Fecal samples from eight healthy adult donors [six female (D1, D3, D5, D6, D7, and D8) and two male (D2 and D4); median age of 29 years old; western style diet] who fulfilled the following inclusion criteria: good general health with no symptoms of chronic or autoimmune disease, a regular eating pattern (3 meals/day, 5 days/week), regular bowel movements (once every 3 days to 3 times per day), no use of immune suppressing drugs, blood thinners, antibiotics, or other medication affecting the gut transit and/or digestion, no intake of pre- and probiotics during the month preceding the sample donation, no regular intestinal discomfort, no operative intestinal interventions, and not being on a weight loss diet were used for the study.

Fecal sample was collected directly after defecation using a sterile container and an anaerobic atmosphere generator (Anaerogen, Thermo Fisher Diagnostics AG, Pratteln, Switzerland) and transported within 1 h into an anaerobic chamber (10% CO₂, 5% H₂, and 85% N₂; Coy Laboratories, Grass Lake, MI, US). A 20% w/v fecal suspension was prepared in sterile anaerobic peptone water (0.1%, pH 7; Thermo Fisher Diagnostics AG). The suspension was homogenized by vortexing in the presence of sterile 5 mm glass beads, let stand for 5 min to allow particles to settle to the bottom. The resulting supernatant was removed with sterile pipettes and transferred to an autoclaved anaerobic Serum flask and used for batch inoculation. The ethics committee of ETH Zurich exempted this study from review because the sample collection procedure was not performed under conditions of intervention, and it was carried out in an anonymized manner.

Brown serum flasks (20 ml, Infochroma AG, Goldau, Switzerland) containing 10 ml of sterile anaerobic Macfarlane medium adapted for batch fermentation conditions [two fold buffering capacity; pH 7.0; composition described in Bircher et al. (13)] were used. The Macfarlane medium was previously developed from the analysis of intestinal content to mimic the human proximal colon chyme of Western adults (14). This medium includes the addition of a filter-sterilized vitamin solution (15), leading to a final vitamin concentration of (ng/ml medium): folic acid, 20; cyanocobalamin, 5; pyridoxine-HCl, 100; 4-aminobenzoic acid, 50; nicotinic acid, 50; biotin, 20; thiamine, 50; riboflavin, 50; phylloquinone, 0.075; menadione, 10; pantothenate, 100 (all purchased from Sigma-Aldrich Chemie GmbH, Buchs, Switzerland). Vitamins were weighted in the dark and the vitamin solution was filtered sterilized through 0.2 μm nylon membrane filter (Infochroma AG), covered by aluminium foil and stored at 4^°C before being supplemented to the autoclaved fermentation medium.

Each fecal microbial suspension was grown in three different media containing distinct levels of cyanocobalamin: (i) Control: MacFarlane medium containing normal vitamin solution (5 ng/ml cyanocobalamin); (ii) NB12: no cyanocobalamin added in the vitamin solution used for the MacFarlane medium and (iii) ExtraB12: MacFarlane medium supplemented with high dose of 2500 ng/ml cyanocobalamin final concentration. No B12 was detected in the MacFarlane medium without B12 and in Control medium by UHPLC-DAD method (measured in triplicate; detection limit 30 ng/ml). Therefore, a competitive enzyme immunoassay (RIDASCREEN FAST Vitamin B12, R-Biopharm, Switzerland) was used to quantify B12 content in MacFarlane medium, with and without vitamin mix, quantified as 8.9 ± 0.1 ng/ml and 4.1 ± 0.4 ng/ml, respectively. The detected B12 in the non-supplemented MacFarlane medium might originate from yeast or meat extract added to the medium. The applied concentrations in our in vitro colon batch (control 5 ng/ml and ExtraB12 2500 ng/ml) are in the range of what we can expect under average diet conditions [average B12 intake of 4.5 μg/day, (16)] or with extreme B12 supplementation [1500 μg per tablet, (10)]. We therefore considered a small intestine absorption of 50% in the case of average dietary intake (17), and 1% in the case of high dose (11, 17). Consequently, 2.25 μg of B12 under average diet conditions and 5940 μg of B12 under extreme supplementation conditions will arrive in the proximal colon per day. Taking a proximal colon volume of 200 ml (18), and a retention time of 8 h (19), resulting in 600 ml proximal colon suspension per day, which corresponds to the average diet condition of 3.75 ng/ml and with extreme B12 supplementation of 2475 ng/ml.

The flasks were inoculated with the fecal suspension (1% v/v), closed with sterile rubber stoppers and aluminium crimp caps N 20 (Macherey-Nagel AG, Oensingen, Switzerland) and incubated anaerobically at 37^°C on a shaker (100 rpm) directly in the anaerobic chamber. After 24 h incubation, 1 ml samples were transferred to a sterile Eppendorf tube and centrifuged at 13,000 × g for 10 min at 4^°C. The supernatant was separated from the pellet and filtered through a 0.45 μm nylon membrane filter into a sterile Eppendorf tube. Pellets and supernatants were stored at –20^°C until further analysis. For each tested condition and fecal microbiota, fermentations were conducted in triplicates.

SCFA (acetate, propionate, butyrate, and valerate), intermediate metabolites (lactate, succinate, and formate), and branched-chain fatty acids (BCFA; isobutyrate and isovalerate) were analyzed using HPLC-RI in fecal and fermentation samples. Briefly, 200 mg of feces were weighted and mixed with 600 μl of HPLC eluent (100 mM H₂SO₄) in a tube containing one glass bead. After homogenization by vortex, samples were centrifuged for 20 min at 4^°C (6000 × g) and the supernatant was filtered through a 0.2 μm filter into a glass vial and closed with a crimp cap. For fermentation samples, 100 μl of filtered supernatant was diluted with 100 μl filter-sterilized pure water in sterile Eppendorf tubes. The tubes were vortexed, and the content was transferred into a Crimp Neck Vial (BGB Analytik AG, Boeckten, Switzerland). Samples were analyzed using a HPLC (Hitachi LaChrom, Merck, Dietikon, Switzerland) equipped with a precolumn SecurityGuard Cartridges Carbo-H (4 × 3.0 mm; Phenomenex Helvetia GmbH, Basel, Switzerland) connected to a Rezex ROA-Organic Acid H + column (300 × 7.8 mm; Phenomenex Helvetia GmbH, Switzerland) and a refractive index detector (20). Samples (40 μl injection volume) were eluted at 40^°C under isocratic conditions with 10 mM H₂SO₄ at a flow rate of 0.4 ml/min. All metabolites were quantified using external standards (Sigma-Aldrich for all, except valeric acid and formic acid from VWR International AG).

The protocol for B12 extraction from fecal suspension (1 ml) prepared from homogenized fecal slurry and from fermentation samples (1 ml) was adapted from Chamlagain et al. (21). The protocol includes the addition of potassium cyanide (KCN) to the samples to convert all B12 forms into the stable cyanocobalamin. In fecal suspension samples, total B12 (intracellular and extracellular) was quantified, while in fermentation samples, both intracellular and extracellular B12 was measured separately. For fecal and extracellular quantification, 1 ml extraction buffer (8.3 mM NaOH and 20.7 mM acetic acid) and 10 μl of 1% KCN solution were added to 1 ml sample in 15 ml falcon tubes (Bioswisstec AG, Schaffhausen, Switzerland) and vortexed for 10 s. For intracellular quantification, pellets were resuspended in 1 ml extraction buffer, transferred into 15 ml falcon tubes, 10 μl 1% KCN solution was added, and the tubes were vortexed for 10 s. Samples were then placed in a water bath at 99^°C for 30 min and afterward ice-cooled for 10 min. Fecal and intracellular samples were centrifuged for 20 min at 4^°C (10000 × g). All samples were filtered through a 0.20 μm nylon membrane filter into brown vials (BGB Analytik AG). All the analyses were carried out under red or dim light conditions to avoid B12 degradation during sample processing.

Cyanocobalamin quantification was performed on a Thermo Scientific Vanquish UHPLC equipped with a Diode Array Detector (DAD) using an Acquity HSS T3 column (2.1 × 1000 mm, 1.8 μm; Waters AG, Baden-Dättwil, Switzerland), connected to ACQUITY UPLC HSS T3 VanGuard Pre-column (100Å, 1.8 μm, 2.1 × 5 mm; Waters AG). The detection was performed at 362 nm. The autosampler was maintained at 12^°C and the column was operated at 30^°C. Injection volume was 10 μl. The separation was performed using the mobile phase acetonitrile (A) and MilliQ water (B) both containing 0.1% formic acid using a gradient flow (0.5 ml/min): 0–0.5 min [A 5% and B 95%]; 0.5–5 min [A 40% and B 60%]; 5–6 min [A 40% and B 60%]; 6–10 min [A 95% and B 5%]; 10–11 min [A 5% and B 95%]; 11–15 min [A 5% and B 95%]. The calibration curve was done by measuring cyanocobalamin standards (Sigma-Aldrich) in the 60–4000 ng/ml range. Based on linearity, the limit of detection was 30 ng/ml.

DNA was isolated from fermentation pellets obtained from 150 mg homogenized fecal samples and from 1 ml of fermentation sample using the FastDNA Spin Kit for Soil (MP Biomedicals, Illkirch-Graffenstaden, France), following the manufacturer’s protocol. Briefly, samples were placed into 2 ml tubes containing Lysing Matrix E and homogenized (40 s, 6.0 m/s) in a FastPrep^® instrument in the presence of MT buffer (lysis solution) and sodium phosphate buffer. Following lysis, samples were centrifuged, and DNA was purified from the supernatant with a silica-based filter in a final volume of 100 μl. DNA was quantified using a Nanodrop^® ND-1000 Spectrophotometer and stored at –20^°C until further analysis.

Quantitative PCR (qPCR) was used to evaluate the abundance of total bacteria in the fecal samples by quantifying the number of copies of the 16S rRNA gene using primers 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 518R (5′-ATTACCGCGGCTGCTGG-3′) (22). Each reaction consisted of 5 μl of 2X SensiFAST SYBR No-ROX Kit master mix (Meridian Bioscience, Cincinnati, OH, USA), 0.5 μl of each forward and reverse primer (final concentration of 0.5 μM each), 1 μl of DNA and 3 μl of nuclease free water. Reactions per sample were performed in triplicates in 96-well plates using a LightCycler 480 qPCR II system (Roche Diagnostics, Rotkreuz, Switzerland) and a two-step program consisting of 3 min of initial denaturation at 95^°C, followed by 40 cycles of 95^°C for 5 s and 65^°C for 30 s. For quantification, a 10-fold dilution series of standards containing a linearized plasmid with the Escherichia coli 16S rRNA gene was included in each run. PCR efficiency (%) was calculated from the slope of the standard curve of each qPCR assay. Assays with an efficiency of 80–110% (slope of 3.2–3.9) were included in the data analysis. The qPCR gene copy number was converted to cell concentration by correcting for the median of five 16S rRNA gene copy number per bacterium based on the Ribosomal RNA Database (23).

The microbial composition of fecal inoculum and batch fermentation samples were analysed using 16S rRNA gene sequencing. Briefly, a two-steps PCR approach was used to target the V4 region using primers 515F (5′-GTGCCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) (Microsynth AG, Balgach, Switzerland) (24). Samples were barcoded using Nextera XT v2 indexes (Illumina, San Diego, CA, USA) and pooled in equimolar concentration. Sequencing was performed with the Illumina MiSeq platform (Genetic Diversity Centre, ETH Zurich, Switzerland) using v2 chemistry supplemented with PhiX 20 pM (10%) and 250 × 2 read length. Raw data were processed using the DADA2 R package (version 1.14.1) (25), to obtain exact Amplicon Sequence Variants (ASVs) using the metabaRpipe R package (26). Briefly, forward and reverse reads were truncated after 260 and 250 nucleotides, respectively. After truncation, reads with expected error rates higher than three and four for forward and reverse reads, respectively, were removed. After filtering, error rate learning, ASV inference and denoising, reads were merged with a minimum overlap of 15 nucleotides. Chimeric sequences were identified and removed. Taxonomy was assigned to ASVs using DADA2 against SILVA database (v138.1) (27, 28). For each fecal donor, enterotype was assigned based on the ASV table at genus level at,¹ and clustered with Jensen-Shannon divergence (jsd) distance matrix. For fermentation samples, phylogenetic relatedness of ASVs, alpha- (Observed and Shannon indexes) and beta-(Unweighted and Weighted Unifrac distance matrices) diversity analyses were performed using the phyloseq (29) and the DivComAnalyses packages (30) in R Studio (31). ASVs were filtered using the filterByExpr function from the edgeR package (32), and differential analysis of ASVs (NB12 vs. Control, ExtraB12 vs. Control, and ExtraB12 vs. NB12) was performed using the limma-voom pipeline (33) from the limma package (34). ASVs that had an absolute log₂ fold-change greater than 1 and P < 0.001, were considered differentially expressed and dotplots were generated using the ggplot2 package (35).

A non-parametric, Wilcoxon statistical analysis was performed to detect significant differences in alpha diversity between treatments for all donors together, and for each donor separated. The same approach was used for the beta diversity distance matrixes, by PERMANOVA analysis. Both analyses were performed via Phyloseq. The statistical analysis for total bacteria, metabolites, and B12 content was done using GraphPad Prism v.9.2.0. An unpaired t-test (P < 0.05, two-tailed) was performed per each donor to compare different treatments, and the Normality of Residuals was validated using the Shapiro–Wilk test. A two-way ANOVA was performed on the absolute concentration for total SCFA, intermediate, BCFA, and for individual metabolites to determine the main effects of donors and treatments, followed by Tukey’s test (P < 0.05) for post-hoc analysis to correct for multiple comparisons.

Microbial composition and B12 content in fecal samples of eight adult healthy donors were first measured. Firmicutes phylum in all donors ranged from 52.2 to 88.6% relative abundance, as measured by relative abundance on 16S data, with D3 containing also a high Bacteroidetes level (39.5% relative abundance; Figure 1A). Lachnospiraceae was the most abundant family in all donors (ranging from 33.2 to 46.3%), except for D3 (26.5%), dominated by Prevotellaceae at 37.8% (Supplementary Figure 1A). Based on 16S data at the genus level, five donors were assigned to Firmicutes (D1, D2, D5, D6 and D8), two to Prevotella (D3 and D4), and one to Bacteroides (D7) enterotype based on (36) (see text footnote 1) (Figure 1B and Supplementary Figure 1B).

The analysis of fecal SCFAs indicated that D2 and D5 exhibited a butyrogenic profile, having an acetate:propionate:butyrate ratio of 60:16:24 and 51:23:26, respectively (Figure 1C and Supplementary Figure 2). All other donors were propionigenic, with higher propionate than butyrate concentrations in the feces (Figure 1C and Supplementary Figure 2). Fecal total B12 content ranged from 684 ng/g in D7 up to 4222 ng/g in D8, with a median at 2716 ng/g, and was below the detection limit in D1 with a detection limit of the UHPLC-DAD method of 30 ng/ml (Figure 1D). No lactate and formate were detected in fecal samples (limit of quantification 1 mM).

Batch fermentations of the eight fecal microbial communities grown in buffered MacFarlane medium without added B12 (NB12), and with B12 added in low (Control) or high (ExtraB12) levels were performed for 24 h at 37^°C under strict anaerobiosis. Extracellular and intracellular B12 were quantified at time 0 (T0) and 24 (T24).

No B12 was detected in Control and NB12 treatment samples for all donors at T0 (Figures 2A,B), while a range from 1860 ± 34 to 2523 ± 141 ng/ml B12 was quantified at T0 in ExtraB12 (Figure 2C). A significant total B12 production was measured for both Control (122 ± 4 to 404 ± 57 ng/ml) and NB12 (152 ± 3 to 564 ± 27 ng/ml) at T24 for all tested donors. Microbiota D3, D5, D6, D7, and D8 had higher levels of intracellular B12 at T24 compared to extracellular (intracellular B12 ranging from 83 ± 9 to 201 ± 15 ng/ml to in Control, from 83 ± 7 to 180 ± 28 ng/ml in NB12, from 188 ± 7 to 370 ± 26 ng/ml in ExtraB12), while D1, D2, and D4 had intracellular B12 content comparatively lower than extracellular (intracellular B12 ranging from 0 to 117 ± 12 ng/ml in Control, from 11 ± 19 to 145 ± 10 ng/ml in NB12, from 31 ± 3 to 506 ± 69 ng/ml in ExtraB12). ExtraB12 treatments showed different responses depending on the donor microbiota after 24 h incubation. A significantly higher concentration of total B12 was observed for D1 and D2, compared to T0, indicating that B12 was produced despite B12 surplus availability. In contrast, a significantly lower content of B12 was quantified at T24 compared to T0 for D6 and D8. For D3, D4, D5, and D7, no significant changes in total B12 content were measured at T24, as compared to T0 (Figure 2C). In all tested conditions, B12 content in D4 was mainly extracellular. For all other donors, a significant higher intracellular B12 content at 24 h was detected in ExtraB12, as compared to NB12 (Supplementary Table 1).

Overall, our data indicate that all gut microbial communities tested could produce B12 after 24 h of fermentation when grown in absence or low amount of B12 in the medium.

Cell numbers were determined by qPCR and microbiota composition was measured at T24 by 16S rRNA sequencing. Total bacteria concentration at T24 were not significantly different among tested conditions and ranged between 10.3 ± 0.15 to 10.8 ± 0.11 log total bacteria in Control samples, 10.5 ± 0.05 to 10.9 ± 0.10 log total bacteria in NB12, and 10.5 ± 0.5 to 11.0 ± 0.2 log total bacteria in ExtraB12 (Supplementary Figure 3).

Furthermore, no significant changes in overall microbiota composition in response to treatments were observed (Supplementary Figure 4), as confirmed by beta diversity distance matrices (Figure 3 and Supplementary Figure 5). Neither NB12 nor ExtraB12 significantly impacted alpha diversity among treatments when grouping all donors together (Supplementary Figure 6) and when considering individual donors, compared to Control (Figure 4). Altogether, B12 availability in the medium did not or only show a small impact on total bacterial growth and microbiota composition and diversity, which was rather determined by the initial gut microbial composition of each donor.

A total of 91 ASVs for all donors changed in abundance at different B12 concentrations, of which 81% belonged to the Firmicutes, and 15% to the Bacteroidetes phylum (Figure 5). Most ASV changes were found in only one microbiota, implying donor-specific changes at the ASV level upon B12 treatments. However, few ASVs responded in more than one microbiota for each of the treatments, as compared to Control. Specifically, for NB12 ASV0332 (Lactobacillus) in D1, D2, D5, and ASV0113 (Barnesiella intestinihominis) in D2 and D5 responded with increased abundance, as compared to Control. In ExtraB12, two ASVs both assigned to Parabacteroides distasonis (ASV0076 and ASV0125) increased in abundance in D7, D8 and D4, D7, respectively, as compared to Control. When NB12 was compared to ExtraB12, none of the ASV were shown to respond in more than one donor (Figure 5).

Overall, observed changes at the individual ASV level did not result in changes in microbiota community structure, as supported by alpha and beta diversity analyses.

B12 is required as a cofactor by gut microbes for various biological processes, including acetate production via the Wood-Ljungdahl pathway, and for the conversion of succinate to propionate. After 24 h, the proportion of total metabolites did not change in NB12 or ExtraB12, as compared to Control, for all tested microbiota (Figure 6), independently of the initial donor fecal metabolite profile (Figure 1C).

When considering all donors together, significant variations in individual metabolite production due to the donor’s effect were reported by two-way ANOVA (Supplementary Table 2), while none of the treatments impacted total metabolite production. Overall, a significant increase in propionate in ExtraB12 was reported by ANOVA, as compared to NB12 (Supplementary Table 3). Furthermore, a higher concentration of succinate was observed in Control, compared to ExtraB12 (Supplementary Table 3).

Donor-dependent metabolic responses were observed in individual donors (Figure 7). However, there were some metabolic changes observed in more than one donor. Briefly, D3 and D4 responded to NB12 with a significant decrease in butyrate (–3.4 ± 0.8 mM and –6.2 ± 0.7 mM, respectively) and isobutyrate (–2.1 ± 0.2 mM, –4.2 ± 0.3 mM), compared to Control. The same donors reported a significant increase in propionate (+ 8.35 ± 1.8 mM, + 4.5 ± 0.5 mM, respectively) when NB12 was compared to ExtraB12 (Figure 7A). On total intermediate metabolites, D3 and D6 showed a significant decrease (–3.16 ± 0.58 and –0.42 ± 0.14 mM, respectively) in response to NB12, compared to Control, while the opposite trend was observed in ExtraB12 for D4 (4.58 ± 0.83 mM) and D8 (5.01 ± 0.88 mM), due to formate increase (Figure 7B). In response to NB12, total BCFA decreased in D4 (–9.54 ± 0.61 mM) and D8 (–2.35 ± 0.49 mM), as compared to control, while in response to ExtraB12, BCFA decreased in D5 (–8.16 ± 1.43 mM) and D8 (–2.93 ± 0.40 mM) (Figure 7C).

To conclude, overall, the treatments with different doses of B12 had no effect on total metabolite production, though minor effects were observed on propionate and succinate production (Supplementary Tables 2, 3 and Figure 7). No lactate was detected in fermentation samples (limit of quantification 1 mM).