3.2.1. The overall structure of StGE2

StGE2 belongs to the /-hydrolase superfamily and is a serine-type hydrolase; its overall architecture follows the three-layer -sandwich hydrolase fold with a Rossmann-fold topology, as revealed by its three-dimensional structure (Fig. 4). However, deviations from the canonical /-hydrolase fold are observed; the canonical fold features a -sheet of eight strands in the core of the structure intercalated between two clusters of -helices (four and two -helices bilaterally; Ollis et al., 1992). The -sheet is expanded by the insertion of two antiparallel -strands at the N-terminus of StGE2, resulting in a twisted -sheet. The latter is sandwiched between two layers containing a total of 18 helices. Eight -helices and two 3₁₀-helices are observed on one side of the -sheet instead of two in the canonical fold, while on the opposite side an octet of -helices and 3₁₀-helices that are equally divided, rather than only four, are recorded. The positions of two of the putative catalytic triad residues, namely Ser213 and Glu236, comply with those suggested by the canonical fold. Detailed analysis of the secondary structure according to PROMOTIF (Hutchinson & Thornton, 1996) as implemented in PDBsum (Laskowski et al., 1997) is presented in Supplementary Table S1.

The three-dimensional structure of StGE2 with one molecule per asymmetric unit follows the monomeric form that StGE2 adopts in solution, as suggested by HPLC gel filtration (data not shown). The latter was confirmed by analysis of the protein interfaces using the PISA server (http://www.ebi.ac.uk/pdbe/prot_int/pistart.html ; Krissinel & Henrick, 2007), revealing the absence of specific interactions that could result in the formation of stable quaternary structures. The overall structure is stabilized by three disulfide bonds, namely Cys31-Cys65, Cys212-Cys347 and Cys244-Cys319.

Residues Ser213, Glu236 and His346 of StGE2 could be proposed as putative catalytic triad residues since their architecture complies with that proposed for the catalytic mechanism of / hydrolases. Specifically, Ser213 is a nucleophilic serine located between strand S6 and helix H6 forming the so-called `nucleophilic elbow', with Ramachandran angles lying in the generously allowed regions (Ollis et al., 1992; Nardini & Dijkstra, 1999; Quevillon-Cheruel et al., 2005). His346 (which acts as an acid/base complementing the nature of the substrate upon binding) bridges strand S9 and helix G6, while Glu236 (which is suggested to adjust the basic character of His346) is situated between strand S7 and helix H7 (Correia et al., 2008). The exact locus and organization of the catalytic triad residues is stabilized by the tight hydrogen-bond interactions that are formed between them. The hydroxyl group of Ser213 interacts with His346 NE2, while Glu236 OE1 is hydrogen-bonded to His346 ND1 (at distances of 2.7 and 2.8 Å, respectively).

A total of 14 ethylene glycol molecules (present in the crystallization medium) and two glycerol molecules (used as cryoprotectant) were also identified in the structure of StGE2. They are predominantly positioned at the interface of the packed monomers, but without providing evidence for specific binding.

3.2.2. Comparison of StGE2 with Cip2_GE from H. jecorina

Comparison of the StGE2 structure with that of Cip2_GE, the only known member of the CE15 family (Pokkuluri et al., 2011) for which a structure has been determined to date, showed that the overall structure of the enzymes is homologous and that they share 56% sequence identity (over the residues observed in the crystal structure, with 97% coverage) and a Z-score of 59 with an r.m.s.d. of 1.1 Å (as determined by the DALI server for monomer B, the closest structural homologue).

However, superposition of the two crystal structures on C atoms (Winn et al., 2011) gave an r.m.s.d. of 3.5 Å, revealing changes in selected secondary-structure elements and in the N- and C-termini. Detailed comparison of the two structures showed that the differences detected at sequence level involve either the insertion/deletion of residues leading to alterations in the secondary structure, the presence/absence of a glycosylation site or the packing of the monomers in the crystal. Specifically, the insertion of two residues led to the formation of an -helix (H5, amino acids 197-200) instead of the 3₁₀-helix observed in the Cip2_GE structure (Figs. 4, 5 and 6). The resulting structure is stabilized through a water-mediated interaction formed between Gln188 and Glu81' from a symmetry-related molecule. Similar changes were recorded for residues 276-278, which form a 3₁₀-helix (G2) that is shorter in length compared with the Cip2_GE structure. Also, residues 289-292 form a 3₁₀-helix (G4) and an -helix (H9) rather than only an -helix as in Cip2_GE, and residues 309-311 form a loop instead of a 3₁₀-helix as in Cip2_GE. All of the aforementioned changes are observed at the interface of the monomers with symmetry-related molecules in the unit cell. Of these, it is the latter modification involving residues 309-311 that is considered to be most significant since it is situated in the vicinity of the catalytic site of StGE2. This is accompanied by a difference in the number of residues forming strand S9 (residues 336-340), which contains one more residue compared with the corresponding strand in the Cip2_GE structure (Figs. 5 and 6) and adopts a slightly more extended conformation. The succeeding residues belong to a tight turn bearing the catalytic triad residue His346 (see §3.2.4).

Additional profound changes were observed; however, these could be attributed to shifts imposed by intermolecular interactions of Cip2_GE monomers (with three molecules in the asymmetric unit) as well as interactions with symmetry-related molecules dictated by crystal packing as in the case of the N- and C-termini (Pokkuluri et al., 2011). Moreover, the residues lining the H12 -helix and the preceding loop region are distorted compared with those in the Cip2_GE structure since the N-linked glycosylation sequence motif (Asn-X-Ser) and the N-acetylglucosamine molecule bound at Asn447 observed in the latter are missing in StGE2 (Asn-X-Ala is the corresponding motif in StGE2; Fig. 6).

Previous studies of the effect of temperature on StGE2 stability showed that StGE2 remains active at 323 K for at least 24 h, maintains more than 70% of its activity at 333-343 K and exhibits half-lives of 22 h at 328 K and 0.5 h at 333 K (Topakas et al., 2010), while Cip2_GE is only quite stable at 313 K, with half-lives of 10 and 2 min at 323 and 333 K, respectively (Li et al., 2007). A comparative analysis of the structures of the thermophilic StGE2 and its mesophilic homologue Cip2_GE was performed to monitor the differences in the structural determinants that could affect their thermostabilities. The typical indicators employed to assess protein thermostability are rather diverse. At the sequence level the protein composition of prolines has been examined, taking into account their limited allowed configurations and the restrictions that they impose on the preceding residue (Prajapati et al., 2007). The primary structure of StGE2 contains 22 proline residues compared with 17 in Cip2_GE. Moreover, the number of thermolabile residues such as Met, Cys, Asn and Gln was considered. StGE2 contains 48 thermolabile residues compared with 54 in Cip2_GE, as is observed in the majority of thermophilic proteins when compared with their mesophile homologues (Kumar et al., 2000). The ratio of Arg/(Arg+Lys) has also been used as a indicator for this purpose. StGE2 exhibits an increased Arg/(Arg+Lys) ratio compared with Cip2_GE (0.56 and 0.5, respectively; Mrabet et al., 1992). Similarly, electrostatic interactions are among the structural features which are known to be shared by thermostable proteins. The structures of StGE2 and Cip2_GE present the same number of salt bridges and cation- interactions, as calculated using the ESBRI server (Costantini et al., 2008) and CaPTURE (Gallivan & Dougherty, 1999).

3.2.3. The overall structures of the unliganded and bound forms of the S213A mutant

The S213A mutant was prepared following a previously developed protocol (Topakas et al., 2010) to shed light on the structure-function relationships of the enzyme and the putative role of the nucleophilic Ser in the substrate-recognition/catalytic mechanism. Determination of its three-dimensional structure at 1.9 Å resolution revealed that overall it remained unchanged with reference to the StGE2 crystal structure except for some areas involved in crystallographic symmetry interactions. Superposition of the S213A mutant atomic coordinates with those of StGE2 on C atoms showed that the two structures differed only slightly by an r.m.s.d. of 0.4 Å. Similarly, the crystal structure of the S213A mutant in complex with MCU indicated that the substrate analogue bound at the catalytic site of the enzyme without affecting the overall structure (r.m.s.d. on C atoms of 0.40 and 0.12 Å with reference to the StGE2 and the S213A mutant structures, respectively; Fig. 6). Ethylene glycol and glycerol molecules were also bound in both structures as in the case of StGE2. Of these, two ethylene glycol molecules were located in the active-site region of the S213A mutant structure (Table 1, Fig. 6; see §3.2.4).

3.2.4. StGE2 catalytic site

Overall, the catalytic site in the monomeric three-dimensional structure of StGE2 lies on the surface of the molecule; it is mostly exposed to the solvent and is not affected by symmetry-related packing interactions. All residues lining the putative catalytic triad face the solvent, with accessible surface areas for Ser213, Glu236 and His346 of 27.1, 2.7 and 75.7 Å², respectively. The catalytic residues Ser213 and Glu236 are not occluded by symmetry-related polypeptide chains, except for those in the vicinity of His346. The StGE2 structure suggests that strands S9 and S7 shape the active site, coordinating His346 and Glu236 towards forming the catalytic triad with Ser213 (Fig. 4). The catalytic triad residues are involved in a network of hydrogen bonds that contribute to the rigid architecture of the catalytic pocket. Focusing on the nucleophilic Ser213, its hydroxyl group interacts directly with His346 NE2, which in turn forms a hydrogen bond between its ND1 atom and Glu236 OE1. The setting is complemented by additional direct and water-mediated polar interactions involving residues Arg214, Gly216, Lys217, Gln235, Phe304 and Asn306 (Supplementary Table S2). This mode of interaction provides structural evidence for the nucleophilic role of Ser213, while His346 is expected to act as an acid-base complementing the nature of the substrate upon binding and Glu236 is expected to play a role as a regulator of the basic character of histidine (Correia et al., 2008).

Mutation of Ser213 to Ala showed that the absence of the hydroxyl group did not affect the active-site architecture. The hydrogen bonds formed between the Ala213 backbone atoms (N and O) and nearby residues, as well as the contacts with His346 and Glu236, were maintained. The loop bearing His346 (residues 344-350) is subjected to shifts ranging from 0.5 to 1.0 Å in the S213A mutant owing to crystal-packing interactions. Despite these shifts, the position of His346 in the catalytic site is preserved (minor shifts of 0.5 Å of all atoms). The disulfide bond formed between Cys347 and Cys212 further enhances the rigidity of the catalytic site by bridging the two adjacent catalytic triad residues (His346 and Ser213, respectively). Additional electron density for two ethylene glycol molecules EDO406 and EDO408 was identified in the S213A mutant structure at the same site. The latter lay at the entrance of the cavity, forming five potential hydrogen bonds to the neighbouring residues Lys217 NZ, Gln259 OE1 and NE2, Glu267 OE2 and Trp310 NE1 (Supplementary Fig. S1). This finding reveals that the catalytic cavity is exposed to the solvent and has the potential to accommodate the substrate by forming favourable binding interactions (Supplementary Tables S3, S4 and S5).

With the aim of advancing our knowledge of the structure-function relationship of StGE2 and elucidating the underlying mechanism of enzyme action, structural studies of the S213A mutant in complex with the substrate analogue MCU were performed. The results revealed that the substrate analogue bound at the catalytic site, forming an extended network of interactions (seven direct and four indirect hydrogen bonds as well as 59 van der Waals interactions) with the active-site residues (Tables 2 and 3, Figs. 7 and 8). The residues participating in the hydrogen-bond network were those observed in the S213A mutant structure with ethylene glycol together with His346 NE2. Two O atoms of the glucopyranose ring, O2 and O3, are aligned with the positions of the ethylene glycol hydroxyl groups shifted by 0.7 and 0.2 Å, respectively (Figs. 7 and 8, Supplementary Fig. S1).

Figure 7 LIGPLOT diagram of MCU interacting with the active-site residue lying in the vicinity. The substrate-analogue bonds are shown in purple, while bonds of the residues lining the site are shown in black. Hydrogen bonds are shown as black dashed lines with distances indicated in Å. Additional residues forming van der Waals interactions with MCU are represented by red semicircles with radiating spokes.

Figure 8 Stereo diagram of the interactions between the S213A mutant and MCU bound at the catalytic pocket as well as the hydrogen bonds formed by the catalytic triad residues. The side chains and backbone atoms of protein residues involved in ligand binding are shown in ball-and-stick representation. Water molecules (w) are depicted as spheres and hydrogen-bond interactions are shown as dotted lines. This figure was prepared with MolScript (Kraulis, 1991) and was rendered with Raster3D (Merritt & Bacon, 1997).

Specifically, the substrate analogue rests in the cavity using the peripheral hydroxyl group O atoms of the glucopyranose ring as an anchor to promote its binding. Both the Gln259 and Glu267 side-chain atoms are hydrogen-bonded to O1, O2 and O3 of the sugar. The conformation of the Glu267 side chain is altered [rotation of its (₁, ₃) dihedral angles by (74°, 127°)] compared with both the S213A mutant and the StGE2 structures, favouring tight binding of the substrate analogue. Trp310 NE1 is also in close contact with O2, while Lys217 NZ interacts with O3. An additional polar contact is formed between Lys217 NZ and O4 at a distance of 3.4 Å. O5 completes this hydrogen-bonding arrangement by taking part in water-mediated interactions with Arg214 NH1 (through water molecules Wat693 and Wat695). O6a, which pertains to the ester bond to be cleaved, is hydrogen-bonded to the NE2 atom of the imidazole ring of His346; this contact contributes to the suggestion that the position of the catalytic triad is rather concrete and the pocket is crafted to impose a `ready for nucleophilic attack' orientation of the substrate regardless of the absence of the nucleophile. To this end, comparison of the StGE2 crystal structure with the complex structure demonstrates that the Ser213 OG atom is indeed facing the ester bond at a distance of 2.2 Å.

The van der Waals interactions of the substrate analogue mainly concern the same palette of residues as those involved in the aforementioned hydrogen bonds. Among these, attention is drawn to the backbone N atom of Arg214 and its counterparts, O4 of the methoxy moiety and O6 of the ester group (at a distance of 3.4 Å), which is also hydrogen-bonded to Ser213 OG, implying that Arg214 is positioned to be part of the so-called oxyanion hole (Figs. 7 and 8).

Superposition of the structures of the unliganded S213A mutant and the complex showed that MCU binds at the same position as previously occupied by one of the two ethylene glycol molecules bound in the catalytic pocket in the unliganded structure. In particular, EDO408 mimics the positions of the C2, O2, C3 and O3 atoms of the glupyranose ring, while EDO406 lies at the vicinity of the site. The complex was prepared using preformed crystals of the S213A mutant; therefore, it appears that the substrate analogue displaced the ethylene glycol molecules from the catalytic pocket upon binding, explaining the quality of the electron density attributed to the ligand (Figs. 3c and 7 and Supplementary Fig. S1).