1 CRYSTAL STRUCTURE OF THE γ-GLUTAMYLTRANSPEPTIDASE PRECURSOR PROTEIN FROM ESCHERICHIA COLI : STRUCTURAL CHANGES UPON AUTOCATALYTIC PROCESSING AND IMPLICATIONS FOR THE MATURATION MECHANISM

CRYSTAL STRUCTURE OF THE γ-GLUTAMYLTRANSPEPTIDASE PRECURSOR PROTEIN FROM ESCHERICHIA COLI: STRUCTURAL CHANGES UPON AUTOCATALYTIC PROCESSING AND IMPLICATIONS FOR THE MATURATION MECHANISM Toshihiro Okada, Hideyuki Suzuki, Kei Wada, Hidehiko Kumagai, and Keiichi Fukuyama From the Department of Biological Sciences, Graduate School of Science, Osaka University, Toyonaka, Osaka 560-0043, Japan, Division of Integrated Life Science, Graduate School of Biostudies, Kyoto University, Sakyo-ku, Kyoto 606-8502, Japan, Research Institute for Bioresources and Biotechnology, Ishikawa Prefectural University, Nonoichi-cho, Ishikawa 921-8836, Japan Running Title: Maturation of γ-Glutamyltranspeptidase Correspondence should be addressed to K. Fukuyama (Phone: +81-6-6850-5422, Fax: +81-6-6850-5425, E-mail: fukuyama@bio.sci.osaka-u.ac.jp)

water and Thr-391 relative to the scissile peptide bond appears suitable for the initiation of autocatalytic processing, as in other members of the N-terminal nucleophile hydrolase superfamily.
γ-Glutamyl-X + H 2 O → Glutamate + X (hydrolysis) γ-Glutamyl-X + X' → γ-Glutamyl-X' + X (transpeptidation) GGT, a member of the N-terminal nucleophile (Ntn) hydrolase superfamily, is an extracellular enzyme that is widely distributed from bacteria to mammals (1)(2)(3) and plays a variety of physiological roles.The most abundant substrates for GGT are glutathione (GSH; γ-glutamyl-cysteinyl-glycine) and GSH-conjugated compounds.GGT catalyses the initial step of the degradation of GSH into constituent amino acids that are then transported into the cell and used as cysteine and nitrogen sources in Escherichia coli, yeast, and mammalian cells (4)(5)(6).In mammals, GGT catalyzes the initial step of the conversion of GSH conjugates into mercapturic acid, which is subsequently excreted into bile and urine (3).
Mature GGT is a heterodimeric enzyme comprising one large (L-) and one small (S-) subunit (1,3).GGT is generated from a precursor protein by post-translational autocatalytic processing (7); other proteins which undergo such processes include Hedgehog proteins (8), pyruvoyl-dependent enzymes (9), and other members of the Ntn hydrolase superfamily (10)(11)(12)(13)(14).During the maturation process of GGT, the scissile bond is hydrolyzed to form the L-and S-subunits.Mutation of the N-terminus of the S-subunit in mature GGT (Thr-391) had significant effects on processing activity (15)(16)(17).In particular, an alanine-substituted mutant, T391A, was isolated as the precursor form only, completely lacking post-translational processing ability.Biochemical studies of T391S and T391C indicated that the post-translational processing of GGT is an intramolecular autocatalytic event, and that Thr-391, which is the catalytic nucleophile in the mature enzyme, is also the catalytic residue for the processing reaction (16,17).Uncoupling of enzymatic and auto-processing activities was verified for Helicobacter pyroli GGT (18).Interestingly, the recently determined crystal structure of E. coli mature GGT showed that the C-terminal region of the L-subunit was distant from the N-terminal region of the S-subunit (19).This result apparently demonstrates that a large conformational change has occurred upon processing.
We report here the crystal structure of E. coli GGT T391A mutant protein that lacks autocatalytic processing ability, refined at 2.55 Å resolution.Structure comparison between the T391A protein and the mature GGT shows that marked structural changes occur during the maturation process in the segment corresponding to the C-terminal region of the L-subunit and in the segment that forms the substrate-binding pocket.Furthermore, the flexible nature of the loop that covers the substrate-binding pocket in the mature GGT in the resting state was demonstrated by the crystallographic analysis of its different crystal form.The molecular mechanism of processing is discussed on the basis of the structure around the scissile bond in the T391A protein.
Crystallization and Data Collection -The ammonium sulfate T391A protein precipitate was dissolved in 50 mM Hepes buffer (pH 7.0), and then desalted by repeated concentration using Vivaspin filter (Sartorius, Goettingen, Germany) and dilution with the buffer.Crystallization conditions were screened with the hanging-drop vapor diffusion method using the PEG/Ion Screen Kit (Hampton Research, Aliso Viejo, USA) and JB Screen Kit (Jena Bioscience, Jena, Germany).The hanging drop was prepared by mixing 1 μl of protein solution (4 mg/ml) with 1 μl of reservoir solution and was equilibrated at 4 °C against 200 μl of reservoir solution.Promising crystals were grown in the drops when either B6 or C1 of JB Screen No. 3 was used as the reservoir solution.Diffraction-quality crystals were produced when the concentrations of PEG 4000 and iso-propanol in the reservoir solution were optimized (18% PEG 4000 and 10% iso-propanol in 0.1 M sodium citrate).The crystals grew in a week to a typical size of 0.1 x 0.1 x 0.2 mm.
Crystals were soaked in cryoprotectant solution, which was prepared by adding PEG 4000 and glycerol to the reservoir solution to final concentrations of 22.5% and 1%, respectively, and flash cooled with a nitrogen gas stream at 100 K. Diffraction data for the T391A protein crystal were collected using synchrotron radiation and the Quantum 315 detector (Area Detector Systems, Poway, USA) at beamline BL41XU, at SPring-8 (Harima, Japan).
The mature GGT in the resting state was crystallized in the monoclinic form (space group P2 1 ; SeMet-GGT-P21) under the condition described previously (19).The crystals were soaked in cryoprotectant solution, which was prepared by adding glycerol to final concentration of 15% to the reservoir solution (20% PEG 4000 and 0.2 M CaCl 2 in 0.1 M Tris-HCl, pH 8.5), and flash cooled in a similar manner as for T391A.Diffraction data were collected using synchrotron radiation and the Jupiter 210 CCD detector (Rigaku/MSC, The Woodlands, TX) at beamline BL38B1, at SPring-8.All diffraction data were processed and scaled using the HKL2000 suite (21).Results of the data collection are provided in Table I.
Structure Determination and Refinement -The structure of the T391A protein was solved by the molecular replacement method with MOLREP in the CCP4 suite (22) using the atomic parameters of the A molecule of the mature GGT (PDB code 2DBU) as the search model, in which residues 375-390 of the L-subunit, which are assumed to take another conformation, were excluded, and Thr-391 was substituted to alanine.Two independent molecules in the asymmetric unit were located by a cross-rotation function and translation search.The F o -F c map calculated after the initial refinement using REFMAC5 (23) at 2.55 Å resolution showed positive continuous density for residues 375-390.These residues occupied different sites in the T391A protein and the mature GGT; residues 385-390 in the T391A protein occupied the site where residues 442-447 are located in the mature GGT.In addition, the electron densities corresponding to residues 29-39, 114-146, 250-265, 341-390, and 430-459 were ambiguous, indicating that these residues are disordered or have different conformations from the model.After removing these residues, several cycles of TLS and restraint refinement with REFMAC5 (24) and manual revision of the model with XTALVIEW/XFIT (25) were performed using 30-2.55Å resolution data.The removed residues were incorporated into the model in the next cycle when the electron density was visible.When only the density for the main-chain atoms was visible, the residue was treated as alanine.Finally, water molecules were picked using XTALVIEW/XFIT and further refinement was carried out with REFMAC5.Except for Lys-382 and Leu-383 in the A molecule, all main-chain atoms and most side-chain atoms in the segment 375-390 could be located in the electron density map.
The structure of SeMet-GGT-P21 was solved by the molecular replacement method in a similar manner as for the T391A protein.The structure was refined at 1.95 Å resolution, in which the residues invisible in the electron density map (residues 438-449 in A molecule and 439-448 in B molecule) were excluded.The stereochemistry of each model was checked using PROCHECK (26) .The secondary structures were defined by DSSP (27).Refinement statistics are presented in Table I.

RESULTS AND DISCUSSION
Overall structure of GGT T391A protein -The T391A mutant protein, which lacks intramolecular autocatalytic processing activity (16), was expressed in E. coli, purified, and crystallized.The crystal structure of the protein was refined to 2.55-Å resolution and R work and R free factors of 0.217 and 0.270, respectively.There are two molecules (A and B) in the asymmetric unit.The conformation of the peptide segment containing Gln-390-Ala-391, which corresponds to the processing site in the wild-type protein, was clearly defined in the electron density map (Fig. 1).The number of segments with high mobility was increased in the T391A protein relative to the mature GGT; both the main-chains and side-chains of 108 residues and the side-chains of 20 additional residues (out of 1112 residues in the A and B molecules) were invisible in the T391A protein whereas only 15 residues were invisible in the mature GGT (19).The A and B molecules of the T391A protein are superimposable with an r.m.s.(root mean square) deviation of 0.39 Å for 494 pairs of corresponding Cα atoms.
The folding of the T391A protein (B molecule) is shown in Fig. 2A.The T391A protein has a stacked αββα-core structure comprising two central β-sheets and surrounding α-helices, similar to the mature GGT (19).Superimposition of the Cα traces of T391A and mature GGT is shown in Fig 2B .Notably, in the mature GGT, the C-terminal residue of the L-subunit was far (ca.36 Å; Fig. 2B) from the N-terminal Thr-391 of the S-subunit (19).Residues 375-390 in the T391A protein took on an extended conformation on the molecular surface.We denote this segment as the P-segment.The P-segments in the T391A protein and in the mature GGT are directed to opposite sides at Ile-378 due to differing ψ values (Fig. 2C; ψ=-45° in the T391A protein and ψ=127° in the mature GGT).
Major structural perturbation was found around the P-segment; significant shifts were observed in residues 194-214, 253-259, 331-353, and 411-416, some of which are residues that form the active site in the mature GGT.Also, invisible residues are located near the P-segment (circled in green in Fig. 2B).Except for the P-segment and the residues near the P-segment, the structure of the T391A protein is similar to that of the mature GGT; 393 pairs of corresponding Cα atoms are superimposable with an r.m.s.deviation of 0.55 Å.
Structural rearrangements accompanied by P-segment displacement -The structures of the P-segments in T391A protein and mature GGT are compared in Fig. 3.The P-segment in the T391A protein shields the active site from solvent (Fig. 3).Displacement of the P-segment upon cleavage of the Gln-390-Thr-391 peptide bond causes the rearrangement of several adjacent segments.The most notable structural change was seen in the segment from Pro-438 to Gly-449 (Fig. 3A).These residues form one side of the substrate-binding site in mature GGT (19), whereas they are disordered in the T391A protein.In other words, when the P-segment extending to Ala-391 in the T391A protein is moved out, the disordered segment 438-449 sits on the site where the P-segment occupied.The 438-449 segment forms a lid of the substrate-binding pocket, and we denote this segment as the lid-loop.We had previously assumed that the P-segment extended near the lid-loop in the precursor protein because a continuous groove was seen on the molecular surface of mature GGT from Ser-375 to Thr-391 (19).The present result, however, clearly demonstrates that the P-segment is replaced by the lid-loop upon processing.
Displacement of the P-segment also rearranges residues 411-416 (Fig. 3B).These residues form one side of the substrate-binding site and residues 482-485 form the other side.When the P-segment is moved out from the pocket, the 411-416 segment shifts markedly (> 1 Å) toward residues 482-485, narrowing the substrate-binding pocket.
In the γ-glutamyl-enzyme intermediate and the product bound form of mature GGT, the length and width of the substrate-binding site are exactly sized to the γ-glutamyl moiety (19).Thus, the movement of residues 411-416 completes the structure of the substrate-binding pocket so that it is exactly able to recognize the γ-glutamyl moiety.
Flexibility of lid-loop -We had previously assumed that the lid-loop (residues 438-449) rigidly shields the active site from solvent based on the observations that Tyr-444 hydrogen bonds to the substrate-binding residue, Asn-411, and that no conformational change in the lid-loop was observed in the reaction intermediate or enzyme-product complexes (19).However, the analysis herein demonstrates that the lid-loop in the T391A protein is disordered, unlike that in the mature GGT and its complex with glutamic acid as well as in γ-glutamyl-enzyme intermediate.In addition, we noted that a Sm 3+ ion occupies the lid-loop site in the samarium derivative of the mature GGT (SeMet-GGT-Sm) (19), indicating that the lid-loop in this derivative is disordered or has another conformation.The refined structure of SeMet-GGT-Sm has shown that the lid-loop was disordered (see Supplemental Data).
The flexible nature of the lid-loop was directly shown by the crystallographic analysis of SeMet-GGT-P21.The electron density in the region corresponding to the loop in SeMet-GGT-P21 is compared with that in GGT-Glu complex (Fig. 4).Residues 438-449 were disordered in SeMet-GGT-P21.Except for these residues, the structure of SeMet-GGT-P21 is identical with that of GGT-Glu; the Cα atoms of the 529 residues of SeMet-GGT-P21 are superimposable on those of GGT-Glu with the r.m.s deviation of 0.71 Å.
The structure of SeMet-GGT-P21 indicates that when the substrate-binding pocket does not bind substrate or product, the lid-loop of the mature GGT is disordered.As noted previously, the F o -F c map for the mature GGT in the orthorhombic crystal (SeMet-GGT) showed a broad electron density in the substrate-binding pocket, suggesting that the pocket is partially occupied by small molecules (19).In this SeMet-GGT crystal, the neighbor molecule is situated near the lid-loop, which may also contribute to fixing its conformation.Together, the structures of the T391A protein, the γ-glutamyl-enzyme-intermediate, and the product bound form of mature GGT suggest that the substrate-binding pocket of mature GGT is open to the solvent for substrate introduction when the lid-loop is in the flexible form and that the pocket is shielded by the lid-loop in the closed form when the substrate is bound to the pocket.
Implications for the mechanism of autocatalytic processing -Site-directed mutational studies of N-terminal nucleophile (Ntn) hydrolases have indicated that processing of the precursor protein proceeds via a rearrangement of the scissile peptide bond into an intermediate ester bond (N-O acyl shift), and that the ester intermediate is subsequently hydrolyzed to form new C-terminus and N-terminus of peptides (10)(11)(12)(13)(14). Time-course studies of in vitro processing of T391C and T391S mutant E. coli GGT proteins have demonstrated that precursor processing is an intramolecular autocatalytic event and that the catalytic nucleophile is the Oγ atom of Thr-391 (17).The structure of the T391A protein, a mimic of the E. coli GGT precursor protein, may provide a structural basis for understanding the mechanism of intermediate ester bond formation.
A close-up view of the autocatalytic processing site of the T391A protein superimposed with the mature GGT model is shown in Fig. 5.The conformations of Ala-391 of the T391A protein and Thr-391 of the mature GGT are very similar.Biochemical studies have suggested that certain base is present that enhances the nucleophilicity of the Oγ atom of Thr-391 (17).The F o -F c map showed that a water molecule (W4) is located near the amide group of Gly-484 in the A molecule of the T391A protein.The distances between W4 and Ser-388 O and W4 and Gly-484 N are 3.2 Å and 3.3 Å, respectively.When the model of mature GGT is superimposed on that of T391A protein, the distance between W4 and Thr-391 Oγ is estimated to be 2.7 Å (Fig. 5).It is likely that rotation around the Cα-Cβ bond of Thr-391 and displacement of W4 occur in the precursor protein so as to optimize the hydrogen bonding geometries of W4 with Ser-388 O, Thr-391 Oγ, and Gly-484 N (see Supplemental Fig. S2).In this model, the Oγ atom of Thr-391 is situated on the carbonyl carbon atom between Gln-390 and Thr-391.These findings suggest that W4 may be the base that enhances the nucleophilicity of Thr-391 Oγ.After the attack of Thr-391 Oγ on the Gln-390 C, the carbonyl carbon likely adopts a tetrahedral arrangement.Interestingly, Gln-390 O is hydrogen-bonded to the Oγ atom of Thr-409 (Fig. 5) in the T391A protein; Thr-409 may help stabilize the orientation of Gln-390 O.The collapse of the tetrahedral arrangement of Gln-390 C shifts the linkage from an amide bond between Gln-390 and Thr-391 to an ester bond between the carbonyl group of Gln-390 and the side chain oxygen atom of Thr-391 (N-O acyl shift).The intermediate ester bond formed by N-O acyl shift is subsequently hydrolyzed, resulting in production of the L-and S-subunits.
It has been reported that a water molecule acts as a base to enhance the nucleophilicity of the catalytic residue during autocatalytic processing of the proteasome β subunit and cephalosporin acylase, members of the Ntn hydrolase family, on the basis of the crystal structures of their precursor proteins (28)(29)(30).We note, however, that the residues involved in processing and the location of the water molecule relative to the scissile peptide bond are variable in these Ntn hydrolases.
The atomic coordinates and structure factors of T391A (code 2E0W), SeMet-GGT-P21 (code 2E0X), and SeMet-GGT-Sm (code 2E0Y) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ (http://www.rcsb.org/).The segments in the T391A protein that correspond to the L-subunit and S-subunit are colored pink and green, respectively, and the P-segment (residue 375-390) is highlighted in orange.Terminal residues that generate invisible segments are labeled.The orange arrow indicates the site at which autocatalytic processing occurs.B, A stereo view of the superimposition of Cα traces of the T391A protein and mature GGT.The structure of mature GGT (A molecule of SeMet-GGT in (19)) was superimposed on that of the T391A protein (B molecule).P-segment residues in the T391A protein and in mature GGT are highlighted in orange and blue, respectively.Residues that had Cα atoms displaced by > 1 Å upon processing are colored yellow.Residues of mature GGT that are invisible in the T391A protein are shown in black.
Regions of invisible residues are circled in green.The distance between Ser-387 C and Thr-391 N in mature GGT is shown.B is rotated by 30° around the vertical axis relative to A. C. A close-up view of the segment Glu-377-Pro-380.A stick model of mature GGT (blue) is superimposed on the T391A protein (orange).The figures were prepared using PYMOL (31).

Fig. 1 .
Fig.1.A stereo view of the F o -F c omit map around the processing site (A molecule).The map was generated on the basis of F c calculated from the model, which was derived from the refinement using REFMAC5(23) omitting residues 385-392 and the water molecule (W4).The map was contoured at the 2.5 σ level.A ball-and-stick model of the T391A protein is overlaid on the map.The arrow indicates the scissile peptide bond that is cleaved in the wild-type precursor protein (Gln-390~Thr-391).The figure was prepared using PYMOL(31).

Fig. 2 .
Fig. 2. The tertiary structure of the T391A protein.A, A ribbon drawing of the T391A protein (B molecule).

Fig. 3 .
Fig.3.The structure of the P-segment.A. A surface drawing of the region around the P-segment of the T391A protein.The surface of the P-segment is omitted for clarity.Ribbon models of the P-segments (residues 375-390) of the T391A protein (orange) and mature GGT (blue) are overlaid on the surface.This figure also shows a ribbon model (blue) of residues 431-458 of mature GGT, including the lid-loop (residues 438-449).All of these residues are disordered in the T391A protein.The 21 residues following Ile-378 extend in different directions in T391A protein and mature GGT.The substrate-binding pocket is colored green.B. A close-up view of the P-segment and neighboring residues in the T391A protein shown as CPK and stick models.B is rotated by 90° relative to A around the vertical axis.Residues 411-416 and 482-485 are shown in gray, and the P-segment is in orange.Residues 411-416 and 482-485 form the sidewalls of the substrate-binding pocket in mature GGT and are involved in the recognition of the γ-glutamyl moiety.A ribbon model of these residues in mature GGT (blue) is superimposed on the T391A protein.The figures were prepared using PYMOL(31).

Fig. 4 .
Fig.4.Comparison of the electron density around the lid-loop sites of mature GGT in the two crystal forms.Shown are the F o -F c omit maps around the lid-loop in SeMet-GGT-P21 (A) and GGT-Glu (B).These maps are viewed from the outside of the molecules.Each map was generated on the basis of F c calculated from each model, which was derived from the refinement using REFMAC5 (23) omitting Thr-391 and the residues 435-452.Both maps were contoured at the 2.5 σ level.Stick model of the Thr-391, and residue 435-452 (colored green) are overlaid on the map.The model of the lid-loop (residue 438-449) was highlighted in orange.Note that the electron density corresponding to the lid-loop in A is invisible.The figure was prepared using PYMOL(31).

Fig. 5 .
Fig.5.A stereo view of the autocatalytic active site.The F o -F c omit map (contoured at the 3 σ level) was generated on the basis of F c calculated from the model, which was derived from the refinement using REFMAC5 (23) omitting W4.Stick models of the T391A protein (green) and mature GGT (blue) are overlaid on the map.The distance between the Cβ atoms in Ala-391 of the T391A protein and in Thr-391 of mature GGT is 0.54 Å. Hydrogen bonds between W4 and the T391A protein are shown as red dashed lines.The blue dotted line indicates the distance between W4 and the Oγ atom of Thr-391 in the mature GGT overlaid on the T391A protein.Numerals show the lengths of these hydrogen bonds in Å units.The figure was prepared using PYMOL(31).