The Structure of a Streptomyces avermitilis α-l-Rhamnosidase Reveals a Novel Carbohydrate-binding Module CBM67 within the Six-domain Arrangement*

Background: α-l-Rhamnosidase hydrolyzes α-linked l-rhamnose from rhamnoglycosides or polysaccharides. Results: The crystal structure of Streptomyces avermitilis α-l-rhamnosidase belonging to glycoside hydrolase family 78 was determined. Conclusion: The l-rhamnose complexed structure revealed the catalytic mechanism of the enzyme and a calcium-dependent carbohydrate-binding module. Significance: Efficient catalysis of an exo-rhamnosidase requires a novel carbohydrate-binding module that binds terminal l-rhamnose sugars. α-l-Rhamnosidases hydrolyze α-linked l-rhamnosides from oligosaccharides or polysaccharides. We determined the crystal structure of the glycoside hydrolase family 78 Streptomyces avermitilis α-l-rhamnosidase (SaRha78A) in its free and l-rhamnose complexed forms, which revealed the presence of six domains N, D, E, F, A, and C. In the ligand complex, l-rhamnose was bound in the proposed active site of the catalytic module, revealing the likely catalytic mechanism of SaRha78A. Glu636 is predicted to donate protons to the glycosidic oxygen, and Glu895 is the likely catalytic general base, activating the nucleophilic water, indicating that the enzyme operates through an inverting mechanism. Replacement of Glu636 and Glu895 resulted in significant loss of α-rhamnosidase activity. Domain D also bound l-rhamnose in a calcium-dependent manner, with a KD of 135 μm. Domain D is thus a non-catalytic carbohydrate binding module (designated SaCBM67). Mutagenesis and structural data identified the amino acids in SaCBM67 that target the features of l-rhamnose that distinguishes it from the other major sugars present in plant cell walls. Inactivation of SaCBM67 caused a substantial reduction in the activity of SaRha78A against the polysaccharide composite gum arabic, but not against aryl rhamnosides, indicating that SaCBM67 contributes to enzyme function against insoluble substrates.

␣-L-Rhamnosidases hydrolyze ␣-linked L-rhamnosides from oligosaccharides or polysaccharides. We determined the crystal structure of the glycoside hydrolase family 78 Streptomyces avermitilis ␣-L-rhamnosidase (SaRha78A) in its free and Lrhamnose complexed forms, which revealed the presence of six domains N, D, E, F, A, and C. In the ligand complex, L-rhamnose was bound in the proposed active site of the catalytic module, revealing the likely catalytic mechanism of SaRha78A. Glu 636 is predicted to donate protons to the glycosidic oxygen, and Glu 895 is the likely catalytic general base, activating the nucleophilic water, indicating that the enzyme operates through an inverting mechanism. Replacement of Glu 636 and Glu 895 resulted in significant loss of ␣-rhamnosidase activity. Domain D also bound L-rhamnose in a calcium-dependent manner, with a K D of 135 M. Domain D is thus a non-catalytic carbohydrate binding module (designated SaCBM67). Mutagenesis and structural data identified the amino acids in SaCBM67 that target the features of L-rhamnose that distinguishes it from the other major sugars present in plant cell walls. Inactivation of SaCBM67 caused a substantial reduction in the activity of SaRha78A against the polysaccharide composite gum arabic, but not against aryl rhamnosides, indicating that SaCBM67 contributes to enzyme function against insoluble substrates.
In a previous study (9), we characterized a GH78 ␣-L-rhamnosidase from Streptomyces avermitilis (SaRha78A), showing that the enzyme hydrolyzed aryl rhamnosides and rhamnosecontaining polysaccharides. In this article, we provide the crystal structure of SaRha78A in apo form and in complex with L-rhamnose. The data revealed the catalytic mechanism of the enzyme and identified a novel non-catalytic carbohydratebinding module (CBM), SaCBM67, the founding member of CBM67. Analysis of the ligand specificity of SaCBM67 showed that the protein module bound L-rhamnose through a calciumdependent mechanism in a short binding cleft. Inactivation of SaCBM67 through mutagenesis or chelation of calcium showed that the CBM made a substantial contribution to the activity of SaRha78A against polysaccharides, but not against aryl rhamnosides.

EXPERIMENTAL PROCEDURES
Crystallization of SaRha78A-Recombinant SaRha78A was expressed in Escherichia coli and purified by histidine tag affinity chromatography as described by Ichinose et al. (9). The purified protein solution in 2 mM Tris-HCl buffer, pH 7.0, containing 20 mM NaCl, was concentrated to 10 mg ml Ϫ1 (A 1 mm 280 ϭ 2.2 units) by ultrafiltration using a YM-30 membrane (Millipore, Billerica, MA), and filtered through a 0.1-m membrane. The protein was crystallized by the sitting-drop vapor diffusion method using a precipitant solution composed of 10% (w/v) PEG6000, 5% (v/v) MPD, and 0.1 M HEPES pH 6.8. Plate-shaped crystals with 500 ϫ 50 ϫ 10 m dimensions were consistently obtained using 50 l of the reservoir solution with a drop consisting of 1 l of protein solution and 1 l of reservoir solution. Selenomethionine-labeled SaRha78A was expressed in LeMaster medium using the methionine-auxotrophic strain B834(DE3) (10) and crystallized in the same condition used for the native enzyme.
Data Collection and Structure Determination-Diffraction experiments for native crystals were conducted at the beamline BL41XU of the synchrotron facility SPring-8, Hyogo, Japan. Crystals were soaked into the reservoir solution containing 13% (v/v) glycerol, scooped in a 0.5-mm nylon loop (Hampton Research, Aliso Viejo, CA), and flash-frozen in a nitrogen stream at 95 K. Diffraction data were collected at a wavelength of 0.97915 Å with a Quantum 210 CCD detector (Area Detector Systems Corp., Poway, CA). Diffraction experiments for the selenomethionine-substituted crystals were conducted at beamline BL-5 of the Photon Factory, High Energy Accelerator Research Organization, Tsukuba, Japan. Diffraction data were collected at a wavelength of 0.97909 Å with a Quantum 315 CCD detector (ADSC). For structural analyses of the enzyme in complex with L-rhamnose (Wako Pure Chemical Industry), SaRha78A crystals were soaked into a drop containing 1% (w/v) L-rhamnose in the precipitant solution for 10 min before the diffraction experiment. Diffraction data of the L-rhamnose complex to 1.9-Å resolution were collected at beamline BL-NW12 of the Photon Factory Advanced Ring. Diffraction data were collected at a wavelength of 1.0000 Å with a Quantum 210 CCD detector (ADSC). All data were integrated and scaled using the program DENZO and SCALEPACK in the HKL2000 program suite (11).
The crystal structure was determined by the single-wavelength anomalous dispersion method using selenomethioninelabeled crystals. In total 12 selenium atom positions were determined, and initial phases were calculated using the program SOLVE/RESOLVE (12,13). The initial model building was conducted by the automodeling program ARP/wARP (14) within the CCP4 program suite (15). Manual model building and molecular refinement were performed using Coot (16) and 18).
For analysis of the L-rhamnose-protein complex, structural determination was conducted using the ligand-free structure as the starting model and the bound L-rhamnose was observed in the difference electron density map. Data collection and refinement statistics are given in Table 1. Stereochemistry of the models was analyzed with the program Rampage (19). Structural drawings were prepared by the PyMol program (DeLano Scientific LLC, Palo Alto, CA).
Binding of SaCBM67 to its target ligands was determined by isothermal titration calorimetry using a MicroCal VP-ITC. Titrations were carried out at 25°C in 20 mM Na-HEPES buffer, pH 7.5, containing 2 mM CaCl 2 , unless otherwise stated. The reaction cell contained protein at 700 -800 M (depending on the ligand), whereas the syringe contained 20 mM of the potential ligands (L-fucose, L-mannose, and L-rhamnose purchased from Sigma). The data were analyzed using MicroCal Origin version 7.0, which yielded the change in enthalpy (⌬H), association constants (K a ) and stoichiometry of binding (n), enabling changes in entropy (⌬S) and Gibbs free energy (⌬G) to be calculated using the equation ϪRTInK a ϭ ⌬G ϭ ⌬H-T⌬S.

RESULTS
Overall Structure of SaRha78A-The crystal structure of SaRha78A was determined by the single-wavelength anomalous dispersion method using selenomethionine crystal data. The structure of the apo form of SaRha78A and the protein in complex with L-rhamnose were determined. Structure refinement statistics are summarized in Table 1. The quality and accuracy of the final structures are indicated by the observation that no residues fell within the disallowed region of the Ramachandran stereochemistry plot. The recombinant SaRha78A molecule is composed of a single polypeptide chain of 1043 amino acids, where the C-terminal His tag, 1031 KLAAALEHH-HHHH 1043 , were derived from the expression vector and purification tag, respectively. The two residues N-terminal Met 1 and Ser 2 and the 11 C-terminal residues, Ala 1033 to His 1043 , were not identified due to lack of electron density. The final model consisted of one SaRha78A molecule in the asymmetric unit bound to a single calcium ion. The ligand free structure contained three Tris molecules, and the L-rhamnose complex structure contained three sodium ions.
The protein forms a multidomain structure comprised of six distinct domains, one ␣-domain and five ␤-domains ( Fig. 1). They were designated as domain N, domain D (designated SaCBM67 as it fulfills a carbohydrate binding function, see below), domain E, domain F, domain A (designated SaRha78 CM as it comprises the catalytic module, see below), and domain C from the N terminus to the C terminus. Domain N (Ala 3 -Gly 114 ) consists of 10 ␤-strands folding into 4-and 3-stranded ␤-sheets and adopt a fibronectin type 3 fold. Domain D (SaCBM67 Pro 133 -Pro 297 ) and domain E (Ala 115 -Ala 132 , Val 298 -Gly 426 , and Glu 444 -Gly 478 ) display a ␤-jellyroll fold consisting of 11 and 13 ␤-strands, respectively (SaCBM67 is described in detail below). Domain E provided two ␤-strands (Pro 427 -Lys 443 ) that were inserted into one of the ␤-sheets of Domain F (Pro 479 -Ala 596 and Pro 427 -Lys 443 ), which also displayed a ␤-jellyroll fold. Domain F consisted of 16 ␤-strands organized into two parallel ␤-sheets.
Domains N, E, F, and C are located around SaRha78 CM . Domains F and C interact with SaRha78 CM through large interfaces, whereas domains N and E also abut onto the catalytic module, but through a small contact area. SaCBM67 protrudes  (Fig. 2). Domain E also superimposed with an equivalent domain in BsRhaB with a root mean square deviation of 1.8 Å, but the equivalent domain is displaced in  BT1001. SaCBM67 were essentially in a different position in SaRha78A and BsRhaB, but was absent in BT1001 (Fig. 2). Crystal Structure of SaRha78A in Complex with L-Rhamnose-To explore the mechanism of enzyme action the structure of SaRha78A in complex with L-rhamnose, its reaction product, was determined. The bound L-rhamnose molecules were observed in two biologically significant regions of the enzyme (Fig. 1A); in the active site of the catalytic module SaRha78 CM (Rha 1501 ) and bound to SaCBM67 (Rha 1511 ). The average B-factors of the cyclic six atoms of the L-rhamnose rings were 20.9 and 17.9 Å 2 for Rha 1501 and Rha 1511 , respectively. The overall structure of the SaRha78A⅐L-rhamnose complex was almost identical to that of ligand-free SaRha78A with a root mean square deviation of 0.21 Å, implying little effect of ligand binding upon the overall structure.
Catalytic Module-Rha 1501 , located in the active site of SaRha78 CM , adopted an intermediate structure between skewboat 5 S 1 and boat 2,5 B conformations (Figs. 1 and 3, A and B), and the bound L-rhamnose was observed in mainly an ␣-anomeric configuration. The bound L-rhamnose was sandwiched between two aromatic residues, Trp 747 and Trp 640 , which thus make extensive hydrophobic interactions with the pyranose ring of the sugar. The pocket topology is completed by additional hydrophobic residues, Trp 695 , Tyr 744 , and Phe 851 . The C-6 methyl group, which comprises the signature feature of L-rhamnose, distinguishing the sugar from L-mannose, was buried in a hydrophobic hollow comprising Trp 695 , Tyr 744 , and Trp 747 . The likely polar contacts between Rha 1501 and the protein are as follows: O1 is within hydrogen bonding distance with Glu 636 O⑀2, and makes water-mediated polar contacts with Glu 636 O⑀1 and Arg 543 N2. O2 hydrogen bonded with Asp 630 O␦2, His 916 N⑀2, and Arg 634 N2, where the closest contact was with Asp 630 (2.4 Å). O3 makes polar contacts with Asp 643 O␦2 and His 916 N⑀2 and, through water-mediated interactions, with Trp 747 N⑀1 and Glu 895 O⑀2. This water molecule was situated in close proximity to the C1 atom of Rha 1501 with a distance of 2.9 Å. O4 hydrogen bonded with Asp 643 O␦1 and Trp 695 N⑀1. Thus, Rha 1501 made eight direct and four solvent-mediated hydrogen bonds with SaRha78A, and all these amino acids are conserved in BsRhaB. The O3 and O4 atoms of Rha 1501 were located at the bottom of the pocket, and the O1 atom was at the entrance and was solvent exposed, explaining why SaRha78A is an exo-acting enzyme that removes L-rhamnose residues from the non-reducing end of oligosaccharide or polysaccharide substrates.
SaCBM67-SaCBM67 displays a ␤-jellyroll fold in which the convex (␤-sheet 1) and concave (␤-sheet 2) ␤-sheets comprise six and five anti-parallel strands, respectively. The order of the ␤-strands is as follows: ␤-sheet 1: ␤-6, ␤-3, ␤-8, ␤-9/␤-1 (forms an interrupted ␤-strand), ␤-11; ␤-sheet 2: ␤-10, ␤-2, ␤-7, ␤-4, ␤-5 (Fig. 1B). Although this fold is typical of the vast majority of CBMs (for review see Ref. 22), unusually, the two sheets are not fully solvent exposed; ␤-sheet 2 is partially occluded by a loop extending from Thr 260 to Lys 282 , whereas the two loops that link SaCBM67 with the rest of the enzyme lie over ␤-sheet 1. Rha 1511 was positioned in a blind canyon interacting primarily with loops connecting ␤-strands ␤-3 and ␤-4, ␤-5 and ␤-6, ␤-7 and ␤-8. A central feature of the L-rhamnose binding site is a calcium ion that makes coordinate bonds with O3 and O4 of the sugar, whereas the metal interacts with the protein through Asp 179 O␦1, Asn 180 O␦1, Asn 228 O␦1, Pro 233 main chain O, and a water-mediated contact with Ser 230 main chain O (Fig. 3, C  and D). The bound L-rhamnose also makes direct hydrogen bonds with Trp 203 N⑀1, Asn 180 N␦2, and Asp 179 O␦2 through O2, O3, and O4 atoms, respectively. It is noteworthy that the peptide linkage between two calcium coordinating residues Asp 179 O␦1 and Asn 180 O␦1 was in cis-configuration. Two water molecules also mediated interactions between L-rhamnose and the protein. The C-6 methyl group pointed toward a small hydrophobic pocket comprising Trp 203 , Pro 233 , Pro 291 , and Trp 292 . The bound L-rhamnose adopted a relaxed 1 C 4 chair conformation with the O1 atom in the ␣-anomeric configuration. Although O2, O3, and O4 pointed at the protein surface, O1 was solvent exposed, indicating that SaCBM67 binds to L-rhamnose residues at the non-reducing termini of complex carbohydrates.
Catalytic Residues of SaRha78A and the Specificity and Function of SaCBM67-To verify the proposed identity of the catalytic residues of SaRha78A, aspartic acid and glutamine mutants of Glu 636 and Glu 895 were constructed. As expected, the activities of all mutants were at least 40 times lower than those of the wild-type enzyme ( Table 2). The k cat /K m values of E636D, E636Q, E895D, and E895Q mutants were 4.66, 11.3, 24.5, and 24.9 mM Ϫ1 s Ϫ1 , respectively. These data support a structural suggestion that Glu 636 and Glu 895 are involved in the enzyme catalysis.
To explore the biological function of SaCBM67, a recombinant form of the protein module was expressed in E. coli and, after purifying by immobilized metal ion affinity chromatography, its binding properties were explored using isothermal titration calorimetry. The results are summarized in Table 3. SaCBM67 bound L-rhamnose with a K a of 7.2 ϫ 10 3 M Ϫ1 and free energy of binding ⌬G of Ϫ5.3 kcal/mol. SaCBM67 did not bind to L-galactose or L-fucose, demonstrating that stereochemistry of the sugar at C4 and/or C2 are important specificity determinants. SaCBM67 bound to L-mannose, albeit with a 2-fold reduction in affinity. The C-6 methyl group of the bound L-rhamnose was partially buried in a hydrophobic pocket, however, there was sufficient solvation in the region of C-6 to accommodate a hydroxyl group. Because L-mannose seldom exists in natural polysaccharides, SaCBM67 binds primarily to L-rhamnose in biological systems.
SaCBM67 did not bind to L-rhamnose in the presence of EDTA, which chelates calcium, demonstrating that the metal ion plays a key role in ligand recognition. Similarly the D179A and N180A mutations, which remove calcium-mediated and direct hydrogen bonds with L-rhamnose, abrogate ligand binding, confirming the importance of calcium in the binding of SaCBM67 to its ligand. To summarize, these data are entirely consistent with the structural data in showing that domain D (hence defined as SaCBM67) in SaRha78A comprises a novel CBM with a biologically relevant specificity for L-rhamnose residues.
To investigate the role of SaCBM67 in the function of SaRha78A, the activities of the wild-type enzyme, in the presence of EDTA, and mutants D179A and N180A were explored. The data (Table 2) showed that inactivation of SaCBM67, through either calcium chelation with EDTA or mutation of key residues, did not influence activity against

TABLE 3 Binding parameters for the recognition of sugars by SaCBM67 and mutants
The concentration of SaCBM67 in the cell was 700 M for L-rhamnose and 800 M for L-mannose. The concentration of both ligands in the syringe was 20 mM.
aryl-rhamnosides, but caused a substantial reduction (ϳ50fold) in activity against the rhamnose-containing polysaccharide composite, gum arabic. These results demonstrate that SaCBM67 plays a central role in the action of SaRha78A against polysaccharides; the mechanism for this catalytic potentiation is discussed below.

DISCUSSION
In this study, we determined the crystal structure of SaRha78A, which was the third example of a GH78 protein. The structure in conjunction with biochemical studies show that the three proteins contain four highly conserved domains. Domain N of SaRha78A comprising a fibronectin type 3 fold is not evident in the other GH78 proteins, whereas the other five domains of the Sterptomyces enzyme are present in BsRhaB or in both BsRhaB and BT1001. Despite the structural similarities of the proteins, the total amino acid identity was less than 15% compared with the other two proteins. These differences in the domain arrangement in the three ␣-L-rhamnosidases points to unusual evolutionary pathways within GH78.
Catalytic Mechanism of GH78 Enzymes-Previous studies have established that GH78 enzymes hydrolyze glycosidic bonds through an acid base-assisted single displacement or inverting mechanism (23). Analysis of the crystal structure of the GH78 enzyme BsRhaB soaked with rhamnose revealed electron density in a deep pocket which, by analogy with related inverting (␣/␣) 6 -barrel glycoside hydrolases, is likely to comprise the active site. Although the electron density was sufficiently large to comprise a sugar, rhamnose, in its relaxed 1 C 4 conformation, could not be modeled into the observed density (7). In the structure of the SaRha78A⅐L-rhamnose complex, an L-rhamnose molecule (Rha 1501 ) was observed in the proposed active site pocket. The modeled L-rhamnose adopted an intermediate conformation between a skewboat 5 S 1 and boat 2,5 B. Previously, in some inverting glycoside hydrolases, including a Clostridium thermocellum GH8 endoglucanase (CtCel8A) that also displays a (␣/␣) 6 fold, the sugar bound at the active site (Ϫ1 subsite) adopted distorted 2 S O / 2,5 B ring conformations (24 -26). Further molecular dynamics studies indicated that the sugar ring in its 2 S O conformation represented the Michaelis complex, which would then adopt a 2,5 B conformation at the transition state, and, finally, the reaction product would display a 5 S 1 conformation (27,28). By analogy, Rha 1501 bound in the active site of SaRha78A appears to be migrating between the transition state and product conformations adopted during the reaction trajectory. Thus, it would appear that a boat 2,5 B is the conformation adopted by the transition state of glycans hydrolyzed by GH78 rhamnosidases. It is interesting to note that although Cui et al. (7) were unable to model an L-rhamnose in its relaxed confirmation into the active site electron density, the authors suggested that this may be because the bound sugar adopted a distorted conformation, as observed here.
Generally, two acidic amino acids, either aspartate or glutamate, are employed by most inverting GHs to catalyze hydrolysis (23,29). One carboxylate protonates the scissile glycosidic oxygen atom and the other coordinates the nucleophilic water molecule. In the SaRha78A⅐L-rhamnose complex, O1 of Rha 1501 was within hydrogen bonding distance with Glu 636 O⑀2 atom. The C-1 hydroxyl group was observed in an ␣-anomeric configuration, and the O1 atom position could be considered as the scissile bond position of the substrate. Therefore, Glu 636 appeared to comprise the catalytic proton donor (acid) of the enzyme. A water molecule makes a strong hydrogen bond with O1 of Rha 1501 , occupying the ␤-anomeric position, and thus is likely to comprise the solvent nucleophile utilized by the "inverting" enzyme. This water molecule makes strong hydrogen bonds with Trp 747 N⑀1 and Glu 895 O⑀2, and thus Glu 895 appears to be the catalytic general base. These catalytic residues and nucleophilic water are conserved in CtCel8A (24) (Fig. 4). Thus, based on the criteria of conservation of fold, catalytic apparatus, and catalytic mechanism, we propose that GH78 be included in clan GH-M that currently contains families GH8 and GH48.
Consistent with the proposed catalytic role of Glu 895 , the glutamate is conserved in the other two GH78 enzymes for which structures are available. Although, Glu 636 , the catalytic acid, corresponds to Glu 572 in BsRhaB, in BT1001 the equivalent residue is Asp 337 . When Glu 572 and Glu 841 in BsRhaB (corresponding to Glu 636 and Glu 895 in SaRha78A) were mutated to glutamine, catalytic efficiency decreased by 4 -5 orders of magnitude, supporting their proposed catalytic function. Furthermore, the observation that BT1001 is catalytically inactive (data not shown) is entirely consistent with the replacement of Glu 636 in SaRha78A (catalytic acid) with an aspartate, which would be too distant from the nucleophilic water to activate the solvent molecule. Reflecting a pH optimum of 6.0 the catalytic acid of SaRha78A requires a pK a modulator which, typically, is a carboxylate residue. The only candidate pK a modulator in SaRha78A is Asp 630 , which is within hydrogen bonding distance with Glu 636 . Consistent with the role of this aspartate is the observation that mutation of the equivalent residue in BsRhaB, Asp 567 , causes a 10 5 -fold reduction in catalytic activity (7).
L-Rhamnose Binding SaCBM67-A DALI search indicated that BsRhaB (PDB code 2OKX) (7), but not BT1001, contains a  (24). Two catalytic residues, the nucleophilic waters in the inverting reaction mechanism, bound sugars at subsite Ϫ1 are shown in stick models.
domain that is structurally equivalent to SaCBM67 with a Z-score of 20.6. SaCBM67 also displays weak structural homology with a CBM32 (Z-score of 5.0, PDB code 2VCA) (30), CBM60 (Z-score of 4.4, PDB 2XFE) (31), CBM36 (Z-score of 4.1, PDB 2DCJ) (32), and CBM35 (Z-score of 3.9, PDB 2W87) (33), which recognize their different ligands through either an exo-(CBM32 and CBM35) or endo-mode (CBM36 and CBM60) of binding. These CBMs all comprise a ␤-jellyroll structure and, in addition to a structural calcium (which is absent in CBM67), contain a second calcium atom that is integral to ligand recognition. In SaCBM67 calcium also plays a key role in L-rhamnose recognition, although the metal binding site is displaced compared with the other CBMs (Fig. 5).
CBM67 members are distributed not only in many bacterial GH78 ␣-L-rhamnosidases, but also in some Basidiomycete lectins, family 1 pectate lyases, peptidases, and proteins of unknown functions. Protein alignment of candidate members of CBM67 (Fig. 6) identified five subfamilies within the constructed phylogenetic tree (supplemental Fig. S1). In addition to SaCBM67, the only member of CBM67 for which a function is known is the lectin from Pleurotus cornucopiae. The lectin, which shows highest affinity against N-acetyl-D-galactosamine   (34,35), contains two CBM67-like sequences in tandem with sequence identities against SaRhaCBM67 of 25 and 35% for the N-and C-terminal modules, respectively. Hemagglutinating activity of this lectin is inhibited by EDTA and restored by calcium chloride, indicating that calcium binding is necessary for sugar binding. The calcium coordinating residues Asp 179 and Asn 228 of SaCBM67 are conserved in the two lectin modules, whereas Asn 180 , which is conserved in the C-terminal module, is replaced by an aspartate in the N-terminal module. Thus, the calcium binding site is conserved in the CBM67 ligand recognition site in the lectin, a putative polysaccharide lyase family 1 pectate lyase from Sorangium cellulosum (GenBank TM accession number CAN95071) and ␣-L-rhamnosidase from Paenibacillus sp. Y412MC10 (ACX62649), and both modules in a putative peptidase S8 and S53 from Pseudoalteromonas atlantica T6c (ABG39167). Although the L-rhamnose binding residues in SaCBM67 are conserved in CBM67 from the Paenibacillus ␣-L-rhamnosidase, suggesting that both proteins bind to L-rhamnose, these amino acids are not retaining in other CBM67 members. In particular, the loop extending from Glu 197 -Gly 202 in SaCBM67, which contains Asp 199 that makes a water-mediated interaction with the bound L-rhamnose, is missing in the Basidiomycete lectin, pectate lyase, and peptidase, which might cause different sugar specificity. Indeed, whereas SaCBM67 binds to L-rhamnose, the protein module does not recognize GlcNAc (data not shown), the ligand recognized by the P. cornucopiae lectin.
Prior to the discovery of the CBM67 family, the only other, nonenzymatic, L-rhamnose-binding proteins were animal lectins for which NMR and crystal structures are available (36,37). These proteins display an ␣/␤-fold and ligand binding is not metal dependent. Thus, the mechanism of ligand recognition in SaCBM67 is distinct from other proteins that bind to the hexose sugar.
SaRha78A is an example of an enzyme that displays an exomode of action in which both the active site of the catalytic module, and the appended CBM, bind to the same terminal sugar. Indeed, the ϳ50-fold, increase in catalytic activity afforded by the CBM is substantially greater than the 2-4-fold enhancement mediated by typical endo-binding CBMs (for review see Refs. 2, 38, and 39). Gum arabic is a highly complex mixture of polysaccharides and glycoproteins, in which the predominant structure is an arabionogalactan in which the ␤-1,3galactan backbone contains branches that are capped with L-rhamnose or L-arabinofuranose residues (40). This presents a structure in which both the GH78 catalytic module and CBM67 of SaRha78 can bind to different terminal L-rhamnose residues of the same polysaccharide molecule. The ensuing avidity effect will result in much tighter binding of SaRha78 to gum arabic, compared with either the CBM or the catalytic module as discrete entities, leading to the observed, CBM-mediated, enhanced catalytic efficiency. This emerging model for how exoacting CBMs potentiate the activity of exo-acting glycosidases is supported by a recent study showing that a CBM, which binds to terminal fructose residues, mediates a substantial increase in the activity of an exo-acting ␤-fructosidase against levan, a highly branched fructose-containing polysaccharide (41).