Probing the Mechanism of Ligand Recognition in Family 29 Carbohydrate-binding Modules*

The recycling of photosynthetically fixed carbon, by the action of microbial plant cell wall hydrolases, is integral to one of the major geochemical cycles and is of considerable industrial importance. Non-catalytic carbohydrate-binding modules (CBMs) play a key role in this degradative process by targeting hydrolytic enzymes to their cognate substrate within the complex milieu of polysaccharides that comprise the plant cell wall. Family 29 CBMs have, thus far, only been found in an extracellular multienzyme plant cell wall-degrading complex from the anaerobic fungus Piromyces equi, where they exist as a CBM29-1:CBM29-2 tandem. Here we present both the structure of the CBM29-1 partner, at 1.5 Å resolution, and examine the importance of hydrophobic stacking interactions as well as direct and solvent-mediated hydrogen bonds in the binding of CBM29-2 to different polysaccharides. CBM29 domains display unusual binding properties, exhibiting specificity for both β-manno- and β-gluco-configured ligands such as mannan, cellulose, and glucomannan. Mutagenesis reveals that “stacking” of tryptophan residues in the n and n+2 subsites plays a critical role in ligand binding, whereas the loss of tyrosine-mediated stacking in the n+4 subsite reduces, but does not abrogate, polysaccharide recognition. Direct hydrogen bonds to ligand, such as those provided by Arg-112 and Glu-78, play a pivotal role in the interaction with both mannan and cellulose, whereas removal of water-mediated interactions has comparatively little effect on carbohydrate binding. The interactions of CBM29-2 with the O2 of glucose or mannose contribute little to binding affinity, explaining why this CBM displays dual gluco/manno specificity.

however, binding is enthalpically driven, and mutagenesis studies suggest that direct hydrogen bonds play an important role in saccharide binding (for example, Refs. 16, 27, and 30). Although, in some cases indirect (water-mediated) hydrogen bonds appear to play a role in the binding of lectins to their ligands (31)(32)(33), the importance of these solvent-mediated interactions in polysaccharide recognition by CBMs is unclear.
Family 29 contains just two members, termed CBM29-1 and CBM29-2, which are components of the Piromyces equi noncatalytic protein NCP1 (34). NCP1 is present in the large, highly efficient, extracellular multienzyme plant cell wall-degrading complex produced by this anaerobic fungus. Both of the CBM29 modules bind to a range of ␤-1,4-linked polysaccharides that include cellulose, xylan, mannan, and glucomannan, although CBM29-1 displays lower affinity for these ligands than CBM29-2. Such flexible ligand recognition targets the anaerobic fungal complex to a range of different components of the plant cell wall and thus plays a pivotal role in the highly efficient degradation of this composite structure by the microbial eukaryote. We described, previously, the crystal structure of CBM29-2 in complex with cellohexaose and mannohexaose (15). Although these structures revealed an open ligand binding cleft providing six sugar binding subsites, they left unanswered questions concerning the structural basis for the difference in affinity between the two family 29 modules and the functional importance of direct and indirect interactions between CBM29-2 and its ligands, particularly with respect to the importance of O2 (which is axial in mannose and equatorial in glucose) as a specificity determinant.
Here we report the crystal structure of CBM29-1 and, by comparison with CBM29-2, provide insight into the structural basis for the different affinities displayed by the two family 29 modules. A mutagenesis approach to interrogate the functional importance of direct and indirect interactions between CBM29-2 and its ligands shows that hydrophobic stacking between aromatic residues and sugar rings dominates the binding affinity, as seen for other Type A and Type B CBMs. Although direct hydrogen bonds between specific amino acids in CBM29-2 and the hydroxyl groups of the sugar ligands make a significant contribution to overall affinity, the interactions between the protein and either the axial or equatorial O2 of mannose and glucose, respectively, contribute little to the overall binding energy. In contrast with studies on some lectins, we show that water-mediated hydrogen bonds contribute very little to ligand recognition by CBM29.

Generation of Mutants of CBM29-1 and CBM29-2
Derivatives of the two CBMs were generated by the QuikChange site-directed mutagenesis kit (Stratagene) according to the manufacturer's instructions, using pVM1 and pVM2 as template DNA. The primers employed in the mutagenesis PCRs are shown in Table 1S. The complete sequences of the DNA encoding the CBM29 mutants were determined by MWG-Biotech (Ebersberg, Germany) using T7 forward and reverse primers to confirm that only the desired mutations had been introduced.
Expression and Purification of CBM29-1 and CBM29-2 CBM29-2 was expressed in E. coli BL21(DE3):pLysS as described previously (15,34); CBM29-1 was produced in E. coli Origami B:pLysS. Briefly, cells were cultured in 1 liter of Luria-Bertoni broth in 2-liter baffled flasks at 30°C and 180 rpm to an A 600 ϳ0.6. Recombinant protein expression was induced by the addition of isopropyl-␤-D-thiogalactopyranoside to a final concentration of 200 M and incubation at 16°C for 16 h. CBM29-2 was purified by immobilized metal ion affinity chromatography using Talon resin (Clontech), anion exchange, and size exclusion chromatography as described by Charnock et al. (15). CBM29-1 was purified using the same procedure as for CBM29-2 except that the protein was eluted from the Talon resin with 5 mM rather than 100 mM imidazole. The CBMs were all purified to electrophoretic homogeneity as judged by SDS-PAGE. The protein concentration was determined from the calculated molar extinction coefficient of CBM29-1 and CBM29-2 at 280 nm, which were 26,860 and 26,150 M Ϫ1 cm Ϫ1 , respectively, and reduced coefficients of 20,340 and 25,560 M Ϫ1 cm Ϫ1 were used for the CBM29-2 tryptophan and tyrosine mutants, respectively.

Ligand Binding Studies
Ligand binding was determined by both affinity gel electrophoresis (AGE) and isothermal titration calorimetry (ITC). AGE was performed as described previously (34) using Konjac high viscosity glucomannan (hvKGM), Carob galactomannan (CGM), and hydroxyethyl cellulose (HEC), which were purchased from Megazyme International (Bray, County Wicklow, Ireland). The polysaccharides were included in the gels at concentrations ranging from 0.125 to 10 mg/ml. K d values were calculated by determining the mobility of the CBM, in the presence and absence of ligand, relative to a nonbinding standard BSA (Sigma) as originally described by Takeo (35). ITC measurements were made at 25 and 10°C following standard procedures (36) using a Microcal Omega titration calorimeter. Proteins were dialyzed, extensively, against 50 mM sodium phosphate buffer, pH 7.0, and the ligand was dissolved in the same buffer to minimize heats of dilution. During a titration experiment the protein sample (100 -800 M), stirred at 300 rpm in a 1.4331-ml reaction cell maintained at 25 or 10°C, was injected with 25 successive 10-l aliquots of ligand comprising 20 mg/ml low viscosity KGM (lvKGM) at 200-s intervals. The molar concentration of CBM29-2 binding sites present in the glucomannan was determined as described previously (37). Integrated heat effects, after correction for heats of dilution, were analyzed by nonlinear regression using a single site binding model (Microcal Origin, version 5.0). The fitted data yield the association constant (K a ) and the enthalpy of binding (⌬H). Other thermodynamic parameters were calculated using the standard thermodynamic equation.

NMR Spectroscopy
To assess the structural integrity of the mutant proteins, one-dimensional NMR analysis was performed on immobilized metal ion affinity chromatography-purified wild type and mutant forms of CBM29-2. Protein samples were ϳ500 M in 10 mM sodium phosphate buffer, pH 6.5, containing 10% D 2 O. NMR spectra were recorded at 30°C on a Bruker DRX-500 spectrometer, and 1 H chemical shifts were referenced to an internal standard of 3-(trimethylsilyl)propionate-2,2,3,3-d 4 at 0.00 ppm. Data were processed by FELIX (Accelrys, Inc., San Diego).

Data Collection and Structure Resolution
CBM29-2 Mutants-Data were collected, to between 2.25 and 1.6 Å resolution, on beamline ID14-EH1 at the European Synchrotron Radiation Facility (ESRF) in Grenoble (France) and beamline PX9.6 at the Daresbury (UK) Synchrotron Radiation Source. Crystals all belong to the orthorhombic space group P4 3 2 1 2. Crystal data and final refinement statistics are reported in Table I. Data were processed and scaled using MOSFLM and SCALA from the CCP4 suite (39) or DENZO and SCALEPACK from the HKL suite (40). Structures were solved by molecular replacement using the program AMoRe (41) using the wild type CBM29-2 structure, in the absence of ligand (1GWK), as the search model. After molecular replacement, maximum likelihood-based refinement of the atomic positions and temperature factors were performed with REFMAC (42). This procedure was interspersed with manual correction using the X-AUTOFIT option within QUANTA (Accelrys, Inc.). Water molecules were first placed automatically with REFMAC/ ARP (42,43) and verified manually.
Data for CBM29-1 in complex with cellohexaose crystals were collected to 2.05 Å (outer shell 12.12-2.05 Å) on beamline ID14-EH2 of the ESRF and processed with the HKL suite. Data have R merge : 6% (49%), a multiplicity of 3.6 (2.6) observations/reflection, and mean I/I of 25 (3.1). The data are 99% (91%) complete. The crystal belongs to space group P2 1 , with cell dimensions a ϭ 40.1, b ϭ 60.1, c ϭ 62.9 Å, and ␤ ϭ 108.3°and has two similarly orientated molecules in the asymmetric unit related by a translation-only vector of 0.5, 0, 0.5. Molecular replacement in this form, using the three-dimensional structure of CBM29-2 (1GWK) as the search model and the program AMoRe (41) DUPS with default settings and data to 3.0 Å resolution, was successful.
After molecular replacement, maximum likelihood-based refinement of the atomic positions and temperature factors was performed with REFMAC (42). At this point the partially refined model was used for molecular replacement of the original uncomplexed form (above) whose data were superior and the CBM29-1 structure refined with REFMAC with manual correction using the X-AUTOFIT option within QUANTA (Accelrys, Inc.). Water molecules were first placed automatically with ARP/REFMAC (42,43) and verified manually.
After the refinement of the native CBM29-1 to 1.5 Å resolution, this model was then used to complete the refinement of CBM29-1 with cellohexaose (PDB accession number 1WCU). Cellohexaose density is visible but is extremely disordered (see "Results" and "Discussion").

FIG. 1. Overlap of unliganded CBM29-1 (blue with yellow ball-andstick) with CBM29-2 mannohexaose complex (red with green ball-andstick).
Of particular interest are Tyr-45 of CBM29-1, which blocks the subsites 1 and 2, and the loop containing Thr-87, which squeezes the substrate binding cleft and could sterically hinder access. Subsites 1 and 6 are indicated for reference. This figure, in divergent (wall-eyed) stereo was drawn with BOBSCRIPT.

RESULTS
Crystal Structure of CBM29-1-The noncatalytic protein, NCP1, of the P. equi multienzyme complex contains two family 29 CBMs arranged in tandem (34). Although both CBMs display similar binding specificities, the C-terminal module, CBM29-2, exhibits higher affinity for the target ligands than the N-terminal CBM29-1 (34). To explore the structural basis for these differences in affinity the crystal structure of CBM29-1, in the presence and absence of cellohexaose, was solved to a resolution of 2.05 and 1.5 Å, respectively (see "Experimental Procedures"). The final refined coordinates have R cryst and R free values of 0.24/0.32 and 0.14/0.16 for "complexed" and native CBM29-1, respectively. The final CBM29-1 structures include residues 4 -143 (corresponding to residues 178 -317 of NCP1); however, despite co-crystallization with cellohexaose, electron density for the hexasaccharide ligand exists as a "smear" precluding the precise modeling of cellooligosaccharides in the binding cleft. With the exception of a movement of the side chain of Tyr-45 (invariant), the complexed and uncomplexed structures of CBM29-1 are extremely similar (Fig. 1).
The structures of CBM29-1 and CBM29-2 reveal significant similarities in the overall folds of the two modules (Fig. 1). Both structures present a classic ␤-jelly roll with five ␤-strands forming each of the two faces. In addition to the differences in the ligand binding surface (described below), there is a disulfide bridge in CBM29-1, between Cys-63 and Cys-141, whereas CBM29-2 does not contain disulfide bridges. This likely explains why CBM29-1 requires expression in E. coli Origami B strain because this allows the formation of disulfide bridges in the cytoplasm of the bacterium (45).
There are, however, some significant differences that may contribute to the lower ligand affinity displayed by CBM29-1. First, Tyr-45 of unliganded CBM29-1 (equating to Tyr-46 of CBM29-2) lies perpendicular to the orientation seen in native and complexed CBM29-2 and CBM29-1 in the presence of cellohexaose such that it blocks the ligand binding surface at subsite 2. Second, there are a number of loop changes, notably in loop 1 (Val-78 to Asn-91 in CBM29-1). In CBM29-2 this loop is both three amino acids shorter (Val-76 to Glu-86), and, more importantly, it lies in an open conformation (although the loop makes crystal packing interactions in three different crystal forms, its conformation is essentially identical, and thus it is likely to adopt this open conformation in solution). In CBM29-1 the loop is much more closed into the binding surface reducing the width at this point by ϳ4 Å. In CBM29-1 the binding cleft is not only narrowed but also presents Thr-87 in such a way that it could clash with ligand bound in subsite 4. The distortion of the aromatic platform occluding subsite 2 and the loop insertion compromising subsite 4 offer a possible explanation for the reduced affinity displayed by CBM29-1 compared to CBM29-2.
Alignment of the residues involved in ligand binding in CBM29-2 (15) with CBM29-1 reveals further minor differences between the two modules. Although CBM29-2 Trp-24, Trp-26, Tyr-46, Arg-112, and Gln-116 are conserved in CBM29-1 (Trp-23, Trp-25, Tyr-45, Arg-116, and Gln-120, respectively), Lys-74, Glu-78, Asp-114, and Ala-118 are replaced with Leu-75, Gln-79, Ile-118, and Gly-122, respectively, in the N-terminal CBM29-1. To explore the functional significance of the amino acid changes in the binding cleft of CBM29-1 and CBM29-2, Gln-79 was replaced with glutamate and Leu-75 with lysine to generate the CBM29-1 mutants Q79E, L75K, and Q79E/L75K, all of which make CBM29-1 more similar to CBM29-2. Gly-122 and Ala-118 were also exchanged to create CBM29-1 G122A and CBM29-2 A118G. Although AGE shows that L75K displays an affinity for hvKGM similar to wild type CBM29-1, both Q79E and Q79E/L75K did not exhibit any binding to the polysaccharide (data not shown). These data indicate that Leu-75 in CBM29-1 corresponding to Lys-74 in CBM29-2, does not make a significant contribution to ligand recognition. By contrast, Gln-79 appears to play a pivotal role in binding of CBM29-1 to glucomannan. The loss in binding mediated by the Q79E mutation suggests that the introduction of the negative charge causes a local perturbation in the binding cleft such that Glu-79 adopts a different position to Gln-79 and is thus unable to make productive interactions with the ligand. The Gly/Ala exchange had no effect on ligand binding in either CBM29-1 or CBM29-2 (data not shown), indicating that these amino acid substitutions have not altered the position of the cognate backbone carbonyl (the carbonyl backbone of Ala-118 interacts with ligand at subsite 1 and 2).
The structure of CBM29-1 thus points to several features that may explain its lower affinity for ligands compared with CBM29-2, including the need for ligand-induced conformational changes and the partial occlusion of the binding cleft by the extended loops in CBM29-1. To probe the role of the con- The upfield region of the one-dimensional spectra obtained from 200 M protein in 10 mM sodium phosphate buffer, pH 6.5, containing 10% D 2 O is shown. Spectra were recorded at 30°C on a Bruker DRX-500 spectrometer referenced to an internal standard of 3-(trimethylsilyl)-propionate-2,2,3,3-d4 at 0.00 ppm. served interactions of CBM29 modules with ligand, a systematic mutagenesis program targeting the direct and indirect hydrogen bonds and the aromatic interactions was initiated. CBM29-2 was chosen for this study because its higher affinity for ligands and crystal structures with both manno-and cello-oligosaccharides.
Contributions to Ligand Affinity in CBM29-2: Structural Integrity of Mutant Proteins-The original crystal structure of CBM29-2 in complex with cellohexaose and mannohexaose (15), revealed residues that are likely to be the key determinants of affinity and specificity (Fig. 2). An alanine scanning mutagenesis approach was therefore used to interrogate the functional significance of the amino acids that make direct, or indirect (water-mediated), interactions with the two ligands and also those conserved aromatics that form a hydrophobic platform in subsites the 2, 4, and 6.
The mutant proteins were purified to electrophoretic homogeneity. The structural integrity of the mutant proteins was initially investigated by NMR spectroscopy (Fig. 3) and in some cases by x-ray crystallography (below). The upfield ends of the different 1 H spectra (below 0.6 ppm) are derived from protons whose signals have been shifted further upfield by the ring currents from aromatic residues. Very minor changes in the positions or orientations of these aromatic rings cause significant changes to the chemical shifts of the ring current-shifted protons. The similarity of the spectra of wild type and mutant proteins therefore suggests that no such changes in ring positions have occurred, and thus the packing in the vicinity of the rings is unaffected by the mutations. Such ring current-shifted protons are generally found in the hydrophobic core and come from widely different parts of the protein sequence. The spectra therefore demonstrate that the hydrophobic core of the proteins is essentially intact and that the CBM29-2 mutants have all folded into a three-dimensional structure that is very similar to the wild type protein.
To assess the effect of these mutations further, the crystal structures of the CBM29-2 mutants Y46A, R112A, K85A, and D114A, were solved in the absence, and K74A and D83A in the presence, of bound cellohexaose. No diffracting crystals of the other mutant proteins were obtained. All of the mutant proteins displayed a gross structure identical to that of the wild type protein. In the Y46A, K74A, D83A, K85A, and R112A mutant proteins the location of residues believed to be important in ligand binding were not affected significantly by the cognate amino acid substitution. In R112A there is a slight change in the position of Lys-85, but this is not thought to be significant because in wild type CBM29-2 this residue adopts multiple conformations. One mutation that does have a significant secondary effect is D114A. This mutation causes a change in the position of Glu-78 (Fig. 4), such that the glutamate occupies the pocket left by the removal of the aspartate side chain of Asp-114 resulting in a ϳ1.5 Å shift in the orientation of the carboxylate because of an 11.6 o rotation around C ␤ .
Both NMR and x-ray crystallography indicate that, with the exception of D114A, all of the mutations of the conserved residues in the CBM29-2 ligand binding cleft have no adverse structural consequences. The biochemical properties of these mutant proteins can thus be investigated with confidence.
Biochemical Properties of CBM29-2 Mutants-The biochemical properties of the CBM29-2 mutants were initially assessed by AGE and ITC. Examples of typical AGE data and the plots used to determine relative affinity are shown in Figs. 5 and 6, respectively; the full data set is displayed in Table II. The binding of the CBM29-2 mutants Y46A, K74A, D83A, and K85A to lvKGM could be quantified accurately by ITC (Table  III and Fig. 7), whereas this method could only provide an estimate of the association constant for Q116A (ϳ3.0 ϫ 10 3 M Ϫ1 ) because of the low affinity displayed by this mutant protein. Both AGE and ITC revealed no binding of W24A, W26A, E78A, and R112A to the polysaccharide.
Binding Properties of the CBM29-2 Mutants of Asp-83, Lys-85, and Asp-114 -The importance of water-mediated interac-tions in the recognition of carbohydrates by CBMs is unclear. The crystal structure of CBM29-2 in complex with either mannohexaose or cellohexaose indicates that Asp-83, Lys-85, and Asp-114 each interacts with ligands via single ordered water molecules (Fig. 2). AGE showed that the D83A and D114A mutations cause a ϳ3and 6 -10-fold decrease in affinity, respectively (Table II). Although the crystal structure of D114A in complex with cellohexaose could not be determined, the structure of the apo form of this protein provides insight into its reduced affinity for the glucose polymer (compared with wild type CBM29-2). It seems likely that the shift in the position of Glu-78 (described above), which plays an important role in ligand binding, will have a dramatic effect on the affinity of the protein for its target saccharides. Although the reduction in affinity through the loss of solvent-mediated hydrogen bonds cannot be completely disregarded, the view that the weak binding displayed by D114A is caused by a shift in the position of Glu-78 is supported by the similar affinities exhibited by E78A and D114A for the polysaccharide ligands HEC, hvKGM, and CGM.
In contrast, the K85A mutant protein displayed a consistent increase in affinity for gluco-configured ligands (Figs. 5-7 and Tables II and III). In previous work we have shown that Lys-85 makes a steric clash with the O2 of glucoside 4, and as a result the side chain of the lysine is forced into multiple conformations beyond C␥ (15). The enhanced affinity displayed by K85A for cellulose-derived ligands presumably reflects the energetic gain resulting from the removal of this steric clash. Surprisingly, K85A also displays a modest increase in affinity for mannose-containing saccharides compared with wild-type CBM29-2 (Tables II and III). Without a crystal structure of K85A in complex with mannohexaose, however, the mechanism for the modest increase in affinity, compared with wild type CBM29-2 for manno-configured ligands, remains unclear. Binding Properties of Mutants of the Aromatic Residues of CBM29-2-Both W24A and W26A display no affinity for ␤-1,4linked gluco-or manno-configured ligands (Tables II and III). Y46A in contrast exhibits reduced, but measurable, affinity for all ligands. These data indicate that the two tryptophan residues play a more important role in ligand binding than Tyr-46. Trp-24, Trp-26, and Tyr-46 stack against the ␣-face of sugars located at subsites 2, 4, and 6, respectively, and clearly these hydrophobic interactions play a key role in ligand binding by CBM29-2. These data are entirely consistent with previous studies, which have also shown that aromatic residues that stack against sugar residues play a critical role in the binding of both Type A and Type B CBMs to their target ligands. (12, 13, 16, 22-28, 30, 46 -48). Indeed, the lesser contribution of tyrosine to these interactions was first noted in family CBM1 by Linder et al. (48), who proposed that the differences reflected the greater hydrophobicity and area of the indole rings of tryptophans compared with the phenolic rings of tyrosines.
Binding Properties of CBM29-2 Mutants of Lys-74, Glu-78, Arg-112, and Gln-116 -The side chains of each of these residues are within direct hydrogen bonding distance of both glucoand manno-configured ligands (Fig. 2). The E78A mutation results in a 10 -20-fold reduction in affinity for all of the polysaccharides tested, whereas Q116A displays an affinity that is 8 -15-fold lower than wild type CBM29-2 (Tables II and III). Glu-78 makes hydrogen bonds with the O3 and O4 of glucose or mannose at subsite 4, whereas Gln-116 interacts with the endocyclic oxygen and O6 of the sugar at subsite 3 and makes an H bond with O2 when pyranoside 3 is mannose. Substitution of Arg-112 with alanine has a very significant effect on ligand binding, reducing affinity for hvKGM by at least 80-fold, whereas binding to CGM and HEC is undetectable. Arg-112 makes direct hydrogen bonds to the O4 and O3 of the pyranoside at subsite 4 and the endocyclic oxygen of a glucoside in subsite 5 (and O2 in the case of mannoside 5). The more extensive interactions mediated by Arg-112 with the target ligands, compared with the other polar residues that make direct hydrogen bonds with the gluco-and manno-configured saccharides, may explain why the R112A mutation causes a more dramatic reduction in affinity compared with E78A or Q116A.
The K74A mutation causes only a modest ϳ2-fold reduction in affinity for all three polysaccharide ligands. Lys-74 interacts with O3 and O2 of mannoside 2 and both conformations of glucose 2 observed in the CBM29-2-cellohexaose complex (15); the basic amino acid interacts with O3 of the more abundant, and O3 and O2 of the less prominent conformer. The small reduction in affinity mediated by the K74A mutation is thus rather surprising because the number and distance of the hydrogen bonds between Lys-74 and the pyranoside at subsite 2 are similar to the corresponding interactions between Glu-78 and Gln-116 and the target ligands. The observation that glu- cose adopts two conformations at subsite 2, and substitution of Tyr-46 that stacks against the sugar ring at this location reduces, but does not completely abrogate, binding suggests that the interaction of the ligands at subsite 2 is generally weaker than at subsites 3, 4, 5, and 6. The other polar residues either interact with the ligands at subsites where aromatic resides are absent or make hydrogen bonds with O4, which is positioned at subsite interfaces. Thus, the interactions between ligand and the polar amino acids Glu-78, Gln-116, and Arg-112 are complementary to the hydrophobic stacking interactions between the sugar polymers and the surface aromatic residues of the protein.
One of the unusual features of CBM29-2 is that the protein makes direct hydrogen bonds with both the equatorial and axial O2 of glucosides and mannosides, respectively, at several subsites (15). The mutagenesis strategy employed here has facilitated the interrogation of the functional significance of these interactions. Lys-74 makes a hydrogen bond with the O2 of both mannoside 2 and glucoside 2 (15), in the complexes with mannohexaose and cellohexaose, respectively, and thus the modest decrease in affinity caused by the K74A mutation indicates that the interaction between the protein and the O2 of the sugar at subsite 2 makes a minor contribution to overall binding energy. The observation that the Q116A mutation results in the loss of a hydrogen bond to the O2 of mannoside 3 (but not O2 of glucoside 3) but causes a similar reduction in affinity for both HEC and CGM, indicates that the interaction between CBM29-2 and the O2 of mannoside 3 is not a key specificity determinant. Overall, these data indicate that the hydrogen bond between CBM29-2 and the O2 of either a mannoside or glucoside at subsite 2, and the O2 of the mannoside at subsite 3 does not make a significant contribution to binding energy. It is possible that Arg-112 may potentiate mannoside binding at subsite 5 (Arg-112 interacts with the O2 of mannoside 5 but not the O2 of glucoside 5), however, the critical interactions made by the guanidino group of this amino acid with other regions of the ligand, as evidenced by the weak binding of R112A to all ligands, precludes analysis of the functional significance of the O2-NH1 hydrogen bond. DISCUSSION The data presented in this study are consistent with previous reports demonstrating the importance of polar residues in the binding of Type B CBMs to gluco-and xylo-configured ligands. Thus, removal of amino acids with carboxylic, amide, or guanidino side chains mediates decreases in affinity that range from ϳ10to 100-fold or more (12,16,27,30). Mutations that cause the largest decreases in affinity normally involve the substitution of amino acids that are buried at the bottom of the binding cleft. For example, the E138A mutant of CBM22-2 from Clostridium thermocellum Xyn10B displays no binding to xylan (27), whereas the introduction of the N120A mutation into CBM6 from C. thermocellum Xyn11A reduced affinity for the xylose polymer 145-fold (16). The buried location of Asn-120 and Glu-138 in the two CBMs would prevent the H bonds, between these residues and the respective ligands, from exchanging with water molecules, and thus the interaction between these amino acids and sugar polymer would likely be tighter than those bonds that are more solvent-exposed. Similarly, in CBM29-2, Arg-112 is more buried than the other polar residues in the binding site, whereas Lys-74 is solvent-exposed, providing a possible explanation for the influence of these amino acids in ligand binding.
In contrast to Type B CBMs, where ligand binding is enthalpically driven, mutagenesis studies on Type A CBMs where a favorable change in entropy dominates binding have shown that polar residues do not make a significant contribution to overall affinity (23,25). At the cusp of Type A and Type B CBMs the modules display very shallow ligand binding sites, and the role of polar residues is variable. In the highly exposed binding site of CBM2b-1 from Cellulomonas fimi the three polar residues Glu-257, Asn-292, and Gln-288 are predicted to make hydrogen bonds with the xylan ligand (28). The triple mutant protein E257A/N292A/Q288A, however, displays only a 2-fold reduction in affinity, although binding is associated with a significant increase in entropy compared with the wild type module, which may reflect more conformational freedom of the bound ligand (28). By contrast, in CBM17, which also contains a very shallow binding site, direct hydrogen bonds with ligand play a significant role in cellulose binding (17).
The mutant protein data presented above also indicate that the solvent-mediated interactions between CBM29-2 and its ligands contribute comparatively little to overall affinity. This is somewhat surprising because the conservation of the position of ordered water molecules in complexes of model lectins FIG. 7. Sample ITC data for the binding of CBM29-2 wild type and mutants to lvKGM. All titrations were carried out at 25°C in 50 mM sodium phosphate, pH 7.5.
with different saccharides has led to the assumption that these indirect interactions are extremely important in ligand binding (32,44). For example, the only apparent difference in the interactions between peanut lectin with lactose and T-antigen, which could explain the 20-fold increase in affinity for the latter sugar, is the presence of two water molecules that make hydrogen bonds with the protein and the N-acetyl group of the T-antigen (31,33).
The view that indirect hydrogen bonds do not play a pivotal role in the binding of CBMs to their target ligands is supported further by the observation that the removal of four polar residues, which make water-mediated interactions with xylan and xylooligosaccharides in CBM15 (a Type B CBM), likewise do not affect affinity greatly (46). It is, however, important to emphasize that solvent-mediated hydrogen bonds between backbone amide and carbonyl groups and the target ligand may contribute more to binding energy than the corresponding interactions with amino acid side chains. Indeed the conformational restriction of these groups in the absence of ligand may well reduce the entropic penalty incurred when they participate in indirect interactions with ligand.
In this report we show that direct hydrogen bonds and hydrophobic stacking interactions play a key role in the binding of family 29 CBMs to both gluco-and manno-configured ligands. These data provide further support for the emerging view that the primary distinguishing feature between Type A and Type B CBMs is the importance of direct hydrogen bonds in polysaccharide binding. In Type A modules polysaccharide recognition is mediated by hydrophobic interactions with hydrogen bonds contributing little to ligand binding, whereas in Type B modules both hydrophobic interactions with aromatic rings and direct hydrogen bonding play critical roles in saccharide recognition. In addition, for the Type B proteins CBM29-2 and CBM15 (46) at least, solvent-mediated hydrogen bonds play no significant role in polysaccharide binding. Interestingly the O2 of gluco-and manno-configured sugar polymers is not a critical specificity determinant, which provides an explanation for the observed plasticity in ligand recognition displayed by the family 29 CBM.