Figure 3

Sphingomyelins respond to insulin resistance. (a) Schematic showing downstream ceramide derivatives sphingomyelins and glycosphingolipids, including hexosylceramides and lactosylceramides. (b) Differences in circulating abundance of hexosylceramides (HexCer) and lactosylceramides (LacCer) and sphingomyelins (SM) between control (Ctrl) and metabolically impaired (Metⁱ) animals pre-diagnosis (pre-Dx) and at time of diagnosis (Dx). Data are shown as the difference in medians divided by the median absolute deviation for each species. Statistical significance based on univariate analysis (two-sample Wilcoxon) is shaded as indicated (unadjusted p values). (c) Lasso logistic regression showing ability of indicated sphingolipids to predict insulin resistance, with significance of regression based predictive p values and accuracy indicated by color gradient. (d) Abundance of circulating hexosylceramide (HexCer) and lactosylceramide (LacCer) species (ng/mL) in control (Ctrl) and metabolically impaired (Metⁱ) animals at time of diagnosis, *p values < 0.05 (unadjusted). (e) Left: heatmap of coefficient of variation (population standard deviation/population mean) of individual sphingomyelin species shown for control (Ctrl) and metabolically impaired (Metⁱ) animals pre-diagnosis (pre-Dx) and at time of diagnosis (Dx). Right: bootstrapped mean difference 95% confidence intervals for sphingomyelins from control (Ctrl) and metabolically impaired (Metⁱ) animals pre-diagnosis (pre-Dx) and at time of diagnosis (Dx) (color coded as indicated). For all data, n = 8 animals per group per time point.