Journal of Structural Biology 190 (2015) 21–30

Contents lists available at ScienceDirect

Journal of Structural Biology journal homepage: www.elsevier.com/locate/yjsbi

Crystal structure and substrate-binding mode of GH63 mannosylglycerate hydrolase from Thermus thermophilus HB8 Takatsugu Miyazaki a, Megumi Ichikawa a, Hitoshi Iino b, Atsushi Nishikawa a, Takashi Tonozuka a,⇑ a b

Department of Applied Biological Science, Tokyo University of Agriculture and Technology, 3-5-8 Saiwai-cho, Fuchu, Tokyo 183-8509, Japan Department of Biological Sciences, Graduate School of Science, Osaka University, 1-1 Machikaneyama-cho, Toyonaka, Osaka 560-0043, Japan

a r t i c l e

i n f o

Article history: Received 5 December 2014 Received in revised form 16 February 2015 Accepted 17 February 2015 Available online 21 February 2015 Keywords: Glycoside hydrolase family 63 Crystal structure (a/a)6-Barrel Trimer Mannosylglycerate Glucosylglycerate

a b s t r a c t Glycoside hydrolase family 63 (GH63) proteins are found in eukaryotes such as processing a-glucosidase I and also many bacteria and archaea. Recent studies have identified two bacterial and one plant GH63 mannosylglycerate hydrolases that act on both glucosylglycerate and mannosylglycerate, which are compatible solutes found in many thermophilic prokaryotes and some plants. Here we report the 1.67-Å crystal structure of one of these GH63 mannosylglycerate hydrolases, Tt8MGH from Thermus thermophilus HB8, which is 99% homologous to mannosylglycerate hydrolase from T. thermophilus HB27. Tt8MGH consists of a single (a/a)6-barrel catalytic domain with two additional helices and two long loops which form a homotrimer. The structures of this protein in complexes with glucose or glycerate were also determined at 1.77- or 2.10-Å resolution, respectively. A comparison of these structures revealed that the conformations of three flexible loops were largely different from each other. The conformational changes may be induced by ligand binding and serve to form finger-like structures for holding substrates. These findings represent the first-ever proposed substrate recognition mechanism for GH63 mannosylglycerate hydrolase. Ó 2015 Elsevier Inc. All rights reserved.

1. Introduction Extremophiles produce various osmolytes to adapt themselves to high-salt environments. These osmolytes are called ‘‘compatible solutes’’ and include sugars, amino acids, and polyols (Kempf and Bremer, 1998; Empadinhas and da Costa, 2011). 2-O-a-DMannopyranosyl-D-glycerate (mannosylglycerate, MG) is a compatible solute initially found in red algae (Colin and Augier, 1939) that also accumulates in thermophilic bacteria and archaea such as Thermus, Rubrobacter, Rhodothermus, and Pyrococcus (Empadinhas and da Costa, 2011). 2-O-a-D-Glucopyranosyl-D-glycerate (glucosylglycerate, GG) is a similar compound produced in many prokaryotes, such as Persephonella marina and actinobacteria, and is a precursor of lipopolysaccharide synthesis in mycobacteria (Empadinhas and da Costa, 2011). Many studies of physiological functions and synthesis of MG and GG have been

Abbreviations: GG, 2-O-a-D-glucopyranosyl-D-glycerate; GGalase, 2-O-a-D-glucopyranosyl-D-galactose hydrolase; GH63, glycoside hydrolase family 63; MG, 2-Oa-D-mannopyranosyl-D-glycerate; MGH, mannosylglycerate hydrolase. ⇑ Corresponding author. Fax: +81 42 367 5705. E-mail address: [email protected] (T. Tonozuka). http://dx.doi.org/10.1016/j.jsb.2015.02.006 1047-8477/Ó 2015 Elsevier Inc. All rights reserved.

published (Borges et al., 2014), but reports on the catabolism of these solutes are less common. Recently, glycoside hydrolases (mannosylglycerate hydrolases, MGHs) active on both MG and GG have been identified in two bacteria [Thermus thermophilus HB27 (Tt27MGH) and Rubrobacter radiotolerans RSPS-4 (RrMGH)] and the plant Selaginella moellendorffii (SmMGH) (Alarico et al., 2013; Nobre et al., 2013). They are classified in glycoside hydrolase family 63 (GH63) in the carbohydrate-active enzymes (CAZy) database (http://www.cazy.org) (Lombard et al., 2014). GH63 proteins are found in archaea, bacteria, and eukaryotes, and the most studied enzyme among them is processing a-glucosidase I (EC 3.2.1.106) (Kalz-Füller et al., 1995; Palcic et al., 1999; Dhanawansa et al., 2002; Faridmoayer and Scaman, 2004, 2005, 2007; Frade-Pérez et al., 2010; Miyazaki et al., 2011). This enzyme specifically hydrolyzes the a-1,2-glucoside linkage of Glc3Man9GlcNAc2, an oligosaccharide precursor of eukaryotic N-linked glycoproteins (Herscovics, 1999; Helenius and Aebi, 2004). The apo structure of processing a-glucosidase I from Saccharomyces cerevisiae (ScCwh41p) has been reported (Barker and Rose, 2013). We previously reported the crystal structure of a glycoside hydrolase YgjK (EcYgjK) from Escherichia coli K12 as the first structure among GH63 members (Kurakata et al.,

22

T. Miyazaki et al. / Journal of Structural Biology 190 (2015) 21–30

2008), and subsequent substrate screening using the combination of glycosynthase reaction and structural analysis revealed that the enzyme was likely a hydrolase (GGalase) specific for 2-O-a-Dglucopyranosyl-D-galactose (Miyazaki et al., 2013). Additionally, the 2.1 Å-resolution apo structure of an uncharacterized protein (TTHA0978; Tt8MGH) from T. thermophilus HB8 was determined by the structural genomics project (Protein Data Bank [PDB]: 2Z07). Although there is no report on the biochemical and structural characterization of Tt8MGH, due to its 99% sequence identity with Tt27MGH, the protein appears to be a glycoside hydrolase active on MG and GG. Although the structure-determined GH63 enzymes share (a/a)6-barrel catalytic domains, their substrate specificities differ, especially in aglycons. To elucidate the relationship between the structure and substrate specificity of MGHs and differences among other GH63 proteins, here we determined the crystal structures of Tt8MGH in apo form and in complexes with glucose (Glc) and glycerate. The present study represents the first proposed substrate recognition mechanism for Tt8MGH with large conformational changes of loops. 2. Materials and methods 2.1. Protein expression and purification

induced with isopropyl-b-D-thiogalactopyranoside at a final concentration of 0.5 mM and cultured for another 6 h at 30 °C. The cells were harvested by centrifugation at 10,000g for 5 min, resuspended in 30 mL 20 mM Tris–HCl buffer (pH 7.5) and then disrupted by sonication. The cell lysate was centrifuged at 10,000g for 20 min to remove insoluble debris, and the soluble fraction was heated at 70 °C for 10 min, followed by centrifugation at 20,000g for 30 min to eliminate the aggregated proteins derived from E. coli. The supernate was applied to a HiPrep Butyl FF 16/10 column (GE Healthcare, Little Chalfont, Buckinghamshire, UK) equilibrated with 20 mM Tris–HCl buffer (pH 7.5) containing 0.4 M ammonium sulfate. The protein was eluted using a decreasing linear gradient of 0.4–0 M ammonium sulfate at a flow rate of 3 mL min1. The fractions containing the protein were collected, dialyzed against 20 mM Tris–HCl (pH 7.5) buffer, and then applied to a HiLoad 16/10 Q-Sepharose HP column (GE Healthcare) equilibrated with the same buffer. The protein was eluted with a linear gradient of 0–0.5 M sodium chloride in the same buffer at a flow rate of 3 mL min1. The purity of the protein was confirmed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS– PAGE) (Supplementary Fig. S1). 2.2. Crystallization, data collection, structure determination, and refinement

The expression plasmid for Tt8MGH (TTHA0978, GenBank: BAD70801.1) was purchased from RIKEN Bioresource Center (Tsukuba, Japan) and was derived from pET-11a vector (Novagen, Madison, WI, USA) subcloned with a gene for Tt8MGH. E. coli BL21 (DE3) was used for overexpression of the gene. Cells harboring the plasmids were cultured at 37 °C in 1 L Luria–Bertani medium containing 50 lg mL1 ampicillin. When the culture reached an optical density of 0.6 measured at 600 nm, it was

Prior to crystallization, the purified protein was concentrated to 10 mg mL1 in 10 mM Tris–HCl (pH 7.5) using the ultrafiltration device Amicon Ultra 10 K (Millipore, Bedford, MA, USA). Tt8MGH was crystallized at 20 °C using the hanging-drop vapor diffusion method in which 1.0 lL protein solution was mixed with an equal volume of crystallization mother liquor containing 20–35% (vol/ vol) 2-methyl-2,4-pentanediol and 100 mM Tris–HCl buffer (pH

Table 1 Data collection and refinement statistics.

Data collection Beamline Wavelength (Å) Space group Cell dimensions a = b (Å) c (Å) Resolution range (Å) Measured reflections Unique reflections Completeness (%) I/r(I) Rmerge Refinement statistics Rwork Rfree RMSD Bond lengths (Å) Bond angles (°) Number of atoms Protein Ligands Chloride ion Water Average B (Å2) Protein Ligands Chloride ion Water Ramachandran plot Favored (%) Outliers (%) a b

Tt8MGH-tris

Tt8MGH-Glc

Tt8MGH-glycerate

PF AR-NW12A 1.0000 R3

PF AR-NE3A 1.0000 R3

PF AR-NE3A 1.0000 P321

92.0a 267.2a 50–1.67 (1.73–1.67)b 524,392 96,185 99.2 (98.2)b 37.0 (6.5)b 0.059 (0.318)b

93.4a 254.1a 50–1.77 (1.83–1.77)b 447,604 80,188 99.3 (100)b 51.3 (10.8)b 0.045 (0.208)b

93.0 190.4 50–2.10 (2.21–2.10)b 250,934 55,067 97.6 (93.6)b 9.3 (5.5)b 0.110 (0.190)b

0.146 0.173

0.159 0.188

0.245 0.280

0.008 1.314

0.010 1.422

0.010 1.333

6871 72 2 674

6714 44 2 548

6796 30 2 265

17.5 25.4 13.3 29.8

23.7 31.2 17.1 31.2

30.4 31.7 21.7 26.5

97.0 0

97.4 0

96.4 0.3

The cell parameters are in hexagonal setting. The values for the highest resolution shells are provided in parentheses.

23

T. Miyazaki et al. / Journal of Structural Biology 190 (2015) 21–30

5.5–6.5). For co-crystallization with Glc or mannose (Man), the protein solution was mixed with each ligand at a final concentration of 10 mM and stored at 4 °C overnight. The Tt8MGH-Glc complex crystals were grown using the hanging-drop vapor diffusion method described above except that 100 mM sodium citrate buffer (pH 4.8–6.0) was used instead of Tris–HCl buffer in a reservoir solution. Screening conditions of co-crystallization with glycerate (Tokyo Chemical Industry, Tokyo, Japan) was performed with a Crystal Screen kit (Hampton Research, Aliso Viejo, CA, USA) supplemented with glycerate, and the crystals were obtained with the solutions No. 1, 3, 6, 9, 11, 17, 23, 24, 25, 29, 40, 41, and 44. The well-diffracted crystals were obtained with a reservoir solution of 0.4–0.7 M potassium sodium tartrate, 0.1 M HEPES–NaOH (pH 7.0–8.0), 50 mM calcium chloride, and 10 mM glycerate. The crystals were cryoprotected with the mother liquor (for ligand-free and Glc-complex crystals) or a mixture of Paratone-N and paraffin oil (1:1, Hampton Research) (for glycerate-complex crystals), and then flash-cooled in a nitrogen stream at 100 K. Diffraction data were collected at PF AR-NW12A or PF AR-NE3A beamlines (Photon Factory, Tsukuba, Japan). All data were processed and scaled using HKL2000 software (Otwinowski and Minor, 1997). The structures were solved by the molecular replacement method using MOLREP in the CCP4 program suite (Collaborative Computational Project Number 4, 1994; Vagin and Teplyakov, 1997) with the coordinate of PDB 2Z07 as a search model. Refinement was performed using REFMAC5 (Murshudov et al., 1997) in the CCP4 program suite, and manual adjustment and rebuilding of the model were performed using COOT (Emsley et al., 2010). Solvent molecules were introduced using ARP/wARP (Perrakis et al., 1999). Validation of the structures was performed using MolProbity (Chen et al., 2010). Data collection and refinement statistics are listed in Table 1. The solvent-accessible surface area was calculated using AREAIMOL in the CCP4 program suite. Analysis of protein–ligand interactions was performed using LIGPLOT (Wallace et al., 1995) and COOT. Protein assembly was evaluated by the Protein Interfaces, Surfaces, and Assemblies (PISA) server (http://www.ebi.ac.uk/pdbe/pisa/) (Krissinel and Henrick, 2007). Figures were prepared using PyMOL (http://www.pymol.org/). The coordinates and structure factors for the ligand-free form (Tt8MGH-Tris), Glc-complex (Tt8MGH-Glc), and glycerate-complex (Tt8MGH-glycerate) have been deposited in the PDB under accession codes 4WVA, 4WVB, and 4WVC, respectively. 2.3. Molecular weight determination The molecular weight of the purified protein was determined by gel filtration chromatography with a Superdex 200 HR 10/30 column (GE Healthcare) at a flow rate of 0.5 mL min1 using the ÄKTApurifier chromatography system (GE Healthcare). The column was equilibrated with 20 mM sodium phosphate containing 0.2 M sodium chloride. Calibration was performed using blue dextran 2000 (2000 kDa), thyroglobulin (669 kDa), ferritin (440 kDa), catalase (232 kDa), lactate dehydrogenase (140 kDa), and albumin (66 kDa). The calibration curve of Kav vs. log molecular weight was prepared using the equation Kav = Ve  Vo/Vt  Vo, where Ve = elution volume for the protein, Vo = column void volume, and Vt = total bed volume.

the crystallization buffer was found at the active site (hence the name Tt8MGH-Tris). Additionally, the structures of Tt8MGH-Glc and Tt8MGH-glycerate were determined at 1.77- and 2.10-Å resolutions, respectively. The crystals of Tt8MGH-Tris and Tt8MGH-Glc belong to the space group R3 and those of Tt8MGH-glycerate belong to the space group P321 (Table 1). Each contains two molecules, named Mol-A and Mol-B, in each asymmetric unit. The whole structure of Tt8MGH and the topology of the secondary structure elements are described in Fig. 1 and Supplementary Fig. S2, respectively. Tt8MGH consists of an (a/a)6-barrel catalytic domain and an inserted subdomain A0 -region (residues 146–234). The protein contains 18 a-helices (H1–H12 and A0 H1–A0 H6) and 10 b-strands (S1–S9 and A0 S1); six of these helices (A0 H1–A0 H6) and one b-strand (A0 S1) are included in the A0 -region. There are three movable loops: between H1 and S1 (residues 23–30, named loop-A); between S2 and S3 (residues 75–96, loop-B); and between A0 H2 and A0 H3 (residues 177–199, loop-C). No electron density for part of loop-B (residues 90–94 of Mol-A and 91–94 of Mol-B) was observed in Tt8MGH-Tris, probably due to its flexibility, and the conformations of loop-B in Mol-A and Mol-B were different from

A

loop-A

loop-A loop-B

loop-C

loop-B

loop-C

B

C

D

3. Results and discussion 3.1. Overall structure of Tt8MGH The crystal structure of Tt8MGH-Tris was determined at 1.67-Å resolution, which is higher than the resolution previously used to determine its structure (PDB: 2Z07, 2.10 Å). A Tris molecule from

Fig.1. Overall structures of Tt8MGH monomers Tt8MGH-Tris Mol-A (A), Tt8MGHTris Mol-B (B), Tt8MGH-Glc Mol-A (C), and Tt8MGH-glycerate Mol-A (D) are shown as ribbon models in stereo. Loop-A, loop-B, and A0 -region (including loop-C) are colored in blue, orange, and magenta, respectively, and the other regions in cyan. Disordered regions are indicated with dotted lines. Ligands are shown as stick models and colored in yellow (Tris), green (Glc), and pink (glycerate).

24

T. Miyazaki et al. / Journal of Structural Biology 190 (2015) 21–30

each other (Fig. 1A and B). Also, parts of loop-B and loop-C in Tt8MGH-Glc (residues 86 and 87) and Tt8MGH-glycerate (residues 91–95 of Mol-A, 87–95 of Mol-B) were disordered, and conformations of loop-A, loop-B, and loop-C were different from each other (Fig. 1C, D and Supplementary Fig. S3) (discussed in Section 3.4). The clear electron density maps for Tris molecules were observed at the active sites in both Mol-A and Mol-B of Tt8MGH-Tris (Supplementary Fig. S4). A glucose molecule was found at the active site in Mol-A of Tt8MGH-Glc, and Tris and glycerate molecules were found in both Mol-A and Mol-B of Tt8MGH-glycerate (Fig. 1C, D and Supplementary Fig. S3). A structural homology search was performed using the DaliLite v.3 server (Holm et al., 2008), and several glycoside hydrolases whose catalytic domains adopt an (a/a)6-barrel fold were identified. As expected, the highest Z scores were observed with the following clan GH-G enzymes: the aforementioned EcYgjK from E. coli (PDB: 3W7S, Z score = 38.3) (Kurakata et al., 2008), ScCwh41p from S. cerevisiae (PDB: 4J5T, Z score = 32.2) (Barker and Rose, 2013), and a GH37 trehalase from E. coli (PDB: 2JJB, Z score = 34.8) (Gibson et al., 2007). GH15 glucoamylase (PDB: 1LF6, Z score = 25.7) (Aleshin et al., 2003), GH94 cellobiose phosphorylase (PDB: 3RSY, Z score = 29.9), and GH94 chitobiose phosphorylase (PDB: 1V7X, Z score = 29.6) (Hidaka et al., 2004) followed. These results are similar to the results for EcYgjK as a template in our previous study (Kurakata et al., 2008). The coordinate of only A0 -region in Tt8MGH was also used as a template for a homology search, but no homologous structure was found. Also, the corresponding regions of EcYgjK and ScCwh41p adopt different folds (Supplementary Fig. S5).

A

3.2. Quaternary structure and inter-subunit interactions The molecular weight values of Tt8MGH are 48.8 kDa, as deduced from the amino acid sequence, and 125.4 kDa, as calibrated by gel filtration chromatography (Fig. 2A and Supplementary Fig. S1), suggesting that the protein forms a dimer or a trimer in solution. The molecular mass of Tt27MGH (which has 99% sequence identity with Tt8MGH) in solution was reported to be 205.8 kDa (Alarico et al., 2013), which is different from the result for Tt8MGH. Mol-A and Mol-B of Tt8MGH-Tris do not directly interact with each other, and appear to form distinct homotrimers consisting of three monomers related by the crystallographic 3-fold axis (Fig. 2B). The trimer is clover leaf-shaped with approximate dimensions of 100  90  50 Å. Furthermore, each monomer of Tt8MGHGlc and Tt8MGH-glycerate forms trimers with the other two monomers related by the crystallographic 3-fold axis, which is isomorphous with that of Tt8MGH-Tris in spite of the different space group (P321) of Tt8MGH-glycerate (Supplementary Fig. S6A). The active sites face the outside of the trimer (Fig. 2C). The total surface area of a monomer is 15220 Å2, whereas the buried interface area is 4470 Å2 (29% of the monomer surface). Twenty-four hydrogen bonds and 18 salt bridges are formed (eight hydrogen bonds and six salt bridges in each interface between two subunits), and 12 amino acid residues are likely to be involved in the hydrophobic interaction (Supplementary Fig. S6B and Table S1), indicating that Tt8MGH monomers bind each other tightly. The PISA analysis using the three crystal structures calculated the DGdiss (the free energy of assembly dissociation) values of 18.6–29.5 kcal mol1.

C 0.5

66 kDa

Kav

0.4 0.3 0.2 0.1

140 kDa 232 kDa 440 kDa 669 kDa

0 104

105

106

Molecular weight (Da)

B

90º

100 Å

90 Å

50 Å

Fig.2. Trimer of Tt8MGH. (A) Calibration curve for molecular weight estimation of Tt8MGH by gel filtration chromatography. Tt8MGH and standards are plotted as open circles and filled diamonds, respectively. (B) Trimers of Mol-A (magenta) generated by the crystallographic symmetry in Tt8MGH-Tris, viewed along (left) and across (right) the trimer axis. (C) Molecular surface model of the trimer of Tt8MGH-Tris. Tris molecules are shown as black and red spheres, and the entrances of the catalytic site are indicated with red arrows.

25

T. Miyazaki et al. / Journal of Structural Biology 190 (2015) 21–30

The positive values indicated that the homotrimer is thermodynamically stable. Thermostability of proteins is often the result of many hydrogen bonds, polar interactions, high packing density, and oligomerization (Villeret et al., 1998; Jaenicke and Böhm, 1998; Kumar et al., 2000; Lokanath et al., 2004; Tanaka et al., 2004). Alarico et al. reported that the optimal temperature for Tt27MGH hydrolysis of MG and GG was 70 °C, and the half-life at 70 °C was 16 h (Alarico et al., 2013). The characterized GH63 enzymes other than mannosylglycerate hydrolases (MGHs) were not reported to be so thermostable and to form oligomers. Considering these data, the trimerization of Tt8MGH is likely to contribute to its stability. 3.3. Complexes with Glc and glycerate, and catalytic site To determine the structures of Tt8MGH in complexes with Glc, Man, and glycerate, which are products of MGH hydrolysis for GG or MG, ligand-soaking and co-crystallization experiments were performed. No diffractable crystal was obtained under the experimental conditions with various concentrations of Man; however, crystals grown with Glc or glycerate were diffracted and analyzed. The electron density for a Glc molecule (Glc 1) was found at subsite 1 (the nomenclature is according to Davies et al., 1997) in Mol-A of Tt8MGH-Glc, but not in Mol-B (Figs. 1C, 3A and Supplementary Fig. S3). A part of loop-C (residues 180–201) in Mol-A is disordered, whereas loop-C in Mol-B was able to be fully modeled. Interestingly, Lys191 on loop-C in Mol-B is inserted into subsite 1, and Lys191 and His192 are likely to neutralize Asp35 and Glu393, respectively (Supplementary Fig. S4C and D). The difference between Mol-A and MolB of Tt8MGH-Glc may be due to its

A

crystal packing environment: loop-C in not Mol-A but Mol-B is in contact with the neighboring molecule, which is likely to prevent loop-C opening and glucose binding (loop-C conformations are discussed in Section 3.4). Glc 1 in Mol-A adopts a 4C1 chair conformation with an a-anomeric configuration, and it forms hydrogen bonds with the side chains of Trp34, Asp35, Gln103, Asp165, Tyr349, and Trp350 as well as the carbonyl oxygen of Gly163 (Fig. 3A). Trp32 is likely to interact hydrophobically with Glc 1. These observations indicate that subsite 1 strictly recognizes the a-glucose residue of the substrate GG. Unfortunately, the complex structure with Man could not be determined; however, we previously determined the structure of EcYgjK in complex with Man (EcYgjK-Man), where two b-mannose molecules (Man 1 and Man +1) occupied subsites 1 and +1 (PDB: 3W7T; Kurakata et al., 2008). Superposition of Tt8MGH-Glc and EcYgjK-Man shows that the amino acid residues of subsite 1 are well conserved, and the orientation of the pyranose rings of Glc 1 and Man 1 is almost identical (Fig. 3B). Alignment of the amino acid sequences of the characterized MGHs shows that the residues at subsite 1 were also fully conserved (Fig. 4). These observations suggest that the subsite 1 of Tt8MGH and other MGHs is acceptable for both Glc and Man residues of its substrates, which matches the kinetic study for MG and GG (Alarico et al., 2013). In Tt8MGH-glycerate, although parts of loop-B in both Mol-A (residues 91–95) and Mol-B (residues 87–95) are disordered, their whole structures are almost entirely identical (root mean square deviation [rmsd] = 0.39 Å). The clear electron densities for a Tris molecule and a glycerate molecule (glycerate +1) are found at subsites 1 and +1, respectively, in both Mol-A and Mol-B (Fig. 3C). Therefore, the following descriptions are based primarily on MolA. Glycerate +1 recognition is provided through interaction with

B W32

D35

W34 E393

D35 D324

E393 E727 (base)

W32 W321 W34 W323 Q103 K391

Q103

G163 G499

G163 Y349 Y679

Y349 D165

W350

C

D165 D501 (acid)

W350 W680

D

W32

D35

Y28 E393

H71

W34 E393

Y28 W32 H71

Y81 Y349

D165 W160

W350

Y81

Y349 W160 D165 R198

R198 Fig.3. Structures of the active sites of Tt8MGH-Glc Mol-A (A and B) and Tt8MGH-glycerate (C and D). (A and C) Glc, Tris, glycerate, and the residues interacting with the ligands are shown as stick models and colored in green, yellow, pink, and white, respectively. Hydrogen bonds are indicated by black dashed lines. The difference Fourier maps are calculated excluding the ligands, and the resulting Fo  Fc omit maps (blue mesh) are contoured at 2.0 r. (B) Structural comparison between subsites 1 of Tt8MGHGlc Mol-A (cyan) and EcYgjK-Man (yellow). Glc 1 and Man 1 are colored in green and magenta, respectively. (D) Superposition of the active sites of Tt8MGH-Glc Mol-A (green) and Tt8MGH-glycerate (pink). The ligands and the residues interacting with them are shown as thick and thin stick models, respectively.

26

T. Miyazaki et al. / Journal of Structural Biology 190 (2015) 21–30

loop-A

H1

loop-B

S1

A’ H 4

S5

S6

H3

S3

A’ H 1

S4

H2

H4

H5

loop-C

A’H2

A’S1

S7

S2

A’ H 3

H6

H7

H8

H1 0

H9

H1 1

S8

S9

H12

Fig.4. Sequence alignment of Tt8MGH and the characterized MGHs. The four sequences, Tt8MGH (GenBank: BAD70801.1), Tt27MGH (GenBank: AAS80962.1), RrMGH (GenBank: AFC76324.1), and SmMGH (GenBank: EFJ37158.1), were aligned by MUSCLE (Edgar, 2004) and the figure was produced with ESPript (Gouet et al., 2003). Residues forming the secondary structures are highlighted above the sequences of Tt8MGH. The identical residues are shown in white with a red background and conservative changes are shown in red with a white background. Residues interacting with Glc 1 and glycerate +1 are marked with green and pink circles, respectively, under the sequences. The catalytic residues are also marked with red stars under the sequences.

the side chains of Trp32, His71, Tyr81, Trp160, Asp165, Arg198, and Tyr349. Glycerate +1 is surrounded by these residues, especially the guanidino group of Arg198 which neutralizes the carboxylate group of glycerate +1, and Tyr81 which forms a hydrogen bond with the carboxylate group. The alignments of the amino acid sequences of the characterized MGHs show that all amino acid

residues involved in the recognition of glycerate +1 are fully conserved with the exception of His71 (Gln in RrMGH) (Fig. 4). Tyr81 and Arg198 were located on loop-B and loop-C, respectively, by which glycerate +1 is covered and rendered inaccessible to solvent. These observations indicate that significant conformational change would be required for substrate entry.

27

T. Miyazaki et al. / Journal of Structural Biology 190 (2015) 21–30

Y81

Y81 Y81

Y81

Y28 Y28

Y28 Y28

Y81

R198

Y81

R198 R198

R198

Fig.5. Stereo view of the conformational changes of the flexible loops. The ligands and the residues involved in substrate recognition located in the loops (Tyr28, Tyr81, and Arg198) are shown as stick models and stick-and-ball models, respectively. Colors are as follows: loop-A, blue (open) and cyan (closed); loop-B, orange (open-1), red (open-2), and yellow (closed); loop-C, purple (open) and slate blue (closed); Glc 1, green; Gly +1, pink.

Superposition of the active sites of Tt8MGH-Glc and Tt8MGH-glycerate is shown in Fig. 3D. The distance between atom O1 of Glc 1 and atom O2 of glycerate +1 is 0.8 Å, which is shorter than the distance between atom O1 of Glc 1 and any other oxygen atoms of glycerate +1. The structural comparison of Tt8MGH and EcYgjK shows that Asp165 and Glu393 are the acid and base catalysts of Tt8MGH, respectively (Fig. 3B). The average distance between the two carboxylate residues in Tt8MGH is 8.1 Å, which is consistent with the proposed distances between those of typical inverting glycoside hydrolases (8 ± 2 Å; calculated based on Mhlongo et al., 2014). These results indicate that the active site of Tt8MGH is suitable for the hydrolysis of a(1 ? 2) linkage of the substrates with inverting mechanism, which was elucidated by 1H-NMR study (Alarico et al., 2013). 3.4. Loop conformations and insights into their function In the present study, we determined that three crystal structures (Tt8MGH-Tris, Tt8MGH-Glc, and Tt8MGH-glycerate), and the conformations of three loops within them (loop-A, loop-B, and loop-C), are different from each other (Fig. 1 and Supplementary Fig. S3). The (a/a)6-barrel catalytic domain of EcYgjK was reported to show the open-to-closed conformational changes along the ligand-binding process (Miyazaki et al., 2013), but no marked difference was observed in the main chain of the (a/a)6-barrel between the structures of Tt8MGH with and without the ligands (rmsd = 0.32–0.39 Å). Loop-A adopts two forms, open and closed forms, and the open form was found only in Tt8MGH-Glc (Figs. 1 and 5). When Glc 1 was superposed onto the closed form of Tt8MGH-glycerate, Tyr28 in loop-A seemed to cap the ligand and shut out solvent (Fig. 3D). Therefore, loop-A is likely to open during substrate entry into the active site. Tyr28 is fully conserved among the characterized MGHs (Fig. 4) and forms hydrogen bonds with the catalytic bases Glu393, Arg198, and Tyr349, which recognize the ligands in the closed form (Fig. 3C). It has been hypothesized that Tyr28 supports the hydrolysis, but it is unclear whether loop-A adopts the closed form after the substrate binding to the catalytic cleft. Loop-B was found to have two distinct open forms (open-1 and open-2) and one closed form (Fig. 5). Tt8MGH-Tris exhibits the open-1 loop-B in Mol-A and the closed form in Mol-B (Fig. 1), whereas Tt8MGH-Glc and Tt8MGH-glycerate show the open-2 and closed forms, respectively. The protein structures without the ligand at subsite +1 may have various forms of loop-B, but glycerate +1 introduced the closed form of loop-B via interaction with

Tyr81, which is completely conserved among the characterized MGHs (Fig. 4). These observations suggested that loop-B is flexible when the substrate is free, and changes to the closed form after substrate binding at the catalytic site. In Tt8MGH-Glc, loop-C of Mol-A is disordered and loop-C of Mol-B is inserted into subsite 1 (closed form), whereas in the other structures determined in the present study, loop-C adopts the open form. The closed form of loop-C was also found in the Tt8MGH structure previously determined (PDB: 2Z07) using crystals grown under the same conditions (citrate buffer). His191 and Lys192, which interact with Glu393 and Asp35, respectively, are not fully conserved among the characterized MGHs (Fig. 4), supporting that the closed form is a crystallographic artifact and may not have an important role in catalysis. Although loop-C of Tt8MGH-Glc Mol-A is disordered, the superposition of Glc 1 onto Tt8MGH-glycerate indicates that there is no steric hindrance between Glc 1 and the open form of loop-C (Fig. 3D). The open form of loop-C holds glycerate +1 via Arg198 as described above. Considering these observations, loop-C seems to adopt the open form when the substrate binds the active site. Tyr349 is located in the proposed YWRXXXW motif, which is highly conserved among GH63 proteins (Kurakata et al., 2008). The corresponding residue Tyr679 of EcYgjK was reported to change its orientation to interact with the catalytic base during substrate binding and was thought to be involved in the catalysis (Miyazaki et al., 2013). Tyr28 (loop-A closed) and Arg198 (loop-C open) form hydrogen bond networks with each other, Tyr349, and the catalytic base Glu393. The loop motions may also correctly position the residues involved in the hydrolysis. It has been reported that Tt27MGH does not hydrolyze a-mannosides and a-glucosides aside from MG and GG (Alarico et al., 2013); similarly, Tt8MGH did not show activity on a-glucobioses (trehalose, kojibiose, nigerose, maltose, and isomaltose) (data not shown). Considering these results, three mobile loops are suggested to function as fingers to catch the substrates MG and GG, and to avoid hydrolyzing ‘wrong’ substrates such as disaccharides. Future mutational experiments and determination of the protein structures in complex with the substrates are required for validation of the catalytic mechanism. 3.5. Comparison between GH63 enzymes and distribution of substrate specificity For the GH63 protein family, the crystal structures of two enzymes, EcYgjK and ScCwh41p, have been reported (Kurakata

28

T. Miyazaki et al. / Journal of Structural Biology 190 (2015) 21–30

A

Q103 /K391 /N453

D35/324/392 W34/323/391

Q103 /K391 /N453

D35/324/392 W32/321/F389 W34/323/391

G163 /499 /566

W32/321/F389

G163 /499 /566

E393/727/771

E393/727/771

-1 +1

H71/D368 /R428

W350/680 /710

-1 +1

H71/D368 /R428

W350/680 /710 W160/496/L563

W160/496/L563

D165/501/568 Y349/679/709

D165/501/568 Y349/679/709

B Subfamily 3 (MGH subfamily)

Subfamily 2 (GGalase subfamily)

Subfamily 1 (processing α-glucosidase I subfamily)

Fig.6. Subfamilies of GH63 proteins based on their substrate specificities. (A) Stereo view of a comparison of the active sites between Tt8MGH (cyan), EcYgjK (green), and ScCwh41p (orange). Residues comprising subsites 1 and +1 are shown as stick models, and Glc 1 and glycerate +1 are indicated as black thin stick models. Residue numbers of EcYgjK and ScCwh41p refer to the literature (Kurakata et al., 2008; Barker and Rose, 2013). (B) Phylogenetic tree of GH63 proteins constructed with MEGA6 (Tamura et al., 2013) using the minimum evolution method. Subfamilies 1, 2, and 3 are highlighted in orange, green, and blue, respectively. The characterized enzymes listed in CAZy and the uncharacterized archaeal protein (GenBank: ABW01842.1) from C. maquilingensis IC-167 are labeled.

et al., 2008; Barker and Rose, 2013). EcYgjK and ScCwh41p are exoglucosidases that hydrolyze the a(1 ? 2)-glucosidic linkages of the disaccharide Glc-a(1 ? 2)-Gal and the distal Glc-a(1 ? 2)-Glc unit of the N-glycan precursor, respectively (Miyazaki et al., 2013; Barker and Rose, 2013). EcYgjK could accept both Glc and Man at subsite 1, but only Glc induced the conformational change of the catalytic domain to enable the catalytic acid to access the glycosidic linkage of the substrate, indicating that EcYgjK might hydrolyze a-glucoside rather than a-mannoside (Miyazaki et al., 2013). The characterized MGHs were reported to hydrolyze both a-glucosidic and a-mannosidic linkages of GG and MG (Alarico et al., 2013; Nobre et al., 2013), and the subsite 1 of Tt8MGH is likely to accept not only Glc but also Man according to its structure

determined in the present study. Therefore, the subsites 1 of Tt8MGH, EcYgjK, and ScCwh41p are specific for Glc/Man, Glc, and Glc, respectively, whereas the subsites +1 are specific for glycerate, galactose (Gal), and Glc, respectively. Superposition of the catalytic sites shows that the residues forming subsites +1 are variable, whereas those forming subsite 1 are completely conserved among the enzymes (Fig. 6A). To investigate the distribution of substrate specificity of GH63 enzymes, a phylogenetic tree was constructed using 474 sequences (without those partially sequenced) and listed in the CAZy database. The residues in uncharacterized GH63 proteins corresponding to the marker residues (comprising the subsites +1 in the characterized GH63 proteins) were examined. Substrate specificity

T. Miyazaki et al. / Journal of Structural Biology 190 (2015) 21–30

of more than half of GH63 proteins in CAZy was expected, and three subfamilies were identified: the processing a-glucosidase I subfamily (subfamily 1; 81 proteins), the GGalase subfamily (subfamily 2; 285 proteins), and the MGH subfamily (subfamily 3; 108 proteins) (Fig. 6B). Subfamilies 2 and 3 consist of only bacterial proteins except for SmMGH, whereas subfamily 1 comprises only eukaryotic proteins. The proteins belonging to subfamily 3 were found in bacterial phyla Actinobacteria, Aquificae, Bacteroidetes, Deinococcus-Thermus, and Proteobacteria as well as the plant S. moellendorffii (Supplementary Fig. S7). Some bacteria belonging to these phyla were reported to possess genes encoding GG or MG synthases (Alarico et al., 2013). Very recently, glucosylglycerate hydrolase from the actinobacteria Mycobacterium hassiacum was studied (Alarico et al., 2014), and had a sequence identity of 36% with Tt8MGH. The gene for SmMGH was thought to be derived from a gene laterally transferred from prokaryotes possessing MGH genes (Nobre et al., 2013). Only one archaeal member, Cmaq_1011 (GenBank: ABW01842.1, sequence identity with Tt8MGH = 28%) from Caldivirga maquilingensis IC-167, is classified in GH63 in the CAZy database, but the corresponding residues of the active site are not fully conserved with MGHs and the protein is not included in the above-noted subfamilies (Fig. 6). A Basic Local Alignment Search Tool (BLAST) search using Tt8MGH as a template found only two hypothetical archaeal MGHs, from Halosarcina pallida (GenBank: ELZ29971.1, sequence identity with Tt8MGH = 39%) and Halogranum salarium (GenBank: EJN60925.1, sequence identity with Tt8MGH = 39%), whose active site residues are fully conserved with MGHs; however, there is no indication that these archaea produce MG or GG. Since many archaea produce MG or GG (Empadinhas and da Costa, 2011), most of the archaea probably possess non-homologous GHs that act on them (Alarico et al., 2013; Borges et al., 2014). All proteins classified into subfamily 3 were predicted to consist of a single (a/a)6-barrel domain and A0 -region except for a hypothetical protein from P. marina (GenBank: ACO04607.1), which has a C-terminal domain of unknown function (DUF547). Conversely, EcYgjK and ScCwh41p possess the N-terminal b-sandwich domains in addition to their barrel domains. Considering that many subfamily 3 proteins are found in thermophilic and deep-lineage bacteria, structures of subfamily 3 proteins are likely to represent prototypes of GH63 proteins, although the evolution of GH63 proteins is still unclear.

4. Conclusion The crystal structures of Tt8MGH in apo form and in complexes with its products (Glc and glycerate) were determined at high resolutions, and its substrate recognition mechanism was proposed. Like other GH63 proteins, Tt8MGH has a single catalytic domain of (a/a)6-barrel; it also has three unique flexible loops, which are not conserved in characterized GH63 enzymes other than MGHs. The structures of Tt8MGH-ligand complexes indicate that the flexible loops are involved in substrate binding and hold the substrate like fingers without open-to-closed conformational changes of the barrel as observed in the bacterial GH63 enzyme EcYgjK. The residues comprising subsite 1 are fully conserved, but those of subsite +1 are variable among the three subfamilies of GH63, resulting in different substrate specificities. These results suggest that GH63 includes the a-glycosidase with especially different aglycon specificities and with different substrate recognition mechanisms. Further experiments such as mutational analysis and co-crystallization with substrates are required to fully elucidate the catalytic mechanisms of MGHs and other GH63 proteins.

29

Acknowledgments We thank Seiki Kuramitsu (Osaka University) for useful comments. We also thank Hiromi Yoshida, Shigehiro Kamitori (Kagawa University), and Akiko Shimizu-Ibuka (Niigata University of Pharmacy and Applied Life Sciences) for help with diffraction data collection. This work was supported in part by a Grant-in-Aid for Scientific Research (T.T., No. 26660277; T.M., No. 25-7279) and a Research Fellowship (T.M.) of the Japan Society for the Promotion of Science. This research was performed with approval of the Photon Factory Advisory Committee, the National Laboratory for High Energy Physics, Tsukuba, Japan (2012G006 and 2014G512). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.jsb.2015.02.006. References Alarico, S., Empadinhas, N., da Costa, M.S., 2013. A new bacterial hydrolase specific for the compatible solutes a-D-mannopyranosyl-(1 ? 2)-D-glycerate and a-Dglucopyranosyl-(1 ? 2)-D-glycerate. Enzyme Microb. Technol. 52, 77–83. Alarico, S., Costa, M., Sousa, M.S., Maranha, A., Lourenço, E.C., Faria, T.Q., Ventura, M.R., Empadinhas, N., 2014. Mycobacterium hassiacum recovers from nitrogen starvation with up-regulation of a novel glucosylglycerate hydrolase and depletion of the accumulated glucosylglycerate. Sci. Rep. 4, 6766. Aleshin, A.E., Feng, P.H., Honzatko, R.B., Reilly, P.J., 2003. Crystal structure and evolution of a prokaryotic glucoamylase. J. Mol. Biol. 327, 61–73. Barker, M.K., Rose, D.R., 2013. Specificity of processing a-glucosidase I is guided by the substrate conformation: crystallographic and in silico studies. J. Biol. Chem. 288, 13563–13574. Borges, N., Jorge, C.D., Gonçalves, L.G., Gonçalves, S., Matias, P.M., Santos, H., 2014. Mannosylglycerate: structural analysis of biosynthesis and evolutionary history. Extremophiles 18, 835–852. Chen, V.B., Arendall 3rd, W.B., Headd, J.J., Keedy, D.A., Immormino, R.M., Kapral, G.J., Murray, L.W., Richardson, J.S., Richardson, D.C., 2010. MolProbity: all-atom structure validation for macromolecular crystallography. Acta Crystallogr. D Biol. Crystallogr. 66, 12–21. Colin, H., Augier, J., 1939. Un glucide original chez les floridées du genre Polysiphonia le D-mannoside de L-glycérate de sodium. C. R. Acad. Sci. 208, 1450–1453. Collaborative Computational Project Number 4, 1994. The CCP4 suite: programs for protein crystallography. Acta Crystallogr. D Biol. Crystallogr. 50, 760–763. Davies, G.J., Wilson, K.S., Henrissat, B., 1997. Nomenclature for sugar-binding subsites in glycosyl hydrolases. Biochem. J. 321, 557–559. Dhanawansa, R., Faridmoayer, A., van der Merwe, G., Li, Y.X., Scaman, C.H., 2002. Overexpression, purification, and partial characterization of Saccharomyces cerevisiae processing alpha glucosidase I. Glycobiology 12, 229–234. Edgar, R.C., 2004. MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res. 32, 1792–1797. Empadinhas, N., da Costa, M.S., 2011. Diversity, biological roles and biosynthetic pathways for sugar-glycerate containing compatible solutes in bacteria and archaea. Environ. Microbiol. 13, 2056–2077. Emsley, P., Lohkamp, B., Scott, W.G., Cowtan, K., 2010. Features and development of Coot. Acta Crystallogr. D Biol. Crystallogr. 66, 486–501. Frade-Pérez, M.D., Hernández-Cervantes, A., Flores-Carreón, A., Mora-Montes, H.M., 2010. Biochemical characterization of Candida albicans a-glucosidase I heterologously expressed in Escherichia coli. Antonie Van Leeuwenhoek 98, 291–298. Faridmoayer, A., Scaman, C.H., 2004. An improved purification procedure for soluble processing a-glucosidase I from Saccharomyces cerevisiae overexpressing CWH41. Protein Expr. Purif. 33, 11–18. Faridmoayer, A., Scaman, C.H., 2005. Binding residues and catalytic domain of soluble Saccharomyces cerevisiae processing a-glucosidase I. Glycobiology 15, 1341–1348. Faridmoayer, A., Scaman, C.H., 2007. Truncations and functional carboxylic acid residues of yeast processing a-glucosidase I. Glycoconj. J. 24, 429–437. Gibson, R.P., Gloster, T.M., Roberts, S., Warren, R.A.J., Storch de Gracia, I., García, A., Chiara, J.L., Davies, G.J., 2007. Molecular basis for trehalase inhibition revealed by the structure of trehalase in complex with potent inhibitors. Angew. Chem. Int. Ed. Engl. 46, 4115–4119. Gouet, P., Robert, X., Courcelle, E., 2003. ESPript/ENDscript: extracting and rendering sequence and 3D information from atomic structures of proteins. Nucleic Acids Res. 31, 3320–3323. Helenius, A., Aebi, M., 2004. Roles of N-linked glycans in the endoplasmic reticulum. Annu. Rev. Biochem. 73, 1019–1049. Herscovics, A., 1999. Importance of glycosidases in mammalian glycoprotein biosynthesis. Biochim. Biophys. Acta 1473, 96–107.

30

T. Miyazaki et al. / Journal of Structural Biology 190 (2015) 21–30

Hidaka, M., Honda, Y., Kitaoka, M., Nirasawa, S., Hayashi, K., Wakagi, T., Shoun, H., Fushinobu, S., 2004. Chitobiose phosphorylase from Vibrio proteolyticus, a member of glycosyl transferase family 36, has a clan GH-L-like (a/a)6 barrel fold. Structure 12, 937–947. Holm, L., Kääriäinen, S., Rosenström, P., Schenkel, A., 2008. Searching protein structure databases with DaliLite v. 3. Bioinformatics 24, 2780–2781. Jaenicke, R., Böhm, G., 1998. The stability of proteins in extreme environments. Curr. Opin. Struct. Biol. 8, 738–748. Kalz-Füller, B., Bieberich, E., Bause, E., 1995. Cloning and expression of glucosidase I from human hippocampus. Eur. J. Biochem. 231, 344–351. Kempf, B., Bremer, E., 1998. Uptake and synthesis of compatible solutes as microbial stress responses to high-osmolality environments. Arch. Microbiol. 170, 319– 330. Krissinel, E., Henrick, K., 2007. Inference of macromolecular assemblies from crystalline state. J. Mol. Biol. 372, 774–797. Kumar, S., Tsai, C.J., Nussinov, R., 2000. Factors enhancing protein thermostability. Protein Eng. 13, 179–191. Kurakata, Y., Uechi, A., Yoshida, H., Kamitori, S., Sakano, Y., Nishikawa, A., Tonozuka, T., 2008. Structural insights into the substrate specificity and function of Escherichia coli K12 YgjK, a glucosidase belonging to the glycoside hydrolase family 63. J. Mol. Biol. 381, 116–128. Lokanath, N.K., Shiromizu, I., Ohshima, N., Nodake, Y., Sugahara, M., Yokoyama, S., Kuramitsu, S., Miyano, M., Kunishima, N., 2004. Structure of aldolase from Thermus thermophilus HB8 showing the contribution of oligomeric state to thermostability. Acta Crystallogr. D Biol. Crystallogr. 60, 1816–1823. Lombard, V., Golaconda Ramulu, H., Drula, E., Coutinho, P.M., Henrissat, B., 2014. The carbohydrate-active enzymes database (CAZy) in 2013. Nucleic Acids Res. 42, D490–495. Mhlongo, N.N., Skelton, A.A., Kruger, G., Soliman, M.E., Williams, I.H., 2014. A critical survey of average distances between catalytic carboxyl groups in glycoside hydrolases. Proteins 82, 1747–1755. Miyazaki, T., Matsumoto, Y., Matsuda, K., Kurakata, Y., Matsuo, I., Ito, Y., Nishikawa, A., Tonozuka, T., 2011. Heterologous expression and characterization of processing a-glucosidase I from Aspergillus brasiliensis ATCC 9642. Glycoconj. J. 28, 563–571.

Miyazaki, T., Ichikawa, M., Yokoi, G., Kitaoka, M., Mori, H., Kitano, Y., Nishikawa, A., Tonozuka, T., 2013. Structure of a bacterial glycoside hydrolase family 63 enzyme in complex with its glycosynthase product, and insights into the substrate specificity. FEBS J. 280, 4560–4571. Murshudov, G.N., Vagin, A.A., Dodson, E.J., 1997. Refinement of macromolecular structures by the maximum-likelihood method. Acta Crystallogr. D Biol. Crystallogr. 53, 240–255. Nobre, A., Empadinhas, N., Nobre, M.F., Lourenço, E.C., Maycock, C., Ventura, M.R., Mingote, A., da Costa, M.S., 2013. The plant Selaginella moellendorffii possesses enzymes for synthesis and hydrolysis of the compatible solutes mannosylglycerate and glucosylglycerate. Planta 237, 891–901. Otwinowski, Z., Minor, W., 1997. Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307–326. Palcic, M.M., Scaman, C.H., Otter, A., Szpacenko, A., Romaniouk, A., Li, Y.X., Vijay, I.K., 1999. Processing a-glucosidase I is an inverting glycosidase. Glycoconj. J. 16, 351–355. Perrakis, A., Morris, R., Lamzin, V.S., 1999. Automated protein model building combined with iterative structure refinement. Nat. Struct. Biol. 6, 458–463. Tamura, K., Stecher, G., Peterson, D., Filipski, A., Kumar, S., 2013. MEGA6: molecular evolutionary genetics analysis version 6.0. Mol. Biol. Evol. 30, 2725–2729. Tanaka, Y., Tsumoto, K., Yasutake, Y., Umetsu, M., Yao, M., Fukada, H., Tanaka, I., Kumagai, I., 2004. How oligomerization contributes to the thermostability of an archaeon protein. Protein L-isoaspartyl-O-methyltransferase from Sulfolobus tokodaii. J. Biol. Chem. 279, 32957–32967. Vagin, A., Teplyakov, A., 1997. MOLREP: an automated program for molecular replacement. J. Appl. Crystallogr. 30, 1022–1025. Villeret, V., Clantin, B., Tricot, C., Legrain, C., Roovers, M., Stalon, V., Glansdorff, N., Van Beeumen, J., 1998. The crystal structure of Pyrococcus furiosus ornithine carbamoyltransferase reveals a key role for oligomerization in enzyme stability at extremely high temperatures. Proc. Natl. Acad. Sci. USA 95, 2801–2806. Wallace, A.C., Laskowski, R.A., Thornton, J.M., 1995. LIGPLOT: a program to generate schematic diagrams of protein–ligand interactions. Protein Eng. 8, 127–134.