Regulatory versus coding signatures of natural selection in a candidate gene involved in the adaptive divergence of whitefish species pairs (Coregonus spp.)

Abstract

While gene expression divergence is known to be involved in adaptive phenotypic divergence and speciation, the relative importance of regulatory and structural evolution of genes is poorly understood. A recent next-generation sequencing experiment allowed identifying candidate genes potentially involved in the ongoing speciation of sympatric dwarf and normal lake whitefish (Coregonus clupeaformis), such as cytosolic malate dehydrogenase (MDH1), which showed both significant expression and sequence divergence. The main goal of this study was to investigate into more details the signatures of natural selection in the regulatory and coding sequences of MDH1 in lake whitefish and test for parallelism of these signatures with other coregonine species. Sequencing of the two regions in 118 fish from four sympatric pairs of whitefish and two cisco species revealed a total of 35 single nucleotide polymorphisms (SNPs), with more genetic diversity in European compared to North American coregonine species. While the coding region was found to be under purifying selection, an SNP in the proximal promoter exhibited significant allele frequency divergence in a parallel manner among independent sympatric pairs of North American lake whitefish and European whitefish (C. lavaretus). According to transcription factor binding simulation for 22 regulatory haplotypes of MDH1, putative binding profiles were fairly conserved among species, except for the region around this SNP. Moreover, we found evidence for the role of this SNP in the regulation of MDH1 expression level. Overall, these results provide further evidence for the role of natural selection in gene regulation evolution among whitefish species pairs and suggest its possible link with patterns of phenotypic diversity observed in coregonine species.

Keywords: Adaptive divergence; Coregonus; gene expression; natural selection; regulatory evolution; speciation.

Figure 1

Normal and dwarf lake whitefish (Coregonus clupeaformis). Normal whitefish (top) commonly exceeds 40 cm in length and 1000 g in weight while dwarf whitefish (bottom) rarely exceeds 20 cm and 100 g.

Figure 2

Putative binding profile of MDH1 regulatory region among coregonine species. Binding probability: highest posterior probability of being bound to a given transcription factor for each of 658-bp upstream of the transcription start site according to binding simulation using Sunflower (Hoffman and Birney 2010), the value of zero was attributed to a given position when the posterior probability of the unbound state was highest. (A) Lake whitefish (C. clupeaformis), two haplotypes, (B) lake cisco (C. artedi), three haplotypes, (C) European whitefish (C. lavaretus), 13 haplotypes, (D) vendace (C. albula), four haplotypes. *Putative binding sites that are unbound (probability = 0) for one or more haplotypes depending on alleles at positions 373 and 374 (Table 3), labeled with putative transcription factor name.

Figure 3

Genetic differentiation of MDH1 5′ regulatory and coding regions between North American and European whitefish species pairs. Based on 781 bp of regulatory sequence (positions 17,590–18,370 in Genbank accession HQ287747) and 807 bp of coding sequence (positions 18,317–24,113 without introns in Genbank accession HQ287747). F_ST values based on overall mean pairwise genetic p-distances computed with HYPHY (Kosakovsky Pond et al. 2005). Negative F_ST estimates were forced to zero. For the 5′ regulatory region, both alleles for each fish were used. For the coding region, mean F_ST for true single nucleotide polymorphisms (SNPs) (excluding paralogous SNPs, see Table 1) was computed by using one copy of each observed haplotype per fish. ‡ Bootstrapped estimator significantly different from zero (P < 0.05, 500 replicates) and probability of a random F_ST greater than the observed value <0.05 (500 permutations). *Probability of a random F_ST greater than the observed value <0.05 (500 permutations).

Figure 4

Predicted ternary structure of whitefish MDH1. Based on whitefish MDH1 protein sequence (Genbank accession ADV02378) of 78% homology with porcine cytoplasmic malate dehydrogenase in the Protein Data Bank (ID = 5MDH) using the 3D-JIGSAW server (v.2.0, http://bmm.cancerresearchuk.org/~3djigsaw/). The graphical representation was created with Pymol v.1.3. (A) Green = amino acid residues of the malate binding domain, (B) orange = amino acid residues of the NAD binding domain, (C) blue = amino acid residues of the dimer interface, (D) amino acid changes due to nonsynonymous substitutions among coregonine populations (see Table 1).

Figure 5

MDH1 expression level as a function of the genotype at SNP –286 in dwarf and normal lake whitefish. Expression level: normalized R/Lowess signal intensity in log₂ from a previous microarray experiment (St-Cyr et al. 2008). Data for 16 fish from Cliff Lake and 15 from Indian Pond, half normals (N), half dwarves (D). Frequency of the T allele: Cliff N = 0.2, D = 0.67 and Indian N = 0.1, D = 0.3. One-way ANOVA between the five groups: P-value = 0.007. *Tukey multiple comparisons of means: Dwarf AT/Dwarf AA P-value = 0.05, Dwarf AT/Normal AA: P-value = 0.007.