D. B. Wilson, Department of Molecular Biology and Genetics, Cornell University, Ithaca, NY 14853, USA Fax: +1 607 255 2428 Tel: +1 607 255 5706 E-mail: firstname.lastname@example.org
Thermobifida fusca exocellulase Cel6B acts by an inverting hydrolysis mechanism; however, the catalytic acid and base residues for this enzyme have not been confirmed. Site-directed mutagenesis and kinetic studies were used to show that Asp274 is the catalytic acid, which is consistent with what is found for other members of family-6 glycoside hydrolases; however, a single catalytic base was not identified. Mutation of all putative catalytic base residues, within 6 Å of the −1/+1 glucose subsites, including the highly conserved Asp226, Asp497 and Glu495, as well as Ser232 and Tyr220, did not reveal a catalytic base, although these residues are all important for activity. We propose a novel hydrolysis mechanism for T. fusca Cel6B involving a proton-transferring network to carry out the catalytic base function.
Thermobifida fusca is an aerobic, filamentous soil bacterium (Actinomycetaceae) that utilizes cellulose as its sole carbon source by producing a mixture of functionally distinct cellulases, which act synergistically. Among them, Cel6B is very important for achieving the maximum activity of synergistic mixtures, although its activity alone is relatively weak on all polysaccharide substrates . T. fusca Cel6B (TfCel6B) is an exocellulase (EC 18.104.22.168) that processively hydrolyzes β-1,4-glycosidic bonds from the nonreducing end of cellulose molecules. The enzyme has higher thermostability and a broader pH optimum than the homologous fungal exocellulase Trichoderma reesei (also known as Hypocrea jecorina) Cel6A (TrCel6A), which is present in most commercial cellulase preparations .
TfCel6B contains two functional domains: a C-terminal family-6 catalytic domain, and an N-terminal family-2a carbohydrate-binding module to which it is linked through a Pro-Ser-rich linker . The three-dimensional structures of the catalytic domains of five family-6 cellulases have been determined. Humicola insolens Cel6A  and TrCel6A  are exocellulases, whereas T. fusca Cel6A (TfCel6A) , H. insolens Cel6B (HiCel6B)  and Mycobacterium tuberculosis Cel6  are endocellulases. The active sites of the exocellulases are enclosed by two long loops forming a tunnel, whereas the corresponding loops in the endocellulases are shorter, opening the active sites. Increasing knowledge of cellulase structures and improvements in modeling software  have greatly facilitated rational protein design for the study of the mechanism of cellulases.
TfCel6B belongs to glycoside hydrolase (GH) family 6; members of this family were shown to catalyze the hydrolysis of cellulose with inversion of the anomeric configuration . In the proposed inverting mechanism, a catalytic base such as a deprotonated Asp or Glu removes a proton from a water molecule, making it a better nucleophile for direct attack on the anomeric C1, breaking the covalent bond between C1 and the glycosidic oxygen, and inverting the linkage from a β to an α configuration, while a catalytic acid assists cleavage by donating a proton to the leaving glycosidic oxygen . However, the actual detailed hydrolysis mechanism for GH-6 enzymes, particularly the existence of a catalytic base, is still in doubt. On the basis of site-directed mutagenesis, Asp392 in Cellulomonas fimi Cel6A, corresponding to TrCel6A Asp401, H. insolens Cel6A (HiCel6A) Asp405, HiCel6B Asp316, TfCel6A Asp265, and TfCel6B Asp497 (Table 1), was proposed to be a classic Brønsted base . However, crystallographic and kinetic studies in TrCel6A suggested that Asp175, not Asp401, was the catalytic base . The HiCel6A D405A and D405N enzymes still retained approximately 0.5–1% activity . Although mutation of Asp316 in HiCel6B to Ala or Asn led to an inactive enzyme , the three-dimensional structure determination showed that Asp316 could only be correctly positioned to act as a base if a conformational rearrangement of the −1 subsite sugar ring occurs . In TfCel6A, Asp265 was not directly involved in hydrolysis, but participated in substrate binding . Therefore, it is interesting to investigate potential catalytic residues in TfCel6B.
Table 1. Amino acids chosen for mutation. The gene sequences were aligned and analyzed using megalign (DNASTAR lasergene).
Glycosyl subsite location
Corresponding residue in:
Increase in pKa
Catalytic base/substrate binding
Recently, activity rescue of catalytic mutants by sodium azide has been demonstrated to be a useful tool for identification of the catalytic base in both retaining  and inverting GHs . This approach distinguished the actual catalytic base from other catalytic residues in the inverting endocellulase T. fusca Cel9A .
We report here the construction and characterization of enzymes with mutations of six highly conserved residues in the active site cleft to investigate their role in the hydrolytic mechanism of TfCel6B.
Cel6B structural model
To choose residues for mutation, a structural model of the Cel6B catalytic domain was built on the basis of the X-ray structures of the HiCel6A catalytic domain (1OCB.pdb) and the TrCel6A catalytic domain (1QK2.pdb), using swiss-model workspace. The reliability of the model was evaluated by the whatcheck program to check a battery of physicochemical constraints . The energy minimization was computed by swiss-model workspace , and the final total energy of the model was −3641 kJ·mol−1.
Selection of amino acids for mutation
All highly conserved Asp and Glu residues, including Asp226, Asp274, Asp497, and Glu495, which are approximately 6 Å away from the −1 and +1 glucose subsites, were mutated to Ala. Figure 1 shows the position of potential Cel6B catalytic residues, and Table 1 shows the corresponding residues in four other family-6 GHs. Asp274 was expected to be the catalytic acid, as the corresponding residue in several family-6 cellulases has been shown to be the catalytic acid [11,13,18]. Asp226 corresponds to a residue in TrCel6A that forms a carboxyl–carboxylate pair with the catalytic acid to increase its pKa . Asp497 is a candidate for a catalytic base, as it is located in the −1 subsite, almost opposite to the putative catalytic acid Asp274. As the structures of exocellulases are known to be somewhat flexible , Glu495, in the −3 subsite, could also be a catalytic base. Ser232 was chosen, as it is near the −1 subsite, hydrogen bonds to Asp226, and therefore might participate in a proton-transferring network, as has been postulated for TrCel6A . The residue corresponding to Tyr220 in TfCel6A (Tyr73) was found to be essential for hydrolysis [21,22]. In retaining enzymes of GH families 33, 34, and 83, a Tyr was showed to act as a catalytic nucleophile .
All mutant enzymes were expressed in the pET plasmid with a yield of 10–12 mg·L−1, and they all behaved similarly to wild-type Cel6B during purification. CD spectra of all mutant enzymes were identical to that of the wild type (data not shown), indicating that the global secondary structure of the mutant proteins remained intact. Mutant enzymes were assayed on bacterial microcrystalline cellulose (BMCC), phosphoric acid-treated swollen cellulose (SC), phosphoric acid-treated cotton (PC), and carboxymethyl cellulose (CMC).
The D274A enzyme could not achieve the target digestion on any polysaccharide substrate, suggesting that Asp274 is essential for catalysis (Table 2). The substrate 2,4-dinitrophenyl-β-d-cellobioside (2,4-DNPC) has an excellent leaving group, which does not require a catalytic acid, so mutation of the catalytic acid does not prevent activity against 2,4-DNPC . The data from the 2,4-DNPC assays fit the Michaelis–Menten equation well. The D274A enzyme hydrolyzed 2,4-DNPC more effectively than the wild type (Table 3).
Table 2. Polysaccharide activity and ligand binding of the TfCel6B enzymes. Activity was calculated at 6% digestion for BMCC, SC and PC, and 1.5% digestion for CMC; the average coefficients of variation were 4, 5, 5.5 and 2.5 for BMCC, SC, PC and CMC, respectively. Kd was determined by fluorescence titration of 53.7 μm enzyme to 1.7 μm of MUG2. ND, not detected.
Activity (μmol cellobiose· min−1·μmol−1 enzyme)
Kd (μm) for MUG2 (× 10−2)
a Target digestion could not be achieved; activity was calculated at 1.5 μm enzyme. b Value is approximate, as titration curve did not fit well, owing to poor binding.
3.6 ± 0.3
57 ± 4
25 ± 1
3.2 ± 0.3
1.5 ± 0.2
∼ 13 000b
73 ± 2
Table 3. 2,4-DNPC kinetics of TfCel6B wild-type and mutant enzymes; 2,4-DNPC at initial concentrations of 20–600 μm was hydrolyzed by 1.5 μm enzyme.
kcat/Km (min−1·μm−1) (× 10−3)
0.34 ± 0.06
2.3 ± 1.9
146 ± 122
0.09 ± 0.04
161 ± 37
0.56 ± 0.28
0.03 ± 0.01
44 ± 13
0.68 ± 0.30
0.11 ± 0.05
6.5 ± 1.3
16.9 ± 8.4
2.26 ± 0.14
1.5 ± 0.2
1507 ± 186
0.006 ± 0.001
214 ± 65
0.03 ± 0.01
The D274A enzyme and its catalytic domain bound more tightly to both BMCC and SC, respectively, than the wild-type enzymes (Fig. 2); the percentage of the D274A enzyme bound to BMCC was about 75%, whereas only 50% of the wild type was bound. A previous study  showed that the fluorescence emission of 4-methylumbelliferyl ligands was strongly quenched upon binding to Cel6B and that the enzyme did not hydrolyze these ligands; therefore, this approach could be used to investigate the ligand-binding affinity of the active sites. Fluorescence titration showed that the D274A enzyme bound 4-methylumbelliferyl β-cellobioside (MUG2) to approximately the same extent as the wild type (Table 2).
The D226A enzyme had very low activity on insoluble cellulose, particularly on BMCC and SC (Table 2). TLC analysis indicated that the wild type completely hydrolyzed cellotetraose, cellopentaose and cellohexaose within 20 min, whereas only traces of products were produced by the D226A enzyme after 16 h (data not shown). Binding to BMCC and SC of the D226A enzyme was higher than that of the wild type (data not shown), and the Kd of the D226A enzyme for MUG2 was only slightly reduced (Table 2), suggesting that activity loss was not caused by loss of substrate binding.
In the absence of a catalytic base, an exogenous nucleophile such as sodium azide can partially rescue enzyme activity . The activity of the D226A enzyme on SC was not improved by sodium azide, even at 2 m (data not shown).
Surprisingly, the D226A enzyme had slightly higher activity than the wild type on CMC. The mutant enzyme also reduced the viscosity of a CMC solution faster than the wild type, although the decrease was much lower than that seen with a typical endocellulase (Fig. 3). TLC analysis and MALDI-TOF MS analysis of the CMC digestion products of the D226A enzyme did not reveal cellobiose, the major product of the wild type, or carboxymethyl cellobiose (Fig. 4). The MALDI-TOF MS spectra showed cellotriose, cellotetraose, cellopentaose, cellohexaose, and their carboxymethyl derivatives. To investigate the production of insoluble reducing sugars from CMC, TLC bands corresponding to the loading spot, cellotriose and cellobiose were eluted, and reducing sugars from each fraction were measured. The majority of reducing sugars produced by the D226A enzyme were found at the loading spot, whereas the wild type produced primarily cellobiose (data not shown). The wild-type and mutant enzymes were assayed on hydroxyethyl cellulose (HEC), which does not contain charged groups as does CMC. The D226A enzyme had several-fold higher HEC activity than the wild type (data not shown). CMC-native gels showed that the D226A enzyme bound CMC as tightly as the wild type (data not shown).
Asp497, Glu495, Ser232 and Tyr220 mutation
The D497A and E495A enzymes showed significantly reduced activity on BMCC and showed smaller reductions on the other substrates (Table 2). The Kd for MUG2 of these enzymes decreased significantly (Table 2). The activity of the D497A enzyme on 2,4-DNPC was nearly 60-fold lower than that of the wild type, whereas its Km was over 90-fold higher (Table 3).
The S232A enzyme retained near wild-type activity on most substrates, but CMC activity was drastically reduced (Table 2). The HEC activity of the S232A enzyme was also lower than that of the wild type (data not shown).
The Y220A enzyme could not reach target digestion on either BMCC or SC, and no PC or CMC activity was detected (Table 2). The enzyme showed a slightly lower Kd for MUG2 than the wild type (Table 2), indicating good binding. However, the kcat of the Y220A enzyme on 2,4-DNPC was approximately 26% of that of the wild type and the Km was increased 70-fold (Table 3).
The activity of none of these four mutant enzymes was rescued by sodium azide (data not shown). To test whether any mutation caused a change in the pKa of the catalytic acid, eliminating activity rescue by sodium azide, the mutant enzymes, except for the Y220A enzyme, owing to its extremely low activity, were normalized by activity at pH 5.5 and assayed for PC activity for 16 h over the pH range from 2 to 12. None of the pH profiles showed a significant difference from the wild type (Fig. 5).
D226A/S232A double mutation
The double mutation knocked out activity on all polysaccharides and slightly decreased ligand binding (Table 2). Binding to BMCC and SC by the mutant enzyme was similar to that of the wild type (data not shown). Excitingly, the CMC activity of the mutant enzyme was partially rescued at low concentrations of sodium azide (Fig. 6).
The results of this study show the essential roles of Asp274 and Tyr220, as mutation of either residue resulted in nearly inactive TfCel6B. Asp274 functions as the catalytic acid, as the D274A mutation increased activity on 2,4-DNPC, which does not require a catalytic acid. A drastic increase in the Km for 2,4-DNPC and a slightly lower Kd for MUG2 support a role for Tyr220 in distortion of the glycosyl unit in subsite −1 rather than in simple binding. The Tyr220 equivalents, Tyr73 and Tyr169, in TfCel6A and TrCel6A, respectively, have been shown to cause substrate distortion in the −1 subsite [22,25,26]. Structures of family-6 GHs [18,27] show no direct hydrogen bond between the residues corresponding to TfCel6B Tyr220 and bound substrates.
Absence of a single catalytic base
TfCel6B lacks a classic Brønsted base, as none of the single mutations in any Asp or Glu residue (D226A, D497A, and E495A), which are within 6 Å of the −1 and +1 subsites, abolished activity on all polysaccharide substrates, and none of the mutants showed activity rescue by sodium azide. Similarly, azide rescue assays on SC for TfCel6A mutations, including all four highly conserved Asp residues (Asp79, Asp117, Asp156, and Asp265), did not show activity rescue (unpublished data). It should be noted that TfCel6A is an inverting endocellulase with short loops, providing a more open active site cleft for substrates and sodium azide; additionally, TfCel6A showed nearly 300-fold higher SC activity than TfCel6B , so even a subtle increase in TfCel6A activity on SC caused by sodium azide could be easily detected.
The retention of nearly 90% of wild-type activity on PC eliminates Asp497 as a Brønsted base, which was suggested for the corresponding Asp392 in C. fimi Cel6A (CfCel6A) . This result is also consistent with the elimination of Asp401 as a catalytic base in TrCel6A . The drastic decrease in 2,4-DNPC and MUG2 binding to the D497A enzyme supports the role of Asp497 in substrate binding, as seen for the TfCel6A D265A mutant . The carboxylate group of TrCel6A Asp401 was seen to interact with the O3 hydroxyl of the glucosyl unit in the −1 subsite, and loss of this interaction might account for decreased binding .
Analysis of the TrCel6A structure indicated that the residue corresponding to Cel6B Glu495 is a key sugar-binding residue . The E495A enzyme bound weakly to MUG2, showing the importance of the hydrogen bonds between Glu495 and the sugar hydroxyl group in the −3 subsite. When the residue was replaced with Asn or Asp, BMCC activity was partially retained .
The only published evidence for a catalytic base in GH family 6 is the loss of activity of the CfCel6A D392A enzyme ; however, sodium azide rescue and substrate binding were not reported for the D392A enzyme, and there was no direct evidence for the correct folding of this enzyme.
The activities of the D226A and S232A enzymes were substrate-specific, in that they reached target digestion on only certain substrates. Although this finding eliminates these residues as single catalytic bases, it does not exclude the participation of these residues in the activation of the catalytic water molecule via a proton-transferring network, which acts as the catalytic base. Depending on the structure of substrates and the position of the network components, their significance in the network for hydrolysis might vary. When both residues were mutated, the network could not function, causing activity loss on all substrates. The success of sodium azide rescue for the double mutant enzyme, but not for the single mutant enzymes, further supports the model where a network of Asp226 and Ser232 acts as the catalytic base. Reducing sugars would not be measured by dinitrosalicylic acid (DNS) if azide adducts were formed, suggesting that the azide ion activates a water molecule, which performs the hydrolysis. On the basis of structural analysis, Asp175 in TrCel6A was suggested to act as the catalytic base via a water chain between Asp175 and Ser181 (Asp226 and Ser232 in TfCel6B, respectively) .
The enhancement of water nucleophility by these two residues may be indirectly controlled by other residues, which may not be located close to the glycosidic oxygen of the scissile bond. The participation of a nonacidic residue in a proton-transferring network was reported in a Bifidobacterium bifidum GH-95 inverting α-fucosidase , where an Asn can serve as an intermediate in the network, leading to activation of a catalytic water molecule.
Cleavage of internal linkages
The unexpected exclusively internal cleavage of CMC by the D226A enzyme is very interesting. This provides the first example of a mutant exocellulase that could hydrolyze a soluble substrate with wild-type activity to produce mainly large, soluble oligosaccharides and insoluble products, but could not hydrolyze crystalline substrate oligosaccharides or produce cellobiose. Under the same experimental conditions, the CMC activity of TfCel6A with the corresponding Asp79 mutation was only 1% of that of the wild type . As high activity was seen on both charged CMC and uncharged HEC, the cleavage on CMC is unlikely to be due to substrate-assisted catalysis . This mechanism has currently been shown only in GH retaining enzymes, cleaving substrates with an acetamido group .
The increase in the amount of polysaccharide products of the D226A enzyme may be explained by the smaller side chain, allowing modified glucose residues to bind in the active site, alowing the mutant enzyme to move along a CMC molecule until it finds a group of unmodified glucose residues, where it can carry out internal cleavage. A study in a GH-18 enzyme  showed that chitinase can processively move along the substrate without hydrolysis. Cleavage probably occurs at a lower rate than with the wild type, but the great increase in potential cleavage sites, due to the ability to move through modified residues, compensates for this. This modification did not change the global conformation of the enzyme, as CD did not reveal any global change; however, a local structural modification cannot be excluded.
In conclusion, the data presented in this article, as well as data obtained from other family-6 cellulases, are consistent with the role of Asp274 as the catalytic acid of TfCel6B, and roles for Glu495 and Asp497 in substrate binding. Tyr220 probably plays an important role in substrate distortion. This enzyme may function via a novel inverting mechanism without the aid of a single Brønsted base residue, which is replaced by a proton-transferring network.
Strains and plasmids
Escherichia coli DH5α and BL21 RPIL DE3 (Agilent Technologies, Santa Clara, CA, USA) were used as the host strains for plasmid extraction and protein expression, respectively. The entire Cel6B gene in plasmid pSZ143  was used as the template for mutagenesis. A plasmid (pTVcd) containing only the catalytic domain of Cel6B was constructed and used as the template to produce the D274A catalytic domain.
Site-directed mutagenesis and enzyme purification
Complementary primers were designed using primerselect lasergene v. 8.0 (DNASTAR, Madison, WI, USA) to incorporate the desired mutations. PCR was performed for 18 cycles at 95 °C for 1 min, 60 °C for 50 s and 68 °C for 7 min, using the QuikChange method (Agilent Technologies). The PCR products were transformed into E. coli DH5α, and mutant plasmids were checked by DNA sequencing (Applied Biosystems Automated 3730 DNA Analyzer, Cornell University Life Sciences Core Laboratories Center, Ithaca, NY, USA). Mutant plasmids with the correct sequence were transformed and expressed in E. coli BL21 RPIL DE3. Wild-type and mutant enzymes were purified using published chromatographic techniques , first on a CL-4B phenyl–Sepharose column, and then on a Q-Sepharose column; enzyme purity was assessed by SDS/PAGE. The concentrations of Cel6B and Cel6Bcd were determined by measurement of absorbance at 280 nm, using extinction coefficients of 115 000 m−1·cm−1 and 87 000 m−1·cm−1, respectively, calculated from the amino acid composition.
As recommended by Ghose , polysaccharide assays were conducted using a series of enzyme concentrations above and below the target digestion for each substrate for a fixed time with saturating substrate. Enzyme activities were determined on 0.25% BMCC, SC, and PC, and 1% CMC. All assays were run in triplicate for 16 h at 50 °C in 50 mm sodium acetate (NaOAc) (pH 5.5). Reducing sugars were measured using DNS , which fits the assay range well and does not have a blank with enzyme. Nanomoles of protein used were plotted versus the A600 nm, and kaleidagraph (Synergy Software, Reading, PA, USA) was used to fit the curve to determine the amount of enzyme required for 6% substrate digestion of BMCC, SC, and PC, and 1.5% digestion of CMC. If the activity was too weak to achieve the target digestion, activity was calculated at a high concentration of enzyme (1.5 μm).
2,4-DNPC was a gift from SG Withers (University of British Columbia, Vancouver, Canada). Reactions were carried out at 50 °C in 50 mm NaOAc (pH 5.5) using 1.5 μm enzyme and initial substrate concentrations of 20, 40, 80, 150 and 600 μm. The change in absorbance at 400 nm, measured for every 10 min minus the blank, was used as the activity for the substrate concentration at the beginning of the next time point. The concentration of 2,4-dinitrophenol was determined at A400 nm, using an extinction coefficient of 10 900 m−1·cm−1 .
Azide rescue assay
Different concentrations of sodium azide, up to 2 m, were added to mixtures of 0.75–1.5 μm enzyme and substrates. Samples were incubated at 50 °C in 50 mm NaOAc (pH 5.5) for 16 h, and reducing sugars were measured with DNS.
BMCC and SC are insoluble, so they can be used for substrate-binding assays. Binding of 4 μm enzyme to 0.1% BMCC or SC was determined in 50 mm NaOAc buffer (pH 5.5) and 10% glycerol. Reactions were incubated for 1 h on a Nutator rocking table (Clay-Adams, Sparks, MD, USA) at 4 °C to limit hydrolysis. The insoluble substrate was separated from the supernatant by centrifugation at 16 000 g for five mins, and the A280 nm of the supernatant was measured to determine the amount of unbound protein. CMC binding was evaluated by the relative migration of enzymes on native gels containing 0.5% CMC.
Viscometric activity was measured according to the method of Irwin et al. .
Fluorescence quenching titration
Dissociation constants, Kd, for the binding of MUG2 to wild-type and mutant enzymes were determined by direct fluorescence titration at 5.5 °C using an Aminco SLM8000C spectrofluorimeter (SLM-Aminco, Urbana, IL, USA) as previously described .
TLC chromatography was performed as previously described .
Spectra of 10 μg·μL−1 protein were recorded from 190 to 290 nm on an Aviv CD400 Spectrometer (AVIV Biomedical, Inc., Lakewood, NJ, USA) at a scanning rate of 1 nm·s−1 at 4 °C.
We gratefully acknowledge the helpful assistance of D. Irwin. This research was supported by the Vietnam Education Foundation and Grant no. DE-FG02-ER15356 from the US Department of Energy.