Long-term intermittent fasting restores function of the gut–liver axis and prevents development of diabetic retinopathy in db/db mice

Constant availability of food can contribute to the pathogenesis of metabolic syndrome and type 2 diabetes. Short term intermittent fasting (IF) can reset the central, light-entrained (suprachiastmatic nucleus) clock and also the peripheral, food-entrained (liver) clock to restore metabolic homeostasis in T2D. We asked if long term IF could prevent development of diabetic retinopathy (DR) in a type 2 diabetes model, the db/db mouse. After 7 months, IF corrected diabetes-induced increases in triglycerides, cholesteryl esters and diglycerides. IF protocol in db/db mice also prevented development of DR. In addition, host frequency and time of food intake affected the gut microbiome composition. IF led to decreased levels of Clostridiales and Akkermansia muciniphila in db/db mice and these changes in flora were accompanied by increased gut mucin, goblet cell number and villus length. Increased levels of Firmicutes in db/db mice on IF supported improved bile acid metabolism. To confirm that the restoration of bile acid function could contribute to the beneficial effects induced by IF on DR, the dual FXR/TGR-5 agonist INT-767 was administered to a second diabetes model, DBA2J mice injected with streptozotocin (STZ) and placed on Western diet (WD). In this model, INT-767 prevented development of DR. These findings support the concept that long-term IF mediates multiple beneficial effect by restoring the gut-liver axis homeostasis. Liver samples from 3 db/db mice per group were collected at zeitgeber (ZT) time ZT9 (daytime) and ZT21 (nighttime) from both the diabetic ad-lib and IF groups and were used for microarray analysis. Total RNA was extracted using miRNeasy Mini kit (Qiagen, Valencia, CA) and standard procedures were used for hybridization, staining and scanning GeneChip® Mouse Gene 2.0 ST Arrays (Affymetrix, Santa Clara, CA). CEL files were analyzed with Expression Console software (Affymetrix) and the statistical analysis was performed with Transcriptome Analysis Console (Affymetrix) and in R (version 2.15.3). A linear model with covariates of long-term IF effect, ZT time effect and feeding/fasting effect was built and the goodness of fit between the full model and null model was tested by likelihood ratio test. False discovery rate (FDR) correction was conducted to adjust for multiple testing. In this study, the genes with FDR less than 0.3 was selected for further post hoc analysis. Only genes with p values for the comparasion between long term IF treatment and ad-lib less than 0.05 and absolute fold-change larger than 1.5 were selected as statistically significantly expressed genes.