Natural variation in cardiac metabolism and gene expression in Fundulus heteroclitus

Abstract

Individual variation in gene expression is important for evolutionary adaptation and susceptibility to diseases and pathologies. In this study, we address the functional importance of this variation by comparing cardiac metabolism to patterns of mRNA expression using microarrays. There is extensive variation in both cardiac metabolism and the expression of metabolic genes among individuals of the teleost fish Fundulus heteroclitus from natural outbred populations raised in a common environment: metabolism differed among individuals by a factor of more than 2, and expression levels of 94% of genes were significantly different (P < 0.01) between individuals in a population. This unexpectedly high variation in metabolic gene expression explains much of the variation in metabolism, suggesting that it is biologically relevant. The patterns of gene expression that are most important in explaining cardiac metabolism differ between groups of individuals. Apparently, the variation in metabolism seems to be related to different patterns of gene expression in the different groups of individuals. The magnitude of differences in gene expression in these groups is not important; large changes in expression have no greater predictive value than small changes. These data suggest that variation in physiological performance is related to the subtle variation in gene expression and that this relationship differs among individuals.

Figure 1

Cardiac metabolism among 16 individuals (identified by numbers) from populations from northern Maine (M) and southern Georgia (G) using three substrates: 5 mM glucose; fatty acid (1 mM palmitic acid bound to bovine serum albumin); and LKA (5 mM lactate, 5 mM each of two ketones: hydroxybutyrate and acetoacetate, and 0.1% ethanol). Two inhibitors of glucose utilization (20 mM 2-deoxyglucose and 10 mM iodoacetate) were included with fatty acid and LKA. Relative changes are based on residuals from log-regression of body mass with metabolic rates and refer to the ratio of the individual value to the overall mean. Principal component analysis among the 16 individuals for all three measures of metabolism was used to order individuals with respect to overall metabolism. The first principal component factor captures 78% of the variation in metabolism and can be thought of as a weighted average for each substrate.

Figure 2

Significant differences in gene expression in a population versus difference relative to the population mean. The x axis shows log₂ values for the ratio of minimum (negative log₂ values) and maximum values (positive log₂ values) relative to population mean. The y axis shows negative log₁₀ values of P values for significant differences in a population (e.g., 4 is equal to a P value of 10⁻⁴. The gray horizontal box encloses genes with nonsignificant differences in a population (P > 0.01). Vertical lines enclose values where minimum and maximum values differ by a factor of <1.5. All P values less than 10⁻¹⁷ are truncated to equal 10⁻¹⁷. Blue circles indicate individuals from northern Maine; red crosses, individuals from southern Georgia.

Figure 3

Hierarchical cluster of metabolic gene expression. Average level of expression for each individual (16 replicates per individual). Individuals (columns) with similar patterns of expression among genes are grouped together, and genes (rows) with similar expression among individuals are grouped together. (a) All 119 metabolic genes. Gene names are supplied in Supplementary Table 2 online. Blue gene trees (left) are shown in b. (b) Select clusters of genes emphasizing alternating patterns of up-down regulations. For a and b, colors are same as in Figure 1. Center correlation with complete linkage was used. Individuals (identified by numbers) are from populations from northern Maine (M) and southern Georgia (G).

Figure 4

Pattern of gene expression arranged according to metabolic rates. (a) Cardiac metabolic rates for the 16 individuals (identified by numbers) from populations from northern Maine (M) and southern Georgia (G; as in Fig. 1). Groups of individuals (based on overall similarities in gene expression; Fig. 3) are indicated. (b–d) Gene expression for glucose metabolic enzymes, TCA enzymes and proteins involved in oxidative phosphorylation (Supplementary Table 5 online). Metabolic enzymes are based on the Kyoto Encyclopedia of Genes and Genomes.

Figure 5

Different patterns of gene expression explain the variation in metabolism. The group-specific (group indicated at the top of each column) relationships between substrate-specific metabolism (y axis) and the most informative principal component (x axis; bold line, filled symbols) are shown. Individuals from the other groups (hollow symbols, thin lines) are included in each graph to illustrate the differences among groups. The most informative principal components (Supplementary Table 6 online) were defined by step-wise regression. The R² value is for the relationship between the principal component listed and metabolism. The P value is given for the final step-wise regression involving one or more principal components (Supplementary Table 7 online). Groups of individuals were as in Figure 3. Blue squares, group 1; red stars, group 2; green circles, group 3. Gly, glucose metabolic enzymes; OxP, oxidative phosphorylation enzymes; TCA, Krebs cycle (TCA) enzymes.

Figure 6

Relative changes versus correlation with metabolism. Relative changes (maximum:minimum ratio; on the x axis) for group 1 (blue squares), group 2 (red circles), group 3 (green diamonds) or all 16 individuals (black triangle) are plotted against the absolute values of the correlation coefficients of gene expression with the three measures of metabolism (a, glucose; b, fatty acid; c, LKA; on the y axis). Relative changes in gene expression in a group are correlated with the variation among individuals with r > 0.91 (i.e., larger variation among individuals with larger maximum:minimum ratios). If genes with large differences in expression are more important, genes with large variation (and, therefore, large relative differences) should have larger correlations with metabolism.