The ketogenic diet influences taxonomic and functional composition of the gut microbiota in children with severe epilepsy

Abstract

The gut microbiota has been linked to various neurological disorders via the gut-brain axis. Diet influences the composition of the gut microbiota. The ketogenic diet (KD) is a high-fat, adequate-protein, low-carbohydrate diet established for treatment of therapy-resistant epilepsy in children. Its efficacy in reducing seizures has been confirmed, but the mechanisms remain elusive. The diet has also shown positive effects in a wide range of other diseases, including Alzheimer's, depression, autism, cancer, and type 2 diabetes. We collected fecal samples from 12 children with therapy-resistant epilepsy before starting KD and after 3 months on the diet. Parents did not start KD and served as diet controls. Applying shotgun metagenomic DNA sequencing, both taxonomic and functional profiles were established. Here we report that alpha diversity is not changed significantly during the diet, but differences in both taxonomic and functional composition are detected. Relative abundance of bifidobacteria as well as E. rectale and Dialister is significantly diminished during the intervention. An increase in relative abundance of E. coli is observed on KD. Functional analysis revealed changes in 29 SEED subsystems including the reduction of seven pathways involved in carbohydrate metabolism. Decomposition of these shifts indicates that bifidobacteria and Escherichia are important contributors to the observed functional shifts. As relative abundance of health-promoting, fiber-consuming bacteria becomes less abundant during KD, we raise concern about the effects of the diet on the gut microbiota and overall health. Further studies need to investigate whether these changes are necessary for the therapeutic effect of KD.

Conflict of interest statement

The authors declare no competing interests.

Figures

Fig. 1

Microbial alpha diversity analysis. From left to right: Total number of observed metagenomic operational taxonomic units (mOTUs), total species richness Chao1, and Shannon’s diversity index of evenness a. Data are presented as follows: center line, median; box limits, upper and lower quartiles; whiskers, 1.5× interquartile range; points, outliers. Dunn’s test of multiple comparisons with Benjamini–Hochberg adjustment: *p < 0.05, **p < 0.01. Ctrl control, Pat patient. White, Controls' time point 1; Ivory, Controls' time point 2; Green, Patients' time point 1; Red, Patients' time point 2. Correlation of alpha diversity to age in patients b with R2 values as indicated

Fig. 2

Microbial beta diversity analysis. Principal component analysis (PCA) of a taxonomic and b functional profiles. Controls before (white circles), Controls after (ivory squares), Patients before (green circles), Patients after (red squares). Taxonomic profiles of individuals at the phylum level c

Fig. 3

Statistical analysis of taxonomic profiles. Significant changes at all taxonomic levels during dietary intervention (KD) using the linear discriminant analysis (LDA) effect size (LEfSe) method; p < 0.05, LDA > 4.0 a. Cladogram of significant changes at all taxonomic levels during KD b. Green: Patients' time point 1, Red: Patients' time point 2

Fig. 4

Taxonomic drivers of functional shifts associated with KD. Taxonomic contributions to the shift in each function are shown as bars for functions enriched pre-KD compared to post-KD a and functions enriched post-KD compared to pre-KD b. Bar length represents the size of the contribution. For each function, the position of the bar indicates the type of contribution. The top (bottom) bars represent contributions from genera with higher (lower) relative abundance in the enrichment group (pre-KD for a, post-KD for b). Bars to the left (right) of the vertical black line represent contributions to decreased (increased) function abundance in the enrichment group. The red diamonds indicate the observed increase in relative median function abundance in pre-KD samples a or post-KD samples b. Colors indicate the genus, with genera from the same phylum sharing similar colors: green for Actinobacteria; orange for Bacteroidetes; teal, blue, and purple for Firmicutes; and red for Proteobacteria. Genus labels within bars are provided for genera with notable contributions where bars were wide enough to accommodate labels

Fig. 5

Analysis of butyrate production potential. RPKM (reads per kilo base per million) values for unique genes of the acetyl-CoA pathway a and the 4-aminobutyrate pathway b for butyrate production. Data are presented as follows: center line, median; box limits, upper and lower quartiles; whiskers, 1.5× interquartile range; points, outliers. Dunn’s test of multiple comparisons with Benjamini–Hochberg adjustment: *p < 0.05, **p < 0.01. Ctrl control, Pat patient. White, Controls time point 1; Ivory, Controls time point 2; Green, Patients time point 1; Red, Patients time point 2. Correlation of RPKM of each gene to age of patients at time point 1 for the acetyl-CoA pathway c and the 4-aminobutyrate pathway d. None of the correlations were significant (p > 0.05). thl acetyl-CoA acetyltransferase (thiolase), bhbd β-hydroxybutyryl-CoA dehydrogenase, cro crotonase, abfH 4-hydroxybutyrate dehydrogenase, 4hbt butyryl-CoA:4-hydroxybutyrate CoA transferase, abfD 4-hydroxybutyryl-CoA dehydratase and vinylacetyl-CoA 3,2-isomerase (same gene)