3‑Hydroxy-Beta-Lactam Inhibitors of Dihydrofolate Synthetase

Abstract

The 3-hydroxy-β-lactam (3-HβL) group derived from the natural product tabtoxinine-β-lactam (TβL), an inhibitor of glutamine synthetase, was repurposed to develop an inhibitor of dihydrofolate synthetase (DHFS). We show that replacement of the carboxyl group of p-amino-benzoic acid (PABA) with a 3-HβL moiety on the chemical scaffold of a folate mimic results in a potent inhibitor of DHFS. Using a combination of in vitro steady-state kinetics, enzyme-coupled assays, and molecular modeling, we validate the essential role of the 3-HβL group in DHFS inhibition. We provide an optimized synthesis of the 3-(p-aminophenyl)-3-HβL component via a sequence of the C–C bond-forming Henry reaction and a β-lactam ring-closing Grignard reaction. We demonstrate full elaboration to an antifolate scaffold via chemical or chemoenzymatic conjugation of the PABA analogue 3-(p-aminophenyl)-3-HβL to a pterin mimic. In this proof-of-concept study, we provide the first evidence that the 3-HβL group can be used as a general pharmacophore for inhibitors of enzymes in the ATP-dependent carboxylate-amine ligase superfamily through carboxylate replacement on substrate scaffolds, which could have broad therapeutic applications.

Tabtoxin is a monobactam phytotoxin produced by plant pathogenic strains of P. syringae and some species of Streptomyces. Tabtoxin is a dipeptide composed of l-Thr at the N-terminus and tabtoxinine-β-lactam (TβL) at the C-terminus. Internalization of the prodrug tabtoxin by dipeptide transporters leads to dipeptidase cleavage and the release of the active component TβL in the cytoplasm. TβL is a monobactam (characterized by its lack of a secondary ring system fused to the β-lactam ring) and acts as an inhibitor of glutamine synthetase (GS). GS catalyzes the addition of ammonia to glutamate, stabilizing a phosphorylated tetrahedral intermediate on the way to the production of glutamine. Rather than acting as a sacrificial substrate to covalently modify its target enzyme (as most β-lactam antibiotics do), TβL uses its β-lactam to position the 3-hydroxy group toward the active site ATP, allowing phosphorylation to occur, but preventing any tetrahedral intermediate breakdown and subsequent product formation (Figure A). , The 3-hydroxy-β-lactam (3-HβL) of TβL, and its unique mechanism of action, is potentially generalizable to the superfamily of ATP-dependent carboxylate-amine ligase enzymes. ,

1.

Reactions catalyzed by ATP-dependent carboxylate–amine ligase enzymes. A) Glutamine synthetase (GS) catalyzed reaction and the mechanism of inhibition by TβL. B) Dihydrofolate synthetase (DHFS) catalyzed reaction and the mechanism of a proposed 3-HβL inhibitor.

One potential target for this type of mechanism-based inhibition is dihydrofolate synthetase (DHFS). DHFS is not endogenous to humans, potentially reducing the off-target effects of DHFS inhibition. DHFS catalyzes the conversion of 7,8-dihydropteroic acid (7,8-DHP) to 7,8-dihydrofolic acid (7,8-DHF). DHFS acts between the two clinically relevant folate biosynthesis enzymes, dihydropteroate synthase (DHPS) and dihydrofolate reductase (DHFR). While structurally distinct, DHFS and GS share mechanistic similarities in the formation of a reactive acyl phosphate intermediate that is derived from a phosphoryl group transfer from ATP to the carboxylic acid substrate (Glu for GS; 7,8-DHP for DHFS). The acyl phosphate is subsequently condensed with an amine nucleophile (NH₃ for GS; l-Glu for DHFS) to form an amide product with the release of P_i (Figure B). The DHFS product 7,8-DHF is subsequently reduced to tetrahydrofolic acid (THF) by DHFR. In E. coli, DHFS also catalyzes the conversion of THF to poly-n-Glu-THF in the same ATP-dependent manner (Figure S1).

DHFS has 9 crystal structures (PDB: 1W78, 1W7K, 3N2A, 3NRS, 3PYZ, 3QCZ, 6K8C, 2VOR, 2VOS) reported in a variety of liganded states. The DHFS structure is composed of two subunits that create a binding pocket, which accommodates the 7,8-dihydropterin moiety of 7,8-dihydropteroic acid (7,8-DHP). While there are FDA-approved inhibitors of DHPS (sulfamethoxazole) and DHFR (trimethoprim) used clinically to treat bacterial infections (Figure S1), there are no approved inhibitors of DHFS. A DHFS inhibitor could replace commonly used DHPS inhibitors for combination therapy with DHFR inhibitors to provide a synergistic drug pair. Two phosphinic acid-based folate mimics have been reported to inhibit DHFS in an ATP-dependent manner similar to glufosinate, a phosphinate inhibitor of GS. − There is one report of simple dihydropteroic acid derivatives with variable pterin-PABA linkages that show weak inhibition of DHFS from N. gonorrhoeae. Hence, we sought to demonstrate that the 3-HβL moiety from tabtoxin could be similarly repurposed to inhibit DHFS through the incorporation of this pharmacophore into an antifolate scaffold (Figure B).

If the unique 3-HβL of TβL can be fitted to a molecule that would bind DHFS, a potential mechanism-based DHFS inhibitor can be synthesized. Given that there are multiple resistance genes to DHPS and DHFR inhibitors circulating in clinics, attacking the folate biosynthesis pathway using a less-targeted enzyme could prove fruitful in combating clinical antibiotic resistance. − Therein lies the hypothesis of this study: a compound utilizing the pterin-binding pocket of DHFS, suitably linked to a 3-(p-aminophenyl)-3-HβL, could produce a tight-binding mechanism-based inhibitor of DHFS. The envisioned pterin-3-HβL (1) underwent retrosynthetic analysis, with the pterin-portion of the molecule to be either chemically or enzymatically ligated as the final step and the 3-(p-aminophenyl)-3-HβL (2) to be produced synthetically (Figure A). Once the forward synthesis of 3-(p-aminophenyl)-3-HβL (2) was completed, it was ligated to natural pterin chemoenzymatically to provide inhibitor 1 and synthetically to two pterin mimics to provide inhibitors 3,4. These compounds were tested and validated as inhibitors of DHFS in vitro. The importance of the 3-HβL pharmacophore was demonstrated through controlled hydrolysis to the corresponding β-amino acid 5, which showed reduced DHFS inhibition in vitro. We performed a variety of biochemical assays and molecular docking to probe the mechanism of DHFS inhibition by the most potent inhibitor (3).

2.

A) Retrosynthetic analysis of a proposed DHFS inhibitor (1) derived from an electrophilic pterin moiety and a nucleophilic 3-(p-aminophenyl)-3-HβL (2). B) The three potential DHFS inhibitors (3–5) that were synthesized in this work.

Methods

All chemicals and solvents were purchased from reputable vendors (e.g., Sigma-Aldrich, Oakwood, Enamine, Thermo Fisher). Folate precursors and metabolites were purchased from Schircks Laboratories. All prep-HPLC was performed using an Agilent/HP 1050 quaternary pump module with an Agilent/HP 1050 MWD module with a Phenomenex Luna 10 μm C18(2) 100 Å column, 250 × 21.20 mm, 10 μm fit with a guard column. All LC-MS was performed on an Agilent 6130 quadrupole LC-MS with a G1313 autosampler or G1367B autosampler, a G1315 diode array detector, and a 1200 series solvent module. A Phenomenex Gemini C18 column, 50 × 2 mm, 5 μm fit with a guard column, was used for all LC-MS separations. Mobile phases for prep-HPLC and LC-MS were 0.1% formic acid in (A) H₂O and (B) CH₃CN, and data were processed using ChemStation software (Agilent). NMR was performed on either an Agilent DD2 600 MHz, an Agilent DD2 500 MHz, or a Varian Unity Plus 300 MHz instrument, and data were processed using MestraNova. Protein concentrations were determined by optical absorbance measurement (A280) using a NanoDrop 2000 UV–vis spectrophotometer (Thermo Scientific). All optical absorption plate readings were performed on either a SpectraMax Plus 384 or a Fisher AccuSkan Go. Data plots and curve fitting were generated by using GraphPad Prism.

Synthesis

Synthetic methods and compound characterization are provided in the Supporting Information included with this publication.

Chemoenzymatic Synthesis Using HPPK/DHPS/DHFR

7,8-Dihydro-6-hydroxymethylpterin-pyrophosphokinase (HPPK; WP_000215139.1), DHPS (WP_000764731.1), DHFR (AAA87976.1), and DHFS (WP_000584546.1) were cloned into a pET28 plasmid encoding an N-terminal His₆ tag (Tables S1–S3). The N-His₆-tagged proteins were overexpressed from E. coli BL21 and purified by Ni-NTA affinity chromatography (Figure S2). The purified enzymes were flash-frozen in liquid N₂ and stored as aliquots at −80 °C for later use. Assay conditions derived from a previous study on DHPS were adapted and used to investigate if DHPS could use 3-(p-aminophenyl)-3-HβL 2 as a substrate. Individual stock solutions of MgCl₂, β-mercaptoethanol (BME), bovine serum albumin (BSA), 6-hydroxymethyl-7,8-dihydropterin, p-amino-benzoic acid (PABA), ATP, and NADPH were made using 100 mM Tris buffer (pH = 7.5) and then stored in the −80 °C freezer until use. Wells of a 96-well plate were flushed with argon and then filled with assay solution in triplicate and initiated by the addition of an enzyme mixture containing HPPK and DHPS. Final assay concentrations were as follows: 5 mM MgCl₂, 20 mM BME, 1 mg/mL BSA, 150 μM 6-hydroxymethyl-7,8-dihydropterin, 150 μM PABA, 300 μM ATP, 1.1 μM HPPK, and 0.65 μM DHPS, with a total well volume of 150 μL. The plate was incubated at 37 °C for 1 h, and then like wells were combined and centrifuge-filtered through a 10K molecular weight cutoff (MWCO) filter to remove any enzyme. The solutions were then analyzed by LC-MS with single-ion monitoring protocols to detect the various pteridine-oxidation states of products. Once the PABA-derived product of DHPS (7,8-DHP) was detected by LC-MS, the assay was run again with the addition of 150 μM of NADPH and 1.1 μM of DHFR. The resulting fast reduction of 7,8-DHP creates a measurable reduction in optical absorbance at 340 nm (NADPH) that only occurs when 7,8-DHP is present (6-hydroxymethyl-7,8-dihydropterin is not a substrate for DHFR). As such, the optical absorbance at 340 nm was read every 10 s for 1 h after initiation by addition of an enzyme cocktail containing HPPK, DHPS, and DHFR. The steady state rate of DHPS production of 7,8-dihydropteroic acid was calculated from the slope of the linear absorbance decrease. Once steady state kinetics of DHPS was established, the assay was rerun with 3-(p-aminophenyl)-3-HβL 2 replacing the PABA at the same concentration. The HPPK and DHPS assay was performed using 3-(p-aminophenyl)-3-HβL 2 as a substrate with LC-MS detection of the corresponding product 1. Inclusion of NADPH and DHFR facilitated the measurement of steady state kinetic data for the DHPS ligation of 6-hydroxymethyl-7,8-dihydropterin pyrophosphate and 3-(p-aminophenyl)-3-HβL 2.

DHFS Steady State Kinetics and Inhibition

7,8-DHP is not commercially available and is unsuitable as a substrate for steady state DHFS kinetic experiments due to competing air oxidation. THF was purchased in a sealed ampule and stored under an inert atmosphere (drybox) until use as a DHFS substrate. To have an optical absorbance signal from which to calculate DHFS steady state kinetics, a coupled enzyme reaction with pyruvate kinase/lactate dehydrogenase (PK/LDH) was chosen to quantify ADP from DHFS activity. , Individual stock solutions of MgCl₂, dithiothreitol (DTT), BSA, glycine, phosphoenolpyruvate (PEP), NADH, ATP, l-Glu, and THF were made using 100 mM Tris buffer (pH = 7.5) and then stored in the −80 °C freezer until use. PK/LDH was bought as a glycerol solution. Wells of a 96-well plate were flushed with argon, then filled with assay solution in triplicate, and initiated by the addition of DHFS. Final assay concentrations were as follows: 10 mM MgCl₂, 5 mM DTT, 1 mg/mL BSA, 50 mM Gly, 150 μM PEP, 150 μM NADH, 100 units/mL PK, 150 units/mL LDH, 150 μM ATP, 100 μM Glu, 50 μM THFA, and 3 μM DHFS, with a final well volume of 150 μL. The plate was incubated at 37 °C for 1 h with the optical absorbance at 340 nm read every 10 s. Each like well was combined, and the solutions were centrifuge-filtered through 10K MWCO filters to remove any enzyme. The solutions were then analyzed by LC-MS with single-ion monitoring protocols to detect the various pteridine-oxidation states of products. Once steady state kinetics and product formation were confirmed, the assay was run with varying concentrations of HβLs 3, 4, and β-amino acid 5, as well as varying amounts of the pterin-3-HβL (1) produced by DHPS. The steady state enzymatic rate was determined for each concentration of compound, and IC₅₀ curves were developed.

Agar Diffusion Assay

E. coli ATCC 25922 was grown in an overnight culture in Luria broth (LB) and then diluted to a McFarland 0.5 standard (OD₆₀₀ = 0.05–0.1). 100 μL portion of this cell suspension was used to inoculate 35 mL of Mueller–Hinton No. 2 (MH II) agar melted and tempered to 47 °C. The inoculated agar was poured into a sterile Petri dish (145 mm × 20 mm; Greiner Bio-One), and once cooled to room temperature, 9 mm wells were cut. Wells were filled with 50 μL DMSO solutions of sulfamethoxazole, trimethoprim, and pterin-3-HβL 3, combinations thereof, as well as a pure DMSO control. The plates were incubated at 37 °C for 18 h, and then, zones of inhibition were measured for each well using an electronic caliper.

Protein Mass Spectrometry

Stock solutions of MgCl₂, dithiothreitol (DTT), BSA, glycine, ATP, and l-Glu were made using 100 mM Tris buffer (pH = 7.5) and then stored in a −80 °C freezer until use. Three independent solutions were made with the following concentrations: 10 mM MgCl₂, 5 mM DTT, 1 mg/mL BSA, 50 mM glycine, 500 μM ATP, and 500 μM Glu. Each solution was then dosed to give 0 μM pterin-3-HβL 3 plus 20 μM DHFS, 40 μM pterin-3-HβL 3 plus 20 μM DHFS, or 400 μM pterin-3-HβL 3 plus 20 μM DHFS. A fourth solution was made using 100 mM Tris buffer (pH = 7.5) and 20 μM DHFS, lacking all cosubstrates and additives. Each solution had a final volume of 200 μL. Each vial was incubated at 37 °C for 1 h and then submitted for protein mass spectrometry analysis.

Three μL of each sample (∼50 pmol of protein) was added to 45 μL of solvent (water with 0.1% of formic acid) and loaded onto a C8 trap column to remove salts and then eluted onto an ESI source with an 8 min acetonitrile gradient. ESI-MS spectra were recorded in the positive-ion mode using a Bruker Maxis II 4G with an m/z range of 250–2900 m/z. Each peak in the LC-MS chromatogram was analyzed for mass spectra and deconvoluted by using ProteinMetrics Intact software.

³¹P NMR

Stock solutions of MgCl₂, dithiothreitol (DTT), BSA, glycine, ATP, ADP, l-Glu, and THF were made using 100 mM Tris buffer (pH = 7.5) and then stored in a −80 °C freezer until use. Individual standard solutions containing 10% by volume of D₂O were made of ATP, ADP, and DHFS. A final solution containing 10% D₂O by volume, with 10 mM MgCl₂, 5 mM DTT, 1 mg/mL BSA, 50 mM Gly, 500 μM ATP, 500 μM Glu, and 250 μM 3-HβL 3, was made, flushed with argon, and cooled. The standard solutions were subjected to ³¹P NMR conditions in an Agilent DD2 500 MHz instrument at 37 °C. The final solution was initiated with the addition of DHFS (final concentration 3 μM) and then immediately subjected to ³¹P NMR conditions in an Agilent DD2 500 MHz instrument at 37 °C. Processing parameters for 31P spectra can be found in Tables S6–S9.

Molecular Docking

Molecular docking was performed using AutoDock Vina (version 4). , The protein template was the DHFS structure from E. coli (PDB: 1W78). Twenty-six compounds were investigated by docking using an AutoDock genetic algorithm against the active site of DHFS with both ADP and ATP bound, and their potential poses were investigated and compared to determine the feasibility of binding and respective binding energies. Ligand interaction maps were assembled using LigPlot+. Protein–ligand complex structures were visualized using PyMOL version 3.0.3 (Schrödinger, LLC).

Results

Synthesis

Synthesis of 3-(p-aminophenyl)-3-HβL 2 is shown in Scheme A and was initiated with copper­(I)- and TEMPO-mediated oxidation of β-keto-ester 6, followed by a Henry reaction to add nitromethane to the α-keto-ester 7. Hydrogenolysis under neutral conditions resulted in a retro-Henry reaction with the elimination of nitromethane and the formation of the reduced β-keto-ester. This problem was solved by performing the hydrogenolysis under acidic conditions, resulting in the reduction of the two nitro moieties simultaneously to yield 9. Selective Boc protection of the aniline in the presence of the primary amine was achieved using Boc₂O in 10% AcOH/dioxane. The primary amine and the ester of 10 were condensed in a tert-butyl magnesium chloride-mediated ring closing, and subsequent deprotection of the aniline provided 3-HβL 2 as the corresponding TFA salt.

1. Synthetic Efforts Toward 3-HβL Inhibitors of DHFS .

^a A) The synthesis of a 3-(p-aminophenyl)-3-HβL warhead 2. B) Ligation of pterin-mimic 12 to 3-(p-aminophenyl)-3-HβL 2 to form 3-HβL inhibitor 3. C) Ligation of pterin-mimic 13 to 3-(p-aminophenyl)-3-HβL 2 to form the 3-HβL inhibitor 4. D) Acid-mediated hydrolysis of the β-lactam ring of 3-HβL 3 to the corresponding β-amino acid 5.

Biologically derived 7,8-dihydropteridines are unstable in air, undergoing oxidation to a fully oxidized form that DHFS will not take as a substrate. As such, two stable pterin-mimic electrophiles derived from DHFR inhibitors methotrexate and trimethoprim were synthesized and ligated to 3-(p-aminophenyl)-3-HβL 2 using S_N2 displacement to form 3 and 4 (Scheme B,C). Finally, the β-lactam ring of 3 was hydrolyzed under acidic conditions to yield a ring-opened version of the potential inhibitor (5) as a negative control to demonstrate the importance of the 3-HβL ring structure to DHFS inhibition (Scheme D). Compounds 3, 4, and 5 were purified by preparative RP-C18 HPLC and lyophilized to provide suitable compounds for biochemical and whole cell assays. Two other synthetic routes to 3-(p-aminophenyl)-3-HβL 2 were attempted. The first route utilized a Norrish–Yang Type II photocyclization of a precursor (p-Cl-phenyl)-β-keto-formamide to create the 3-HβL ring. While this photocyclization was successful, we were unable to convert 3-(p-Cl-phenyl)-3-HβL to the desired 3-(p-aminophenyl)-3-HβL 2 via a Buchwald–Hartwig cross-coupling strategy (Figure S3). The second route attempted the addition of an aryl anion to an α-keto-β-lactam to form the corresponding 3-HβL ring. This route required protection of the β-lactam nitrogen, which proved to be difficult to remove (Figure S4).

Chemoenzymatic Synthesis of DHFS Inhibitors

With 3-(p-aminophenyl)-3-HβL 2 and its three pterin derivatives 3, 4, and 5 in hand, biochemical assays were developed to test their suitability as DHFS inhibitors. Benchtop synthesis of 7,8-dihydropteroate derivatives was not feasible due to rapid air oxidation during handling. Therefore, we developed a chemoenzymatic route to rapidly prepare compound 1 and use this in DHFS inhibition assays (Figure A). Performing enzyme-catalyzed reactions under an argon atmosphere enabled 6-hydroxymethyl-dihydropterin to be used as an enzyme substrate for DHPS with minimal competing air oxidation. DHPS is known to use sulfa-drugs as a competing substrate, leading to the formation of product inhibitors. This known substrate promiscuity was leveraged to test the ability of folate biosynthetic enzymes (HPPK and DHPS) to assemble a pterin-3-HβL molecule 1 utilizing 6-hydroxymethyl-dihydropterin and 3-(p-aminophenyl)-3-HβL 2 as substrates.

3.

Chemoenzymatic preparation of pterin-3-HβL 1 and the initial validation of DHFS inhibition. A) Chemoenzymatic route toward pterin-3-HβL 1 using HPPK/DHPS-catalyzed reactions and its DHFR-reduced form 14. B) LC-MS data showing the appearance of the molecular ion corresponding to the pterin-3-HβL 1 [M + H]⁺ ion (m/z = 356) when HPPK and DHPS are introduced. C) Steady-state enzyme kinetics for the production of PABA-derived product 7,8-DHP versus 3-(p-aminophenyl)-3-HβL 2 derived product pterin-3-HβL 1. Plot represents the consumption of NADPH as monitored by A₃₄₀ over time for the DHFR-catalyzed reduction of the HPPK/DHPS-derived product. D) Use of chemoenzymatically prepared pterin-3-HβL 1 as an inhibitor of DHFS-catalyzed conversion of THF to THF-Glu. The graph represents the apparent enzyme velocity for DHFS measured using a coupled enzyme assay for the conversion of THF to THF-Glu with DHFS, PK, and LDH. PK and LDH convert the DHFS-produced ADP to ATP with consumption of PEP and NADPH allowing for measurement of steady-state kinetics by monitoring A₃₄₀.

Reactions containing HPPK, DHPS, ATP, 6-hydroxymethyl-dihydropterin, and 3-(p-aminophenyl)-3-HβL 2 produced a strong peak in the LC-MS chromatogram corresponding to the expected molecular weight of the pterin-3-HβL 1 [M + H]⁺ molecular ion (m/z 356) (Figure B). Control reactions lacking HPPK and DHPS did not produce this unique metabolite. We conclude from this result that 3-(p-aminophenyl)-3-HβL 2 can serve as a substrate for DHPS similar to the misincorporation of the sulfa drugs. To compare the relative rates of substrate-to-product conversion for the natural substrate PABA and unnatural substrate 3-(p-aminophenyl)-3-HβL 2, we measured the apparent enzyme velocities of DHPS-catalyzed reactions under steady-state conditions. We coupled product formation to DHFR-catalyzed reduction of the products and monitored the decrease in optical absorbance at 340 nm, corresponding to the consumption of NADPH. Consistent with the results from the LC-MS analysis of HPPK/DHPS reactions, DHPS was able to accept 3-(p-aminophenyl)-3-HβL 2 as a substrate to produce the corresponding pterin-3-HβL 1 in significant quantities, although at a lower relative rate of production as compared to the PABA-derived product 7,8-DHP (Figure C).

Inhibition of DHFS

We used an in vitro PK/LDH-coupled assay to quantify the ADP-production in DHFS-catalyzed reactions. Because we could not purchase or make 7,8-DHP, we could not directly assay against the initial glutamylation reaction forming 7,8-DHF. Instead, we used fully reduced THF as the substrate and assay for poly glutamylation. First, we investigated DHFS inhibition by chemoenzymatically generated compound 1. We were unable to accurately determine the concentration of compound 1 due to spontaneous air oxidation and the small analytical quantities generated. However, we were still able to show dose dependent inhibition of DHFS by varying volumes of unknown concentration from the same stock solution, resulting in clear reductions in apparent enzyme velocity corresponding to increased inhibitor concentrations (Figure D).

Next, we investigated the dose-dependent inhibition of DHFS by synthetic 3-HβLs 3 and 4 and the control compound 5. A dose-dependent inhibitory response was shown for pterin-3-HβL 3 with an apparent IC₅₀ of 2.6 ± 1.6 μM (Figure A). We were unable to accurately fit the kinetic data from HβL 4 and hydrolyzed β-lactam 5 (the error associated with calculated IC₅₀ values was higher than the values themselves); however, we did observe a trend of decreasing enzyme velocity with increasing concentrations of compounds 4 and 5 that is suggestive of weak inhibition (Figure B,C). Pterin-3-HβL 3 produced 107 ± 39% inhibition of DHFS at 56 μM, while HβL 4 showed 34 ± 11% at 1024 μM, and hydrolyzed compound 5 achieved 35 ± 13% inhibition at 56 μM. Pterin-3-HβL 3 appears to be the best DHFS inhibitor, suggesting that DHFS favors the pteridine mimic from methotrexate over the 2,4-diaminopyrimidine from trimethoprim.

4.

In vitro inhibition of E. coli DHFS using the DHFS/PK/LDH coupled assay monitoring for inhibition of THF to Glu-THF. A) IC₅₀ curve for pterin-3-HβL 3 (apparent IC₅₀ of 2.6 ± 1.6 μM). B) IC₅₀ curve for pterin-3-HβL 4 (no accurate fit). C) IC₅₀ curve for hydrolyzed pterin-3-HβL 5 (no accurate fit). D) Relative % inhibition of compounds 3 (56 μM), 4 (1024 μM), and 5 (56 μM).

We validated the PK/LDH-coupled DHFS inhibition assay using LC-MS to detect the product formation from inhibitor-treated reactions (Figure C). We used single ion monitoring for the [M + H]⁺ molecular ion corresponding to Glu-7,8-DHF (m/z 573; observed), which consistently appeared as the dominant product ion from air oxidation of the Glu-THF product (m/z 575; not observed). LC-MS analysis of samples taken from these kinetic assays showed almost complete inhibition of Glu-7,8-DHF formation at the highest doses of pterin-3-HβL 3 (Figure A). Conversely, while hydrolyzed pterin-3-HβL 5 showed some decrease in product formation at the highest dose, it was clearly unable to completely stop DHFS from catalyzing its reaction (Figure B). These observations are consistent with the steady state enzyme inhibition results, which showed pterin-3-HβL 3 achieving near full DHFS inhibition at highest doses (Figure ).

5.

Extracted ion chromatograms (EICs) from LC-MS analysis of DHFS enzymatic conversion of THF to Glu-7,8-DHF (m/z = 573 for the [M + H]+ molecular ion). A) The near complete inhibition of DHFS-catalyzed product formation was observed with the highest dose of pterin-3-HβL 3 tested (56 μM). B) Decrease in the extracted ion counts of DHFS-catalyzed product formation in the presence of hydrolysis product 5 at 56 μM. C) Reaction scheme for the conversion of THF to Glu-THF followed by air oxidation to Glu-DHF. The graphs depict a representative EIC trace for reactions performed in triplicate as independent trials. Individual traces are presented as a stacked plot for visualization with an increasing offset of 10% of the maximum value on both the x- and y-axes.

Bacterial Growth Inhibition Studies

We used the Kirby–Bauer agar diffusion assay to test for bacterial growth inhibition by our synthetic antifolates. Both trimethoprim and sulfamethoxazole showed zones of growth inhibition for E. coli ATCC 25922, with the increasing size of the inhibition zone for higher doses (Figure S5). Pterin-3-HβL 3 showed no zone of inhibition beyond the DMSO control and showed no synergistic effects in combination with either trimethoprim or sulfamethoxazole. We attribute this lack of activity to a lack of bacterial cell permeability, which is common for antifolates including methotrexate. , Given that we could demonstrate the incorporation of 3-(p-aminophenyl)-3-HβL 2 into the natural folate biosynthetic pathway to generate pterin-3-HβL 1 (Figure ), we also attempted broth microdilution assays with E. coli ATCC 25922 treated with compound 2 as a potential prodrug for generating DHFS inhibitor 1 inside the cell. However, these experiments failed to show any bacterial growth inhibition or synergy with trimethoprim (data not shown).

Probing the Mechanism of DHFS Inhibition

The β-lactam motif presents an opportunity for covalent inhibition of DHFS through direct acylation of nucleophilic amino acid side chains. Should a phospho-3-HβL form in the DHFS active site, it could undergo assisted elimination to form an electrophilic quinone imine methide capable of alkylating amino acid side chains (Figure A). To investigate covalent DHFS inhibition, we incubated recombinant DHFS from E. coli with increasing concentrations of pterin-3-HβL 3 in the presence of ATP. We analyzed these samples by high-resolution intact protein mass spectrometry and searched for covalent adducts by monitoring the shift in apparent molecular weight of the DHFS molecular ions. We observed a strong molecular ion corresponding to the parent DHFS enzyme with an N-terminal hexahistidine tag appearing at m/z 47,495.5 Da (predicted 47,496.85; Figure S6). We did not observe any shift in apparent molecular weight with the inclusion of 3-HβL 3 and ATP. We also performed ³¹P NMR on similarly prepared samples and searched for potential ³¹P-signals corresponding to a phosphorylated inhibitor. We observed an increase in the ratio of P_i/ATP in samples containing inhibitor 3 under complete reaction conditions relative to the controls. A weak ³¹P-signal was also detected at ∼6.49 ppm that was unique to the experimental sample (Figure B; top spectra). This chemical shift is slightly downfield from the known ³¹P-signals for phosphoserine and phosphothreonine (∼5 ppm under similar conditions), but given the unique structure of the proposed phosphoester, the chemical shift is difficult to predict. , While the ³¹P-spectra are suggestive of a phosphorylation event, we were unable to directly detect any phosphorylated inhibitor species via LC-MS analysis of these samples. This could be due to instability of phospho-3-HβL 3 or in-source fragmentation during LC-MS analysis (Figure A).

6.

A) Proposed mechanism for the formation of a quinone imine methide intermediate and P_i from a 3-phospho-HβL inhibitor 3 followed by hydration to reform pterin-3-HβL inhibitor 3. B) ³¹P NMR (202 MHz, D₂O, pH 7.5) spectra of reaction mixtures containing DHFS, pterin-3-HβL 3, and ATP (top spectra) along with relevant control reactions (bottom three spectra). The P_i signal is set to a chemical shift of 0 ppm. The ³¹P resonance signals for ATP and ADP are labeled. An apparent new ³¹P resonance is observed at 6.49 ppm exclusively for the complete reaction mixture containing all of the necessary components (top spectra).

To further explore the potential for the formation of a phosphorylated inhibitor, we performed molecular docking of potential phosphorylated tetrahedral intermediates and potential phosphorylated inhibitors against a template DHFS structure from E. coli (PDB: 1W78) (Figure ). The template DHFS structure contains bound ligands for the reactive activated ester intermediate, 7,8-DHP acyl phosphate, and ADP (Figure A,B). We validated a molecular docking protocol using AutoDock Vina by redocking 7,8-DHP and comparing the ligand fit to the published crystal structure, which closely resembled the observed ligand conformation for 7,8-DHP acyl phosphate. We docked the putative tetrahedral intermediate, both R and S enantiomers, that would form after the reaction of 7,8-DHP acyl phosphate with the α-amino group of an l-Glu nucleophile (Figure C,D). We found that the tetrahedral intermediate with S stereochemistry was best accommodated in our docked structures (with the R stereochemistry ejecting the tetrahedral carbon into solvent space). There is one crystal structure of DHFS from Y. pestis with an occupied Glu-binding pocket (PDB entry 3QCZ). We found that the best fit for the Glu-side chain in our docked ligands occupied the crystallographically observed Glu-binding pocket. We docked an additional 24 compounds against the E. coli DHFS structural model. Docked structures include the reaction substrate, the tetrahedral intermediate, synthetic inhibitors, and theoretical inhibitors designed by using a structure-guided approach. Additional theoretical inhibitor structures were docked and are presented in Figure S7. The docked structures with the best fit were examined for protein–ligand interactions compared with 7,8-DHP (Figure ). Our hypothesis that the pterin-binding pocket can be used to anchor a potential inhibitor was bolstered by the docking results, with 24 of the 26 structures docked with the pterin mimic overlaying the 7,8-DHP pterin. Potential phosphorylation at the 3-hydroxy position results in a 0.7 kcal/mol increase in binding energy as compared to that of the free hydroxyl group (Figure A,B). The hydrolyzed 3-HβL ring of compound 5 binds in the same general conformation as the intact 3-HβL of compound 3, but with a 0.7 kcal/mol lower binding energy (Figure C). When the 3-hydroxyl group is phosphorylated, the conformation is maintained but with a reduction of 0.6 kcal/mol binding energy compared to the ring-closed form (Figure D). The stereochemistry at the C3 position of the 3-HβL proved pivotal when substituents are added to the 3-HβL. In particular, when the glutaric acid group of a modeled Glu is bound to N1 of 3-HβL and the 3-hydroxyl group is phosphorylated, there is a difference in calculated binding energy of 2.1 kcal/mol between the R and S (at the C3 position) structures. The docked R configuration orients the phosphate group toward the Mg²⁺ ions, while the S configuration at C3 of the 3-HβL ring forces this group out toward solvent rather than toward the Mg²⁺ ions, showing that stereochemistry at the C3 position is critical for facilitating the favorable phosphate–magnesium binding interactions (Figure E). The highest binding energy came from the phosphorylated inhibitor structure, with a l-glutaric acid group stemming from the N1 and R configuration at the C3 position (Figure E).

7.

7,8-Dihydropteroate phosphate and ADP bound to DHFS (PDB entry 1W78). A) Zoom in on the active site showing the pterin binding pocket anchoring the PABA acyl phosphate intermediate (cyan) toward the ADP (gray) and the two Mg²⁺ ions (magenta). B) The predicted active site interactions between phospho-7,8-DHP and DHFS. C) Zoom in on the active site showing the pterin binding pocket anchoring the docked tetrahedral intermediate (cyan) overlaid with the docked 7,8-DHP. D) The predicted active site interactions between the docked S enantiomer of the tetrahedral intermediate and DHFS. Docking was performed against PDB 1W78 as a template using AutoDock Vina.

8.

Selected compounds showing docked structures overlaid with 7,8-DHP in the active site of DHFS. A) The R enantiomer of pterin-3-HβL 3. B) The R enantiomer of pterin-3-HβL 3 with a phosphorylated 3-hydroxyl group. C) The R enantiomer of hydrolyzed 3-HβL 5. D) The R enantiomer of hydrolyzed 3-HβL 5 with a phosphorylated 3-hydroxyl group. E) The R enantiomer of phosphorylated pterin-3-HβL 3 with a glutaric acid linked to the N1 position of the β-lactam.

Discussion

In this work, we sought to repurpose 3-HβL as a general inhibitor of ATP-dependent carboxylate-amine ligases using DHFS as a model system. DHFS is also an attractive antibacterial target, given that it could serve as a replacement for the traditional antifolate target of the sulfa drugs, DHPS, which suffers from established clinical resistance. − , We modeled our study after the natural GS inhibitor TβL, a glutamate antimetabolite that mimics the phosphorylated transition state of the enzyme-catalyzed condensation of Glu γ-carboxy phosphate with ammonia (Figure ). We hypothesized that replacement of the PABA-carboxyl group of 7,8-DHP (or a structural mimic) with 3-HβL would result in a mechanism-based transition state inhibitor of DHFS following a mechanism analogous to the inhibition of GS by TβL (Figure ). We validated this hypothesis by synthesizing a pterin-linked 3-aryl-3-HβL (3) that exhibited potent DHFS inhibition in vitro. We validated the importance of the β-lactam ring by direct comparison to the hydrolyzed analog 5. ³¹P NMR and molecular docking were used to support a model for mechanism-based inhibition, where inhibitor phosphorylation at C3 of the 3-HβL generates a tetrahedral product that resembles the tetrahedral intermediate of the enzyme-catalyzed reaction (Figure ). This study provides the first proof-of-principle that the 3-HβL pharmacophore can be repurposed to inhibit diverse enzymes in the ATP-dependent carboxylate-amine ligase superfamily.

Other GS inhibitor “warheads,” including sulfoximines and phosphinates, have been repurposed to inhibit ATP-dependent carboxylate-amine ligases, including d-Ala-d-Ala ligase (DDL), MurC through MurF, Glu-Cys ligase, and DHFS. The 3-HβL pharmacophore offers several advantages over sulfoximines and phosphinates, including additional sites for functionalization (N1, C3, C4), control of stereochemistry, breadth of synthetic methods available for synthesizing β-lactams, , and charge neutrality (anionic phosphinates fail to permeate the bacterial cell envelope). Inhibitor design for DHFS requires consideration of the three-substrate (ATP, 7,8-DHP, and l-Glu) enzyme reaction mechanism. The reaction trajectory of the amine and carboxylate components must be reflected in the stereochemistry of the chiral tetrahedral intermediate, which resembles the transition states for the nucleophilic acyl substitution. Molecular docking suggests that the R-stereochemistry at C3 of the 3-HβL best aligns a 3-phospho-HβL toward the bis-Mg²⁺ catalytic center while positioning the pterin and Glu components in the structurally characterized binding cavities (Figure B). 3-HβL from TβL is ideal for mimicking the NH₃ nucleophile in the GS-catalyzed reaction, but it is likely that the l-Glu glutaric acid side chain is important for optimal DHFS binding. This glutaric acid group could be appended to N1 or C4 of 3-HβL, with N1 being favored by molecular docking (Figure E).

The aniline-3-HβL component synthesized in this work puts the 3-hydroxyl group in a benzylic position. The para-aniline activates the benzylic position, which provides potential access to a reactive quinone imine methide. The DHFS-catalyzed phosphorylation of a 3-aryl-3-HβL could promote phosphate elimination from the benzylic position, leading to quinone imine methide formation (Figure A). This hypothesis is supported by the observation that during the synthesis of the aniline-3-HβL 2, the hydrogenolysis of the nitro-aniline 8 spontaneously eliminates nitromethane from the benzylic position in a retro-Henry reaction, which can be suppressed by protonation of the aniline amino group by conducting the hydrogenolysis under acidic conditions.

Hydrolysis of a quinone methide resulting from the elimination of P_i from phospho-3-aryl-3-HβL could potentially regenerate the parent 3-aryl-3-HβL in solution, resulting in a futile cycle of ATP hydrolysis (Figure A). We provide some evidence for this via the observation that the ratio of P_i/ATP increases in the presence of DHFS, inhibitor 3, and ATP relative to that of control reactions (Figure B). We analyzed denatured DHFS-catalyzed reactions by LC-MS for phosphorylated inhibitors, covalent adducts, thiol adducts (DTT from buffer), and other potential breakdown products but did not observe any such compounds. In all cases, we only observe the parent 3-aryl-3-HβL inhibitor (3), which could arise from in-source fragmentation, the suggested quinone imine methide mechanism, or might suggest that the DHFS inhibition mechanism by 3-aryl-3-HβLs is reversible. There is one report of reversible DHFS inhibitors based on dihydropteroic acid analogues that showed weak in vitro inhibition in the 200–500 μM range. We observed more potent in vitro inhibition of E. coli DHFS with an apparent IC₅₀ of 2.6 ± 1.6 μM for pterin-3-HβL 3, suggesting that 3-HβL provides a significant enhancement in binding and inhibition. Further exploration of structure–activity relationships is needed to probe the inhibition mechanism. One interesting analogue to synthesize would be a 3-aryl-3-HβL inhibitor replacing the para-aniline with a methylene linker as found in trimethoprim to suppress the potential C3-phosphate elimination (Figure A).

The lack of whole cell activity suggests that the methotrexate-derived synthetic pterin-3-HβL 3 fails to permeate the E. coli cell envelope (Figure S5), which is a common challenge for antifolates. , We explored the possibility of using aniline-3-HβL 2 as a prodrug where DHPS would accept the PABA mimic as a substrate to form the corresponding pterin-3-HβL 1 in the cytoplasm of the bacterial cell, as observed for the sulfa drugs. While we did demonstrate that this chemoenzymatic conversion is possible in vitro (Figure A,B), aniline-3-HβL 2 still failed to show whole cell antibacterial activity against E. coli (not shown). Prodrug approaches using PABA mimics have been shown to have antifolate activity against Mycobacterium tuberculosis, so the potential for this therapeutic approach might be species-dependent.

Given that the relative rate of PABA incorporation by DHPS was significantly faster than the rate for aniline-3-HβL 2 (Figure C), we turned to the synthetic production of the complete DHFS inhibitor scaffolds. Pterin-3-HβL 3 is a tight-binding inhibitor of DHFS, and it has this capacity despite the fully oxidized nature of the pterin mimic, which is known to reduce affinity for enzymes in the folate pathway. DHFS naturally prefers the half-reduced pterin form of the 7,8-DHP substrate in the initial Glu condensation reaction and prefers the fully reduced pterin form of THF in the poly-Glu condensation steps. A fully reduced form of the pterin-3-HβL 3 might improve the DHFS binding affinity and inhibition potency. Indeed, the two previously studied DHFS inhibitors based on aryl phosphinates showed nanomolar activity when their respective pterin was partially reduced prior to the assay. We attempted chemical reduction of pterin-3-HβL 3 with ascorbic acid but found these conditions to be incompatible with the PK/LDH ADP detection system used in this work, as ascorbic acid inhibits LDH. ,

The pterin mimic derived from trimethoprim proved unsuitable to produce a DHFS inhibitor (Figure B,D), potentially indicating that the size of the pterin mimic and spacer is vitally important to position 3-HβL correctly in the DHFS active site. Docking of these trimethoprim-derived HβLs supported this hypothesis, as the pterin-mimic finds the pterin-binding pocket, but the spacer is too short to properly orient 3-HβL near the two active site Mg²⁺ atoms (Figure S7), instead of pushing the 3-HβL out toward solvents. Clearly, the choice of pterin mimic is an important consideration when designing DHFS inhibitors, and it cannot be assumed that DHFR inhibitors such as trimethoprim can be repurposed as DHFS inhibitor scaffolds. The importance of 3-HβL itself is shown when the ring is hydrolyzed, and the inhibitory activity drastically decreases (Figure C,D). The potential zwitterionic nature of the hydrolyzed β-lactam compound 5 might also indicate that the electronics near the active site magnesium ions play a vital role in binding. The β-lactam ring of pterin-3-HβL 3 was found to be stable in the reaction buffer used for in vitro DHFS inhibition assays, according to LC-MS analysis. Hence, we conclude that the 3-HβL ring enhances binding and inhibition of DHFS.

Conclusion and Future Directions

The 3-HβL of TβL presents a potential general warhead to inhibit the superfamily of ATP-dependent carboxylate-amine ligase enzymes. The 3-HβL was successfully installed on a molecule that binds to DHFS and inhibits its enzymatic activity. The 3-HβL moiety itself was shown to be important for DHFS binding, but the reasons for this importance still need to be elucidated. The lack of whole cell bacterial growth inhibition for the newly discovered DHFS inhibitor is a roadblock for current inhibitor design. Further investigation into cell permeability will be required to guide inhibitor modifications that increase potency in whole cell assays. With the knowledge that the 3-HβL warhead is a potent pharmacophore for GS and DHFS, this inhibitor strategy could be extended to related enzyme families. While there are a multitude of ATP-dependent carboxylate-amine ligase enzymes, one target with high potential is DDL. , While there is an inhibitor of DDL used clinically to treat tuberculosis, d-cycloserine, it is dose-limited by toxicity. − Similar to GS and DHFS, phosphinate inhibitors of DDL have also been reported. , As such, a 3-HβL-containing compound could potentially inhibit DDL due to its similarity in mechanism to both GS and DHFS, as well as its endogeny to bacteria similar to DHFS.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsbiomedchemau.5c00036.

• Synthetic procedures, Figures S1–S33, Tables S1–S9, and NMR spectra (PDF)

The authors declare the following competing financial interest(s): A provisional patent application (18/297,368) describing these inhibitors has been filed through the Washington University in St. Louis Office of Technology Management.