Crystal Structures and Site-directed Mutagenesis of a Mycothiol-dependent Enzyme Reveal a Novel Folding and Molecular Basis for Mycothiol-mediated Maleylpyruvate Isomerization*

Mycothiol (MSH) is the major low molecular mass thiols in many Gram-positive bacteria such as Mycobacterium tuberculosis and Corynebacterium glutamicum. The physiological roles of MSH are believed to be equivalent to those of GSH in Gram-negative bacteria, but current knowledge of MSH is limited to detoxification of alkalating chemicals and protection from host cell defense/killing systems. Recently, an MSH-dependent maleylpyruvate isomerase (MDMPI) was discovered from C. glutamicum, and this isomerase represents one example of many putative MSH-dependent enzymes that take MSH as cofactor. In this report, fourteen mutants of MDMPI were generated. The wild type and mutant (H52A) MDMPIs were crystallized and their structures were solved at 1.75 and 2.05Å resolution, respectively. The crystal structures reveal that this enzyme contains a divalent metal-binding domain and a C-terminal domain possessing a novel folding pattern (αβαββα fold). The divalent metal-binding site is composed of residues His52, Glu144, and His148 and is located at the bottom of a surface pocket. Combining the structural and site-directed mutagenesis studies, it is proposed that this surface pocket including the metal ion and MSH moiety formed the putative catalytic center.

Mycothiol, also known as MSH 4 and chemically 1D-myo-inosityl-2-(N-acetyl-L-cysteinyl)-amido-2-deoxy-␣-D-glucopyranoiside (1)(2)(3)(4), is the major low molecule mass thiol in many groups of Gram-positive bacteria such as coryneform bacteria, mycobacteria, and streptomycetes (5,6). These bacteria synthesize MSH but lack GSH molecule that plays important roles in many physiological processes. It is believed that MSH functions similarly to GSH in many microbial activities (6). However, the understanding of MSH physiological function was limited to detoxification of reactive oxygen/alkalating species and to protection of pathogens such as Mycobacterium tuberculosis from host cell defense systems (1,7,8). Very recently, a novel physiological role of MSH in assimilation of aromatic compounds was described and an MSH-dependent maleylpyruvate isomerase (MDMPI) was identified in Corynebacterium glutamicum (9). This MDMPI catalyzes the conversion of maleylpyruvate (substrate) to fumarylpyruvate (product) (Fig. 1).
BLAST searches with MDMPI sequence against GenBank TM and other protein data bases revealed that MDMPI is not homologous to any functionally identified proteins but showed significant identities (27-36%) to a range of conserved hypothetical proteins from the genomes of Streptomyces coelicolor, Streptomyces avermitilis, Propionibacterium acnes, and Nocardia farcinica (10) (Fig. 2A). Earlier research on mycothiol biochemistry and biosynthesis has been conducted with mycobacteria such as M. tuberculosis, with the ultimate aim of defining new targets against tuberculosis, the resurgence of which has been a growing health concern in both developed and developing nations. The understanding of MSH-dependent enzyme structure is the key to the characterization of MSH-dependent enzymes that are helpful for the development of new drugs against diseases caused by Mycobacterium spp.
To elucidate the catalytic mechanism of MDMPI and to provide clues to find more MSH-dependent enzymes, we have determined the crystal structure of MDMPI from C. glutamicum at a resolution of 1.75Å. The structure reveals that the MDMPI contains two domains: A metal binding domain binds a divalent metal cofactor (Zn 2ϩ ) and an extended domain con-tains totally new folding. The site-directed mutagenesis reveals 9 of the 14 conserved amino acid residues are essential for activity. One mutant (H52A) was crystallized. By combination of structural and biochemical results, a putative catalytic center is proposed. So far as we know this is the first structure of an MSH-dependent enzyme, as well as an enzyme involved in gentisate/3-hydroxybenzoate assimilation that is widely existed in bacteria and participate in microbial geocycling of elements of aromatic compounds in nature.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Plasmids, and Culture Conditions-Bacterial strains and plasmids used in this study are listed in supplemental Table S1. Escherichia coli strains were grown aerobically on a rotary shaker (150 rpm) at 37°C in LB broth or on LB plate with 1.5% (w/v) agar. When needed, kanamycin at 50 g ml Ϫ1 for E. coli was used.
Expression and Purification of MDMPI and Its Mutants-Various derivatives of plasmid pET28a (supplemental Table S1) that harbor MDMPI or its mutants were constructed by methods described previously (10). Recombinant proteins were expressed in E. coli BL21 (DE3) cells by inducing with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside when the culture reached an A 600 of 0.6 -0.8. The culture was harvested after growing for additional 8 h at 25°C. Protein was purified using a nickel affinity column and a Superdex 75 gel-filtration column (GE Healthcare). The prepared protein sample was checked for purity by SDS-PAGE and by gel-filtration chromatography, which showed only one protein of 28 kDa corresponding to the molecular mass of MDMPI and its mutants. After changing buffer to 10 mM Tris (pH 8.0) by repeatedly concentration and dilution using Amicon Ultra-4 concentrators (molecular weight cutoff 5000), the purified protein was finally concentrated to 17 mg/ml and stored at Ϫ80°C.
Crystallization and Data Collection-The hanging-drop vapordiffusion method was used for crystallization. The drop was formed by mixing 1 l of protein solution (17 mg/ml) with an equal volume of reservoir solution containing 1.2 M (NH 4 ) 2 SO 4 , 12% glycerol, and 0.1 M Tris (pH 8.5). A 1.75 Å resolution native data set was collected utilizing Rigaku RAXIS IV image plate detector at Institute of Biophysics (Chinese Academy of Sciences). A bromide derivative was prepared by soaking the crystal in cryoprotecting solution containing 1.2 M (NH 4 ) 2 SO 4 , 25% glycerol, 1 M NaBr, and 0.1 M Tris (pH 8.5) for 10 min. Single-wavelength anomalous dispersion data were collected at the Beijing Synchrotron Radiation Facility at the peak wavelength of bromine (0.920 Å) with 1°o scillation and an exposure of 15 s per image. A total of 600°of data were collected to a resolution of 1.95 Å with an average of 20-fold redundancy.
The crystallization of mutant H52A was performed similarly to that of native MDMPI. The drops were composed of 1 l of protein solution (15 mg/ml) and 1 l of reservoir solution containing 1.4 M (NH 4 ) 2 SO 4 , 11% glycerol, and 0.1 M Tris (pH 8.5). A 2.05 Å resolution data set for this mutant was collected at the Institute of Biophysics (Chinese Academy of Sciences).
All the diffraction data (Table 1) were integrated using DENZO and scaled using SCALEPACK (11).
Phasing, Structural Determination, Refinement, and Analysis-Five bromide heavy atom sites in the asymmetric unit were determined using PHENIX via the single-wavelength anomalous dispersion method (12,13). After phase improvement an initial model was built using PHENIX with the final figure of merit 0.65 for eight bromide sites. ARP/wARP was used to automated model rebuilding with the sequence of MDMPI (14,15). The model was subsequently subjected to annealing refinement by CNS (16) using 50-1.75 Å data, and the traditional crystallographic refinement was carried out using CNS (16). After several cycles of refinement and model rebuilding a final model with an R factor of 0.21 and an R free factor of 0.22 was obtained, which consisted of 240 residues and 272 water molecules. We modeled Zn 2ϩ into the density and the B-factor of Zn 2ϩ has been refined to 26. Crystallographic statistics showed the quality of the overall structure was very good. The N-terminal His-tag fragment was missed in the electron density map due to its location on the molecule surface, and the C-terminal residue 241 was also missed.
The structure of mutant H52A was solved by molecular replacement with MolRep (17) using the MDMPI wild type structure as a search model. After several cycles of refinement and model rebuilding, a final model with an R-factor of 0.20 and an R free -factor of 0.22 was obtained.
A structure appraisal of the final model of MDMPI and H52A using PROCHECK (18) indicated that the main-chain torsion angles for both molecules corresponded well with allowed values (96.6% in the most favored regions of the Ramachandran plot of MDMPI and 96.2% in the most favored regions of H52A). A summary of the refinement statistics is given in Table 1.
X-ray Fluorescence Scan Analysis-To determine what kind of metal ion binds to the protein, fluorescence scan analysis of the same crystal used in data collection was performed at beamline, Photon Factory (Tsukuba, Japan). The wavelength ranges of the experiments were selected near the K-edge of Zn 2ϩ , Ni 2ϩ , Cu 2ϩ , Mn 2ϩ , Fe 3ϩ , and Ca 2ϩ , respectively.
Protein Data Bank Accession Numbers-The coordinates and structure factors for MDMPI wild type and mutant H52A have been deposited in the RCSB Protein Data Bank under the accession numbers of 2NSF and 2NSG, respectively.

RESULTS AND DISCUSSION
Overall Structure of the MDMPI-The MDMPI of C. glutamicum was purified from recombinant E. coli and its crystal structure was resolved at 1.75 Å. The MDMPI is a monomer and belongs to ␣/␤ folding type (Figs. 2B and Fig. 3). Two domains were clearly observed: An N-terminal domain largely composed of ␣-helices and a C-terminal domain characterized by shorter stretches of ␣-helices and ␤-sheets. The core of the N-terminal domain consists of five long (Ϸ20 residues) ␣-helices (␣1, ␣2, ␣3, ␣4, and ␣5). Also included in the core domain is a metal ion (Fig. 3). The C-terminal domain shows a novel fold with the pattern of ␣␤␣␤␤␣ fold, which contains three long loops, three ␣-helices (␣7, ␣8, and ␣9), and three ␤-strands (␤3, ␤4, and ␤5). It is noteworthy that 42% of the amino acid residues in the C-terminal domain are located at the three long loops, with the longest loop containing 20 residues (residues 221-241) and are not part of the core C-terminal domain structure (Fig. 3).
The C-terminal Domain Represents a Novel Folding Pattern and Is Necessary for Enzymatic Activity-The folding topology of the C-terminal domain of MDMPI is novel. Structural similarity searches of MDMPI C-terminal domain utilizing the DALI server (20) showed few matches with significant Z score and small root mean square deviation values. The best match was a putative sterol carrier protein (PDB code 1WFR) with Z value of 3.0. The 67 residues of this protein share 15% sequence identity with the C-terminal domain of MDMPI and can be superimposed to this domain with a root mean square deviation of 3.3 Å.
The C-terminal domain is necessary for the enzyme activity and structural stability. When the C-terminal domain was deleted, the truncated MDMPI did not show any enzyme activity (Table 2) and existed as inclusion bodies in E. coli cells (data not shown). Furthermore, site-directed mutagenesis revealed  (Fig. 4). These two amino acid residues (Arg 222 and Asp 151 ) are highly conserved among the protein family ( Fig. 2A). Replacement of either of them with Ala residue resulted in complete loss of enzyme activity. Thus, it is proposed that the interaction between these two residues is vital to enzyme activity and structure stability.

The MDMPI Is a Metal-dependent Protein and Contains a Metal-binding Motif, His 52 -Glu 144 -His 148 -Structural analysis revealed that
MDMPI is a metal-dependent protein with the metal binding site located in the N-terminal domain. The metal atom is coordinated by His 52 (NE2), Glu 144 (O⌭1), His 148 (NE2), two water molecules and one glycerol molecule (Fig. 5A). The x-ray fluorescence scan gave clear absorption jump at the spectrum over K-edge of Zn 2ϩ and Ni 2ϩ (data not shown), which demonstrated that this was a mixture of Zn 2ϩ and Ni 2ϩ . Coordination of Zn 2ϩ and Ni 2ϩ indicated that either could be coordinated in a tetragonal bipyramidal geometry (21). Since in the whole expression and purification course, we did not add any zinccontaining reagent, it suggested that Zn 2ϩ was integrated into the metal-ion binding site in vivo. In addition, previous enzyme activity assays showed that Zn 2ϩ (1 mM) stimulated isomerase activity up to 121%, while Ni 2ϩ (1 mM) did not affect activity (9). Therefore, it is deduced that the natural MDMPI contains Zn 2ϩ ion instead of Ni 2ϩ ion. It is further deduced that Zn 2ϩ was partially replaced by Ni 2ϩ during purification of MDMPI with a nickel-affinity chromatography.
The contribution of the three conserved residues, namely His 52 , Glu 144 , and His 148 , to the binding of metal atom was examined by individual mutation of each amino acid residue. Enzyme activity assays indicated that all mutants (H52A, FIGURE 3. Stereo view of the overall structure of MDMPI. The overall structure of MDMPI consists of two domains that are connected by two loops and one ␣-helix (␣6). In the N-terminal domain (residues 1-152), the five long ␣-helices (␣1 to ␣5), which form the core part, are shown in purple and pink, and the metal ion is shown in green, short ␣-helices (␣ ϩ 1, ␣ ϩ 2, and ␣ ϩ 3) are shown in yellow, and ␤-sheets (␤1 and ␤2) are shown in blue. The C-terminal domain (residues 163-241) includes helices ␣7 and ␣9, a three-strand mixed ␤-sheet (␤3, ␤4, and ␤5), which is connected by a shorter helix ␣8. The C-terminal domain consisted of two subdomains. The ␣-helices subdomain includes two ␣-helices (␣7 and ␣9), and the other subdomain includes three ␤-sheets (␤3, ␤4, and ␤5) and ␣8. These two subdomains are connected by three loops. A long loop (residues 221-241) is located at the terminal of the C-terminal domain. The image was created by PyMol.

FIGURE 2. Alignment of MDMPI from C. glutamicum and its homologous proteins (A) and topology of MDMPI (B).
Conserved amino acid residues are highlighted by red, including the metal-binding residues (His 52 , Glu 144 , and His 148 ) and residues for structure stability (Gly 43 , Asn 151 , Pro 162 , Gly 210 , Arg 222 , Gly 221 , Gly 232 , Pro 234 , and Pro 238 ). Amino acid residues that were mutated for generation of mutant MDMPIs in this study are indicated by asterisk. The sequence alignment was carried out with the EBI server ClustalW (23) and represented by ESPript (24). The abbreviations are as follows: Cg, C. glutamicum; Ce, Corynebacterium efficiens; WA46, Streptomyces sp. strain WA46; Nf, Nocardia farcinica; Sc, S. coelicolor; Sa, S. avermitilis.
E144A, and H148A) were inactive. Moreover, the crystal structure of mutant H52A at resolution of 2.05Å revealed that there was no metal ion bound to the protein (although nickel-affinity chromatography was used in the purification process). The metal atom position at the native MDMPI molecule was occupied by one glycerol moiety (Fig. 5B). Compared with the native MDMPI, the mutant MDMPI (H52A) has the following structural changes: 1) the main chain of the loop (amino acid residues 36 -45) moved toward the inner; 2) the side chain of Trp 44 at the loop is much closer to the former metal binding site; 3) the side chain of Glu 144 had double conformations; and 4) conformational change took place at the side chain of Arg 82 (Fig. 6).
Probably it is the absence of the metal ion that resulted in the significant conformation change around the proposed metal binding region.
Previously, it was proposed that MDMPI represents a novel family of proteins (9), but the structures of this novel family were unknown. According to sequence alignment, the metal binding site, namely residues His 52 -Glu 144 -His 148 , are highly conserved throughout the whole family ( Fig. 2A), and these binding residues are located in the surface pocket between the ␣2 (His 52 ) and ␣5 (Glu 144 and His 148 ) in N-terminal domain (Fig. 5A).
Biochemical and Structural Identification of Catalytic Center-In total fourteen amino acid residues were targeted for mutation (Table 2), according to sequence alignment and the above structure analysis. Some mutant MDMPIs such as H52A, E144A, H148A, D151A, and R222A that completely (100%) lost enzymatic activity were described above. The other mutant MDMPIs include three (H48A, C61A, E85A) that did not lose activity, two that showed significant loss of activity (S78A, 7%; R82A, 30%; activity remained), and four (Y76A, R141A, W44A, N56A) that demonstrated a complete loss of activity (Table 2). Structure modeling indicated that Tyr 76 , Arg 82 , Ser 78 , Arg 141 , Trp 44 , and Asn 56 are located at the metal-binding region. Moreover, eight of these amino acid residues, including His 52 , Trp 44 , Asn 56 , Tyr 76 , Arg 82 , Arg 141 , Glu 144 , and His 148 , formed a surface pocket (Fig. 7). The metal ion (Zn 2ϩ ) and its binding site (His 52 , Glu 144 , and His 148 ) are located at the bottom of the surface pocket. Thus, it is proposed that these highly conserved sites (Trp 44 , Asn 56 , Tyr 76 , Arg 82 , and Arg 141 ) that surrounding the metal binding site provides the hydrophilic environment for MDMPI catalysis. The surface pocket including the metal atom    Conformation differences were found at the side chains of Arg 82 and Trp 44 . In H52A the metal ion was missing and a glycerol occupied the position of the metal ion, and the location of SO 4 2Ϫ also changed. Furthermore, there is no obvious conformation changes observed, when compared with that of the wild type. The image was created by PyMol.