Calorie restriction improves metabolic state independently of gut microbiome composition: a randomized dietary intervention trial

Abstract

Background: The gut microbiota has been suggested to play a significant role in the development of overweight and obesity. However, the effects of calorie restriction on gut microbiota of overweight and obese adults, especially over longer durations, are largely unexplored.

Methods: Here, we longitudinally analyzed the effects of intermittent calorie restriction (ICR) operationalized as the 5:2 diet versus continuous calorie restriction (CCR) on fecal microbiota of 147 overweight or obese adults in a 50-week parallel-arm randomized controlled trial, the HELENA Trial. The primary outcome of the trial was the differential effects of ICR versus CCR on gene expression in subcutaneous adipose tissue. Changes in the gut microbiome, which are the focus of this publication, were defined as exploratory endpoint of the trial. The trial comprised a 12-week intervention period, a 12-week maintenance period, and a final follow-up period of 26 weeks.

Results: Both diets resulted in ~5% weight loss. However, except for Lactobacillales being enriched after ICR, post-intervention microbiome composition did not significantly differ between groups. Overall weight loss was associated with significant metabolic improvements, but not with changes in the gut microbiome. Nonetheless, the abundance of the Dorea genus at baseline was moderately predictive of subsequent weight loss (AUROC of 0.74 for distinguishing the highest versus lowest weight loss quartiles). Despite the lack of consistent intervention effects on microbiome composition, significant study group-independent co-variation between gut bacterial families and metabolic biomarkers, anthropometric measures, and dietary composition was detectable. Our analysis in particular revealed associations between insulin sensitivity (HOMA-IR) and Akkermansiaceae, Christensenellaceae, and Tanerellaceae. It also suggests the possibility of a beneficial modulation of the latter two intestinal taxa by a diet high in vegetables and fiber, and low in processed meat.

Conclusions: Overall, our results suggest that the gut microbiome remains stable and highly individual-specific under dietary calorie restriction.

Trial registration: The trial, including the present microbiome component, was prospectively registered at ClinicalTrials.gov NCT02449148 on May 20, 2015.

Keywords: Gut microbiome; Intermittent calorie restriction; Obesity; Overweight; Weight loss.

Conflict of interest statement

The authors declare that they have no competing interests.

© 2022. The Author(s).

Figures

Fig. 1

Modified CONSORT diagram showing the flow of participants in the HELENA Trial. ICR – intermittent calorie restriction; CCR – continuous calorie restriction; CTR – control group; CONSORT – consolidated standards of reporting trials

Fig. 2

Overview of the study and intervention effects on the gut microbiome composition. a Graphical overview of the HELENA Trial design showing details of all data collected over the course of the trial. In total, 150 overweight/obese adults were randomly assigned to the ICR (5 days eucaloric diet, 2 days of ~25% energy requirement), CCR (daily reduction of ~20% of caloric intake) or CTR group, 147 of them provided stool samples at baseline. Anthropometric measurements were performed for all participants. Blood samples were collected at all study timepoints for the assessment of plasma/serum concentrations of routine biomarkers. Dietary information was collected at baseline, week 12, and week 50. The amount of weight loss was significantly higher among ICR and CCR participants compared to CTR participants at all study timepoints, as previously published by Schübel et al. [22]. Differences in weight change across intervention groups at each timepoint was assessed using linear mixed models adjusted for age and sex, with participant identifiers as random effects in the model. A significant difference, i.e., p < 0.05 is indicated by (*). ICR – intermittent calorie restriction; CCR – continuous calorie restriction; CTR – controls. b The gut microbiota was predominantly composed of bacteria belonging to Clostridia followed by Bacteroidia as shown by stacked bar charts of taxonomic classes across timepoints and subdivided by intervention group. c Non-metric multidimensional scaling (NMDS) plots of Bray-Curtis dissimilarity distances based on ASV relative abundances between samples after 12 weeks of the intervention. Each point represents the microbial community of a sample, and the colour indicates the intervention group to which it belongs. NMDS plots do not show clustering by intervention groups and PERMANOVA does not indicate a statistically significant difference between groups. d Intra-individual variability (same subject compared across time points) was significantly lower compared with inter-individual variability (different subjects, same group and different subjects, different groups). Technical replicates in this study showed a very low variation in Bray-Curtis dissimilarity distances based on ASV relative abundances, attesting to a high technical reproducibility. Statistical significance as indicated above boxplots was assessed by Wilcoxon signed-rank test. e There was no significant intervention effect on overall gut microbiota alpha diversity (Shannon Index) across all intervention timepoints. All boxplots show the interquartile ranges (IQRs) as boxes, with the median as a black horizontal line and the whiskers extending up to the most extreme points within 1.5-fold IQR. Additional file 3: Source data 2

Fig. 3

Changes in Lactobacillales abundance and processed meat intake, and associations between changes in Lactobacillales abundance and processed meat intake. a The relative abundance of the order Lactobacillales increased in the ICR group after the 12-week intervention but returned to levels similar to baseline at the end of follow-up as shown by boxplots (defined as in Fig. 3); statistical significance was assessed by Wilcoxon signed-rank test. b Participants in the ICR group significantly reduced their intake of processed meat during the 12-week intervention period, although intakes increased again at follow-up. In contrast, there was no significant change in the consumption of processed meat in either the CCR or CTR groups as shown by boxplots. c The changes in the abundance of Lactobacillales after 12 weeks were significantly associated with the changes in processed meat intake (Spearman correlation, q = 0.02). Scatter plot showing Spearman correlation between changes in the relative abundance of Lactobacillales and changes in processed meat intake. d The genus Streptococcus, which belongs to the order Lactobacillales, also significantly co-varied with the changes in processed meat intake after the 12-week intervention (q = 0.02). e Abundance changes in Lactobacillales and its member genus Streptococcus were significantly associated with changes in processed meat intake after 12 weeks. Bar length indicates q value (FDR-corrected p-value) and effect size is color-coded. Additional file 4: Source data

Fig. 4

Association of weight loss with anthropometric, clinical markers, metabolic biomarkers and the gut microbiome. a Weight loss was significantly associated with decreases in body fat compartments (e.g., VAT and SAT), and routine biomarkers (e.g., cholesterol and LDL), but not with changes in gut microbial and small-molecule metabolites, with the exception of choline. Only variables significantly associated with weight loss are labelled (red dots). Measured metabolites (unlabelled due to a lack of association with weight loss) aside routine biomarkers such as those of glucose metabolism, e.g., insulin, comprised bile acids, SCFAs, TMAO, and its precursors, amino acids, and acylcarnitines, but also inflammatory cytokines (Additional file 5: Source data 4). Variables shown by dark red spots were significantly associated with weight loss also at follow-up, i.e., week 50. b The changes in weight after 12 weeks were not significantly associated with any individual bacterial taxon across all taxonomic levels, after FDR correction (all taxonomic levels included in the plot) c There was no differential association of weight loss with gut microbial alpha diversity across all timepoints. Boxplots show changes in alpha diversity over time according to quartiles of weight loss (Q1—quartile 1, Q2—quartile 2, Q3—quartile 3, and Q4—quartile 4, see Fig. 2 for definition of boxplots). Noticeably, metabolic improvements based on the changes in anthropometric and body composition parameters as well as blood concentrations of some biomarkers were greater among participants in Q4 who achieved the highest amount of weight loss after 12 weeks (See Additional file 2: Table S8). As also indicated by the volcano plot in a, there was a significant linear association between weight loss and changes in d visceral adipose tissue, e liver fat, f leptin, g gamma-glutamyl transferase (GGT), h choline, and i cholesterol after the 12-week intervention period. j Baseline abundances, rather than intervention group or weight loss accounted for the largest variation in post-intervention relative abundance of bacteria (here shown at phylum level). Bar plot showing the variations in phyla relative abundance (see colour key) after 12 weeks of intervention accounted for by age at recruitment, sex, baseline phyla abundance, and weight loss (%). Models were generated separately for weight loss and intervention, while partial R-squares of age, sex, and baseline abundances are from the model including weight loss. Additional file 5: Source data 4

Fig. 5

Association of baseline gut microbiota composition with weight loss. A higher abundance of the genus Dorea at baseline may be associated with a difficulty in losing excess weight. a Volcano plot showing the difference in the relative abundance of Dorea between participants in the highest quartile of weight loss and those in the lowest at baseline. Baseline Dorea abundances were significantly higher among participants in the lowest quartile of weight loss. b Baseline Dorea abundances rather than Prevotella-to-Bacteroides ratio or gut microbiota richness was predictive of post-intervention weight loss. AUROC showing how the aforementioned parameters were predictive of weight loss. c The relative abundance of Dorea at baseline was positively associated with weight loss, i.e., a higher abundance corresponds to a lower amount of weight loss. A scatter plot showing the association between the relative abundance of Dorea at baseline and weight loss after 12 weeks. Additional file 6: Source data 5

Fig. 6

Association of the microbiome with anthropometric measurements, metabolic biomarkers, and dietary intake across timepoints. Several core families within the gut microbiome significantly co-vary with some of the clinical biomarkers, body composition measures, and dietary intake assessed in the trial. a Heatmap showing the strength of association calculated irrespective of weight loss quartile or outcome. Phylum affiliation for each family is indicated as a colour strip on the left. Significance for each association after FDR correction is indicated by asterisks (q < 0.1 (*), q < 0.05 (**), and q < 0.01 (***)). Asterisks colored yellow were no longer significant after adjustment for weight loss in the LME models. b Many of the associations shown in a were consistent, even after adjusting for weight loss in the LME models. The yellow points represent the associations for which significance was lost after adjustment for weight loss. c–i Scatterplots for selected significant associations based on the LME models in a. Log-transformed relative abundances of gut bacterial families were corrected for participant-specific offsets using the regression intercepts. The regression slopes for each weight loss quartile has been shown for associations that were significantly influenced by the degree of weight loss (refer to legend) (raw LME model plots without individual-specific abundance correction are shown in Figure S7). j Correlation network between bacterial families and anthropometric measurements, clinical markers and food and energy intake based on the data from a. The network includes bacterial families with at least one significant association to any of the aforementioned parameters and other families that show significant correlation across LME model coefficients (rows of the heatmap in a) with any of those families (Spearman correlation q < 0.05, edge thickness proportional to rho). Edges between families and the parameters are included if the absolute effect size (estimated by the LME) exceeded 0.075 (significant associations are indicated by stronger edges). Additional file 7: Source data 6.