Divergence in drought‐response traits between sympatric species of Mimulus

Abstract Differential adaptation to local environmental conditions is thought to be an important driver of speciation. Plants, whose sedentary lifestyle necessitates fine‐tuned adaptation to edaphic conditions such as water availability, are often distributed based on these conditions. Populations occupying water‐limited habitats may employ a variety of strategies, involving numerous phenotypes, to prevent and withstand desiccation. In sympatry, two closely related Mimulus species—M. guttatus and M. nasutus—occupy distinct microhabitats that differ in seasonal water availability. In a common garden experiment, we characterized natural variation within and between sympatric M. guttatus and M. nasutus in the ability to successfully set seed under well‐watered and drought conditions. We also measured key phenotypes for drought adaptation, including developmental timing, plant size, flower size, and stomatal density. Consistent with their microhabitat associations in nature, M. nasutus set seed much more successfully than M. guttatus under water‐limited conditions. This divergence in reproductive output under drought was due to differences in mortality after the onset of flowering, with M. nasutus surviving at a much higher rate than M. guttatus. Higher seed set in M. nasutus was mediated, at least in part, by a plastic increase in the rate of late‐stage development (i.e., fruit maturation), consistent with the ability of this species to inhabit more ephemeral habitats in the field. Our results suggest adaptation to water availability may be an important factor in species maintenance of these Mimulus taxa in sympatry.

. Although adaptation to different habitats is thought to be one of the most important drivers of speciation (Sobel, Chen, Watt, & Schemske, 2010), in most cases, little is known about the ecological factors involved or the particular phenotypes that contribute to divergence.
In contrast, with avoidance and tolerance strategies, plants prevent drought-induced senescence by increasing water-use efficiency (e.g., via a decrease in stomatal conductance) or though physiological changes (e.g., osmotic adjustment, root growth). Because these strategies involve diverse mechanisms and suites of traits, adaptation to dry soils is often accompanied by dramatic phenotypic changes, which can have important consequences for reproductive isolation between closely related sympatric species. For example, a shift to earlier flowering-a hallmark of drought escape-can lead to phenological isolation (Fishman, Sweigart, Kenney, & Campbell, 2014;Franks & Weis, 2009;Martin, Bouck, & Arnold, 2005). Despite the potential importance of water availability as an axis of plant divergence, there are few detailed characterizations of drought adaptation between closely related species that grow sympatrically (Dunning et al., 2016;Eckhart, Geber, & McGuire, 2004;Geber & Eckhart, 2005).
Here, we focus on divergence in drought response traits between the yellow monkeyflowers Mimulus guttatus and M. nasutus.
Mimulus guttatus is a phenotypically and genetically diverse, primarily outcrossing species that occupies wet soils across western North America (Wu et al., 2008). Mimulus nasutus is a highly selfing species that diverged recently (~200KYA) from M. guttatus (Brandvain, Kenney, Flagel, Coop, & Sweigart, 2014). The two species are largely allopatric, but sympatric populations of M. nasutus and annual ecotypes of M. guttatus are not uncommon throughout their shared range. In addition to their divergent mating systems and associated floral traits (Fishman, Kelly, & Willis, 2002), the two species show clear ecological differentiation, with M. nasutus flowering earlier and tending to occupy microhabitats that dry out sooner than M. guttatus (Kiang & Hamrick, 1978). This shift to earlier flowering in M. nasutus is caused, at least in part, by an ability to flower under much shorter day lengths (<10 hr) than M. guttatus, which often requires at least 14 hr of daylight to initiate reproduction (Friedman & Willis, 2013;Kooyers et al., 2015). When the two species co-occur, divergence in critical photoperiod (Fishman et al., 2014), and in flowering phenology more generally, is a major barrier to interspecific mating (Kenney & Sweigart, 2016;Kiang & Hamrick, 1978;Martin & Willis, 2007). Despite this strong barrier, hybridization between sympatric populations of M. guttatus and M. nasutus can be substantial (Kenney & Sweigart, 2016) and there is clear evidence of ongoing interspecific introgression (Brandvain et al., 2014;Kenney & Sweigart, 2016;Sweigart & Willis, 2003).
How, then, are these two Mimulus species maintained in the face of considerable gene flow? In a previous study (Kenney & Sweigart, 2016), we began to address this question by focusing on populations of M. guttatus and M. nasutus that have come into secondary contact at Catherine Creek (CAC), a gradually sloping, rocky meadow with streams and seeps that flow down into the Columbia River Gorge ( Figure 1). Edaphic conditions, including water availability, are highly F I G U R E 1 One of the streambeds of Catherine Creek (a) in May, with hybrid individuals and putative Mimulus guttatus (b) and M. nasutus (c) growing in close proximity, often within one meter heterogeneous at this site, and although the two Mimulus species often grow within a meter of each other, they are found in somewhat distinct microhabitats: M. nasutus occurs in patches of moss in and around flowing streams that dry up in late spring, whereas M. guttatus grows in deeper seeps that stay wet through spring and into summer. The two species also flower asynchronously at the CAC site (Kenney & Sweigart, 2016) due, in part, to divergence at two major genetic loci for critical photoperiod (Fishman et al., 2014).  (Kenney & Sweigart, 2016). There is also ongoing introgression at CAC, mostly from M. nasutus into M. guttatus (Brandvain et al., 2014), including at one of the two mapped critical photoperiod loci (Kenney & Sweigart, 2016

| Plant lines and growth conditions
To characterize natural variation in drought response within and be-

| Experimental design
We grew plants under two distinct watering treatments to examine plant responses to variation in soil moisture conditions. Following

| Plant trait measurements
To investigate variation within and between Mimulus species for response to water limitation, we quantified a number of drought-related traits under each watering regime. All temporal values were numbered relative to Day 0, when seeds were transferred into growth chambers following stratification.

| Developmental timing
We

| Lifetime maternal fitness
We obtained survival and maturation rates by daily inspection of plants. For most M. guttatus plants that survived to flowering, we marked and hand-pollinated one flower, on its first day of stigma receptivity, from the first or second flower pair with pollen donated from the IM767 edge plants (in some cases, we were unable to perform pollinations before plants dropped their corollas; these individuals were dropped from our analyses). IM767, an inbred line derived from the allopatric Iron Mountain population, was used as the common pollen donor as it is likely to be roughly equally differentiated from all CAC samples (pairwise nucleotide diversity, π s , is ~ 5% between IM and CAC plants, see Brandvain et al., 2014).
Following initial pollinations, the few flowers that remained receptive were hand-pollinated a second time to ensure pollen was not limiting. For most M. nasutus plants that survived to flowering, we marked one or two flowers from the first or second flower pair, and allowed them to self-fertilize (in some cases, we marked flowers from later pairs; these individuals were dropped from our analyses). From these marked flowers, we measured an individual's seed production on a per fruit basis. We note that hand pollination in M. guttatus versus self-pollination in M. nasutus might contribute to species differences in seed production. Nevertheless, variation in seed set due to treatment or species × treatment will be readily detectable.

| Rosette diameter
Using calipers, we measured the rosette diameter of plants at their widest points on Day 25.

| Floral traits
For most plants that survived to flowering, we measured the corolla length and width of one marked flower on the first or second flower pair (in some cases, plants dropped their corollas before measurements could be taken) on the day it was recorded as flowering. We Note: Standard error and sample size given in parentheses. Models: Rosette diameter, corolla length, corolla width, days to bud, days bud to flower, and seeds per fruit: "group" (fixed effect with "line" nested within it) and "treatment" (fixed effect), main effects, "group × treatment," interaction effect "block," random effect. Days flower to fruit: "group" (fixed effect) and "treatment" (fixed effect), main effects, "group × treatment," interaction effect, "block," random effect. Stomatal density: "group" (fixed effect with "line" nested within it) and "block," random effect. Letters indicate Tukey-Kramer grouping for each trait following ANOVA a Stomatal Density was only measured on a subset of plants grown under well-watered conditions. measured corolla length as the distance from the base of the calyx to the end of the longest petal when hand straightened and corolla width as the distance between the widest point of the bottom petal lobes.

| Stomatal density
For plants that survived to Day 52 with healthy green tissue (those in the well-watered treatment), we made a pressing of the abaxial surface of the largest, fully expanded leaf using GE Clear 100% Silicone Caulk (General Electric). We taped these pressings to slides and examined them under a light microscope. For each leaf, we randomly selected four fields of view at 1000x magnification and counted the number of stomata; we took the average of these four values to compute stomatal density (number of stomata per field of view).

| Data analysis
To included "group" (fixed effect) and "treatment" (fixed effect) as main effects, "group × treatment" as an interaction effect, and "block" as a random effect (LSMs for annual M. guttatus in the dry-down treatment could not be estimated from models including a nested "line" term due to small sample size of this group). Because it was measured only under well-watered conditions, the model estimating stomatal density included only "group" (fixed effect, "line" nested within it) and "block" as a random effect. These ANOVAs were run using the lmerTest package in R v. 3.2.3 using a Satterthwaite approximation to account for different variances among groups. We determined significance using a Bonferroni correction of α = .006 (to correct for multiple comparisons) and performed post hoc Tukey-Kramer HSD tests (p < .05) on all significant effects.
To investigate variation in seed set within groups, we used JMP 13.0 (SAS Institute) to perform a two-way ANOVA to calculate LSMs for each plant line; the model included "line" (fixed effect) and "treatment" (fixed effect) as main effects and a "line × treatment" interac-

| RE SULTS
Our simulated drought treatment had clear and consistently negative impacts on Mimulus growth and fitness, but the effects were not uniform across the three groups (perennial M. guttatus, annual M. guttatus, and annual M. nasutus). Rosette diameter, flower size (corolla width and length), and seed production were all strongly reduced under dry drown conditions (Table 2), but the extent of the reduction in flower size and seed production varied dramatically among groups (i.e., we observed significant "group × treatment" interactions in Table 3). As previously documented (Wu et al., 2010), perennial M. guttatus performed particularly poorly: none of the 36 plants exposed to drought-like conditions survived to produce any flowers ( TA B L E 3 Hierarchical ANOVA results for rosette diameter, corolla width, and seed set using a Satterthwaite approximation including "group" (fixed effect with "line" nested within it) and "treatment" (fixed effect) as main effects, "group × treatment" (interaction effect), "block," random effect. Significance determined using a Bonferroni correction of α = 0.006    Figure 4). In contrast, very few M. nasutus plants died after they had produced a mature flower, suggesting this species has diverged for traits that promote fruit maturation even under severe water limitation.
One key question is which phenotypes might explain this differ- nasutus under well-watered (black) and dry-down (gray) conditions. Days are numbered relative to Day 0, when seeds were transferred into growth chambers following stratification associated with higher water use efficiency) than leaves from M. guttatus (Table 2).
Restricting our focus to just the sympatric taxa at CAC, it is clear that interspecific differences in drought response become more pronounced later in the life cycle. Under the long days of our experiment, the two species' flowering phenologies were almost entirely overlapping, regardless of treatment ( Figure 5). The one exception to this pattern is that CAC M. guttatus budded slightly earlier (i.e., less than a day on average) under dry-down than under well-watered conditions (hazards ratio = 0.19, p = .0011; Figure 5). However, this very small head start in M. guttatus seems to have made little difference in terms of fitness: even the earliest flowering CAC M. guttatus usually died before making mature fruits or producing seeds (Figures 4 and 6). In CAC M. nasutus, on the other hand, drydown seed production was negatively correlated with flowering time (F = 6.14, p = .0166, Figure 6), showing that this species experiences selection for early flowering in water-limited environments.
In contrast to flowering time, we observed striking differences between CAC M. guttatus and M. nasutus in fruit maturation rates under dry-down conditions (Figure 4). Remarkably, M. nasutus fruits matured more than 12 days earlier under simulated drought than under well-watered conditions (hazards ratio = 12.50, p < .0001, Figure 5). The late-stage drought response in CAC M. guttatus was very different: among the few plants that survived to produce fruit, maturation occurred only one day earlier than among their well-watered counterparts (hazards ratio = 1.00, p < .0001, Figure 5). Taken  plastic drought response will require additional investigation, but F I G U R E 6 Under simulated drought, the effect of flowering time on seed production varies between sympatric Mimulus species. In a multiple linear regression (quadratic and cubic regression showed lower support), days to bud and plant line were significant predictors of seeds production in M. nasutus ("days to bud": F = 6.14, p = .0166; "line": F = 7.18, p = .0004; "days to bud × line": F = 0.61, p = .6145; F 7,49 = 4.95, p = .0003, R 2 = .41), but not in M. guttatus (F 19,129 = 0.64, p = .8708, R 2 = .09) it is likely a key component of species divergence in microhabitat adaptation at CAC.

| D ISCUSS I ON
Although an increase in developmental rate is a hallmark of the drought escape strategy (Ludlow, 1989) and/or additional environmental variables (e.g., greenhouse temperature, light intensity). Because of this earlier flowering, we also began our dry-down treatment nine days sooner than in Wu et al. (2010).  (Ivey & Carr, 2012;Wu et al., 2010), which might be due to CAC-specific traits. Alternatively, the difference might be explained by variation among experimental conditions: our dry-down treatment was applied earlier than that of Wu et al. (2010) and was likely more severe than the simulated drought used by Ivey & Carr (2012).  (Chaves, Maroco, & Pereira, 2003), but plants might also be able to avoid the negative consequences of drought by accumulating stores of nutrients when water is plentiful and/ or reallocating carbohydrate resources during initial water deficits (Kooyers, 2015). This adaptive response to drought has been well documented in cereal crops (Palta, Kobata, Turner, & Fillery, 1994;Schnyder, 1993;Yang, Zhang, Huang, Zhu, & Wang, 2000) and has also been observed in the Mediterranean annual Lupinus albus, which diverts resources from stems to seed pods as soon as it senses drought (Rodrigues, Pacheco, & Chaves, 1995). Going forward, if we are to achieve a more mechanistic understanding of divergence in drought response between CAC M. nasutus and M. gutattus, future experiments should investigate a more comprehensive set of physiological, leaf, and whole-plant traits.
In addition to elucidating the mechanisms of drought response within Mimulus species, our study provides important insight into the role of differential habitat adaptation in species divergence.
Our results suggest that a simple shift in critical photoperiod (from long-to short-day flowering) would be insufficient for CAC Mimulus to succeed in microhabitats that dry out sooner in the season. Soil moisture is highly heterogeneous at CAC and although short-day flowering might enable some plants to complete reproduction before they experience any water limitation, other individuals are likely to occupy patches that impose significant drought stress. As we have seen, CAC M. nasutus alone is able to cope with such conditions, surviving longer and speeding up its development to produce many more seeds than CAC M. guttatus. The picture emerging from this and previous studies (Ivey & Carr, 2012;Wu et al., 2010) is that habitat divergence between M. nasutus and M. guttatus is complex, involving many traits, both constitutive and plastic. Although some of the key traits involved might be genetically simple (e.g., critical photoperiod: Fishman et al., 2014), the microhabitat isolation we observe at CAC is likely to involve changes at many loci. M. guttatus flowers early, it is unable to overcome water deficits to set seed. This result might help explain why one of the two photoperiod loci remains highly divergent between species (Kenney & Sweigart, 2016), even in the face of considerable interspecific gene flow. Consistent with the idea that differentially adapted loci contribute to reproductive isolation between species, we find evidence of selection against M. nasutus ancestry across the M. guttatus genome at CAC (Brandvain et al., 2014;Kenney & Sweigart, 2016). Our work here sets the stage for future experiments to map the genetic basis of key ecological traits and fitness across the complex and variable environments of CAC, an approach that holds great promise for understanding how the process of abiotic adaptation can contribute to speciation.