Crystal structure of mannanase 26A from Pseudomonas cellulosa and analysis of residues involved in substrate binding.

The crystal structure of Pseudomonas cellulosa mannanase 26A has been solved by multiple isomorphous replacement and refined at 1.85 A resolution to an R-factor of 0.182 (R-free = 0.211). The enzyme comprises (beta/alpha)(8)-barrel architecture with two catalytic glutamates at the ends of beta-strands 4 and 7 in precisely the same location as the corresponding glutamates in other 4/7-superfamily glycoside hydrolase enzymes (clan GH-A glycoside hydrolases). The family 26 glycoside hydrolases are therefore members of clan GH-A. Functional analyses of mannanase 26A, informed by the crystal structure of the enzyme, provided important insights into the role of residues close to the catalytic glutamates. These data showed that Trp-360 played a critical role in binding substrate at the -1 subsite, whereas Tyr-285 was important to the function of the nucleophile catalyst. His-211 in mannanase 26A does not have the same function as the equivalent asparagine in the other GH-A enzymes. The data also suggest that Trp-217 and Trp-162 are important for the activity of mannanase 26A against mannooligosaccharides but are less important for activity against polysaccharides.

Mannan is a major matrix polysaccharide in the cell walls of angiosperms, and comprises a backbone of ␤-1,4-linked mannose units (in glucomannans the backbone consists of mannose and glucose residues), which are decorated with galactose and acetate residues depending on its origin (1). Endo-␤1,4-mannanases (mannanases), which play a pivotal role in the cleavage of the backbone of mannans and glucomannans (2), have been isolated from plants (3), anaerobic (4) and aerobic fungi (5), and eubacteria (6,7). Hydrophobic cluster analysis and comparison of the sequences of these enzymes have been used to assign the mannanases to glycoside hydrolase (GH) 1 families 5 and 26. 2 Both enzyme families cleave the glycosidic bonds by a double displace-ment mechanism (9 -11). Family 26 consists mainly of mannanases and includes Pseudomonas cellulosa mannanase 26A (Pc Man26A). Endo-acting cellulases and mannanases are the major enzymes in family 5; recently the structure of a family 5 mannanase from Thermomonospora fusca (Tf Man5A) was solved (12). Family 5 glycoside hydrolases are members of the 4/7superfamily of glycoside hydrolases (also known as clan G-HA (13,14)), and it has been suggested that family 26 glycoside hydrolases (GH26) also belong to this superfamily (11). Three highly conserved residues in this superfamily are an adjacent Asn-Glu pair at the end of ␤-strand 4 and a Glu at the end of ␤-strand 7, although GH26 mannanases have histidine substituted in place of the asparagine. The importance of the substitution of an asparagine for a histidine in GH26 enzymes is unclear.
To determine whether GH26 enzymes are members of the GH-A clan, and to investigate the role of amino acids in the active site of these glycoside hydrolases, the three-dimensional structure of Pc Man26A was solved and a series of substrate-binding site mutants characterized.

Enzyme Expression, Purification, and Crystallization-Pc
Man26A expressed in Escherichia coli strain BL21 (Novagen) containing the plasmid pDB1 was purified to homogeneity in a single step using anion exchange chromatography (11). Single crystals were grown by the hanging drop method initially using Hampton Crystal Screen Kits to search for appropriate conditions. DNA Sequencing-The carboxyl-terminal 49 amino acids of Pc Man26A did not fit the electron density map, and the sequence of the 3Ј-end of the Pc Man26A gene (man26A) was therefore redetermined using an ABI dye-terminator kit and an ABI373 automatic sequencer. pDB1, which comprises the region of man26A encoding the mature mannanase cloned into NdeI/SalI-restricted pET21a (Novagen; Ref. 11), served as the template DNA and the vector's T7 terminator sequence as the primer. The data clearly showed that the previously determined sequence of man26A (7), was incorrect and that there was an additional C at position 1090 (nucleotide 1 is the A of the ATG initiation codon), resulting in a change in the reading frame at the 3Ј-end of the gene. The revised sequence of mature Pc Man26A, which is shown in Table I, could easily be fitted to the electron density map.
Mutants were initially identified by the presence of primer-derived restriction sites in the plasmids generated by this method. To ensure that only the desired mutations had been incorporated into the mannanase gene, the complete sequences of the mutated forms of man26A were determined. The mutants were all expressed at high levels, and circular dichroism spectra revealed that the mutants had a secondary structure similar to native Man26A.
Assays-The sources of the Man26A substrates used in this study were described previously (11) with the exception of mannotriose and mannohexaose, which were both obtained from Megazyme International Ltd. Protein concentration was determined by measuring A 280 (15) using a molar extinction coefficient of 93,600 M Ϫ1 cm Ϫ1 . Assays used to measure Man26A activity against 2,4-dinitrophenyl-␤mannobioside (2,4-DNPM 2 ), carob galactomannan, and azo-carob galactomannan were as described previously (11). To evaluate the activity of the mannanase against mannooligosaccharides, 0.02-460 nM enzyme (depending on the substrate used and the activity of the Man26A derivative) were incubated with 2-14 M substrate in 50 mM sodium phosphate, 12 mM citrate buffer, pH 6.5, for up to 500 min. At regular intervals, an 0.5-ml aliquot was removed, the enzyme was inactivated by boiling for 10 min, and the mannooligosaccharides in the samples were quantified by HPLC as described previously (11). The progress curves of oligosaccharide cleavage were used to determine the k cat /K m of the reaction using the following equation described by Matsui et al. (16), where k ϭ (k cat /K m )[enzyme], t represents time, and [S 0 ] and [S t ] represent substrate concentration at times 0 and t, respectively. The purity and conformational integrity of the mutant enzymes were assessed by SDS-polyacrylamide gel electrophoresis and circular dichroism spectroscopy, respectively, as described previously (11). Data Collection-High-resolution native data were collected using the EMBL BW7B beam line at the DORIS storage ring, DESY ( ϭ 0.8374 Å) equipped with a MAR345 image plate. Data from the monomethyl mercury derivative were collected using the EMBL X31 beam line with a MAR300 image plate at a wavelength of ϭ 1.00621 Å to optimize the anomalous signal. Data from the uranyl acetate derivative were collected in-house using a MacScience rotating anode generator with double mirrors and nickel filter ( ϭ 1.5418 Å) equipped with a DIP1030 image plate. The data were reduced and scaled using DENZO and SCALEPACK (17), and further calculations were made using the CCP4 program suite (18).
Structure Solution-A mercury derivative was prepared by soaking a crystal in 1 mM methyl mercury chloride for 8 h. A uranyl actetate derivative was prepared using 10 mM uranyl acetate, and the crystal was soaked for 2 h. The isomorphous and anomalous difference Pattersons for the mercury derivative revealed that the methyl mercury chloride had bound predominantly to a single site. The uranyl sites were located using cross-phased difference Fouriers. MLPHARE (19) was used for the refinement of the heavy atom sites and for the calculation of the protein phases that were improved using DM (20).
Model Building and Refinement-WARP (21) was used to automatically trace the chain of Pc Man26A and facilitate the building of a model using the graphics program O (22). Subsequent refinement used ARP (23) and REFMAC (24) to refine atomic positions and B-factors. The model was validated using PROCHECK (25).
Superimposition of Protein Structures and Modeling-Superimposition of Tf Man5A and of P. cellulosa xylanase 10A (Pc Xyn10A) on Pc Man26A was achieved using the program O (22) as was modeling of the enzyme-substrate complex.

RESULTS AND DISCUSSION
Crystallographic Analysis-Good quality single crystals were grown by the hanging drop method using a reservoir of 26% polyethylene glycol monomethyl ether 550 with 0.012 M zinc sulfate and 0.1 M MES buffer at pH 6.5. These crystals are tetragonal, space group P4 1 or P4 3 , with a ϭ 93.2 Å, c ϭ 54.8 Å and with a single molecule in the asymmetric unit. The tetragonal crystals can be cryocooled to 100 K directly for data collection because of their high content of low molecular weight polyethylene glycol monoethyl ether.
The results of the data collection, phasing, and refinement are presented in Table II. A single monomethyl mercury site and 12 uranyl sites were used in calculating the protein phases. The space group ambiguity was resolved by comparing the occupancies of the heavy atoms during heavy atom refinement in both space groups, and the space group was established as P4 1 . The final model comprises 337 residues in three protein chains, four zinc ions, and 446 water molecules. One of the zinc ions mediates a crystal contact. The final R-factor and R-free are 18.2 and 21.1% at 1.85 Å resolution, and the overall Gfactor from PROCHECK is ϩ0.20 (Table II contains  There is no evidence in the electron density map for residues 361-391, but SDS-polyacrylamide gel electrophoresis analysis of dissolved crystals showed the protein to have the correct molecular mass. This sequence is therefore present in the crystal but is highly mobile or disordered. Also disordered in the crystal are: the eight residues 324 to 331; the five aminoterminal residues, Arg-39 -Lys-43; and the four carboxyl-terminal residues, Leu-420 -Lys 423. The final protein model therefore comprises three chains (Pro-44 -Ile-323, Leu-332-Trp-360, and Thr-392-Thr-419). Adventitious binding of zinc occurs close to the active site of Pc Man26A, a result of the inclusion of 12 mM zinc sulfate in the crystallization conditions. Photon-induced x-ray emission analysis (26) of purified Pc Man26A showed that in buffers lacking zinc there is no zinc bound to the enzyme. The presence of zinc had no significant effect on the catalytic activity or stability of the enzyme.
Description of the Crystal Structure of Pc Man26A and Comparison with Tf Man5A and Other GH-A Enzymes-The ␣-carbon trace of the Pc Man26A polypeptide revealed the classic (␣/␤) 8 -barrel architecture (Fig. 1), as first seen in triose phosphate isomerase (27), with the major axis running from ␤1 to ␤5, typical of clan GH-A enzymes ( Fig. 2; Ref. 13). The disordered regions described above comprised the extreme aminoand carboxyl-terminal sequences and the loops between ␤7 and ␣7 (residues 324 -331) and ␤8 and ␣8 (residues 361-391). The putative acid/base catalyst (Glu-212) is at the end of ␤4, the nucleophile catalyst (Glu-320) is at the end of ␤7, Glu-212 and Glu-320 are 5.5 Å apart, and the spatial location of these amino acids is identical to the catalytic residues of other clan GH-A enzymes. Pc Man26A and, by inference, all other GH26 enzymes are therefore demonstrated to be members of clan GH-A. In addition, the ␤-bulges on strands 4 and 7 adjacent to the active site amino acids are very similar in clan GH-A enzymes (28), and Pc Man26A has the single ␤-bulge on strand 7 involving the nucleophile and a double ␤-bulge on strand 4 involving the histidine and acid/base (Fig. 1). The effect of the double ␤-bulge at the end of ␤4 is to cause the histidine (or asparagine in, for example, GH10 and GH5 enzymes) to stack against the acid/base. The overall location, geometry, and presentation of the active site amino acids in Man26A are therefore very similar in other clan GH-A enzymes. The detailed interactions involving the active site amino acids are similar (Fig. 3, A and B (Fig. 3, A and B).  II Data collection, phasing statistics, refinement details, and model stereochemistry The various crystallographic parameters are defined as follows: R merge (I), ¥͉I i Ϫ ͗I͉͘/¥I i , where I i is the intensity of the i-th observation, ͗I͘ is the mean intensity of the reflection, and the summation extends over all data; R iso , ¥͉F PH Ϫ F p ͉/¥͉F P ͉, the mean relative isomorphous difference between the native protein (F P ) and the derivative (F PH ) data; R Cullis , ¥ʈF PH ϩ F P ͉ Ϫ F H ͉/¥͉F PH Ϫ F P ͉, where F H is the calculated heavy atom structure factor contribution; phasing power, ͗F H ͘/͗E͘, where E is the root mean square lack of closure; R,¥͉F obs Ϫ F calc ͉/¥͉F obs ͉, where F obs and F calc represent, respectively, the observed and calculated structure factor amplitudes.   (12), is shown in Fig. 4. Loop Flexibility-It is interesting to note that the two loops that are poorly defined or missing in Pc Man26A, loops 7 and 8, are also flexible or disordered in the family 5 mannanase. However, in Pc Man26A, loop 8 is 31 residues in length compared with only five residues in the family 5 enzyme.
Functional Importance of Amino Acids in the Distal Region of the Substrate Binding Cleft-The crystal structure of Pc Man26A suggested that Trp-162 could form an important hydrophobic stacking interaction with the saccharide unit located at the Ϫ2 subsite of the enzyme. The W162A mutant displayed biochemical properties similar to native Pc Man26A when using carob galactomannan or azo-carob galactomannan as the substrates (Table III). In contrast, W162A was 95-, 70-, and 30-fold less active than the wild type mannanase against mannotriose, mannotetraose, and mannohexaose, respectively. It is clear that Trp-162 is critical for the productive binding of mannooligosaccharides but not for the productive binding of the polysaccharide mannan. The polysaccharide will be more conformationally restricted because of the extensive hydrogen bonding within the mannan. The likely role of Trp-162 is therefore to ensure that oligosaccharides bind in the correct orientation and conformation for cleavage. Similar observations have been made for Pc Xyn10A subsite Ϫ2 and ϩ1 mutants (31,32) and Ϫ6 subsite mutants of the Saccharomyces cerevisiae amylase (33) on their respective oligosaccharide and polysaccharide substrates. That binding interactions distal to the active site have a more profound influence on the cleavage of oligosaccharides than on polysaccharides is thus an emerging theme in polysaccharide hydrolases.
His-143 is highly conserved in GH26 enzymes and is located at the Ϫ2 subsite. The catalytic properties of H143A showed that the mutant enzyme exhibited very low activity against all substrates tested, both mannans and mannooligosaccharides. It would appear, therefore, that His-143 plays an important role in substrate binding of both polysaccharides and oligosaccharides at the Ϫ2 subsite, probably through hydrogen bonds between the imidazole ring and the sugar hydroxyl groups. Trp-156 in the active site of Man26A (Fig. 4) is more remote from the substrate-binding site. The catalytic properties of W156A were very similar to the wild type mannanase, revealing that this aromatic residue does not play a critical role in substrate binding or catalysis. This finding is consistent with the enzyme containing only four subsites extending from Ϫ2 to ϩ2.
Functional Importance of Trp-217-Substrate modeling of Pc Man26A suggests that Trp-217 forms a hydrophobic stacking interaction with a mannopyranose unit at the ϩ1 subsite, which is consistent with the lack of activity of W217A against both mannotriose and mannotetraose (Table III). In contrast the capacity of W217A to cleave mannans was only approximately 4-fold lower than the wild type enzyme. These data are similar to the results obtained for W162A and indicate that modification of either the Ϫ2 or ϩ1 subsite has a larger influence on oligosaccharide cleavage than polysaccharide hydrolysis.
Functional Importance of W360A-The structure of Pc Man26A indicates that Trp-360 interacts with the sugar located at the Ϫ1 subsite. The observation that W360A exhibited no catalytic activity against any of the substrates evaluated (Table III) underlines the importance of the Ϫ1 subsite in binding and catalysis. The interaction between Trp-360 and the mannose moiety is likely to be critical for transition state stabilization. The importance of Trp-360 in enzyme function is further illustrated by the fact that this residue is highly conserved in clan GH-A enzymes, and substitution of this amino acid causes a substantial reduction in the catalytic activity of the respective enzymes. Although the data presented in this paper indicate that Trp-360 plays an absolutely critical role in binding and catalysis, this interpretation must be viewed with some caution. It is possible that the loss of the aromatic residue caused a distortion in the structure of the Ϫ1 subsite, which also contributed to the loss in activity. Unfortunately, as crystals of W360A have not be obtained, the influence of this mutation on the integrity of the Ϫ1 subsite cannot be assessed.
Functional Importance of His-211-The data presented in Table III revealed that the H211A mutation caused a 100-and 700-fold decrease in activity against carob galactomannans and mannotetraose, respectively, but only a 6-fold reduction in activity against 2,4-DNPM 2. The imidazole ring of His-211, by forming a hydrogen bond with the carboxylic group of Glu-320, plays an important role in maintaining the position of the catalytic nucleophile. The loss in activity through the removal of His-211 is therefore likely to be due to a slight repositioning of the catalytic nucleophile such that it attacks the C-1 of the mannopyranose residue less efficiently than in the wild type enzyme.
The mutation of the equivalent residue to His-211 (Asn-127) in Pc Xyn10A led to a large reduction in k 3 but not k 2 against substrates with good leaving groups (31,32). This effect is caused by the loss of a hydrogen bond between Asn-127 and the C-2 hydroxyl of the sugar at the Ϫ1 subsite. The fact that the H211A mutation does not cause a large reduction in K m against 2,4-DNPM 2 suggests that the histidine does not hydrogen bond to the C-2 hydroxyl of mannose. Thus, the role of the residue immediately preceding the catalytic acid/base residue is different between GH26 enzymes and other members of the 4/7superfamily (clan GH-A).
The positions of ND1 and NE2 in His-211 are equivalent to the positions of OD1 and NE2 of the asparagine, and thus replacing His-211 with an asparagine residue might be anticipated to have little influence of the activity of Pc Man26A. However, H211N was much less active than wild type Pc Man26A (Table III). In clan GH-A enzymes the asparagine forms a hydrogen bond with the histidine, equivalent to Asp-   Table III showed that the k cat of Y285A against carob galactomannan and 2,4-DNPM 2 was substantially reduced, as was the k cat /K m for mannotetraose and the K m against 2,4-DNPM 2 . It is possible that the Y285A mutation caused a slight change in the position of the nucleophile, which did not significantly influence its capacity to attack the C-1 of the mannose residue at the Ϫ1 subsite. However, subsequent deglycosylation occurs very slowly because this step requires proton abstraction from Glu-212, and the activated water molecule will be located further from the glycosyl-enzyme ester linkage. This interpretation of the properties of Tyr-285 are consistent with studies on Pc Xyn10A in which the mutation of some amino acids that influence the position of the catalytic nucleophile had a greater influence on k 3 than k 2 for substrates that had good leaving groups (32). The low activity against mannan reflects the poorer leaving group of this substrate, which requires protonation of the glycosidic oxygen to elicit the k 2 and k 3 steps of the reaction.
The removal of the catalytic nucleophile (E320A) of Pc Man26A reduced the catalytic activity of the enzyme 10 4 -fold, which is not as great as other nucleophile mutants. A possible explanation for the relatively high activity displayed by E320A is that another amino acid is functioning as the catalytic nucleophile; Asp-283, which is approximately 3.5 Å away from Glu-320, is the only possible candidate. The observation that the double mutant E320A/D283A showed no detectable catalytic activity against all of the substrates evaluated (a decrease of Ͼ10 7 ) supports the view that Asp-283 can function as an alternative nucleophile when Glu320 has been removed.
Analysis of the Products Generated by Mutants of Pc Man26A-The products generated by the Pc Man26A mutants when incubated with mannotetraose indicate that H211A, Y265A, and W162A hydrolyze mannotetraose predominantly to mannobiose, similar to the wild type protein. In contrast, W217A generates mannose, mannobiose, and mannotriose in the ratio of 1:2.5:1 (Fig. 5). These results suggest that mannotetraose does not form productive complexes with the enzyme by occupying exclusively the Ϫ2 to ϩ2 subsites but cleaves the substrate with virtually equal efficiency when occupying either four (Ϫ2 to ϩ2) or three subsites (Ϫ2 to ϩ1; Trp-217 is located in the aglycone region of the active site). If Trp-217 only influenced substrate binding at the ϩ1 subsite, mannotetraose would still interact with the ϩ2 resulting in the generation of two mannobiose molecules. Thus it would appear that the W217A mutation has compromised substrate binding at both aglycone subsites, suggesting that the ϩ1 and ϩ2 sites act in synergy to bind substrate. Thus, loss of binding at the ϩ1 subsite compromises the utilization of binding energy at the ϩ2 subsite in the distortion of the mannopyranose ring of the sugar at the Ϫ1 subsite from a chair configuration in its ground state into its transition state conformation. Support for this latter view is provided by a recent study (34) showing that extensive synergy occurred between the aglycone subsites of Pc Xyn10A.
Conclusions-The data presented in this report show that the GH26 enzyme Pc Man26A has a classic (␣/␤) 8 barrel struc-ture in which the catalytic acid/base and nucleophile are at the ends of ␤4 and ␤7, respectively. Thus, this enzyme and, by inference, all GH26 enzymes are members of the 4/7-superfamily of glycoside hydrolases (clan GH-A). Site-directed mutagenesis studies have identified several residues at the active site that play an important role in substrate binding and catalysis.
Trp-360 appears to play an even more important role in catalysis than the two catalytic residues. This result underlines the importance of the Ϫ1 subsite for binding and hydrolysis. His-211 is shown to occur at an equivalent location but to have a different function than the corresponding asparagine in all other GH families that belong to clan GH-A. The results also show that Trp-217 dominates substrate binding in the aglycone region of the active site. Finally, the data reveal that substrate interactions distal from the active site, at subsites Ϫ2 and ϩ1, have a more profound influence on the binding and cleavage of mannooligosaccharides than on mannan, an emerging theme in cleavage by polysaccharide hydrolases.