Identification of a 2-aminobenzimidazole scaffold that potentiates gram-positive selective antibiotics against gram-negative bacteria

Abstract

The development of novel therapeutic approaches is crucial in the fight against multi-drug resistant (MDR) bacteria, particularly gram-negative species. Small molecule adjuvants that enhance the activity of otherwise gram-positive selective antibiotics against gram-negative bacteria have the potential to expand current treatment options. We have previously reported adjuvants based upon a 2-aminoimidazole (2-AI) scaffold that potentiate macrolide antibiotics against several gram-negative pathogens. Herein, we report the discovery and structure-activity relationship (SAR) investigation of an additional class of macrolide adjuvants based upon a 2-aminobenzimidazole (2-ABI) scaffold. The lead compound lowers the minimum inhibitory concentration (MIC) of clarithromycin (CLR) from 512 to 2 μg/mL at 30 μM against Klebsiella pneumoniae 2146, and from 32 to 2 μg/mL at 5 μM, against Acinetobacter baumannii 5075. Preliminary investigation into the mechanism of action suggests that the compounds are binding to lipopolysaccharide (LPS) in K. pneumoniae, and modulating lipooligosaccharide (LOS) biosynthesis, assembly, or transport in A. baumannii.

Graphical Abstract

Small molecule adjuvants have the potential to expand the spectrum of gram-positive selective antibiotics to encompass gram-negative bacteria. We identifiy a series of 2-aminobenzimidazole (2-ABI) macrolide adjuvants with activity against Acinetobacter baumannii and Klebsiella pneumoniae. Preliminary mechanism of action studies suggests that adjuvants bind to lipopolysaccharide in K. pneumoniae, and modulate lipooligosaccharide biosynthesis, assembly, or transport in A. baumannii.

Introduction

Antibiotic resistance is one of the most serious threats to modern medicine. Infections caused by drug-resistant bacteria have significant consequences including higher patient morbidity and mortality, and increased medical costs.^[1] In 2019, the Centers for Disease Control and Prevention (CDC) reported that each year multidrug-resistant (MDR) bacterial infections account for over 35,000 deaths, and incur $4.6 billion in healthcare costs.^[2] The divestment of pharmaceutical companies from this area has further compounded the antibiotic resistance crisis.^[1]

Six bacterial species, termed the ESKAPE pathogens (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter species), are responsible for the majority of hospital acquired infections (HAIs), and have become the most concerning species in the context of antibiotic resistance.^[3] Four of these species are gram-negative, for which treatment options are particularly limited. The carbapenem class of antibiotics has long been used as a treatment for these gram-negative infections, but increasing resistance has led clinicians to resort to more poorly tolerated treatments such as polymyxins.^[4] Novel treatments to combat gram-negative infections are acutely needed.

Infections caused by gram-negative bacteria are considerably harder to treat than those caused by gram-positive bacteria. Gram-negative bacteria are intrinsically resistant to many antibiotic classes including: macrolides, glycopeptides, lipopeptides and oxazolidinones due to the presence of the outer membrane (OM) and efflux pumps.^[5] The outer leaflet of the OM is comprised of lipopolysaccharide (LPS) or lipooligosaccharide (LOS) molecules that pack tightly together and are stabilized by divalent cations, creating a barrier that is difficult to penetrate.^[5] Typically, drug uptake by gram-negative cells is limited to small (less than ~600 Da), hydrophilic molecules that access the cell through porins.^[5]

One approach to combating infections caused by gram-negative bacteria is the use of small molecule adjuvants.^[6] Adjuvants that potentiate otherwise gram-positive selective antibiotics against gram-negative bacteria would considerably expand the options available to clinicians.^[7] Known potentiators of gram-positive selective antibiotics typically cause physical disruption of the OM, facilitating access of antibiotics to their intracellular targets. SPR741, a cationic peptide that is structurally similar to polymyxin B, binds and compromises the integrity of the OM, and potentiates azithromycin (AZM) activity against MDR Enterobacteriaceae.^[8] Liproxistatin-1 and MAC-0549481 also bind to LPS and disrupt the integrity of the OM, and potentiate various gram-positive active antibiotics against Escherichia coli, K. pneumoniae, A. baumannii, and P. aeruginosa.^[9]

We have previously reported three classes of adjuvants that potentiate macrolide antibiotics against gram-negative ESKAPE pathogens (representative structures of each class are depicted in Figure 1). The bis-2-aminoimidazole (bis-2-AI) 1 lowers the AZM minimum inhibitory concentration (MIC) 256-fold to 1 μg/mL against P. aeruginosa PAO1 at 40 μM.^[10] Aryl 2-AI 2 lowers the clarithromycin (CLR) MIC from 32 μg/mL to 2 μg/mL at 10 μM against A. baumannii AB5075,^[11] while second generation analogues 3 and 4 exhibit improved activity and drop the CLR MIC to 1 μg/mL and 0.25 μg/mL at 7.5 μM, respectively. Compound 5, the lead compound from our most recently reported class of macrolide adjuvants, dimeric 2-AIs, exhibits a broader range of activity against A. baumannii, K. pneumoniae, P. aeruginosa, and E. coli. Against A. baumannii AB5075, 5 lowers the CLR MIC 16-fold to 2 μg/mL at 1.5 μM.^[12] Investigation of the mechanism of action of compound 2 in A. baumannii indicated that it likely alters LOS presentation and/or biosynthesis, rather than binding and physically disrupting the OM.^[11]

Figure 1.

Structures of previously reported macrolide adjuvants with activity against gram-negative ESKAPE pathogens.^10-13

We herein report a screen of our in-house marine-alkaloid derived library against K. pneumoniae that identified a fourth class of macrolide adjuvants based on a 2-aminobenzimidazole (2-ABI) scaffold. A structure activity relationship (SAR) study identified two lead compounds that potentiate CLR 256-fold at 30 μM against K. pneumoniae. We determined the activity of the lead compounds in combination with other gram-positive selective antibiotics, and against other gram-negative bacterial species, and report preliminary mechanism of action studies that suggest the compounds physically interact with the OM in K. pneumoniae, and disrupt LOS formation in A. baumannii.

Results and Discussion

We previously identified a scaffold for the development of macrolide adjuvants against P. aeruginosa through a screen of our internal library of nitrogen-dense marine-alkaloid derived small molecules.^[10] We used this approach to identify a scaffold with macrolide adjuvant activity against a MDR strain of K. pneumoniae (KP2146). In the initial screen (Table S1) ~200 compounds were tested at 60 μM in combination with AZM. In the combination screen, control wells were included that contain compound only, which allows for the identification of compounds that exhibit standalone toxicity at 60 μM. Any compound that exhibited standalone toxicity at 60 μM was subject to MIC determination, followed by retesting with AZM at 30% the compound’s MIC. AZM was chosen as the representative macrolide for the initial screen because KP2146 displays moderate resistance to AZM with an MIC of 64 μg/mL, while KP2146 is highly resistant to CLR, and erythromycin (ERY), (MICs ≥ 512 μg/mL). There are no reported clinical breakpoints for macrolide antibiotics against K. pneumoniae, and the breakpoints for the few gram-negative species that have been determined fall between 1 – 16 μg/mL^[13], therefore we aimed to uncover an adjuvant that lowers the AZM MIC by a minimum of four-fold (to ≤ 16 μg/mL). From the initial screen, compound 6^[14] was found to lower the AZM MIC from 64 to 16 μg/mL at 60 μM. The stand-alone MIC of 6 was subsequently determined to be >200 μM.

Following the identification of compound 6, structurally similar compounds from the library that did not form part of the initial screen were examined, however no compounds with increased activity were identified (Table S1). Compound 6 is based on the marine alkaloid bromoageliferin, and its core is comprised of a 2-AI fused to a cyclohexane ring. Previously we have investigated the 2-ABI heterocycle as a more synthetically accessible alternative to the bromoageliferin core^{[15, 16]}, and therefore next opted to determine the activity of 2-ABI containing small molecules from the library. As in the initial screen, compounds were assayed at 60 μM, unless otherwise stated, in combination with AZM (Table S2). This follow up screen returned two active compounds, 7 and 8,^[16] (Figure 2) that have essentially equivalent (MICs within two-fold) activity to compound 6, both dropping the AZM MIC to 8 μg/mL at 60 μM (Table S2). Neither compound exhibits anti-bacterial activity alone against KP2146 (MICs > 200 μM).

Figure 2:

Structure of initial hit compound 6 with rationale for screening of 2ABIs (blue dashed box). Structures of hits from screen of 2ABI containing small molecules, compounds 7 and 8.

Compounds 7 and 8 both possess a 2-ABI-amide moiety; in compound 7, the amide is linked via a methylene unit to a cyclopentyl ring, while compound 8, the amide is linked via a methylene unit to a 4-chlorophenyl tail. We first resynthesized compounds 7 and 8 and confirmed activity, then designed and synthesized a series of analogues in which the 2-ABI and amide linker were retained, and the tail region varied, to investigate the SAR of the 2-ABI scaffold.

The synthetic route to access these analogues is outlined in Scheme 1. Utilizing a previously reported procedure^{[16, 17]}, 4-nitro-o-phenylenediamine was cyclized with cyanogen bromide under reflux, followed by Boc protection with Boc anhydride, and subsequent reduction of the nitro group with hydrogen and 5% palladium on carbon to yield 2-ABI 9. Compound 9 was then acylated using commercially available acid chlorides, or coupled with commercially available carboxylic acids using 1-ethyl-3-(3-dimethylaminopropyl)caboiimide (EDC). Removal of the BOC groups with a 1:1 mixture of trifluoroacetic acid (TFA) and dichloromethane (DCM), followed by conversion to the hydrochloric acid salts delivered 7, 8, and 10-39 (Scheme 1).

Scheme 1:

Synthesis of 2-ABI analogues 7, 8, and 10-39. Reagents and conditions (a) BrCN, H₂O/CH₃CN, reflux, 4 h.; (b) Boc₂O, DMAP, THF, 25 °C, 18 h.; (c) H₂, 5% Pd/C, MeOH, 25 °C, 16 h.; (d) ArCOCl, TEA, DCM, 25 °C, 16 h.; (e) ArCOOH, EDC, DMAP, 25 °C, 16 h.; (f): TFA, DCM, 25 °C, 3 h, then 6 M HCl, MeOH, 25 °C, 2 min.

The standalone MIC of all synthesized compounds against KP2146 was determined, followed by the MICs of AZM, CLR, and ERY, in the presence or absence of each compound at 30% MIC (Table 1). Most compounds returned standalone MICs of ≥200 μM (Table S3), with compounds 23, 31, 32, 37, 38, and 39 returning MICs of 100 μM, and compound 36 returning an MIC of 50 μM.

Table 1:

Macrolide potentiation data for 2-ABI analogues tested at 30% stand-alone MIC against K. pneumoniae KP2146.

		MIC (μg/mL) [fold-change]
Cmpd	Conc. (μM)	AZM	CLR	ERY
-	-	64	512	> 512
7	60	8 [8]	64 [8]	> 256 [2]
8	60	8 [8]	64 [8]	256 [>2]
10	60	32 [2]	64 [8]	> 256 [2]
11	60	16 [4]	32 [16]	256 [>2]
12	60	16 [4]	64 [8]	128 [>4]
13	60	64 [0]	256 [2]	> 256 [2]
14	60	64 [0]	>256 [< 2]	> 256 [2]
15	60	64 [0]	>256 [< 2]	> 256 [2]
16	60	8 [8]	32 [16]	128 [>4]
17	60	2 [32]	0.5 [1024]	8 [>64]
18	60	1 [64]	0.25 [2048]	16 [>32]
19	60	64 [0]	>256 [< 2]	> 256 [2]
20	30	64 [0]	>256 [< 2]	> 256 [2]
21	60	1 [64]	0.125 [4096]	4 [>128]
22	30	32 [2]	0.5 [1024]	> 256 [2]
23	60	1 [64]	1 [64]	16 [>32]
24	30	1 [64]	0.25 [2048]	16 [>32]
25	60	4 [16]	16 [32]	64 [>8]
26	30	8 [8]	8 [64]	64 [>8]
27	60	64 [0]	>256 [< 2]	> 256 [2]
28	60	32 [2]	8 [64]	256 [>2]
29	60	4 [16]	0.5 [1024]	32 [>16]
30	60	2 [32]	0.5 [1024]	16 [>32]
31	60	32 [2]	>256 [< 2]	> 256 [2]
32	60	4 [16]	0.5 [1024]	32 [>16]
33	60	16 [4]	32 [16]	128 [>4]
34	60	1 [64]	0.125 [4096]	4 [>128]
35	60	2 [32]	0.25 [2048]	32 [>16]
36	30	4 [16]	0.5 [1024]	16 [>32]
37	15	64 [0]	1 [64]	> 256 [2]
38	30	8 [8]	1 [64]	> 256 [2]
39	60	8 [8]	1 [512]	32 [>16]

All adjuvant MICs = 200 μM, tested at 60 μM unless otherwise stated. MIC = 100 μM, tested at 30 μM. MIC = 50 μM, tested at 15 μM.

Compounds 7 and 8 moderately potentiate CLR, both reducing the MIC from 512 μg/mL to 64 μg/mL (eight-fold reduction), and exhibit little activity with ERY, reducing the MIC from >512 μg/mL to >256 and 256 μg/mL respectively. Lengthening the alkyl linker of compound 7 by one methylene unit (compound 10) leads to complete abrogation of activity (adjuvant concentration and potentiation are summarized in Table 1), while we had already observed in the 2-ABI library screen (Table S2) that complete removal of the alkyl liker also abrogates activity, as does increasing or decreasing the size of the cycloalkane (with AZM). Cycloalkane analogues were therefore not explored further, and we focused our attention on analogues of compound 8. Removing the linker (11) or increasing the length by one methylene unit (12) results in comparable activity to compound 8. Changing the position of the chloro substituent from 4- to 3- (13) or 2- (14) leads to abrogation of activity, as does replacing the 4-chloro substituent with a 4-fluoro substituent (15). The 4-bromo analogue of compound 8 (16) exhibits comparable activity to 8.

We have previously reported highly potent adjuvants that harbor an aryl group with a di-halogen substitution pattern,^[18] therefore we probed if these substitutions would enhance activity with the 2-ABI scaffold. Dichloro analogues of compound 8, compounds 17 (3,5-dichloro) and 18 (3,4-dichloro) both exhibit increased activity with all three macrolides, most compellingly CLR. Compound 17 reduces the MICs of AZM, CLR, and ERY to 2 μg/mL, 0.5 μg/mL, and 8 μg/mL respectively, while compound 18 reduces the MICs to 1 μg/mL, 0.25 μg/mL, and 16 μg/mL respectively. The fluoro (19) and trifluoromethyl (20) analogues of compound 17 are inactive, however the bromo analogue (21) exhibits comparable to increased activity to compound 17, reducing the MICs of AZM, CLR, and ERY to 1 μg/mL, 0.125 μg/mL, and 16 μg/mL respectively. Mixed halogen analogues of compound 18 display varied levels of activity, with the 3-chloro-4-trifluoromethyl analogue (22) being the most potent, reducing the MICs of AZM, CLR, and ERY to 1 μg/mL, 0. 25 μg/mL, and 16 μg/mL respectively, and the 3-fluoro-4-trifluoromethyl analogue (23) reducing the MICs of AZM, CLR, and ERY to 1 μg/mL, 1 μg/mL, and 16 μg/mL respectively. The 3-trifluoromethyl-4-chloro analogue (24) reduces the CLR MIC to 0.5 μg/mL but does not potentiate AZM or ERY, while the 3-trifluoromethyl-4-fluoro analogue (25) only moderately reduces the MIC of all three macrolides. An analogue of 17 possessing an additional methylene unit (26) is inactive.

Turning to compounds lacking the methylene unit, the fluoro (27) and bromo (28) analogues of compound 11 are inactive, with the exception that 28 exhibits moderate activity with CLR, reducing the MIC to 8 μg/mL. The 3,5-dichloro compound (29) exhibits comparable activity to its methylene containing analogue 17 with AZM and CLR, and reduced activity with ERY, and similarly the 3,4-dichloro compound 30 exhibits comparable activity to 18. The 3,5-difluoro compound 31 is inactive, while the 3,5-dibromo compound 32 exhibits slightly reduced activity compared to its methylene containing analogue 21. A 3,5-dimethyl analogue (33) designed as a steric isostere of 29 is inactive. The 3,5-bistrifluoromethyl analogue 34 is highly active, reducing the MICs of AZM, CLR, and ERY to 1 μg/mL, 0.125 μg/mL, and 4 μg/mL respectively. Again, mixed halogen 3,4-disubstituted compounds display varied levels of activity. In the absence of a methylene unit, the 3-trifluoromethyl-4-fluoro (35) and 3-trifluoromethyl-4-chloro analogues (36) are the most active, with 35 reducing the MICs of AZM, CLR, and ERY to 2 μg/mL, 0.25 μg/mL, and 32 μg/mL respectively, and 36 reducing the MICs of AZM, CLR, and AZM to 4 μg/mL, 0.5 μg/mL, and 16 μg/mL respectively. The 3-chloro-4-trifluoromethyl (37) and 3-fluoro-4-trifluoromethyl (38) analogues potentiate only CLR to an appreciable degree, both reducing the MIC to 1 μg/mL. Finally, the tri-halogenated 39 exhibits comparable activity to 29.

Overall, the most potent activity was observed with CLR, and to further probe the activity of the lead compounds, we next determined dose response activity (Table S4). At 30 μM six compounds (21, 29, 30, 32, 34 and 39) lower the CLR MIC to ≤ 2 μg/mL, with 21, 32, and 34 lowering the MIC to ≤ 0.5 μg/mL. Most compounds are completely inactive at 15 μM, with only compounds 29 (4 μg/mL), 30 (8 μg/mL), and 32 (32 μg/mL) potentiating CLR at this concentration. Compound 37, which has an MIC of 50 μM lowers the MIC to 1 μg/mL at 15 μM but is inactive at 10 μM.

To determine if the 2-ABI scaffold has broader gram-negative activity, we examined activity against an MDR-strain of A. baumannii (AB5075) (Tables 2 and S5). Twenty-three 2-ABIs lower the CLR MIC to ≤ 2 μg/mL. A follow-up dose-response study on the active compounds (Table S6) revealed that compounds 29 and 30 lower the MIC of CLR from 32 μg/mL to 2 μg/mL (16-fold reduction) at 10 μM, while three compounds (32, 37, and 39) lower the CLR MIC to ≤ 2 μg/mL at 7.5 μM. Compound 39 is the most potent overall against AB5075, lowering the CLR to 2 μg/mL at 5 μM. For their cross-species activities, compounds 29 and 39 were chosen as lead adjuvants for further examination.

Table 2:

Macrolide potentiation data for 2-ABI analogues tested at 30% stand-alone MIC against A. baumannii AB5075.

		MIC (μg/mL) [fold-change]
Cmpd	Conc. (μM)	AZM	CLR	ERY
-	-	32	32	32
7	60	16 [2]	32 [0]	16 [2]
8	60	16 [2]	4 [8]	16 [2]
10	60	8 [4]	4 [8]	16 [2]
11	60	8 [4]	0.5 [64]	8 [4]
12	60	16 [2]	1 [32]	8 [4]
13	60	16 [2]	16 [2]	16 [2]
14	60	16 [2]	16 [2]	16 [2]
15	60	32 [0]	32 [0]	32 [0]
16	60	16 [2]	2 [16]	8 [4]
17	60	16 [2]	0.25 [128]	8 [4]
18	60	16 [2]	0.5 [64]	8 [4]
19	60	32 [0]	32 [0]	32 [0]
20	60	8 [4]	0.25 [128]	4 [8]
21	60	16 [2]	0.25 [128]	4 [8]
22	60	8 [4]	0.5 [64]	8 [4]
23	60	8 [4]	0.5 [64]	8 [4]
24	60	8 [4]	0.125 [256]	4 [8]
25	60	8 [4]	0.5 [64]	4 [8]
26	60	16 [2]	0.5 [64]	8 [4]
27	60	32 [0]	32 [0]	32 [0]
28	60	16 [2]	0.5 [64]	8 [4]
29	60	4 [8]	0.5 [64]	4 [8]
30	60	8 [4]	0.5 [64]	8 [4]
31	60	32 [0]	8 [4]	32 [0]
32	60	8 [4]	0.25 [128]	8 [4]
33	60	16 [2]	1 [32]	4 [8]
34	60	8 [4]	0.25 [128]	4 [8]
35	60	16 [2]	0.25 [128]	4 [8]
36	60	4 [8]	0.25 [128]	2 [16]
37	30	4 [8]	0.5 [64]	2 [16]
38	60	8 [4]	0.25 [128]	4 [8]
39	60	2 [16]	0.5 [64]	2 [16]

All adjuvant alone MICs = 200 μM, tested at 60 μM unless otherwise stated. MIC = 100 μM, tested at 30 μM. MIC = 50 μM, tested at 15 μM.

We next examined lead adjuvants 29 and 39 with CLR against a panel of 12 K. pneumoniae strains, and 26 A. baumannii strains. The panel of A. baumannii strains was obtained from the Walter Reed Army Institute of Research (WRAIR), and is a diverse group that represents all major clades and many minor clades that are clinically relevant.^[19] Eight of the K. pneumoniae strains were obtained from the University of Maryland, while the other four were obtained from the American Type Culture Collection (ATCC). Against the 26 A. baumannii strains, CLR MICs range from 16 – 64 μg/mL. In the presence of 15 μM 29 or 10 μM 39, CLR MICs for 25 of the 26 isolates are lowered to 0.5 – 4 μg/mL or 0.5 – 2 μg/mL, respectively (Table S7). Against A. baumannii strain AB3917 however, both compounds lower the CLR MIC only two-fold to 32 μg/mL. Eleven of the 12 K. pneumoniae strains exhibit a highly resistant CLR phenotype, returning MICs of 1024 – 2048 μg/mL. One strain (KP700603) returned a CLR MIC of 256 μg/mL. At 30 μM, 29 and 39 lower the CLR MIC to 0.5 – 2 μg/mL and 0.125 – 2 μg/mL, respectively (Table S8). Both adjuvants appear highly active with CLR against diverse A. baumannii and K. pneumoniae strains.

We further investigated the spectrum of activity of compounds 29 and 39 by examining activity against the other gram-negative ESKAPE pathogens: P. aeruginosa (PAO1) and Enterobacter spp. (E. cloacae 2468), in addition to E. coli (EC25955), and Salmonella enterica serovar Typhimurium (JSG210). Against each bacterium, with the exception of compound 39 in PAO1, both compounds have MICs ≤ 100 μM, with the lowest 25 μM (39 against EC25922). Neither compound exhibits CLR potentiation activity against PAO1, E. cloacae 2468, or JSG210 at 30% their MIC. Compound 29 lowers the MIC of CLR against EC25922 from 32 to 0.125 μg/mL at 15 μM. No activity was observed for compound 39 at 7.5 μM against EC25922 (Table S9).

Next, we sought to uncover whether adjuvants 29 and 39 potentiate additional gram-positive selective antibiotics against K. pneumoniae and A. baumannii. Antibiotics investigated were vancomycin (VAN), a large glycopeptide antibiotic typically used in the treatment of methicillin resistant S. aureus (MRSA) that has little to no permeation through the gram-negative OM^[20]; linezolid (LNZ), an oxazolidinone antibiotic used in the treatment of various gram-positive infections that can permeate the gram-negative OM due to its small size, but is subject to efflux rendering it ineffective^[21]; rifampin (RIF), a large rifamycin antibiotic unable to penetrate the gram-negative cell that is typically used in combination with isoniazid and pyrazinamide for treatment of tuberculosis and some gram-positive infections^[22]; and daptomycin (DAP), a lipopeptide antibiotic typically used in the treatment of MRSA and vancomycin-resistant enterococci (VRE) that also has minimal permeation through the gram-negative OM.^[23] Activity was defined by breakpoint concentrations in Staphylococcus spp.: ≥ 16 μg/mL (VAN), ≥ 8 μg/mL (LNZ), ≥ 4 μg/mL (RIF), ≥ 1 μg/mL (DAP).^[13]

K. pneumoniae KP2146 is highly resistant to VAN, LNZ, and DAP, with MICs >512 μg/mL, and exhibits moderate resistance to RIF, with an MIC of 32 μg/mL. At 60 μM, the highest concentration tested, 29 and 39 only lower the MICs of VAN and LNZ by 32-fold and eight-fold, respectively (Table 3), and do not potentiate DAP. Both compounds lower the RIF MIC well below 2 μg/mL; compound 29 lowers the MIC from 32 μg/mL to 781 ng/mL (4096-fold reduction), while compound 39 lowers the MIC to 0.0625 μg/mL (512-fold reduction), at 60 μM. Lowering the concentration to 30 μM (the active concentration with CLR), leads to a loss of activity with VAN, LNZ, and DAP, but with RIF, activity was retained (MICs of 0.0625 μg/mL and 0.125 μg/mL for 29 and 39 respectively).

Table 3:

Activity of compounds 29 and 39 with additional gram-positive selective antibiotics against K. pneumoniae KP2146 and A. baumannii AB5075.

	MIC (μg/mL) [fold-change]
	KP2146					AB5075
Cmpd	Conc. (μM)	VAN	LNZ	RIF	DAP	Conc. (μM)	VAN	LNZ	RIF	DAP
-	-	> 512	> 512	32	>512	-	512	256	4	>512
29	60	32 [>16]	128 [>4]	0.00781 [4096]	>512 [0]	60	4 [128]	128 [2]	0.03125 [128]	512 [0]
	30	>512 [0]	256 [>2]	0.0625 [512]	>512 [0]	10	256 [2]	128 [2]	0.5 [8]	>512 [0]
39	60	32 [>16]	128 [>4]	0.0625 [512]	256 [>2]	60	4 [128]	128 [2]	0.0625 [64]	128 [>4]
	30	64 [>8]	>512 [0]	0.125 [256]	>512 [0]	5	64 [8]	128 [2]	0.125 [32]	>512 [0]

AB5075 is also highly resistant to VAN, LNZ, and DAP, with MICs of 512 μg/mL, 256 μg/mL, and >512 μg/mL respectively, but susceptible to RIF (MIC of 4 μg/mL), with respect to the breakpoint concentrations in Staphylococcus spp.^[13] Compounds 29 and 39, both lower the VAN MIC from 512 μg/mL to 4 μg/mL at 60 μM. Neither compound potentiates LNZ against AB5075 and both exhibit minimal activity with DAP, lowering the MIC four-fold. Despite AB5075 being sensitive to RIF, both compounds still lower the MIC of RIF upwards of 64-fold at 60 μM (to 0.03125 μg/mL and 0.0625 μg/mL for 29 and 39 respectively). At their active concentrations for CLR (10 μM and 5 μM), activity is lost for both VAN and DAP, but retained for RIF with MICs of 0.5 (eight-fold reduction) with 29 and 0.125 (32-fold reduction) with 39 (Table 3).

Time-kill curves were constructed for compounds 29 and 39 with CLR against KP2146 and AB5075 (Figure S1) to determine if the combination is acting synergistically, as well as if the combination is bacteriostatic or bactericidal. Synergy is defined as ≥ 2log₁₀ decrease in the number of colony forming units (CFUs) following treatment with the combination compared to antibiotic alone.^[24] After 24 h, combinations of 29 and 39 at 30 μM with 8 or 2 μg/mL CLR against KP2146 showed synergy. Against AB5075, 29 (10 μM) and 39 (5 μM) in combination with 8 μg/mL CLR exhibited synergy. When combined with 2 μg/mL CLR, 29 does not exhibit synergy at 24 h, while 39 does.

A reduction in CFU/mL of 99.9% (≥ 3log₁₀) from the starting inoculum is defined as bactericidal, while a reduction of less than 3log₁₀ CFU/mL is defined as bacteriostatic.^[25] The macrolide class of antibiotics inhibits protein synthesis, stalling growth and reproduction of the bacterium, and is bacteriostatic.^[26] As expected, each combination of compound and CLR, in both bacterial species, returned bacteriostatic activity. Compound 39 did show early growth toxicity, but at 24 h the CFUs were equivalent to the control.

To investigate the mechanism of action of the 2-ABI adjuvants, we first determined the effect of the lead compound 29 on the MIC of colistin (COL). Activity of COL is driven by membrane disruption through binding LPS (K. pneumoniae)/LOS (A. baumannii),^[4] and the 2-AI adjuvant 2, which is hypothesized to alter LOS presentation and/or biosynthesis, antagonizes COL against AB5075, so we sought to determine if the 2-ABI adjuvants are acting in a similar manner.^[11] In KP2146 (COL susceptible), no change in COL MIC was observed for compound 29 at 60 μM (Table S10), while in AB5075 (also COL susceptible) the COL MIC is raised from 1 μg/mL to 16 μg/mL in the presence of 60 μM 29. In some A. baumannii strains COL resistance can be achieved through loss of LOS, however K. pneumoniae does not exhibit this phenomenon.^{[27, 28]} Resistance to COL can also occur via mutations in the genes encoding the PmrAB (AB and KP) or PhoPQ (KP only) two-component systems, or by acquisition of a plasmid borne mobile colistin resistance gene (mcr1-10,), both of which lead to modifications of the lipid A component of LPS/LOS, which decreases the affinity for COL for the OM.^[29],[30] Compound 29 does not affect the COL MIC against highly COL-resistant strains of K. pneumoniae (KPB9) and A. baumannii (AB4106). Against strains that harbor mcr-1; 29 does not affect the COL MIC of AB17978^mcr−1 (64 μg/mL), but moderately lowers the COL MIC against KPF2210291^mcr−1 (from 16 μg/mL to 4 μg/mL).^[31] Activity against the parent (COL susceptible) strains mirrors that against AB5075 and KP2146, in that the MIC is raised against AB17978 (from 2 to 8 μg/mL) and unaffected against KPF2210291 (1 μg/mL).

We also determined the activity with CLR against a small panel of six COL-resistant strains (two COL-resistant A. baumannii, two COL-resistant K. pneumoniae, plus the parent strains of mcr-1 containing strains Table S11). With the exception of AB4016, which is highly CLR resistant (MIC >1024 μg/mL), compound 29 exhibits potent activity against all strains, lowering the CLR MIC to ≤ 2 μg/mL at 60 μM. Against AB4106, compound 29 lowers the CLR MIC to 128 μg/mL.

We next studied the effect of the addition of MgCl₂ or exogenous LPS (K. pneumoniae) or LOS (A. baumannii) had on the adjuvant activity of 29. Divalent cations like Mg²⁺ stabilize interactions between adjacent LPS or LOS molecules in the OM, and displacement of these ions leads to increased permeation.^[32] If the mechanism by which 29 enhances CLR activity involves physical disruption of the OM leading to divalent cation displacement, we would expect that MgCl₂ supplementation would cause a decrease in activity.^{[4, 32]} If 29 is directly binding LPS or LOS, exogenous LPS or LOS will compete for adjuvant binding, and a decrease in activity would be expected.

In KP2146, the presence of 10 or 20 mM MgCl₂ leads to an increase in the CLR MIC (from 512 μg/mL to 2048 μg/mL), while adjuvant activity of 29 is completely lost in the presence of 20 mM MgCl₂, and greatly diminished in the presence of 10 mM MgCl₂, only lowering the CLR MIC to 512 μg/mL (four-fold reduction) both at 60 μM (Table S12). In AB5075, 20 mM MgCl₂ does not impact the CLR MIC, and effects a more moderate loss in adjuvant activity than that observed in K. pneumoniae, with the MIC of CLR being lowered to 4 μg/mL (as compared to 0.5 μg/mL in the absence of MgCl₂) (Table S12). To investigate the effect of exogenous LPS, we used the K. pneumoniae strain KP15380, for which LPS is commercially available. In the absence of exogenous LPS, compound 29 lowers the CLR MIC from 128 g/mL to 0.25 μg/mL in this strain (Table S13). The presence of 0.5 mg/mL exogenous LPS does not affect the CLR MIC but abolishes adjuvant activity. We subsequently isolated LOS from AB5075 and found that addition of this LOS does not affect the CLR MIC or adjuvant activity (Table S13). Taken together, these results indicate that compound 29 is most likely directly binding LPS in K. pneumoniae but acting through a different mechanism in A. baumannii.

We next used a BODIPY-cadaverine displacement assay^[33], to examine binding of compound 29 to LPS/LOS from each species. When BODIPY-tagged cadaverine binds to LPS/LOS, fluorescence is quenched, and displacement of the BODIPY-cadaverine probe leads to an increase in fluorescence,^[33] as observed with COL (2 μg/mL) (Figure 3). With LPS from KP15380, compound 29 effects a dose dependent increase in fluorescence intensity, with intensity at higher concentrations of 29 (100 μM), similar to that observed for COL. An inactive compound, 27, effects no increase in fluorescence intensity. With LOS from AB5075, a dose response effect was not observed for compound 29. These results further support our hypothesis that 29 is working through separate mechanisms against the two species.

Figure 3:

BODIPY-cadaverine LPS binding assay. (A) KP15380 LPS (10 μg/mL) or (B) AB5705 LPS (10 μg/mL) and BODIPY-cadaverine (10 μM). Untreated (dark blue), 2 μg/mL COL (med blue) or increasing concentrations of compounds 27 (light blue), 29 (light green). Error bars represent standard error.

In A. baumannii, to further investigate the necessity of LOS for adjuvant activity we generated LOS deficient mutants of AB5075 as previously reported.^[28] The LOS deficient mutant is highly sensitive to CLR (MIC 0.015 μg/mL), and resistant to COL (MIC 128 μg/mL) (Table S14). In the presence of 60 μM 29 there is no change in CLR MIC (Table S14), indicating the requirement for LOS for adjuvant activity.

From initial mechanism of action studies, it appears that in the two bacterial species, K. pneumoniae and A. baumannii, 2-ABI adjuvants are exerting their CLR potentiation activity through differing mechanisms. Against K. pneumoniae, our results indicate that adjuvant 29 binds LPS and loses adjuvant activity in the presence of either exogenous MgCl_2, or exogenous LPS. In a BODIPY-cadaverine assay with LPS from K. pneumoniae, compound 29 increases fluorescence in a dose-dependent manner and achieves BODIPY-cadaverine displacement comparable to the known LPS binder COL at the highest concentration employed (100 μM). This same effect is not observed with LOS from A. baumannii. An increase in resistance to COL in A. baumannii when treated with 29, and a loss of CLR potentiation activity against an LOS deficient strain of A. baumannii indicates that 29 may be operating in a similar manner to compound 2, achieving potentiation of CLR through disrupting LOS formation in A. baumannii.^[11]

Conclusion

In summary, we have identified a series of 2-ABI containing small molecules that potentiate macrolide antibiotics against gram-negative bacteria. Lead compounds 29 and 39 lower the CLR MIC in AB5075 to 2 μg/mL (16-fold reduction) at concentrations (10 μM and 5 μM respectively) that are comparable to in vitro susceptibility levels of the clinically used adjuvant clavulanic acid^[34] (10 μM). In KP2146, both compounds lower the MIC of CLR to 2 μg/mL (256-fold) at 30 μM, and are also active against other gram-negative species. Additionally, 29 and 39 moderately potentiate vancomycin, and strongly potentiate rifampin The compounds display little microbicidal activity alone and are bacteriostatic in combination with CLR. Mechanism of action studies suggest that 2-ABI adjuvants are acting through distinct pathways between bacterial species. In K. pneumoniae, we hypothesize adjuvants are directly binding LPS based on the loss of activity of 29 upon addition of MgCl₂ or exogenous LPS, and the displacement of the BODIPY probe from LPS. In A. baumannii, compound 29 antagonizes COL and does not potentiate CLR in an LOS-deficient AB5075 mutant, indicating it may be affecting LOS biosynthesis or assembly similar to compound 2. Genetic studies are currently ongoing and these, along with proteomic studies and pull-down assays, will be used to further elucidate the mechanism of action and target(s) of this class of adjuvants. The 2-ABI scaffold represents the expansion of the possibility of combination therapy with gram-positive selective antibiotics for use against hard-to-treat gram-negative infections.