Pharmacological validation of dihydrofolate reductase as a drug target in Mycobacterium abscessus

ABSTRACT The Mycobacterium abscessus drug development pipeline is poorly populated, with particularly few validated target-lead couples to initiate de novo drug discovery. Trimethoprim, an inhibitor of dihydrofolate reductase (DHFR) used for the treatment of a range of bacterial infections, is not active against M. abscessus. Thus, evidence that M. abscessus DHFR is vulnerable to pharmacological intervention with a small molecule inhibitor is lacking. Here, we show that the pyrrolo-quinazoline PQD-1, previously identified as a DHFR inhibitor active against Mycobacterium tuberculosis, exerts whole cell activity against M. abscessus. Enzyme inhibition studies showed that PQD-1, in contrast to trimethoprim, is a potent inhibitor of M. abscessus DHFR and over-expression of DHFR causes resistance to PQD-1, providing biochemical and genetic evidence that DHFR is a vulnerable target and mediates PQD-1’s growth inhibitory activity in M. abscessus. As observed in M. tuberculosis, PQD-1 resistant mutations mapped to the folate pathway enzyme thymidylate synthase (TYMS) ThyA. Like trimethoprim in other bacteria, PQD-1 synergizes with the dihydropteroate synthase (DHPS) inhibitor sulfamethoxazole (SMX), offering an opportunity to exploit the successful dual inhibition of the folate pathway and develop similarly potent combinations against M. abscessus. PQD-1 is active against subspecies of M. abscessus and a panel of clinical isolates, providing epidemiological validation of the target-lead couple. Leveraging a series of PQD-1 analogs, we have demonstrated a dynamic structure-activity relationship (SAR). Collectively, the results identify M. abscessus DHFR as an attractive target and PQD-1 as a chemical starting point for the discovery of novel drugs and drug combinations that target the folate pathway in M. abscessus.

T reatment of lung disease caused by the opportunistic pathogen Mycobacterium abscessus delivers unsatisfactory cure rates despite long-term use of multidrug regimens with a macrolide, clarithromycin (CLR) or azithromycin, as the cornerstone (1).Treatment of M. abscessus is further complicated by the fact that the three subspecies, namely, M. abscessus subsp.abscessus, M. abscessus subsp.massiliense, and M. abscessus subsp.bolletii, present with differential drug susceptibility (2).The M. abscessus drug discovery pipeline is thin, and there is a need for novel target-lead couples to populate the early preclinical space (3).
Dihydrofolate reductase (DHFR) is a ubiquitous enzyme that catalyzes the NADPHdependent conversion of dihydrofolate to tetrahydrofolate, involved in subsequent metabolic reactions such as thymidylate and purine nucleotide biosynthesis.DHFR is exploited as a target for anti-bacterial, anti-parasitic, anti-fungal, and anti-cancer drugs (4).Cotrimoxazole (Bactrim), the synergistic combination of DHFR inhibitors trimetho prim and sulfamethoxazole (SMX), with the latter targeting dihydropteroate synthase [(DHPS) the first enzyme in the folate pathway], is used for the treatment of a range of Gram-negative and Gram-positive bacterial infections (5).However, there are no DHFR inhibitors in clinical use against infections caused by mycobacteria.Interestingly, cotrimoxazole is recommended for the treatment of M. abscessus lung disease (1,6) although the combination shows limited if any activity in vitro (7)(8)(9).
DHFR is encoded by dfrA (MAB_3090 c) in M. abscessus and is genetically essential, i.e., loss-of-function mutations due to transposon insertion prevent growth (10).However, evidence that pharmacological inhibition of DHFR with a small molecule results in effective inhibition of growth is lacking.Here, our objective was to determine whether M. abscessus DHFR is a vulnerable target and to identify a starting point for drug discovery.
We have previously shown that collections of anti-tuberculosis compounds provide a rich source for the identification of M. abscessus actives (11)(12)(13).Several screening campaigns yielded attractive hits against M. tuberculosis DHFR (14)(15)(16)(17)(18), including the pyrrolo-quinazoline PQD-1 with an MIC of 0.5 µM (19).Here, we show that PQD-1 also displays attractive potency against M. abscessus.We use PQD-1 as a tool compound to carry out biochemical, in silico docking, genetic, synergy, and structure-activity relationship (SAR) studies and to demonstrate activity against the subspecies of M. abscessus and a panel of clinical isolates.

PQD-1 inhibits the recombinant M. abscessus DHFR enzyme
To determine whether the DHFR inhibitors are active against M. abscessus act via DHFR, the recombinant protein from M. abscessus ATCC 19977 was generated and purified (Fig. S1 and S2) for biochemical inhibition assays.PQD-1, trimetrexate, and WR99210 showed low nanomolar IC 50 (concentration inhibiting half-maximal enzyme activity) against the isolated protein (Fig. 2; Table 1).The other tested anti-bacterial and anti-parasitic DHFR inhibitors were ~1,000-fold less active in the enzymatic assay, mirroring their lack of whole cell activity (Table 1).This included trimethoprim (Fig. 2; Table 1), suggesting that the intrinsic resistance of M. abscessus to this antibacterial is due to poor on-tar get activity.Interestingly, the two anti-cancer pteridines, namely, methotrexate and pralatrexate, had sub-micromolar IC 50 against the isolated target (Table 1), indicative of good target engagement despite their lack of activity against whole cells (Table 1), pointing to a non-target-related intrinsic resistance mechanism against these molecules, such as poor uptake, active efflux, and/or intrabacterial metabolism.

PQD-1 interacts favorably with the active site of M. abscessus DHFR
To gain structural insight into the binding mode of PQD-1, molecular modeling studies were carried out by docking PQD-1 into the crystal structure of M. abscessus DHFR (PDB code: 7K6C).In the conformation with the lowest binding free energy (−7.65 kcal/mol, Fig. 3), the nitrogen atoms of the 2,4-diamino substitution in the pyrimidine ring of PQD-1 form three hydrogen bonds with the oxygen atoms of the active site residues, i.e., Ile8, Ile99, and Asp30 (30).The nitrogen atoms in the pyrimidine heterocycle form favorable charge and salt bridge interactions with Asp30.Furthermore, the aromatic pyrrolo-quinazoline and benzene moieties form hydrophobic interactions with Ala10, Val57, and Leu60 and π-π staking interactions with Phe34 (Fig. 3).Docking of trimetho prim resulted in fewer interactions (Fig. 3) and a weaker docking score of −5.57kcal/mol, consistent with the lower inhibitory activity of trimethoprim against M. abscessus DHFR (Fig. 2; Table 1).As expected, trimethoprim shared PQD-1 binding residues interacting with the diamino-pyrimidine ring, a conserved pharmacophore in DHFR inhibitors (Fig. 1) (4).

Over-expression of DHFR in M. abscessus causes resistance to PQD-1
To provide functional evidence that PQD-1 exerts its whole cell activity by inhibition of bacterial DHFR, the M. abscessus ATCC 19977 dfrA gene was over-expressed in M. abscessus ATCC 19977 using the hsp60 promoter carried by plasmid pMV262, and the impact on PQD-1 susceptibility was determined.Over-expression of the DHFR gene caused strong resistance to PQD-1 (Fig. 4A) indicating that PQD-1 acts via the folate pathway and likely inhibits DHFR.

Spontaneous M. abscessus resistance to PQD-1 maps to thymidylate synthase ThyA
We and others have shown that one of the mechanisms of how spontaneous resist ance against DHFR inhibitors in M. tuberculosis can emerge is via missense mutations in the TYMS ThyA, a folate pathway enzyme downstream of DHFR, rather than mis sense mutations in the direct target DHFR (32)(33)(34)(35).TYMS ThyA catalyzes the reductive methylation of dUMP to dTMP utilizing methylenetetrahydrofolate (mTHF) as the methyl donor and reductant in the reaction, yielding dihydrofolate, the substrate of DHFR, as a by-product (Fig. 4B).Missense mutations in TYMS ThyA are thought to reduce enzymatic activity and thus the amount of dihydrofolate produced.This in turn affects the growth inhibitory activity of DHFR inhibitors (35).To determine whether this mechanism of resistance against DHFR inhibitors in M. tuberculosis is conserved in M. abscessus, we isolated spontaneous resistance mutants.Three independent cultures of M. abscessus ATCC 19977 were plated on Middlebrook 7H10 agar containing 33× MIC 90 (100 µM) of PQD-1, the lowest concentration that suppressed the growth of wild-type colonies on a solid medium.Resistance colonies emerged at a frequency of ~10 −7 /CFU.Three resistant strains from each selection experiment were randomly picked for further analyses.MIC determinations revealed a similar, high-level resistance to PQD-1 (Table 2; Fig. S3).Targeted Sanger sequencing of thyA and dfrA revealed missense mutations in ThyA in all nine strains and no polymorphisms in dfrA (Table 2).Complementation of a PQD-1 resistant thyA mutant strain with a copy of wild-type M. abscessus ATCC 19977 thyA increased susceptibility to PQD-1 and thus partially reverted the phenotype (Fig. S4).These results suggest that spontaneous resistance against PQD-1 can emerge via missense mutation in thyA, similar to what was observed for DHFR inhibitors active against M. tuberculosis.
In the past decade, two non-canonical folate enzymes, RibD (a second functional DHFR) and ThyX (a second thymidylate synthase, Fig. 4B) were found to be involved in the mechanisms of action and resistance of antifolates in M. tuberculosis (31,34).The M. abscessus genome harbors potential homologs of these genes (MAB_2976, MAB_3085 c) (10).To determine whether the PQD-1 resistant thyA mutant strains harbor additional mutations in these loci, which may contribute to their PQD-1 resistance, the nine strains were subjected to whole genome sequencing (Table S1).Additional consistent polymorphisms were not detected.Importantly, no polymorphisms were detected in the ribD or thyX coding or upstream sequences (Table S1).This suggests that the observed reduced PDQ-1 susceptibility in the nine resistant strains is solely due to the mutations in thyA.It is to be noted that the result of our PQD-1 resistance analysis does not exclude the possibility that characterization of a larger number of PQD-1-resistant M. abscessus strains may reveal additional resistance mechanisms due to mutations in ribD, thyX, or other genes.

PQD-1 analogs reveal a dynamic SAR against M. abscessus
To explore the SAR around the PQD-1 scaffold, nine analogs were retrieved from the Merck & Co., Inc., Rahway, NJ, USA, compound archive.MIC against M. abscessus and IC 50 against its recombinant DHFR enzyme were determined.The analogs, which mostly retained the pyrrolo-quinazoline moiety of PQD-1 and carried modifications in the eastern part of the scaffold (Table 3), showed a range of potency (Table 3) suggesting dynamic SAR and thus suitability of PQD-1 as a starting point for chemical optimization.Thus, the liabilities of PQD-1, briefly inadequate metabolic stability, and poor selectivity towards the human enzyme (19), can now be addressed in a lead optimization pro gram aimed at improving pharmacokinetic properties and introducing specificity for the mycobacterial enzyme.Whether the liabilities can be removed while retaining on-target activity remains to be determined.

PQD-1 synergizes with the dihydropteroate synthase inhibitor sulfamethoxa zole in M. abscessus
In other bacteria, DHFR inhibitors typically synergize with sulfonamide inhibitors of DHPS (5).Growth inhibition experiments showed that sulfamethoxazole alone is active against M. abscessus whole cells, with an MIC 50 of 12 µM (MIC 90 >100 µM).To determine whether PQD-1 and sulfamethoxazole synergize against M. abscessus, checkerboard growth inhibition experiments were carried out, revealing a fractional inhibitory concentration index (FICI) of 0.3, indicative of synergy between PQD-1 and sulfamethoxazole (Fig. 4C).This result paves the way for the development of a Bactrim-like synergistic drug combination for the treatment of M. abscessus infections.

PQD-1 is active against M. abscessus subspecies reference strains and a panel of clinical isolates
To determine whether the attractive activity of PQD-1 against the type strain M. abscessus subsp.abscessus ATCC 19977 was retained against M. abscessus subspecies, growth inhibition activities were measured against M. abscessus subsp.massiliense and bolletii reference strains and a panel of clinical isolates.PQD-1 retained activity against this strain collection indicating that PQD-1 is broadly active against the M. abscessus complex and providing epidemiological validation of the DHFR-PQD-1 target-lead pair (Table 4).
In summary, we have shown that M. abscessus DHFR is a vulnerable target and that the intrinsic resistance of M. abscessus to trimethoprim is due to poor activity against the M. abscessus DHFR enzyme.Biochemical data, complemented by docking and genetic studies, indicate that PQD-1 exerts whole cell anti-M.abscessus activity by inhibiting DHFR, thus pharmacologically validating the target.Screening against a panel of M. abscessus strains shows that target vulnerability is epidemiologically conserved, confirming DHFR/PQD-1 as an attractive target-lead couple.Dynamic SAR around the PQD-1 scaffold and synergy with the DHPS inhibitor sulfamethoxazole support a dual targeting approach of the folate pathway in M. abscessus, similar to the broadly successful Bactrim antibiotic.

Bacterial strains, media, and culture conditions
M. abscessus subsp.abscessus ATCC 19977 was purchased from the American Type Culture Collection.M. abscessus subsp.bolletii CCUG 50,184T and M. abscessus subsp.massiliense CCUG 48,898T was purchased from the Culture Collection University of Goteborg.M. abscessus subsp.abscessus Bamboo was provided by Wei Chang Huang (Taichung Veterans General Hospital, Taichung, Taiwan).M. abscessus subsp.abscessus K21 was provided by Sung Jae Shin (Department of Microbiology, Yonsei University College of Medicine, Seoul, South Korea) and Won-Jung Koh (Division of Pulmonary and Critical Care Medicine, Samsung Medical Center, Seoul, South Korea).The clinical M isolates of M. abscessus were provided by Jeanette W. P. Teo (Department of Laboratory Medicine, National University Hospital of Singapore).

Growth inhibition assay
Dose-response growth inhibition was measured as described previously (37).Briefly, for each inhibitor a two-fold dilution series starting at the desired highest concentration was dispensed onto flat-bottom 96-well plates using a D300e Digital Dispenser (Tecan).A 200-µL mid-log-phase culture of OD 600 nm = 0.05 was dispensed to each drug-containing well.Untreated control wells were included on each plate.Plates were sealed with parafilm, stored in boxes with wet paper towels, and incubated for 3 days at 37°C with shaking.Growth was monitored by measuring OD 600 nm using a Tecan Infinite 200 Pro microplate reader (Tecan).Day 0 values were subtracted from the corresponding end-point values, and percentage growth was calculated by dividing the growth value in the drug-containing well by the average growth value of the untreated control wells and multiplying by 100.Dose-response curves were generated by plotting drug concentra tions versus percentage growth using GraphPad Prism 9.  equipment.Non-reduced intact mass data were obtained on an Agilent electrospray ionization time-of-flight (ESI-TOF) mass spectrometer (6230) by injecting 1.5 µg of the final product.Further analysis was carried out with the Agilent BioConfirm software.

DHFR inhibition assay
Enzyme inhibition assays were performed according to the manufacturer's instructions using a DHFR assay kit purchased from Sigma-Aldrich (CS0340, Sigma-Aldrich Inc.val ues are defined as drug concentrations ), with minor modifications.In brief, 120-µM NADPH was mixed with 0.6 µg/mL M. abscessus DHFR, and 50 µL of the resulting mix was dispensed onto 96-well plates.For each drug/compound, a twofold dilution series starting at twice the desired highest concentration was dispensed onto the 96-well plates using a Tecan D300e Digital Dispenser.The reaction was started by adding 50 µL of 100-µM DHF to each well to give a final volume of 100 µL per well with final NADPH, DHFR, DHF, and DMSO concentrations of 60 µM, 0.3 µg/mL, 50 µM, and 0.25%, respectively.For each experiment, drug-free and enzyme-free controls were included in which DMSO was added to the reaction mix instead of test drugs and DHFR, respectively.All the reagents were prepared with 1 × DHFR assay buffer provided by the kit.The reaction progress was monitored by measuring the decrease in the absorbance (340 nm) at 30-second intervals for 10 minutes at room temperature using a Tecan Infinite 200 Pro microplate reader (Tecan).A graph of absorbance values versus time was plotted, and the slope of the graph over the linear range was taken to represent the velocity of the reaction.Percentage activity was calculated by dividing the slope of the inhibited enzyme by the average slope of the uninhibited enzyme and multiplying by 100.The IC 50 values were calculated by fitting the data to a non-linear regression sigmoidal dose-response curves (variable slope) using GraphPad Prism 9.5.1.

In silico docking
In silico docking was carried out by BOC Sciences Inc., NY, USA.AutoDock-GPU, an OpenCL-accelerated version of AutoDock4 running on GPUs, was used for molecular docking experiments (38).The crystal structure of M. abscessus DHFR (PDB code: 7K6C) was used as a receptor.AutoDock used the Lamarckian genetic algorithm and the empirical free energy scoring function and typically provided reproducible docking results for ligands.During the docking process, the coordinate files of protein and ligand were created, which included polar hydrogen atoms, partial charges, atom types, and information on the articulation of flexible molecules.The grid parameter file of the binding pocket was created with the grid center 1.786, 1.766, 58.555 (x, y, z in Å), and dimensions 30 × 30 × 30 Å referring to the binding sites of the ligand in 7K6C.
A conformation search algorithm was used to explore the conformational states of a flexible ligand, using the grid maps to evaluate ligand-protein interactions at each point in the docking simulation.The docking results were clustered to identify similar conformations.The conformation with the lowest binding free energy was identified as the most probable conformation.The interactions between the protein-ligand complex were mapped by PyMOL.

Over-expression of DHFR in M. abscessus
Wild-type M. abscessus ATCC19977 dfrA was cloned into plasmid pMV262 (39) to overexpress the gene under the control of the constitutive hsp60 promoter as described previously (16).The coding sequence of dfrA was PCR amplified (Phusion high-fidelity DNA polymerase, Fisher Scientific) from M. abscessus ATCC19977 genomic DNA with primers CCGGGATCCATGACGGGAACCATCGGG (BamHI, forward) and CCGGAATTCTCAC CCGTCGACTTTCCG (EcoRI, reverse), inserted in frame (BamHI-EcoRI), and transformed into E. coli TOP10.Positive clones were identified by PCR after which the integrity of the construct was confirmed by Sanger sequencing (Azenta Life Sciences).The purified plasmid DNA was electroporated into M. abscessus ATCC 19977 as previously described

FIG 2
FIG 2 Activity of selected DHFR inhibitors against recombinant M. abscessus ATCC 19977 DHFR enzyme.Enzyme inhibition dose-response curves are shown for (A) PQD-1, (B) trimetrexate (TMX), (C) WR99210, and (D) trimethoprim (TMP).The data were fitted to a non-linear regression curve using the variable slope model, and the IC 50 values were calculated using GraphPad Prism 9. Percent activity values are the means of three independently performed experiments, and error bars indicate standard deviations.

FIG 3
FIG 3 In silico model of PQD-1 and trimethoprim binding to M. abscessus ATCC 19977 DHFR (PDB code: 7K6C).(A, C) Optimized poses of PQD-1 (yellow sticks) and trimethoprim (green sticks) with interactions with DHFR active site residues.DHFR is colored in cyan, with the residues in the binding pockets shown as cyan sticks.(B, D) Two-dimensional presentation of the key interactions of PQD-1 (B) and trimethoprim (D) with their binding pockets shown in (A, C).The hydrogen bonds, salt bridges, and hydrophobic and π-π stacking interactions are depicted as green, orange, pink, and warm pink dashed lines, respectively.

FIG 4
FIG 4 Mechanism of action of PQD-1.(A) Effect of DHFR over-expression on PQD-1 susceptibility of M. abscessus ATCC 19977.Cultures were treated for 3 days with PQD-1 (top) or TMX (bottom), the second most potent M. abscessus inhibitor identified in the initial whole cell screen (Table 1) and confirmed as a biochemical inhibitor of M. abscessus DHFR (Fig. 2).Percent growth values are the means of three independently carried out experiments, and error bars indicate standard deviations.(B) Schematic of mycobacterial folate pathway and role of thymidylate synthase (TYMS) ThyA.DHPS, inhibited by SMX, converts p-amino benzoic acid (pABA) into dihydropteroate (DHP), which in turn is converted to dihydrofolate (DHF).DHFR, inhibited by PQD-1, reduces DHF to tetrahydrofolate (THF), which is converted into methyl-tetrahydrofolate (mTHF).The TYMS ThyA catalyzes the reductive methylation of deoxyuridine-monophosphate (dUMP) to deoxythymidine-monophosphate (dTMP) utilizing mTHF as the methyl donor and reductant in the reaction, yielding DHF, the substrate of DHFR, as a by-product.ThyX is a second thymidylate synthase catalyzing the reductive methylation of dUMP to dTMP utilizing mTHF only as the methyl donor (hence generating THF instead of DHF) and NADPH as the reductant.The pathway is according to Hajian et al. (31).(C) Plot of the fractional inhibitory concentrations (FIC) of PQD-1 (top) and TMX (bottom) versus DHPS inhibitor SMX.The FIC index (FICI), calculated as (MIC A combi /MIC A alone ) + (MIC B combi /MIC B alone ), indicates synergy if <0.5.The FICs were calculated based on the MIC 50 values.The experiments were carried out three times independently in duplicate, and the presented data are one representative example.

TABLE 3 1 a
Dynamic SAR of PQD-1 analogs against M. abscessus ATCC 19977 and its recombinant DHFR enzyme MIC 50 and MIC 90 values are defined as compound concentrations causing 50% and 90% growth inhibition, respectively, compared to untreated control and are the means of three independent experiments.b IC 50 values are defined as compound concentrations causing 50% inhibition of M. abscessus DHFR enzyme activity compared to untreated control and are the means of three independent experiments.

TABLE 1
Activity of DHFR inhibitors against M. abscessus ATCC 19977 and its recombinant DHFR enzyme MIC 50 and MIC 90 values are defined as drug concentrations causing 50% and 90% growth inhibition, respectively, compared to untreated control and are the means of three independent experiments.
a b IC 50 values are defined as compound concentrations causing 50% inhibition of M. abscessus DHFR enzyme activity compared to untreated control and are the means of three independent experiments.c NA, not applicable.

TABLE 2
Characterization of spontaneous PQD-1 resistant M. abscessus ATCC 19977 strains a Exp.#, independent selection experiments with independently grown culture batches; wt, wild type; PQD R , PQD-1 resistant strain; TMX, trimetrexate; CLR, clarithromycin; nt/aa, nucleotide sequence polymorphism and associated amino acid substitution.MIC 90 , drug concentrations causing 90% growth inhibition compared to untreated control.The MIC experiments were carried out three times independently, and the mean values are shown (see Fig.S3for dose-response curves). a -pressure homogenization.DHFR protein was purified with Ni-NTA resin and eluted in 20-mM Tris-HCl, 300-mM NaCl, 250-mM imidazole, 0.5-mM TCEP, and 10% glycerol, at pH 7.5.Then, the buffer was exchanged into a final storage buffer with 20-mM Tris-HCl, 150-mM NaCl, 0.5-mM TCEP, and 10% glycerol, at pH 7.5.The concentration of the purified DHFR was measured by the Bradford method.Non-reduced and reduced [100-mM dithiothreitol (DTT)] SDS-PAGE was performed with Bio-Rad

TABLE 4
Activity of PQD-1 against M. abscessus subspecies reference strains and clinical isolates MIC 90 values are defined as drug concentrations causing 90% growth inhibition compared to untreated control and are the means of two independent experiments.TMX, trimetrexate; CLR, clarithromycin.