Structural Basis for the Specificity of the Reducing End Xylose-releasing Exo-oligoxylanase from Bacillus halodurans C-125*

Reducing end xylose-releasing exo-oligoxylanase from Bacillus halodurans C-125 (Rex) hydrolyzes xylooligosaccharides whose degree of polymerization is greater than or equal to 3, releasing the xylose unit at the reducing end. It is a unique exo-type glycoside hydrolase that recognizes the xylose unit at the reducing end in a very strict manner, even discriminating the β-anomeric hydroxyl configuration from the α-anomer or 1-deoxyxylose. We have determined the crystal structures of Rex in unliganded and complex forms at 1.35–2.20-Å resolution and revealed the structural aspects of its three subsites ranging from –2 to +1. The structure of Rex was compared with those of endo-type enzymes in glycoside hydrolase subfamily 8a (GH-8a). The catalytic machinery of Rex is basically conserved with other GH-8a enzymes. However, subsite +2 is blocked by a barrier formed by a kink in the loop before helix α10. His-319 in this loop forms a direct hydrogen bond with the β-hydroxyl of xylose at subsite +1, contributing to the specific recognition of anomers at the reducing end.


EXPERIMENTAL PROCEDURES
Crystallography-Expression and purification of the wild-type and E70A mutant enzymes were previously reported (1). Both types of enzyme were crystallized under the conditions described elsewhere (12). The xylose complex (WT-xylose) was prepared by co-crystallization using a reservoir solution containing 10 mM xylose. The xylobiose complex of the E70A mutant enzyme (E70A-xylobiose) was prepared by co-crystallization using a reservoir solution containing 10 mM xylotriose, and the crystals that grew in 6 days were used for data collection. Diffraction data were collected using a charge-coupled device camera on the BL-5A station at the Photon Factory and the NW-12 station at the Photon Factory AR, High Energy Accelerator Research Organization (KEK), Tsukuba, Japan. The crystals were flash-cooled in a stream of liquid nitrogen at 100 K. Diffraction images were indexed, integrated, and scaled using the HKL2000 program suite (13). Initial phases of the native structure were obtained by the molecular replacement method, using the structure of psychrophilic endo-␤-1,4-xylanase from P. haloplanktis (Protein Data Bank code 1H14) as a search model. Molecular replacement was performed with MOLREP (14) in the CCP4 program suite (15). The program ARP/wARP (16) was used for automatic model building. Visual inspection of the models was performed using XtalView (17). Crystallographic refinement was carried out using CNS1.1 (18). The data collection and refinement statistics are summarized in Table  I. The complex structures were solved by starting from the refined native structure. The figures were prepared using SPOCK (19), Raster3D (20), MOLSCRIPT (21), and XtalView.
Site-directed Mutagenesis and Enzyme Assay-Site-directed mutagenesis for H319A was performed using the PCR overlap extension method (22). The following mutagenic oligonucleotide primer was used (the mismatched bases are underlined): 5Ј-GAG AAA TCA TTG GCC CCT GTC GGA CTG AT-3Ј. Preparation, purification, and activity measurement of the mutant enzyme were carried out as described previously (1). As to the HPLC analysis of the anomeric form of the products, the concentrations of the substrate (X3) and the enzyme (H319A mutant) were 50 mM and 19.3 M, respectively.

Crystallography-
The unliganded crystal structure of the wild-type Rex (WT-native) was solved by means of molecular replacement, using the crystal structure of pXyl. The complex structure of the wild-type Rex with xylose (WT-xylose) was determined by a co-crystallization method. In the WT-xylose structure, the electron density of a xylose molecule was found at subsite ϩ1, following the definition of the subsites of CelA is the intensity of the ith measurement of reflection h and ͗I(h)͘ is the average value over multiple measurements.
b Calculated using a test data set; 5% of total data randomly selected from the observed reflections. (Fig. 1a). Although the density peaks were relatively low and the temperature factors were relatively high (Table I), the density was clear in shape, so we could confidently determine the orientation and conformation of the xylose molecule. Further attempts to obtain other types of complex structures (i.e. co-crystallization of the wild-type Rex in 10 mM xylobiose or a mixture of 5 mM xylose and 5 mM xylobiose) resulted in a similar density to that in the case of WT-xylose (a xylose molecule was bound at subsite ϩ1; data not shown). When the wild-type Rex crystals were soaked in 10 mM xylotriose for a short time (10 s), density peaks for sugar units ranging from subsite Ϫ2 to ϩ1 were observed. However, the density at subsite Ϫ1 was ambiguous and appeared to reflect a mixture of several conformations, probably involving a fraction of the cleaved substrate. The second type of complex structure was obtained by co-crystallizing the inactive E70A mutant with xylotriose (E70A-xylobiose). In the substrate binding cleft of E70A-xylobiose, the clear density of a xylobiose unit at subsites Ϫ2 and Ϫ1, as well as a glycerol at subsite ϩ1, was observed ( Fig. 1b). There are two possibilities that explain the discrepancy between the co-crystallized reagent (xylotriose) and the density observed for xylobiose: (a) xylotriose had been cleaved during the crystal formation (6 days), although the hydrolytic activity of the E70A mutant is 10 Ϫ4 orders lower that of the wild-type; and (b) xylotriose is bound at subsites Ϫ3 to Ϫ1, but a xylose moiety at the nonreducing end could not be detected due to high mobility or disorder. All three structures contain a metal ion and a glycerol, both of which bind at a crystal packing interface far from the active site (Fig. 2b). The metal ion was assigned as a nickel ion (Ni 2ϩ ) because a Ni 2ϩ -nitrilotriacetic acid-agarose column was used for purification, and no other candidate was included in the reservoir solution. Refined temperature factors of the nickel ion in the three structures were within 8.9 -17.6 Å 2 (Table I), and the refined F o Ϫ F c maps were almost flat (data not shown). The nickel ion was tetrahedrally coordinated by Gln 27 , Gln 30 , Asp 253Ј , and His 259Ј (symmetry-related residues are indicated by primes). The glycerol molecule was held by the main-chain atoms of Gln 192 , Tyr 247 , and Asp 248 and the side chain of Trp 123Ј . Glycerol was absolutely required for crystallization (12). A polypeptide chain extending from Glu 6 and Pro 381 was modeled in WT-xylose and E70A-xylobiose, whereas Glu 45 and Thr 46 were not included in WT-native because of local disorder.
Overall Structure-The structure of Rex comprises a disordered (␣/␣) 6 barrel similar to that of pXyl, whereas CelA has a less disordered (␣/␣) 6 barrel with a circular cross-section (Fig.  2). The root mean square deviations as to pXyl and CelA are 1.8 Å for 357 residues and 2.3 Å for 335 residues, respectively. There are four free cysteine residues located inside the molecule. This characteristic is usual for intracellular enzymes. On the other hand, two cysteine residues of pXyl, which is an extracellular enzyme, form a disulfide bond (9). The molecular surfaces of the three GH-8a enzymes are shown in Fig. 3. The substrate-binding cleft of CelA is clearly larger than those of the other two enzymes. Interestingly, subsite ϩ2 of Rex is blocked by a barrier at the upper side of the cleft (described below). In contrast, the other two endo-GHs have a long cleft spanning the molecule, which can accommodate a long polysaccharide chain.
Complex Structures-WT-native and the two complex structures (WT-xylose and E70A-xylobiose) of Rex were almost identical (root mean square deviation for C␣ atoms Ͻ0.26 Å and for all atoms Ͻ0.44 Å between all pairs of the three Rex structures). There was no large conformational change on substrate binding like on other GH-8a enzymes. For example, substrate binding on CelA induces only small structural changes in the protein, mostly slight reorientations of aromatic and polar side chains in contact with the substrate (11). However, substrate binding on Rex induced slight movements in two regions (Thr 62 -Asn 64 and Gly 355 -Arg 357 ) around subsite ϩ1 (Fig. 4a). The main chain moves toward the substrate in both of the complex structures. In the WT-xylose structure, the side chain of Arg 357 swings to form a specific interaction with the xylose at subsite ϩ1, accompanied by movement of the adjacent Asn 356 residue. The side chain of Arg 357 was ordered in WT-xylose, whereas it was disordered in the other two structures.
The xylose molecule in WT-xylose takes on the 4 C 1 conformation, and all of the oxygen atoms (O-1 to O-5) form direct or water-mediated hydrogen bonds with protein atoms (Fig. 5a). His 319 forms one of the two direct hydrogen bonds present between xylose and the Rex protein, which recognizes the ␤-hydroxyl group at the O-1 position. A stacking interaction of Tyr 360 with the ␤-face of the xylose also occurs. This residue can sterically interfere binding of an ␣-anomer. When an ␣-glucose molecule was superimposed onto the ␤-xylose, the ␣-anomeric hydroxyl oxygen atom was located at distances of 2.7 and 2.5 Å from the O␤ and O␥ atoms of Tyr 360 , respectively (see Supplemental Material). Actually, electron density for only the ␤-anomer was observed, probably because of these interactions. In the E70A-xylobiose complex, the xylobiose molecule takes on the 4 C 1 conformation at both of the two sugar rings and the ␤-anomeric configuration at the reducing end. The glycerol molecule is located at almost the same position as subsite ϩ1 and forms a water-mediated hydrogen bond with the xylobiose (Fig. 5b). Although the positions of the three hydroxyl oxygen atoms of the glycerol are slightly different from those of xylose at subsite ϩ1, they interact with several residues at subsite ϩ1 and cause movement of the main chain in the two regions described above. All of the oxygen atoms form direct or watermediated hydrogen bonds with the protein, and the xylose ring at subsite Ϫ2 of the xylobiose is stacked with Trp 112 at the ␤-face.
Subsites- Fig. 4b shows composite superimpositioning of the subsites of the three GH-8a enzymes: the WT-xylose structure of Rex combined with xylobiose in the E70A-xylobiose structure (gray), the pXyl D144N mutant structure (Protein Data Bank code 1H14) combined with a xylose molecule at subsite ϩ4 in the complex structure (Protein Data Bank code 1H12) (green), and the CelA E95Q mutant structure complexed with cellopentaose (Protein Data Bank code 1KWF) (yellow). Both the cellopentaose molecule (subsites Ϫ3 to ϩ2; ball-and-stick model) and the partially bound product molecule (cellotriose at subsites ϩ1 to ϩ3) are shown. Xylose and xylobiose in Rex approximately overlap with the glucose units at subsites Ϫ2 to ϩ1 in CelA.
At subsite Ϫ3 of CelA, Trp 205 forms a stacking interaction, and the side chain of Arg 204 forms a hydrogen bond with the O-3 atom. However, there is no subsite here in pXyl and Rex. In Rex, the Trp is substituted by Ile 188 , and there is no corresponding residue with the Arg, losing specific interactions to a sugar moiety at this site. Rex exhibits significantly higher K m for xylooligasaccharides of longer than xylotetraose (1). There is, however, no steric hindrance at subsite Ϫ3, and the subsite is highly accessible to solvent. We could not find any convincing structural factor that can make the binding effect on this subsite negative. Perhaps the side chain of Glu 190 , which is solvent-exposed and somewhat disordered in the crystal structure, may interfere with the sugar binding in an extended conformation (Fig. 4b). At subsite Ϫ2, the bound xylose/glucose groups of Rex and CelA almost completely overlap. The stacking Trp residue (Trp 112 /Trp 124 /Trp 132 in Rex, pXyl, and CelA) is conserved in GH-8a enzymes. The O-2, O-3, and O-5 atoms at subsite Ϫ2 form water-mediated hydrogen bonds with Tyr 244 and Tyr 198 (Fig. 5b).
A significant kink is observed between sugar rings at subsites Ϫ1 and ϩ1, the two sugar rings being twisted to become almost perpendicular to each other. Catalytically important residues are rather concentrated around this region. Glu 70 (Glu 78 /Glu 94 in pXyl and CelA) has been identified as the proton donor (1). The catalytic base residue (Asp 263 /Asp 281 /Asp 278 in Rex, pXyl, and CelA) holds a water molecule through hydrogen bonds together with a conserved Tyr residue (Tyr 198 / Tyr 203 /Tyr 215 ). The water is thought to correspond to the nucleophilic water, which is activated by the catalytic base residue (11). In the CelA-cellopentaose complex structure, the third catalytically important residue (Asp 128 /Asp 144 /Asp 152 ) forms bifurcated hydrogen bonds with the O-2 and O-3 atoms of the glucose unit at subsite Ϫ1 with the 2,5 B conformation (11). The interaction is critical to stabilize the sugar ring in a strained boat conformation, which is thought to be a prerequisite for the inverting hydrolytic mechanism through a transition state with the oxocarbenium ion, the planarity of the atoms C-5, O-5, C-1, and C-2 facilitating the formation of a partial double bond between O-5 and C-1 (11). In the E70Axylobiose structure, Asp 128 does not form any interaction with the ligand. However, the side chain of Asp 128 is positioned similarly with that of Asp 152 of CelA, and it would also stabilize the xylose ring at subsite Ϫ1 if the ring takes on a boat conformation. In summary, the catalytic mechanism of Rex seems to be basically conserved with other GH-8a enzymes.
At subsite ϩ1, the xylose molecule in Rex overlaps better with a glucose unit of the product cellotriose molecule in CelA, rather than that of the uncleaved cellopentaose molecule. For CelA, the product molecule at subsites ϩ1 to ϩ3 is thought to represent a possible first step during which the leaving group rotates slightly and shifts away from the reaction center, the stacking interactions being preserved (11). The xylose molecule in WT-xylose of Rex also seems to correspond to a product molecule being released. Subsite ϩ1 of Rex is unique compared with that of other GH-8a enzymes, because the residues at this subsite (Asp 61 , Asn 64 , Arg 68 , Ser 262 , His 319 , and Arg 357 ) are not conserved in CelA and pXyl. However, the stacking Tyr residue (Tyr 360 /Tyr 381 / Tyr 372 in Rex, pXyl, and CelA) is conserved and fixes the sugar unit at an approximate position.
The most notable difference between Rex and the other two GH-8a enzymes is the blockage of subsite ϩ2 by the kink in the loop before ␣ 10 at Ser 317 -Pro 320 (Fig. 4b). His 319 is directly hydrogen-bonded with the ␤-hydroxyl of the xylose at subsite ϩ1, contributing to the discrimination of the anomers at the reducing end. Leu 318 blocks subsite ϩ2 with its long side chain together with His 319 . A proline residue (Pro 320 ) is present only in Rex, and the main chain trace bends at an almost right angle at this position. The bent loop structure seems to be intrinsically stable, because there is no conformational difference between the unliganded and liganded structures. On the other hand, in pXyl and CelA, this loop is located away from the substrate. Instead, a Tyr residue in the loop before ␣ 12 (Tyr 378 / Tyr 369 in pXyl and CelA) forms subsite ϩ3.
Specificity for Xylosides-The order of preference of Rex for xylo-/glucooligosaccharides is XXX Ͼ Ͼ GXX Ͼ XXG Ͼ Ͼ GXG (where G represents glucose and X is xylose), and no detectable activity is observed for XGG, GGX, and XGX (1). This indicates that subsite Ϫ1 strictly requires a xyloside, and the other two subsites also exhibit a certain level of selectivity. At the putative C-6 position of subsite Ϫ1 in Rex, there is a long side chain of the Arg 68 residue that interferes with binding of a glucoside. Arg at FIG. 4. The subsites of Rex and GH-8a enzymes. a, stereoview of superimposition of the three structures of Rex around the substrate-binding site. The WT-native, WT-xylose, and E70A-xylobiose structures are shown in yellow, gray, and green, respectively. b, stereoview of superimpositioning of Rex (gray), pXyl (green), and CelA (yellow). The structures were superimposed using the atoms of the two catalytic residues and the third catalytically important residue (Asp 128 / Asp 144 /Asp 152 in Rex/pXyl/CelA). The xylobiose (Ϫ2 to Ϫ1) and xylose (ϩ1) in Rex, the xylose in pXyl (ϩ4), and the cellopentaose in CelA (subsites Ϫ3 to ϩ2) are shown as a ball-and-stick model. The partially bound product molecule in CelA (cellotriose at subsites ϩ1 to ϩ3) is shown as thin sticks. Residue names and numbers are labeled in the order of Rex/pXyl/ CelA with each color. this position is conserved in pXyl (Arg 76 ), whereas CelA has the Val 93 residue. Instead, a guanidium group of Arg 84 comes from another position in CelA. Therefore, the selectivity for xylose/ glucose at subsite Ϫ1 seems to be caused by these Arg residues. On the other hand, the selectivity at subsites Ϫ2 and ϩ1 is not structurally obvious. At subsite Ϫ2, the main chain atom of Pro 125 -Ala 126 and the side chain of Ile 188 appear to sterically clash against an O-6 group. At subsite ϩ1, the catalytic base residue (Asp 263 ) appears to cause a steric hindrance. Actually, the catalytic base of CelA (Asp 278 ) is rotated at its carboxyl end to avoid a steric clash. The residues in the distal area from the catalytic base are not conserved in CelA and the other two GH-8a enzymes. Molecular surface representation also shows that this area of CelA comprises a relatively wide open cleft (Fig. 3).
Kinetic Analysis of H319A Mutant-In order to investigate the importance of the His 319 residue in the discrimination of the anomers at the reducing end, we examined the enzymatic activity of the H319A mutant on X3 and X3-de substrates as well as the kinetic parameters of X3 hydrolysis (Table II). The H319A mutant exhibited significant decrease in the specificity to the anomeric hydroxyl group (activity ratio for X3/X3-de ϭ 1.6), compared with the wild type (activity ratio for X3/X3-de ϭ 31). The mutation caused a 14.3-fold increase in the K m value and a 7.8-fold decrease in the k cat value for X3 hydrolysis. As shown in Fig. 6, however, the H319A mutant produced ␤-anomer of X1 and ␣-anomer of X2 from X3 in the reaction for 1 min. Furthermore the ␣-anomer of X3 was the predominant anomer remaining in the reaction. These results indicate that the H319A mutant still discriminates the ␤-anomer from the ␣-anomer at the reducing end like the wild-type enzyme does, whereas the catalytic efficiency for the ␤-anomer is drastically decreased. The effect of the H319A mutation is consistent with the structural feature of Rex. Steric crash at the side chain of Tyr 360 remains in the H319A mutant, interfering with the binding of ␣-anomer. However, loss of the direct hydrogen bond of His 319 leads to a drastic decrease in the activity toward the X3 substrate. The remaining two water-mediated hydrogen bonds held by Asn 64 , Arg 357 , and Asp 61 (Fig. 5a) may be responsible for the slight preference of the H319A mutant for X3 against X3-de. DISCUSSION To date, the crystal structures of two types of reducing endspecific CBHs (Cel7A and Cel48A) have been determined, and similar mechanisms for processing of the substrate are proposed (7,23).   the cleaved disaccharide (cellobiose), the polysaccharide chain slides for the next processing position. The tunnels of Cel7A from Hypocrea jecorina (formerly known as CBH I from Trichoderma reesei) is formed by extra loops covering the cleft of the basal ␤-jelly roll fold, in comparison with other endo-type GH-7 cellulases having an open cleft (23). The tunnels of GH-48 CBHs from Clostridia (Cel48A; known as CelF and CelS) are formed by an extra ␤-domain on a (␣/␣) 6 barrel core, which is structurally similar to those of GH-8 enzymes (7,24).
It is not clear to what extent GH-74 oligoxyloglucan reducing end-specific cellobiohydrolase is specific for the reducing end sugars of substrates, because no complex structures are yet available. However, an insertion/deletion of a loop comprising about 10 amino acid residues has been implied to be the determinant for the exo-and endo-types of GH-74 enzymes (25).
Two GH-13 enzymes (maltodextrin glucosidase from E. coli K12 and an amylolytic enzyme from T. maritima) are also known to be reducing end-specific monomer-forming exo-glycosidases (2, 3). However, they can also release p-nitrophenol from p-nitrophenyl maltooligosaccharides. Therefore, these enzymes are not specific for the group released from the reducing end. Although the crystal structures of these enzymes remain unknown, they probably possess a cleft similar to that of Rex, which is blocked at subsite ϩ2.
The structural basis for the reducing end specificity of Rex has been revealed to be rather simple; subsite ϩ2 is blocked by a kink in the loop before ␣ 10 , and His 319 forms a specific hydrogen bond with the ␤-hydroxyl of a reducing end xyloside at subsite ϩ1. Moreover, Asp 61 forms another direct hydrogen bond at this subsite. These two direct hydrogen bonds make the subsite stronger than the other two, the latter having few direct hydrogen bonds unless a xylose ring at subsite Ϫ1 takes on an unfavorable boat conformation. Accordingly, all of our attempts at co-crystallization with the wild-type enzyme with various xylooligosaccharides resulted in similar structures to that of the WT-xylose; a xylose was bound at subsite ϩ1. Conversion of endo-type GH-8 enzymes to exo-enzymes like Rex could be accomplished by means of a limited number of mutations around subsite ϩ1. The alkalophilic bacterium B. halodurans may have adopted such a strategy for evolving an intracellular xylooligosaccharide-degrading enzyme (Rex) from an endo-type GH-8 enzyme.
As clearly seen in Fig. 4a, the O-2 and O-3 atoms of the xylose at subsite Ϫ2 point into solvent, whereas those atoms at subsites Ϫ1 and ϩ1 are blocked, suggesting that Rex can hydrolyze xylooligosaccharides decorated at xylose n Ϫ 2 (when n is the released xylose at the reducing end). Recently, it was shown that the major decorated xylooligosaccharides products generated by GH-10 xylanases contain arabinose or glucuronic side chains at the nonreducing end (26 -28). The xylanolytic mechanism of Cellvibrio japonicus, a mesophilic bacterium, has been proposed according to the localization and the specificities of the related enzymes (28). In the mechanism, xylan is initially solubilized by the three secreted endoxylanases (Xyn10A, Xyn11A, and Xyn11B), followed by processing with the cellbound endoxylanase (Xyn10C), ␣-arabinofuranosidase (Abf51), and ␣-glucuruonidase (GlcA67) into short undecorated xylooligosaccharides to incorporate into periplasm. The xylooligosac-charides are degraded in periplasm by the specific endoxylanase (Xyn10D) and transported into cytoplasm as xylobiose. Finally, the xylobiose was converted into xylose in the cytoplasm by intracellular ␤-xylosidases. The xylanolytic system of B. halodurans, an alkalophilic bacterium, is supposed to be different from that of C. japonicus. In the system, xylan is degraded into xylooligosaccharides having an ␣-arabinofuranosyl or ␣-glucuronyl residue at their nonreducing ends by the two extracellular alkalophilic endoxylanases (29,30), and the resultant oligosaccharides are directly transported into the cytoplasm (1). The reducing end specificity of Rex might be useful to hydrolyze such oligosaccharides by the concurrent action with the intracellular ␣-arabinofuranosidases and ␣-glucuronidase (1), to produce smaller undecorated xylooligosaccharides, which are suitable substrates for ␤-xylosidases.