Weight gain in anorexia nervosa does not ameliorate the faecal microbiota, branched chain fatty acid profiles, and gastrointestinal complaints

Abstract

The gut microbiota not only influences host metabolism but can also affect brain function and behaviour through the microbiota-gut-brain axis. To explore the potential role of the intestinal microbiota in anorexia nervosa (AN), we comprehensively investigated the faecal microbiota and short-chain fatty acids in these patients before (n = 55) and after weight gain (n = 44) in comparison to normal-weight participants (NW, n = 55) along with dietary intake and gastrointestinal complaints. We show profound microbial perturbations in AN patients as compared to NW participants, with higher levels of mucin-degraders and members of Clostridium clusters I, XI and XVIII and reduced levels of the butyrate-producing Roseburia spp. Branched-chain fatty acid concentrations, being markers for protein fermentation, were elevated. Distinct perturbations in microbial community compositions were observed for individual restrictive and binge/purging AN-subtypes. Upon weight gain, microbial richness increased, however perturbations in intestinal microbiota and short chain fatty acid profiles in addition to several gastrointestinal symptoms did not recover. These insights provide new leads to modulate the intestinal microbiota in order to improve the outcomes of the standard therapy.

Figure 1. Energy and macronutrient intake of anorexia nervosa patients (AN) before (T1) and after (T2) weight gain in comparison to normal-weight participants (NW).

The daily intakes of energy (A), fibre (sum of cellulose + non-cellulosic polysaccharides + lignin) (B), protein (C), fat (D) and carbohydrates (E), and the energy percentages derived from protein (F), fat (G) and carbohydrates (H) in the diet are shown for NW and AN at T1 and T2. A p-value < 0.01 was considered as statistically significant. **Indicates a p < 0.01, ***indicates a p < 0.001.

Figure 2. Severity scores for Gastrointestinal symptoms (GIS) in anorexia nervosa patients (AN) before (T1) and after (T2) weight gain and in comparison to normal-weight participants (NW).

(A) Upper GIS and lower GIS for all groups are presented. (B) Significant changes of GIS between T1 and T2 in AN are given. A complete overview of GIS between T1 and T2 in AN is given in Supplement 2. (C) The heatmap shows Spearman correlations between ANT1 versus NW and between ANT2 and NW, respectively for GIS. Negative correlations (in red) indicate increased Severity-Scores of the specific symptoms in AN compared to NW. Positive correlations (in blue) indicate decreased Severity-Scores of the specific symptoms in AN compared to NW. All p-values were false discovery rate (FDR)-adjusted. A FDR < 0.05 was considered as statistically significant. **Indicates a FDR < 0.01 and ***a FDR < 0.001.

Figure 3. Taxonomy at phylum and genus level for anorexia nervosa patients (AN) before (T1) and after (T2) weight gain compared to normal-weight participants (NW).

(A) An overview of the mean relative abundance at phylum level is given for taxa being present in at least 25% of all participants. Significant differences between the groups at phylum level are presented separately in box-whiskers for Bacterioidetes (B), Firmicutes (C), Actinobacteria (D) and Verrucomicrobia (E). (F) Relative abundance of Firmicutes and Bacteroidetes per AN and NW are shown. Subjects are sorted according to the relative abundance of Firmicutes. (G) The heatmap shows significant Spearman correlations between ANT1 versus NW and between ANT2 and NW, respectively at genus level. Positive correlations (in blue) indicate decreased levels of the specific taxon in AN compared to NW. Negative correlations (in red) indicate increased levels in AN compared to NW. (H) Significant differences within AN from T1 to T2 are presented. (B–E,G,H) The p-values were false discovery rate (FDR)-adjusted. In order to account for the beta error a FDR < 0.15 was considered as statistically significant. A FDR < 0.15 was found for all displayed taxa. *Indicates a FDR < 0.15 and **a FDR < 0.05.

Figure 4. The core microbiome of anorexia nervosa patients (AN) before (T1) and after (T2) weight gain compared to normal-weight participants (NW).

(A) The Venn diagram shows the core microbiome of AN at T1 and T2 and of NW. For each of the groups the core microbiome was defined as the OTUs that were observed in at least 90% of the participants. Bacterial taxa belonging only to the core of ANT1, ANT2 or NW are listed in respectively red, blue and green. (B) The fraction of samples in the core microbiome of AN at T1 and T2 and of NW are presented. The number of OTUs part of the core microbiome defined as the OTUs that were observed in a minimal fraction of subjects ranging from 75% to 100%. Lines represent the mean number of OTUs in the core (based upon 10 rarefactions), surrounding areas represent the 95% confidence intervals.

Figure 5. Alpha and beta diversity metrics in anorexia nervosa patients (AN) before (T1) and after (T2) weight gain compared to normal-weight participants (NW).

The number of observed OTUs (A), Chao 1 index (B) and Shannon index (C) for all groups are shown. The p-values of all metrics were Bonferroni adjusted. A p < 0.05 was considered as statistically significant. *Indicates a p < 0.05 and **indicates a p < 0.01. Unweighted UniFrac (D) and Bray Curtis (E) distances were calculated within AN subjects (average pairwise distance in microbiota composition between samples of the same AN patient at T1 and T2) and between subjects (average pairwise distance between samples of different subjects) for the AN groups at T1 and T2 and for the NW group, respectively. A p-value < 0.001 was considered as statistically significant. ***Indicates a p < 0.001.

Figure 6. Principal Coordinate Analyses (PCoAs) based on Bray-Curtis distance and assortment of gut microbial communities into enterotypes for anorexia nervosa patients before weight gain (ANT1) and normal-weight participants (NW).

(A) PCoA, colored according to NW (green), ANT1 (red), (B) PCoA, colored according to microbial richness (Chao1), (C) PCoA, colored according to Bacteroides abundance, (D) PCoA, colored according to Prevotella abundance. Plots (B–D) are colored using a gradient ranging from low (red) to high (cyan-blue). (E) Shown are between-class analysis visualizations of enterotypes, as identified by PAM clustering, with closed dots and squares representing individual AN patients or NW participants, respectively and numbered white rectangles marking the center of each enterotype. The box-whiskers present the relative abundances of the three bacterial taxa that are mainly driving the separation of the enterotypes.

Figure 7. Distance-based Redundancy Analysis (db-RDA) plot showing the relationship of age and disease status to the microbial community structure.

The plots represents a dbRDA ordination based upon the Bray-Curtis distance including (A) anorexia nervosa patients (AN) at baseline (T1) and normal-weight participants and with age and disease status as explanatory variables or (B) AN at baseline at T1 and with age and disease status, gastrointestinal symptoms, BMI, total energy intake and fibre, fat and protein intake as explanatory variables. Age and anorexia subtype were the only significant explanatory variables. Binge purging = Binge/Purging AN-type, Restrictive = Restrictive AN-type.