Human Gut Faecalibacterium prausnitzii Deploys a Highly Efficient Conserved System To Cross-Feed on β-Mannan-Derived Oligosaccharides

ABSTRACT β-Mannans are hemicelluloses that are abundant in modern diets as components in seed endosperms and common additives in processed food. Currently, the collective understanding of β-mannan saccharification in the human colon is limited to a few keystone species, which presumably liberate low-molecular-weight mannooligosaccharide fragments that become directly available to the surrounding microbial community. Here, we show that a dominant butyrate producer in the human gut, Faecalibacterium prausnitzii, is able to acquire and degrade various β-mannooligosaccharides (β-MOS), which are derived by the primary mannanolytic activity of neighboring gut microbiota. Detailed biochemical analyses of selected protein components from their two β-MOS utilization loci (F. prausnitzii β-MOS utilization loci [FpMULs]) supported a concerted model whereby the imported β-MOS are stepwise disassembled intracellularly by highly adapted enzymes. Coculturing experiments of F. prausnitzii with the primary degraders Bacteroides ovatus and Roseburia intestinalis on polymeric β-mannan resulted in syntrophic growth, thus confirming the high efficiency of the FpMULs’ uptake system. Genomic comparison with human F. prausnitzii strains and analyses of 2,441 public human metagenomes revealed that FpMULs are highly conserved and distributed worldwide. Together, our results provide a significant advance in the knowledge of β-mannan metabolism and the degree to which its degradation is mediated by cross-feeding interactions between prominent beneficial microbes in the human gut.

IMPORTANCE Commensal butyrate-producing bacteria belonging to the Firmicutes phylum are abundant in the human gut and are crucial for maintaining health. Currently, insight is lacking into how they target otherwise indigestible dietary fibers and into the trophic interactions they establish with other glycan degraders in the competitive gut environment. By combining cultivation, genomic, and detailed biochemical analyses, this work reveals the mechanism enabling F. prausnitzii, as a model Ruminococcaceae within Firmicutes, to cross-feed and access b-mannan-derived oligosaccharides released in the gut ecosystem by the action of primary degraders. A comprehensive survey of human gut metagenomes shows that FpMULs are ubiquitous in human populations globally, highlighting the importance of microbial metabolism of b-mannans/b-MOS as a common dietary component. Our findings provide a mechanistic understanding of the b-MOS utilization capability by F. prausnitzii that may be exploited to select dietary formulations specifically boosting this beneficial symbiont, and thus butyrate production, in the gut.
T he human distal gut supports a densely populated microbial community that extends the metabolic capabilities lacking in the host genome (1). In particular, recalcitrant glycans that are resistant to human digestive enzymes are broken down by the colonic microbiota to monosaccharides and further fermented into host-absorbable short-chain fatty acids (SCFAs). Microbial-borne SCFAs serve critical functions both as energy sources and regulators of inflammation, cell proliferation, and apoptosis (2). Therefore, catabolism of complex dietary carbohydrates reaching the distal part of the gastrointestinal tract has a central role in shaping the structure and metabolic output of the human gut microbiota and, in turn, host health status (3).
Members of the Gram-positive Firmicutes and the Gram-negative Bacteroidetes phyla constitute the majority of the bacteria found in this ecosystem (4), individual species of which have evolved different strategies to harvest energy from the available dietary glycans (5). Within the Bacteroidetes, Bacteroides spp. have been extensively investigated with respect to carbohydrate degradation, and they are considered generalists, displaying broad plasticity for glycan utilization (5). Bacteroides spp. are particularly notable for dedicating large proportions of their genome to carbohydrate utilization, organizing genes coding for functionally related carbohydrate active enzymes (CAZymes) and transport and regulatory proteins into polysaccharide utilization loci (PULs) (6). Despite variations in the polysaccharides they target, the key feature of a PUL is the presence of one or more TonB-dependent receptor (SusC homolog) and a contiguous substrate-binding lipoprotein (SusD homolog). Compared with Bacteroides, Firmicutes encode a lower proportional number of CAZymes and are thought to be nutritionally specialized for selected glycans (1,5). Recently, species within the Firmicutes phylum have been shown to organize cohorts of genes encoding glycan utilization systems into loci and being primary degraders of common dietary carbohydrates (7)(8)(9). Firmicutes typically utilize glycan-specific ATP-binding cassette (ABC) transporters, which mediate high-affinity capture of oligosaccharides via their extracellular solute-binding proteins (SBPs) (5).
Faecalibacterium prausnitzii, a member of the Ruminococcaceae family within the Firmicutes phylum, is one of the three most abundant species detected in the human gut microbiota and one of the main sources of butyrate in the colon (10). A growing body of evidence recognizes the crucial role played by F. prausnitzii populations in maintaining local and systemic host health, as they are often found to be less abundant in individuals affected by colorectal cancer (11) and certain forms of inflammatory disorders, including alternating-type irritable bowel syndrome (IBS), inflammatory bowel diseases, celiac disease, obesity and type 2 diabetes, appendicitis, and chronic diarrhea (12,13). In addition, studies in mouse models have demonstrated that both cell and supernatant fractions of F. prausnitzii reduce the severity of acute, chronic, and low-level chemically induced inflammations (14,15). F. prausnitzii also contributes to colonic epithelial homeostasis by stimulating the production of mucin O-glycans and by maintaining appropriate proportions of different cell types of the secretory lineage (16). Collectively, these aforementioned properties make F. prausnitzii a potential novel health-promoting probiotic (17), and interventions aimed at increasing the representation of these butyrate-producing bacteria may be used to confer protection against several intestinal disorders.
A common component of the human diet are b-mannans. These complex plant glycans are found in high concentrations as naturally occurring dietary fibers in certain nuts, beans, legume seeds, tomato seeds, coconut, and coffee beans (18). In addition, mannan hydrocolloids, including guar gum and carob galactomannan (CGM) as well as konjac glucomannan (KGM), are widely used in the food industry to improve the rheological properties of processed products (19). The constant exposure of the gut bacterial community to dietary mannans is consistent with the finding that b-mannan metabolism is one of the core pathways in the human gut microbiota (20). Structurally, b-mannans display source-related diversity with respect to the presence of b-1,4linked mannosyl and glycosyl residues, a-1,6-linked galactosyl groups and acetyl decorations at positions O-2, O-3, and/or O-6 (18). PULs degrading homopolymeric mannan and galactomannan have been described in the glycan generalists Bacteroides fragilis and Bacteroides ovatus, respectively (21,22). We recently reported the characterization of a novel b-mannan utilization locus conferring Roseburia intestinalis, a model for the clostridial cluster XIVa (Lachnospiraceae) Firmicutes, with the ability to ferment this fiber through to butyrate via a selfish mechanism (7). b-Mannan degradation was proven to be initiated by an endo-acting multimodular GH26 enzyme localized on the cell surface; the resulting oligosaccharides are imported intracellularly through a highly specific ABC transporter, and completely depolymerized to their component monosaccharides by an enzymatic cocktail containing carbohydrate esterases, b-glucosidases, and phosphorylases (7).
Although F. prausnitzii has been described as an efficient degrader of host-derived and plant glycans (23), the ability of this important butyrate-producing microbe to utilize dietary b-mannans has received little attention. In a previous study, we reported that wood-derived acetylated galactoglucomannan stimulates the proliferation of F. prausnitzii populations in a pH-controlled batch culture fermentation system inoculated with healthy adult human feces (24). However, the molecular mechanism underlining b-mannan utilization by F. prausnitzii in the human gut has not been explored thus far.
In this study, we describe and biochemically characterize components of two loci that mediate acquisition and catabolism of b-mannooligosaccharides (b-MOS) by the human gut isolate F. prausnitzii SL3/3. Together, these data allowed us to outline a pathway for dietary b-MOS deconstruction and saccharification to monosaccharides through cross-feeding with Bacteroides and Roseburia species, which contributes to the ecology of b-mannan utilization in the gut ecosystem. Remarkably, we show that the binding proteins that confer b-MOS capture in F. prausnitzii targeted ligands with stronger affinity than that of Bacteroides species, thus providing F. prausnitzii with the ability to cross-feed on the b-MOS available in the environment with high efficiency.

RESULTS
F. prausnitzii possess two loci for utilization of b-mannan breakdown products. Based on the similarity with the components from the previously characterized b-mannan utilization system in R. intestinalis (7), genes encoding enzymatic activities required to catabolize mannans were identified within two putative mannan utilization loci (MULs) in F. prausnitzii SL3/3 (Fig. 1a). The large MUL from F. prausnitzii (FpMULL) consists of 14 genes encoding nine enzymes, the components of an ABC transporter, a predicted LacI-type transcriptional regulator (TR), and a hypothetical protein (Fig. 1a). The enzymes encoded by FpMULL include an a-galactosidase belonging to the glycoside hydrolase (GH) family 36 (FpGH36), two carbohydrate esterases (CEs [FpCE2 and FpCE17]), a GH113 (FpGH113), one epimerase (FpMep), a b-1,4-mannooligosaccharide phosphorylase (FpGH130_2), a mannosylglucose phosphorylase (FpGH130_1), a phosphomutase (FpPmm) and a GH1 isomerase (FpGH1). In addition, genes encoding two predicted GH3 b-glucosidases (FpGH3A and FpGH3B) were identified. These two genes are located in a different locus in the genome, hereafter referred to as FpMULS, and are likely to be involved in (galacto)glucomannan turnover. Based on known activities within GH families, the b-1,4-mannan backbone is predicted to be hydrolyzed by extracellular GH26, GH5, and/or GH134 enzymes (see www.cazy.org). However, no gene coding for such an enzyme was identified in the genome of F. prausnitzii SL3/3. In addition, endo-b-1,4-mannanase activity was originally reported for two GH113 enzymes (see www.cazy.org), although we demonstrated that a GH113 within the mannan utilization locus of R. intestinalis is a reducing end mannose-releasing exo-oligomannosidase. A gene encoding a GH113 was detected in the FpMULL (Fig. 1a).
Based on the results of a genomic context analysis and in silico prediction of a signal peptide, FpGH113 would function as an intracellular mannanase or mannosidase; thus, its enzymatic function could not be assigned before an in-depth biochemical characterization (see later results for FpGH113).
Genomic comparisons showed that homologous systems to the FpMULL and FpMULS occur in other sequenced Faecalibacterium members with high percentages of identity (Fig. 1a). Comparison of the gene organization and protein sequence also revealed various levels of rearrangements and moderate protein homology with the two b-mannan utilization loci from R. intestinalis. Examination of the regions flanking the FpMULL of F. prausnitzii SL3/3 showed the presence of genes encoding plasmid mobilization-related proteins, including a cell invasion protein, a relaxase MobA/VirD2, and a DNA ligase (Fig. 1a). Interestingly, R. intestinalis L1-82 genome harbors a similar region, including genes coding for the same plasmid-related components, suggesting that the origin of FpMULL could be the result of horizontal transfer through bacterial conjugation within colonic microbes. Further comparisons revealed that the genes located upstream and downstream of the FpMULL of F. prausnitzii M21/2 and six other sequenced F. prausnitzii strains code for an incomplete prophage, including one or two relaxases and an integrase (Fig. 1a). On the basis of these results, we hypothesize that phage-related horizontal gene transfer was an alternative mechanism for the acquisition of this cluster at the same point in the evolutionary history of these strains. Orthologues of both MULL and MULS, with some rearrangements, were identified in Subdoligranulum variabile CIM:MAG 1127, suggesting that mannan utilization could be a metabolic feature shared with other Ruminococcaceae members.
To further understand the distribution of the two MULs within human-associated F. prausnitzii strains, we surveyed the publicly available metagenome data from a total of 2,441 individuals from regions with distinct geography (North America, Europe, China, and Japan) and dietary patterns (Fig. 1b). Overall, 26.35% of the subjects harbor the two FpMULs identified in this study, while 37.70% carry the FpMULL, irrespective of the nationality or health state. When examined for frequency within single data sets, different cohorts and nationalities exhibited differing trends. The two FpMULs were most common in the European (up to 40.75% of the subjects), Chinese (up to 48.27%), and Japanese (20.31%) metagenomes, whereas their prevalence was lower in North American (10.68%) metagenomes. Among the two hunter-gatherer populations, the Yanomami and Hadza, we detected the presence of only FpMULL in one Yanomami and two Hadza individuals, indicating that these microbiomes may be able to degrade galactomannan derived from tubers that are part of their diet (25).
F. prausnitzii grows efficiently on b-mannooligosaccharides. Growth studies showed that F. prausnitzii SL3/3 failed to grow on KGM and CGM (Fig. 2a), likely reflecting the absence of a surface b-1,4-endomannanase required to generate suitable b-MOS Depolymerization by Faecalibacterium prausnitzii ® b-MOS for import into the cell. This hypothesis was confirmed by growing F. prausnitzii on both substrates predigested with a GH26 b-1,4-endomannanase from R. intestinalis (Fig. 2a). To assay for oligosaccharide generation and/or uptake, we used high-performance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD) and determined the concentration of b-MOS in the initial and spent supernatant from F. prausnitzii cultures (Fig. 2b). Only polymeric b-mannan was observed in the spent supernatant after growth of the bacterium on KGM and CGM, demonstrating that F. prausnitzii does not display surface b-1,4-endomannanase activity. In contrast, F. prausnitzii was able to take up and utilize CGM-and KGM-derived b-MOS while mannose and mannobiose (M2) are seemingly untouched.
Taken together, these data support the concept that the two MULs are being expressed and the resulting proteins orchestrate the degradation of different b-MOS. To determine the biochemical basis for b-MOS import and deornamentation, the specificity of the b-MOS-binding protein, FpGH36, FpGH113, and the two CEs was determined. A model for catabolism of CGM-and KGM-derived b-MOS is presented in Fig. 3.
FpMOBP is a binding protein specific for b-MOS. In a previous study, we have shown that the binding protein RiMnBP of R. intestinalis is part of an ABC transporter that confers uptake of b-MOS with a degree of polymerization between 3 and 6 (7). The FpMUL harbors the components of an ABC transporter similar to that of R. intestinalis, whereby FpMOBP is a 48-kDa protein sharing 48% identity with RiMnBP. In order to confirm the expected role of FpMOBP in capturing b-MOS, we assessed the binding of the recombinantly produced protein to mannohexaose using isothermal titration calorimetry (ITC). FpMOBP bound to mannohexaose with a K d (dissociation constant) of 189 6 1.4 mM (DG = 25.08 6 0.01 kcal/mol; DH = 219.8 6 0.28 kcal/mol; TDS = 14.7 6 0.14 kcal/mol; n = 0.8; corresponding thermograms are shown in Fig. S1a in the supplemental material). FpMOBP did not show any appreciable binding to cellohexaose (Fig. S1b), demonstrating the specificity of FpMOBP toward mannopyranosyl-linked ligands. Together, these data demonstrate that FpMOBP is part of an ABC transporter specific for b-MOS.
Removal of a-galactosyl and acetyl substitutions from b-MOS. FpGH36 is a predicted intracellularly localized 79-kDa enzyme with two GH36 domains, located at the N and C termini of the protein, as well as an internal melibiase domain. The F. prausnitzii GH36 domains were all similar to those found in well-characterized a-galactosidases, with AgaB from the thermophilic bacterium Geobacillus stearothermophilus being the closest structurally characterized homolog (44% identity) (27). FpGH36 showed 42% identity to RiGH36 from R. intestinalis (Fig. 1). FpGH36 hydrolyzed a-1,6galactose side chains from CGM-derived b-MOS (Fig. 4c) and Gal 2 Man 5 ( Fig. S1a and b), exhibiting minor activity against polymeric galactomannan (Fig. 4c). This is consistent with the sequential activity of FpGH36 on internalized galacto-b-MOS in vivo.
We have previously shown that in R. intestinalis the complete removal of acetyl substitutions on the b-MOS backbone is achieved through the complementary action of two esterases, where RiCE2 attacks acetyl groups on either the 3-O, 4-O, or 6-O position, while RiCE17 attacks acetyl groups on the 2-O position (28). To explore whether F. prausnitzii employs a similar mechanism, KGM was prehydrolyzed with RiGH26 to generate glucomanno-oligosaccharides (GMOS) that were subsequently incubated Once intracellular, the a-galactosidase GH36 and the reducing end mannose-releasing exo-oligomannosidase GH113 process the galactomannooligosaccharides into galactose, mannose, and mannotriose. The glucomannooligosaccharides are processed by the acetylesterases CE2 and CE17, two b-glucosidases (GH3), and the reducing end mannose-releasing exo-oligomannosidase GH113 into acetate, mannose, glucose, and mannotriose. The mannotriose is hydrolyzed by the b-1,4-mannooligosaccharide phosphorylase GH130_2 into mannose-1phosphate and mannobiose, which is then epimerized to mannosyl-glucose by an epimerase. A b-1,4-mannosylglucose phosphorylase GH130_1 phosphorolyses mannosyl-glucose into mannose-1-phosphate and glucose. The mannose-1-phosphate is converted into mannose-6-phosphate by a mannose phosphate mutase (Pmm) and further isomerized into fructose-6-phosphate by a GH1. This product, together with the other liberated monosaccharides and acetate, enters glycolysis that generates pyruvate, some of which is converted into butyrate. Proteins characterized in this study are indicated by asterisks.  with FpCE2 (32% amino acid sequence identity with RiCE2) and FpCE17 (46% amino acid sequence identity with RiCE17). Matrix-assisted laser desorption ionization2time of flight mass spectrometry (MALDI-ToF MS) analysis of products released from GMOS revealed that the two enzymes mediated the complete removal of acetylations when added together, while a partial deacetylation was observed when the substrate was treated with each of the enzymes separately (Fig. 4d). To explore the extent to which this strategy for complete substrate deacetylation is conserved in Firmicutes, we exploited the transacetylation specificity of R. intestinalis esterases (28) to generate acetylated mannohexaoses (AcM 6 ) and tested the activity of FpCE2 and FpCE17 on these substrates. FpCE2 was able to deacetylate only the RiCE2-generated AcM 6 , thus demonstrating that this enzyme removes 3-O-, 4-O-, and 6-O-acetylations (Fig. 4e). FpCE17 was effective on RiCE17-generated AcM 6 and displayed no activity on RiCE2-generated AcM 6 , thus showing that FpCE17 exclusively removes the axially oriented 2-O-acetylations (Fig. 4e). Taken together, these results prove that the F. prausnitzii esterases have the same acetylation site specificity as their corresponding enzymes in R. intestinalis.
To further characterize the two F. prausnitzii esterases, we evaluated their activity both on a commercial substrate, i.e., para-nitrophenyl (pNP) acetate, and on a natural substrate, i.e., RiGH26 hydrolyzed acetylated galactoglucomannan (AcGGM). When tested on pNP acetate, both FpCE2 and FpCE17 were most active at pH 6.75 (Fig. 4f). Deacetylation rate measurements on RiGH26-prehydrolyzed AcGGM at pH 6.75 and 35°C, conditions that prevent acetyl migration, at equal enzyme loadings (50 nM) indicated that FpCE2 releases acetate approximately four times faster than FpCE17 (Table 1). When combined, using 25 nM of each esterase, the deacetylation rate, k cat , and specific activity were approximately two-fold higher than the values from treatments with FpCE17 and two-fold lower than the values from treatments with the FpCE2 when used on its own, respectively ( Table 1).The reduced resulting rate of deacetylation suggests that the esterases are not acting synergistically but may rather be competing for the substrate, a behavior previously reported in cocktails of multiple enzymes for lignocellulose hydrolysis (29).
Melting curves for both enzymes in buffers at pH 5.0 to 8.0 were obtained using a protein thermal shift assay ( Fig. 4g and h). Both FpCE2 and FpCE17 displayed an irreversible thermal unfolding transition, which is consistent with their multidomain structure (28,30). FpCE17 was stable up to 73°C, with the highest observed melting temperature at pH 6.0; its lowest observed melting temperature was 58°C at pH 5. For FpCE2, the unfolding took place at higher temperature, with a melting point of 62°C at pH 6.0 and a highest melting point of 73°C at pH 7.0 and 8.0 (Fig. 4g and h); its lowest observed melting temperature was 56°C at pH 5.0.
Studies of substrate specificities have shown that acetyl esterases are able to efficiently catalyze the transfer of an acetyl group from a donor, such as vinyl acetate, to precise positions of an oligosaccharide with the generation of highly specific esterified   (28). Consistent with that notion, we found that both F. prausnitzii esterases were able to transacetylate mannotriose (M 3 ) and mannotetraose (data not shown). To further test the preferred substrate for the esterases, we incubated either FpCE17 or FpCE2 with a mix of M 3 , cellotetraose (G 4 ), and xylopentaose (X 5 ) and used vinyl acetate as an acetyl donor. MALDI-ToF MS analysis of products generated by these reactions showed that the esterases transferred the acetyl group only to M 3 (Fig. 4i), thus confirming the manno-oligosaccharide specificity of FpCE17 and FpCE2. Cocultivation of F. prausnitzii with primary b-mannan degraders. The data presented above suggest that F. prausnitzii has a sufficiently complex enzymatic toolbox to benefit from the uptake of b-MOS liberated in the surrounding environment by other gut microbes. To test this hypothesis and evaluate the competitiveness of this strain in the utilization of b-MOS, we cocultured F. prausnitzii with two keystone commensal organisms for b-mannan utilization, namely, the Gram-negative Bacteroidetes B. ovatus strain V975 and the Gram-positive Firmicutes R. intestinalis strain L1-82. F. prausnitzii grew in the presence but showed poor growth in the absence of B. ovatus on intact KGM (Fig. 5a). The optical densities obtained when F. prausnitzii was grown in monoculture in the no-carbon source control (Fig. 5b) were similar to those obtained in KGM, suggesting that the microbe is not able to utilize this glycan on its own. Notably, the maximum optical density at 650 nm (OD 650 ) of the coculture (OD 650 = 0.55) appeared higher in the b-mannan polymer than those observed for B. ovatus in single culture (OD 650 = 0.49), indicating that syntrophic growth exists between these two populations under these conditions (Fig. 5a). The two strains showed comparable growth on glucose (Fig. 5c). F. prausnitzii is a butyrate producer, while carbohydrate fermentation by B. ovatus results in the production of propionate (5). Therefore, comparing differences in butyrate levels between the single F. prausnitzii culture and coculture may provide evidence as to whether cross-feeding of b-mannan breakdown products by F. prausnitzii occurred. Butyrate concentrations were significantly increased (P = 0.001) in the coculture compared to the monococulture in KGM (Fig. 5d), the coculture in minimal medium with no carbon source (Fig. 5e) or glucose (Fig. 5f), which suggests that F. prausnitzii can effectively compete for b-MOS generated by the cell-surface-exposed endo-mannanase BoMan26B from B. ovatus (31). This effect required the presence of living B. ovatus cells, as no evidence of an increase of butyrate levels was detected when F. prausnitzii was cogrown with a heat-treated B. ovatus culture (Fig. S3). Population estimates using quantitative PCR (qPCR) showed that, while no growth was observed when F. prausnitzii was cultivated singly on KGM, it grew well in the coculture supplemented with KGM (Fig. 5g), thus indicating cross-feeding activity by F. prausnitzii. Both F. prausnitzii and B. ovatus displayed similar growth on glucose (Fig. 5h), suggesting that the bacteria shared the available carbon source and maintained coexistence. When F. prausnitzii was cocultured with R. intestinalis in KGM, in the absence of a carbon source or in glucose, the growth curves appeared very similar to when R. intestinalis was cultured on its own (Fig. 5i to k). As both bacteria produce butyrate, we compared the value observed in the coculture to the sum of butyrate concentration in both single cultures. No significant increase (P . 0.05) of butyrate concentrations was observed in the coculture compared to the single cultures in KGM (Fig. 5l), minimal medium (Fig. 5m), or minimal medium plus glucose (Fig. 5n). Intriguingly, while no growth was observed in monococulture in KGM, F. prausnitzii displayed growth in coculture with R. intestinalis (Fig. 5o), thus demonstrating cross-feeding behaviors. When growing in coculture on glucose, R. intestinalis dominated the culture and outcompeted F. prausnitzii (Fig. 5p). Notably, cocultivation of F. prausnitzii with either B. ovatus (Fig. 5b) or R. intestinalis (Fig. 5j) without any carbon source resulted in no increase in the overall levels of butyrate in the cocultures (Fig. 5e and m). These data indicate the specific effect of KGM degradation products to support F. prausnitzii growth and exclude the possibility that this microbe is cross-feeding on bacterial-derived components (such as capsular polysaccharides).

DISCUSSION
Biochemical work presented herein demonstrates that two MULs support the ability of F. prausnitzii to utilize b-MOS. b-MOS from diet are highly variable with respect to sugar composition and linkages, and the F. prausnitzii enzymatic apparatus is adapted to deal with this diversity. Our findings show that b-MOS are bound by FpMOBP at the cell surface and subsequently imported intracellularly; here, they are further saccharified by FpGH113 and degalactosylated and deacetylated by the combined action of FpGH36, FpCE2, and FpCE17 (Fig. 3). By analogy with the model described for R. intestinalis (7), putative b-glucosidases of GH3 may confer the removal of terminal glucose residues in gluco-b-MOS prior to depolymerization of the remaining linear b-MOS by the activity of a putative mannooligosaccharide phosphorylase (FpGH130_2) into mannobiose. Mannobiose is subsequently epimerized into mannosyl-glucose by a putative epimerase, FpMep, and phosphorolysed by FpGH130_1 into glucose and mannose-1-phosphate, similar to the pathway described for Ruminococcus albus (32). As shown for R. intestinalis In panels g, h, o, and p, the number of 16S rRNA gene copies per milliliter of culture, obtained using qPCR, at the start (T0) and end (T48) of the experimental run were expressed as T48/T0 to obtain fold changes which were then log 2 transformed. In all panels, the data are means with standard deviations of a minimum of three replicates. For calculation of SCFA concentrations, the values measured in uninoculated M2, glucose, and KGM media were subtracted from the sample values. Statistically significant differences for butyrate concentration were determined with the Mann-Whitney test using SCFA data. b-MOS Depolymerization by Faecalibacterium prausnitzii ® (7), the end products of this pathway enter glycolysis either directly (glucose) or after being converted to mannose-6-phosphate and fructose-6-phosphate by the phosphomannose mutase FpPmm and the isomerase FpGH1, respectively. Released galactose and mannose are converted to galactose-1-phosphate (via the Leloir pathway) and mannose-6-phosphate (by a hexokinase) before entering glycolysis (7). The pyruvate generated from glycolysis is then partly converted to butyrate (33). A comparative genomic analysis revealed that the FpMULs are widespread and highly conserved among human gut-associated F. prausnitzii (Fig. 1a). The presence of genes associated with conjugation and phage-related events in the flanking regions suggests that the FpMULL was acquired through horizontal gene transfer from other gut bacteria, as previously observed for PULs identified in commensal Bacteroides genomes (34).
Members of the dominant Bacteroides genus, such as B. ovatus, as well as Firmicutes species, like Roseburia, that possess GH26 endomannanases have been described as the keystone bacteria for b-mannan degradation in the gut (7,21,31). In contrast, F. prausnitzii may only access oligosaccharides, released by these primary degraders, which can be imported without the need for extracellular enzymatic cleavage. In this context, we demonstrate that b-MOS are indeed released into the culture medium by B. ovatus during cogrowth on KGM and that F. prausnitzii is capable of efficiently competing for and utilizing these oligosaccharides (Fig. 5g). Similarly, we observed evidence of cooperative growth between R. intestinalis and F. prausnitzii when cocultured in KGM (Fig. 5o). We recently demonstrated the competitiveness of R. intestinalis on b-mannan when in coculture with B. ovatus during growth on AcGGM (7), and highlighted a pivotal role of a transport protein (RiMnBP) within the uptake system, which exhibited strong binding to short b-MOS (degree of polymerization [DP], 3 to 6) with different side chain decoration patterns (7). Similar observations were recently reported for the Bifidobacterium genus, with B. animalis subsp. lactis ATCC 27673 outcompeting B. ovatus during growth on galactomannan (35). Notably, the two B. animalis binding proteins, BlMnBP2 and BlMnBP1, mediated high-affinity capture of b-MOS with preference to oligosaccharides with a DP of 3 to 4 and K d values in the 70 to 80 nM range (35). In contrast, the SusD-like b-MOS-binding proteins from B. ovatus displayed binding to M 6 (K d value of 1.8 6 0.2 mM) (31) with about 10-fold lower affinity than that of FpMOBP, 53-fold lower affinity than that of RiMnBP, and about 900-fold lower affinity than that of BlMnBP2 and BlMnBP1. The results shown in Fig. 5g and o show that the growth of F. prausnitzii was much more pronounced when in cocultures with B. ovatus (Fig. 5g) (with an average 16S rRNA gene copies per ml of culture of 5.91E18 F. prausnitzii and 1.34E109 B. ovatus, corresponding to a ratio of 2.3:1 B. ovatus-F. prausnitzii) than in F. prausnitzii-R. intestinalis cocultures (Fig. 5o) (with an average 16S rRNA gene copies per ml of culture of 1.24E18 F. prausnitzii and 6.94E18 R. intestinalis, corresponding to a ratio of 5.6:1 R. intestinalis-F. prausnitzii). This is consistent with the fact that F. prausnitzii FpMOBP is more efficient in capturing oligosaccharides from the "weaker" binding of SusD-like b-MOS-binding protein of B. ovatus than from the R. intestinalis RiMnBP. Thus, the differential transporter affinity to b-MOS provides a possible rationale for the trophic interactions established by F. prausnitzii with Bacteroides, enabling efficient capture of communally available nutrients within synthetic consortia, and potentially in natural gut communities. On the other hand, the binding affinity of MnBPs plausibly gives a reason as to why cross-feeding on KGM was present to a lower extent in the R. intestinalis and F. prausnitzii coculture and why the affinity of RiMnBP to M 6 (K d value of 33.75 6 0.95 mM) was five times stronger than that of FpMOBP. This difference is likely to be crucial for selfish resource capture by the keystone b-mannan primary degrader R. intestinalis (7) with minimal loss of valuable nutrients in a competitive environment.
To understand the capacity of human populations to derive nutrition from b-MOS, we surveyed 2,441 publicly available human metagenomes and revealed that MULs closely related to those of F. prausnitzii are widely distributed throughout human populations (Fig. 1b). Indeed, we did not observe correlation with any particular population or nation, consistent with the fact that dietary b-MOS/b-mannan are a ubiquitous component of the human diet. The FpMULs were more common than population restricted traits like red algal porphyran degradation, known to be confined to a small cohort of Japanese subjects and absent in the microbiome of Western individuals (34), but they were less common than Bacteroides-associated PULs for degradation of plant cell wall xyloglucan (92% of samples) (36), mixed-linkage b-glucans (92.5% of samples) (37), b-(1, 3)-glucans (59% of samples) (38), and yeast a-mannans (62% of samples) (39). Moving beyond the human microbiota, we detected two analogous FpMULs in a F. prausnitzii strain found in the porcine gut microbiota (40). Proteomic analysis identified FpMUL-encoded proteins being more abundant in pigs fed a diet supplemented with 4% AcGGM, thus providing evidence that these analogous MULs are employed by F. prausnitzii inhabiting environments beyond the human gut (40).
In conclusion, biochemical and microbiological data presented in this study illustrate that F. prausnitzii possesses an extensive enzymatic apparatus that targets b-MOS released by neighboring colonic bacteria. ITC data provide evidence that the external recognition machinery is tailored for the capture of b-MOS with stronger affinity than Bacteroides. This is in line with the fact that, when in coculture, F. prausnitzii showed cross-feeding behaviors with B. ovatus, whose own b-MOS uptake requires a SusD-like protein that binds oligosaccharides with about 10-fold-lower affinity than that of FpMOBP. Furthermore, this study in conjunction with a previous report (7) points to a competitive mechanism of b-MOS/b-mannan utilization in the gut microbiota where keystone Lachnospiraceae members like R. intestinalis have developed a sophisticated "selfish" uptake and degrading system to minimize sharing resources with Bacteroides and Ruminococcaceae species such as F. prausnitzii.
Overall, our study contributes toward the understanding of cross-feeding mechanisms deployed by a beneficial commensal microorganism to interact with dietary b-mannan. Significantly, these findings could help to design prebiotic/symbiotic formulations that are optimized for selective manipulation of gut microbiome functions in ways that promote human health and beyond.

MATERIALS AND METHODS
Substrates. All glycan stocks were prepared at 10 mg/ml in double-distilled water (ddH 2 O) and sterilized by filtration using a 0.22-mm membrane filter (Sarstedt AG & Co., Germany).
Bacterial strains and culture conditions. Faecalibacterium prausnitzii SL3/3, Roseburia intestinalis L1-82, and Bacteroides ovatus V975 were routinely cultured with CO 2 at 37°C in M2 medium containing 30% clarified rumen fluid supplemented with 0.2% (wt/vol) glucose, soluble potato starch, and cellobiose (GSC) (41). These bacterial isolates were from stocks held by the authors (S. H. Duncan, Rowett Institute of Nutrition and Health, Aberdeen, United Kingdom), and all are of human fecal origin (23,42,43). Growth measurements on individual substrates were performed in M2GSC medium containing a single carbohydrate at 0.2% (wt/vol) final concentration using 96-well plates in a Don Whitley MACS-VA500 workstation (80% N 2 , 10% H 2 , and 10% CO 2 ). Growth was assessed by measuring the absorbance at 650 nm (optical density at 650 nm [OD 650 ]) at 2-h intervals for up to 24 h using an Epoch 2 microplate reader (BioTek, VT, USA). The competition assays of F. prausnitzii and either B. ovatus or R. intestinalis were conducted by growing the strains as described above in the presence of 0.2% (wt/vol) konjac glucomannan. The strains inoculated into M2 medium with no added carbohydrate source and M2 medium with 0.2% (wt/vol) glucose were included as negative and positive controls, respectively. Five-microliter samples of overnight bacterial cultures from both strains were used to inoculate the wells (final volume of 200 ml). The cocultures were incubated at 37°C anaerobically, and growth was followed by measuring the OD 650 for 24 to 48 h. Samples were collected at the end of the experiment, cells were pelleted for subsequent DNA extraction, and the supernatant (500 ml) was used for SCFA analysis. All growth experiments were performed in triplicate. b-MOS Depolymerization by Faecalibacterium prausnitzii ® DNA extraction and quantitative PCR. After the growth experiments involving single cultures and cocultures were completed, cell pellets were obtained from triplicate wells for each condition tested (600 ml) and the bacterial inoculum (600 ml). The cell pellets were resuspended in 978 ml sodium phosphate buffer and 122 ml MT buffer and stored at -70°C until DNA extraction, which was performed using the FastDNA SPIN kit for Soil (MP Biomedicals, Irvine, CA, USA) following the manufacturer's instructions. The extracted DNA from the mono-and coculture experiments was analyzed using quantitative PCR as described by Chung et al. (44) with some modifications. Herring sperm (HS) DNA (5 ng/ml) (Promega, Madison, WI, USA) was used to dilute the samples and 16S PCR products of reference strains to obtain a 10-fold dilution series for standard curves. A master mix containing iTaq Universal SYBR green Supermix (Bio-Rad, Hemel Hempstead, UK), 10 mM primer pairs (500 nM final concentration), and molecular biology grade water was prepared. The master mix (8 ml) together with 2 ml of either a diluted sample (1 ng/ml) or standard, giving a total volume of 10 ml, was distributed into the wells of 384-well plates. The plates were then sealed with optical seals (Bio-Rad). Samples were amplified with universal primers (UniF and UniR) against total bacteria and specific primers targeting Roseburia spp. (44), Bacteroides spp. (44), and F. prausnitzii. Specific primers against F. prausnitzii (100.85% 6 1.53% efficiency) were designed as part of this study (see Table S2 in the supplemental material). The Bio-Rad CFX384 real-time system was used for the amplification of samples and standards using the protocol described previously with annealing temperatures appropriate for each primer set (Table S2). The reactions were performed in duplicate, and the number of 16S rRNA gene copies per milliliter of culture for each bacterial strain in the mixes tested were determined using information from the standard curves. Negative controls containing only herring sperm (HS) DNA were used to determine the detection limit.
Bioinformatics, cloning, expression, and purification of recombinant proteins. F. prausnitzii SL3/ 3's FpMULL and FpMULS were identified by tBLASTn-based homology searches using the amino acid sequences of the proteins from the previously characterized R. intestinalis b-mannan degradation system as the query (7). The genes encoding mature forms of the proteins described in this study were amplified from the F. prausnitzii SL3/3 genomic DNA (BioProject accession number PRJNA39151) by PCR using appropriate primers (Table S1). All primers were designed to amplify constructs to exclude predicted signal peptides (predicted by the SignalP v4.1 server [45]). PCR products were generated using the Q5 High-Fidelity DNA polymerase (New England BioLabs, United Kingdom) with 50 ng genomic DNA as the template. The PCR products were cloned into pNIC-CH (Addgene plasmid 26117) by ligation-independent cloning (46). All constructs were designed to harbor a C-terminal His 6 tag fusion in the translated recombinant peptide, although for FpGH36, His tag translation was prevented by the introduction of one stop codon at the end of the open reading frame. Successful generation of constructs was verified by sequencing (Eurofins, UK). Plasmids harboring the gene of interest were transformed into chemically competent Escherichia coli BL21 STAR cells (Invitrogen), and an overnight preculture was inoculated to 1% in 500 ml tryptone yeast extract (TYG) containing 50 mg/ml kanamycin, followed by incubation of the fresh culture for 16 h at 25°C. Protein overexpression was induced by adding isopropyl b-D-thiogalactopyranoside (IPTG) to a final concentration of 200 mM. Recombinant protein production continued overnight at 25°C, after which the cells were collected by centrifugation. FpCE17, FpCE2, FpGH113, and FpMOBP were purified by immobilized metal ion affinity chromatography (IMAC). To this aim, the harvested cell pellet was resuspended in binding buffer (20 mM sodium phosphate [pH 7.4], 500 mM sodium chloride, 5 mM imidazole) and lysed using a Vibracell ultrasonic homogenizer (Sonics and Materials, USA). The cell debris was pelleted by centrifugation, and the supernatant was loaded onto a 5-ml HisTrap IMAC HP nickel Sepharose column (GE Healthcare), using an ÄKTA Pure chromatography system (GE Healthcare). The target His-tagged protein was eluted using a linear gradient of 0 to 100% elution buffer (20 mM sodium phosphate [pH 7.4], 500 mM sodium chloride, 500 mM imidazole) over 16 column volumes. FpGH36 was purified by hydrophobic interaction chromatography (HIC). FpGH36-containing cell pellet was resuspended in a buffer with 1.5 M ammonium sulfate and lysed as described above. The cell-free supernatant was loaded onto a 5-ml HiTrap Phenyl FF (GE Healthcare), and protein was eluted by using a linear reverse gradient to 100 mM NaCl over 90 min at a flow rate of 2.5 ml/min. After IMAC or HIC purification, eluted protein fractions were pooled, concentrated using a Vivaspin 20 centrifugal concentrator (10-kDa molecular weight cutoff), and applied to a HiLoad 16/600 Superdex 75 pg gel filtration column (GE Healthcare). Pure protein samples were dialyzed against 10 mM Tris-HCl (pH 7.0) and concentrated as described above. Protein purity was determined by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis. Protein concentrations were determined using the Bradford assay (Bio-Rad, Germany).
Activity assays. Unless otherwise stated, enzyme reaction mixtures contained 10 mM sodium phosphate (pH 5.8) and 0.1 mg/ml substrate. Reaction mixtures were preheated (37°C for 10 min) in a Thermomixer C incubator with a heated lid (Eppendorf), before addition of the enzyme to 1 mM (in a final reaction mixture volume of 100 ml) for further incubation (up to 16 h) at 37°C and 700 rpm. All experiments were performed in triplicate.
Carbohydrate analysis using high-performance anion-exchange chromatography. Oligo-and monosaccharides were analyzed by high-performance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD) on a Dionex ICS-3000 system operated by Chromeleon software version 7 (Dionex). Sugars were loaded onto a CarboPac PA1 2 Â 250-mm analytical column (Dionex, Thermo Scientific) coupled to a CarboPac PA1 2 Â 50-mm guard column kept at 30°C. Depending on the analytes, the following gradients were used. The system was run at a flow rate of 0.25 ml/min. For manno-oligosaccharides, the elution conditions were as follows: for 0 to 9 min, 0.1 M NaOH; for 9 to 35 min, 0.1 M NaOH with a 0.1 to 0.3 M sodium acetate (NaOAc) gradient; for 35 to 40 min, 0.1 M NaOH with 0.3 M NaOAc, and for 40 to 50 min 0.1 M NaOH. Commercial mannose and manno-oligosaccharides (DP, 2 to 6) were used as external standards.
Acetate release measurements using high-performance liquid chromatography. Acetate content in the samples was analyzed on an RSLC Ultimate 3000 (Dionex, USA) high-performance liquid chromatograph (HPLC) using a REZEX ROA-Organic Acid H1 300 Â 7.8-mm ion exclusion column (Phenomenex, USA). The injection volume was 5 ml, and separation was conducted at 65°C, with isocratic elution using 0.6 ml/min of 5 mM H 2 SO 4 as the mobile phase. The UV detector was set at 210 nm. Data collection and analysis were carried out with the Chromeleon 7.0 software (Dionex).
SCFA analysis. SCFA concentrations were measured using a gas chromatograph analyzer equipped with a flame ionization detector (GC-FID) as described previously (47). Following derivatization of the samples using N-tertbutyldimethylsilyl-N-methyltrifluoroacetamide, the samples were analyzed on a Hewlett-Packard 6890 gas chromatograph equipped with a silica capillary column using helium as the carrier gas. Quantification of SCFA in the chromatograms was determined based on the retention times of the respective SCFA standards (Sigma-Aldrich, United Kingdom) at concentrations ranging between 5 and 30 mM.
Isothermal titration calorimetry. Binding of mannohexaose and cellohexaose to FpMOBP was measured at 25°C in 50 mM sodium phosphate (pH 6.5) using either a MicroCal ITC 200 microcalorimeter or a MicroCal VP-ITC system. To assess the binding to mannohexaose using a MicroCal ITC 200 microcalorimeter, FpMOBP in the sample cell (2.5 mM) was titrated by a first injection of 0.5 ml followed by 19 2-ml injections of carbohydrate ligand (2.5 mM) with 120 s between injections. To evaluate the binding to cellohexaose using a MicroCal VP-ITC system, FpMOBP in the sample cell (22.5 mM) was titrated by a first injection of 2 ml followed by 29 6-ml injections of carbohydrate ligand (2.5 mM) with 180 s between injections. Thermodynamic binding parameters were determined using either the MicroCal Origin software (version 7.0) or the VPviewer2000 software (version 2.6).
FpCE2 and FpCE17 optimal pH. The pH optima for FpCE17 and FpCE2 were assessed by incubation of the enzymes with 0.5 mM 4-nitrophenyl (pNP) acetate (Sigma-Aldrich, Germany) at 25°C using 50 mM sodium phosphate buffer at pHs ranging from 5.0 to 8.0. Due to the difference in deacetylation rate of pNP acetate by the two enzymes, 1 nM FpCE2 and 0.1 mM FpCE17 were used in these experiments. Standard plots of 4-nitrophenol (p-nitrophenol; Sigma-Aldrich) were prepared at each pH. The experiments were conducted in triplicate in a volume of 100 ml of sample mixture in 96-well microtiter plates. The reaction was followed by measuring the absorbance at 405 nm at 1-min intervals for 10 min using a Microplate reader (BioTek, USA).
Protein thermal shift assay. The thermal stability of FpCE17 and FpCE2 was examined using the Protein Thermal Shift kit (ThermoFisher, USA) by measuring fluorescence in a real-time PCR system (Applied Biosystems, USA). A final concentration of 0.1 mg/ml of FpCE17 and FpCE2 was used in 50 mM sodium phosphate buffers at pH 5.0 to 8.0 and mixed with ROX dye according to the kit protocol. The melting temperature was executed in four replicates with temperatures from 25 to 99°C in 1% increments. The data were processed using the StepOne software (Applied Biosystems, USA).
Transesterification reactions. Transesterification of oligosaccharides was conducted using vinyl acetate (Thermo Scientific, USA) as acetate donors. Enzymes (1 mg/ml final concentration) were mixed with 1 mg/ml oligosaccharides, and a volume of vinyl acetate corresponding to 50% of the sample volume was added. The samples were incubated in a thermomixer (Eppendorf, Norway), shaking at 600 rpm, at ambient temperature overnight, then kept at -20°C until frozen. The vinyl acetate, which remained in liquid phase on top of the frozen aqueous phase, was discarded; enzyme deactivation and carbohydrate precipitation were achieved by adjusting the aqueous phase to 80% (vol/vol) ethanol with ice-cold 96% ethanol. Enzymes were removed through filtration using a 1-ml Amicon Ultracel 3-kDa ultrafiltration device (Merck KGaA, Germany). The samples were then dried using an Eppendorf Concentrator plus (Eppendorf, Norway) at 30°C, and the material was dissolved in 100 ml dH 2 O.
Comparative genomics analysis. Searches for the presence of MULL and MULS in other publicly available F. prausnitzii genomes were conducted using a similar strategy as described previously (7). Briefly, the identification of similar MULs in strains other than F. prausnitzii SL3/3 was done using BLASTN and the Gene Ortholog Neighborhood viewer on the Integrated Microbial Genomes website (https://img.jgi.doe.gov) using the sequences of the genes coding for FpMOBP (FPR_17280), FpGH113 (FPR_17310), and FpGH3 (FPR_09740) as the search homolog and the default threshold E value of 1e25. If this generated a hit, we repeated the process with the adjacent gene to verify that the locus was found in the identified strain. Then, the amino acid identities between each F. prausnitzii SL3/3 MULL-MULS RefSeq annotated protein and the hits identified in other F. prausnitzii strains were determined by BLASTP-based analyses. Finally, we compared the genomic regions surrounding each orthologous MUL for gene conservation and amino acid identities.
Analysis of human gut metagenomic data sets for the presence of MULs. Available cohorts of human gut metagenomic sequence data (National Center for Biotechnology Information projects: PRJNA422434 (48) (58) were searched for the presence of MUL nucleotide sequences from F. prauznitzii MULL (17.5 kb) and F. prauznitzii MULS (5.5 kb) using the following workflow. Each MUL nucleotide b-MOS Depolymerization by Faecalibacterium prausnitzii ® sequence was used separately as a template, and then Magic-BLAST (59) v1.5.0 was used to recruit raw Illumina reads from the available metagenomic data sets with an identity cutoff of 97%. Next, the alignment files were used to generate a coverage map using bedtools (60) v2. 29.0 to calculate the percentage coverage of each sample against each individual reference. We considered a metagenomic data sample to be positive for a particular MUL if it had at least 70% of the corresponding MUL nucleotide sequence covered.
Data availability. All data supporting the findings of this study are available within the article and supplemental material.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.   We declare that we have no competing interests.