Weight-loss maintenance is accompanied by interconnected alterations in circulating FGF21-adiponectin-leptin and bioactive sphingolipids

Abstract

Weight loss is often followed by weight regain. Characterizing endocrine alterations accompanying weight reduction and regain may disentangle the complex biology of weight-loss maintenance. Here, we profile energy-balance-regulating metabokines and sphingolipids in adults with obesity undergoing an initial low-calorie diet-induced weight loss and a subsequent weight-loss maintenance phase with exercise, glucagon-like peptide-1 (GLP-1) analog therapy, both combined, or placebo. We show that circulating growth differentiation factor 15 (GDF15) and C16:0-C18:0 ceramides transiently increase upon initial diet-induced weight loss. Conversely, circulating fibroblast growth factor 21 (FGF21) is downregulated following weight-loss maintenance with combined exercise and GLP-1 analog therapy, coinciding with increased adiponectin, decreased leptin, and overall decrements in ceramide and sphingosine-1-phosphate levels. Subgroup analyses reveal differential alterations in FGF21-adiponectin-leptin-sphingolipids between weight maintainers and regainers. Clinically, cardiometabolic health outcomes associate with selective metabokine-sphingolipid remodeling signatures. Collectively, our findings indicate distinct FGF21, GDF15, and ceramide responses to diverse phases of weight change and suggest that weight-loss maintenance involves alterations within the metabokine-sphingolipid axis.

Keywords: FGF21; GDF15; GLP-1 receptor agonist; adiponectin; ceramide; cytokines; exercise; leptin; obesity; sphingolipids.

Graphical abstract

Figure 1

Circulating GDF15 transiently increases upon diet-induced weight loss whereas FGF21 decreases following weight-loss maintenance with combined exercise and GLP-1RA therapy (A) Schematic overview of the study design. (B and C) Levels of and changes in serum FGF21 and GDF15 as measured by ELISA in fasting blood samples. Time-course data are presented as estimated means ±95% confidence limits. Fold changes are expressed as the log₂ fold change relative to baseline (week −8) and are presented as observed individual values with estimated means ±95% confidence limits. Constrained linear mixed models were used to estimate within- and between-treatment differences. ∗Significant within-treatment change (p < 0.05).

Figure 2

Circulating adiponectin and leptin profiles are ameliorated following weight-loss maintenance with exercise, GLP-1RA, and their combined treatment Levels of and changes in serum adiponectin (A) and leptin (B) as measured by ELISA and radioimmunoassay (RIA), respectively, in fasting blood samples. Time-course data are presented as estimated means ±95% confidence limits. Fold changes are expressed as log₂ fold change relative to baseline (week −8) and are presented as observed individual values with estimated means ±95% confidence limits. Constrained linear mixed models were used to estimate within- and between-treatment differences. ∗Significant within-treatment change (p < 0.05).

Figure 3

Individual ceramide and S1P species are reduced following weight-loss maintenance with GLP-1RA alone or combined with exercise (A) Schematics of sphingolipid metabolism pathway. Italic numbers indicate the individual lipid species identified for each class of sphingolipids. (B and C) Volcano plot and heatmap showing changes, expressed as the log₂ fold change relative to baseline (week −8), in plasma levels of individual sphingolipid species in response to the 8-week low-calorie diet. Gray color in heatmap denotes non-significant changes. (D–G) Volcano plots showing within-treatment changes, expressed as the log₂ fold change relative to baseline (week −8), in plasma levels of individual sphingolipid species in response to the 52-week weight-loss maintenance treatments. (H) Venn diagrams showing the overlap between treatment-induced significant changes in plasma sphingolipids. (I) Heatmap showing changes in plasma levels of individual sphingolipid species, expressed as log₂-transformed fold changes relative to baseline (week −8), in response to the 52-week treatment with placebo (P), exercise (E), liraglutide (L), and combined exercise and liraglutide (C). Constrained linear mixed models were used to estimate within-treatment changes and between-treatment differences. Gray color denotes non-significant within-treatment changes. ∗Significant between-treatment difference (p < 0.05). (J) Hierarchical clustering of changes in individual sphingolipid species in response to the weight-loss maintenance phase. Each column represents a study participant.

Figure 4

Differential alterations in FGF21, adiponectin, leptin, and individual sphingolipid species in weight maintainers and weight regainers (A) Schematics of the weight “maintainers” and “regainers” subgroup analysis. (B–E) Levels of and changes in circulating metabokines in weight maintainers and weight regainers. Time-course data are presented as estimated means ±95% confidence limits. Changes were computed as log₂-transformed fold changes (Log₂FC) at the end of treatment (week 52) relative to baseline (week −8) and are presented as observed individual values with estimated means ±95% confidence limits. (F) Changes in circulating sphingolipids in weight maintainers and weight regainers. Changes were computed as log₂-transformed fold changes (Log₂FC) at the end of treatment (week 52) relative to baseline (week −8). Linear mixed models were used to estimate within-group changes and between-group differences. ∗Significant within-group change (p < 0.05). ^#Significant between-group difference (p < 0.05).

Figure 5

Wide range of metabokine-sphingolipid associations Correlations between baseline (week −8) levels of and changes (log₂-transformed fold change at the end of treatment (week 52) relative to baseline (week −8) in circulating metabokines and sphingolipids. For this analysis, all four weight-maintenance treatment groups were pooled together. Only significant correlations (p < 0.05) are shown (gray color denotes non-significant correlations).

Figure 6

Remodeling of the circulating metabokine-sphingolipid profile is associated with changes in cardiometabolic health outcomes Associations between changes in circulating metabokine-sphingolipid levels and markers of cardiometabolic health. Changes were computed as the log₂-transformed fold change at the end of treatment (week 52) relative to baseline (week −8). For this analysis, all four weight-maintenance treatment groups were pooled together. Linear regression models adjusted for treatment group, age, gender, and baseline value of the outcome variable were used to calculate standardized β regression coefficients. Only significant associations (p < 0.05) are shown (gray color denotes non-significant associations). BW, body weight; FM, fat mass; FFM, fat-free mass; W/H, waist-to-hip ratio; GLU, fasting glucose; INS, fasting insulin; CHOL, total cholesterol; HDL, high-density lipoprotein cholesterol; LDL, low-density lipoprotein cholesterol; VLDL, very-low-density lipoprotein cholesterol; TGL, triglycerides; ALAT, alanine aminotransferase; hs-CRP, high-sensitivity C-reactive protein; SBP, systolic blood pressure; DBP, diastolic blood pressure; RHR, resting heart rate.