Preservation of conjugated primary bile acids by oxygenation of the small intestinal microbiota in vitro

ABSTRACT Bile acids play a critical role in the emulsification of dietary lipids, a critical step in the primary function of the small intestine, which is the digestion and absorption of food. Primary bile acids delivered into the small intestine are conjugated to enhance functionality, in part, by increasing aqueous solubility and preventing passive diffusion of bile acids out of the gut lumen. Bile acid function can be disrupted by the gut microbiota via the deconjugation of primary bile acids by bile salt hydrolases (BSHs), leading to their conversion into secondary bile acids through the expression of bacterial bile acid-inducible genes, a process often observed in malabsorption due to small intestinal bacterial overgrowth. By modeling the small intestinal microbiota in vitro using human small intestinal ileostomy effluent as the inocula, we show here that the infusion of physiologically relevant levels of oxygen, normally found in the proximal small intestine, reduced deconjugation of primary bile acids, in part, through the expansion of bacterial taxa known to have a low abundance of BSHs. Further recapitulating the small intestinal bile acid composition of the small intestine, limited conversion of primary into secondary bile acids was observed. Remarkably, these effects were preserved among four separate communities, each inoculated with a different small intestinal microbiota, despite a high degree of taxonomic variability under both anoxic and aerobic conditions. In total, these results provide evidence for a previously unrecognized role that the oxygenated environment of the small intestine plays in the maintenance of normal digestive physiology. IMPORTANCE Conjugated primary bile acids are produced by the liver and exist at high concentrations in the proximal small intestine, where they are critical for proper digestion. Deconjugation of these bile acids with subsequent transformation via dehydroxylation into secondary bile acids is regulated by the colonic gut microbiota and reduces their digestive function. Using an in vitro platform modeling the small intestinal microbiota, we analyzed the ability of this community to transform primary bile acids and studied the effect of physiological levels of oxygen normally found in the proximal small intestine (5%) on this metabolic process. We found that oxygenation of the small intestinal microbiota inhibited the deconjugation of primary bile acids in vitro. These findings suggest that luminal oxygen levels normally found in the small intestine may maintain the optimal role of bile acids in the digestive process by regulating bile acid conversion by the gut microbiota.

synthesized and conjugated to either taurine or glycine in the liver, secreted into the duodenum, and almost completely (~95%) reabsorbed in the terminal ileum (3)(4)(5)(6).It has been previously found that conjugation of 1° bile acids serves two purposes, increasing aqueous solubility and preventing passive diffusion of the bile acids out of the gut lumen (5,7,8).In addition, conjugation has been found to increase the rate of lipolysis by bile acids, which indicates that deconjugation impacts lipid metabolism (9).Therefore, the presence of conjugated, 1° bile acids is required for sufficient digestion to occur in the small intestine.Indeed, the disruption of bile acid synthesis or function can lead to nutritional malabsorption (10).
Bile acids that are not reabsorbed in the terminal ileum (~5%) and enter the colon are rapidly deconjugated and converted into secondary (2°) bile acids through actions of the resident gut microbiota that have genes for bile salt hydrolases (BSHs), which perform deconjugation as well as bile acid-inducible (BAI) genes that convert 1° into 2° bile acids (6,7,11,12).Although microbial metabolism of bile acids via BSH and BAI genes is a robust functionality associated with the colonic microbiota, these enzymatic reactions are much less characteristic of the small intestinal microbiota (13)(14)(15).Up until now, this had been attributed to the distinct taxonomic differences in community structure and resulting function of the small intestinal versus colonic microbiota, which are largely shaped by their physiological environments (13,15).Yet, due to the complications in harvesting samples from the small intestines of subjects in vivo and the high degree of reported inter-subject variability of the microbial community, only circumstantial evidence is available to support the supposition that the lack of bile acid metabolism by the small intestinal gut microbiota is primarily due to community composition (14,(16)(17)(18)(19).In general, the role of the physiological environment on microbial metabolism has not been considered.Specifically relevant to the gut microbiota in the proximal small intestine is the high physiological level of oxygen (O 2 ) that has been measured to be 40 Torr, which shapes the community and favors the growth of facultative anaerobes (13,20).
Taking advantage of the flexibility and control only attained through the application of an in vitro platform, we developed a model of the small intestinal gut microbiota and used it to study inter-subject variability, in terms of structure, function, and nutrient utilization, and how this variability impacts the community's response to physiological levels of O 2 .Importantly, we were able to analyze the role of both community structure and physiological levels of O 2 on the deconjugation and conversion of bile acids.Our results provide evidence that both of these both factors contribute to preserving bile acid function in the small intestine, highlighting the importance of the physiological environment in shaping the functionality of the gut microbiota.

Recapitulating in vivo reports, structure of the in vitro small intestinal communities, and their response to O 2 , displayed strong inter-subject variability
We began by developing an in vitro model of the small intestinal microbiota comprising four computer-controlled cultivars programmed to mimic the physiological conditions of the small intestine.Ileostomy effluent harvested from four different subjects was used as the inocula representing the small intestinal gut microbiota.Each ileostomy community harbored a unique taxonomic composition, recapitulating the inter-subject diversity of the gut microbiota previously reported from in vivo studies (Fig. S1) (14,16).Within the cultivars, mature communities were first developed under anaerobic conditions (anoxic conditions) and then 5% O 2 was infused at a constant rate to mimic physiological levels of the small intestinal environment (oxic conditions) (21,22).Samples harvested during both anoxic and oxic conditions were subjected to 16S rRNA gene sequencing to determine community structure.All sequencing results are publicly available on NCBI (PRJNA1017510).
Following inoculation of the in vitro model, robust microbial communities developed, which were substantially higher in density compared to their respective inoculums.Once maximal community density was attained for each unit, the total bacterial load remained static until the termination of the experiment and was unresponsive to the infusion of 5% O 2 (Fig. 1A).Despite the inter-subject variability of the inoculums, alpha diversity, evaluated in terms of richness and evenness (Shannon's index), decreased in all units following cultivation, except for richness for Unit 1, which remained at a level similar to that of the inoculum (Fig. 1B).Infusion of 5% O 2 did not produce any significant changes to richness but did significantly lower Shannon's diversity in Unit 1 (q = 0.0004677) and increased this metric for Unit 3 (q = 0.0122) (Fig. 1C).This indicated that the composition of the respective inocula was the driving factor in the community's response to the addition of O 2 .
Supporting the conclusion that infusion of 5% O 2 did not have a unilateral impact on community diversity but only produced minimal, unitspecific effects, the results of principal coordinate analysis (PCoA) of unweighted UniFrac distance portrayed a level of high variability across all units, in which the anoxic and oxic-exposed communities did not group separately from each other (Fig. 2A).It should be noted here that the PCoA figures were generated with all units combined and then separated by unit.Therefore, the fact that the communities would not overlap in the graph further indicated their divergence from each other.By contrast, visualization by PCoA for weighted UniFrac distances clearly illustrated a strong divergence in community structure driven by the infusion of O 2 (q = 0.027, PERMANOVA with FDR correction), where the anoxic and oxic-exposed communities were grouped separately from each other for two of the four units (Units 1 and 2) (Fig. 2B).This indicates that infusion of 5% O 2 did elicit restructuring of the community in terms of relative abundance for some of the units (Fig. 2B) but not a gain or loss of taxa (Fig. 2A) for all units.The response to O 2 in Fig. 2B was personalized, suggesting that these alterations were dependent upon the original structure of the community.
To provide granularity on the community response to O 2 , 16S rRNA gene amplicon sequencing was used to classify taxa as obligate or facultative anaerobes, revealing that O 2 infusion corresponded with a significant enhancement in the ratio of facultative anaerobes in Units 1-3 and a decrease to this category in Unit 4 (Fig. 3A) (23).The incongruency of response between Units 1-3 and Unit 4 was most likely driven by inter-subject differences in the starting communities, as the ileostomy effluents used FIG 1 Alpha diversity for communities in terms of (A) bacterial load, (B) number of ASVs, and (C) Shannon's index.Statistical significance between anoxic and oxic conditions was determined using a Welch two-sided t-test with FDR correction and significant differences (q < 0.05) are indicated with an asterisk.
to inoculate Units 1-3 were predominantly obligate anaerobes, whereas the ileostomy sample used to inoculate Unit 4 was dominated by facultative anaerobes.Specifically, the composition of the ileostomy samples was highly variable, but for Units 1-3, all were predominantly Firmicutes (91.2%, 74.5%, and 71.0%, respectively), followed by Proteobacteria or Actinobacteria, and some representation of Bacteroidetes (Fig. S1).Conversely, the ileostomy sample used to inoculate Unit 4 was equally Firmicutes and Proteobacteria (49.6% and 49.6%, respectively), with a large representation of Enterobac teriaceae and Klebsiella (Fig. S1).
When cultured under anoxic conditions, Firmicutes was dominant in all four units, although the prevalent genera within Firmicutes were highly unit dependent (Fig. 3B).Representatives of Bacteroidetes (Bacteroides and Parabacteroides) were only found in Unit 2, which also had the highest levels of Bifidobacterium following cultivation.Regardless of inter-subject differences, infusion of 5% O 2 led to major structural remodeling of all the communities.Most notably, there was a substantial increase in Proteobacteria for Units 1-3.In Units 1 and 2, 5% O 2 drove a significant increase in levels of Escherichia/Shigella, from 0.7% and 3.8% relative abundance under anoxic conditions up to 50.9% and 41.9%, respectively, in oxic conditions.Such a high increase in the abundance of a single taxon was most likely responsible for the large diver gence between the anoxic and oxic communities for Units 1 and 2 observed in the weighted PCoA analysis (Fig. 2B).For Unit 3, there was a large percentage of Klebsiella pneumonia already present under anoxic conditions, which further increased in the oxic environment, although this did not achieve statistical significance.Yet, this increase in Proteobacteria was not observed for Unit 4, most likely due to the absence of Proteobac teria during cultivation under anoxic conditions and the dominance of the community by Enterococcus when inoculated into the cultivar.The lack of substantial increase to Proteobacteria would explain why divergence between communities was less apparent for Units 3 and 4 in the weighted PCoA analysis (Fig. 2B).

FIG 3
Community structure was altered due to perturbation by oxygen causing a shift in (A) the ratio of obligate to facultative anaerobes, * indicates a statistically significant difference (Student's t-test) between anoxic and oxic conditions for obligate and facultative anaerobes (p < 0.05), and (B) relative abundance of the dominant taxa.Only taxa present at >1% abundance in at least one sample are depicted.Statistically significant changes between anoxic and oxic conditions were determined using a linear model in MaAsLin2, and differences with q < 0.05 are underlined, boldfaced, and italicized.
There were no taxa identified that responded significantly between the anoxic and oxic conditions in all four units, most likely due to the compositional differences between communities (Fig. 3B).However, when switched to oxic conditions Clostridium sensu stricto 1 decreased significantly for Units 2, 3, and 4, and while this also decreased in Unit 1, it failed to achieve statistical significance.Ruminococcus gnavus group decreased significantly for Units 1 and 2, decreased in Unit 3, although not significantly, and was not detected in Unit 4. Escherichia/Shigella statistically increased for Units 1, 2, and 3 but was not detected in Unit 4. While it is impossible to decipher exactly why these specific taxa increased or decreased in this study, it is possibly due to their sensitivity to oxygen.It is known that both Clostridium sensu stricto 1 and Ruminococcus gnavus are strict anaerobes, and these decreased in response to O 2 , whereas Escherichia/Shi gella is a facultative anaerobe and increased in response to O 2 (24)(25)(26).It is also possible that complex, synergistic interactions occurred between taxa in this community.Together, the results of 16S rRNA gene amplicon sequencing analysis demonstrated large inter-subject variability in both alpha and beta diversity, which extended into the response of these communities to O 2.

Levels of short chain fatty acid production, amino acid consumption, and response to O 2 varied depending on the initial community
The results of 16S rRNA gene sequencing found that there was large inter-subject variability between the initial community compositions and the response of these communities to infusion with 5% O 2 .Whether or not this variation would be observed in the functional output and nutritional utilization was assessed by the quantification of short chain fatty acids (SCFAs) (Table S1), which are the end products of microbial fermentation, and levels of free amino acids (AAs) (Table S2), which can be released through protein degradation or used by the community as a source of nitrogen/carbon (27,28).Although the same source of nutrients was provided, the levels of total SCFAs were variable between units, ranging from 43.5 to 19.7 mmol/L (Fig. 4A).Similarly, the levels of branched-chain fatty acids (BCFAs), which are the end products of amino acid or protein fermentation, were variable between units, ranging from 2.3 mmol/L down to 0 (Fig. 4A).Perturbation with oxygen significantly reduced total SCFAs in Unit 1, with no statistical effect on Units 2, 3, or 4. For total BCFAs, although the addition of 5% O 2 reduced the levels for all units, this change was only significant for Unit 4.
The most prominent SCFA detected in all communities was acetic acid followed by either propanoic acid or butanoic acid, and no pentanoic acid was present (Fig. 4B).Although, the levels of all three of these SCFAs were variable between communities, a significant reduction in response to 5% O 2 for all three was observed only in Unit 1 (Fig. 4B).Since the production of SCFAs is a result of anaerobic fermentation, it is notable that Unit 1 had the greatest reduction in obligate anaerobes with a proportional increase in facultative anaerobes (Fig. 3).Additionally, Unit 2 had the highest levels of propanoic acid, which corresponded with a high abundance of Bacteroides in the community.As Bacteroides are known to produce propanoic acid, this represents a potential link between structure and function observed only in Unit 2 (29).
Similar to the results of SCFA analysis, when free AA levels were quantified as a representation of nutrient utilization, the profiles generated under anoxic and oxic conditions depicted large inter-subject differences between communities (Fig. 5).Whereas Unit 1 presented with low levels of almost all AAs, except for taurine and alanine, Unit 4 had robust levels of multiple AAs measured, regardless of the addition of 5% O 2 .The high level of AA utilization in Unit 1 was most likely the reason that Unit 1 had the highest levels of BCFAs under anoxic and oxic conditions (Fig. 4A).In total, the results of both SCFA production and AA metabolism by these communities showed that inter-subject variability dominated over the effects of oxygenation.

Regardless of inter-subject variability, there was a lack of bile acid conversion observed with O 2 which inhibited deconjugation events
Bile acids were provided to the community during feeding every 8 h as a mixture of primary (1°) and secondary (2°) species at an approximate 3:1 ratio, and the levels within the communities were quantified to determine the extent of deconjugation and dehydroxylation that occurred (Table S3).The ratio of 1° to 2° bile acids remained similar to the input for all four units, regardless of anoxic or oxic conditions; how ever, there was some inter-subject variability observed (Fig. 6A).This lack of microbial mediated conversion of bile acids from their 1° to 2° forms was opposite of what has been previously reported for the colon gut microbiota cultured in vitro, where almost complete conversion of bile acids occurred by the end of the fermentation cycle (30).These results demonstrate that our in vitro model recapitulated the in vivo situation in which bile acids are preserved in their active form (1°) within the small intestine and only undergo conversion to their 2° forms when they reach the colon, under normal, biological circumstances.

FIG 4
Function output as determined through quantification of (A) total SCFAs and total BCFAs, and (B) acetic, propanoic, and butanoic acids each unit under anoxic and oxic conditions.Statistical significance between anoxic and oxic conditions was determined using a Welch's two-sided t-test and a q value < 0.05 is indicated with an asterisk.
While there were no statistically significant differences in the amounts of total 1° bile acids for any of the communities (unconjugated and conjugated forms combined), infusion of 5% O 2 significantly increased the levels of conjugated 1° bile acids for all communities except Unit 1, in which levels were higher, but this did not reach statistical significance (Fig. 6B).Conversely, under anoxic conditions, there were statistically higher levels of unconjugated 1° bile acid.
Looking at bile acids with higher resolution, we quantified the levels of the 1° bile acids, cholic acid (CA) and chenodeoxycholic acid (CDCA) and their conjugated forms.For CA, we saw increased levels of conjugated forms in the presence of 5% O 2 , which were FIG 5 Nutrient utilization was determined through quantification of free amino acid levels for each unit under anoxic and oxic conditions.Statistical significance between anoxic and oxic conditions was determined using a Welch's two-sided t-test and a q value < 0.05 is indicated with an asterisk.statistically reduced under anoxic conditions, where a statistical increase in deconjuga ted forms was observed (Fig. 6C).This pattern was also observed for CDCA in Units 3 and 4, where the conjugated forms were statistically increased during infusion with O 2 and reduced under anoxic conditions (Fig. 6C).For Units 1 and 2, the levels of CDCA and its conjugated forms were too low to determine significance.
In total, these results showed that the infusion of O 2 into the small intestinal communities, designed to mimic the oxygenated environment of the proximal small intestine (22), which is responsible for most nutrient absorption, consistently reduced the deconjugation of 1° bile acids without any conversion into 2° bile acids-alterations known to enhance the small intestinal digestive process.Unlike SCFA production and AA metabolism, the effects of oxygen on bile acid composition exceeded the high levels of inter-subject variability in both ileostomy inoculums and overall composition of each cultivar unit at steady state.

DISCUSSION
Studies on the small intestinal gut microbiota continue to uncover the importance of this community to the digestive process and human health (1,31).Yet, progress in this area has been slow, primarily due to technical challenges in obtaining samples from the small intestinal GIT and the large inter-subject variability reported.This has left a large gap in knowledge on the small intestinal microbiota in terms of its structure, function, nutrient utilization, and how these factors are shaped by the physiological FIG 6 Bile acid quantification in terms of (A) the percentage of primary (1°) and secondary (2°) bile species, and; levels of (B) total conjugated and unconjugated 1° bile acid species; (C) cholic acid, conjugated and unconjugated, and; D) chenodeoxycholic acid, conjugated and unconjugated.GCA, glycocholic acid; TCA, taurocholic acid; CA, cholic acid; GCDA, glycochenodeoxycholic acid; TCDA, taurochenodeoxycholic acid, and; CDA, chenodeoxycholic acid.Statistical significance between anoxic and oxic conditions was determined using a Welch's two-sided t-test with FDR correction, and a q value < 0.05 is indicated with an asterisk.
environment, such as the presence of O 2 .Even less clearly defined was the ability of these microbes, or lack thereof, to metabolize bile acids in the small intestine, which are critical to digestive physiology and serve as signaling molecules with an array of biological activities (22,32).In the present study, an in vitro model was applied to cultivate the small intestinal microbiota of four different subjects, and the results of 16S rRNA gene amplicon sequencing and metabolomics were combined to address these gaps in knowledge.
The results of 16S rRNA gene amplicon sequencing analysis clearly showed that the in vitro small intestinal communities maintained the high levels of inter-subject variability of their respective inoculums, similar to what has been reported in vivo (19,(33)(34)(35)(36)(37).Furthermore, we found that the inter-subject variability in community structure between units extended to functional output and nutrient utilization, demonstrated here by variable results between units in total quantification of SCFAs and AAs.In addition, the response of each community to infusion with 5% O 2 was highly dynamic, both in terms of structure and function.While the observed variability reported here was expected based on existing knowledge of the small intestinal dynamics in vivo, the novelty and significance of our results lie within the observed consistencies that eclipsed the effect of inter-subject variability.
Upon cultivation, we found a consistent increase in biomass for all communities developed, which occurred to the same extent both with and without the infusion of 5% O 2 .Bacterial density in vitro was higher than that of the inocula used, and even higher than the estimated biomass of the proximal or distal small intestine, or what may occur during small intestinal bacterial overgrowth (13,38).The observed increase in biomass here was an artifact of in vitro cultivation, as the in vitro model lacks a number of factors that would contribute to the maintenance of low density in vivo, such as the production of bacteriocins or release of other secreted immune peptides (39)(40)(41).While density increased unilaterally following in vitro cultivation, there was a concurrent and consistent decrease in evenness, as evaluated using the Shannon index.Unlike density, for evenness, the extent of decrease was influenced by the infusion of O 2 for two of the four communities tested.This demonstrated that, while the decrease in evenness was consistent, the magnitude of change in evenness elicited by O 2 depended on community composition.
In terms of beta-diversity, there was large variability in the inoculums, the commun ities that developed in vitro, and the response of these communities to infusion with O 2 .This is evidenced in the PCoA plot depicting unweighted UniFrac distances and in the heatmap showing taxonomic structure.Despite this variability, infusion with O 2 elicited a significant shift in structure for all four units, which is clearly shown in the PCoA plot depicting weighted UniFrac distances.In this case, the addition of O 2 was not affecting the presence or absence of taxa, but the ratio in which they were coexisting.These results are similar to previous findings, which increased luminal oxygenation in the gut altered the structure of the microbiota, favoring expansion of aerobic or facultative taxa (21,42,43).Here, we observed that alterations occured due to O 2 infusion and an increase in facultative taxa in three of four communities, particularly from Escherichia/Shi gella, a member of phylum Proteobacteria.It was noted that a few obligate anaerobes did statistically increase in abundance in the presence of O 2 .Specifically, Clostridium sensu stricto 13 and Flavinofractor in Unit 1 and Parabacteroides and Erysipelatoclostridium in Unit 2. This indicated that complex, synergistic interactions may have occurred in which O 2 utilization by the facultative anaerobes allowed the obligate anaerobes to expand, similar to what has been previously reported (44).
Function of the communities tested here was markedly variable in SCFA production and AA utilization, as well as the effect of O 2 on these metrics.However, phenocopying normal mammalian physiology, there was a consistent lack of 1° to 2° bile acid conver sion observed here, regardless of O 2 infusion.Bile acids, which are produced by the liver and play a role in the emulsification, solubilization, and absorption of dietary lipids, are modified by the gut microbiota through deconjugation via BSHs and dehydroxylation via genes in the BAI operon (8,(45)(46)(47).Our observation that the small intestinal communi ties did not convert 1° bile acids aligns with previous reports that have found the small intestine to be dominated by 1° bile acid species and the colon by 2° bile acid species, further supporting the divergent functional roles of the small intestinal versus colonic microbiota (19,48).
An unanticipated finding of this study was a consistent correlation between infusion with O 2 and conjugation of 1° bile salts.Under anaerobic conditions, we found a significant increase in deconjugated bile acids, indicating that infusion of O 2 inhibited deconjugation events.Concurrently, we found a consistent increase in the levels of taurine, most likely the result of deconjugation of taurine-conjugated bile acids.Our results align with previous findings that the levels of conjugated bile acids were lower in the colon compared to the small intestine (3,19).Since deconjugation of 1° bile acids is a required step before dehydroxylation and conversion to the 2° form, preventing deconjugation, in effect, stalled conversion (49).Our observation here indicates that the physiological environment of the small intestine, specifically the presence of low levels of O 2 , mediates microbial bile acid metabolism by preventing deconjugation.Considering the corresponding change in structure observed, it can be considered that the lack of deconjugation was most likely due to the outgrowth of taxa from the phylum Proteobacteria upon O 2 infusion in three of four units, as these are not known to carry a large repertoire of BSH genes (12).Under anoxic conditions, which are more similar to the physiology of the colon, we saw that deconjugation of bile acids occurs at a higher rate, which would then allow for dehydroxylation to occur via the BAI operon and conversion to take place.
Given the significant impact of the small intestinal microbiota on the host, as well as the significant difficulties associated with obtaining human small intestinal samples, in vitro models can offer unique insight into this community's response to external stimuli, such as dietary compounds and xenobiotics.Here, we applied an in vitro model to study the small intestinal gut microbiota, analyzing both community structure and function and its response to the physiological parameter O 2 .Taken together, our results provide new insights into the dynamics of bacterial physiology in response to chang ing environmental conditions relevant to the mammalian intestine within the context of distinct complex communities.Further advancing our understanding of bile acid function, we found that despite high levels of inter-subject variability, there was an observed lack of bile acid conversion from 1° to 2° forms, which was not affected by the presence of O 2 .However, physiological concentrations of O 2 , normally found in the proximal small intestine, reduced the deconjugation of primary bile acids, which would preserve their functionality in the digestive process.These results provided evidence that the community structure of the small intestinal microbiota is shaped by the physiological environment and the presence of O 2 , which together inhibit bile acid metabolism.This is a novel finding that is important for bile acid functionality and is relevant to digestive physiology.

Human subjects
The sample small intestinal cultivars were inoculated using a baseline ileal effluent collected from healthy volunteers with either a diverting or an end ileostomy for an indication other than Crohn's disease and enrolled in a crossover feeding study titled "Enteral Nutritional Therapy and the Gut Microbiota and their Metabolites in Ileostomy Contents" (EMMI, IRB 825368).

Cultivar operation
Bioflow 320 cultivars were purchased from Eppendorf (Hamburg, Germany) and assembled according to the manufacturer's specifications.During the experiment, all cultivars were maintained at a temperature of 37°C with agitation set to 100 rpm.The pH was kept at 7.4 ± 0.1, using 1 M NaOH and CO 2 gas.There was a constant sparging of gas at a rate of 1 L/min, with the ratio of N 2 :O 2 :CO 2 dependent on the experimental phase.The initial volume of the cultivars was 1 L of a 70:30 ratio of defined medium: pancre atic juice (30,50).Defined medium was Adult M-SHIME growth medium with starch purchased from ProDigest (Ghent, Belgium), which contains the following ingredients in g*L -1 : arabinogalactan (1.2), pectin (2.0), xylan (0.5), resistant starch (4.0), glucose (0.4), yeast extract (3.0), peptone (1.0), mucin (3.0), and cysteine (0.5) (30,50).Pancreatic juice was made fresh every 2-3 days and contained 12.5 g*L -1 NaHCO 3 (Sigma-Aldrich, St. Louis, MO, USA), 6 g*L -1 oxgall bile (Becton-Dickinson, Franklin Lakes, NJ, USA), and 0.9 g*L -1 pancreatin (Sigma-Aldrich, St. Louis, MO, USA) (30).The cultivars were connected using silicon tubing (Cole Parmer, Vernon Hills, IL, USA) to a supply of defined medium and pancreatic juice to provide nutrition and biliary-pancreatic enzymes and to a urinary drainage bag (Becton, Dickinson and Company, Franklin Lakes, NJ, USA) to collect waste (30,50).Ileostomy samples (1 g) were homogenized at 10% (wt/vol) in phosphate buffer within an anaerobic chamber, and 10 mL of the resuspended inoculums was added to each bioreactor.The cultivars were grown overnight, 16 h, with temperature, pH, agitation, and gas flow.Following overnight growth, the cultivars were maintained with daily feeding cycles in which the cultivar was provided fresh medium and pancreatic juice.Every 8 h, the volume of the cultivar was reduced to 800 mL, and 200 mL of a 70:30 defined medium: pancreatic juice mixture was added (51,52).

Experimental design
The experiment consisted of four BioFlo 320 cultivars (Eppendorf, Hamburg, Germany), each inoculated with an individual sample from a unique human subject (approximately 1 g of frozen ileostomy effluent).Following inoculation, communities were maintained for 14 days under anaerobic conditions to develop a mature community (50,53).During anaerobic culturing, sparging was performed using 100% N 2 , supplemented with CO 2 as needed to maintain pH.On day 15, the cultivars were switched to microaerobic (5 Torr O 2 ) conditions and maintained for an additional 14 days (22,54).During microaerobic growth, sparging was performed using 5% O 2 (40 Torr) and 95% N 2 .The N 2 gas was supplemented with CO 2 to maintain pH.Samples were collected daily in the afternoon, at the end of the feeding cycle (approximately 7.5 h after feeding).Following inoculation, samples were harvested representing biological replicates on days 10, 13, and 14 to represent the anoxic conditions, while samples representing the oxic conditions were harvested on days 24, 27, and 28.Samples were used for 16S rRNA amplicon sequencing, qPCR, SCFA analysis, bile salt quantification, and amino acid detection.Data generated from the biological replicates under either anoxic or oxic conditions were averaged together and presented as the average with standard deviation.

16S rRNA sequencing
DNA was extracted using the QIAcube extraction machine (Qiagen, Hilden, Germany) using a DNeasy 96 PowerSoil Pro QIAcube HT kit (Qiagen, MD, USA) following the manufacturer's protocol.Briefly, 1.5 mL samples taken directly from the cultivars were centrifuged at 4°C for 10 minutes at 5,000 × g, and the supernatant was removed from the pellet.The resulting bacterial pellet was frozen at −80°C until use.Pellets were homogenized using the Bead Ruptor Elite (Omni International, GA, USA), and samples were extracted using the QIAcube HT (Qiagen, MD, USA).Extracted DNA was subjected to 16S rRNA gene sequencing targeting the V3-V4 region using 2 × 300 bp chemistry on an Illumina Miseq following the manufacturer's guidelines (31,55).Briefly, primers were ordered from Integrated DNA Technologies (Coralville, IA, USA), 16S Amplicon PCR Forward Primer 5′-TCGTCGGCAGCGTCAGATGTGTATAAGAGAC AG CCTACGGGNGGCWGCAG-3′, 16S Amplicon PCR Reverse Primer 5′-GTCTCGTGG GCTC GGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC-3′, and 2× KAPA HiFi HotStart Ready Mix (Fisher Scientific, Hampton, NH, USA) were used to amplify variable regions.PCR product clean-up was performed using AMP XP beads (VWR International, Radnor, PA, USA).NexteraXT Indexes (Illumina, San Diego, CA, USA) were applied to generate indexed libraries, which were denatured and mixed with 25% PhiX (Illumina) as an internal control and 8 pM loaded on a V3 reagent cartridge (Illumina).

Liquid chromatography/mass spectrometry
Reactor samples for metabolomics were centrifuged at 13,000 × g for 5 minutes at the time of collection, and the supernatant was stored at −80°C until analysis.Bile acids were quantified as previously described (48).Briefly, samples were thawed, vortexed to homogenize, and centrifuged twice at 13,000 × g for 5 minutes.The supernatant was analyzed on a Waters Acquity uPLC System with a Cortecs UPLC C-18+ 1.6 µm 2.1 × 50 mm column and a Qda single quadrupole mass detector.The levels of individual bile acid species were quantified against standard curves using commercially available standards (Fisher Scientific; Steraloids, Newport, RI, USA and Cayman Chemical Company, Ann Arbor, MI, USA).Amino acids were quantified as previously described (70).Briefly, samples were thawed, vortexed to homogenize, and centrifuged twice at 13,000 × g for 5 minutes.Amino acids in the supernatant were derivatized using the Waters AccQ-Tag Ultra Amino Acid Derivatization Kit and analyzed using the Waters UPLC AAA H-Class Application Kit (Waters Corporation, Milford, MA, USA) according to the manufacturer's instructions.Short chain fatty acids were quantified using a GC-MS (QP2010 Ultra, Shimadzu Scientific, Columbia, MD, USA) as described previously (30).Briefly, samples were centrifuged at 5,000 × g for 10 minutes at 4°C.The supernatant was harvested and filtered through a 0.2 µm PES filter (Corning, Corning, NY, USA) for SCFA detection.All of the GC vials were randomized prior to placing on the GCMS, and each sample was run in triplicate.Chromatographs were integrated based on the best m/z for each compound and quantified using an internal standard and calibration of the pure compounds in concentrations between 1.25 and 5,000 ppm.

FIG 2
FIG 2 Visualization of community structure as portrayed with PCoA featuring (A) unweighted and (B) weighted UniFrac distances.