Climate drives adaptive genetic responses associated with survival in big sagebrush (Artemisia tridentata)

Abstract A genecological approach was used to explore genetic variation for survival in Artemisia tridentata (big sagebrush). Artemisia tridentata is a widespread and foundational shrub species in western North America. This species has become extremely fragmented, to the detriment of dependent wildlife, and efforts to restore it are now a land management priority. Common‐garden experiments were established at three sites with seedlings from 55 source‐populations. Populations included each of the three predominant subspecies, and cytotype variations. Survival was monitored for 5 years to assess differences in survival between gardens and populations. We found evidence of adaptive genetic variation for survival. Survival within gardens differed by source‐population and a substantial proportion of this variation was explained by seed climate of origin. Plants from areas with the coldest winters had the highest levels of survival, while populations from warmer and drier sites had the lowest levels of survival. Survival was lowest, 36%, in the garden that was prone to the lowest minimum temperatures. These results suggest the importance of climatic driven genetic differences and their effect on survival. Understanding how genetic variation is arrayed across the landscape, and its association with climate can greatly enhance the success of restoration and conservation.

Genecological studies (sensu Turesson, 1923) are a primary means for assessing how landscape and spatial variation causes genetic differences among plant populations. For example, cold temperatures have been found to be an important factor affecting adaptation in many temperate plant species, influencing how and when plants grow to maximize water availability or avoid freezing (Richardson et al., 2014;St. Clair, Mandel, & Vance-Borland, 2005). Responses to the environment can then be used to create geographically delineated seed zones to help ensure that seed can be transferred among areas of similar climate, increasing odds of appropriately adapted vegetation and successful restoration (Bower, St. Clair, & Erickson, 2014). Recent genecology studies have found that winter minimum temperatures are a primary agent of genetic variation in conifers (Rehfeldt, 2004;Rehfeldt & Jaquish, 2010;, native grasses (R. C. Johnson et al., 2012;St. Clair, Kilkenny, Johnson, Shaw, & Weaver, 2013), and shrubs (Horning, McGovern, Darris, Mandel, & Johnson, 2010;Richardson et al., 2014). However, sufficient information on climatic adaptations is lacking in the majority of native grasses, forbs, and shrubs used for restoration purposes.
Artemisia tridentata (big sagebrush) is a widespread desert shrub that dominates much of western North America. Three predominant subspecies of A. tridentata are defined by ecological niches affected principally by climate, soil type, and soil depth (McArthur, 1994).
Common gardens or variation in subspecies habitats has revealed genetic divergence across local gradients, such as across hillslopes or topographic positions (Graham, Freeman, & McArthur, 1995;Wang, McArthur, & Freeman, 1999;Wang, McArthur, Sanderson, Graham, & Freeman, 1997) that can be at least partly explained by differences in climate (Kolb & Sperry, 1999). Local adaptation within (Brabec et al., 2015) and among (Brabec, Germino, & Richardson, 2016) subspecies of A. tridentata was evident from survival analysis of a small number of populations (four and 11, respectively). These considerations and previous findings suggest a high likelihood of local climate adaptation in big sagebrush, although a comprehensive, range-wide assessment is needed. As a foundational species in the Great Basin ecosystems (Prevéy, Germino, Huntly, & Inouye, 2010), it is host to sagebrushobligate wildlife, such as the greater sage-grouse (Centrocercus urophansianus), and is of conservation concern (Knick & Connelly, 2011). Millions of hectares of sagebrush have been eliminated during the past century due to change in land use (e.g., grazing), exotic plant invasion, and increased frequency of wildfire (Miller et al., 2011).
Restoration of A. tridentata ecosystems through seeding or planting has occurred on millions of hectares in the western United States in recent decades and continues to be a management priority. Restoration success of sagebrush is contingent on a number of factors including weather, soils, seed mix composition, and seeding and planting techniques (Hardegree et al., 2011). Genetic adaptation, however, is by far the most critical factor for long-term resiliency of these ecosystems, underscoring the need to better understand how climate affects genetic variation in this species.
In the current study, we assess adaptive genetic variation within and between 55 source-populations of A. tridentata using commongarden experiments at three sites. The use of common gardens allows researchers to test the genetic basis for phenotypic differences among a relatively large number of plant populations. In genecology studies, geospatial patterns of these genetic traits are then related to plant source climates using regression and bioclimatic modeling. In this study, we address two main questions: (i) How is survivorship influenced by genetics, environment and/or their interactions? (ii) What are the climate patterns that affect survivorship? These results will be fundamental in understanding adaptive genetic variation and fitness in A. tridentata, which are essential to restoration in a changing climate.

| Source seed collection
Artemisia tridentata includes three predominant and widely recognized subspecies and two cytotype (ploidy) levels. Subspecies tridentata (basin big sagebrush) is diploid or tetraploid, grows in lower elevations with deep soil, and exhibits rapid growth and tall stature.
Subspecies wyomingensis (Wyoming big sagebrush) is tetraploid, occurs in relatively warm and dry areas with shallow soils, and exhibits slow growth and a shorter stature. Subspecies vaseyana (mountain big sagebrush) is diploid or tetraploid, occurs in cooler and relatively mesic conditions at higher elevations, and has a compact growth form (McArthur & Sanderson, 1999). Tetraploids are polyphyletic, deriving from multiple events within and between subspecies (Richardson, Page, Bajgain, Sanderson, & Udall, 2012). Seeds from each of the three subspecies of A. tridentata were collected from 55 sites, hereafter referred to as populations, in the fall of 2009 (Table S1). A population typically consisted of half-sib families collected from eight to ten parents. These seeds were collected at random over an area approximately 1 ha. Subspecies were identified by morphology (height and stature), ultraviolet iridescence (following the methods of McArthur, Welch, & Sanderson, 1988), cytotype, molecular genetic analyses (Richardson et al., 2012), and volatile organic compounds (Jaeger, Runyon, & Richardson, 2016). Cytotype was typically uniform within a source-population, but could vary among subspecies (Table S1), with the exception of wyomingensis that is only tetraploid.

| Experimental design
A five-year common-garden experiment was designed with three sites to understand how genetic and environmental factors contribute to A. tridentata survival (Table 1; Fig. S1). Garden sites were chosen based on the following criteria: (i) climates that were representative of each of the subspecies, (ii) adjacent to weather stations, and (iii) accessibility for data collection and maintenance. The garden in Ephraim, Utah, USA, is a relatively cold basin site with a dry climate typical of ssp.
tridentata; the garden in Majors Flat, Utah, USA, is a higher-elevation mountainous site with a mesic climate typical of ssp. vaseyana; and the garden in Orchard, Idaho, USA, is a warm plain site with a dry climate typical of ssp. wyomingensis. Weather data at the common gardens were obtained from adjacent weather stations (USDA, Natural Resources Conservation Service and Forest Service).
Seeds were collected from 55 wild source-populations found throughout the species distribution (Table S1; Fig. S1). After collection, seeds were cleaned and placed into a freezer until sowing. Replicate seeds from each population were sown into six-inch cone-tainers with a custom soil mix for native plants, established in a greenhouse for approximately 3 months, and cold hardened outside for 2 weeks. Due to differences in climate (Table 1), the Ephraim and Orchard gardens were planted in late April, and the Majors Flat garden was planted in early June 2010. The germination and planting dates for Majors Flat seedlings were delayed to ensure plants were approximately the same age when transplanted. Seedlings were planted in a completely randomized design in a lightly tilled plot that had an above and below-ground fence to discourage ungulate and rabbit herbivores. Due to variation in seed yield and germination, populations in each garden ranged from 1 to 11 individuals with an average of 7.8 plants per population (Table S1). Plant spacing was 1.5 m between rows and 1 m within rows, and a border row was planted around the perimeter to minimize edge effects. To ensure establishment, plants were watered periodically during the first growing season in 2010.

| Data collection
Phenotypic measurements focused on A. tridentata survival in the three gardens. A plant was considered dead when all foliage was missing and no re-greening occurred during the growing season. Mortality for each experimental plant was recorded in each garden an aver- Climate for each source-population was utilized to determine how source climate affects survival. Geographic coordinates and elevation for each source-population were used to estimate climate variables from thin plate splines using ANUSPLIN v 4.1 (Hutchinson, 2000) based on normalized monthly means from years 1961 to 1990 with a 0.0083 o resolution (~1 km 2 ) (Crookston & Rehfeldt, 2012;Rehfeldt, 2006). A total of 40 climate variables were examined for this study including 18 yearly and season temperature and precipitation indices and 22 interactions among different variables that can be important to plant adaptation (Table S2)

| Data analysis
Data were first analyzed to assess significance of genetics, environment, and their interaction for survival in A. tridentata across the three common gardens. This analysis was performed using a generalized linear mixed model with a quasi-binomial error distribution on survival proportion of each source-population across the three gardens. A multivariate response variable with number of survivors and the number of dead for each source-population was used to account for differences in sample sizes. Estimates of genetic effects were specified as fixed effects and included plant subspecies:cytotype group, and source-population climate variables.
Estimates of environment and genetics by environment were specified as random effects and included garden and the interaction of garden and subspecies:cytotype group. Model was fit using the glmer function from the lme4 package (Bates, Mächler, Bolker, & Walker, 2015) with the objective to find the model with the highest predictive value, combined with the fewest number of climate variables.
To do this, methods similar to R. C. Johnson et al. (2012) were used.
More specifically, a forward selection procedure was used to select the source-population climate variables that best explain variation for survival in the model and variables were no longer added when not significant (p > .05). Significance for each of the effects was calculated based on likelihood-ratio chi-square tests. Conditional and marginal R 2 values were calculated with the rsquared.glmm function (Nakagawa & Schielzeth, 2013); conditional R 2 describes the variation explained by both fixed and random effects; and a marginal R 2 describes fixed effects alone. To provide a more accurate model, only populations with more than two samples were evaluated, resulting in n = 152 populations among the three gardens. This model is also referred to as the genecological model and is used in the mapping discussed below.
Next, a more detailed survival analysis was performed to determine the rate of survival within the Ephraim garden. To do this, the survival package in R (Therneau & Lumley, 2014) was utilized to fit a parametric survival model. This method accounts for the interval and T A B L E 1 Geographic and climatic attributes of the three common gardens  the slope of the two climate variables and wyomingensis). We focused on subspecies wyomingensis because it is the subspecies of greatest conservation concern and restoration investment. Further, distributions of subspecies tridentata and wyomingensis can often be sympatric at this spatial scale, with the distinction that tridentata is usually controlled by local topographic features such as soil depth and higher moisture features (e.g., fence lines, roadways, washes) that provide additional moisture and are difficult to model (Still & Richardson, 2015).

| RESULTS
Survival of A. tridentata varied in the three common gardens as a function of environment, genetics, and their interaction. Survival was significantly different between gardens, indicating an effect of environment on A. tridentata survival ( 2 1 = 15.417, p < .001; Table 2). Survival at Ephraim was much lower than the other two gardens; 36% of A. tridentata at Ephraim survived compared to 85% and 78% at Majors Flat and Orchard, respectively (Fig. S2) and marginally significant SMRP (summer precipitation, July and August), were found to best explain variation in survival ( 2 1 = 32.764, p < .001 and 2 1 = 3.712, p = .054, TDIFF and SMRP, respectively; Table 2).
Mapping the generalized linear mixed genecological model re- To further investigate the low rate of survival at Ephraim, we performed a detailed survival analysis for this garden. Survival increased at a log-normal rate with a scale of 0.649, that is, the rate of mortality decreased over time (Figure 2; Table S3). Rate of survival was significantly different for tetraploid vaseyana (V4x) than for the other subspecies:cytotype groups ( 2 4 = 52.6, p < .001). Specifically, survival time of tetraploid vaseyana (V4x) was 0.445 times shorter than the baseline tetraploid tridentata (T4x) (e −0.810 ; Table S3). The median survival of tetraploid vaseyana (V4x) was 23 months (i.e., only a 50% chance of survival at 23 months), and there was less than 10% chance of survival at 48 months. In contrast, tetraploid tridentata (T4x) had a median survival of 47 months (Table S4)

| Adaptive genetic responses to climate
Growth in A. tridentata occurs primarily in the spring, while soil moisture levels are high (Germino & Reinhardt, 2014;Schlaepfer, Lauenroth, & Bradford, 2014), yet minimum temperatures are often physiologically limiting and even lethal (Brabec et al., 2016). Strong differences in avoidance and tolerance were evidence among seedlings of the subspecies:cytotype groups of A. tridentata related to their mortality and suggested the importance of freezing response as an axis for adaptive differentiation (Brabec et al., 2016). At the landscape level, selection pressures of different climates have potentially generated differences in phenology and growth strategies for A. tridentata.
One hypothesis would be that interior regions, with a continental climate (greater summer-winter temperature extremes), support big sagebrush populations that have a growth strategy that confers resistance, tolerance, or avoidance of freezing temperatures through F I G U R E 1 Genecological projection of survivorship in big sagebrush in three common gardens. Common-garden locations shown as black circles. Log-link transformed coefficients of slopes and intercepts of the generalized linear mixed model were used to project probability of survival into the niche boundaries of wyomingnesis big sagebrush (Still & Richardson, 2015). Climate predictors included in the model are TDIFF and SMRP. Areas that have greater temperature differences between summer and winter and wetter summers have higher probability of survival (blue), while areas with more moderated temperatures and drier summers have lower probability of survival (red). TDIFF: temperature difference between mean temperature in the coldest month and mean temperature in the warmest month; SMRP: summer precipitation (July and August). Hillshade background is based on the US Geological Survey Digital Elevation Model physiological mechanisms or deferment/delay of growth until after the spring frost period. In more moderate climates to the south and west, we further hypothesize that A. tridentata initiates growth earlier in spring to capture spring soil moisture, resulting in greater risk of exposure to freezing in the gardens we evaluated. Below, we discuss how results from our analyses support this hypothesis.
Results from the generalized linear mixed model indicated that source-population climate is a strong predictor of survival in A. tridentata. The final generalized linear mixed genecological model ex- Our results support previous research that also indicated the adaptive importance of minimum temperatures (Bower et al., 2014;Erickson, Mandel, & Sorensen, 2004;Horning et al., 2010;Johnson, Sorensen, Bradley St Clair, & Cronn, 2004;Richardson et al., 2014;St. Clair et al., 2013). Although A. tridentata is generally thought of as being cold tolerant, winterkill has been documented in natural populations (Walser et al., 1990) and common gardens of seedlings at relatively warm sites (Brabec et al., 2016).
Other studies of A. tridentata also found differences in freeze tolerance associated with source location, such as greater freezing acclimation for seedlings from higher-elevation seed sources (Loik & Redar, 2003).
While we found differences in survival rate among plant subspecies:cytotype groups, these differences were highly influenced by source-population. Post hoc comparisons of our survival analysis found that only tetraploid vaseyana (V4x) was significantly different from other subspecies:cytotype groups. This may be attributable to source-population origins rather than true taxonomical differences, considering how many tetraploid vaseyana (V4x) were from the southwest United States (e.g., CAV3, CAV4, NVV3). Previous work has suggested that A. tridentata subspecies are adapted to different habitats (Mahalovich & McArthur, 2004;McArthur, 1994;Wang et al., 1997), which can often be defined on local spatial scales. However, our findings suggest that within subspecies, populations are principally adapted to clines in continentality (Table 2).

| Microclimate variation influences survival
Survival of A. tridentata differed substantially between commongarden sites (  (Franco & Nobel, 1989;Loik & Nobel, 1993;Shreve, 1911). Further, research in A. tridentata has found low survival in vaseyana in areas where a typically high snow pack was absent, yet high survival in areas with a snow cover present (Hanson, Johnson, & Wight, 1982). Further, Loik and Redar (2003) suggested that A. tridentata seedlings at higher elevations may have less exposure to freezing due to greater prevalence of snow. Consistent with this, Brabec et al. (2016) found that subspecies vaseyana had the least and wyomingensis had the greatest physiological avoidance and resistance to freezing among subspecies, and they proposed the seemingly ironic differences were due to greater insulating snow cover at higher elevation during winter and spring. These results suggest that selection for cold tolerance may relate to local-scale microclimate variation, which we propose to be an important topic for further investigation.

| Applications
Future climate scenarios predict an increase in global mean temperatures and an increase in extreme weather patterns. As the climate that A. tridentata is adapted to is displaced, plants will need to migrate to new areas, a slow task for a sessile organism (Shaw & Etterson, 2012).
The movement of plant species in response to rapid climatic warming will frequently be slower than phenotypic and adaptive genetic changes required to adjust to the novel climate. Due to population differentiation, the effects of climate change are likely to vary throughout a species range (Davis, Shaw, & Etterson, 2005). While locations subject to frequent cold air pooling are not likely to escape regionally increasing temperatures, they may act as microrefugia against the amplified temperature trends and variations (Dobrowski, 2011). Through genecological modeling of future climate scenarios, it is likely that as the climate changes, A. tridentata will not be genetically suited to the environment in which it currently grows, resulting in extirpation of some populations.
Movement of seed populations to areas outside of their adaptive breadth can have a negative impact on fitness (Hereford, 2009).
Restoration of A. tridentata after fires or other disturbances has been a management priority over the past decade, yet success of these efforts has been varied Knutson et al., 2014).
Planting A. tridentata seed outside their adaptive breadth could result in unsuccessful establishment and/or low fitness and provide an opportunity for invasive species encroachment leading to a loss of species diversity and ecosystem degradation (reviewed in Dumroese, Luna, Richardson, Kilkenny, & Runyon, 2015). Previous research has shown that sagebrush species/subspecies are adapted to different ecological niches such as elevation and soil type (McArthur, 1994).
Moving sagebrush populations to different climatic or edaphic conditions is therefore not recommended (Mahalovich & McArthur, 2004).
Our work is the first step to creating climate-based guidelines for the transfer of seed for restoration purposes in A. tridentata. Being able to map traits relating to climate across the landscape is one practical advantage of genecology, which is particularly useful in highly heterogeneous environments such as western North America. This research will be the first step to delineate seed transfer zones, guidelines to ensure that seed used for restoration is adapted to the site.