Structure–activity relationship studies of [1,2,5]oxadiazolo[3,4-b] pyrazine-containing polymyxin-selective resistance-modifying agents

Abstract

Multidrug-resistant (MDR) Gram-negative bacteria are an urgent and rapidly spreading threat to human health with limited treatment options. Previously, we discovered a novel [1,2,5]oxadiazolo[3,4-b]pyrazine-containing compound (1) that selectively re-sensitized a variety of MDR Gram-negative bacteria to colistin, one of the last-resort antibiotic. Herein, we report the structure–activity relationship studies of compound 1 that led to the discovery of several more potent and/or less toxic resistance-modifying agents (RMAs). Further evaluation of these RMAs showed that they were effective in a wide range of MDR bacteria. These results demonstrated these compounds as a novel class of RMAs and may be further developed as therapeutic agents.

Introduction

Antimicrobial resistance (AMR) is one of the leading global health threats and is spreading at an accelerated pace.¹ It is estimated that, by 2050, we will not have any effective antibiotics available if no new antibiotics are developed.² Multidrug resistant (MDR) Gram-negative bacteria, such as Pseudomonas aeruginosa and Acinetobacter baumannii, are of great concern. Their infections have very limited treatment options, resulting in high mortality and morbidity, especially in immuno-compromised patients.^3–5 Colistin, a cationic antimicrobial peptide of the polymyxin family, is one of the last-resort antibiotics for the treatment of Gram-negative bacteria infections.⁶ However, bacteria have evolved resistance mechanisms to colistin that are also rapidly spreading.

Colistin is an amphiphilic polypeptide that contains both a poly-cationic macrocycle and a lipophilic tail. It is attracted to the negatively charged Lipid A on the outer membrane of Gram-negative bacteria via electrostatic interactions, and then inserts its lipophilic tail into the membrane to cause cell lysis. Bacteria have developed resistance to colistin by reducing the negative charges of Lipid A via addition of phosphoethanolamine (pEtN) and/or 4-amino-arabinose (Ara4N). These modifications are most common in A. baumannii and P. aeruginosa, respectively.^7,8 pEtN modification has spread to Enterobacteriaceae spp. via the mobilized colistin resistance (mcr) plasmids.⁹.

Resistance-modifying agents (RMAs) are a promising approach to fight against resistance bacteria.^10–14 RMAs can further expand the life span of existing antibiotics with well-studied antimicrobial spectra and safety profile. In addition, they often target non-essential genes or gene products and have low resistance potential. Several small molecules have been reported to resensitize bacteria expressing mcr-1 gene to colistin.^15–19 Recently, we also discovered a highly selective and active RMA, N⁵,N⁶-di-m-tolyl-[1,2,5]oxadiazolo[3,4-b]pyrazine-5,6-diamine (1, Fig. 1), that selectively resensitizes MDR Gram-negative bacteria to polymyxin family antibiotics.²⁰ Its closely related analog, compounds 2 (Fig. 1), showed similar but lower RMA activity. Herein, we report our follow-up structure–activity relationship (SAR) studies of this family of compounds. In addition to the RMA activity, the antibacterial and mammalian toxicity of the most potent RMAs were also evaluated.

Fig. 1.

Structures of two RMAs for colistin-resistant Gram-negative bacteria.

A variety of analogs of 1 and 2 bearing different substituents of the anilines, 5a–s and 6a-p (Scheme 1), were synthesized using procedures reported by Childress²¹ and Salamoun,²² respectively. The synthesis began with the reaction of commercially available diaminofurazan (3), oxalic acid, and aqueous 10% HCl solution, followed by chlorination with PCl₅/POCl₃ to afford dichloride 4 (Scheme 1). Dichloride 4 was then treated with various anilines in the presence of triethylamine in THF and refluxed overnight to provide bisanilino derivatives 5a–p. To synthesize analogs 6a-p, 4 was reacted with the aniline at room temperature followed by hydroxylation with aqueous potassium hydroxide and then quenching with dilute hydrochloric acid. For asymmetrical bisanilino derivatives, intermediate 4 was treated with 1.0 equivalent of meta-trifluoromethoxy aniline and triethylamine in THF at 0–25 °C for 1 h followed by the addition of the second anilines. The resulting mixture was then refluxed overnight to produce 5q-s.

Scheme 1.

Reagents and conditions: (a) oxalic acid, 10% HCl, reflux, 4 h, 72%; (b) PCl₅, POCl₃, reflux, 2 h, 75%; (c) R¹C₆H₄NH₂, Et₃N, THF, reflux, overnight, (d) R¹C₆H₄NH₂, THF, Et₃N, 0–25 °C, 1 h; KOH, 25 °C, 2 h, 30–60%; (e) meta-trifluoromethoxy aniline, Et₃N, THF, 0–25 °C, 1 h; (f) R²C₆H₄NH₂, Et₃N, THF, reflux, overnight, 20–40%.

All compounds were tested for their respective minimum resensitizing concentrations (MRCs) for colistin against E. coli. AR0493, a MDR strain from CDC and FDA Antibiotic Resistance Bank (https://wwwn.cdc.gov/ARIsolateBank/). Compounds with good RMA activity with MRCs < 1 μg/mL were also tested for their minimum inhibitory concentrations (MICs).²³ For the symmetrical bisanilino series, 5a–p, electron-donating alkyl and alkoxy groups (entries 5a–g, Table 1) at various positions of the phenyl rings were investigated. Only 5d with para-Me and 5f with meta-OMe substitutions showed similar RMA activity as 1 with meta-Me substitutions. Substitutions at the ortho positions significantly reduced the RMA activity. Substitutions on the phenyl groups using electron-withdrawing groups (e.g., F, Cl, and OCF₃) appeared to benefit the RMA activity in many cases. Among these three groups, trifluoromethoxy groups with the strongest electron-withdrawing capability and highest hydrophobicity showed the best RMA activity overall, and the analogs with ortho substitutions were still the worst within each series. Three analogs with meta-Cl, meta-OCF₃, and para-OCF₃ (i.e., 5 l, 5o, 5p), respectively, showed the best RMA activity with MRCs of 0.25 μg/mL and no antibacterial activity on their own with MICs > 64 μg/mL, the highest concentration tested. Three asymmetric bisanilino analogs (5q–s, Table 1) were synthesized based on the SARs of the monoanilino analogs 6a–p (vide infra). However, none of them showed improved RMA activity.

Table 1.

Biological evaluation of the bisanilino analogs of 1.

Entry	R1	R2	MRCa	MICb	GI50c	SId
5a	H	H	0.5	>64	4.4	8.8
5b	o-Me	o-Me	8	–	–	–
5c (1)	m-Me	m-Me	0.5	>64	15	30
5d	p-Me	p-Me	0.5	>64	11	22
5e	o-OMe	o-OMe	>64	–	–	–
5f	m-OMe	o-OMe	2	–	–	–
5g	p-OMe	p-OMe	>64	–	–	–
5h	o-F	o-F	4	–	–	–
5i	m-F	m-F	0.5	>64	20	40
5j	p-F	p-F	0.5	>64	9.4	19
5k	o-Cl	o-Cl	8	–	–	–
5l	m-Cl	m-Cl	0.25	>64	11	44
5m	p-Cl	p-Cl	0.5	>64	28	56
5n	o-OCF3	o-OCF3	>64	–	–	–
5o	m-OCF3	m-OCF3	0.25	>64	4.3	17
5p	p-OCF3	p-OCF3	0.25	>64	5.7	23
5q	m-OCF3	H	1	–	–	–
5r	m-OCF3	m-Me	0.5	>64	12	24
5s	m-OCF3	o-OMe	>64	–	–	–

– Not tested.

^aMRC values are in μg/mL in the presence of 2 μg/mL of colistin.

^bMIC values are in μg/mL.

^cGI₅₀ values are half maximum inhibitory concentrations in μg/mL.

^dSI (Selective Index) indicates the ratio of GI₅₀ and MRC values.

To evaluate the mammalian toxicity of these analogs, we tested them in cell viability assay using human cervical adenocarcinoma HeLa cells. Their respective half maximal growth inhibition concentration (GI₅₀) was calculated using KaleidaGraph. The selectivity index (SI) for each compound was also calculated as the ratio of their GI₅₀ and MRC. Compared with the lead compound, 1 (5c), three analogs (i.e., 5j, 5l and 5m), showed improved SIs. Although two analogs bearing trifluoromethoxy substitutions (i.e., 5n and 5o) are more potent RMAs, they were also more toxic than compound 1, resulting in reduced SIs. Therefore, we selected two bisanilino analogs, 5l and 5m, that bear chlorines at the meta- and para-positions, respectively, of the phenyl groups for further characterization studies.

For the monoanilino analogs, 6a-p, their RMA activities appeared to be considerably (>10 folds) lower than their corresponding bisanilino analogs. Only two compounds, 6e (ortho-OMe, Table 2) and 6o (meta-OCF₃, Table 2) showed slightly improved activity with MRCs of 4 μg/mL, when compared to the parent compound 2 (vide supra). Their mammalian toxicities were also evaluated, and SIs calculated. The results showed that 6e bearing an ortho-OMe group has very weak toxicity in mammalian cells with a GI₅₀ > 100 μg/mL, the highest concentration tested, while 6o has a moderate SI. Considering the moderate RMA activity and potentially low metabolic stability of 6e, it was not selected for further studies.

Table 2.

Biological evaluation of the monoanilino analogs 6a–p.

Entry	R1	MRCa	MICb	GI50c	SId
6a	H	>64	–	–	–
6b	o-Me	64	–	–	–
6c (2)	m-Me	8	–	–	–
6d	p-Me	64	–	–	–
6e	o-OMe	4	>64	>100	>25
6f	m-OMe	16	–	–	–
6g	p-OMe	>64	–	–	–
6h	o-F	64	–	–	–
6i	m-F	64	–	–	–
6j	p-F	32	–	–	–
6k	o-Cl	16	–	–	–
6l	m-Cl	16	–	–	–
6m	p-Cl	32	–	–	–
6n	o-OCF3	>64	–	–	–
6o	m-OCF3	4	>64	24	6
6p	p-OCF3	16	–	–	–

– Not tested.

^aMRC values are in μg/mL in the presence of 2 μg/mL of colistin.

^bMIC values are in μg/mL.

^cGI₅₀ values are half maximum inhibitory concentrations in μg/mL.

^dSI (Selective Index) is the ratio of GI₅₀ and MRC values.

With two potent analogs 5l and 5m in hand, we next explored the scope of their colistin-potentiating activity in 8 different MDR Gram-negative bacteria with diverse genetic backgrounds and resistance mechanisms (Table 3). All strains were obtained from CDC and FDA Antibiotic Resistance Bank. For example, E. coli AR-0493 and K. pneumoniae AR-0497 carry the plasmid that contains the most prevalent mcr-1 gene. E. coli AR-0538, Salmonella Typhimurium AR-0539 and AR-0635 contain the mcr-2, 3, and 4 genes, respectively. These genes encode the phosphoethanolamine transferase, which was originated form the chromosome of A. baumannii, such as AR-0310. P. aeruginosa, on the other hand, contains genes that are responsible for the Ara4N modification of Lipid A. These strains are all resistant to colistin with their MICs in the range of 4–32 μg/mL. When used at 1 μg /mL, compounds 5l and 5m reduced the colistin MIC to 1–2 μg/mL in strains that contain the mcr genes. A. baumannii AR-0310 appears more sensitive to these two compounds and the MICs of colistin was reduced to 0.25 μg/mL. Interestingly, both compounds are also effective in two P. aeruginosa strains and reduced the MICs of colistin in both AR-0239 and AR-0257 by 4 folds. These results suggest that the oxadiazolopyrazine-containing compounds do not inhibit the activity of pEtN transferase directly, but resensitize these resistant bacteria to colistin via a novel mechanism.

Table 3.

MICs of colistin in the absence or presence of two best RMAs in various MCR bacterial strains.

Bacterial strain	Resistance gene	Colistin MICa
		-	(+51) b	(+5m)b
E. coli AR-0493	mcr-1	8	1	1
E. coli AR-0538	mcr-2	8	1	2
S. Typhimurium AR-0539	mcr-3	8	1	2
S. Typhimurium AR-0635	mcr-4	32	1	1
K. pneumoniae AR-0497	mcr-1	8	1	1
A. baumannii AR-0310		32	0.25	0.25
P. aeruginosa AR-0239		4	1	1
P. aeruginosa AR-0257		8	2	2

^aColistin MIC in μg/mL.

^bColistin MIC in μg/mL, in the presence of 1 μg/mL of compound indicated, unless otherwise noted.

Both series of compounds had previously been reported as mitochondrial protonophores uncouplers by Santos and coworkers.^21–22 They are weak lipophilic acids that uncouple oxidative phosphorylation from ATP production by transporting protons across the inner membrane of mitochondria into the mitochondrial matrix. The bisanilino compounds (e.g., 1) were more potent in the cellular oxygen consumption assay. However, they suffer from high hydrophobicity and poor aqueous solubility. A para-OCF₃ analog of the monoanilino compound 2 possesses more favorable pharmacological properties, albeit with slightly lower activity in the cellular oxygen consumption assay. The SARs of these compounds as RMAs and those as mitochondria protonophore uncouplers are similar in certain aspects, but different in others. For example, analogs with higher pKa and hydrophobicity are preferred and the bisanilino analogs are much more potent than their corresponding monoanilino analogs. However, monoanilino analog 6p bearing a para-OCF₃ substitution was one of the most potent uncouplers both in vitro and in mice, yet with poor RMA activity. Bisanilino analog 5p bearing para-OCF₃ substitutions was a poor uncoupler, but with very potent RMA activity. Taken together, these comparisons suggest that these compounds may potentiate colistin via a novel mechanism, other than shuffling protons across the membrane.

In summary, we have synthesized a variety of both bisanilino and monanilino [1,2,5]oxadiazolo[3,4-b]pyrazine-containing compounds and evaluated their antibacterial and colistin-resensitizing activity, as well as the mammalian toxicity. The SARs of both series of compounds were established and suggested that the bisanilino analogs are significantly more potent RMAs, and the analogs with electron-withdrawing and hydrophobic groups on the meta- or para-positions of the phenyl rings are beneficial to the RMA activity, while substitutions at the ortho-positions have deleterious effect. Two bisanilino analogs bearing meta-Cl and para-Cl substitutions, respectively, were identified with improved activity or mammalian toxicity. Further evaluation of these in a wide range of MDR Gram-negative bacteria showed that they are effective RMAs in all species and strains tested and were able to lower the MICs of colistin to its clinical breakpoint, 2 μg/mL or lower. Since P. aeruginosa is resistant to colistin by Ara4N modification of Lipid A, the results suggested that these compounds do not inhibit the pEtN transferase directly and may potentiate colistin via a novel mechanism. Further mechanistic studies and development of this novel class of RMA as antibiotic adjuvant are ongoing and will be report in due course.