**4.4. Tests of predictions #4 and #5**

Here the goal is to test whether TFs can be implicated in the tradeoff using TF-linked markers in the crosses, and then to conduct further genetic association analyses using markers within the implicated candidate TFs. Previous molecular work allows one to locate markers on a *B. stricta* linkage map that are linked to the candidate TFs in the crosses. Chromosomal painting and end sequencing have shown that there are large syntenic blocks that align between the *A. thaliana* and *B. stricta* genomes [41]. These Arabidopsis blocks have been located in the *B. stricta* genome. This can now also be verified with the recently sequenced *B. stricta* genome.

Analyses within the candidate genes could be conducted in populations of related and unrelated individuals. The analyses of related individuals could use F2 and F3 mapping populations, but there would be confounding effects of linkage and linkage disequilibrium (LD) from other unknown genes and alleles. Genotyping in the F2 generation also allows for genetic association analyses in the F3 field experiments. Significant associations using unre‐ lated individuals would help to eliminate confounding effects of linkage and linkage disequi‐ librium (LD) from other unknown genes and alleles. Another set of markers such as microsatellite markers (e.g., [29, 36]) would need to be used to identify and control population genetic structure among the unrelated individuals [33].

For marker development within genes of *B. stricta*, identification of genetic variation (e.g., single nucleotide polymorphisms – SNPs) in marker-implicated candidate TFs can be per‐ formed by first sequencing the TFs and their promoter regions from the diverged populations. Alignment of the diverged sequences would identify any polymorphisms. The recently sequenced *B. stricta* genome allows one to readily design primers to amplify and then sequence the genes.

To test whether polymorphic genes are under selection, statistical tests of synonymous to nonsynonymous substitution ratios in coding regions would be performed. Comparing dN/dS ratios allows for detection of positive or negative selection, albeit conservatively. The MacDonald-Kreitman test is more liberal for detecting deviation from the neutral model of molecular evolution [42].

### **4.5. Test of prediction #6**

To determine whether drought tolerance and defensive QTL co-localize on the *B. stricta* genome or whether there are epistatic interactions between QTL from different traits, linkage mapping would need to be conducted in F2 lab and F3 field experiments. Genotyping conducted in the F2 generation can also be used for successful mapping among F3 sib families [43]. Linkage analysis is often extended to incorporate information from several markers, called multi-point or interval mapping; therefore, composite interval mapping followed by multiple interval mapping using QTL-Cartographer will be used.

#### **4.6. Test of prediction #7**

If natural selection has acted upon genetic variation in signaling pathways for more stable or frequent expression of traits, this should be detectable in a comparison of the high- and lowelevation populations for trait responses and gene expression. For example, the populations could be compared in a double challenge experiment in which plants are first drought-stressed and then fed upon by generalist insect herbivores. One would predict that drought stress responses would take precedence over herbivore-induced defense responses in close wild relatives of Arabidopsis, reflecting the crosstalk between ABA and JA signaling that has been well documented in Arabidopsis. But we would predict that this plastic tradeoff would vary between high and low diverged populations and that this would be reflected in differential gene expression involving the candidate signaling pathways. To examine the expression levels of candidate TFs, one could use qPCR, but genome-wide gene expression using RNAseq would also allow the assessment of other genes as well.
