Comparison of the structural changes in two cellobiohydrolases, CcCel6A and CcCel6C, from Coprinopsis cinerea – a tweezer-like motion in the structure of CcCel6C


T. Tonozuka, Department of Applied Biological Science, Tokyo University of Agriculture and Technology, 3-5-8 Saiwai-cho, Fuchu, Tokyo 183-8509, Japan
Fax: +81 42 367 5705
Tel: +81 42 367 5702


The basidiomycete Coprinopsis cinerea produces five cellobiohydrolases belonging to glycoside hydrolase family 6 (GH6). Among these enzymes, C. cinerea cellulase 6C (CcCel6C), but not C. cinerea cellulase 6A (CcCel6A), can efficiently hydrolyze carboxymethyl cellulose and is constitutively expressed in C. cinerea. In contrast, CcCel6A possesses a cellulose-binding domain, and is strongly induced by cellobiose. Here, we determined the crystal structures of the CcCel6A catalytic domain complexed with a Hepes buffer molecule, with cellobiose, and with p-nitrophenyl β-d-cellotrioside (pNPG3). A notable feature of the GH6 cellobiohydrolases is that the active site is enclosed by two loops to form a tunnel, and the loops have been demonstrated to open and close in response to ligand binding. The enclosed tunnel of CcCel6A–Hepes is seen as the open form, whereas the tunnels of CcCel6A–cellobiose and CcCel6A–pNPG3 adopt the closed form. pNPG3 was not hydrolyzed by CcCel6A, and bound in subsites +1 to +4. On the basis of this observation, we constructed two mutants, CcCel6A D164A and CcCel6C D102A. Neither CcCel6A D164A nor CcCel6C D102A hydrolyze phosphoric acid-swollen cellulose. We have previously determined the crystal structures of CcCel6C unbound and in complex with ligand, both of which adopt the open form. In the present study, both CcCel6A and CcCel6C mutants were identified as the closed form. However, the motion angle of CcCel6C was more than 10-fold greater than that of CcCel6A. The width of the active site cleft of CcCel6C was narrowed, owing to a tweezer-like motion.

The coordinates and structure factors described in this article have been deposited in the Protein Data Bank under the accession codes 3VOG, 3VOH, 3VOI, 3VOJ, and 3VOF


Coprinopsis cinerea cellulase 6A


Coprinopsis cinerea cellulase 6C


glycoside hydrolase family 6


Humicola insolens cellulase 6A


Humicola insolens cellulase 6B


Hypocrea jecorina cellulase 6A


Protein Data Bank


p-nitrophenyl β-d-cellotrioside


Cellulases belonging to glycoside hydrolase family 6 (GH6) are important for biomass conversion [1–3]. Members of this family employ an inverting mechanism for glucosidic bond cleavage [4,5], and, in the CAZy database (, endoglucanases (EC3.2.1.4) and cellobiohydrolases (EC3.2.1.91) are classified as GH6 cellulases [6]. Endoglucanases randomly hydrolyze internal β-1,4-glucosidic linkages of cellulose, whereas cellobiohydrolases cleave the nonreducing ends of cellulose and cellooligosaccharides to produce mostly cellobiose [7–9]. The crystal structures of GH6 cellobiohydrolases from two ascomycetes, Hypocrea jecorina (formerly known as Trichoderma reesei) cellulase 6A (HjeCel6A) [10] and Humicola insolens cellulase 6A (HinCel6A) [11] have been determined. The catalytic domains of HjeCel6A and HinCel6A consist of a distorted seven-stranded β/α-barrel, and a notable feature is that the active sites are enclosed by the N-terminal and C-terminal loops (designated as loop-1 and loop-2, respectively, in this article), which form a tunnel [12–14]. The two loops have been demonstrated to open and close in response to ligand binding [13,14]. Eight substrate-binding subsites from −4 to +4 are present in the active site cleft of HjeCel6A and HinCel6A [15,16]. In contrast, and despite displaying high sequence similarity to HjeCel6A and HinCel6A, the structure of the fungal endoglucanase H. insolens cellulase 6B (HinCel6B) shows that the active sites are present in a cleft formed by a C-terminal loop deletion coupled with the peeling open of an N-terminal loop [17]. Bacterial GH6 endoglucanases also possess an open active site, which is more accessible to the ligands [18–20].

Many reports are available on crystal structures of GH6 enzymes from ascomycetes, but only one from a basidiomycete, our previous report on the structure of Coprinopsis cinerea cellulase 6C (CcCel6C) [21]. The basidiomycete C. cinerea (formerly known as Coprinus cinereus) possesses five genes that encode GH6 enzymes, which are designated as C. cinerea cellulase 6A (CcCel6A), C. cinerea cellulase 6B, CcCel6C, C. cinerea cellulase 6D, and C. cinerea cellulase 6E [22]. We have previously determined the enzymatic properties and the structure of CcCel6C, which is produced by C. cinerea at low and constitutive levels. Although CcCel6C is classified as a cellobiohydrolase, its activity and structural characteristics are unique in that the enzyme hydrolyzes carboxymethyl cellulose, which is a poor substrate for typical cellobiohydrolases [23]. CcCel6C shows higher sequence identity to the endoglucanase HinCel6B (43% identity) than to HjeCel6A (36%) and HinCel6A (39%), and lacks a cellulose-binding domain, which is necessary for the hydrolysis of crystalline cellulose and is present in most of the GH6 cellobiohydrolases, including HjeCel6A and HinCel6A [22]. The crystal structure of CcCel6C shows that the enclosed tunnel is wider than that of HjeCel6A and HinCel6A, and this open tunnel is likely to contribute to the unique features of CcCel6C [21].

In the present study, we determined the crystal structures of the catalytic domain of CcCel6A in complex with Hepes, cellobiose, and p-nitrophenyl β-d-cellotrioside (pNPG3). There is high sequence identity between CcCel6A and HjeCel6A (48%), and between CcCel6A and HinCel6A (52%), which contain an N-terminal cellulose-binding domain, and unlike the case with the CcCel6C gene, the presence of cellobiose strongly induces the transcription of the CcCel6A gene. Therefore, CcCel6A is a major enzyme involved in the degradation of crystalline cellulose in C. cinerea [22]. Sequence identity between CcCel6C and the catalytic domain of CcCel6A is 48%, although the same organism produces the two enzymes. On the basis of the comparison of the ligand complex structures of CcCel6A, we constructed CcCel6A and CcCel6C mutants to elucidate their conformational changes. The crystal structures of these mutants indicated that the conformational change in CcCel6C was much more drastic than that in CcCel6A, although we previously predicted that the conformational change in CcCel6C would be less favorable.

Results and Discussion

Overall structure of the catalytic domain of CcCel6A

The crystal structures of the catalytic domain of native CcCel6A (hereafter referred to as CcCel6A) and the CcCel6A–cellobiose and CcCel6A–pNPG3 complexes were determined. The crystals belonged to space group P212121 (Table 1), with one molecule in the asymmetric unit. In the native structure, a Hepes molecule was found in the catalytic center, and the structure is therefore denoted here as CcCel6A–Hepes. The electron density (2Fo − Fc) maps for these structures contoured at 1σ show continuous density for all main chain atoms. The overall structure of CcCel6A is shown in Fig. S1. As in other fungal GH6 cellobiohydrolases [10,11], the catalytic domain of CcCel6A consisted of a seven-stranded β/α-barrel fold and contained two disulfide bridges (Cys165–Cys224 and Cys355–Cys402).

Table 1.   Data collection and refinement statistics. The values for the highest-resolution shells are given in parentheses.
 CcCel6A–HepesCcCel6A–cellobioseCcCel6A–pNPG3CcCel6A D164ACcCel6C D102A
Data collection
 BeamlinePF-AR NW12APF-AR NW12APF-AR NW12APF-AR NW12ASPring-8 BL26B2
 Space groupP212121P212121P212121P212121P212121
Cell dimensions
 a (Å)46.045.646.045.949.9
 b (Å)72.572.171.571.879.3
 c (Å)104.9104.7101.1100.595.0
Resolution range (Å)50–1.45 (1.50–1.45)50–2.40 (2.49–2.40)50–2.00 (2.07–2.00)50–2.29 (2.37–2.29)50–1.60 (1.66–1.60)
Measured reflections613 99268 757120 516107 772368 328
Unique reflections61 61013 66622 34915 43250 422
Completeness (%)97.6 (92.7)95.1 (97.6)95.1 (96.8)99.9 (99.7)99.9 (99.9)
I/σ(I)41.4 (5.1)21.5 (5.1)29.5 (8.3)37.8 (12.3)27.2 (4.1)
Rmerge0.053 (0.356)0.084 (0.298)0.091 (0.320)0.084 (0.216)0.076 (0.396)
Refinement statistics
 Bond lengths (Å)0.0100.0090.0100.0090.009
 Bond angles (°)1.341.
Ramachandran plot (molprobity)
 Favored (%)97.596.997.296.796.5
 Outliers (%)00000

An important feature of GH6 cellobiohydrolases is that the active site is located within an enclosed tunnel, and the conformational change in the two tunnel-forming loops, loop-1 and loop-2, is induced by binding of the substrate [13,14]. The structural features of CcCel6A were similar to those found in other GH6 cellobiohydrolases. The Cα backbone of CcCel6A–Hepes was superimposed with that of CcCel6A–cellobiose and CcCel6A–pNPG3 (Fig. 1). The enclosed tunnel of CcCel6A–Hepes was wide, and was designated as the open form. In contrast, the backbones of CcCel6A–cellobiose and CcCel6A–pNPG3 were almost identical, and the tunnels adopted a narrower conformation, which was designated as the closed form (Table 2). The conformational change from the open form to the closed form of CcCel6A occurred in a manner essentially identical to that of HinCel6A. Large loop movements in HinCel6A have been described for residues 183–188 and 412–416, and similar movements were observed in the corresponding regions of CcCel6A: in residues 167–172 in loop-1 and residues 395–399 in loop-2. Average B-values for the loops of both open and closed forms (residues 167–172 and 395–399 of CcCel6A–Hepes, 19.7 and 23.2 Å2; CcCel6A–cellobiose, 41.8 and 43.6 Å2; CcCel6A–pNPG3, 17.8 and 20.0 Å2) were higher than those for whole proteins (CcCel6A–Hepes, 13.3 Å2; CcCel6A–cellobiose, 35.9 Å2; CcCel6A–pNPG3, 17.0 Å2), but electron densities for the two loops were well defined. In HinCel6A, a concerted movement of residues 230–245 has been reported by virtue of the presence of a disulfide bond between Cys181 and Cys240 [14], and the conformational change in the corresponding residues, 214–229, of CcCel6A was also observed, owing to the corresponding disulfide bond between Cys165 and Cys224.

Figure 1.

 Comparison of the Cα backbones of CcCel6A–Hepes (blue), CcCel6C–cellobiose (green), and CcCel6C–pNPG3 (red). Black arrow: a large loop movement observed for residues 167–172 in loop-1. White arrow: glucose located on the surface of CcCel6A.

Table 2.   Structural statistics.
 CcCel6A–HepesCcCel6A–cellobioseCcCel6A–pNPG3CcCel6A D164ACcCel6C D102A
Loop statusOpenClosedClosedClosedClosed
Number of atoms
Average B2)
Soaked saccharideCellobiosepNPG3pNPG3
Modeled ligand or magnesium/subsite number or position/average B2)Hepes/–1 to +1/21.8Cellobiose/–3 to −2/38.9
Cellotriose/+1 to +3/40.1
Glucose/molecular surface/61.9
pNPG3/+1 to +4/19.3
Magnesium/molecular surface/24.1

Ligand-bound structures of CcCel6A

A Hepes molecule was found to bind to the active site of CcCel6A. Electron density for the Hepes molecule was clearly observed in subsites −1 to +1 (Fig. 2A). Despite this binding, CcCel6A adopted the open form. An oxygen atom in the sulfonate group of Hepes directly formed a hydrogen bond with atom Nε2 of His255 (Figs 3A and S1B). A hydroxyl group of Hepes formed a hydrogen bond with atom Oγ of Ser293. Although multiple amino acids of CcCel6A were observed to participate in the interaction with Hepes, no interaction between Hepes and loop-1 (residues 167–172) was found (Fig. 3A).

Figure 2.

 Stereo views of the Fo − Fc omit maps of (A) Hepes bound to CcCel6A, (B) cellobiose and cellotriose bound to CcCel6A, (C) glucose located on the surface of CcCel6A, (D) pNPG3 and the hexacoordinate magnesium ion bound to CcCel6A, and (E) glucose bound to CcCel6C D102A. The difference Fourier maps were calculated with exclusion of the ligands, and the resultant Fo − Fc omit maps were contoured at 2.0σ. The subsites are labeled from −3 to +4.

Figure 3.

 Stereo views of the conformational change in loop-1 of CcCel6A. (A) Overlays of the ligands in CcCel6A–Hepes (open form, blue), CcCel6A–cellobiose (closed form, green), and CcCel6A–pNPG3 (including the magnesium ion; closed form, red). Four important residues, Asp164, Ser170, His255, and Ser293, and the Cα backbones of loop-1 are indicated. (B) Comparison of the residues in and around loop-1 of CcCel6A–Hepes (open form, blue) and CcCel6A D164A (closed form, orange). Hydrogen bonds are shown as dashed lines. In addition to the side chains of amino acids, the main chain carbonyl oxygen atoms of Asp/Ala164 and Ala167 are indicated.

In the structure of CcCel6A–cellobiose, a cellobiose molecule was observed in subsites −3 to −2 (average B = 38.9 Å2) (Fig. 2B). Another molecule, which bound to subsites +1 to +3, was modeled as cellotriose (average B = 40.1 Å2) (Table 2). A similar observation has been reported in HinCel6A–cellobiose, in which there are two HinCel6A molecules in an asymmetric unit. A saccharide bound in the active site of both copies of HinCel6A has been modeled as glucose (subsite −2), and another saccharide has been identified as cellotetraose (subsites +1 to +4) or cellotriose (subsites +1 to +3) [14]. We observed that the structure of CcCel6A–cellobiose adopts the closed form, and Asp164 and Ser170 in loop-1 directly interact with these saccharides (Fig. 3A). We further observed that atom Oδ1 of Asp164 forms a hydrogen bond with atom O3A of cellotriose, and atom Oγ of Ser170 forms hydrogen bonds with atoms O1′ and O2′ of cellobiose. Weak electron density was seen in subsite −1, but the electron density seen in subsite −1 did not convincingly resemble a glucose residue (Fig. 2B).

A glucose molecule was found on the molecular surface of CcCel6A (Fig. 1). The omit map showed that the electron density for the glucose molecule (average B = 61.9 Å2) was weaker than that for cellobiose and cellotriose (Fig. 2C). It is likely that the glucose molecule located on the surface was an artefact, because the molecule was present in a crystal contact, and only a stacking interaction between Tyr358 and the hexose ring appeared to be important for the binding as calculated with the program ligplot [24]. The corresponding residue in HinCel6A was identified as Ile375, indicating that Tyr358 of CcCel6A is not conserved.

There appeared to be one pNPG3 molecule in CcCel6A–pNPG3. Electron density for pNPG3 was observed in subsites +1 to +4 (Fig. 2D). CcCel6A was identified as a closed form, and atom Oδ1 of Asp164 in loop-1 formed a hydrogen bond with atom O3A of pNPG3. In CcCel6C–pNPG3, two pNPG3 molecules, in subsites −4 to −1 and in subsites +1 to +4, have been found in the active site [21], whereas, in the present study, no density for pNPG3 was found in the minus subsites of CcCel6A. In addition, instead of the saccharide, a hexacoordinate magnesium ion was found in subsite −1 (Fig. 2D). Water molecules in the hexacoordinate ion interacted with Asp164 and Ser170 in loop-1, and also with Tyr158, Asp164, Ser170, Ser293, Lys382 and Asp388 via hydrogen bonds (Fig. 4). There were two magnesium ions in the structure, and the other one was located on the molecular surface near Gly309.

Figure 4.

 Schematic drawing of the amino acids interacting with pNPG3. The subsite numbers are labeled, and the hexacoordinate magnesium ion is illustrated. Symbols: white circle, oxygen atom; black circle, carbon atom; gray circle, nitrogen atom; and dashed line, hydrogen bond. The residues involved in hydrophobic interactions are illustrated.

To test whether CcCel6A hydrolyzes pNPG3, a mixture of CcCel6A and pNPG3 was incubated and analyzed by TLC. CcCel6A did not hydrolyze 2 mmpNPG3 (Fig. S2A), whereas it did hydrolyze 2 mm cellotriose to produce glucose and cellobiose under the same conditions (Fig. S2A,B). The crystal structures of HinCel6A complexed with various ligands have indicated that the cellulose chain initially binds to the plus subsites, and the cellulose chain then slides through the active center tunnel until it occupies subsites −2 and −1 [16]. It is likely that pNPG3 is unable to slide through the active center tunnel. It has been reported that GH6 enzymes hydrolyze p-nitrophenyl and 4-methylumbelliferyl oligosaccharides poorly [25,26], but HjeCel6A hydrolyzes cellotriose and 4-methylumbelliferyl β-d-cellotrioside, and the turnover numbers for these substrates are identical [27].

Structures of CcCel6A D164A and CcCel6C D102A

The structure of CcCel6A–pNPG3 was identified as a closed form, and only one residue located in loop-1, Asp164, directly formed a hydrogen bond with pNPG3 (Fig. 3A). This result suggests that Asp164 may play an important role in the closure of loop-1 upon substrate binding. In the present study, Asp164 of CcCel6A and the corresponding residue, Asp102, of CcCel6C were replaced by Ala, and the structures of the mutants, CcCel6A D164A and CcCel6C D102A, were determined.

The structure of CcCel6A D164A was almost isomorphous with that of wild-type CcCel6A, and the rmsd between CcCel6A–Hepes and CcCel6A D164A was 0.594 Å. Although no saccharides were added to the crystallization solution, and the crystallization conditions for CcCel6A–Hepes and CcCel6A D164A were almost identical, CcCel6A D164A adopted a closed form. No Hepes molecule was observed in the active site. The structures of the open form of CcCel6A–Hepes and the closed form of CcCel6A D164A were superimposed. In the open form of CcCel6A–Hepes, atom Oδ1 of Asp164 formed hydrogen bonds with atom Oη of Tyr158, and atoms Oδ2 of Asp210 and Oδ2 of Asp164 formed hydrogen bonds with atom Oδ2 of Ser170; thus, Asp164 appeared to be anchored by many hydrogen bonds (Fig. 3B). In contrast, the side chain of the mutated residue of CcCel6A D164A was not able to form hydrogen bonds, and a shift of 0.6 Å in the position of the Cα atom towards the active site was observed. A new hydrogen bond between the main chain carbonyl oxygen atoms of residue 164 and Ala167 was generated, and the side chain of Ser170 was directed towards subsite −1 (Fig. 3B).

The CcCel6C D102A crystal belonged to space group P212121, whereas the wild-type CcCel6C crystal was assigned to space group P1. The unit cell parameters of the two crystals were entirely different. Before the diffraction data were collected, the crystal of CcCel6C D102A was soaked in a solution of pNPG3, with the same experimental procedure as that used for CcCel6A–pNPG3; however, only one glucose residue was observed in subsite −2 (abbreviated as Glc −2) (Fig. 2E). CcCel6C D102A also adopted a closed conformation. The essence of the motion of CcCel6C is similar to that of CcCel6A; the large loop movements of CcCel6C were observed in loop-1 and loop-2 (mainly residues 98 to 111 and residues 335 to 347, respectively), and a concerted movement (mainly residues 154–169; denoted as loop-1′) was observed, owing to the presence of a disulfide bond between Cys103 and Cys164. In CcCel6C D102A, Ser108, a residue corresponding to Ser170 of CcCel6A, was pointed to the active site, and atom Oγ of Ser108 formed a hydrogen bond with atom O2 of glucose found in subsite −2 (Fig. S3A).

A tweezer-like motion in the structure of CcCel6C

The structural difference between wild-type CcCel6C and the D102A mutant was compared with that between wild-type CcCel6A and the D164A mutant (Fig. 5A,B). In our previous study, the crystals of CcCel6C were soaked in a solution of cellobiose and pNPG3; however, no structural differences were observed between the unliganded CcCel6C [Protein Data Bank (PDB) ID 3A64], CcCel6C–cellobiose (PDB ID 3A9B), and CcCel6C–pNPG3 (PDB ID 3ABX) [21], unlike in the case of CcCel6A. It is now apparent from our current study that the D102A mutation is responsible for the change in the conformation of CcCel6C. The conformational change in CcCel6C is, however, much more drastic than that in CcCel6A, and the reason why we previously did not observe the closed form of CcCel6C is probably that the motion of CcCel6C is too large to occur in the crystal packing.

Figure 5.

 Conformational changes in CcCel6C and CcCel6A. (A) Comparison of the Cα backbones of unliganded CcCel6C (open form; PDB ID 3A64; cyan) and CcCel6C D102A (closed form, magenta). The angles produced by Asp115 Cα–Ser236 Cα and Asn340 Cα–Ser236 Cα are shown. (B) Comparison of the Cα backbones of CcCel6A–Hepes (open form, blue) and CcCel6A D164A (closed form, orange). The angles produced by Ala177 Cα–Ser293 Cα and Ser394 Cα–Ser293 Cα are shown. (C) Stereo views of the Cα backbones of unliganded CcCel6C (open form) and CcCel6C D102A (closed form). Module-1 (open form, pink; closed form, red), module-2 (open form, cyan; closed form, blue) and hinge residues (green) are shown. Module-2 of the two structures was superimposed.

Superimposition of the backbones of the unliganded CcCel6C and CcCel6C D102A showed that the structure of CcCel6C behaved like a pair of tweezers. The hinge motion was analyzed with the program rigidfinder at the molmovdb server ( [28]. CcCel6C was roughly divided into two rigid body modules, module-1 (residues 12–24, 58–235, 256–281, and 309–327) and module-2 (residues 26–56, 237–254, 283–307, and 329–383), and the conformational change appeared to be centered on Lys25, Pro57, Ser236, Ala255, Gln282, Arg308, and Lys328 (Fig. 5C). For the whole protein, the rmsd between unliganded CcCel6C and CcCel6C D102A was 0.983 Å, whereas those for module-1 and module-2 were 0.538 and 0.402 Å, respectively.

To compare the motion, angles produced by the two vectors, Asp115 Cα–Ser236 Cα and Asn340 Cα–Ser236 Cα, of CcCel6C were chosen, and the differences in the two angles between the unliganded CcCel6C and CcCel6C D102A were calculated with program geomcalc of ccp4 [29]. The angles of unliganded CcCel6C and CcCel6C D102A were 64.0° and 53.7°, respectively; thus, the motion angle is 10.3° (Fig. 5A). The same calculations were carried out for CcCel6A–Hepes and CcCel6A D164A, with the corresponding atoms, Ala177 Cα–Ser293 Cα and Ser394 Cα–Ser293 Cα. The angles of CcCel6A–Hepes and CcCel6A D164A were 56.2° and 55.4°, respectively; thus, the motion angle was 0.8° (Fig. 5B). These results indicated that the motion angle of CcCel6C was more than 10-fold greater than that of CcCel6A. In HinCel6A, the motion angle produced by the corresponding three atoms (Cα atoms of Ala193, Ala309, and Thr411) between the open form (unliganded form; PDB ID 1BVW) and the closed form (cellobiose complex; PDB ID 2BVW) [14] was calculated to be 5.3°. The motion angle of HjeCel6A produced by the corresponding three atoms (Cα atoms of Ala188, Ala304, and Ser407) between the ‘most open’ form (chain A of the D175A mutant; PDB ID 1HGW) [30] and the ‘most closed’ form (chain A of the Y169F mutant; PDB ID 1QJW) [13] was calculated to be 0.5°. These findings indicated that the conformational change in CcCel6C is much more drastic than that in CcCel6A and other GH6 cellobiohydrolases with known structures.

The most significant movement in HinCel6A occurred at sequence AAASNG in loop-1 [14]; the corresponding region in CcCel6C is AKASDG (residues 105–110). We have described in the previous report how the conformational change in loop-1 of CcCel6C seems to be less favorable because of the bulky Lys (Lys106), and the enclosed tunnel of CcCel6C is not completely ‘enclosed’ [21]. In the present study, however, the comparison of the structures of unliganded CcCel6C and CcCel6C D102A showed that the distance between atom Cα of Lys106 and atom Cα of Phe344 in the open form was 10.3 Å, whereas the corresponding distance in the closed form was 6.1 Å, and that Lys106 seemed to function as a cover of the active site cleft to form the ‘enclosed’ tunnel (Fig. 6A). In the structures of the open and the closed forms of CcCel6A, the distances between the corresponding residues (Ala168 and Tyr398) were estimated to be 8.1 Å (CcCel6A–Hepes) and 5.8 Å (CcCel6A D164A), respectively, suggesting that the motion is less drastic (Fig. 6B).

Figure 6.

 Comparison of the open and the closed forms of CcCel6C (A) and CcCel6A (B). Stereo views of the lid of the active site tunnel are shown. Unliganded CcCel6C (open form, cyan) and CcCel6C D102A (closed form, magenta) (A), or CcCel6A–Hepes (open form, blue) and CcCel6A D164A (closed form, orange) (B), are overlaid. The distances of dashed gray lines (open form) and dashed black lines (closed form) are described in the text.

It is noteworthy that the width of the active site cleft, particularly in subsites −1 to −2, of CcCel6C was narrowed because of a tweezer-like motion. For example, the distance between atom Cα of Tyr96 and atom Cα of Glu332 of unliganded CcCel6C was 15.8 Å, whereas that between atom Cα of Tyr96 and atom Cα of Glu332 of CcCel6C D102A was 14.6 Å (Fig. S3A). The corresponding distances (between atom Cα of Tyr158 and atom Cα of Glu386) of the open form (CcCel6A–Hepes) and the closed form (CcCel6A D164A) of CcCel6A were 15.5 and 15.0 Å, respectively (Fig. S3B). This result showed that, in the closed form, the width of the active site cleft of CcCel6C was similar to, or rather narrower than, that of CcCel6A. The Glc −2 residue found in CcCel6C D102A was compared with the corresponding glucose residue in the open form CcCel6C–pNPG3. In CcCel6C D102A, in addition to the formation of a hydrogen bond between Ser108 and atom O2 of Glc −2 as described above, atom Oγ of Ser63 formed a hydrogen bond with atom O4 of Glc −2, owing to the narrowing of the active site cleft (Fig. S3B). In contrast, the corresponding glucose residue in subsite −2 did not interact with these two Ser residues in the open form CcCel6C–pNPG3.

Comparison with HjeCel6A

A similar experiment to that described in the previous section has been performed in HjeCel6A [13,30]. The conformational change in HjeCel6A is more complicated, and four states (most closed, more open, even more open, and most open) of the loop have been identified [13]. The corresponding residue, Asp175, of HjeCel6A (Table 3) has been replaced by Ala, but, unlike CcCel6A D164A and CcCel6C D102A, HjeCel6A D175A adopts a ‘most open’ form, whereas the unliganded form of wild-type HjeCel6A is observed as a ‘more open’ form [30]. Asp221 (equivalent to Asp210 of CcCel6A) has also been replaced by Ala, and HjeCel6A D221A is identified as a ‘most closed’ form [13,30]. The structural features of HjeCel6A D221A are similar to those of CcCel6A D164A and CcCel6C D102A; the hydrogen bond cluster of Asp175 is broken, and Ser181 (equivalent to Ser170 of CcCel6A) is pointed into the active site.

Table 3.   Some key residues described in the text and the corresponding residues in other GH6 enzymes. TfuCel6A, Thermobifida fusca Cel6A; TfuCel6B, Thermobifida fusca Cel6B; CfiCel6A, Cellulomonas fimi Cel6A; CBH, cellobiohydrolase; EG, endoglucanase.
Subsite −2Asp125Ser63Asp137Asp139Ser54Ala43Gly186Gly179
Subsite −2Tyr158Tyr96Tyr169Tyr174Tyr86Tyr73Tyr220Tyr210
Subsite +1, loop-1Asp164Asp102Asp175Asp180Asp92Asp79Asp226Asp216
Subsite −1, loop-1Ser170Ser108Ser181Ser186Ser98Ser85Ser232Gly222?
Catalytic acidAsp210Asp150Asp221Asp226Asp139Asp117Asp226Asp252
Subsite +2His255His195His266His271Asn183His159His326His290
Subsite −1Tyr295Tyr238Tyr306Tyr311Tyr223Tyr191Tyr367Tyr321
Subsite −1Arg341Arg284Arg353Arg357Arg269Arg221Arg428Arg349
Molecular surfaceTyr358Arg301Ile371Ile375Asn286Ser235Ala468Arg362
Subsite −2Lys382Lys328Lys395Lys399Lys310Lys259Lys491Lys386
Subsite −1Asp388Asp334Asp401Asp405Asp316Asp265Asp497Asp392

The enzymatic activities of CcCel6A D164A and CcCel6C D102A were measured, and no activity was detected on phosphoric acid-swollen cellulose for either mutant. A conserved Asp (Asp210 in CcCel6A, Asp150 in CcCel6C, and Asp221 in HjeCel6A; Table 3) has been identified as the catalytic acid residue of GH6 enzymes, whereas it is still debated which residue may act as the catalytic base of GH6 enzymes [31–34]. An Asp in Cellulomonas fimi cellulase 6A (formerly known as CenA), Asp392 (Asp388 in CcCel6A, Asp334 in CcCel6C, and Asp401 in HjeCel6A), was first proposed as the catalytic base [31]. However, Asp175 in HjeCel6A has been shown to act as catalytic base indirectly in what has been called a Grotthus mechanism [30], and recent reports support this catalytic mechanism [32,33]. In HjeCel6A, Asp401 undergoes hydrogen bond acceptor interactions with the nearby side chains of Tyr306, Arg353, and Lys395 [30], and the same interactions were observed in both CcCel6A (Tyr295, Arg341, and Lys382) and CcCel6C (Tyr238, Arg284, and Lys328). Therefore, the structural studies of CcCel6A and CcCel6C are found to be compatible with the indirect Grotthus catalytic mechanism of GH6 enzymes.

This study shows that the motion angles for the open-to-closed conformational change of CcCel6A and CcCel6C are markedly different, although the mechanisms of the conformational changes are essentially identical. CcCel6A has been indicated to be the most important enzyme for the hydrolysis of crystalline cellulose, and the expression of CcCel6A is induced by cellobiose [22]. In contrast, CcCel6C is produced at low and constitutive levels, and the enzyme does not possess a cellulose-binding domain. Therefore, we proposed that the physiological role of CcCel6C is to hydrolyze amorphous cellulose to produce cellobiose, which is the inducer of CcCel6A expression [22,23]. The enzymatic activities of CcCel6A and CcCel6C for phosphoric acid-swollen cellulose were similar; however, CcCel6C hydrolyzes the chemically modified form of cellulose, carboxymethyl cellulose [23]. To elucidate the role of the enclosed tunnel, surface models of CcCel6C, CcCel6A and HjeCel6A in complexes with a cellulose chain were compared. In the open form, areas of cellulose chain covered by the tunnel were different among the three enzymes (Fig. 7). In contrast, in the closed form, the tunnel lengths were almost identical, and the tunnel-forming loops, loop-1 and loop-2, covered subsites +1, −1, and −2 in all the three models, regardless of their tunnel lengths in the open form (Fig. 7). The results of the present study suggest that, in the open form of CcCel6C, the wide tunnel is suitable for the binding of a substrate such as carboxymethyl cellulose, whereas the narrow tunnel is required to cleave the β-1,4-glucosidic linkage in the closed form. Phylogenetic analysis of GH6 enzymes demonstrated that the lengths of loops that form the active site tunnel in cellobiohydrolases vary among subfamilies [35]. Further studies are needed to elucidate the relationships among conformational change, physiological role and phylogenetic position of GH6 cellobiohydrolases.

Figure 7.

 Surface models of the enclosed tunnel of CcCel6C (A), CcCel6A (B), and HjeCel6A (C). The open and closed forms are shown in the left and right panels, respectively. The cellulose chain (red) was constructed by combining the ligands of the reported structures (PDB IDs 1OCB and 1QJW), and overlaid on the surface models; the glucose residues in subsites −3 to +2, from the top to the bottom of each panel, are indicated. Amino acids in loop-1 and loop-2 (CcCel6C, residues 105–110 and 341–345; CcCel6A, residues 167–172 and 395–399; HjeCel6A, residues 178–183 and 408–412) are shown in yellow.

Experimental procedures

Construction of the CcCel6A and CcCel6C mutants

Escherichia coli JM109 was used as the host strain for DNA manipulation. The expression plasmids of the catalytic domains of CcCel6A and CcCel6C were constructed with the pET21a plasmid (Merck, Darmstadt, Germany), as described previously [23,36]. Site-directed mutagenesis was performed with the QuikChange Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA, USA). To construct the expression plasmid of CcCel6A D164A, the oligonucleotide 5′-GAC TTG CCT GAC CGT GCA TGC GCT GCC GCA GCC-3′ and its complementary oligonucleotide were used as primers, and the expression plasmid of CcCel6A was used as a template for the reaction. To construct the expression plasmid of CcCel6C D102A, the oligonucleotide 5′-CTG CCC GAT CGT GCA TGC GCTGCC AAG GCA-3′ and its complementary oligonucleotide were used as primers, and the expression plasmid of CcCel6C was used as a template for the reaction. The amplified products were transformed into E. coli JM109 and verified by DNA sequencing.

Measurement of enzymatic activity and protein

The enzymatic activity was measured essentially as previously described [23]. The enzymes were used at a final concentration of 1.0 μm for substrate hydrolysis, and were incubated for 6 h with 0.5% (w/v) phosphoric acid-swollen cellulose in 50 mm Mes/NaOH buffer (pH 7.0). The amount of reducing sugar was measured by the Somogyi–Nelson method, with glucose as a standard. To determine whether CcCel6A hydrolyzes pNPG3 and cellotriose, CcCel6A (1.75 mg·mL−1) was incubated at 30 °C with pNPG3 or cellotriose (2 mm) in 100 mm Mes buffer (pH 7.0). The samples were separated by TLC on Kieselgel 60 (Merck) with ethyl acetate/acetic acid/water (3 : 2 : 1, v/v), as previously described [23]. For cellotriose, the glucose produced was also quantified by the glucose oxidase method with a Glucose Test kit (Wako Pure Industries, Osaka, Japan). The protein concentration was determined by measurement of the absorbance at 280 nm, using the molar extinction coefficient calculated by the expasy protparam server (

Enzyme preparation and crystallization

The expression, purification and crystallization of wild-type CcCel6A, CcCel6A D164A and CcCel6C D102A were performed with the same procedure as used for wild-type CcCel6C [36]. Briefly, the enzymes were fused with a His-tag, produced in E. coli BL21(DE3) cells, and then purified with Ni2+–nitrilotriacetic acid agarose (Qiagen, Hilden, Germany). The enzymes were crystallized at 20 °C with the hanging-drop vapor-diffusion method, in which 1 μL of protein solution (20 mg·mL−1) was mixed with the same volume of well solution (100 mm Hepes/KOH, pH 7.0–7.5, 20–25% polyethylene glycol 8000, 50–200 mm magnesium acetate). The obtained crystals were transferred to a cryosolution of 20% glycerol in the well solution, and frozen in a nitrogen stream. The crystals of CcCel6A–pNPG3 and CcCel6C D102A–pNPG3 were obtained by soaking the complexes in the well solution containing 60 mmpNPG3 for 5 min. The crystal of CcCel6A–cellobiose was obtained by soaking the complex in the same well solution containing 220 mm cellobiose for 5 min. For both pNPG3 and cellobiose, the solution containing the ligands also acted as a cryoprotectant. The diffraction data were collected at beamline PF-AR NW12 at Photon Factory, Tsukuba, Japan, and at beamline BL26B2 at Spring-8, Hyogo, Japan. The data were processed and scaled with the program hkl2000 (Table 1) [37].

Model building and refinement

The structures were solved by molecular replacement with the program molrep in the ccp4 suite [29]. Models of HinCel6A (PDB ID 1BVW; for determination of the CcCel6A structure) [11] and wild-type CcCel6C (PDB ID 3A64; for determination of the CcCel6C D102A structure) [21] were employed as search models. To determine the structure of CcCel6A, automated model building was performed with the program arp/warp [38]. The refinement was carried out with the program refmac in the ccp4 suite. Manual adjustment and rebuilding of the model were carried out with the program coot [39], and the models for the ligands were built from both 2Fo − Fc and Fo − Fc electron density maps. Solvent molecules were introduced with the program arp/warp. Validation of the structures was performed with the molprobity server [40]. Refinement statistics are listed in Table 1. In Ramachandran plots, 96.5–97.5% of the residues were shown to be in the favored regions, and no residues were identified as outliers. Figures were generated with pymol ( and ligplot [24]. The rmsd was calculated with the SSM Superpose tool in coot.


We would like to thank T. Fujii for her technical assistance, and Y. Liu for his useful advice. We also thank the reviewer for helpful comments. This research was supported, in part, by the Green Biomass Research for Improvement of Local Energy Self-Sufficiency Program of the Ministry of Education, Culture, Sports, Science, and Technology of Japan. This research was performed with the approval of the Photon Factory Advisory Committee (2010G001) of the National Laboratory for High Energy Physics, Tsukuba, Japan. The synchrotron-radiation experiments were also performed on BL26B2 at SPring-8 with the approval of the Priority Program for Disaster-Affected Quantum Beam Facilities (2011A1881), the Japan Synchrotron Radiation Research Institute.