We identified open reading frames (ORF) in our reference transcriptome using the program getorf in emboss (European Molecular Biology Open Software Suite). We kept the longest open-ended ORF (minimum length of 300 nucleotides) as the most probable translated region of the gene. We used paml (Phylogenetic Analysis by Maximum Likelihood [59], which accounts for codon usage bias and GC content, to calculate the rate of non synonymous mutations scaled by the rate of synonymous mutations (dN/dS) in a maximum likelihood framework (run-mode = 0, CodonFreq = 2, model = 2). For dN/dS calculations, we first identified the consensus (major) allele for each site for each species and then annotated these sites into a single consensus sequence per gene per species. We discarded genes where either dNor dS equalled zero, keeping only those dN/dS ratios that could be calculated with confidence.