Insulin and leptin oscillations license food-entrained browning and metabolic flexibility

Abstract

Timed feeding drives adipose browning, although the integrative mechanisms for the same remain unclear. Here, we show that twice-a-night (TAN) feeding generates biphasic oscillations of circulating insulin and leptin, representing their entrainment by timed feeding. Insulin and leptin surges lead to marked cellular, functional, and metabolic remodeling of subcutaneous white adipose tissue (sWAT), resulting in increased energy expenditure. Single-cell RNA-sequencing (scRNA-seq) analyses and flow cytometry demonstrate a role for insulin and leptin surges in innate lymphoid type 2 (ILC2) cell recruitment and sWAT browning, since sWAT depot denervation or loss of leptin or insulin receptor signaling or ILC2 recruitment each dampens TAN feeding-induced sWAT remodeling and energy expenditure. Consistently, recreating insulin and leptin oscillations via once-a-day timed co-injections is sufficient to favorably remodel innervated sWAT. Innervation is necessary for sWAT remodeling, since denervation of sWAT, but not brown adipose tissue (BAT), blocks TAN-induced sWAT remodeling and resolution of inflammation. In sum, reorganization of nutrient-sensitive pathways remodels sWAT and drives the metabolic benefits of timed feeding.

Keywords: CP: Metabolism; ILC2; browning; circadian; insulin; intermittent fasting; leptin; oscillations; time-restricted feeding.

Figure 1.. TAN feeding entrains insulin and leptin oscillations and rewires metabolic programs in sWAT

(A) Scheme depicting the twice-a-night (TAN) feeding intervention. Bulk RNA-seq analysis was done at the indicated time points in sWAT of C57BL/6J male mice fed ad libitum (ad-lib) or TAN for 6 months (n = 5). Period-wide pathway enrichment (BIOCARTA) for the top 100 upregulated genes and a chord diagram for significantly enriched pathways as a function of time are shown. Fold enrichment is represented as a heatmap, and *FDR (false discovery rate) < 0.05. (B) UpSet plot of samples in (A) depicts intersection sets for significantly enriched genes across the six time points. Each connection represents a distinct combination between six (Zeitgeber) ZT time points and the number of overlapping gene sets. (C) Enrichment network using the Reactome database for significantly upregulated genes in TAN-fed mice at early postprandial (12:00 a.m.) and late postprandial (4:00 a.m.) time points. Node size represents the enrichment score. Node color represents the significance. Salmon, FDR < 0.05; lemon, p < 0.05. The network edge represents the betweenness for genes associated with representative pathways. (D–F) Serum insulin (D) and leptin (E) at six time points from ad-lib or TAN-fed mice housed at ambient temperature in a 12 h/12 h light/dark cycle (D and E) and in a 24 h dark cycle (F), i.e., total darkness (TD). Feeding windows are indicated by salmon-colored boxes; n = 4–10. Every plot across 24 h shows the mean value (dots) per time point ± SEM. See also Figure S1.

Figure 2.. TAN feeding associates with improved metabolic flexibility in sWAT

(A) Bulk RNA-seq of sWAT at six ZT time points from ad-lib and TAN-fed mice indicated in Figure 1A (n = 4). Radar charts for period-wide oscillations in the expression of the top 5 metabolic elasticity genes in sWAT from ad-lib and TAN-fed mice are shown. Circle diameter denotes FPKM (fragments per kilobase of transcript per million mapped reads) values. (B) Period-wide cluster map of the top 20 metabolic elasticity genes in sWAT of ad-lib or TAN-fed mice (n = 4). Z-score-normalized values are plotted and implemented for hierarchical clustering. Sky blue denotes upregulation, and lemon yellow denotes downregulation. (C) Period-wide oscillations in gene expression determined by qPCR for representative genes in each cluster. The x axis denotes time. The y axis is FPKM value. (D) Correlation coefficient network diagrams of 20 metabolic elasticity genes from sWAT of ad-lib or TAN-fed mice at 12:00 p.m. Nodes represent individual genes. The edge denotes the number of significant positive and negative correlations. Edge thickness signifies Pearson’s correlation coefficient values. Red and green indicate positive or inverse correlations in gene expression. (E) Protein-protein interaction networks (bipartite) generated for metabolic elasticity gene sets using STRING. The subnetwork is selected on confidence score(>900) threshold. Degree and betweenness are denoted by node size and color. (F) Gene set enrichment analysis for metabolic elasticity genes at 12:00 p.m. using the KEGG database. Node size represents the enrichment score, while node color signifies the p value (<0.05) of the functional enrichment network. (G) Sankey bubble chart representation of metabolic elasticity genes and their associated enriched pathways. A corresponding bubble chart for pathway enrichment depicting gene ratio (x axis), −log₁₀ (p) (bubble color), and gene count (bubble size) is shown. Feeding windows are indicated by salmon-colored boxes.

Figure 3.. TAN feeding leads to cellular and functional remodeling of sWAT

(A) sWAT weight (wt) of regular chow diet (RD)- or high-fat diet (HFD)-fed C57BL/6J male mice subjected to ad-lib (n = 10) or TAN feeding (n = 5) for 6 months. (B) Representative images of adipocytes generated via light-sheet microscopy and 3D reconstruction of whole sWAT (inguinal-dorsal) from C57BL/6J male mice subjected to ad-lib or TAN feeding on RD for 5 months (n = 5). Adipocyte volumes are represented in color scale (μm³) with larger adipocytes in red and smaller adipocytes in purple. Graph shows the mean adipocyte size. See Videos S1 (ad-lib) and S2 (TAN) for 3D constructions of sWAT from ad-lib and TAN-fed mice. (C) Representative H&E and immunostaining of sWAT from C57BL/6J male mice subjected to ad-lib or TAN feeding for 6 months showing adipocyte size (top) and UCP1 positivity (bottom). (D) Representative images and quantification of CD31-positive blood vessels generated via light-sheet microscopy of whole sWAT (inguinal-dorsal) from C57BL/6J male mice subjected to ad-lib or TAN feeding on RD for 5 months. White rectangles are magnified in insets below. See Videos S3 (ad-lib) and S4 (TAN) for 3D constructions of sWAT from ad-lib and TAN-fed mice (n = 5). (E) Representative images and quantification of tyrosine hydroxylase (TH)-positive nerves generated via light-sheet microscopy of whole sWAT (inguinal-dorsal) from C57BL/6J male mice fed ad-lib or TAN on RD for 6 months (n = 5). (F) Area under the curve (AUC) for oxygen consumption rate (OCR) in sWAT and BAT of C57BL/6J male mice subjected to ad-lib (RD n = 5, HFD n = 5) or TAN (RD n = 4, HFD n = 5) feeding on RD or HFD for 6 months (n = 4–5). (G) AUC for OCR in sWAT, BAT, and eWAT of Con or Ucp1^+/− male mice fed ad-lib or TAN and maintained in thermoneutrality (30°C) for 6 months (n = 5). (H) AUC for OCR in sWAT collected from ad-lib and TAN-fed RD mice at the six indicated time points (n = 5). Feeding windows are salmon-colored boxes. Values are mean ± SEM. Dot plots show individual values (dots) and mean (line). n.s., non-significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Two-way ANOVA and Tukey-corrected (A, F, G) and two-tailed unpaired Student’s t test (B, D, E). Magnifications bars are shown. See also Figure S2.

Figure 4.. Innervation-dependent sWAT remodeling and metabolic flexibility in TAN mice

(A) Scheme showing innervated control (Con) or sWAT denervated (DNV^sWAT) C57BL/6J male mice fed ad-lib or TAN on HFD for 3 months. (B and C) Weights of sWAT (g/g body wt) (B) and eWAT (g/g body wt) (C) from Con or DNV^sWAT C57BL/6J male mice fed ad-lib (n = 6 Con, n = 5 DNV^sWAT) or TAN (n = 6 Con, n = 6 DNV^sWAT) on HFD for 3 months. (D and E) AUC for OCR in sWAT (D) and BAT (E) from Con or DNV^sWAT C57BL/6J male mice fed ad-lib (n = 6 Con, n = 4 DNV^sWAT) or TAN (n = 5 Con, n = 6 DNV^sWAT) on HFD for 3 months. (F) qPCR for metabolic elasticity genes in sWAT from Con or DNV^sWAT C57BL/6J male mice fed ad-lib (n = 6 Con, n = 5 DNV^sWAT) or TAN (n = 6 Con, n = 6 DNV^sWAT) on HFD for 3 months. (G) Scheme showing innervated control (Con) or BAT denervated (DNV^BAT) C57BL/6J male mice fed ad-lib (n = 6 Con, n = 7 DNV^BAT) or TAN (n = 6 Con, n = 7 DNV^BAT) on HFD for 3 months. (H and I) Weights of sWAT (g/g body wt) (H) and eWAT (g/g body wt) (I) from Con or DNV^BAT C57BL/6J male mice fed ad-lib or TAN on HFD for 3 months (n = 6). (J and K) AUC for OCR in sWAT (J) and BAT (K) from Con or DNV^BAT C57BL/6J male mice fed ad-lib (n = 5–6 Con, n = 5–6 DNV^BAT) or TAN (n = 5–6 Con, n = 5–6 DNV^BAT) on HFD for 3 months. (L) Representative F4/80 (red) staining in eWAT from control (Intact^BAT) and DNV^BAT C57BL/6J male mice fed ad-lib (n = 9 Con, n = 4 DNV^BAT) or TAN (n = 9 Con, n = 6 DNV^BAT) on HFD for 3 months. Quantification for number of F4/80⁺ cells/field (one section/mouse observed with 5× original magnification) is shown. Arrowheads indicate F4/80⁺ crown-like structures. (M) Representative F4/80 (red) staining in eWAT from control (Intact^eWAT) and DNV^sWAT C57BL/6J male mice fed ad-lib (n = 5 Con, n = 4 DNV^sWAT) or TAN (n = 6 Con, n = 6 DNV^sWAT) on HFD for 3 months (n = 4–9). Quantification for number of F4/80⁺ cells/field (one section/mouse observed with 5× original magnification) is shown. Dot plots show individual values (dots) and mean (line); n.s., non-significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Two-way ANOVA and Tukey-corrected. See also Figure S3.

Figure 5.. scRNA-seq analyses reveal ILC2 cell recruitment in innervated sWAT of TAN mice

(A) t-distributed stochastic neighbor embedding (t-SNE) plots showing cell clustering of sWAT stromal vascular fractions (SVF) from C57BL/6J male mice fed ad-lib or TAN on RD for 5 months (n = 4 each group). (B) t-SNE subclustering of immune cells from sWAT SVF from C57BL/6J male mice fed ad-lib or TAN on RD for 5 months (n = 4). (C) Volcano plot showing up- and downregulated genes in sWAT ILC2 cells in TAN-fed mice compared to ad-lib mice (n = 4). (D and E) Representative contour plots for abundance (D) and quantification for ILC2 cells (E) in sWAT SVF from C57BL/6J male Con or DNV^sWAT mice fed ad-lib (n = 6 Con, n = 4 DNV^sWAT) or TAN (n = 6 Con, n = 5 DNV^sWAT) on HFD for 3 months. (F) t-SNE subclustering of adipocyte progenitor cells (APCs) from sWAT SVF of C57BL/6J male mice fed ad-lib or TAN on RD for 5 months (n = 4). (G) Identity analysis for IL-33 levels in APCs in sWAT SVF from C57BL/6J male mice fed ad-lib or TAN on RD for 5 months (n = 4 each group). (H) IL-33 protein levels in sWAT from C57BL/6J male mice fed ad-lib (n = 6) or TAN (n = 5) on RD for 5 months. (I–K) Volcano plots of downregulated (green) and upregulated (blue) genes from APC2 (I), APC (J), and APC1 cells (K) in sWAT SVF of TAN-fed mice compared to ad-lib mice (n = 4 each group). (L and M) Representative contour plots of APCs in sWAT SVF from ad-lib and TAN-fed mice (L) and quantification for percentage APC population in sWAT SVF from ad-lib (n = 9 Con, n = 5 DNV^sWAT) and TAN-fed (n = 10 Con, n = 5 DNV^sWAT) or DNV^sWAT mice (n = 4–6). (N) Volcano plot of down- (green) and upregulated (blue) genes from bulk RNA-seq of sWAT of TAN-fed mice compared to ad-lib mice (n = 4 each group). (O) Bubble plot of Reactome showing the top 14 upregulated pathways. Bubble size represents number of genes per pathway, the y axis represents the percentage of enrichment, and bubble color represents −log₁₀ p. Arrows highlight metabolism-related pathways (n = 4 each group). Dot plots show individual values (dots) and mean (line); n.s., not significant; *p < 0.05. Two-way ANOVA and Tukey-corrected (E and M) and two-tailed unpaired Student’s t test (H). See also Figure S4.

Figure 6.. Insulin and leptin oscillations and ILC2 cells drive sWAT remodeling in TAN mice

(A) Model for feeding-induced insulin and leptin oscillations driving ILC2 recruitment to stimulate WAT remodeling. (B and C) Scheme (B) and body composition (C) (fat and lean mass [relative to body wt]) for Con or leptin^KO (Ob/Ob) male mice fed ad-lib (n = 6 Con, n = 6 Ob/Ob) or TAN (n = 6 Con, n = 6 Ob/Ob) on HFD (Con) or RD (Ob/Ob) for 3 months. (D) sWAT weight (g/body wt) from Con or Ob/Ob male mice fed ad-lib (n = 6 Con, n = 6 Ob/Ob) or TAN (n = 6 Con, n = 6 Ob/Ob) on HFD (Con) or RD (Ob/Ob) for 3 months. (E) AUC for OCR in sWAT from Con or Ob/Ob male mice fed ad-lib (n = 5 Con, n = 6 Ob/Ob) or TAN (n = 6 Con, n = 6 Ob/Ob) on HFD (Con) or RD (Ob/Ob) for 3 months (n = 5–6). (F) Representative contour plots/quantification for percentage GATA-3⁺;ST2⁺ ILC2 cells in sWAT from Con or Ob/Ob male mice fed ad-lib (n = 6 Con, n = 6 Ob/Ob) or TAN (n = 6 Con, n = 6 Ob/Ob) on HFD (Con) or RD (Ob/Ob) for 3 months. (G and H) Representative H&E (G) and F4/80-positive (H) cells in eWAT from Con or Ob/Ob male mice fed ad-lib or TAN on HFD (Con) or RD (Ob/Ob) for 3 months. Arrows (H) highlight crown-like structures. (I) Scheme showing InsR^KO mice and low-dose streptozotocin (STZ)-injected insulin-deficient C57BL/6J male mice (generated as depicted) and their corresponding sex- and age-matched controls. (J) AUC for OCR in sWAT from mice in (I). Con or InsR^KO mice were fed ad-lib (n = 8 Con, n = 5 InsR^KO) or TAN (n = 6 Con, n = 5 InsR^KO) on RD for 5 months. (K) AUC for OCR in sWAT from mice in (I). Con or STZ-injected mice were fed ad-lib (n = 12 Con, n = 12 STZ) or TAN (n = 12 Con, n = 13 STZ) on RD for 5 months. (L) AUC for OCR in BAT from mice in (I). Con or InsR^KO mice were fed ad-lib (n = 5 Con, n = 5 InsR^KO) or TAN (n = 5 Con, n = 5 InsR^KO) on RD for 5 months. (M) AUC for OCR in BAT from mice described in (I). Con or STZ-injected mice were fed ad-lib (n = 11 Con, n = 9 STZ) or TAN (n = 10 Con, n = 13 STZ) on RD for 5 months. (N) Quantification for percentage GATA-3⁺;ST2⁺ ILC2 cells in sWAT SVF from Con, InsR^KO, and Il33^KO mice fed ad-lib or TAN for 5 months (n = 9 in Con and n = 5 each in InsR^KO and Il33^KO groups). (O) AUC for OCR in sWAT from Con and Il33^KO mice fed ad-lib or TAN on RD for 5 months (n = 5 each group). (P) ELISA for IL-33 protein levels in sWAT (pg/μg of total protein) from Con and InsR^KO male mice fed ad-lib or TAN for 3 months (n = 5 each group). Dot plots show individual values (dots) and mean (line); n.s., not significant; *p < 0.05, **p < 0.01. Two-way ANOVA and Tukey-corrected. See also Figure S5.

Figure 7.. Reconstituting insulin and leptin oscillations recapitulates sWAT remodeling in mice

(A) Plan for re-creating circulating insulin and leptin surges in C57BL/6J male mice via once-a-day injections of insulin (0.16 IU/day) (n = 10) or leptin (5 mg/kg/day) (n = 9) or both (Co-T_x) (n = 9) at 9:30 p.m. for 1.5 months. Con mice received vehicle (n = 16). Co-T_x was administered in Con or DNV^sWAT (n = 5) mice. All groups were pair-fed to Co-T_x Con mice. (B) Serum insulin and leptin levels from mice described in (A) at the indicated time points in response to an acute injection of insulin (0.16 IU/day) or leptin (5 mg/kg/day). (C) Daily average food intake in mice described in (A) across 47 days of experiment (n = 5–15). (D) Fat and lean mass (normalized to body weight) for groups defined in (A) (n = 5–16). (E and F) sWAT and eWAT weight (g/body weight) for groups in (A) (n = 5–16). (G and H) AUC for OCR in sWAT (G) and BAT (H) for groups in (A) (n = 5–16). (I) IL-33 protein levels (pg/mg total protein) in sWAT of Con (n = 7), insulin (n = 7), leptin (n = 7), Co-T_x (n = 8), and DNV^sWAT Co-T_x (n = 5) mice. (J) Percentage of GATA-3⁺;ST2 ILC2 cells in sWAT SVF of Con (n = 7), insulin (n = 8), leptin (n = 8), Co-T_x (n = 8), and DNV^sWAT Co-T_x (n = 5) mice. (K) Representative H&E (top) and UCP1 (bottom) staining in sWAT of each group as in (A). (L and M) Representative immunoblot for indicated proteins in sWAT (L) of each group as in (A) and their corresponding quantifications (M) (n = 5–13). Ponceau is the loading control. (N) Model for feeding-induced insulin and leptin oscillations, wherein insulin oscillations increase mitochondrial mass and OXPHOS, while leptin oscillations recruit ILC2 cells to cooperatively promote sWAT browning and metabolic flexibility. Dot plots show individual values (dots) and mean (line). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. One-way ANOVA and Tukey-corrected (C, D, E, F, G, H, I, J, and M). See also Figure S6.