Mucin-driven ecological interactions in an in vitro synthetic community of human gut microbes Authors: Maryse D Berkhout, Athanasia Ioannou, Carol de Ram, Sjef Boeren, Caroline M Plugge, Clara Belzer Published: 2024 ABSTRACT AbstractSpecific human gut microbes inhabit the outer mucus layer of the gastrointestinal tract. Certain residents of this niche can degrade the large and complex mucin glycoproteins that constitute this layer and utilise the degradation products for their metabolism. In turn, this microbial mucin degradation drives specific microbiological ecological interactions in the human gut mucus layer. However, the exact nature of these interactions remains unknown. In this study, we designed and studied an in vitro mucin-degrading synthetic community that included mucin O-glycan degraders and cross-feeding microorganisms by monitoring community composition and dynamics through a combination of 16S rRNA gene amplicon sequencing and qPCR, mucin glycan degradation with PGC-LC–MS/MS, production of mucin-degrading enzymes and other proteins through metaproteomics, and metabolite production with HPLC. We demonstrated that specialist and generalist mucin O-glycan degraders stably co-exist and found evidence for cross-feeding relationships. Cross-feeding on the products of mucin degradation by other gut microbes resulted in butyrate production, hydrogenotrophic acetogenesis, sulfate reduction and methanogenesis. Metaproteomics analysis revealed that mucin glycan degraders Akkermansia muciniphila, Bacteroides spp. and Ruminococcus torques together contributed 92% of the total mucin O-glycan degrading enzyme pool of this community. Furthermore, comparative proteomics showed that in response to cultivation in a community compared to monoculture, mucin glycan degraders increased carbohydrate-active enzymes whereas we also found indications for niche differentiation. These results confirm the complexity of mucin-driven microbiological ecological interactions and the intricate role of carbohydrate-active enzymes in the human gut mucus layer.Keywords: Akkermansia, CAZymes, gut microbiota, mucin, synthetic community Introduction A growing body of literature highlights the human gut mucosa as a niche for a specific microbiome that benefits host health (Zoetendal et al. 2002; Eckburg et al. 2005; Hong et al. 2011; Zhang et al. 2014; Li et al. 2015). However, a functional understanding of this mucosal microbiome as an ecological network is still lacking. The large intestinal epithelium is protected from direct contact with gut microbes by this mucus layer, which consists of an inner and outer mucus layer. The inner mucus layer is impenetrable to bacteria, but the outer mucus layer is more loose and forms a niche for specific gut microbes (Johansson et al. 2008). Some of these microbes can degrade the mucin glycans that constitute this layer (Glover et al. 2022). In turn, their presence and activity influence host immune and metabolic responses (Suriano et al. 2022). Mucin glycans are complex glycoproteins that consist of a protein core and branched O-glycan extensions, which are composed of galactose (Gal), N-acetylglucosamine (GlcNAc) and N-acetylgalactosamine (GalNAc), and capped by sialic acid (Neu5Ac), fucose (Fuc) or sulfate groups (Tailford et al. 2015; Corfield 2018). In order to degrade these complex and structurally diverse glycoproteins, an extensive palette of extracellular mucin O-glycan targeting enzymes has evolved within the mucosal microbiome (Martens et al. 2008; Pudlo et al. 2015; Tailford et al. 2015; Glover et al. 2022; Berkhout et al. 2022a). Several human gut commensals are capable of mucin glycan degradation. This includes commensal bacteria from the genera Bifidobacterium (Hoskins et al. 1992; Ruas-Madiedo et al. 2008), Bacteroides (Salyers et al. 1977; Sato et al. 2020), Prevotella (Wright et al. 2000), Akkermansia (Derrien et al. 2004) and Ruminococcus (Hoskins et al. 1992). Some of these mucin glycan degraders (MD) are specialists, such as Akkermansia muciniphila and Ruminococcus gnavus, which dedicate most of their carbohydrate-degrading machinery towards mucin (van Passel et al. 2011; Crost et al. 2013). Other mucin glycan degraders target a wider array of carbohydrates that are available in the human gut and are therefore considered generalists, such as Bifidobacterium bifidum (Turroni et al. 2010) and Bacteroides thetaiotaomicron (Martens et al. 2008). Furthermore, the mucosal layer houses microbes that can cross-feed on the products of mucin degradation to produce beneficial short-chain fatty acids (SCFAs) for uptake by the host (Mahowald et al. 2009; Wrzosek et al. 2013; Belzer et al. 2017; Bunesova et al. 2018; Nishiyama et al. 2018; Pichler et al. 2020). We hypothesise that mucin glycan degradation occurs in an ecological network of mucin glycan degraders and cross-feeders that mutually express the complete repertoire of enzymes necessary for degradation and utilisation of mucin O-glycans. It is crucial to understand the gut mucosal microbial ecological networks and their functions to appreciate their abilities and potential to modulate host health and cellular responses. Therefore, we studied a synthetic community (SynCom) of mucosal microbes that was grown with mucin as the main carbon and energy source in a continuously operated fermentor system to evaluate microbial ecological network structure and function. This approach was chosen for several reasons: As opposed to diet-derived glycans, mucin O-glycans form a constantly available energy source for the gut microbiota that can utilise them (Koropatkin et al. 2014). By supplying mucin continuously, we simulate its constant availability in the gut. By supplying solely mucin to the synthetic community, we focus exclusively on the mucin-driven ecological interactions. Furthermore, in this in vitro system growth conditions can be controlled and modulated, and sampling is convenient and can be frequent (Shetty et al. 2022). Finally, applying a defined microbial community allows replication of experiments and the selection of well-studied mucosal residents with a function of interest (Berkhout et al. 2022b). To evaluate how microbes build an ecological network in a mucosal environment, we set up a mucin-degrading synthetic community (MDSC) in bioreactors. The synthetic community consisted of three proposed guilds: mucin glycan degraders (MD), secondary degraders (SD) and hydrogen consumers (HC). We monitored microbial relative abundance, the degradation of mucin O-glycans, metabolite production and protein expression of MDSC. Our main finding is that single microbes express different mucin glycan degrading enzymes in single culture compared to during growth in a community and that within this in vitro microbial synthetic community mucin sustains multiple mucin-degrading species and cross-feeders with overlapping function. The resulting microbial network produces the enzymes to efficiently break down mucin and can be used in follow-up experiments to further evaluate microbe-mucus interactions. Results Design and set-up of a mucin glycan degrading synthetic community in bioreactors We set up a 15-species mucin-degrading synthetic community (MDSC) to grow in continuous anoxic bioreactors with porcine gastric mucin as the main carbon and energy source. MDSC consisted of known residents of the human gut mucosal layer. We selected mucin glycan degraders (MD), secondary degraders (SD) and hydrogen consumers (HC) based on presence in the human gut mucosal layer, relationship with the host, culturability, availability of the genome and fit into one of the proposed guilds. MDSC was designed to incorporate functional redundancy in terms of mucin glycan degradation capacity and included both mucin specialists and generalists. The community included seven well-characterised MD: specialists Akkermansia muciniphila, Ruminococcus gnavus and Ruminococcus torques, and generalists Bacteroides caccae, Bacteroides fragilis, Bacteroides thetaiotaomicron and Phocaeicola vulgatus. The secondary degraders (SD) guild consisted of non-mucin degrading residents of the gut mucosal layer that can cross-feed on the released products of mucin degradation (Belzer et al. 2017). The SDs in this study were producers of the beneficial SCFA butyrate, namely Anaerostipes caccae, Anaerobutyricum hallii, Agathobacter rectalis, Faecalibacterium duncaniae and Roseburia intestinalis. The third proposed guild included hydrogen (H2) consumers Desulfovibrio piger, Methanobrevibacter smithii and Blautia hydrogenotrophica. Hydrogen is a major end product of fermentation in the gut, and its accumulation may slow further fermentation and SCFA production (Smith et al. 2019). The three groups of hydrogen consumers in the human gut were represented in MDSC: sulfate reducing bacterium (SRB) Desulfovibrio piger, methanogen Methanobrevibacter smithii and acetogen Blautia hydrogenotrophica. Community development MDSC achieved a stable phase in terms of microbial composition, and metabolite and protein production in the three replicate bioreactors (F1-F3) between t = 72–120 h. From t = 72 h onwards, the OD600nm (Supplementary Fig. 2A) and the total bacterial abundance as determined by qPCR (Supplementary Fig. 2B) were stable. Furthermore, we assessed community stability and replicability during this stable phase based on microbial relative abundance. For replicates, no significant differences in community composition were found by PERMANOVA (F = 0.58, P value = 0.81, Supplementary Table 4A). To assess community composition and stability over time during the stable phase, PERMANOVA was applied to only t = 72-120 h. No significant differences in community composition were found during the stable phase (F = 0.77, P value = 0.67, Supplementary Table 4B). Another indication of stable phase was stable metabolite production. The major metabolites that were produced by the MDSC were ethanol and SCFAs acetate, propionate and butyrate (Supplementary Fig. 3A–D). The ratio between the major SCFAs during the stable phase (t = 72–120 h) was butyrate:propionate:acetate = 1:1:5.7 (Supplementary Fig. 3E). Metaproteomic analysis confirmed stability of MDSC in terms of protein presence and abundance. Pairwise comparison by FDR-corrected t-tests on log-transformed LFQ values of all proteins detected in each timepoint of the stable phase revealed no or a negligible number of significantly differently detected proteins between timepoints (Supplementary Fig. 4A–F, Supplementary Fig. 5A–C). Therefore, t = 72–120 h was considered as the stable phase and will be the focus of further analyses. During the stable phase (t = 72–120 h), the community was dominated by mucin glycan degraders (Fig. 1) (mean relative abundance during stable phase = 0.92). In particular, A. muciniphila (0.54) and Bacteroides spp. (0.30) displayed a high relative abundance, followed by Ruminococcus spp. (0.07) (Fig. 1). Although the secondary degraders had a mean relative abundance of 0.10 after the batch phase (t = 6 h), their combined mean relative abundance decreased to 0.013 during the stable phase. Still, A. caccae and F. duncaniae were present in all three fermentors at all timepoints. A. hallii was present in F2 and F3 at all timepoints but could not be detected after t = 96 h in F1. A. rectalis and R. intestinalis could not be detected after the batch phase (t = 6 h). Hydrogen consumers were consistently present with a mean relative abundance of 0.056 during the stable phase. The most abundant hydrogen consumer was acetogen B. hydrogenotrophica, followed by sulfate reducer D. piger. Methanogen M. smithii was not detected for all timepoints in all fermentors, probably due to its low relative abundance (mean relative abundance during stable state 0.0012) and growth rate. At t = 48 h, a contamination with Bacillus thuringiensis occurred in all three fermentors (Supplementary Text 3) and remained part of MDSC at all timepoints (mean relative abundance during stable phase 0.010). Relative abundance of MDSC species over time based on a combination 16S rRNA gene amplicon sequencing and qPCR. Relative abundances of B. caccae, B. fragilis, B. thetaiotaomicron, and P. vulgatus were determined by species-specific qPCR. Data is normalised for 16S rRNA gene copy number per species. MDSC was collectively capable of utilising all O-glycans from porcine stomach mucin. This includes a variety of mucin O-glycan core structures, which are mainly extended by GlcNAc and Gal. The main terminal structure was fucose, but there were also sialylated and sulfated O-glycans (Supplementary Fig. 6). PGC-LC–MS/MS analysis revealed that all O-glycans that were measured in the medium at t = 0 h were completely or substantially utilised during the stable phase (t = 96 h and t = 120 h, Supplementary Fig. 6). In summary, the consortium is replicable and the stable phase was reached after t = 72 h, which is supported by stability in terms of community composition, metabolite production, mucin O-glycan degradation and microbial proteins. The community consisted of representatives of all three proposed trophic guilds and included seven mucin glycan degraders, two secondary degraders and two hydrogen consumers. Therefore, the stable phase was used for further analysis. Five mucin glycan degraders account for most mucin-degrading CAZymes and sulfatases Metaproteomics analysis revealed that MDSC produced a broad palette of annotated mucin-targeting enzymes during the stable phase (t = 72 h, t = 96 h and t = 120 h), which included peptidases that cleave the protein backbone, CAZymes that cleave the mucin O- glycans and sulfatases that remove sulfate groups from O-glycans (Supplementary Table 5). The main contributors to the CAZyme and sulfatase pool were A. muciniphila, B. caccae, B. fragilis, B. thetaiotaomicron and R. torques (Fig. 2, Supplementary Fig. 7). These five organisms together accounted for 92% of sialidases, fucosidases, galactosidases, hexosaminidases and sulfatases in MDSC. By contrast, a limited number of mucin-targeting enzymes were detected for P. vulgatus and R. gnavus. Furthermore, a small number of mucin-targeting CAZymes were detected for several non-mucin glycan degraders (Fig. 3). Number of fucosidases, galactosidases, hexosaminidases, sialidases, and sulfatases detected in MDSC during the stable phase, coloured by the organism that produced them. Detection of different enzyme categories per species. A circle around the species code signifies that a gene coding for this enzyme type was present in the genome (that was used as a database for the analysis of proteomics data) of this species. A filled circle indicates that this enzyme was detected in our metaproteomic analysis. Am = A. muciniphila; Bc = B. caccae; Bf = B. fragilis; Bt = B. thetaiotaomicron; Pv = P. vulgatus; Ac = A. caccae; Fd = F. duncaniae; Ah = A. hallii; Ar = A. rectalis; Ri = R. intestinalis; Bh = B. hydrogenotrophica; Dp = D. piger; Ms = M. smithii. MDSC produced more CAZymes than solely mucin O-glycan targeting enzymes during growth on mucin glycoproteins. In total, 264 CAZymes from 48 different glycoside hydrolase (GH) families were detected (Fig. 4). This included for example mannanases and mannosidases that potentially target the mannose-rich mucin N-glycans which form a minority of mucin glycans (Gunning et al. 2013; Luis and Hansson 2023). The majority of the mannose-targeting enzymes were produced by B. thetaiotaomicron (Supplementary Fig. 8, Supplementary Table 5). Overall, the most well-represented GH families (≥10 proteins identified in dataset) were GH2, GH13, GH20, GH29 and GH92. With the identification of 88 GHs, B. thetaiotaomicron showcased the largest palette of glycan-targeting enzymes (Supplementary Table 5). Other organisms with a wide variety of GHs included A. muciniphila [39], B. caccae [52] and B. fragilis [30]. Glycosyl hydrolases (GHs) per GH family detected in the synthetic community. In addition to glycan-targeting enzymes, MDSC also synthesised peptidases which potentially target the mucin protein backbone. Overall, 208 peptidases that could be assigned to a MEROPS peptidase family (Rawlings et al. 2018) were identified (Supplementary Fig. 9). More than half of these peptidases belonged to the metallopeptidase (M) family. 43 MEROPS families were represented in this community, of which M3, M16, M18, M20, M24, M41, S9 and S41 were identified most (≥10 proteins identified in dataset). The organisms with the largest number of peptidases were mucin glycan degraders A. muciniphila [21], B. caccae [44], B. fragilis [28] and B. thetaiotaomicron [37]. MDSC produces enzymes for the metabolism of mucin monosaccharides After release of a monosaccharide from a mucin glycan by a CAZyme, this monosaccharide can either be metabolised by the microorganism that released the monosaccharide, or it can be scavenged by another community member. Therefore, we assessed the proteins related to mucin O-glycan monosaccharide metabolic pathways within MDSC and compared the results to the genomic capacity of the individual members of MDSC (Fig. 5A–C). Fucose is most commonly degraded via the fuculose-1-phosphate pathway (Fig. 5A). Within MDSC, fucose isomerase (FucI) was detected for most mucin glycan degraders [5/7], whereas fuculose kinase (FucK) was only detected for A. muciniphila. Although there are more MDSC members that encode a fuculose phosphate aldolase (FucA), it was only detected for R. gnavus (Fig. 5A). Lactaldehyde reductase (FucO) was detected for several mucin glycan degraders and secondary degraders within MDSC. Finally, we detected propanediol utilisation (Pdu) proteins for R. gnavus and A. hallii (Supplementary Table 6, Supplementary Table 7). Sialic acid and GalNAc metabolism were rare in MDSC. Although N-acetylneuraminate lyase (NanA) was detected for some members of MDSC, other enzymes of the Nan cluster for sialic acid utilisation were not observed (Fig. 5B). For GalNAc metabolism, only D-galactosamine 6-phosphate deaminase/isomerase (AgaS) was detected for R. torques. By contrast, proteins for the metabolism of GlcNAc were amply detected. N-acetylglucosamine-6-phosphate deacetylase (NagA) was detected for five mucin glycan degraders, secondary degrader F. duncaniae and hydrogen consumer B. hydrogenotrophica, and glucosamine-6-phosphate deaminase (NagB) was found for all mucin glycan degraders, A. caccae, F. duncaniae and B. hydrogenotrophica (Fig. 5B). Enzymes from the Leloir pathway for galactose utilisation were also widely detected in MDSC (Fig. 5C). Galactose mutarotase (GalM) was detected for five mucin glycan degraders, galactokinase (GalK) for all mucin glycan degraders and B. hydrogenotrophica, UDP-galactose-4-epimerase (GalE) for six mucin glycan degraders, three secondary degraders and B. hydrogenotrophica, and glucose-1-phosphate-uridylyltransferase (GalU) was identified for A. muciniphila, two secondary degraders and D. piger. Mucin monosaccharide metabolism by MDSC. A) Fucose, B) sialic acid, and GlcNAc, C) galactose. A circle around the species code signifies that this enzyme type was present in the genome (that was used as a database for the analysis of proteomics data) of this species. A filled circle indicates that this enzyme was identified in our proteomics analysis. Am = A. muciniphila; Bc = B. caccae; Bf = B. fragilis; Bt = B. thetaiotaomicron; Pv = P. vulgatus; Ac = A. caccae; Fd = F. duncaniae; Ah = A. hallii; Ar = A. rectalis; Ri = R. intestinalis; Bh = B. hydrogenotrophica; Dp = D. piger; Ms = M. smithii. Butyrate production indicates cross-feeding by secondary degraders The secondary degraders in MDSC are unable to degrade mucin glycans in monoculture. Therefore, secondary degraders in this consortium must rely on cross-feeding to sustain themselves. As the secondary degraders were the only microbes of MDSC that possess the Bcd-Etf gene for butyrate production (Supplementary Table 7), we hypothesised that successful cross-feeding by these bacteria would result in the production of butyrate, which we indeed observed albeit in small concentrations (<3 mM) (Supplementary Fig. 3). Furthermore, we observed that secondary degraders A. caccae and F. duncaniae remained part of the community during the complete experiment in all three replicate bioreactors. Additionally, A. hallii was present for the duration of the fermentation in two out of three bioreactors. In MDSC, the Bcd-Etf protein that is responsible for the conversion of crotonyl-CoA to butyryl-CoA (Louis and Flint 2009; Vital et al. 2014) was detected for all five secondary degraders (Supplementary Table 6). This indicated that these secondary degraders indeed cross-fed on the products of mucin degradation by primary degraders, which resulted in the production of butyrate. Hydrogen is scavenged primarily through acetogenesis and sulfate reduction MDSC contained three organisms that can utilise the H2 gas that is produced during the fermentation of mucin, namely acetogen B. hydrogenotrophica, sulfate reducer D. piger and methanogen M. smithii. In MDSC, B. hydrogenotrophica expressed the complete Wood-Ljungdahl pathway (WLP) for reduction of CO2 to acetate (Fig. 6). The enzymes of both branches, including key enzymes CODH, FDH and ACS, were all observed except for methylenetetrahydrofolate dehydrogenase (Supplementary Table 6). Furthermore, B. hydrogenotrophica also produced enzymes for the conversion of acetyl-CoA to acetate. Altogether, B. hydrogenotrophica expressed the WLP and generated acetate from H2 and CO2. We hypothesised that SRB D. piger would be able to reduce sulfate released from mucin by MDSC to hydrogen sulfide (H2S), as was shown previously (Gibson et al. 1988; Rey et al. 2013). Indeed, D. piger produced all three key enzymes for the reduction of sulfate in MDSC (Supplementary Table 6, Fig. 6). Finally, H2 could be removed in MDSC through hydrogenotrophic methanogenesis by M. smithii that expressed the complete hydrogenotrophic methanogenesis pathway (Fig. 6), which is striking considering the low relative abundance of M. smithii in MDSC (Fig. 1). The methanogen also produced a formate dehydrogenase, allowing formate as a substrate for this pathway as well (Supplementary Table 6). Overall, metaproteomics analysis revealed that acetogenesis, sulfate reduction and methanogenesis were three pathways for hydrogen scavenging in MDSC. Hydrogen scavenging pathways in MDSC: Acetogenesis by B. hydrogenotrophica (Bh), sulfate reduction by D. piger (Dp), and methanogenesis by M. smithii (Ms). A circle around the species code signifies that this enzyme type was present in the genome (that was used as a database for the analysis of proteomics data) of this species. A filled circle indicates that this enzyme was identified in our proteomics analysis. Mucin glycan degraders upregulate CAZymes in response to cultivation in a community compared to cultivation in monoculture To compare the response of individual microbes to the presence of other mucin glycan degraders, we compared proteins detected in MDSC to proteins detected in monocultures of MDSC members. Comparison of normalised protein LFQs between pooled monocultures and community samples yielded 567 significantly differently regulated proteins in the pellet and 548 significantly differently regulated proteins in the supernatant (S0 = 2 and FDR < 0.05, Supplementary Table 8, Supplementary Fig. 10). Active growth of mucin glycan degraders in a mucin glycan degrading community compared to bacterial monoculture on mucin induced a significant increase of several mucin O-glycan degrading CAZymes (Fig. 7). In particular, hexosaminidases were upregulated during growth in the community by A. muciniphila (GH20), B. fragilis (GH109), B. thetaiotaomicron (GH20) and R. torques (GH20, GH84, GH89), but A. muciniphila downregulated a GH84 hexosaminidase. Additionally, galactosidases were upregulated in community by A. muciniphila (GH2), B. caccae (GH2), B. thetaiotaomicron (GH2, GH27), P. vulgatus (GH2) and R. torques (GH2). Furthermore, during growth in the community, A. muciniphila downregulated a GH33 sialidase in the community, whereas B. thetaiotaomicron and R. torques upregulated a GH33 sialidase. Moreover, R. torques upregulated a GH29 fucosidase and a GH95 fucosidase, whereas R. gnavus downregulated a GH29 fucosidase in the community. Finally, B. thetaiotaomicron upregulated a sulfatase in the community. Differential detection of mucin-targeting glycosyl hydrolases by mucin glycan degraders in community compared to monoculture. A broader arsenal of GHs was upregulated during growth in community. For example, Bacteroides spp. increased numerous proteins related to their starch utilization system (Sus). Moreover, significantly upregulated proteins during growth in the community included a GH3 glucosidase, a GH18 putative secreted endoglycosidase, a GH30 glucosylceramidase, a GH31 xylosidase, a GH38 mannosidase, a GH76 mannanase, a GH97 glucosidase, a putative GH92 mannosidase and a GH125 from B. thetaiotaomicron and a GH1 glucosidase, GH31, GH36 and GH51 from R. torques. By contrast, a B. fragilis GH3 glucosidase, a B. thetaiotaomicron GH3 glucosidase and a P. vulgatus GH63 were downregulated in the community. Various proteins related to mucin monosaccharide metabolism were significantly regulated during growth in the community. For fucose metabolism, R. gnavus upregulated FucU and A. muciniphila upregulated FucI. Galactose metabolism was upregulated by R. torques, through upregulation of proteins GalM, GalK and GalE. GalM was also upregulated by B. thetaiotaomicron. Additionally, R. torques upregulated NagA from the GlcNAc metabolism pathway. For GalNAc and sialic acid metabolism no significantly regulated proteins were identified. For peptidases, differences in protein detection between community and single culture were observed as well. More specifically, B. thetaiotaomicron upregulated an M60 and M16 peptidase in the community, whereas B. caccae upregulated an S9 peptidase and R. gnavus upregulated a C39 peptidase. Furthermore, peptidases that were downregulated were an M3 of A. muciniphila, an S1 and an M50 of B. thetaiotaomicron, an M18 of R. gnavus and an S14 from R. torques. Several other notable differentially regulated proteins were found. For example, A. muciniphila upregulated glutamate decarboxylase in the community, whereas B. fragilis and B. thetaiotaomicron downregulated this protein in the presence of others. Additionally, several proteins that are potentially related to microbe-microbe interactions were significantly differently regulated. For example, A. muciniphila upregulated a type II/III secretion system protein and B. fragilis upregulated a type IV secretion system protein in the community. Furthermore, R. torques upregulated a beta-lactamase, whereas B. fragilis downregulated its putative beta-lactamase. P. vulgatus downregulated a tetracycline resistance protein. Discussion This study examined a synthetic mucin-degrading microbial community (MDSC), which consisted of mucin glycan degraders, secondary degraders and hydrogen consumers, to model the microbial ecological network of the human gut mucus layer. We cultivated the fifteen species that made up this community in triplicate bioreactors with porcine gastric mucin as the carbon source. During the stable phase, we monitored microbial community composition, metabolite production, mucin O-glycan degradation and protein production. Thus, we modelled an ecological network of the human gut mucosal layer with an in vitro synthetic community approach combined with metaproteomic analysis of mucin-targeting enzymes and downstream metabolism. We successfully assembled a synthetic mucin-degrading community consisting of mucin glycan degraders and cross-feeders. As expected, the mucin glycan degrader guild dominated, with mucin specialist A. muciniphila as the most abundant species. All seven mucin glycan degraders co-existed, so the four generalist species were not outcompeted by the mucin specialists. This is in agreement with previous studies, including a study that describes in vitro co-existence of mucin specialist A. muciniphila and generalist B. thetaiotaomicron (Kostopoulos et al. 2021) and an in vivo study that showed co-existence of A. muciniphila, P. vulgatus and R. gnavus in mucosal biopsies of healthy controls (Png et al. 2010). Furthermore, a mice study demonstrated co-existence of A. muciniphila, B. caccae and B. thetaiotaomicron during a fibre-free diet, which stimulated utilisation of mucin glycans (Desai et al. 2016). A recent effort found that mucin degradation by R. torques enhanced growth of B. thetaiotaomicron in an in vitro coculture through cross-feeding (Schaus et al. 2024). A. muciniphila and Ruminococcus spp. often display an inverse relationship wherein A. muciniphila is positively associated with health and, even though they are considered commensal microbiota, Ruminococcus spp. are associated with a disease state (Png et al. 2010; Qiu et al. 2022). As the relative abundance of Ruminococcus spp. in MDSC is low compared to the dominant A. muciniphila, it can be assumed that the balance between A. muciniphila and Ruminococcus spp. in MDSC resembles a more healthy state of the mucosal microbiome, which was the purpose of this study. Taken together, these findings suggest that the complexity of mucin induces expression of distinct strategies of microbial mucin glycan degradation by different mucin glycan degraders (Crost et al. 2016; Kim et al. 2023; Shuoker et al. 2023; Schaus et al. 2024), which allows co-existence. We detected a wealth of bacterial enzymes that are crucial for mucin O-glycan degradation. Mucin O-glycans are often capped by sialic acid, fucose and/or a sulfate group and removal of these terminal monosaccharides is regarded as an essential step for microbial mucin O-glycan degradation (Tailford et al. 2015; Luis et al. 2021; Shuoker et al. 2023). Therefore, sialidase, fucosidase and sulfatase expression is crucial for access to mucin O-glycans. In MDSC, five out of seven mucin glycan degraders produced all three of these enzyme categories (Fig. 3). Furthermore, six out of seven mucin glycan degraders produced galactosidases and hexosaminidases, which release galactose, GlcNAc and GalNAc subunits from mucin O-glycans. This broad expression of mucin O-glycan degrading CAZymes suggests that all seven mucin glycan degraders actively participated in mucin glycan degradation, which sustained them in this highly competitive community. Furthermore, we detected bacterial enzymes that could degrade the N-glycans that are present in low quantities in mucin (Luis and Hansson 2023), such as mannosidases and mannanases. The majority of these mannose-releasing enzymes were produced by B. thetaiotaomicron (Supplementary Fig. 8, Supplementary Table 5). This may pave a unique niche for B. thetaiotaomicron or Bacteroides spp. during growth on mucin. However, this work focused on mucin O-glycan degradation, as O-glycosylation is dominant over N-glycosylation in both porcine gastric mucin (data not shown) and human MUC2 (Luis and Hansson 2023). Several CAZymes that were identified as crucial in mucin degradation by previous studies were detected here. First of all, we observed sulfatase BT1636 (Q8A789) from B. thetaiotaomicron which plays a crucial role in this organism’s access to mucin by removing 3S-Gal (Luis et al. 2021). Furthermore, key fucosidases Amuc_0392 (GH29) and Amuc_1120 (GH95), and key GH33 sialidases Amuc_0625 and Amuc_1835 were detected in our dataset (Shuoker et al. 2023). However, Amuc_1547, which is the defining member of the GH181 family of sialidases, was not identified in our dataset. This may be explained by its specificity towards the sialyl-T-antigen (Shuoker et al. 2023). Supplementary Table 5 lists references to additional previously characterised CAZymes. Additionally, we compared proteins of mucin glycan degraders in the community to monocultures, which showed that in general, growth in a community led to a relatively increased abundance of CAZymes. This was not limited to only mucin O-glycan targeting CAZymes including fucosidases, sialidases, galactosidases and hexosaminidases, but also included other GHs such as glucosidases and mannosidases. This is in accordance with a transcriptomics study that compared A. muciniphila and B. thetaiotaomicron in coculture to their respective monocultures (Kostopoulos et al. 2021). However, in the present study there were also indications of niche differentiation within mucin glycans. For example, B. thetaiotaomicron and R. torques upregulated a sialidase, whereas A. muciniphila downregulated a sialidase in the community. Similarly, R. gnavus downregulated a fucosidase in the community, whereas R. torques upregulated fucosidases. These opposite effects could indicate that in response to the presence of other microbes with similar mucin O-glycan degradation capacities, a microbe’s niche within the mucin glycan is confined to release and/or degradation of specific monosaccharide(s). In addition to mucin glycan degrading enzymes, we detected a wealth of peptidases that potentially target the mucin protein backbone through metaproteomics. The main contributors to the peptidase pool were Bacteroides spp. and A. muciniphila. We detected four peptidases from A. muciniphila and one peptidase from B. thetaiotaomicron that were previously characterised as O-glycopeptidases that specifically cleave the mucin peptide backbone, namely Amuc_1438, Amuc_1514, Amuc_0627 and Amuc_0908, and BT4244 (Noach et al. 2017; Shon et al. 2020; Medley et al. 2022; Taleb et al. 2022). Furthermore, this mucin degrading M60 peptidase BT4244 was significantly upregulated in the community compared to monoculture, along with an M16 peptidase from B. thetaiotaomicron. This could indicate that B. thetaiotaomicron upregulates its mucin protein degrading machinery to compete with A. muciniphila. However, future studies should assess the degradation of the mucin protein in community and monocultures to determine microbial ecological interactions that are specifically driven by the mucin protein. Mucin glycan degradation by MD in MDSC sustained several cross-feeders. On a metaproteomic level, we found indications for potential substrate and metabolite cross-feeding. The extracellular action of CAZymes during growth on mucin can lead to the release of monosaccharides and small oligosaccharides as a substrate for cross-feeding by other community members, as has been demonstrated before in smaller scale synthetic communities (Belzer et al. 2017; Pichler et al. 2020; Shuoker et al. 2023; Schaus et al. 2024). An example of potential substrate cross-feeding in MDSC is GlcNAc utilisation by F. duncaniae and B. hydrogenotrophica, as Nag proteins for GlcNAc metabolism were detected, but no hexosaminidases were identified. However, the Nag proteins were also widely detected for mucin glycan degraders, indicating that these bacteria themselves benefited from the release of GlcNAc. Furthermore, mucin-driven metabolite cross-feeding can occur during growth on mucin, as has been shown before (Mahowald et al. 2009; Wrzosek et al. 2013). An example of potential metabolite cross-feeding within MDSC is A. hallii, for which Pdu proteins were detected but not the fuculose-1-phosphate pathway for fucose utilisation. Pdu proteins are required for the conversion of 1,2-propanediol to propionate (Reichardt et al. 2014), so this observation might be an indication for cross-feeding of A. hallii on 1,2-propanediol that was produced by the metabolism of fucose by other MDSC members. We detected a limited number of Nan proteins for the utilisation of sialic acid within MDSC. This could indicate that sialic acid utilisation in this community is low, which can be explained by the lower relative abundance of sialic acid in porcine gastric mucin compared to human MUC2 (Fancy et al. 2024) (Supplementary Fig. 6) (0.5%–1.5% bound sialic acid according to the manufacturer) or by a preference for other available substrates by this community. The lower availability of sialic acid might have influenced, for example, R. gnavus, whose mucin degradation strategy largely depends on sialic acid metabolism (Bell et al. 2019). Similarly, a low number of proteins related to GalNAc metabolism were identified. This may be explained by the fact that GalNAc metabolism is not widely encoded within MDSC (Supplementary Table 7). This is in accordance with a genomic study that found GalNAc metabolism in only 76 out of 397 analysed human gut microbiota genomes (Ravcheev and Thiele 2017). In MDSC, only F. duncaniae and R. torques were expected to encode the genes that are necessary for GalNAc utilisation by this genomic study. In the present study, only AgaS was detected for R. torques, indicating possible utilisation of GalNAc as a unique niche within MDSC for this bacterium. We included three hydrogen consumers in this community to remove H2 that is produced by other community members during fermentation. A previous in vitro study showed that acetogen B. hydrogenotrophica, sulfate reducer D. piger and methanogen M. smithii can co-exist and grow together in the presence of H2 and sulfate (Wang et al. 2023). In our experiment, B. hydrogenotrophica and D. piger co-existed, whereas M. smithii was not consistently observed. D. piger is the sole sulfate reducer in this community, allowing it to utilise sulfate that is liberated from mucin by mucin glycan degraders, as has been demonstrated previously (Willis et al. 1996; Rey et al. 2013). This is confirmed by the detection of all proteins that it requires for sulfate reduction. Additionally, for B. hydrogenotrophica a near-complete WLP for the utilisation of H2 and CO2 for acetogenesis was detected. B. hydrogenotrophica can also grow heterotrophically on sugars that are present in this experiment (Bernalier et al. 1996). Therefore, it is likely that in this community it is not completely dependent on the WLP to provide for itself. Even though M. smithii was not consistently observed in all fermentors at all timepoints, its proteins that are involved in methanogenesis were identified, indicating that this organism was performing methanogenesis. Overall, the main pathways for the removal of H2 in MDSC were acetogenesis and sulfate reduction. The main secondary degraders in MDSC were A. caccae and F. duncaniae. They were consistently present throughout the continuous fermentation and we detected enzymes related to the production of butyrate for these species. Both were shown to cross-feed on the products of mucin glycan degradation by A. muciniphila in a previous study (Belzer et al. 2017), so it can be assumed that similar interactions occurred in MDSC. The ratio between the major SCFAs during the stable phase of MDSC was butyrate:propionate:acetate = 1:1:5.7, whereas the typical ratio of SCFAs in the lumen of the colon is butyrate:propionate:acetate = 1:1:3 (Cummings et al. 1987), indicating that acetate production is relatively high. The lower butyrate concentration can be explained by a lack of substrates that secondary degraders can utilise, as in MDSC they depend on mucin degradation by others and produce butyrate by cross-feeding, whereas they can produce butyrate by directly fermenting dietary fibers in vivo. However, this in vitro ratio does not take absorption by the host into account. The synthetic community approach allowed us to focus on mucin-driven ecological interactions, without interference of other substrates in the gut and select microbes of interest to create a microbial network. The in vitro chemostat approach enabled thorough control of the continuous growth conditions, frequent sampling and replication. Combining the in vitro synthetic community set-up with metaproteomics yielded unique insights in the mucin-driven ecological interactions of the human gut microbiota. However, this approach also has limitations. First, there is no host component that mimics mucin-driven host–microbe interactions, which are of importance given the proximity of these microbes to the host epithelium in vivo and are reviewed elsewhere (Ottman et al. 2017; Cai et al. 2020; Crost et al. 2023). Approaches to incorporate a host cell component in an in vitro mucin-driven microbiological experiment include cocultivation of the microbes with a human cell line, with an organ-on-a-chip device or using gut organoids (Elzinga et al. 2019). Furthermore, there was no possibility for the microbes to attach to a surface in the bioreactors, which were stirred constantly. Physical attachment might be crucial for some ecological interactions in the mucus layer. A potential solution would be the application of mucin-coated carriers that allow attachment of bacteria (Van Den Abbeele et al. 2013; Liu et al. 2023). The third major limitation is the use of pig gastric mucin as primary carbon and energy source. Although pig gastric mucin is rather similar to human intestinal mucin in composition, pig gastric mucin has a lower degree of sialylation and it is more heavily sulfated than human intestinal mucin (Quintana et al. 2018). This may have resulted in differences in the microbial community composition, as, for example, the higher degree of sulfation may have benefitted sulfate reducer D. piger. In summary, we established a synthetic microbial mucin glycan degrading community and provided further evidence to the hypothesis that mucin glycan degradation occurs in a network of cooperating human gut microbes. Our metaproteomic analysis revealed the richness of this community in terms of mucin-targeting enzymes, downstream metabolism and cross-feeding interactions (Fig. 8). Future studies could reduce complexity by decreasing the number of species to mimic specific interactions and/or selecting a more simple experimental approach such as batch fermentations in anoxic vials. Complexity could be increased by increasing the number of species through rational selection, or by selecting a more complex system that includes an attachment and/or host component. Furthermore, synthetic mucin glycan degrading microbial communities could be exposed to different mucin types to model community behaviour under different conditions, including disease states. Overall, this study suggests that mucin degradation is a collective process that is executed by the human gut microbiota, which includes ecological interactions such as cross-feeding, co-existence and competition. Overview of mucin degradation within MDSC. Mucin glycan degraders apply carbohydrate-active enzymes and sulfatases to release monosaccharides and sulfate groups from mucin (A). Next, saccharides can be utilised by the mucin glycan degraders themselves or cross-feeding organisms (B), which results in the production of SCFAs. Hydrogen that is produced during the fermentation of mucin is mainly consumed through acetogenesis and sulfate reduction (C). Materials & methods Strains and cultivation Mucin glycan degrading bacteria (MD) included in this study were Akkermansia muciniphila DSM 22959T, Ruminococcus gnavus ATCC 29149T, Ruminococcus torques ATCC 27756T, Bacteroides caccae DSM 19024T, Bacteroides fragilis DSM 2151T, Bacteroides thetaiotaomicron DSM 2079T and Phocaeicola vulgatus ATCC 8482T (previously Bacteroides vulgatus). Secondary degraders (SD) included in this study were Anaerostipes caccae DSM 14662T, Anaerobutyricum hallii DSM 3353T (previously Eubacterium hallii), Agathobacter rectalis ATCC 33656T (previously Eubacterium rectale), Faecalibacterium duncaniae DSM 17677T (previously Faecalibacterium prausnitzii) and Roseburia intestinalis DSM 14610T. Hydrogen consuming organisms (HC) were Blautia hydrogenotrophica DSM 10507T, Desulfovibrio piger DSM 749T and Methanobrevibacter smithii DSM 11975. All strains were cultivated in a general, multipurpose, bicarbonate-buffered mineral salts freshwater anoxic medium that was composed and prepared as described previously (Plugge 2005), and enriched with 1 g/L yeast extract. The media were autoclaved at 121 °C for 20 min if mucin was not present and at 121 °C for 60 min if mucin was present. The media were reduced with cysteine-HCl (5% v/v of an autoclaved anoxic solution containing 20 g/L cysteine-HCl). Type III Mucin from porcine stomach (Sigma-Aldrich, St. Louis, MO, USA) was used in all experiments (0.5% w/v) and prepared according to Supplementary Text 1. Precultures Precultures were grown in anoxic serum bottles containing the medium described above and with carbon source(s) and gas phase required for their growth (Table 1). The additives were added after autoclavation from sterile anoxic stock solutions. After 0.5% (v/v) inoculation, the precultures were incubated for 24 h at 37 °C without shaking (exception: M. smithii was incubated for 3 days). The OD600nm was measured directly after inoculation and after incubation for normalisation purposes. Table 1Additions to the medium and gas phase for the preculture per strain. Strain Additives Gas phase a Akkermansia muciniphila 5 mg/mL crude mucinN2/CO2 80/20 Bacteroides caccae 5 mg/mL crude mucinN2/CO2 80/20 Bacteroides fragilis 5 mg/mL crude mucinN2/CO2 80/20 Bacteroides thetaiotaomicron 5 mg/mL crude mucinN2/CO2 80/20 Phocaeicola vulgatus 5 mg/mL crude mucinN2/CO2 80/20 Ruminococcus gnavus 20 mM glucose2 g/L peptoneN2/CO2 80/20 Ruminococcus torques 5 mg/mL crude mucin2 g/L peptoneN2/CO2 80/20 Anaerostipes caccae 20 mM glucose2 g/L peptoneN2/CO2 80/20 Faecalibacterium duncaniae 20 mM glucose2 g/L peptone20 mM acetateN2/CO2 80/20 Anaerobutyricum hallii 20 mM glucose2 g/L peptone20 mM acetateN2/CO2 80/20 Agathobacter rectalis 20 mM glucose2 g/L peptone20 mM acetateN2/CO2 80/20 Roseburia intestinalis 20 mM glucose2 g/L peptone20 mM acetateN2/CO2 80/20 Blautia hydrogenotrophica 20 mM glucoseH2/CO2 80/20 Desulfovibrio piger 20 mM Na2SO42 g/L peptoneH2/CO2 80/20 Methanobrevibacter smithii 6 g/L Brain Heart Infusion (BHI)6 g/L Proteose peptone1 g/L Yeast extract1 g/L Na-acetate2 g/L Na-formateH2/CO2 80/20 aBottles were pressurised at 1.5 atm. Additions to the medium and gas phase for the preculture per strain. aBottles were pressurised at 1.5 atm. Fermentor operation The experiments were performed with the DASGIP Parallel Bioreactor System (Eppendorf, Hamburg, Germany) (Supplementary Fig. 1). First, the bioreactors were filled with 408 mL medium containing 0.5% (w/v) purified mucin as in Supplementary Text 1, leaving space for 72 mL for inoculation at a later stage. Before adding the reducing agent, the medium was sparged at 3 sL/h with N2/CO2 (80/20). The temperature was set to 37 °C, the medium was stirred at 150 rpm, the pH was set to 7 and kept stable by connecting 1 M KOH as base and 1 M HCl as acid. When these conditions were achieved, the fermentors were operated overnight without inoculation to check sterility as a negative control. Optical density was measured and samples were taken for 16S rRNA gene amplicon sequencing to confirm that no growth had taken place during this phase (data not shown). After sterility was confirmed, 72 mL of inoculum was added to the triplicate fermentors (F1-F3). The inoculum was prepared by proportionally mixing precultures based on the increase of their optical density (∆OD600nm). All replicate fermentors were inoculated with the same inoculum. Then, the fermentors were operated in batch for 6 h. After the batch phase, the feed was switched on to pump fresh medium at a rate of 20 mL/h. The feed originated from a tank containing 10 L of sterile anoxic medium that was kept anoxic by sparging it at 6 sL/h with N2. The waste outflow of the fermentors was set to pump at 30 mL/h at the level corresponding to a medium volume of 480 mL, so that the medium volume never exceeded 480 mL and the medium was completely refreshed in 24 h. Samples for OD600nm measurements, 16S rRNA gene amplicon sequencing and qPCR, and HPLC were taken at t = 0 h, t = 6 h, t = 24 h, and every 24 h thereafter. Samples for metaproteomics were taken at t = 72 h, t = 96 h and t = 120 h, and samples for analysis of mucin glycans were taken at t = 0 h, t = 96 h and t = 120 h. Figure 9 presents an overview of the timeline of the experiment. Experimental overview. The inoculum was prepared from fresh precultures and used to inoculate the triplicate fermentors (F1-F3) at t = 0 h. From t = 0 h–6 h, the fermentors were operated in batch mode. From t = 6 h onwards, fermentors were operated in continuous mode and sampled every 24 h for community composition (16S rRNA gene amplicon sequencing and qPCR) and metabolite production (HPLC). During the stable phase (t = 72 h–120 h), additional samples were taken to assess protein production (metaproteomics) and mucin O-glycan degradation (PGC-LC–MS/MS). 16S rRNA gene amplicon sequencing To determine the relative microbial composition, 16S rRNA amplicon sequencing was applied. First, genomic DNA was isolated from pelleted samples using the FastDNA SPIN Kit for Soil according to the manufacturer’s instructions (MP Biomedicals, Irvine, CA, USA). The isolated DNA was diluted to 20 ng/μL. The V4 region of the 16S rRNA gene was amplified in triplicates by PCR using sample-specific barcoded primers (forward primer: 515-F 5′-GTGYCAGCMGCCGCGGTAA (Parada et al. 2016), reverse primer: 806-R 5′-GGACTACNVGGGTWTCTAAT (Apprill et al. 2015)). Each reaction contained 0.35 μL Phusion Hot start II DNA polymerase (2 U/μL), 0.7 μL of the primer set (10 μM), 0.7 μL dNTPs (10 mM), 0.7 μL template DNA (20 ng/μL) and 25.5 μL nuclease-free water. The PCR protocol was as follows: the initial step was at 98 °C for 30 s, followed by 25 cycles of 98 °C for 10 s, 50 °C for 10 s and 72 °C for 10 s and the final step was 72 °C for 7 min. The 291 bp PCR products were verified on a 1% agarose gel. The triplicates were then pooled and purified using CleanPCR beads (CleanNA, Waddinxveen, The Netherlands) according to the manufacturer’s instructions. The DNA concentrations of the purified PCR products were measured by Qubit dsDNA BR assay (Invitrogen, Waltham, MA, USA) and pooled equimolarly. Pooling of the samples was done by combining 200 ng of each sample in one tube. Finally, the library was concentrated to 50 μL with CleanPCR beads (CleanNA) and sent to Novogene (Cambridge, UK) for 16S rRNA gene amplicon sequencing on NovaSeq 6000 (Illumina). The library included several negative controls: samples from each fermentor at the start of operation, samples from each fermentor after overnight operation shortly before inoculation, two negative controls for DNA isolation and one PCR negative control. Additionally, the library included several positive controls: two mocks of known composition and five mocks of known composition that specifically include strains used in this experiment. qPCR The total absolute bacterial abundance was determined by qPCR of the 16S rRNA gene. Furthermore, as 16S rRNA gene amplicon sequencing yielded reliable results up to genus level and not to species level, we applied species-specific qPCR to determine the ratio between B. caccae, B. fragilis, B. thetaiotaomicron and P. vulgatus. For Ruminococcus spp., reliable species-level classification was obtained with 16S rRNA gene amplicon sequencing. A standard curve was generated with DNA of known concentration from the relevant strain. For the determination of total bacteria, a standard of A. muciniphila DNA was used. Standards were prepared by extracting DNA from pure cultures using the FastDNA SPIN Kit for Soil (MP Biomedicals). Subsequently, the 16S rRNA gene was amplified by PCR with primers 27-F (5′-AGAGTTTGATCMTGGCTCAG) and 1492-R (5′-TACGGYTACCTTGTTACGACTT) (0.2 μM) by adding 1 μL of template DNA to a 49 μL mix of 0.2 μM 27-F and 1492-R primers, Green Phusion HF Buffer (ThermoFisher Scientific, Waltham, MA, USA), 2 mM dNTPs mix (Promega, Madison, WI, USA) and 1 U Phusion HS II DNA polymerase (ThermoFisher Scientific). The PCR programme was as follows: 98 °C for 10 min, then 35 repetitions of 98 °C for 20 s, 52 °C for 20 s and 72 °C for 45 s, with a final step of 72 °C for 5 min. The 1465 bp product was verified on a 1% agarose gel. PCR products were purified using the GeneJET PCR Purification Kit (ThermoFisher Scientific) and their concentrations were measured fluorometrically using the Qubit dsDNA BR assay (Invitrogen). Copy numbers of the 16S rRNA gene per μL were calculated as follows: the measured DNA concentration was divided by the PCR fragment length of 1,465 bp. This was multiplied by 0.91242424 × 1012. To generate the standard curves, 10× dilutions were made ranging from 108 to 101 copy numbers/μL. For qPCR, the samples were diluted to 1 ng/μL. The qPCR was performed in triplicate. Each reaction contained a 9 μL mix of 4.5 μL Sybr Green Master Mix (Bio-Rad, Hercules, CA, USA) and 0.2 μM primers (Supplementary Table 1), to which 1 μL of the diluted template DNA was added in a 384-wells plate. After sealing with optical sealing tape, the qPCR protocol was run according to Supplementary Table 1 (Bio-Rad CFX384). Species-specific primers did not cross-react with other MDSC species (data not shown). Relative and absolute abundance data processing First, quality of 16S rRNA gene amplicon sequencing results was assessed with FastQC (Andrews 2010). To process the 16S rRNA gene amplicon sequencing data, NG-Tax 2 was used with default settings and a forward and reverse read length of 100 bp (Poncheewin et al. 2020). Taxonomy was assigned with the SILVA 138.1 reference database (Quast et al. 2013). Sequences that were not identified up to genus level were BLASTed to assign the genus. qPCR data were analysed with CFX Manager 3.1 (Bio-Rad). If one of the replicates had a Cq standard deviation of >0.35, the most deviating replicate was removed from the data (Gürtler et al. 2018). The mean SQ value of the replicates was then multiplied by the original DNA concentration of the sample to obtain the copy number/μL in the original sample. When targeting specific species of the Bacteroides genus, the copy number/μL was divided by the number of 16S rRNA gene copies in the genome of the respective species (Supplementary Table 2) to obtain the absolute abundance (cell count). Next, the absolute abundance of all MDSC species was determined by multiplying the total bacteria copy number/μL by the relative abundance of each species and dividing by 16S rRNA gene copy number of that species. This method allows us to study relative abundance based on cell count as in (Jian et al. 2020). Altogether, the combination of 16S rRNA gene amplicon sequencing and qPCR on synthetic community data allowed us to obtain 16S rRNA gene copy number corrected species-level relative abundances. Downstream analyses were performed in R (4.3.1) using the packages phyloseq (McMurdie and Holmes 2013), microbiome Lahti and Shetty 2017) and vegan (Oksanen et al. 2022). To assess differences in community composition between replicates and timepoints, PERMANOVA (permutational multivariate analysis of variance, vegan package) was applied when the assumption of homogeneous dispersion was met. For visualisation, we used ggplot2 (Wickham 2016). HPLC High-performance liquid chromatography (HPLC) was performed to determine liquid metabolite production by MDSC. The supernatant of samples taken from the fermentors was first treated with K4Fe(CN)6 and ZnSO4 according to the Carrez method to avoid that mucin would damage the column (Carrez 1909; Acker et al. 1967). To 500 μL sample, 250 μL cold Carrez I solution (0.1 M K4Fe(CN)6.3H2O) was added and the mixture was vortexed. Next, 250 μL Carrez II solution (0.2 M ZnSO4.7H2O) was added and the mixture was vortexed again. Afterwards, the sample was centrifuged at 21,000 g for 5 min. The supernatant was used for analysis by HPLC. Samples and standards were mixed 1:1 with 0.1 M H2SO4. A Shimadzu LC2030C HPLC (Kyoto, Japan) equipped with a HI-PLEX H column (Agilent, Santa Clara, CA, USA) was used to measure liquid metabolites at a flow rate of 1 mL/min with 0.01 N H2SO4 at 45 °C. Data analysis was performed with Chromeleon (ThermoFisher Scientific). Proteomics For metaproteomic analyses, 2 mL samples from fermentors were taken and stored in LoBind tubes (Eppendorf). To obtain monoculture control samples, equal amounts of cells of each monoculture were combined. Samples were then pelleted at 5,000 g at 4 °C for 2 min to separate pellet and supernatant. The supernatant was removed and proteins in the supernatant were precipitated with a 4× volume of acetone and pelleted at 4,000 g at 4 °C for 2 min, and the precipitate was resuspended in 100 μL Tris–HCl (pH 8) with protease inhibitor (1 tablet of protease inhibitor (ThermoFisher Scientific Pierce Protease Inhibitor Mini Tablets EDTA free) in 10 mL of 10 mM Tris HCl pH 8). The original pellet was washed once with Tris–HCl (pH 8), pelleted again at 5,000 g at 4 °C for 2 min and finally resuspended in 100 μL Tris–HCl (pH 8) with protease inhibitor. Cells were lysed with cup sonicator Q500 (QSonica, Newtown, CT, USA) by sonication for 30 min alternating a 15 s pulse and a 15 s break at 100% amplitude. Protein reduction was performed by adding 10 μL 150 mM DTT (dithiothreitol) to 100 μL sample and incubating the mixture shaking at 45 °C for 30 min. Then, proteins were unfolded by adding 330 μL of 8 M urea in 100 mM Tris–HCl. Directly after, 44 μL 200 mM acrylamide was added for alkylation and the sample was incubated shaking at 21 °C for 30 min. The reaction was stopped by adding 7 μL 10% trifuoroacetic acid (TFA). To each sample, 10 μL of a 50%/50% mix of SpeedBeads 65152105050250 and 45152105050250 (GE Healthcare, Chicago, Il, USA) was added. Additionally, 1248 μL of acetonitrile was added and the bead mixtures were incubated shaking at room temperature for 20 min. Afterwards, the mixtures were placed on a magnetic rack. When the beads were on the magnet, the liquid was removed and the beads were washed with 1 mL 70% ethanol, followed by a wash with 100% acetonitrile. Then, 100 μL of 5 ng/μL sequencing grade trypsin in 50 mM ammonium bicarbonate was added to the beads. After 2 h of incubation shaking at room temperature, another 10 μL of 50 ng/μLtrypsin in ammonium bicarbonate was added. After overnight incubation shaking at room temperature, the digestion was stopped by adding 5 μL 10% TFA. The samples were placed on a magnetic rack and the liquid was transferred to a fresh LoBind tube. Remaining beads were removed by centrifuging at 10,000 g for 5 min and transferring the supernatant to a fresh LoBind tube. A Concentrator Plus (Eppendorf) was used to concentrate the sample to less than 50 μL. 1 mL/L HCOOH was used to adjust volumes to exactly 50 μL. Samples were analysed with nLC–MS/MS (reversed phase nano-LC (Vanquish Neo, ThermoFisher Scientific) and Orbitrap Exploris 480 (ThermoFisher Scientific)) as described before (Feng et al. 2022). The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (Vizcaíno et al. 2016) partner repository with the dataset identifier PXD051481. Analysis of proteomics data The nLC–MS/MS data were analysed with MaxQuant (v2.0.3.0) (Cox and Mann 2008). The reference database consisted of proteomes of MDSC members as downloaded from Uniprot. Where possible, the reference proteome was used (Supplementary Table 3). For M. smithii, we translated the genome that we previously sequenced (Plugge et al. 2024), as for this strain there was no proteome available. Monosaccharide metabolism pathways were analysed as follows: KofamKOALA (Aramaki et al. 2020) was applied to assign KEGG Orthologs (KO-identifiers) to the proteins that were identified by metaproteomics analysis. A protein was considered present if it occurred at least in three samples during the stable phase, so that it was at minimum present in either one fermentor during the three timepoints of the stable phase or in all three fermentors during one timepoint. The same tool was used to obtain a list of KO-identifiers for the proteomes that were used as database in MaxQuant, as a proxy for genomic capacity of the organisms in the community. Then, genetic capacity and expression of key proteins in the monosaccharide metabolism of MDSC could be assessed through the presence or absence of the corresponding KO-numbers in this dataset. For metabolisms that are already included in the KEGG mapper such as butyrate production and sulfate reduction, the KEGG Mapper Color tool (Kanehisa et al. 2022) was applied for similar purposes as described above. Community samples and pooled monoculture samples differed in organism abundances. Therefore, we corrected for species relative abundance through normalisation of expression values to be able to compare community and pooled monoculture samples (Kleiner 2017). This was done by dividing the LFQ of each protein of each species by the total LFQ of that species and multiplying by 1 × 1011. Perseus (v1.6.2.1) (Tyanova et al. 2016) was applied to perform further annotations and generate volcano plots. For volcano plots, normalised LFQ values were log2 transformed and student’s t tests were performed with FDR < 0.05 and S0 = 2. Carbohydrate-active enzymes (CAZymes) were categorised based on GOMF annotation in Perseus and KEGG orthology and manually curated with consultation of Uniprot and the CAZy database (Drula et al. 2022). Mucin O-glycan analysis Analysis of mucin O-glycan degradation was performed through O-glycan release using reductive β-elimination followed by PGC-LC-MS/MS analysis as described in Supplementary Text 2. Supplementary Material Acknowledgments We would like to acknowledge Hauke Smidt and Tom Schonewille for welcoming us in the UNLOCK facilities to perform bioreactor experiments. Furthermore, we would like to thank Steven Aalvink, Sara Moreno Paz, Sabine van Oossanen and Reinier Egas for their assistance in running the bioreactors. Contributor Information Maryse D Berkhout, Laboratory of Microbiology, Wageningen University and Research, Stippeneng 4, Wageningen 6708 WE, The Netherlands. Athanasia Ioannou, Laboratory of Microbiology, Wageningen University and Research, Stippeneng 4, Wageningen 6708 WE, The Netherlands. Carol de Ram, Laboratory of Food Chemistry, Wageningen University and Research, Bornse Weilanden 9, Wageningen 6708 WG, The Netherlands. Sjef Boeren, Laboratory of Biochemistry, Wageningen University and Research, Stippeneng 4, Wageningen 6708 WE, The Netherlands. Caroline M Plugge, Laboratory of Microbiology, Wageningen University and Research, Stippeneng 4, Wageningen 6708 WE, The Netherlands. Clara Belzer, Laboratory of Microbiology, Wageningen University and Research, Stippeneng 4, Wageningen 6708 WE, The Netherlands. Author contributions Maryse Berkhout (Conceptualization [equal], Data curation [lead], Formal analysis [lead], Investigation [lead], Methodology [lead], Visualization [lead], Writing—original draft [lead], Writing—review & editing [equal]), Athanasia Ioannou (Data curation [supporting], Formal analysis [supporting], Methodology [supporting], Writing—review & editing [supporting]), Carol de Ram (Formal analysis [supporting], Methodology [supporting], Visualization [supporting], Writing—review & editing [supporting]), Sjef Boeren (Formal analysis [supporting], Methodology [supporting], Writing—review & editing [supporting]), Caroline Plugge (Conceptualization [lead], Funding acquisition [equal], Investigation [equal], Supervision [equal], Writing—review & editing [lead]), and Clara Belzer (Conceptualization [lead], Funding acquisition [lead], Investigation [equal], Project administration [lead], Supervision [lead], Writing—review & editing [lead]). Funding This work was supported by the Netherlands Ministry of Education, Culture and Science and the Dutch Research Council (NWO) through the Soehngen Institute of Anaerobic Microbiology (SIAM) Gravitation Grant [grant number 0.24.002.002]. Conflict of interest statement None declared. Data availability The 16S rRNA gene amplicon sequencing dataset generated and analysed during the current study is available in the ENA repository, under project accession number PRJEB72538. The metaproteomics dataset that was generated and analysed in this study is available in the PRIDE repository under accession number PXD051481. References Associated Data Supplementary Materials Data Availability Statement The 16S rRNA gene amplicon sequencing dataset generated and analysed during the current study is available in the ENA repository, under project accession number PRJEB72538. The metaproteomics dataset that was generated and analysed in this study is available in the PRIDE repository under accession number PXD051481.