Ketogenic diet therapy for pediatric epilepsy is associated with alterations in the human gut microbiome that confer seizure resistance in mice

Abstract

The gut microbiome modulates seizure susceptibility and the anti-seizure effects of the ketogenic diet (KD) in animal models, but whether these relationships translate to KD therapies for human epilepsy is unclear. We find that the clinical KD alters gut microbial function in children with refractory epilepsy. Colonizing mice with KD-associated microbes promotes seizure resistance relative to matched pre-treatment controls. Select metagenomic and metabolomic features, including those related to anaplerosis, fatty acid β-oxidation, and amino acid metabolism, are seen with human KD therapy and preserved upon microbiome transfer to mice. Mice colonized with KD-associated gut microbes exhibit altered hippocampal transcriptomes, including pathways related to ATP synthesis, glutathione metabolism, and oxidative phosphorylation, and are linked to susceptibility genes identified in human epilepsy. Our findings reveal key microbial functions that are altered by KD therapies for pediatric epilepsy and linked to microbiome-induced alterations in brain gene expression and seizure protection in mice.

Keywords: CP: Microbiology; CP: Neuroscience; epilepsy; ketogenic diet; microbiome; seizure.

Figure 1.. Transfer of the clinical KD-associated gut microbiota from pediatric epilepsy patients to mice confers resistance to 6-Hz seizures

(A and B) (A) Experimental schematic and (B) 6-Hz seizure thresholds for mice inoculated with pre-KD and post-KD human microbiota (one-way ANOVA with Tukey’s, n = 13–15 mice/patient sample). (C) Average seizure thresholds of recipient mice per patient sample (two-tailed, unpaired Welch’s t test. n = 10 patients/group). (D) Latency to exploration for all pre-KD (n = 140) and post-KD (n = 141) recipient mice. (E) Average current for remission seizures (two-tailed, unpaired Welch’s t test. n = 10 patients/group). Data are displayed as mean ± SEM. ***p < 0.001, ****p < 0.0001, ^####p < 0.0001 (within-patient mouse recipients).

Figure 2.. The clinical KD-associated human microbiome exhibits functional alterations that are phenocopied in seizure-protected recipient mice

(A) Differentially abundant metagenomic pathways (p < 0.10, where *p < 0.05) in post-KD vs. pre-KD (n = 10/condition; where each recipient n reflects average of five mice per donor sample). Red font indicates changes in same direction in post-KD donors and recipients. (B) Metagenomic pathways differentially abundant in the same direction in post-KD donors and recipients. (C) Beta-hydroxybutyrate and glucose in donor and recipient feces and serum (two-tailed Wilcoxon; n = 10/condition, where each recipient n reflects average of five mice per donor sample). (D) Top 25 enriched chemical subclasses for differentially abundant fecal metabolites (p < 0.05, matched-pairs Student’s t test, n = 10/condition). Red font indicates differential chemical subclasses shared across human and mouse. Orange asterisks indicate additional chemical subclasses relevant to KD based on literature. (E) Top 25 enriched Small Molecule Pathway Database pathways for differentially abundant fecal metabolites (p < 0.05, matched-pairs Student’s t test, n = 10/condition). Data are displayed as mean ± SEM. *p < 0.05; ***p < 0.001; n.s., not statistically significant.

Figure 3.. Seizure resistance in mice inoculated with the post-KD microbiota is associated with alterations in the brain transcriptome

(A) Biological Process Gene Ontology (GO) of differentially expressed genes (DEGs) (p < 0.05) in post-KD vs. pre-KD mouse hippocampus (n = 10/condition, where each n is pooled from six mice per donor sample). (B) Top 25 DEGs ranked by p value with minimum log2 fold change (log2FC) > 2. log2FC was z score normalized by column. (C) Protein interaction network (enrichment score >0.7) of DEGs that appeared in both GO and STRING network analyses. (D) Functional enrichment of top subnetwork clusters. If log2FC > 3 or <−3, the value is listed next to the node name.

Figure 4.. Multi-omic network analysis identifies key microbial genomic pathways and microbially modulated metabolites associated with differential expression of hippocampal transcripts

(A) Co-occurrence network and weighted key drivers constructed from human donor fecal metagenomics and metabolomics (top) and mouse recipient fecal metagenomics, fecal metabolomics, serum metabolomics, hippocampal transcriptomics, and frontal cortical transcriptomics datasets (bottom). Red text indicates pathways, metabolites, or genes that were differentially regulated (p < 0.05) in post-KD vs. pre-KD (n = 10/condition, where each recipient n is average of five or six mice). (B) Top epilepsy GWAS genes that map onto mouse hippocampal and frontal cortical DEGs.