Whole-body modelling reveals microbiome and genomic interactions on reduced urine formate levels in Alzheimer's disease

Abstract

In this study, we aimed to understand the potential role of the gut microbiome in the development of Alzheimer's disease (AD). We took a multi-faceted approach to investigate this relationship. Urine metabolomics were examined in individuals with AD and controls, revealing decreased formate and fumarate concentrations in AD. Additionally, we utilized whole-genome sequencing (WGS) data obtained from a separate group of individuals with AD and controls. This information allowed us to create and investigate host-microbiome personalized models. Notably, AD individuals displayed diminished formate microbial secretion in these models. Additionally, we identified specific reactions responsible for the production of formate in the host, and interestingly, these reactions were linked to genes that have correlations with AD. This study suggests formate as a possible early AD marker and highlights genetic and microbiome contributions to its production. The reduced formate secretion and its genetic associations point to a complex connection between gut microbiota and AD. This holistic understanding might pave the way for novel diagnostic and therapeutic avenues in AD management.

Keywords: Alzheimer’s disease; co-metabolism; constraint-based modelling; formate; host-microbiome; metabolic modelling; metabolomics; microbiome; pathways.

Figure 1. 1D-NMR metabolomic data on urine samples from the DELCODE study cohort support altered metabolism in AD, including formate metabolism.

A: Descriptive statistics for the analysed DELCODE samples. *p-value derived from one-factorial ANOVA, ^#p-value derived from Fisher’s exact test. SCD = subjective cognitive decline, MCI = mild cognitive impairment, AD = Alzheimer’s disease, BMI = body mass index. B-C: Box plots for formate and fumarate (both creatinine normalised) over the four study groups. P-values were derived from multivariable regressions adjusting for age, sex, and BMI using heteroscedastic robust standard errors. SCD = subjective cognitive decline. MCI = mild cognitive impairment. AD = Alzheimer’s disease.

Figure 2. Descriptive statistics for the analysed samples and their corresponding Qiita-derived microbiome models, and diversity analysis on microbial relative abundance.

A: Samples and model characteristics of the Wisconsin cohort studies. AD = Alzheimer’s disease, SD = Standard deviation, ^ap-value from Welch t-tests, ^bp-value from Fisher’s exact test. Full results can be found in Table S03. B: Boxplot of gOTU log-ratio analysis on microbes, whose models’ relative abundances were found to be altered between healthy and dementia-AD participants, p-value from Welch t-tests. The full results of the relative abundance analysis can be found in Table S04.

Figure 3. Evaluation of host-microbiome involvement in urine secretion of microbe-derived metabolites.

A: Breakdown of maximum sex-specific urine formate secretion highlighting the presence of microbiome-host co-metabolism. B: Microbiome average metabolites secretion when investigating for maximum urinary formate secretion and predicted urine formate secretion values after diet addition of the metabolites in male and female germ-free WBMs. C: Average metabolite microbiome secretion when the host-microbiome WBMs were interrogated for the maximum urine secretion, and urine secretions when the specific metabolites were unconstrained in the diet for the germ-free models.

Figure 4. Cellular metabolism involved in the production of formate including reactions found responsible for host-microbiota co-metabolism.

Dotted lines represent diet constituent involvement in the overall formate production, circled reactions when deleted together largely reduced the formate urinary production. Metabolites names (abbreviations are given in VMH IDs, www.vmh.life, ): 10fth=10-formyl-THF; 3pg=3-phosphoglyceric acid; chol=choline; cys_L=L-cysteine; dmgly=dimethylglycine; fald=formaldehyde; for=formate; glc_D=D-glucose; gly=glycine; glyb=betaine; glyc=glycerol; glyc3p=glycerol-3phosphate; gthrd=GSH; his_L=L-histidine; lac_D=D-lactate; lkynr=L-kynurenine; meoh=methanol; methf=5,10-methenyl-THF; mlthf=5,10-methylene-THF; mma=methylamine; orn=ornithine; sarcs=sarcosine; ser_L=L-serine; thf=tetrahydrofolate; trp_L=L-tryptophane; tyr_L=L-tyrosine. Reactions’ names can be found in Table S07. Complete KO analysis can be found in Table S08. Differential gene expression (DEG) and differential protein abundance (DEP) analysis of association to AD-related phenotypes can be found in Table S09.