**4. Approach and Discussion**

An Ecological Genomics approach would satisfy the need for experimental and molecular genetics to evaluate the hypothesis in several of the predictions. Ecological Genomics (EG) is an interdisciplinary approach in the biological sciences that seeks to find the genes underlying species interactions in natural habitats and to study the evolutionary forces that have shaped these genes, their expression patterns, and the phenotypes that they encode [22-26]. As such, nucleic acid sequencing, forward and reverse genetic tools, comparative methods, and other molecular techniques are required to find the relevant genes, to establish databases of candi‐ date genes, and to study their expression patterns and allelic variants.

tolerance. Candidate genes in Arabidopsis for the evolutionary constraint include specific TFs that are involved in gene regulation within and among signaling pathways (Figure 1). For example, if there exists genetic variation in TFs MYC2 or AF2 that natural or artificial selection could act upon to increase drought stress tolerance traits for more stable expression, we predict that [1] a glucosinolate defense response would also change because these TFs help to regulate

**1.** Defense and stress tolerance phenotypes will be genetically correlated in family structured

**2.** Genetically diverged populations from different elevations will also be diverged for defense and abiotic stress tolerance traits; neither population will have high values of both

**3.** Defense allocation and stress tolerance phenotypes will not segregate independently from

**4.** Defense and stress tolerance genetic covariation will be associated with markers linked to candidate transcription factors (TFs) that regulate both defense and drought tolerance pathways (AtMYC2 [At1g32640], AtMYB2 [At2g47190], and a NAC TF AtAF2

**5.** DNA sequence variation of candidate regulatory genes (TFs) implicated in prediction #3 will also correlate with the defense and stress tolerance tradeoff in (a) the segregating crosses and in (b) unrelated individuals (linkage disequilibrium association analysis). Comparison of DNA sequence of coding regions will also show molecular evidence for

**6.** There may be other regions of the genome besides those containing the candidate TFs that also simultaneously affect defense and stress tolerance phenotypes and thus would also provide molecular evidence for the tradeoff. For whole genome marker analysis away from candidate TFs, defense and stress tolerance phenotypes will either co-locate in

**7.** Gene expression of candidate genes in genetically diverged populations will reflect the evolution of more stable expression of defense and stress tolerance response pathways.

An Ecological Genomics approach would satisfy the need for experimental and molecular genetics to evaluate the hypothesis in several of the predictions. Ecological Genomics (EG) is an interdisciplinary approach in the biological sciences that seeks to find the genes underlying

one another in extended generation crosses between diverged populations.

evolution (McDonald-Kreitman Test and dS/dN ratios).

linkage mapping or their QTL will show negative epistasis.

both pathways, and [2] other pleiotropic or epistatic effects would reduce fitness.

**3. Predictions**

common garden experiments.

692 Abiotic and Biotic Stress in Plants - Recent Advances and Future Perspectives

traits simultaneously.

[At5g08790]).

**4. Approach and Discussion**

Although many model species (species whose genomes have been completely sequenced and that have broad interest from molecular biologists) meet and exceed these molecular criteria, often they lack the attributes for ecological studies. For example, the natural distribution of *A. thaliana* is in Eurasia, making it difficult for study by North American researchers. Further‐ more, there are still relatively few model species. With the advent of affordable high-through‐ put next-generation sequencing (NGS), however, sequence information is becoming available for more species and populations. As of June 2015, for example, there were 6,653 completely sequenced genomes, but only a small fraction of these are eukaryotes (http://www.genome‐ sonline.org). The GOLD website also listed 60,631 genome sequencing projects (June, 2015), but only 9,059 or 15% were eukaryotes. Thus, although NGS is helping the field of EG to move away from model species, especially, for example, in evolutionary studies [27], there is still a need to understand the attributes that constitute an ideal model organism for EG studies.

Feder and Mitchell-Olds [22] listed the criteria for an ideal model species in Evolutionary and Ecological Functional Genomics (EEFG), which is synonymous with EG. The criteria stated that there needs to be: (1) a co-operative community of researchers from different disciplines that share resources and information; (2) the tools to find genes and study their variation within and among species; (3) natural, undisturbed habitats such that genetically diverged popula‐ tions can be studied for local adaptation; (4) molecular data on sequences and chromosomal maps for marker development and mapping, cis and trans regulatory regions identified, and gene function and its fitness consequences known under natural conditions; and, finally, (5) the ability to study the ecological consequences of natural genetic sequence variation in the genes for evolutionary inferences. Thus, access to NGS alone does not necessarily make a species ideally suited for EG study.

*Boechera stricta*, a close wild relative of Arabidopsis, satisfies many of these criteria and is an emerging ecological model species that inhabits environments that differ substantially in drought stress, herbivore community, and other abiotic and biotic conditions [28]. The selfing rate of *B. stricta* in the northern portion of its geographic range is 0.95 [29], enabling the creation of experimental advanced generation hybrids for forward ecological genetic studies (e.g., [30]). The genome of *B. stricta* has also been recently sequenced (http://phytozome.jgi.doe.gov/pz/ portal.html#!info?alias=Org\_Bstricta).

Forward genetics for finding candidate genes in ecology include population genomics, association mapping, linkage mapping, and transcriptomics [31, 32]. Population genomics identifies outlier marker loci in the statistical analysis of population genetic parameters, but there is no knowledge of associated phenotypes. Association and linkage mapping both include measurement of phenotypes, but a distinct advantage of association mapping is that no pedigree is required and that allelic variants representative of naturally occurring popula‐ tions can be used in analyses [26, 33]. Association mapping can be conducted on unrelated individuals because it is based on general inherent linkage disequilibrium (LD). When markers for association analysis are developed from candidate genes, significant associations may actually identify relevant genes. We assume that candidate genes in Arabidopsis could often be studied with success under natural conditions in close relatives in the genus *Boechera*. Unidentified causal genes in LD with significant candidate markers are unlikely in *B. stricta* where LD decays rapidly, within 10 kb [34]. If present, population substructure (i.e., stratifi‐ cation, admixture, or inbreeding) must be controlled for genetic association analyses because of confounding effects on LD that can lead to false positives [33].

#### **4.1. Tests of prediction #1**

Using family structured quantitative genetic analyses, previous studies have examined the genetic correlation between defense and stress tolerance of prediction #1. This work has been conducted on a close wild relative of Arabidopsis in the genus *Boechera*. *Boechera stricta* is a genetically diverse, diploid, predominantly self-fertilizing species that occurs at higher elevations throughout western North America in natural habitats [29, 34]. The phenotype of an individual of any species is determined by genetic and environmental factors (P = G + E), and these factors and thus the phenotypes vary among individuals within populations (VP = VG + VE). If the phenotypes are measured from in a common garden experiment, environmental variation is eliminated so that the phenotypic and genetic variations are equal (VP = VG). Among full-siblings of clonal or self-fertilizing species such as *B. stricta*, the genetic variation (VG) measured from a common environment can be used for evolutionary inferences because all allele combinations within and among loci are inherited without change. In accordance with prediction #1, the negative genetic correlation between glucosinolate (GS) toxin defense allocation and stress tolerance associated with range limits of *B. stricta* has been observed using these methods in the field and lab five times previously [2, 35-37]. Stressors involved in the tradeoff have included drought, nutrient deficiency, and change in plant community structure across the range boundary (suggesting competition) and the multivariate stress of the range boundary itself. In these studies, it was hypothesized that these tradeoffs occurred because of antagonistic crosstalk between abscisic acid (ABA) stress tolerance and jasmonic acid/ethylene (JA/ET) or salicylic acid (SA) defense-signaling pathways. Circumstantial evidence implicating the pathways in the tradeoff comes from experimental ABA soil inoculations that depended on endogenous GS level [35]. These conclusions are based mainly on correlative methods and circumstantial evidence. Experimental and molecular genetics and more direct measures of pathway components are needed.

#### **4.2. Tests of prediction #2**

For gene mapping in *B. stricta*, a replicated cross has been conducted between populations from the Big Horn Mountains, Wyoming, and the Black Hills, South Dakota. These are geographically isolated and genetically diverged populations [36, 38]. The populations are located at different ends of the altitudinal range of *B. stricta* (Big Horns 3,000 m, Black Hills 1,700m) and thus the sites differ by several environmental factors. The populations have diverged for glucosinolate content and stress tolerance traits such as root:shoot ratio as predicted; neither population had high values of both traits (Figure 2).

**Figure 2.** Genetic variation in GS levels (F18, 701 = 7.101, P < 0.001) (left) and root:shoot ratios (F5, 251 = 3.576, P = 0.004) (right) between genetically diverged populations. Data are from a growth chamber experiment. Residuals controlled for plant size, thus GS levels reflect content.

#### **4.3. Tests of prediction #3**

individuals because it is based on general inherent linkage disequilibrium (LD). When markers for association analysis are developed from candidate genes, significant associations may actually identify relevant genes. We assume that candidate genes in Arabidopsis could often be studied with success under natural conditions in close relatives in the genus *Boechera*. Unidentified causal genes in LD with significant candidate markers are unlikely in *B. stricta* where LD decays rapidly, within 10 kb [34]. If present, population substructure (i.e., stratifi‐ cation, admixture, or inbreeding) must be controlled for genetic association analyses because

Using family structured quantitative genetic analyses, previous studies have examined the genetic correlation between defense and stress tolerance of prediction #1. This work has been conducted on a close wild relative of Arabidopsis in the genus *Boechera*. *Boechera stricta* is a genetically diverse, diploid, predominantly self-fertilizing species that occurs at higher elevations throughout western North America in natural habitats [29, 34]. The phenotype of an individual of any species is determined by genetic and environmental factors (P = G + E), and these factors and thus the phenotypes vary among individuals within populations (VP = VG + VE). If the phenotypes are measured from in a common garden experiment, environmental variation is eliminated so that the phenotypic and genetic variations are equal (VP = VG). Among full-siblings of clonal or self-fertilizing species such as *B. stricta*, the genetic variation (VG) measured from a common environment can be used for evolutionary inferences because all allele combinations within and among loci are inherited without change. In accordance with prediction #1, the negative genetic correlation between glucosinolate (GS) toxin defense allocation and stress tolerance associated with range limits of *B. stricta* has been observed using these methods in the field and lab five times previously [2, 35-37]. Stressors involved in the tradeoff have included drought, nutrient deficiency, and change in plant community structure across the range boundary (suggesting competition) and the multivariate stress of the range boundary itself. In these studies, it was hypothesized that these tradeoffs occurred because of antagonistic crosstalk between abscisic acid (ABA) stress tolerance and jasmonic acid/ethylene (JA/ET) or salicylic acid (SA) defense-signaling pathways. Circumstantial evidence implicating the pathways in the tradeoff comes from experimental ABA soil inoculations that depended on endogenous GS level [35]. These conclusions are based mainly on correlative methods and circumstantial evidence. Experimental and molecular genetics and more direct measures of

For gene mapping in *B. stricta*, a replicated cross has been conducted between populations from the Big Horn Mountains, Wyoming, and the Black Hills, South Dakota. These are geographically isolated and genetically diverged populations [36, 38]. The populations are located at different ends of the altitudinal range of *B. stricta* (Big Horns 3,000 m, Black Hills 1,700m) and thus the sites differ by several environmental factors. The populations have diverged for glucosinolate content and stress tolerance traits such as root:shoot ratio as

predicted; neither population had high values of both traits (Figure 2).

of confounding effects on LD that can lead to false positives [33].

694 Abiotic and Biotic Stress in Plants - Recent Advances and Future Perspectives

**4.1. Tests of prediction #1**

pathway components are needed.

**4.2. Tests of prediction #2**

The segregating F2 populations from the crosses can be used to test prediction #3 that the defense and stress tolerance traits will not be inherited independently of one another. F2s from more than one cross allow for broader inference and increased statistical power [39]. Because each genotype in the F2 segregating generation cannot be replicated, drought treatments in the lab would need to be imposed on all plants after the plants were first monitored for performance (e.g., growth) under controlled watering conditions. Measures of drought tolerance could include relative growth (before and after drought treatments), leaf mass area (LMA), water use efficiency (carbon isotope ratio), and root:shoot ratio. Several measures of stress tolerance increase the probability of detecting the tradeoff, and provide a general assessment of stress tolerance [40]. Glucosinolate analysis should be conducted on leaf tissue at the time of drought stress.

Of course, the goal should be to test for the tradeoff in the field across the low-elevation range boundary using the segregating generation of the experimental crosses. However, this requires replication of each F2 genotype that could be compared within and just outside the range boundary. In the F3 generation, there is replication of each F2 lineage. Common garden experiments of F3 families could be established within and across low-elevation range boundaries. Previous field transplant experiments [2, 35, 36] have established the areas just across the low-elevation range boundary as stressful in terms of several correlated abiotic and biotic stressors, which manifest as slower growth and survivorship. Large sample sizes would guarantee the detection of effects if they exist, and would also allow for assessment across multiple years and the possibility of ecological gradient manipulations. For example, removal treatments of candidate competitors, such as *Lithophragma* parviflorum [36], could be included within and outside the range. Competitive interactions also induces ABA [36] and therefore might also induce the tradeoff.
