Enzymatic Adaptation of Bifidobacterium bifidum to Host Glycans, Viewed from Glycoside Hydrolyases and Carbohydrate-Binding Modules.

Certain species of the genus Bifidobacterium represent human symbionts. Many studies have shown that the establishment of symbiosis with such bifidobacterial species confers various beneficial effects on human health. Among the more than ten (sub)species of human gut-associated Bifidobacterium that have significantly varied genetic characteristics at the species level, Bifidobacterium bifidum is unique in that it is found in the intestines of a wide age group, ranging from infants to adults. This species is likely to have adapted to efficiently degrade host-derived carbohydrate chains, such as human milk oligosaccharides (HMOs) and mucin O-glycans, which enabled the longitudinal colonization of intestines. The ability of this species to assimilate various host glycans can be attributed to the possession of an adequate set of extracellular glycoside hydrolases (GHs). Importantly, the polypeptides of those glycosidases frequently contain carbohydrate-binding modules (CBMs) with deduced affinities to the target glycans, which is also a distinct characteristic of this species among members of human gut-associated bifidobacteria. This review firstly describes the prevalence and distribution of B. bifidum in the human gut and then explains the enzymatic machinery that B. bifidum has developed for host glycan degradation by referring to the functions of GHs and CBMs. Finally, we show the data of co-culture experiments using host-derived glycans as carbon sources, which underpin the interesting altruistic behavior of this species as a cross-feeder.


Introduction
Bifidobacteria are Gram-positive obligate anaerobes that represent common inhabitants of the gastrointestinal tract of animals [1]. In humans, it is well known that a bifidobacteria-rich microbiota (so-called, bifidus-flora) is formed in the intestinal tracts of breast-fed infants. These "first colonizers" are believed to confer various beneficial effects on human health, including protection against pathogen infection [2,3], nutritional supplementation [4], reduced inflammation [5,6], and the development of the immune system [7]. Notably, a recent research revealed that bifidus-flora formation is associated with years) by quantitative-PCR. With the exception of centenarians, B. bifidum was detected at all ages, but their prevalence was lower (28.3%) than that of the B. longum group (B. longum and B. infantis) (88.1%). A similar pattern was observed for the Bifidobacterium catenulatum group. Nagpal et al. [45] have examined the early-life dynamics of the bifidobacterial population in 76 full-term vaginally-born Japanese infants. They showed that B. bifidum became prevalent (~60%) six-months after birth and became less prevalent (~20%) after 3 years. B. breve was detected in 71.4% of children under 3 years old and its prevalence was consistently high in individuals younger than 10, but the number of B. breve cells decreased with age and was very scarce past the age of 50. The prevalence of B. breve also decreased during the transition from childhood to adulthood.
Collectively, B. breve and B. longum are, in general, more abundantly detected than B. bifidum during the first three years of life. The prevalence of B. breve is, however, largely limited to infants. B. longum is widely distributed across age groups and frequently found in adult fecal samples. B. bifidum is also widely distributed across age groups, but with a lesser prevalence than the B. longum group. B. adolescentis and the B. catenulatum group are also abundant in adult intestinal microbiota. This transition of bifidobacterial species in human intestinal microbiota with age is primarily due to changes in diet and possibly host glycans (described below). Considering the age-related compositional changes and nutritional adaptation, Bifidobacterium species can be classified into at least four groups: the HMO-dedicated group (B. infantis), the dietary fiber-adapted group (B. adolescentis, B. catenulatum group, and other adult-type species except B. longum), the both HMO and dietary fiber-adapted group (B. breve and B. longum), and the both HMO and mucin-adapted group (B. bifidum). Thus, as the sole known mucin-degrader within the genus Bifidobacterium, B. bifidum is uniquely positioned in the context of carbohydrate utilization strategies. This may be the reason it is detected in a wide range of age groups, except for centenarians. It is known that intestinal mucin production becomes low in elderly adults [46]. Therefore, the nutritional switch, as well as host glycan availability, may have an impact on bifidobacterial prevalence and abundance.

Structures of Human Milk Oligosaccharides and Mucin O-Glycans
"HMOs" is a collective term for the mixture of more than 100 structurally different free oligosaccharides (degree of polymerization ≥ 3) produced in the mammary gland [20,47]. The synthesis of HMOs begins with the modification or the elongation of lactose (Galβ1-4Glc). Fucosylation occurs at the C2 position of Gal or the C3 position of Glc of lactose by the activity of fucosyltransferases (FUTs) to form 2 -fucosyllactose (2 -FL) and 3-fucosyllactose (3-FL), respectively. Sialylation with N-acetylneuraminic acid occurs at the C3 or C6 position of Gal of lactose to form 3 /6 -sialyllactose (SL). The elongation or branching of lactose is initiated with the addition of N-acetylglucosamine (GlcNAc) onto the C3 position of Gal of lactose to form lacto-N-triose II, then subsequent addition of Gal onto the C3 position or the C4 position of GlcNAc occurs to form the type-1 chain unit (Galβ1-3GlcNAc or lacto-N-biose I, LNB)-containing lacto-N-tetraose (LNT, Galβ1-3GlcNAcβ1-3Galβ1-4Glc), or the type-2 chain LacNAc unit (Galβ1-4GlcNAc)-containing lacto-N-neotetraose (LNnT, Galβ1-4GlcNAcβ1-3Galβ1-4Glc). LNT and LNnT can then be subjected to further modification and/or elongation. GlcNAc branching also occurs at the C6 position of the Gal residue of Lac after the synthesis of lacto-N-triose II, and the branch is further elongated by the addition of Gal in a similar manner. When compared to the milk oligosaccharides of other mammals, HMOs are characterized by the richness of type-1 disaccharide units [20]. It should be noted that the genetic background of mothers significantly influences the structural variety of HMOs. For example, non-secretor FUT2-deficient mothers (Se-negative, 0% to 50%, depending upon ethnic background) do not produce H-antigen (Fucα1-2Gal-OR)-containing oligosaccharides such as 2 -FL or lacto-N-fucopentaose (LNFP) I in their milk [16], and FUT3-deficient mothers with the Lewis negative phenotype produce HMOs with differential amounts of Lewis antigen structures [48].
While HMOs are present as a free form in the aqueous phase of milk, mucin O-glycans are covalently attached to the polypeptide backbone through N-acetylgalactosaminyl Ser/Thr residues.
O-Glycosylation on mucin glycoproteins densely occurs in regions that are rich in the amino acid residues of Pro/Thr/Ser (mucin domain) [49], and the resulting overall molecular structures of mucin glycoproteins are thought to be shaped like bottle brushes. O-Glycans are synthesized by sequential addition of monosaccharides by the orchestrated action of glycosyltransferases in mucin-producing cells, i.e., goblet cells, and they are classified into several groups based on their "core" types (primarily core 1, core 2, core 3, and core 4, as reviewed by [50][51][52]). Therefore, the structure of mucin O-glycans is highly heterogeneous and it is also dependent on which mucin molecules they are attached to [53,54]; however, they still share a structural similarity in part with HMOs. O-Glycans contain many type-1/2 LacNAc units as terminal and internal structures, and fucosylation/sialylation as terminal modifications ( Figure 1). The structural similarities between HMOs and mucin O-glycans, which may reflect their overlapping function as a defense from pathogens by acting as anti-adhesive decoys [55], can also serve as prime targets for decomposition by glycan-degrading enzymes of B. bifidum to assimilate sugars. glycoproteins are thought to be shaped like bottle brushes. O-Glycans are synthesized by sequential addition of monosaccharides by the orchestrated action of glycosyltransferases in mucin-producing cells, i.e., goblet cells, and they are classified into several groups based on their "core" types (primarily core 1, core 2, core 3, and core 4, as reviewed by [50][51][52]). Therefore, the structure of mucin O-glycans is highly heterogeneous and it is also dependent on which mucin molecules they are attached to [53,54]; however, they still share a structural similarity in part with HMOs. O-Glycans contain many type-1/2 LacNAc units as terminal and internal structures, and fucosylation/sialylation as terminal modifications ( Figure 1). The structural similarities between HMOs and mucin O-glycans, which may reflect their overlapping function as a defense from pathogens by acting as anti-adhesive decoys [55], can also serve as prime targets for decomposition by glycan-degrading enzymes of B. bifidum to assimilate sugars.

Glycoside Hydrolases Involved in the Degradation of HMOs/Mucin O-Glycans
Biochemical studies of glycoside hydrolases (GHs) from B. bifidum have revealed the mechanism of adaptation to HMOs and mucin O-glycans. The reports on a GH95 1,2-α-L-fucosidase (AfcA) from B. bifidum JCM 1254 by Katayama et al. [56] and a galactose operon containing a gene encoding a GH112 galacto-N-biose (GNB)/LNB phosphorylase from B. longum JCM 1217 by Kitaoka et al. [57] were the key findings that revived attention on HMOs by linking them to bifidobacterial biology, long after the first notion, proposed in 1954 by Gauhe et al. [58], that milk oligosaccharides
Sialic acid residues are frequently modified with O-acetylation [71]. One of the GH33 sialidases, SiaBb1, contains an esterase-domain in its polypeptide sequence and was shown to remove the acetyl group from 9-O-acetylated N-acetylneuraminic acids [65]. Additionally, sulfated Gal and GlcNAc residues are present on intestinal mucin O-glycans [72]. We have recently identified a GH20 enzyme, BbhII, as a sulfoglycosidase that can remove terminal 6-sulfated GlcNAc residues from mucin [61].
Intracellular GH112 GNB/LNB phosphorylase acts on both GNB and LNB, converting them into galactose-1-phosphate and N-acetylhexosamine (GalNAc and GlcNAc, respectively) [57,73]. GNB is released from mucin core 1 by the action of extracellular GH101 endo-α-N-acetylgalactosaminidase (EngBF) [74], and LNB is formed not only from type-1 HMOs but also from mucin O-glycans by the action of LnbB. GNB and LNB are known to be imported by the GNB/LNB transporter, an ATP-binding cassette-type (ABC) transporter [75,76]. GH129 α-N-acetylgalactosaminidase (NagBb) is able to act on α-N-acetylgalactosaminyl Ser (Tn antigen) to liberate GalNAc and Ser [26,77]. GalNAc-Ser/Thr can be generated from core 3 or core 4 mucin O-glycans after complete trimming by concerted actions of GHs and peptidases. NagBb does not have a signal peptide, indicating that the hydrolysis of GalNAc-Ser/Thr takes place in the cytosol after the uptake of GalNAc-Ser/Thr into the cell by an unknown transporter.

Enzymatic Adaptation to HMO-and Mucin O-Glycan Degradation from the View of the "Spectra" of GHs and CBMs
The complete genomic sequence of B. bifidum was determined for the strain JCM 1255 by Hattori's group [78] and for the strain PRL2010 by Ventura's group [25], and, currently, a total of eleven sequences are available in public databases (NCBI genome assembly database, as of October 2019). A comparison of the annotated GH genes with those of other human-associated Bifidobacterium highlights its distinct mechanism that is exclusive to the assimilation of HMOs/mucin O-glycans. At the genus level, the total number of GHs ranges from 40 to 60 in the genomes of Bifidobacterium; however, the "GH spectrum", or the distribution of GH families and the cellular localization of those GHs, is species-dependent and should reflect its carbohydrate preference and distinct assimilation mechanisms. As visualized in Figure 2, B. bifidum [strains PRL2010, LMG 13195 (= JCM 7004), and TMC 3115] has 14 extracellular GHs involved in HMO/mucin O-glycan degradation, as predicted in silico, whereas B. infantis ATCC 15697 (= JCM 1222) and B. breve UCC2003 are equipped with the intracellular counterpart enzymes (e.g., GH33 sialidase and GH95 1,2-α-l-fucosidase). Indeed, B. infantis ATCC 15697 (= JCM 1222) has been shown to assimilate HMOs but not mucin that was supplemented in vitro [25]. Furthermore, it is worth noting that although some GHs, including GH33 sialidases, GH2 β-galactosidase, and GH20 β-N-acetylhexosaminidase, are also able to act on complex N-glycans, B. bifidum does not contain N-glycan-specific enzymes such as GH38 α-mannosidase or GH18/GH85 endo-β-N-acetylglucosaminidase. Among infant-gut associated Bifidobacterium, B. breve and B. infantis possess such N-glycan-specific GHs and are reportedly able to grow on media containing N-glycosylated proteins [79]. B. bifidum, B. breve, and B. infantis are all adapted to HMO utilization. However, in terms of the degradation of glycan chains attached to glycoproteins, B. breve and B. infantis are specific to N-glycans, while B. bifidum prefers O-glycans to perhaps avoid competition with other Bifidobacterium species.  Interestingly, the number of carbohydrate-binding modules (CBMs) in B. bifidum is also higher than that of other species: 20-23 in B. bifidum vs. six in B. infantis ATCC 15697, eight in B. longum JCM 1217, and nine in B. breve UCC2003 (Figure 3). In general, the CBMs are present in tandem with the peptide chains of the GH domains and serve to enhance the catalytic activities of GHs by increasing their affinity toward the polymeric carbohydrates [80]. In particular, B. bifidum has 13 CBM32 domains that are specifically associated with the recognition of Gal/Lac/LacNAc and their derivatives [81], possibly to enhance catalytic activities for mucin O-glycans. The quantity of CBMs is outstanding within the genus Bifidobacterium. This could represent the evolutionary adaptation toward mucins and the basis for carbohydrate-preference of B. bifidum. Interestingly, Nishiyama et al. [42] found that SiaBb2 from B. bifidum can interact with α2,6-linked sialic acid and the blood group A trisaccharide moieties by the GH33 sialidase domain, not the CBM domain, suggesting that there are more unidentified protein sequences involved in the interaction with glycans.   These HMO/mucin-related GHs are distributed separately within the B. bifidum genomes, rather than being clustered [25]. However, HMO/mucin supplementation induces the expression of some of these genes [25,61,82]. Although the mechanism of the transcriptional regulation remains unclear, a conserved 24 bp sequence is found upstream of these genes, which suggests the presence of a common transcriptional regulator(s) with a shared ligand(s), presumably a sugar metabolite(s) [25].

Cross-Feeding of the Oligosaccharide Degradants among Bifidobacterial Community
A repertoire of the transporters may also be an important factor that dictates the carbohydrate availability for bacteria. Although the GNB/LNB [75,76], fucosyllactose [12,24,83,84], and LNnT transporters [85] have been identified in Bifidobacterium, our knowledge regarding HMO/mucin-related transporters remains limited. However, Turroni et al. [86] reported that the genome of B. bifidum PRL2010 contains 25 genes predicted to encode components of a transport system dedicated to carbohydrate uptake. These are classified into different groups: eight genes of ABC-type transporters, 12 genes for a component of the phosphoenolpyruvate-phosphotransferase systems, one major intrinsic protein, and four secondary carriers. The number of genes predicted to encode the proteins for carbohydrate uptake in B. bifidum PRL2010 is relatively small when compared with the numbers of such genes in other species of human-associated Bifidobacterium. For example, B. infantis ATCC 15697 has 68 genes predicted to encode proteins for carbohydrate uptake, including 20 of the solute-binding protein family 1 of the ABC transporter [9], while B. bifidum PRL2010 has only 4 of the solute-binding proteins [86], which is a stark contrast with the numbers of extracellular GHs (22 extracellular GHs in B. bifidum PRL2010 vs. 10 in B. infantis ATCC 15697). The differences in the diversity and richness of sugar transporters and extracellular GHs may thus be complementary to carbohydrate acquisition strategies within the bifidobacterial community, which simultaneously allows for the differentiation of accessible carbohydrates amongst the community members. Further studies determining the substrate specificity of transporters will reveal how these transporters serve to define the adaptation strategy of each Bifidobacterium species for diet-derived and/or host-derived carbohydrates.

Cross-Feeding of the Oligosaccharide Degradants among Bifidobacterial Community
Regarding the bifidobacterial population in breast-fed infant guts, B. breve and B. longum are generally more abundant than B. infantis and B. bifidum. This observation is not paralleled with their in vitro growth ability in HMO-containing media, i.e., B. infantis and B. bifidum grow avidly with HMOs as a sole carbon source and obtain high cell density, but B. breve and B. longum show very limited growth in single culture experiments [87]. This apparent inconsistency between the abundance in the intestinal tracts and abundance in vitro suggest that there is a more complicated mechanism underlying the bifidus-flora formation in infant guts.
The HMO consumption behavior of B. bifidum was analyzed by in vitro culture experiments followed by a high performance liquid chromatography analysis of the sugars in the corresponding spent media. The results reveal the release of monosaccharides (Fuc, Gal, and Glc) and the disaccharides (Lac and LNB) into the media [28,87]. B. bifidum grew well on HMOs; however, high amounts of Fuc and Gal remained unconsumed, even after prolonged culture. The leftover HMO mono-and disaccharidic breakdown products in the spent media were observed for four different strains of B. bifidum, and thus it appears to be a common characteristic of this species [28]. B. infantis is able to consume almost all HMOs by directly internalizing them with their highly abundant ABC transporters, and degrading them using cytoplasmic GHs [9,24]. Most of the B. breve strains catabolize limited species of HMOs, such as LNT and LNnT through direct uptake and intracellular degradation [85,88]. B. longum also shows limited HMO utilization ability, and, in most cases, it consumes LNT only. However, recent studies revealed that several strains of B. breve and B. longum have the ability to assimilate 2 -FL, 3-FL, lactodifucotetraose (LDFT), and LNFP I using specific ABC transporters and intracellular GHs [12,24,83]. Interestingly, the corresponding transporter (fucosyllactose transporter) was shown to be enriched in breast-fed infant guts [24].
Tannock et al. [89] reported that the proportion of B. bifidum within the total Bifidobacterium population was positively correlated with the elevated abundance of the genus Bifidobacterium within the total microbiota of breast-fed infant guts, suggesting that B. bifidum plays an important role for an efficient bifidus-flora formation. Given the capability of B. bifidum to liberate sugars into the spent media (the surrounding environments), it was suggested that cross-feeding may occur to mediate the increased abundance of total Bifidobacterium. Indeed, it was shown that sialic acid and other sugars liberated by B. bifidum can be utilized by B. breve in in vitro co-cultures and in vivo mouse models [68,90,91], though B. breve does not grow in single culture experiments in media supplemented with mucin as a sole carbon source. Additionally, a correlation with the higher abundance of Bifidobacterium in the total microbiota was observed not only for B. bifidum in breast-fed infants, but also for B. longum, B. adolescentis, and B. catenulatum group in human subjects with a wide age range [44]. Interestingly, some B. longum and B. adolescentis strains were shown to have the capability to serve as cross-feeders of plant-derived polysaccharide degradants [92][93][94]. These observations suggested that intragenus cross-feeding might continue in the bifidobacterial community after weaning. As for extracellular GHs targeting plant-derived polysaccharides, B. longum JCM 1217 possesses eleven extracellular GH43 β-xylosidase/α-l-arabinofuranosidases (Figure 2). B. adolescentis ATCC 15703 and B. catenulatum DSM 16992 (= JCM 1194) have two extracellular GH13 (amylase family) and one extracellular GH121 (β-l-arabinobiosidase), respectively.
A recent study suggested that cross-feeding of the HMO/mucin degradation products is not limited to the bifidobacterial community. A Cutibacterium avidum isolate from infant feces was shown to produce a higher amount of propionate when co-cultured with B. bifidum in yeast extract, casitone and fatty acid (YCFA) medium containing HMOs than when cultured alone in the same medium, suggesting inter-genus utilization of B. bifidum-released sugar metabolites [95].

Co-Culture Experiments
The co-culture experiments in our previous study showed that B. bifidum supported the growth of other bifidobacterial species. B. longum 105-A strain grew in medium containing 1% HMOs as the sole carbon source, but only when it is co-cultured with B. bifidum [28] (Figure 4A,B). The data were reproduced from [28]. The supportive growth effect of B. bifidum for B. longum was also observed in medium containing 1% porcine gastric mucin (PGM) as a sole carbon source ( Figure 4C,D). To examine how LNB and GNB disaccharides liberated from HMOs and mucin O-glycans affect the growth of B. longum by cross-feeding from B. bifidum, we disrupted the gltA gene that encodes the solute-binding protein of the GNB/LNB transporter of B. longum [75,76] and made the knockout strain compete against the wild-type strain. Unexpectedly, in the growth competition assays in the presence of B. bifidum, the colony-forming units of ∆gltA strain at 12 h culture were comparable with that of the wild-type in the media containing either HMOs or mucin ( Figure 4E,F). These results indicate that the utilization of GNB/LNB was not prioritized by B. longum in the presence of many other sugars, such as Gal, GlcNAc, or lactose released from HMOs/mucin O-glycans under our experimental conditions. It should be noted that the growth ability on LNB was completely abrogated by the disruption of gltA (Supplementary Figure S1), which is well-conserved among the infant gut-associated Bifidobacterium [11,96].  A,B) or spectinomycin (C-F) resistance gene on a plasmid (pBFS38 or pBFO2, respectively) [97] was used for monitoring the growth. Colony-forming units (CFU) of B. longum 105-A were determined by spreading the serial dilution of the cultures on the agar plates containing the antibiotics (closed circles), while the CFU of total Bifidobacterium was determined using the agar plates without antibiotics (open circles). The data used in (A) was obtained from our previous study [28]. (E,F): The growth competition between wild-type and ΔgltA mutant strains of B. longum 105-A in the presence of B. bifidum JCM 1254. Lactose (Lac), HMOs, or PGM was used as the sole carbon source. The ratio of the growth on HMOs (E) or PGM (F) that was normalized by that on Lac was compared. The ΔgltA mutant was transformed with pBFS38 [97] carrying the chloramphenicol resistance gene, and CFU was determined on agar plates containing the antibiotics.

Conclusions
B. bifidum resides in the gastrointestinal tract of humans of a wide age range, yet generally does not dominate the intestinal microbiota, like B. longum and B. breve do during infancy. Instead of being the dominant species in the gut ecosystem, however, it influences the microbiota composition  [97] was used for monitoring the growth. Colony-forming units (CFU) of B. longum 105-A were determined by spreading the serial dilution of the cultures on the agar plates containing the antibiotics (closed circles), while the CFU of total Bifidobacterium was determined using the agar plates without antibiotics (open circles). The data used in (A) was obtained from our previous study [28]. (E,F): The growth competition between wild-type and ∆gltA mutant strains of B. longum 105-A in the presence of B. bifidum JCM 1254. Lactose (Lac), HMOs, or PGM was used as the sole carbon source. The ratio of the growth on HMOs (E) or PGM (F) that was normalized by that on Lac was compared. The ∆gltA mutant was transformed with pBFS38 [97] carrying the chloramphenicol resistance gene, and CFU was determined on agar plates containing the antibiotics. Further evidence for cross-feeding between B. bifidum and other species was provided by fecal culture experiments [28]. We collected fecal samples from healthy Japanese infants, children, and adults, and then incubated the samples in media supplemented with Glc or HMOs as a sole carbon source, with and without the addition of B. bifidum strains. After 24 h of culture, we quantified the abundance of Bifidobacterium species other than B. bifidum by quantitative-polymerase chain reaction (qPCR). The results indicate that the addition of B. bifidum markedly increased both the abundance and ratio of endogenous Bifidobacterium species other than B. bifidum in the samples. Importantly, this enhancement was observed only for HMO-containing media but not for Glc-containing media. Additionally, in the presence of deoxyfuconojirimycin (DFJ), a potent inhibitor for α-fucosidases (GH29 1,3-1,4-α-l-fucosidase and GH95 1,2-α-l-fucosidase), the bifidus-flora formation-promoting effect exerted by B. bifidum was abrogated. Since most of the abundant HMOs, such as 2 -FL, 3-FL, LDFT, LNFP, and LNDFH are modified with fucose, the removal of the fucose residues should be a critical step for other GHs to further degrade the sugars. Taken together, B. bifidum can enrich the bifidobacterial population and abundance by acting as a cross-feeder of the HMO-and mucin O-glycan degradants produced by its membrane-associated GHs with CBMs.

Conclusions
B. bifidum resides in the gastrointestinal tract of humans of a wide age range, yet generally does not dominate the intestinal microbiota, like B. longum and B. breve do during infancy. Instead of being the dominant species in the gut ecosystem, however, it influences the microbiota composition through the altruistic cross-feeding of host glycan degradants. Apart from the microbe to microbe interactions, recent studies have revealed that B. bifidum has immunomodulatory activity by inducing regulatory T-cells by microbe-host interaction with its cell-surface polysaccharides [98]. The production of indole lactic acid, an AhR ligand, by infant gut-associated Bifidobacterium species, including B. bifidum, was also recently reported [8,99]. Further investigation is needed to understand the physiology of this species in the gut ecosystem and its role in host interactions.

Disruption of gltA in B. longum 105-A
A markerless gene disruption of gltA (BL105A_1604) was carried out by the methods described previously [24].
The upstream and downstream regions of gltA were amplified by PCR from the genome of B. longum 105-A (JCM 31944) [101] using the primer pairs of Pr-15/16 (5 -cggtacccggggatcatctctact ccttcgtagtgaaatc-3 and 5 -caggcatgcaagctttaacatgcggtgtccccgttg-3 ) and Pr-11/12 (5 -tatatatgagtactgatgcgaccacgcccggaatg-3 and 5 -cgagtcgctcgaattcattaacggttagggttccttccag-3 ), respectively. The underlined sequences represent 15 bp-extensions required for In-Fusion cloning (Clontech Laboratories, Mountain View, CA, USA). The amplified fragments were ligated with the replicon of the conditional replication plasmid pBS423 ∆repA so that the marker gene (spectinomycin resistance) was sandwiched between the upstream and downstream regions of gltA [24,102]. Subsequent integration into the genome and excision of the marker gene from the genome were conducted as described previously [24]. Disruption at the targeted site was confirmed by genomic PCR using the primer pair Pr-36/37 (5 -gtcgccgaagaagttcaccttg-3 and 5 -ttctggaacctgtcttcttgattgc-3 ) and Western Blot analysis using anti-GltA antibodies [87] (Supplementary Figure S1).

Co-Culture of B. longum with B. bifidum
The wild-type (WT) strain of B. longum 105-A with a plasmid carrying the spectinomycin resistance gene on a plasmid (pBFO2) [97] was cultured in basal medium [87] supplemented with 1% porcine gastric mucin (PGM, Sigma-Aldrich, MO, USA) as a carbon source in the presence and absence of B. bifidum JCM 1254. During incubation, aliquots were collected, serially diluted, and spread on GAM agar plates (Nissui Pharmaceutical, Co., Ltd., Tokyo, Japan) supplemented with and without 30 µg/mL of spectinomycin. The colonies appearing on the antibiotic-containing plates were attributed to B. longum cells, while those formed on antibiotic-free plates were assumed to represent the sum of B. longum and B. bifidum cells. When the ∆gltA derivative of B. longum 105-A was competed against its parental WT strain (with pBFO2 carrying the spectinomycin resistance gene [97]) for the growth on PGM in the presence of B. bifidum, pBFS38 with the chloramphenicol resistance gene [97] was used as the marker of the ∆gltA strain. WT and ∆gltA derivative were also cultured for competition in basal medium supplemented with 1% HMOs as a carbon source. During incubation, aliquots were collected, serially diluted, and spread on GAM agar plates containing the respective antibiotics to determine the CFUs of each strain. Spectinomycin and chloramphenicol were used at the final concentrations of 30 and 2.5 µg/mL, respectively.