Study participants were selected from the Pathobiology of Prediabetes in a Biracial Cohort (POP-ABC) [24, 25]. Participants were eligible for the POP-ABC study if they were 18–65 years, self-reported as non-Hispanic Black (African American) or non-Hispanic White (European American) and had biological parent(s) with type 2 diabetes. Additional participants were included from the POP-ABC normative cohort comprising African American European American adults without a family history of diabetes. All participants underwent a screening 75-gram oral glucose tolerance test (OGTT) to document normal fasting plasma glucose (FPG; <100 mg/dL [5.6 mmol/L]) and/or normal glucose tolerance (2-h plasma glucose [2hPG] <140 mg/dL [7.8 mmol/L]) [17, 18]. Intercurrent illness, hyperglycemia, a history of diabetes, or exposure to medications known to alter glucose metabolism were among the exclusion criteria [24, 25]. A written informed consent was obtained from all participants before enrollment in the POP-ABC study [24, 25].

For the present report, we selected African American participants (N = 129) and European American participants (N = 111), matched in age (within 5 years) and sex, who had completed assessment of insulin sensitivity and insulin secretion at enrollment [20, 21]. Plasma specimens obtained from the participants at enrollment were analyzed for sphingolipid levels in the ongoing Ceramides and Sphingolipids as Predictors of Incident Dysglycemia (CASPID) study. The CASPID study protocol (IRB Approval #21-07936-FB) was approved by the University of Tennessee Health Science Center (UTHSC) Institutional Review Board and carried out at the UTHSC General Clinical Research Center (GCRC).

Participants arrived at the GCRC after fasting overnight for baseline assessments. These included a structured medical interview and general physical examination, measurement of height, weight, waist circumference, and OGTT. The body mass index (BMI) was calculated as weight in kilograms divided by the height in meters squared. Additional assessments included measurement of insulin sensitivity and insulin secretion.

Ahead of the baseline visit, participants received written instructions to continue with their normal diet, refrain from strenuous exercise and alcohol consumption for a day prior, and to avoid smoking the morning of the OGTT. After an overnight fast, venous blood specimens were sampled from participants before (0 min) and at 30 and 120 min after drinking 75 g of flavored glucose solution (Trutol 75; Custom Laboratories, Baltimore, MD) [24, 25].

Whole body insulin sensitivity was assessed using the hyperinsulinemic euglycemic clamp method of DeFronzo et al. [26]. In brief, participants were fasted overnight for approximately 12 h before undergoing the clamp studies at the GCRC. A primed, continuous intravenous infusion of regular insulin (2 mU/kg/min; 12 pmol/kg/min) was administered for 180 min through intravenous cannulas in both arms, while maintaining blood glucose level at ∼100 mg/dL (5.6 mmol/L) with a variable rate dextrose (20%) infusion. Glucose and insulin measurements were carried out on arterialized blood samples obtained every 10 min during the procedure. During the final 60 min of insulin infusion (steady state), the total insulin-stimulated glucose disposal rate (M) was calculated and corrected for the steady-state plasma insulin levels to derive the final insulin sensitivity index (Si-clamp), as previously described [24–26]. Fasting glucose and insulin levels were used to calculate the homeostasis model assessment of insulin resistance (HOMA-IR) (HOMA2 calculator. https://homa-calculator.informer.com/2.2/) based on Matthews et al. [27].

Acute insulin secretory response to glucose (AIRg) was assessed using the frequently sampled intravenous glucose tolerance test (FSIGT), as previously described [24, 25]. In brief, after an overnight fast, participants arrived at the GCRC for the FSIGT. An intravenous dextrose bolus (25g) was administered, with arterialized blood sampling for the measurement of glucose and insulin levels at −30, 0, 2, 3, 4, 5, 7, and 10 min relative to dextrose bolus. The AIRg was computed as the mean incremental insulin concentration from 3 to 5 min after the dextrose bolus [23, 25, 28]. Fasting glucose and insulin levels were used to calculate the homeostasis model assessment of beta cells (HOMA-B) (HOMA2 calculator. https://homa-calculator.informer.com/2.2/) based on Matthews et al. [27]. The disposition index, a measure of the insulin secretion corrected for ambient insulin sensitivity, was calculated as AIRg x Si-clamp [29].

Plasma glucose levels during OGTT and hyperinsulinemic euglycemic clamp were measured at the bedside using the YSI glucose analyzer (Yellow Spring Instruments Co., Inc., Yellow Spring, OH). Plasma insulin levels were measured in our Endocrine Research Laboratory with a solid phase, two-site sequential chemiluminescent immunometric assays on the Siemens Immulite analyzer (Siemens Healthcare Diagnostics Products, Ltd., Camberley, Surrey, UK). The within-batch variation coefficient for the insulin assay was <5%.

Targeted lipidomic analysis was performed in baseline plasma specimens obtained at enrolment in the POP-ABC study. The plasma specimens preserved in EDTA were stored at −80°C until the lipidomic analysis. A total of 58 selected sphingolipids of various classes (ceramides, monohexosyl-ceramides, sphingomyelins, sphingosine and dihydro-sphingosine-1-phosphate) were assayed using liquid chromatography-tandem mass spectrometry (LC-MS/MS). The measurements were carried out at the Lipidomics Core at Virginia Commonwealth University, Richmond, Virginia, using established protocols [19, 20]. The individual sphingolipid species were validated and quantified using a Shimadzu LC-20 AD binary pump system coupled to a SIL-20AC autoinjector, and a DGU20A3 degasser coupled to an ABI 4000 quadrupole/linear ion trap (QTrap) (Applied Biosystems), operating in triple quadrupole mode. Using the retention times and m/z ratio, the 58 SPL species were identified and quantified by comparing the target lipid ion of interest with the normalization of quantitated ion abundances [30, 31].

Data were reported as means ± SD. The significance of differences between defined groups was analyzed using t-test for continuous variables and chi squared test for categorical variables. Linear regression models were used to assess the relationship between individual sphingolipid species and measures of adiposity, insulin secretion, and insulin sensitivity. Significance level was generally set as P < 0.05 (two-tailed). For analysis of the association of 58 individual sphingolipid species with insulin sensitivity or insulin secretion, a Bonferroni-corrected significant P value of <0.0009 was adopted, to minimize false discovery from multiple comparisons. The analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC).

The study population (N = 240; 150 women, 90 men) comprised 129 African American (non-Hispanic Black) and 111 European American (non-Hispanic White) participants. Table 1 shows their baseline characteristics. The mean age was 43.1 ± 12.0 y, BMI 29.4 ± 6.23 kg/m², fasting plasma glucose 91.4 ± 6.91 mg/dL, and 2-hr plasma glucose 123 ± 26.3 mg/dL (Table 1). The mean fasting plasma glucose was lower, and the BMI and waist circumference were higher, in African American vs. European American participants. However, there were no ethnic differences in 2-hr plasma glucose or total plasma levels of sphingolipids, ceramides, monohexosyl ceramides, sphingomyelins, and sphingosine (Table 1).

Nominally significant associations were observed between several sphingolipid species and whole-body insulin sensitivity (Si-clamp) (Supplementary Table S1). Plasma levels of long-chain ceramide C18:1 (r = −0.22, P = 0.0062) and long-chain sphingomyelin C18:1 (r = −0.23, P = 0.0046) were inversely associated with insulin sensitivity (Supplementary Table S1). In contrast, plasma levels very-long-chain ceramide C26:0 (r = 0.20, P = 0.014), monohexosylceramide C34:0 (r = 0.30, P = 0.0002) and sphingomyelin C28 – C34 species were positively associated with insulin sensitivity (Supplementary Table S1). No other circulating sphingolipid species (including ceramides, monohexosylceramides, sphingomyelins or sphingosines) reached even nominally significant association with insulin sensitivity in our study population.

After correction for multiple testing, plasma levels of very-long-chain monohexosylceramides C34:0 (r = 0.30, P 0.0002), very-long-chain sphingomyelins C28:1, C28:0, C30:1, C30:0, C32:0, and C34:0 (r = 0.31–0.35, P = 0.0004 - <0.0001) were positively associated with insulin sensitivity (Table 2; Supplementary Table S1). Notably, there was no significant association between plasma levels of total ceramides, total monohexosylceramides, sphingosine, or total sphingolipids and insulin sensitivity (Supplementary Table S1). In contrast, plasma levels of total very-long-chain sphingomyelins were significantly associated with insulin sensitivity (r = 0.35, P < 0.0001) (Supplementary Table S1).

Compared to insulin sensitivity, fewer individual sphingolipid species showed nominally significant associations with insulin secretion (Supplementary Table S2). Inverse associations with AIRg were observed for the very-long-chain sphingomyelins C28:0 (r = −0.25, P = 0.0003), C30:1(r = −0.27, P < 0.0001), and C30:0 (r = −0.20, P = 0.0044) (r = −0.16, P = 0.024) that had showed positive associations with insulin sensitivity (Supplementary Tables S1, S2). Weaker inverse associations with AIRg were observed for some very-long-chain ceramide species (r = −0.14 to −0.16, P = 0.05 - <0.018) and very-long-chain monohexosylceramide C28:1 (r = −0.16, P = 0.024) (Supplementary Table S2). In contrast, positive associations with AIRg were observed for plasma levels of the long-chain sphingomyelins C18:1 (r = 0.23, P = 0.0009) and C18:0 (r = 0.22, P = 0.0012) that were inversely associated with insulin sensitivity (Supplementary Tables S1, S2).

Unlike the findings regarding insulin sensitivity, most of the associations between individual sphingolipid species and AIRg lost statistical significance after Bonferroni correction for multiple testing, except for the association with very-long-chain sphingomyelins C28:0 and C30:1 (Table 2; Supplementary Table S2). Plasma levels of total very-long-chain sphingomyelins were significantly associated with AIRg (r = −0.27, P < 0.0002). In contrast, total plasma levels of sphingolipids and major species (ceramides, monohexosylceramides, sphingosine) were not significantly associated with insulin secretion (Supplementary Table S2).

Plasma levels of very-long-chain sphingomyelins C28:0, C28:1, C30:1 showed concurrent associations with insulin sensitivity (r = 0.31–0.34; P = 0.0004 - <0.0001) and AIRg (r = −0.20 to −0.27; P = 0.004 - <0.0001) (Figure 1; Supplementary Tables S1, S2).

Plasma levels of sphingomyelin C16:0 showed nominal associations with HOMA-IR (r = 0.18, P = 0.038) and HOMA-B (r = 0.23, P = 0.009), as did sphingomyelin C34:0 with HOMA-B (r = 0.18, P = 0.036. However, the associations were not significant after correction for multiple testing. The other sphingolipid species assessed in the present study showed no association with HOMA-IR or HOMA-B.

There were no significant differences by sex or race/ethnicity in the measured sphingolipid species. Further, the observed associations between sphingolipid species and insulin sensitivity and insulin secretion were consistent across race/ethnicity and sex.

Table 3 shows some metabolic correlates of the individual plasma sphingolipid species that exhibited significant associations with insulin sensitivity or insulin secretion. In general, the very-long-chain sphingomyelins and very-long-chain monohexosylceramide C34:0 that were positively associated with insulin sensitivity showed inverse correlations with waist circumference and BMI (Figure 2; Table 3). The very-long-chain monohexosylceramide C34:0 and other very-long-chain sphingomyelin species also were inversely correlated with fasting plasma glucose, 2-h plasma glucose or hemoglobin A1c levels (Table 3).