Fat-Secreted Ceramides Regulate Vascular Redox State and Influence Outcomes in Patients With Cardiovascular Disease

Abstract

Background: Obesity is associated with increased cardiovascular risk; however, the potential role of dysregulations in the adipose tissue (AT) metabolome is unknown.

Objectives: The aim of this study was to explore the role of dysregulation in the AT metabolome on vascular redox signaling and cardiovascular outcomes.

Methods: A screen was conducted for metabolites differentially secreted by thoracic AT (ThAT) and subcutaneous AT in obese patients with atherosclerosis (n = 48), and these metabolites were then linked with dysregulated vascular redox signaling in 633 patients undergoing coronary bypass surgery. The underlying mechanisms were explored in human aortic endothelial cells, and their clinical value was tested against hard clinical endpoints.

Results: Because ThAT volume was associated significantly with arterial oxidative stress, there were significant differences in sphingolipid secretion between ThAT and subcutaneous AT, with C16:0-ceramide and derivatives being the most abundant species released within adipocyte-derived extracellular vesicles. High ThAT sphingolipid secretion was significantly associated with reduced endothelial nitric oxide bioavailability and increased superoxide generated in human vessels. Circulating C16:0-ceramide correlated positively with ThAT ceramides, dysregulated vascular redox signaling, and increased systemic inflammation in 633 patients with atherosclerosis. Exogenous C16:0-ceramide directly increased superoxide via tetrahydrobiopterin-mediated endothelial nitric oxide synthase uncoupling and dysregulated protein phosphatase 2 in human aortic endothelial cells. High plasma C16:0-ceramide and its glycosylated derivative were independently related with increased risk for cardiac mortality (adjusted hazard ratios: 1.394; 95% confidence interval: 1.030 to 1.886; p = 0.031 for C16:0-ceramide and 1.595; 95% confidence interval: 1.042 to 2.442; p = 0.032 for C16:0-glycosylceramide per 1 SD). In a randomized controlled clinical trial, 1-year treatment of obese patients with the glucagon-like peptide-1 analog liraglutide suppressed plasma C16:0-ceramide and C16:0-glycosylceramide changes compared with control subjects.

Conclusions: These results demonstrate for the first time in humans that AT-derived ceramides are modifiable regulators of vascular redox state in obesity, with a direct impact on cardiac mortality in advanced atherosclerosis. (The Interaction Between Appetite Hormones; NCT02094183).

Keywords: C16:0-ceramide; adipose tissue; cardiovascular disease; metabolomics; sphingolipids; vascular redox state.

Conflict of interest statement

Funding Support and Author Disclosures This study was supported by the Novo Nordisk Foundation Tripartite Immunometabolism Consortium Award (NNF15CC0018486), the British Heart Foundation (FS/16/15/32047, PG/13/56/30383, RG/17/10/32859 and RG/F/21/110040) and British Heart Foundation Chair award (CH/16/1/32013), British Heart Foundation Centre of Research Excellence award (RG/13/1/30181), the National Institute for Health Research Oxford Biomedical Research Centre, and the Swedish Heart Lung Foundation (HLF 20180290). The authors have reported that they have no relationships relevant to the contents of this paper to disclose.

Copyright © 2021 The Authors. Published by Elsevier Inc. All rights reserved.

Figures

Graphical abstract

Figure 1

Study Design A schematic diagram depicts the goals, the patient cohorts, and the research methodologies of each study. AT = adipose tissue; CABG = coronary artery bypass graft; NADPH = reduced nicotinamide adenine dinucleotide phosphate; NOS = nitric oxide synthase; SPL = sphingolipid.

Figure 2

Metabolomics Profiling of AT Secretome (A) Heatmap showing the differential secretion patterns for the top 63 differential metabolites (p ≤ 0.001) in thoracic AT (ThAT) versus subcutaneous AT (ScAT) secretome (study 2). (B) Pathway enrichment results for the differential metabolites (study 2). The bar chart shows the pathways with false discovery rate– and Holm-adjusted p values <0.05, ordered on the basis of their impact and colored on the basis of the percentage of up-regulated metabolites in ThAT secretome versus ScAT secretome. (C) Significantly enriched metabolic pathways in the ThAT secretome (study 2) of obese patients (n = 31) versus lean counterparts (n = 17). The node color is based on p value, and the node radius is determined on the basis of pathway impact values. (D) Fold change of relative abundance of all measured sphingolipid species containing an 18:1-sphingoid backbone comparing significant enrichment of ceramides in ThAT versus ScAT secretome (study 2). Results are presented as median ± interquartile range; ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗p < 0.001, ThAT versus ScAT. (E) Heatmap representing average amounts of sphingolipids in ThAT secretome (study 2) of obese (n = 31) and lean (n = 17) patients. (F) Sphingolipid pathway. Cer = ceramide; Cer1P = ceramide-1-phosphate; DhCer = dihydroceramide; GlcCer = glycosylceramide; 3KetoSpa = 3-ketosphinganine; LacCer = lactosylceramide; SM = sphingomyelin; Spa = sphinganine; Sph = sphingosine; Sph1P = sphingosine-1-phosphate; tRNA = transfer ribonucleic acid; other abbreviations as in Figure 1.

Figure 3

Differential Synthesis and Secretion Patterns of Ceramides in ThAT Versus ScAT Ceramides concentrations were measured in ThAT and ScAT (study 2; n = 96) and their secretome (study 2; n = 96) by liquid chromatography–tandem mass spectrometry. Ceramide secretion (A) and synthesis (B) pattern in ThAT versus ScAT. Box plots represent median (interquartile range); ∗∗p < 0.05 (false discovery rate adjusted), ThAT versus ScAT (calculated using the Wilcoxon signed rank test). Abbreviations as in Figure 2.

Figure 4

Human Primary Adipocytes Produce SPLs and Secrete Them via EVs SPL intracellular levels in differentiated adipocytes and pre-adipocytes isolated from ThAT biopsies (n = 10 patients) and relative abundance per class (A to D). Adipocyte extracellular vesicles (EVs) (n = 5) were pooled for western blot analysis and were positive for EV markers (E). Adipocyte EVs showed typical morphology by transmission electron microscopy (F). The abundance of each SPL species in EVs released from adipocytes is presented in G to L. Bars in A to D indicate mean ± SEM; bars in G to L indicate median ± interquartile range; ∗p < 0.05 (nominal) and ∗∗p < 0.01 (false discovery rate adjusted) for mature adipocytes versus pre-adipocytes (calculated using Wilcoxon signed rank tests). Abbreviations as in Figures 1 and 2.

Figure 5

Relationship Between ThAT Ceramides and Vascular Redox State Unsupervised clustering of study 2 patients (n = 48) on the basis of the amounts of SPLs in AT stratified the patients into 3 clusters per tissue: ThAT clusters (Th1, n = 11; Th2, n = 25; Th3, n = 12) and ScAT clusters (Sc1, n = 11; Sc2, n = 26; Sc3, n = 11). Relevant heatmaps showing the intensities of secreted and produced SPLs in the clusters of ThAT (A) and ScAT (B). High SPL synthesis in ThAT (C,D), but not in ScAT (E,F), were related with high vascular superoxide (O2.−) production in both internal mammary arteries (IMAs) and saphenous veins (SV) obtained from the same patients. Similarly, impaired vasorelaxations of human vessels in response to acetylcholine (ACh) were observed in patients with high SPLs in ThAT (G), while there was a similar (but borderline) trend for SPLs in ScAT (H) from the same patients, suggesting that high AT SPL levels are related with reduced nitric oxide (NO) bioavailability in vessels. On the contrary, NO-independent vasorelaxations in response to sodium nitroprusside (SNP) vasorelaxations were not related with SPL levels in either ThAT (I) or ScAT (J). P values in C to F were calculated using the Kruskal-Wallis test. The p values in G to J were calculated using 2-way analysis of variance with dose × group interaction terms. RLU = relative light units; other abbreviations as in Figures 1 and 2.

Figure 6

Circulating Cer16:0 Associated With Vascular Redox State Univariate correlations between plasma levels of 10 quantified ceramides, vascular redox state, and risk profile of patients in study 1 (A). Plasma C16:0-ceramide (Cer16:0) levels were significantly related with ceramide levels in ThAT but not ScAT (B). Association between plasma Cer16:0 levels and serum high-sensitivity C-reactive protein (hsCRP) levels (C; n = 351), O2.− production in human IMAs (D; n = 309), endothelial NO oxide uncoupling determined by the NG-nitro-l-arginine methyl ester (L-NAME)–inhibitable O2.− (E; n = 269), and reduced endothelial NO bioavailability measured by the ex vivo vasorelaxations in response to ACh (F; n = 128). There was no association between plasma Cer16:0 and endothelium-independent vasorelaxations of human vessels to SNP (G; n = 123). P values were calculated using the Kruskal-Wallis test (B–D) or 2-way analysis of variance with dose × group interaction terms (F,G). BMI = body mass index; HDL = high-density lipoprotein; HOMA-IR = homeostasis model assessment of insulin resistance; LDL = low-density lipoprotein; other abbreviations as in Figures 1, 2, and 5.

Figure 7

Effects of Exogenous Cer16:0 on Redox State and eNOS Coupling of Human Aortic Endothelial Cells Confocal microscope images of NBD-Cer6:0 (green, A). Cer16:0 treatment significantly increased Cer16:0 in immortalized human aortic endothelial cells (teloHAECs) (B; n = 11), resulting in increased intracellular O2.− production in teloHAECs (C; n = 9) and primary human aortic endothelial cells (HAECs) (D, n = 5 different donors). Representative teloHAECs stained with dihydroethidium (DHE) show increased intensity of O2.− fluorescence in treated cells (E; n = 10). Treatment of teloHAECs (F; n = 9) and primary HAECs (G; n = 5) with Cer16:0 increased the L-NAME-inhibitable O2.−, via reduction of intracellular tetrahydrobiopterin (BH4) (H; n = 5) and BH4/total biopterin ratio (I; n = 5). Treatment of teloHAECs (n = 7 or 8) with Cer16:0 reduced endothelial NO synthase (eNOS) phosphorylation on activation site Ser1177 (J) and increased its phosphorylation at the inhibitory site Thr495 (K). Although there was no effect of Cer16:0 on AKT phosphorylation at site Ser473 (L), it induced significant reduction of protein phosphatase 2 (PP2A) phosphorylation at its inhibitory site Tyr307 (M). The Cer16:0-induced reductions in p-eNOSSer1177/eNOS ratio was reversed by 10 μM LB100, a PP2A inhibitor (N; n = 6). ∗p < 0.05 versus control (DMSO <1%); all presented as median (interquartile range). Abbreviations as in Figures 1, 2, 5, and 6.

Figure 8

Plasma Cer16:0 and GlcCer16:0 Are Modifiable Risk Biomarkers Kaplan-Meier estimates of cumulative probability of cardiac death by years of follow-up and levels of circulating C16:0-ceramide (Cer16:0) (A), C16:0-glycosylceramide (GlcCer16:0) (B), Cer16:0/C22:0-ceramide (Cer22:0) (C), and Cer16:0/C24:0-ceramide (Cer24:0) (D) and estimates of noncardiac death by years of follow-up and levels of circulating Cer16:0 (E), GlcCer16:0 (F), Cer16:0/Cer22:0 (G), and Cer16:0/Cer24:0 (H). Log-rank significance values are presented for the plots. (I) Adjusted hazard ratios (HRs) with 95% confidence intervals (CIs) derived from Cox regression after correction for age, sex, and smoking status.

Figure 8

Plasma Cer16:0 and GlcCer16:0 Are Modifiable Risk Biomarkers Kaplan-Meier estimates of cumulative probability of cardiac death by years of follow-up and levels of circulating C16:0-ceramide (Cer16:0) (A), C16:0-glycosylceramide (GlcCer16:0) (B), Cer16:0/C22:0-ceramide (Cer22:0) (C), and Cer16:0/C24:0-ceramide (Cer24:0) (D) and estimates of noncardiac death by years of follow-up and levels of circulating Cer16:0 (E), GlcCer16:0 (F), Cer16:0/Cer22:0 (G), and Cer16:0/Cer24:0 (H). Log-rank significance values are presented for the plots. (I) Adjusted hazard ratios (HRs) with 95% confidence intervals (CIs) derived from Cox regression after correction for age, sex, and smoking status.

Figure 9

Plasma Cer16:0 and GlcCer16:0 Are Modifiable Risk Biomarkers (A) A schematic representation of the randomized controlled trial design. (B) A heat map showing the significant differential metabolites between day 0 and week 8. Box plots represent changes in GlcCer16:0 (C,D) and Cer16:0 (E,F) body mass index–normalized concentrations in patients who were randomized to a low-calorie diet or liraglutide treatment (1.2 mg/day) for 52 weeks, following 8 weeks (baseline) of low-calorie diet–induced weight loss. Box plots represent median (interquartile range). Abbreviations as in Figure 8.

Central Illustration

Proposed Mechanistic Effects of Fat-Derived C16:0-Ceramide Dysfunctional thoracic adipose tissue in obese cardiac patients secretes higher levels of C16:0-ceramide (Cer16:0), increasing its circulating levels. Acting via endothelial nitric oxide synthase signaling, circulating Cer16:0 enhances systemic inflammation, amplifies oxidative stress, and reduces endothelium-dependent vasorelaxation, eventually leading to adverse cardiac outcomes including death.