Research article

Structural and enzymatic characterization of a glycoside hydrolase family 31 α-xylosidase from Cellvibrio japonicus involved in xyloglucan saccharification

Johan Larsbrink, Atsushi Izumi, Farid M. Ibatullin, Azadeh Nakhai, Harry J. Gilbert, Gideon J. Davies, Harry Brumer


The desire for improved methods of biomass conversion into fuels and feedstocks has re-awakened interest in the enzymology of plant cell wall degradation. The complex polysaccharide xyloglucan is abundant in plant matter, where it may account for up to 20% of the total primary cell wall carbohydrates. Despite this, few studies have focused on xyloglucan saccharification, which requires a consortium of enzymes including endo-xyloglucanases, α-xylosidases, β-galactosidases and α-L-fucosidases, among others. In the present paper, we show the characterization of Xyl31A, a key α-xylosidase in xyloglucan utilization by the model Gram-negative soil saprophyte Cellvibrio japonicus. CjXyl31A exhibits high regiospecificity for the hydrolysis of XGOs (xylogluco-oligosaccharides), with a particular preference for longer substrates. Crystallographic structures of both the apo enzyme and the trapped covalent 5-fluoro-β-xylosyl-enzyme intermediate, together with docking studies with the XXXG heptasaccharide, revealed, for the first time in GH31 (glycoside hydrolase family 31), the importance of a PA14 domain insert in the recognition of longer oligosaccharides by extension of the active-site pocket. The observation that CjXyl31A was localized to the outer membrane provided support for a biological model of xyloglucan utilization by C. japonicus, in which XGOs generated by the action of a secreted endo-xyloglucanase are ultimately degraded in close proximity to the cell surface. Moreover, the present study diversifies the toolbox of glycosidases for the specific modification and saccharification of cell wall polymers for biotechnological applications.

  • enzymology
  • hemicellulose
  • plant cell wall
  • saccharification
  • xyloglucan


The enzymatic degradation of cellulose and matrix glycans of the plant cell wall into simple sugars provides energy for organisms from all kingdoms of life and, as such, represents a key aspect of the global carbon cycle [1]. This process is complex, with an implicit requirement for an enormous diversity of monosaccharide- and linkage-specific glycoside hydrolases and carbohydrate lyases [2]. With the rising desire to convert the vast amounts of plant polysaccharides available into liquid fuels and value-added products, significant attention continues to be focused on improving understanding of the mechanisms of microbial plant cell wall degradation [3].

Cellvibrio japonicus (previously Pseudomonas fluorescens subsp. cellulosa) is a saprophytic Gram-negative bacterium, which was first isolated from Japanese soil [4]. The bacterium has been shown to utilize the major plant polysaccharides cellulose, mannan and xylan and, as such, is a good model for the study of bacterial plant polysaccharide degradation [5]. Previously, the genome of C. japonicus was sequenced, and not surprisingly, a vast array of carbohydrate-degrading enzymes were identified [6]. In contrast with anaerobic species, such as the Clostridia, the enzymes of C. japonicus that target polysaccharides integral to the plant cell walls do not form large membrane-attached cellulosome complexes [7], but are instead either secreted into the medium or attached to the cell membrane as lipoproteins. Indeed, approximately one-third of the encoded carbohydrate-degrading enzymes have been predicted to be lipoproteins [6], which may suggest that C. japonicus has developed a strategy involving the ultimate degradation of plant cell wall-derived oligosaccharides in close proximity to the cell, as has been shown, for example, in the case of xylan utilization [8,9].

As a soil saprophyte, an abundant potential nutrient source for C. japonicus, apart from cellulose, is the group of non-crystalline cross-linking glycans commonly known as hemicelluloses. Among the most prevalent of these are the XGs (xyloglucans), which are present in the primary cell wall of essentially all land plants. For instance, XGs may account for up to 20% of the total carbohydrate content (dry weight) of the primary cell wall in dicots and non-graminaceous monocots, whereas more limited amounts are found in grass species [10]. XGs are a family of heterogeneous polysaccharides, which are typified by a β(1→4)-D-glucan backbone substituted with α(1→6)-D-xylopyranosyl moieties at regular intervals. The basic repeating motifs may be further decorated with galactopyranose, arabinofuranose, fucopyranose and O-acetyl moieties in a tissue- and species-dependent manner (Figure 1) [11,12]. Thus several enzymatic activities are required to degrade xyloglucan into its constituent monosaccharides. Although the endogenous enzymes responsible for the mobilization of XGs in plants have been previously studied in some detail [1320], the elucidation of the enzymes responsible for XG degradation by micro-organisms such as C. japonicus is currently in its infancy [21].

Figure 1 Structures of XG and CjXyl31A substrates

(A) General structure of primary cell wall xyloglucans showing variable branching. (B) Aryl glycosides for −1 subsite specificity determination. (C) XGO used for extended specificity determination. XG fragments are named according to [28].

α(1→6)-Linked xylopyranose represents the second most abundant sugar in XGs, the cleavage of which is essential for the ultimate liberation of glucose from the polysaccharide backbone. α-Xylosidases have thus far only been identified in GH31 (glycoside hydrolase family 31) (; [2]), which also contains diverse enzymes active on α-glucosidic bonds, such as α-glucosidases, sucrose-isomaltases, α-glucan lyases and isomaltosyltransglucosidases. Whole-genome analysis of C. japonicus has suggested that this organism produces only a single GH31 α-xylosidase, CjXyl31A [6], which is thus likely to be a key player in XG degradation by this organism. We describe in the present paper the biochemical characterization of CjXyl31A together with the three-dimensional structure of the enzyme. Detailed kinetic data indicated a strict exo-specificity for non-reducing-end xylose moieties and increased specificity for extended non-galactosylated XGOs (xylogluco-oligosaccharides). In harness, structural analysis revealed that this specificity was affected, in part, by a unique PA14 domain insert in the N-terminal domain of CjXyl31A that endowed the active-site pocket with an extended substrate-binding platform. Taken together, these data provide exquisite molecular insight into the enzymatic features that enable C. japonicus to utilize the naturally abundant plant cell wall polysaccharide XG.


Curve fitting was performed using Origin 8 (OriginLab). Ultrapure water was used in all experiments and refers to water purified on a Milli-Q system (Millipore) with a resistivity of ρ>18.2 MΩ·cm.

General analytical methods


For MALDI–TOF (matrix-assisted laser desorption ionization–time-of-flight)-MS, a Biflex IV controlled by FlexControl 2.4 (Bruker Daltonics) was used, with 2,5-dihydroxy benzoic acid used as the matrix. ESI (electrospray ionization)-MS on oligosaccharides was performed as described previously [22]. LC (liquid chromatography)-ESI-MS was used for protein molar mass determination as described previously [23].

NMR spectroscopy

NMR spectroscopic data were recorded on a Bruker Avance 400 instrument operating at 400.1 MHz for 1H, using the residual solvent signal as a reference. Chemical shifts (δ) are reported in p.p.m. and coupling constants (J) are given in Hertz (Hz).

Native-gel electrophoresis

Oligomerization of Xyl31A was investigated by gel electrophoresis under native conditions, using a NuPAGE Novex 7% Tris-acetate gel (Invitrogen) with 25 mM Tris and 192 mM glycine (pH 8.3) as a running buffer. Samples were loaded on to the gel in 50 mM Tris/HCl, 10% (v/v) glycerol and 0.05% Bromophenol Blue (pH 8.6), and the molecular masses were estimated by comparison with a high-molecular-mass marker for Native PAGE (HMW Native Marker Kit, GE Healthcare). Proteins were visualized using PageBlue™ Protein Staining Solution (Fermentas).

HPAEC-PAD (high-performance anion-exchange chromatography with pulsed amperometric detection)

Oligo- and mono-saccharides were analysed on a Dionex ICS-3000 HPLC system operated by Chromelion software version 6.80 (Dionex) using a Dionex Carbopac PA200 column. Solvent A was water, solvent B was 1 M sodium hydroxide and solvent C was 1 M sodium acetate. Depending on the analytes, the following gradients were employed. Gradient A: 0–5 min, 10% solvent B and 2% solvent C; 5–12 min, 10% B and a linear gradient from 2–30% C; 12–12.1 min, 50% B and 50% C; 12.1—13 min, an exponential gradient of NaOH and NaOAc (sodium acetate) back to initial conditions; and 13–17 min, initial conditions. Gradient B: 0–5 min, 10% B and 0% C; 5–10 min, 10% B and a linear gradient from 0–30% C; 10–10.1 min, 50% B and 50% C; 10.1–12.5 min, an exponential gradient of NaOH and NaOAc back to initial conditions; and 12.5–15 min, initial conditions. Gradient C: 0–4 min, 10% B and 6% C; 4–17 min, 10% B and a linear gradient from 0–25% C; 17–17.1 min, 50% B and 50% C; 17.1–18 min, an exponential gradient of NaOH and NaOAc back to initial conditions; and 18–22 min, initial conditions.

Enzyme substrates

pNP (p-nitrophenyl) α-D-xylopyranoside, α-D-glucopyranoside, α-D-galactopyranoside and β-L-arabinopyranoside were purchased from Sigma. Tamarind seed XG and the derived oligosaccharides XXXG (Glc4Xyl3; see [24] for oligosaccharide nomenclature), isoprimeverose [X, Xylp-α(1→6)-Glcp] and borohydride-reduced xylosyl-cellobiose [XGol, Xylp-α(1→6)-Glcp-β(1→4)-Glcp] were purchased from Megazyme (catalogue numbers O-X3G4, O-IPRM and O-XCBIR respectively). XLLG (Glc4Xyl3Gal2) and XXXGXXXG (Glc8Xyl6) were prepared from tamarind XG as described previously [22,25].

The oligosaccharide XX (Glc2Xyl2) was prepared from tamarind XG. XG (200 g) was vigorously stirred in 2 l of water at 65 °C to obtain a homogeneous mixture. The mixture was cooled to room temperature (22 °C) and ammonium acetate buffer (1 M, pH 4.5) was added until the desired pH was reached. Digestion was performed by the addition of 1 g of the xgl1 endoglucanase (Dyadic NL) and the reaction proceeded at room temperature for 4 days with stirring. The reaction was stopped by heating at 80 °C for 30 min. The resulting suspension was centrifuged and the supernatant was collected, concentrated and precipitated using ethanol [4 equivalents (v/v)]. The precipitate was collected and dissolved in water, followed by lyophilization. A selected amount (4 g) of the XGO mixture was acetylated and purified by column chromatography as described previously [25], to obtain acetylated XX (0.15 g). The structure was verified by NMR (Supplementary Figure S1 at and ESI-MS (m/z calculated for C46H62O31Na: 1133.3173; found: 1133.1969). The deacetylation was performed as described previously [25], and the molar mass was verified by ESI-MS (m/z calculated for C22H38O19Na: 629.1905; found: 629.2047). Following lyophilization, the resulting powder was dissolved in water for biochemical assays.

Bioinformatic analysis

The phylogenetic tree of GH31 was derived from the protein sequences of biochemically characterized entries in CAZy (carbohydrate-active enzymes) together with the sequence of Xyl31A. The predicted catalytic domains of the genes were aligned using MUSCLE [26], and the alignment was used to construct the phylogenetic tree using the maximum likelihood method with the program PhyML [27]. The reliability of the tree was analysed by bootstrap analysis of 100 resamplings of the dataset. The tree was displayed using MEGA4 [28].

Cloning of Xyl31A

The open reading frame encoding Xyl31A (GenBank® accession number ACE86259.1) was amplified by PCR from the genomic DNA of C. japonicus Ueda107 using Phusion polymerase (Finnzymes) and the following primers (Thermo Fischer Scientific): 5′-CACCATGTTATCGGCACATCAGTG-3′ and 5′-TTAACGCGGGCGTTTAATGC-3′ with the forward primer incorporating the CACC overhang needed for TOPO® cloning. The PCR fragments were cloned into the pENTR/SD/D-TOPO entry vector (Invitrogen) according to the manufacturer's instructions. Chemically competent TOP10 cells were transformed with the cloning reaction and grown overnight. Plasmid DNA was extracted using the MiniPrep kit (Qiagen), and sequenced (Eurofins MWG Operon) to identify positive clones. The constructs were recombined, using the LR Clonase mix II (Invitrogen), into the pET-DEST42 destination vector (Invitrogen) (providing a C-terminal His6 tag) according to the manufacturer's instructions.

Recombinant gene expression and protein purification

Plasmids harbouring the Xyl31A gene were transformed into Escherichia coli BL21(DE3) by electroporation, and the resulting transformants were grown in Terrific Broth containing 50 μg/ml ampicillin at 37 °C to a D600 of 0.5–0.8. Gene expression was induced by the addition of IPTG (0.5 mM isopropyl β-D-galactopyranoside), and protein production was carried out at 25 °C overnight. The cells were collected by centrifugation at 3800 g for 10 min at 4 °C, and were resuspended in buffer A [20 mM sodium phosphate (pH 7.4), 0.5 M NaCl and 20 mM imidazole]. The cells were passed twice through a French Press and centrifuged at 27000 g for 45 min. The supernatant was loaded on to a Ni-Sepharose fast-flow column (GE Healthcare) using an ÄKTA FPLC (GE Healthcare) and washed thoroughly with buffer A. Xyl31A was eluted with a linear gradient of 0–100% of buffer B [20 mM sodium phosphate (pH 7.4), 0.5 M NaCl and 1.0 M imidazole]. The eluted protein was washed and concentrated with 50 mM sodium phosphate (pH 7.5) with 10 kDa cut-off Amicon Ultra centrifugal filters (Millipore). The protein was further purified by SEC (size-exclusion chromatography) using a HiPrep 26/60 Sephacryl S-300 column (GE Healthcare) using 50 mM sodium phosphate (pH 7.5).

Antibody generation and purification

Antibodies were raised against recombinant Xyl31A in rabbits (AgriSera AB). The immunization was as follows. Immunization 1: 200 μg of antigen (10 mg/ml) and FCA (Freund's complete adjuvant); immunization 2, 3 and 4 (1, 2 and 3 months later respectively): 100 μg of antigen and FCI (Freund's incomplete adjuvant). The final bleeding was carried out 11 days after immunization 4.

Xyl31A was coupled to a 1 ml HiTrap NHS-activated HP Column (GE Healthcare) according to the manufacturer's instructions. The column was washed using an ÄKTA FPLC (GE Healthcare) with several column volumes of PBST [PBS (pH 6.9) containing 0.05% Tween 20]. To 4.5 ml of antiserum containing antibodies against Xyl31A was added 0.5 ml of 10×PBS and the solution was applied to the Xyl31A-coupled HP column. The column was washed with several column volumes of PBST. Antibodies bound to the column were eluted with 200 mM glycine and 1 mM EDTA (pH 2.5) in 0.5 ml fractions, to wells containing 50 μl of 1 M Tris and 75 μl of 10×PBS (pH 6.9) to neutralize the eluent. The purified antibodies were passed through a PD10 column packed with His-tagged albumin-binding protein coupled to an NHS matrix to remove His-tag-specific antibodies.

X-ray crystallography

Initial crystallization conditions were screened using several kits and needle-shaped crystals were obtained from 20% (w/v) PEG [poly(ethylene glycol)] 4000, 0.1 M Mes (pH 6.5) and 5 mM NiCl2. Ultimately, two crystallization conditions, using the hanging-drop vapour diffusion method, were optimized. Crystal form 1 grew from protein in 10 mM Mes (pH 6.5) and 0.15 M NaCl with 40% pentaerythritol propoxylate (5/4 PO/OH), 0.1 M Bis/Tris (pH 7.0) and 5 mM NiSO4 at 20 °C. Crystals of form 1 were of poor quality, as reflected in the processing statistics. Crystal form 2 grew from protein in 10 mM Mes (pH 6.5), 0.15 M NaCl and 5 mM NiSO4 with 25% PEG mME 550 (PEG monomethyl ether 550) and 0.1 M Bis/Tris (pH 7.0) at 20 °C. The 5-fluoro-β-D-xylopyranosyl-enzyme intermediate was obtained by using crystal form 2 after soaking with approximately 2 mM 5FαXylF (5-fluoro-α-D-xylopyranosyl fluoride) [29], a kind gift from Professor Stephen Withers (Department of Chemistry, University of British Columbia, Canada) for approximately 1 h.

Crystals of form 1 were cryoprotected using the reservoir solution alone, whereas form 2 required elevation of the PEG mME 550 concentration to 40%. Results for form 1 and 2 were collected on beamline ID14-2 and beamline ID29 of the ESRF (European Synchrotron Radiation Facility) respectively. The results were integrated using the HKL suite and MOSFLM/CCP4 [30]. The results processing and structure-refinement statistics are shown in Table 1.

View this table:
Table 1 Activity of CjXyl31A on various substrates at pH 6

Substrate structures are shown in Figure 1(B) ΔΔG‡ was calculated using the formula: ΔΔG‡=−RTln[(kcat/Km XGO)/(kcat/Km XXXG)].

The structure was solved by the molecular-replacement method using the MalA structure (Sulfolobus solfataricus α-glucosidase, PDB code 2G3M), which was selected by BALBES [31], and the initial model was constructed using RESOLVE [32] and ARP/wARP [33] by using the 2.3 Å (1 Å=0.1 nm) data from crystal form 1. Manual corrections were made using COOT [34]. All other computing used the CCP4 suite [30], unless explicitly stated. The co-ordinate file and monomer library description for 5/4 PO/OH was prepared using PRODRG [35] and PHENIX [36], whereas those for the trapped 5-fluoro-β-D-xylopyranosyl-enzyme intermediate were generated using the CCP4 suite. B-factors for all structures were refined with PHENIX by using TLS (translation libration screw-motion) refinement with 5 TLS groups prepared by TLS motion determination website ( The final structures were validated by MolProbity [37] ( Structure similarity searches were performed with DALI [38].

For docking analysis of the XXXG oligosaccharide, a library file containing oligosaccharide bond and torsion angle information was prepared using the PRODRG server ( from PDB code 2CN3 [22]. Subsequently, the non-reducing-end xylosyl unit of XXXG was superimposed on to the 5-fluoro-β-D-xylopyranosyl group in the glycosyl-enzyme structure using the Superpose Ligands extension in COOT, followed by manual optimization of glycosidic torsional angles to relieve obvious steric clashes.

Stopped assay for hydrolysis of pNP glycosides

Hydrolysis of pNP glycosides (2 mM) by Xyl31A was performed in 100 μl reaction volumes at 25 °C in 50 mM citrate buffer (pH 6) using typically 2–10 μM enzyme. The reactions were stopped by addition of 100 μl of 0.2 M NaCO3. Released pNP was measured in a Cary50 UV–visible spectrophotometer (Varian) at 410 nm, using 1-cm-path length cuvettes (pNP anion molar absorption coefficient used for calculations was 18500 M−1·cm−1 [39]).

pH-rate profile

Measurements of the pH-dependence of Xyl31A were carried out using pNP-α-Xyl (2 mM) and the stopped assay. The following buffers were used (50 mM): sodium citrate (pH 3–6.5), sodium phosphate (pH 6–8), glycylglycine (pH 7.5–9) and glycine (pH 9–9.5).

Temperature stability

To monitor the temperature stability of Xyl31A, the enzyme was incubated at different temperatures with aliquots taken out at different time points (5, 10, 20, 30, 60, 120 and 240 min). The aliquoted enzyme was used in the stopped assay to determine the remaining activity.

Specificity for XGO substrates

XGOs (XXXG, XLLG, XX and X) were incubated with Xyl31A (2.5 nM for longer XGOs, 2 μM for isoprimeverose) in 50 μl reaction volumes containing 50 mM citrate buffer (pH 6) at 25 °C and the reactions were stopped by the addition of 1 μl of 5 M NaOH. Product formation was assayed using HPAEC-PAD, with gradient A used for XXXG, XLLG and XX and Gradient B for X. For the XGol substrate, the BCA (bicinchoninic acid) reducing sugar assay was used [40, 41]. Reaction volumes of 250 μl were made in 50 mM citrate buffer (pH 6) with 20 nM Xyl31A, and stopped by the addition of an equal amount of BCA reagent solution. Colour was developed by incubation at 80 °C and measured at 560 nm. Standard curves were made using xylose (Sigma).

XGO regiospecificity determination

To fully degrade XXXG to cellobiose, Xyl31A was used together with a β-glucosidase (Clostridium thermocellum Glucosidase 1A; NZYTech). The reactions were carried out in 200 mM citrate buffer (pH 6), at 37 °C for Xyl31A and 55 °C for Glc1A, until the reactions had gone to completion. First, Xyl31A was added to hydrolyse the xylose moiety from the non-reducing-end, followed by denaturation by boiling. The same procedure was carried out with the β-glucosidase, hydrolysing the now exposed glucose moiety. The procedure was cycled until cellobiose was obtained. Product formation was analysed by ESI-MS. To further analyse the regiochemistry of Xyl31A, XXXG2 was used as a substrate with subsequent analysis by MALDI–TOF MS.

Cellular localization of Xyl31A

C. japonicus Ueda107 was grown in 250 ml of M9 minimal medium [42] containing tamarind seed XGOs (0.4%) as the sole carbon source at 30 °C for 24 h until a D600 of 1, corresponding to a late-exponential phase [43]. The cells were collected by centrifugation at 4400 g for 10 min at 4 °C and the supernatant was decanted and concentrated approximately 25-fold using 10 kDa cut-off Amicon Ultra centrifugal filters (Millipore).

Periplasmatic proteins were collected by osmotic shock. The cells were washed with 10 ml of 50 mM Tris/HCl (pH 7.7) and collected by centrifugation at 4400 g for 10 min at 4 °C. The pellet was resuspended in 50 ml of 30 mM Tris/HCl, 20% (w/v) sucrose and 1 mM EDTA (pH 8.0), and the cells were incubated at room temperature for 10 min. The cells were collected by centrifugation at 4400 g for 15 min at 4 °C. Ice-cold (50 ml) 5 mM MgSO4 was added and the cells were incubated on ice for 10 min. The cells were centrifuged at 14000 g for 10 min at 4 °C and the supernatant collected.

The cell pellet were resuspended in 50 mM sodium phosphate buffer (pH 7.4) and passed twice through a French Press. The lysate was centrifuged at 5000 g for 10 min at 4 °C to remove debris. The supernatant was centrifuged at 100000 g for 1 h at 4 °C and the supernatant containing the soluble proteins was collected. The pellet, containing the membrane fraction was resuspended in 100 mM sodium carbonate buffer (pH 9) to remove trapped soluble proteins and/or weakly membrane-associated proteins and centrifuged again at 100000 g for 1 h at 4 °C. The pellet was resuspended in 1 ml of 50 mM sodium phosphate buffer (pH 7.4). Western blot analysis was used to visualize the content of Xyl31A in the secreted, periplasmatic, soluble intracellular and membrane fractions. The purified anti-Xyl31A antibodies were used as the primary antibody, and anti-rabbit IgG (Sigma) coupled to alkaline phosphatase was used as the secondary antibody, followed by visualization using SIGMAFAST™ BCIP® (5-bromo-4-chloroindol-3-yl phosphate)/NBT (Nitro Blue Tetrazolium) tablets (Sigma) dissolved in water.

Depolymerization of XG by secreted enzymes of C. japonicus

The secreted protein fraction from the cellular localization studies, above, was screened for xyloglucanase activity using XG as the substrate. The fraction was washed with 50 mM sodium phosphate buffer (pH 7.4) using 10 kDa cut-off Amicon Ultra centrifugal filters (Millipore) to remove traces of XGOs from the growth medium.

Reactions consisted of 10 μl of the protein fraction (0.04 mg/ml, determined by the Bradford method with a BSA standard curve) added to 145 μl of 1.72 g/l XG in 69 mM ammonium acetate buffer (pH 5.5) and incubation at 37 °C for 10, 20, 30, 120, 275 min and 48 h. The reactions were stopped by boiling for 10 min, with subsequent lyophilization. The samples were redissolved in 350 μl of DMSO and heated at 80 °C for 10 min. HPSEC (high-performance size-exclusion chromatography) analysis was used to monitor the depolymerization of XG, as described previously [44] though using two PLgel 10 μm of mixed B columns and a guard column (Agilent) in tandem kept at 70 °C. An 11 point pullulan standard curve (molecular mass 180–1660000 Da) was used for molecular mass estimation. HPAEC-PAD was used for product analysis of the XGOs formed after the full depolymerization of XG, using gradient C.


Primary structure analysis of CjXyl31A and GH31 phylogeny

Previous analysis of the genome of C. japonicus predicted a vast array of polysaccharide-degrading enzymes, of which CjXyl31A was annotated as the sole putative α-xylosidase belonging to GH31 in the CAZy classification [6]. Re-analysis of the primary structure indicated that CjXyl31A had the highest protein sequence identity with the S. solfataricus GH31 α-xylosidase XylS (GenBank® accession number CAB99206.1), with values of 27, 39 and 35% for the N-terminal catalytic and C-terminal domains respectively. Also, an approximately 150 amino-acid long insert in the CjXyl31A N-terminal domain, which is not present in biochemically characterized GH31 enzymes, was identified. A BLAST (basic local alignment search tool) search reveals that this sequence (amino acids 238–384) belongs to the PA14 superfamily (named after the protective antigen of anthrax toxin [45]), which is known to form inserts in a wide variety of enzymes and has been proposed to be involved in substrate binding [46]. Sequence analysis has previously indicated that the PA14 domain is also found in members of glycoside hydrolase families GH2, GH3, GH10 and GH20, in addition to GH31 [46]. However, little is presently known about the tertiary structure and function of PA14 domains in these enzymes. As a lone example, a recent report on the structure–function analysis of the Kluyveromyces marxianus β-glucosidase has highlighted that the PA14 domain in that GH3 enzyme occludes the active-site pocket, thus conferring a preference for disaccharide substrates [47]. In CjXyl31A, the PA14 domain is likewise positioned close to the active site, but in contrast appears to be involved in the extended recognition of longer xylogluco-oligosaccharides, as discussed below.

To delineate the diverse α-glucoside- and α-xyloside-processing activities in GH31, a phylogenetic tree was constructed using only the (β/α)8 catalytic domain of all members of GH31 whose enzyme specificity has previously been determined (results from the CAZy database [2]), together with CjXyl31A (Figure 2). In this phylogeny, a number of activities toward α-glucosidic bonds stood out clearly as distinct clades, including the α-1,4-glucan lyases, the isomaltosyl- and 6-α-glucosyltransglycosylases, and the sucrase isomaltases. However, notably, the eight α-xylosidases characterized to date did not clearly segregate, thus suggesting that α-xylosidase activity has arisen on separate occasions from glucospecific enzymes. In particular, the known plant α-xylosidases [GenBank® accession number BAA99366.1 (Oryza sativa), GenBank® accession number AAD37363.1 (Arabidopsis thaliana) and GenBank® accession number CAA10382.2 (Tropaeolum majus)] cluster with plant α-glucosidases, whereas the only other characterized eukaryotic α-xylosidase from Aspergillus nidulans (GenBank® accession number EAA62085.1) is most similar to the bacterial α-xylosidases (Figure 2). The promiscuity of the A. thaliana α-xylosidase/α-glucosidase [19] further highlights that small differences in protein sequence can modulate substrate discrimination based on the presence or absence of the C6 hydroxymethyl group on the pyranose ring [48] (see Figure 1B). Indeed, the E. coli GH31 α-xylosidase YicI has been converted into an α-glucosidase by site-directed mutagenesis [49]. Given the comparatively limited data available to date, it is clear that phylogenetic groupings do not unambiguously predict catalytic function in GH31, thus motivating further structural and functional characterization of CjXyl31A.

Figure 2 Phylogenetic tree of all biochemically characterized enzymes of GH31, including CjXyl31A

The genes are named by their GenBank® accession numbers. Activities are designated as follows: grey circle, α-glucosidase; star, α-xylosidase; diamond, transglycosylase; triangle, α-glucan lyase; black circle, sucrase-isomaltase. Accession numbers from eukaryotes are in black, those from bacteria and archaea are in grey.

It is worth noting here that an earlier suggestion [29] that bacterial GH31 members invariantly contain the consensus sequence KTDFGE surrounding the conserved catalytic nucleophile (underlined) appears to not be borne out in the analysis of larger datasets. CjXyl31A, in particular, contains the sequence WLDAVE, which is also similar to that of the well-characterized archeal α-xylosidase, SsXylS (WLDASE, [50], see Supplementary Figure S3 at Of the 15 biochemically characterized bacterial genes used to construct the phylogenetic tree showing in Figure 2, only E. coli YicI (GenBank® accession number AAC76680.1) and Lactobacillus pentosus XylQ (GenBank® accession number AAC62251.1) contain the exact KTDFGE sequence, whereas four bacterial isomaltosyltransglucosidases (GenBank® accession numbers BAB88401.1, BAD34979.1, BAB88403.1 and BAC54957.1) possess the similar KTDFGg consensus sequence; the remainder of the bacterial enzymes in the present analysis contain the consensus sequence WnDmnE.

Heterologous expression and protein purification

CjXyl31A bearing a C-terminal His6 tag was produced in E. coli BL21(DE3) cells; the positioning of the purification tag was chosen based on a previous report which indicated that the N-terminal portion of GH31 enzymes may influence substrate binding [29]. The recombinant protein was purified by IMAC (immobilized metal-ion-affinity chromatography) followed by SEC to remove low molar mass contaminants. SEC data, confirmed by native PAGE on individual fractions, showed the protein being in mainly monomeric form (approximately 75%), with minor fractions in dimeric, tetrameric and hexameric forms. The electrophoretically pure (SDS/PAGE gel, results not shown) pooled fractions from the SEC was used for all subsequent analyses. LC-ESI-MS indicated that the CjXyl31A so produced had a molar mass of 113644±3 Da (calculated 113643.2), corresponding to the His-tagged protein less a 23 residue signal peptide. The LipoP 1.0 Server [51] indicated that this was probably a lipoprotein signal peptide, which would anchor the enzyme to the plasma membrane through the sulfhydryl group of Cys23 after cleavage by signal peptidase II [52]. CjXyl31A was, however, produced in good yields (200 mg/l) as a soluble protein in E. coli, with Ser24 as the N-terminus after signal-peptide processing.

Substrate specificity of the −1 (glycone) subsite

A range of pNP α-glycopyranosides were used to screen the substrate specificity of CjXyl31A (Figure 1B). In keeping with the original prediction of activity [6] and sequence similarity to SsXylS (Figure 2), CjXyl31A exhibited the highest activity toward pNPαXyl (Table 1). With the corresponding glucoside, activity was several orders of magnitude lower, and no activity was observed on pNPαGal or pNPβ-L-Arap substrates. Compared with the E. coli α-xylosidase YicI [29], CjXyl31A had a 6-fold lower kcat/Km value on pNPαGlc (YicI 6.010−3mM−1·s−1 and CjXyl31A 1.2×10−3 mM−1·s−1) and a 10-fold higher kcat/Km value on pNPαXyl (YicI 0.18 mM−1·s−1 and CjXyl31A 1.5 mM−1·s−1), which indicates a significantly higher specificity for xylose (i.e. discrimination against the 6-hydroxymethyl substituent) in the −1 subsite [53] of CjXyl31A.

pH-dependence and thermostability studies

The pH-dependence of catalysis by CjXyl31A was assayed using pNPαXyl. A classic bell-shaped pH profile was obtained (Figure 3), with an optimum at pH 6.5 and apparent pKa values of 4.3 and 8.8, which probably correspond with the ionization of the catalytic nucleophile and acid/base residues respectively. The thermal stability of CjXyl31A was assayed by incubating the enzyme at different temperatures and assaying remaining activity after defined time intervals (Figure 3). At 35 °C and 45 °C, CjXyl31A exhibits good stability under the solution conditions used, with approximately 15% activity loss after 4 h at both temperatures. At higher temperatures (≥55 °C), the protein loses activity rapidly, as might be expected for an enzyme from a mesophile.

Figure 3 pH profile and thermostability of CjXyl31A

(A) The buffers (50 mM) used in the pH-rate profile were citrate (pH 3–6.5; squares), phosphate (pH 6.5–8.5; circles), glycylglycine (pH 7.5–9; triangles) and glycine (pH 9–9.5; inverted triangles). (B) For the thermostability study, the enzyme was incubated at different temperatures (35 °C, circles; 45 °C, squares; 55 °C, triangles and 65 °C, diamonds), with aliquots of enzyme taken at different time intervals. The remaining activity was used to determine the level of thermal denaturation. pNP-α-Xyl (2 mM) was used as the substrate for both the pH-rate profile and thermostability measurements.

Specificity toward XGOs

In Nature, α-linked xylose is most prevalent in the plant cell wall polysaccharide XG [10]. To investigate the potential of CjXyl31A to participate in XG degradation, assays were performed on a small library of XGOs obtained by enzymatic hydrolysis (Figure 1C). CjXyl31A showed a clear preference for longer substrates, with the heptasaccharide XXXG yielding the highest kcat/Km value, which was similar to the tetrasaccharide XX (Table 1). On the trisaccharide alditol XGol, the kcat/Km value was approximately one-third of that for XXXG. The disaccharide X [Xylp-α(1→6)-Glcp; Figure 1A] was hydrolysed at a significantly lower rate than the other XGOs and had a 100-fold higher Km value compared with XXXG. Galactosylation also reduced the specificity of CjXyl31A for oligosaccharide substrates; the kcat/Km value for the nonasaccharide XLLG was 3-fold lower than for XXXG, due to a significantly higher Km value of the former.

Product analysis by ESI-MS indicated that for all of the XGO substrates, only one xylose residue was released (full results not shown; see e.g. XXXG, Table 2). Similarly, MALDI–TOF-MS analysis of the reaction products following incubation of CjXyl31A with XXXGXXXG (Figure 1C) indicated that a single xylosyl unit was cleanly released from the tetradecasaccharide (Figure 4), presumably from the extreme non-reducing-end. The strict non-reducing-end exo-activity of CjXyl31A was confirmed by sequential incubation of the heptsaccharide XXXG with CjXyl31A and Clostridium thermocellum exo-β-glucosidase Glc1A, incorporating a heat-inactivation step between each enzyme treatment. Thus XXXG was degraded by CjXyl31A to yield only GXXG, which could then by hydrolysed to XXG by CtGlc1A, which could again be cleaved by CjXyl31A to GXG, and so on, to yield cellobiose after the last xylosidase treatment (Table 2). Taken together, these results demonstrate that CjXyl31A is unable to access internal xylose residues in xyloglucan polymers, even when they are followed by an unbranched glucose residue in the backbone, as in XXXGXXXG. This behaviour is similar to that of previously characterized α-xylosidases from plants and archea, which require a β-glucosidase for the full degradation of, e.g. the XXXG motif [15,16,50]. Analysis of the three-dimensional structure of CjXyl31A, reported below in apo and substrate-complexed forms, provides a clear rationalization for this exquisite regiospecificity.

View this table:
Table 2 Product analysis following sequential rounds of XXXG hydrolysis by CjXyl31A and CtGlc1A
Figure 4 MALDI–TOF-MS analysis of the XXXGXXXG tetradecasaccharide and its hydrolysis product GXXGXXXG obtained by incubation with CjXyl31A

a.u., arbitrary unit.

In the context of biotechnological applications, including plant biomass saccharification, CjXyl31A displays catalytic constants for the hydrolysis of the artificial substrate pNPαXyl (Table 1) similar to that of other GH31 α-xylosidases. Thus the kcat and Km values of mesophilic enzymes for pNPαXyl typically fall in the ranges of 0.1–0.6 s−1 and 0.1–1 mM respectively (see [29] and references therein), whereas kcat values for the thermophilic α-xylosidase from S. sulfataricus may approach 3–10 s−1 in the temperature range 65–85 °C [50] (assuming enzyme saturation at [S]=37 mM used in that study). However, whereas CjXyl31A is more selective for pNPαXyl over pNPαGlc than the well-characterized E. coli YicI [29] (see above), the C. japonicus enzyme is a much poorer degrader of the minimal disaccharide isoprimeverose (kcat of 2.9 s−1 and Km of 55 mM; Table 1) than either its E. coli (kcat of 51 s−1 and Km of 0.55 mM; [54]) or S. sulfataricus (specific activity approximately 20 s−1 at [S]=100 mM and 65 °C, [50]) homologues. Direct comparison with plant α-xylosidases is hampered by a paucity of detailed kinetic analyses [19,20]. Thus the kinetic data (Table 1) suggest that CjXyl31A is best suited to technical uses that involve the breakdown of larger XGOs.

Overall three-dimensional structure of CjXyl31A

To illuminate the structural bases for catalysis and the potential role of the PA14 domain, the three-dimensional structure of the xylosidase was solved. Crystals of CjXyl31A were obtained in the hexagonal space group P6322 with one molecule in the asymmetric unit. The structure was solved by molecular replacement using the structure of the S. solfataricus α-glucosidase MalA (PDB code 2G3M). The refined structure of the apo enzyme, the 5-fluoro-β-xylosyl-enzyme intermediate and the pentaerythritol propoxylate complex were solved at 2.6, 2.5 and 2.3 Å resolution respectively (Table 3). As described above, MS analysis indicated that the predicted signal sequence of 24 amino acids was processed during protein expression. In addition, no electron density was visible for the first 20 N-terminal residues of the mature protein in any of the structures. In total, 31 C-terminal residues of the apo structure and 32 C-terminal residues of the complex structures were also unresolved and are presumably disordered.

View this table:
Table 3 X-ray diffraction and structure refinement statistics for CjXyl31A

The CjXyl31A monomer structure has a complex modular architecture which is difficult to define unambiguously, but which may be considered as five domains consisting of two N-terminal ‘lobes’, a central catalytic domain and two C-terminal modules (Figure 5). The central catalytic domain forms an (β/α)8-barrel consisting of residues 412–791 and with two significant ‘inserts’ beyond the standard barrel termed Ins1 (insert region-1) at residues 512–556 and Ins2 (insert region-2) at residues 586–613. The (β/α)8-barrel contains the catalytic apparatus, as will be described subsequently in light of a covalent intermediate complex and by virtue of similarities to related GH31 and glycoside hydrolase clan GH-D enzymes of known three-dimensional structure. The N-terminal region clearly forms two essentially discrete lobes. The first, primarily N-terminal, lobe is defined by residues 45–237, which form a β-sandwich domain. The topology of this domain is complicated by the fact that the three-dimensional structure of the whole domain also involves residues 392–411 when they return from the excursion that forms the second N-terminal domain. This latter domain, also itself a β-sandwich lobe, is an excursion between two of the strands of the N-terminal domain. This inserted domain, defined approximately by residues 238–391, is classified as a PA14 domain (described previously, and in more detail below). The C-terminal domains follow a more regular topology defined by two discrete sequential β-sandwich domains approximately delineated by residues 793–870 and 871–988 respectively. Interface analysis using PISA [55] suggests that none of the potential interfaces are significant in terms of a stable higher oligomeric complex.

Figure 5 Structures of the Cellvibrio japonicus Xyl31A and E. coli YicI 5-fluoroxylosyl enzyme intermediates

Global structures of CjXyl31A (A) and YicI (B), in wall-eyed stereo, highlighting the distinct PA14 insert of CjXyl31A (purple) in the N-terminal domain (green). The homologous domains of both enzymes are coloured as follows: N-terminal domains, green; catalytic domains, cyan; loop inserts in the catalytic domains, blue; C-terminal domains, yellow and red. The catalytic amino acid residues are shown in ball-and-stick (dark grey) with the nucleophile bound to the 5F-β-Xyl ligand (yellow). (C) Electron density map of 2Fo−Fc around 5F-β-Xyl, including the nucleophilic water, shown at the 1.7σ level. (D) Diagram of the interactions around the 5F-β-Xyl ligand. The arrow indicates the distance in Ångströms between the anomeric carbon of eq-5FαXF and water molecule bound to Asp659. The Figure was prepared using PyMOL and ChemDraw.

The most striking difference between CjXyl31A and other known GH31/clan GH-D tertiary structures is the unusual N-terminal PA14 domain, the function of which is not generally known. The PA14 domain derives its name from the N-terminal cleaved domain of anthrax protective antigen [45], with which the corresponding residues in CjXyl31A display a RMSD (root mean square deviation) of 2.3 Å over 105 matched Cα positions (DALI [38] Z score=8.6). The domain is also similar to a number of other sugar-binding domains, such as the mannose-binding flocculation Flo5 protein [56] (PDB code 2XJ2, Z score 9.1 with 119 Cα overlapping, RMSD 2.5 Å). Most intriguingly, this domain shares high structural similarity with a number of other domains such as observed in a family GH3 β-glucosidase [47] (PDB code 3ABZ, DALI Z score 13.1, with 132 Cα overlapping, RMSD 2.7Å), the family GH2 β-glucuronidase [57] (PDB code 3K46, Z score 9.1, 118 Cα overlapping, RMSD 3.5Å), family GH2 β-mannosidases (PDB code 2JE8, Z score 7.9, 113 Cα overlapping, RMSD 3.2Å), as well as other similar domains from both β-glucosidase and β-glucuronidases. The PA14 domain observed in CjXyl31A is, thus far, unique to GH31 enzymes of known structure; it is not, for example, found in E. coli YicI (Figure 5). Given the proximity of the PA14 domain to the active-site pocket of CjXyl31A, we speculate that this domain plays a role in the binding of oligosaccharides, contributing to the preference for longer XGOs, as will be described below.

Three-dimensional structure of a trapped covalent glucosylenzyme intermediate

Crystals of form 2 (see the Experimental section) were soaked in the mother liquor supplemented with approximately 2 mM 5FαXylF. Electron density for a covalently linked 5-fluoro-β-xylosyl residue was clearly seen in the −1 subsite with the anomeric carbon atom making a covalent bond to OD2 atom of Asp582 with a 1.4 Å distance (Figure 5). The ring conformation refined to a 1S3 (skew boat) conformation, consistent with that observed in structures of α-glycosidases from GH31 and clan GH-D described previously [29,58]. Asp582 and Asp659 are predicted to act as the catalytic nucleophile and acid/base residues, as previously defined for this sequence family [29]. Asp470, His740, Asp659 and Arg642 all make hydrogen bonds to hydroxy groups of the 5-fluoro-β-xylosyl ester, whereas Phe692 and Trp471 make van der Waals contacts with the ligand. A water molecule binding to Asp659 lies 3.1 Å from the C1 atom of the intermediate and would appear to be perfectly poised to act as the hydrolytic water in the reaction mechanism. Crystal form 1 also yielded a structure with the precipitant pentaerythritol propoxylate bound in the catalytic pocket. The environment of the pocket is almost the same as that in apo and intermediate structures except, that Trp347 and Trp471 are flipped away from the inside of the pocket by approximately 20° and 80° respectively. Pentaerythritol propoxylate settles between Trp347 and Trp542 and has a van der Waals contact to Trp471. Together, these results suggest plasticity of the binding site.

Relationship of the three-dimensional structure to observed kinetics and digestion patterns

Previous studies on α-xylosidases from T. majus, A. thaliana and S. solfataricus have shown the selective hydrolysis of one xylose moiety from the non-reducing end of XGOs, without observed activity on internal xylose moieties [16,18,50]. To date, the three-dimensional structure has been determined for five members of GH31, with YicI from E. coli being the only α-xylosidase [29]. These five enzymes all display a highly similar fold, with a central catalytic (α/β)8-barrel flanked by N- and C-terminal β-sheet domains of unknown functions, although the N-terminal domain has been suggested to be involved in substrate binding [29].

In the case of CjXyl31A, substrate specificity has two facets: (1) this strictly exo-acting enzyme is specific for a terminal unsubstituted xyloside and (2) this xyloside must lie at the extreme-reducing-end of a XGO; there is no activity on the internal xylosides of substrates such as XXXG and XXXGXXXG. The specificity for an unmodified terminal xylose moiety is provided, as with other GH31 enzymes, by the constellation of residues that interact directly with the −1 subsite xyloside. Trp580, Trp656 and Phe692 together provide a steric barrier at the end of the active centre pocket, in which Asp470, His740, Asp659 and Arg642 make hydrogen bonds to the −1 subsite xyloside (described above in light of the xylosyl-enzyme intermediate structure; see also Figure 5). There is thus no scope, simply in terms of sterics, to accommodate a xyloside with appended Gal or Gal-Fuc moieties, as found in oligosaccharides such as XLLG and XLFG (Figure 1). The pocket itself, however, is more than merely a blunt-ended −1 subsite. The POCKETFINDER server ( estimates that the catalytic crevice has minimum dimensions of ~11 Å depth and 8 Å width, although the binding surface could certainly be more extended as the mouth of the canyon expands into a larger estuary.

Our attempts to soak/co-crystallize with various XGOs did not yield a complexed structure, which, in the case of soaking, may partially reflect a two-fold axis of crystallographic symmetry that resulted in the active sites of two monomers facing each other. To circumvent this paucity of direct experimental data, a simple manual model of the Michaelis complex between CjXyl31A and XXXG was constructed based upon the three-dimensional structures of a known XGO/endo-xyloglucanase complex [22] and the glycosyl-enzyme intermediate described above. As shown in Figure 6, interactions with the XXXG substrate extend beyond the active-site pocket to the extended face provided by the PA14 domain. Another key feature of the model is that the O4 atom of the +1 glucoside moiety is also enrobed by a tight surface comprising, notably, Arg642 and the loop of residues 582–587, whose distal end begins one of the unusual insertions in the classical (β/α)8-barrel. Likewise, Trp542, from the second of the unusual insertions, lies under the potential +1/+2 boundary and may also play a role in discriminating against ‘endo’ attack. Together these residues provide the steric features that prevent binding of a xyloside from an ‘internal’ glucose moiety and demand that only the xylose on the extreme-non-reducing-end glucoside is accommodated. The data further suggest that if an internal xylose-cleaving enzyme is desired, homologues should be sought with truncations in these regions.

Figure 6 Molecular surface of CjXyl31A with docked heptasaccharide XXXG

Global view (A) and detailed view (B) of the active site in wall-eyed stereo. The colour scheme is identical with that used in Figure 5: N-terminal domain, green; PA14 domain, pink; catalytic domain, cyan; loop insert in the catalytic domain, blue; and C-terminal domains, yellow and red. The Figure was prepared using PyMOL.

The unique PA14 domain lies on the side of the substrate-binding cleft and could, potentially, both facilitate binding of longer oligosaccharides and perhaps also play a role in productive binding both of the backbone of the XG oligosaccharides (Trp347, for example, lies on the floor of the binding canyon such that it may interact with a +2 subsite glucoside), while also providing numerous hydrogen-bonding residues that may interact productively with appended xyloside moieties. Such proposals are clearly speculative, but it is fascinating to see that other similar PA14 domains in unrelated glycosidases have also been recruited to modulate the oligosaccharide specificity, such as in Bacteroides thetaiotaomicron Man2A where the domain serves to block off an otherwise ‘endo’ active centre groove to provide the chain-end specificity demanded by a β-mannosidase [59]. Similarly, the PA14 domain of the family GH3 β-glucosidase has recently been show to modify the exo/endo behaviour of that enzyme [47]. Clearly, our understanding of the potential roles of PA14 domains is only in its infancy, due to the limited number of structure–function studies presently available. However, given the wide distribution of this domain across GH families [46], we anticipate further focused studies on the importance of PA14 in carbohydrate recognition and catalysis.

The physiological role of CjXyl31A in the scavenging of XG by C. japonicus in the natural environment

The detailed protein structural and enzyme kinetic data presented here lay the initial groundwork for understanding the mechanism by which the soil saprophyte C. japonicus scavenges XG in its habitat. As highlighted in the Introduction section, complete saccharification of the plant polysaccharide probably involves the concerted action of one or more endo-acting xyloglucanases capable of generating shorter oligosaccharide fragments (e.g. EC [21], together with a battery of exo-acting glycosidases, including CjXyl31A, to release individual monosaccharides for cellular uptake.

Thus far no XG-specific endo-glucanases have been described from C. japonicus, nor is there any general information on the ability of the organism to degrade XG polymers. Consequently, C. japonicus was grown on a minimal medium containing a mixture of tamarind seed XGOs as the sole carbon source, in an attempt to induce the production of XG-active enzymes. Subsequent analysis of the extracellular medium obtained during the late-exponential growth phase indeed indicated the presence of endo-xyloglucanase activity. Specifically, HPSEC and HPAEC-PAD analysis (Supplementary Figure S3) indicated that this crude preparation was able to hydrolyse tamarind seed XG to the four component oligosaccharides, XXXG, XLXG, XXLG, and XLLG (see Figure 1 for structures) by specific cleavage at the unbranched glucosyl unit. Although the isolation and further characterization of the individual enzyme(s) responsible for this activity are beyond the scope of this paper, these results clearly demonstrate that C. japonicus has the capacity to generate its own XGOs for further utilization. It is also noteworthy that the extracellular fraction appears to be essentially devoid of β(1→2)-galactosidase and α(1→6)-xylosidase activities, as evidenced by the peak areas of the component oligosaccharides, which are similar to those obtained via digestion with purified endo-xyloglucanases [22,25,60].

The observation that CjXyl31A is encoded with a predicted lipoprotein signal peptide (see above) provided a further tantalizing clue about the potential mechanism by which C. japonicus systematically disassembles XG. To verify the membrane localization of CjXyl31A in the native organism, Western blot analysis was performed on total membrane, soluble intracellular, periplasmatic and secreted protein fractions (Figure 7). The analysis indicates that CjXyl31A was most abundant in the total membrane fraction, but was also found in significant quantity in the soluble intracellular fraction, and, to a lesser extent, in the periplasm. In keeping with the HPLC data (Supplementary Figure S3), essentially no CjXyl31A was found in the extracellular medium; trace amounts, detectable as a very faint band on the Western blot after prolonged staining, may be attributable to cell lysis late in the exponential growth phase. Localization of CjXyl31A on the outer membrane is consistent with previous studies on membrane-bound glycosidases of C. japonicus [8,9]. Furthermore, the absence of an aspartic acid residue at position +2, this residue is a serine in CjXyl31A, is also suggestive of outer membrane localization [51].

Figure 7 Western blot analysis of C. japonicus fractions using a polyclonal antibody preparation raised against recombinant CjXyl31A

The protein contents of the fractions were normalized prior to loading the lanes: 1, membrane; 2, soluble intracellular; 3, periplasmatic; 4, secreted; and X, recombinant CjXyl31A produced in E. coli. The SeeBlue® Plus2 ladder (Invitrogen) was used. Note that the phosphorylase marker at 97 kDa faded during the Western blot protocol.


Taken together, the results of the present study provide an emerging picture of XG utilization by C. japonicus, in which soluble, secreted endo-xyloglucanase activity liberates oligosaccharides that are further degraded to monosaccharides close to the cell surface (Figure 8). Notably, this model of step-wise hydrolysis of XGOs by non-reducing-end-specific exo-glycosidases also rationalizes the low activity that CjXyl31A exhibits toward the disaccharide isoprimeverose [Xylp-α(1→6)-Glc; X in Figure 1C], which would not generally be encountered; the shortest oligosaccharide hydrolysed by CjXyl31A would be xylosyl-cellobiose (Table 2), the alditol of which (XGol) is cleaved with essentially the same selectivity as the parent oligosaccharides XXXG and XLLG (Table 1). Although the identities of the α(1→2)-L-fucosidase, β(1→2)-galactosidase(s) and β(1→4)-glucosidase necessary for the complete saccharification of XGOs are presently unknown, as is their potential to synergize with each other and with one or more unidentified sugar transporters, this model suggests fertile ground for future study. The detailed enzyme structure–function analysis in the present study has given new insight into the mechanisms of XG degradation by C. japonicus, yet there is still much information missing to fully describe the pathways of polysaccharide utilization by this soil saprophyte.

Figure 8 Working model of XG utilization by C. japonicus

Secreted endo-xyloglucanase activity and outer-membrane-anchored CjXyl31A demonstrated in the present study are indicated. Hypothetical exo-glycosidases are denoted by (?). The order of interaction of these exo-glycosidases and monosaccharide membrane transporters is also speculative.


Johan Larsbrink performed all enzyme kinetic and biochemical analyses, critically analysed structural data, performed phylogenetic analyses and was responsible for the assembly of the complete manuscript, including major writing contributions. Atsushi Izumi performed all crystallography and related data analysis, and made major contributions to the writing of the manuscript. Farid Ibatullin synthesized all of the non-commercially-available oligosaccharide substrates used in this study, with the exception of the XX disaccharide substrate, which was synthesized by Azadeh Nakhai who also wrote the experimental section concerning this synthesis. Harry Gilbert analysed and provided essential sequence data from the C. japonicus genome, Harry Brumer and Gideon Davies devised the study, supervised data acquisition, analysed data and made major contributions to the writing of the manuscript. All authors contributed to the review and improvement of the final version of the manuscript prior to publication.


Work in Stockholm and Newcastle was funded through the European Union WoodWisdom-Net ERA-NET project ( FibreSurf: New biotechnological tools for wood fibre surface modification and analysis by The Swedish Research Council Formas and UK Forestry respectively. Supplemental funding from Formas via CarboMat - The KTH Advanced Carbohydrate Materials Consortium and the Wallenberg Wood Science Center is also gratefully acknowledged. Work in York was funded by the Biotechnology and Biological Sciences Research Council (BBSRC) [grant numbers BB/I014802/1 and BB/E001696/1]. G.J.D. is a Royal Society/Wolfson Research Merit Award recipient.


We especially thank Professor Pedro M. Coutinho (UMR6098, CNRS/Universités de Provence/Université de la Méditerranée, Marseille, France) for his kind help with access to information on biochemically characterized enzymes from the CAZy database. Dr Gustav Sundqvist and Dr Chunlin Xu are thanked for assistance with MS analyses (AlbaNova University Center, Stockholm, Sweden).

Abbreviations: BCA, bicinchoninic acid; CAZy, carbohydrate-active enzymes; ESI, electrospray ionization; ESRF, European Synchrotron Radiation Facility; 5FαXylF, 5-fluoro-α-D-xylopyranosyl fluoride; GH31, glycoside hydrolase family 31; HPAEC-PAD, high-performance anion-exchange chromatography with pulsed amperometric detection; HPSEC, high-performance size-exclusion chromatography; LC, liquid chromatography; MALDI–TOF, matrix-assisted laser desorption ionization–time-of-flight; NaOAc, sodium acetate; PBST, PBS (pH 6.9) containing 0.05% Tween 20; PEG, poly (ethylene glycol); PEG, mME 550, PEG monomethyl ether 550; pNP, p-nitrophenyl; RMSD, root mean square deviation; SEC, size-exclusion chromatography; TLS, translation libration screw-motion; XG, xyloglucan; XGO, xylogluco-oligosaccharide


View Abstract