 95
from predictable and unpredictable environments provides complimentary metrics to predict 96
populations’ capacity to respond to changing selective pressures. These metrics can be used to 97
guide seed transfer recommendations and aid in determining both the initial risk of transfer 98
across environments and the likelihood populations will adapt once established. 99
Geum triflorum Pursch., is an early season perennial forb associated with remnant prairie 100
habitat across much of the Great Plains of North America (Hamilton and Eckert 2007, Yoko et 101
al., 2020). The Great Plains are critically imperiled due to habitat loss associated with land 102
conversion, fragmentation, and urban expansion and thus are important habitats for restoration 103
efforts (Hoekstra et al., 2005; Gascoigne et al., 2011; Comer et al., 2018; Wimberly et al., 2018; 104
Bengtsson et al., 2019). Populations of G. triflorum also persist as isolated ‘islands’ across alvar 105
habitats scattered throughout the Great Lakes and into Manitoba, Canada (Hamilton and Eckert 106
2007; Yoko et al., 2020). Alvars are habitats characterized by a thin layer of soil over limestone 107
bedrock that harbor a unique assemblage of plants largely disjunct from the core of their 108
distribution (Hamilton and Eckert 2007). Alvars experience extreme, but predictable annual 109
fluctuation in water availability from flooding in the spring to early summer desiccation (Catling 110
and Brownell 1995; Hamilton et al., 2002; Yoko et al., 2020). In contrast, while prairies 111
experience flooding and drought, compared to the predictable interannual extremes of the alvar 112
ecoregion, the onset of these events is less predictable. In addition, the deep, organically rich soil 113
characterizing prairie ecoregions provides a buffer to extreme water fluctuations (Anderson, 114
2006). Thus, we suggest the alvar ecoregion reflects a ‘predictable’ history of selection, whereas 115
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

6
the prairie ecoregion reflects an ’unpredictable’ history of selection in response to changing 116
water availability. These ecoregions provide an ideal system to evaluate the role predictability of 117
the environment may play in influencing the amount and distribution of genetic variance for 118
phenotypic traits. Physiological traits, including stomatal size and density along with water-use 119
efficiency (WUE) are expected to vary between prairie and alvar ecoregions. Given the 120
importance of stomatal traits and WUE to plant persistence, examining how environment of 121
origin has influenced variation in these traits will inform seed transfer recommendations. 122
Using a common garden experiment of maternal seed families for G. triflorum sourced 123
from both prairie and alvar ecoregions, we evaluated the role source environment has had on the 124
distribution of physiological trait variation linked to plant water use. We quantified ecoregional 125
differentiation for each trait and tested for correlations between functional traits and climate of 126
origin for all sampled populations. Lastly, we quantified standing genetic variance for stomatal 127
traits, including estimates of heritability and evolvability. Specifically, we ask 1) do 128
physiological traits exhibit ecoregional differences, 2) is there a relationship between 129
physiological trait variation and source climate, and 3) does the history of selection associated 130
with seed source environment structure the distribution of additive genetic variance and the 131
heritability or evolvability of physiological traits? We predict alvar ecoregions will exhibit 132
smaller, but more numerous stomata in addition to greater WUE relative to prairie populations. 133
We also expect populations from the alvar ecoregions will have greater heritability for 134
quantitative traits relative to prairie populations due to the history of selection associated with 135
predictable changes in water availability. An understanding of how differentiation in 136
physiological traits evolve and the role selection may play in shaping the distribution of heritable 137
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

7
trait variation will be valuable for predicting the response of seeds to restored environments and 138
estimating their longer-term evolutionary potential. 139
140
MATERIALS AND METHODS 141
Field sampling – In the spring of 2015, seed from 19 populations of Geum triflorum Pursch. 142
(Rosaceae) spanning much of its distribution were collected, including 11 populations from the 143
Great Lake alvar ecoregion (GLA), two from the Manitoba alvar ecoregion (MBA), and six from 144
the midwestern prairie ecoregion (PRA, Fig. 2). Forty individual seed heads, each representing a 145
maternal family, were harvested approximately every two meters along a 100 m transect within 146
each population (as in Hamilton and Eckert 2007). In addition to field collections, three seed 147
populations with known provenance were provided by commercial growers from within the 148
prairie ecoregion and incorporated. Two populations were provided by commercial growers (SD-149
PMG, MN-PMG), and one from the United States Department of Agriculture collected near 150
Pullman, WA (WA-BLK). In total, twenty-two populations were sampled across much of the 151
species’ distribution for inclusion in the common garden experiment (Fig. 2). 152
Common garden experiment – On November 7th, 2015, seeds were planted in a greenhouse at 153
North Dakota State University in Fargo, ND. Using a half-sib design, 12 seeds from each of ten 154
maternal seed families from each population were planted across 12 randomized complete 155
blocks. In addition, two individual seeds from each commercial collection were planted in each 156
block, for a total of 24 seeds per commercial seed population. In May 2016, surviving seedlings 157
(58% of the total, Yoko et al., 2020) were transferred to a permanent field common garden 158
location at Minnesota State University at Moorhead's Ecoregional Science Center (46.86913N, -159
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

8
96.4522W) in Moorhead, MN. The randomized-complete block design was maintained, and a 160
full description of the establishment and maintenance of the common garden experiment can be 161
found in Yoko (2020). 162
Measurement of physiological traits 163
Stomatal Traits – In July 2016, using mature plants from the field common garden experiment, 164
both abaxial (lower) and adaxial (upper) leaf surfaces were assessed for stomatal trait variation. 165
Using a thin layer of Newskin "liquid bandage," two randomly selected leaves per individual 166
were sampled to quantify lower and upper leaf surface stomatal trait variation (N=650, 417 GLA, 167
91 MBA, 142 PRA). 'Liquid bandage' leaf impressions were mounted onto slides and 168
photographed using a Zeiss Stereo Discovery (V8) digital microscope (Carl Zeiss Microscopy, 169
LLC, Thornwood, NY, USA) with a Canon Rebel T3 E0S 1100D digital camera (Canon Virginia 170
Inc., Newport News, VA, USA). Photographs were standardized to a 0.32 x 0.42 mm grid and 171
were subsequently analyzed for stomatal trait variation using ImageJ software (v1.52a, National 172
Institutes of Health, USA). Stomatal traits evaluated included
guard cell length (GCL,
μ
m), 173
which is a proxy for stomatal size, stomatal density (SD, mm2), which represents the number of 174
stomata per unit leaf area, and area of the leaf occupied by stomata, stomatal area index (SAI, 175
mm2). Measurements are reported independently for both abaxial (ab) and adaxial (ad) leaf 176
surfaces. Individual stomatal size measurements represent an average of three guard cells per 177
individual. Stomatal density was calculated by dividing the total number of stomata per slide by 178
the area of the grid. The stomatal area index is reported as the product of average guard cell 179
length and stomatal density (Bertel et al., 2017). 180
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

9
Water-use efficiency -- In May 2018, leaf samples were harvested to estimate intrinsic water-use 181
efficiency (WUE). Foliar carbon isotopes were analyzed because they provide an ability to assess 182
water-use efficiency over the lifetime of a leaf (Farquhar et al., 1989). We sampled leaves from 183
approximately five individuals per population (53 GLA, 9 MBA and 31 PRA). Leaf tissue was 184
oven-dried at 55

over 24 hours and then homogenized into a fine powder using a TissueLyser 185
II (Qiagen, Hilden, Germany). Between 4-5 mg of homogenized leaf tissue were weighed and 186
placed into a tin capsule (Costech, Valencia, CA, USA) for 13C isotope analysis using a 187
continuous-flow isotope ratio mass spectrometer (Sercon Ltd., Cheshire, UK) at UC Davis Stable 188
Isotope Facility (Davis, CA, USA). The reported

13C values are expressed relative to the Vienna 189
Pee Dee Belemnite. 190
Statistical analysis 191
Assessing genetic differences across ecoregions – To test for differentiation in stomatal traits and 192
WUE associated with ecoregion of origin, we used an analysis of variance (ANOVA). All traits 193
were first assessed for normality and homogeneity of variance. Following the ANOVA, a 194
posthoc Tukey honest significance test was performed to identify significant pairwise differences 195
between ecoregions. All statistical tests were performed in R (R Core Team, 2018). 196
The role of climate to population trait variation – To evaluate the relationship between climate 197
of origin and its contribution to physiological trait variation, we extracted annual climatic 198
variables, representing the most recent 30-year averages spanning 1980-2010, from ClimateNA 199
v.5.50 using source population latitude, longitude, and elevation (Wang et al., 2016). Climate 200
variables were highly correlated; therefore, we performed a principal component analysis (PCA) 201
to reduce multicollinearity across traits using R (R Core Team, 2018). PC1 explained 47% of the 202
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

10
climatic variation across populations, and PC2 explained 30%. Given the first two PC axes 203
explained over 77% of the variation across traits, subsequent analyses reflect only PC1 and PC2. 204
(Supporting Table 1). 205
To test whether differences in stomatal traits and WUE could be explained by climatic 206
variation summarized as PC1 and PC2, we fit linear regressions using the lm function in R (R 207
Core Team, 2018). We tested three models using PC1 and PC2 as predictor variables. Two 208
models included each PC as a single predictor, and the third model included both factors as 209
additive predictors. The best-fitting model was identified as that with the lowest Akaike 210
Information Criterion (AIC, Supporting Table 1). 211
Ecoregion-specific genetic variance for stomatal traits –Additive genetic variance (VA), broad-212
sense heritability (H2), narrow-sense heritability (h2), and evolvability (CVA) were estimated for 213
all stomatal traits. As only a subset of individuals were assessed for WUE, it was not possible to 214
estimate genetic variance components for this trait. Phenotypic variance attributable to 215
ecoregion, population, family, and block were evaluated using a linear mixed-effect model and 216
the lmer_alt function within the afex package in R (Singmann et al., 2021):
217


  

 

 
,
 

 
,
 

where


is the phenotypic variance attributable to the random effect of the ith block,


is the 218
random effect of the jth population,

,
is the random effect of the kth family within the ith 219
population,


 
,
is the interaction effect of block x family, and


is the random error. We 220
calculated the additive genetic variance (VA) as: 221


 2.5  
,
222
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

11
Where

,
is the random effect of the kth family within the ith population. Traditionally, 223
narrow-sense heritabilities use a coefficient of relationship (p=1/4) to account for a half-sib 224
design like that used in our experiment. However, we modified our model based on Ahrens et al., 225
(2020) to account for the potential mixed mating system of G. triflorum adopting a coefficient of 226
relationship of p=1/2.5 to generate conservative values. 227
Broad-sense heritability (H2) was calculated as follows using variance components from the 228
mixed model: 229





 
,
 

 
,


 

 
,
 

 
,
 

230
The variance components associated with the kth family within the ith population and the block x 231
family (


 
.
) interaction effect was subsequently used to estimate narrow-sense heritability 232
and the evolvability. Narrow-sense heritability (h2) for leaf surface traits was estimated using the 233
following equation: 234


2.5  

,



,

 

,
 

 


where


is the narrow-sense heritability,


,

is the family within population component of 235
variance,


,
 

is the variance in the interaction effect of block and family (


 
.

, and 236



is the error component
 


representing individual variance. Standard errors for h2 237
were calculated by dividing the pooled standard deviations of the model by the trait sample size. 238
In addition to broad and narrow-sense heritability estimates, we included estimates of 239
evolvability to standardize comparisons of additive genetic variance as a means for comparison 240
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

12
across traits and ecoregions (Hansen and Houle 2008; Hansen et al., 2011; Cava et al., 2019). We 241
calculated evolvability as follows: 242


 2.5  

,

χ
where


,

is the family component of variance, and
χ
is the mean of the trait under evaluation. 243
RESULTS 244
Physiological trait differentiation 245
Using an ANOVA, we tested for differences in stomatal traits and WUE for populations 246
of G. triflorum sourced from distinct ecoregions. Significant trait differentiation was observed 247
between ecoregions for all traits (Table 1). Populations from both GLA and MBA ecoregions 248
exhibited reduced guard cell length for abaxial (GLA=27.1 µm± 0.1 µm; MBA=26.5 µm ± 0.2 249
µm) and adaxial (GLA=26.5 µm ± 0.1 µm; MBA=26.1 µm ± 0.2 µm) leaf surfaces relative to 250
PRA values (abaxial 28.2 µm ± 0.2 µm and adaxial, 28.1 µm ± 0.2 µm, Fig 3A-B). While 251
populations sourced from both GLA and MBA did not differ significantly in abaxial stomatal 252
density, both had greater overall stomatal density (260.8 mm2 ± 3.7 mm2; 246.3 mm2 ±7.5 mm2 253
respectively) relative to populations sourced from the PRA ecoregion (PRA=206.6 mm2 ±5.8 254
mm2, Fig3C). Interestingly, all three ecoregions differed significantly from each other for adaxial 255
stomatal density, with MBA exhibiting the greatest (MBA=193.3 mm2 ± 5.3 mm2), followed by 256
GLA (GLA=176.6 mm2 ± 2.6 mm2), and PRA (PRA=141.32 mm2 ± 4.1 mm2, Fig3D). For 257
stomatal area index (SAI) significant ecoregional differences were observed across both leaf 258
surfaces (Fig3E-F). Populations sourced from the GLA ecoregion had, on average, the largest 259
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

13
abaxial stomatal area index (GLA=5.28 mm2 ± 01 mm2), followed by MBA (MBA=4.87 mm2 ± 260
0.1 mm2), with PRA exhibiting the lowest (PRA=4.30 mm2 ±0.1 mm2). For adaxial stomatal area 261
index, MBA exhibited the greatest SAI (MBA=5 mm2± 0.1 mm2), followed by GLA (GLA=4.7 262
mm2 ±0.1 mm2), while PRA consistently exhibited reduced SAI for both leaf surfaces relative to 263
alvar ecoregions (PRA=3.93 mm2 ±0.1 mm2). Plants sourced from GLA exhibited higher water 264
use efficiency (GLA=-29.4 ± 0.1) relative to MBA (-29.7 ±0.2), although there was no 265
significant difference between the two alvar ecoregions. In contrast, plants sourced from the 266
PRA ecoregion exhibited significantly reduced water use efficiency relative to both GLA and 267
MBA (-30.1 ±0.1) (Fig4). 268
Physiological trait variation associated with climate of origin 269
Using a PCA based on 30-year climate averages we assessed the distribution of climate 270
variation for populations sourced from prairie and alvar ecoregions (Table S1). The first two 271
axes of the PCA explained over 77% of the variation, with PC1 explaining 47% and PC2 272
explaining 30%. The climatic variables that exhibited the highest loadings on PC1 were all 273
associated with temperature and photoperiod, including mean annual temperature (MAT), 274
growing degree days above 18°C (DD_18), and variables related to the frost-free period (eFFP 275
and FFP). Climate variables with the highest loadings on PC2 were related to water availability, 276
including climate-moisture deficit (CMD) and annual heat moisture index (AHM) which indicate 277
the amount of water available for plant uptake. A linear regression with combinations of PC1 and 278
PC2 was used to predict stomatal trait and WUE variation for plants grown in the common 279
garden. The model that best predicted physiological trait variation was assessed by comparing 280
AIC scores, with the chosen model exhibiting the lowest AIC value. In nearly all cases PC2 was 281
the best predictor for stomatal traits and WUE (Table S2), indicating water availability likely 282
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

14
explains variance in physiological traits for G. triflorum. Populations sourced from regions of the 283
species’ range that experience increased climate moisture deficit and higher annual heat moisture 284
indices were associated with increased abaxial and adaxial guard cell length (r2=0.39, P<0.01; 285
r2=0.39, P<0.01 respectively, Fig5 A-B). In contrast, greater climate moisture deficit and higher 286
annual heat moisture associated with the PC2 axis predicted reduced stomatal density (r2=0.40, 287
P<0.01; r2=0.37, P<0.01 respectively, Fig5 C-D). This indicates that populations from 288
ecoregions with reduced water availability on average produced fewer, but larger, stomata. 289
Similarly, both abaxial and adaxial stomatal area indices decreased with larger values of PC2 290
(r2=0.41, P<0.01; r2=0.33, P<0.01, respectively, Fig5 E-F). Finally, WUE decreased across the 291
PC2 axis (r2=0.37, P<0.01, Fig. 6), indicating the control over water-use decreases under 292
increased CMD and AHM for G. triflorum populations. 293
Distribution of additive genetic variance for stomatal traits 294
Additive genetic variance (VA) provides an estimate of the amount of genetic variation 295
available for selection to act upon (Falconer and Mackay, 1996). Ecoregion-specific estimates 296
for additive genetic variance (VA) of stomatal traits were quantified using the half-sibling design 297
(Table 2). In the common garden, Great Lake alvar individuals exhibited the greatest VA for 298
adaxial and abaxial guard cell length (0.60, 0.50), stomatal density (425, 353) and for abaxial 299
stomatal density (0.10). These results suggest that for G. triflorum, the majority of standing 300
genetic variation for stomatal traits is associated with the Great Lake alvar ecoregion. 301
Heritability of stomatal traits 302
We estimated both broad-sense and narrow-sense heritabilities for traits across eco-303
regions to understand how history of selection may influence the distribution of phenotypic trait 304
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

15
variance. Broad-sense heritability (H2) accounts for all genetic components of total phenotypic 305
variance and was calculated for each ecoregion. Estimates of stomatal trait H2 ranged from 0.09-306
0.9 (Table 2). H2 was comparable for GLA and MBA individuals for abaxial and adaxial GCL 307
(GLA= 0.9±0.28, MBA= 0.8±0.2; GLA= 0.9±0.32, MBA= 0.8±0.21, respectively), and PRA 308
exhibited the lowest H2 for abaxial and adaxial GCL (0.09±0.05; 0.12±0.1, respectively). Across 309
all three ecoregions, large variances were observed around H2 estimates for abaxial and adaxial 310
stomatal density traits limiting our ability for comparison of stomatal density H2 across 311
ecoregions. Finally, abaxial and adaxial stomatal area index showed similar trends to H2 312
estimates for GCL, where individuals sourced from the alvar ecoregions had comparable H2 313
estimates (GLA= 0.86±0.18, MBA= 0.9±0.12; GLA= 0.85±0.16, MBA= 0.84±9.15, 314
respectively) and were greater than individuals sourced from the PRA ecoregion (0.34±0.09, 315
0.33±0.12). These results suggest physiological trait differences associated with ecoregions are 316
likely attributable to genetic effects. 317
While broad sense heritabilities includes total genetic variance and provide an 318
understanding of total genetic effects contributing to a trait phenotype, narrow-sense 319
heritabilities (h2) account for the proportion of genetic variance attributed to additive effects, 320
providing an understanding of potential response to selection. There was substantial variability in 321
narrow sense heritabilities estimated for physiological traits across ecoregions (Table 2). On 322
average, individuals from the GLA ecoregion exhibited greater h2 for all traits, excluding 323
stomatal area index, which was greatest for individuals sourced from the PRA ecoregion (Table 324
2). Narrow-sense heritability was similar between GLA (0.43±0.10) and MBA (0.39±0.20) for 325
abaxial guard cell length and h2 for individuals sourced from the GLA ecoregion were relatively 326
consistent across leaf surfaces (adaxial GLA=0.39±0.20). However, for individuals sourced 327
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

16
from the MBA ecoregion there was no heritability observed for adaxial guard cell length, likely 328
reflecting a lack of maternal families for this ecoregion. For abaxial guard cell length, GLA 329
exhibited the highest h2 (0.43±0.10), followed by MBA (0.39±0.20) and PRA, which exhibited 330
the lowest degree of h2 (0.22±0.15). Additionally, GLA exhibited the greatest h2 for adaxial 331
guard cell length traits (0.39±0.10), followed by PRA (0.21± 0.15). Across all three ecoregions, 332
large variances were observed for heritabilities for abaxial and adaxial stomatal density, 333
hindering our ability to make ecoregion comparisons for stomatal density. For abaxial stomatal 334
area index, PRA exhibited the greatest h2 (0.17±0.06), followed by GLA (0.1±0.04), however, no 335
h2 was observed for MBA individuals as seen with adaxial guard cell length. Lastly, MBA had 336
the greatest h2 for adaxial stomatal area index (0.52±0.15), followed by PRA (0.44±0.10), and 337
GLA exhibited the lowest (0.41±0.06). These values suggest a proportion of the ecotypic 338
differences observed between regions are likely attributable to additive genetic effects with some 339
variance across ecoregions. 340
Evolvability of stomatal traits 341
To determine whether the adaptive capacity of stomatal traits differs across ecoregions 342
we calculated evolvability. While alvar and prairie ecoregions exhibited eco-region differences 343
in VA, evolvability did not vary by ecoregion (Table 2). This suggests, that while the amount of 344
additive genetic variance is greater for alvar ecoregions, the per-generation change expected due 345
to any given selection coefficient is similar across ecoregions. Although evolvability did not vary 346
by ecoregion it did vary across traits (Table 2). Abaxial and adaxial guard cell length had the 347
lowest evolvabilities (0-0.04), and adaxial stomatal area index exhibited the greatest (0.15-0.18), 348
suggesting that the per generation change will be greater for stomatal area index traits than for 349
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

17
guard cell length. Overall, evolvabilities for all traits were low ranging from 0- 0.18 indicating 350
that the expected per generation change in these traits is likely limited (Table 2). 351
DISCUSSION 352
Selection associated with environment can influence the distribution of genetic variance 353
underlying traits across a species’ range. Here, by examining populations of G. triflorum sourced 354
from distinct ecoregions with contrasting predictability in water availability, we observed 355
substantial differentiation in physiological traits that could impact recommendations for seed 356
transfer across environments. Climate factors associated with varying water availability strongly 357
predicted physiological trait variation across ecoregions, indicating that environment has likely 358
contributed to the evolution of trait differences. Populations sourced from the alvar ecoregion 359
exhibited increased stomatal density but reduced stomatal size and greater water use efficiency 360
relative to prairie populations when grown in a common environment. This suggests that plants 361
sourced from alvar ecoregions may have evolved increased control over water use. In addition, 362
additive genetic variance for physiological traits was greater for populations sourced from the 363
environmentally predictable alvar ecoregions relative to those sourced from the prairie 364
ecoregion. Heritability estimates suggest the alvar populations exhibit increased genetic control 365
over the phenotypic expression of physiological traits. However, estimates of evolvability 366
suggest that exposure to varying selection coefficients may lead to limited change in traits over 367
generations across ecoregions. Thus, our results suggest that while the environment contributes 368
to the evolution of genetic differences across ecoregions and the distribution of genetic variance 369
in traits important to adaptation, the adaptive capacity overall of G. triflorum may be limited 370
range wide. Combined, the evolution of genetic differences may lead to environment-trait 371
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

18
mismatches following movement of seed across ecoregions and populations may have limited 372
capacity to buffer the fitness consequences of mismatches via plasticity. 373
Physiological trait differentiation associated with seed source environment 374
Physiological traits often exhibit differentiation associated with environment of origin 375
(Dudley, 1996; Didiano et al., 2016; Dittberner et al., 2019; Galliart et al., 2018; Ramirez-376
Valiente et al., 2018). Here we observed that when grown in a common environment, 377
populations sourced from the alvar ecoregion exhibited, on average, smaller and more numerous 378
stomata relative to populations sourced from prairie ecoregions. In addition, alvar populations 379
exhibited greater intrinsic WUE relative to prairie populations suggesting that physiological 380
traits are differentiated across ecoregions. Alvar environments exhibit annual cycles of extreme 381
variation in water availability; from flooding in the spring to early summer desiccation (Catling 382
and Brownell, 1995, Yoko et al., 2020;). Thus, variation across ecoregions may reflect the 383
evolution of physiological traits required to maintain fitness under seasonal extremes in water 384
availability. Many, but small stomata may enable plants to respond rapidly to varying extremes 385
within the alvar ecoregion (Drake et al., 2013). Previous studies have shown the evolution of 386
traits in response to water stress (Anderson et al., 2011; Wadgymar et al., 2016) or have directly 387
linked physiological trait variation and water-availability to source environment (Dudley, 1996; 388
Didiano et al., 2016; Dittberner et al., 2019; Galliart et al., 2018; Ramirez-Valiente et al., 2018). 389
Our results demonstrate that the environment of seed source may contribute to the evolution of 390
phenotypic differences in physiological traits, which could impact fitness if seed is transferred to 391
a new environment during restoration. 392
393
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

19
Stochastic evolutionary processes may also contribute to trait differentiation observed for 394
populations of G. triflorum. Previous studies suggest that alvar populations were likely founded 395
from an expansion of the prairie ecoregion during the warming Hypsithermal but have 396
subsequently become isolated following the consequent cooling period (Hamilton and Eckert, 397
2007). Thus, genetic differences may have accumulated across alvar populations due to 398
stochastic processes associated with isolation and reduced connectivity relative to prairie 399
populations (Lande 1992, Young et al., 1996). If lack of gene flow or drift following isolation 400
were the primary mechanisms contributing to differentiation, we would expect geographically 401
proximal MBA populations to be more similar to PRA populations, where there is a common 402
history and high probability of gene flow between ecoregions that would limit the evolution of 403
trait differences. However, our results suggest that Great Lake and Manitoba alvar populations 404
are more similar to each other, suggesting that selection associated with environment has likely 405
driven the evolution of physiological trait differences among ecoregions. 406
Climate of origin predicts physiological trait variation 407
Using climate associated with population origin we performed a PCA to identify those 408
climate variables that structure population variation across the range of G. triflorum. We found 409
ecoregions were differentiated primarily by temperature (PC1) and water availability (PC2). 410
While PC1 explained the most variation between ecoregions, it did not predict physiological trait 411
variation. However, we did observe a relationship between PC2 and physiological trait variation. 412
Using the PC2 axis, we observed greater annual climate moisture deficit (CMD) and annual heat 413
moisture index (AHM) were associated with fewer, but larger stomata and lower WUE 414
characteristic of the prairie ecoregion. Previous studies found similar patterns where drier 415
conditions led to reduced stomatal control impacting plant water use (Didiano et al., 2016; Guo 416
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

20
et al., 2017). Yoko (2020) suggested that stomatal traits and WUE in G. triflorum were likely 417
under strong divergent selection due to ecoregional differences in water availability. 418
Interestingly, other traits examined by Yoko (2020) may be related to climatic variables along 419
the PC1 axis. For example, Yoko (2020) found that prairie populations invest more energy 420
towards resource allocation than alvar populations, which may be related to variables observed 421
along the PC1 axis. 422
423
Distribution of genetic variance across ecoregions 424
The amount of additive genetic variance in fitness-related traits is proportional to the 425
amount of genetic variance available for selection to act upon (Kulbaba et al., 2019). Here, we 426
found that individuals sourced from the alvar ecoregions, which exhibit predictable seasonal 427
extremes in water availability, exhibited the greatest amount of additive genetic variance for 428
stomatal traits (Table 2). Temporally varying, but predictable environments like those featured in 429
the alvar ecoregion likely favor the maintenance of additive genetic variance (Levins, 1963; 430
Baythavong, 2011; Kulbaba et al., 2021). As such, estimates of VA for prairie populations, which 431
experience unpredictable changes in water availability, may reflect selection for plasticity 432
(Baythavong, 2011; Kulbaba et al., 2021). Increased estimates of VA from the alvar ecoregion for 433
stomatal traits also suggest that these populations may harbor greater capacity to respond to 434
selection (Kulbaba et al., 2019). However, it is important to note the variance in VA estimates 435
across ecoregions may reflect variance in the number of families included for each regional 436
estimate of trait variance in our common garden experiment (GLA=60; MBA=12; PRA=18). 437
Fewer maternal families evaluated in the MBA and PRA region may lead to conservative 438
estimates of VA. 439
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

21
440
Heritability of stomatal traits 441
Quantifying heritability provides insight into the degree to which trait variation is largely 442
mediated by genetic or environmental effects. We predicted that heritable trait variation would 443
be greater within alvar environments as they experience environmental heterogeneity that is 444
predictable. Interestingly, broad-sense heritabilities for stomatal traits ranged from 0.09-0.9 and 445
narrow-sense heritability estimates ranged from 0.1-0.5 and were similar across ecoregions 446
(Table 2). This suggests that while the genetic effects for some traits is substantial, there is also a 447
substantial proportion of variance attributable to environmental variance. Indeed, unpredictable, 448
heterogeneous environments may select for the maintenance of plasticity to ensure plant 449
resilience to change (Chevin et al., 2010; Reed et al., 2010; Ghalambor et al., 2007; Baythavong, 450
2011; March-Salas et al., 2019). As maternal effects are strongest in first generation seedlings 451
and the first year of growth (Donohue, 2009), it is possible that heritability estimates in our study 452
would decrease in a second generation. However, the perennial life-history and time to produce 453
seed did not facilitate the inclusion of a second generation. To limit the potential effect of the 454
maternal environment we evaluated trait variation following at least six months of establishment 455
in the field as previous studies have indicated the impact of the maternal environment may 456
diminish over time (Donohue, 2009). However, given these caveats, our estimates of both broad 457
and narrow-sense heritability likely represent upper bounds (Falconer and Mackay, 1996). 458
459
Understanding the heritability of traits remains important in the context of restoration, 460
where the degree to which traits are mediated by genetic or environmental effects can be used to 461
inform seed transfer guidelines (Broadhurst et al., 2008; Espeland et al., 2017; Bucharova et al., 462
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

22
2017). Here, we observed that individuals sourced from the alvar ecoregion exhibited greater 463
broad-sense heritability relative to individuals from the prairie ecoregion. This may suggest that 464
ecotypic differences that may have arisen in one environment may impact expression of 465
phenotypes in novel restored environments. Where there is a difference between seed source and 466
transferred environment an increased probability of environmental mismatch may reduce fitness 467
in the restored environment (Reed et al., 2010; March-Salas et al., 2019). Despite this, narrow-468
sense heritability estimates indicate that populations may be able to produce a plastic response to 469
the environment, potentially mitigating negative effects associated with seed transfer and climate 470
change (Arntx and Delph, 2001). 471
Evolvability of stomatal traits 472
Populations require sufficient genetic variation for selection to act upon for adaptation to 473
occur (Shaw & Etterson, 2001, Jump & Peñuelas, 2005; Cotto et al., 2017). We estimated 474
evolvability for stomatal traits based on the standardization of additive genetic variance and 475
noted that all estimates were close to zero with little to no differences across ecoregions (Table 476
2). This suggests that populations used in this experiment may have limited capacity to respond 477
to selection. This is a concern in the context of restoration, which will require seed transferred to 478
a new environment to adapt. In G. triflorum, limited evolvability in traits associated with water 479
use could lead to adaptational lags when seed is transferred across environments. Reduced 480
evolvability may leave populations more susceptible to demographic declines (Shaw & Etterson, 481
2001, Jump & Peñuelas, 2005; Cotto et al., 2017). As restorations are multi-species, we advocate 482
for studies that quantify the distribution of genetic variation and evolvability for traits important 483
to adaptation. In this way we may predict long-term evolutionary potential of seed populations 484
used in restoration. Finally, as the work presented here was conducted in a common garden in the 485
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

23
prairie region, we urge caution when extrapolating our results into other systems and recognize 486
that a reciprocal transplant experiment allows for the evaluation of additive genetic variance in 487
alvar environments and quantification of plasticity. 488
CONCLUSIONS 489
Identifying how the environment influences the evolution of ecotypes is important to 490
development of seed transfer guidelines. For G. triflorum populations, we observed ecoregional 491
differentiation for physiological traits and variation in the distribution of genetic variation. This 492
suggests different seed source populations may exhibit varying evolutionary trajectories that 493
could impact seed transfer decisions. Thus, minimizing environmental differences when 494
transferring seed across environments may be necessary where genetic differences exist among 495
seed sources. However, sourcing local seed may not be enough to create restored populations 496
capable of withstanding climate change (Broadhurst et al., 2008; Bucharova et al., 2018; 497
Espeland et al., 2017). By evaluating heritable genetic variation for traits important to adaptation, 498
it may be possible to quantify the effect selecting seed for restoration beyond local sources will 499
have to long-term adaptive potential. 500
ACKNOWLEDGEMENTS 501
The authors thank Jon Sweetman, Chad Stratilo, Mary Vetter, Rebekah Neufeld, Tyler Stadel, 502
Steve Travers, and the Nature Conservancy of Canada for help with initial field sampling to 503
establish the common garden experiment. In addition, we thank Nick Hugo, Zoe Portlas, Naomi 504
Hegwood, Storm Nies and Zeb Yoko for assistance in the field, and specifically acknowledge 505
Stephen Johnson and Alexis Pearson for help with making stomatal trait impressions in the 506
common garden. This work was supported by a Frank J. Cassel Undergraduate Research Award 507
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

24
to K.L.V. and a new faculty award from the office of North Dakota Experimental Program to 508
Stimulate Competitive Research (ND-EPSCoR NSF-IIS-1355466) to J.A.H. 509
510
DATA AVAILABILITY STATEMENT 511
All data and scripts associated with this manuscript are available on GitHub 512
(https://github.com/KateLVolk/AJB-common-garden-physiology). 513
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

25
REFFERENCES 514
Ahrens, C. W., M. E. Andrew, R. A. Mazanec, K. X. Ruthrof, A. Challis, G. Hardy, M. Byrne, et 515
al. 2020. Plant functional traits differ in adaptability and are predicted to be differentially 516
affected by climate change. Ecology and evolution 10: 232–248. 517
Anderson, J. T., A. M. Panetta, and T. Mitchell-Olds. 2012. Evolutionary and ecological 518
responses to anthropogenic climate change. Plant Physiology 160: 1728–1740. 519
Anderson, J. T., J. H. Willis, and T. Mitchell-Olds. 2011. Evolutionary genetics of plant 520
adaptation. Trends in Genetics 27: 258–266. 521
Anderson, R. H., S. D. Fuhlendorf, and D. M. Engle. 2006. Soil nitrogen availability in tallgrass 522
prairie under the fire–grazing interaction. Rangeland Ecology & Management 59: 625–631. 523
Arntz, M. A., and L. F. Delph. 2001. Pattern and process: evidence for the evolution of 524
photosynthetic traits in natural populations. Oecologia 127: 455–467. 525
Aspinwall, M. J., D. B. Lowry, S. H. Taylor, T. E. Juenger, C. V Hawkes, M. V Johnson, J. R. 526
Kiniry, and P. A. Fay. 2013. Genotypic variation in traits linked to climate and aboveground 527
productivity in a widespread C4 grass: evidence for a functional trait syndrome. New 528
Phytologist 199: 966–980. 529
Baythavong, B. S. 2011. Linking the spatial scale of environmental variation and the evolution of 530
phenotypic plasticity: Selection favors adaptive plasticity in fine-grained environments. 531
American Naturalist 178: 75–87. 532
Becklin, K. M., J. T. Anderson, L. M. Gerhart, S. M. Wadgymar, C. A. Wessinger, and J. K. 533
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

26
Ward. 2016. Examining plant physiological responses to climate change through an 534
evolutionary lens. Plant physiology 172: 635–649. 535
Bengtsson, J., J. M. Bullock, B. Egoh, C. Everson, T. Everson, T. O’Connor, P. J. O’Farrell, et 536
al. 2019. Grasslands—more important for ecosystem services than you might think. 537
Ecosphere 10. 538
Bertel, C., P. Schönswetter, B. Frajman, A. Holzinger, and G. Neuner. 2017. Leaf anatomy of 539
two reciprocally non-monophyletic mountain plants (Heliosperma spp.): does heritable 540
adaptation to divergent growing sites accompany the onset of speciation? Protoplasma 254: 541
1411–1420. 542
Braasch, J. E., L. N. Di Santo, Z. J. Tarble, J. R. Prasifka, and J. A. Hamilton. 2021. Testing for 543
evolutionary change in restoration: A genomic comparison between ex situ, native, and 544
commercial seed sources of Helianthus maximiliani. Evolutionary applications 14: 2206–545
2220. 546
Broadhurst, L., M. Driver, L. Guja, T. North, B. Vanzella, G. Fifield, S. Bruce, et al. 2015. 547
Seeding the future - the issues of supply and demand in restoration in Australia. Ecological 548
Management and Restoration 16: 29–32. 549
Bucharova, A., Bossdorf, O., Hölzel, N., Kollmann, J., Prasse, R., & Durka, W. 2019. Mix and 550
match: regional admixture provenancing strikes a balance among different seed-sourcing 551
strategies for ecological restoration. Conservation Genetics 20: 7–17. 552
Bucharova, A., W. Durka, N. Hölzel, J. Kollmann, S. Michalski, and O. Bossdorf. 2017. Are 553
local plants the best for ecosystem restoration? It depends on how you analyze the data. 554
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

27
Ecology and Evolution 7: 10683–10689. 555
Catling, Paul M., and V. R. B. 1995. A review of the alvars of the Great Lakes region: 556
distribution, floristic composition, biogeography and protection. Canadian field-naturalist 557
109: 143–171. 558
Cava, J. A., N. G. Perlut, and S. E. Travis. 2019. Heritability and evolvability of morphological 559
traits of Savannah sparrows (Passerculus sandwichensis) breeding in agricultural grasslands. 560
PloS one 14: e0210472. 561
Chevin, L.-M., R. Lande, and G. M. Mace. 2010. Adaptation, plasticity, and extinction in a 562
changing environment: towards a predictive theory. PLoS biology 8: e1000357. 563
Comer, P. J., J. C. Hak, K. Kindscher, E. Muldavin, and J. Singhurst. 2018. Continent-scale 564
landscape conservation design for temperate grasslands of the Great Plains and Chihuahuan 565
Desert. Natural areas journal 38: 196–211. 566
Cotto, O., J. Wessely, D. Georges, G. Klonner, M. Schmid, S. Dullinger, W. Thuiller, and F. 567
Guillaume. 2017. A dynamic eco-evolutionary model predicts slow response of alpine 568
plants to climate warming. Nature Communications 8: 1–9. 569
Crow, T. M., S. E. Albeke, C. A. Buerkle, and K. M. Hufford. 2018. Provisional methods to 570
guide species

specific seed transfer in ecological restoration. Ecosphere 9: e02059. 571
Crowe, K. A., and W. H. Parker. 2008. Using portfolio theory to guide reforestation and 572
restoration under climate change scenarios. Climatic Change 89: 355–370. 573
Didiano, T. J., M. T. J. Johnson, and T. P. Duval. 2016. Disentangling the effects of precipitation 574
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

28
amount and frequency on the performance of 14 grassland species. PLoS One 11: 575
e0162310. 576
Dittberner, H., A. Korte, T. Mettler

Altmann, A. P. M. Weber, G. Monroe, and J. de Meaux. 577
2018. Natural variation in stomata size contributes to the local adaptation of water

use 578
efficiency in Arabidopsis thaliana. Molecular ecology 27: 4052–4065. 579
Donohue, K. 2009. Completing the cycle: maternal effects as the missing link in plant life 580
histories. Philosophical Transactions of the Royal Society B: Biological Sciences 364: 581
1059–1074. 582
Drake, P. L., R. H. Froend, and P. J. Franks. 2013. Smaller, faster stomata: scaling of stomatal 583
size, rate of response, and stomatal conductance. Journal of experimental botany 64: 495–584
505. 585
Dudley, S. A. 1996. Differing selection on plant physiological traits in response to environmental 586
water availability: a test of adaptive hypotheses. Evolution 50: 92–102. 587
Espeland, E. K., N. C. Emery, K. L. Mercer, S. A. Woolbright, K. M. Kettenring, P. Gepts, and 588
J. R. Etterson. 2017. Evolution of plant materials for ecological restoration: insights from 589
the applied and basic literature. Journal of Applied Ecology 54: 102–115. 590
Etterson, J. R., and R. G. Shaw. 2001. Constraint to adaptive evolution in response to global 591
warming. science 294: 151–154. 592
Falconer, D. S. 1996. Introduction to quantitative genetics. Pearson Education India. 593
Galliart, M., N. Bello, M. Knapp, J. Poland, P. St Amand, S. Baer, B. Maricle, et al. 2019. Local 594
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

29
adaptation, genetic divergence, and experimental selection in a foundation grass across the 595
US Great Plains’ climate gradient. Global Change Biology 25: 850–868. 596
Gascoigne, W. R., D. Hoag, L. Koontz, B. A. Tangen, T. L. Shaffer, and R. A. Gleason. 2011. 597
Valuing ecosystem and economic services across land-use scenarios in the Prairie Pothole 598
Region of the Dakotas, USA. Ecological Economics 70: 1715–1725. 599
Ghalambor, C. K., J. K. McKay, S. P. Carroll, and D. N. Reznick. 2007. Adaptive versus 600
non

adaptive phenotypic plasticity and the potential for contemporary adaptation in new 601
environments. Functional ecology 21: 394–407. 602
Gilbert, A. L., and D. B. Miles. 2019. Antagonistic responses of exposure to sublethal 603
temperatures: adaptive phenotypic plasticity coincides with a reduction in organismal 604
performance. The American Naturalist 194: 344–355. 605
Guo, C., Ma, L., Yuan, S., & Wang, R. 2017. Morphological, physiological and anatomical traits 606
of plant functional types in temperate grasslands along a large-scale aridity gradient in 607
northeastern China. Scientific reports 7: 1–10. 608
Hamilton, J. A., and C. G. Eckert. 2007. Population genetic consequences of geographic 609
disjunction: A prairie plant isolated on Great Lakes alvars. Molecular Ecology 16: 1649–610
1660. 611
Hamilton, J., S. Flint, J. Lindstrom, K. Volk, R. Shaw, and M. Ahlering. 2020. Evolutionary 612
approaches to seed sourcing for grassland restorations. New Phytologist 225: 2246–2248. 613
Hansen, T. F., and D. Houle. 2008. Measuring and comparing evolvability and constraint in 614
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

30
multivariate characters. Journal of evolutionary biology 21: 1201–1219. 615
Hansen, T. F., C. Pélabon, and D. Houle. 2011. Heritability is not evolvability. Evolutionary 616
Biology 38: 258–277. 617
Havens, K., P. Vitt, S. Still, A. T. Kramer, J. B. Fant, and K. Schatz. 2015. Seed sourcing for 618
restoration in an era of climate change. Natural Areas Journal 35: 122–133. 619
Hoekstra, J. M., T. M. Boucher, T. H. Ricketts, and C. Roberts. 2005. Confronting a biome 620
crisis: global disparities of habitat loss and protection. Ecology letters 8: 23–29. 621
Jump, A. S., and J. Penuelas. 2005. Running to stand still: adaptation and the response of plants 622
to rapid climate change. Ecology letters 8: 1010–1020. 623
Kawecki, T. J., & Ebert, D. 2004. Conceptual issues in local adaptation. Ecology letters 7: 1225–624
1241. 625
Kittelson, P. M., S. Wagenius, R. Nielsen, S. Qazi, M. Howe, G. Kiefer, R. G. Shaw, and N. J. B. 626
Kraft. 2015. How functional traits, herbivory, and genetic diversity interact in Echinacea: 627
Implications for fragmented populations. Ecology 96: 1877–1886. 628
Kulbaba, M. W., S. N. Sheth, R. E. Pain, V. M. Eckhart, and R. G. Shaw. 2019. Additive genetic 629
variance for lifetime fitness and the capacity for adaptation in an annual plant. Evolution 73: 630
1746–1758. 631
Kulbaba, M. W., Z. Yoko, and J. A. Hamilton. 2021. Chasing the fitness optimum: temporal 632
variation in the genetic and environmental expression of life-history traits for a perennial 633
plant. bioRxiv. 634
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

31
Lande, R. 1992. Neutral theory of quantitative genetic variance in an island model with local 635
extinction and colonization. Evolution 46: 381–389. 636
Levins, R. 1963. Theory of fitness in a heterogeneous environment. II. Developmental flexibility 637
and niche selection. The American Naturalist 97: 75–90. 638
Lowry, D. B., M. C. Hall, D. E. Salt, and J. H. Willis. 2009. Genetic and physiological basis of 639
adaptive salt tolerance divergence between coastal and inland Mimulus guttatus. New 640
Phytologist 183: 776–788. 641
March-Salas, M., M. van Kleunen, and P. S. Fitze. 2019. Rapid and positive responses of plants 642
to lower precipitation predictability. Proceedings of the Royal Society B 286: 20191486. 643
Maricle, B. R., K. L. Caudle, K. J. Lindsey, S. G. Baer, and L. C. Johnson. 2017. Effects of 644
extreme drought on photosynthesis and water potential of Andropogon gerardii (big 645
bluestem) ecotypes in common gardens across Kansas. Transactions of the Kansas 646
Academy of Science 120: 1–16. 647
McKay, J. K., C. E. Christian, S. Harrison, and K. J. Rice. 2005. “How local is local?”—a review 648
of practical and conceptual issues in the genetics of restoration. Restoration Ecology 13: 649
432–440. 650
McMillan, C. 1959. The role of ecotypic variation in the distribution of the central grassland of 651
North America. Ecological Monographs 29: 286–308. 652
Olsen, J. T., K. L. Caudle, L. C. Johnson, S. G. Baer, and B. R. Maricle. 2013. Environmental 653
and genetic variation in leaf anatomy among populations of Andropogon gerardii (Poaceae) 654
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

32
along a precipitation gradient. American Journal of Botany 100: 1957–1968. 655
Picotte, J. J., D. M. Rosenthal, J. M. Rhode, and M. B. Cruzan. 2007. Plastic responses to 656
temporal variation in moisture availability: consequences for water use efficiency and plant 657
performance. Oecologia 153: 821–832. 658
Ramírez

Valiente, J. A., N. J. Deacon, J. Etterson, A. Center, J. P. Sparks, K. L. Sparks, T. 659
Longwell, et al. 2018. Natural selection and neutral evolutionary processes contribute to 660
genetic divergence in leaf traits across a precipitation gradient in the tropical oak Quercus 661
oleoides. Molecular Ecology 27: 2176–2192. 662
Reed, T. E., R. S. Waples, D. E. Schindler, J. J. Hard, and M. T. Kinnison. 2010. Phenotypic 663
plasticity and population viability: the importance of environmental predictability. 664
Proceedings of the Royal Society B: Biological Sciences 277: 3391–3400. 665
Singmann, H., Bolker, B., Westfall, J., Aust, F., Ben-Shachar, M.S., Hojsgaard, S., Fox, J., 666
Lawrence, M.A., Mertens, U., Love, J., Lenth, R. , Christensen, R. H. B. 2021. Package 667
‘afex’. 668
VanWallendael, A., D. B. Lowry, and J. A. Hamilton. 2022. One hundred years into the study of 669
ecotypes, new advances are being made through large-scale field experiments in perennial 670
plant systems. Current Opinion in Plant Biology 66: 102152. 671
Wang, T., Hamann, A., Spittlehouse, D., & Carroll, C. 2016. Wang T, Hamann A, Spittlehouse 672
D, Carroll C. Locally downscaled and spatially customizable climate data for historical and 673
future periods for North America. PloS one 11. 674
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

33
Wang, T., G. A. O’Neill, and S. N. Aitken. 2010. Integrating environmental and genetic effects 675
to predict responses of tree populations to climate. Ecological applications 20: 153–163. 676
Wimberly, M. C., D. M. Narem, P. J. Bauman, B. T. Carlson, and M. A. Ahlering. 2018. 677
Grassland connectivity in fragmented agricultural landscapes of the north-central United 678
States. Biological Conservation 217: 121–130. 679
Yoko, Z. G., K. L. Volk, N. A. Dochtermann, and J. A. Hamilton. 2020. The importance of 680
quantitative trait differentiation in restoration: landscape heterogeneity and functional traits 681
inform seed transfer guidelines. AoB Plants 12: plaa009. 682
Young, A., T. Boyle, and T. Brown. 1996. The population genetic consequences of habitat 683
fragmentation for plants. Trends in ecology & evolution 11: 413–418. 684
685
686
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

34
Table 1
. Summary of ANOVA for stomatal traits and
WUE comparing differences associated with ecoregion of
origin for Geum triflorum seedlings planted in a common
environment. Bold values are significant P<0.001
Trait df SS F P
GCLab 2.00 1.5E+02 19.15 <0.001
GCLad 2.00 2.7E+02 29.83 <0.001
SDab 2.00 2.7E+05 29.12 <0.001
SDad 2.00 1.6E+05 35.29 <0.001
SAIab 2.00 8.7E+01 27.59 <0.001
SAIad 2.00 6.7E+01 24.75 <0.001
WUE 2.00 9.1E+00 12.12 <0.001
Note: GCLab, abaxial guard cell length; GCLad, adaxial guard cell length;SDab,
abaxial stomatal density; SDad, adaxial stomatal density; SAIab,stomatal area index;
WUE, water-use efficiency
687
688
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

1
Table 2. Ecoregional-specific broad-sense heritabilities (H2), additive genetic variance (V
A
), narrow-sense heritabilities (h2) and evolvability (CV
A
) f
PRA MBA GLA
Trait H2 V
A
h2 CVA H2 VA h2 CVA H2 VA
GCLab 0.09±0.05 0.42 0.22±0.15 0.04 0.8±0.2 0.51 0.39±0.20 0.04 0.9±0.28 0.60
0.
GCLad 0.12±0.1 0.42 0.21± 0.15 0.04 0.8±0.21 0.00 0.0±0.0 0.00 0.9±0.32 0.50
0.
SDab 0.32±4.35 168 0.13±2.93 0.10 0.9±9.1 239.00 0.13±4.19 0.10 0.86±10.12 425.00
0.
SDad 0.4±4.17 240 0.37±3.49 0.17 0.9±5.8 252.00 0.27±4.35 0.13 0.85±7.1 353.00
0.
SAIab 0.34±0.09 0.07 0.17±0.06 0.10 0.9±0.12 0.00 0 0.00 0.86±0.18 0.10
0
SAIad 0.33±0.12 0.2 0.44±0.10 0.18 0.84±9,15 0.30 0.52±0.15 0.17 0.85±0.16 0.19
0.
Note: GCLab, abaxial guard cell length; GCLad, adaxial guard cell length;SDab,abaxial stomatal density; SDad, adaxial stomata
l density; SAIab,stomatal areaindex
; WUE, water
-
use efficiency; PRA, prairie; MBA, Manitob
alvar
689
690
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

1
Fig 1. Predictive scenarios for how the distribution of heritable genetic variation could be 691
influenced by environmental predictability. Under greater environmental predictability, the 692
distribution of heritable genetic variation should increase (blue line) while decreasing variation 693
attributed to plasticity (red line). 694
Fig 2. Collection sites of G. triflorum populations. Green points represent Great Lake Alvar 695
(GLA) populations, blue points represent Manitoba Alvar populations (MBA), and yellow points 696
represent Prairie populations (PRA). The red star shape indicates the common garden location. 697
Fig 3. Box plots indicate regional differences in physiological traits associated with water-use, 698
including abaxial guard cell length (A), adaxial guard cell length (B), abaxial stomatal density 699
(C), adaxial stomatal density (D), abaxial stomatal area index (E), and adaxial stomatal area 700
index (F). The horizontal line in the box plot indicates the median, and white diamonds indicate 701
the mean. Boxplots with the same letter are not significantly different based on Tukey’s 702
comparison of means (alpha = 0.05). 703
Fig 4. Box plot indicates regional differences in water-use efficiency (WUE). The horizontal line 704
in the box plot indicates the median, and white diamonds indicate the mean. Boxplots with the 705
same letter are not significantly different based on Tukey’s comparison of means (alpha = 0.05). 706
Fig 5. Relationships between stomatal traits and principal component 2 (PC2), including abaxial 707
guard cell length (A), adaxial guard cell length (B), abaxial stomatal density (C), adaxial 708
stomatal density (D), abaxial stomatal area index (E), and adaxial stomatal area index (F). Data 709
points represent population-level averages for each region. Lines depict the shape of the 710
association between PC2 and trait values surrounded by a 95% confidence shading. The 711
significance of each relationship is indicated in the top right corner of each graph (p<0.01). 712
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

2
Fig 6. Relationships between water-use efficiency (WUE) and principal component 2 (PC2). 713
Data points represent population-level averages for each region. Lines depict the shape of the 714
association between PC2 and trait values surrounded by a 95% confidence shading. The 715
significance of the relationship is indicated in the top right corner of each graph (p<0.01). 716
717
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

54°N
50°N
46°N
42°N
76°W80°W84°W88°W92°W96°W100°W104°W108°W112°W116°W
Region
Great Lakes Alvar
Manitoba Alvar
Prairie
Common Garden
0 500 1,000250 Km ¯
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

Aa a b
20
25
30
35
GLA MBA PRA
Abaxial guard cell length (µm)
Ba a b
20
25
30
35
GLA MBA PRA
Adaxial guard cell length (µm)
Ca a b
100
200
300
400
500
GLA MBA PRA
Abaxial stomatal density (mm²)
Da b c
100
200
300
GLA MBA PRA
Adaxial stomatal density (mm²)
Ea b c
2.5
5.0
7.5
10.0
GLA MBA PRA
Abaxial stomatal area index (mm²)
Fa b c
2.5
5.0
7.5
10.0
GLA MBA PRA
Adaxial stomatal area index (mm²)
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

a ab b
High WUE
Low WUE
−31
−30
−29
−28
−27
GLA MBA PRA
Carbon isotope composition
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

R2= 0.39**
A
25
26
27
28
29
30
−2 0 2 4 6
Abaxial guard cell length (µ m)
R2= 0.38**
B
26
28
30
−2 0 2 4 6
Adaxial guard cell length (µ m)
R2= 0.40**
C
100
150
200
250
300
−2 0 2 4 6
Abaxial stomatal density (mm2)
R2= 0.37**
D
100
150
200
250
300
−2 0 2 4 6
Adaxial stomatal density (mm2)
R2= 0.38**
E
3
4
5
6
7
−2 0 2 4 6
PC 2
Abaxial stomatal area index (mm2)
R2= 0.33**
F
3
4
5
6
7
−2 0 2 4 6
PC 2
Adaxial stomatal area index (mm2)
GLA
MBA
PRA
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint

R2= 0.37**
High WUE
Low WUE
−31
−30
−29
−28
−2 0 2 4 6
PC 2
Carbon isotope composition
GLA
MBA
PRA
.CC-BY 4.0 International licenseavailable under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprintthis version posted February 12, 2022. ; https://doi.org/10.1101/2022.02.11.480132doi: bioRxiv preprint