Identification and Biochemical Characterization of Pyrrolidinediones as Novel Inhibitors of the Bacterial Enzyme MurA

Reem K Fathalla; Wolfgang Fröhner; Chantal D Bader; Patrick D Fischer; Charlotte Dahlem; Deep Chatterjee; Sebastian Mathea; Alexandra K Kiemer; Haribabu Arthanari; Rolf Müller; Mohammad Abdel-Halim; Christian Ducho; Matthias Engel

doi:10.1021/acs.jmedchem.2c01275

. Author manuscript; available in PMC: 2023 Nov 10.

Published in final edited form as: J Med Chem. 2022 Oct 21;65(21):14740–14763. doi: 10.1021/acs.jmedchem.2c01275

Identification and Biochemical Characterization of Pyrrolidinediones as Novel Inhibitors of the Bacterial Enzyme MurA

Reem K Fathalla ¹, Wolfgang Fröhner ¹, Chantal D Bader ^2,³, Patrick D Fischer ^1,^4,⁵, Charlotte Dahlem ⁶, Deep Chatterjee ⁷, Sebastian Mathea ⁷, Alexandra K Kiemer ⁶, Haribabu Arthanari ^4,⁵, Rolf Müller ^2,^3,⁸, Mohammad Abdel-Halim ⁹, Christian Ducho ^1,^*, Matthias Engel ^1,^*

PMCID: PMC9989942 NIHMSID: NIHMS1874881 PMID: 36269107

Abstract

To develop novel antibiotics, targeting the early steps of the cell wall peptidoglycan biosynthesis seems to be a promising strategy that is still under-utilized. MurA, the first enzyme in this pathway, is targeted by the clinically used irreversible inhibitor fosfomycin. However, mutations in its binding site can cause bacterial resistance. We herein report a series of novel reversible pyrrolidinedione-based MurA inhibitors that equally inhibit wild type (WT) MurA and the fosfomycin-resistant MurA C115D mutant, showing an additive effect with fosfomycin for the inhibition of WT MurA. For the most potent inhibitor 46 (IC₅₀ = 4.5 μM), the mode of inhibition was analyzed using native mass spectrometry and protein NMR spectroscopy. The compound class was non-toxic against human cells and highly stable in human S9 fraction, human plasma and bacterial cell lysate. Taken together, this novel compound class might be further developed towards antibiotic drug candidates that inhibit cell wall synthesis.

Graphical Abstract

graphic file with name nihms-1874881-f0001.jpg

Introduction

The rapidly rising issue of bacterial resistance to many known antibiotics has necessitated the discovery of antibacterial drug classes with novel mechanisms of action.^1-4 Peptidoglycan is an essential part of prokaryotic cell walls, playing an integral role in preserving the integrity and structure of bacterial cells.^5,6 Consequently, its biosynthesis has been the target of many efforts for antibacterial drug discovery.^7-9 The early cytoplasmic steps of peptidoglycan biosynthesis in particular have recently been considered with growing interest as potential targets, since the older, more established β-lactam antibiotics mainly acting in the bacterial periplasm, which is more accessible to small molecules, suffer from extensive bacterial resistance.^10-12 The early steps of peptidoglycan synthesis are performed by the Mur family of enzymes, comprising six intracellular enzymes (MurA to F) that furnish the main cytosolic peptidoglycan precursor uridine-5-diphosphate-N-acetylmuramyl-pentapeptide (Park's nucleotide)^7,9,10

MurA is the first enzyme of this cascade and catalyzes the conversion of UDP-N-acetylglucosamine (UDP-GlcNAc, UNAG) into UDP-N-acetylglucosamine enolpyruvate (UNAGEP) through addition of an enolpyruvate moiety from the co-substrate phosphoenolpyruvate (PEP, Figure 1).^10,13,14 It is an essential enzyme that is conserved across both Gram-positive and Gram-negative bacteria with no mammalian homolog.^15,16 Hence, inhibitors directed against bacterial MurA have potential to act as broad spectrum antibiotics without affecting a related enzyme in humans. At present, the only approved antibiotic targeting MurA is fosfomycin, which has been in clinical use since the early 1970s for the treatment of uncomplicated urinary tract infections and pediatric gastrointestinal infections resulting from Shiga toxin-producing Escherichia coli (STEC) in Japan.^9,17-19 Fosfomycin acts as an analogue of PEP that competitively inhibits MurA by covalent alkylation of Cys115 (E. coli numbering), the main catalytic residue involved in PEP binding (Figure 1).^19-22 However, there is a high incidence of resistance to fosfomycin through decreased uptake,^23,24 enzymatic modification of the antibiotic,^25,26 MurA overexpression,^27,28 and finally mutation of the key Cys residue to Asp. This MurA mutant is naturally present in several bacteria such as Mycobacterium tuberculosis (M. tuberculosis).^29,30

Figure 1. — A. MurA-catalyzed reaction. B. Mechanism for the inhibition of WT MurA by fosfomycin.

Screening campaigns to identify MurA inhibitors (Figure 2) have been performed by various groups and led to the discovery of several low micromolar inhibitors including avenaciolides (I),²⁰ tulipalines (II),³¹ the sesquiterpene lactone cnicin (III),^32,33 the cyclic disulfide RWJ-3981 (IV),³⁴ and ebselen (V).³⁵ These inhibitors usually possess highly electrophilic moieties that target MurA by irreversible interactions in its active site through several mechanisms ranging from reaction of the Cys key residue in a Michael addition such as in the case of I, and II, (avenaciolides and tulipalines respectively)^20,31,36 to formation of a suicide inhibitor of the MurA active site via covalent binding to UNAG with III (cnicin).^32,33 V (ebselen) targets MurA through covalent binding to the active-site Cys key residue,³⁵ and IV (RWJ-3981) acts by irreversible non-covalent attachment near the PEP binding site.^37,38 While targeting the catalytic Cys residue in the active site results in effective inhibition of MurA, this mechanism also implies that most of these compounds are inactive against the naturally occurring MurA C115D mutant.^31,32,35,37 In addition, several of these inhibitors have been shown to react non-specifically with thiol-containing compounds such as dithiothreitol (DTT) and also the cellular antioxidant glutathione (GSH).^20,31,35,37 Moreover, the highly electrophilic structures also raise concerns about their specificity and safety as they have been shown to undergo several off-target interactions such as the inhibition of glutamate transport in rat liver mitochondria (III, avenaciolides)^39-41 and the nonspecific inhibition of several bacterial targets in addition to failing to protect mice from death in a lethal Staphylococcus aureus (S. aureus) infection model (IV, RWJ-3981).³⁷ Therefore, it is of great importance to develop new scaffolds that inhibit MurA specifically and have proven safety on mammalian cells.

In a high-throughput screening campaign, Miller and coworkers identified benzothioxalones (Figure 2, VI) as MurA inhibitors, displaying IC₅₀ values against MurA between 0.25 and 51 μM. The compounds were also found to form covalent adducts due to reaction of Cys115 with the cyclic thiocarbonate unit.⁴² Consequently, binding was not detected with the C115D mutant MurA protein. Additionally, the cyclic thiocarbonate motif might react with other enzymes having a reactive cysteine in the active site, similar to the aforementioned Michael acceptor-type inhibitors.

Another class of reported inhibitors are pyrazolidinediones (Figure 2, VII, and VIII) that act as dual MurA/B inhibitors with MurB (i.e., the subsequent enzyme from the peptidoglycan biosynthesis cascade) as their main target.^6,43-45 We had an in-house library of aryl-substituted pyrrolidinediones available that had formerly provided moderately active inhibitors of active atypical protein kinase C (PKC).⁴⁶ The structural similarity of this compound class to the aforementioned pyrazolidinedione-type MurA/B inhibitors prompted us to initiate a screening campaign for MurA inhibition with some selected pyrrolidinediones from our in-house library. In this study, we report the unprecedented identification of a diarylpyrrolidinedione scaffold that inhibits MurA without any reaction with the key active-site Cys115 residue. Therefore, inhibitory activities towards the E. coli MurA C115D mutant and the wild type (WT) MurA enzyme were identical. As this new class of MurA inhibitors also showed no toxicity to human cell lines, it overcomes many limitations of previously reported MurA inhibitors and might serve as a useful starting point for the development of novel antibiotics targeting MurA.

Results and Discussion

Chemistry

The synthesis of all target pyrrolidinediones 1–46 (Table 1) was carried out using an efficient three-component one-pot procedure (Scheme 1) that had been described by Merchant and Schiff.^47,48 A major advantage of this method was that all three required reagents were usually commercially available in a large variety. For a first series of target compounds, several amines (1–7a, 14a, and 15a) were reacted with aldehydes (1–7b, 14b, and 15b) to furnish the respective imines (1–7c, 14c, and 15c), which were converted to pyrrolidinediones 1–7, 14 and 15 with tricarbonyl reagents (1–7e, 14e, and 15e) generated by Claisen condensation with diethyl oxalate. For a second series of target pyrrolidinediones 8–13 and 16–46, the same protocol was applied using aminobenzothiazoles as amine components (8–13a, and 16–46a). The depicted structure of pyrrolidinediones 1–7, 14 and 15 with residues R¹, R² and R³ (Scheme 1, Table 1) will subsequently be referred to as ‘general structure’ of the pyrrolidinedione scaffold when structure-activity relationship data are discussed (vide infra).

Table 1.

Structures and substitution patterns of the synthesized pyrrolidinedione derivatives.


No.	R¹/R⁴	R²	R³	No.	R¹/R⁴	R²	R³
1		4-Cl Ph	Et	24	5,6-diMe	Biph	Et
2		4-Cl Ph	Et	25	6-Me	4-OH Ph	Et
3	Biph	Ph	Et	26	6-Me		Et
4	4-Cl Ph	4-Cl Ph	Et	27	6-Me		Et
5	Ph	Biph	Et	28	6-Me		Et
6		Biph	Et	29	6-Me		Et
7		Biph	Et	30	6-Me		Et
8	H	Ph	Et	31	6-Me		Et
9	4-Me	Ph	Et	32	6-Me		Et
10	6-Me	Ph	Et	33	6-Me		Et
11	4-Cl	Ph	Et	34	6-Me		Et
12	6-Cl	Ph	Et	35	6-Me		Et
13	O-Phe	Ph	Et	36	6-Me		Et
14		Ph	Et	37	6-Me		Et
15		Ph	Et	38	6-Me		Et
16	H	Biph	Et	39	6-Me		Et
17	6-Br	Biph	Et	40	6-Me		Et
18	6-CF₃	Biph	Et	41	4-Me	Ph	tBu
19	6-Me	Biph	Et	42	6-Me	Ph	tBu
20	6-OMe	Biph	Et	43	6-Cl	Ph	tBu
21	4-Cl	Biph	Et	44	6-Me		EtOH
22	4-OMe	Biph	Et	45	6-Me		Me
23	4,6-diF	Biph	Et	46	6-Me		tBu

Open in a new tab

Scheme 1. — Synthesis of pyrrolidinediones **1–46** (see Table 1 for exact structures).

Biological Evaluation

Inhibition of MurA

We started the screening campaign with some selected compounds (Table 2) from our in-house library of aryl-substituted pyrrolidinediones against E. coli MurA. For the initial screening, we selected a small set of compounds covering the chemical diversity present in the positions R¹ and R² of the 'general structure' (vide supra, Table 1). The activity of MurA was tested using a standard malachite green assay measuring the amount of released inorganic phosphate using malachite green and sodium molybdate. Together, the phosphate and these reagents form a green complex that is then quantified spectrophotometrically.⁴⁹ When tested at a fixed concentration of 20 μM in this in vitro MurA assay, we observed inhibitory activity for several compounds, dependent on the type of the substituents (Table 2). In particular, compound 7 showed a remarkable activity with an IC₅₀ of 5 μM, thus being equipotent to the established MurA inhibitor fosfomycin under the assay conditions – notwithstanding that the IC₅₀ values are only comparable to a limited extent because of the irreversible binding mode of fosfomycin. Altogether, the aryl-substituted pyrrolidinedione system proved to be a new scaffold for the development of MurA inhibitors.

Table 2.

Results from the screening of the initial set of aryl-substituted pyrrolidinediones against MurA.


Compound	R¹	R⁵	Inhibition at 20 μM ± SD, (IC₅₀ (μM) ± SD)^a
1		Cl	34 ± 6%
2		Cl	15± 9%
3	Biph	H	19 ± 1%
4	4-Cl Ph	Cl	14 ± 3%
5	Ph	Ph	13 ± 6%
6		Ph	40 ± 6%
7		Ph	78 ± 6% (5.1 ± 0.4)
fosfomycin	-----	-----	78 ± 3% (5.3 ± 1.7)

Open in a new tab

IC₅₀ values were determined for all compounds that showed more than 50% inhibition at 20 μM

In the following rounds of screening, we explored the effect of changing one structural feature of the scaffold while preserving the others to derive structure-activity relationships (SAR) for each feature. As observed with 7, a 6-chlorobenzothiazole unit conveyed the strongest MurA inhibition from all the different aryl moieties in the N1-position (R¹ in the general structure). Therefore, further benzothiazole motifs at this position were explored in the next round of screening, in combination with either a plain phenyl or a biphenyl side chain at position 2 (R² in the general structure, Table 3).

Table 3.

MurA inhibition of benzothiazole-substituted pyrrolidinediones (variations in position R¹ of the general structure).


Compound	R⁴	R⁵	Inhibition at 20 μM (± SD)	IC₅₀ (μM) (± SD)^a
8	H	H	0	n.d.
9	4-Me	H	0 ± 2%	n.d.
10	6-Me	H	26 ± 4%	n.d.
11	4-Cl	H	18 ± 4%	n.d.
12	6-Cl	H	14 ± 7%	n.d.
13	6-OPh	H	53 ± 8%	17 ± 3
14	-----	-----	14 ± 6%	n.d.
15	-----	-----	36 ± 3%	n.d.
16	H	Ph	45 ± 1%	n.d.
17	6-Br	Ph	52 ± 2%	21 ± 1
18	6-CF₃	Ph	70 ± 5%	6.1 ± 0.7
19	6-Me	Ph	76 ± 4%	8.5 ± 1.9
20	6-OMe	Ph	51 ± 4%	22 ± 8
21	4-Cl	Ph	73 ± 3%	7.0 ± 0.7
22	4-OMe	Ph	31 ± 3%	25 ± 2
23	4,6-diF	Ph	76 ± 5%	7.5 ± 0.4
24	5,6-diMe	Ph	85 ± 2%	5.8 ± 0.7

Open in a new tab

IC₅₀ values were determined for all compounds that showed more than 50% inhibition at 20 μM; n.d. = not determined.

The results of the second round of screening revealed that the presence of a biphenyl moiety (R⁵ = phenyl) was a major determinant of activity (compounds 16 -14, Table 3); compound 13 was the only compound lacking the additional phenyl at R⁵ to show pronounced activity (IC₅₀ = 17 μM), probably because the 6-phenoxy substituent at R¹ compensated for the absence of the biphenyl system by establishing additional aromatic interactions. Among the biphenyl compounds, an additional gain in activity of up to five-fold was achieved by the substituent R¹ at the benzothiazole. Generally, a methyl or chloro substituent at the 4- or 6-position of the benzothiazole was favorable for MurA inhibition as seen with compounds 7 (Table 2), 19 and 21 (Table 3). Bulkier substituents such as bromo or methoxy groups were slightly detrimental to the activity as demonstrated when comparing compounds 17 (6-Br) and 20 (6-OMe) with compound 19 (6-Me).

This strong dependency of the activity on the aromatic moiety at R² led us to explore a broader range of differently substituted phenyl, various simple heteroaryl and extended biaryl and naphthyl derivatives (Table 4). The 6-methylbenzothiazole motif at position R¹ was kept constant in this series.

Table 4.

MurA inhibition of benzothiazole-aryl-substituted pyrrolidinediones (variations in position R² of the general structure).


Compound	R²	Inhibition at 20 μM (± SD)	IC₅₀ (μM) (± SD)^a
25		23 ± 13%	n.d.
26		21 ± 1%	n.d.
27		14 ± 2%	n.d.
28		29 ± 2%	n.d.
29		44 ± 3%	n.d.
30		43 ± 4%	n.d.
31		54 ± 4%	15 ± 3
32		74 ± 2%	9.5 ± 0.3
33		12 ± 4%	n.d.
34		57 ± 5%	17 ± 3
35		36 ± 3%	n.d.
36		37 ± 4%	n.d.
37		75 ± 3%	9.9 ± 1.2
38		94 ± 2%	4.8 ± 0.2
39		72 ± 1%	9.9 ± 2.1
40		50 ± 6%	22 ± 5

Open in a new tab

IC₅₀ values were determined for all compounds that showed more than 50% inhibition at 20 μM; n.d. = not determined.

This round of screening confirmed that a more extended aromatic system at R², e.g. the biaryl moieties in compounds 19 (Table 3), 38 and 39 (Table 4), enhanced the inhibitory activity against MurA. As seen with the 5-aryl-furan-2-yl derivatives 36–38, further lipophilic substituents on the phenyl ring modulated the activity. The disubstitution pattern in 38 led to the strongest increase of potency, however, this was at the cost of increased lipophilicity, which in general was the main problem with the extended aryl systems at R². Unfortunately, the more polar pyridazine analog 35 was much less active, possibly indicating the correlation of potency with the degree of lipophilicity.

In the last round of screening, the ethyl ester side chain in position R³ of the general structure was changed to other alkyl esters (Table 5). It was found that a t-butyl ester led to slightly or even pronounced increases of inhibitory activity (41 vs. 9, 42 vs. 10, and 43 vs. 12, Table 5). This is most clearly demonstrated when comparing compounds 9 and 41, where even in absence of an extended aryl system in position R², compound 41 showed an IC₅₀ value of 11 μM. However, we still observed a dependency on the substitution pattern of the benzothiazole unit, as there was only weak additive enhancement of activity for compounds 42 and 43 over compounds 10 and 12, respectively. In the case of the rather active 5-aryl-furan-2-yl derivatives, the ester side chain also had a noticable influence on the activity (compounds 44–46, Table 5). Changing the ethyl to the more hydrophilic 2-hydroxy ethyl ester in compound 44 led to a significant decrease in activity, suggesting that the alkyl residue interacts with MurA in a hydrophobic environment. The dependency of potency of the derivative on the size of the ester moiety is confirmed by the fact that changing the ethyl ester of 37 (Table 4) to a methyl congener (45, Table 5) resulted in decreased inhibition. We achieved maximum MurA inhibition with the bulkier t-butyl ester in compound 46 (Table 5) as the most active derivative in the series alongside 38 (Table 4), which had been found to be too lipophilic (cf. Table S3 for more information regarding the lipophilic properties and ligand efficiency of some selected compounds).

Table 5.

MurA inhibition of benzothiazole-substituted pyrrolidinediones with different ester moieties (variations in position R³ of the general structure).


Compound	R⁴	R³	Inhibition at 20 μM (± SD)	IC₅₀ (μM) (± SD)^a
41	4-Me	tBu	67 ± 5%	11 ± 1
9 ^b	4-Me	Me	0 ± 2%	n.d.
42	6-Me	tBu	38 ± 2%	n.d.
10 ^b	6-Me	Me	26 ± 4%	n.d.
43	6-Cl	tBu	38 ± 14%	n.d.
12 ^b	6-Cl	Me	14 ± 7%	n.d.
44	Me	2-OH-Et	43 ± 5%	n.d.
45	Me	Me	59 ± 5%	16 ± 5
46	Me	tBu	88 ± 3%	4.5 ± 0.5

Open in a new tab

IC₅₀ values were determined for all compounds that showed more than 50% inhibition at 20 μM

values for compounds 9, 10, and 12 were taken from Table 2 for comparison; n.d. = not determined.

Taken together, the results obtained for pyrrolidinediones as potent MurA inhibitors with a clear SAR demonstrated that the re-screening of a compound library for activities against microbial targets can be a viable option in the discovery process for potential antibacterial agents. Compound 46 was chosen as the best-in-class derivative of the series for more detailed subsequent studies. Keeping in mind that the compounds are structurally similar to the published pyrazolidinedione type of dual MurA/MurB inhibitors (e. g., VII and VIII in Figure 2), we also tested compound 46 in a specific MurB assay. However, we found only a weak activity against MurB (46% inhibition at 20 μM), confirming MurA as being the major target of compound 46 in the early steps of peptidoglycan synthesis.

Compound 46 showed equal potency against the clinical MurA C115D mutant

With respect to the frequent occurrence of drug-resistant mutants, it is advantageous for new inhibitors of established bacterial targets to address alternative binding sites on such a target. Therefore, we tested compound 46 for its inhibitory properties in vitro towards the C115D MurA mutant, which shows resistance against fosfomycin.³⁰ Remarkably, we obtained an IC₅₀ value of 4.9 μM, which was identical to the IC₅₀ value against WT MurA (4.5 μM, Table 5). This demonstrated that 46 inhibited MurA via an alternative binding mode or a different binding site, and that the pyrrolidinedione scaffold might have the potential to break fosfomycin resistance.

Reversibility of MurA binding of compound 46

Compound 46 was tested for its binding reversibility to MurA, using fosfomycin as a control. When employed in a dilution assay, 46 was shown to be a reversible binder, as it lost its MurA inhibitory activity when the assay mixture was diluted after an initial preincubation period with the enzyme. This was different from the irreversible inhibitor fosfomycin which retained its inhibitory effect even after the 1:50 dilution step (Figure 3). For these experiments, three assay conditions were used: a 1:50 diluted assay (A) for which the assay mixture with MurA, UNAG and the respective compound (10 μM initial concentration) was diluted to reach a 0.2 μM final concentration of the inhibitor; a concentrated assay (B) with a 10 μM final concentration of both compounds; and a directly diluted assay (C) without a preincubation step with a final compound concentration of 0.2 μM.

Figure 3. — Assays on the reversibility of binding of fosfomycin and compound 46 to MurA. The data shown are the mean of three independent experiments, and the error bars represent ± SD. A. Diluted assay with a final compound concentration of 0.2 μM; B. Concentrated assay performed under the same conditions as the diluted assay, but compound concentrations were adjusted to reach a final concentration of 10 μM; C. The assay conditions from A were directly reproduced, but without a preincubation step.

Analysis of potential cooperative effects of compound 46 and fosfomycin

The observation that the potency of 46 was not reduced with the fosfomycin-insensitive C115D mutant of MurA may suggest that both compounds might target different sites. To investigate this further, we looked into a potential synergism of inhibition between both inhibitors. Indeed, compound 46 showed an additive effect with fosfomycin when they were applied together in the enzymatic assay (Figure 4). This additive effect disappeared, but the inhibition by 46 was retained, when the same assay was performed using the fosfomycin-resistant C115D MurA mutant. This demonstrated the potential of a combination treatment approach with a pyrrolidinedione MurA inhibitor (such as 46) and fosfomycin and also confirmed that the binding site of 46 did not significantly overlap with that of fosfomycin. More importantly, simultaneous blocking of MurA at two different sites might impede the occurrence of resistant mutants in a clinical setting, which is a major advantage for our suggested combination.

Figure 4. — Additive effect of the coadministration of compound 46 and fosfomycin on WT MurA. Fosfomycin had no effect on C115D MurA while compound 46 retained its activity. The data shown are the mean of three independent experiments, and the error bars represent ± SD; final concentrations of both compounds were 2 μM.

Investigation of the mode of inhibition of compound 46

We first performed an enzyme kinetics study of the binding of 46 to MurA. While a strict Michaelis-Menten model probably could not be fully applied to MurA (as the enzyme has two substrates), we attempted to study the interplay of the binding of 46 and the concentration of UNAG in the assay. This was done by modifying the MurA malachite green assay, which is an endpoint assay, to a continuous assay measured over a time period of 30 minutes. For each tested concentration of 46, four different concentrations of UNAG were examined (25, 100, 250 and 363 μM). The Malachite green absorbance vs. time was plotted for each UNAG concentration, and the slope was calculated for each resultant line. The Lineweaver-Burk plot (Figure 5) was then obtained by plotting the reciprocal of the slope (representing 1/V_o) vs. the reciprocal of the UNAG concentration (representing 1/S). The lines for each concentration of 46 were almost parallel to each other, indicating an uncompetitive binding.⁵⁰ So far, the enzyme kinetics data seemed to support a binding of 46 to the MurA-UNAG complex, thus preventing the formation of the MurA product UNAGEP.

Figure 5. — Lineweaver-Burk plot for the relationship of MurA activity and UNAG concentration at different concentrations of 46. The data shown are the mean of three independent experiments, and the error bars represent ± SD. Four different concentrations of UNAG were tested: 25, 100, 250, and 363 μM. The measurements were taken at 5, 10, 20, and 30 min.

However, we contemplated that the enzyme kinetics data might also be consistent with the need for UNAG to 'unlock' the enzyme for binding of 46. According to previous reports, recombinantly overexpressed MurA was isolated in the 'closed' UNAM- and PEP-bound form, which persisted throughout the purification procedures (Figure S1A).^22,51,52 It was further reported that increasing concentrations of UNAG could displace the UNAM, forming a binary complex which can enter the catalytic cycle by additionally binding PEP or partially exist in equilibrium with the 'open' conformational state of the enzyme (Figure S1B);^22,51,52 the latter would then be competent for binding of 46. UNAG might be bound or not to the MurA complex with 46, which would not be distinguishable by the enzyme kinetics analysis. Therefore, we decided to use an alternative experimental approach, i.e., non-denaturing ('native') ESI-MS, to analyze the mode of inhibition with respect to UNAG. Native MS preserves the noncovalent interactions between enzymes and their binding partners.⁵³ In addition, denaturing ('non-native') ESI-MS of the protein was also performed to obtain the molecular weight of the ligand-free protein.

The obtained results from the native MS experiments (Table 6) reveal that MurA had indeed been expressed in the closed UNAM- and PEP-bound form, showing an equivalent mass to that ternary complex in addition to sodium (Table 6, entry 1, and Supporting Information, Figure S2A). Upon incubation with UNAG, this complex apparently opened up and the mass of bound UNAG in addition to inorganic phosphate (P_i) was observed (entry 2, and Figure S2B). The presence of inorganic phosphate could be explained by the initial enzymatic reaction that would occur when the intrinsically bound PEP molecule is converted with the added UNAG to furnish inorganic phosphate as a by-product. Subsequently, in absence of additional PEP, P_i is not released while the enzyme binds another UNAG, most likely leading to the formation of the observed complex. Two further native MS experiments involved the addition of 46 to the enzyme in both the absence and presence of UNAG. In the absence of UNAG, we again observed the closed enzyme complex with UNAM and PEP (Table 6 entry 3, and Figure S2C). In contrast, when the enzyme was first preincubated with UNAG for 15 minutes prior to addition of 46, we observed the additional masses of bound 46, the enzymatic product P_i as before, and in addition, an acetate ion from the dialysis buffer (entry 4, and Figure S2D). This finding led us to reconsider the hypothesis of uncompetitive inhibition of compound 46, as the mass of the expected enzyme-UNAG-46 complex was not detected. Retrospectively, the results from the kinetics experiments may be explained by the necessity of the presence of UNAG for unlocking the enzyme structure before the binding of 46 can occur. This means that compound 46 is not simply an uncompetitive MurA inhibitor, but rather a more complex mechanism of action is involved in its binding.

Table 6.

Results from non-denaturing (native) ESI-MS on MurA-ligand complexes.

#	Conditions	M_av (Da)^a	ΔM (Da)^b	Bound ligands	M_{bound ligands} (Da)
1	MurA	48423	870	PEP, UNAM, Na⁺	870
2	MurA + UNAG	48257	704	UNAG, P_i	705
3	MurA + 46	48401	847	PEP, UNAM	847
4	MurA + UNAG + 46	48232	679	46, P_i, OAc⁻	679

Open in a new tab

M_av is the average detected mass by native MS

The mass of the ligand-free protein (determined by non-native MS) was 47553 Da.

Binding analysis by NMR spectroscopy

Numerous attempts to obtain an X-ray co-crystal structure of MurA bound to inhibitor 46 were ultimately unsuccessful (data not shown). This might have resulted from MurA being intrinsically expressed in a closed PEP-UNAM bound form, as it had been reported before⁵¹, and is in agreement with our native MS experiments. This closed ligand-bound conformation is very robust to dilution and does not convert into the open (ligand-free) conformation except in the presence of the enzyme substrate UNAG. Attempts to form crystals of the ligand-free apo enzyme were also unsuccessful, and therefore, we ultimately could not obtain a co-crystal with 46 bound to MurA. The X-ray crystallography experiments only confirmed that UNAM and PEP were tightly bound to overexpressed MurA, where the crystals corresponded to PDB entry 3SWD (data not shown).⁵¹

Hence, to elucidate the binding mode of inhibitor 46 to MurA in comparison to fosfomycin, we attempted protein-detected NMR spectroscopy in the presence and absence of UNAG, with 46 and fosfomycin. U-¹⁵N-labelled wild type MurA and C115D MurA were expressed in E. coli, and ¹⁵N-¹H TROSY (transverse relaxation-optimized spectroscopy) HSQC spectra were recorded with 200 μM protein. The addition of UNAG at 500 μM resulted in significant chemical shift perturbations (CSPs) of a subset of peaks (Supporting Information, Figure S3). WT MurA (in the presence of 500 μM UNAG) was incubated with 500 μM fosfomycin or 500 μM compound 46 (WT MurA+UNAG+fosfomycin and WT MurA+UNAG+46, Figure 6). When comparing the CSPs and peak intensity reductions (arising from binding in a fast or intermediate exchange regime, respectively) of WT MurA+UNAG+fosfomycin with WT MurA+UNAG+46, it was noted that a similar subset of peaks was affected by small molecule binding in both cases, indicating binding of both compounds to a similar region of the protein. However, it should be noted that not all peaks affected by fosfomycin are affected by compound 46 and vice versa, suggesting that the binding sites meet in a certain region but do not superimpose.

Figure 6. — ¹⁵N-¹H TROSY HSQC spectra of WT MurA bound to UNAG (red), overlaid with WT MurA (UNAG-bound) in the presence of fosfomycin (green, top spectrum) or compound 46 (blue, bottom spectrum). Zoom in on selected regions highlights peaks that undergo chemical shift perturbations (CSPs).

When comparing the spectra for WT MurA and C115D MurA in the presence of UNAG (WT MurA+UNAG and C115D MurA+UNAG), the majority of CSPs were found to be identical, except for some differences which identified the peaks representing the region around Cys115 in WT MurA (Supporting Information, Figure S4). Additionally, in accordance with the results from enzyme kinetics, X-ray crystallography, and native MS data, we observed no CSPs caused by 46 in the absence of UNAG (WT MurA+46), confirming that specific interaction by this inhibitor with MurA requires the open form of the enzyme (Figure S5).

To further investigate the residues involved in the binding of inhibitor 46, the same set of experiments was repeated with MurA C115D (C115D MurA+UNAG+fosfomycin and C115D MurA+UNAG+46, Figure 7). Firstly, it was noted that fosfomycin was unable to bind to the C115D mutant. Secondly, compound 46 was still capable of binding MurA C115D (as shown by CSPs and peak intensity reduction, Figure 7), indicating that cysteine-115 is not crucial for the binding of compound 46.

Figure 7. — ¹⁵N-¹H TROSY HSQC spectra of MurA C115D bound to UNAG (black), overlaid with MurA C115D (UNAG-bound) in the presence of fosfomycin (green, top spectrum) or compound 46 (orange, bottom spectrum). Zoom in on selected regions highlights peaks that undergo chemical shift perturbations (CSPs).

Potential binding mode of 46 with E. coli MurA

Next, we used molecular docking to derive a potential binding mode for 46 that would be in accordance with all experimental data. Our results from the kinetic analysis, native MS experiments and NMR spectroscopy clearly indicated that the UNAG-bound conformation of MurA is also the form to which 46 binds. Furthermore, as binding of 46 occurred in the absence of UNAG but included a phosphate ion formed from the co-substrate PEP, we reasoned that the binding site of 46 may overlap with that of UNAG but not of PEP. Based on these considerations, we selected an available X-ray co-crystal structure of E. coli MurA with UNAG and covalently bound fosfomycin for docking simulations, with the latter serving as a mimic of the PEP-derived phosphate (PDB code: 1UAE),⁵⁴ assuming that the phosphate group of fosfomycin would be coordinated by the same residues. The most plausible binding mode was obtained with the (S)-enantiomer of 46 (Figure 8). It largely overlapped with the binding site of UNAG, with the pyrrollidinedione ring occupying the space of the ribose, capturing lysine 160 that normally interacts with the uracil carbonyl. Since the pyrrolidinedione enol tautomer is slightly acidic, the charge-supported interaction with Lys160 might strongly enhance the binding affinity. The proposed binding mode was the only one that explained the clear contribution of the tert-butyl ester to the potency (vide supra), as only in this pose the tert-butyl moiety was in contact with a cluster of hydrophobic side chains, consisting of Val161, Pro289, Val327, and the Lys160 methylene groups, thus acting as a 'plug' closing part of the entry site of the binding pocket (see Figure S1C for further illustration). Interestingly, the terminal 4-chlorophenyl ring was predicted to replace the salt bridge of the UNAG phosphate with Arg120 by a cation-π interaction. This might be a key interaction, explaining why a similar subset of HSQC-NMR CSPs were observed for the binding of 46 and fosfomycin: both inhibitors stabilize the active, open conformation of the enzyme, capturing the large, mobile active site loop via coordination of the key loop residue Arg120 from two sides. This prediction was also consistent with the additive effect of 46 and fosfomycin on the enzymatic activity that suggested that 46 and fosfomycin can bind simultaneously. Notably, the extended biaryl system at position R² of the general structure was not only important for the interaction with Arg120, but also stabilized the biologically active conformation of the 46 (S)-enantiomer by intramolecular CH-π interactions. This conformational pre-organization might reduce the entropic penalty upon binding to the pocket; a largely rigid ligand seems particularly necessary here as the binding site is framed by four flexible loop motifs that must be stabilized. Finally, the outer face of the benzothiazole moiety was positioned in a lipophilic environment that was not filled by UNAG (Supporting Information, Figure S6), formed by Pro298, Val327, Phe328, and the methyl group of Thr304. The latter showed hydrophobic interactions with the methyl substituent in the 6-position of the benzothiazole (indicated by a green dashed line in Figure 8), potentially explaining the positive impact of this substitution on the potency.

Figure 8. — Predicted binding mode (by molecular docking) of 46 ((S)-enantiomer) in the X-ray co-crystal structure of UNAG and fosfomycin with MurA from *E. coli* (PDB code: 1UAE).⁵⁴ Inhibitor 46 (orange sticks) was predicted to fill the UNAG binding pocket completely, also including regions not occupied by UNAG. Amino acid residues (cyan) involved in direct or in hydrophobic interactions are labeled. Distances (green numbers) are given in Å; fm: covalently bound fosfomycin; blue dashed lines: H-bonds; brown dashed lines: CH-π and cation-π interactions.

Compound 46 is non-cytotoxic to human cells

Since the investigated pyrrolidinediones had initially been envisioned as human PKC inhibitors, it was necessary to test their toxicity to human cell lines.⁴⁶ A standard MTT assay revealed that compounds 7 and 46 were non-toxic when applied to HepG2 (human liver carcinoma cell line) and the non-tumor cell line MRC-5 (human lung fibroblasts, Figure 9). This highlights the fact that our most potent novel MurA inhibitors have no significant toxic effect on human cells, in line with their relatively low potency toward PKC.⁴⁶

Figure 9. — MTT assay result for compounds 7 and 46 indicating no cytotoxicity on both MRC-5 and HepG2 cell lines. The data shown are the mean of three independent experiments, and the error bars represent ± SD. Eight compound concentrations up to 100 μM (solubility in medium was verified) were tested and MTT assays were performed after 24 hours.

Antibacterial activities

All compounds were tested for their antibacterial activities against E. coli ΔtolC as a Gram-negative strain and S. aureus Newman as a Gram-positive strain. It was decided to quantify antibacterial activities as IC₉₀ values (i.e., concentrations for 90% growth inhibition) instead of MIC values (i.e., minimal inhibitory concentrations, usually obtained in series of two-fold dilutions) as this would enable a more precise quantification, in particular when moderate activities are also taken into account. Several compounds showed a moderate activity against S. aureus, including 46 with an IC₉₀ value of 40 μM (Supporting Information, Table S1). In contrast, no activity was found against the growth of the E. coli ΔtolC strain, which is efflux-deficient due to a deletion in the gene encoding the TolC efflux pump. We hypothesized that the compounds' inactivity against Gram-negative bacteria might be owed to limited cellular uptake. Due to the high stability of 46, compound degradation in the bacterial cell could be excluded (vide infra).

Polymyxin B nonapeptide (PMBN) is a weak antibacterial agent that lacks bactericidal activity of its own, but is still able to bind to the LPS (lipopolysaccharide) of Gram-negative bacteria and sensitize them to hydrophobic moieties by permeabilizing the outer membrane of the bacteria.^55,56 PMBN was added in the assays for antibacterial activity to test if the diminished activity of 46 against E. coli ΔtolC was due to limited cellular uptake. It was added in increasing concentrations up to 6 μg/mL (significantly less than the concentrations often reported).^55,56 Fosfomycin and ampicillin were included as controls to test their effect as established antibiotics that efficiently pass the outer membrane of E. coli. A drastic decrease in the IC₉₀ value of 46 was observed that was dependent on the concentration of PMBN, reaching 2.3 μM at 6 μg/mL of PMBN, a value ca. 3-fold lower than that determined for ampicillin under identical conditions (IC₉₀ = 6.8 μM, Supporting Information, Table S2). We observed no significant changes in the IC₉₀ values of both control antibiotics with increasing concentrations of PMBN, thus confirming that the improvement of the antibacterial activity of 46 was due to its enhanced entry into bacterial cells (Figure 10). This experiment also proved that 46 acted efficiently once it reached the interior of the cell. Poor penetration of the bacterial cell membrane might also be responsible for the low activity against S. aureus (cf. above). Some congeners that showed lower IC₉₀ values at least with S. aureus (Table S1) seemed to indicate that modification of the ester function (1, 37 and 45) could be crucial to enhance the cellular uptake in future optimization trials with the scaffold.

Figure 10. — Antibacterial activities (IC₉₀ values) of 46, fosfomycin, and ampicillin against *E. coli* Δ*tolC* in the presence of increasing PMBN concentrations (0, 3, 4.5, and 6 μg/mL). The data shown are the mean of two independent experiments, and the error bars represent ± SD. *IC₉₀ for compound 46 at 0 μg/mL PMBN was >40 μM.

Stability in biological media

The stability of MurA inhibitor 46 was assessed in human S9 liver microsomal fraction. As 46 contains a potentially labile ester function, we also tested the compound's stability in human plasma and in bacterial cell lysate (E. coli ΔtolC). Compound 46 did not show any degradation after a two-hour incubation period with human S9 fraction, and neither after four hours in either human plasma or bacterial lysate (Table 7). It thus exhibited a pronounced stability in a variety of biological media.

Table 7.

Stability of MurA inhibitor 46 in various biological media.^a

Compound	% degradation in S9 fraction^b	t_1/2 in S9 fraction (min)	% degradation in human plasma^c	t_1/2 in human plasma (min)	% degradation in bacterial lysate^c	t_1/2 in bacterial lysate (min)
46	0 ± 8	>120	0 ± 8	>240	0 ± 2%	>240
reference compound^d	89 ± 10	5.9	100	20	-----	-----

Open in a new tab

The data shown are the mean of two independent experiments, each performed in duplicates

measured after 2 h

measured after 4 h

reference compound for the S9 fraction assay was testosterone, and tetracaine was used for the human plasma stability assay.

Conclusions

In this study, we report substituted pyrrolidinediones as a new class of inhibitors of the bacterial enzyme MurA. The overall most potent of these compounds was inhibitor 46 with an IC₅₀ value of 4.5 μM against wild type MurA. Remarkably, 46 inhibited the clinically relevant fosfomycin-resistant MurA mutant C115D with equal potency (IC₅₀ = 4.9 μM). Moreover, it showed an additive effect when tested with fosfomcyin, suggesting that with optimized derivatives of 46, a combination therapy might be conceivable.

The binding of 46 was demonstrated to occur only with the conformationally open form of MurA, unlocked by the presence of the substrate UNAG and rendering the active site accessible. The precise binding mode could not be robustly established due to significant challenges in the crystallography trial. However, based on native MS experiments and the additive inhibition with fosfomycin, it could be inferred that 46 binds in the absence of UNAG but in the presence of fosfomycin or inorganic phosphate. Furthermore, NMR spectroscopic data indicated that the binding of 46 occurs in a region adjacent to or partially overlapping with the region also targeted by fosfomycin. However, there is no competition for the same binding pocket. The ability of compound 46 to still bind and inhibit the MurA C115D mutant rules out any direct interaction of 46 with the covalent binding partner of fosfomycin, cysteine-115.

The reported novel MurA inhibitors were not optimized for uptake into bacterial cells yet, but at least for some compounds we found growth inhibition of S. aureus with IC₉₀ values ranging from 18 to 40 μM. Apart from that, the new class of MurA inhibitors fulfilled other important requirements for antibiotics: It showed no toxicity against human cell lines, as exemplified at least by compounds 7 and 46. Moreover, Compound 46 was found to be highly stable in human plasma, human liver microsomes (S9 fraction) and bacterial cell lysate. Altogether, we provided evidence that the newly discovered substituted pyrrolidinediones are promising candidates for further development as MurA inhibitors with potential antibacterial activity.