Muscle very long-chain ceramides associate with insulin resistance independently of obesity

Abstract

Lipids, in particular ceramides and diacylglycerols (DAGs), are implicated in insulin resistance (IR), however their precise roles remain unclear. Here, we leverage natural genetic variation to examine muscle lipids and systemic IR in 399 Diversity Outbred Australia mice fed either chow or a high-fat diet. Adipose tissue mass was significantly associated with 55% of muscle lipid features and whole-body insulin sensitivity, with DAGs as the only lipid class enriched in this association. To disentangle the contribution of adiposity and muscle lipids to whole-body insulin sensitivity, we employed two independent approaches: (1) a linear model correcting muscle lipid features for adipose tissue mass to assess their relationship with insulin sensitivity, and (2) stratifying mice into insulin sensitivity quartiles within adiposity bins. Both revealed that very long-chain ceramides, but not DAGs, were linked to IR. RNA sequencing and proteomics from the same muscles further associated these very long-chain ceramides with cellular stress, mitochondrial dysfunction, and protein synthesis. Meanwhile, DAGs correlated with leptin gene expression in skeletal muscle, suggesting they originate from contaminating adipocytes rather than myocytes per se. We propose that many muscle lipids, including DAGs, associate with muscle and systemic IR due to accumulation of adipose tissue rather than directly influencing muscle insulin sensitivity. By addressing the relationship between adiposity and metabolic state, we identified very long-chain muscle ceramides as being highly associated with IR independently of adiposity.

Keywords: Ceramides; Insulin resistance; Lipids; Skeletal muscle.

Figure 1

Metabolic traits across 399 Diversity Outbred Australia (DOz) mice fed high-fat diet or chow. (A) Male DOz mice were either fed chow (n = 203) or a high-fat diet for 8 (n = 80) or 15 weeks (n = 116). The chow fed group was aged-matched to the 8 weeks HFD group. (B & C) Fat mass (B) and fasting blood insulin (C) across all mice. Boxplots show traits grouped by diet and scatter plots show each trait ordered by blood insulin (lowest to highest). Insert shows the relationship between fat mass and blood insulin. For both traits, chow <8 weeks < 15 weeks significantly different. (D) HOMA-IR across diet groups; chow <8 weeks < 15 weeks significantly different. (E) Fasting blood glucose across diet groups; chow group significantly different from HFD groups only. (F) Relationship between fasting blood insulin and HOMA-IR.

Figure 2

Target lipidomics in quadriceps from 399 DOz male mice. (A) Relative (top) and absolute (bottom) abundance of lipid classes quantified by targeted lipidomics. Both are ordered by abundance of absolute total phosphatidylcholine abundance (lowest to highest). (B) Dimension reduction by multidimensional scaling (MDS) of all lipid features. (C) Bicor correlation coefficient of all lipid features to fat mass, fasting blood insulin, HOMA-IR, blood insulin at 15 min during glucose tolerance test and fasting blood glucose.

Figure 3

Linear modelling showing relationship between HOMA-IR and each lipid feature. (A) Main effect of HOMA-IR on individual lipid features from linear models accounting for diet and time of diet. (B) Main effect of HOMA-IR on individual lipid features corrected for overall fat mass from linear models accounting for diet and time of diet. Size of data points are sized to the absolute change in HOMA-IR.

Figure 4

Stratified quartile-based assessment of muscle lipids by HOMA-IR across adiposity bins. (A) Relationship between adiposity and HOMA-IR. Two adiposity bin of 10 percentage points highlighted. Within each bin, 1st and 4th quartile (Q) of HOMA-IR are shown. (B) HOMA-IR across all adiposity bins. Number above indicates the total number of animals (Q1 + Q4). Significant difference between Q1 and Q4 in all adiposity bins. (C) Comparison of ceramides (Cers) and diacylglycerols (DAGs) abundance between Q1 and Q4 relative to Q1 across all adiposity bins. (D) Overall average difference between Q1 and Q4 across all adiposity bins. Significance tested by one-sample t-test (null hypothesis: log2 fold-changes centers around 0 (log2FC = 0)). (E) Selected pathways associated with very long-chain ceramides from both transcriptomic (n = 92) and proteomic (n = 197) data. (F) The relationship between total DAG abundance and leptin gene expression in skeletal muscle. (G) The relationship between very long-chain ceramides and leptin gene expression. Error bars are standard error over mean (SEM).