Fasting elicits gut microbiome signature changes that extend to type 1 diabetes patients

Abstract

The gut microbiome has been linked to the pathogenesis of type 1 diabetes (T1D), identifying it as a promising therapeutic target. Nutritional interventions, which are an effective way to modulate the gut microbiome, thus show potential to be applied as complementary therapies for T1D. One particular dietary intervention, prolonged therapeutic fasting, has been shown to ameliorate symptoms of several autoimmune diseases, while also modifying the gut microbiota composition of healthy populations. It is unclear, however, how the gut microbiota of patients suffering from diseases of autoimmunity will respond to fasting. In this pilot study, we investigate the effects of prolonged fasting on the gut microbiome of T1D patients: Fasting substantially changed the composition and structure of the T1D gut microbiome so that it converged with that of non-diabetic controls immediately post fasting. Moreover, a comparison with a population of patients suffering from Multiple Sclerosis revealed substantial overlap in post-fasted microbiome changes and a remarkable consistency with published data of non-autoimmune populations, indicating that fasting leads to signature microbiome changes that are independent of host health status and disease type. A correlation analysis between fasting-mediated microbiota modifications and changes in clinical parameters revealed several significant associations between the Oscillospiraceae and Lachnospiraceae families and cholesterol and blood pressure changes in the T1D cohort, corroborating previous studies reporting on these associations in non-diabetic subjects. In conclusion, the observed fasting-mediated microbiome signature suggests that nutrient availability is a major disease-independent factor in shaping gut microbiome composition, likely driven by the need for metabolic diversification of microbial nutrient acquisition. The corresponding clinical associations highlight the need to investigate if these fasting-driven changes in the reported taxa are causally linked to the recorded clinical benefits of therapeutic fasting and what importance fasting as an additional therapeutic intervention might have to improve long term conditions in people with T1D.

Keywords: autoimmunity; fasting; gut bacteria; gut microbiome; multiple sclerosis; nutrient availability; therapeutic fasting; type 1 diabetes.

Figure 1

Overview of study design for FAMED1 study. (A) Prolonged fasting for patients with type 1 diabetes mellitus (FAMED1) is a non-randomized controlled pilot study that enrolled a total of 30 people. (B) Timeline of stool sample collection. Figure created with BioRender.com.

Figure 2

Post-fasting gut microbiome composition of T1D patients and controls converging. Principal Coordinates Analysis (PCoA) depicting bacterial β-diversity measured by Bray-Curtis dissimilarity. Statistical differences between groups were calculated by PERMANOVA. Ellipses represent 95% confidence intervals. (A) FAMED1 intragroup changes over time. Baseline (d0), post-fast (d7), follow-up (d150). (B) FAMED1 between-group differences at indicated time points. Statistical differences between groups calculated by PERMANOVA. Ellipses represent 95% confidence intervals.*** p < 0.0001, ** p < 0.001, * p < 0.01. p-values are adjusted for multiple comparisons using the Benjamini-Hochberg method.

Figure 3

Fasting induces short-term significant shift in microbial composition. (A) PCoA plots showing Bray-Curtis dissimilarity by individual patients longitudinally. (B) Volatility analysis measuring distance between time points for individual patients using a matrix of Bray-Curtis distances. Fed-fed referring to distance between d0 and d150 (baseline and follow-up) and fed-fast referring to distance between d0 and d7 (baseline and post-fast). (C) PCoA plots showing Bray-Curtis dissimilarity grouped by nutritional state (“fed” considering samples from d0 and d150, “fasted” considering samples from d7). Statistical differences between groups calculated by PERMANOVA. Ellipses represent 95% confidence intervals. (D) Heatmap of differentially abundant taxa shown as the log₂(fold-change) between time points for each significant genus. ***p < 0.0001, **p < 0.001, *p < 0.01. p-values are adjusted for multiple comparisons using the Benjamini-Hochberg method.

Figure 4

Microbiome differential abundance changes are correlated with changes in clinical parameters over time. (A) Heatmap colored by Spearman rank correlation coefficients (Rho) assessing association of significantly changed genera identified in Figure 3D with changes in clinical variables for the FAMED1 study. Cells colored in grey indicate that the standard deviation in one of the variables is zero, therefore a correlation coefficient cannot be determined. *p < 0.05. p-values are adjusted for multiple comparisons using the Holm method. (B) Correlation plots for the taxa significantly associated with clinical parameters identified in Figure 4A . Linear graphs highlighted using a fat stroke display the intervention group with the significant association. (C) Table displaying top ten most significant associations provided for reference.

Figure 5

Differentially abundant genera form a microbiome fasting signature shared between the autoimmune diseases MS and T1D. (A, B) PCoA plots depicting bacterial β-diversity measured by Bray-Curtis dissimilarity considering all study participants from two studies (A) at baseline (d0) grouped by disease and (B) at all time points grouped by nutritional status with “fed” referring to combined samples from baseline and follow-up. Statistical differences between groups calculated by PERMANOVA. Ellipses represent 95% confidence intervals (C) Heatmap of the common differentially abundant taxa shown as the log₂(fold-change) between time points for each genus that was significantly changed in both studies (same data as 3D and S4B but with standardized scale). ***p < 0.0001, **p < 0.001, *p < 0.01. p-values are adjusted for multiple comparisons using the Benjamini-Hochberg method. (D) Venn-Diagram of genera that are significantly changed immediately post-fasting compared to baseline (d0) in each group and change in relative abundance of the 7 conjointly changed taxa over time for all study participants. (E) Venn-Diagrams of common ASVs that are significantly changed in the respective group and time point.