Cell Surface Xyloglucan Recognition and Hydrolysis by the Human Gut Commensal Bacteroides uniformis

ABSTRACT

Xyloglucan (XyG) is a ubiquitous plant cell wall hemicellulose that is targeted by a range of syntenic, microheterogeneous xyloglucan utilization loci (XyGUL) in Bacteroidetes species of the human gut microbiota (HGM), including Bacteroides ovatus and B. uniformis. Comprehensive biochemical and biophysical analyses have identified key differences in the protein complements of each locus that confer differential access to structurally diverse XyG side chain variants. A second, nonsyntenic XyGUL was previously identified in B. uniformis, although its function in XyG utilization compared to its syntenic counterpart was unclear. Here, complementary enzymatic product profiles and bacterial growth curves showcase the notable preference of BuXyGUL2 surface glycan-binding proteins (SGBPs) to bind full-length XyG, as well as a range of oligosaccharides produced by the glycoside hydrolase family 5 (GH5_4) endo-xyloglucanase from this locus. We use isothermal titration calorimetry (ITC) to characterize this binding capacity and pinpoint the specific contributions of each protein to nutrient capture. The high-resolution structure of BuXyGUL2 SGBP-B reveals remarkable putative binding site conservation with the canonical XyG-binding BoXyGUL SGBP-B, supporting similar roles for these proteins in glycan capture. Together, these data underpin the central role of complementary XyGUL function in B. uniformis and broaden our systems-based and mechanistic understanding of XyG utilization in the HGM.

IMPORTANCE The omnipresence of xyloglucans in the human diet has led to the evolution of heterogeneous gene clusters in several Bacteroidetes species in the HGM, each specially tuned to respond to the structural variations of these complex plant cell wall polysaccharides. Our research illuminates the complementary roles of syntenic and nonsyntenic XyGUL in B. uniformis in conferring growth on a variety of XyG-derived substrates, providing evidence of glycan-binding protein microadaptation within a single species. These data serve as a comprehensive overview of the binding capacities of the SGBPs from a nonsyntenic B. uniformis XyGUL and will inform future studies on the roles of complementary loci in glycan targeting by key HGM species.

INTRODUCTION

The human distal colon is home to a complex consortium of microbes, collectively referred to as the human gut microbiota (HGM). This community is dominated by species from phyla Bacteroidetes, Firmicutes, and Actinobacteria and is fueled by complex carbohydrates, primarily present in the gut in the form of dietary and host cell surface glycans (1, 2). The saccharification of these glycans into monosaccharides via anaerobic fermentation by the HGM has a range of benefits for the host, including the production of metabolites such as short-chain fatty acids (SCFAs) that are absorbed by the gut endothelium (3). As the utilization of glycans in the distal gut supports a range of species and is indispensable for the health of the host, the deconstruction of these glycans by the microbiota is a process requiring in-depth understanding and characterization.

Bacteroidetes species are especially adept at dietary glycan metabolism, owing to the notable enrichment of carbohydrate-active enzymes (CAZymes) in their genomes (4, 5). A distinguishing feature of Bacteroidetes CAZyme arsenals is the colocalization of the corresponding genes into coregulated clusters called polysaccharide utilization loci (PULs) (reviewed in references 6–8). The number and diversity of CAZymes within a given PUL are typically commensurate with the complexity of the target glycan, resulting in a broad collection of PULs highly tailored for the deconstruction of chemically diverse substrates (9, 10). Species encoding large numbers of CAZymes (upwards of 300 in some cases) (4) are thus referred to as “generalists,” as their CAZyme profiles grant them access to a wide range of substrates (11–13). In contrast, species with narrow, often highly specific CAZyme profiles are glycan “specialists,” i.e., degraders of a small subset of highly specific and/or rare glycans, e.g., mucin O-glycans and dietary algal polysaccharides (13–16).

While the CAZyme composition of individual PULs varies based on the target polysaccharide, these multiprotein systems also share a set of protein components which work in concert to bind and degrade the target polysaccharide. Each PUL contains a collection of cell surface and periplasmic CAZymes, an outer membrane TonB-dependent transporter (TBDT), one or more surface glycan-binding proteins (SGBPs), and an inner membrane carbohydrate sensor/regulator (typically a hybrid two-component system [HTCS]) (6). The TBDT and SGBP-A (SusC and SusD homologs, respectively) (17) tightly associate to form an active transport complex for glycan fragments produced by the cell surface endo-acting CAZyme(s) (18, 19). SGBPs-A (SusD homologs) are based on a tetratricopeptide repeat (TPR) fold that is strongly conserved among PULs (20–24). PULs also typically encode one or more additional, nonhomologous SGBPs that assist in substrate capture at the cell surface. These SGBPs, which lack sequence similarity across even closely related PULs, are designated SGBP-B and SGBP-C (or “SusE-positioned” and “SusF-positioned“ proteins, respectively) based on their genetic organization downstream of the tandem TBDT/SGBP-A pair (24–29).

One of the most common families of dietary plant cell wall polysaccharides is the xyloglucans (XyGs). XyGs are found in the cell walls of all terrestrial plants and comprise a β-(1,4)-glucan backbone regularly decorated with α-(1,6)-xylopyranosyl residues. To simplify the representation of XyG structures, unbranched Glcp residues are given the symbol G, and branched [α-(1,6)-Xylp]-β(1,4)-Glcp disaccharide units are given the symbol X (Fig. 1A) (30). Based on the general repeating structures, the two types of XyGs commonly found in plants are known as “XXXG type” and “XXGG type.” The branching xylosyl residues can be further decorated with galactopyranosyl, fucopyranosyl, or arabinofuranosyl residues, in a tissue- and species-dependent manner, to give L, F, and S units, respectively; other side chains are also known (30).

FIG 1.

Xyloglucan and xyloglucan utilization loci (XyGUL) in commensal HGM Bacteroides species. (A) XXXG-type xyloglucan. Monosaccharides are represented by Consortium for Functional Glycomics notation (https://www.ncbi.nlm.nih.gov/glycans/snfg.html). (B) B. ovatus and B. uniformis XyGUL architectures. Arrows indicate gene orientation, and locus tags denote PUL boundaries. Proteins/domains are labeled as follows: GH, glycoside hydrolase; SGBP, surface glycan-binding protein; TBDT, TonB-dependent transporter; HTCS, hybrid two-component system; and SASA, sialic acid-specific acetylesterase. Hypothetical proteins are shown in white, and genes with functions unrelated to polysaccharide utilization are shown in gray.

We previously employed a systems-based approach to characterize a XyG utilization locus (XyGUL) from the human gut symbiont Bacteroides ovatus ATCC 8483 (31) (BoXyGUL) (Fig. 1B). In addition to a TBDT/SGBP-A pair, SGBP-B, and HTCS, this locus encodes a complement of CAZymes commensurate with the saccharification of dicot arabinoxyloglucan, viz.: a vanguard cell surface endo-xyloglucanase (GH5_4) and a suite of periplasmic GHs comprising two α-l-arabinofuranosidases (GH43), two β-glucosidases (GH3), a β-galactosidase (GH2), and an α-xylosidase (GH31). Bioinformatics identified the presence of syntenic XyGULs, i.e., those with highly similar genetic content and arrangement, in other commensal HGM Bacteroidetes species. Importantly, these studies indicated that XyGULs are essentially ubiquitous in human gut metagenomes, commensurate with the regular intake of XyG in dietary fruits and vegetables (31).

One such syntenic XyGUL was identified in B. uniformis ATCC 8492 (here referred to as BuXyGUL1) (31). Yet this locus encodes two notable differences in protein composition, which confer additional specificity for this species on the following divergent XyG structures: a GH95 α-l-fucosidase allows for the removal of fucosyl residues from dicot fucogalactoxyloglucan, and a novel sequence-divergent “SusF-positioned” SGBP-C specific for xyloglucan oligosaccharides (XyGOs) (Fig. 1B) (32). Such variations suggest that the protein complement of syntenic XyGULs mediates differential access of their parent species to individual XyG subtypes, e.g., fucogalactoxyloglucan versus arabinogalactoxyloglucan.

Unique among known Bacteroides species, B. uniformis also encodes an expanded, nonsyntenic XyGUL (here referred to as BuXyGUL2), which is upregulated on tamarind galactoxyloglucan (tamXyG) (31). In addition to the CAZymes required to deconstruct XXXG-type dicot fucogalactoxyloglucan (GH5, GH3, GH2, GH31, and GH29), BuXyGUL2 also encodes GH97 and GH42 members (Fig. 1B), with predicted α-glycosidase/α-galactosidase and β-galactosidase/α-l-arabinopryanosidase activities, respectively. Interestingly, the locus also encodes a multimodular GH43 with a predicted acetylesterase catalytic domain, which may aid in accessing acetylated XyG in natural substrates (30).

Building upon our previous analysis of BuXyGUL1 (32), we present here the initial biochemical characterization of BuXyGUL2, with a particular focus on the xyloglucan hydrolysis product distribution of the vanguard GH5 endo-xyloglucanase and the binding of the cognate SGBP-A and SGBP-B to xyloglucan fragments. We also present the high-resolution structure of BuXyGUL2 SGBP-B and discuss the predicted molecular determinants of specific ligand recognition by this protein. Together, our results outline the contributions of BuXyGUL2 in accessing a range of XyG-based ligands. We also compare and contrast the binding modes present across syntenic and nonsyntenic XyGUL from the HGM, thereby expanding our understanding of XyG utilization in the human gut.

RESULTS

Product distribution of BuGH5B and differential growth on XyGOs from homologous endo-xyloglucanases.

We recently reported that the cell surface GH5_4 endo-xyloglucanases from the syntenic B. ovatus and B. uniformis XyGULs produce distinct distributions of XyGOs from tamXyG (32). Specifically, BoGH5 cleaves tamXyG at unbranched G residues to produce primarily a canonical mixture of XXXG, XLXG, XXLG, and XLLG as limit-digest products. In contrast, BuGH5A is able to cleave the XyG backbone regardless of side chain branching to produce a mixture of shorter XyGOs (32) (Fig. 2A; Fig. S1 in the supplemental material). Here, we observe that BuGH5B, from the nonsyntenic XyGUL2, generates a panel of products comprising a hybrid of those produced by BoGH5 and BuGH5A. These data suggest that although BuGH5B has some capacity to accommodate xylosyl-substituted residues (X) in the −1 subsite, resulting in shorter XyGO products, it is comparably proficient at accepting unbranched G residues in this position (for a further detailed discussion of endo-xyloglucanase backbone specificity, see references 33 and 34). We note that this promiscuity is comparatively rare in GH5_4 enzymes (34) yet is shared in BuGH5A and BuGH5B (Fig. 2A).

FIG 2.

Production and utilization of tamXyGOs by B. ovatus and B. uniformis. (A) HPAEC-PAD analysis of limit-digest hydrolysis products of tamarind xyloglucan (tamXyG) by the GH5_4 endo-xyloglucanases from B. ovatus XyGUL (BoGH5) and B. uniformis XyGUL1 (BuGH5A) and XyGUL2 (BuGH5B). Products identified by MALDI-TOF MS are labeled and denoted with stars. (B) Growth profiles of B. ovatus (blue) and B. uniformis (orange) on tamXyG and GH5_4 XyGOs as sole carbon sources. Curves represent the mean OD₆₀₀ from three technical replicates.

The generation of distinct product profiles by the B. ovatus and B. uniformis endo-xyloglucanases, including the unusual production of both canonical Glc₄-based and shorter XyGOs by BuGH5B, suggests that each species may preferentially utilize a distinct set of oligosaccharides for growth. However, as these sets overlap, it is plausible that products generated by one species may be accessible to the other. To investigate this, we cultured B. ovatus and B. uniformis on XyGO mixtures produced by each GH5 endo-xyloglucanase. Previous studies have shown that both species are capable of utilizing XXXG-type tamXyG as a sole carbon source and that each XyGUL is upregulated when the bacterium is grown on this substrate (31). Here, we noted intriguing differences in the ability of each species to access both their own and exogenously produced tamXyGOs (Fig. 2B). B. ovatus grew on XyGOs produced by all three GH5 members but curiously displayed reduced growth on the XyGOs produced by its own endo-xyloglucanase, BoGH5, thus suggesting a preference for either the full-length XyG polysaccharide or smaller oligosaccharides such as those produced by B. uniformis. Similarly, B. uniformis showed comparable growth on XyG polysaccharide and Glc₄-based XyGOs from BoGH5 and BuGH5B, but limited growth on the smaller BuGH5A XyGOs.

BuXyGUL2 SGBPs preferentially bind larger XyGOs.

Having established relationships between cell surface product production and subsequent growth profiles, we next sought to investigate the role of cell surface glycan capture by the BuXyGUL2 SGBPs in this process. BuXyGUL1 and BuXyGUL2 encode three and two predicted SGBPs, respectively (Fig. 1B). Recently, we utilized isothermal titration calorimetry (ITC) to confirm the binding of BuXyGUL1 SGBP-B and BuXyGUL1 SGBP-C to tamXyG and variably galactosylated XyGOs (32). However, the ligand-binding profiles of the SGBPs from BuXyGUL2, as well as BuXyGUL1 SGBP-A, were previously unstudied.

Whereas BuXyGUL2 SGBP-A was readily produced in Escherichia coli, the full-length BuXyGUL2 SGBP-B protein was recalcitrant to recombinant production. Therefore, we cloned, overexpressed, and purified a truncated version in which the predicted N-terminal domain A, a predicted non-XyG-binding Ig-like domain, was removed (Fig. 3A). ITC analyses confirmed the binding of BuXyGUL2 SGBP-A and SGBP-B to XyG (Table 1; Fig. S2 and S3), with absorption equilibrium constant (K_a) values consistent with those previously observed for XyG-binding SGBPs (24, 31, 32). Interestingly, BuXyGUL2 SGBP-A exclusively bound Glc₄-based XyGOs produced by BoGH5, with little to no binding to XyGOs produced by BuGH5A and BuGH5B, which contain exclusively or predominantly shorter oligosaccharides, respectively (Table 1, cf. Fig. 2A). The lack of significant binding (at the limit of detection of ITC, the K_a value was ∼10³ M⁻¹) to the broad range of products produced by BuGH5B reflects the reduced relative abundance of the preferred, larger Glc₄-based XyGOs in this mixture (Fig. 2A). BuXyGUL2 SGBP-B displayed the highest binding affinities for full-length tamXyG and BoGH5 and BuGH5B XyGOs (K_a, ∼10⁵ M⁻¹) compared to those observed for XyGOs produced by the actions of BuGH5A (K_a, ∼10⁴ M⁻¹).

FIG 3.

Structure of BuXyGUL2 SGBP-B at 1.45 Å resolution. (A) Domain architecture of BuXyGUL2 SGBP-B. Amino acid boundaries are indicated. (B) Overall structure of BuXyGUL2 SGBP-B with transparent surface and domains indicated. Residues believed to form the aromatic platform necessary for ligand binding are shown. (C) Comparison of the putative ligand-binding site from BuXyGUL2 SGBP-B (violet, residues in bold) with the XyGO₂-binding site of BoXyGUL SGBP-B (white, residues in brackets) (PDB ID 5E7G) (24). (D) Sequence alignment between the carbohydrate-binding domains of SGBPs-B from BoXyGUL and BuXyGUL2. Predicted ligand-binding residues from BuXyGUL2 SGBP-B are denoted with pink circles, with those structurally conserved in BoXyGUL SGBP-B shown in bold.

TABLE 1.

Thermodynamic parameters for the binding of xyloglucan substrates to BuXyGUL SGBPs^e

Ligand	No.	Liganda (M) of:			Ka (M−1) of:			ΔG (kcal·mol−1) of:			ΔHf (kcal·mol−1) of:			TΔS (kcal·mol−1) of:
		BuXyGUL1 SGBP-A	BuXyGUL2 SGBP-A	BuXyGUL2 SGBP-B	BuXyGUL1 SGBP-A	BuXyGUL2 SGBP-A	BuXyGUL2 SGBP-B	BuXyGUL1 SGBP-A	BuXyGUL2 SGBP-A	BuXyGUL2 SGBP-B	BuXyGUL1 SGBP-A	BuXyGUL2 SGBP-A	BuXyGUL2 SGBP-B	BuXyGUL1 SGBP-A	BuXyGUL2 SGBP-A	BuXyGUL2 SGBP-B
XyG	1	NBb	6.8 (±0.2) × 10−4	6.8 (±0.1) × 10−4	NB	2.0 (±0.3) × 105	1.6 (±0.1) × 105	NB	−7.22	−7.09	NB	−23.0 ± 0.47	−11.5 ± 0.14	NB	−15.8	−4.43
BoGH5 XyGOs	1	1.1 (±0.04) × 10−3	8.2 (±0.5) × 10−4	7.4 (±0.4) × 10−4	6.8 (±1.2) × 104	4.4 (±0.8) × 104	1.8 (±0.6) × 105	−6.61	−6.34	−7.18	−8.56 ± 0.41	−15.1 ± 0.45	−3.31 ± 0.15	−1.95	−8.74	3.88
BuGH5A XyGOs	1	Weakc	Weak	1.3 (±0.6) × 10−3	Weak	Weak	5.8 (±0.8) × 104	Weak	Weak	−6.50	Weak	Weak	−12.6 ± 0.28	Weak	Weak	−6.14
BuGH5B XyGOs	1	2.8 (±0.4) × 10−4	NB	2.3 (±0.7) × 10−3	4.0 (±3.6) × 104	NB	2.4 (±0.5) × 105	−3.39	NB	−7.33	−11.4 ± 1.41	NB	−10.1 ± 0.26	−8.01	NB	−2.76
XXXG	1	NB	NDd	NB	NB	ND	NB	NB	ND	NB	NB	ND	NB	NB	ND	NB
XLLG	1	Weak	NB	NB	Weak	NB	NB	Weak	NB	NB	Weak	NB	NB	Weak	NB	NB
(XXXG)2	1	Weak	1.94 (±0.62) × 10−3	9.6 (±4.6) × 10−4	Weak	2.9 (±1.9) × 104	6.2 (±4.1) × 103	Weak	−6.08	−5.18	Weak	−14.8 ± 1.23	−26.5 ± 1.84	Weak	−8.71	−21.4
(XLLG)2	1	1.1 (±0.1) × 10−3	ND	8.0 (±0.6) × 10−4	2.7 (±0.4) × 104	ND	7.6 (±2.0) × 104	−6.04	ND	−6.66	−6.54 ± 0.16	ND	−4.50 ± 0.19	−0.508	ND	2.16

^aCalculated during ITC model fitting when n = 1.

^bNB, no binding.

^cWeak binding represents K_a of >500 M⁻¹.

^dND, not determined.

^eErrors calculated during ITC model fitting.

^fΔH, enthalpy change; TΔS, entropy change.

Consistent with previous biophysical and structural analyses of hemicellulose- and β-glucan-specific SGBPs-B (24, 32, 35), we observed a weak association of BuXyGUL2 SGBP-A and SGBP-B to single Glc₄-based products (XXXG, XLLG). Robust binding was only observed for Glc₈-based products; each protein displayed the ability to bind purified (XXXG)₂, with BuXyGUL2 SGBP-B also displaying binding to the galactosylated (XLLG)₂ (Table 1; Fig. S2 and S3).

BuXyGUL2 SGBP-B shares significant structural homology with BoXyGUL SGBP-B.

To investigate the structural features responsible for XyG binding by BuXyGUL2 SGBP-B, we pursued structural characterization of this protein via X-ray crystallography. Diffraction-quality crystals of truncated BuXyGUL2 SGBP-B with the N-terminal domain removed enabled the determination of the high-resolution (1.45 Å) structure of this protein via molecular replacement, using BoXyGUL SGBP-B (PDB ID 5E7G) (24) as the search model (Table 2). BuXyGUL2 SGBP-B shares the extended multimodular architecture typical of SGBPs-B characterized to date (24, 26, 36) (see also PDB ID 3ORJ). Like BoXyGUL SGBP-B, BuXyGUL2 SGBP-B comprises two immunoglobulin-like (Ig-like) domains (domain B, residues 129 to 220, and domain C, residues 224 to 304) and a predicted C-terminal XyG-binding domain (domain D, residues 305 to 485) (Fig. 3A). The Ig-like domains B and C superimpose closely with one another (root mean square deviation [RMSD], 0.89 Å over 54 Cα pairs), as well as with the equivalent domains in BoXyGUL SGBP-B (RMSD, 0.50 Å over 77 Cα pairs and 0.90 Å over 62 Cα pairs, respectively) (see discussion of protein homology below). The predicted N-terminal domain A (residues 1 to 123), which was omitted to facilitate recombinant production, likely has significant structural homology with the equivalent Ig-like domain in BoXyGUL SGBP-B (PDB ID 5E7G) (24), due to 73% sequence identity.

TABLE 2.

Data collection and refinement statistics for the crystal structure of BuXyGUL2 SGBP-B^a

BuXyGUL2 SGBP-B data category	Value
Data collection
Beamline	APS 23-ID-D
Wavelength (Å)	1.033
Space group	C121
Cell dimensions
a, b, c (Å)	119.9, 66.7, 58.0
α, β, λ (°)	90, 115.2, 90
Resolution (Å)	34.85–1.45 (1.51–1.45)
Rmerge (%)	0.052 (1.52)
Rmeas (%)	0.057 (1.66)
CC1/2b	0.999 (0.43)
I/σI	15.9 (0.9)
Completeness (%)	99.2 (97.9)
Multiplicity	6.6 (5.9)
Refinement
Resolution (Å)	34.85–1.45
No. of reflections (work/free)	72,030/3,567
Rwork/Rfree (%)	0.178/0.202
No. of atoms
Protein	2,827
Ligand	5
Water	417
Avg B factor (Å2)
Protein	29.2
Ligands	38.6
Water	40.3
RMSDs
Bond length (Å)	0.007
Bond angle (°)	0.87
Ramachandran statistics
Favored (%)	96.61
Outliers (%)	0
PDB code	7K44

^aNumbers in brackets refer to values from the highest-resolution shell.

^bCC_1/2, correlation coefficient between intensities of crystallographic random half data sets.

The C-terminal domain is the site of ligand binding in essentially all characterized SGBPs-B (24–27, 36). Despite our best efforts, we were unable to obtain crystals of oligosaccharide-bound BuXyGUL2 SGBP-B. However, we observe high degrees of structural overlap with the equivalent domain in XyGO₂-bound BoXyGUL SGBP-B (PDB ID 5E7G [24]; RMSD, 0.81 Å over 150 Cα pairs), including the sequential and structural conservation of the concave surface of the top β-sheet residues and the three residues forming the aromatic platform necessary for coordination of the XyGO₂ glucosyl backbone by BoXyGUL SGBP-B (24) (W330, Y363, and W364) (Fig. 3B to D). This conservation strongly supports a similar role for the equivalent residues in BuXyGUL2 SGBP-B (W321, Y354, and W356). Notably, a fourth, nonconserved aromatic residue (W353) extends this platform, possibly aiding to coordinate tamXyG and the Glc₈-based (XXXG)₂ and (XLLG)₂, the preferential binding of which was quantified by ITC.

Several nonaromatic accessory residues surrounding the binding site have been implicated in stabilizing the xylosyl side chains in the XyGO₂ complex structure of BoXyGUL SGBP-B, including a trio of aromatic residues (Y396, F414, and Y466) (24). These are not sequentially or structurally conserved in BuXyGUL2 SGBP-B (Fig. 3B and C). Instead, these positions in BuXyGUL2 SGBP-B are occupied by nonaromatic residues (Q361, E408, and Q460, respectively), which might coordinate branching sugar residues via hydrogen-bonding interactions. Additional nonaromatic residues (E456, R401) may contribute in a similar fashion, given their proximity to the binding site, as might a single tyrosine residue (Y403) in this area.

SGBPs-B are distinguished by similar three-dimensional tandem domain arrangements despite generally very poor sequence similarity (24–26, 35, 36). A DALI search returned, in addition to BuXyGUL2 SGBP-B, multimodular SGBPs-B with distinct binding specificities (Table S1). Despite the low sequence identity of the full-length proteins, a global structural alignment of BuXyGUL2 SGBP-B with each structural homolog notably reveals the topographical conservation of domains C and D (Fig. 4A to C). Indeed, crystallographic and mutagenesis studies have identified that these domains are essentially fused in the absence of flexible linker sequences (24, 26, 37) (see also PDB ID 3ORJ). This conserved domain architecture and rigidity juxtapose the N-terminal domains A and B, which have been observed to adopt a variety of confirmations relative to domains C and D in other SGBPs-B (Fig. 4A, see also references 24, 26, 35, and 36).

FIG 4.

Overlay of BuXyGUL2 SGBP-B with DALI-generated structural homologs. BuXyGUL2 SGBP-B is overlaid with xyloglucan-binding SGBP-B from the B. ovatus XyGUL (RMSD, 6.4 Å) (PDB ID 5E7G) (24), predicted xylan-binding SGBP-B from B. ovatus (RMSD, 7.1 Å) (PDB ID 3ORJ) (Joint Centre for Structural Genomics, unpublished data) (B), and mannan-binding SGBP-B from the Prevotella bryantii mannan utilization locus (RMSD, 5.2 Å) (PDB ID 6D2Y) (42) (C). In each case, BuXyGUL2 SGBP-B is colored by domain (Ig-like domain B, green; Ig-like domain C, blue; carbohydrate-binding domain, violet). Structural homologs are shown in white in two orientations with respect to the central vertical axis.

BuXyGUL SGBPs-A exhibit divergent glycan-binding and structural features.

Our previous analysis of BuXyGUL1 focused on the characterization of the highly divergent SGBP-B and SGBP-C (32). As such, the SGBP-A (SusD homolog) from BuXyGUL1 was previously uncharacterized. Hence, we characterized the binding of this protein to XyG-based ligands using ITC to enable comparison of the roles of the SGBPs-A from each BuXyGUL. In contrast to BuXyGUL2 SGBP-A, BuXyGUL1 SGBP-A curiously displayed no binding to full-length tamXyG (Table 1; Fig. S4). Instead, BuXyGUL1 SGBP-A showed equal affinities for XyGOs produced by BoGH5 and BuGH5B (based on Glc₄) and for (XLLG)₂ comprising a Glc₈ backbone. Robust binding was not observed for BuGH5A-generated XGOs, which are based on shorter backbones.

Whereas BuXyGUL2 SGBP-A shares high sequence identity with the BoXyGUL SGBP-A (24) (68%; Table S2), BuXyGUL1 SGBP-A has a notably divergent sequence (∼24% identity with BuXyGUL2 SGBP-A or BoXyGUL SGBP-A), including large sequence insertions and deletions (Table S2; Fig. S5). The sequence alignment also reveals minimal conservation of ligand-binding residues from BoXyGUL SGBP-A; only two aromatic residues implicated in substrate stacking interactions (W283 and W306) are conserved across all three proteins. An overlay of the XyGO₂-bound crystal structure of BoXyGUL SGBP-A (PDB ID 5E76) (24) with a Phyre2 homology model of BuXyGUL1 SGBP-A clearly shows that their global lack of primary sequence conservation corresponds to poor structural conservation across the canonical tetratricopeptide repeat-containing SusD-like fold (RMSD, 2.16 Å), particularly of the loop regions encompassing the binding site (Fig. S6A). The modeled binding site of BuXyGUL1 SGBP-A is notably deficient in the set of planar surface-exposed aromatic residues that typically form the extended binding platform and is further encumbered by additional loops not observed in the BoXyGUL SGBP-A XyGO₂ structure. These structural deviations, in particular the loop extensions, likely explain the inability of BuXyGUL1 SGBP-A to bind the full-length polysaccharide yet still accommodate shorter substrates. This can be contrasted with the structural homology model of BuXyGUL1 SGBP-A, which, due to its high sequence similarity to BoXyGUL SGBP-A (PDB ID 5E76) (24), reveals excellent conservation of the binding site residues and surrounding loops (Fig. S6B).

DISCUSSION

SGBPs play crucial roles in nutrient capture by Bacteroidetes of the HGM through sequestration of cognate glycans. Advances in anaerobic culturing and genomic sequencing of HGM species have produced a vast wealth of information on species-specific glycan utilization potential encoded by highly specialized PULs (reviewed in references 7 and 8). The notable synteny of PULs targeting common dietary glycans (32, 35, 38, 39) has revealed both conserved and unique proteins required for the utilization of these substrates, including diverse SGBPs. To fully understand PUL function, defining the role of each of these proteins in the recognition of unique polysaccharide signatures through careful biochemical and structural dissection is required. Here, we extended this approach to characterize the SGBPs of a distinct nonsyntenic xyloglucan utilization locus in B. uniformis, BuXyGUL2, and examined their roles in glycan capture versus counterparts in the canonical BoXyGUL and BuXyGUL1.

Commensurate with its role as part of a complex with the TBDT (SusC homolog) at the cell surface (18, 19), the SGBP-A of BuXyGUL2 bound XyG and longer XyG fragments with affinities similar to BoSGBP-A, which represents the only characterized XyG-binding SusD homolog to date (24). Indeed, sequence analysis and structural homology modeling revealed a high degree of binding site conservation among these proteins (Fig. S5 and S6 in the supplemental material).

Surprisingly, however, BuXyGUL1 SGBP-A failed to bind full-length tamXyG, but did bind XyG oligosaccharides. In this context, it is important to note that reverse genetic and mutagenesis studies on PULs targeting various polysaccharides have revealed that glycan binding by the cognate SGBP-A is not always required for bacterial growth, especially when a functional SGBP-B is present (17, 21, 22, 24–27). Indeed, in a limited number of cases, native SGBPs-A that do not bind glycans are known (35, 40–42). Regardless, the presence of an SGBP-A is structurally indispensable as the “lid” in the TBDT/SGBP-A “pedal-bin” complex (18, 19). TBDT/SGBP-A complexes showcasing captured glycans have revealed that ligand-binding residues may also be contributed by the exterior-facing TBDT barrel loops (18, 19). Thus, the limited binding capacities of BuXyGUL1 SGBP-A, which likely results from significant binding-site differences with BuXyGUL2 SGBP-A and BoSGBP-A (Fig. S5 and S6), is complemented by BuXyGUL1 SGBP-B, SGBP-C (32), and perhaps the cognate TBDT itself.

Like BuXyGUL2 SGBP-A, BuXyGUL2 SGBP-B binds the XyG polysaccharide and XyGOs with good affinity (K_a, ∼10⁴ to 10⁵ M⁻¹) (Table 1). The high-resolution crystal structure of BuXyGUL2 SGBP-B rationalizes the ligand specificity of this protein. The conservation of several glycosyl-coordinating aromatic residues with BoXyGUL SGBP-B (Fig. 3B), as well as the presence of an additional network of nonaromatic residues, suggests that these proteins share considerable similarities in XyGO binding. In particular, the long linker regions between these domains and the resulting “beads-on-a-string” configuration suggest a high degree of flexibility. This flexibility likely enables both CAZymes and SGBPs-B to adapt to complex substrate topologies, both at the polymer (polysaccharide) and particle (digesta) scales, thus facilitating coordinated capture and cleavage of complex glycans (28, 43).

Indeed, cell surface GHs from Bacteroidetes likewise often contain N-terminal spacer domains (PFAM PF13004), which, likewise, do not have catalytic or binding function (26, 31, 35, 44). Similar spacer functions have been attributed to noncatalytic modules from CAZymes across a variety of other systems, including the β-sandwich domains from large, multimodular pneumococcal and clostridial virulence factors (45, 46). In each case, the extended conformational heterogeneity of these enzymes may allow access to a larger pool of extracellular ligands, which would be otherwise inaccessible without the presence of the flexible stalk.

Together, our data suggest that BuXyGUL2 is poised to capture and import large XyGOs and that BuXyGUL2 SGBP-B plays an important role in this process. The collaborative glycan utilization by both BuXyGUL1 and BuXyGUL2 may provide B. uniformis with a distinct growth advantage in the competitive environment of the HGM, but the precise interplay of these two loci is presently unclear. XyG elicits major shifts in microbiota composition in vitro, with B. uniformis displaying particularly strong enrichment in the presence of fucogalactoxyloglucans compared to other species (47). Both B. uniformis XyGULs are also upregulated in the presence of the minimal galactoxyloglucan from tamarind seeds, but to different extents (31). The CAZyme complement of BuXyGUL1 is consistent with an ability to fully saccharify both fuco- and arabinogalactoxyloglucan, especially due to the presence of GH95 (α-fucosidase) and GH43 (α-arabinofuranosidase) members, respectively (31, 32). In BuXyGUL2, these functions may be fulfilled, respectively, by corresponding GH29 and GH43 members. At the same time, the richer, currently uncharacterized, cohort of GHs in BuXyGUL2 suggests that B. uniformis may access other complex or novel XyGs (48). As the development of molecular tools for the genetic manipulation of HGM species expands, reverse-genetic experiments involving the selective deletion of components from each BuXyGUL will continue to inform the discernment of each locus for particular XyGs. To enable this, however, the ability to purify individual XyG variants to homogeneity will be essential.

In conclusion, our study also adds to the growing catalogue of species whose genomes contain homologous but complementary PUL pairs, which each target specific substructures within structurally diverse polysaccharide families, e.g., xylans and mannans (37, 49). With the continued advancement of genome sequencing and the production of high-resolution metagenomic data sets from the microbiota of those consuming a variety of diets, we anticipate that the identification and systems-based analysis of these dual PUL systems will become more prevalent. As new CAZyme families continue to be uncovered in these PULs, their characterization may also signal the presence of yet undiscovered XyG structures from diverse plant sources (48). The outcomes of these studies will be essential for our complete understanding of the complex glycan-microbe and microbe-microbe interactions comprising the HGM.

MATERIALS AND METHODS

Carbohydrates.

Tamarind seed XyG (Megazyme) was used for bacterial growth profiling. Oligosaccharides were produced from tamarind kernel powder XyG (Innovasynth) as previously described (31). For each digestion, 20 g XyG was dissolved in 500 ml water and stirred at 95°C for 5 h. The mixture was cooled prior to the addition of 2 mg of the corresponding recombinant GH5 endo-xyloglucanase (BoGH5, BuGH5A, or BuGH5B), followed by incubation at 37°C to completion (assessed by HPAEC-PAD as described below). Insoluble material was removed by centrifugation, and solubilized XyGOs were filtered using 0.45-μm Durapore membranes (MilliporeSigma). The solution was concentrated under reduced pressure by rotary evaporation, followed by lyophilization. (XXXG)₂ and (XLLG)₂ were produced as previously described (50).

Analytical carbohydrate chromatography.

Xylogluco-oligosaccharides (XyGOs) produced by the digestion of tamarind XyG by B. ovatus and B. uniformis GH5 endo-xyloglucanases were analyzed by high-performance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD) using an ICS-5000 system (Dionex). Oligosaccharides were separated on a CarboPac PA200 analytical column (3 mm by 250 mm; Dionex) equipped with a CarboPac PA200 guard column (3 mm by 50 mm; Dionex) as previously described (32). Matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS) was used to identify individual oligosaccharides, using a standard mixture of XXXG, XXLG/XLXG, and XLLG as previously described (32, 51).

Bacterial growth profiling.

All growth and cleavage experiments were conducted at 37°C in an anaerobic chamber (Coy Lab Products, Grass Lake, MI, USA) in an atmosphere consisting of 80% N₂, 10% CO₂, and 10% H₂. The type strains B. ovatus ATCC 8483 and B. uniformis ATCC 8492 were grown in chopped meat broth (CMB) overnight. Cells were pelleted and washed in 2× minimal media (MM) (prepared as per reference 52) prior to resuspension in phosphate-buffered saline (PBS), pH 7.4. Growth experiments were performed in Falcon 300-μl flat-bottomed 96-well microtiter plates. Wells were loaded with 100 μl each of inoculum in PBS (or 2× MM for negative controls) and 10 mg/ml carbohydrate. Plates were sealed using Breathe-Easy gas-permeable membranes (Sigma-Aldrich), and absorbance at 600 nm was measured every 10 min for 72 h. Data were processed using Gen5 software (BioTek).

Cloning.

The open reading frames encoding BuXyGUL2 GH5, BuXyGUL1 SGBP-A, BuXyGUL2 SGBP-A, and domains B to D from BuXyGUL2 SGBP-B were amplified via PCR from B. uniformis ATCC 8492 genomic DNA (purified as per reference 32) using Q5 high-fidelity polymerase (NEB) (nucleotides 27,596 to 28,969 from BACUNI_03803, 46,661 to 48,340 from BACUNI_00316, 24,285 to 25,907 from BACUNI_03800, and 25,948 to 27,405 from BACUNI_03801, respectively). Primers were designed to remove the native signal peptides from the protein sequences (53) (Table 3). The resulting PCR products were cloned into either pET28a or pMCSG53 for ligation-independent cloning (54), providing an N-terminal His₆ tag to the protein product. Successful cloning was verified by colony PCR and sequencing (Genewiz).

TABLE 3.

Primer sequences used for recombinant protein production

Gene (protein)	Primer	Primer sequence (5′→3′)a
BACUNI_03803 (BuGH5B)	5B_forward (NdeI)	GGAGCTCATATGGACGATAAAAAGGAGCTGAAAATCATC
	5B_reverse (XhoI)	GGAGCTCTCGAGTTAATTCCGTGGATATTTAACCGG
BACUNI_00316 (BuXyGUL1 SGBP-A)	1A_forward	TACTTCCAATCCAATGCATGTACCGACAGTTTTCTCGATG
	1A_reverse	TTATCCACTTCCAATGTTATTACCATCCGGGATTCTGCAC
BACUNI_03800 (BuXyGUL2 SGBP-A)	2A_forward	TACTTCCAATCCAATGCATGCAGCGACTCATTTTTGGAAC
	2A_reverse	TTATCCACTTCCAATGTTATTATTCACTGAATTCGTATGCTACC
BACUNI_03801 (BuXyGUL2 SGBP-B)	2B_forward	TACTTCCAATCCAATGCACATGTACTCTCGCCCGCAC
	2B_reverse	TTATCCACTTCCAATGTTATTATTGTATCTTGACAACTCGGAAATTATC

^aRestriction endonuclease sites for ligation into pET28 are italicized, and universal sequences used for insertion into pMCSG53 via ligation-independent cloning are in bold.

Recombinant expression and protein purification.

The vectors encoding BACUNI_03803 (B. uniformis XyGUL2 GH5B), BACUNI_00316 (B. uniformis XyGUL1 SGBP-A), BACUNI_03800 (B. uniformis XyGUL2 SGBP-A), and BACUNI_03801 (B. uniformis XyGUL2 SGBP-B) were transformed into chemically competent BL21(DE3) or Rosetta (DE3) Escherichia coli expression cell lines and then cultured to mid-exponential phase (optical density at 600 nm [OD₆₀₀], 0.6 to 0.8) in lysogeny broth (LB) supplemented with kanamycin (50 μg/ml), chloramphenicol (35 μg/ml), and/or ampicillin (50 μg/ml) at 37°C (200 rpm). Overexpression was induced by the addition of 0.5 mM isopropyl β-d-thiogalactopyranoside (IPTG), and cultures were further grown at 37°C for 4 h (BuXyGUL2 SGBP-B) or 16°C for 18 h (BuGH5B, BuXyGUL1 SGBP-A, and BuXyGUL2 SGBP-A). Cells were harvested by centrifugation, and the His₆-tagged protein was purified by immobilized metal ion affinity chromatography using cobalt-based Talon resin (TaKaRa Bio), with bound proteins eluted with 100 mM imidazole. Prior to crystallization experiments, BuXyGUL2 SGBP-B was further purified by size exclusion chromatography using Superdex 75 resin (GE Life Sciences) in 20 mM HEPES, pH 7.4, and 100 mM NaCl. Purity was assessed by SDS-PAGE. The catalytic domains of BoXyGUL GH5 and BuXyGUL1 GH5 (BuGH5A) were cloned, recombinantly produced, and purified as per references 31 and 32).

Isothermal titration calorimetry.

ITC of glycan binding by SGBPs was performed in 20 mM Na-HEPES buffer (pH 7.4) at 25°C using a MicroCal PEAQ-ITC (Malvern Panalytical). All proteins and carbohydrates were prepared in this buffer by buffer exchange or direct dissolution, respectively. Proteins (30 to 50 μM) were loaded into the sample cell, and the syringe was loaded with 2.5 mg/ml tamarind XyG (Megazyme), 1 mM oligosaccharides of known molecular mass, or 10 mg/ml XyGO mixtures. Previously determined molar proportions were used to calculate the average molar masses of tamarind XyGO mixtures (55). An initial injection of 0.2 μl was followed by 18 injections of 2 μl spaced 150 s apart, with an injection duration of 4 s. Data were analyzed using the MicroCal PEAQ-ITC analysis software (Malvern Panalytical). Data were fit to a standard one-site binding model (n = 1), with ligand concentration as a variable. In this way, experimental ligand concentrations were calculated via ITC (Table 1).

Crystallography and structure solution.

All crystallization studies were performed at 298 K using the sitting drop vapor diffusion method. BuXyGUL2 SGBP-B was crystallized at 31 mg/ml in 0.2 M magnesium chloride, 0.1 M Tris, pH 8.5, and 25% (wt/vol) polyethylene glycol 3350 (PEG 3350). Crystals were cryoprotected using crystallization solution supplemented with 20% ethylene glycol prior to data collection.

Diffraction data were collected at 100 K at the Advanced Photon Source (Argonne National Laboratory) beamline 23-ID-D (Table 2). Data were processed using autoPROC (56) (uses XDS) (57), Pointless (58), Aimless (59), and CCP4 (60). Phasing by molecular replacement was performed using Phaser (61), with B. ovatus XyGUL SGBP-B (PDB ID 5E7G) (24) as a search model. Initial models were built using Phenix.autobuild (62), and subsequent model building and refinement were performed using Coot (63) and Phenix.refine (64). Model validation was performed using Rampage (65), SFCHECK (66), and PROCHECK (67).

Data availability.

Coordinates for BuXyGUL2 SGBP-B have been deposited in the Protein Data Bank (PDB) under the accession code 7K44. All biochemical data have been reported in the manuscript.