Heritability of fat distributions in male mice from the founder strains of the Diversity Outbred mouse population

Figure 1. Illustration of MRI parameters in mice. Representative examples of axial T1 spin echo and Dixon scans, as well as segmentation of fat measurements, are shown for the abdomen, thorax and pericardium, neck and tongue. Subcutaneous fat is shown in pink/red, internal fat (i.e., visceral abdominal, intra-neck, intrathoracic, and tongue) in yellow, and pericardial fat in blue.

Figure 2. Heritability estimates across phenotypes. Estimates of the relative proportion of phenotypic variability explained by genetic factors (heritability; h2) and nongenetic factors (e.g., environmental) are shown across all phenotypes both (A) unadjusted and (B) controlling for body weight. Nearly all traits show high heritability in unadjusted analyses. While most estimates are reduced when adjusting for body weight, a number of traits maintain high heritability; the estimate for the second principal component (PC2) remains unchanged in adjusted analyses.

Figure 3. Comparison of pericardial fat and tongue fat volumes. The distribution of (A) pericardial fat and (B) tongue fat are shown across the eight founder strains. Vertical error bars represent the observed mean ± standard deviation. The NZO/HILtJ and C57BL/6J strains had significantly more pericardial fat than all other strains. We observed a significant heritability estimate of 0.621 (95% CI: 0.503, 0.740) for pericardial fat. This estimate remained significant after controlling for body weight (0.439 [0.287, 0.590]). For tongue fat, the NZO/HILtJ again demonstrated significantly more fat than other strains, but the PWK/PhJ strain, rather than C57BL/6J, also demonstrated significantly higher volume than all strains except NZO/HILtJ. Tongue fat volume had high heritability (0.765 [0.658, 0.872]), which remained significant when controlling for body weight (0.577 [0.385, 0.769]). Thus, both pericardial fat and tongue fat show significant heritability adjusting for body weight.

Figure 4 Graphical summaries of principal components analysis. Graphs summarizing the results of the principal components analysis of individual fat distribution measures are shown, including (A) the cumulative variance explained, (B) a factor loading plot, and (C) a scatter plot. As shown in the plot of cumulative variance explained, the first two principal components explain >80% of the variability in fat distribution measures. The factor loading plot illustrates the relative correlations of each fat distribution with the two components, showing that all measures are positively correlated with PC1, whereas only intra-neck fat and pericardial fat are positively correlated with PC2. Finally, the scatter plot shows the values for individual mice of each strain.

Figure 5. Comparison of principal components derived from individual fat distributions. Values of principal components are shown across the founder strains. Vertical error bars represent the observed mean ± standard deviation. Both components significantly differed among strains (P < 0.0001). The obese NZO/HILtJ strain had significantly higher values of PC1 compared to all other strains, reflecting the positive correlations between PC1 and all fat volume measures (see also Table 1). Results suggest this component is highly heritable, with an h2 (95% CI) of 0.837 (0.771, 0.903). Heritability is reduced by ∼50% when adjusting for body weight (0.439 [0.271, 0.608]). For PC2, which was most strongly correlated with intra-neck fat and pericardial fat volumes, the C57BL/6J mice have higher values than all other strains. This second component has high heritability (0.643 [0.532, 0.754]), which remains nearly identical after controlling for body weight (0.655 [0.548, 0.761]). Thus, results support a genetic influence on fat distributions, independent of general obesity.