Microbiome-derived carnitine mimics as previously unknown mediators of gut-brain axis communication

Abstract

Alterations to the gut microbiome are associated with various neurological diseases, yet evidence of causality and identity of microbiome-derived compounds that mediate gut-brain axis interaction remain elusive. Here, we identify two previously unknown bacterial metabolites 3-methyl-4-(trimethylammonio)butanoate and 4-(trimethylammonio)pentanoate, structural analogs of carnitine that are present in both gut and brain of specific pathogen-free mice but absent in germ-free mice. We demonstrate that these compounds are produced by anaerobic commensal bacteria from the family Lachnospiraceae (Clostridiales) family, colocalize with carnitine in brain white matter, and inhibit carnitine-mediated fatty acid oxidation in a murine cell culture model of central nervous system white matter. This is the first description of direct molecular inter-kingdom exchange between gut prokaryotes and mammalian brain cells, leading to inhibition of brain cell function.

Fig. 1. MALDI-MSI on brain and gut sections from C57BL/6 GF and SPF mice.

(A) Hematoxylin and eosin-stained sections (top) and MALDI-MSI images (bottom) of brain tissue. MALDI-MSI identified a peak at m/z 160.133 that was absent in GF mice but present in discrete locations in the brains of SPF mice. (B) Bar plot of relative abundance of m/z 160.1 in different regions of the SPF brain and the average across the whole brain. (C) This metabolite was also present in the SPF colon but absent in the GF colon. a.u., arbitrary units. Annotated brain regions: Cb, cerebellum; arb, arbor vitae; MO, medulla oblongata; P, pons; MB, midbrain; cc, corpus callosum; fr, fasciculus retroflexus. The heatmap intensity bar shows the color scale from low levels of the molecule (black/dark blue) to high levels of the molecule (pink/white). Scale bars, 1 mm.

Fig. 2. DESI-MSI was used to screen intestinal bacteria for the production of metabolites of interest.

Bacterial cultures were grown on solid FAB agar, and colonies were resuspended in PBS to OD₆₀₀ (optical density at 600 nm) of approximately 0.3. Two microliters was allowed to air-dry on a slide before MSI. Bacterial strains tested: (A) C. symbiosum LM19R, (B) C. symbiosum LM19B, (C) C. clostridioforme LM41A, (D) C. symbiosum LM42D, (E) Bifidobacterium animalis LM33, (F) Lactobacillus animalis LM31, (G) Propionibacterium spp. YM23, (H) Clostridium difficile LM27, (I) Enterococcus faecalis YM13, (J) Bacteroides fragilis NCTC 9343 (type strain), (K) C. clostridioforme NCTC 11224 (type strain), (L) C. clostridioforme NCTC 7155 (type strain), (M) blank control, (N) blank control, (O) C. symbiosum LM19R, and (P) Escherichia coli F18.

Fig. 3. ¹H-¹H COSY of crude extracts showing the occurrence of three related trimethylammonium derivatives.

Cross peaks have been colored-coded accordingly, while lines shown on the spectrum represent the correlation between neighboring coupling protons in the respective derivative: 3M-4-TMAB (green) and 4-TMAP (red) as well as the demethylated congener, GBB (black).

Fig. 4. Conversion of MALDI-MSI into binary images showing the high-intensity (black) and low-intensity (gray) regions of the ion images of m/z 160.1 (3M-4-TMAB/4-TMAP) and m/z 162.1 (carnitine).

The overlap of these regions was then calculated as the percentage of pixels in the high-intensity region of the m/z 160.1 image that were also high intensity in the m/z 162.1 image. In addition, Pearson’s correlation coefficient between the two images was calculated for each tissue across three biological replicates (SPF1, SPF2, and SPF3), with a coefficient of 1 indicating perfect colocalization and −1 indicating no colocalization whatsoever (SPF1 overlap 92.85714, Pearson’s correlation coefficient 0.921547; SPF2 93.20652, 0.906855; SPF3 63.20542, 0.609946).

Fig. 5. Quantitation of 3M-4-TMAB and 4-TMAP in the mouse brain was performed using DESI-MSI of m/z 160.133.

(A) An equimolar mix of 3M-4-TMAB and 4-TMAP standards was prepared and spotted on a GF brain section at various concentrations. (B) The m/z 160.133 ion intensity of each spot from the MSI results was used to generate a concentration curve against the amount of standard in each spot. (C) The concentration curve was used to calculate the average endogenous concentration of 3M-4-TMAB and 4-TMAP across the whole brain and across the high abundance area, the corpus callosum, and the hippocampus region. The high abundant region is outlined in (A). The results show two technical replicates for GF and SPF brain sections.

Fig. 6. OCR was used as an indicator of FAO in the presence of 3M-4-TMAB and 4-TMAP.

Oxidation of palmitate is significantly decreased in primary murine myelinating CNS white matter cultures in the presence of carnitine and either 3M-4-TMAB, 4-TMAP, or a combination of 3M-4-TMAB and 4-TMAP at the indicated concentrations. Etomoxir, an irreversible inhibitor of FAO, was used as a control. The OCR values shown are relative to the BSA and palmitate condition. Each symbol represents the relative mean value for one experiment (n = 4 independent experiments). The neural precursors used to generate the model of CNS white matter were obtained by pooling dissociated spinal cord cells from all embryos from a single pregnant female mouse. A one-sample t test was used to test the statistical significance against the relative value 1. P values ≤0.05 were considered significant. *P < 0.05 and ****P < 0.0001. Bars represent mean ± SD.