2.1 Chemicals, reagents, instruments, and quality control fecal samples
Standards were purchased from CHEM SERVICE, Kanto Chemical (Tokyo, Japan), TCI (Tokyo, Japan), SIGMA-Aldrich (STL, USA), ICN (CA, USA), Fluka (NC, USA), FUJIFILM Wako (Osaka, Japan), Cayman Chemical (MI, USA), and others. Ribitol used as an internal standard was purchased from Wako. Myristic acid was from the Fiehn GC/MS Metbolomics standards kit (Agilent, 400505). N-Methyl-N-(trimethylsilyl)trifluoroacetamide (MSTFA) and methoxyamine were purchased from SIGMA-Aldrich. Quality control (QC) fecal samples were prepared by mixing 10 freeze-dried human fecal samples.
2.2 Evaluation of SCFA volatility with or without fecal co-treatment
To verify whether SCFA samples volatilize during the extraction and derivatization procedures, we tested the effects of fecal co-treatment on measurement of SCFAs. Acetic acid, propionic acid, and butyric acid were used for the validation. The analysis compared the relative peak areas of 2,500 µM standard solutions of each SCFA with and without the addition of QC feces, and the relative area of the samples added with QC feces was calculated by subtracting the relative area of the QC feces alone. Standard solutions of each SCFA were analyzed in two forms: as salts and in their free states. The extraction and derivatization procedures are described below.
2.3 Extraction and derivatization of metabolites from fecal samples
The experimental procedure is shown in Fig. 1. Fecal samples were freeze-dried for at least 24 hours using a VD-800R freeze-dryer (TAITEC, Japan). They were then subjected to vigorous shaking with 3.0 mm zirconia beads using a Shake Master (Biomedical Science, Japan) to rupture the cell membrane (1,500 rpm., 10 minutes). After disruption, 10 mg of the samples were weighed, and 0.1 mm zirconia/silica beads were added. Then, 600 μL of a water:methanol (1:2) solution containing ribitol as an internal standard was added. The mixture was then shaken again for 10 minutes and centrifuged (20 °C, 4,600 x g, 10 min). Then, 300 μL of supernatant was transformed to Bond Elut C18 (Agilent 12102058) and dispensed into a 2 mL tube. The hydrophilic compound was extracted with 500 μl of water, and the hydrophobic compound with 600 μl of methanol. The mixture was dried using a centrifugal concentrator under reduced pressure (40 °C, 10 hours). After drying, 20 μl of methoxyamine hydrochloride (20 mg/mL, pyridine solution) was added and the mixture was incubated under 30 °C for 90 min. Then, 70 μl of N-Methyl-N-trimethylsilyl trifluoroacetamide (MSTFA) and 10 μl of myristic acid-d27, serving as an internal standard for retention time locking, were added. The TMS derivatization was performed at 37°C for 30 minutes.
2.4 GC/MS analysis
The GC/MS analysis was performed using an Agilent 5977B GC/MS with the split-splitless inlet, equipped with electron impact (EI) ion source. A DB-5MSUI fused-silica capillary column (30 m, 0.25 mm, 0.5 um, 122-5536UI) was utilized to separate the derivatives. The column thickness generally used in metabolomics is 0.25 μm, but 0.5 μm was employed because the retention time of SCFAs is buried in the peak of the TMS derivatization reagent when 0.25 μm is used. Hydrogen was used as a carrier gas at a constant flow rate of 1.5 mL/min through the column. One microliter of the sample was injected in the split mode at a ratio of 1:50. The solvent delay time was set to 1.8 min. The initial oven temperature was held at 30 °C for 2.76 min, ramped to 100 °C at a rate of 9 °C/min, to 180 °C at a rate of 18 °C/min, to 310 °C at a rate of 2 °C/min, to 325 °C at a rate of 18 °C/min, and finally held at 325 °C for 8.27 min. The temperatures of the injector, transfer line, and EI ion source were set to 250 °C, 300 °C, and 250 °C, respectively. The electron energy was 70 eV, and mass data was collected in a sim/scan mode (m/z 50-700).
2.5 Quantitative validation of compounds
To evaluate the quantitative performance of our method, we assessed the linearity, linearity range, recovery rate, lower limit of detection (LLOD), and precision. The compounds used to validate these assessment items are shown in Supplementary Table 1. Verified compounds include SCFAs, BAs, AAs, sugars, as well as vitamins and indole compounds.
Calibration curves for each metabolite were created by plotting the peak area ratio of the analyte to the internal standard (ribitol) against the analyte concentrations. The calibration curves were prepared using a QC fecal sample to account for potential of volatilization of SCFAs. We employed 10% QC samples, prepared by diluting QC samples to one-tenth of their original concentrations. This dilution strategy enables a precise examination of SCFA’s behavior at lower concentrations, providing insights into the method’s sensitivity and specificity across the analyte concentration range (Table 1). Linearity was assessed by the coefficient of determination (R2) for the linear regression between concentration and relative peak area of each metabolite. Good linearity was defined as R2 being greater than 0.95. The signal-to-noise (S/N) ratio for determining the limit of detection (LOD) was set to a minimum value of 3. For analytes producing multiple peaks, the peak with the highest intensity was generally selected.
To evaluate precision and recovery, a standard mixture of metabolites was spiked into a 100% single feces sample for mimicking the composition of actual fecal samples without any dilution. This approach ensures that our analysis reflects conditions as close to natural fecal samples as possible. For evaluating precision, Ten repeated measurements were conducted once, and five repeated measurements performed twice on separate days. The precision of metabolites was evaluated in terms of intra-day and inter-day variability and expressed as relative standard deviation (RSD) %). The precision was deemed acceptable if less than 15% [26].
The recovery was determined by comparing the concentration of feces spiked with mixed standard solutions before extraction. The recovery rate was calculated as
(pre-extraction concentration / post-extraction concentration) × 100
at each spiking level. The acceptable recovery rate ranged from 70 to 130% [26].
Compounds that met all of these criteria were considered to have passed quantitative validation.
2.7 Mouse samples and animal treatment
Male C57BL/6J mice, 6 weeks old (n = 4), were obtained from CLEA Japan, Inc. (Tokyo, Japan) and were fed a mixed diet consisting of AIN-93G (EP Trading, Tokyo, Japan), CE-2 (CLEA Japan, Tokyo, Japan) and D12492 (EP Trading, Tokyo, Japan).After acclimation, AIN-93G (Control), D12492 (High Fat), and CE-2 (High MAC) were each fed every week (Fig. 4A). At the end of each week, mouse fecal samples were collected and stored at −80 ◦C for further analysis.
Processed and having metabolites extracted similarly to human fecal samples, calibration curves for the concentration and relative area of each metabolite in the mouse study were created using a baseline of 10% mouse QC feces.
2.8 Microbiome analysis from mouse fecal samples
The extraction and measurement of fecal microbial DNA were performed as previously described [27]. Briefly, the fecal samples were initially lyophilized and shaken vigorously using a Shake Master. Samples were then suspended in DNA extraction buffer containing 400 μL of a 1% w/v SDS/TE (10 mM Tris-HCl, 1 mM EDTA; pH 8.0) solution, and fecal samples in the buffer were further shaken with 0.1 mm zirconia/silica beads using a Shake Master (1,500 rpm, 5 min). After centrifugation (20 °C, 17,800 x g, 10 min), bacterial DNA was extracted using an automated DNA extraction machine according to the instruction manual (GENE PREP STAR PI-480). After DNA extraction, the V1–V2 variable region of the 16S rRNA gene was amplified using the bacterial universal primers 27F-mod (5′-AGRGT TTGATYMTGGCTCAG-3′) and 338R (5′-TGCTGCCTCCC GTAGGAGT-3′) with Tks Gflex DNA Polymerase (Takara Bio Inc., Japan) [28]. Amplicon DNA was sequenced using MiSeq (Illumina, USA), according to the manufacturer’s protocol.
2.9 Bioinformatics analysis
For 16S rRNA gene-based microbiome analysis, QIIME2 (version 2019.10) was used [29]. Primer bases were trimmed using cutadapt (option: –p-discard-untrimmed) [30]. Sequence data were processed using the DADA2 pipeline for quality filtering and denoising (options: –p-trunc-len-f 230 –p-trunc-len-r 130) [31]. Contamination by the human genes was checked by mapping the filtered output sequences, and no contamination was found. The filtered output sequences were assigned to taxa using the “qiime feature-classifier classify-sklearn” command with the default parameters [32]. Silva SSU Ref Nr 99 (version 132) was used as the reference database for taxonomic assignment. Alpha and beta diversities were calculated using “qiime phylogeny align-to-tree-mafft-fasttree” and “qiime diversity core-metrics-phylogenetic” commands with the sampling depth set to the lowest read numbers.
2.10 Statistical analysis
All statistical analyses were performed using Python scripts (version 3.7.6). For beta-diversity analysis, microbiome unweighted/weighted UniFrac distance and metabolome spearman correlation distance were used. Distance matrices were visualized via principal coordinate analysis (PCoA) analysis. Each metabolite value was standardized by centering to a mean of 0 and dividing by the standard deviation (z-score) of each metabolite. Z-score was obtained by normalization among all samples. Spearman rank correlation coefficient was used to validate the associations between gut bacteria and metabolites (scipy version 1.5.2). Visualization was performed using Cytoscape 3.10.2 © software based on Spearman correlations between gut bacteria and intestinal metabolites.