Fig. 3: Identification of genes associated with heat tolerance in common garden and wild-caught lizards.

A One hundred and thirty lizards from heterogeneous treatments (114 wild-caught, 16 common garden) were used to identify co-expression modules of the Anolis skeletal muscle transcriptome. Weighted gene correlation network analysis (WGCNA) revealed seven regulatory modules that define the full regulatory architecture of the skeletal muscle transcriptome: Module 1 (black), Module 2 (blue), Module 3 (brown), Module 4 (green), Module 5 (red), Module 6 (turquoise), and Module 8 (yellow)⁶⁰. Bars represent the number of genes within each module. Inset is a representation of connectivity among co-expression modules. Circle sizes represent the relative size of each module (circle area is proportional to log-transformed number of genes within each module) and line thickness represents the relative strength of connectivity between modules (absolute value of pairwise Pearson’s correlation among eigengene values). B Violin plot of gene significance scores (GS) for CT_MAX (strength of association between gene expression and CT_MAX). Dots indicate individual genes assigned to each module. Colored dots within each module indicate genes with significant phenotypic correlations after multiple testing correction (GS.q < 0.05). This subset of genes made up the focal data set of the current study. C To validate the candidate gene set, two additional network analyses were conducted for wild-caught (n = 114) and common garden lizards (n = 16), separately. Gene significance (GS) scores were calculated for candidate genes independently for each group. Genes identified as negatively correlated with CT_MAX (negative regulators) in the full data set are indicated in blue; genes positively correlated with CT_MAX in the full data set (positive regulators) are indicated in red. Black dots in the center of each violin plot are GS scores for each individual gene. Gray lines present the directionality of GS score change for each gene between common garden and wild-caught animals. There was a significant expression-phenotype correlation between wild-caught and common garden data (linear model, R²: 0.601; p « 0.001). Relative rank of gene significance was highly conserved across sets (Spearman’s rank correlation ρ: 0.75, p « 0.001). Negative regulators displayed no bias in gene significance between groups (paired t-test, p = 0.26). Positive regulators displayed higher gene significance values in the common garden data set than the wild-caught data set (paired t-test, p « 0.001). The gene–gene correlations between the wild-caught and common garden data sets was small but statistically significant in each case (positive regulators: R²: 0.04, p < 0.0001; negative regulators R²: 0.03, p = 0.003).