INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Thermomonospora fusca is a filamentous soil bacterium that degrades most plant cell wall polymers, including cellulose. T. fusca secretes six different cellulases, including two exocellulases (E3 and E6), three endocellulases (E1, E2, and E5), and one unusual endocellulase (E4). E4 is a 90.4-kDa protein and consists of four domains: an N-terminal 51.4-kDa family 9 catalytic domain, a family IIIc cellulose binding domain (CBD), a fibronectin III-like domain, and a C-terminal family II CBD (Fig. 1). The gene for E4 has been sequenced (17), and the corrected sequence is in GenBank (accession no. L20093). E4 is unique in that it has relatively high activity on bacterial microcrystalline cellulose (BMCC) and synergizes with both classes of exocellulases and with endocellulases E2 and E5 (16). The other endocellulases from T. fusca do not synergize with each other (16). E4 retains more than 70% of its activity from pH 4.7 to pH 10.1. 

View larger version (29K): [in this window] [in a new window]   FIG. 1.   Schematic diagrams of the E4 domain combinations used in this study. aa, amino acids.

The crystal structure of a 68-kDa fragment of E4 (E4-68) has been solved (22) and consists of an (/)₆ barrel catalytic domain joined with only limited flexibility to a family IIIc CBD, which has an antiparallel -sandwich fold. These domains interact in two loop regions so that the shallow catalytic cleft is aligned with the flat face of the family IIIc CBD in such a way that a cellulose strand could bind along both domains (22). Tormo et al. (24) have determined the crystal structure of the Clostridium thermocellum family IIIa CBD from the cellulosomal scaffoldin subunit and proposed that it is this flat surface which interacts with cellulose. The structure of the catalytic domain of a homologous family 9 protein, C. thermocellum CelD, has been solved and includes a small N-terminal domain that is not present in E4 (18). The role of this domain is unknown, and it is not aligned with the active cleft of the catalytic domain.

The enzyme-product complex structure observed after the soaking of cellopentaose (CP) into E4-68 crystals was modeled as two overlapping binding modes of cleaved CP, each at 50% occupancy (22). These complexes identified the active center of the enzyme and showed the locations of six glucosyl binding sites numbered, from the nonreducing end, 4 to +2. Family 9 enzymes cleave cellulose with inversion at the anomeric carbon (11). Glu424 was identified as the active acidic residue, while Asp55 and Asp58 could both act as a base to deprotonate the nucleophilic water. All of these residues are conserved in family 9 proteins (17). The observed cleavage product clearly showed the  configuration of the anomeric carbon at the point of hydrolysis (22). The crystal structure also showed an induced fit, a closing of amino acid residues over the 3, 2, and 1 glucosyl binding sites, which was dependent on the occupancy of the +1 site (22). The open conformation appears to be complementary to a linear (unkinked) substrate, while the closed conformation shifts Glu424 within hydrogen bonding distance of the scissile oxygen and forces the substrate [especially Glc(2) and Glc(1)] toward the nucleophilic water (22).

In the work described here, we have studied the activities and properties of four combinations of the E4 domains, as well as the Asp55Cys mutants of two of these constructs. These data are discussed, along with previous data, in an effort to understand the mechanism of cellulose degradation by E4.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Cloning of E4-68, E4-68 Asp55Cys, E4-51, E4-51 Asp55Cys, and E4-74. Standard nucleic acid techniques were used as described by Sambrook et al. (23). DNA sequence analysis was performed by the Cornell Biotechnology Resource Center using a Perkin Elmer/Applied Biosystems automated sequencer. The strains used were Escherichia coli DH5 (13) and Streptomyces lividans TKM31 (19).

Production of E4 proteins from S. lividans. Each S. lividans transformant was grown and harvested as previously described (26), except that cultures were harvested after only 41 to 48 h. The supernatant, containing 1.2 M (NH₄)₂SO₄, was loaded onto a CL-4B phenyl-Sepharose column and eluted as previously described (26). Only the fractions containing the desired protein (as judged by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and activity assays) were combined and applied to a Q-Sepharose column. E4 contamination by endogenous S. lividans endocellulase (26) was avoided by using a short fermentation time and carefully selecting the phenyl-Sepharose fractions to be combined. Elution from Q Sepharose was with 0.01 M bis-Tris (pH 5.4)-10% glycerol buffer containing NaCl gradients (0.18 to 0.4 M for E4-68, 0.18 to 0.5 M for E4-74, 0 to 0.4 M for E4-51, and 0.2 to 0.6 M for E4-90). The purest fractions for each protein were combined and concentrated as previously described (22) and stored in buffer containing 10% glycerol.

Protein concentration and cellulase activity assays. Protein concentrations were determined by measuring A₂₈₀ by using extinction coefficients calculated from the predicted amino acid compositions. Filter paper (FP; Whatman no. 1; 8.7 mg/ml), BMCC (Cellulon; Weyerhaeuser; 2.5 mg/ml), acid-swollen cellulose (SW; 2 mg/ml) (8), and carboxymethyl cellulose (CMC; Sigma Low Viscosity; 10 mg/ml) assays were performed as previously described (16), in 0.4-ml volumes at 50°C for 16, 16, 2, and 2 h, respectively. The reducing sugar produced was determined by the dinitrosalicylic acid method (12, 20). The buffer was 0.05 M Na acetate (pH 5.5) with 15 mM CaCl₂ added for filter paper reactions. All assays were done in triplicate by using increasing amounts of the E4 proteins. Cellulase synergism and the distribution of reducing ends between filter paper and supernatant were measured as described by Irwin et al. (16).

Calculation of cellulase activity. Cellulase activity assays are nonlinear, at least partly due to substrate heterogeneity. To compare the activities of the different E4 proteins, we assayed them for a fixed amount of time by using various amounts of protein and calculated the activities at a fixed amount of reducing sugar produced. The assays were set up to produce between 0.1 and 0.6 µmol of reducing sugar per reaction. The dinitrosalicylic acid standard curve was made by using cellobiose as the reducing sugar. The nanomoles of protein used (X) versus the micromoles of reducing sugar (Y) were plotted and fitted to the equation Y = m1 · X/(m2 + X) by using the program KaleidaGraph (Synergy Software) (m1 and m2 are constants determined from the curve fit). This curve was used to determine the amount of enzyme required to produce the arbitrary value of 0.424 µmol of reducing sugar in the reaction (or 1.06 µmol/ml). If this target amount of product could not be produced, then the activity was calculated from the micromoles of cellobiose produced by 0.6 nmol of enzyme (1.5 nmol/ml).

Binding assays. Binding assays were done in siliconized 1.5-ml Eppendorf tubes by using 0.05 M Na acetate buffer (pH 5.5)-10% glycerol. Binding assays contained 0 to 3 mg of BMCC per ml and 1 nmol of the desired E4 protein per ml. Tubes were rotated end over end at room temperature for 1 h and centrifuged for 5 min. The supernatants were filtered through 0.45-µm-pore-size CA micro-spin filters (Lida) which had been pretreated with 300 µl of bovine serum albumin at 1 mg/ml and then rinsed with buffer. The amount of E4 remaining unbound was determined by CMC assays or by protein concentration (A₂₈₀).

Viscosity assays. Viscosity assays were done by using a size 100 Ostwald-Fenske viscometer at 50°C in 0.05 M Na acetate buffer (pH 5.5). Enzyme (100 µl) was added to 10 ml of 0.3% Hercules CMC 4H1F. The time of outflow was measured at intervals. Samples for reducing sugar assays were assayed by the bicinchoninate method (7, 10).

Identification of hydrolysis products by TLC. E4 proteins were reacted with SW (6.3 mg/ml) in 12.5 mM Na acetate (pH 5.5) at 50°C in a total volume of 1.2 ml. After preincubation for 5 min, enzyme was added and mixed and an aliquot was withdrawn immediately and frozen on dry ice; further aliquots were withdrawn at the time points indicated (see Fig. 3) and frozen. The aliquots were analyzed by thin-layer chromatography (TLC) as previously described (5, 17).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Effect of each CBD on activity and thermostability. The different E4 domain constructs are diagrammed in Fig. 1, and their specific activities on different cellulose substrates are given in Table 1. E4-90 has the highest activity of any T. fusca cellulase on BMCC, while the two constructs which lack the family IIIc CBD, E4-74, and E4-51, have almost no activity on BMCC. Only E4-90 could reach the target of 4.5% digestion on FP, illustrating that both the family II and IIIc CBDs are necessary for degradation of this crystalline substrate. The other constructs had only minimal FP activity, which did not increase with the addition of further enzyme. The CMC activities show that the family II CBD is not required but there is a fourfold decrease in the activity of the proteins lacking the family IIIc CBD. E4-51, which has no binding domain at all, has only 3% of the activity of E4-90 on SW. In comparison, the catalytic domains of two other T. fusca endocellulases, E2 and E5, retain 62 and 95% of their SW activity, respectively. The E4-68 Asp55Cys mutant has low, but not zero, activity, which agrees well with its proposed role as one of the two catalytic bases. The activities of E4-90 reported here are substantially higher than those reported earlier (16); this may be due to the slightly different purification methods and the use of 10% glycerol in the final purification buffers.

Processivity. Although we do not have an assay to measure processivity directly, one indication of the degree of processivity is to compare the ratio of the reducing sugar ends in the supernatant (soluble) to the reducing sugar ends left in a filter paper circle (insoluble). We have previously shown that there is a clear distinction between endo- and exocellulases (16) in that the soluble/insoluble ratio is about 2.2 for endocellulases E2 and E5 and 12 to 23 for exocellulases (16). E4-90 is unusual in that its ratio falls between values for exo- and endocellulases (Table 1). The soluble/insoluble ratio of the E4 constructs decreases with the removal of the family II CBD and decreases much more with the removal of the family IIIc CBD. The effect is most dramatic when both CBDs are removed, giving 63% insoluble products.

Reduction of the viscosity of a CMC solution. E4-51 and E4-74 required a fivefold higher enzyme concentration than did E4-90 or E4-68 to reduce the viscosity of a CMC solution (Fig. 2A). E4 is a much less active endocellulase than E5 in viscosity experiments but is as active as E2. The specific fluidity as a function of the amount of reducing sugar produced showed no significant difference between the different E4 constructions (Fig. 2B).

View larger version (22K): [in this window] [in a new window]   FIG. 2.   Viscosity reduction of CMC. Time of flow of buffer is t0; time of flow of CMC solution is t. The concentration of E4-90, E4-68, E2, and E5 was 0.005 nmol/ml, and the concentration of E4-74 and E4-51 was 0.025 nmol/ml.

Substrate binding in the catalytic cleft and analysis of products. TLC analysis of the products from a time course of hydrolysis of SW by E4-90 is shown in Fig. 3. This experiment clearly shows that the initial cleavage of SW produces cellotetraose (CTet) almost exclusively. This same TLC pattern was seen for E4-68, while for E4-51, the activity was so low that only CTet could be seen on the TLC plate (data not shown). Over time, CTet is cleaved to cellobiose or to cellotriose and glucose in roughly equal amounts and does not accumulate (as previously shown [17]). Barr et al. (1) showed that CP labeled with ¹⁸O at the reducing-end anomeric carbon was cleaved by E4 with the ¹⁸O distributed as follows: 11% cellotriose and 89% cellobiose or 20% CTet and 80% glucose. This fits well with the crystallographic data showing that CP binds in the E4 cleft a large majority of the time in one of two modes: in 3 to +2 or in 4 to +1. To form the low-frequency products, about 5% of the time CP binds with its nonreducing end in the 2 position and 10% of the time with its nonreducing end in the 1 position. Since the catalytic domain only has sites 4 to +2, this suggests there are some binding interactions on the flat face of the family IIIc CBD which is aligned with the catalytic cleft.

View larger version (140K): [in this window] [in a new window]   FIG. 3.   Time course of products released by E4-90 digestion of SW. CB, cellobiose; Ctri, cellotriose.

Cellulose binding properties and effect of an Asp55Cys mutation. The binding of the different constructs to BMCC is shown in Fig. 4. E4-90 and E4-74 bind equally well to BMCC, and yet E4-74 has very little activity on FP or BMCC. Surprisingly, E4-68 does not bind to BMCC, even though it retains 32% of the E4-90 BMCC activity. In contrast, the E4-68 Asp55Cys mutant demonstrates significant binding to BMCC but E4-51 Asp55Cys does not bind at all. The inability of E4-68 to bind to BMCC suggests that once a strand has been hydrolyzed, its binding becomes very weak. A rationale for this result could be that tight binding is not desirable if the role of the family IIIc CBD is to guide the cellulose chain toward the catalytic cleft to aid in processivity.

View larger version (22K): [in this window] [in a new window]   FIG. 4.   Binding of the E4 constructs to BMCC.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

Gal et al. (9) have studied C. cellulolyticum CelG, which contains a family 9 catalytic domain and a family IIIc CBD, as well as a dockerin domain specific to cellulosomal hydrolases. The amino acids of the catalytic (CelGcat) and CBD (GST-CBDCelG) domains show 51 and 39% similarity to E4. Constructs of these two domains alone and also of CelGS, which is missing only the dockerin domain, were made. Neither CBDCelG nor CelGcat was able to bind to SW, Avicel, or BMCC, which is in agreement with the results presented here. CelG and CelGS were able to bind to Avicel but had weaker binding than the family IIIa CBD of miniCipC₁ (21).

The C. thermocellum scaffoldin family IIIa Cip-CBD (24) serves to bind the cellulosome to cellulose, and only one CBD has been found in any scaffoldin. An alignment based on the three-dimensional crystal structures of the CipA CBD and the E4 family IIIc CBD is shown in Fig. 5. Yaron et al. (25) have mutated to Ala 10 residues of CipA considered likely to affect cellulose binding. Mutation of any one of the residues participating in the planar aromatic strip (Asp56, His57, Tyr67, Arg112, and Trp118) decreased binding significantly (3- to 10-fold). On the other hand, mutation of the "anchoring" residues (Asn10, Asn16, Gln110, Ser12, and Ser133), which are proposed to participate in hydrogen bonding to cellulose, did not reduce the affinity significantly. The Cip CBD aromatic strip residues are not conserved in the E4 family IIIc CBD. The E4 family IIIc CBD conserved residues suggested by Sakon et al. (22) to be well aligned to interact with a cellulose chain leading into the E4 active cleft are also noted in Fig. 5, and only three of the seven are somewhat conserved in CipA. While the CipA scaffoldin CBD has a relatively lengthy 11-residue beta 4 strand (Gln51-Ile61) terminating in a hairpin turn including a three-residue beta 4' strand (24), E4 has only a three-residue beta 4 strand (Thr514-Ser516) (22). Although these two molecules have similar overall topologies, the family IIIc CBDs clearly play a role very different from that of the family IIIa CBDs.

View larger version (29K): [in this window] [in a new window]   FIG. 5.   Alignment of the E4 family IIIc CBD with the CipA scaffoldin CBD based on superposition of the three-dimensional structures (22, 24) by the strategy of Chothia and Lesk (6) using a 3Å cutoff to define structurally equivalent residues (denoted by black lines between the sequences). Residues where the structures diverge in their C placement are bracketed to emphasize that the alignment in these regions is not meaningful. Gaps are indicated by hyphens. The  strands are labeled above or below each sequence. The E4 residues which are double underlined, Asn470, Glu478, Lys480, Arg557, Glu559, Gln561, and Arg563, are conserved in family IIIc and lie in a region (colored red in Fig. 5a of reference 22) that could interact well with an extended cellulose chain from the active site. CipA residues which were mutated to Ala by Yaron et al. (25) are double underlined, and those which resulted in 3- to 10-fold lower binding to cellulose are marked by arrows. Residues marked by asterisks have the same C orientation, but the directions of their side chains are markedly different; the CipA Arg112 side chain is involved in a salt bridge with an Asp, while the corresponding E4 Arg563 is not. Bayer et al. (2) have identified six well-conserved residues of family III CBDs that form a shallow groove on the opposite side of the molecule from the "binding" residues; these are represented by vertical lines between the two sequences.

Based on the results presented in this paper and the crystal structure, we hypothesize that after E4 makes an initial cleavage (either endo or exo), the cellulose strand is processed, from the nonreducing end, with the help of the family IIIc binding domain to refill the six glucosyl sites of the catalytic cleft and then CTet is cleaved from the nonreducing end. E4 is probably able to give synergy with exocellulases because it can bind and cleave internal sites, while it is able to give synergy with other endocellulases because of its processive activity. Analysis of the activities of the various E4 constructs on different substrates provides clues to the roles of the different domains. The constructs without the family IIIc CBD are only 25% as efficient in CMC hydrolysis. There is no evidence from the viscosity experiments that CMC hydrolysis is processive. Probably, the carboxymethyl groups of CMC prevent processivity. This suggests that one role of this domain is simply to help attract a CMC chain to the vicinity of the catalytic cleft.

Pairwise comparisons of SW hydrolysis by proteins with a family IIIc CBD versus those without (E4-90 versus E4-74 and E4-68 versus E4-51) show 8.8- and 8.6-fold differences in activity, respectively. Similar comparisons of proteins with and without the family II CBD (E4-90 versus E4-68 and E4-74 versus E4-51) show a 3.7-fold decrease in both cases. Loss of both CBDs (E4-90 versus E4-51) results in a 32-fold difference, illustrating that the effects of the two domains are multiplicative.

Constructs containing the family IIIc CBD produce much more soluble product after filter paper hydrolysis than do those without this domain. This implies that E4 processivity requires this domain. After hydrolysis, the E4 residues shown to close over the substrate (22) move back to their open position, and the strand could directly dissociate from E4. The function of the type IIIc CBD, in part, must be to hold the enzyme onto the strand during this event and lessen the chance of dissociation. Yet, this CBD must also allow for enzyme migration along the strand, for the cellulose chain must travel the distance of four glucosyl residues to fully occupy the catalytic cleft before the next cleavage.

It is clear that CTet is readily cleaved further, since it does not accumulate. The CTet molecules would also compete with cellulose chains for the catalytic cleft. On the other hand, the cellulose chains would have many more potential (albeit weak) binding sites along the family IIIc CBD, making their binding in the cleft more favorable. CP is cleaved far more rapidly than CTet; in separate reactions with the oligosaccharides, after 15 min, there was no remaining CP in the reaction with E4, while after 2 h, about one-third of the CTet was not yet hydrolyzed (17).

The BMCC activities demonstrate that the family IIIc CBD plays a role in disrupting crystalline cellulose, as the activities increase 40- to 66-fold when this domain is present. BMCC, although crystalline, has a much more open structure than Avicel (3, 4), which would allow a relatively large protein such as E4 better access to this substrate. In contrast, hydrolysis of filter paper requires both binding domains. The liberation of a crystalline cellulose chain from its neighbors to make it available for hydrolysis requires time for a number of hydrophobic and hydrogen bonds to break and reform. It makes sense that having an anchor (the family II CBD) to keep the enzyme close to the solid substrate would be advantageous.

These experiments show the power of combining structural studies with enzymology and DNA techniques. They also illustrate that the more we learn about how enzymes function the more complicated and interesting the story becomes.