Recognition and hydrolysis of noncrystalline cellulose.

Cellulase Cel5A from alkalophilic Bacillus sp. 1139 contains a family 17 carbohydrate-binding module (BspCBM17) and a family 28 CBM (BspCBM28) in tandem. The two modules have significantly similar amino acid sequences, but amino acid residues essential for binding are not conserved. BspCBM28 was obtained as a discrete polypeptide by engineering the cel5A gene. BspCBM17 could not be obtained as a discrete polypeptide, so a family 17 CBM from endoglucanase Cel5A of Clostridium cellulovorans, CcCBM17, was used to compare the binding characteristics of the two families of CBM. Both CcCBM17 and BspCBM28 recognized two classes of binding sites on amorphous cellulose: a high affinity site (K(a) approximately 1 x 10(6) M(-1)) and a low affinity site (K(a) approximately 2 x 10(4) M(-1)). They did not compete for binding to the high affinity sites, suggesting that they bound at different sites on the cellulose. A polypeptide, BspCBM17/CBM28, comprising the tandem CBMs from Cel5A, bound to amorphous cellulose with a significantly higher affinity than the sum of the affinities of CcCBM17 and BspCBM28, indicating cooperativity between the linked CBMs. Cel5A mutants were constructed that were defective in one or both of the CBMs. The mutants differed from the wild-type enzyme in the amounts and sizes of the soluble products produced from amorphous cellulose. This suggests that either the CBMs can modify the action of the catalytic module of Cel5A or that they target the enzyme to areas of the cellulose that differ in susceptibility to hydrolysis.

Many enzymes that hydrolyze insoluble polysaccharides contain discrete carbohydrate-binding modules (CBMs) 1 that can increase the effective enzyme concentration on the polysaccharide surface (1,2). CBMs are similar to lectins in that they can discriminate between different polysaccharides, and exhibit binding affinities in the micromolar range. Like their cognate catalytic modules, CBMs are classified into families of related amino acid sequences (afmb.cnrs-mrs. fr./ϳcazy/CAZY/index.html). A particularly interesting group of CBMs are those that recognize cellulose, which is arguably the most important polysaccharide on the planet because of its source and abundance, it is the main structural polysaccharide in plant cell walls.
Although the cellulose molecule is a simple polymer of up to 10,000 glucose residues linked by ␤-1,4-glucosidic bonds, natural cellulose is relatively resistant to enzymatic hydrolysis. Chains longer than six glucose residues are insoluble. In natural cellulose many molecules associate in parallel to form fibrils, which in turn associate to form fibers. Cellulose is heterogeneous: current biophysical techniques used to study the structure of cellulose show regions of highly ordered cellulose chains (crystalline cellulose), pseudo-ordered chains (paracrystalline cellulose), and disordered chains (so called "amorphous" or noncrystalline cellulose) (3). It is thought that the varied structure of cellulose contributes to its recalcitrance to complete hydrolysis by individual enzymes; the efficient degradation of cellulose is only achieved by the concerted action of complex microbial enzyme systems. It is known that some CBMs bind preferentially to the crystalline regions, others to the amorphous regions, and that the two types do not compete when binding (4). Those binding amorphous but not those binding crystalline cellulose can also bind soluble cellooligosaccharides. Surprisingly, some CBMs known to be specific for noncrystalline cellulose do not compete when binding, suggesting that they are discriminating between different regions of the amorphous cellulose and making it evident that the interaction of CBMs with noncrystalline cellulose may be more complex than originally thought (5). It is significant in this context that the three-dimensional structure of amorphous cellulose is still unclear.
Cellulase Cel5A from Bacillus sp. 1139, like several other enzymes in this family, contains a family 5 catalytic module and two CBMs, one from family 17 the other from family 28, both of which bind to amorphous cellulose. The combination of similarity in amino acid sequence and secondary structure analyses clearly established the structural relationship between family 17 and family 28 CBMs (8). However, the amino acid residues involved in binding are poorly conserved between these two families (8) suggesting that the CBMs might have different binding characteristics that could shed further light on the heterogeneity of noncrystalline cellulose. This paper shows that family 17 and family 28 CBMs do not compete for sites on noncrystalline cellulose, emphasizing the presence of cellulose chains with physical presentations that are distinguishable by the CBMs. Furthermore, the discrimination of these cellulose regions by the CBMs in Cel5A appears to influence cellulose hydrolysis by this enzyme. TM PH101; FMC International, Little Island, County Cork, Ireland), Oregon Green TM succinimidyl ester (Molecular Probes, Eugene, OR), cellohexaose (Seikagaku, Tokyo, Japan), and 2Ј,4Ј-dinitrophenyl-␤-D-cellobioside (Sigma) were obtained from the sources indicated. Regenerated cellulose (RC) was prepared by phosphoric acid treatment of Avicel TM as reported previously (6). Bacterial microcrystalline cellulose (crystalline cellulose I) was prepared from cultures of Acetobacter xylinum (ATCC 23769) as described previously (7).

Chemicals-Microcrystalline cellulose (Avicel
Purification of CBMs-Gene fragments encoding CBMs were cloned, expressed, and the products purified as described elsewhere (8,9). The CBMs used in this study were CcCBM17, the C-terminal module from Clostridium cellulovorans Cel5A (10); BspCBM28, the C-terminal module from Bacillus sp. 1139 Cel5A; and BspCBM17/CBM28, the tandem of C-terminal CBMs from Bacillus sp. 1139 Cel5A. The concentration of purified proteins was determined by UV absorbance (280 nm) using calculated molar extinction coefficients (11)  Binding Studies-Isothermal titration calorimetry was performed essentially as described previously (6) using a MCS isothermal titration calorimetry (MicroCal Inc., Northampton, MA). BspCBM17/CBM27 (wild-type or W453/W500 mutant) was at pH 7.0 in 50 mM K-phosphate buffer. Titrations were performed by injecting 10-l samples of cellohexaose solution (5 mM in K-phosphate buffer saved from the dialysis of BspCBM17/CBM27) isothermal titration calorimetry sample cell (volume ϭ 1.3528 ml) containing 275 M BspCBM17/CBM27. Heats of dilution of the titrant were assessed by titration of the carbohydrates into buffer lacking CBM; these were found to be negligible. Two or three independent titration experiments were performed. Stoichiometries, enthalpies, and equilibrium association constants were determined by fitting the corrected data to a bimolecular interaction model. Adsorption isotherms using RC or Avicel TM were obtained as described previously (4,12). Data were analyzed by nonlinear regression of the data to a Langmuir-type two binding site model (Equation 1), v ͓F͔ .
This model does not account for a fraction of inaccessible lattice units, so v was calculated using estimates of the functional lattice unit concentration from the fits of Equation 2. Goodness-of-fit of the models to the data was assessed by runs tests of the residuals and p value analysis. CBMs were labeled with Oregon Green TM 514 carboxylic acid succinimidyl ester, which labels the amine groups of lysines and the N terminus. This was done according to the instructions from the sup-plier. Fluorescence isotherms were obtained as described previously (4,5). Competition displacement assays were done as follows. A constant concentration of RC or Avicel TM (1 or 10 mg/ml, respectively) was mixed with a constant concentration of Oregon Green TM -labeled CBM. The concentration of Oregon Green TM -CBM chosen was sufficient to nearly or completely saturate the sites bound with high affinity. Unbound Oregon Green TM -CBM at equilibrium was quantified by fluorescence spectroscopy in the presence of increasing concentrations of an unlabeled competing CBM and reference to a standard curve relating Oregon Green TM -CBM concentration to fluorescence intensity: bound Oregon Green TM -CBM ϭ total Oregon Green TM -CBM Ϫ unbound Oregon Green TM -CBM. Displacement and, therefore, binding competition was determined graphically by plotting the concentration of bound Oregon Green TM -CBM versus the concentration of competing CBM.
Cloning, Production, and Purification of Cel5A and Mutants-Escherichia coli BL21(DE3) was used for cloning and protein purification. The plasmid used was pET28a (Novagen). E. coli cultures were grown routinely in TYP medium (16) containing 50 g of kanamycin ml Ϫ1 , at 37°C for plasmid preparation or 30°C for protein production.
Plasmid preparation, bacterial transformation, and agarose gel electrophoresis were described previously (16). Restriction endonucleases were used as recommended by the suppliers. DNA fragments were purified from the agarose gels with the Qiaex II kit (Qiagen). Oligodeoxynucleotide primers were synthesized by the Nucleic Acid and Protein Service (NAPS) Unit of the University of British Columbia, using an Applied Biosystems model 380A DNA synthesizer, and purified by extraction with 1-butanol. DNA was sequenced by the NAPS using the AmpliTaq dye termination cycle sequencing protocol and an Applied Biosystems model 377 sequencer.
The DNA fragment corresponding to nucleotides 91-2287 (encoding amino acids 31-762) of the cel5A gene was amplified from Bacillus sp. 1139 genomic DNA (prepared as described previously (8)) by PCR. This was inserted via introduced 5Ј and 3Ј NheI and HindIII restriction endonuclease sites, respectively, into appropriately digested pET28a (Novagen, Madison, WI). This construct was verified by DNA sequencing. The remaining Cel5A mutants (see text) were constructed using standard gene cloning techniques.
E. coli BL21(DE3) containing the pET28aCel5A constructs were grown in 2.5 liters of TYP medium containing 50 g of kanamycin ml Ϫ1 at 30°C. The cultures were induced with 0.1 mM isopropyl-1-thio-␤-Dgalactopyranoside, either overnight at low optical density (0.1 A 600 ) or 4 h at high optical density (0.6 A 600 ). The cells were harvested and resuspended in 30 ml of 20 mM Tris-HCl buffer, pH 7.9, 0.5 M NaCl, and ruptured by two passages through a French pressure cell. Debris in the cell extracts was removed by centrifugation for 30 min at 27,000 ϫ g at 4°C. Proteins were purified from the clarified cell extract by immobilized metal affinity chromatography according to the manufacturer's protocols (Novagen, Madison, WI). The concentrations of purified proteins were determined by UV absorbance (280 nm) and calculated molar extinction coefficients (11) of 145,800 M Ϫ1 cm Ϫ1 , 134,420 M Ϫ1 cm Ϫ1 , 102,740 M Ϫ1 cm Ϫ1 , and 91,360 M Ϫ1 cm Ϫ1 for Cel5A, Cel5A17 (Ϫ) , Cel5A⌬28, and Cel5A17 (Ϫ) ⌬28, respectively.
Cellulose Hydrolysis-The activity of Cel5A and its mutants on RC was assayed at pH 7 and 37°C by measuring the liberated reducing sugars with the hydroxybenzoic acid hydrazide reagent (17). The reaction mixture contained 20 nmol of enzyme and 20 mg of RC in 20 ml of 50 mM sodium citrate buffer, pH 7.0, with 0.02% sodium azide (i.e. 1 M enzyme and 1 mg ml Ϫ1 RC). Duplicate samples were incubated at 37°C for 96 h with slow end-over-end mixing on a tube roller. Samples of 0.3 ml were removed at intervals, then centrifuged twice for 5 min at 10,000 ϫ g. Samples of 10 to 100 l of the supernatants were assayed for reducing sugars (17). Reducing sugar concentrations were obtained by reference to a standard curve prepared with glucose.
Heat-inactivated samples were analyzed by analytical high performance liquid chromatography using a CarboPac PA-100 column. Samples were loaded onto the column in 10 mM citrate buffer and eluted with a gradient of 0 -100 mM sodium acetate in 100 mM sodium hydroxide. Eluted carbohydrates were detected by pulsed amperometry. Sample peaks were identified and quantified by comparison with the elution profiles of cellooligosaccharide standards.

RESULTS
Adsorption of CcCBM17 and BspCBM28 to Cellulose-Neither CcCBM17 nor BspCBM28 bound to highly crystalline cellulose I (bacterial microcrystalline cellulose) or highly crystalline cellulose II (mercerized cotton fibers) (not shown) but both bound to RC and Avicel TM (Fig. 1, a and b). The adsorption isotherms were inconsistent with a single binding site Langmuir isotherm. Scatchard plots were concave confirming the complex nature of the binding (Fig. 1, c-f). Descending isotherms showed the binding of CcCBM17 and BspCBM28 to be completely reversible (not shown).
CcCBM17 and BspCBM28 bind to soluble cellooligosaccharides (8,9). Therefore, they probably bind to individual chains in noncrystalline regions of insoluble cellulose but not to crystalline cellulose I and II. They have up to five glucose-binding subsites (8,9) so the molecules to which CcCBM17 and BspCBM28 bind can be considered as a linear array of potentially overlapping binding sites. This could result in nonlinear Scatchard plots, as could the presence of multiple, independent classes of binding sites with differing affinities. Therefore, the binding isotherms for CcCBM17 and BspCBM28 were evaluated by nonlinear regression against binding models for noncooperative binding to two or more classes of binding sites (Equation 1), noncooperative binding with steric exclusion of overlapping binding sites (Equation 2), and cooperative binding with steric exclusion of overlapping binding sites (Equations 3 and 4). The fits to the modified McGhee-von Hipple models (Equations 2, 3, and 4) were poor and were rejected (Fig. 1). In contrast, fits using the two-binding site model (Equation 1) were better (Fig. 1). Whereas it is possible that there were more than two classes of binding sites, fits to three-or four-binding site models were not statistically better, so they were rejected on the basis that binding of this complexity could not be resolved with data of this accuracy.
There was little difference between the association constants for CcCBM17 and BspCBM28 binding to RC and Avicel TM . Values of ϳ1 ϫ 10 6 M Ϫ1 (⌬G ϳ8.3 kcal/mol) were obtained for the high affinity association constant (K a1 ) and ϳ2 ϫ 10 4 M Ϫ1 (⌬G ϳ5.9 kcal/mol) for the low affinity K a2 (Table I). It must be noted that the K a values and binding site densities ([N] o ) for the low affinity interactions are only estimates because the data do not extend to saturation of these binding sites. The association constants were 2-5-fold higher than originally reported for these CBMs, probably because of the use of a two-binding site model to fit the data, rather than the one-site model that was used originally. In general, the density of the low affinity sites ([N 2 ] o ) was 2-10-fold higher than that of the high affinity sites ([N 1 ] o ) for a given CBM and cellulose preparation ( larger for CcCBM17 indicating a significant difference in the way the two CBMs recognize cellulose. Neither CBM (20 M) adsorbed to RC (0.5 mg/ml) in the presence of excess cellopentaose (1 mM), a specific competitor for the CBM-binding site. This ratio of CBM concentration to RC was sufficient to saturate the high affinity sites and extend into the low affinity interactions. Scatchard plots of isotherms obtained in the presence of bovine serum albumin, a noncellulose binding protein used as a "blocking" agent, and using fluorescently tagged Oregon Green TM -CBMs to quantify free and bound protein also showed two classes of binding sites (not shown). This suggests specific binding to both classes of binding site, although a potential contribution of CBM-CBM interactions at the cellulose surface cannot be entirely excluded.
Competition Binding-BspCBM28 did not compete with CcCBM17 for binding to Avicel TM or RC (Fig. 2). Likewise, CcCBM17 was unable to compete with BspCBM28 for binding to sites on Avicel TM or RC. Adsorption isotherms with labeled CBMs showed that the fluorescent label did not interfere with binding (not shown). As expected, unlabeled CBM competed with the same labeled CBM for binding (data not shown). Therefore, there appear to be separate and distinct high affinity binding sites for CcCBM17 and BspCBM28 in noncrystalline cellulose. The quantities of protein required to do these experiments precluded a similar approach to examining competition for the low affinity class of sites.
In CcCBM17 tryptophan residues 88 and 135 are crucial to cellulose binding (9). Analogous tryptophan to alanine mutations at residues 453 and 500 in the CBM17 module of BspCBM17/CBM28 to give BspCBM17 (Ϫ) /CBM28 (residue numbers correspond to that of the whole Cel5A enzyme) resulted in a stoichiometry of 1.1 (Ϯ0.1) for binding cellohexaose in contrast to a stoichiometry of 2 mol of cellohexaose to 1 mol of wild-type BspCBM17/CBM28. The binding parameters of BspCBM17 (Ϫ) /CBM28 were virtually identical to those for the isolated BspCBM28 module (K a ϭ 3.90 (Ϯ0.05) ϫ 10 4 M Ϫ1 and ⌬H ϭ Ϫ58.24 (Ϯ1.0) kJ/mol for BspCBM17 (Ϫ) /CBM28 and K a ϭ 4.04 (Ϯ0.05) ϫ 10 4 M Ϫ1 and ⌬H ϭ Ϫ62.24 (Ϯ0.5) kJ/mol for BspCBM28 (8)   3). The kinetics of 2Ј4Ј-dinitrophenyl-␤-D-cellobioside hydrolysis were the same, within experimental error, for wild-type Cel5A and the three mutants. Therefore, the catalytic activity of the enzyme was unaffected by the mutations (not shown). The initial rates of reducing sugar release from RC were: (Fig. 4A). The enzyme concentrations used (1 M) were such that only high affinity sites were bound. The rates of hydrolysis slowed with time until at ϳ20 h all four enzymes were working at similar rates (indicated by constant differences between Cel5A, Cel5A17 (Ϫ) , Cel5A⌬28, and Cel5A17 (Ϫ) ⌬28; Fig. 4B). None of the reactions went to completion; only ϳ20 -25% of the insoluble cellulose was converted to soluble sugars. After 48 h of incubation the addition of fresh RC to hydrolysis reactions resulted in a burst of reducing sugar release for all four Cel5A derivatives (not shown) indicating that the wild-type and mutant enzymes were stable beyond 48 h of incubation. Furthermore, although the rates of sugar release in this burst were ϳ3-fold lower than at the initial time points, the relative rates of the enzymes remained consistent (not shown).

DISCUSSION
The Binding Sites for CcCBM17 and BspCBM28 Are Not the Same-CcCBM17 and BspCBM28 are unable to bind to crystalline preparations of cellulose and, therefore, must recognize only the regions of RC and Avicel TM that are not crystalline. The crystal structure of CcCBM17 revealed a binding site architecture well designed to accommodate only individual glucan chains providing a rationale for this binding specificity (9). Examples of CBMs in families 4, 6, 17, 28, and 29, now form a class of cellulose-specific modules that bind glucan chains rather than crystalline surfaces (8, 18 -22). The presence of high and low affinity binding sites for CcCBM17 and BspCBM28 in noncrystalline cellulose is a new and unexpected finding. It suggests multiple different physical presentations of cellulose chains that are sufficiently different to be discriminated by CBMs. Differences in physical presentation of a cellulose chain may involve chain conformation, the presence of local microstructure (e.g. proximity of other chains), or even differences in solvation. Distinct binding sites are also present on crystalline cellulose: CfCBM2a from Cellulomonas fimi Xylanase 10A recognized high affinity and low affinity binding sites in bacterial microcrystalline cellulose (12). In this case, heterogeneity in the structure of crystalline cellulose was proposed to influence binding.
CcCBM17 and CfCBM4-1 from C. fimi Cel9B bind to distinct sites in noncrystalline preparations of cellulose because, like CcCBM17 and BspCBM28, they do not compete for binding (5). This led to the proposal that cellulose chains in noncrystalline cellulose could have two or more different "physical forms" or "substructures." Our observations are unlikely to be an artifact of unseen oddities in the binding equilibria, such as the apparent irreversibility of CfCBM2a (5), because 1) for both CBMs the binding was completely reversible, 2) cellopentaose was as an effective binding competitor, and 3) each CBM competed with a fluorescently tagged itself for binding. Furthermore, there was a 2-3-fold disparity in the densities ([N] o ) of the high affinity binding sites for the two CBMs (Table I). Because these CBMs are a similar size (190 amino acids), similar shape (␤sandwich fold), and bind optimally to cellohexaose (8,21) differences in the "footprint" size of the CBMs on a cellulose molecule cannot explain the differences in capacity. Thus, there is compelling evidence that CcCBM17 and BspCBM28 recognize different high affinity sites in the noncrystalline regions of Avicel TM and RC. We are again led to the conclusion that noncrystalline cellulose contains chains with sufficiently different physical presentations to be discriminated by CBMs.
Noncrystalline cellulose has traditionally been referred to as amorphous cellulose, literally shapeless cellulose, because we fail to detect regular structural elements. It is well known that insoluble and soluble polysaccharides adopt conformations with energetic minima (23). Even the conformation of cellooligosaccharides in solution is stabilized by intramolecular hydrogen bonds to a consistent helical axis between 2-and 3-fold (24). Accordingly, it is likely that the molecules in noncrystalline regions of cellulose can adopt different conformations of minimum energy.
The conformation of oligosaccharides is known to play a role in the binding specificity of CBMs. Family 4 CBMs can discriminate between the loose, looping helical molecules of ␤-1,3glucans and the linear molecules of cellulose (25) and the position in space of a single tryptophan residue in family 2 CBMs determines specificity for crystalline cellulose or the FIG. 4. Hydrolysis of regenerated cellulose. Panel A, release of reducing sugar from RC by Cel5A (f), Cel5A17 (Ϫ) (q), Cel5A⌬28 (OE), and Cel5A17 (Ϫ) ⌬28 () measured by the HBAH reducing sugar assay (17). Error bars represent the standard deviation of readings performed in quadruplicate. Panel B, data normalized to the activity of the catalytic module alone (Cel5A17 (Ϫ) ⌬28) and expressed as a fraction of the maximum of that sugar released by Cel5A17 (Ϫ) ⌬28, i.e. the difference between the sample and Cel5A17 (Ϫ) ⌬28 at any time divided by the reducing sugars released by Cel5A17 (Ϫ) ⌬28 at 96 h.
3-fold helix of xylan (26). Discrimination of different cellulose chain conformations in noncrystalline cellulose may explain the observations made with CcCBM17, BspCBM28, and Cf-CBM4-1. However, the conformations of cellooligosaccharides bound in crystals of CcCBM17 and CfCBM4-1 are essentially identical to one another and to that of cellopentaose in solution (9,25). The only apparent difference in how these CBMs bind cellooligosaccharides is the depth of their binding sites and the orientations of the chains in the binding sites. CcCBM17 has a very shallow groove that binds the cellooligosaccharide with the planes of the sugar rings approximately parallel to the protein surface (9), whereas CfCBM4-1 has a deep groove that binds the cellooligosaccharide edge-on (25). It is possible that the conformations of cellooligosaccharides in the crystals may not represent the conformation of molecules in noncrystalline cellulose that are optimally bound by CBMs. It is significant that both of these CBMs bind noncrystalline cellulose with affinities approximately an order of magnitude larger than their affinities for cellooligosaccharides (9,27). Nonetheless, these structural observations do suggest that the presentation of the cellulose chains, e.g. orientation of chains relative to potentially interfering microstructures, rather than their conformation may play a role in how CBMs bind them.
Fine-tuning Cellulose Recognition with Tandem CBMs-Unfortunately, despite considerable effort, BspCBM17 could not be independently produced and characterized for comparison with BspCBM28. CcCBM17 was chosen as a substitute for BspCBM17 because it is well characterized and it has 55% amino acid identity and 70% similarity with BspCBM17 (Fig.  6). Most of the sequence divergence between CcCBM17 and BspCBM17 is in loop regions removed from the binding site (Fig. 6). All of the amino acid side chains comprising the cellulose-binding site of CcCBM17 are conserved in BspCBM17 and the residues in CcCBM17 that interact directly with the car- bohydrate are identical in the two molecules (Fig. 6). Thus, we believe that the structures and binding functions of BspCBM17 and CcCBM17 are very similar making CcCBM17 an appropriate substitute for BspCBM17. The following conclusions are based on this assumption. Although unlikely, it is formally possible that undetected subtle differences in BspCBM17 and CcCBM17 do lend them different specificity for noncrystalline cellulose.
The high affinity of BspCBM17/CBM28 relative to its constituent modules (20-fold larger K a s than the individual modules) is another example of cooperativity between the modules in tandem CBMs that leads to nonadditive enhancements in affinity, termed avidity (20,28,29). This must result from the simultaneous interaction of the covalently linked modules with binding sites in the cellulose that are proximal in three-dimensional space. A two-step mechanism for the adsorption of tandem CBMs to crystalline cellulose is appropriate to describe the high affinity interaction, K a1 , of BspCBM17/CBM28 with noncrystalline cellulose (29) (Fig. 7). Using the high affinity binding constants of CcCBM17 and BspCBM28 as estimates of K a, 17 and K a, 28 , the unimolecular fractions of K a1 , K u, 17 and K u, 28 can be calculated to be ϳ10 (dimensionless units). This equilibrium strongly favors the doubly bound species (both modules of the tandem bound) over the singly bound species 10 to 1 until binding sites with an arrangement appropriate for occupation by both modules of BspCBM17/CBM28 are depleted (i.e. [N 1 ] o ϭ 3.5 mol/g RC). The [N 1 ] o for BspCBM17/CBM28 compares remarkably well with [N 1 ] o for BspCBM28 binding to RC suggesting that the high affinity interaction of BspCBM17/CBM28 is limited by the number of CBM28-binding sites. Only some 40% of the high affinity CBM17-binding sites participate in this interaction. [N 2 ] o of BspCBM17/CBM28 was very close to the concentration of the remaining 60% (or ϳ5.0 mol/g cellu-lose) of the high affinity CBM17-binding sites. Furthermore, the affinity of CcCBM17 for RC was identical to K a2 of BspCBM17/CBM28. Presumably, the second class of binding sites (K a2 ) for BspCBM17/CBM28 represents the formation of species bound only by its CBM17 module to a high affinity CBM17-binding site.
The biological significance of this form of binding cooperativity in CBMs is unclear. We have previously suggested that in many cases this may be a mechanism for compensating the loss of affinity at higher temperatures in thermophilic and hyperthermophilic CBMs (20). This is unlikely in this case, because the source organism is mesophilic, and in other cases of cooperativity in CBMs that originate from mesophilic organisms Residues identical in BspCBM17 are colored blue, similar residues are colored pink, and variant residues in khaki. The bound cellotetraose (gray and red) and residues that interact with it (color corresponds to conservation as above) are shown in a ball and stick representation. The bound Ca 2ϩ is shown as a green sphere. This figure was prepared with MOLMOL (39). Panel B shows an amino acid alignment of CcCBM17 and BspCBM17. Conserved residues, identical or similar, corresponding to those in panel A are highlighted. Aromatic amino acids involved in the binding of CcCBM17 to cellotetraose are indicated above the sequences by ࡗ. Polar amino acids involved in binding are indicated by q. Asterisks (*) mark blocks of 10 amino acids. This alignment was prepared using ClustalW (40). (28,30). BspCBM17/CBM28 is effectively concentrated to a specific region through its high affinity for proximal CBM17 and CBM28-binding sites. The low affinity sites for BspCBM17/ CBM28 will only be occupied when the high affinity arrangement is completely occupied. Thus, an alternative function of tandem CBMs may be to "fine-tune" binding specificity for cellulose chains, as in this case, or other glycans with a particular physical presentation.
Do the CBM17 and CBM28 Modules of Cel5A Influence Cellulose Hydrolysis?-The presence of the CBM17 and CBM28 modules in wild-type Cel5A clearly enhances the ability of the enzyme to release soluble sugars from noncrystalline cellulose. However, the effect is subtle, producing only up to a 2-fold enhancement in the initial rate of sugar release and a ϳ40% increase in overall soluble sugar yield relative to Cel5A lacking functional CBMs. This is consistent with the modest 2-3-fold decreases in activity on insoluble xylan, ␤-1,3-glucans, or cellulose when a CBM is removed from its cognate catalytic module (31)(32)(33)(34)(35). Occasionally, larger decreases in catalytic activity are observed (36 -38).
The enzyme loading on the cellulose at the initiation of hydrolysis can be predicted using knowledge of the individual CBM and tandem binding properties. 99% of Cel5A and 74% of Cel5A17 (Ϫ) are predicted to be bound to the region where CBM17-and CBM28-binding sites are in close proximity, which will be called region A. 89% of Cel5A⌬28 would be distributed ϳ40:60 between region A and the region comprising only CBM17-binding sites, which will be called region B. Considering only bound quantities of enzyme, one would expect differences in the initial rates of cellulose hydrolysis (within the first 2-3 h) that are approximately proportional to the amounts of bound enzyme, i.e. relative activities of Cel5A Ͼ Cel5A⌬28 Ͼ Cel5A17 (Ϫ) . All were better than Cel5A17 (Ϫ) ⌬28 supporting the general importance of the CBMs to Cel5A activity. However, Cel5A17 (Ϫ) had better initial rates of releasing reducing sugars than Cel5A⌬28 (Fig. 4), mostly attributable to glucose release (Fig. 5B), which is not consistent with predictions based only on bound quantities of enzyme. Although the difference is admittedly subtle, we suggest that region A may have greater susceptibility to hydrolysis than region B resulting in the observed differences in enzyme activity. The finetuning of cellulose recognition by the CBM tandem appears to refine noncrystalline cellulose hydrolysis by Cel5A.
Biological Implications-Binary (crystalline versus noncrystalline) paradigms of cellulose structure are inadequate with respect to the enzymology of cellulose recognition and hydrolysis. Mounting evidence indicates that CBMs discriminate distinct binding sites within noncrystalline cellulose implying that this form of cellulose is not truly structureless as it was thought to be. Indeed, this form of cellulose, which is more appropriately called noncrystalline cellulose instead of amorphous cellulose, appears to have cellulose chains with different and distinct physical presentations. Discrimination of these physical presentations by the CBMs of cellulases has clear implications on the hydrolytic activities of these enzymes.