The bile acid chenodeoxycholic acid associates with reduced stroke in humans and mice

Abstract

Bile acids are liver-derived signaling molecules that can be found in the brain, but their role there remains largely unknown. We found increased brain chenodeoxycholic acid (CDCA) in mice with absent 12α-hydroxylase (Cyp8b1), a bile acid synthesis enzyme. In these Cyp8b1^-/-, and in Wt mice administered CDCA, stroke infarct area was reduced. Elevated glutamate-induced excitotoxicity mediated by aberrant N-methyl-D-aspartate receptor (NMDAR) overactivation contributes to neuronal death in ischemic stroke. We found reduced glutamate-induced excitotoxicity in neurons from Cyp8b1^-/- and CDCA-treated Wt mice. CDCA decreased NMDAR-mediated excitatory post-synaptic currents by reducing over-activation of NMDAR subunit GluN2B in Wt brains. Synaptic NMDAR activity was also decreased in Cyp8b1^-/- brains. Expression and synaptic distribution of GluN2B were unaltered in Cyp8b1^-/- mice, suggesting CDCA may directly antagonize GluN2B-containing NMDARs. Supporting our findings, in a case-control cohort of acute ischemic stroke patients, we found lower circulatory CDCA. Together, our data suggest that CDCA, acting in the liver-brain axis, decreases GluN2B-containing NMDAR overactivation, contributing to neuroprotection in stroke.

Keywords: bile acids; chenodeoxycholic acid; excitotoxicity; stroke.

Fig. 1

∼84% of brain bile acids are hepatic in origin. (A) Schematic of the neutral BA synthesis pathway. The neutral bile acid synthesis pathway is the major pathway of BA synthesis and occurs in the liver. Cytochrome P450 Family 8 Subfamily B Member 1 (CYP8B1) generates the primary BA cholic acid (CA). In the absence of CYP8B1, CA is reduced, and the primary BA chenodeoxycholic acid (CDCA) is increased. Before secretion from the liver, primary BAs are conjugated with either glycine or taurine. (B) Brain secondary BAs, which are derived solely from intestinal gut bacterial function, are reduced in WT mice administered antibiotics. (C) The primary bile acid cholic acid (CA) is decreased, and tauro conjugated CDCA (TCDCA), as well as its derivatives tauro α-muricholic acid (TαMCA) and tauro β-muricholic acid (TβMCA) trended toward increase in brain upon antibiotic administration. (D) Plasma secondary BAs are reduced in antibiotics-administered Wt mice. (E) CA is decreased, and TCDCA, TαMCA, and TβMCA trended toward increase in antibiotics-administered plasma. (F) Brain BAs are strongly correlated with plasma BAs (r_s = 0.84) post antibiotic administration. (G) Antibiotic administration does not alter brain mRNA expression of the BA synthesis enzymes Cyp8b1, Cyp27a1 and Cyp7b1. mRNA expression was normalized to Gapdh. Data are shown as average ± SEM, and assessed using Mann-Whitney U-tests. ND = not detected. r_s = Spearman correlation coefficient.

Fig. 2

Increased brain CDCA and TUDCA levels in Cyp8b1^−/− mice. Generation of Cy8b1^−/− mice, showing (A) Genotyping PCR of Cy8b1^+/+, Cy8b1^+/−, and Cy8b1^−/− mice, and (B) quantitative RT-PCR of the Cyp8b1 transcript from the liver of Cy8b1^+/+ and Cy8b1^−/− mice, showing no Cyp8b1 expression in the Cy8b1^−/− mice. Data are shown as average ± SEM of n = 5 mice and were normalized to ribosomal protein L37 (Rpl37). Increased (C) CDCA, (D) TUDCA, and (E) non-12α-hydroxylated BAs, and decreased (E) 12α-hydroxylated BAs, in Cyp8b1^−/− mouse plasma. Increased (F) CDCA, (G) TUDCA, and (H) non-12α-hydroxylated BAs, and decreased (H) 12α-hydroxylated BAs in Cyp8b1^−/− brains. Data are average ± SEM, and analyzed using Mann-Whitney U-tests. BA, bile acid; CDCA, chenodeoxycholic acid; TUDCA, tauro-ursodeoxycholic acid.

Fig. 3

CDCA reduces focal ischemia and glutamate-induced neurotoxicity. (A) Reduced ischemic lesion area in Cyp8b1^−/− mice. (B) Reduced media LDH levels, and (C) reduced 89 kDa PARP-1 fragment in Cyp8b1^−/− primary cortical neurons treated with glutamate. (D) CDCA reduces LDH release by glutamate-treated Wt neurons. (E) TUDCA does not significantly decrease glutamate-induced LDH release in Wt neurons. Experiments for (D and E) were performed simultaneously, therefore share controls (vehicle and 30 μM glutamate). (F) Reduced ischemic lesion area in CDCA-administered Wt mice. Data are average ± SEM. Data in (A and B) were analyzed using students t-tests. Data in (C–F) were analyzed using Mann-Whitney U-tests. CDCA, chenodeoxycholic acid; TUDCA, tauro-ursodeoxycholic acid.

Fig. 4

Decreased NMDAR-mediated EPSCs upon CDCA elevation. (A) Representative traces and quantitation of AMPAR- and NMDAR-mediated EPSCs in Wt mouse acute brain slices showing CDCA decreases NMDAR/AMPAR ratio. (B) Representative traces of input/output response of AMPAR mediated current recorded from CA1 neurons of acute hippocampal slices from Wt mice treated with CDCA, and quantified input/output responses of AMPAR mediated current, showing that CDCA does not significantly modulate AMPA-mediated excitatory post-synaptic currents. (C) Representative traces of AMPAR- and NMDAR-mediated EPSCs, showing reduced NMDAR/AMPAR-mediated EPSC ratio in Cyp8b1^−/− acute brain slices. (D) Reduced NMDAR-mediated EPSCs in Cyp8b1^−/− mice. Cyp8b1^+/+ average, dark circle; Cyp8b1^−/− average, dark square. (E to G) No differences in GluN1, GluN2A and GluN2B expression in Cyp8b1^−/− mice. Calnexin control is shared in (E and F) since GluN1 and GluN2A were run on the same gel. Data are average ± SEM. Data in (A) are analyzed using one-way ANOVA followed by Kruskal-Wallis post test. Data in (B) are analyzed using Two-way ANOVA followed by Tukey’s multiple comparisons tests. Data in (C) are analyzed using Mann-Whitney U-test. Data in (E–G) are analyzed using student’s t-tests. (C and D): Cyp8b1^+/+ n = 15 from 3 mice; Cyp8b1^−/− n = 13 from 2 mice. AMPAR, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor; CDCA, chenodeoxycholic acid; NMDAR, N-methyl-D-aspartate receptor.

Fig. 5

CDCA reduces NMDAR-GluN2B response, with no effects on synaptic GluN2B concentrations. (A) CDCA reduces evoked NMDAR-mediated EPSC decay time in acute hippocampal slices, but not evoked AMPAR-mediated EPSC decay time. Insets show representative traces. The CDCA amplitude (red) is scaled to control (black). Scale bar, 200 ms. (B) Representative traces of total NMDAR (black) and GluN2A (red) response from hippocampal neurons showing CDCA reduces relative GluN2B response. GluN2A response was obtained with Ro25-6981, a GluN2B-specific inhibitor, and GluN2B response was derived by subtracting GluN2A from total NMDAR response. (C) Representative post-embedding immunogold electron micrographs of GluN2B (Scale bar: 0.25 μm), showing (D) unaltered gold particle density (GDP) in the post-synaptic density (PSD), per mouse, and per synapse. (E) Unaltered extra synaptic membrane (ESM) GPD, per mouse, and per synapse. (F) Unaltered PSD length, per mouse, and per synapse. Data are average ± SEM. Data in (A and B), and quantitation per synapse in (D–F) were analyzed using Mann-Whitney U-tests. Quantitation per mouse in (D–F) were analyzed using independent samples T-tests. CDCA, chenodeoxycholic acid; NMDAR, N-methyl-D-aspartate receptor; GPD, gold particle density; AMPAR, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor; EPSC, excitatory postsynaptic current.

Fig. 6

Circulatory CDCA is reduced in human stroke patients. (A) Analysis workflow for the study. Plasma CDCA and its conjugates were profiled in Chinese stroke patients and nonstroke controls using liquid chromatography-mass spectrometry (LC-MS/MS). Distribution of CDCA levels were examined across the two groups and variation patterns were compared with other bile acid levels by assessing odds ratios. Odds ratios were then estimated with adjustment for covariates. Finally, contribution by CDCA to stroke was compared with the covariates. (B) Distribution of CDCA variation across stroke and control subjects. Individuals with the lowest tertile of CDCA level were associated with a heightened odds of 36 for stroke. (C) Forest plot for the univariate and multivariate logistic fits across different bile acid species in stroke and control subjects. Covariates used for adjustments (smoking history, hypertension, and fasting glucose) were selected based on two iterations of stepwise backward regression (P < 0.05 for all covariates). Log odds ratios were computed directly across binary groups for categorical variables. Continuous variables were first converted into tertiles, and log odd ratios were estimated across the first tertile versus the second and third tertiles. Nineteen individuals were excluded from this analysis because of unavailability of covariate measurements. (D) Forest plot for multivariate logistic fits for CDCA and the covariates. The values represented in this plot corresponded to the multivariate logistic model for CDCA in plot C (labeled with obelisk †). CDCA, chenodeoxycholic acid.

Supp Fig 1

Supp Fig 2

Supp Fig 3

Supp Fig 4