2.1 Strains

A. muciniphila (DSM 22959), B. thetaiotaomicron (DSM 2079), and F. prausnitzii (DSM 17677) were ordered from DSMZ (German Collection of Microorganisms and Cell Cultures GmbH) and stored in 20% glycerol and PBS solution (pH 6.8) at −80°C. The F. prausnitzii genome was recently proposed as Faecalibacterium duncaniae sp. nov. (DSM 17677) (Sakamoto et al., 2022).

2.2 Preparation of yeast extract, casitone, fatty acids, and mucin culture medium (YCFAM)

The composition of the medium (if not stated otherwise) was as follows (per litre): 10 g casitone (Tryptone Plus, Sigma-Aldrich), 2.5 g yeast extract (NuCel 545, Bio Springer), 4 g l-cysteine HCl (Alfa Aesar), 1 mg resazurin sodium salt (Acros Organics), 2.93 g K₂HPO₄ (Sigma-Aldrich), 4.65 g KH₂PO₄ (Sigma-Aldrich), 0.9 g (NH₄)₂SO₄ (Alfa Aesar), 0.9 g NaCl (Acros Organics), 0.28 g KOH (Fluka), 1 g NaHCO₃ (Sigma-Aldrich), 10 mg hemin (Sigma-Aldrich), 10 µg biotin (Merck), 10 µg cobalamin (Fisher Bioreagents), 30 µg para-aminobenzoic acid (Sigma-Aldrich), 150 µg pyridoxamine (Carbosynth), 50 µg folate (Alfa Aesar), 0.09 g MgSO₄·7H₂O (Alfa Aesar), 0.09 g CaCl₂·H₂O (ApliChem). The medium was completed with a 1 mM concentration of acetaldehyde (Fluka) and SCFA, including acetate (Honeywell), propionate (Sigma-Aldrich), valerate (Alfa Aesar), isovalerate (Acros Organics), isobutyrate (Acros Organics), formate (Fluka), lactate (Sigma-Aldrich), l-malate (Fluka), citrate (Sigma-Aldrich), and fumarate (Fluka). Mucin from porcine stomach type III (Sigma-Aldrich) was used as a substrate and included at 1.25 g L⁻¹. SCFAs, vitamins, and hemin solution were filter-sterilized and freshly added the day before the experiment. Mucin was sterilized at 115°C for 5 min before being added to the medium. The pH of the medium was adjusted to 7.1 using a 3 M NaOH solution. The final medium, called YCFAM, was kept overnight in an anaerobic chamber (COY box, Coy Laboratory Products Inc.) filled with 5% H₂, 10% CO_2, and 85% N₂ to remove traces of oxygen.

2.3 Culture growth conditions

Precultures were grown overnight in an anoxic YCFAM medium with 0.5% (w/v) glucose. Before inoculation, the cell culture stock was centrifuged at 14,000g for 1 min and resuspended in an anoxic YFCAM medium at a final OD (600 nm) of 0.62. All fermentations were performed under anaerobic conditions in microcalorimetry ampoules (2 mL working volume) containing YCFAM medium inoculated with bacterial culture (1% v/v). After inoculation, the ampoules were hermetically sealed and lowered into isothermal microcalorimeter channels (TAM III and TAM IV-48, TA Instruments) set at 37°C. For each passage, 1% (v/v) of the culture (20 µL) from the prior batch was transferred into a new ampoule containing fresh YCFAM medium (1.98 mL). The growth profile was determined by continuously measuring the heat difference between the IMC control temperature (37°C) and the heat released by the bacterial metabolic activity. For the consortium fermentation, bacteria were prepared as described above. Heat curves from three biological replicates, from three independent IMC vials, were used to determine the mean and standard deviation of all the growth rates. Unless mentioned otherwise, the YCFAM medium was inoculated with 1% (v/v) of each bacterial culture. Each passage was performed as described above. At the end of the fermentation, for both individual strains and the consortium, the supernatant was collected after centrifugation (14,000g, 5 min, 4°C), culture composition was determined, and OD was measured.

2.4 Quantification of extracellular metabolites

Samples were centrifuged at 14,000g for 5 min at 4°C. The supernatant was transferred into a new tube and filtered through 0.2 µm syringe filters (Millipore Millex-LG filters 13 mm Philic PTFE 0.2 µm non-sterile SLLGH13NK; Millipore Corp.). The concentrations of acetate, citrate, butyrate, formate, isobutyrate, isovalerate, lactate, malate, propionate, and succinate were determined using high-performance liquid chromatography (Waters 2695 HPLC system; Waters Corporation). The elution was done isocratically with 0.005 M H₂SO₄ solution at 0.6 mL min⁻¹ using a 300 × 7.8 mm Aminex HPX-87H column (Bio-Rad), with precolumn 30 × 4.6 mm Micro-Guard Cation H Refill Ca (Bio-Rad). A 50/50 milliQ/MeOH (v/v) seal wash, 10/90 acetonitrile/milliQ (v/v) needle wash, autosampler temperature of 10°C, column temperature of 35°C, and injection volume of 20 µL were used. Refractive index (RI) and ultraviolet (UV) detectors (210 nm) (Waters Corporation) were used to detect and quantify metabolites using external standard curves. Standards were prepared by performing serial dilutions with milliQ water: 1x, 2x, 4x, 8x, 16x, 32x, 64x, and 128x.

2.5 DNA extraction and 16S rRNA NGS

The extraction of genomic DNA was done from the biomass pellet using a Sigma-Aldrich GenElute Bacterial Genomic DNA kit and diluted to 10 ng µL⁻¹. For library preparation, the V4 hypervariable region of the 16S rRNA gene was amplified with forward primer F515 5′-GTGCCAGCMGCCGCGGTAA-3′ and reverse primer R806 5′-GGACTACHVGGGTWTCTAAT-3′ (Caporaso et al., 2011) following a standard library preparation pipeline. The libraries were pooled at equimolar concentration and sequenced using the iSeq. 100 Sequencing System (Illumina) following a 2 × 150 cycles paired-end sequencing protocol. Identification and quantification of the reads were done using BION-META software (McDonald et al., 2016). The number of reads for each strain was normalized by the copy number of the 16S rRNA gene (3, 5, and 6 copies for A. muciniphila, B. thetaiotaomicron, and F. prausnitzii, respectively) (Stoddard et al., 2015).

2.6 Free amino acid quantification

The concentration of free amino acids was determined as described by Kivima et al. (2021) with a few modifications. Samples were thawed, vortexed and centrifuged (21,000g, 5 min, RT). Centrifuged samples were filtrated (Millipore Millex-LG filters 13 mm Philic PTFE 0.2 µm non-sterile SLLGH13NK; Millipore Corp.) and the supernatant was diluted 10 times with milliQ water. The free amino acids from the diluted supernatant were derivatized, separated with an AccQ-Tag Ultra RP 1.7 µm (2.1 × 100 mm) column and quantified using a Waters UPLC-PDA system (Waters). For derivatization, 6-aminoquinolyl-N-hydroxysuccinimidyl carbamate (Synchem UG & Co) was used. The elution was carried out using a gradient of AccQTag Ultra eluents A and B (Waters). The data were processed with Empower 2 software (Waters).

2.7 Genome-scale metabolic model (GEM)

2.8 Data analysis

Quantification of extracellular metabolites was carried out with Empower software (Waters Corporation). Principal component analysis (PCA) was performed using R studio (v4.2.0) on centered and scaled data. For other analyzes, data were processed using Excel v20.04 (Microsoft Corporation). Figures were prepared using Inkscape (v1.2.1).

3.1 Specific and reproductive growth pattern

In the current study, A. muciniphila, B. thetaiotaomicron, and F. prausnitzii were initially characterized in monoculture and then in the consortium, with the prior data sets used as references for the consortium.

The characterization of the strains started by assessing their growth stability over several passages. A passage was defined by the transfer of 1% (v/v) of the preceding bacterial culture into fresh YCFAM medium, plus the fermentation time (usually 24 h). Passages were conducted daily and under strictly anaerobic conditions over 7 days. Numbers 1–7 are used in figures to refer to the passages, first to last. The growth kinetics of individual strains and the consortium were monitored using an isothermal microcalorimeter. The heat flow (W) data measured over the seven passages were summarized using a PCA (Figure 1). According to the PCA, the number of passages needed to reach stability (i.e. reproducibility in heat flow curves between passages), varied between individual strains and the consortium. The consortium required three passages to reach stability, while two passages were sufficient for B. thetaiotaomicron and F. prausnitzii monocultures (Figure 1b–d). At least four passages were needed before A. muciniphila monoculture showed stabilized growth kinetics (Figure 1a). From the fifth passage onward, the stability of the heat flow measurement was observed under all conditions. The heat flow from the fifth to seventh passages was used, after transformation into heat (J), to calculate growth rates. The existence of a linear relationship between heat and biomass was additionally confirmed before any growth rate calculations were made (Table S2).

The growth kinetics of A. muciniphila were characterized by a rapid increase of heat flow in the first 10 h after inoculation with a specific growth rate of 0.57 ± 0.02 h⁻¹ (Figure S1A,B). This specific growth rate was slightly higher than the highest growth rate in basal medium reported by Noora et al. (2017) (0.41 ± 0.05 h⁻¹). This difference could be attributed to the presence of yeast extract and casitone in our medium, as well as the higher concentration of mucin, which has been shown to trigger and modulate gene expression (Liu et al., 2021). The heat flow curves of B. thetaiotaomicron and F. prausnitzii showed multiple phases (Figure S1A,C,D). These phases were likely caused by metabolic switches triggered by the depletion of the most favorable substrate(s), thus reducing the growth rate at each new phase. For both B. thetaiotaomicron and F. prausnitzii, the highest growth rate occurred in the first few hours after inoculation (0.88 ± 0.06 h⁻¹ and 1.00 ± 0.08 h⁻¹, respectively) followed by a second growth phase (0.20 ± 0.01 h⁻¹ and 0.21 ± 0.01 h⁻¹, respectively) and a third phase (0.07 ± 0.002 h⁻¹ and 0.09 ± 0.004 h⁻¹, respectively). Prior genomic sequencing studies confirmed the presence and expression of mucin-associated hydrolases and catabolizing pathways in B. thetaiotaomicron (Glover et al., 2022; Martens et al., 2011; Ravcheev et al., 2013). Moreover, Martens et al. (2011) already observed complicated polyphasic growth profiles with B. thetaiotaomicron. In mineral medium (ZBW1), B. thetaiotaomicron growth rate was 0.015 ± 0.002 h⁻¹. This growth rate was seven times lower than the lowest measured in our YCFAM medium. This difference was most certainly explained by the consumption of other components (casitone, yeast extract) present in the YCFAM medium. In Lopez-Siles et al. (2012) study, F. prausnitzii was unable to grow with mucin as a carbon source while in our YCFAM medium, its growth rate ranged from 0.09 ± 0.004 h⁻¹ to 1.00 ± 0.08 h^-1. The comparison between our and Lopez-Siles et al. (2012) medium showed that ours contained an additional six SCFAs, had a higher concentration of l-cysteine and included phosphate buffers. Interestingly, growth rates obtained in the YCFAM medium were more consistent with those reported in the presence of mucin-derived monosaccharides, such as N-AcGam, glucosamine, or glucose (Duncan et al., 2002; Lopez-Siles et al., 2012).

After individually characterizing each strain, A. muciniphila, B. thetaiotaomicron, and F. prausnitzii were co-inoculated in the YCFAM medium, starting with the same initial cell density. The three-strain consortium growth was monitored by IMC over seven passages and compared to the individual strains (Figure S1A). In accordance with PCA (Figure 1d), the robustness of the consortium composition improved with the number of passages. From the fourth passage, the consortium heat flow curves were comparable to the combination of the curves from individual strains (Figure S1A). To confirm that the heat flow released by the consortium originated from the metabolic activities of all three strains, their presence was confirmed by 16S rRNA sequencing (Figure 2b). B. thetaiotaomicron, A. muciniphila, and F. prausnitzii represented 52 ± 2%, 34 ± 2%, and 13 ± 1% of the final consortium, respectively (Table S3).

As a further step in consortium characterization, the growth rate of individual strains within the consortium was determined. The abundance of each strain within the consortium was quantified hourly over 24 h using 16S rRNA amplicon sequencing. These abundances were used to reconstruct the growth curve of individual strains from the total heat generated by the consortium. Regardless of the strain, the growth rates calculated from the reconstructed curves resembled those obtained with monoculture cultivations (above paragraph and Figure 2c). In this condition, the results suggested that B. thetaiotaomicron, A. muciniphila, and F. prausnitzii stably coexist in the YCFAM medium.

3.2 Evaluation of consortium resilience through growth kinetics and 16S rRNA NGS

In their work, Allison and Martiny (2008) defined resilience as the ability of a community to quickly recover from a disturbance whether by growth, physiological, or genetic adaptation. Recent works suggested that pH, medium composition, taxonomic group, and species initial abundance influence community stability over serial passages (Goldford et al., 2018; Segura Munoz et al., 2022; Zegeye et al., 2019).

To assess the resilience of our consortium, the effect of species’ initial abundance was tested across three combinations (Figure 3). In each combination, one strain was inoculated at a concentration a hundred times lower than the other two strains. The composition of the consortia was evaluated at the end of each passage. Resilience was achieved when the growth kinetics and composition of a combination closely resembled the reference (Figure 3a).

The growth kinetics were significantly altered from the reference consortium when the initial concentration of BT and AM were lowered (Figure 3a, first two panels). A similar growth kinetics to the reference was retrieved after the second and fifth passages, respectively. Similarly to B. thetaiotaomicron, two passages were also sufficient for F. prausnitzii (Figure 3a, third panel). In coherence with the heat flow, a return to the former consortium composition was also obtained by 16S rRNA sequencing at the end of the fifth passage (Figure 3b [passage 5] and Figure 2b [passages 5–7]). The composition of the consortia from passages 1–5 is reported in Table S4. This return could not be achieved in the event of drastic competition for growth-dependent substrate(s).

3.3 A slight interaction is suggested by changes in organic acid concentrations

Although the three strains proved able to thrive in the consortium, some metabolic changes may have been necessary to maintain the stability. To evaluate this hypothesis, the production and consumption of extracellular organic acids were quantified at the end of monoculture and consortium cultivations (Table S5). As reported in the literature, F. prausnitzii was characterized by the secretion of butyrate and formate (Figure 4) (D'hoe et al., 2018; Lopez-Siles et al., 2012; Wrzosek et al., 2013). However, the consumption of acetate by F. prausnitzii was not confirmed in our study. B. thetaiotaomicron and A. muciniphila showed similar excretion profiles, with significant production of acetate and propionate, and consumption of malate (Das et al., 2018; Noora et al., 2017; van der Ark et al., 2018). The secretion of both isovaleric and isobutyrate and the consumption of lactate were measured only with B. thetaiotaomicron. The consortium had a significantly higher concentration of acetate, butyrate, and succinate compared to the combination of concentrations produced by individual strains (Figure 4; Table S5). The higher production of butyrate in the consortium was not correlated with the expected decrease in acetate concentration suggested by Wrzosek et al. (2013). However, at the end of the consortium fermentation, the concentration of acetate was eight times higher compared to F. prausnitzii monoculture in YCFAM, which could have enhanced butyrate production by F. prausnitzii (Figure 4).

3.4 Free amino acid consumption in the consortium is mainly influenced by B. thetaiotaomicron

In addition to organic acids, the concentrations of 20 free amino acids were assessed at the end of the fifth, sixth, and seventh passages.

The difference in secretion and consumption of free amino acids between individual strains and consortium was examined using PCA and including the initial medium composition as a reference. The first and second components (PCs) explained 99.9% of the variability (Figure S2). The consortium and B. thetaiotaomicron were separated from F. prausnitzii, A. muciniphila and the medium samples in PC2. This result suggested that in the consortium the free amino acids consumption in the consortium was mainly driven by B. thetaiotaomicron. At the end of A. muciniphila and F. prausnitzii cultivations, the concentration of the free amino acids was comparable to the reference, suggesting either a negligible or no consumption of any free amino acids (Table S6). This observation differed greatly from the net consumption and excretion of 10 amino acids by F. prausnitzii reported by Heinken et al. (2014). In addition, our data did not confirm threonine consumption by A. muciniphila as predicted by Noora et al. (2017) and reported by van der Ark et al. (2018) (Table S6). At the end of B. thetaiotaomicron cultivation, over half of the measured free amino acids had increased in the medium (Table S6). Phenylalanine and tryptophan concentrations remained unchanged. In comparison to the reference, the abundances of proline and glutamine were higher. Furthermore, aspartate and asparagine were depleted. Data reported by Catlett et al. (2020) suggested that B. thetaiotaomicron uses l-asparagine hydrolase (EC 3.5.1.1; BT_0526, BT_2404, BT2757) to convert asparagine to aspartate. Aspartate can then enter the TCA cycle as oxaloacetate (EC 2.6.1.1; BT_2415), resulting in the generation of ATP and cofactor (NADPH). From the overflow in the TCA cycle, the succinate that is produced by succinyl-CoA ligase (EC 6.2.1.5; BT_0788, BT_0787) is either secreted and/or mobilized in the respiratory chain to generate the proton gradient required for ATP production (Catlett et al., 2020). In addition to asparagine and aspartate, serine was also consumed and can be converted into pyruvate (EC 4.3.1.17; BT_4678).

3.5 Metabolic modeling suggests distinct growth-limiting substrates for individual strains

In this study, FBA was used to identify strain-specific and shared carbon sources. It was assumed that all strains could uptake and catabolize amino acids and that F. prausnitzii was the only strain unable to cleave, and thus consume, mucin-derived monosaccharides. A. muciniphila, B. thetaiotaomicron, and F. prausnitzii genome-scale metabolic models (GEMs) consisted of 1308, 1491, and 1137 metabolic reactions, respectively (Table S1). GEMs were constrained using experimental rates (Table S7, Section 2.7.1). The analysis looked for the flux distributions for which the production of biomass, i.e., specific growth rate, was the most optimized (Tables S8–S10). It is important to mention that an average experimental growth rate was calculated and used to constrain GEMs for B. thetaiotaomicron and F. prausnitzii. Growth of A. muciniphila was predicted to be mainly sustained by threonine and N-AcGam consumption (Figure 5a). Threonine was predicted to enter the TCA cycle as succinyl-CoA to generate ATP. In addition, threonine was converted to glycine and serine. These predictions were not confirmed experimentally (Table S6).

B. thetaiotaomicron was predicted to consume preferentially aspartate and N-AcGam (Figure 5b). As described in Section 3.3 for B. thetaiotaomicron, aspartate entered the TCA cycle to generate ATP, leading to the excretion of succinate and propionate. For both A. muciniphila and B. thetaiotaomicron, N-AcGam entered glycolysis as fructose 6-phosphate in three metabolic reactions. The energy (ATP) required to grow was generated by glycolytic enzymes (pyruvate kinase PYK, phosphoglycerate kinase PGK) and TCA enzyme (succinyl-CoA synthetase SUCOAS) (Tables S8–S10). In A. muciniphila metabolism, ATP was actively used by formate tetrahydrofolate ligase (FTHFLi) and phosphofructokinase (PFK). For B. thetaiotaomicron, the prediction showed that ATP turnover was highest for N-acetylglucosamine kinase (ACGAMK) and glucose-1-phosphate adenylyl transferase (GLGC). In the absence of evidence suggesting that F. prausnitzii could cleave mucin and thus consume any of the monosaccharides, FBA predicted serine to be the main carbon source for F. prausnitzii (Figure 5c). Serine was predicted to enter central carbon metabolism after being converted by l-serine ammonia-lyase (SERD_L) to pyruvate. This result was consistent with the data reported in Auger et al. (2022), showing a higher expression of serine/threonine transporter (SstT) at an early stage of fermentation. However, the depletion or consumption of serine by F. prausnitzii was not experimentally confirmed (Table S6). Acetate kinase reaction (ACKr) was predicted as an almost unique ATP source for F. prausnitzii, with one ATP produced per acetate catabolized. The limited use of the glycolysis pathway to generate ATP was due to the absence of mucin-derived monosaccharides consumption.

The following step aimed to determine which of the predicted carbon sources were essential to sustain growth and which were potentially shared by species (Table S11). The depletion of a substrate from the medium was mimicked by setting the consumption rate to zero while optimizing biomass production. In the case of A. muciniphila, the depletion of N-AcGam, asparagine, or threonine entirely stopped the growth, while the other substrates had only a slight effect on the predicted growth rate (Figure 5d). These results were coherent with a previous study where N-AcGam and l-threonine were essential for the growth of A. muciniphila in a minimal medium (van der Ark et al., 2018). Although asparagine was not shown as an essential nutrient, its consumption was confirmed in a BHI medium supplemented with mucin (Liu et al., 2021). B. thetaiotaomicron, the most robust strain, had the widest range of substrates and was slightly affected by the depletion of most of its carbon sources (Figure 5e). Only the depletion of aspartate or N-AcGam had a significant effect on its growth rate. For F. prausnitzii, serine, threonine, and uracil turned out to be essential for its growth in this specific condition (Figure 5f). Interestingly, the omission of serine, threonine, and uracil in the defined medium used by Heinken et al., (2014) still allowed the growth of F. prausnitzii. Nonetheless, the growth in this medium was described by the authors as poor and inconsistent (lower than 0.13 h⁻¹). Although predicting fluxes using a complex medium is controversial, our results suggested that threonine and N-AcGam were the common substrates that strains could compete for within the consortium via which they could alter each other's growth.