Sphingolipids are thought to promote skeletal muscle insulin resistance. Deoxysphingolipids (dSLs) are atypical sphingolipids that are increased in the plasma of individuals with type 2 diabetes and cause β-cell dysfunction in vitro. However, their role in human skeletal muscle is unknown. We found that dSL species are significantly elevated in muscle of individuals with obesity and type 2 diabetes compared with athletes and lean individuals and are inversely related to insulin sensitivity. Furthermore, we observed a significant reduction in muscle dSL content in individuals with obesity who completed a combined weight loss and exercise intervention. Increased dSL content in primary human myotubes caused a decrease in insulin sensitivity associated with increased inflammation, decreased AMPK phosphorylation, and altered insulin signaling. Our findings reveal a central role for dSL in human muscle insulin resistance and suggest dSLs as therapeutic targets for the treatment and prevention of type 2 diabetes.
Deoxysphingolipids (dSLs) are atypical sphingolipids elevated in the plasma of individuals with type 2 diabetes, and their role in muscle insulin resistance has not been investigated.
We evaluated dSL in vivo in skeletal muscle from cross-sectional and longitudinal insulin-sensitizing intervention studies and in vitro in myotubes manipulated to synthesize higher dSLs.
dSLs were increased in the muscle of people with insulin resistance, inversely correlated to insulin sensitivity, and significantly decreased after an insulin-sensitizing intervention; increased intracellular dSL concentrations cause myotubes to become more insulin resistant.
Reduction of muscle dSL levels is a potential novel therapeutic target to prevent/treat skeletal muscle insulin resistance.
Introduction
Skeletal muscle is responsible for ∼80% of whole-body insulin-stimulated glucose disposal and is consequently a key site of insulin resistance, a critical feature in the pathophysiology of type 2 diabetes (1). Insulin resistance can be promoted by lipids in muscle and was initially associated with triglycerides, but further investigations revealed a central role for other bioactive lipids, such as diacylglycerols (DAGs) and ceramides (2–5). Additionally, our research group and several others have demonstrated that skeletal muscle sphingolipids play an important role in insulin resistance in the pathogenesis of type 2 diabetes (6–8).
Sphingolipids are a complex and diverse family of lipids, which are important structural components of eukaryotic cell membranes and function as mediators of complex signaling pathways involved in inflammation, apoptosis, and cell proliferation and differentiation (9). Sphingolipid de novo biosynthesis starts with the condensation of palmitoyl-CoA and L-serine catalyzed by the enzyme serine palmitoyltransferase (SPT) to form 3-ketosphinganine, subsequently reduced to sphinganine. Several different fatty acids can be acylated to the sphinganine backbone, resulting in the formation of ceramide, the key intermediate in the biosynthesis of all complex sphingolipids, in which the 1-hydroxyl position can be further modified via linkage with carbohydrates or phosphate (10).
In addition to L-serine, mammalian SPTs can also catalyze, to a lesser extent, the condensation of L-alanine or glycine to palmitoyl-CoA (11,12), giving rise to atypical 1-deoxysphingolipids (dSLs). dSLs lack the hydroxyl group in the C1 position and, therefore, cannot be metabolized to complex sphingolipids, such as sphingomyelins, glycosphingolipids, or sulfatides, nor can they undergo turnover via phosphorylation of sphingosine and formation of sphingosine-1-phospate (13). Compared with their canonical equivalents (e.g., ceramides), dSLs are characterized by a higher hydrophobicity that prevents them from efficiently mixing within lipid bilayers, with a significant negative impact on membrane integrity and stability (14).
Plasma dSLs are significantly increased in several metabolic and neurological disorders and are particularly elevated in individuals with type 2 diabetes and metabolic syndrome (15,16). Plasma dSL levels have been suggested as predictive biomarkers for the development of type 2 diabetes in at-risk populations (17), and 1-deoxydihydroceramides (dDHCers), specifically, have been found to be significantly increased in plasma of individuals with diabetic neuropathy (18). Furthermore, 1-deoxysphinganine (dSA), the dSL backbone, has been shown to be toxic to pancreatic β-cells (19) and neurons (20) where it appears to compromise cytoskeletal stability by metabolism to dDHCers. Exogenous administration of dSA to C2C12 myotubes in vitro reduces myoblast viability, induces apoptosis, and reduces insulin-stimulated glucose uptake, further substantiating the effects of circulating dSLs on peripheral tissues (21).
Given the importance of sphingolipids to muscle insulin resistance, it is likely that tissue dSLs could also play an important role in regulating muscle insulin sensitivity. However, dSLs have never been measured in human skeletal muscle, and their relationship to muscle insulin resistance has never been investigated. The primary objective of this study was to quantify dSLs in skeletal muscle in individuals spanning a wide range of insulin sensitivity and metabolic health and to assess changes in muscle dSLs after an insulin sensitizing lifestyle intervention. Additionally, we developed a protocol to alter dSL content in primary human myotubes in vitro to directly evaluate how increased dSL muscle levels can affect insulin sensitivity.
Research Design and Methods
Cross-Sectional Study
The methods and procedures for the cross-sectional human study have already been reported in detail in a previous publication by our laboratory (7). Briefly, lean sedentary control participants, endurance trained athletes, and sedentary participants with obesity with and without type 2 diabetes were recruited for the study after obtaining informed consent. After screening visits and an overnight fast, participants underwent a muscle biopsy from midway between the greater trochanter of the femur and the patella, followed by a standard hyperinsulinemic-euglycemic clamp. Diet was not controlled to avoid confounding acute dietary changes on muscle lipid composition. The data shown in this article were obtained using skeletal muscle from a subset of the original study participants, and the participant demographic characteristics and plasma marker values are summarized in Table 1. Individuals with type 2 diabetes were excluded from the study if they used insulin and/or thiazolidinediones. All other medications were permissible but were washed out for 2 weeks prior to metabolic testing. These medications included metformin (n = 4), sitagliptin (n = 2), sitagliptin/metformin (n = 2), glyburide (n = 1), glipizide (n = 1), and liraglutide (n = 1). Fasting and 2-h oral glucose tolerance test (OGTT) plasma glucose was measured from all consented participants to screen for type 2 diabetes during preliminary testing, and participants were classified into study groups based on criteria set by the American Diabetes Association. Participants with obesity with fasting plasma glucose levels ≥125 mg/dL and/or 2-h OGTT values ≥200 mg/dL were placed in the type 2 diabetes study group, while participants with obesity with fasting plasma glucose values <125 mg/dL and/or 2-h OGTT values <200 mg/dL were included in the obese study group. This study was approved by the Colorado Multiple Institutional Review Board at the University of Colorado Anschutz Medical Campus (protocol no. 10-0443).
Participant demographic characteristics and plasma markers
Variable . | Lean control individuals . | Athletes . | Individuals with obesity . | Individuals with type 2 diabetes . |
---|---|---|---|---|
Participants, n | 14 | 13 | 12 | 10 |
Female | 7 | 5 | 5 | 4 |
Male | 7 | 8 | 7 | 6 |
Age (years) | 42.7 ± 1.9 | 42.1 ± 1.5 | 42.4 ± 1.6 | 46.1 ± 1.8 |
BMI (kg/m2) | 22.1 ± 0.7 | 23.1 ± 0.6 | 34.4 ± 0.7¥# | 33.9 ± 2.1¥# |
Percent body fat | 23.2 ± 2.4 | 17.4 ± 1.6 | 35.5 ± 2.3¥# | 34.8 ± 2.9¥# |
2-h OGTT glucose (mg/dL) | 94.1 ± 3.6 | 72.3 ± 6.1 | 102.0 ± 6.4 | 311.5 ± 25.2¥#§ |
Fasting glucose (mg/dL) | 89.8 ± 1.7 | 86.8 ± 2.6 | 93.2 ± 2.2 | 179.0 ± 14.7¥#§ |
Fasting insulin (μU/mL) | 8.8 ± 1.3 | 7.2 ± 0.7 | 17.0 ± 2.1¥# | 19.4 ± 2.7¥# |
Free fatty acids (μEq/L) | 624.2 ± 80.0 | 562.4 ± 65.0 | 505.3 ± 40.3 | 568.5 ± 40.7 |
Triglycerides (mg/dL) | 97.9 ± 16.3 | 67.9 ± 5.2 | 130.3 ± 16.6# | 157.6 ± 23.0¥# |
TNF-α (pg/mL) | 1.2 ± 0.2 | 0.8 ± 0.1 | 1.3 ± 0.2 | 1.2 ± 0.2 |
IL-6 (pg/mL) | 1.5 ± 0.2 | 1.1 ± 0.4 | 1.5 ± 0.2 | 2.3 ± 0.4 |
Glucose infusion rate (mg/kg/min) | 8.5 ± 0.8#§* | 12.2 ± 0.9¥§* | 5.4 ± 0.7¥#* | 1.9 ± 0.6¥#§ |
Variable . | Lean control individuals . | Athletes . | Individuals with obesity . | Individuals with type 2 diabetes . |
---|---|---|---|---|
Participants, n | 14 | 13 | 12 | 10 |
Female | 7 | 5 | 5 | 4 |
Male | 7 | 8 | 7 | 6 |
Age (years) | 42.7 ± 1.9 | 42.1 ± 1.5 | 42.4 ± 1.6 | 46.1 ± 1.8 |
BMI (kg/m2) | 22.1 ± 0.7 | 23.1 ± 0.6 | 34.4 ± 0.7¥# | 33.9 ± 2.1¥# |
Percent body fat | 23.2 ± 2.4 | 17.4 ± 1.6 | 35.5 ± 2.3¥# | 34.8 ± 2.9¥# |
2-h OGTT glucose (mg/dL) | 94.1 ± 3.6 | 72.3 ± 6.1 | 102.0 ± 6.4 | 311.5 ± 25.2¥#§ |
Fasting glucose (mg/dL) | 89.8 ± 1.7 | 86.8 ± 2.6 | 93.2 ± 2.2 | 179.0 ± 14.7¥#§ |
Fasting insulin (μU/mL) | 8.8 ± 1.3 | 7.2 ± 0.7 | 17.0 ± 2.1¥# | 19.4 ± 2.7¥# |
Free fatty acids (μEq/L) | 624.2 ± 80.0 | 562.4 ± 65.0 | 505.3 ± 40.3 | 568.5 ± 40.7 |
Triglycerides (mg/dL) | 97.9 ± 16.3 | 67.9 ± 5.2 | 130.3 ± 16.6# | 157.6 ± 23.0¥# |
TNF-α (pg/mL) | 1.2 ± 0.2 | 0.8 ± 0.1 | 1.3 ± 0.2 | 1.2 ± 0.2 |
IL-6 (pg/mL) | 1.5 ± 0.2 | 1.1 ± 0.4 | 1.5 ± 0.2 | 2.3 ± 0.4 |
Glucose infusion rate (mg/kg/min) | 8.5 ± 0.8#§* | 12.2 ± 0.9¥§* | 5.4 ± 0.7¥#* | 1.9 ± 0.6¥#§ |
Data are mean ± SEM unless otherwise indicated. P values were calculated by one-way ANOVA, with P < 0.05 considered significant. IL-6, interleukin-6; TNF-α, tumor necrosis factor-α.
Significantly different than lean individuals.
Significantly different than athletes.
Significantly different than individuals with obesity.
Significantly different than individuals with type 2 diabetes.
Intervention Study
Twenty-one sedentary obese individuals with normal glucose tolerance, prediabetes, or newly diagnosed type 2 diabetes (Table 2) completed a 12-week combined weight loss and exercise training intervention. This study was approved by the Colorado Multiple Institutional Review Board at the University of Colorado Anschutz Medical Campus (protocol no. 13-1551).
Intervention study participant demographic characteristics and plasma markers
Variable . | Preintervention . | Postintervention . |
---|---|---|
Participants, n | 21 | |
Female | 18 | |
Male | 3 | |
Age (years) | 45.9 ± 1.9 | |
BMI (kg/m2) | 34.9 ± 0.9 | 31.3 ± 0.9* |
Percent body fat | 41.4 ± 1.3 | 38.1 ± 1.5* |
2-h OGTT glucose (mg/dL) | 143 ± 19 | |
Fasting glucose (mg/dL) | 101.1 ± 6.6 | 91.9 ± 2.3 |
Fasting insulin (μU/mL) | 15.1 ± 1.2 | 10.5 ± 1.1* |
Free fatty acids (μEq/L) | 583 ± 26 | 508 ± 19* |
Triglycerides (mg/dL) | 124 ± 12 | 113 ± 7 |
TNF-α (pg/mL) | 0.90 ± 0.15 | 0.68 ± 0.14* |
IL-6 (pg/mL) | 1.7 ± 0.15 | 1.59 ± 0.18 |
Glucose infusion rate (mg/kg/min) | 3.6 ± 0.3 | 5.6 ± 0.5* |
Variable . | Preintervention . | Postintervention . |
---|---|---|
Participants, n | 21 | |
Female | 18 | |
Male | 3 | |
Age (years) | 45.9 ± 1.9 | |
BMI (kg/m2) | 34.9 ± 0.9 | 31.3 ± 0.9* |
Percent body fat | 41.4 ± 1.3 | 38.1 ± 1.5* |
2-h OGTT glucose (mg/dL) | 143 ± 19 | |
Fasting glucose (mg/dL) | 101.1 ± 6.6 | 91.9 ± 2.3 |
Fasting insulin (μU/mL) | 15.1 ± 1.2 | 10.5 ± 1.1* |
Free fatty acids (μEq/L) | 583 ± 26 | 508 ± 19* |
Triglycerides (mg/dL) | 124 ± 12 | 113 ± 7 |
TNF-α (pg/mL) | 0.90 ± 0.15 | 0.68 ± 0.14* |
IL-6 (pg/mL) | 1.7 ± 0.15 | 1.59 ± 0.18 |
Glucose infusion rate (mg/kg/min) | 3.6 ± 0.3 | 5.6 ± 0.5* |
Data are mean ± SEM. IL-6, interleukin-6; TNF-α, tumor necrosis factor-α.
P < 0.05 by two-tailed paired Student t test.
Participants gave written informed consent and underwent preliminary screening, including a fasting blood draw, health and physical examination, a standard 75-g OGTT, and a DEXA scan. Individuals with normal glucose tolerance, prediabetes, or newly diagnosed type 2 diabetes were recruited and verified using OGTT at the initial screening. Normal glucose tolerance was defined as fasting glucose <100 mg/dL, with postprandial glucose <140 mg/dL 2 h after a 75-g OGTT. Prediabetes was defined as fasting glucose between 100 and 125 mg/dL and/or postprandial glucose between 140 and 199 mg/dL 2 h after a 75-g OGTT. Participants with type 2 diabetes had a fasting glucose >125 mg/dL and/or postprandial glucose >200 mg/dL 2 h after a 75-g OGTT. Maximal work capacity (VO2max) was measured using a standard Balke treadmill test. After an overnight fast, participants underwent a vastus lateralis skeletal muscle biopsy from midway between the greater trochanter of the femur and the patella, followed by a standard hyperinsulinemic-euglycemic clamp with insulin infused at 40 mU/m2/min, as previously described (7). Participants were weight stable in the 6 months prior to the study, and food was provided to each participant, following standard American Heart Association recommendations, for 7 days before the insulin clamp.
After the first insulin clamp and muscle biopsy visit, participants entered a combined 12-week low-calorie diet and exercise training intervention. For the weight loss intervention, participants received a low-calorie diet consisting of five portions of a meal replacement product (Health One; Health Nutrition Technology Inc.), providing 890 kcal/day, 75 g of protein, 15 g fat, 110 g carbohydrate, and 100% of the dietary reference intake of all vitamins, minerals, and micronutrients, and four to five servings of fruits and vegetables per day. To promote gall bladder contraction and reduce the risk of gallstone formation, participants were asked to consume 2 teaspoons of vegetable oil per day. Noncaloric beverages were allowed, but no other food intake was permitted. Study participants were seen weekly in one-on-one sessions with a registered dietitian for help with the diet and to receive nutritional counseling. Participants also underwent supervised endurance exercise training using well-described procedures used by the Nutrition and Obesity Research Center energy balance core (22). Participants were asked to attend four 60-min supervised exercise sessions per week. Each session included a short warm-up period, 40–50 min of endurance exercise, and a cool-down period. The exercise program consisted primarily of brisk walking or jogging and was supplemented with rowing, stepping, or elliptical exercise. Individualized exercise prescriptions considered the fitness level of the participant, preferences regarding type of exercise, and any orthopedic limitations. Participants wore a heart rate monitor that captured and stored heart rate data throughout the exercise sessions. After 3 months, participants transitioned to a weight maintenance diet consisting of three servings of meal replacement per day along with one meal of typical foods for 2 weeks before postintervention testing. During the postintervention testing, participants repeated the DEXA and VO2max tests and the muscle biopsy and hyperinsulinemic-euglycemic clamp. The insulin clamp was performed at least 48 h from the last exercise bout.
Metabolomics
Metabolomic analysis of underivatized amino acids on frozen skeletal muscle biopsies from the cross-sectional study participants was performed using high-performance liquid chromatography and high-resolution quadrupole orbitrap mass spectrometry (MS) by the Metabolomic Core Facility at the University of Colorado Anschutz Medical Campus, as previously described (23).
Cell Culture
Primary human skeletal muscle cells were generated in our laboratory using muscle biopsies from lean men and women with insulin sensitivity from the cross-sectional study described above. Cells were grown on collagen-coated plates in low-glucose DMEM containing 0.25 mg/mL fetuin, 0.5 mg/mL BSA, 0.025 mg/mL gentamicin, 0.125 μg/mL amphotericin B, 0.01 μg/mL recombinant human epidermal growth factor, 0.39 μg/mL dexamethasone, 1× GlutaMAX, 10% FBS, and 2% penicillin/streptomycin at 37°C in 5% CO2. Before seeding the cells, plates were collagen coated for 1 h at room temperature by adding a diluted solution (50 μg/mL) of collagen type I (from rat tail) in 0.02 N acetic acid at a final concentration of 5 μg/cm2 of growth area. When cells reached ∼90% confluence, they were differentiated into myotubes by replacing the regular growth media with low-glucose DMEM containing 0.5 mg/mL BSA, 0.025 mg/mL gentamicin, 0.125 μg/mL amphotericin B, 2% horse serum, and 20 μg/mL L-carnitine. Cells were plated on 12-well plates for glycogen synthesis assay and on 6-well plates for lipid analysis by liquid chromatography-tandem MS (LC-MS/MS). Assays were performed during day 7 of differentiation, and only cells between passages 3 and 4 were used.
Cell Treatment
Primary myotubes on their 5th day of differentiation were treated for an additional 48 h with differentiation media prepared with minimum essential medium (MEM) without L-serine, L-alanine, and glycine (M0446; Sigma-Aldrich) supplemented with different concentrations of the three amino acids. Control cells (L-serine/L-alanine/glycine 1X) received concentrations normally found in the regular DMEM (35.6 μg/mL L-alanine, 30 μg/mL glycine, and 42 μg/mL L-serine); 71.2 μg/mL L-alanine, 60 μg/mL glycine, and 4.2 μg/mL L-serine were used to achieve enhanced dSL synthesis (L-alanine/glycine 2X, L-serine 0.1X). Media were replaced once after 24 h.
Insulin-Stimulated Glycogen Synthesis Assay
On day 7 of differentiation, primary myotubes were serum starved for 3.5 h in serum-free MEM containing 20 μg/mL L-carnitine and L-serine/L-alanine/glycine at the different concentrations described above. Cells were then incubated for 1 h at 37°C in serum-free MEM with L-carnitine and different concentrations of amino acids containing 2 μCi/mL D-[U-14C]-glucose in the presence or absence of 100 nmol/L insulin, and glycogen synthesis assay was performed according to the protocol previously described by Schmitz-Peiffer et al. (24).
Lipid Extraction and Analysis by LC-MS/MS
Muscle samples were collected in 500 μL of water, and adherent primary myotubes were washed three times with ice cold PBS and scraped off the wells using 500 μL of water. Samples were homogenized using a bead mill homogenizer for 2 min at 25 Hz. After removing 20 μL for protein measurements, 450 μL of muscle or cell homogenate were diluted to 750 μL with water, and 900 μL of methanol was added. After addition of the appropriate internal standard cocktail, lipids were extracted using a modified methyl-tert-butyl ether extraction protocol, as previously described (25).
LC-MS/MS analysis and quantitation of dSLs and typical sphingolipids was conducted as previously described by our laboratory (18,26). Data were acquired using the software program Analyst 1.6.2 and processed using MultiQuant 3.0.3 software (Sciex). All lipid standards were purchased from Avanti Polar Lipids, Inc.
Reverse-Phase Protein Array
Primary human skeletal muscle cells in control or increased dSL conditions (∼2 × 106) were harvested in reverse-phase protein array (RPPA) lysis buffer (T-PER Tissue Protein Extraction Reagent in 0.5 mol/L NaCl; Thermo Fisher Scientific) containing protease and phosphatase inhibitors and analyzed using high-throughput antibody-based RPPA performed using ≥240 validated antibodies by the Antibody-Based Proteomics Core at Baylor College of Medicine, directed by Dr. Shixia Huang, as previously described (27).
Immunoblot Analysis
Human primary myotubes (∼2 × 106 cells), in control or enhanced dSL conditions, were protein extracted and separated by SDS-PAGE using standard techniques. All primary and secondary antibodies were purchased from Cell Signaling Technology, except for anti–phospho-IRS-1 (Tyr612) (MilliporeSigma). The signal was detected using an Odyssey CLx Infrared Imaging System and analyzed using Image Studio software (LI-COR).
Statistical Analysis
Statistical analysis was performed using GraphPad Prism 9.4.0 (GraphPad Software). Data are presented as mean ± SEM. Differences in normally distributed data between groups were analyzed using a one-way ANOVA. Two-way ANOVA was used to compare two independent variables within multiple groups. When significant differences were detected, groups were compared using two-tailed Student t tests. Differences in dSLs before and after intervention, as well as alterations in myotubes in response to treatment, were evaluated using a paired Student t test. Nonnormally distributed data were log-transformed prior to analysis. Significant differences in individual lipid species between groups were adjusted for multiple comparisons using the Benjamini-Hochberg procedure. Significant relationships between individual lipids and insulin sensitivity were determined using Pearson correlation coefficient and were adjusted for multiple comparisons using the Benjamini-Hochberg procedure. P < 0.05 was considered significant.
Data and Resource Availability
The data sets generated in this study are available upon request from the corresponding author. No applicable resources were generated during the study.
Results
Skeletal Muscle dSLs Scale to Insulin Sensitivity in Humans
We evaluated skeletal muscle dSL content in endurance-trained athletes, lean individuals, and individuals with obesity and type 2 diabetes (Table 1) by LC-MS/MS. We were able to measure dSL species in whole-muscle homogenates from ∼15 mg of tissue. Although our analytical platform included dSA, 1-deoxysphingosine; 1-deoxyceramide, dDHCer, 1-deoxymethylceramide; and 1-deoxymethyldihydroceramide subspecies, only dDHCers were detected. Muscle from participants with obesity and type 2 diabetes had a significantly higher content of total dDHCer compared with endurance-trained athletes and lean control participants, with athletes showing the lowest concentrations (Fig. 1A). Interestingly, total dDHCers were significantly inversely related to insulin sensitivity, as measured in the same participants via an hyperinsulinemic-euglycemic clamp (Fig. 1B). Analysis of individual dDHCer species showed that C18:0 dDHCer was the most abundant dSL in muscle, and C18:0, C22:0, and C24:1 species were significantly increased in participants with insulin resistance (Fig. 1C). Although C16:0 dDHCer was not elevated significantly in participants with obesity and type 2 diabetes (Fig. 1B), its increased levels were still significantly related to insulin resistance (Fig. 1D). All three species showed a significant inverse correlation to insulin sensitivity (Fig. 1E–G). These results provide evidence of an inverse relationship between dDHCer levels and insulin sensitivity, suggesting a possible role for these lipids in the development of muscle insulin resistance.
dSL content in human skeletal muscle scales to insulin sensitivity. A: Total dSL content in endurance-trained athletes (n = 13), lean control individuals (n = 14), individuals with obesity with (n = 10) and without (n = 12) type 2 diabetes, as measured by LC-MS/MS. The P values were calculated by one-way ANOVA. B: Pearson correlation coefficients between total dSL content and insulin sensitivity, as measured by hyperinsulinemic-euglycemic clamp. C: Individual dDHCer species concentration in endurance-trained athletes, lean control individuals, and individuals with obesity with and without type 2 diabetes. The P values were calculated by two-way ANOVA. D–G: Pearson correlation coefficients between individual dSL species and insulin sensitivity for C16:0 dDHCer (D), C18:0 dDHCer (E), C22:0 dDHCer (F), and C24:1 dDHCer (G). Data are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001.
dSL content in human skeletal muscle scales to insulin sensitivity. A: Total dSL content in endurance-trained athletes (n = 13), lean control individuals (n = 14), individuals with obesity with (n = 10) and without (n = 12) type 2 diabetes, as measured by LC-MS/MS. The P values were calculated by one-way ANOVA. B: Pearson correlation coefficients between total dSL content and insulin sensitivity, as measured by hyperinsulinemic-euglycemic clamp. C: Individual dDHCer species concentration in endurance-trained athletes, lean control individuals, and individuals with obesity with and without type 2 diabetes. The P values were calculated by two-way ANOVA. D–G: Pearson correlation coefficients between individual dSL species and insulin sensitivity for C16:0 dDHCer (D), C18:0 dDHCer (E), C22:0 dDHCer (F), and C24:1 dDHCer (G). Data are mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001.
Whole-Muscle L-Alanine Levels Correlate With dSL Muscle Content
dSLs are synthesized when SPT favors the condensation of palmitoyl-CoA with L-alanine or glycine rather than L-serine, so we used high-resolution orbitrap MS to quantify L-alanine, glycine, and L-serine in muscle homogenates from endurance-trained athletes, lean individuals, and individuals with obesity and type 2 diabetes. Glycine data from one lean participant was removed from the analysis with outlier values that were 14 times the mean and 54 SDs away from the mean. As shown in Fig. 2A, no significant changes in L-serine and glycine muscle content were observed between groups, while L-alanine was significantly increased in lean participants and participants with obesity and type 2 diabetes compared with athletes. L-alanine levels were also significantly higher in participants with type 2 diabetes compared with lean. dDHCers are the only dSL species we could detect in skeletal muscle, and they are exclusively derived from palmitoyl-CoA condensation with L-alanine (13). Skeletal muscle L-alanine concentrations showed significant positive correlations with total dDHCer levels (Fig. 2B), with C18:0 dDHCer, the most abundant dSL detected in skeletal muscle, having the strongest correlation (Fig. 2C). Interestingly, the ratio between L-alanine and L-serine muscle concentrations showed significant positive correlations with both total and C18:0 dDHCers (Fig. 2D and E). Taken together, these data suggest a potential relationship between availability of muscle L-alanine and changes in the relative ratio between L-alanine and L-serine and increased dSL content in individuals with insulin resistance.
dSL content correlates with L-alanine and L-alanine/L-serine concentration ratio in skeletal muscle. A: L-serine, L-alanine, and glycine content in skeletal muscle homogenates from endurance-trained athletes, lean control individuals, and individuals with obesity with and without type 2 diabetes, as determined by high-resolution orbitrap MS. B–E: Correlations between L-alanine concentrations and total muscle dDHCer (B) and C18:0 dDHCer (C) levels and between L-alanine/L-serine ratio and total muscle dDHCer (D) and C18:0 dDHCer (E) levels using the Pearson correlation coefficient. Data are mean ± SEM. *P < 0.05, ***P < 0.001 by two-way ANOVA. AU, arbitrary unit; T2D, type 2 diabetes.
dSL content correlates with L-alanine and L-alanine/L-serine concentration ratio in skeletal muscle. A: L-serine, L-alanine, and glycine content in skeletal muscle homogenates from endurance-trained athletes, lean control individuals, and individuals with obesity with and without type 2 diabetes, as determined by high-resolution orbitrap MS. B–E: Correlations between L-alanine concentrations and total muscle dDHCer (B) and C18:0 dDHCer (C) levels and between L-alanine/L-serine ratio and total muscle dDHCer (D) and C18:0 dDHCer (E) levels using the Pearson correlation coefficient. Data are mean ± SEM. *P < 0.05, ***P < 0.001 by two-way ANOVA. AU, arbitrary unit; T2D, type 2 diabetes.
dSL Whole-Muscle Content Decreases After a Combined Weight Loss and Exercise Training Intervention
To evaluate whether dSLs could be related to changes in muscle insulin sensitivity in vivo, we measured dSL content by LC-MS/MS in muscle biopsies collected before and after a 12-week insulin-sensitizing intervention in 21 participants with obesity and normal glucose tolerance, prediabetes, or newly diagnosed type 2 diabetes (Table 2). The 12-week intervention resulted in a 13.7 ± 3.0% increase in VO2max, 10.7 ± 1.0% decrease in body weight, and a 66.1 ± 11.7% increase in insulin sensitivity, as measured by a hyperinsulinemic-euglycemic clamp. After the intervention, total dDHCer content in skeletal muscle homogenates was significantly reduced (Fig. 3A). When individual dDHCer species were compared pre- and postintervention, most dDHCer subspecies (C16:0, C18:0, C20:0, C22:0, and C24:0) were significantly decreased (Fig. 3B). Taken together, these results show that muscle dSL content can be altered in humans through weight loss and exercise intervention and decreases in parallel with increased insulin sensitivity, consistent with a role for muscle dSL in insulin resistance.
dSL content in skeletal muscle from individuals with obesity and normal glucose tolerance, prediabetes, or newly diagnosed type 2 diabetes decreases after an insulin sensitizing intervention. A: Total dSL content was evaluated by LC-MS/MS in skeletal muscle homogenates from 21 individuals before and after a combined weight loss and exercise training intervention. B: Individual dDHCer species content comparisons pre- and postintervention. Data are mean ± SEM. *P < 0.05, **P < 0.01 by two-tailed paired Student t test.
dSL content in skeletal muscle from individuals with obesity and normal glucose tolerance, prediabetes, or newly diagnosed type 2 diabetes decreases after an insulin sensitizing intervention. A: Total dSL content was evaluated by LC-MS/MS in skeletal muscle homogenates from 21 individuals before and after a combined weight loss and exercise training intervention. B: Individual dDHCer species content comparisons pre- and postintervention. Data are mean ± SEM. *P < 0.05, **P < 0.01 by two-tailed paired Student t test.
Increased Myotube dSL Concentration Causes Insulin Resistance In Vitro
We developed an experimental protocol to alter dSL levels in human primary myotubes by modifying media L-serine, L-alanine, and glycine content. We replaced our regular media for the last 2 days of differentiation with MEM lacking L-serine, L-alanine, and glycine and added back these amino acids at different concentration combinations. In control cells, we added back the same amino acid concentrations present in our regular media (L-serine/L-alanine/glycine 1X), while we increased dSL synthesis by adding two times the amounts of L-alanine and glycine and keeping L-serine concentration low at 10% of the amount present in regular DMEM (L-alanine/glycine 2X, L-serine 0.1X). As shown in Fig. 4A, treatment of cells with L-alanine and glycine 2X and L-serine 0.1X increased the total dSL content approximately fourfold. The increased dSL synthesis resulted in a significant decrease in insulin sensitivity, measured using the percent increase in insulin-stimulated glycogen synthesis (Fig. 4B). Similar to what was observed in whole muscle, the only dSL species detected in myotubes was dDHCer, but in this case, the most abundant metabolite was C16:0 dDHCer, followed by C18:0 and C24:1 dDHCer. All the subspecies detected were significantly increased compared with the control condition (Fig. 4C).
Modulation of media amino acid concentrations alters dSL content and insulin sensitivity in human primary myotubes in vitro without affecting typical sphingolipid and DAG cell content. A: Total dSL content in myotubes treated for 48 h in differentiation media containing L-serine/L-alanine/glycine 1X or L-alanine/glycine 2X and L-serine 0.1X. B: Relationship between total dSL concentration and insulin sensitivity, as measured by insulin-stimulated glycogen synthesis. C: Effect of media amino acid content on individual dSL species concentrations. D and E: Effect of media amino acid content on total sphingolipids (ceramides [Cer], dihydroceramides [DHCer], glucosyl-ceramides [GluCer], and lactosyl-ceramides [LacCer]) and total DAGs (1,2- and 1,3-DAGs). Data are mean ± SEM (n = 5), with two technical replicates each. *P < 0.05, **P < 0.01, ***P < 0.001 by two-tailed paired Student t test.
Modulation of media amino acid concentrations alters dSL content and insulin sensitivity in human primary myotubes in vitro without affecting typical sphingolipid and DAG cell content. A: Total dSL content in myotubes treated for 48 h in differentiation media containing L-serine/L-alanine/glycine 1X or L-alanine/glycine 2X and L-serine 0.1X. B: Relationship between total dSL concentration and insulin sensitivity, as measured by insulin-stimulated glycogen synthesis. C: Effect of media amino acid content on individual dSL species concentrations. D and E: Effect of media amino acid content on total sphingolipids (ceramides [Cer], dihydroceramides [DHCer], glucosyl-ceramides [GluCer], and lactosyl-ceramides [LacCer]) and total DAGs (1,2- and 1,3-DAGs). Data are mean ± SEM (n = 5), with two technical replicates each. *P < 0.05, **P < 0.01, ***P < 0.001 by two-tailed paired Student t test.
Importantly, in these experimental conditions, the amino acid treatment did not increase skeletal muscle cell content of typical sphingolipids (ceramides, dihydroceramides, glucosylceramides, and lactosylceramides) (Fig. 4D) and DAGS (1,2- and 1,3-DAG) (Fig. 4E), which are known to impact insulin sensitivity, supporting a direct role for dSLs in causing insulin resistance in muscle cells. To evaluate the impact of altering the amino acid profile to increase dSL content on overall metabolic status of the cells, we performed additional untargeted metabolomics profiling, MS analysis of serine-containing phosphatidylserines, and thiobarbituric acid reactive substances assay on myotubes. The description of these experiments can be found in the Supplementary Materials, and the data are summarized in Supplementary Figs. 1–3. We could not detect significant differences in the metabolic profile, phospholipids, and lipid peroxidation when comparing cells in control vs enhanced dSL conditions. These data suggest that altering the media concentration of L-alanine, glycine, and L-serine did not result in global metabolic changes that can explain the insulin resistance observed.
Myotube dSL Concentration Modulates Proteins Involved in Glucose Transport and Insulin Signaling
To evaluate whether increased concentrations of dSL in skeletal muscle cells affected protein and phosphoprotein expression profiles, RPPA analysis was performed on lysates from control cells and myotubes containing increased levels of dSL. When dSL levels were augmented by manipulating media L-serine, L-alanine, and glycine content, we observed a significant decrease in cell c-Met expression (Fig. 5A). c-Met is the receptor for the hepatocyte growth factor, and the hepatocyte growth factor-Met axis has been shown to augment IRS-1 signaling and promote skeletal muscle glucose uptake and metabolism (28–30).
dSL myotube content affects c-Met expression and basal and insulin-stimulated cell signaling. A: Effect of media supplementation with L-serine/L-alanine/glycine 1X or L-alanine/glycine 2X and L-serine 0.1X on primary myotube c-Met expression, as measured by high-throughput RPPA. Data are mean ± SEM (n = 4), with three technical replicates each. B: Effects of increased dSL concentrations on phosphorylation of downstream effectors of insulin signaling in basal and insulin-stimulated conditions, as measured via immunoblot analysis. C: Representative insulin signaling blot. Data are mean ± SEM (n = 6). *P < 0.05, **P < 0.01 by two-tailed paired Student t test. B, basal; IS, insulin stimulated.
dSL myotube content affects c-Met expression and basal and insulin-stimulated cell signaling. A: Effect of media supplementation with L-serine/L-alanine/glycine 1X or L-alanine/glycine 2X and L-serine 0.1X on primary myotube c-Met expression, as measured by high-throughput RPPA. Data are mean ± SEM (n = 4), with three technical replicates each. B: Effects of increased dSL concentrations on phosphorylation of downstream effectors of insulin signaling in basal and insulin-stimulated conditions, as measured via immunoblot analysis. C: Representative insulin signaling blot. Data are mean ± SEM (n = 6). *P < 0.05, **P < 0.01 by two-tailed paired Student t test. B, basal; IS, insulin stimulated.
To evaluate changes in activation of key players in insulin signaling, we also performed an immunoblot analysis on cell lysates from myotubes differentiated in the same experimental conditions that caused a significant decrease in insulin-stimulated glycogen. In response to insulin, increased concentrations of dSL did not cause any significant changes in phosphorylation of Akt (Ser473), and glycogen synthase kinase 3β (GSK-3β) (Ser9) relative to control conditions. Surprisingly, phosphorylation of IRS-1 (Tyr612) was significantly decreased (Fig. 5B). A representative gel for insulin signaling in vitro is shown in Fig. 5C. These in vitro findings suggest that increased dSL concentrations may decrease c-Met amplification of IRS-1 signaling to attenuate cell insulin sensitivity.
dSL Content in Human Primary Myotubes Affects Inflammatory Kinase Activation
RPPA analysis revealed significant changes in the activation of two protein kinases in myotubes grown in enhanced dSL conditions. IκB kinase (IKK) α/β phosphorylation (Ser176/180) (Fig. 6A) was significantly increased, while a significant decrease in AMPK phosphorylation of the α-subunit (phosphorylated on T172) (Fig. 6B) was observed compared with control conditions. Interestingly, both these kinases have been reported in the literature as involved in the modulation of insulin resistance (31,32).
Phosphorylation of AMPK and IKKAα/β is modulated by dSL concentrations in primary myotubes. A and B: Effect of media supplementation with 1X L-serine/L-alanine/glycine or 2X L-alanine/glycine and 0.1X L-serine on primary myotube p-AMPK and p-IKKα/β activation, as measured as phosphorylated protein increase by high-throughput RPPA. Data are means ± SEM (n = 4), with three technical replicates each. C: Effects of increased dSL concentrations on phosphorylation of downstream effectors of inflammatory signaling, as measured by RPPA. Data are mean ± SEM (n = 4), with three technical replicates each. *P < 0.05 by two-tailed Student t test.
Phosphorylation of AMPK and IKKAα/β is modulated by dSL concentrations in primary myotubes. A and B: Effect of media supplementation with 1X L-serine/L-alanine/glycine or 2X L-alanine/glycine and 0.1X L-serine on primary myotube p-AMPK and p-IKKα/β activation, as measured as phosphorylated protein increase by high-throughput RPPA. Data are means ± SEM (n = 4), with three technical replicates each. C: Effects of increased dSL concentrations on phosphorylation of downstream effectors of inflammatory signaling, as measured by RPPA. Data are mean ± SEM (n = 4), with three technical replicates each. *P < 0.05 by two-tailed Student t test.
We did not observe significant differences in phosphorylation of proinflammatory signaling through p44/42 mitogen-activated protein kinase (MAPK) (Erk1/2) (Thr202/Tyr204), nuclear factor-κB (NF-κB) p65 (Ser536), stress-activated protein kinases (SAPK)/c-Jun N-terminal kinases (JNK) (Thr183/Tyr185), and p38 MAPK (Thr180/Tyr182) (Fig. 6C), as measured via immunoblot. These results suggest that elevated intracellular dSL concentrations in myotubes could promote the development of insulin resistance through greater IKKα/β and decreased AMPK activation.
Discussion
dSLs are a novel class of sphingolipids that are elevated in plasma from individuals with type 2 diabetes. In vitro results revealed that dSL decreases glucose-induced insulin secretion (19), promotes mitochondrial dysfunction in cancer and neuronal cells (33,34), reduces myoblast viability and insulin-stimulated glucose uptake, and induces apoptosis when administered to C2C12 myotubes (21). However, their role in human skeletal muscle is unknown. The main findings of this study reveal that skeletal muscle dSL levels are elevated in individuals with obesity and type 2 diabetes, inversely correlate to insulin sensitivity, and are significantly decreased in vivo after an insulin sensitizing lifestyle intervention. Furthermore, increased dSLs cause insulin resistance in vitro in human primary myotubes. Together, these data are the first to implicate dSLs as key contributors to skeletal muscle insulin resistance in humans.
One of the initial goals of this study was to determine the identities of dSL species and their concentration in human skeletal muscle from individuals spanning a wide range of insulin sensitivity and evaluate their relationship to insulin resistance. We were able to show that dSL can be measured in human skeletal muscle and scale inversely to insulin sensitivity. Our methodology allowed us to measure dSL species present in muscle, different from most studies in the literature that hydrolyzed the molecule and only measured the sphingoid base in plasma and cells (11,16,17). Although our analytical platform is comprehensive and includes most of the possible dSL species, only dDHCers were detected in muscle, with C18:0 dDHCer being the most abundant individual dSL. This is very likely related to the increased skeletal muscle expression of the highly C18:0-specific ceramide synthase 1 enzyme observed in obesity-associated insulin resistance (35). These findings are in agreement with previous reports on mouse neurons, retina, nervous tissue, and various aging tissues where dDHCers were the main dSLs detected when intact dSLs were analyzed (20,36–38). The undetectable levels of 1-deoxyceramides in muscle could be explained by a lower expression in this tissue of fatty acid desaturase type 3, a desaturase that specifically introduces a double bond in the Δ14Z position of the deoxysphingoid base (39). To our knowledge, dSLs have never been measured in human skeletal muscle prior to this study, and our report clearly shows that not only are dSLs significantly increased in muscle tissue of individuals with insulin resistance but that dDHCer muscle levels are also inversely correlated to insulin sensitivity. These data are consistent with a direct role for these atypical sphingolipids in the modulation of insulin sensitivity in humans.
SPT is the critical limiting step in the synthesis of sphingolipids. Specific mutations in genes encoding for SPT can cause a shift in preference from L-serine to L-alanine and have been identified in patients affected by two very rare hereditary diseases characterized by increased circulating dSL levels, hereditary sensory and autonomic neuropathy 1 and macular telangiectasia type 2 (36,40,41). Plasma dSLs have been reported as major players in diabetes, diabetic neuropathy, and metabolic syndrome and are suggested as potential plasma biomarkers (15–18), but no SPT defects have been associated to these specific conditions. However, dSLs can be synthesized to a lesser extent by wild-type SPT (12), and detectable levels of dSLs are always observed in plasma from healthy individuals (11,36). It has been suggested that a change in the relative ratio between L-alanine and L-serine cell levels could lead to increased concentrations of dSL (42,43). The significant correlations we observed between muscle L-alanine and, more importantly, L-alanine/L-serine ratio and dDHCer concentrations, corroborate and expand previously published data supporting L-alanine relative to L-serine changes playing a role in causing changes in dSL, although it is still unclear whether the effects observed are simply due to increased substrate availability, proximity of L-alanine to SPT, or changes in the enzyme specificity. Importantly, these data strengthen the potential of promising studies that showed L-serine supplementation through the diet could be effective in reducing dSL in diabetic mice (44,45), as well as ameliorate clinical outcomes in both mice and humans affected by hereditary sensory and autonomic neuropathy 1 (46–48).
Sphingolipid skeletal muscle content, in particular ceramide and glucosylceramide, has been associated with muscle insulin resistance, although contrasting reports are present in the literature (49,50), which may be explained in part by a lack of specific evaluation of lipid degradation and turnover, subcellular lipid localization, and species differences. Moreover, changes in ceramide muscle levels frequently do not explain improvement in insulin sensitivity observed after physical activity in cross-sectional studies, although there seems to be a clear correlation between decreased skeletal muscle ceramides and improved insulin sensitivity after exercise training (50–52). The data presented in this study clearly show that total dSL muscle content and most individual dSL subspecies significantly decreased after individuals with obesity and normal glucose tolerance, prediabetes, or newly diagnosed type 2 diabetes completed a combined weight loss and exercise training intervention that caused a substantial increase in insulin sensitivity. Given the nature of the study, we cannot determine the relative contribution of the two lifestyle interventions to the dSL changes observed. Interestingly, C18:0 dDHCer, the main dSL detected in muscle, showed the most significant decrease postintervention, supporting previous findings from our laboratory and others who suggested a unique relationship between stearate-containing sphingolipid species and insulin resistance (6,35,53). Combined, the data obtained from both the cross-sectional and longitudinal studies included in this work strongly support a direct relationship between skeletal muscle dSL and insulin resistance and could help to explain some of the ambiguity in the literature surrounding changes in sphingolipids and insulin resistance.
To study the direct effects of increased dSL content on insulin sensitivity, we developed an in vitro protocol to alter dSL content in primary myotubes by manipulating media concentrations of L-serine, L-alanine, and glycine and evaluated insulin sensitivity via an insulin-stimulated glycogen synthesis assay. Our data provide the first evidence for a direct effect of dSL in decreasing insulin sensitivity in primary human muscle cells and strongly support a causal role of dSL in promoting muscle insulin resistance. To facilitate data interpretation, we specifically chose L-serine, L-alanine, and glycine concentrations that increased myotube dSL levels without changing intracellular levels of other lipids that are known to affect insulin sensitivity, such as ceramides and DAGs. The intracellular dSL concentrations we achieved in vitro were ∼10 times the amount we measured in whole muscle, and the observed effects on insulin sensitivity reflect a more acute exposure compared with the chronic environment of the skeletal muscle. Nonetheless, our in vitro results establish proof of concept for dSLs decreasing muscle insulin sensitivity.
The in vitro protocol we developed was suitable to conduct experiments to evaluate mechanisms through which dSLs could cause insulin resistance. Increased levels of dDHCers in cell membranes have significant effects on membrane integrity and stability (14), mostly because of extreme hydrophobicity and inability to self-organize in bilayers. Exogenous dSA, the deoxysphingoid base precursor of dDHCer, has been shown to localize preferentially to the mitochondria in cultured cells, causing mitochondrial fragmentation and dysfunction (33). dSA can induce toxicity in primary neurons and islets by altering endoplasmic reticulum function, Ca2+ handling, cytoskeleton organization, and activation of Rac-1, c-Jun N-terminal kinase, and Erk1/2 (19,20,34). Our data corroborate the signaling potential of dSLs by showing that dSLs can decrease muscle cell c-Met content, which has been shown to activate the insulin receptor and promote IRS-1 signaling (29). Our data are consistent with decreased c-Met activation of IRS-1 signaling, as both decreased c-Met content and attenuated IRS-1 (Tyr612) phosphorylation in response to insulin were found in myotubes with elevated dSL content compared with the control condition. However, decreased IRS-1 signaling did not result in reduced downstream Akt or GSK-3β phosphorylation in response to insulin. Similar data have been found in insulin resistant states in humans and rodents and could be explained by only a portion of IRS-1 signaling being required for Akt activation, subcellular compartmentalization of Akt phosphorylation for insulin signaling, and/or reduced insulin-stimulated translocation of Akt to the plasma membrane (54–56). Our data also suggest that dSLs activate muscle IKKα/β phosphorylation to activate NF-κB and decrease AMPK phosphorylation, both of which are associated with insulin resistance (57,58). We did not observe differences in p65 (Ser536) phosphorylation between treatments, suggesting that dSL does not alter the NF-κB complex nuclear translocation or proteolytic degradation (59). Combined, our data suggest that elevated dSLs decrease insulin signaling, activate the NF-κB inflammatory response, and decrease AMPK phosphorylation, which results in decreased insulin sensitivity in primary myotubes.
There are a few limitations to note about this study. It is likely that analyzing whole muscle would lead to an underestimation of the complexity of how these sphingolipids affect insulin resistance. Therefore, performing experiments to identify the subcellular localization of these lipids could reveal preferential sites of accumulation and more-specific mechanisms through which they promote insulin resistance, providing better support for future potential treatments and interventions. Furthermore, we did not find a significant difference in dSL concentration between participants with obesity with and without type 2 diabetes despite significant differences in insulin sensitivity. This could simply be due to lack of power, but these data also support that insulin resistance is multifactorial and suggest that there are other factors promoting decreased insulin sensitivity in muscle between these two groups. Finally, although the media L-serine, L-alanine, and glycine supplementation protocol we used to enhance dSL synthesis in vitro did not alter intracellular levels of typical sphingolipids or significantly change the overall myotube metabolic profile, phosphatidylserine content, and lipid peroxidation, we cannot completely exclude that this treatment affected other factors within the cells that could change myotube response to insulin.
In summary, our studies reveal a key role for dSL in promoting skeletal muscle insulin resistance. The findings suggest that the reduction of muscle dSL concentration is a novel therapeutic target to prevent and/or treat type 2 diabetes.
Clinical trial reg. no. NCT02043405, clinicaltrials.gov
This article contains supplementary material online at https://doi.org/10.2337/figshare.22666174.
Article Information
Acknowledgments. The authors thank Dr. Colleen McKenna from the Division of Endocrinology, Metabolism, and Diabetes at the University of Colorado Anschutz Medical Campus for invaluable help with designing and preparing the graphical abstract. The authors thank the University of Colorado School of Medicine Metabolomics Core for contributions to this article. The authors also thank Dr. Xuan Wang and Dr. Zhongcheng Shi from the Baylor School of Medicine Antibody-Based Proteomics Core/Shared Resource for technical assistance with the RPPA analysis and Dr. Christian Coarfa and Dr. Sandra L. Grimm from the Baylor School of Medicine Antibody-Based Proteomics Core/Shared Resource for RPPA data processing and normalization.
Funding. This work was supported by CCTSI 2019 CO-Pilot grant CO-M-19-74 to S.Z. and partially supported by the National Institutes of Health General Clinical Research Center grant RR-00036, the National Institute of Diabetes and Digestive and Kidney Diseases grant R01DK089170 to B.C.B., the Colorado Nutrition Obesity Research Center grant P30DK048520, and the American Diabetes Association grant 1-14-CE-05 to B.C.B. The Metabolomics Core Facility is supported by the Colorado Cancer Center support grant P30CA046934. The Antibody-Based Proteomics Core is supported by the CPRIT Core Facility Support Award RP210227, the National Cancer Institute Cancer Center Support Grant P30CA125123, and the National Institutes of Health S10 Instrument Award S10OD028648 to Shixia Huang.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. S.Z. designed and performed all the in vitro experiments and all dSL analyses, analyzed and interpreted data, and wrote the manuscript. K.A.Z.B. performed sphingolipid and DAG analyses and edited the manuscript. D.E.K. and A.G. performed immunoblot analyses and edited the manuscript. L.P. helped to design the studies, provided medical oversight, performed all biopsies, and edited the manuscript. A.K. performed participant testing. B.C.B. designed the studies, analyzed and interpreted data, and helped write the manuscript. S.Z. and B.C.B. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Prior Presentation. Parts of this study were presented in abstract form at the 78th Scientific Sessions of the American Diabetes Association, Orlando FL, 22–26 June 2018.