Rational inhibitor design for Pseudomonas aeruginosa salicylate adenylation enzyme PchD

Pseudomonas aeruginosa is an increasingly antibiotic-resistant pathogen that causes severe lung infections, burn wound infections, and diabetic foot infections. P. aeruginosa produces the siderophore pyochelin through the use of a non-ribosomal peptide synthetase (NRPS) biosynthetic pathway. Targeting members of siderophore NRPS proteins is one avenue currently under investigation for the development of new antibiotics against antibiotic-resistant organisms. Here, the crystal structure of the pyochelin adenylation domain PchD is reported. The structure was solved to 2.11 Å when co-crystallized with the adenylation inhibitor 5′-O-(N-salicylsulfamoyl)adenosine (salicyl-AMS) and to 1.69 Å with a modified version of salicyl-AMS designed to target an active site cysteine (4-cyano-salicyl-AMS). In the structures, PchD adopts the adenylation conformation, similar to that reported for AB3403 from Acinetobacter baumannii. Graphical abstract


Introduction
Pathogenic organisms compete with their biological hosts for metal nutrients, particularly iron. The host limits free metals in an effort to starve bacteria of needed metals, a phenomenon termed nutritional immunity [1]. To circumvent nutritional immunity, bacteria have developed metal acquisition systems that include synthesis and secretion of metal chelators that scavenge metal from the host and are then imported back into the bacteria. Pyochelin is a metallophore produced by the antibiotic-resistant pathogen Pseudomonas aeruginosa, and while capable of binding a variety of metals, pyochelin's highest affinity is for Fe 3+ [2,3]. Highlighting the nutritional importance of pyochelinmediated iron uptake are early publications demonstrating that pyochelin stimulates bacterial growth in murine infections [4] and that pyochelin is needed for P. aeruginosa virulence [5].
Like many natural products, the metallophore pyochelin is synthesized by a non-ribosomal peptide synthase (NRPS). NRPS modules are multidomain proteins (adenylation domain, peptidyl carrier domain, condensation domain) that construct small peptides in an assembly-line fashion. The adenylation domain activates each new amino acid prior to its integration into the growing peptide chain by catalyzing the formation of an aminoacyl-AMP. Following activation, the peptidyl carrier domain transfers the activated product to the condensation domain where it is joined to the growing peptide chain. This process is repeated with additional NRPS modules until the full molecule is assembled. Pyochelin is assembled from three precursor molecules: one salicylate and two cysteines. The biosynthesis is achieved by three NRPS modules, two accessory enzymes, and one stand-alone tailoring enzyme [6]. PchD is a stand-alone adenylation enzyme and is part of the initiation module of pyochelin biosynthesis.
The role of PchD is twofold. In the adenylation reaction, salicyl-AMP is formed from salicylate and ATP, releasing pyrophosphate (Fig. 1A). Next, the thiol of the phosphopantetheinyl (Ppant) prosthetic group of the N-terminal peptidyl carrier protein domain of PchE performs a nucleophilic attack on the carboxyl ester of salicyl-AMP, releasing AMP and transferring the salicyl moiety to the Ppant tether, and priming the first module in the biosynthetic pathway. A homolog to PchD is also found in the major pathogen Mycobacterium tuberculosis (MbtA) which acts as a similar salicylate-AMP ligase in the biosynthetic pathway for the siderophore mycobactin [7]. Both PchD and MbtA are part of the family of adenylating enzymes that, in addition to including adenylation domains from NRPS modules, also encompasses firefly luciferase and acyl and aryl-CoA synthetases and is, therefore, termed the ANL superfamily (Acyl-CoA synthetases, NRPS adenylation domains, and Luciferase enzymes) [8]. Previously solved structures have captured other NRPS adenylation domains in either the adenylation conformation or in the thioester-forming conformation [9]. These structures serve as a template for identifying the catalytic relevance of the structures described here and of other adenylation domain structures.
P. aeruginosa and M. tuberculosis have been listed by the Centers for Disease Control and Prevention as serious threats due to their increasing antibiotic resistance [10], which led to the hypothesis that NRPS adenylation domains such as PchD and MbtA are attractive targets for the development of new antibiotics [11]. Noting that mechanistically related adenylation domains (i.e. aminoacyl tRNA-synthetases) were inhibited by analogues of their acyl-AMP intermediates, Finking et al. first described the use of sulfonyladenosine (AMS) inhibitors for NRPS adenylation domains [12]. Ferreras et al. subsequently pioneered the development of inhibitors against the NRPS adenylation domains in M. tuberculosis (MbtA), Yersinia pestis (YbtE; Y. pestis is the causative agent of plague), and Pseudomonas aeruginosa (PchD) [13]. They synthesized and tested the inhibitor 5′-O-(N-salicylsulfamoyl)adenosine (salicyl-AMS) (Fig. 1B) and demonstrated nanomolar K i values for MbtA, YbtE and PchD and micromolar IC 50 values for growth inhibition of M. tuberculosis and Y. pestis. Since that time, salicyl-AMS has been the focus of multiple structure-activity studies [14][15][16][17][18][19][20][21][22][23]. The work presented here describes a modification to salicyl-AMS that exploits the presence of an active site cysteine residue near the salicyl ring of the inhibitor. We hypothesized that the extension of an electrophile off the salicyl ring would provide the opportunity for covalent binding between the cysteine in the active site and the newly introduced electrophile. Using a method based on the one developed by Ferreras [13], a nitrile group was introduced on C4 of the aryl group to produce 4-cyano-salicyl-AMS (Fig. 1B). The apo-structure of the adenylation domain MbtA from Mycobacterium smegmatis, which shares 69.2% sequence identity with M. tuberculosis MbtA, has been determined previously [24]. The structures presented here represent the P. aeruginosa adenylation domain PchD with either salicyl-AMS or 4-cyano-salicyl-AMS. In contrast to the MbtA structure which was solved in an apparent non-catalytic conformation, the PchD structures presented here demonstrate the catalytically relevant adenylation conformation as well as important active site contacts with the ligands.

Cloning, expression, and purification of pchD
The pchD gene was amplified from Pseudomonas aeruginosa PAO1 genomic DNA by polymerase chain reaction by use of Herculase polymerase (Agilent) supplemented with 6-10% DMSO and 3% glycerol. The forward primer (5′-TAT ATT CAT ATG ACT TCC TCG CCC GTC ACC-3′) includes an NdeI site (underlined), whereas the reverse primer (5′-TTA TAT GGA TCC TCA TGC GCG GGC CTC CAG-3′) contains a BamHI site (underlined). The amplified 1665 basepair fragment was digested with NdeI and BamHI and ligated into the pET28b plasmid (Novagen) digested with the same enzymes. The resulting plasmid encodes the gene for the PchD protein with an N-terminal histidine tag. BL21(DE3) pLysS E. coli (Invitrogen) containing the PchD expression plasmid were grown in LB broth containing 50 μg/mL kanamycin at 37 °C with shaking (225 rpm). When OD 600 was ~0.7, the temperature was reduced to 30 °C and protein expression was induced with the addition of isopropyl β-d-thiogalactopyranoside to a final concentration of 200 μM. The cells were harvested by centrifugation (6000×g, 10 min, 4 °C) after ~4 h. The cell pellet was resuspended in 5 mL of 25 mM Tris-HCl pH 8, 500 mM NaCl per 2 L of culture. Cells were disrupted by use of a French pressure cell (35,000 psi), and cellular debris was removed by centrifugation (12,000×g, 1 h, 4 °C). The supernatant was applied to a chelating Sepharose fast-flow column (Amersham Biosciences) charged with nickel chloride and pre-equilibrated in 25 mM Tris-HCl pH 8, 500 mM NaCl, 50 mM imidazole (buffer A). PchD protein eluted at 100 mM imidazole in a linear gradient of 50-300 mM in buffer A. The pooled fractions were applied to a Superdex 200 sizeexclusion column (Amersham Biosciences) equilibrated with 25 mM Tris-HCl pH 8, 150 mM NaCl, 10% glycerol, 2 mM DTT. The fractions containing PchD were pooled and concentrated by use of an Amicon stirred cell with a YM-10 membrane to 17 mg/mL as determined by the Bradford assay and stored at −80 °C.

Synthesis of salicyl-AMS and 4-cyano-salicyl-AMS
All chemical reagents were purchased from commercial suppliers and used without further purification. Flash column chromatography was performed on silica gel (40-63 mm) from Sorbent Technologies. Separation was performed with a Teledyne Isco CombiFlash Rf.
5′-O-(N-salicylsulfamoyl)adenosine (salicyl-AMS). The synthesis of salicyl-AMS was performed as previously described in Ferreras et al. [13]. The spectroscopic data of the purified product agreed with literature values.

Data collection and processing
PchD salicyl-AMS co-crystal diffraction data (1° oscillation images for a total of 270°) were collected at the Stanford Synchrotron Radiation Laboratory (Stanford, CA) beamline 7-1 with a wavelength of 1.1271 Å at 100 K. The exposure time per frame was 60 s with an attenuation of 0% and a crystal to detector distance of 200 mm. The data were indexed and scaled with XDS to 2.11 Å [25]. The crystals were assigned to the space group C2 with unit cell

Structure solution and refinement
Molecular replacement calculations were performed using Phaser [26] in the PHENIX [27] program suite. The model used for this structure was dihydroxybenzoate adenylation domain (DhbE) from Bacillus subtilis (PDB 1MDB) with waters and ligand removed [28]. For the salicyl-AMS structure molecular replacement calculations yielded a clear solution with a log likelihood gain of 624.78 and a TFZ score of 16.7. Model building and refinement were performed using Coot [29] and Phenix refine [30] and waters were placed by Phenix refine, corrected manually and verified using a 2mF o -F c electron-density map contoured at 1.5σ following a round of refinement. The salicyl-AMS was generated in eLBOW [31] and placed using LigandFit [32,33]. The final model includes one molecule with residues 14-201 and 206-540, one salicyl-AMS molecule, and 139 water molecules. Geometry analysis was performed by MolProbity [34]. The salicyl-AMS structure was used as the molecular replacement model for the 4-cyano-salicyl-AMS structure  [29] and Phenix refine [30]. The 4-cyanosalicyl-AMS was generated in eLBOW [32] and placed using LigandFit [32,33]. Waters were placed by Phenix refine and validated manually with each round of refinement. The final model includes one molecule with residues 13-542, one 4-cyanosalicyl-AMS molecule, and 292 water molecules. A comparison of structures and calculation of RMSD values were performed using PDBeFold [35]. Structure figures were generated in PyMOL (PyMOL Molecular Graphics System, version 2.0, Schrödinger, LLC).

Results and discussion
Previous attempts to crystalize the N-terminally his-tagged PchD in the absence of a ligand were unsuccessful [36] but crystallization was easily achieved when either salicyl-AMS or 4-cyano-salicyl-AMS were present. Additional confirmation that the correct ligands were produced and co-crystalized was demonstrated retrospectively by the high-resolution structures, despite sub-optimal purity for the 4-cyano-salicyl-AMS synthesis. The co-crystal of PchD and salicyl-AMS diffracted to 2.11 Å resolution and X-ray diffraction data were collected on Stanford Synchrotron Radiation Lightsource beamline 7-1 ( Fig. 2A; Table 1). The co-crystal of PchD and 4-cyano-salicyl-AMS diffracted to 1.69 Å resolution and X-ray diffraction data were collected on Stanford Synchrotron Radiation Lightsource beamline 12-2 ( Fig. 2B; Table 1). Molecular replacement was performed using the dihydroxybenzoate adenylation domain (DhbE) from Bacillus subtilis (PDB 1MDB). PchD is comprised of a large N-terminal subdomain (A core ) consisting of residues 1-439 and a smaller C-terminal subdomain (A sub ) consisting of residues 441-542 with residue K440 designating the hinge location that allows a range of motion between the two domains (Fig. 2). The A core subdomain is made of two αβαβα regions that form a sandwich and a β-barrel. The A sub of PchD is comprised of a central three stranded β-sheets surrounded by three helices. Regardless of the ligand bound, the two structures are in identical conformation with an rmsd of 0.15 Å over 523 Cα. Drake et al. crystallized two complete NRPS modules, each trapped in a different stage of the catalytic cycle [9]. The adenylation domains of these two modules provide useful comparisons for determining the catalytic conformation of PchD. The NRPS module from Acinetobacter baumannii, implicated in biofilm formation, has been termed AB3403.
AB3403 was co-crystalized with Mg·ATP and glycine and demonstrates the adenylation conformation shown in Fig. 3A (PDB 4ZXI) [9]. The EntF NRPS module from Escherichia coli for the production of the siderophore enterobactin was co-crystalized with a mechanism-based inhibitor to capture the thioester-forming conformation and is shown in Fig. 3B (PDB 5T3D) [9].
Visualization of these two states provided confirmation that the motion of the A sub adenylation domain is important for catalysis [8,9]. PchD overlays very closely with the adenylate-forming conformation of the AB3403 adenylation domain with an rmsd of 2.03 Å over 438 Cα (Fig. 4A). However, when the same comparison is made to the thioesterforming conformation of the EntF adenylation domain, the rmsd is similar (2.10 Å) but for only 364 Cα (Fig. 4B). The large difference in α-carbons for the comparison is due to the rotation in the A sub between the adenylation-forming and thioester-forming conformations. Separate alignments of the A core and A sub provide a more complete picture. Comparing these two regions for PchD and AB3403, the A core has an rmsd of 2.10 Å over 344 Cα while the A sub has an rmsd of 1.49 Å over 93 Cα. When the same regions are aligned between PchD and EntF the A core has an rmsd of 2.05 Å over 356 Cα and the A sub has an rmsd of 1.86 Å over 89 Cα. The presence of either the salicyl-AMS ligand or the 4-cyano-salicyl-AMS ligand was an absolute requirement for crystallization of PchD, likely because it locks down the highly mobile A sub subdomain into the adenylate-forming conformation.
The structure of a homologous free-standing adenylation enzyme MbtA from M. smegmatis was published by Vergnolle et al. [24]. Figure 4C shows an overlay of MbtA and PchD demonstrating remarkable similarity of the two enzymes except in the conformation of the A sub (rmsd 1.07 Å for 409 Cα). Because both PchD and MbtA use salicylate as a substrate, it is unsurprising that they are so similar. This similarity is further evident by separate comparisons of the A core and A sub regions because this eliminates the differences in orientation of the A sub . For the A core there is an rmsd of 1.05 Å for 406 Cα and for the A sub the rmsd is 1.30 Å for 96 Cα. The conformation of the MbtA A sub does not conform to either the adenylation conformation (Fig. 4C) or the thioester conformation (Fig. 4D). Because the MbtA structure was crystallized without a ligand, it is possible that the orientation of the A sub maximized crystal contacts making the conformation an artefact of crystallography.
Analysis of the salicyl-AMS-bound PchD active site revealed that Cys250 is located near to the salicylate ring of salicyl-AMS (Fig. 5). Therefore, a salicyl-AMS inhibitor derivative was designed and synthesized with a cyano group at C4 that could extend towards Cys250 with the expectation of covalent bond formation by the electrophilic warhead would convert the potent inhibitor into a targeted covalent inhibitor [37]. As anticipated, the nitrile group of 4-cyano-salicyl-AMS and Cys250 were in close proximity (3.8 Å). The presence and location of the cyano group were verified by calculating a polder map (Fig. 5B). Surprisingly, while a slight change in the orientation of the Cys250 side chain might allow for bond formation, the covalent bond was not observed in the crystal structure. The absence of demonstrates the thioesterforming conformation and was co-crystalized with the ligand serine adenosine vinylsulfonamide (magenta). The A sub of MbtA (wheat; A core : orange) does not align with either the C adenylation or the D thioester conformation bond formation may indicate the need for a change in the orientation of the cyano group relative to Cys250. Alternatively, the active site may alter the pKa of the cysteine such that deprotonation for bond formation is not favored. An increase in the pH of the crystallization conditions may facilitate bond formation.
DhbE is a free-standing adenylation domain in the biosynthetic pathway for the bacillibactin siderophore in Bacillus subtilis. The substrate for DhbE is 2,3-dihydroxybenzoic acid (DHB) which differs from salicylate only in the presence of a second hydroxyl group at the C3 position. The structure of DhbE bound to its adenylated product DHB-AMP was solved by May et al. and served as the molecular replacement search model for both the PchD structures reported here and for the MbtA structure [28]. The amino acid composition of the active site for DHB adenylation domains and salicylate adenylation domains is remarkably similar. In trying to identify which active site residues confer substrate specificity, May et al. hypothesized that two positions distinguish between DHB and salicylate. They noted that the Cys250 position (described above) was conserved in salicylate adenylation domains but that in DHB adenylation domains this same position is a serine which is needed to make room for the additional hydroxyl group on DHB. The second position is Ile347 in PchD (Leu250 in MbtA). In  shown in yellow overlayed with both salicyl-AMS PchD (salmon inhibitor, midnight blue structure) and 4-cyano-salicyl-AMS bound PchD (orange inhibitor, skyblue structure). Residues conferring specificity for salicylate are Cys250 and Ile347. For the DhbE structure these positions are a conserved serine and a valine which are hypothesized to more readily accommodate the second hydroxyl of DHB   Fig. 7 Contacts between active site residues and the inhibitor DhbE this position is a valine, again, to accommodate the second hydroxyl functional group (Fig. 6).
The finding that PchD would not crystallize except in the presence of an inhibitor compound indicates that important contacts are formed between the compound and active site residues and serve to lock the enzyme in the adenylating conformation (Fig. 7). Lysine 529 of the A sub is within 3 Å of both the sulfamate oxygen and the sugar oxygen and likely helps coordinate this region of the compound. Phenylalanine 247 likely forms a hydrophobic region for the salicylate ring, but is too distant (4.3-4.9 Å) to form pi-pi stacking interactions. The P-loop region, the phosphatebinding motif in ATP-binding enzymes, encompasses residues 200-210 (SGGT TGT PKLI) [28]. Though important for binding the native substrate (ATP), the P-loop appears to have little involvement in coordinating the 4-cyano-salicyl-AMS inhibitor. The P-loop remained quite mobile based on elevated B-factors in this region and broken density in the map. In the salicyl-AMS structure, the P-loop is even more disordered with insufficient density to fully build the loop.

Conclusions
Using the potent salicyl-AMS inhibitor originally designed by Ferreras et al. [13], the structure of the stand-alone salicylate adenylation domain from Pseudomonas aeruginosa was determined to 2.11 Å. This is the first structure of an adenylation domain bound to the salicyl-AMS inhibitor and serves as an important visual model for making future modifications to continue to develop salicyl-AMS as a therapeutic drug. Indeed, the initial structure led to the design and synthesis of a proposed targeted covalent inhibitor, 4-cyanosalicyl-AMS, and a subsequent structure was determined to 1.69 Å. However, the novel inhibitor did not react with the active site cysteine as anticipated. Importantly, both structures were trapped in the catalytically relevant adenylation conformation due to the presence of the inhibitor. This work furthers the continuing story of targeting the adenylation domains of non-ribosomal peptide synthases for antimicrobial drug discovery, with an eye toward therapeutic intervention for infections caused by Pseudomonas aeruginosa, Mycobacterium tuberculosis, and Yersinia pestis.