Fecal short-chain fatty acids vary by sex and amyloid status

Abstract

Introduction: Short-chain fatty acids (SCFAs), produced by gut microbes, influence Alzheimer's disease (AD) pathology in animals. Less is known about SCFAs and AD in humans. We profiled feces of adults along the AD continuum to investigate gut microbiome and SCFA associations with AD pathology and cognition.

Methods: We measured SCFAs and bacterial abundances in fecal samples from 287 participants in the Wisconsin Alzheimer's Disease Research Center and Wisconsin Registry for Alzheimer's Prevention. We performed regressions examining associations between SCFAs or gut microbes and AD pathology and cognition.

Results: Fecal propionate, isovalerate, and propionate-producing bacteria are inversely associated with amyloid status. Mediation analysis found that propionate mediates sex-specific associations between SCFAs and cerebrospinal fluid biomarkers. SCFA levels are associated with slower cognitive decline.

Discussion: These results link SCFAs and propionate-producing bacteria with AD. This may inform efforts to leverage diet and specific bacteria to boost SCFA production and potentially ameliorate AD progression.

Highlights: Fecal SCFAs link to lower amyloid burden and microglial activation, notably in females. SCFA-producing gut microbes have reduced abundance in amyloid-positive participants. Fecal propionate mediates relationships between gut microbes and amyloid status. SCFAs are associated with slower plasma pTau₂₁₇ accumulation in females. SCFAs are associated with slower cognitive decline.

Keywords: Alzheimer's disease; amyloid; bacteria; cerebrospinal fluid biomarkers; gut microbiome; metabolites; microbiota; short‐chain fatty acids.

FIGURE 1

Analysis dataset selection per study goal. CONSORT flow diagram indicating selection criteria for inclusion in each analysis goal. A+, amyloid status determined with ¹¹C‐PiB PET, CSF Aβ₄₂/Aβ₄₀, or plasma pTau₂₁₇. CSF, cerebrospinal fluid; PET, positron emission tomography; SCFA, short‐chain fatty acids measured in fecal samples; T+, tau positive status measured with ¹⁸F‐MK6240 PET neuroimaging.

FIGURE 2

Amyloid relationships with SCFA abundance. (A) Fecal SCFA levels compared across A+ and A− participants and (B) forest plot with odds ratios depicting A+ risk as a function of SCFA level and modified by male sex. From left to right per panel, a scatter plot depicts individual participant values, a box plot shows median SCFA level per amyloid status, and a density plot shows the data distribution and highlights the mean SCFA level per amyloid status. SCFA levels in CU participants are depicted with circles, and in MCI or dementia participants with triangles. For each SCFA, abundances are normalized on the median abundance represented in the sample. Odds ratios for main SCFA effects and SCFA‐by‐sex interactions were derived from the logistic regression model Amyloid status ∼ SCFA + sex + SCFA‐by‐sex + age + APOE ε4 carrier status + diagnosis + time between biomarker measurement and fecal collection. To allow for visual comparison of mean effects, the 95% confidence interval upper limits were annotated for coefficients with wide confidence intervals. A ± , amyloid positive or negative status determined with ¹¹C‐PiB PET, CSF Aβ₄₂/Aβ₄₀, or plasma pTau₂₁₇. APOE, apolipoprotein E; CU, cognitively unimpaired; MCI, mild cognitive impairment; SCFA, short‐chain fatty acid; UL, upper limit of 95% confidence interval.

FIGURE 3

SCFAs significantly associated with CSF biomarkers of amyloid pathology and microglial activation in cognitively unimpaired participant subset. (A) Forest plot with SCFA and SCFA‐by‐male sex effect coefficients on CSF biomarker outcomes. Effects were estimated with the regression model CSF biomarker ∼ SCFA + sex + SCFA‐by‐sex + age + APOE ε4 carrier status + BSS + CSF biomarker batch. (B–D) CSF biomarker levels residualized on model covariates (age, sex, SCFA level, APOE ε4 carrier status, BSS, CSF biomarker batch) to demonstrate (B, C) Aβ₄₂/Aβ₄₀ relationships with isobutyrate and isovalerate and (D) sTREM2 relationships with butyrate. (E–H) Aβ₄₂/Aβ₄₀ residualized on model covariates (age, sex, APOE ε4 carrier status, BSS, CSF biomarker batch) to demonstrate amyloid relationships with (E) acetate, (F) butyrate, (G) isobutyrate, and (H) isovalerate main effects. Aβ, amyloid beta; APOE, apolipoprotein E; BSS, Bristol Stool Score; GFAP, glial fibrillary acidic protein; NfL, neurofilament light chain protein; pTau₁₈₁, tau phosphorylated at amino acid 181; SCFA, short‐chain fatty acid; sTREM2, soluble triggering factor expressed on myeloid cells 2; tTau, total tau; YKL‐40, chitinase‐3‐like protein 1.

FIGURE 4

Effects of SCFAs on longitudinal trajectories of pTau₂₁₇ and cognition in cognitively unimpaired subset. (A) Effects of SCFAs on longitudinal trajectories of plasma pTau₂₁₇ or (B) cognition (PACC3) by multiple linear regression: plasma pTau₂₁₇ or cognition ∼ SCFA + sex + SCFA‐by‐male sex + age² + APOE ε4 carrier status + BSS. For visualization, SCFA levels (normally distributed) are grouped into even tertiles: high, medium, and low. APOE, apolipoprotein E; BSS, Bristol Stool Score; PACC3, Preclinical Alzheimer Cognitive Composite score (three‐item version); pTau₂₁₇, tau phosphorylated at amino acid 217; SCFA, short‐chain fatty acid.

FIGURE 5

Distributions of propionate‐ and butyrate‐producing bacteria among cognitively unimpaired participants are shifted toward amyloid‐negative status. The distribution of log transformed odds ratios (μ) for metagenome‐assembled genomes (MAGs) with genetic pathways for (A and C) propionate production and (B and D) butyrate production in A+ participants compared to A−, for producer MAGs versus non‐producer MAGs (labeled propionate or butyrate positive or negative by color), as assessed by a Wilcoxon rank‐sum test for log‐transformed odds ratios, indicating the effect of each MAG on amyloid status (negative values = A−, positive values = A+). Effects of MAGs on amyloid status were assessed by logistic regression: Amyloid status ∼ MAG + sex + age + APOE ε4 carrier status + BSS. (C and D) Top to bottom per panel: density plot shows data distribution, highlighting mean MAG log‐transformed odds ratio per group; scatter plot depicts log‐transformed odds ratio for each MAG. Propionate‐positive and butyrate‐positive MAGs were classified based on gene content. A ± , amyloid positive or negative status determined with ¹¹C‐PiB PET, CSF Aβ₄₂/Aβ₄₀, or plasma pTau₂₁₇. APOE, apolipoprotein E; BSS, Bristol Stool Score; CSF, cerebrospinal fluid; pTau₂₁₇, tau phosphorylated at amino acid 217; MAG, metagenome‐assembled genome.

FIGURE 6

MAGs are significantly associated with CSF biomarkers of AD, neuroinflammation, and neurodegeneration in the cognitively unimpaired participant subset. Propionate‐ and butyrate‐producing MAGs, as well as MAG‐by‐male sex interaction terms, have effects on CSF (A) Aβ₄₂/Aβ₄₀ and (B) sTREM2, as assessed by multiple linear regression: CSF biomarker ∼ MAG + sex + MAG‐by‐male sex + age + APOE ε4 carrier status + BSS + CSF biomarker batch. P values shown are before false discovery rate correction. Aβ₄₂/Aβ₄₀, ratio of amyloid beta 42 to amyloid beta 40. APOE, apolipoprotein E; BSS, Bristol Stool Score; CSF, cerebrospinal fluid; GFAP, glial fibrillary acidic protein; MAG, metagenome‐assembled genome; NfL, neurofilament light chain protein; pTau₁₈₁/Aβ₄₂, ratio phosphorylated tau 181 to amyloid beta 42; sTREM2, soluble triggering factor expressed on myeloid cells 2; tTau, total tau; YKL‐40, chitinase‐3‐like protein 1; SCFA, short‐chain fatty acid.

FIGURE 7

SCFA mediation of relationship between MAGs and AD biomarkers. (A) Table of significant mediation relationships. Mediation was assessed by causal mediation analysis, accounting for the covariates sex, age, APOE ε4 carrier status, BSS, and CSF biomarker batch. (B) Representative schematic for mediation analysis framework. (C) Propionate mediates the relationship between Paraprevotella clara and A+ status. Propionate mediates the relationship between propionate‐producing MAGs and (D) CSF pTau₁₈₁/Aβ₄₂ and (E) CSF GFAP. Effects of SCFAs on CSF biomarkers were controlled for sex, age, APOE ε4 status, and BSS. Propionate‐producing and butyrate‐producing MAGs were classified based on gene content. Yellow = propionate‐producing MAG. A+, amyloid‐positive status determined with ¹¹C‐PiB PET, CSF Aβ₄₂/Aβ₄₀, or plasma pTau₂₁₇; Aβ₄₂/Aβ₄₀, ratio of amyloid beta 42 to amyloid beta 40. APOE, apolipoprotein E; B, butyrate producer; bm, biomarker; BSS, Bristol Stool Score; CSF, cerebrospinal fluid; GFAP, glial fibrillary acidic protein; MAG, metagenome‐assembled genome; P, propionate producer; pTau₁₈₁/Aβ₄₂, ratio of phosphorylated tau 181 to amyloid beta 42; SCFA, short‐chain fatty acid; sTREM2, soluble triggering factor expressed on myeloid cells 2.