Structural characterization of glycinamide-RNase-transformylase T from Mycobacterium tuberculosis

ABSTRACT Enzymes from the purine salvage pathway in Mycobacterium tuberculosis (Mtb) have been regarded as an attractive target for the development of anti-bacterial drugs. Although this pathway has not been extensively studied in Mtb, it has been identified as essential for growth and survival. Glycinamide-RNase-transformylase T (PurT) is found only in some specific bacteria including Mtb and utilizes ATP-dependent ligation to catalyze the formylation of 5′-phosphoribosyl-glycinamide (GAR) in the third reaction of the de novo purine salvage pathway. In the study, we determined the crystal structure of MtbPurT at a resolution of 2.79 Å. In contrast to Pyrococcus horikoshii OT3 PurT (phBCCPPurT), MtbPurT exhibits an “open” conformation, which results in a broader ATP-binding pocket and thus might facilitate the entry and exit of the cofactor. Additionally, active site superposition with E.coli PurT (EcPurT) showed that residues involved in the ATP-binding site in MtbPurT exhibited structural similarity but had notable difference in the GAR-binding site. The loop 383-389 in MtbPurT was much shorter and shifted 5.7 Å away from the phosphate of the GAR substrate. The different GAR-binding mode might result in a large conformational change in MtbPurT, and would provide a possible opportunity for anti-TB drug development.


Introduction
Tuberculosis (TB) is considered a serious global health issue threatening humans worldwide, claiming approximately 2 million lives each year [1]. It is a chronic disease caused by a pathogenic bacterium named Mycobacterium tuberculosis (Mtb) and spreads from person to person through the air. TB generally infects the lungs but can also infect other parts of the body [2][3][4]. Current TB therapy involves a regimen of four vaccines for effective control in developed countries [5,6]. However, these drugs still have the limitations of high cost and prolonged administration period in developing countries. In particular, the prevalence of multidrug resistance continues to increase at an alarming rate, resulting in morbidity and mortality [5]. Therefore, exploring new drugs and vaccines against latent bacteria would provide a better insight into the resistance of Mtb strains and achieve effective control of TB in a short period, particularly in developing countries.
Nucleoside pathways can be a good source of metabolic energy and are considered a prospective target for new potential drug leads. The enzymes involved in this pathway usually have distinct characteristics from those of human counterparts. One of the attractive pathways for the development of nucleoside analogs with anti-TB activity is the purine salvage pathway [7,8]. Purine salvage enzymes are useful in the treatment of TB infection because of their capacity to permit the metabolism of nucleoside analogs to active compounds [9][10][11]. To date, several homologues to enzymes involved in the purine salvage pathway have been identified based on the genome sequence of Mtb H37Rv [12]. However, little is known about purine metabolism in Mtb; thus, we need to investigate the enzymes involved in the salvage of purine nucleosides. A comprehensive understanding of those enzymes would provide insight into the identification of nucleoside analogs and affect the Mtb strain.

Protein expression and purification
The open reading frame sequence encoding PurT (Rv03894) was amplified from the genomic DNA of Mtb H37Rv by polymerase chain reaction (PCR). The PCR amplified minigene (Gene ID: 886032, bases 1-1260 nt) was purified, digested with EcoRI and XhoI and cloned into the same restriction sites of pProExHta (Novagen, Madison, Wisconsin, USA). Clones were selected by PCR and restriction mapping. The plasmid was transformed into E.coli strain BL21 (DE3) for protein expression. The E.coli BL21 (DE3) transformant was cultured in 1000 ml of fresh LB medium (10 g of Bacto tryptone, 5 g of yeast extract and 10 g of NaCl per litre of solution) containing ampicillin (100 μg/ml) at 37°C. When OD 600 ∼0.8, protein expression was induced by 0.0625 mM isopropyl-β-dthiogalactoside (IPTG) at 37°C for 4 h. The cells were harvested by centrifugation at 6,520 g for 1 h at 4°C, washed with phosphate buffered saline (PBS) twice and then stored at −80°C.
The harvested cells were resuspended in lysis buffer (50 mM Tris-HCl, pH 7.5, 250 mM NaCl, 5 mM imidazole) and homogenized by sonication. The suspended lysate was centrifuged at 15,930 g for 1 h to remove cell debris. The clear supernatant was filtered (pore diameter 0.45 µm; Sartorius, Goettingen, Germany) and applied to a column of nickel-NTA beads (Qiagen, Hilden, Germany) pre-equilibrated with lysis buffer. The column was loaded with 70 column volumes of slurry and then washed with 15 column volumes of washing buffer (50 mM Tris-HCl pH 7.5, 250 mM NaCl, 30 mM imidazole), then eluted with 5 column volumes of elution buffer (50 mM Tris-HCl pH 7.5, 10 mM NaCl, 300 mM imidazole). Fractions containing MtbPurT were pooled and concentrated in 50 mM Tris, pH 8.0 by ultrafiltration using a Centriprep YM-10 (Millipore Corporation, Bedford, MA, USA). The MtbPurT was further purified by ion-exchange chromatography on an HQ20 column (Perseptive Biosystems, Foster City, CA, USA) preequilibrated with 50 mM Tris buffer, pH 8.0. The protein was eluted at ∼0.35 M NaCl and concentrated by ultrafiltration (Centriprep YM-10, Millipore Corporation, Bedford, MA, USA). MtbPurT was finally purified by gel-filtration chromatography with a Superdex 200 10/300 GL column (GE Healthcare, Piscataway, NJ, USA) in gel-filtration buffer (50 mM Tris-HCl pH 8.0,100 mM NaCl). The fractions containing MtbPurT protein were collected, exchanged into HEPES buffer (10 mM HEPES, pH 7.5) by ultrafiltration (Centriprep YM-10), and subsequently snap-frozen in liquid nitrogen at a −80°C.
Pyrococcus horikoshii OT3 phosphoribosylglycinamide formyl transferase (phBCCPPurT, Gene ID: 1444201, bases 1-1293 nt) was custom synthesized and cloned into the pET28a vector (Novagen). The pET28a-phBCCPPurT construct was further confirmed by sequencing and then transformed into E.coli strain BL21 (DE3) for protein expression and purification as above.

Crystallization and data collection
Purified MtbPurT was concentrated to 8 mg/ml in 10 mM HEPES pH 7.5, and then screened for crystallization. Initial crystallization was carried out with Index Screen, PEGRx Screen and Crystal Screens I Figure 1. Scheme of the reactions catalyzed by two GAR transformylases, PurT and PurN. In the third step of the de novo purine nucleotide biosynthesis pathway, PurT uses formate and ATP to catalyze GAR to N-formyl-GAR (FGAR), while PurN uses 10formyltetrahydrofolate in this reaction. and II from Hampon Research (California, USA), Wizard Screens I, II, Cryo I and Cryo II from Emerald Biostrucures (Bainbridage Island, Washington, USA), JBScreen Basic HTS I, II from Jena Bioscience (Jena, Germany), Classics Suite I, II, JCSG Core I, II, III, IV, PHClear Suite I and II from Qiagen (Hilden, Germany) using the microbatch crystallization method at 18°C. After optimal crystallization, rod-shaped crystals were successfully obtained by the hanging drop vapour diffusion method. The crystallization experiments were conducted at 18°C, and the reservoir solution contained 0.06 M sodium acetate, pH 4.6, 6.0% PEG4000, 32% glycerol. For data collection the crystals were equilibrated in the reservoir solution adding 10% ethylene glycol to the well solution, flash cooled and stored in liquid nitrogen. Diffraction data were collected at 100 K by a Dectris Eiger X 16M detector at Shanghai Synchrotron Radiation Facility (SSRF) beamline BL17U1. All diffraction images were indexed, integrated and scaled using the programme XDS package [21].

Structural determination and refinement
Structure of MtbPurT was determined by the molecular replacement methods with the programme PHASER in CCP4i [22], and structure with PDB code 1EYZ (42% identity) was used as a search model. An initial model of MtbPurT was manually built using Coot (Crystallographic Object-Oriented Toolkit) [23], and structural refinement was performed using the PHENIX programme [24]. The quality of the model was validated periodically during the model building/refinement process with the MolProbity programme [25]. The structural figures were prepared by the PyMOL [26]. All structural refinement statistics are listed in Table 1.

Isothermal titration calorimetry assay
The isothermal titration calorimetry (ITC) assay was performed on a NANO ITC 2G at 37°C. Briefly, the proteins (phBCCPPurT, MtbPurT and its mutations) were dissolved in PBS buffer and injected into the sample cell at the concentration of 100 μM, and then titrated with aliquots of 2 mM ATPγS or CTP solution. Data analysis were performed by NanoAnalyze software package (TA Instruments, New Castle, USA) using Independent Model.

Site-directed mutagenesis
Primers for the MtbPurT mutation (Table 2) were designed and commercially synthesized. MtbPurT mutants were amplified by site-directed mutagenesis (KOD plus) according to the manufacturer's protocol. The sequences of the mutants were confirmed through DNA sequencing. The expression and purification of mutants were performed as described above for native MtbPurT.

Sequence and structure alignment
Sequence alignment was performed by MUSCLE software [27,28], and was illustrated and generated through a website ESPript 3.0 [29]. The structural alignment was performed by DALI Server [30]. The MtbPurT/ATP/GAR model was generated based on

Structure determination of MtbPurT
The MtbPurT crystal diffracted at 2.79 Å resolution at the SSRF and belonged to a space group P3 1 21. Its 3D-structure was determined by molecular replacement and refined to a R work /R free of 0.217/0.242. The structure of MtbPurT agrees well with the crystallographic data and expected geometric values. In the Ramachandran plot, 90.0% of the residues were in the most favourable region, and all of the residues were in the allowed region. There were two MtbPurT molecules in the asymmetric unit. Data collection and refinement statistics are summarized in Table 1. Additional attempts to crystallize complex MtbPurT/ATP or MtbPurT/ADP were also made, unfortunately, no crystals grew.

Overall architecture of MtbPurT
The molecular weight of the MtbPurT protein was estimated to be ∼90 kDa from gel-filtration chromatography, and the monomer protein was about 44 kDa from SDS-PAGE (Figure 2(A)). Therefore, the MtbPurT protein exists as a dimer in solution. The final structure of MtbPurT is well ordered, and electron densities for all residues were clearly interpretable, except in four parts (residues 111-113, 175-180, 189-198 and 205-215). The overall architecture of MtbPurT forms a tight dimer in the asymmetric unit and the two monomers show a structure similarity to each other (Figure 2(B)). According to the MtbPurT structure, the monomer of MtbPurT consists of a small lid domain and a large α/β domain (Figure 2 (C)). The small lid domain comprises one α helix (α7) and three β strands (β6, β7 and β8). The large domain mainly contains two α/β domains, α (1-6)/β

Comparison with apo-phBCCPPurT
To date, the structures of the apo form of Pyrococcus horikoshii OT3 (phBCCP) and its complex with ADP and the structures of E.coli PurT (EcPurT) complexed with ADP, ATP, ATP/GAR, AMPPCP and ATPγS complex have been successfully solved [19,32]. Figure 3 (A) shows the structural superposition of the subunits of apo-MtbPurT and apo-phBCCPPurT (PDB code: 2CZG, r.m.s.d: 1.30 Å). In contrast to phBCCPPurT, loop 252-255 between β11 and β12 in MtbPurT shifted by 13.1 Å away from the cofactor binding site. This structural rearrangement in apo-MtbPurT represents an "open" conformation. However, the apo-phBCCPPurT exhibits a "close" conformation. As a result, the pocket of MtbPurT for the cofactor binding has a much broader opening with a width of 11.6 Å (distance between Ser182 and Asp251) than 7.3 Å (distance between Ser172 and Asp239) for phBCCPPurT (Figure 3(B)). As PurT may cause ATP hydrolysis (Figure 1), we utilized a non-hydrolysable ATP analog (ATPγS) to determine the ATP-binding affinity for MtbPurT in this study. The isothermal titration calorimetry (ITC) assay was performed. As shown in Figure 3(C), phBCCPPurT binds ATPγS with an apparent dissociation constant (K d ) of 41.83 ± 7.21 μM. And MtbPurT recognizes ATPγS with a relatively low binding affinity, with the K d value of 207.41 ± 8.02 μM. No binding affinity with CTP was detected for MtbPurT (Figure 3(D)).

Structural similarity with the ATP-grasp fold family
We compared the crystal structure of MtbPurT with other known three-dimensional protein structures in the Protein Data Bank using DALI Server [33].  Figure  4(C)) reveals that the ATP-binding motif of these proteins aligns well at the C-terminal and belongs to the ATP-grasp superfamily. The superimposition of these structures with MtbPurT revealed that the amino acid residues for ATP binding share a surprising similarity. The corresponding residues contributing to this motif in phBCCPPurT-ADP are Arg121, Lys162, Glu202, Glu203, Ile205, Glu210 and Glu291 (Figure 4 (A)); those in FtPurK-AMPPNP are Lys138, Asp145, Gly146, Glu175, Val178, Glu183, Glu247 and Glu259 (Figure 4(B)); and those in LlPC-ADP are Lys116, Lys157, Gly163, Gly164, Glu199, Lys200, Ile202 and Gln231 (Figure 4(C)). According to the sequence alignment, most residues are conserved for the binding with ATP-analogs (Figure 4(D)).

A model of MtbPurT in complex with the cofactor ATP and substrate GAR
It has been previously shown that the active site of PurT could be divided into two subsites, referred to the ATP-binding site and the GAR-binding site [38]. The ATP-binding site is positioned in the C terminus, while the GAR-binding site is located in the N   terminus. To analyze the active binding pocket of MtbPurT, we generated a model of MtbPurT/ATP/ GAR based on the superposition of the EcPurT/ATP/ GAR (PDB code: 1KJ8, r.m.s.d: 1.01 Å, Figure 5(A)). The model shows that the ATP molecule is primarily located on the pocket surrounded by the A-and Cdomains. The amino acid residues surrounding the ATP-binding site share a remarkable degree of structural similarity to those of EcPurT ( Figure 5(B)). In the MtbPurT/ATP/GAR model, the backbone carbonyl oxygen of Ser218 and the peptidic NH group of Val220 could contact the adenine ring of ATP, and these two amino acids are highly conserved in members of the ATP-grasp fold family. In addition, residue Glu217 also interacts with the adenine ring of ATP, which is consistent with Glu195 in EcPurT. Residue Arg131 could generate a hydrogen bond to the phosphate group of ATP. Most of the residues for the ATP-binding in MtbPurT are highly conserved as the EcPurT structure. To validate our structural findings, we constructed four mutants of MtbPurT, each carrying an amino acid substitution in the ATP-binding pocket.
The binding affinity was also tested with ATPγS by ITC. The results were shown in Figure 5(C). All four mutants (R131A, E217A, S218A and V220A) lost their ATP-binding capability, indicating essential roles for these residues in binding with ATP. However, a long loop 175-180 between β7 and β8 involved in the interaction with the phosphate group of ATP is disorder, which needs to be investigated further.
In EcPurT, the side chain of Glu82 forms hydrogen bonds with the hydroxyl groups of the GAR ribose, and the carboxylate side chain of Asp286 interacts with the amino group of GAR. Based on the sequence alignment, their corresponding residues Val96 and Gly312 in MtbPurT have short side chains that no longer form hydrogen bonds with the GAR ribose. In EcPurT, the side chains of residues Arg362 and Lys355 interact with the phosphoryl group of the GAR, and the guanidinium group of Arg363 contacts the carbonyl oxygen of GAR. However, the corresponding residues Arg386 and Gly387 in MtbPurT show no hydrogen bonds with GAR. With their active sites superposed, the residues Arg362, Arg363 and Lys355 in EcPurT were positioned in a long loop with 13 amino acids (residue 352-364), while Arg386 and Gly387 in MtbPurT were located in a short loop with 7 amino acids (residue 383-389) ( Figure 5(D)). A distance of 5.7 Å is observed between these two loops and the short loop in MtbPurT is shifted away from the phosphoryl group of the GAR. Therefore, the structural differences suggested that the GAR binding would result in a large conformational change in MtbPurT, and the catalytic mechanism of GAR should be different from that of EcPurT.

Discussion
Structural comparisons of MtbPurT using DALI Server indicated that it was structurally similar to the ATPgrasp fold proteins with their characteristic three domains, and thus MtbPurT is categorized as a member of the ATP-grasp superfamily as in the previous studies [13,39]. Each MtbPurT monomer contains three structures referred to as the A-, B-and C-domains. Structural analysis of MtbPurT indicated that the mycobacterial enzyme closely resembles the other bacterial PurT enzymes in terms of both overall fold and active site structure [19,32]. One of the most interesting results from a close inspection of the superimposed structure of apo-MtbPurT over apo-phBCCPPurT structures is the difference in one loop around the phosphate groups of ATP. The loop 252-255 in apo-MtbPurT adopts a position that defines an "open" conformation of the active site in the absence of ATP, in contrast with the "closed" conformation in phBCCPPurT. This rearrangement of residues and formation of the "open" conformation presents a broader binding pocket in the active site, although MtbPurT exhibited similar ATP-binding affinity compared with phBCCPPurT. This noticeable difference in the ATP-binding pocket between Mtb and phBCCP PurTs was reflected in the biochemical structure-activity relationship studies and would provide crucial information for the design of more specific inhibitors.
Since PurT was found to catalyze the formylation of GAR using a catalytic mechanism requiring ATP and formate, it was of interest to understand how ATP and GAR bind to the protein. Therefore, we modelled both ATP and GAR into the MtbPurT structure to seek an explanation for their binding mode. Model of MtbPurT/ATP/GAR complex was generated based on the superimposition of EcPurT/ATP/GAR. In the MtbPurT/ATP/GAR model, the putative ATP-binding pocket in MtbPurT was almost identical to other PurTs and the residues forming the ATP-binding pocket were well conserved among the PurTs. These residues such as Ser218, Val200, Glu217 and Arg131, form a hydrophobic pocket and play an important role in interacting with the ATP. In addition, the model of MtbPurT/ATP/GAR shows that GAR is located on the pocket surrounded by the A-and Cdomains. With the GAR-binding sites superposed, the residues of MtbPurT surrounding the GAR-binding pocket showed a distinct difference from those of EcPurT. In particular, loop 383-389 in MtbPurT is much shorter than that in EcPurT (loop 352-364) and shifted 5.7 Å away from the phosphate of the GAR substrate; thus, no hydrogen bonds with GAR were found. In fact, we performed docking calculations using AUTODOCK 4.2 [40] and tried to model the GAR into MtbPurT. However, no output files were produced. Therefore, subtle differences in the loops that mediate substrate binding would markedly alter the MtbPurT conformation and change the GAR binding site shape and contacts. The loop closure is a critical requirement for the binding of the GAR substrate and the subsequent catalysis.
In conclusion, the molecular structure of MtbPurT has now been resolved at 2.79 Å resolution. PurT is found only in some eubacteria, including Mtb. Therefore, compounds that target MtbPurT are attractive drug development options. This wealth of information might play an important knowledge base for the design of potential anti-TB agents in further. The inhibitors of MtbPurT could contribute a possible opportunity and serve as useful lead compounds for anti-TB drug development.