Chemicals were obtained from Alfa Aesar (Kandel, Germany), Carl Roth (Karlsruhe, Germany), Chem-Lab (Zedelgem, Belgium), Keyser & Mackay (Brussels, Belgium), Merck (Overijse, Belgium), Oxoid (Merelbeke, Belgium), Sigma-Aldrich (Overijse, Belgium), and VWR (Leuven, Belgium), as elaborated below.

During the present study, several inoculation strategies were compared. This involved the use of faecal inocula from healthy human adults collected and prepared as a 20% (w/v) solution of the fecal sample in anaerobic phosphate buffer as previously described (Molly et al., 1993), ileostomy effluents and a synthetic consortium comprising 12 bacterial species (Table 1). Samples derived from human donors were collected according to the ethical approval of the University Hospital Ghent with reference number B670201836585.

Glycerol stocks were prepared for each strain separately. Grown cultures were harvested during exponential growth and 0.5 ml culture was mixed with 0.5 ml glycerol-based cryoprotectant (50% v/v glycerol (86%) [Merck], 0.5 g L⁻¹ l-cysteine-HCl [Merck], 10 g L⁻¹ trehalose [Sigma-Aldrich], 3 g L⁻¹ tryptic soy broth (TSB) [Oxoid], 18 g L⁻¹ NaCl [Chem-Lab], 1 ml L⁻¹ resazurin [Alfa Aesar] stock solution (2% w/v) and 50% (v/v) dH₂O). For strict anaerobic strains, the glycerol-based cryoprotectant was boiled for 1′ to reduce oxygen content prior to autoclaving. Glycerol stocks were stored at −80°C until usage.

All experiments were performed through modification of the M-SHIME® technology previously described by Van den Abbeele et al. (2012). Setups were autoclaved prior to inoculation to assure sterility of the vessels. Furthermore, reactors were continuously homogenized through stirring, maintained at 37°C, and kept anaerobically by flushing with N₂ for 15 min, three times per day.

Quantification of SCFAs, including acetate, propionate, butyrate, and branched SCFAs (BSCFA; i.e., isobutyrate, isovalerate, and isocaproate) was performed as previously reported by Ghyselinck et al. (2020). Lactate concentrations were determined on supernatant aliquots (centrifuged for 5 min at 7,690 × g) using a commercially available enzymatic assay kit according to the manufacturer’s instructions. For the screening experiment and reproducibility experiment, the enzymatic assay kit of Roche was used (Roche Diagnostics, Machelen, Belgium) while the R-Biopharm kit was used for lactate quantification of the optimization experiment and incorporation experiment (R-Biopharm, Darmstadt, Germany). Concentrations of the bile salts taurocholic acid (TCA), glycocholic acid (GCA), taurodeoxycholic acid (TDCA), and glycodeoxycholic acid (GDCA) were determined through reversed-phase high pressure liquid chromatography (RP-HPLC; Hitachi Chromaster, Hitachi, Brussels, Belgium) with a diode array detector (DAD; VWR), using a reversed-phase C18 column (Hydro-RP, 4 μm, 80 Å, 250 × 4.6 mm, Synergi, Phenomenex BV, Utrecht, The Netherlands). The mobile phase at a flow rate of 0.7 ml/min consisted of acetonitrile (eluent A; VWR), and ultrapure HPLC-grade H₂O at a pH of 2 (eluent B; VWR), with the following gradient: 0.0 min, 30% A and 70% B; 70 min, 90% A and 10% B; 71 min, 30% A and 70% B; and 75 min, 30% A and 70% B. The bile salts were detected through DAD at a wavelength of 210 nm. Sample preparation involved centrifugation of samples at 7,690 × g for 5 min before storage at −20°C. Afterwards, 500 μl of supernatant was added to 500 μl of methanol following centrifugation at 7,690 × g for 5 min. Finally, the supernatant was filtered through a PTFE filter (0.2 μm; VWR) prior to injection (50 μl) onto the column. Quantification of samples was performed by using external standards.

For bacterial community analysis, DNA was extracted as previously described by Boon et al. with some minor modifications (Boon et al., 2003). Homogenization was performed using a Beadblaster device (Benchmark Scientific, Edison, NJ, United States), which was conducted twice for 40 s at 6.00 m/s with a cooling period of 5 min between shakings. As an exception, DNA from samples of the final experiment was extracted via the ZymoBIOMICS 96 MagBead DNA Kit (Zymo Research, Irvine, CA, United States) at the KingFisher Flex Purification System (Thermo Fischer Scientific, Waltham, MA, United States) according to the manufacturer’s instructions. Luminal DNA originated from pellets obtained from 1 ml sample, while mucosal DNA was extracted from 0.25 g mucin alginate agar (ileum) or mucin agar (colon).

Targeted microbial quantification was determined through quantitative polymerase chain reaction (qPCR) on a QuantStudio 5 Real-Time PCR system (Applied Biosystems, Forster City, CA, United States). Standard curves were generated from a 10-fold dilution series ranging from 10⁶ gene copies/μl to 10 gene copies/μl. Except for qPCRs targeting Streptococcaceae and Veillonellaceae – which used PCR product –, plasmid DNA was used to generate standard curves. Each sample was run in technical triplicate and outliers were removed when standard deviation between replicates exceeded 0.5. qPCRs for following groups were performed as previously described: Lactobacillus spp. (Furet et al., 2009), Bifidobacterium spp. and Eubacterium rectale/Clostridium coccoides (Rinttilä et al., 2004), Bacteroidetes (Guo et al., 2008), Faecalibacterium prausnitzii (Sokol et al., 2009), Veillonellaceae (Rinttilä et al., 2004), Enterobacteriaceae (Nakano et al., 2003), Streptococcaceae (van den Bogert et al., 2011), Enterococcaceae (Rinttilä et al., 2004), and Akkermansia muciniphila (Collado et al., 2007). Illumina 16S rRNA gene amplicon libraries were generated and sequenced at BaseClear BV. In short, barcoded amplicons from the V3-V4 region of 16S rRNA genes were generated using a 2-step PCR. 10 ng genomic (g)DNA was used as template for the first PCR with a total volume of 50 μl using the 341F (5′-CCTACGGGNGGCWGCAG-3′) and the 785R (5′-GACTACHVGGGTATCTAATCC-3′) primers appended with Illumina adaptor sequences. PCR products were purified, and the sizes of the PCR products were checked on Fragment analyzer (Agilent, Santa Clara, CA, United States) and quantified by fluorometric analysis. Purified PCR products were used for the second PCR in combination with sample-specific barcoded primers (Nextera XT index kit; Illumina, San Diego, CA, United States). Subsequently, PCR products were purified, checked on a Fragment analyzer (Agilent) and quantified, followed by multiplexing, clustering, and sequencing on an Illumina MiSeq with the paired-end (2×) 300 bp protocol and indexing. The sequencing run was analyzed with the Illumina CASAVA pipeline (v1.8.3) with demultiplexing based on sample-specific barcodes. The raw sequencing data was processed by removal of sequence reads of too low quality (only “passing filter” reads were selected) and discard of reads containing adaptor sequences or PhiX control with an in-house filtering protocol. A quality assessment on remaining reads was performed using the FASTQC quality control tool version 0.10.0. For data processing, Illumina-paired reads were merged into single reads (pseudoreads) through sequence overlap, after removal of the forward and reverse primers. Chimeric pseudoreads were removed and remaining reads were aligned to the RDP 16S gene databases. Based on the alignment scores of the pseudoreads, the taxonomic depth of the lineage is based on the identity threshold of the rank; Species 99%, Genus 97%, Family 95%, Order 90%, Class 85%, Phylum 80%.

Total bacterial cells were quantified via flow cytometry to convert proportional data obtained via 16S rRNA gene targeted Illumina sequencing to quantitative data (multiplication). Samples were 1:1-diluted in cryoprotectant and stored at −80°C until analysis according to Hoefman et al. (2012). Upon staining with 0.01 mM SYTO24 (Life Technologies Europe NV, Merelbeke, Belgium) at room temperature for 15 min in the dark, samples were analyzed on a BD Facsverse (BDBiosciences, Merelbeke, Belgium) using the high flowrate setting and bacteria were separated from medium debris and signal noise by applying a threshold level of 200 on the SYTO channel. Flow cytometry data were analyzed using FlowJo, version 10.5.0.

Statistically significant differences between acetate, propionate, butyrate, lactate, and bile salt concentrations were determined for each experiment separately. Significant differences on bacterial abundances were determined on absolute levels for luminal samples and relative levels for mucosal samples for the respective experiments. Statistical analysis was performed in Excel 2011 (Microsoft, Redmond, WA, United States). Two-sided t-tests were applied for comparisons between different vessels during statistical analysis of all experiments. Benjamini-Hochberg false discovery rate (FDR) was applied (with FDR = 0.05) to correct for multiplicity issues as described previously (Lee and Lee, 2018). Principal component analysis (PCA) was performed with CLUSTVIS (biit.cs.ut.ee/clustvis/) using the proportional 16S rRNA gene targeted Illumina sequencing data of the 15 most abundant families in the lumen across all simulated communities supplemented with taxonomic families representing consortium genera but not being part of the top 15 most abundant families.

During an initial screening experiment, four test conditions were evaluated in parallel and compared to a control colon simulation. Reactors 1 and 2 were inoculated with a synthetic consortium, while reactors 3 and 4 were inoculated with ileostomy effluent and faecal sample, respectively. Reactor 2 was distinguished from reactor 1 in that it was subjected to a longer retention time (8 h instead of 4 h). Both the inoculation strategies and retention time affected microbial colonization during the ileum simulations in terms of microbial composition and activity.

First, the validity of a qPCR panel in which 10 taxonomic groups were targeted was evaluated using an ileostomy effluent and faecal sample (Supplementary Table S1). This confirmed the presence of distinct microbial community compositions in both samples, in correspondence with literature, i.e., enrichment of Veillonellaceae, Streptococcaceae, and Enterococcaceae (and Lactobacillaceae depending on the intake of probiotics) in ileostomy samples in contrast to higher levels of Enterobacteriaceae, Akkermansiaceae, Bifidobacteriaceae, Bacteroidetes, Lachnospiraceae, and F. prausnitzii in the faecal sample (Supplementary Table S1), thus confirming that the qPCR panel allowed to make representative statements on whether a microbial community is more ileum or colon-like.

This allowed to conclude that the steady-state communities of the screening experiments upon 2 weeks of colonization were largely affected by the inoculation strategy, with more ileum-like communities being obtained upon inoculation with the synthetic consortium. Indeed, the latter (upon applying a transit time of 4 h) resulted in a community dominated by Veillonellaceae [8.73 ± log(copies/ml)], Streptococcaceae [8.94 log(copies/ml)], Enterococcaceae [9.16 log(copies/ml)] and to a lesser extent, Lactobacillaceae [6.81 log(copies/ml)] (Table 2). Strikingly, incubating the same consortium at increased retention time (8 h) induced dominance by Enterococcaceae [9.52 log(copies/ml)] and Lactobacillaceae [7.53 log(copies/ml)] at the expense of Veillonellaceae [3.50 log(copies/ml)] and Streptococcaceae [3.46 log(copies/ml)]. Further, inoculation with ileostomy effluent or faecal slurry increased abundances of colon-associated taxonomy groups such as Enterobacteriaceae [6.79 and 7.15 log(copies/ml), respectively], Akkermansiaceae [4.45 and 4.38 log(copies/ml), respectively], Bifidobacteriaceae [8.38 and 8.40 log(copies/ml), respectively], Bacteroidetes [7.83 and 7.96 log(copies/ml), respectively], C. coccoides/E. rectale [8.64 and 8.90 log(copies/ml)], and F. prausnitzii [5.91 and 7.41 log(copies/ml)]. The latter more closely resembled the stable community composition in the control colon vessel.

These drastic differences in microbiota composition significantly affected microbial activity. In a first aspect, inoculation with the synthetic consortium significantly lowered total SCFA levels in the reference condition (38.6 ± 1.1 mM) and even further upon increased retention time (6.5 ± 0.2 mM) as compared to the reference conditions inoculated with ileostomy effluent and faecal slurry (64.5 ± 1.1 mM and 61.7 ± 2.3 mM, respectively), while total SCFA levels were the highest in the control colon (69.4 ± 1.9 mM; Figure 2A). In a second aspect, steady-state lactate levels were similarly low among all reference conditions and control colon, which might indicate efficient cross-feeding of lactate to, e.g., propionate and/or butyrate. In contrast, the prolonged retention time significantly increased lactate, thereby suggesting impaired cross-feeding (Figure 2B). In addition, kinetic sampling during 8 h upon receiving fresh nutritional medium on day 14 of the experiment revealed an initial boost in lactate production followed by lactate consumption and a concomitant production of acetate and propionate for the reference consortium condition (Figure 3). Although to a lesser extent, this kinetic profile was also observed for the reference conditions inoculated with ileostomy effluent and faecal slurry. In contrast, entrance of nutritional medium into the condition with extended retention time increased lactate production but lacked the consequential consumption of lactate and concomitant acetate and propionate production (corresponding with the low Veillonellaceae levels upon increased retention time). The control colon showed a modest increase in acetate, propionate, and butyrate concentrations over time while no increase in lactate level was observed.

Furthermore, different conditions affected bile salt metabolism both between the test conditions and as compared to the control colon (Figures 2C–F). The latter showed complete deconjugation of TCA, GCA, TDCA, and GDCA, while said bile salts were only partially deconjugated in the ileal vessels. Lowest bile salt conversions were observed upon inoculation with the synthetic consortium, and more specific, in combination with an increased retention time.

Overall, from all tested conditions, the synthetic consortium inoculation under reference conditions was selected as most closely representing the in vivo ileal community (Booijink et al., 2010; Zoetendal et al., 2012; El Aidy et al., 2015; Cieplak et al., 2018; Seekatz et al., 2019).

Next, the selected reference condition from the screening experiment was performed in triplicate to validate reproducibility of the simulation.

Luminal SCFA profiles were stable over time (87, 86, and 88% for replicates 1, 2, and 3, respectively) and highly reproducible (95%) across the multiple replicate vessels of the selected condition (Figure 4A). Lactic acid levels were 1.40 ± 0.69 mM on average and did not differ significantly between replicate simulations (data not shown). 16S rRNA gene targeted Illumina sequencing demonstrated a nearly identical community composition within the replicate vessels at different timepoints, thereby confirming both stability and reproducibility of the bacterial community (Figure 4B). Luminal communities were dominated by the genera Streptococcus (41.2 ± 2.5%), Veillonella (30.7 ± 2.0%), Enterococcus (11.2 ± 2.7%), and Clostridium (8.4 ± 1.3%). The genera Lactobacillus (2.6 ± 0.8%), Blautia (1.6 ± 0.4%), Faecalibacterium (0.7 ± 0.2%), and Prevotella (0.1 ± 0.1%) were present at lower abundances.

Total cell concentrations were estimated from qPCR measurements based on a post-hoc comparison between flow cytometry and qPCR analyses. Based on this correlation, qPCR analysis during the screening experiment pointed out that levels above 10⁹ cells/ml in the ileum simulations were obtained, which was considered too high versus the in vivo situation where levels in the range of 10⁸ cells/ml have been reported (Booijink et al., 2007). Therefore, with the aim of reducing the bacterial load in the ileum simulation, adapted ileal nutritional medium (30%) was compared to the initial ileal nutritional medium (100%) in two identically configured vessels. Importantly, bacterial cell quantification through flow cytometry demonstrated an approximately 3-fold reduction in total bacterial density upon administration of the adapted ileal nutritional medium (2.53 and 7.58 * 10⁸ cells/ml at the beginning and end of the feeding cycle, respectively) as compared to the initial ileal nutritional medium (6.65 and 19.95 * 10⁸ cells/ml at beginning and end of the feeding cycle, respectively; Figure 5D).

This decreased cell density did not alter microbial activity or community composition. Indeed, while the adapted medium (30%) significantly (p = 2 * 10⁻⁸) reduced total SCFA concentration, it did not alter the acetate/propionate/butyrate ratio (approximately 55/45/0; Figure 5A). Lactate was significantly (p < 10⁻⁵) higher – but still minimal – in the adapted medium (0.025 ± 0.003 mM) than the initial medium (0.007 ± 0.002 mM). 16S rRNA gene targeted Illumina sequencing revealed a similar microbial composition in both the luminal and mucosal regions between both nutritional media (Figures 5B,C). Lowered nutrient levels resulted in a mild proportional increase of Clostridium (12.5% vs. 6.8%) and Enterococcus (16.7% vs. 8.7%) at the expense of Streptococcus (46.7% vs. 55.6%) and Veillonella (21.8% vs. 26.7%). Hence, where streptococci and Veillonellae accounted for more than 80% of the total community in the vessel receiving the 100% nutritional medium, these taxonomic groups represented approximately 65% of the community in the vessel receiving the adapted ileal nutritional medium (30%). Furthermore, Lactobacillus (0.2% vs. 0.2%), Blautia (0.1% vs. 0.1%), Faecalibacterium (<0.1% vs. <0.1%) and Prevotella (<0.1% vs. <0.1%) were not impacted. Together, genera comprised in the synthetic consortium accounted for more than 98% of the simulated community (98.0 and 98.2% for adjusted and initial nutritional medium, respectively). The missing approximately 2% relative abundance is represented by Lactococcus (1.2% vs. 1.3% for adjusted and initial nutritional medium, respectively), Hungatella (0.20% vs. 0.15%, for adjusted and initial nutritional medium, respectively) and a myriad of low-abundant genera (<0.1%; data not shown).

During the final experiment, the ileal simulation was integrated into the established M-SHIME® model which simulates colonic microbiota. For this, two arms for each of two donors were run in parallel. Per donor, one arm comprised of an ileum, proximal colon, and distal colon, whereas the second arm consisted of a proximal and distal colon only.

As illustrated by a PCA, based on the 15 most abundant families across the lumen of all reactors combined with families to which the genera of the ileal consortium belong but are not within the top 15 most abundant families, colon-region specific communities colonize the M-SHIME® model (Figure 6). Integration of an ileum simulation in the M-SHIME® maintained this colon-region specificity, thereby indicating that a preceding ileum does not disturb subsequent simulated colonic communities. Moreover, inclusion of the ileum microbiota contributed to an increased diversity (Figure 7) as well as an increased species richness (data not shown) in the in vitro gut model for both donors, reflecting more what has been reported happing in vivo. Furthermore, it was observed that distal colons more closely resembled the faecal inocula when preceded by an ileum as compared to distal colons without preceding ileum (Figure 6).

Although integration of an ileum reactor increased initial lactate levels in the proximal (0.84 ± 0.14 mM to 2.20 ± 0.06 mM) and distal (0.22 ± 0.04 mM to 0.32 ± 0.01 mM) colons, lactate was more rapidly consumed in ileum-preceded colon reactors. SCFA analysis confirmed both ileum simulations to show nearly identical SCFA profiles (Figure 8). The total SCFA concentrations were 13.5 ± 0.7 mM and 13.9 ± 0.5 mM, and the acetate/propionate/butyrate ratios were 50/49/1 and 54/45/2 in the ileum vessels of reactor units 1 and 2, respectively. Incorporation of an ileum vessel did not impact the total SCFA concentration in proximal (54.8 ± 3.8 mM vs. 53.1 ± 6.3 mM with or without preceding ileum, respectively) or distal colon vessels (73.0 ± 4.2 mM vs. 72.5 ± 2.1 mM with and without preceding ileum, respectively). However, preceding ileum simulations impacted the proportional SCFA profiles in the subsequent colon compartments. For donor 1, the preceding ileum increased butyrate concentration in both colon vessels at the expense of acetate and propionate (SCFA ratios changed from 58/22/21 to 51/17/32 for the PC, and from 67/21/12 to 62/18/19 for the DC). Nonetheless, for donor 2, SCFA ratios remained constant (from 52/14/34 to 54/11/35 for the PC, and from 63/17/20 to 64/16/20 for the DC).

Regarding microbial composition, steady-state communities in the lumen and mucus of both ileum simulations consisted nearly entirely of genera included in the synthetic consortium (98.8 ± 0.4% and 99.3 ± 0.2%, and 97.3 ± 0.3% and 98.8 ± 0.2% for the lumen and mucus of ileum vessel of unit 1 and 2, respectively), indicating absence of colonization by contaminating strains (Figures 9A–C). In the ileum vessels of reactor units 1 and 2, stable luminal communities were dominated by the genera Veillonella (30.6 ± 4.6% and 24.3 ± 6.8%, respectively), Streptococcus (13.0 ± 5.0% and 27.8 ± 7.0%, respectively), and Enterococcus (55.1 ± 1.6% and 38.5 ± 1.1%, respectively).

Integration of an ileum simulation into the M-SHIME® technology did not affect bacterial densities in subsequent proximal and distal colon reactors. Moreover, incorporation of the ileum simulation did not majorly affect the luminal bacterial community composition in neither the proximal nor distal colon (Supplementary Table S1). Minor changes included, for donor 1, a reduction in Bacteroidaceae (PC), Selenomonadaceae (PC), Synergistaceae (DC), and Akkermansiaceae (DC), while the levels of Enterococcaceae (PC and DC) and Streptococcaceae (PC and DC) were increased. Although not significantly, Ruminococcaceae levels were increased in the PC upon incorporation of an ileum simulation. For donor 2, Streptococcaceae and Enterococcaceae were increased in both the PC and DC, while the Ruminococcaceae were significantly increased in the PC and significantly decreased in the DC. Changes in mucosal community compositions were less pronounced.