THE JOURNAL OF BIOLOGICAL CHEMISTRY 0 1992 by The American Society for Biochemistry and Molecular Biology, Inc.

The Adsorption of a Bacterial Cellulase to Crystalline Cellulose*

Vol. 267, No. 10, Issue of April 5, pp. 6743-6749, 1992 Printed in U. S. A.

and Its Two Isolated Domains

(Received for publication, December 5, 1991)

Neil R. Gilkes$s, Eric JervisSII, Bernard Henrissat 11, Bahar Tekant$, Robert C. Miller Jr.$, R. Antony J. Warren$, and Douglas G. Kilburn$ From the $Department of Microbiology, the TDepartment of Chemical Engineering, University of British Columbia, Vancouver, British Columbia V6T 123. Canada and the IlCentre de Recherches sur les Macromolecules Vegetales, Centre National de la Recherche Scientifique, F-38041 Grenoble, France

CenA is a bacterial cellulase @-1,4-glucanase) com- prised of a globular catalytic domain joined to an ex- tended cellulose-binding domain (CBD) by a short linker peptide. The adsorption of CenA and its two isolated domains to crystalline cellulose was analyzed. CenA and CBD*PTcenA (the CBD plus linker) adsorbed rapidly to cellulose at 30 O C , and no net desorption of protein was observed during the following 16.7 h. There was no detectable adsorption of the catalytic domain. Scatchard plots of adsorption data for CenA and for CBD*PTcenA were nonlinear (concave upward). The adsorption of CenA and CBD*PTcen~ exceeded 7 and 8 pmol/g cellulose, respectively, but saturation was not attained at the highest total protein concentrations employed. A new model for adsorption was developed to describe the interaction of a large ligand (protein) with a lattice of overlapping potential binding sites (cellobiose residues). A relative equilibrium associa- tion constant (Kr) of 40.6 and 45.3 liter-g cellulose” was estimated for CenA and CBD*PTcenA, respectively, according to this model. A similar K , value (33.3 li- ter-g”) was also obtained for Cex, a Cellulomonas fimi enzyme which contains a related CBD but which hy- drolyzes both B1,4-xylosidic and ~-1,4-glucosidic bonds. It was estimated that the CBD occupies approx- imately 39 cellobiose residues on the cellulose surface.

Microorganisms use a consortium of lytic enzymes to con- vert cellulose and hemicelluloses to soluble sugars. The cel- lulases often act synergistically and become strongly adsorbed to their insoluble substrate. A structural basis for the adsorp- tion of bacterial cellulases has been provided by our studies on an endo-/3-1,4-glucanase (CenA)’ from Cellulomom fimi (1-9). This enzyme is composed of two domains which are

* This work was supported by the Natural Sciences and Engineer- ing Research Council of Canada. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accord- ance with 18 U.S.C. Section 1734 solely to indicate this fact.

§ T o whom correspondence should be addressed. Fax: 604-822- 6041.

The abbreviations used are: CenA, C. fimi endo-/3-1,4-glucanase A (for previous nomenclature, see Ref. 5); BMCC, bacterial micro- crystalline cellulose; CBD, cellulose-binding domain; CBD. PTcenA, the cellulose-binding domain of C. fimi endo-P-1,4-glucanase A, plus the 22 amino-proximal amino acids of the proline- and threonine- rich interdomain linker peptide (see Fig. 1C); CenB, C. f imi endo-p- 1,4-glucanase B; Cex, C. f imi exo-p-1,4-glucanase and @-1,4-xylanase; P T box, the proline- and threonine-rich interdomain linker peptide of CenA and other C. fimi P-1,4-glucanases; SDS-PAGE, sodium dodecyl sulfate-polyacrylamide gel electrophoresis; 1, liter(s).

arranged in the shape of a tadpole (8). The extended tail region can bind independently to cellulose when isolated by proteolytic cleavage or genetic manipulation; catalytic func- tions are contained in the globular head. The two domains are joined by a 23-amino acid proline- and threonine-rich linker (the P.T box) (2, 3, 10). The same general structural and functional organization has been demonstrated in cellu- lases from the fungus Trichoderma reesei (11, 12).

The tail region, or CBD, of CenA is amino-terminal and contains 111 amino acids. CBDs with similar primary struc- tures are located at the carboxyl termini of C. fimi Cex (an exo-/3-1,4-xylanase and -/3-1,4-glucanase) and CenB (an endo- p-1, 4-glucanase) (2, 7, 13, 14); related structures have since been noted at the amino or carboxyl termini of cellulases from three other bacterial genera (13) and in a spore germination- specific cellulase from Dictyostelium discoideum. (15). Fungal CBDs have a rather different primary structure (11). Other CBDs, which are not obviously related to those from C. fimi, are evident in several bacterial enzymes (16-19).

The widespread occurrence of CBDs implies that enzyme adsorption plays an important role in cellulose degradation and that the mechanism of adsorption must be understood in order to describe this process fully. The adsorption of C. fimi- type CBDs also has practical significance: our experiments with heterologous fusion proteins containing C. fimi CBDs have shown that these domains provide an effective means to purify and immobilize foreign proteins on cellulose matrices (9, 20-22). Accordingly, we have analyzed the adsorption of CenA and its isolated CBD to crystalline cellulose on a quantitative basis. The adsorption of Cex was also examined. This is the first such investigation for any bacterial cellulase. The analysis of adsorption to cellulose is complicated by the nature of the substrate. We argue here that the surface of crystalline cellulose presents a two-dimensional array of over- lapping potential binding sites and, consequently, that the analysis of all cellulase adsorption data requires a different approach from the simple Scatchard treatment normally em- ployed.

EXPERIMENTAL PROCEDURES

Materials-CenA was purified from Escherichia coli JMlOl (pUC18-1.6 c e d ) and transferred to 50 mM potassium phosphate, pH 7.0 (phosphate buffer), containing 0.01% NaN3, as described previously (5). The isolated CenA catalytic domain (p30) was pre- pared from CenA by digestion with C. fimi protease, followed by size- exclusion chromatography in phosphate buffer containing 0.01% NaN3, as described (8). CBD. PTcenA was purified from E. coli JMlOl (pUC18-CBDPT) as described below. Details of plasmid construction are given in Fig. 1. Cex was purified from E. coli C600 (pUC12- l . l(PTIS)) as described previously (5). BMCC was prepared from cultures of Acetobacter xylinum (ATCC 23769) grown on peptone/ yeast extract/glucose medium (23). One-1 cultures in 25 x30-cm

6743

6744 Adsorption of a Cellulase to Crystalline Cellulose covered trays were incubated at 30 "C for 7 days, without shaking; the cellulose pellicles produced on the surface of the medium were then harvested. The pellicles from 10 cultures were cut into 1-cm squares, washed extensively with HzO, and extracted with 2 1 of 4% NaOH at 4 "C for 24 h. Extraction was repeated five times with fresh NaOH. After washing in HzO, the cellulose was added to 1 1 of 2.5 N HCI and heated under reflux for 1 h. The mixture was cooled and then homogenized in a blender operated at full speed for 20 s. The cellulose microfibrils were recovered by filtration on a glass fiber filter and resuspended in fresh 2.5 N HCI. Heating under reflux was repeated two more times for 30 min each. Acid was removed by extensive washing with HzO; the purified microfibrils (BMCC) were recovered by centrifugation at 10,000 X g for 10 min. Aliquots of BMCC suspended in HzO were withdrawn for dry weight determina- tion. The total yield of BMCC was 6.4 g. BMCC was resuspended in phosphate buffer containing 0.01% NaN3 to a concentration of 2 mg/ ml and stored at 4 "C.

Production and Purification of CBD.PTc<,-E. coli JMlOl (pUC18-CBD.PT) (Fig. 1) was grown at 37 "C in a 14-1 Chemap fermenter (Chemap, Volketswil, Switzerland) fitted with pH and oxygen electrodes. Ten 1 of medium (0.5% yeast extract, 0.3%

glucose, and 0.01% sodium ampicillin) were inoculated with 0.5 1 of an overnight culture. The pH in the fermenter was maintained at 7.0 by automated addition of 5 N HCI or 0.5 N HCI. Dissolved oxygen was maintained at ~ ( 0 2 ) = 0.1 atmospheres by automated regulation of agitation speed and oxygen supplementation of the sparger inlet gas. Antifoam C (Sigma) was added automatically, as required. The culture was grown to an optical density (600 nm) of 40.0 using a fed batch procedure (24) by supplementation of glucose and yeast extract, as required. Isopropyl-P-D-thiogalactoside was then added to a final concentration of 0.1 mM to induce synthesis of CBD.PT~,.A. Fed batch fermentation was continued, with glycerol and yeast extract supplementation, until the culture reached an optical density of 50; cells were then harvested using a Sharples type T1-P centrifuge (Penwatt Corp., Warminster, PA). The total yield of cells (wet weight) was 560 g. Cells were stored frozen at -70 "C.

The cell pellet was thawed and resuspended in 2 1 of phosphate buffer containing 0.02% NaN3, 3 mM Na EDTA, 1 pM phenylmeth- ylsulfonyl fluoride, and 0.1 p~ pepstatin A. Cells were ruptured in a French pressure cell and the cell lysate centrifuged at 27,000 X g for 30 min to remove cellular debris. Streptomycin sulfate was added to a final concentration of 1.5%, and the cell extract was stirred at 4 "C for 48 h. Precipitate was removed at 24 and 48 h by centrifugation at 150,000 X g for 30 min; the clarified extract was stored at -70 "C. CBD.PTc..A was isolated from this extract by cellulose affinity chro- matography at 4 "C using a 5 X 60-cm column containing 250 g of CF1 cellulose (Whatman Inc., Clifton, NJ). The column was con- nected to an fast protein liquid chromatography system (Pharmacia LKB Biotechnology, Uppsula, Sweden) and equilibrated with phos- phate buffer containing 0.02% NaN3. The column flow rate was 1 ml/ min. The extract was loaded onto the column and eluted successively with 1.8 1 of phosphate buffer containing 1 M NaCl and 0.02% NaN3, and 1.5 1 of phosphate buffer containing 0.01% NaN3, to remove unbound E. coli proteins. CBD. PT~,,A, which remained bound to the column, was eluted with phosphate buffer containing 7 M guanidine hydrochloride. Elution of protein was monitored by absorbance at 280 nm. The guanidine hydrochloride eluate was concentrated to 10 ml by ultrafiltration using a YM5 membrane (Amicon Corp., Danvers, MA) and then diluted to 200 ml with 2 mM NH40H, pH 9.4. The concentration of guanidine hydrochloride was reduced to approxi- mately 2 p~ by repeated concentration and dilution with 2 mM NH,OH. The final volume of the concentrated desalted guanidine hydrochloride eluate was reduced to 6.4 ml by ultrafiltration, as above. The protein concentration of this partially purified CBD.PTcenA preparation was estimated at approximately 53 mg/ml, using an extinction coefficient (280 nm, 1 mg/ml; 1 cm) of 2.85 (see below). Final purification was by anion-exchange chromatography using a Mono Q HR 5/5 column (Pharmacia) equilibrated with 2 mM NHIOH, pH 9.4, operated at a flow rate of 1.0 ml/min (Fig. 2). The partially purified protein was chromatographed in 10-mg aliquots. CBD. PTc..A was recovered in the column flow-through; contaminating proteins remained bound to the column and were removed with a salt gradient (0.0-0.5 M sodium acetate, pH 9.8). The flow-through frac- tions from 21 chromatographic runs were pooled (total volume = 392 ml) and concentrated to 20 ml by ultrafiltration using a YM3 mem- brane (Amicon). Purification was monitored by SDS-PAGE (7); molecular mass standards (Fig. 3) were myosin, rabbit muscle (212

KHzPO4, 0.3% KZHP04, 0.1% (NH4)Z HPO4, 0.02% MgS04, 0.2%

kDa); @-galactosidase, E. coli (130 kDa); phosphorylase b, rabbit muscle (97.4 kDa); albumin, bovine (68 kDa); pyruvate kinase, rabbit muscle (57.2 kDa); glutamate dehydrogenase, bovine liver (53 kDa); ovalbumin (45 kDa); alcohol dehydrogenase, equine liver (41 kDa); glyceraldehyde-3-phosphate dehydrogenase, rabbit muscle (36 kDa); carbonic anhydrase, bovine erythrocyte (29 kDa); trypsin inhibitor, soybean (20.1 kDa); and cytochrome c, equine heart (12.4 kDa). The final yield of CBD. PTc..A was 161 mg. The concentrated preparation was lyophilized in 1.0-ml aliquots and stored frozen at -20 "C.

Estimation of Adsorption Parameters-Adsorption assays were done at 30 "C in 2-ml microcentrifuge tubes. Tubes contained 1.5 mg of BMCC, 1.5 ml of phosphate buffer, 0.01% NaN3, and 1.1-32.2 p~ CenA, p30, CBD. PTcen~, or Cex (see legend to Fig. 4). Tube contents were continually mixed by rotation at 2 rpm. After equilibration for 16 h, BMCC and bound protein were removed by centrifugation at 20,000 X g for 5 min. Centrifugation of supernatants was repeated twice to ensure removal of all particulate material; the concentration of unbound protein, [ F l , was then determined from the absorbance at 280 nm. The bound protein concentration, [B] , was determined from the difference between the initial protein concentration and [FJ. The far-UV absorbance method of Scopes (25) was used to estimate extinction coefficients at 205 nm for CenA, p30, and CBD. PTc,,A. These values were used to calculate the absolute concentra- tions of solutions of these proteins that were then used to estimate extinction coefficients at 280 nm (1 mg/ml; 1 cm). The extinction coefficients for CenA and p30 (2.64 and 2.38, respectively) were reported previously (8); these values are in good agreement with those predicted from the tryptophan and tyrosine contents of the proteins (26). The molecular masses for CenA and p30 (43.8 and 29.6 kDa, respectively) were calculated from primary structures deduced from DNA sequence analysis (3). The extinction coefficient (280 nm) estimated for CBD. PTc,,A was 2.85; the predicted value is 2.71, using a molecular mass of 14.1 kDa (calculated from the deduced CBD. PTc,,A primary structure, see Fig. 1C). The extinction coefficient (280 nm) estimated for Cex was 1.61; the predicted value is 1.73, using a molecular mass of 47.3 kDa.

Modeling of Adsorption and Analysis of Adsorption Data-The adsorption of a ligand (CenA, CBD.PT~.,A, or Cex) is assumed to be an equilibrium reaction in which a single ligand reacts with one or more of the repeating cellobiose lattice units on the surface of crys- talline cellulose. At equilibrium, the adsorption reaction is described by the following

where [B] is the concentration of bound ligand (mo1es.g cellulose"), [ F J is the concentration of free ligand (molar), [ I V l is the concentra- tion of available binding sites (mo1es.g cellulose"), and KO is the equilibrium association constant (liters. mol").

If a single ligand interacts with only one lattice unit and there are no positive or negative co-operative effects,

[NI = IN01 - [BI (2)

where [No] is the concentration of binding sites in the absence of ligand. Substitution of Equation 2 into Equation 1 and rearrangement yields the Langmuir equation

However, the dimensions of the CBD (8, 10) greatly exceed the dimensions of the repeating cellobiose lattice unit on the cellulose surface (27); therefore, the ligand must occupy several lattice units. If a binding site is larger than one lattice unit, the surface must be considered as an array of overlapping potential binding sites. Under these conditions, [ I V l is described by a probability function which depends not only on [B] but also on the configuration of bound ligands on the cellulose surface (see "Discussion"). To avoid this complication, it is appropriate to consider adsorption at only very low values of [B] where ligands are spaced such that any two nearest ligands do not exclude the binding of a third ligand. Under these conditions, Equation 2 may be rewritten as follows

[N] = [NO] - a[Bl (4)

where a is the number of lattice units occupied by a single ligand

Adsorption of a Cellulase to Crystalline Cellulose 6745

molecule. Substitution of Equation 4 into Equation 1 and re- arrangement yields the following

Data obtained at low ligand concentrations may be fitted to this equation. It is convenient to plot the adsorption data using the double- reciprocal form of equation ( 5 )

1 l l a -=-.-+- [BI k m o l [FI [No1

(6)

which emphasises data for the lower concentration range. The slope (l/K.[NO]) and intercept (a/[No]) of a plot of l/[B] uersus l / [ q were estimated by fitting a straight line through data points for low values of [B] . Unique solutions for KO, [No], or a cannot be obtained from this analysis but a relative equilibrium association constant, K, (li- ters .g cellulose"), where

K, = [NOIKO (7)

can be used to compare the affinities of various related ligands for a given preparation of cellulose (i.e. when [No] is constant). A doubly weighted least squares analysis was used because error occurs in both the 1/[B] and l/[q dimensions. In this method, residuals are weighted along both axes when minimizing the sum of squares errors between the fitted line and data points.

Other Procedures-Cellulose hydrolysis was estimated from the release of soluble reducing sugar determined with dinitrosalicylic acid reagent, with reference to D-glucose (1). The theoretical PI of CBD. PTcenA was calculated from its primary structure, as described (28). Plasmid construction used standard recombinant DNA methodology.

RESULTS

Purification of CenA, p30, and CBD.PTced and Cex for Adsorption Analyses-Purification of the nonglycosylated form of CenA, synthesized from recombinant C. fimi DNA in E. coli (Fig. 3, lune 3), has been described (5). This enzyme is cleaved by a C. fimi protease between Thr-165 and Val-166 (i.e. at the carboxyl terminus of the P.T box) into two fragments (5,7). The stable 29.7-kDa carboxyl-terminal frag- ment (p30) comprises the CenA catalytic domain; this can be purified from the digest by size-exclusion chromatography (Fig. 3, lune 2 ) . The corresponding amino-terminal fragment, which comprises the CBD plus the P.T box, is produced in nonstoichiometric proportions, presumably because it is sus- ceptible to further proteolysis (5). Therefore, an alternative method for the production of this polypeptide was adopted. The cenA gene was manipulated in vitro to remove the region encoding the catalytic domain (Fig. U). The resulting gene fragment was ligated into pUC18 to give the recombinant plasmid pUC18-CBD.PT (Fig. 1B) which was used to trans- form E. coli JM101. The cenA gene fragment on this plasmid encodes the polypeptide CBD-PTcenA (i.e. the entire CBD, plus the P-T box lacking Thr-165), as shown in Fig. 1C. CBD. PTcenA was purified from an E. coli JMlOl (pUC18- CBD. PT) cell extract by cellulose affinity chromatography (5) and anion-exchange chromatography (Fig. 2). CBD. PTcenA bound very weakly to the anion-exchange column a t pH 9.4 (i.e. 3.2 pH units above its theoretical PI), presumably because of its low charge density (2). However, since contam- inating proteins were more strongly bound, the chromato- graphic step was effective. The purified CBD . PTc,,A prepa- ration was homogeneous, as judged by SDS-PAGE (Fig. 3, lane 1 ) . Its apparent molecular mass (20.0 kDa), relative to standard proteins, was greater than the molecular mass pre- dicted from its primary structure (14.1 kDa). The P .T box was previously shown to cause anomolous electrophoretic migration (7). p30 was also purified to apparent homogeneity (Fig. 3, lune 2 ) ; a minor (<1% total) 30-kDa contaminant was evident in the purified CenA preparation (Fig. 3, lune 3 ) . Cex

A

C 1 P U S T A A O A A P Q C 1 P T P .

ATG TO... m G C C G C O U G G C G G O X G X T ~ ... . . ~ O X * u j o x ~ ~ o

$ 6 4

FIG. 1. Construction of the expression vector pUC18- CBD-PT and structure of the CBD-PT gene. The construction of the expression vector by a three-fragment ligation is shown in A . Fragment i was obtained from pUC18 by digestion with EcoRI and KpnI. A 1.6-kilobase SstI-SstI fragment containing c e d (6) was excised from pUC18 1.6 c e d and used to prepare single-stranded DNA in M13 mp18. A 24-base primer (5' CGT CGG CGT GGG GGT GGG GGT CGG 37, complementary to nucleotides 472-495 of cenA (3), was hybridized to the single-stranded DNA and extended. The single-stranded overhang was removed with mung bean nuclease and the double-stranded DNA cut with EcoRI to give fragment ii; the region encoding the CBD is shown stkpled, and the regions encoding the P.T box (blunt end), and leader peptide is shown in black. Fragment iii was a synthetic blunt end-KpnI linker containing an internal EcoRI site. Ligation of fragments i, ii, and iii gave plasmid pUC18 CBD.PT ( B ) . The functional orientations of the p-lactamase (Amp') and lacZ' genes are indicated by arrows. The structure of the CBD.PT gene is shown in C. The gene encodes the CBD, plus the PT box missing its COOH-terminal amino acid residue (Thr-165). The encoded amino acid sequence is shown in single-letter code, numbered from Met-1 of the leader peptide. Leader peptide processing occurs between Ala-28 and Ala-29 and between Ala-31 and Ala-32 (5).

was purified to apparent homogeneity, as described previously (5).

Analysis of the Adsorption of CenA, CBD . PTC,,,..,, p30, and Cex to BMCC-The kinetics of the adsorption of CenA and CBD.PTcenA to BMCC at 30 "C are shown in Fig. 4 (inset). High and low total protein concentrations, relative to the concentration range used to measure adsorption isotherms, were tested. Equilibration was complete within the shortest experimentally feasible incubation time (0.2 min), a t both concentrations. There was no net desorption of either protein during the following 16.7 h. There was no detectable adsorp- tion of p30. Hydrolysis of BMCC after an 18-h incubation with 18.3 PM CenA amounted to 2.7% of the total cellulose, as determined by the release of soluble reducing sugar; 0.8% hydrolysis was obtained with 18.3 PM p30. No hydrolysis was detected with CBD . PTc,,A at the same molar concentration.

The equilibrium adsorption isotherms for CenA, CBD. PTcenA, and p30 (1.1-32.2 p~ total protein) are shown in Fig. 4 (main panel). The absence of p30 adsorption found in the kinetic experiment was confirmed. Saturation of BMCC by CenA and CBD.PTc,,A was approached but not attained at the highest total protein concentrations used. This failure to

6746 Adsorption of a Cellulase to Crystalline Cellulose

................ ]m

................................... .. 0

o i 5 3 J 4 5

time (mill)

FIG. 2. Anion-exchange chromatography of CBD-PTC,~. Partially purified CBD. PT (10 mg in 10 ml of 2 p~ NH,OH, pH 9.4) was loaded onto an anion-exchange column at 0 min and chromato- graphed as described under “Experimental Procedures.” The major chromatographic peak, which appeared in the column flow-through, contained purified CBD. PT&.A. Contaminating proteins bound to the column and were removed with a 0-100% buffer B (0.5 M am- monium acetate, pH 9.8) gradient.

1 2 3 4

FIG. 3. SDS-PAGE analysis of CBD.PTC,.A, p30, and CenA. Purified CBD.PTC,.A ( l a n e I ) , p30 (lane 2) , and CenA ( l a n e 3) were analyzed on a gel containing 12.5% acrylamide. Each lane was loaded with 10 pg of protein. Arrowheads on the right indicate the positions of the molecular mass standards (212, 130, 116, 97.4, 68, 53, 45, 41, 36, 29, 20, and 12.4 kDa; see “Experimental Procedures”) shown in lane 4.

reach saturation was emphasized when the same data was plotted in semi-logarithmic form ( [B] uersus log [Fj), as shown in Fig. 5a. Scatchard plots of the data for CenA and CBD. PTcenA (Fig. 5b) were nonlinear (concave upward), indicating a complex interaction of these proteins with BMCC.

A relative equilibrium association constant (K,) of 40.5 1 g cellulose” was estimated for the adsorption of CenA from the limiting slope of a plot of 1/[B] uersus l/[q (Fig. 6a). Esti- mation of K, is based on a model of overlapping potential binding sites, each comprised of multiple lattice units (cello- biose residues) on the cellulose surface, as described under “Experimental Procedures” and discussed below. The corre- sponding K, value estimated for CBD PTcen* (Fig. 6b) was

The adsorption of Cex to BMCC was also examined. Ad- sorption at low protein concentrations only was analyzed, to permit estimation of K, (Fig. 6c). The kinetics of Cex adsorp- tion within this concentration range were comparable with CenA (data not shown). The K, estimated for Cex was 33.3 1. g”. K, estimates are summarized in Table I. Estimates for K, (the equilibrium association constant) and a (the number of lattice sites occupied by one ligand) are also listed. The K,

45.3 1. g-1.

7m 7.0

3 - 6.0 m

v) 0

-

4.0

- 3.0

2.0

- 1.0 m 0.0

.- W

- v) al

a v

t 4-

“T- ””l

TUE

-1.0 1 0.0 10.0 20.0 30.0

[F] (pM ligand)

FIG. 4. Adsorption of CenA and its isolated domains to BMCC. The main panel shows the equilibrium adsorption isotherms ( [ B ] uersw [q) for CenA (A), CBD*PTcenA (W), and p30, the catalytic domain (+). Adsorption assays were done at 30 “C, as described under “Experimental Procedures.” The initial protein concentration range was 1.1-27.3 p~ for CenA, 1.1-32.2 p~ for CBD.PTC.,A, and 2.3- 12.2 p~ for p30. Each data point is the mean of six replicates; standard errors in two dimensions are indicated by vertical and horizontal bars. The inset shows the kinetics of adsorption of 3.4 pM (A) or 18.3 pM (A) CenA, 3.4 pM (W) or 18.3 pM (0) CBD.PTcenA, and 18.3 pM p30 (0). Each data point is the mean of six replicates; standard errors are indicated by Vertical bars.

9.0 ,

1 1 a L

7.0

8.0

w 5.0

m 4.0 .- c

**

.01 1 1 10 100 [F] (PM ligand)

40

- 30 -

- Y - 10 m -

0

0.0 2.0 4.0 8.0 8.0 [Bl (prnoles llpand Q cellulose.~)

FIG. 5. Analyses of the adsorption of CenA and CBD*PTC..A to BMCC. a shows a semi-logarithmic plot ( [ B ] uersw log[q) of the adsorption data for CenA (A) and CBD.PTcenA (W) from Fig. 4. b shows Scatchard plots ( [ B ] / [ q uersw [ B ] ) of the same data. Curued lines were fitted to data points for CenA (-) and CBD. P T c e n ~ (- - -) by least squares regression analysis. In both types of plot, the standard errors in two dimensions are indicated by vertical and horizontal bars.

Adsorption of a Cellulase to Crystalline Cellulose 6747

0.1 I I I I 0.0 10.0 20.0

V[FI (pU llpand“) 30.0

1.0 , I

0.0 10.0 20.0 30.0 VlFl (pM Ilpand”)

0.2 1 4 I

0.1 I I I

0.0 10.0 20.0 30.0 VlFl (pU Ilpand“)

FIG. 6. Double-reciprocal plots of adsorption data for CenA and CBD-PTC.,~ and Cex. The adsorption data for CenA and CBD. PTC,,* from Fig. 4 are shown in a and b, respectively; data for Cex are shown in c. Data are plotted in double-reciprocal form (1/ [B] uersw l/[Fl), with standard errors in both dimensions indicated by vertical and horizontal bars. The K, and a/[No] values listed in Table I were estimated from the limiting slopes (straight lines) of these plots and their intercepts on the l/[B] axis, respectively, ac- cording to Equations 6 and 7 under “Experimental Procedures.” In each case, the slope was obtained by fitting lines through data points for the five lowest values of [B].

and a values were estimated from plots of l/[B] uersus l / [ q according to equation (6), “Experimental Procedures,” using an estimated value for [No] of 101 pmol of lattice residues/g of BMCC, as discussed below.

DISCUSSION

Many cellulases bind tightly to cellulose. The recent dis- covery that several cellulases contain an independent domain which mediates binding (4, 5, 11) enables the adsorption to be analyzed on a more detailed molecular basis. C. fimi CenA provides a useful model system for such analysis. Pure enzyme and its isolated CBD are easily prepared from E. coli express- ing recombinant C. fimi DNA, whereas the isolated catalytic domain is conveniently obtained from a proteolytic digest of the whole enzyme. Moreover, the approximate sizes and shapes of the enzyme and its domains are known from small- angle x-ray scattering studies (8).

The suitability of the cellulose substrate used in adsorption analyses was also considered. BMCC has certain advantages

TABLE I Adsorption parameters for the binding of CenA, CBD .PTCenA, and

Cer to BMCC The parameters (f accumulated standard error) were calculated

from adsorption data plotted in double reciprocal form (Fig. 6), as described under “Experimental Procedures.” The values for K. and a were calculated using [No] = 101 pmol lattice residues.g cellulose”, as detailed under “Discussion.”

Ligand K, K. a l [ N ~ l a

1.g” 1.pmot’ g. pmot’ mol. molt’ CenA 40.5 f 3.3 0.401 f 0.032 0.325 f 0.014 32.9 f 1.4

45.3 f 2.1 0.449 f 0.020 0.357 f 0.002 39.2 f 0.2 CBD. PTCenA

Cex 33.3 f 5.5 0.330 f 0.051 0.276 f 0.113 27.9 f 11.4

over less defined microcrystalline cellulose preparations, no- tably Avicel, usually employed in analyses. BMCC consists of a uniform aqueous suspension of cellulose fibrils prepared from the cellulose pellicle produced by A. xylinum cultures (23). The suspension contains bundles of microfibrils, 20-50 nm wide and several pm long (29, 30). The microfibrils are highly crystalline, allowing structural analysis by x-ray or electron diffraction. They consist of cellulose I, i.e. extended /3-1,4-glucan chains arranged in parallel orientation (31). By contrast, Avicel is a relatively heterogeneous cellulose prepa- ration obtained from wood fibres by partial acid hydrolysis followed by spray drying of the washed slurry. It contains both crystalline and disordered (so-called “paracrystalline” or “amorphous”) components resulting in a crystallinity of about 50% (32-34); electron microscopy reveals a mixture of rod- shaped particles and irregular aggregates (35). Porosity meas- urements show intraparticulate pores (1-10 nm) and inter- particulate voids ( 2 5 pm) (35). The complications of substrate heterogeneity and surface accessibility associated with Avicel can be largely avoided by the use of BMCC.

The adsorption of CenA, CBD . PTCen,+, and Cex to BMCC was rapid. Much slower equilibration times have been re- ported for the adsorption of T. reesei cellulases to Avicel (36, 37), which may reflect the time required for their diffusion into Avicel pores as well as specific differences in adsorption kinetics. BMCC was hydrolyzed very slowly by CenA approx- imately 2.7% of the substrate was hydrolyzed during 18 h incubation at high (18.3 p ~ ) CenA concentrations. As ex- pected, no hydrolysis was observed with CBD . PTcenA. Hy- drolysis by the isolated CenA catalytic domain was approxi- mately 30% of that determined with an equimolar concentra- tion of intact enzyme. We previously noted a similar decrease in the rate of Avicel hydrolysis following removal of the CBD (5). There was no detectable hydrolysis of BMCC by Cex.

Plots of [B]/[r;l uersus B (Scatchard) or 1/[B] uersus l / [ q for the adsorption of CenA, CBD.PTc.,A, and Cex were nonlinear. Deviation from linearity (concave upward) in a Scatchard plot is usually interpreted as indicative of two or more classes of binding sites and/or ligand-ligand interaction (negative co-operativity); graphical analysis alone cannot dis- tinguish or resolve these possibilities (38). These interpreta- tions are generally valid for small ligands interacting with independent binding sites; however, many biological macro- molecules are large ligands which occupy several lattice resi- dues on their substrates. In such cases, the substrate should be considered as an array or lattice of overlapping potential binding sites. As previously noted for a one-dimensional lat- tice, the binding of any large noninteracting (i.e. nonco- operative) ligand to an array of overlapping potential sites inevitably results in a nonlinear (concave upward) Scatchard plot for which classical analysis is inapplicable (39). The dimensions of the CBD of CenA (8, 10) greatly exceed those

6748 Adsorption of a Cellulase to Crystalline Cellulose

of the repeating cellobiose lattice units exposed on the surface of bacterial cellulose microfibrils (27), so it is reasonable to assume that adsorption involves interaction with (or at least masking of) more than one lattice unit. Consequently, the surface of the cellulose crystal can be viewed as a two- dimensional lattice comprised of an array of overlapping potential binding sites (see Fig. 4 of Ref. 27). As noted for a one-dimensional overlapping binding site model (39, the con- centration of free ligand binding sites, [w, is a function, not only of the number of ligands already bound, but also of their distribution on the lattice. Unlike the one-dimensional case, i t is not possible, using simple conditional probabilities, to derive an analytical expression for [w with a two-dimensional lattice. However, if the size and configuration of the ligand- lattice interaction is known precisely, a value for [ N I could be estimated by a simulation process in which the lattice surface is randomly sampled for available sites at any degree of lattice saturation. An alternative approach, in the absence of definitive information on the ligand-lattice interaction, is to consider adsorption only at low ligand concentrations where the probability of any two ligands excluding a third potential binding site is very low. Under these conditions, a relative equilibrium association constant, K,, can be deter- mined from the limiting slope of adsorption data plotted in the form 1/[B] uersus l/[Fl, according to Equation 6, "Exper- imental Procedures," so that the affinities of related ligands for a given cellulosic substrate can be compared.

The K, values estimated for the absorption of CenA and CBD PTc,,A by this method indicate that both ligands have comparable affinities for BMCC (Table I). This is consistent with the negligible adsorption of the isolated CenA catalytic domain p30 (Fig. 4) and is in contrast to T. reesei cellobioh- ydrolase I in which binding to cellulose appears to be mediated by both the CBD and the catalytic domain (11). A comparable K, value was also obtained for Cex (Table I). Small-angle x- ray scattering analysis shows that Cex has a size and shape comparable with CenA,' and semi-quantitative analysis has shown that, like CenA, adsorption to cellulose is mediated only by its CBD (5). CBDcex has approximately 50% sequence identity to CBDc,,A (2) and shows similar adsorption param- eters (Table I). It is noteworthy that although Cex was first recognized as a /3-1,4-glucanase, it has a 50-fold greater cata- lytic activity for xylosidic linkages (40). The significance of a CBD on a /3-1,4-xylanase is not yet clear.

Absolute values for K, (the equilibrium association con- stant) and a (the number of lattice sites occupied by a single ligand) can be determined if [No] is known, according to Equation 6. It is not possible to determine [No] precisely because of the inherent heterogeneity of cellulose prepara- tions, including BMCC. However, it can be approximated from available experimental data and used to determine whether the size of a is consistent with the dimensions of CBD. PTc,,A determined by small-angle x-ray scattering analysis (8, 10). Bacterial cellulose is synthesized as a long ribbon composed of 50-80 aggregated microfibrils (29). These ribbons are partially disrupted during BMCC preparation into bundles of microfibrils with cross-sections of the order of 40 nm ( l i 0 face) by 15 nm (110 face) (29, 31). Because T. reesei (41) and C. fimi3 cellulases appear to adsorb preferentially to the 110 face of algal cellulose crystals, it is appropriate to consider adsorption to the 110 face alone for equilibrium adsorption data obtained at low ligand concentrations. The repeating lattice unit (cellobiose residue) exposed at the 110 face has a spacing of 0.53 x 1.04 nm (31, 42). Given a density

N. R. Gilkes and M. Schmuck, unpublished results. B. Henrissat and N. R. Gilkes, unpublished results.

of 1.5 g/cm3 for crystalline cellulose (43), BMCC comprised of 40 X 15-nm microfibril bundles has approximately 101 pmol of 110 face lattice residues/g at maximal dispersion. Using this value of [No] to calculate a, we find that a single molecule of CBD.. T C e n A occupies approximately 39 110 face lattice residues (Table I). This is consistent with the dimen- sions of CBDC,,A determined by small-angle x-ray scattering analyses since the area occupied by one CBD molecule (26.25 nm2 when modelled as a triangle, base = 3.5 nm, height = 15 nm (8)) corresponds to about forty-eight lattice residues. Similar a values were calculated for CenA and Cex (Table I). The K, estimated for CBD.PTc,,A using this [No] value is 0.449 1.mol lattice residues"; the K,, values for CenA and Cex are in the same range (Table I). These values are consistent with the tight binding to cellulose shown by hybrid proteins in which CBDC,,A or CBDcex is fused to a heterologous protein by genetic manipulation (9, 20-22).

The 110 faces of BMCC comprised entirely of 40 X 15-nm microfibril bundles have a total capacity of 2.6 pmol of CBD . PTc,,A .gW1 when packing is optimized, given a = 39 lattice residue. This corresponds to approximately one-third of the observed amount of CBD.PTcenA bound as saturation is ap- proached, i.e. 2 8 pmol. g" (Fig.4). A further 5.6 pmol of CBD . PTc,,A could be accommodated on the 1 i O faces (surface area = 8.9 X lo5 cmZ.g") of these bundles for a total of 8.2 pmol. g-'. This value agrees closely with the observed value and indicates that both the 110 and 1 i O faces of the microfibril bundles must be filled at saturation. The pattern of adsorption of gold-labeled CBD.PTcenA to bacterial and algal cellulose seen at high ligand concentration3 supports this view. How- ever, optimal packing would not occur if ligands attach ran- domly to the cellulose surface, as proposed in the above model. This suggests that, as the cellulose surface becomes filled, some ligands bind by making only partial contact. The occur- rence of so-called "dangling ligands" (39) is plausible, since we have demonstrated that a truncated form of CenA, in which 64 amino acids are removed from the NH2 terminus of the CBD (i.e. 58% of the total CBD) still binds to cellulose, albeit with reduced affinity (7).

This discussion emphasizes the complexities inherent in the analysis of the adsorption of cellulases, or their isolated CBDs, to cellulose, even when the substrate surface is mod- elled as an idealized highly crystalline form. Theoretical analysis has shown that the binding of a large ligand to an array of overlapping potential binding sites on a one-dimen- sional substrate inevitably results in a nonlinear (concave upward) Scatchard ( [ B ] / F uersus[B]) data plot (39), and it is evident that similar considerations apply to a two-dimen- sional surface. Nonlinearity is compounded by the occurrence of two nonequivalent cellulose crystal faces, if their respective affinities differ; also by the probable occurrence of partial ligand binding and electrostatic repulsion between ligands as these surfaces become filled. It is apparent that previous adsorption experiments, all of which are further confounded by the use of very heterogeneous cellulosic substrates such as Avicel, comprised of both crystalline and paracrystalline re- gions, cannot be realistically analyzed by a simple one- or two-site model based on a Langmuir adsorption isotherm. Further refinement of the model presented here requires more precise information on the structure of the CBD and on the geometry of its interaction with the cellulose surface.

The functional significance of the adsorption to cellulose has not been determined, but it is noteworthy that other enzymes involved in the hydrolysis of insoluble polymeric carbohydrates (e.g. amylases and chitinases) also contain independent substrate binding domains (44,45). The mecha-

Adsorption of a Cellulase

nism of adsorption is also unknown but all C. fimi-type CBDs have a high content of aromatic amino acids, including a characteristic motif of four invariant tryptophan residues (7, 13). Other CBDs from bacteria and fungi, which show no obvious similarity to the C. firni-type in primary structure, are also rich in aromatic residues (11). Interactions between these residues and the cellulose surface may mediate adsorp- tion. Aromatic residues have been shown to play an important role in the binding of E. coli maltose binding protein to maltodextrins (46).

Acknowledgments-We thank Emily Kwan, Judy Alimonte, and David Nordquist for expert technical assistance. The enthusiastic support and stimulating discussion provided by Henri Chanzy is gratefully acknowledged.

REFERENCES 1. Gilkes, N. R., Langsford, M. L., Kilburn, D. G., Miller, R. C., Jr.,

and Warren, R. A. J. (1984) J. Biol. Chem. 2 5 9 , 10455-10459 2. Warren, R. A. J., Beck, C. F., Gilkes, N. R., Kilburn, D. G.,

Langsford, M. L., Miller, R. C., Jr., O'Neill, G. P., Scheufens, M., and Wong, W. K. R. (1986) Proteins 1,335-341

3. Wong, W. K. R., Gerhard, B., Guo, Z. M., Kilburn, D. G., Warren, R. A. J., and Miller, R. C., Jr. (1986) Gene (Amst.) 4 4 , 315- 324

4. Langsford, M. L., Gilkes, N. R., Singh, B., Moser, B., Miller, R. C., Jr., Warren, R. A. J., and Kilburn, D. G. (1987) FEBS Lett.

5. Gilkes, N. R., Warren, R. A. J., Miller, R. C., Jr., and Kilburn, D. G. (1988) J. Biol. Chem. 263,10401-10407

6. Guo, Z., Arfman, N., Ong, E., Gilkes, N. R., Kilburn, D. G., Warren, R. A. J., and Miller, R. C., Jr. (1988) FEMS Microbwl.

7. Gillres, N. R., Kilburn, D. G., Miller, R. C., Jr., and Warren, R. A. J. (1989) J. Biol. Chem. 264,17802-17808

8. Pilz, I., Schwarz, E. Kilburn, D. G., Miller, R. C., Jr., Warren, R. A. J., and Gilkes, N. R. (1990) Biochem. J . 2 7 1 , 277-280

9. Greenwood, J. M., Gilkes, N. R., Kilburn, D. G., Miller, R. C., Jr., and Warren, R. A. J. (1989) FEBS Lett. 244 , 127-131

10. Shen, H., Schmuck, M., Pilz, I., Gilkes, N. R., Kilburn, D. G., Miller, R. C., Jr., and Warren, R. A. J. (1991) J. Biol. Chem.

11. Tomme, P., van Tilbeurgh, H., Pettersson, G., van Dame, J., Vanderkerckhove, J., Knowles, J., Teeri, T., and Claeyssens, M. (1988) Eur. J. Biochem. 170 , 575-581

12. Abuja, P. M., Pilz, I., Claeyssens, M., and Tomme, P. (1988) Biochem. Biophys. Res. Commun. 156,180-185

13. Gilkes, N. R., Henrissat, B., Kilburn, D. G., Miller, R. C., Jr., and Warren, R. A. J. (1991) Microbiol. Reu. 5 5 , 303-315

14. Meinke, A., Braun, C., Gilkes, N. R., Kilburn, D. G., Miller, R. C., Jr., and Warren, R. A. J. (1991) J. Bacteriol. 173,308-314

15. Giorda, R., Ohmachi, T., Shaw, D. R., and Ennis, H. L. (1990) Biochemistry 29,7264-7269

16. McGavin, M., and Forsberg, C. W. (1989) J. Bacteriol. 171 ,

17. Jauris, S., Rucknagel, K. P., Schwarz, W. H., Kratzsch, P.,

226,163-167

Lett. 49 , 279-283

226,11335-11340

3310-3315

to Crystalline Cellulose 6749

Bronnenmeier, K., and Staudenbauer, W. L. (1990) Mol. & Gen. Genet. 223,258-267

18. Durrant, A. J., Hall, J. Hazlewood, G. P., and Gilbert, H. J. (1991) Biochem. J. 273 , 289-293

19. Py, B., Bortoli-German, I., Haiech, J., Chippaux, M., and Barras, F. (1991) Protein Eng. 4 , 325-333

20. Ong, E., Gilkes, N. R., Warren, R. A. J., Miller, R. C., Jr., and Kilburn, D. G. (1989) BioTechnology 7,604-607

21. Ong, E., Greenwood, J. M., Gilkes, N. R., Kilburn, D. G., Miller, R. C., Jr., and Warren, R. A. J. (1989) Trends Biotechnol. 7 ,

22. Ong, E., Gilkes, N. R., Miller, R. C., Jr., Warren, R. A. J., and Kilburn, D. G. (1991) Enzyme Microb. Technol. 13,59-65

23. Hestrin, S. (1963) in Methods in Carbohydrate Chemistry (Whis- tler, R. L., ed) Vol. 3, pp. 4-9, Academic Press, New York

24. Eppstein, L., Shevitz, J., Yang, X."., and Weiss, S. (1989) BioTechnology 7,1178-1181

25. Scopes, R. K. (1974) Anal. Biochem. 59,277-282 26. Cantor, C. R., and Schimmel, P. R. (1980) in Biophysical Chem-

istry, Part 2, pp. 380-381, W. H. Freeman, San Francisco 27. Henrissat, B., Vigny, B., Buleon, A., and Perez, S. (1988) FEBS

28. Skoog, B., and Wichman, A. (1986) Trends Anal. Chem. 5 , 82-

29. White. A. R.. and Brown. R. M.. Jr. (1981) Proc. Natl. Acad. Sci.

239-243

Lett. 231,177-182

83 , . ,

u. S . A. 78,1047-1051 30. Henrissat. B., and Chanzv. H. (1986) in Cellulose Structure.

Modification and Hydroljsis (Young, R. A., and Rowell, R. M.; eds) pp. 337-347, John Wiley and Sons, New York

31. Kuga, S., and Brown, R. M., Jr. (1987) J. Electron Microsc. Technol. 6,349-356

32. Kulshreshtha, A. K., and Dweltz, N. E. (1973) J. Polym. Sci. Polym. Phys. Ed. 11,487-497

33. Lee, S. B., Shin, S. H., Ryu, D. D. Y., and Mandels, M. (1982) Biotechnol. Bioeng. 24 , 2137-2153

34. Ooshima, H., Sakata, M., and Harano, Y. (1983) Biotechnol. Bioeng. 25,3103-3114

35. Marshall, K., and Sixsmith, D. (1974) Drug Deuel. Commun. 1 ,

36. Tomme. P.. Heriban. V.. and Claevssens. M. (1990) Biotechnol. 51-71

, I

Lett. 12,'525-530 I ~ I~~

37. StHhlbere. J.. Johansson. G.. and Pettersson. G. (1991) Bio- , ,

TechGbgy9, 286-290

(Creighton, T. E., ed) pp. 25-54, IRL Press, New York

, ,

38. Klotz, I. M. (1989) in Protein Function: A Practical Approach

39. McGhee, J. D., and von Hippel, P. H. (1974) J. Mol. Biol. 8 6 ,

40. Gilkes, N. R., Claeyssens, M., Aebersold, R., Henrissat, B., Meinke, A., Morrison, H. D., Kilburn, D. G., Warren, R. A. J., and Miller, R. C., Jr. (1991) Eur. J . Biochem. 202,367-377

41. Chanzy, H., Henrissat, B., andVuong, R. (1984) FEBSLett. 172 ,

42. Gardner, K. H., and Blackwell, J. (1974) Biopolymers 13, 1975-

43. Meyer, K. H., and Misch, L. (1937) Helu. Chin. Acta 20 , 232-

44. Svensson, B. H., Jespersen, H., Sierks, M. R., and MacGregor,

45. Kuranda, M. J., and Robbins, P. W. (1991) J. Bwl. Chem. 266 ,

46. Spurlino, J. C., Lu, G.-Y., and Quiocho, F. A. (1991) J. Biol.

469-489

193-197

2001

244

A. (1989) Biochem. J. 264,309-311

19758-19767

Chem. 266,5202-5219