ABSTRACT
Environmental gradients can drive adaptive evolutionary shifts in plant resource allocation among growth, reproduction, and herbivore resistance. However, few studies have attempted to connect these adaptations to the underlying physiological and genetic mechanisms. Here, we evaluate potential mechanisms underlying a coordinated locally adaptive shift between growth, reproduction, and herbivore defense in the yellow monkeyflower, Mimulus guttatus. Through manipulative laboratory experiments we found that gibberellin (GA) growth hormones may play a role in the developmental divergence between coastal perennial and inland annual ecotypes of M. guttatus. Further, we detected an interaction between a locally adaptive chromosomal inversion, DIV1, and GA addition. This finding is consistent with the inversion contributing to the evolution of growth form via an interaction with the GA pathway. Finally, we found evidence that the DIV1 inversion is partially responsible for a coordinated shift in the divergence of growth, reproduction, and herbivore resistance traits between coastal perennial and inland annual M. guttatus. The inversion has already been established to have a substantial impact on the life-history shift between long-term growth and rapid reproduction in this system. Here, we demonstrate that the DIV1 inversion also has sizable impacts on both the total abundance and profile of phytochemical compounds involved in herbivore resistance.
INTRODUCTION
One of the fundamental tenants of evolutionary biology is that adaptation of organisms to specific environmental conditions inevitability results in a fitness trade-off. Trade-offs often manifest in the form of a cost, such that organisms that become adapted to one set of environmental conditions will be at a disadvantage in alternative environments (Futuyma and Moreno 1988; Whitlock 1996). The idea of trade-offs involved in adaptation and ecological specialization has been borne out in a wide range of evolutionary scenarios, including predator-prey relationships and host-races formation in insect herbivores (Futuyma and Moreno 1988; Kawecki 1998; Svanback and Eklov 2003; Forister et al. 2012). A common source of ecological specialization and consequent trade-offs is the process of local adaptation across environmental gradients (Kawecki and Ebert 2004; Hereford 2009).
Local adaption across environmental gradients can lead to shifts in the allocation of resources to long-term growth (survival) and reproduction (fecundity; Clausen and Hiesey 1958; Lowry 2012; Friedman and Rubin 2015). Those shifts in life-history strategy along environmental gradients can also have major impacts on allocation to herbivore defense (Hahn and Maron 2016). However, there appears to be key differences in how resources are allocated across environmental gradient for interspecific and intraspecific comparisons. Interspecific variation in plant species typically fits well with the resource allocation hypothesis (Coley et al. 1985), where low resource environments tend to be composed of slower growing better defended species while high resource environments promote faster growing poorly defended plants (Endara and Coley 2011). In contrast, intraspecific variation along environmental gradients is far less consistent and more often than not contradicts the predictions of the resource allocation hypothesis (Hahn and Maron 2016). One common pattern for intraspecific plant variation is a positive relationship between the length of the growing season along an environmental gradient and the level of herbivore resistance (Hahn and Maron 2016; Kooyers et al. 2017). This pattern could be driven by plants having more time to allocate resources to leaf production and defense, greater herbivore pressure in habitats with longer growing seasons, and/or a greater apparency of plants with longer growing seasons (Feeny 1976; Mason and Donovan 2015; Hahn and Maron 2016; Kooyers et al. 2017). Allocation to reproduction (fecundity) frequently trades off with constitutive and/or induced herbivore resistance (Agren and Schemske 1993; Heil and Baldwin 2002; Strauss et al. 2002; Stowe and Marquis 2011; Cipollini et al. 2014).
Achieving an evolutionary optimum in how resources are allocated to growth, reproduction, and defense will depend on the nature of all environmental challenges faced by each local population (Rhoades 1979; Rausher 1996; Hamilton et al. 2001; Strauss et al. 2002; Stamp 2003; Karban 2011; Cipollini et al. 2014; Jensen et al. 2015; Smilanich et al. 2016). Despite the development of multiple ecological and evolutionary hypotheses that postulate a relationship between growth, reproduction, and resistance to herbivores (Feeny 1976; Coley et al. 1985; Rhoades 1979; Herms and Mattson 1992; Strauss et al. 2002; Stamp 2003, Fine et al. 2006; Agrawal et al. 2010; Cipollini et al. 2014; Hahn and Maron 2016), these hypotheses do not make any predictions about the underlying molecular mechanisms that mediate these relationships. The genetic mechanisms responsible for trade-offs among growth, reproduction, and resistance are just beginning to be elucidated in model systems (Lorenzo et al. 2004; Yang et al. 2012; Kerwin et al. 2015; Campos et al. 2016; Havko et al. 2016; Major et al. 2017; Howe et al. 2018; Rasmann et al. 2018), but have yet to be evaluated in the evolutionary context of local adaptation.
Recent studies have shown that changes in the allocation of resources to growth versus resistance are made through a set of interacting gene networks (Karzan and Manners 2012; Huot et al. 2014; Campos et al. 2016; Havko et al. 2016). Jasmonates (JA) are key regulatory hormones involved in the response of plants to herbivore attack (Zhang and Turner 2008; Havko et al. 2016). While JA production increases herbivore defenses, it also inhibits plant growth through interactions with other gene networks (Zhang and Turner 2008; Yan et al. 2007; Karzan and Manners 2012; Yang et al. 2012). For example, the interactions of JAZ genes with DELLA genes in the signaling pathway of Gibberellin (GA) growth hormones are thought to play a key role in mediating resource allocation (Yang et al. 2012; Hou et al. 2013; Havko et al. 2016).
Here, we focus on understanding the physiological and genetic mechanisms underlying shifts in allocation to growth, reproduction, and defense for local adapted populations of the yellow monkeyflower Mimulus guttatus. The availability of soil water is a key driver of local adaptation in the M. guttatus species complex (Hall and Willis 2006; Lowry et al. 2008; Ferris et al. 2017). The coastal habitats of California and Oregon have many wet seeps and streams that are maintained year-round as a result of persistent summer oceanic fog and cool temperatures (Hall and Willis 2006; Lowry et al. 2008; Lowry and Willis 2010). All coastal populations of M. guttatus that reside in those habitats have a late-flowering life-history strategy. These coastal perennial populations thus, make a long-term investment in growth over reproduction (Hall and Willis 2006; Lowry et al. 2008; Hall et al. 2010; Baker and Diggle 2011; Baker et al. 2012). That investment in growth manifests through the production of many vegetative lateral stolons, adventitious roots, and leaves in coastal perennial plants. In contrast, the vast majority of nearby inland populations of M. guttatus in the coastal mountain ranges reside in habitats that dry out completely during summer months. These inland populations have evolved a rapid growth drought escape annual life-history strategy. Instead of investing in vegetative lateral stolons, the axillary branches of inland plants are mostly upcurved and typically produce flowers quickly (Lowry et al. 2008; Lowry and Willis 2010; Friedman et al. 2015). Further, inland plants invest less into the production of leaves before flowering than coastal perennials (Friedman et al. 2015). It should be noted that a smaller number of inland populations in the Coast Ranges do reside in rivers and perennials seeps and have a perennial life-history. Farther inland, perennial populations are more common, especially in high elevation streams and hot springs of the Sierra and Cascade Mountains (Oneal et al. 2014).
While perennial populations invest more into vegetative growth than reproductive growth, they also invest more heavily in defending their vegetative tissues. Perennial populations have higher levels of both constitutive and induced defensive phenylpropanoid glycoside (PPG) compounds than the annual populations when grown in a common environment (Holeski et al. 2013). This pattern of highly defended plants in wetter habitats with long growing seasons versus poorly defended plants in dry habitats with short growing seasons is consistent with an optimal defense strategy: Greater allocation of resources to herbivore resistance is favored in long growing season habitats by a greater abundance of herbivores and a lower cost of producing defensive compounds (Kooyers et al. 2017).
Two major QTLs (DIV1 and DIV2) and many minor QTLs control key traits involved in local adaptation to perennial and annual habitats within the M. guttatus species complex (Hall et al. 2006; Lowry and Willis 2010; Hall et al. 2010; Friedman and Willis 2013; Friedman et al. 2015). DIV1 has the largest effect on the most traits and has thus been more extensively studied than DIV2. DIV1 is a large paracentric chromosomal inversion that plays a pivotal role in the annual versus perennial life-history divergence described above (Lowry and Willis 2010). The inversion is at minimum 6.3 Mbp in length along linkage group 8 (LG8) and contains at least 785 annotated genes. In hybrids, DIV1 has a major effect on growth rate including the adaptive flowering time phenotype, explaining 21% to 48% of the divergence between inland annual and coastal perennial parents (Lowry and Willis 2010). In addition to flowering time, the DIV1 inversion has major effects on multiple traits involved in the evolutionary shift from more allocation of resources to long-term growth versus reproduction (Lowry and Willis 2010). These traits include the production of lateral stolons, adventitious roots, and leaf size (Lowry and Willis 2010; Friedman et al. 2015). Recent outlier analysis of coastal perennial and inland annual populations identified candidate genes in the gibberellin pathway that may underlie a pleiotropic shift in allocation between growth and reproduction (Gould et al. 2017).
In this study, we evaluate the role of GA in the divergence of growth morphology and herbivore resistance between perennial and annual ecotypes of M. guttatus. We then test whether there is an interaction between the DIV1 inversion and GA addition, which would be consistent with the inversion contributing to the evolution of shifts in allocation between long-term growth and reproduction via effects on the GA pathway. Finally, we examined whether the DIV1 inversion is responsible for the shift in allocation between reproduction and defense that has been broadly observed for populations of M. guttatus that vary in growing season length (Lowry et al. 2008; Holeski et al. 2013; Kooyers et al. 2017).
METHODS
Plant material
For comparisons among ecotypes, we utilized single family population accessions derived from five coastal perennial, four inland annual, and two inland perennial populations of M. guttatus (Fig. 1). The locations from where population accessions were collected are listed in Table S1. Previous population structure analyses found that coastal perennial populations of M. guttatus are more closely related to each other than they are to the inland populations (Lowry et al. 2008; Twyford and Friedman 2015). Thus, the coastal populations collectively constitute a distinct locally adapted ecotype (Lowry 2012). In contrast, population structure between inland annuals and inland perennial populations is generally low (Twyford and Friedman 2015). However, particular regions of the genome, including an adaptive chromosomal inversion (DIV1, discussed below) are more differentiated between inland annuals and perennials (Oneal et al. 2014; Twyford and Friedman 2015). We therefore consider inland annuals and inland perennials as different ecotypes as well.
To understand the phenotypic effects of the DIV1 inversion, Lowry and Willis (2010) previously created near-isogenic lines (NILs). The NILs are the product of crosses between inbred lines from the coastal perennial SWB population and the nearby inland annual LMC population. F1 hybrids were recurrently backcrossed to both of their respective parents for four generations. Heterozygous fourth generation backcrosses were then self-fertilized to produce two types of NILs: 1) Individuals that were homozygous for the introgressed allele of DIV1 (Introgression-NILs) and 2) Individuals that were homozygous for the DIV1 allele of the genetic background (henceforth referred to as Control-NILs). Comparisons between Introgression-NILs and Control-NILs are ideal for testing inversion function because their genetic backgrounds are nearly identical, but they are homozygous for opposite DIV1 alleles.
The effects of GA application on plant growth among ecotypes
To evaluate whether coastal perennial and inland annual plants differ in their response to GA addition, we conducted a greenhouse experiment with accessions derived from five coastal and four inland populations. Seeds were sown in Suremix soil (Michigan Grower Products Inc., Galsburg, MI) and stratified at 4° C for two weeks. After stratification, pots were moved to the Michigan State University Greenhouses. Temperature was set in the greenhouse room to 22° C days/18° C nights. Plants were grown in 16-hour days and 8-hour nights, where supplemental lighting was used during the full day period. Seedlings were transplanted to individual 3.5-inch square pots filled with Suremix soil. Transplanted seedlings were randomized across the experiment and randomly assigned to a GA treatment group or a mock control group. After transplantation, plants were sprayed five times each, every other day, with 100mM GA3 (GA treatment) or DI water (mock). This spray volume amounts to ∼3.5 mL of volume.
To standardize the developmental time point at which plant traits were quantified, we measured the following traits on the day of first flowering (anthesis): Total number of nodes on the primary shoot, lengths and widths of the first three internodes, length and width of the corolla of the first open flower, plant height, the total number of adventitious roots at the first node of all branches, total number of stoloniferous nodes sprouting adventitious roots, total number of aerial branches, total number of stolons, length of the longest aerial branch, length of the longest stolon, and the length and width of the longest leaf at the second node. Ten days after first flower, we quantified the same traits as at first flower, with the following exceptions: Length and width of corollas were not quantified, but we did count the total number of flowers.
Results were analyzed with JMP 12.2.0 (SAS Institute, Cary, NC). To gain a general understanding of the effect of GA addition on coastal perennial, inland annual, and inland perennial ecotypes, we conducted principle components analyses using all traits measured 10 days after flowering plus the width and length of corollas measured at flowering. We did not use the other traits measured at first flower to avoid including repeated measures in the principle components analysis. We saved the first three PCs from the analysis of all individuals. To understand the effects of accession, ecotype, and GA treatment on PCs, we fit standard least squares models. Each of the three PCs were modeled as response variables to the following factors and interactions: accession (nested within ecotype), ecotype (coastal perennial, inland annual, inland perennial), treatment, accession x treatment, and ecotype x treatment. Following our PC analysis, we fit the same model for individual traits.
Interactions of GA application with the adaptive DIV1 inversion
We grew coastal (S1) and inland (L1) parental inbred lines along with the NILs derived from those lines in a fully randomized design in the Michigan State University greenhouses, with 16-hours of supplemental lighting. We focused on the effect of the inversion in the coastal perennial genetic background, as a previous study had shown that the effect of the inversion had the greatest effect in the perennial genetic background (Friedman 2014). Following transplantation, seedlings were sprayed with GA or a mock water treatment every other day and traits were quantified in the same way as for the comparing population accessions.
To establish how trait variation of the coastal and inland parental lines was influenced by the GA treatment, we conducted a PC analysis with the parental lines. As for the analysis with multiple population accessions (above), we conducted principle components analyses using all traits measured 10 days after flowering plus the width and length of corollas measured at flowering. Models were fit with the first three PCs and individual traits as response variables to the following factors: line, treatment, and the line x treatment interaction. Following analysis of the parental lines, we analyzed the effects of the inversion in the perennial genetic background NILs. As above, we first conducted a PC-based analysis and saved the first three PCs. We then fit the first three PCs and individual traits as response variables for the following factors: inversion type, treatment, and the inversion type x treatment interaction.
Effects of the DIV1 inversion on resistance compound concentrations
To evaluate the effects of the DIV1 inversion on the production of herbivore resistance compounds, we conducted an experiment using the inversion NILs. Seeds were stratified, germinated, and transplanted following the same protocols as in the previous two experiments. In contrast to the GA NIL experiment, we grew outbred NILs which were created by intercrossing independently derived NILs. As for the GA experiment, we focused our study on the effect of the inversion in the perennial genetic background. The details of how outbred NILs were generated can be found in Lowry and Willis (2010). Here, we used the outbred NILs made by intercrossing S2 and S3 derived coastal genetic background NILs. We used intercrosses between L2 and L3 for the inland parents and between S2 and S3 for the coastal perennial parent comparisons.
To ensure that enough leaf tissue was available for analyses, outbred NILs and outbred parents were allowed to flower prior to the collections for PPG quantification. Collected leaf tissue was lyophilized for two days and then shipped to Northern Arizona University for analyses. We ground the leaf tissue using a 1600 MiniG (Spex, Metuchen, New Jersey). Extractions were conducted in methanol, as described in Holeski et al. 2013, 2014. We quantified PPGs using high performance liquid chromatography (HPLC), via an Agilent 1260 HPLC (Agilent Technologies, Santa Clara, California) with a diode array detector and Poroshell 120 EC-C18 analytical column (4.6 x 250mm, 2.7µm particle size) maintained at 30°C. HPLC run conditions, as were described in Kooyers et al. (2017). We calculated concentrations of PPGs as verbascoside equivalents, using a standard verbascoside solution (Santa Cruz Biotechnology, Dallas, Texas), as described in Holeski et al. (2013, 2014). We compared the concentrations of total PPGs and individual PPGs with one-way ANOVAs fit in JMP 12.2.0. Post-hoc Tukey Tests were used to compare means of parental and NIL classes.
RESULTS
The effects of GA application on plant growth among ecotypes
Consistent with previous studies (Hall et al. 2006; Lowry et al. 2008; Lowry and Willis 2010; Oneal et al. 2014), there were large differences in morphology between coastal perennial, inland annual, and inland perennial ecotypes. There were also differences between the inland perennial ecotype and the two other ecotypes. Most traits (18 out of 20) heavily loaded (Loadings > 0.40) onto the first PC (Eigenvalue = 8.258; Table S2). The ecotype effect was highly significant for PC1 (F2,226 = 333.64, P < 0.0001; Table 1). Within ecotype, there was a significant effect of accession on PC1 (F8,226 = 20.12, P < 0.0001), and there was a significant effect of the GA treatment on PC1 (F1,226 = 25.75, P < 0.0001). While there was a significant accession x treatment effect on PC1 (F8,226 = 3.36, P = 0.0012), the treatment x ecotype effect was not significant (P > 0.05). In contrast to PC1, both of the interactions were significant for PC2 (Eigenvalue = 3.463; accession x treatment: F8,226 = 8.18, P < 0.0001; ecotype x treatment: F2,226 = 6.20, P = 0.0024; Fig. 2A) and PC3 (Eigenvalue = 2.01; accession x treatment: F8,226 = 2.92; P = 0.0040, ecotype x treatment: F2,226 = 41.64, P < 0.0001; Fig. 2B). Second and third internode length, plant height, the total number of adventitious roots at the first nodes, total number of stoloniferous nodes sprouting adventitious roots, total number of aerial branches, total number of stolons, and the length of the longest aerial branch all heavily loaded (> 0.40) onto PC2 (Table S2). Third internode length, total number of aerial branches, total number of stolons, and leaf width heavily loaded onto PC3 (Table S2).
The results of the individual traits analyses were largely consistent with the principle components analyses. Ecotype had a significant effect on all traits except for third internode length. There was a significant accession effect for every trait that we measured in the experiment. The GA treatment had a significant effect on 16 out of 19 traits at first flower and 12 out of 18 traits measured at 10 days after first flower (Table 1). The most pronounced changes of the plants in response to the GA treatment was an increase in plant height and a conversion of lateral branches from adventitious root-making stolon branches into unpcurved aerial branches (Fig. 2C). As the perennials typically have more stolons with adventitious roots than the annuals, they were generally more obviously affected by the GA treatment. This effect was captured through the significant ecotype x treatment interactions on plant height. It should be noted that the effect of GA varied across accessions within ecotype. The coastal PGR accession is the tallest coastal accession with the fewest stolons (Table 1; Fig. 2C). Thus, it was affected the least by the GA treatment. In contrast, OPB is a short prostrate coastal accession and was dramatically affected by the GA treatment (Fig. 2C). Among the inland annual accession, SWC showed the greatest response to the GA treatment in terms of height and aerial branch formation.
Interactions of GA application with the adaptive DIV1 inversion
Consistent with previous studies (Lowry and Willis 2010; Friedman 2014), the coastal perennial (SWB S1) and inland annual (LMC L1) lines were highly divergent in morphological traits and the two lines were differentiated strongly along PC1 (F1,168 = 416.81, P < 0.0001). Similar to the accession analyses (above), the line x treatment interaction was not significant for PC1, but was highly significant for PC2 (F1,168 = 85.57, P < 0.0001) and PC3 (F1,168 = 13.54, P < 0.0001). The line x treatment interactions was also significant for 16 out of the 19 traits measured at flowering and 12 out of the 18 traits measured 10 days after flowering. Overall, the coastal perennial line (S1) responded more strongly to GA treatment, just as we found across coastal and inland populations more generally (above).
The DIV1 chromosomal inversion is one of many loci responsible for divergence between the annual and perennial ecotypes. Thus, main effects and interactions in the NILs were expected to be subtler than for the parental lines. As in previous studies (Lowry and Willis 2010; Friedman 2014), the inversion had significant effects on morphology, with highly significant main effects on PC1 (F1,173 = 35.44, P < 0.0001; Table 2, S3) and PC3 (F1,173 = 37.55, P < 0.0001). The GA treatment had significant effects on PC1 (F1,173 = 16.76, P < 0.0001) and PC2 (F1,173 = 489.50, P < 0.0001). There were weak, but significant, line x treatment interactions for PC2 (F1,173 = 4.13, P = 0.0437; Fig. 3A) and PC3 (F1,173 = 4.17, P = 0.0427). While the line x treatment effect on the morphological PCs was marginal, the effect was greater for some of the individual traits (Fig. 3; Table 2).
Effects of the DIV1 inversion on resistance compound concentrations
We quantified the concentrations of seven PPGs (Table 3). Consistent with our previous observations (Holeski et al. 2013), the coastal perennial parental (SWB) plants produced 2.5 times more total PPGs than the inland annual parental (LMC) plants (F1,31 = 51.03; P < 0.0001; Table 3; Fig. 4). There were also significant differences for six out of seven of the PPGs between the coastal perennial (SWB) and inland annual (LMC) parental lines.
Analysis of the DIV1 NILs revealed that the introgressed region containing the inversion had major effects on foliar concentrations of PPGs. Control NILs that were homozygous for the coastal orientation of the DIV1 inversion produced 35% higher concentrations of total PPGs than the introgression NILs, which were homozygous for the inland DIV1 orientation (F1,87 = 22.70; P < 0.0001). In addition, the DIV1 locus had significant effects on four out of the seven individual PPGs. Interestingly, the control NILs had higher concentrations of conandroside and mimuloside, but lower concentrations of calceolarioside A and unknown PPG10, than the introgression NILs (Table 3). Thus, the DIV1 inversion influences both the total concentration of PPGs as well as the composition of suites of these PPGs.
DISCUSSION
In this study, we identified a potential genetic mechanism underlying a coordinated evolutionary shift between growth, reproduction, and herbivore resistance in the M. guttatus species complex. We found that GA has the potential to play a role in the divergence between coastal perennial and inland annual ecotypes of M. guttatus, with coastal and inland plants responding differently to the addition of GA for a number of phenotypic traits associated with shifts between long-term growth and reproduction. Further, we detected an interaction between the locally adaptive DIV1 inversion and GA addition, which is consistent with the inversion contributing to the evolution of growth form by modulating the GA pathway. Finally, we found evidence that the DIV1 inversion contributes to the trade-off between growth, reproduction and resistance. The coastal orientation of the DIV1 inversion causes plants to allocate more to long-term growth and herbivore resistance over rapid reproduction than for the inland inversion orientation. We discuss these findings in the context of the broader literature below.
Environmental gradients and the evolution of growth, reproduction, and defense traits
Studies of intraspecific variation among natural populations adapted to different soil water availability regimes provide an excellent opportunity to understand how the abiotic environment influences the relative allocation of resources by plants to growth and constitutive/induced resistance. Soil water is one of the most limiting factors for plants on Earth (Whittaker 1975; Bohnert et al. 1995; Bray 1997) and can drastically differ in availability among seasons (Cowling et al. 1996), which in turn influences plant resource allocation (Juenger 2013). The timing of soil water availability can dictate the length of the growing season. One major evolutionary strategy for seasonally low water availability is to allocate resources primarily to growth and reproduction to achieve an early flowering, drought escape life-history (Ludlow 1989; Juenger 2013; Kooyers 2015). Beyond selection on plants, soil moisture gradients can drive the abundance of herbivores, which in turn exert their own selective pressures (Kooyers et al. 2017).
In M. gutattus, evolutionary shifts across a soil moisture gradient drives changes in the allocation not only between growth and reproduction, but also for herbivore resistance (Lowry et al. 2008; Holeski et al. 2013; Kooyers et al. 2017). The phenotypic differences between coastal perennial and inland annual populations is likely driven by multiple selective pressures that are tied to the soil water availability gradient between coastal and inland habitats. Inland annual habitats generally dry out very quickly at the end of the spring, which leaves little time for a plant to reproduce before being killed by the summer drought. Further, the short growing season may also prevent the growth of sizable herbivore populations, which would explain the low level of leaf damage in fast drying inland annual habitats (Lowry et al. 2008; Kooyers et al. 2017). In contrast, the year-round soil water availability of coastal habitats means that plants growing there have much more time to allocate resources to vegetative growth and herbivore resistance. In addition, wet coastal habitats can build up a considerable load of herbivores, which is likely reflected by much greater leaf damage and early season mortality in these habitats (Lowry et al. 2008; Lowry and Willis 2010). For intraspecific differences among populations, the strength of herbivore pressure is thought to be a key driver of plant resistance (Hahn and Maron 2016). It should be noted that a fair amount of leaf damage in coastal habitats may also be due to oceanic salt spray (Boyce 1954; Ahmed and Wainwright 1976; Griffiths 2006; Lowry et al. 2009). Future manipulative field experiments are needed to partition out the relative contributions of herbivory and salt spray to leaf damage for M. guttatus in coastal habitats.
Our findings in M. guttatus are likely to have implications for intraspecific variation in many other plant species as well. There are many studies that have found similar developmental differences between coastal and inland populations as we have found for M. guttatus (reviewed in Lowry 2012). Given the commonality of coastal plants investing more heavily in lateral vegetative branches versus inland populations investing primarily in upright flowering branches, we predict that coastal population of plants will generally be more highly defended than inland populations, particularly in Mediterranean climates with steep soil moisture gradients.
The role of pleiotropy and linkage
The results of this study and previous studies (Lowry and Willis 2010; Friedman 2014; Friedman et al. 2015) collectively demonstrate that adaptive chromosomal inversion DIV1 contributes to the shift in allocation between long-term growth, short-term fecundity, and herbivore resistance. An outstanding question is whether this coordinated shift is due to genetic linkage or pleiotropy. Chromosomal inversions are thought to evolve as adaptation “supergenes,” which can trap multiple linked adaptive loci through their suppression of meiotic recombination (Dobzhansky 1970; Kirkpatrick and Barton 2006; Schwander et al. 2014; Wellenreuther and Bernatchez 2018). Thus, the fact that the DIV1 inversion contributes to the evolution of multiple phenotypes could be result of adaptive alleles at multiple linked loci being held together in tight linkage by the chromosomal inversion. Alternatively, a single gene within the inversion could have pleiotropic effects on all of the phenotypic changes.
Two other recent studies have also found potential pleiotropic effects of genes on allocation to reproduction and herbivore resistance. Rasmann et al. (2018) found that NILs containing genetic variants of the Flowering Locus C (FLC) gene in Cardamine hirsute are responsible for a trade-off between early flowering and herbivore resistance in terms of glucosinolate production. Kerwin et al. (2015) found that there was a positive correlation in Arabidopsis thaliana between glucosinalate production and flowering time for mutant alleles of genes in the glucosinolate biosynthetic pathway. Overall, both of these studies identified the same trade-off of rapid reproduction versus herbivore resistance that we found in our study, although mediated through independent genetic mechanisms.
A hormonal basis of a coordinated shift in the evolution of growth, reproduction, and herbivore resistance?
The finding that coastal perennial and inland annual plants respond differentially to GA is consistent with the role of this hormone playing a role in their evolutionary divergence. A recent outlier analysis of coastal perennial and inland annual populations of M. gutattus found that the gene GA20-oxidase2 (GA20ox2) was a major allele frequency outlier between the ecotypes within the vicinity of the DIV1 inversion (Gould et al. 2017). This gene is a strong potential candidate gene for a pleiotropic shift in allocation between growth and reproduction. GA-oxidases are involved in the evolution of dwarfed coastal populations of A. thaliana (Barboza et al. 2013) and played a key role in the development of dwarfed Green Revolution rice and barley (Sasaki et al. 2002; Jia et al. 2009). Further, the DELLA gene, GAI, is located within the vicinity of another growth regulatory QTL (DIV2). GAI was also an allele frequency outlier between coastal and inland populations (Gould et al. 2017). Friedman (2014) found that the DIV1 and DIV2 loci interact epistatically. Thus, it would not be surprising if the genetic changes that underlie both QTL are in the same molecular pathway. Further, the negative antagonism between the GA and JA hormone pathways via the DELLA-JAZ signaling node (Havko et al. 2016; Guo et al. 2018; Howe et al. 2018) suggests a direct mechanism by which trade-offs between growth, reproduction, and resistance could easily evolve. Future functional genetic studies will be needed to determine whether these genes in fact are involved in adaptive shifts between growth, reproduction, and defense underlying local adaptation in this system.
While we saw a greater response to GA addition in coastal plants and observed an interaction between GA addition and the inversion, other hormones could also play a role or even be the ultimate cause of the divergence between the coastal perennial and inland annual ecotypes. Three major classes of hormones, Auxins, Brassinosteroids and Gibberellins, are all associated with growth phenotypes like those that differ between coastal perennial and inland annual ecotypes of M. guttatus (Ross and Quittenden 2016; Unterholzner et al. 2016). These hormones interact in multiple ways, which have yet to be fully elucidated, to result in shifts in growth/reproduction phenotypes. Future functional studies in M. guttatus will be needed to identify the ultimate causative mechanisms underlying the dramatic shift in allocation between annual and perennial populations.
Conclusions and future directions
There are numerous evolutionary and ecological models that make predictions on the evolution of relationships between growth, reproduction, and herbivore resistance. While recent meta-analyses have found that some models have moderate support (Endara and Coley 2011), there are numerous exceptions and many models appear to not be well supported at all (Stamps et al. 2003; Hahn and Maron 2016; Smilanich et al. 2016). The reasons that these models do not hold up are often attributed to vast variation in the extrinsic environmental factors that exert selective pressures on plant populations, broad variation in life-history among plant species, and differences between interspecific and intraspecific variation (Stamps et al. 2003; Hahn and Maron 2016; Smilanich et al. 2016). Less well appreciated are the molecular genetic mechanisms that underlie shifts in allocation between growth, reproduction, and herbivore resistance (Kerwin et al. 2015; Rasmann et al. 2018). The nature of the gene networks responsible for shifts in allocation may also be very important for whether or not particular systems will conform to a given evolutionary or ecological model. Future research should focus on uncovering the molecular mechanisms that underlie the evolution of growth, reproduction, and defense trade-offs in natural populations and integrate predictions from those mechanisms into ecological models.
AKNOWLEGEMENT
We would like to thank Sol Chavez for assisting with the quantification of PPGs. Seed collections were originally made possible by permission from the state parks of Oregon and California. Funding for this research was provided by Michigan State University through a startup package to DBL.