Catabolic Network of the Fermentative Gut Bacterium Phocaeicola vulgatus (Phylum Bacteroidota) from a Physiologic-Proteomic Perspective Open Access

Members of the phylum Bacteroidota stand out by their plethora of glycan-sensing, -binding and -hydrolyzing enzymes [1, 4] qualifying them for efficient carbohydrate break down, and rationalizing their prominent abundance in the microbiome of the human gut, in particular in the anoxic colon [5, 7], as well as in the gut of diverse animals [8]. Notably, the capacity of carbohydrate-active enzymes at the disposal of human gut Bacteroidetes is partly expanded by transfer from marine bacteria [9, 10]. Bacteriodetes anaerobically degrade (plant-derived) polysaccharides into short-chain fatty acids (primarily butyrate, acetate, and propanoate) [7], thereby serving not only the host but also driving interspecies metabolic interactions within the gut microbiome [11, 12]. Changes in the composition of the gut microbiome, along with its intricate balancing of diet and fermentation products, can promote health, modulate the immune development, or cause diseases [13, 20]. Next to its nutritional and health services to humans, the gut microbiome can possibly be exploited for enzymes and metabolic processes for a sustainable production of bulk chemicals (e.g., succinate and propane-1,2-diol) required for industrial chemical synthesis [21].

The Bacteroidetes member Phocaeicola vulgatus (formerly Bacteroides vulgatus [22]) is a particularly abundant member of the human gut microbiome [23, 24]. Several studies with P. vulgatus demonstrated its multi-facetted effect on human health and microbial interspecies relations: (i) Overproduction of succinate by, e.g., P. vulgatus increases the probability of disease incidents [17]. (ii) Proteases of P. vulgatus are linked to severity of ulcerative colitis [25]. (iii) Decrease of the P. vulgatus: Lactobacillus spp. ratio correlates with obesity in children [26]. (iv) Release of fucose during extracellular degradation of mucin by P. vulgatus nourishes other gut bacteria [27]. (v) Decomposition of the complex pectic polysaccharide rhamnogalacturonan II by the well-studied gut bacterium Bacteroides thetaiotaomicron releases among others apiose, which in turn is utilized by P. vulgatus [28].

Considering the relevance of P. vulgatus for the metabolic homeostasis in the human gut, only a limited number of studies on its physiology have been reported. Glucose is degraded via glycolysis to acetate and formate; in addition, branching off from phosphoenolpyruvate (PEP) also succinate and subsequently propanoate can be formed [29], as known from other gut bacteria [30]. l-Galactose was proposed to be converted via a 3-step pathway to d-tagaturonate [31], while degradation of d-galactose is apparently unclear. Degradation of pentoses was suggested to proceed via the phosphoketolase pathway [32]. Furthermore, the disaccharides maltose and lactose [33] and the polysaccharide xylan [12] were demonstrated to serve as growth substrates. Pudlo et al. [34], using a custom phenotyping array, observed growth of 33 different P. vulgatus strains with several carbohydrates including fructose, fucose, rhamnose, and mannose. Growth stoichiometry and pathway analyses have previously been reported for Prevotella copri [35, 36], which has recently been reclassified as Segatella copri [37].

The present study aims at advancing the physiological understanding of P. vulgatus to better appreciate its metabolic contributions to the gut microbiome’s modus operandi and to assess its potential for future biotechnological applications. For this purpose, we combined two lines of investigations. First, 14 different growth-supporting carbohydrates with partly unknown degradation routes were selected to determine the substrate-specific range of growth parameters, including product range and yields. Second, the catabolic network underlying these degradation capacities in conjunction with the starch utilization system (Sus)-like and respiratory systems was reconstructed by comprehensive proteogenomic analyses.

The carbohydrates to be investigated were selected to achieve a good balance between the different compound types and natural sources (in particular components of plant-derived polysaccharides), as well as diversity of presumptive degradation pathways: (i) the two pentoses xylose (aldose; monomer of xylan) and arabinose (aldose; component of arabinoxylan); (ii) the six hexoses glucose (aldose; monomer of e.g., cellulose, starch, and α-cyclodextrin), galactose (aldose; monomer of galactan in hemicellulose), fructose (ketose), mannose (aldose; monomer of mannanes in hemicellulose), rhamnose (deoxy sugar; component of heteropolymers), and fucose (deoxy sugar; e.g., component of mucin); (iii) the uronic acid glucuronate (component of, e.g., xanthan); (iv) the three disaccharides maltose (two α1→4-linked glucose moieties), lactose (α1→4-linked galactose and glucose), and sucrose (α,β1→2-linked glucose and fructose); (v) the oligomer α-cyclodextrin (six α1→4-linked glucose moieties); and (vi) the polymer xylan (abundant plant cell wall heteropolysaccharide with d-xylose as main monomer).

To avoid memory effects and to ensure comparability across the 14 selected substrate conditions, P. vulgatus was adapted to anaerobic growth with each of them over five passages. Then, the actual stoichiometric growth experiments were conducted comprising 3 replicates per substrate. In each case, growth was monitored by measuring the optical density (OD) and by determining the total cell count (TCC) and cellular dry weight (CDW). This was paralleled by HPLC-based profiling of substrate depletion and fermentation product formation, as well as by recording the related decrease of pH. Growth curves are illustrated in Figure 1 and online supplementary Figure S1 (for all online suppl. material, see https://doi.org/10.1159/000536327), and the determined growth parameters are compiled in Table 1 and online supplementary Table S1. In the following, results for glucose, rhamnose, xylan, and glucuronate are detailed, as a representative cross-section of the 14 tested substrate conditions with respect to differences in substrate type, growth performance, and fermentation product range. To further facilitate comparability, the compound depletion/formation data are related to their carbon content, i.e., features as mM_Carbon (mM_C). The growth stoichiometric analyses were complemented by energetic perspectives (Table 1; online suppl. Table S1, S2), which were gained by balancing the total of reactions from the fermentation network (see below) that consume or produce ATP and reducing equivalents. Furthermore, potential respiratory energy conservation with reducing equivalents not utilized during formation of fermentation products was considered.

Utilizing glucose (Fig. 1a), P. vulgatus grew with a short doubling time of 1.6 h and reached after ∼9.7 h an OD_max of ∼1.8, which corresponds to 7.7 × 10⁹ cells/mL or ∼0.4 g_CDW/L and reflects yields of 60 mg_CDW/mmol_S and ∼1.9 mM_ATP/mM_S, respectively (Table 1). The fermentation product range was dominated by acetate (14.3 mM_C) and succinate (22.0 mM_C) followed by formate (2.3 mM_C) and very low levels of propanoate (0.4 mM_C). Formation of fermentation products correlated to a drop of pH from 7.3 to 6.5.

Growth with rhamnose (Fig. 1b) occurred with a doubling time of 3.7 h and yielded after 15.8 h an OD_max of ∼0.6, ∼4.3 × 10⁹ cells/mL, ∼0.4 g_CDW/L, and an ATP yield of ∼1.3 mM_ATP/mM_S. Strikingly, the fermentation product range was clearly dominated by propane-1,2-diol (33.8 mM_C), followed by markedly lower levels of acetate (11.0 mM_C) and only traces of succinate and formate. Across the studied substrate conditions, formation of propane-1,2-diol was observed in addition only with fucose (16.1 mM_C), albeit at more balanced shares of formed fermentation products (online suppl. Fig. S1d). Agreeing with the dominance of propane-1,2-diol among the fermentation products upon growth with rhamnose, the pH decreased from 7.3 merely to 6.9.

Application of xylan (Fig. 1c) as growth substrate facilitated a doubling time of 4.6 h and resulted after 16.9 h in an OD_max of ∼0.8, ∼7.9 × 10⁹ cells/mL, ∼0.4 g_CDW/L, and an ATP yield of ∼20.1 mM_ATP/mM_S. The fermentation product range was characterized by similar shares of acetate (11.3 mM_C) and succinate (10.8 mM_C), followed by lower levels of formate and propanoate, and yielded a pH drop from 7.3 to 6.7. Since the depletion of xylan could not be analytically traced, the starting concentration, which is related to the in total provided carbon (90 mM_C), served as basis for calculations; 90 mM_C were uniformly applied for all substrate conditions.

Growth of P. vulgatus with glucuronate (Fig. 1d) was slowest among all tested substrate conditions evidenced by a doubling time of 14.3 h. After 49 h of growth, an OD_max of ∼0.8, ∼9.5 × 10⁹ cells/mL, ∼0.45 g_CDW/L, and an ATP yield of ∼3.6 mM_ATP/mM_S were achieved. The fermentation product range was rather unique due to the almost exclusive dominance by acetate (28.1 mM_C) accompanied by only traces of succinate and propanoate. During the course of growth, the pH dropped from 6.9 to 6.6. As mentioned for xylan before, 90 mM_C of glucuronate starting concentration formed the basis for calculations.

Growth curves of P. vulgatus utilizing either of the other 10 tested carbohydrates (xylose, arabinose, galactose, fructose, mannose, fucose, maltose, lactose, sucrose, and α-cyclodextrin) are provided in online supplementary Figure S1a‒j and the respective growth parameters, and biomass and product yields are compiled in Table 1 and online supplementary Table S1. Comparing the biomass yields attained during growth with glucose, maltose, or α-cyclodextrin revealed a gradual increase (0.06, 0.12, and 0.27 g_CDW/mmol_S, respectively), which fairly well reflects the multiples of glucose moieties constituting these three compounds (1, 2, and 6, respectively). Similarly, the biomass yield with xylan was considerably higher than with its monomer xylose (0.14 vs. 0.05 g_CDW/mmol_S). When relating these biomass yields to mM_C, however, these differences are essentially evened out, as expected. The biomass yields with the two further tested disaccharides lactose and sucrose (both 0.09 g_CDW/mmol_S) were somewhat lower than with maltose (0.12 g_CDW/mmol_S), albeit considerably higher than with all tested monosaccharides (ranging from 0.01 g_CDW/mmol_S for fucose to 0.06 g_CDW/mmol_S for the hexoses glucose, fructose, and mannose). The ATP yields calculated for all 14 tested carbohydrates covered a broad range, when related to mM_S: from ∼1.3 mM_ATP/mM_S (rhamnose) to ∼20.1 mM_ATP/mM_S (xylan). However, when related to mM_C, the substrate-dependent differences of ATP yields were expectedly far less pronounced, ranging from ∼0.2 mM_ATP/mM_C (rhamnose) to ∼0.6 mM_ATP/mM_C (glucuronate/xylan).

The substrate-dependent variations of the product yields attained by P. vulgatus were significantly more pronounced than described before for the biomass yields. Acetate was with few exception (e.g., lactose) the dominant fermentation product (P) with yields ranging from 0.31 mM_P/mM_S (with fucose) via 0.94 mM_P/mM_S (with glucuronate) to 2.13 mM_P/mM_S (with xylan). Succinate was observed as second most abundant fermentation product showing highest yields among monosaccharides with glucose (0.39 mM_P/mM_S), among disaccharides with sucrose (0.86 mM_P/mM_S), and overall with α-cyclodextrin (1.90 mM_P/mM_S). Formate was detected under all substrate conditions, albeit at lower yields, which was among monosaccharides highest with glucose (0.16 mM_P/mM_S) and overall with xylan (1.37 mM_P/mM_S). Propanoate was detected across all tested substrate conditions but consistently with lowest yields, e.g., 0.01 mM_P/mM_S with glucose and at its highest levels 0.51 mM_P/mM_S with xylan. Propane-1,2-diol was exclusively detected during growth with rhamnose and fucose, reaching product yields of 1.03 and 0.40 mM_P/mM_S, respectively, while the other products were formed at lower levels. The closure of the carbon balance ranged from 80.6% (lactose) to 99.5% (mannose) with a median of 90.5% (online suppl. Table S1).

The proteomes of the substrate-adapted cells of P. vulgatus were profiled by separately analyzing the soluble and membrane-enriched protein fractions (three replicates each per substrate condition), yielding in total the identification of 1,400 different proteins (34.9% of the 4,011 proteins predicted from the genome). The coding genes of the identified proteins were found to be essentially equally distributed across the genome of P. vulgatus (Fig. 2a, left panel). The specific functional modules studied here, viz. fermentation network, Sus-like, and respiratory systems, comprised 56, 98, and 33 individual proteins, respectively, which combined account for 13.4% of the covered proteome (4.7% of predicted proteome) (Fig. 2a, right panel). A principal component analysis of the entire proteomic dataset revealed a tight clustering according to substrate adaptation condition and the respective replicates (Fig. 2b), underpinning the reproducibility of the consecutively applied cultivation and proteomic approaches. The generated differential proteomic data were used to reconstruct the fermentation network and transport (Sus-like) systems for the studied growth substrates as well as the transmembrane complexes potentially involved in respiratory energy conservation.

The fermentation network for the 14 different studied substrates was investigated by integrating the generated substrate-specific differential proteome profiles with known literature data as well as previous and current manual annotations of the genome of P. vulgatus. The reconstructed network consists of 56 predicted proteins, all of which could be identified. The general architecture of the fermentation network is schemed in Figure 3, with supporting details on the substrate-specific and shared parts provided in online supplementary Figure S2a‒n and differential abundance profiles of assigned proteins illustrated in Figure 4.

The fermentation network is arranged around the following interconnected central modules: glycolysis, oxidative pentose cycle, methylmalonyl-CoA pathway, and a partial TCA cycle. The protein constituents of these central modules were essentially detected at rather high abundances under all 14 tested substrate conditions and their coding genes are distributed across the genome of P. vulgatus. The substrate-specific peripheral reactions sequences can be divided roughly into four groups: first, xylan, xylose, and arabinose feed via the oxidative pentose phosphate cycle at the level of glyceraldehyde-3-phosphate into the lower part of glycolysis. Second, α-cyclodextrin, maltose, lactose, sucrose, galactose, glucose, and fructose are channeled at various levels into the upper part of glycolysis. Third, glucuronate is converted by a specific reaction sequence to glyceraldehyde-3-phosphate and pyruvate of glycolysis. Fourth, the compound-specific routes for the degradation of rhamnose and fucose converge at the level of lactaldehyde.

The proteins assigned to the peripheral degradation routes channeling arabinose, xylan, and xylose into the pentose phosphate pathway were detected with high substrate specificity (except for Tkt), while this was not observed with the routes channeling mannose, lactose, and galactose into glycolysis. It stands to reason that substrate-specific regulation of the respective degradation modules is mediated by sensory/regulatory systems on the transcriptional level.

According to the aforementioned substrate-specific network modulation, a broad collection of transcriptional regulators could be predicted for P. vulgatus (online suppl. Table S5) based on the RegPrecise database [38] combined with manual searches. However, their assignment to individual degradation modules remains in parts ambiguous since their coding genes typically do not co-localize with their cognate “catabolic” genes and they mostly escaped the applied proteomic window due to their typically small size and low abundance. Some examples are provided in the following. While no homologs of the known glucuronate-specific UxuR and ExuR regulators from Escherichia coli [39] are predicted, the KdgR repressor known in the plant-pathogen Erwinia chrysanthemi to regulate “pectinolysis” genes [40] could be identified in glucuronate-adapted P. vulgatus. Even though the coding genes for the fucose-specific FucKR regulator from B. thetaiotaomicron [41] and the rhamnose-specific RhaR regulator from E. coli [42] could be predicted, they were not detected. Notably, the genome of P. vulgatus harbors 16 homologs of arabinose/xylose-specific AraC regulators from E. coli [43, 44], which were, however, not detected in the present study. Instead, a predicted AraR regulator was detected in arabinose-adapted cells and a XylR regulator in xylose/xylan-adapted cells of P. vulgatus, both of which have been reported to be widespread among Bacteroides and AraR in B. thetaiotaomicron to be indeed arabinose responsive [45, 46]. Next to the degradation module-specific transcriptional regulators, P. vulgatus possesses two well-known global regulators of carbohydrate metabolism, which were detected under all of the tested substrate conditions (online suppl. Table S5): (i) the Mlc protein negatively controls MalT, the transcriptional activator of the maltose regulon [47]; (ii) the cAMP-activated Crp protein for general activation of transcription of various genes related to carbohydrate catabolism [48].

The five different fermentation products of P. vulgatus are formed by a suite of specific as well as shared reaction sequences (Fig. 3). The route for the formation of propanoate and succinate branches off from the lower section of glycolysis at the level of PEP, which is fed into the reverse TCA-cycle and therefrom into the methylmalonyl-CoA pathway (online suppl. Fig. S2m, n). In case of rhamnose and fucose, oxaloacetate is apparently also formed via PEP since this is energetically more favorable than the direct formation from pyruvate (genes present). Propane-1,2-diol is formed from lactaldehyde by lactaldehyde reductase (FucO) (online suppl. Fig. S2i), which is aptly detected with highest abundances in rhamnose- and fucose-adapted cells. Lactate can in principle be formed either from pyruvate (online suppl. Fig. S2l) by lactate dehydrogenase (Ldh) or from lactaldehyde by aldehyde dehydrogenase (AldA); both enzymes were detected only at low abundances and sporadically, agreeing with the general non-detection of lactate. Formation of formate and acetate (online suppl. Fig. S2k) relies on the conversion of pyruvate to acetyl-CoA, which can in principle occur via two alternatives: oxidative decarboxylation by pyruvate:ferredoxin oxidoreductase (Pfo) or formate release by pyruvate-formate lyase (PflB). The proteomic data hint at a more prominent role of Pfo since it displays 1.3- to 53.0-fold higher peptide counts across the tested substrate conditions, as compared to PflB. This is also consistent with the markedly lower formation of formate as compared to that of acetate. Formation of acetate from acetyl-CoA is then achieved by cooperation of phosphate transacetylase (Pta) and acetate kinase (AckA).

The Sus, shown in Figure 5b, is well known for mediating import of glycan-derived mono- and oligomers across the outer membrane, prior to their import across the cytoplasmic membrane via respective primary and/or secondary transporters. P. vulgatus possesses 67 predicted polysaccharide utilization loci (PULs), which is well in the range of other Bacteroides members (Table 2). The genes related to the PULs for xylan and α-cyclodextrin degradation (PULs 1 and 38, respectively) are each co-localized on the genome of P. vulgatus and their protein products were substrate-specifically identified (Fig. 5a). Furthermore, Figure 5a displays the adjacent susC17D17 genes, the encoded proteins of which were detected under all 14 studied substrate conditions. In accord with the high number of PULs, the genome of P. vulgatus encodes a total of 352 carbohydrate-activating enzymes (CAZymes), with the largest share (204) contributed by glycoside hydrolase family (Table 2).

The genome of P. vulgatus encodes for a total of 196 Sus-like system-related proteins (2 TonB, 111 SusC, 81 SusD, 1 SusE, and 1 SusR); coding genes for SusF were not predicted. Furthermore, 8 proteins of the glycoside hydrolase 97 (glycoside hydrolase) family could be detected as well as 4 proteins of the GH13 family. However, the exact assignment of SusA/G and SusB proteins within the GH97 and GH13 families, respectively, is currently unclear. From these predicted proteins, a total of 95 could be identified (1 TonB, 4 SusA/G, 1 SusB, 57 SusC, 31 SusD, and 1 SusE) and assigned to PULs (online suppl. Table S4). Only 13 of the identified Sus-related proteins could not be assigned to PULs and are provided as additional data (FAIRDOMHub). In most cases, the abundance of identified Sus-like system-related proteins is apparently independent of the substrate used for adaptation. Notably, however, xylan-adapted cells harbor the largest diversity of Sus-like system-related proteins.

The genome of P. vulgatus and the current proteomic dataset were specifically examined to retrieve potential uptake systems for the tested growth substrates. Accordingly, 19 transporters (excluding SusC paralogs) were predicted from the genome (online suppl. Tables S3, S5), which have assumed specificities for maltose (1), lactose (1), glucose (3), galactose (2), xylose (2), arabinose (2), l-fucose (3), l-rhamnose (1), glucuronate (1), and others (6), which to the largest parts are tentatively assigned to the MFS-family of transmembrane secondary transport. Notably, only a single hint on an ABC-type (primary) transporter (BVU_RS19895) was obtained. Robust proteomic evidence was only obtained for a l-fucose:H⁺ symporter (FucP1, BVU_RS06945), an arabinose transporter (AraP, BVU_RS11685), and xylose/xylooligosaccharide transporters (XylE1/XynT, BVU_RS20420/00225), which were specifically detected in fucose-, arabinose-, and xylan-adapted cells of P. vulgatus, respectively. While these results suggest secondary transporters to mediate the uptake of the studied carbohydrates, their substrate specificity vs. promiscuity remains unclear at present.

Despite the metabolic specialization of P. vulgatus on fermentation, the present proteogenomic study revealed the presence of several ion motive force generating transmembrane electron transfer systems. They are illustrated in Figure 6a with the protein abundance profiles of the individual subunits compiled in online supplementary Figure S3. Contrasting most of the other catabolic processes studied here, the genes related to these systems are organized in distinct operon-like structures (Fig. 6b).

All subunits of the Na⁺-pumping Nqr-complex (NqrA-F) could be identified, while this was only the case for the NuoB subunit of the H⁺-pumping NDHI-complex (complex I, NuoA‒L). This implicated a more prominent role of the Nqr-complex for P. vulgatus under the studied cultivation conditions. Both complexes (Nqr and NDHI) transfer electrons from the cytoplasmic NADH pool to the membrane-embedded quinone pool. The latter can then provide electrons to respiratory fumarate reductase (FrdABC) and cytochrome oxidase (CydAB), the subunits of both of which were identified and in case of Frd even under all studied conditions. This suggests that P. vulgatus can, in principle, utilize fumarate and O₂ as terminal electron acceptors. Furthermore, the genome of P. vulgatus encodes a Na⁺-pumping Rnf complex (RnfA-EG), most subunits of which were identified, albeit not under all studied cultivation conditions. While 3 Na⁺/H⁺-antiporters (BVU_RS15380/16245/19570) are predicted, they were not detected in the current proteomic dataset. Finally, P. vulgatus possesses two Na⁺/H⁺-pumping ATP-synthases one of the F-type and the other of the V-type. Most subunits of the F-type ATP synthase were detected under all studied conditions, which was not observed for the V-type ATP synthase. Thus, one may assume that the F-type ATP synthase is physiologically more relevant under the studied conditions.

The prominent role of Bacteroides members and particularly P. vulgatus in the gut and their contribution to human health by their nutritional specialization in and fermentative decomposition of dietary fiber-derived (poly)saccharides is well appreciated [e.g., 4, 49-51]. Likewise, fitness profiling of a barcoded transposon mutant library of B. thataiotaomicron across several dozen substrate conditions revealed substrate-specific contributions of different PULs [52]. Moreover, recent studies implicate intricate (metabolic, genetic) interactions of P. vulgatus with other members of the gut microbiome: e.g., the yeast Candida albicans [53]; the acetate-consuming, butyrate-producing Bacillota member Anaerostipes caccae, previously assigned to the genus Eubacterium [54, 55]; the food-born (often chicken) pathogenic epsilonbacterium Campylobacter jejuni [27]; and the highly abundant bacteriophage crAssphage [56, 57]. Still, the current understanding of the condition-dependent fluctuations of the fermentation spectrum and its underlying metabolic network in P. vulgatus is limited, a knowledge gap the present study is addressing. While comparisons of the present findings with previous reports on P. vulgatus and other Bacteroides spp. are obvious, they are somewhat hampered by differences in applied growth conditions. Nevertheless, the here reconstructed fermentation network is in good agreement with previous observations on diet-specific fitness determinants of B. thetaiotaomicron and B. cellulosilyticus obtained upon feeding experiments with laboratory mice systems [58].

Against the backdrop of the aforementioned nutritional specialization of P. vulgatus, the present study demonstrates that this bacterium’s fermentation spectrum is significantly and in parts distinctively shaped by the provided carbohydrate source, as exemplarily discussed in the following. With glucose as growth substrate, P. vulgatus mainly formed acetate and succinate, as previously observed with S. copri (formerly P. copri) [35], but at markedly lower product yields: 8.0 vs. 22.7 mmol acetate/g_CDW and 6.2 vs. 22.7 mmol succinate/g_CDW, respectively. While formation of lactate could not be observed in this study, recently a 200-fold increased lactate dehydrogenase activity and redirecting of the carbon flow from glucose toward lactate in P. vulgatus upon homologous overexpression of the ldh gene was reported [59].

With rhamnose and fucose, the characteristic fermentation product of P. vulgatus was propane-1,2-diol, as previously reported for Prevotella ruminicola (formerly Bacteroides ruminicola) [60, 61], E. coli [62], as well as for Salmonella typhimurium and Klebsiella pneumoniae [63]. With rhamnose, but not with fucose, P. vulgatus formed equimolar concentrations of propane-1,2-diol as observed before in the case of S. typhimurium and K. pneumoniae with both of these sugars [63]. Notably, the products yield of P. vulgatus achieved under the here applied conditions was ∼5-fold higher as previously reported for P. ruminicola (0.48 vs. 0.09 g_P/g_S) [60]. The low level of succinate formation by P. vulgatus during utilization of rhamnose or fucose is mirrored by the reduced abundance level of PEP carboxylase (PckA). Formed acetate reflects activity of acetate kinase (AckA) as the only opportunity for substrate level phosphorylation along the fermentation route, which in turn explains the lowest ATP yield among the tested substrates. Key to the pathway in P. vulgatus is the intermediate lactaldehyde and its reduction by lactaldehyde dehydrogenase (FucO) to propane-1,2-diol. This is presumably also the case for the other aforementioned propane-1,2-diol-forming fermenting bacteria, but not for the prominent gut bacterium B. thetaiotaomicron, since it does not possess the relevant genes despite its capacity to utilize fucose [64]. The FucO protein of P. vulgatus shows high sequence similarity to its counterpart of E. coli (∼58% identity) and B. thetaiotaomicron (∼66% identity). The crystal structure of the E. coli enzyme revealed the key residues for binding of Fe²⁺ [65], which are also conserved in FucO of P. vulgatus (Fig. 3b). Notably, propane-1,2-diol supports metabolism of intestinal inflammation inducing S. enterica serovar typhimurium, which indicates metabolic interactions between members of the gut microbiome and gut pathogens [66].

In the case of galactose utilization by P. vulgatus, a previous study proposed a reaction sequence for converting l-galactose via d-tagaturonate to glyceraldehyde and pyruvate, with the involved enzymes biochemically characterized [31]. The present differential proteomic study with d-galactose-adapted cells (online suppl. Fig. S2e) indicated involvement of the known Leloir pathway [67], i.e., via d-galactose-1-phosphate to glucose-6-phosphate. By contrast, arabinose and xylose utilization by P. vulgatus via the pentose phosphate pathway, as suggested by the differential proteomic data (online suppl. Fig. S2a, j), is in accord with previous genomic predictions for B. thetaiotaomicron [38]. The conspicuous observation with glucuronate-utilizing P. vulgatus was the essentially exclusive formation of acetate as fermentation end product. This is obviously due to the fact that the used pathway provides net ATP synthesis only via acetate-forming acetate kinase (Fig. 3) and agrees well with earlier findings obtained with B. thetaiotaomicron grown in glucuronate-limited chemostats [68].

In the case of growth with xylan, P. vulgatus relies on a PUL (PUL 1) comprising 8 genes (6 encoded proteins were identified), which is rather compact compared to other large (up to 32 genes) PULs [1], but occurs also in uncultured human gut Bacteroides strains [69]. This could reflect the less complex structure of the applied xylan from corn cob, which is constituted by 95% of xylooligosaccharides. The ability to grow with xylan is a rare feature among Bacteroides members and was only observed with B. cellulosilyticus, B. eggerthii, B. ovatus, and B. xylanisolvens [34]. Furthermore, most Bacteroides species cannot utilize glucuronate, or, if they do, achieving only low growth [34]. A recent study using gnotobiotic mice colonized with a 13-member community of human gut microbiota revealed that knockdown of B. cellulosilyticus resulted in strongly increased gene expression for PUL 1 of P. vulgatus, possible to sustain glycan utilization despite the community disturbance [70].

The overwhelming detection of Sus-like systems in P. vulgatus not only during growth with xylan, but also with all other tested sugar substrates, agrees with the apparent uptake strategy of Bacteroides spp. [1] but nevertheless is contrasted by the broad range of transporters encoded in the genome of P. vulgatus and other Bacteroides spp. [71]. As compared to the other tested substrates, xylan gave rise to a more diverse fermentation spectrum with propanoate and formate formed next to the abundant succinate and acetate. This is in accord with xylan-limited chemostat-based studies inoculated with adult fecal microbiota and revealing P. vulgatus among the prevalent species during the experiment [72]. Interest in xylan metabolism in the human gut is expected to continue since this polysaccharide is not only an abundant component of dietary fibers but also regarded as prebiotic, i.e., essential gut nutrient [73].

The anaerobic transmembrane electron transport chain of P. vulgatus apparently centers around fumarate respiration as recently reported for S. copri [34]. In S. copri, targeted transcript profiling suggested the NDHI complex as well as the Nqr complex to equally deliver electrons to the quinone pool coupled to the generation of a Na⁺/H⁺ gradient. The present proteomic data for P. vulgatus suggested an outstanding role of the Nqr complex (NADH oxidizing) complemented by the Rnf complex (Fd_red oxidizing), while the NDHI complex probably plays a subordinate role only. In accord, the Nqr complex was reported in Bacteroides fragilis to contribute by 65% to the NADH:quinone oxidoreductase activity [74]. A similar role of the Nqr complex was also reported for Prevotella bryantii B14 and other Prevotella spp. [75]. Moreover, in case of P. bryantii a direct interaction with fumarate reductase, forming the so-called sodium-translocating NADH:fumarate oxidoreductase supercomplex, was recently demonstrated [76]. A noteworthy observation with respect to the Rnf complex was that Fd_red-oxidizing RnfB and NAD⁺-reducing RnfC subunits were detected with highest abundances in deoxy sugar-adapted P. vulgatus, possibly providing an additional contribution to the Na⁺-gradient. Na⁺-dependent activity of the Rnf complex was previously demonstrated with Bacteroides fragilis [77].

As in S. copri, the F-type ATP synthase is probably responsible for ATP synthesis driven by proton/sodium motive force since the subunits of the less complex V-type ATP synthase were detected in P. vulgatus at lower abundances and under fewer tested conditions. Nevertheless, it remains elusive at present, which of the two ATP synthases of P. vulgatus contributes to what extent to either ATP synthesis or in reverse direction to generation of proton/sodium motive force [78].

The detection of O₂-reducing (to H₂O) CydAB in the presence of several of the tested substrates, despite the applied anoxic cultivation conditions, may indicate readiness of P. vulgatus to respire away traces or pulses of O₂. Such a strategy is also known from other anaerobic bacteria, such as sulfate reducers [79]. While the oxygen gradient across the intestinal mucosa is tightly controlled to maintain redox homeostasis in the gut, it can be disturbed due to dietary or disease-related conditions leading to increased pO₂ and thereby to an impact on the Bacteroidota population [80, 81].

The here obtained insights into the substrate-specific fermentation spectrum and its underlying metabolic network of P. vulgatus may serve as valuable basis to develop P. vulgatus as non-standard platform organism for the sustainable production of bulk chemicals. For example, the major fermentation products succinate, propane-1,2-diol, and acetate of P. vulgatus are currently produced at amounts of 2.5 × 10⁵t, 1.9 × 10⁵t, and 1.7 × 10⁷t per year, respectively, on a global scale [82, 84]. In case of propane-1,2-diol, sustainable microbial production may be particularly beneficial, since the current bulk synthesis proceeds via high temperature hydrolysis of propylene oxide, which is toxic (carcinogenic, mutagenic) and produced from fossil fuels in an environmentally harmful process. Accordingly, microbial biosynthesis of propane-1,2-diol and strategies for respective metabolic engineering represent a timely topic in biotechnology [85, 86]. Against this backdrop, the markedly enhanced lactate fermentation recently reported by Lück and Deppenmeier [58] provides a proof-of-principle for genetic engineering of the fermentation network of P. vulgatus in various product directions.

Phocaeicola vulgatus ATCC 8482 (DSM 1447) [87] was obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ; Braunschweig, Germany) and stored at −80°C in glycerol stocks. P. vulgatus was exclusively cultivated in a defined mineral medium as previously described by others [34, 88], with the following modification: instead of l-cysteine, 5 mM Na₂S₂O₃ as sulfur source and 6 mM Na₂S as reductant were added to avoid an additional carbon source. The medium was adjusted to pH 7.0–7.3. Growth substrates were added from sterile stock solutions, using sterile, N₂/CO₂-flushed syringes. All chemicals used were of analytical grade. Cultivation was performed under anoxic conditions in rubber-stopper-sealed glass bottles, which had an N₂/CO₂ (90:10) headspace and were incubated stationary at 37°C. Two types of controls were run for each growth experiment to ascertain sterility of the medium: first, with inoculum but without substrate and second, without inoculum but with substrate. Purity control of cultures was conducted by streaking cells on YM Agar and nutrient agar plates and by microscopic examination (Axiostar; Zeiss AG, Göttingen, Germany).

Prior to any growth experiment, cells of P. vulgatus from the stock cultures were revived via dilution series (up to 10⁻⁶) in glucose-containing mineral medium, in order to remove residual glycerol. For adaptation to a given substrate, revived cells were then transferred to 80 mL of mineral medium containing 90 mM_C of the respective substrate (in 100 mL flasks). Applying this setting, four passages were done using actively growing cells (∼¾ OD_max) of a current passage to inoculate the consecutive one with a starting OD of 0.05. The 5^th passage was conducted in 500-mL flasks containing 400 mL medium to further adapt the cells to the culture volume used in the main experiments. The 90 mM_C of substrate used for the respective adaptation was adjusted by the following additions to the medium: 15 mM fructose, fucose, galactose, glucose, glucuronate, mannose, or rhamnose; 18 mM arabinose or xylose; 7.5 mM lactose, maltose, or sucrose; 2.5 mM α-cyclodextrin; and 2.6 mM xylan.

All growth experiments, viz., for stoichiometric as well as proteomic analyses, were conducted with substrate-adapted cells (see above) in 500-mL flasks containing 400 mL medium.

Samples were retrieved with sterile, N₂/CO₂-flushed syringes either at higher resolution for monitoring OD, pH, and extracellular substrate/product concentrations (via HPLC) or at lower resolution for the determination of CDW and TCC. In the first case, sample volumes of 5 mL were retrieved per time point and divided into one 1-mL (OD) and two 2-mL (pH and HPLC analyses) aliquot(s). OD was measured at 660 nm (UVmini-1240 spectrophotometer; Shimadzu, Duisburg, Germany), the pH with a pH electrode (pH meter: 766 Calimatic, Knick, Berlin, Germany; pH sensor: InLab Routine, Mettler Toledo, Gießen, Germany), and the depletion of individual growth substrate versus formation of their respective fermentation products by means of HPLC (see below). Samples for HPLC analyses were centrifuged (20,800 × g, 10 min, 4°C), the supernatants transferred to new micro reaction tubes, and stored at −20°C. In the second case, samples were retrieved at ¼ OD_max, ½ OD_max, ¾ OD_max, and OD_max and for TCC also directly after inoculation. Sample volumes were 1 mL for TCC and 5–40 mL (depending on OD) for CDW. TCC was determined by means of a Thoma counting chamber using two different sample dilutions (Glaswarenfabrik Karl Hecht GmbH & Co. KG, Sondheim vor der Rhön, Germany). Samples for CDW determination were centrifuged (25,200 × g, 10 min, 4°C), the supernatants discarded, the resulting cell pellets washed in 5 mL of 50 mM ammonium acetate buffer, ultimately resuspended in 300 μL of the same buffer, transferred into pre-dried and weighed micro reaction tubes, and incubated at 60°C until constant weight was reached.

Per substrate condition, five parallel cultures were run, with two of them incubated until OD_max to ascertain reproducibility of growth compared to the aforementioned stoichiometric experiments. The other three parallel cultures were harvested at ½ OD_max by centrifuging the complete 400 mL culture broths (14,300 × g, 30 min, 4°C), followed by resuspending each pellet in 100 mL washing buffer (100 mM Tris/HCl, 5 mM MgCl₂ × 6 H₂O, pH 7.5) and anew centrifugation under the same conditions. The resultant pellets were each resuspended in 1 mL wash buffer and transferred to micro reaction tubes, which were then centrifuged (20,800 × g, 10 min, 4°C). The final pellets were immediately shock frozen in liquid N₂ and stored at −80°C until further treatment for proteomic analyses.

Substrate depletion and product formation during growth of P. vulgatus were quantified in cell-free culture samples via HPLC. Prior to analyses, stored samples (see above) were thawed, diluted (1:1) in membrane-purified water, and filtered (pore size, 0.2 µm, regenerated cellulose; Chroma Globe, Kreuzau, Germany). For HPLC analyses, an UltiMate 3000 system (Thermo Fisher Scientific, Germering, Bavaria, Germany) was equipped with an RI detector (Shodex RI-101; Showa Denko GmbH, München, Germany) for sugars or with an UV detector (DAD-3000; Thermo Fisher Scientific) for fumarate. Compound separation was achieved with an Eurokat column (300 × 8 mm, 10 μm bead size; Knauer, Berlin, Germany), except for rhamnose, arabinose, fucose, and sucrose, which were analyzed via a Rezex ROA-Organic Acid H⁺ column (300 × 7.8 mm; Phenomenex Ltd., Aschaffenburg, Germany) to achieve improved chromatographic separation. The HPLC system was operated at 70°C (Eurokat column) or 20°C (Rezex column), always applying 0.5 mM H₂SO₄ as isocratic eluent administered at a flow rate of 0.5 mL/min. The retention times of the targeted compounds (in alphabetic order) were as follows: acetate at 20.4 min, arabinose at 15.7 min, formate at 19.0 min, fructose at 14.8 min, fucose at 16.1 min, galactose at 13.8 min, glucose at 13.4 min, lactose at 11.0 min, maltose at 11.5 min, mannose at 14.1 min, propane-1,2-diol at 23.2 min, propanoate at 23.4 min, rhamnose at 14.7 min, succinate at 15.9 min, sucrose at 11.6 min, and xylose at 14.0 min. Since the retention times of propanoate and propane-1,2-diol lie very close together, detection of propane-1,2-diol was additionally corroborated by standard addition. The limit of detection was 25 µM and the dynamic range extended to 10 mM. For all sampling time points and substrate conditions, three biological replicates with two technical replicates each were measured.

The ATP yields were calculated, by balancing the substrate-specific consumed versus generated ATP molecules and reducing equivalents (e.g., Fd_red) in the oxidative (from substrate to PEP/pyruvate) and reductive halves (from PEP/pyruvate/lactaldehyde to fermentation products) of the fermentation network (Fig. 3) and the generated ion gradients (H⁺ and Na⁺) across the membrane as well as the resulting ATP synthesis via ATPases by the respiratory system. In case of the reductive part, the experimentally quantified ratio of fermentation products per substrate (online suppl. Table S1) allowed to determine the relative shares of the product-forming pathways in the total fermentation spectrum and thereby to determine the overall balances of ATP and reducing equivalents. Reducing equivalents formed in the oxidative part, but not consumed in the reductive part according to the quantified fermentation spectra, were used for estimating potential respiratory energy conservation. Here, also Fd_red formed by Pfo was considered. As suggested by the present proteogenomic analyses, Fd_red- and NADH-driven Na⁺-pumping could be mediated by the Nqr and Rnf complexes, respectively. Since the ATP/ion-coupling ratios of the F- and V-type ATP synthases in P. vulgatus are currently unknown, the two most divergent known ratios of F-type ATP synthases in heterotrophic bacteria framed our calculations (online suppl. Table S2), as previously described [89]: 3.3 ions per ATP in E. coli versus 4.3 ions per ATP in Alkalihalobacillus pseudofirmus OF4 [90]. To account for biomass formation of P. vulgatus beyond the determined CDW, the elemental composition (C₁H_1.79O_0.44) of related S. copri (formerly P. copri) as reported by Franke und Deppenmeier [34] was considered.

The performed proteomic analyses were based on the separation of soluble and membrane protein-enriched fractions of P. vulgatus. Initially, a given cell pellet was thawed on ice prior to resuspending it in 500 μL of shotgun lysis buffer (7 M urea, 2 M thiourea, 30 mM Tris/HCl, pH 8.5). The cell suspension was then transferred into a matrix tube, which was filled with 0.25 g of an even mixture of 0.1- and 1-mm zirconium beads, and cell breakage was achieved by 3 rounds of bead beating at 6 m/s for 30 s (Fast-Prep-24 5G; MP Biomedical, Eschwege, Germany) with in-between incubation for 90 s on ice. Each sample (without the beads) was then ultra-centrifuged twice (104,000 × g, 60 min, 10°C). The resultant supernatant, which represents the fraction of soluble proteins, was then immediately shock frozen in liquid N₂ and stored at −80°C until further analyses. The two pellets generated during the two previous rounds of ultracentrifugation were each resuspended in 300 μL of membrane lysis buffer (100 mM Tris/HCl, pH 7.5, 2 mM MgCl₂ × 6 H₂O, 10% wt/vol glycerol, 0.5 mM DTT), treated with 35 mL ice-cold sodium carbonate solution (100 mM) by stirring for 1 h in an ice bath. Following ultra-centrifugation (200,000 × g, 60 min, 4°C) and discarding of the supernatant, the pellet was resuspended in 400 μL membrane lysis buffer and ultra-centrifuged again. The resultant pellet was resuspended in 150 μL of 1% wt/vol SDS and incubated at 95°C for 10 min with constant shaking (600 rpm). Following centrifugation (20,800 × g, 20 min, 20°C), the supernatant, which represented the membrane protein-enriched fraction, was immediately shock frozen in liquid N₂ and stored at −80°C until further analysis.

The protein concentrations in the prepared soluble and membrane protein-enriched fractions were determined according to Bradford [91] and using the RCDC protein assay, respectively (Bio-Rad Laboratories, Munich, Germany). Further processing was performed essentially as previously described [92].

In the case of the soluble protein fraction, 50 µg protein was diluted with urea buffer (8 mM urea, 0.4 mM NH₄HCO₃) to a final volume of 50 µL. After adding 45 mM DTT, the protein solution was incubated at 55°C for 30 min in the dark (Thermomixer comfort; Eppendorf AG, Hamburg, Germany). After cooling to ambient temperature, 100 mM iodoacetamide was added and incubation in the dark was continued for additional 15 min. Then, 143 μL water (HPLC grade) and 1 μg trypsin (Serva, Heidelberg, Germany) were added, followed by an overnight incubation at 37°C.

In the case of the membrane protein-enriched fraction, samples were decomplexed using 12% SDS mini gels (10 × 7 cm; Bio-Rad), which were loaded with 10 µg protein per lane and stained with Coomassie brilliant blue after electrophoresis. Then, each gel lane was cut into 7 slices, each of which was further cut into small pieces (∼1 mm³). The latter was subjected to washing, reduction, alkylation, and tryptic digest as described [92]. Peptide mixtures generated from both subcellular fractions were shock frozen in liquid N₂ and stored at −80°C until MS analyses.

Tryptic peptides were separated by nanoLC (UltiMate 3000 nanoRSLC; Thermo Fisher Scientific) applying the following setup: trap column (C₁₈, 2 cm × 100 μm, 5-µm bead size; Thermo Fisher Scientific), separation column (C₁₈, 25 cm × 75 µm, 2-µm bead size; Thermo Fisher Scientific), and a 280-min (shotgun) or 90-min (membrane, per gel fraction) linear gradient of acetonitrile (0–80% vol/vol). The eluent was continuously analyzed by an online-coupled ion trap mass spectrometer (amaZon speed ETD; Bruker Daltonics GmbH, Bremen, Germany) using settings as described by Zech et al. [92]. Per substrate conditions, six replicates (3 times each soluble and membrane protein-enriched fraction) were measured, giving rise to a total of 84 analyzed samples.

For protein identification, the ProteinScape platform (version 3.1; Bruker Daltonik GmbH), an in-house Mascot server (version 2.3; Matrix Science Ltd., London, UK), and the translated genome of P. vulgatus [93] were used. A target decoy strategy with described settings [94] was applied.

The manual reconstruction of the catabolic network for the selected substrates was based on the published genomes of P. vulgatus [93] and other Bacteroides spp. [22], pathway predictions provided by the KEGG [95] and MetaCyc [96] databases, relevant primary literature, as well as detailed comparisons with the here generated differential proteomic dataset. For the analysis of PULs and in particular Sus-like systems, the PUL [97] and CAZy [98] databases were used. Search for relevant transporters and regulators mediating substrate uptake and transcriptional regulation of degradation modules, respectively, benefitted from the transporter classification database (TCDB) [99] and the RegPrecise database [100] as well as relevant literature. To explore potential functions of hypothetical proteins, BLAST analyses via UniProt [101] were conducted. The BRENDA database [102] was consulted for EC numbers and gene names. MetaboMaps [103] was used to generate heatmaps of differential protein abundance profiles. For analysis of genomic context and for visualization of the genome circle, the Artemis Comparison Tool (version 18.2.0) [104] was used. Pairwise sequence alignment of FucO from P. vulgatus, B. thetaiotaomicron, and E. coli was conducted using the tool “global alignment: Needle (EMBOSS)” [105].

An ethics statement was not required for this study type; no human or animal subjects or materials were used.

The authors have no conflicts of interest to declare.

This study was supported by the Federal Ministry of Education and Research (Bundesministerium für Bildung und Forschung) within the framework of the collaborative BaPro project.

R.R. conceived the study; U.C., S.-T.V., and P.L. conducted the cultivation experiments; U.C. and S.-T.V. evaluated the physiological data and calculated the (growth and other) parameters; S.S. operated the HPLC analyses; M.G. and L.W. performed the proteomic analyses; U.C. and S.-T.V. performed the proteogenomic reconstruction of the catabolic network; R.R., U.C., and S.-T.V. wrote the manuscript; and all authors agreed to the final version of the manuscript.

Urte Clausen and Sören-Tobias Vital contributed equally to this study.

Supplemental material is available online only. The proteomic dataset and calculations were deposited at FAIRDOMHub (https://fairdomhub.org/data_files/6878?version=1). Further inquiries can be directed to the corresponding author.