Substrate preference of an ABC importer corresponds to selective growth on β-(1,6)-galactosides in Bifidobacterium animalis subsp. lactis

Bifidobacteria are exposed to substantial amounts of dietary β-galactosides. Distinctive preferences for growth on different β-galactosides are observed within Bifidobacterium members, but the basis of these preferences remains unclear. We previously described the first β-(1,6)/(1,3)-galactosidase from Bifidobacterium animalis subsp. lactis Bl-04. This enzyme is relatively promiscuous, exhibiting only 5-fold higher efficiency on the preferred β-(1,6)-galactobiose than the β-(1,4) isomer. Here, we characterize the solute-binding protein (Bal6GBP) that governs the specificity of the ABC transporter encoded by the same β-galactoside utilization locus. We observed that although Bal6GBP recognizes both β-(1,6)- and β-(1,4)-galactobiose, Bal6GBP has a 1630-fold higher selectivity for the former, reflected in dramatic differences in growth, with several hours lag on less preferred β-(1,4)- and β-(1,3)-galactobiose. Experiments performed in the presence of varying proportions of β-(1,4)/β-(1,6)-galactobioses indicated that the preferred substrate was preferentially depleted from the culture supernatant. This established that the poor growth on the nonpreferred β-(1,4) was due to inefficient uptake. We solved the structure of Bal6GBP in complex with β-(1,6)-galactobiose at 1.39 Å resolution, revealing the structural basis of this strict selectivity. Moreover, we observed a close evolutionary relationship with the human milk disaccharide lacto-N-biose-binding protein from Bifidobacterium longum, indicating that the recognition of the nonreducing galactosyl is essentially conserved, whereas the adjacent position is diversified to fit different glycosidic linkages and monosaccharide residues. These findings indicate that oligosaccharide uptake has a pivotal role in governing selectivity for distinct growth substrates and have uncovered evolutionary trajectories that shape the diversification of sugar uptake proteins within Bifidobacterium.

The human gut microbiota (HGM) 4 is increasingly acknowledged as a major contributor to human health as well as a modulator of host metabolism (1) and immune-homeostasis (2). Specific microbiota signatures are associated with serious disease states, including inflammatory disorders (3), obesity, type 2 diabetes (4), and colorectal cancer (5), underscoring the crucial impact of a balanced HGM composition. Diet is a major external factor affecting the composition of the microbiota (6), and oligo-as well as polysaccharides that are nondigestible to the human host are instrumental in shaping this community (7). The abundance and the diversity of dietary glycans have driven the evolution of different microbial strategies to harvest energy from these metabolic resources. Dominant commensals from the Bacteroides genus are considered generalists that target a broad range of glycans via outer membrane-bound extracellular enzymes (7,8). Other taxonomic groups (e.g. bifidobacteria) typically possess fewer extracellular enzymes compared with Bacteroides. Instead, bifidobacteria rely on effective transport systems, mainly ATP-binding cassette (ABC) transporters, to access oligosaccharides that are either naturally present in diet (9) or produced by hydrolytic enzymes from other HGM members, thereby facilitating cross-feeding (10,11). Genomes of bifidobacteria typically encode a multitude of oligosaccharide-specific ABC transporters (12), which mediate high-affinity capture of oligosaccharides via their extracellular lipid-anchored solute-binding proteins (SBPs). Uptake of oligosaccharides by these transporters is likely to confer an important advantage in the competitive human gut niche, particularly in the colon, which harbors the highest bacterial density. Despite this key role, our insight into oligosaccharide uptake in the human gut niche remains limited.
␤-Galactosides are ubiquitous in the human diet (e.g. in both human and animal milk) (13 and in fruits and vegetables, where they are omnipresent in primary cell wall pectic polysaccharides (14)). Oligomeric ␤-galactosides of different origin have been shown to selectively promote members of the Bifidobacterium genus that harbor many probiotic health-promoting strains (15). Dietary ␤-galactosides display source-dependent diversity with respect to degree of polymerization (DP), glycosidic linkages, and the presence of other monosaccharides. ␤-Galactosides are also considered prime prebiotic dietary supplements (galactooligosaccharides (GOS)) produced by transglycosylation of lactose, partially due to their bifidogenic effect. Notably, commercial GOS preparations typically comprise tens of compounds of DP 2-6 and contain ␤-(1,3)-, ␤-(1,4)-, and ␤-(1,6)-galactosidic linkages (16). Only some of these oligosaccharides are metabolized by bifidobacteria in a species-and strain-specific manner (17)(18)(19). Knowledge about ␤-galactoside preferences of bifidobacteria largely stems from characterization of ␤-galactosidases (20) or from growth and oligosaccharide consumption profiles (19). The role of oligosaccharide transporters in ␤-galactooligosaccharide utilization selectivity has received less attention, with two recent studies having addressed the identification and binding properties of ␤-galactooligosaccharide transport proteins in Bifidobacterium breve (21,22).
We have previously identified a transcriptionally induced operon during growth of the probiotic strain Bifidobacterium animalis subsp. lactis Bl-04 on GOS (a commercial ␤-galactoside mixture). This operon encodes a transcriptional regulator, a ␤-galactosidase of glycoside hydrolase family 42 (BlGH42A), and an ABC transporter (23). The ␤-galactosidase revealed a broad specificity, hydrolyzing ␤-(1,6)-, ␤-(1,3)-, and ␤-(1,4)galactosidic bonds with a modest 5-fold preference for ␤-(1,6)galactobiose (␤6Gal2) as compared with the ␤-(1,4) isomer (24). Here, we report the molecular characterization of the ␤-galactoside-specific SBP (Bal6GBP) associated with the ABC transporter. In contrast to the relatively promiscuous ␤-gal, Bal6GBP displayed selectivity exceeding 3 orders of magnitude for ␤6Gal2 as compared with the ␤-(1,4) isomers, which was explained by the structure of Bal6GBP in complex with ␤6Gal2. The high selectivity of Bal6GBP is reflected in a dramatic difference in growth profiles of B. animalis subsp. lactis Bl-04 showing several hours of lag phase on less preferred substrates. This was shown to be attributed to the inefficient uptake of ligands that are less preferred by the binding protein.
Structural analyses allowed tracking the evolutionary relationship with the lacto-N-biose (human milk-derived) binding protein from Bifidobacterium longum subsp. longum (25), which shares only 26% sequence identity with Bal6GBP. Despite the low primary structure identities, these two binding proteins share a similar architecture of the side chains interacting with the nonreducing galactosyl moiety but diverge in the recognition of the penultimate carbohydrate ring from the nonreducing end.
Altogether, the present study constitutes a unique case enabling the comparison of ligand selectivity of an ABC-associated transport protein and the intracellular glycoside hydrolase mediating uptake and breakdown, respectively, of the same substrates. These findings underscore the leading role of ABCmediated glycan transport in governing the growth preferences of core gut microbiota members on ␤-galactosides. This insight paves the way to tailored editing of specific members of the microbiota based on uptake profiles in future interventions to promote health and to alleviate disease.

Overall three-dimensional structure of Bal6GBP in complex with the preferred ligand
The crystal structure of Bal6GBP in the free open conformation was solved to a resolution of 1.94 Å by SeMet single anomalous dispersion phasing. Coordinates of the separate domains from the open conformation were used as molecular replacement (MR) models to solve the structure of Bal6GBP in complex with the preferred ligand ␤6Gal2 to 1.39 Å (Table 3).
Bal6GBP adopts a classical SBP-fold (cluster B type SBP according to structural classification (26)), which comprises two ␣/␤ domains of different sizes interconnected by three hinge regions with the ligand binding site located at the domain interface (Fig. 3A). The smaller domain 1 (residues 1-144 and 311-362) is formed by six ␣-helices and five centrally positioned ␤-strands. Domain 2 (residues 149 -306

Table 2
Binding energetics of Bal6GBP to ␤6Gal2 and ␤6Gal4 at 25°C and pH 6.5 using ITC a Average of three experiments. The deviation of the stoichiometry from 1:1 is likely due to inaccuracy in the ligand concentration, which makes the thermodynamic parameters from this measurement of this ligand less reliable. longum ATCC 15697 (PDB code 2Z8D-F; z-score ϭ 42.0; RMSD ϭ 2.8 Å; 26% sequence identity) as the structurally closest related homologue to Bal6GBP (25). The structure of GL-BP was determined in complex with galacto-N-biose (PDB code 2Z8E), lacto-N-biose (PDB code 2Z8D), and lacto-N-tetraose (PDB code 2Z8F).
The crystal obtained for phasing from the SeMet-labeled protein belongs to the same space group as the ligand-bound form, but with a different ␤ angle and slightly different cell lengths ( Table 3). The structure of the open form of the protein was determined from the SeMet protein crystals, which revealed high flexibility between domain 1 and domain 2, resulting in an open ligand-free form and a closed ligand-bound form (29). The transition between these two states involves a rotation angle of 32.3°between the domains of Bal6GBP, as estimated with the DynDom server (30) (Fig.  S1). Thus, closure of the protein involves a rigid-body movement of domains with changes mainly occurring in the hinge region.

␤-(1,6)-Galactobiose-binding site in Bal6GBP
Bal6GBP in complex with ␤6Gal2 showed a well-defined density of the ligand that primarily assumes the ␤-configuration. The ligand is almost completely engulfed in the largely solvent-inaccessible binding site (Fig. S2). The nonreducing galactosyl stacks onto Trp-275 from domain 2, which defines position 1. A total of 12 potential hydrogen bonds recognize ␤6Gal2: three from domain 1, six from domain 2, and three from the hinge region (Table S2 and Fig. 3B). Five of the six water molecules that coordinate the ligand also mediate hydrogen-bonding networks to residues in the binding pocket. All hydroxyl groups of the galactosyl at position 1 are recognized by polar interactions to Gln-96, Glu-98, Asp-145, Asn-279, Gly-310, and Ser-311. This correlates well with the strict requirement for ␤-galactosyl at position 1 for recognition by Bal6GBP. By contrast, only the C2-OH and C4-OH of the EDITORS' PICK: ␤-Galactoside capture in B. animalis subsp. lactis galactosyl at position 2 are recognized by direct polar contacts to Asn-74, Glu-98, Asp-196, and Asn-308. Glucosyl can be accommodated at position 2, albeit with 140-fold reduced affinity for the disaccharide allolactose as compared with ␤6Gal2 (Table 2), probably due to the recognition of the axial C4-OH in ␤6Gal2 by Asn-74 and Glu-98, which also interacts with the C2-OH from the reducing end galactosyl unit. Both hemiacetal ring oxygens in ␤6Gal2 form hydrogen bonds with indole nitrogens of tryptophans (Trp-47 and Trp-254; Fig. 3B).

Transporter preference is reflected in the ␤-galactobiose isomer growth profiles of B. animalis subsp. lactis Bl-04
Only two loci were previously observed to be highly up-regulated during growth on galactooligosaccharides, the canonical lactose metabolism locus encoding a LacS transport system and the ABC system characterized in the present study (23), suggesting that this ABC system is the only ␤-galactoside-specific locus. To evaluate the impact of the preference of Bal6GBP on growth of B. animalis sp. lactis Bl-04, the bacterium was cultivated for 48 h on the ␤-(1,3), ␤-(1,4), or ␤-(1,6) structural isomers of ␤-galactobiose as a carbon source (Fig. 4A). Starting from the same inoculum, the most rapid growth was on ␤6Gal2 with no observable lag phase. Conversely, growth on ␤3Gal2 and ␤4Gal2 commenced after lag phases of 9 and 15 h, respectively. We carried out another growth experiment to verify whether the growth delay on less preferred ligands was related to inefficient transcriptional response or due to poor uptake. The growth in this case was performed on 0.5% ␤6Gal2, 0.2% each of ␤4Gal2 and ␤6Gal2, or 0.45 and 0.05% of ␤4Gal2 and ␤6Gal2, respectively (all w/v) (Fig. 4B). The growth level was proportional to the amount of the preferred ␤6Gal2 after 40 h, although the final one-point OD measurement before the culture harvest at about 48 h was higher (data not shown) compared with the 40-h sample, consistent with the delayed growth pattern observed on ␤4Gal2 (Fig. 4A). The reason for using the lower concentration in the 1:1 galactobiose mixture was to have higher sensitivity in the uptake assay for monitoring uptake during the exponential phase. The uptake preference was evaluated by analysis of supernatants from a culture growing on the same equimolar mixture of ␤4Gal2 and ␤6Gal2 as in Fig. 4B. The preferred ␤6Gal2 was depleted after 12 h, whereas consumption of ␤4Gal2 was only observed after 24 h and was not completed after 72 h of growth (Fig. 4C). These experiments suggest that the uptake preference is a key factor that governs the onset and the extent of growth on both pure substrates and mixtures of those.

Transporter proteins as sensitive probes of the evolution of microbial glycan preferences in the gut niche
The evolution of bifidobacteria has been shaped by the acquisition of genes that support the adaptation to the digestive tracts of animals that provide parental care, especially mammals (31,32). The evolution of unique glycan utilization capabilities by Bifidobacterium members underscores the impact of metabolic specialization as a driving force for species differentiation and niche adaptation (12). Considerable acquisition of carbohydrate transporters, especially of the ABC type, has been proposed in Bifidobacterium (12). The evolution of oligosaccharide transporter specificities to support the adaptation of EDITORS' PICK: ␤-Galactoside capture in B. animalis subsp. lactis bifidobacteria to distinct niches in host guts has not received sufficient attention. Transport proteins are suitable probes of the evolution of glycan metabolic preferences due to their superior selectivities as demonstrated in the present study. Biochemical and structural characterization of carbohydrate transporters is key to enable the assignment of metabolic preferences based on genome sequence data. To illustrate this, we performed structural and multiple sequence alignments of Bal6GBP and GL-BP, the homologue conferring uptake of the mucin-derived GNB and the human milkderived LNB from B. longum sp. longum (25), together with closely related sequences that cluster with these two characterized proteins in the phylogenetic analysis (Fig. S3). The aromatic platform, which is the hallmark of carbohydrate recognition by SBPs at position 1, is strictly conserved in both the primary and tertiary structural alignment together with the four residues recognizing the nonreducing ␤-galactosyl (Fig. 5, C and D). This striking conservation unveils the evolutionary pressure to retain the functionality of binding subsite 1 of these two proteins despite the divergence of primary structure (26% sequence identity). Another tryptophan (Bal6GBP Trp-254) is also fully conserved in the sequence alignment (Fig. 5A), but this residue serves two different roles at position 2: aromatic stacking onto the N-acetyl hexosamine ring in GL-BP, whereas in Bal6GBP, the indole ring is orthogonal to the galactosyl plane and forms a polar interaction with the hemiacetal oxygen.
A key difference is the substitution of Trp-47 in Bal6GBP with Arg-49 in GL-BP. Importantly, Arg-49 recognizes the N-acetyl group unique to GL-BP substrates via direct and water-mediated hydrogen bonds (Fig. 5B). This residue together with Glu-110 also engages the galactosyl at position 1 (ring oxygen and C2-OH) and the C4-OH of the N-acetyl hexoseamine at position 2. This arginine is strictly conserved in close homologs of GL-BP, thereby constituting a specificity signature. Another facet underpinning the divergence of GL-BP toward LNB/GNB recognition entails amino acid changes that reduce the volume of side chains flanking the bulky N-acetyl group to enable its accommodation.
The structure of Bal6GBP also gives an explanation for the extreme preference for the longer ␤-(1,6)-galactosidic bond that allows higher rotational freedom between the two monosaccharide rings as compared with ␤4Gal2, with ␤3Gal2 having intermediate rotational freedom. The rotational flexibility of ␤6Gal2 allows the positioning the sugar planes at an angle of about 140°, a conformation that is enforced by Glu-98. This residue locks the orientation of the galactosyl at position 2 through a bidentate polar interaction that connects the galac-  (Fig. 3B). Notably, the position of Glu-98 is fixed by two highly conserved tyrosines (Tyr-78 and Tyr-100) that are from rigid ␣-helices, which is likely to be important for the high selectivity for ␤6Gal2. The preference for galactosyl at position 2 is mediated by a polar interaction between the axial C4-OH and Asn-74. A 140-fold weaker affinity is observed when the axial C4-OH is replaced by an equatorial one in allolactose, possibly due to the loss of the hydrogen bond to Asn-74. The glycosidic bond to an axial OH group in ␤4Gal2 restricts the rotation, rendering this ligand more rigid. Therefore, binding of ␤4Gal2 is associated with a large penalty due to conformational rearrangements of the protein and a strained conformation of the ligand. Comparing Bal6GBP with the ␤-galactosyl-(1,4)-lactose-specific homolog from B. breve (21) reveals that the hydrophobic platform and six of seven residues that bind the galactosyl at position 1 are also conserved in the primary structure (cluster 5 in Fig. S3). Strikingly, however, the double tyrosine motif and Asn-74 discussed above are variable, highlighting the sequence divergence underpinning the evolution of transporters for recognition of different ␤-galactosides.

The selectivity of transport binding proteins correlates to rapid growth and competitiveness on preferred substrates
Dietary glycans are instrumental in shaping the composition of the HGM (7,8). For bacteria that lack extracellular polysaccharide-degrading enzymes, high-affinity capture and transport of oligosaccharides is crucial to cross-feed on glycan olig-omers produced by enzymes of primary degraders. The early dominance of bifidobacteria in the human gut relies partially on their arsenal of ABC transporters that confer the uptake of abundant human milk galactosides (33,34). The competitiveness of bifidobacteria on preferred substrates for their ABC transporters was recently demonstrated in B. animalis subsp. lactis in co-culture with the human gut symbiont Bacteroides ovatus during growth on raffinose (9). This transporter is highly conserved in adult human gut bifidobacterial species, consistent with the stimulation of bifidobacteria on ␣-raffinose family ␣-galactosides from legumes. Similar observations were recently reported in other taxa (e.g. competitive growth of the xylan-degrading symbiont Roseburia intestinalis grown on a preferred ligand for its xylooligosaccharide ABC transporter in a mixed culture with B. ovatus) (35). Most bifidobacteria lack extracellular xylanases but are able to cross-feed on shorter arabinoxylooligosaccharides produced by other primary degraders using an efficient ABC system (10,11), further supporting the pivotal role of ABC transporters in bifidobacterial adaptation as secondary degraders that efficiently target oligosaccharides.
Bifidobacteria selectively ferment different dietary ␤-galactosides (e.g. B. animalis subsp. lactis grows well on ␤-(1,6)galactosyl lactose, but not on the ␤-(1,4) version of this trisaccharide (both from bovine milk)) (36). Recent studies have identified an ABC transporter that mediates the uptake of ␤-(1,6)-galactosyl-lactose from B. breve and characterized the preference of the associated SBP (22,37). Notably, B. breve was EDITORS' PICK: ␤-Galactoside capture in B. animalis subsp. lactis able to grow comparably well on the ␤-(1,3)-, ␤-(1,4)-, and ␤-(1,6)-galactosyl-lactose (GL) isomers. Different ABC transporters appear to mediate this utilization, and the closest homolog of Bal6GBP (WP_003828494.1; amino acid sequence identity 76%) showed the highest affinity for ␤-(1,6)-GL (K D ϭ 0.357 M). This protein has a markedly different selectivity as compared with Bal6GBP, based on a decrease in affinity of only 16-fold for ␤-(1,4)-GL and 460-fold lower affinity for ␤-(1,3)-GL (22). The shared preference for ␤-(1,6)-galactosides, but large selectivity differences, are consistent with these SBPs populating different branches in the same clade in the phylogenetic tree (clade 1, Fig. S3). The notable difference in selectivity despite the relatedness of these homologues highlights the need for more structural and biochemical data for reliable assignments of transport proteins from this niche. The second closest homolog (WP_003828303.1; sequence identity 45%) is specific for only ␤-(1,4)-galactosides with the highest affinity for ␤4Gal2 (K D ϭ 23.7 nM) followed by ␤-(1,4)-GL (K D ϭ 10.4 M) (21). ␤-(1,6)-Galactoside motifs are present in plant cell wall pectic arabinogalactans (types I and II) and arabinogalactan proteins (type II) (38,39). We have identified homologs (ϳ80% identity) of galactan ␤-(1,3)-galactosidase (40) and an endo-␤-(1,6)-galactanase (41) in the Human Microbiome Project database (42), suggesting that ␤-(1,6)-galactosides are likely to be accessible degradation products from especially type II arabinogalactan in the diet. The about 100-fold lower selectivity of the B. breve closest homolog for the ␤-(1,6)galactosides may suggest an adaptation to bovine milk, whereas the strict selectivity of Bal6GBP may suggest an adaptation to different ligands (e.g. from pectin). The selectivity of Bal6GBP for ␤6Gal2 (1630-fold preference over ␤4Gal2) demonstrates the role of oligosaccharide transporters in defining metabolic preferences as compared with intracellular enzymes, which may display more promiscuity (24). The preference of Bal6GBP correlates to rapid growth on the preferred ligand, whereas relative delay of several hours was observed on less-preferred ligands (Fig. 4A). The preferential depletion of ␤6Gal2 before ␤4Gal2 (Fig. 4C) provides compelling evidence that the large delay in the onset of growth on ␤4Gal2 is attributable to the inefficient uptake. The impact of the poor uptake is amplified by the lower affinity of the intracellular ␤-gal. The observed correlation between the selectivity of transport systems and early growth efficiency is likely to be crucial for resource allocation in the competitive human gut niche.
In conclusion, the study gives an unprecedented insight into the selectivity of transport proteins as compared with intracellular enzymes targeting the same substrates and the large impact of this selectivity on bacterial growth patterns. Access to the first structure of a ␤6Gal2-binding protein uniquely allowed analyses of the structural elements that support the evolutionary divergence of bifidobacterial ␤-galactoside transport proteins, which is otherwise challenging due to low primary structure conservation. Altogether, these findings promote our understanding of the structural and functional aspect of oligosaccharide transport in prevalent core microbiota.

Carbohydrate ligands and chemicals
The carbohydrates tested for binding interactions with Bal6GBP are listed in Table S1 and were 94 -99% pure. All other chemicals were of analytical grade.

SPR
SPR-binding analysis was performed using a Biacore T100 (GE Healthcare). Bal6GBP (0.1 mg ml Ϫ1 ) in 10 mM sodium acetate, pH 4.0, was immobilized on a CM5 chip by random amine coupling (GE Healthcare) to 2385-3448 response units. Initial binding screening was conducted at a flow rate of 30 l min Ϫ1 and 10 mM carbohydrate ligand at 25°C with association and dissociation steps of 60 and 150 s, respectively. Subsequently, saccharides that displayed binding were analyzed EDITORS' PICK: ␤-Galactoside capture in B. animalis subsp. lactis using 10 concentrations in the range of 0.05-10 K D in at least duplicates. The pH and temperature dependence of ␤-(1,4)galactobiose binding to Bal6GBP was analyzed at eight temperatures (15-40°C) and at nine pH values (3.6 -8.2) in 20 mM phosphate citrate buffer. The BIA Evaluation software (GE Healthcare) was used to analyze the data as described previously (10) to determine dissociation constants (K D ) as well as association (k on ) and dissociation rate (k off ) constants where possible.

ITC
Binding thermodynamics of Bal6GBP (8 -600 M) and preferred galactosides were measured at 25°C in 20 mM phosphate citrate, pH 6.5, using an iTC 200 microcalorimeter (MicroCal). The titration was initiated by a 0.3-l injection followed by 19 2-l injections of ligand. The concentrations used were 80 M ␤6Gal2 injected into 8 M Bal6GBP and 2.6 (or 2.4 in the duplicate experiment) mM ␤6Gal4 injected into 222 (or 246 in the dupilcate) M Bal6GBP. A one-binding site model was fitted to the ITC data after baseline adjustment and correction for the dilution enthalpy using Origin software 7.0552 supplied with the instrument to determine the equilibrium association constant (K A ), the stoichiometry of binding (N), and the molar binding enthalpy (⌬H).

Crystallization, data processing, and protein structure determination
Crystallization of Bal6GBP was initially tested with the purified protein in 10 mM MES buffer, pH 6.5, at a concentration of 10 mg ml Ϫ1 and 2.5 mM ␤6Gal2 with a reservoir from the JCSGϩ screen (Molecular Dimensions) set up using a Mosquito liquid-handling robot (TTP Labtech) with 200 nl of protein mixed with a 100-nl reservoir in MRC 3-well plates, and after several months at room temperature (297 K), a crystal was obtained in C12 (reservoir consisting of 10% PEG 1000 and 10% PEG 8000) that diffracted beyond 1.4 Å resolution. Data were collected at the I911-3 beamline at MAX-II (44), and a total of 390 images (195°of data) were collected to a maximum resolution of 1.5 Å on the edge of the square detector, which impacted the completeness of the data in the outer shell ( Table 3). The SeMet-labeled protein was crystallized in MRC 2-well plates using an Oryx 8 liquid-handling robot (Douglas Instruments) at a concentration of 20 mg ml Ϫ1 . Standard screens JCSGϩ, PACT, and Morpheus (Molecular Dimensions) were initially carried out. Drops consisting of 200 nl of protein and 100 nl of reservoir or 150 nl of each were set up and kept at room temperature (around 297 K). Crystals appeared under several conditions. Data were collected on a crystal obtained with the Morpheus screen, condition E4 (12.5% PEG 1000, 12.5% PEG 3350, 12.5% MDP, 30 mM each di-, tri-, tetra-, and pentaethyleneglycol with 0.1 M MES/imidazole at pH 6.5) at the ESRF on beamline ID29 (45) at a wavelength of 0.979 Å and a temperature of 100 K. A total of 360°were collected over 2400 images to a resolution of 2 Å. For both data sets, the diffraction data were processed and scaled with XDS (46) using XDSAPP (47). Both native and SeMet Bal6GBP crystals belong to space group P2 1 but are not isomorphous. Phenix.AutoSol (48) found a total of 19 selenium sites with a figure of merit of 0.382, and the density-modified map was used in phenix.autobuild (49) to produce a preliminary model for the SeMet-labeled protein. This model was split into domain 1 and domain 2 and used in molecular replacement with phaser. After automated model building with phenix.autobuild, the model was completed using Coot (50) and phenix.refine (51) and analyzed with Molprobity (52) (see Table  3). The Bal6GBP structures with ␤6Gal2 (PDB code 6H0H) and in the open SeMet form (PDB code 6Q5G) were submitted to the Protein Data Bank.

Galactobiose uptake preference of B. animalis sp. lactis Bl-04
Growth of Bl-04 was performed as previously described (9). Freeze-dried B. animalis subsp. lactis Bl-04 was grown anaerobically overnight at 37°C in Bifidus Selective Medium, and a 2% inoculum of the overnight culture was similarly propagated twice in de Man, Rogosa, and Sharpe (MRS) broth supplemented with 0.5 g liter Ϫ1 L-cysteine, before 2% (v/v) of this culture was used to inoculate a fresh culture in the modified MRS medium supplemented with either 0.5% (w/v) ␤3Gal2, ␤4Gal2, or ␤6Gal2 in total volumes of 200 l in a 96-well microtiter plate. In another experiment, the same procedure was used to inoculate 0.5% (w/v) ␤6Gal2 or mixtures of ␤6Gal2 and ␤4Gal2 (0.20% ϩ 0.20% or 0.05% ϩ 0.45%, respectively). Growth was continued for 48 h at 37°C and monitored at A 600 by the 2030 Multilabel Reader Victor TM X4 (PerkinElmer Life Sciences). The growth was performed in five replicates in both cases.
Preferential culture supernatant utilization of the ␤6Gal2 and ␤4Gal2 isomers by Bl-04 was analyzed by high-performance anion-exchange chromatography with pulsed amperometric detection (Dionex, Sunnyvale, CA) in a timedependent manner of a growth experiment, performed in Eppendorf tubes, otherwise as described above with a 0.5% (w/v) equimolar mixture of ␤4Gal2 and ␤6Gal2, and aliquots were collected for 72 h. An injection volume of 10 l onto a Carbopac PA200 column and a mobile phase (0.35 ml min Ϫ1 ) of constant 100 mM NaOH and a gradient of sodium acetate of 0 -30 min at 0 -100 mM and 30 -35 min at 100 -400 mM was used. Standards and samples were diluted 10ϫ in 100 mM NaOH before analysis, and standards of the two isomers (0.58 mM) were used to identify corresponding peaks in the chromatograms.

Bioinformatic analyses
A BLAST was performed with the native Bal6GBP sequence (GenBank TM accession number ACS45862.1) as the query against all proteins of the Bifidobacterium genus. The 795 resulting protein sequences (length 350 -550 amino acids) with maximum scores Ն100 were aligned by MAFFT with default settings (53). From these sequences, 40 and 123 protein sequences clustered together with Bal6GP and GL-BP (the GNB/LNB binder from B. longum sp. longum), respectively, in a neighbor-joining consensus phylogenetic analysis (S3) performed in Geneious version 11.1.5 (54) including bootstrap resampling with 1000 replicates. Amino acid sequence logos of the residues involved in ligand binding were generated by TEXshade (55) for the orthologue groups.