Phosphopantetheinyl transferase binding and inhibition by amidino-urea and hydroxypyrimidinethione compounds

Owing to their role in activating enzymes essential for bacterial viability and pathogenicity, phosphopantetheinyl transferases represent novel and attractive drug targets. In this work, we examined the inhibitory effect of the aminido-urea 8918 compound against the phosphopantetheinyl transferases PptAb from Mycobacterium abscessus and PcpS from Pseudomonas aeruginosa, two pathogenic bacteria associated with cystic fibrosis and bronchiectasis, respectively. Compound 8918 exhibits inhibitory activity against PptAb but displays no activity against PcpS in vitro, while no antimicrobial activity against Mycobacterium abscessus or Pseudomonas aeruginosa could be detected. X-ray crystallographic analysis of 8918 bound to PptAb-CoA alone and in complex with an acyl carrier protein domain in addition to the crystal structure of PcpS in complex with CoA revealed the structural basis for the inhibition mechanism of PptAb by 8918 and its ineffectiveness against PcpS. Finally, in crystallo screening of potent inhibitors from the National Cancer Institute library identified a hydroxypyrimidinethione derivative that binds PptAb. Both compounds could serve as scaffolds for the future development of phosphopantetheinyl transferases inhibitors.

www.nature.com/scientificreports/ Altogether these findings indicate that PptT fulfils all the requirements to be a good drug target in the search for new anti-tuberculosis drugs 8 . Initiatives to inhibit PptT have emerged with the National Cancer Institute (NCI) Development Therapeutics Program (DTP) Open Compound collection (http:// dtp. nci. nih. gov) using a combination of virtual screening, structure similarity-based searching, enzyme inhibition dose response with the MBP-PptT fusion protein 9 , and whole-cell inhibition of M. tuberculosis growth 10 . Thirty-six molecules were identified from primary and secondary screenings, with IC 50 in the range 0.4-314 µM. As a general rule, none of these compounds displayed efficient whole-cell inhibition of M. tuberculosis. On the other hand, the discovery of the 8918 amidino-urea compound (1-[(2,6-diethylphenyl)-3-N-ethylcarbamimodoyl]urea; compound 1 in Fig. 1b) that kills M. tuberculosis by inhibiting PptT in vitro and in a mouse model, with no toxicity to human cells, contributed to redefine PPTases as major contributors to the development of antimycobacterial agents 11 . This motivated the recent development of a simple high-throughput colorimetric assay to directly screen for inhibitors of PptT 12 .
Mycobacterium abscessus and Pseudomonas aeruginosa represent two other human pathogens, involved in cystic fibrosis and bronchiectasis 13 , for which PPTases may represent attractive drug targets. M. abscessus produces two PPTases, which display high sequence identity with the PPTases from M. tuberculosis. Given the conservation of genes and biosynthetic pathways involved in the synthesis of lipids in mycobacteria, it is likely that the PptT ortholog from M. abscessus, PptAb, is required for the activation of diverse PKSs and NRPSs involved in the production of compounds such as mycolic acids, glycopeptidolipids and siderophores 14,15 . P. aeruginosa has a unique PPTase, PcpS, which was shown to be essential for the growth of the bacteria 16 . PcpS was found to activate the acyl carrier protein (ACP) component of the FAS system and also to act on NRPSs responsible for the assembly of pyoverdin and pyochelin, two siderophores required for the virulence of the bacteria 16,17 .

Materials and methods
Chemistry. (1-[(2,6-diethylphenyl)-3-N-ethylcarbamimodoyl]urea (8918; compound 1 in Fig. 1b) was prepared according to Ballinger et al 11 . To a solution of 1-ethylguanidine sulfate (300 mg, 1.62 mmol, 1.0 equiv) in acetone (6.5 mL) was added dropwise a 10% aqueous KOH solution (1.8 mL, 2.0 equiv). After stirring at RT for 5 min, 1,3-diethyl-2-isocyanatobenzene (304 mg, 1.8 mmol, 1.1 equiv) was added dropwise and the resulting solution allowed to reflux overnight. After cooling and removing of the solvents under a stream of nitrogen, deionized water (7 mL) was added to the crude reaction mixture and the pH adjusted to 7.0 using 2 N aqueous HCl. A white precipitate was collected by vacuum filtration and purified by chromatography on silica gel (ethyl acetate-petroleum ether 1:9 to 2:8) to give the title compound (154 mg, 37% yield) as a white solid. All analyses agreed with the data reported in the literature.
Production and purification of recombinant proteins. PptAb and S2105A ACP were expressed and purified as previously described 18 . The pcpS gene was amplified by PCR from P. aeruginosa genomic DNA and cloned between the NdeI and HindIII restriction sites of the IPTG-inducible expression vector pET28a. The resulting plasmid expresses PcpS with an N-terminal His 6 tag followed by a recognition site for thrombin. Frozen  www.nature.com/scientificreports/ cells expressing the full-length PcpS were used to start an overnight 50 mL LB-kanamycin (35 µg mL −1 ) culture at 37 °C prior to inoculation in baffled flasks containing 2 L of the same media. Cells were allowed to grow for approximately 2 h at 37 °C before temperature was dropped from 37 to 30 °C. When OD 600 reached 0.5-0.7, cells were induced with IPTG at a final concentration of 0.7 mM and grown for an additional 4 h prior to harvesting by centrifugation at 4,000 × g for 30 min. To eliminate residual LB medium, the cell pellets were resuspended in 40 mL of 100 mM HEPES pH 7.5, 150 mM NaCl prior to centrifugation at 4,000 × g for 20 min and storage at -80 °C. The pelleted cells were resuspended in 60 mL buffer A (50 mM HEPES pH 7.5, 150 mM NaCl, 1 mM MnCl 2 , 10% glycerol). After sonication (4 cycles of 30-s pulse, 50% amplitude, and power 6 W) and centrifugation at 30,000 × g for 30 min, the lysate was loaded on a gravity-flow column containing 2 mL of Talon Superflow Metal Affinity Resin (GE Healthcare). The His 6 -tagged PcpS was eluted from the column with buffer A supplemented with 120 mM imidazole. After elution, the His 6 tag was removed by adding thrombin during 2 h dialysis against buffer C (25 mM HEPES pH 7.5, 50 mM NaCl, 5 mM MgCl 2 ) at 4 °C. Dialysis product was then loaded again on metal affinity beads to remove free His 6 -tag and His 6 -tagged PcpS. The 10-mL protein flow-through was then concentrated to 5 mL and injected into a HiLoad Superdex 75 16/60 (GE Healthcare) preequilibrated with buffer C. After elution, the fractions of interest were supplemented with 2 mM CoA, pooled and concentrated to the desired concentration for crystallization.
Activity/inhibition assay. Inhibition was assessed using the previously described activity assay 19   For obtaining the complex with 8918, PptAb-CoA crystals were soaked 20 h in a 5 mM solution of compound. Soaking with higher concentrations led to crystal cracking. For the crystallographic screening of compounds from the National Cancer Institute (NCI) Development Therapeutics Program (DTP) Open Compound collection, crystallization of PptAb-CoA was performed at 12 °C on a plate pre-coated with 1 μL of a 10 mM solution of compound in DMSO, as described in the 'dry' co-crystallization approach 23 . Crystals were directly cryo-cooled in liquid nitrogen. Data were collected at beamline ID30B at ESRF (compound 8918) and XALOC at ALBA (NCI compounds), and processed using XDS 24 and CCP4 25 . Structures were initially refined using Pipedream 26 followed by alternating refinement cycles with BUSTER 27 and Phenix 28 in order to allow for anisotropic B-factor refinement. Coot 29 was used for model building and corrections. Refinement dictionaries for all compounds were generated using MarvinSketch (Marvin 20.8, 2020, ChemAxon) and the Grade server 30 . Crystallization and structure determination of PcpS-CoA. Crystals of PcpS-CoA were obtained at 12 °C using the hanging drop method by mixing an equivalent volume of freshly prepared protein solution at 6 mg mL −1 and of reservoir solution (0.1 M Na citrate pH 5.6, 10% PEG 4 K, 10% propanol). Crystals were directly cryo-cooled in liquid nitrogen. Data were collected at beamline XALOC at ALBA, and processed using XDS 24 and Phenix 31 . Structures were solved by molecular replacement using the X-ray structure of PptT (PDB code 4U89) as a template in Phaser 32 . Iterative cycles of manual model building in Coot 29 and refinement procedures using Phenix refine 28 were applied until convergence. www.nature.com/scientificreports/ reservoir solution (0.1 M Na Cacodylate pH 6.5, 30% PEG 8 K, 0.2 M (NH 4 ) 2 SO 4 ) and directly frozen in liquid nitrogen. Data were collected at beamline PX1 at SOLEIL, and processed using XDS 24 and Phenix 31 . Structures were solved by molecular replacement using the X-ray structure of the PptAb-ACP complex (PDB code 6RCX) as a template in Phaser 32 . Refinement was conducted as described for PcpS-CoA.

Results and discussion
Inhibition of PPTase activity and bacterial growth by 8918. To check whether 8918 could inhibit the conversion of apo-ACP to holo-ACP by PptAb and PcpS, we resynthesized and evaluated this compound using an in vitro assay 18,19 . We found that 8918 inhibits ACP conversion by PptAb at nM concentrations and that total inhibition was achieved in the µM range ( Fig. 2a and Supplementary Figs. S1 and S2). In contrast, no marked inhibition of PcpS could be observed even at high concentration of 8918 (5 µM; Fig. 2a and Supplementary Fig. S3), in accordance with previous work where 8918 was inactive against P. aeruginosa 11 . Here, we verified the lack of inhibitory activity of 8918 against P. aeruginosa by measuring the reduction of a tetrazolium salt by metabolically active cells into a colored formazan compound (Fig. 2b). We also assessed the impact of 8918 on the growth of two so far non-tested mycobacterial species, M. bovis BCG and M. abscessus. Surprisingly, although an inhibition by 8918 could be observed in the case of M. bovis BCG (Fig. 2b), in a manner reminiscent of what was observed in the case of M. tuberculosis 11 , no marked inhibition could be observed in the case of M. abscessus at concentration up to 128 μM (Fig. 2b). Consistently, a time-kill kinetics study showed that 8918 has no bactericidal activity against M. abscessus at concentrations up to 128 μM ( Supplementary Fig. S4). It has been shown that the inhibitory effect of 8918 against M. tuberculosis is potentiated by the PptH (phosphopantetheinyl hydrolase) enzyme encoded by the rv2795c gene 11 . PptH catalyzes the hydrolysis of the Ppant arm from holo-CP substrates and resistance to 8918 was in the majority associated with loss-of-function PptH mutations 11 . As other species in the Mycobacterium genus 33 , both M. bovis BCG and M. abscessus possess a PptH ortholog with identity score to the M. tuberculosis enzyme of 99 and 80%, respectively. We checked that the corresponding protein sequences also share the same active site residues, especially those proposed for the catalytic reaction, as PptH 33 . Then, the lack of inhibitory effect on the growth of M. abscessus does not seem to be linked to an inactive phosphopantetheinyl hydrolase ortholog. One hypothesis to explain this could be attributed to the low permeability of the M. abscessus cell envelope and consequently to a lower penetration of the inhibitor in bacteria.
Crystallographic analysis of 8918 binding to PptAb. Despite the absence of inhibition on M. abscessus growth, and since 8918 was able to inhibit ACP conversion in our in vitro assay, we wanted to confirm binding to PptAb using X-ray crystallography. Crystals of PptAb prepared in the presence of CoA and Mn 2+ were soaked in a 5 mM solution of 8918 and the structure of the complex was solved at 1.40 Å (Supplementary  Table S1 and Supplementary Fig. S5a; Fig. 3a). Interestingly, binding of compound 8918 to PptAb displays the same pattern as that observed for PptT (Fig. 4a,c). Three hydrogen bonds involving the catalytic base Glu153, one hydrogen bond formed with the main chain oxygen atom of Leu167, and van der Waals contacts established with Lys152, Tyr156, Lys157, Trp166, Gly168, Phe169, and CoA are found. It is noteworthy that a second 8918 molecule was found at the surface of PptT 11 , which is not the case for PptAb.  Fig. 1). This figure was generated using Image Lab Version 6.  Table S1 and Supplementary Fig. S5b; Fig. 3b). Since PptAb was incubated with 4 mM 8918 prior to mixing with ACP, the inhibitor does not prevent the ACP domain and PptAb from interacting. In fact, the ACP domain intervenes very little in the interface. Only the residue Trp2129 is involved in binding through van der Waals contact and, by sitting on top of the diethylphenyl moiety of 8918, causing the displacement of the inhibitor along its longitudinal axis by about 0.9 Å (Fig. 4b). Despite this, three of the four hydrogen bonds observed for the enzyme alone are preserved (Figs. 3 and 4c).

The crystal structure of PcpS explains the lack of inhibition by 8918.
In the course of this study, we also resolved the crystal structure of PcpS in the presence of CoA at 2.40 Å resolution (Fig. 5a). PcpS crystallizes in space group P1 with four molecules in the asymmetric unit (Supplementary Table S1). However, analysis of the protein interfaces using PISA calculations as implemented on the EMBL-EBI server 34  www.nature.com/scientificreports/ sequence alignment indicates that PcpS have only 23 and 21% sequence identity with PptAb and PptT, respectively, whereas sequence identity between the latter amounts 65% (Supplementary Fig. S6). Despite low sequence identity, PcpS shares a high degree of structural conservation with the mycobacterial PPTases. For instance, superposing the structure of PcpS-CoA and that of PptAb-CoA (PDB entry 6QWU 18 ) led to an RMSD value of 1.8 Å for 192 common Cα atoms. The main secondary structural elements that constitute the core of the PPTases are preserved whereas notable differences appear in the N-terminal part, which is longer in the case of PcpS, and in insertion and deletion elements, which are found in loops or short additional helices at the periphery of the protein (Fig. 5a and Supplementary Fig. S6). The cofactor in PcpS occupies the same position and configuration as those found for CoA in the PptAb structure with the Ppant arm perfectly visible in electron density (Supplementary Fig. S5c). The importance of pH, and in particular acidic pH, for the stabilization of the Ppant arm has been demonstrated 18 , and with that respect it should be noted that PcpS crystallizes in the presence of citrate buffer at pH 5.6. The chemical environments of cofactor and cations in PcpS and PptAb also share strong conservation with 55% sequence identity and 66% similarity among the 29 residues making up the ligand-binding site. Major differences are found at Lys72/Val90 (PptAb/PcpS numbering), Lys75/Asp93, Gly76/Arg94, Gln77/ Ala95, Trp166/Arg188, Gly168/Tyr190, and Asp171/His193 (Fig. 5b,c). The Lys75 → Asp93, Gln77 → Ala95, Gly168 → Tyr190, and Asp171 → His193 PptAb to PcpS replacements do not seem to have any structural and chemical consequences, and it is noteworthy that the change Gly168 → Tyr190 compensates the loss of Trp at position 188. The Lys72 → Val90 substitution induces the loss of a hydrogen bond with CoA, which is more than compensated by the change of Gly76 in PptAb to Arg94 in PcpS (Fig. 5c). Tight binding of CoA to PcpS is completed by hydrogen bonds between Glu175, probably the PcpS catalytic base, to another oxygen of the β-phosphate and to the nitrogen atom of the β-alanine (Fig. 5b). Differences are also observed in the coordination of the M 2+ ions. Two Mn 2+ ions are found in the PptAb-CoA structure whereas only one Mg 2+ has been identified looking at all four protein molecules constituting the PcpS-CoA asymmetric unit (Fig. 5c). The presence of an arginine residue in position 94 and to a lesser extent in position 188 makes the PcpS active site less accessible, at least once the fixation of the cofactor is engaged. Despite several attempts, either by  www.nature.com/scientificreports/ co-crystallization or by soaking, no structure of a complex between PcpS and 8918 could be obtained. In fact, the lack of in vitro inhibition and of in crystallo binding for PcpS can be explained by the structure. Indeed, Arg94 in PcpS, found at the same position as Gly76 in PptAb ( Supplementary Fig. S6), a residue that takes part to the CoA-binding site, would cause a steric clash with the diethylphenyl group of the inhibitor (Fig. 5c). Although on the surface of the protein, the side chain of Arg94 offers very little flexibility, engaged as it is in hydrogen bonds with CoA (Fig. 5b). The fact that the Ppant of CoA is perfectly ordered in the structure of the complex with PcpS, which is not the case for PptAb-CoA-8918 ± ACP complexes ( Supplementary Fig. S5) is in accordance with the finding that 8918 inhibits its target by displacing the Ppant arm, which decreases but does not abolish catalytic activity in the case of PptT 11 .

Crystallographic screening of NCI compounds with PptT inhibitory effect.
To check whether or not NCI compounds identified against PptT 10 could also bind to PptAb, all 36 molecules were screened using the 'dry' co-crystallization approach 23 and 24 of them gave crystals, 21 of which could be exploited for structure determination. Despite the high resolution of the collected data sets, only one compound, P-62 (5-[(4-chlorophenyl)methyl]-6-[2-(dimethylamino)ethylimino-methyl]-2-sulfanylidene-1H-pyrimidin-4-one; compound 2 in Fig. 1b) was found to bind to PptAb in crystallo (Supplementary Tables S1 and S2; Supplementary Fig. S5d). P-62 binding to PptAb displays significant similarity with that of 8918 (Fig. 6). The position and conformation of CoA and of all residues defining the ligand-binding site are conserved. One exception is for Glu153, whose side chain is flipped by 90° around the χ2 dihedral angle making polar interactions with one of the two Mn 2+ ion. This ion is displaced by 2.0 Å when compared to its position in the PptAb-CoA-8918 complex. Residues Tyr156 (whose side chain is also flipped), Trp166 and Phe169 found to establish van der Waals contacts with 8918 also do so with P-62. The involvement of Leu167 main chain in hydrogen bonding to the ligand is also conserved. As for 8918, there is no electron density for the Ppant arm of CoA ( Supplementary Fig. S5d), which is also disordered in the structure. Surprisingly no inhibition of the in vitro transfer by PptAb could be observed with P-62 ( Supplementary Fig. S7) and P-62 neither affected cell viability of M. abscessus up to a 128 µM concentration ( Supplementary Fig. S4).  www.nature.com/scientificreports/ In conclusion, molecules from the NCI database and an amidino-urea compound with proven inhibitory activity against PptT from M. tuberculosis were tested against M. abscessus PptAb in vitro and in broth culture. Although none of the NCI compounds exhibits an inhibitory activity against PptAb, inhibition of PptAb in the presence of compound 8918 could be achieved in the µM range. Nevertheless, whole cell based inhibition experiments of M. abscessus in broth culture also emphasize how M. abscessus represents an "antibiotic nightmare" 35,36 . Indeed, unlike M. tuberculosis for which the MIC 90 of compound 8918 varies from 0.56 to 3 μM on clinical isolates, it has no impact on the growth of M. abscessus in broth culture at concentrations up to 128.0 μM, probably as a result of its inability to penetrate the cell envelope. Interestingly, binding of the NCI compound P-62 and 8918 to PptAb occupy the same region with interactions with conserved residues. Altogether, despite the limitations, this study shows that compounds of the amidinourea and hydroxypyrimidinethione type represent interesting molecular scaffolds for the rational design of more efficient inhibitors.