Báo cáo khoa học: Structures of Phanerochaete chrysosporium Cel7D in complex with product and inhibitors ppt

Structures of Phanerochaete chrysosporium Cel7D in complex with product and inhibitors Wimal Ubhayasekera 1 , Ine ´ s G. Mun ˜ oz 1, *, Andrea Vasella 2 , Jerry Sta ˚ hlberg 1 and Sherry L. Mowbray 1 1 Department of Molecular Biology, Swedish University of Agricultural Sciences, Uppsala, Sweden 2 Laboratory of Organic Chemistry, ETH, Ho ¨ nggerberg, Zu ¨ rich, Switzerland Cellulose is the most abundant polymer on earth. It has been estimated that as much as 15% of all atmo- spheric carbon dioxide is fixed yearly, resulting in vast quantities of plant biomass, mostly as a complex mix- ture of cellulose and lignin [1]. The recycling of this carbon is critically dependent on the action of micro- bial organisms, primarily fungi and bacteria. An understanding of the processes at work is obviously of enormous environmental importance. The enzymes involved are also useful for applications that include, among others, their use in commercial laundry pow- ders, as well as in the de-inking of recycled paper and the synthesis of fine chemicals. Cellulases, the enzymes that hydrolyse cellulose, have been broadly characterized as cellobiohydrolases (1,4-b-d-glucan cellobiohydrolase, EC 3.2.1.91) and Keywords Cellulase; cellobiohydrolase; glycoside hydrolase; Trichoderma reesei; Phanerochaete chrysosporium Correspondence J. Sta ˚ hlberg, Department of Molecular Biology, Swedish University of Agricultural Sciences, Biomedical Centre, PO Box590, SE-751 24 Uppsala, Sweden Fax: +46 18 536971 Tel: +46 18 471 4566 E-mail: Jerry.Stahlberg@molbio.slu.se *Present address Structural Biology and Biocomputing Programme, Spanish National Cancer Centre (CNIO), Melchor Ferna ´ ndez Almagro 3, 28029 Madrid, Spain (Received 6 December 2004, revised 15 February 2005, accepted 22 February 2005) doi:10.1111/j.1742-4658.2005.04625.x The cellobiohydrolase Pc_Cel7D is the major cellulase produced by the white-rot fungus Phanerochaete chrysosporium, constituting 10% of the total secreted protein in liquid culture on cellulose. The enzyme is classified into family 7 of the glycoside hydrolases and, like other family members, catalyses cellulose hydrolysis with net retention of the anomeric carbon configuration. Previous work described the apo structure of the enzyme. Here we investigate the binding of the product, cellobiose, and several inhibitors, i.e. lactose, cellobioimidazole, Tris ⁄ HCl, calcium and a thio- linked substrate analogue, methyl 4-S-b-cellobiosyl-4-thio-b-cellobioside (GG-S-GG). The three disaccharides bind in the glucosyl-binding subsites +1 and +2, close to the exit of the cellulose-binding tunnel ⁄ cleft. Pc_Cel7D binds to lactose more strongly than cellobiose, while the oppos- ite is true for the homologous Trichoderma reesei cellobiohydrolase Tr_Cel7A. Although both sugars bind Pc_Cel7D in a similar fashion, the different preferences can be explained by varying interactions with nearby loops. Cellobioimidazole is bound at a slightly different position, displaced 2A ˚ toward the catalytic centre. Thus the Pc_Cel7D complexes provide evidence for two binding modes of the reducing-end cellobiosyl moiety; this conclusion is confirmed by comparison with other available structures. The combined results suggest that hydrolysis of the glycosyl-enzyme intermedi- ate may not require the prior release of the cellobiose product from the enzyme. Further, the structure obtained in the presence of both GG-S-GG and cellobiose revealed electron density for Tris at the catalytic centre. Inhibition experiments confirm that both Tris and calcium are effective inhibitors at the conditions used for crystallization. Abbreviations GG-S-GG, methyl 4-S-b-cellobiosyl-4-thio-b-cellobioside; IBTG, o-iodo-benzyl-b- D-thio-glucoside; Pc_Cel7D, cellobiohydrolase Cel7D from Phanerochaete chrysosporium; PDB, Protein Data Bank; pNP-Lac, p-nitrophenyl-b- D-lactoside; Tr_Cel7A, cellobiohydrolase Cel7A from Trichoderma reesei. 1952 FEBS Journal 272 (2005) 1952–1964 ª 2005 FEBS endoglucanases (1,4-b-d-glucan glucanohydrolase, EC 3.2.1.4) [2]. Cellobiohydrolases tend to act proces- sively from the end of a cellulose chain, that is, they cleave off a number of cellobiose units in succession before the enzyme is released [3,4]. Endoglucanases cut cellulose at random positions within the chains, thus creating new ends from which cellobiohydrolases can work. Efficient degradation of cellulose requires a synergistic balance between the two types of activities. Cellulases and other glycoside hydrolases have been classified into structurally related families, based on sequence homology as well as the patterns of hydro- phobic residues [5,6]. To date nearly 100 glycoside hydrolase families are defined in the CAZY database (http://afmb.cnrs-mrs.fr/CAZY/). Efficient cellulose- degrading fungi generally have at least one member of glycoside hydrolase family 7. The enzymes in this family perform hydrolysis with net retention of the anomeric configuration, in a double-displacement mechanism through a covalent glycosyl-enzyme inter- mediate [7,8]. Most, but not all, members have a small cellulose-binding module connected to the catalytic module by a presumably flexible linker. The catalytic core of this family is a b-sandwich composed of two large, mainly antiparallel, b-sheets packed onto each other (Fig. 1). A long cellulose-binding site is defined by loops on one face of the sandwich. It has been demonstrated, for this and some other structural famil- ies, that a very important difference between an endo- glucanase and a cellobiohydrolase is the size of such loops. In a cellobiohydrolase, they are generally lon- ger, and form a tunnel that encloses the catalytic resi- dues. Substrate usually reaches the active site by threading itself in from the end of the tunnel. In con- trast, an endoglucanase has shorter loops that define a more open binding cleft, and allow more direct access of an intact cellulose chain. Among the fungi that have a family 7 cellobiohydro- lase, it is the major enzyme in the cellulase mixture secreted. The first member of the family for which the structure was determined was the cellobiohydrolase of Trichoderma reesei (a clonal derivative of Hypocrea jecorina), Tr_Cel7A, formerly called CBH 1 [9]. Three acidic residues (Glu212, Asp214 and Glu217) were shown to be responsible for cleavage of the cellulose chain. Further studies allowed a complete mapping of cellulose binding along the 50 A ˚ -long active site tunnel [10,11]. Tr_Cel7A binds 10 glucosyl units in subsites )7 to +3 (numbering starts from the point of glycosi- dic bond cleavage, between )1 and +1; negative num- bers indicate the nonreducing end of the cellulose chain, and positive numbers, the reducing end [12]). The +1 and +2 sites are often designated as the ‘product sites’, as they bind the cellobiose unit that will -7 -4 -2 -1 +1 +2 Fig. 1. Binding of disaccharides to Pc_Cel7D. Overall structure of Pc_Cel7D with cellobiose bound in the +1 ⁄ +2 sites. Backbone of the enzyme’s catalytic domain is coloured terracotta, aromatic side chains that form cellulose-binding subsites are green, and the three acidic residues involved in catalysis are red. Cellobiose is indicated by a ball-and-stick model coloured light blue. Numbers indicate the position of some of the glucosyl-binding subsites. W. Ubhayasekera et al. Cel7D ⁄ saccharide complex structures FEBS Journal 272 (2005) 1952–1964 ª 2005 FEBS 1953 be cleaved off at the reducing end of the chain (Fig. 1). These sites are placed close to the exit of the binding site cleft⁄ tunnel, which should simplify the subsequent release of the disaccharide. However, prod- uct inhibition is commonly observed for cellobiohydro- lases [13–16]. Structures are also known for three endoglucanases of family 7, from T. reesei [17], Humicola insolens [18] and Fusarium oxysporum [19]. The active sites of these enzymes are very similar and the enzymes are believed to use the same catalytic mechanism as the cellobio- hydrolases. As expected, the loops flanking the cellu- lose-binding cleft in each case are significantly shorter, leaving the active sites completely open to solvent. Cellobiohydrolase Cel7D (previously called CBH 58) is the major cellobiohydrolase produced by the basi- diomycete Phanerochaete chrysosporium under most growth conditions [20]. We recently solved the struc- ture of Pc_Cel7D, and showed that it is similar to Tr_Cel7A [21]. The catalytic residues were identified as Glu207, Asp209 and Glu212. Nearly all interacting residues of Tr_Cel7A are conserved, which suggested that Pc_Cel7D would bind cellulose in much the same way. However, several deletions make the binding tunnel slightly more open in Pc_Cel7D. A recent comparative study revealed striking differ- ences in the activity on insoluble model substrates: although Pc_Cel7D had only slightly higher activity on cellotetraose, it hydrolysed amorphous and bacterial microcrystalline cellulose eight times and 4.4 times fas- ter, respectively, than Tr_Cel7A. Enzyme kinetics on p-nitrophenyl lactoside gave similar k cat values for the two enzymes; however, Pc_Cel7D showed a threefold higher K m (and hence threefold lower k cat ⁄ K m ) as well as reduced cellobiose inhibition (eight times higher K i ) [22]. Furthermore, estimation of specificity constants (k cat ⁄ K m ) for dinitrophenyl-cellooligosaccharides with 2–5 glucose units, pointed at differences between the enzymes in the relative contribution of intrinsic bind- ing energy to catalysis at subsites )3to)5. Another study revealed differences in the binding specificity for cellobiose and lactose, presumably at the product sites +1 ⁄ +2. While Tr_Cel7A prefers binding of cellobiose to lactose, the opposite is true for Pc_Cel7D [23]. As part of global efforts to replace fossil fuels with renewable energy sources, cellulases have received increasing attention as a possible means of converting cellulosic biomass to fermentable sugars for ethanol production [24]. However, the enzyme cost is a critical factor, and improvements in the efficiency of the pro- cess will directly influence whether such ‘bioethanol’ can effectively compete with petroleum [25]. The major industrial source of cellulase enzymes at present is T. reesei [26]. Deletion of individual cellulase genes in T. reesei showed that Tr_Cel7A was rate limiting in the degradation of crystalline cellulose in the fungal system [27]. Understanding the molecular details of how the Cel7 enzymes work thus lies at the heart of finding the best solution in future applications. In the present paper, we report three structures of Pc_Cel7D in complex with disaccharides: the product (cellobiose) and two inhibitors (lactose and cellobio- imidazole). These structures provide a picture of two different glycosyl binding modes, as well as explaining the differences in affinity between the two natural sugars. A structure obtained in the presence of cellobi- ose, methyl 4-S-b-cellobiosyl-4-thio-b-cellobioside (GG- S-GG), Tris ⁄ HCl and calcium revealed that Tris binds in the active site. In kinetic studies, we show that Pc_Cel7D is in fact inhibited by both Tris and calcium at the concentrations used in the crystallization; this is the first report of such behaviour within the family. Results Overall structures Deglycosylated Pc_Cel7D catalytic module was crystal- lized in the presence of two natural disaccharides, cell- obiose and lactose, as well as with cellobioimidazole, a compound that mimic the transition state of some cel- lulases [28]. The crystals were isomorphous with pre- vious ones [21] and complete diffraction data sets to 1.7 A ˚ resolution or better could be collected using syn- chrotron radiation. In all three cases, clear electron density was found for the bound ligand prior to its inclusion in the models (Fig. 2). Statistics relating to the diffraction data and the final refined models are summarized in Table 1. Each model contains the com- plete catalytic module of Pc_Cel7D (residues 1–431), an N-acetylglucosamine residue bound to Asn286, one molecule of the respective ligand and a number of bound waters. The protein structures are very similar to each other and to the published structure of Pc_Cel7D {Protein Data Bank (PDB) [29] entry code 1GPI [21]} with overall r.m.s. differences of 0.2–0.3 A ˚ when all Ca atoms are compared pair-wise. Binding of cellobiose to Pc_Cel7D Product inhibition in Pc_Cel7D is consistent with the observed binding of cellobiose in the +1 ⁄ +2 (pro- duct) sites of Pc_Cel7D (Figs 1, 2A and 3A). The nonreducing end of the disaccharide is in the +1 site; this glucosyl unit shows the ‘classical’ stacking on a tryptophan residue that is a feature in many proteins Cel7D ⁄ saccharide complex structures W. Ubhayasekera et al. 1954 FEBS Journal 272 (2005) 1952–1964 ª 2005 FEBS interacting with carbohydrates [30]. The hydrophobic B-face of the sugar thus makes a number of nonpolar contacts with the indole ring of Trp373 (Fig. 3A). The interactions on the opposite (A) face of this carbohy- drate unit are more polar. O2 interacts with the main- chain carbonyl oxygen of Asp248. O3 and O4 are linked to the catalytic acid Glu212 via hydrogen bonds with water. In addition, the guanidino group of Arg240 forms hydrogen bonds with O5 and O6. The electron density for the side-chain atoms of Arg240 is slightly weaker than average. Both the structural set- ting and the density suggest that the interactions of Arg240 with sugar compete with a salt link to Asp248, and a hydrogen bond to Gln172. Fewer interactions are seen in subsite +2, and electron density of this glycosyl unit is also somewhat weaker than that observed for the +1 sugar. The observed interactions are on the same side of the cleft as those in the +1 subsite. The guanidino group of Arg391 is within hydrogen bonding distance to O1, O5 and O6 of the sugar. O1 also interacts with Asp336, and O6 with a solvent molecule. There is, however, no aromatic stacking in the +2 subsite. Both glucosyl rings adopt a regular 4 C 1 chair, i.e. a favourable conformation in solution. The planes of the two sugar units have opposite orientations, with torsion angles (/ ¼ )78°, w ¼ +120°) that deviate slightly from those observed in the small-molecule crystal structure of cellotetraose ()93, +96, and )93, +86) [31]. Of the inter-residue interactions that stabil- ize cellulose chains, only the O3 i+1 –O5 i hydrogen bond is present; that between O6 i+1 and O2 i is lack- ing. The less common gauche-gauche conformation of the exocyclic C6–O6 bond is apparently stabilized by its interaction with Arg391, which is preferred to an intramolecular one with O2. The sugar is tightly sand- wiched between the walls of the product sites by the interactions with protein described above. The hydro- xyls along one edge of the disaccharide point into the binding cleft, where several water molecules are found; hydroxyls along the other edge point out toward the bulk solvent. Binding of lactose to Pc_Cel7D Lactose is an effective competitive inhibitor of Pc_Cel7D (Table 2). The only chemical difference between this disaccharide and cellobiose is the confi- guration at C4 in the galactosyl unit: the hydroxyl group is equatorial in cellobiose, and axial in lactose. As might be expected, lactose binds in the +1 ⁄ +2 subsites in a manner very similar to that described +2 +1 -1-2 -3 -4 -5 A cellobiose B lactose C cellobioimidazole D GG-S-GG + TRIS + cellobiose Fig. 2. Electron density for ligands bound in Pc_Cel7D. Final 2F o -F c maps contoured at 1 r, are shown for (A) cellobiose (B) lactose (C) cello- bioimidazole and (D) GG-S-GG, Tris and cellobiose. The numbers for the glucosyl-binding subsites indicate the location within the substrate binding tunnel of Pc_Cel7D. W. Ubhayasekera et al. Cel7D ⁄ saccharide complex structures FEBS Journal 272 (2005) 1952–1964 ª 2005 FEBS 1955 above for cellobiose (Figs 3B and 4A). However, the axial placement of O4 allows the disaccharide to make a direct hydrogen bond to Arg240. The guanidino group of that side chain has rotated slightly, so that it interacts with O4 and O5, instead of O5 and O6 (Fig. 3B). At the same time, Arg240 can make more favourable interactions with Gln172 and Asp248. O6 now appears to lack a fixed hydrogen-bonding part- ner. The interactions of the glucosyl unit in the +2 site are the same as those observed for cellobiose. However, the electron density for both sugar and protein in the immediate area is significantly better than that observed for cellobiose (Fig. 2A and B), and the temperature factors for the ligand are corre- spondingly lower (Table 1). These differences provide a structural basis for the observation that Pc_Cel7D binds more tightly to lactose than to cellobiose (Table 2). Binding of cellobioimidazole in Pc_Cel7D Monosaccharide-derived imidazoles such as cellobio- imidazole (Fig. 2C), feature an sp 2 -hybridized ano- meric centre and a charge distribution that mimics the transition state of some exoglycosidase reactions [32,33]. Cellobioimidazole is apparently not sufficiently compatible with the transition state of Pc_Cel7D to promote its binding in the catalytic substrate sites )2 ⁄ )1. The preferred binding is instead in sites +1 ⁄ +2, as was seen for the other two disaccharides (Fig. 3). The cellobioimidazole is, however, shifted more than 2 A ˚ along the cleft, toward the catalytic centre (Fig. 4A). The glucosyl unit in the +1 site continues to make good stacking interactions with Trp373, although it now lies more directly against the six-membered ring of the indole. O2 maintains the hydrogen bond to 248- O, but the sugar oxygen is displaced  1A ˚ from its position in the cellobiose and lactose complexes; this is the only one of the polar interactions that is preserved in this site. The hydrogen-bonding capacity of O3 is saturated by interactions with the side chains of His223, Asp209 and Glu212, and with a solvent mole- cule. O4 also makes a hydrogen bond with Glu212, but O6 appears to have no hydrogen-bonding partner in the enzyme. Like the sugars of cellobiose and lac- tose, the glucosyl unit in cellobioimidazole adopts a regular 4 C 1 chair conformation. In the +2 site, the glucoimidazole ring makes hydrogen bonds to Arg240 and Arg391, as well as to several solvent molecules. Under the crystallization conditions (pH 7.0), only a small fraction of the cello- bioimidazole will be protonated (pK a  6.1 [34]), and charge–charge repulsion is apparently not a problem. Unlike the glucosyl and galactosyl units, the glucoimi- dazole moiety cannot adopt a 4 C 1 chair conformation because of its C1–N5 double bond. The most favour- able solution conformation is the 4 H 3 half-chair observed in the crystal and NMR structures of glucoimidazole alone, although several other confor- mations are also possible within its pseudo-rotational sequence [35]. In the complex with Pc_Cel7D, the A cellobiose 212 391 240 207 373 C cellobioimidazole His223 212 207 209 240 373 Arg391 Asp336 Asp248 Arg240 Gln172 Glu207 nucleophile Asp209 Trp373 Arg256 Glu212 acid/base B lactose +2 +1 Fig. 3. Interactions between the disaccharides and Pc_Cel7D. Hydrogen-bonding interactions are shown for (A) cellobiose (B) lactose and (C) cellobioimidazole. In each case, the protein is shown with gold carbon atoms, while those of the ligand are yellow. Hydrogen bonds are indi- cated by cyan-coloured ‘bubbled’ lines. Water molecules interacting with protein and ligand are small light-blue spheres. Cel7D ⁄ saccharide complex structures W. Ubhayasekera et al. 1956 FEBS Journal 272 (2005) 1952–1964 ª 2005 FEBS conformation is closest to an envelope form, with C3 out of plane. The electron density (Fig. 2C) and temperature fac- tors (Table 1) for cellobioimidazole, as well as its good interactions with protein, suggest that it should be an effective cellobiohydrolase inhibitor. No affinity meas- urements are yet available for this compound with Pc_Cel7D, but the related Tr_Cel7A is indeed inhibited by cellobioimidazole, although, at least at pH 5.7, bind- ing is weaker than for cellobiose (Table 2) [36]. Due to the close proximity of the imidazole ring to the guanidi- no groups of arginines 256 and 391, it seems likely that the ligand will be a better inhibitor at pH 7 or higher, when the glucoimidazole ring is unprotonated, than at lower pH. More biochemical data are obviously needed. Binding of thio-linked substrate analogue GG-S-GG and cellobiose Crystals of Pc_Cel7D were soaked with a combination of cellobiose and the thio-linked sugar GG-S-GG, in hopes of obtaining a complex that included sugars bound in both ends of the active-site cleft. As seen from the electron density in Fig. 2D, the product sites +1 ⁄ +2 were again completely occupied with disac- charide, in the position observed with cellobiose alone. Table 2. Selected binding and kinetic constants for Pc_Cel7D and Tr_Cel7A. Enzyme K d , cellobiose (l M) K d , lactose (l M) K i , cellobio-imidazole (l M) K m , pNP-Lac (l M) k cat , pNP-Lac (s )1 ) Pc_Cel7D 115 a 77 a NA 5100 b 0.17 b Tr_Cel7A 20 c 310 c 130 d 900 b 0.10 b a K d values at pH 5.0 and 25 °C from displacement chromatography experiments with Pc_Cel7D immobilized on silica, as published by Hen- riksson et al. [23]. b Michaelis–Menten kinetic parameters at pH 5.0 and 25 °C [23]. c K d at pH 5.0 and 25 °C determined by protein differ- ence spectroscopy [16]. d Non-competitive K i at pH 5.7 and 30 °C from inhibition experiments using 2-chloro-nitrophenyl b-lactoside as substrate [36]. NA, data not available. Table 1. Data collection and refinement statistics. The space group was C2. Statistics for the highest resolution shell are given in paren- theses. A stringent boundary Ramachandran plot was used [47]. Data collection statistics were taken from TRUNCATE [48]. Other values for the refined structures were calculated using MOLEMAN2 [49]. Data collection Complex Cellobiose Lactose Cellobioimidazole Cellobiose, Tris, GG-S-GG Environment ESRF, ID14 : 4 ESRF, ID14 : 4 ESRF, ID14 : 2 ESRF, ID14 : 3 Wavelength 0.9370 0.9322 0.9330 0.9310 Cell dimensions 87.4, 46.8, 99.5 86.9,46.7, 99.1 87.4, 46.6, 98.5 86.47, 46.47, 98.36 (A ˚ , °) b ¼ 103.0° b ¼ 102.8° b ¼ 102.6° b ¼ 102.49 Resolution (A ˚ ) 50–1.70 (1.73–1.70) 50–1.60 (1.63–1.60) 43–1.70 (1.79–1.7) 96.321–1.7 (1.79–1.70) Unique reflections 43 328 50 272 39 785 50967 Average multiplicity 4.3 3.4 2.3 3.6 Completeness (%) 99.6 (93.1) 99.3 (99.6) 93.2 (92.4) 96.6 R merge 9.8 (32.1) 7.2 (23.8) 5.2 (19.4) 10.9 <I⁄ rI > 11.0 (5.8) 14.8 (8.1) 14.0 (3.8) 11.6 (4.3) Refinement Number of reflections 40 782 (100.0) 47 274 (100.0) 37 334 (100.0) 37 329 (96.32) (completeness,%) Resolution range (A ˚ ) 40.0–1.70 39.2–1.61 36.3–1.70 95.35–1.72 R-factor ⁄ R-free (%) 17.2 (20.2) 15.7 (20.2) 17.8 (22.7) 17.1 (23.24) Number of protein atoms (Average B, A ˚ 2 ) 3198 (24.7) 3198 (22.2) 3198 (17.5) 3198 (20.6) Number of water molecules (Average B, A ˚ 2 ) 106 (27.9) 278 (31.2) 180 (21.9) 389 (30.8) Number of N-acetyl-glucosamine atoms 14 (37.2) 14 (29.1) 14 (26.5) 14 (34.0) (Average B, A ˚ 2 ) e Number of ligand atoms (Average B, A ˚ 2 ) 23 (33.7) 23 (24.0) 25 (16.9) 31 (24.2) r.m.s bond length (A ˚ ) 0.027 0.010 0.006 0.013 r.m.s. bond angle (°) 2.08 1.51 1.21 1.50 No. Ramachandran plot outliers (%) 4 (1.1) 3 (0.8) 3 (0.8) 3 (0.8) W. Ubhayasekera et al. Cel7D ⁄ saccharide complex structures FEBS Journal 272 (2005) 1952–1964 ª 2005 FEBS 1957 The other side of the cleft contains additional electron density not observed in the other Pc_Cel7D structures, which may indicate binding of the longer sugar at low occupancy. Although the shape of the density in sites )4 and )5 bears strong resemblance to glucose rings, on the whole it is too weak to allow unambiguous modelling of this ligand within the active site. The low occupancy is almost certainly due to the limited crystal soaking time used (2 min); the crystals deteriorated at longer soaking times. Co-crystallization could not be used because the long exposure time would lead to hydrolysis of the O-glycosidic bonds in the ligand. We are currently seeking other solutions to this problem. However, clear electron density in the immediate vicinity of the catalytic residues of this structure was only compatible with Tris, among the known crystal- lization reagents. Inhibition by Tris had not been reported previously for Cel7 enzymes. Inhibition experiments The discovery of Tris density in the active site promp- ted us to undertake a systematic study of the com- ponents of the crystallization solution. Inhibition experiments at pH 7.0 using p-nitrophenyl lactoside as substrate (summarized in Fig. 5) showed that both 10 mm Tris ⁄ HCl and 5 mm CaCl 2 individually inhibit Pc_Cel7D. As there was no inhibition with 10 mm NaCl, the Tris and the calcium ions are the inhibiting species. Comparison with ligand binding in Tr_Cel7A To date, five structures have been published for the related cellobiohydrolase of T. reesei (Tr_Cel7A) with carbohydrates bound in sites +1 ⁄ +2: the wild-type enzyme with the inhibitor o-iodo-benzyl-b-d-thio- glucoside (IBTG; PDB entry 1CEL [9]), an inactive B A C Fig. 4. Superposition of sugar residues bound in the +1 ⁄ +2 sites of Pc_Cel7D and Tr_Cel7A. Selected residues of Pc_Cel7D are shown with gold carbon atoms and of Tr_Cel7A with blue carbon atoms. (A) Superposition of the three disaccharides as bound by Pc_Cel7D. Cellobiose (magenta), lactose (lilac) and cellobioimidazole (cyan) are shown as ball-and-stick representations. Only cellobioimidazole enters directly into the active site, forming hydrogen bonds to the catalytic acid Glu212. The interactions of O2 with 248-O in the +1 site, and Arg391 with O6 in the +2 site, are the only direct polar interactions found in all three complexes (Results). Two additional conserved interactions are shown, in which water molecules medi- ate links between sugar hydroxyls and protein residues at the deep- est point of the binding cleft. The O6 hydroxyl in subsite +2 thus also interacts with a water molecule bound to Asp251 OD2. A sec- ond water molecule links Thr221-OG1 to O2 in site +1. These inter- actions hold O2 of the +1 subsite and O6 of the +2 subsite in very similar positions in all three complexes. (B) Cellobiose binding near the catalytic residues in Pc_Cel7D (magenta ligand) is shown together with the complex of cellobiose with Tr_Cel7A (yellow lig- and). In this binding mode there is room for water (pale green spheres) between the O4 hydroxyl of the hexose in site +1 and the catalytic acid ⁄ base (Glu212 in Pc_Cel7D). (C) Superposition of avail- able cellobiohydrolase structures with sugars bound in the +1 ⁄ +2 sites highlighting the existence of two discrete binding modes. The bound sugar residues are colour-ramped using a rainbow, with blue indicating the position closest to the point of cleavage, and red, that farthest away. The identity of each complex is indicated by coloured boxes. Conserved interactions involving water are also shown. Cel7D ⁄ saccharide complex structures W. Ubhayasekera et al. 1958 FEBS Journal 272 (2005) 1952–1964 ª 2005 FEBS E212Q mutant with cellobiose (3CEL [10]), cellotetra- ose (5CEL [11]) and cellopentaose (6CEL [11]), and a second inactive mutant (E217Q) with cellobiose (together with cellohexaose bound in sites )7to)2; 7CEL [11]). The catalytic domains of Tr_Cel7A and Pc_Cel7D share 55% amino-acid sequence identity, which provides a good basis for detailed comparison of the two enzymes. The complex of Tr_Cel7A with cellobiose (3CEL) can be superimposed on that of Pc_Cel7D using a cut-off of 0.7 A ˚ , giving an r.m.s. difference of 0.4 A ˚ for 267 matching Ca atoms ( 60% of the total). The most similar portions of the enzymes represent the core b-sheet structure as well as the highly con- served residues of the active site. Cellobiose is bound in an equivalent position in the two enzymes (< 0.8 A ˚ difference for all atoms); the position of O6 within the +2 site is identical (Fig. 4B). Although this Tr_Cel7A structure actually represents an inactive mutant, the mutated residue (equivalent to Pc_Cel7D’s Glu207) is not directly involved in lig- and binding, and does not appear to complicate the comparison. The most significant differences between the two complexes result from a deletion in one of the active-site loops of Pc_Cel7D (Fig. 6). The role of Arg240 in binding to O5 and O6 in the +1 site is thus assumed by Arg251 in Tr_Cel7A. The main- chain atoms of Arg251 are  3A ˚ away from those of Arg240, but the functional guanidino groups are similarly placed and serve a similar purpose in the two enzymes. However, Arg251 in Tr_Cel7A is sup- ported by a better local network of hydrogen bonds, including Thr246 and Asp259. The longer loop in Tr_Cel7A also provides one additional direct hydro- gen bond to the ligand: Thr246 interacts with O6 in the +1 site. On the other hand, a deletion in Tr_Cel7A results in the loss of an interaction at O1 in site +2, which can be provided by Asp336 in Pc_Cel7D (Fig. 6). All of the other interactions that Pc_Cel7D makes with cellobiose are found intact in the complex with Tr_Cel7A. Better hydrogen bond- ing, together with the more enclosed Tr_Cel7A active site, provides a reasonable explanation for why cello- biose binds more tightly to this enzyme than to Pc_Cel7D (Table 2), and so there is less product inhibition in the latter enzyme. Fig. 6. Comparison of the Pc_Cel7D and Tr_Cel7A structures near the exit of the cellulose-binding tunnel. The complex of Pc_Cel7D with cellobiose is in gold and coral carbons, while Tr_Cel7A is in green. The sugar is embraced in Tr_Cel7A by the 245–250 loop (seen at the upper left), but not in Pc_Cel7D that has a six-residue deletion here. Pro258 in Tr_Cel7A may hold Arg251 out of reach for O4 of lactose. At the bottom of the figure it is seen that the inser- tion of Asp336 in Pc_Cel7D results in differences in main-chain con- formation and provides an additional hydrogen bond with the reducing-end hydroxyl of a bound cellulose chain. B A Fig. 5. Inhibition of Pc_Cel7D by Tris and other compounds. Absorbance at 400 nm was measured after a 30-min incubation of Pc_Cel7D with pNP-Lac at 30 °C, pH 7.0, as described in Experi- mental procedures. (A) h, No inhibitor; ·,10m M NaCl; +, 10 mM Tris ⁄ HCl; –, 5 mM CaCl 2 ;*,5mM CaCl 2 and 10 mM Tris ⁄ HCl. (B) h, No inhibitor; n, 0.1 m M cellobiose; s, 0.1 mM cellobiose and 10 m M Tris ⁄ HCl; e, 0.1 mM cellobiose, 10 mM Tris ⁄ HCl and 5 mM CaCl 2 . W. Ubhayasekera et al. Cel7D ⁄ saccharide complex structures FEBS Journal 272 (2005) 1952–1964 ª 2005 FEBS 1959 There is presently no structure available for Tr_Cel7A in complex with lactose, but some points seem clear in a comparison of the two enzymes. In Pc_Ce7D, the flexibility of the Arg240 side chain is an important factor in binding of the various disaccha- rides. In the lactose complex particularly, Arg240 has adapted its conformation to give a good hydrogen- bonding network that includes the axial O4 of the galactosyl unit. Possibilities are different for the Arg251 in Tr_Cel7A, since both steric and hydrogen- bonding options will be altered locally. In Pc_Cel7D, lactose also has an additional H-bond from O1 to Asp336 in the +2 site. Apparently, the combined differences do not favour the binding of lactose over cellobiose by Tr_Cel7A (Table 2). Discussion Our structural results provide a good framework for understanding why Pc_Cel7D binds lactose more tightly than cellobiose, and why Tr_Cel7A exhibits the opposite behaviour. Furthermore, the distinct mode of binding exhibited by the complex of Pc_Cel7D with cellobioimidazole prompted a closer look at the avail- able structural data on binding in the +1 and +2 sites in these enzymes (Fig. 4C). The nonreducing end of the disaccharide in each of the available complex struc- tures is in the +1 site, and the reducing end in the +2 site. However, two quite distinct binding modes are clearly present. In one scenario, the hexose in site +1 is close to the active centre, with its O4 hydroxyl bound to the catalytic acid Glu212 (equivalent to 217 in Tr_Cel7A). This type of binding is found in the Pc_Cel7D ⁄ imidazole complex, and in the complexes of Tr_Cel7A with IBTG, cellopentaose, and cellobiose + cellohexaose (1CEL, 6CEL, 7CEL). In the second mode, the sugar is shifted  2A ˚ away from the cata- lytic centre, leaving room for a water molecule between the sugar and the catalytic acid. This type of binding is observed for cellobiose and lactose in Pc_Cel7D, and for complexes of cellobiose or cellotetraose with Tr_Cel7A (3CEL and 5CEL, respectively). The loca- tion of O2 in the +1 site, and of O6 in the +2 site, is very similar in all structures; the two primary binding modes appear to result from a pivoting motion around these points. As the disaccharide moves away from the catalytic residues, the sugar in the +1 site moves out toward the bulk solvent, and the sugar in the +2 site moves deeper into the cleft ⁄ tunnel of the enzymes. Processive hydrolysis of cellulose requires that the enzyme can slide along a cellulose chain. The tunnels of Pc_Cel7D and Tr_Cel7A are wide enough to allow the passage of the chain, apparently without need for conformational changes in the protein [11,21]. When the reducing end of the chain has passed the active centre and entered into the product binding sites, the glucose residue at site )1 still has sufficient space to remain in the most stable 4 C 1 chair conformation. However, in order for hydrolysis to take place the )1 glucosyl must approach the catalytic nucleophile at the bottom of the catalytic centre; this requires a flip from the chair into the boat conformation, and a concomit- ant bending of the cellulose chain at this position. Our observations hint at events near the active site during catalysis. We propose that the docking mode where the disaccharide unit is placed immediately at the active site (as observed, e.g. for the complex of Pc_Cel7D with cellobioimidazole) represents a ‘cut’ mode. We will refer to the other docking position, that where the disaccharide is slightly further away (as for the cellobiose and lactose complexes), as the ‘slide’ mode. For the cellulose chain to be cleaved, it must first dock in the ‘cut’ mode, and the )1 glucosyl must flip from a chair to a boat conformation. The O3 hydroxyl in site +1 then points down towards the deepest part of the cleft, and interacts directly with the acid ⁄ base Glu212-OE1. It is now also within hydrogen-bonding distance of the catalytically important Asp209 and His223. The other carboxylate oxygen of Glu212 hydrogen-bonds to O4. In a true enzyme–substrate complex, this O4 hydroxyl would actually be the gly- cosidic oxygen that links the sugar in site +1 to that in site )1; the hydrogen bond between Glu212 and O4 is suggestive of the protonation of the glycosidic oxy- gen in the transition state. In the transition state, the reducing end of the cellulose substrate (i.e. the cello- biosyl unit in sites +1 ⁄ +2) remains bound in the ‘cut’ mode, with the glycosidic oxygen bound to Glu212. Once the cellulose chain is cleaved, the cellobiose product can remain bound, but pivots into the ‘slide’ mode. Now, the positions previously occupied by the O3 and O4 hydroxyls are filled by water molecules to which O3 and O4 bind. The water molecule that lies between O4 of the hexose in site +1 and the acid ⁄ base Glu212 is compatible with the existence of the inter- mediate, and well positioned to perform a nucleophilic attack on its anomeric carbon. Therefore, it seems possible that cleavage of the intermediate is followed by release of the product, rather than the product necessarily leaving prior to hydrolysis, as has always been proposed. Indeed, the cellobiose could be essential to proper positioning of the catalytic water. After the water attack, the glucosyl unit at the new reducing end of the cellulose chain (that in the )1 site) will have a new hydroxyl group in the b-configuration. When the Cel7D ⁄ saccharide complex structures W. Ubhayasekera et al. 1960 FEBS Journal 272 (2005) 1952–1964 ª 2005 FEBS sugar unit flips back from the boat conformation to an ordinary 4 C 1 chair, this would be expected to create steric problems that would force the cellobiose product (rather than the longer and more firmly bound cellulose chain on the substrate side) to leave the site. Thus the energy released when the high-energy glycosyl-diester bond is cleaved would provide kinetic energy used to drive product release. Such a mechanism might explain why cellobiose is a much weaker inhibitor of Tr_Cel7A ( 80-fold difference in K i ) when it acts processively on cellulose than with soluble substrates [37]. One might expect that the enzyme had evolved for optimal interactions with the cellobiosyl unit in the ‘cut’ mode, in order to maximize transition-state stabil- ization. Since cellobiose clearly makes more favourable interactions with Pc_Cel7D in the ‘slide’ mode, stabil- izing the transition state cannot be the sole considera- tion for this enzyme. A key residue in stabilizing the ‘slide’ mode in Pc_Cel7D, Asp336, is located at the very end of the binding cleft ⁄ tunnel (Fig. 6). The cor- responding residue is deleted in Tr_Cel7A. An align- ment of 53 nonredundant sequences retrieved from the ProDom server (http://prodes.toulouse.inra.fr/prodom/ 2004.1/html/home.php) indicates that family 7 endo- glucanases lack this structural motif, although it is rather well conserved in the cellobiohydrolases. In 31 out of 41 cellobiohydrolase sequences, the segment has the same length, and the aspartate is conserved. In another five the length is conserved, but not the aspar- tate. In two of these, the aspartate is replaced with glu- tamate that might play a similar role. Four sequences are shorter (by one residue), including three enzymes of Trichoderma species (T. reesei, T. viride, T. harzia- num) and one from Thermoascus aurantiacus (Swiss- Prot accession code Q96UR5). There is also a single Cel7 sequence with a one-residue insertion; this seg- ment includes two glutamates, but no aspartate (Lep- tosphaeria maculans Cel2, Q9P8K7). We anticipate that the differences will be indicative of different kin- etic properties in the respective enzymes. Among the enzymes that can stabilize the ‘slide’ mode in this way, one might also expect a reduction of transglycosylation activity, since the product would be too distant to per- form the reverse reaction. The Pc_Cel7D structure was used previously for homology modelling of the other five family 7 iso- enzymes in P. chrysosporium [21]. We predicted that the catalytic properties of Pc_Cel7C, E and F would be very similar to those of Pc_Cel7D ( 80% identity), while Pc_Cel7A and B were expected to be more distinct (66% identity). Re-evaluation of the models with the present structural data indicates that the two residues (Arg240 and D336) implicated as important for binding in the product sites are conserved in C, E and F, but not in A and B isozymes. Arg240 is replaced by Ser (in A) or Ala (in B), while Asp336 is replaced by Glu (in A) or Gly (in B). The differences would be expected to reflect different kinetic proper- ties, and possible endoglucanase activity. Our data also provide the new information that Pc_Cel7D is inhibited by both Tris and calcium. As these conditions resemble those in the crystallization solutions used here, it is not surprising that Tris is observed in the active site of the complex with GG-S- GG, although the tetrasaccharide is observed at only low occupancy (Fig. 2D). The position of Tris near the nucleophile Glu207 and its partner Asp209, as well as the acid ⁄ base Glu212, provides a clear explanation for the inhibition. Re-inspection of previous data con- firmed that Tris is not present in the structures with apo enzyme or the disaccharides alone, indicating that some degree of synergy exists in its binding with the thio-linked sugar. Comparison of the catalytic-site regions suggests that these compounds will also bind to and inhibit Tr_Cel7A. Such inhibition has not pre- viously been reported for enzymes in glycoside hydro- lase family 7, although unrelated proteins with similar catalytic sites, such as the family 13 amylase [38] are known to be inhibited by Tris. Although the amylase has a completely different structure, based on a (b ⁄ a) 8 barrel, it too is a retaining enzyme that binds Tris in the )1 site in the immediate vicinity of the nucleophile and catalytic acid. Experimental procedures Preparation of protein, crystallization and data collection Intact Cel7D protein from P. chrysosporium was the kind gift of Gunnar Johansson, Department Biochemistry, Uppsala University. Preparation of the deglycosylated cata- lytic module of Cel7D has been described previously [21]. Hanging-drop vapour diffusion experiments included 18 mgÆmL )1 protein, 10 mm Tris ⁄ HCl pH 7.0, 5 mm CaCl 2 , 15–22.5% polyethylene glycol 5000 and 12% glycerol. Sin- gle-soaking experiments of Cel7D crystals were performed with 10 mm cellobiose, 10 mm lactose (Sigma, St. Louis, MO, USA) and 5 mm cellobioimidazole {(5R,6R,7S,8S)-6- (b-d-glucopyranosyloxy)-5,6,7,8-tetrahydro-5-[(hydroxy)methyl] imidazo[1,2-a] pyridine-7,8-diol [36]}, respectively. A dou- ble-soak experiment was performed with 10 mm cellobiose followed by 0.5 mm of the thio-linked cellotetraoside, GG- S-GG. Soaking time of the shorter ligands was almost 10 min, but even after 1 day these crystals were stable. However, in the case of the GG-S-GG soaks, crystals were W. Ubhayasekera et al. Cel7D ⁄ saccharide complex structures FEBS Journal 272 (2005) 1952–1964 ª 2005 FEBS 1961 [...]... 1Z3T, 1Z3V, and 1Z3W for the cellobiose, lactose and cellobioimidazole complexes, respectively Additional cellulase structures were obtained from the Protein Data Bank, and were aligned and compared with the present ones using the programs lsqman [45] and o [44] Figures were prepared using o and molray [46] Inhibition studies Inhibition of intact Pc _Cel7D was tested at pH 7 with 2 lm enzyme in 50 mm sodium... Several rounds of rebuilding using the program o [44] and the placement of water, a covalently bound residue of N-acetylglucosamine and ligands into the electron density, resulted in the four structures described here Statistics after crystallographic refinement are summarized in Table 1 Coordinates for the final models, and the corresponding structure factor data, have been deposited in the PDB with entry... N-acetylamino imidazopyridines and their evaluation as inhibitors of glycosidases Synthesis (Stuttgart) 1459–1468 35 Granier T, Panday N & Vasella A (1997) Structureactivity relations for imidazo-pyridine-type inhibitors of beta-d-glucosidases Helvetica Chimica Acta 80, 979– 987 1963 Cel7D ⁄ saccharide complex structures 36 Vonhoff S, Piens K, Pipelier M, Braet C, Claeyssens M & Vasella A (1999) Inhibition of. .. 2096–2100 32 Panday N, Canac Y & Vasella A (2000) Very strong inhibition of glucosidases by C (2)-substituted tetrahydroimidazopyridines Helvetica Chimica Acta 83, 58–79 33 Heightman TD & Vasella AT (1999) Recent insights into inhibition, structure, and mechanism of configuration-retaining glycosidases Angewandte Chemie-Int Edition 38, 750–770 34 Panday N & Vasella A (1999) Synthesis of glucose- and mannose-derived... calculations [39,40] during refinement Initial phases were obtained using CCP4 [41] using the protein coordinates for apo Pc _Cel7D (PDB entry code 1GPI [21]) Refinement was carried out with REFMAC5 [42,43] and included rigidbody refinement as the first step As with previous data sets, the complex data were rather anisotropic, and the most successful refinement strategy made use of Babinet’s bulk solvent correction... Cambridge, UK) and microsoft excel Acknowledgements We are grateful to Dr Gunnar Johansson, Department of Biochemistry, Uppsala University, for providing us 1962 W Ubhayasekera et al with Pc _Cel7D protein, to Prof Hugues Driguez, CNRS-CERMAV, Grenoble, France, for providing the GG-S-GG ligand, and to Sabah Mahdi and Gunnar Berglund for some of the crystallization and data collection work Financial support... 500 lL of 0.2 m NaOH Absorbance was measured at 400 nm in a Beckman DU 640 spectrophotometer Controls were included to account for possible background absorbance of the enzyme, substrate and inhibitor solutions The amount of 4-nitrophenol released from pNP-Lac was calculated using an extinction coefficient of 16590 m)1Æcm)1 Results were fit using nonlinear regression with the programs ultrafit (Biosoft,... Comparison of the specificities of the cellobiohydrolases isolated from Penicillium pinophilum and Trichoderma reesei Biochem J 261, 819–825 17 Kleywegt GJ, Zou JY, Divne C, Davies GJ, Sinning I, ˚ Stahlberg J, Reinikainen T, Srisodsuk M, Teeri TT & Jones TA (1997) The crystal structure of the catalytic core domain of endoglucanase I from Trichoderma reesei at 3.6 A resolution, and a comparison with related... Schulein M, Withers SG & Davies GJ (1998) Crystal structure of the family 7 endoglucanase I (Cel7B) from Humicola insolens at 2.2 A resolution and identification of the catalytic nucleophile by trapping of the covalent glycosyl-enzyme intermediate Biochem J 335, 409–416 19 Sulzenbacher G, Schulein M & Davies GJ (1997) Structure of the endoglucanase I from Fusarium oxysporum: Native, cellobiose, and 3,4-epoxybutyl... Cellobiohydrolases from Phanerochaete chrysosporium: Crystal structure of the catalytic module ˚ of CBH58 (Cel7D) at 1.32 A resolution and homology models of the isozymes J Mol Biol 314, 1097–1111 FEBS Journal 272 (2005) 1952–1964 ª 2005 FEBS Cel7D ⁄ saccharide complex structures 22 von Ossowski I, Stahlberg J, Koivula A, Piens K, Becker D, Boer H, Harle R, Harris M, Divne C, Mahdi S et al (2003) Engineering the exo-loop . Structures of Phanerochaete chrysosporium Cel7D in complex with product and inhibitors Wimal Ubhayasekera 1 , Ine ´ s G. Mun ˜ oz 1, *, Andrea Vasella 2 ,. compared pair-wise. Binding of cellobiose to Pc _Cel7D Product inhibition in Pc _Cel7D is consistent with the observed binding of cellobiose in the +1 ⁄ +2 (pro- duct)