X4 Modules Represent a New Family of Carbohydrate-binding Modules That Display Novel Properties*

The hydrolysis of the plant cell wall by microbial glycoside hydrolases and esterases is the primary mechanism by which stored organic carbon is utilized in the biosphere, and thus these enzymes are of considerable biological and industrial importance. Plant cell wall-degrading enzymes in general display a modular architecture comprising catalytic and non-catalytic modules. The X4 modules in glycoside hydrolases represent a large family of non-catalytic modules whose function is unknown. Here we show that the X4 modules from a Cellvibrio japonicus mannanase (Man5C) and arabinofuranosidase (Abf62A) bind to polysaccharides, and thus these proteins comprise a new family of carbohydrate-binding modules (CBMs), designated CBM35. The Man5C-CBM35 binds to galactomannan, insoluble amorphous mannan, glucomannan, and manno-oligosaccha-rides but does not interact with crystalline mannan, cellulose, cello-oligosaccharides, or other polysaccharides derived from the plant cell wall. Man5C-CBM35 also potentiates mannanase activity against insoluble amorphous 50 m dissolved in the same to minimize heats of Where appropriate, TCEP, calcium, or EDTA at final concentrations of 1, 5, and 10 m M , respectively, were added Abf62A-CBM35 and carbohy- experiment, a 1.4331-ml successive of ligand comprising polysaccharide (5–25 mg/ml) or oligosaccharide (5–15 m M ), at 200-s intervals. The apo form of Abf62A-CBM35 was titrated against 3 m M calcium chloride to determine whether cal- cium bound independently to the protein. Prior to the titration the cell was washed out with 5 m M EDTA, followed by an extensive Chelex- treated buffer wash to remove any traces of the metal ion from the machine. Integrated heat effects, after correction for heats of dilution, non-linear site-binding Fitted data yield the association constant ( K A ) and the enthalpy of binding ( (cid:8) H ). Other thermodynamic parameters were calculated using the standard thermodynamic equa- tion: (cid:4) RT ln K A (cid:5) (cid:8) G (cid:5) (cid:8) H (cid:4) T (cid:8) S . The c values (product of the association constant (cid:3) the concentration of the acceptor (cid:3) the number of binding sites on the acceptor) were 3–30. At least two independent titrations were performed for each ligand tested. The molar concentra- tion of Man5C-CBM35 binding sites present in galactomannan and glucomannan was determined by altering the concentration of polysac- charide used for regression of the isotherm until the fit yielded a value of 1 for n (number of binding sites on each molecule of CBM). The assumption that n (cid:5) 1 was based on the oligosaccharide titrations, which all displayed a stoichiometry of 1:1. For analysis of the xylan binding data with Abf62A-CBM35, this approach could not be used because the CBM did not interact with xylo-oligosaccharides. Instead, the binding data was fitted by treating the CBM in the sample cell as the ligand and the polysaccharide as the acceptor. This gives accurate values for K A and (cid:8) H , but not n (in this case the number of binding sites on each molecule of polysaccharide).

The plant cell wall comprises the most abundant source of renewable carbon on the planet. This extensive resource is made available to the biosphere through the action of microbial glycoside hydrolases, which are thus of considerable biological and industrial importance. Plant cell walls are composed of a complex network of polysaccharides that are highly inaccessible to enzyme attack (1). Glycoside hydrolases that degrade the plant cell wall are generally modular enzymes comprising catalytic and non-catalytic modules that are joined via flexible linker sequences. Many of these non-catalytic modules bind to specific oligo-and polysaccharides derived from the plant cell wall and are thus defined as carbohydrate-binding modules (CBMs, 1 Ref. 2). By localizing the appended catalytic module onto the surface of the (mainly) insoluble polysaccharide substrates, CBMs potentiate the activity of glycoside hydrolases against these composite structures (3,4). Thus, CBMs play a pivotal role in the capacity of glycoside hydrolases to degrade the plant cell wall.
Based on sequence similarities, CBMs have been grouped into families (afmb.cnrs-mrs.fr/CAZY/, Ref. 5). Currently there are 34 CBM families, 32 of which contain modules from prokaryotic enzymes whereas only families 1 and 29 contain fungal proteins (afmb.cnrs-mrs.fr/CAZY/, Ref. 5). Three-dimensional structures of representatives of over half of the CBM families demonstrate that these proteins generally adopt a ␤-jelly roll fold (2). Structural data has also shown that the topology of the ligand-binding site of CBMs varies. In Type A CBMs, which interact with the flat surfaces crystalline polysaccharides such as cellulose, the binding site comprises a hydrophobic planar surface that contains a linear strip of exposed aromatic amino acids (6 -9), while in Type B CBMs, which interact with individual polysaccharide chains, the ligand is accommodated within a cleft of varying depth that extends the length of the protein (10 -13). In contrast to lectins, there is a low density of hydrogen bonds between CBMs and their target saccharides (2). While lectins contain multiple binding sites that interact with mono-or disaccharides of complex carbohydrates (14), CBMs generally contain a single binding site that accommodates 5-6 saccharide units (13,15). Although all Type A or B CBMs characterized to date are monovalent (contain only one binding site), co-operativity between multiple CBMs in a single enzyme can lead to significant increases in affinity compared with the individual modules (16 -18). Ligand specificity in Type A CBMs is generally invariant, while in the Type B modules polysaccharide recognition is variable within a family and normally reflects the catalytic activity of the enzyme from which it is derived (10 -13).
In addition to known CBMs, plant cell wall hydrolases often contain non-catalytic modules of unknown function. This is exemplified by three plant cell wall-degrading enzymes from Cellvibrio japonicus, Xyn10B (xylanase), Abf62A (arabinofuranosidase), and Est1A (acetyl xylan esterase) that contain an identical ϳ150-amino acid module, termed X4, whose role in enzyme function is unclear (19,20). This X4 module is also present in a mannanase (Man5C) from the same organism (21). To understand the mechanism by which glycoside hydrolases attack the complex composite structure that comprises the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence should be addressed. Tel.: 44-0-191-2226962; Fax: 44-0-191-2228684; E-mail: H.J.Gilbert@Newcastle.ac.uk. 1 The abbreviations used are: CBMs, carbohydrate-binding modules; Abf62A, arabinofuranosidase 62A; Est1A, acetyl xylan esterase 1A; Man5C, mannanase 5C; GH, glycoside hydrolase family; DGM, degalac-tosylated mannan; AGE, affinity gel electrophoresis; GST, glutathione S-transferase; IMAC, immobilized metal affinity chromatography; ITC, isothermal titration calorimetry; HPLC, high performance liquid chromatography; ORF, open reading frame; TCEP, Tris(2-carboxyethyl)phosphine. plant cell wall requires knowledge of the function of all the components of these modular enzymes. In this report we show that the X4 module of unknown function in four C. japonicus enzymes is also present in a range of other glycoside hydrolases. The two different Cellvibrio X4 modules bind to the polysaccharide that is the substrate of the cognate enzyme, and thus these proteins represent a new family of CBMs, designated CBM35. The CBM appended to Abf62A requires calcium for binding, a feature that has not been observed previously in CBMs.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Culture Conditions, and Plasmids-The Escherichia coli strains Tuner (Novagen), C41 (DE3, a gift from Prof. A. R. Fersht at the Medical Research Council, Cambridge, UK), JM83 (DE3), and JM83 were used in this study. The plasmid vectors used were pGEX-2T (Amersham Biosciences), pRSET (Invitrogen), pET16b, and pET22b (Novagen), while the recombinant plasmids and the proteins they encode are shown in Fig. 1. To generate pDB1, encoding Abf62A-CBM35, the region of the cognate gene, abf62A comprising nucleotides 457-945 was amplified by PCR using the primers: 5Ј-GCGGGATC-CTCTTCTGCATCCAGTGTGGCC-3Ј and 5Ј-GCGGAATTCTTAACTT-GATGAGGAGGATGACG-3Ј, which contain BamHI and EcoRI restriction sites, respectively, and the resultant DNA fragment was cloned into BamHI/EcoRI-restricted pRSET. The plasmid pDB2, encoding GST-Abf62A-CBM35, was produced by cloning the same region of abf62A present in pDB1 into BamHI/EcoRI-digested pGEX-2T. The plasmid pGP1, encoding Man5C-CBM35, was generated by amplifying nucleotides 589 -981 of man5C using the primers 5Ј-GCGCATATGAT-GGCAGTACCGGAAGGC-3Јand 5Ј-GCGCTCGAGGGCTGGCGAGCG-GATGGTC-3Ј, which contain NdeI and XhoI restriction sites, respectively, and the resultant DNA was cloned into similarly restricted and pET22b. Construction of pDH25 encoding the glycoside hydrolase family 5 (GH5) catalytic module of Man5C was described previously (21). The recombinant plasmid pDH26, which encodes Man5C-CBM35-GH5, was generated by amplifying the region of man5C comprising nucleotides 589 -2495 using the primers 5Ј-GCGCATATGATGGCAGTACCG-GAAGGC-3Ј and 5Ј-GCGGGATCCTTATTGCATCAGCGACCGG-3Ј, which contain NdeI and BamHI restriction sites, respectively. The amplified DNA was digested with NdeI and BamHI and cloned into similarly restricted pET16b. In pDB1, pDH25, and pDH26 the encoded protein module is appended to an N-terminal His tag, while in pGP1 the encoded CBM35 module has a C-terminal His tag. In pDB2 the encoded Abf62A-CBM35 is fused to the C terminus of GST.
Expression and Purification of C. japonicus Proteins in E. coli-To generate the proteins encoded by recombinants of pET16b and pRSET, the E. coli strains Tuner or C41 harboring pDH26 and pDB1, respectively, were cultured in LB supplemented with 50 g/ml ampicillin (1000 ml in 2-liter conical baffle flasks) at 37°C and 180 rpm to mid-exponential phase (OD 600 nm ϳ0.6). The culture was then cooled to 16°C, before expression of CBM35 was induced by addition of isopropyl ␤-thiogalactopyranoside (IPTG) to a final concentration of either 0.1 mM for C41 cells or 0.2 mM for Tuner cells and incubation at 16°C for a further 15 h. The cells were harvested by centrifugation at 4500 ϫ g for 10 min at 4°C and resuspended in one-fiftieth volume 20 mM Tris/HCl buffer, pH 8.0, containing 300 mM NaCl before lysis by sonication and centrifugation (25,000 ϫ g) for 15 min at 4°C to produce cell-free extract. The CBM35 proteins were purified from cell-free extract by immobilized metal affinity chromatography (IMAC) as described previously (22) using Talon TM resin (Clontech). To generate Man5C-CBM35, JM83 (DE3) harboring pGP1 was grown, and recombinant protein was purified as described above except that the expression of the recombinant protein was induced at 37°C for 3 h using a final concentration of 1 mM IPTG. To produce GST-Abf62A-CBM35, E. coli JM83 containing pDB2 was cultured as described above except that expression of the recombinant protein was induced at 30°C for 4 h using a final IPTG concentration of 0.5 mM. The GST fusion protein was purified from cell-free extract by glutathione-Sepharose (Amersham Biosciences) affinity chromatography as described previously (23). Protein concentration was determined from the calculated molar extinction coefficients at 280 nm, which were 60,700 M Ϫ1 cm Ϫ1 , 19,700 M Ϫ1 cm Ϫ1 , 30,440 M Ϫ1 cm Ϫ1 , 126,670 M Ϫ1 cm Ϫ1 and 96,230 M Ϫ1 cm Ϫ1 for GST-Abf62A-CBM35, Abf62A-CBM35, Man5C-CBM35, Man5C-CBM35-GH5, and GH5 alone, respectively.
Sources of Sugars Used-All oligosaccharides and polysaccharides were purchased from Megazyme International, except glucuronoxylan, birchwood xylan, oat-spelt xylan, and hydroxyethyl cellulose (HEC), which were obtained from Sigma. Bacterial microcrystalline cellulose, acid swollen cellulose, and the soluble and insoluble fractions of oatspelled xylan were prepared as described previously (24). To generate insoluble degalactosylated mannan (DGM), 0.5 g of carob galactomannan (high viscosity, galactose/mannose ϭ 1:4) was treated with 10 units of C. japonicus ␣-galactosidase Gal27A (25) in 10 ml of 50 mM sodium phosphate buffer, pH 7.0, at 37°C for 18 h. Removal of the galactose side chains caused the mannan chains to aggregate and become insoluble. The precipitated mannan was washed extensively in distilled water. The monosaccharide content of the insoluble DGM polysaccharide, determined by acid hydrolysis followed by HPLC, revealed that the ratio of galactose/mannose was 1:50.
Affinity Gel Electrophoresis (AGE)-AGE was performed as described previously (26) using oat-spelt xylan (arabinose/xylose ϭ 1:10), rye arabinoxylan (high viscosity; arabinose/xylose ϭ 1:1), 4-O-methylglucuronoxylan, birchwood xylan (glucuronic acid/xylose ϭ 1:10; Ref. 27), barley ␤-glucan (medium viscosity), hydroxyethylcellulose, konjac glucomannan (low viscosity; glucose/mannose ϭ 2:3), carob galactomannan (low viscosity; galactose/mannose ϭ 1:4), debranched ␣1,5-arabinan (sugar beet), ␤1,4-galactan (potato), and rhamnogalacturonan (soy bean) as ligands. The ratio of sugars in each polysaccharide was provided by the manufacturer. Briefly, the continuous gels contained 7.5% (w/v) acrylamide in 25 mM Tris, 250 mM glycine buffer, pH 8.3. For ligand-containing gels, glycan was added to the separating gel mixtures to 0.01-5 mg/ml prior to polymerization. Native polyacrylamide gels, with and without ligand, were polymerized at the same time and were electrophoresed in the same gel tank. The proteins (5 g) were electrophoresed at 25°C and 10 mA/gel in gels with and without ligand for 2 h. GST was used as a negative, non-interacting control. Proteins were visualized by Coomassie Blue staining. The migration distances of the CBMs and the reference protein were measured from the bottom of the protein bands evident on the gels, and these data were used to determine the dissociation constants (K D ) from plots of 1/(R 0 Ϫ r) versus 1/C according to the affinity equation shown in Equation 1, where r is the relative migration distance of the CBM in the presence of ligand in the gel, R 0 is the relative migration distance of the free CBM in the absence of ligand, R C is the relative migration distance of the complex at high excess of ligand where all CBM molecules are fully complexed, C is the concentration of the ligand in the gel, and K D is the dissociation constant of CBM for the macromolecular ligand. K D values were determined as the inverse of the absolute value of the intercept on the abscissa of data plotted according to the affinity equation. All migration distances of the CBMs were measured relative to the migration of the reference protein GST.

Creation of the Apo Form of Abf62A-CBM35-Purified
Abf62A-CBM35 at a concentration of 400 M in 50 mM sodium HEPES buffer, pH 8.0, was passed through a 20 ϫ 1.5-cm column containing a 10-ml bed volume of Chelex-100 (Sigma) under the flow of gravity. The buffer used for ITC was treated in the same way to remove any traces of calcium.
Gel Filtration Chromatography-Gel filtration was performed using a HiLoad 16/60 Superdex 75 column (Amersham Biosciences) attached to a Bio-Rad Biologic HR FPLC system. The column was calibrated using Sigma gel filtration molecular weight markers ranging from 12.4 to 66 kDa. Blue dextran (2 MDa) was used to determine the column void volume. Protein (1 ml) was loaded on the column and run at 1 ml/min in 50 mM sodium HEPES buffer, pH 8.0, for 120 min. Where appropriate the reducing agent Tris(2-carboxyethyl)phosphine (TCEP) was added to the sample and the running buffer at a final concentration of 1 mM.
Isothermal Titration Calorimetry (ITC)-ITC measurements were made at 25°C using a Microcal Omega titration calorimeter. The Man5C-CBM35 and Abf62A-CBM35 modules were dialyzed extensively against 50 mM sodium HEPES buffer, pH 8.0, and the ligands were dissolved in the same buffers to minimize heats of dilution. Where appropriate, TCEP, calcium, or EDTA at final concentrations of 1, 5, and 10 mM, respectively, were added to Abf62A-CBM35 and carbohydrate ligand prior to ITC. During a titration experiment, the protein sample (150 -600 M), stirred at 300 rpm in a 1.4331-ml reaction cell maintained at 25°C, was injected with 25-50 successive 10-l aliquots of ligand comprising polysaccharide (5-25 mg/ml) or oligosaccharide (5-15 mM), at 200-s intervals. The apo form of Abf62A-CBM35 (150 M) was titrated against 3 mM calcium chloride to determine whether calcium bound independently to the protein. Prior to the titration the cell was washed out with 5 mM EDTA, followed by an extensive Chelextreated buffer wash to remove any traces of the metal ion from the machine. Integrated heat effects, after correction for heats of dilution, were analyzed by non-linear regression using a single site-binding model (Microcal Origin, version 5.0). Fitted data yield the association constant (K A ) and the enthalpy of binding (⌬H). Other thermodynamic parameters were calculated using the standard thermodynamic equation: ϪRTlnK A ϭ ⌬G ϭ ⌬H Ϫ T⌬S. The c values (product of the association constant ϫ the concentration of the acceptor ϫ the number of binding sites on the acceptor) were 3-30. At least two independent titrations were performed for each ligand tested. The molar concentration of Man5C-CBM35 binding sites present in galactomannan and glucomannan was determined by altering the concentration of polysaccharide used for regression of the isotherm until the fit yielded a value of 1 for n (number of binding sites on each molecule of CBM). The assumption that n ϭ 1 was based on the oligosaccharide titrations, which all displayed a stoichiometry of 1:1. For analysis of the xylan binding data with Abf62A-CBM35, this approach could not be used because the CBM did not interact with xylo-oligosaccharides. Instead, the binding data was fitted by treating the CBM in the sample cell as the ligand and the polysaccharide as the acceptor. This gives accurate values for K A and ⌬H, but not n (in this case the number of binding sites on each molecule of polysaccharide).
Binding to Insoluble Polysaccharides-The binding of the CBM35 proteins to insoluble polysaccharides (acid-swollen cellulose, bacterial microcrystalline cellulose, DGM, ivory nut mannan, and insoluble oatspelt xylan) was determined qualitatively using SDS-PAGE. Pure protein (100 g in 20 mM Tris/HCl buffer, pH 8.0) was mixed with 2 mg of polysaccharide in a final volume of 100 l. Tubes were incubated on ice for 1 h, with regular gentle mixing before being centrifuged at 13,000 ϫ g for 1 min and the supernatant, containing unbound protein, carefully removed. The polysaccharide pellet was then washed in 100 l of the same buffer, before being resuspended in 50 l of 10% (w/v) SDS and boiled for 10 min to dissociate any bound protein. Controls with protein but no polysaccharide were included to insure that no precipitation occurred during the assay period. Bound and unbound fractions were analyzed by SDS-PAGE using a 12.5% (w/v) polyacrylamide gel.
Enzyme Assays-The mannanase activity of the Man5C derivatives Man5C-CBM35-GH5 and GH5 was evaluated by HPLC with mannohexaose, ivory nut mannan or insoluble DGM as substrates using equal concentrations of protein. The Man5C derivatives (50 nM) were incubated with 5 mg/ml ivory nut mannan or insoluble DGM in 50 mM sodium phosphate/12 mM citrate (PC) buffer, pH 6.5 at 37°C for up to 5 h in a total volume of 0.5 ml. At regular time intervals, a 40-l aliquot was removed, the enzyme was inactivated by boiling for 10 min, and mannotetraose, the primary reaction product, was quantified by HPLC following the method of Hogg et al. (21). To determine the rate of mannohexaose hydrolysis, the hexasaccharide (0.6 mM) was incubated with 40 nM enzyme in PC buffer, pH 6.5 at 37°C for up to 30 min in a total volume of 0.4 ml, and the release of mannotetraose was quantified as described above.

RESULTS AND DISCUSSION
Identification of the CBM35 Family of Protein Modules-Previous studies identified a protein module, originally termed X4, comprising ϳ150 amino acids in three C. japonicus enzymes, Xyn10B (formerly XynB), Abf62A (formerly XynC), and Est1A (formerly XynD) that are involved in the hydrolysis of xylan. All three enzymes contain an identical N-terminal region that comprises a typical family 2a CBM joined via a serine-rich linker to the X4 module (19,20). As the X4 modules in these Cellvibrio enzymes are shown to bind polysaccharides (see below), henceforth these sequences will be designated as a new family of CBMs (family 35, CBM35). When the primary structure of Abf62A-CBM35 was used to query databases using BLAST, a number of sequences that display similarity to this sequence were identified in enzymes that modify carbohydrates, including glycoside hydrolases and lyases that attack the hemicellulosic and pectic polysaccharides, respectively, within the plant cell wall. Based on sequence similarities the CBM35 modules can be group into three clads containing modules derived from xylan/pectin-modifying enzymes, mannanases, and isomalto-oligosaccharide-modifying enzymes, respectively. The alignment of the CBM35s and phylograms of this family are displayed in Fig. 2. It should be noted that while CBM35 constitutes a discrete protein family, it displays a distant relationship with family 6 CBMs. The observation that CBM35s lack the three aromatic residues (e.g. Trp-92, Tyr-33, and Trp-39 in CBM6 -2 from Cellvibrio mixtus Cel5A), which play a pivotal role in ligand recognition by CBM6 proteins (28), provides further support for the view that CBM35 and CBM6 comprise discrete families of CBMs.
Expression of the CBM35 Modules-To determine the function of CBM35s in glycoside hydrolases we have focused on two enzymes from C. japonicus, Man5C and Abf62A (20,21). The mannanase comprises a family 5 and family 10 CBM, which are joined by typical serine-rich linker sequences, and a CBM35 module abutted to the GH5 catalytic module. The region of the arabinofuranosidase and mannanase genes encoding the respective CBM35s were cloned into E. coli expression vectors, and the encoded proteins, fused to a His tag supplied by the vector, were produced in soluble form in the host bacterium. The two proteins were purified by IMAC. Purified Abf62A-CBM35 contained both monomeric and dimeric species as judged by non-reducing SDS-PAGE and gel filtration; however, the addition of reducing agent converted the protein to its monomeric form (Fig. 3), suggesting that dimerization was the result of an interchain disulfide bond. The subsequent biochemical characterization of Abf62-CBM35 was carried out in the presence of reducing agent (unless stated otherwise) to ensure that it was maintained in its monomeric form.
Family 35 CBMs Bind to Polysaccharides-The biochemical properties of the two CBM35 modules were initially evaluated using AGE. As a result of its high pI, Abf62A-CBM35 was fused to glutathione S-transferase (GST-Abf62A-CBM35) to ensure migration into non-denaturing gels. Examples of the affinity gels and subsequent plots used to quantify binding are displayed in Figs. 4 and 5, respectively, and the full data set is presented in Table I. The CBM35 module from Man5C (Man5C-CBM35) binds to galactomannan and glucomannan but does not interact with substituted or unsubstituted xylans, soluble ␤-linked glucose polymers, rhamnogalacturonan, or the pectin side chains arabinan and galactan. The capacity of Man5C-CBM35 to bind insoluble polysaccharides was evaluated using SDS-PAGE to monitor bound and unbound protein.
Example data are shown in Fig. 6 and the full data set presented in Table I. Man5C-CBM35 does not bind to Avicel, acid-swollen cellulose, bacterial crystalline cellulose, or ivory nut mannan, but does associate with DGM, an insoluble form of mannan generated by the enzymic removal of galactose side chains from galactomannan. The polysaccharide chains in FIG. 2. Alignment of CBM35 modules (panel A) and phylograms (panels B and C) DGM are likely to associate in a less ordered fashion than those found in crystalline mannans and are therefore analogous to amorphous cellulose where discrete polysaccharide chains are able to interact with CBMs that contain a ligand binding cleft (18). Thus, Man5C-CBM35 appears to exhibit features typical of Type B CBMs, which accommodate individual polysaccharide chains in a binding site that displays a cleft topology (10,15), but are unable to interact with the flat surfaces of highly crystalline ligands such as ivory nut mannan.
Analysis of the polysaccharide binding properties of the Abf62A-CBM35 module were carried out using GST-CBM35 for soluble ligands and the His-tagged CBM35 module for insoluble polysaccharides. The data show that the protein binds to both soluble (Figs. 4 and 5) and insoluble oat-spelt xylan ( Fig.  6) but displays very weak affinity for soluble arabinoxylan from rye ( Fig. 5 and Table I). AGE analysis also revealed slight retardation of Abf62A-CBM35 by glucuronoxylan, although binding was too weak to quantify ( Fig. 4 and Table I). The protein does not interact with birchwood xylan, soluble, or insoluble forms of cellulose, ␤-glucan, galactomannan, glucomannan, or pectins (Table I). These data demonstrate that Abf62A-CBM35 is a xylan-specific CBM that interacts preferentially with unsubstituted forms of the polysaccharide. The results presented above reveal that both Abf62A-CBM35 and Man5C-CBM35 display polysaccharide binding properties and thus justify the reclassification of X4 modules as family 35 CBMs (CBM35).
The Use of ITC to Measure Ligand Binding of Man5C-CBM35-ITC was used to measure binding of Man5C-CBM35 to both polysaccharide and oligosaccharide ligands. Examples of these titrations are presented in Fig. 7, and the full data set is displayed in Table II. The thermodynamics of the interaction of Man5C-CBM35 with polysaccharides and oligosaccharides is enthalpy-driven with the change in entropy making an unfavorable contribution to ligand binding. This pattern of energetics is typical of the binding of proteins to soluble saccharides (12,14,18,29). The CBM binds to manno-oligosaccharides with a stoichiometry of 1:1 displaying maximal affinity for mannopentaose and mannohexaose. Binding of Man5C-CBM35 to mannotetraose and mannotriose was too weak to accurately quantify by ITC, although the estimated K A values were 2.7 ϫ 10 3 M Ϫ1 and 5 ϫ 10 2 M Ϫ1 , respectively (data not shown). The K A of Man5C-CBM35 for mannopentaose, substituted at the 3 and 4 position (from the reducing end) with ␣-1,6-linked galactosyl residues (6 3 ,6 4 -␣-D-galactosyl-mannopentaose), was estimated to be 1.3 ϫ 10 3 M Ϫ1 (affinity was too low to determine K A accurately by ITC), which is considerably lower than for unsubstituted mannopentaose. This indicates that the side chains have a detrimental effect on binding. Without a structure, it is difficult to predict the regions of the mannan-binding CBM that are unable to accommodate galactosyl side chains, although the poor affinity for 6 3 ,6 4 -␣-D-galactosyl-mannopentaose (compared with mannopentaose) indicates that the ␣-1,6linked decoration is likely to form a steric clash with the protein when located in both subsites 3 and 4. The observations that the affinity of Man5C-CBM35 for galactomannan is similar to mannopentaose and at saturation approximately six mannose units constitute a single CBM binding site indicate that Man5C-CBM35 is able to bind tightly to the vast majority of the polysaccharide. As 25% of the mannose residues are decorated with galactose (manufacturer's data), it would appear that side chains can be accommodated in several subsites, providing the decoration does not occur on adjacent backbone sugars. In general there is little information on the binding of CBMs to decorated mannans. Charnock et al. (15) showed that CBM29-2, from Piromyces equi NCP1, is able to bind to galactomannan more tightly than to mannohexaoase, although the side chains in 6 3 ,6 4 -␣-D-galactosyl-mannopentaose cause a 10fold reduction in affinity. The crystal structure of the protein in complex with mannohexaose reveals that alternate subsites in the protein are able to accommodate side chains substituted at the O6 position on the mannose backbone, which explains why the CBM binds weakly to ligands that are decorated on adjacent mannose residues. In addition, Boraston et al. (11) show that a mannan-binding family 27 CBM (CBM27) from the Thermotoga maritima mannanase Man5A is able to fully saturate carob galactomannan even though it binds extremely poorly to 6 3 ,6 4 -␣-D-galactosyl-mannopentaose. The three-dimensional structure of CBM27 in complex with the decorated ligand and mannopentaose reveals that a galactosyl residue can be accommodated in subsites 1, 2, 3, and 5 but sterically clashes in subsite 4 (11). While a similar topology in the ligandbinding cleft of the Man5C-CBM35 may explain its capacity to bind galactomannan, the observation that the side chains in 6 3 ,6 4 -␣-D-galactosyl-mannopentaose cause a much larger (87fold) reduction in affinity, compared with mannopentaose, in CBM27 than in Man5C-CBM35 points to possible differences in the mechanism of ligand binding in these two modules.
While Man5C-CBM35 binds to glucomannan with an affinity similar to galactomannan, the protein does not interact with glucose homopolymers such as soluble or insoluble cellulose, or cellohexaose. At saturation, 7-8 sugar units in the glucomannan backbone constitutes a single binding site for the CBM, indicating that the protein is able to bind to most of the polysaccharide, although the degree of coverage is slightly less than for galactomannan (see above). As 40% of glucomannan consists of glucose moieties (manufacturer's data), it is apparent that this sugar can be accommodated at several subsites in Ϫ ϩ a Abf62A-CBM35 construct used in assessing binding to soluble polysaccharides was GST-Abf62A-CBM35, while Abf62A-CBM35 was used to evaluate binding to insoluble polysaccharide (see Fig. 1  Man5C-CBM35. It would appear, therefore, that several subsites in Man5C-CBM35 are able to bind to glucose or mannose residues, and the protein is able to interact with either an axial (mannose) or equatorial (glucose) O2 at these locations. This promiscuity in ligand recognition is similar to CBM29-2, which is able to bind gluco-or manno-configured sugars at each of its six subsites, and in two of these subsites the same amino acid is able to interact with an equatorial or axial O2 (15). In contrast to CBM29, the inability of Man5C-CBM35 to bind to a homopolymer of glucose indicates that either the interaction with an axial O2 is a critical element of sugar binding in at least one subsite, and/or an equatorial O2, by making a steric clash with the protein at one or more subsites, prevents the CBM35 from binding to cellohexaose or cellulose. CBM27 is also able to bind glucomannan but is unable to interact with cellopentaose or insoluble regenerated cellulose (11), again suggesting that glucose can bind at selected subsites but is precluded from others. The crystal structures of CBM27 in complex with mannose-based ligands reveal that the protein makes hydrogen bonds with the axial O2 of mannose at subsites 2, 3, and 4. While the equatorial O2 of glucose can be tolerated in subsites 1, 2, and 5, steric clashes prevent the sugar from binding at subsites 3 and 4 explaining why the protein does not bind to cellulose. A similar selectivity for mannose at specific subsites can be invoked to explain why Man5C-CBM35 binds to glucomannan but not cellulose or cello-oligosaccharides.
Man5C-CBM35 Potentiates Mannanase Activity Against Insoluble Amorphous Mannan-To evaluate whether Man5C-CBM35 potentiates mannanase activity, derivatives of Man5C comprising the GH5 catalytic module (GH5) and GH5 fused to CBM35 (Man5C-CBM35-GH5) were expressed in E. coli, and the catalytic activity of these proteins was evaluated. Both derivatives of Man5C display similar activities against mannohexaose and insoluble ivory nut mannan; however, Man5C-CBM35-GH5 hydrolyzes insoluble DGM five times faster than GH5 alone (data not shown). Addition of the Man5C-CBM35 to GH5 in trans (as discrete proteins) in various ratios ranging from 1:1 to 10:1 (CBM35/GH5) did not increase the activity of the mannanase against any of the substrates evaluated (data not shown). These results suggest that the activity of the mannanase is compromised by restricted access to insoluble DGM, but that the enzyme is able to rapidly access soluble substrates. CBM35, by bringing the catalytic module in the Man5C derivative CBM35-GH5 into intimate and prolonged association with DGM, increases enzyme access to the substrate leading to more efficient catalysis. The CBM does not potentiate mannanase activity in trans, indicating that the module does not mediate its affect by disrupting the interchain interactions in mannan, which is in contrast to some CBM2a proteins that enhance cellulase action by disrupting the surface of crystalline cellulose, leading to an increase in substrate access (30,31). The inability of CBM35 to improve the activity of the catalytic module of Man5C against ivory nut mannan is consistent with the observation that the CBM does not bind to the crystalline polysaccharide. Overall, these data demonstrate that Man5C-CBM35 displays properties similar to several cellulose (crys-talline and non-crystalline) and xylan-binding CBMs, which have also been shown to enhance the catalytic activity of appended glycoside hydrolases against insoluble polysaccharides by increasing enzyme-substrate proximity (3,4,32).
Calcium Mediates Binding of Abf62A-CBM35 to Xylan-For ITC studies, the CBM35 module linked to a His tag was titrated with oat-spelt xylan in HEPES buffer. In the presence of 5 mM calcium the protein bound to xylan; however, when the divalent ion was replaced with 10 mM EDTA no interaction between Abf62A-CBM35 and the polysaccharide was evident. (Fig. 8). It should be noted that when Abf62A-CBM35, which had not been treated with reducing agent (exists as a dimer/ monomer), was titrated with oat-spelt xylan the affinity was similar to the monomeric form of the CBM. However, aggregation of the polysaccharide occurred implying that the two xylan-binding sites in the disulfide-mediated dimer cross-link individual xylan chains (data not shown). Xylan aggregation by CBMs has also been demonstrated by the three linked CBM6 modules in Clostridium stercorarium xylanase 11A (18). This phenomenon is also well established in the lectin field where multiple binding sites on the proteins mediate cross-linking of complex multivalent carbohydrates (14).
To investigate the role of the calcium in ligand-binding in more detail, an apo form of Abf62A-CBM35 was produced by treating the protein with Chelex (see "Experimental Procedures"). ITC shows that the apo form of the CBM does not interact with oat-spelt xylan in HEPES buffer; however, the protein binds tightly to the polysaccharide in the presence of 5 mM CaCl 2 ( Fig. 8 and Table III). The ⌬H and T⌬S values for the binding of the CBM35 to oat-spelt xylan are negative, similar to Man5C-CBM35 and other CBMs that interact with soluble polysaccharides (10,11,15). Titration of the apo form of Abf62A-CBM35 with calcium demonstrates that the protein binds tightly to the divalent metal ion ( Fig. 8 and Table III). These data indicate that Abf62A-CBM35 displays an absolute requirement for calcium when binding to xylan, and furthermore, the protein is able to interact with the metal ion in the absence of the polysaccharide.
Previous studies have shown that CBMs from families 4, 6, 9, and 22 contain one or more calcium ions located at sites remote from the ligand binding cleft, suggesting a structural role for the metal (10,29,(33)(34)(35). In support of this view, removal of calcium from both cellulose and xylan-binding CBM4s reduced the temperature at which the proteins unfolded by 8°C (34) and 23°C (35), respectively. However, the loss of this metal in both family 4 and 22 CBMs did not influence ligand binding. This report therefore provides one of the first examples of calcium playing a direct role in the binding of a CBM to its target ligand. Indeed, the demonstration that the metal ion also mediates the binding of a family X9 CBM to xylan 2 suggests that the involvement of calcium in the association of CBMs with their target ligands may be a common phenomenon. Although the role of calcium in the interaction of CBMs 2 A. B. Boraston, personal communication.

X4 Modules Represent a New Family of CBMs
with polysaccharides and oligosaccharides has not been extensively studied, the importance of this metal in the binding of lectins to carbohydrates is well established (14,18). Abf62A-CBM35 Only Binds to Poorly Substituted Xylans-To further investigate the interaction of Abf62A-CBM35 with carbohydrates, ITC was performed using 4-O-methylglucuronoxylan and xylohexaose as ligands (data not shown). No significant binding was observed with either of the sugars tested, confirming that the protein targets extended unsubstituted regions of xylan.
The ligand specificity of Abf62A-CBM35 is in sharp contrast to the other xylan-binding CBMs described to date, which are all able to interact with substituted and unsubstituted forms of the hemicellulose with similar affinities (10,12,18). Substituted xylans contain arabinose and/or 4-O-methylglucuronic acid groups attached to O2 and O3s of the xylose backbone. Structural studies on CBM15 from C. japonicus Xyn10C complexed with xylopentaose have revealed how xylan side chains can be accommodated. Six of the ten C2-OH and C3-OH groups in the pentasaccharide are solvent-exposed and therefore FIG. 8. ITC data showing the interaction Abf62A-CBM35 with oat-spelt xylan and calcium. The upper parts of the panels show the raw binding heats, the lower parts are the integrated binding heats minus the control heats of dilution fitted to a single-site binding model. In all cases 25 mg/ml xylan and 150 M protein were used. Calcium was at 5 mM when included in titrations with xylan and was at 3 mM (in the syringe) when titrated into apoAbf62A-CBM35. capable of being substituted without clashing with the protein; in only one subsite is the CBM unable to interact with a decorated xylose moiety (12). The restricted ligand specificity of Abf62A-CBM35 indicates that the O2 and O3 hydroxyls of bound ligand are rarely solvent-exposed. Within this context it is perhaps worth noting that Abf62A-CBM35 can interact with oat-spelt xylan but not birchwood xylan, even though both polysaccharides have a similar level (ϳ10%) of decoration. This may be due to the location of the different monosaccharide substituents in each of these xylans. The arabinose side chains in oat-spelt xylan are attached mainly to O3, whereas 4-Omethylglucuronic acid moieties (found in birchwood xylan and glucuronoxylan) are linked to O2 (36). It is possible, therefore, that the binding site of Abf62A-CBM35 can accommodate O3linked sugars, but cannot tolerate a substituent at O2 in any of its binding subsites. The large reduction in affinity observed with rye arabinoxylan can be explained by the presence of arabinose decorations on both O2 and O3 of the xylose backbone of this polysaccharide (36). The capacity of Abf62A-CBM35 to bind unsubstituted xylan but not to interact with xylohexaose is also in contrast to the other xylan-binding CBMs described to date, which display similar affinities for xylohexaose and xylan indicating that the binding site in these proteins can accommodate up to six xylose residues (10,12,29). It is possible that Abf62A-CBM35 contains a binding cleft that is considerably longer than other CBMs and thus requires ligands with a degree of polymerization significantly larger than 6 to form a productive complex. Interestingly, CBM29-2 from P. equi NCP1 also displays a preference for polysaccharides over oligosaccharides, with an affinity for galactomannan ϳ37-fold higher than that for mannohexaose (15). Although the mechanism for this specificity is unclear, it appears therefore that a preference for polysaccharides over oligosaccharides may be a more common feature of CBMs than had previously been recognized. The specificity of Abf62A-CBM35 for poorly decorated regions of xylan points to a unique targeting role for this protein module, although the biological relevance of this specificity (when the CBM is appended to enzymes that remove the side chains from this polysaccharide) is currently unclear. CONCLUSIONS This report shows that the X4 family of non-catalytic modules comprises a novel family of CBMs, designated CBM35. The CBM35s bind to the target substrate of the appended catalytic module, suggesting that the ligand specificity of this family reflects the enzyme from which they are derived. This report also demonstrates, for the first time, that a mannan-binding CBM is able to potentiate mannanase activity against insoluble mannan, similar to the capacity of CBMs that recognize cellulose and xylan to enhance the enzymic hydrolysis of insoluble forms of their target ligands. Abf62A-CBM35 displays several features that are unique not only within xylan-binding CBMs but within CBMs in general. Thus, although it is well established that many lectins require calcium to bind to their target carbohydrates, Abf62A-CBM35 represents one of the very first examples of the divalent metal ion playing a pivotal role in the interaction of a CBM with its ligand. Another unique feature of Abf62A-CBM35 is that the module interacts specifically with highly unsubstituted xylose polymers that have a dp Ͼ6, while all other xylan-binding CBMs described to date recognize small xylo-oligosaccharides and are able to bind, with similar affinity, to both decorated and poorly substituted xylans.