Reshaping lipid metabolism with long-term alternate day feeding in type 2 diabetes mice

Impact of IF on the bioenergetics in diabetic mice
Diabetic-IF mice did not exhibit weight loss (Fig. 2A) or any changes in body fat (Fig. 2B), but they had a reduction in lean mass (Fig. 2C). Control- IF gained fat mass. Both groups on IF ate more food during their feeding phase (Supplemental Fig. 1A), but diabetic IF mice did not differ in their activity levels with treatment, in contrast to control IF mice that showed higher activity with IF (Supplemental Fig. 1B).

Mice (n = 8) from each cohort (D-AL: diabetic-ad-lib, D-IF: diabetic-intermittent fasting, C-AL: control-ad-lib, and C-IF: control-intermittent fasting) were assessed for body composition and indirect calorimetry after a 48-hr acclimation period followed by 48 hr. of data collection every 10 min. (A) Gross body weight (gr), (B) percentage of body fat, and (C) percentage of lean mass analyzed using the EchoMRI Analyzer. (D) Energy expenditure, (E) respiratory exchange ratio (RER), (F) carbohydrate utilization, and (G) lipid utilization were assessed with the TSE systems LabMaster Metabolism Research Platform. Two-way ANOVA was used for (A–C), *p < 0.05, ****p < 0.0001. ANOCOVA was used for (D–G).
Metabolic rate, measured as EE, showed that diabetic-IF exhibited a significant reduction of EE (D-IF vs D-AL p < 0.001) (Fig. 2D), which was not observed in the C-IF group. Despite this, the RER (Fig. 2E), which indicates substrate utilization, was increased in both IF groups (D-IF vs D-AL p < 0.001, C-IF vs C-AL p = 0.0126). An increase in carbohydrate utilization was observed in both IF groups compared to their AL counterparts (D-IF vs. D-AL p < 0.001, C-IF vs. C-AL p = 0.01) (Fig. 2F), but lipid utilization remained unchanged (Fig. 2G). Collectively, these results support that IF in diabetic mice did not result in weight loss, possibly due to the reduction in EE observed; however, the D-IF mice showed improved RER and carbohydrate utilization.
Impact on glucose homeostasis and insulin sensitivity
We then examined whether IF improved glucose homeostasis in the diabetic cohort. D-IF mice had reduced glycated hemoglobin (Fig. 3A), improved random glucose (Fig. 3B), reduced fasting glucose (Fig. 3C) and better glucose tolerance (Fig. 3D, E). HOMA-IR, a surrogate for insulin resistance, suggests that IF improved insulin sensitivity in the diabetic cohort (Fig. 3F). To examine if IF increased insulin production, pancreatic sections were stained for insulin immunoreactivity. D-IF mice presented significantly more insulin-positive cells compared to D-AL mice suggesting that beta cells are generating more insulin (Fig. 3G, H). Basal levels of insulin in the blood, however, were similar between D-AL and D-IF (Fig. 3I). Collectively, this data show that chronic IF improved glucose homeostasis in db/db mice by increased insulin sensitivity and glucose clearance.

Mice (n = 6-8) from D-AL: diabetic -ad-lib, and D-IF: diabetic-intermittent fasting) were assessed for (A) glycated hemoglobulin, (B) random glucose, and (C) fasting glucose measured after a 6-hour fast during the day. (D) Glucose tolerance test (GTT) was performed following a 16 hr fast and intraperitoneal injection of 2 g/kg lean mass of D-glucose solution, and (E) area under the curve for GTT was calculated. (F) HOMA-IR was calculated with fasting glucose and insulin levels in blood taken from mice during midday after 16 hour fast. (G) Insulin-positive cells in fixed pancreatic sections (5 sections per mouse). (H) The percentage of insulin-positive area/ pancreatic area was calculated. (I) Circulating basal insulin levels in blood. T-tests, asterisks indicate statistical difference *p < 0.5, **P < 0.01, ****p < 0.0001.
Impact on white adipose tissue (WAT)
We next asked whether IF could correct the dysfunction and inflammation typically observed in WAT in mice with metabolic syndrome36. Adipocytes from D-IF mice were of similar size as adipocytes from D-AL mice (Fig. 4A, B). Consistent with WAT insulin resistance, D mice had decreased adiponectin levels, which were not restored by IF (Fig. 4C). Expression levels of key genes that regulate lipolysis (lipe, pnpla2) and genes related to fatty acid synthesis (fasn and acaca) were examined (Fig. 4D). As expected, lipe was increased during the fasting phase of the IF regimen in the diabetic cohort but pnpla2. No changes were observed with fasting in fasn and acaca. In contrast, these genes were significantly increased in the control IF cohort. Collectively, these data suggest that lipolysis was increased in the D-IF mice during fasting, but overall lipid metabolism remained dysregulated in WAT.

Epididymal fat and blood samples were collected from mice (n = 5) from each cohort (D-AL: diabetic-ad-lib, D-IF: diabetic-intermittent fasting, C-AL: control-ad-lib, and C-IF: control-intermittent fasting) after 6 months of IF for further assessment. Samples were all collected at the same time of the day (A) H&E staining of fixed WAT. B adipocyte size shown in bars was measured as adipocyte area. C adiponectin levels in plasma measured by ELISA. D mRNA gene expression of lipe, pnpla2, fasn, and acaca in total RNA isolated from WAT at midday. E Mac-2 staining in fixed WAT and (F) counts of Mac2 positive crown like structures per 1mm2 WAT area. G mRNA expression of adgre1 (F480), as another indicator of macrophages. H mRNA gene expression of genes associated with M1 macrophages (atgax, nos2, tnfa, il-1b, and il-6) and genes associated with M2 macrophages (arg1, mrc2, mrc1, clec10a, and Il-10). I mRNA expression of chemokine Ccl2 in WAT. J Numbers of circulating monocytes in blood identified by flow cytometry. Two-way ANOVA: asterisks indicate statistical difference *p < 0.5, **P < 0.01, ***p < 0.001, ****p < 0.0001.
We next examined markers of WAT inflammation. Crown-like structures, identified as Mac2 staining, a hallmark of proinflammatory processes in WAT, showed a trended decrease in D-IF compared to D-AL mice (Fig. 4E, F). The mRNA expression of the F480 antigen, adgre1, confirms significant reduction of macrophages in the D-IF fasted cohort (Fig. 4G). To examine the polarization of WAT macrophages, we measured mRNA expression levels of M1 gene signature (atgax (Cd11c), nos2, tnfα, il-1β and il-6) and M2 (arg1, mrc2, mrc1, clec10a and il-10) (Fig. 4H). D-AL mice had increased expression of atgax, tnfa, arg1, and il10 compared to controls, and IF treatment, especially during the fasting period, reduced them. Gene expression for inflammatory cytokines, tnfα, il-1β and il-6, were increased during the feeding period in the IF treatment, however, they were significantly suppressed with fasting, implying that food availability was linked to an inflammatory state. From the M2 markers, similar regulation based on food availability was observed for mrc1 and il10, with increased expression during the feeding period and suppression during fasting.
We then examined whether this temporal regulation of inflammation was linked with the kinetics of monocyte infiltration to WAT. To accomplish this, we examined the expression of ccl2 mRNA, a chemokine that dictates inflammatory monocyte trafficking into tissues. Cccl2 mRNA expression was significantly increased in WAT under diabetic conditions and remained elevated during the feeding period but was significantly suppressed during the fasting period in D-IF mice (Fig. 4I). Similarly, the levels of circulating CCR2+ monocytes in both IF cohorts were depleted from the circulation during the fasting period of IF (Fig. 4J). Altogether, these results suggest chronic IF during fasting showed that one lipolytic gene was increased, suggesting at least some restoration of the physiological response to fasting; however, gene expression for other enzymes did not change supporting persistent metabolic dysfunction. Inflammation of WAT only occurred in diabetic mice and showed that WAT had fewer resident macrophages and inflammation mostly during the fasting period.
IF affects sphingolipid metabolism in the gut microbiome
Previously, we demonstrated that IF cohorts had a unique microbiota signature with increased microbial species that generated secondary bile acids such as TUDCA, which reduced inflammation in the retina and prevented diabetic retinopathy12. Using this microbiome data, we performed a predictive functional microbiome analysis and demonstrated that the fasting microbiome was enriched in bacterial species predicted to regulate energy metabolism, phosphatase and tensin homolog (PTEN), sphingolipid metabolism, N-glycan biosynthesis, and RNA degradation. In contrast, bacteria predicted to participate in iron transport, phosphotransferase and phenylalanine metabolism were reduced (Fig. 5A–C). This data brings attention to the importance of including changes in microbiota when assessing host metabolic responses. Further emphasizing this point is the knowledge that bacterial sphingolipids can be processed via mammalian sphingolipid pathways37.

D-AL: diabetic -ad-lib, D-IF: diabetic-intermittent fasting (Feed/Fast). 16S rRNA sequences from samples collected at the same time of the day were selected for the analysis. Samples were collected at 6 months of IF treatment (n = 5). (A) heatmap of KEGG pathways from 16S ribosomal RNA sequences from faeces isolated in D-AL, D-IF in feeding and in fasting (daytime) indicate clustering based on the feeding/fasting cycle. (B) Linear discriminant analysis effect size with logLDA score <1.5. (C) averaged abundance counts for each KEGG pathway shown in B.
IF results in unique blood metabolic signature in diabetic mice
The complexity of interpretation of microbiome studies leads us to perform a global metabolomics analysis on blood samples isolated during the day (ZT5) in the four cohorts. The presence of diabetes affected 422 metabolites. IF treatment affected 178 metabolites (Fig. 6A and Supplemental Table 1). When D-IF-Feed mice were compared to D-AL, beta-oxidation of medium and very-long chain fatty acids, sphingolipid metabolism and general fatty acid biosynthesis was reduced, while bile acid, steroid and glycerolipid biosynthesis pathways were increased (Fig. 6B). When D-IF-Fasting mice were compared to D-IF-Feeding, as expected oxidation of medium and very long chain fatty acids was upregulated, and many biosynthesis pathways were reduced (Fig. 6B). Both D and C groups utilized beta-oxidation during fasting, as 3-Hydroxybutyrate (BHBA) was significantly elevated compared to non-fasted samples and returned to a lower level during the re-feeding period (Fig. 6C). D-AL mice are expected to have increased lipid utilization, but while BHBA was not found to be as elevated in the AL stage compared to the IF-fast state, acetyl-carnitines were increased (Supplemental Fig. 2A–C) and, with long-term IF, they were significantly reduced in the fed state and only increased in the fasted state (Supplemental Fig. 2C), indicating an improvement in fatty acid metabolism with IF. The D-IF fasted mice had elevated levels of long-chain fatty acids in the blood, supporting active lipolysis from WAT (Fig. 6D). Moreover, the metabolomics analysis reinforced that sphingolipid metabolism is affected by IF (Fig. 6E–I). Total ceramide species, including hexylceramides, and ceramides containing C16 and C18 acyl chains, were significantly reduced in D-IF compared to D-AL, especially at the fast state. Collectively, these data strongly support that lipid metabolism was dramatically impacted by IF in the diabetic cohort.

Plasma samples were collected from mice (n = 5) from each cohort (D-AL: diabetic-ad-lib, D-IF: diabetic-intermittent fasting, C-AL: control-ad-lib and C-IF: control-intermittent fasting) after 6 months of IF at midday (ZT5). Global metabolomics analysis performed by Metabolon Inc. (A) summary of statistical analysis of the identified metabolites using a global untargeted metabolomics platform. (B) Enrichment analysis of the metabolites that were found to be differentially expressed in control and diabetes during their response to IF-feeding and IF-fasting. (C) levels of 3-hydroxyburate, (D) levels of long chain fatty acids and (E–H) levels of components of sphingolipid metabolism. Data were normalized to sample volume extracted (vol normalized). Two-way ANOVA: asterisks indicate statistical difference *p < 0.5, **P < 0.01, ***p < 0.001, ****p < 0.0001. p < 0.05.
Impact on liver lipid composition
The liver is central to both lipid and glucose metabolism, we therefore examined the effects of IF on liver pathology and lipid content. Liver steatosis scores26 were significantly elevated compared to control and they showed a trend toward improvement with IF (Fig. 7A, B). Using mass spectrometry for both polar and non-polar lipids32, we observed striking changes in lipid content with the IF treatment (Fig. 7C and Supplemental Table 2) D-IF mice in the fed state show the appropriate reduction in lipid species compared to D-AL reinforcing the notion that IF results in better lipid handling by the diabetic liver. Livers from D-AL mice had increased levels of most lipid species compared to healthy control mice apart from sphingomyelins (Fig. 7D). An influx of lipids is observed in the diabetic livers during the fasting state, possibly because of increased WAT lipolysis, however, these lipids were vastly different compared to D-AL (Fig. 7E). All phospholipids, oxidized lipids, and hexylceramides were significantly reduced, especially in the fed state, but they remained reduced during fasting (Supplemental Fig. 3), despite the influx of lipids in the liver. Moreover, total triglycerides (Fig. 7G), total diglycerides (Fig. 7H), cholesterol esters (Fig. 7I) and total hexyl-ceramides (Fig. 7H) were dramatically reduced in the D-IF fed compared to D-AL mice. A detailed comparison of the glycosylated ceramide species suggests reductions of those containing C24:0 and C24:1 long-chain fatty acids (Fig. 7H). Taken together, alternate feeding/fasting cycles significantly impacted lipid metabolism in the diabetic liver, reduced levels of all classes of lipids, oxidized lipids, and ceramides, and dramatically improved lipid composition in the fed state, indicative of improved insulin sensitivity.

Liver samples were collected from mice from each cohort (D-AL: diabetic -ad-lib, D-IF: diabetic-intermittent fasting, C-AL: control-ad-lib and C-IF: control-intermittent fasting) after 6 months of IF. (A) H&E staining of fixed liver tissue. (B) Liver steatosis scores (n = 5). One-Way ANOVA *p < 0.05, ***p < 0.001. (C) Heatmap of lipid classes identified by high resolution/accurate mass spectrometry from liver lipid extracts. Samples (n = 3) collected at 6 time points throughout day /night. Data was normalized to sample weight extracted, and z-scores were mapped. (D) Significant differential lipid species in D-AL vs C-AL and D-IF-Feeding vs D-AL. (E) Significant differential lipid species between D-AL vs D-IF-Fasting. Summary of all (F) triglycerides, (G) diglycerides, (H) cholesteryl esters (CE), and (I) hexosylceramides and abundance of individual HexCer molecular species. Data are shown after log2 transformation of the fold change of D-AL over C-AL (open bars) and D-IF fed over D-AL mice (closed bars). Lipid species are presented as total fatty acid carbons: total double bonds. For (B–I) data represent daytime (ZT9).
IF has pleiotropic effects on liver lipid metabolism
To better understand the mechanisms responsible for the changes we observed in the D-IF liver, we examined gene expression under feeding and fasting conditions and compared the changes to D-AL. Using microarray studies of livers, we found a total of 650 genes that were differentially expressed in pairwise comparisons demonstrating a higher than two-fold change (Supplemental Table 3). The majority of differentially expressed genes (503 genes) were driven by the feed/fast cycle. In contrast, only 95 differentially expressed genes were observed when both groups were in the fed state. Heatmap (Fig. 8A) and PCA analysis (Fig. 8B) confirmed that most differences were driven by the feeding/fasting cycle. The D-AL liver is closer to the D-IF liver during the feeding cycle than the fasting showing fewer gene changes (Fig. 8C) than when comparing D- IF fed to fasting (Fig. 8D).

Liver samples were collected from mice (n = 3) from each cohort (D-AL: diabetic -ad-lib, D-IF: diabetic-intermittent fasting, C-AL: control-ad-lib and C-IF: control-intermittent fasting) after 6 months of IF. Liver microarray analysis was done for diabetic samples (n = 3/cohort) collected at ZT9. (A) heatmap and cluster analysis of the differential genes in livers from diabetic mice after IF compared to AL conditions. (B) Principal Component Analysis of differentiated genes. (C) Volcano plots showing genes that were significantly changed in IF-fed vs. AL conditions. (D) Volcano plots showing genes that were significantly changed in Diabetic IF-fed vs. IF fasted conditions. (E) Comparative ingenuity pathway analysis of upstream regulators of the differentially expressed genes. (F) Graphical summary of long-term effects of IF on function of liver in db/db mice. Left panel: D-AL mice exhibit insulin resistance, impaired glucose tolerance, decreased insulin sensitivity, dysbiosis, increased inflammation, and ectopic lipid accumulation. Right panel: In the absence of weight loss, D-IF mice experienced enhanced glucose tolerance and insulin sensitivity reflected in increased carbohydrate utilization and reduced lipid accumulation in the liver during feeding. IF-treated diabetic mice showed reduced inflammatory pathways during the fasting state. Functional prediction analysis of the gut microbiome showed that D-IF mice had enrichment in bacterial pathways involved in sphingolipid metabolism. Plasma and liver lipidomics demonstrated an overall reduction of lipids, in particular ceramide levels and oxidized lipids. Oxidation of fatty acids was dramatically reduced, indicating increased insulin sensitivity, better mitochondrial metabolism and reduced reactive oxygen species resulting in improved overall lipid metabolism.
D-IF during the fed state resulted in upregulation of liver genes that belonged to pathways of long-chain fatty acid metabolic processes, fatty acid metabolism, PPAR signalling, arachidonic and eicosanoid metabolism, regulation of phagocytosis, adaptive immunity, neutrophil degradation lipid transport and distribution (Obpa2a, Atp8b4), and immune regulation (Sirpb1, Sirbp2) compared to D-AL. Genes that were downregulated in D-IF-feed mice compared to D-AL mice were related to peroxisomal and fatty acid metabolism- especially those of long-chain fatty acids (Hao2, Acot3, Acot4, Acaa1b, Me1, Mogat1, Cyp2a4, Cyp2a22)- as well as cholesterol metabolic processes (Ces1e, Lepr, Tsku). Collectively, these data indicate that D-IF mice during the fed state did not need to rely as much on perixosomal oxidation and lipid oxidation compared to the D-AL mice.
When comparing D-IF in the fed state to the fasting state (Fig. 8D), D-IF mice showed downregulation of genes related to fatty acid metabolism (Elovl3, Crat, Cyp2a4, Cyp2b13, Ces1e, Acot1, Acot2, Acot3, Acot4, Acaa1b, Hao2, Pdk4, Plin5, Eci3, Insig2, Ehhadh, Cyp8b1), PPAR signaling (Aqp7, Cy4a10, Angptl4, Ehhadh, Acaab1, Cyp4a31, Cyp4a32) and lipid peroxidation (Crat, Pdk4, Hao2, Plin5, Eci3, Ehhadh, Acaa1b). On the other hand, D-IF fed mice showed increased expression of inflammatory genes, belonged to immune cell activation, migration and inflammation and cholesterol metabolism (Cyp51, Fasn, Hmgcr, Hmgcs1, Lss, Scd2, Soat1, Sqle, Srefb1, Msm1, Pmvk3, Acsl3, Lbr).
To identify the cascade of upstream regulators that could explain the observed gene changes, we used the Ingenuity upstream regulator analysis in IPA (Fig. 8E). When comparing D-IF in the fed state to D-AL mice, transcriptional factor proteins that were predicted to be upstream regulators included those for insulin and growth factor signaling (PI3K, STAT3, HNF4A), cytokine signaling (NFκΒ, STAT1, STAT4, STAT5B), immune cell differentiation (GATA2) and factors influencing the circadian clock (HNF4A, BHLE40). On the other hand, transcriptional factor proteins PPARA and transcriptional repressor ZBTB16 were predicted to be downregulated. The reduction of PPARA in the fed state is in accordance with our lipidomic and transcriptomics analysis showing that the liver of D- IF mice had a significant reduction of fat oxidation supporting improving carbohydrate utilization. However, inflammatory signaling remained elevated, like what was observed in the white adipose tissue, suggesting that inflammatory pathways were suppressed only during the fasting phase. Long-term IF is predicted to upregulate key lipid regulators such as PPARA, PPARG, PPARD, and PPARGC1A. On the other hand, insulin signaling was reduced during fasting in the absence of nutrients. Key predicted lipid master regulators were validated by qRT-PCR (Supplemental Fig. 4). Altogether, these data show that long-term imposition of restricted feeding/fasting cycles provides transcriptional flexibility linked to nutrient availability, such as genes linked to insulin signaling, inflammation, and lipid and cholesterol metabolism.
link