6-Bromoindirubin-3′-oxime derivatives are highly active colistin adjuvants against Klebsiella pneumoniae

Abstract

Multidrug resistant (MDR) bacterial infections have become increasingly common, leading clinicians to rely on last-resort antibiotics such as colistin. However, the utility of colistin is becoming increasingly compromised as a result of increasing polymyxin resistance. Recently we discovered that derivatives of the eukaryotic kinase inhibitor meridianin D abrogate colistin resistance in several Gram-negative species. A subsequent screen of three commercial kinase inhibitor libraries led to the identification of several scaffolds that potentiate colistin activity, including 6-bromoindirubin-3′-oxime, which potently suppresses colistin resistance in Klebsiella pneumoniae. Herein we report the activity of a library of 6-bromoindirubin-3′-oxime analogs and identify four derivatives that show equal or increased colistin potentiation activity compared to the parent compound.

Multidrug resistant (MDR) bacterial infections have become increasingly common, leading clinicians to rely on last-resort antibiotics such as colistin.

Introduction

Increasing and widespread antimicrobial resistance among pathogenic bacteria represents a considerable threat to human health.¹ Many multi-drug resistant (MDR) strains isolated from nosocomial settings are opportunistic pathogens that can easily bypass a patient's compromised immune system and cause severe, life-threatening infections.² One such opportunistic pathogen, Klebsiella pneumoniae, is a member of the notorious ESKAPE pathogens (Enterococcus faecium, Staphylococcus aureus, K. pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter species), that are named for their ability to “escape” antibiotic therapy and are responsible for many of the MDR infections encountered in hospital settings.³K. pneumoniae is implicated in hospital-acquired urinary and soft tissue infections, pneumonia and sepsis.⁴ Resistance to several antibiotic classes including polymyxins, quinolones, aminoglycosides and β-lactams has been reported,⁵ while carbapenem resistant strains of K. pneumoniae (and other Enterobacteriaceae) are considered an “urgent threat” by the Centers for Disease Control and Prevention (CDC).⁶

Colistin was approved for clinical use in the 1950s, but fell out of favor due to nephrotoxic side effects and the emergence of less-toxic antibiotic options.^7,8 However, as a result of increasing antibiotic resistance, coupled with the lack of new antibiotic classes that are active against Gram-negative bacteria, clinicians have turned to the use of colistin as a treatment of last resort for MDR Gram-negative infections.^9,10 Colistin permeabilizes the Gram-negative outer membrane by binding to lipopolysaccharide (LPS) and displacing divalent cations that help stabilize LPS. This binding/displacement causes an increase in outer membrane permeability that allows colistin access to the cytoplasmic membrane. Colistin can then bind LPS in the cytoplasmic membrane, causing subsequent disruption that leads to cytoplasmic leakage and ultimately cell death.^11,12 Colistin resistance in K. pneumoniae is mediated by chromosomal mutations in the PmrAB and PhoPQ two-component systems, which effect modification of the lipid A component of LPS with cationic residues such as phosphoethanolamine (PmrAB) and aminoarabinose (PhoPQ).^13–15 These modifications reduce the overall negative charge of LPS and reduce the affinity for cationic polymyxin antibiotics, which ultimately renders the bacteria resistant to this antibiotic class.¹² More recently, the emergence of the plasmid-borne mobile colistin resistance genes (mcr-1-10),^16,17 which also mediate cationic modification of lipid A, have provided a mechanism for rapid dissemination of polymyxin resistance into the general pathogen pool through horizontal gene transfer.

Given the well-documented lack of new antibiotics for Gram-negative bacteria,^18,19 exploration of alternative methods to combat MDR bacteria are warranted. As an alternative to developing new antibiotics, our lab and others have been actively developing small molecules that target antibiotic resistance mechanisms.²⁰ Such molecules, termed antibiotic adjuvants, typically possess limited to no standalone antibiotic activity, yet when paired with conventional antibiotics deliver a powerful combination therapy that can effectively kill MDR bacteria. Recently, we reported that analogs of the eukaryotic kinase inhibitor meridianin D (1)²¹ potentiate the effects of colistin against colistin resistant bacteria and in the process restore colistin susceptibility. Based upon this, we screened commercially available kinase inhibitor libraries for additional scaffolds that break colistin resistance and identified a number of active compounds including IMD-0354 2, sorafenib 3, AR-12 4 and 6-bromoindirubin-3′-oxime 5 (Fig. 1).²² Subsequent analog synthesis identified additional salicylanilide derivatives²³ as well as benzimidazole derivatives²⁴ (representative structures 6 and 7) that are more active than IMD-0354.

Fig. 1. Structure of kinase inhibitor adjuvant: meridianin D analog 1; IMD-0354 2; sorafenib 3; AR-12 4; 6-bromoindirubin-3′-oxime 5; salicylanilide derivative 6; benzimidazole derivative 7.

With the success of augmenting the adjuvant activity of IMD-0354, we posited that a similar analog approach would also augment the activity of 6-bromoindirubin-3′-oxime 5. When tested at 5 μM (1.8 μg mL⁻¹), 6-bromoindirubin-3′-oxime 5 reduces the colistin minimum inhibitory concentration (MIC) from 512 μg mL⁻¹ to 1 μg mL⁻¹ against the colistin resistant K. pneumoniae strain B9 (KPB9), an MIC that is under the current colistin intermediate/susceptibility breakpoint of 2 μg mL⁻¹.²⁵ Indirubin is a 3,2′-bisindole isomer of indigo and was discovered as a strong inhibitor of cyclin-dependent kinase by blocking the ATP binding pocket.²⁶ Indirubins have also been shown to act as potent glycogen synthase kinase-3β inhibitors.²⁷ Herein, we report the synthesis and activity of 33 indirubin derivatives based upon 6-bromoindirubin-3′-oxime. We discovered eight compounds that possess comparable activity to the parent compound. At 5 μM (1.8 μg mL⁻¹), our most active compound from this series reduces the colistin MIC against K. pneumoniae B9 (KPB9) to 0.5 μg mL⁻¹, and also lowers colistin MICs against multiple colistin resistant K. pneumoniae strains as well as Escherichia coli and K. pneumoniae harboring mcr-1.

Result and discussion

Synthesis and evaluation of indirubin library

We initiated this study by synthesizing 6-bromoindirubin-3′-oxime 5 and confirming its activity against K. pneumoniae B9 (Table 1). Once confirmed, we first set out to investigate the structure–activity relationship (SAR) of this scaffold, initially focusing on the brominated oxindole where structural diversity was introduced as depicted in Scheme 1A. Analogs were synthesized following the previously established synthetic procedure outlined in Scheme 1B.²⁶ The synthesis of indirubins was based on the dimerization of an appropriately substituted isatin with various 3-acetoxyindoles, followed by refluxing with hydroxylammonium chloride in pyridine.²⁶ 3-Acetoxyindoles were prepared via a two-step procedure that started with reaction of the appropriate indole starting material with iodine and potassium iodide in the presence of sodium hydroxide in methanol/water (9 : 1). The crude product was then treated with silver acetate in acetic acid at 90 °C that, following purification, delivered the target 3-acetoxyindoles.

Initial screen of indirubin analogs for colistin resistance suppression.

Compound	MIC (μM)	Colistin MICa (μg mL−1)	Compound	MIC (μM)	Colistin MICa (μg mL−1)
—	—	512		—	512
5	>200	1(512)	17	>200	8(64)
12	>200	>16	18	>200	>16
13	>200	>16	19	>200	>16
14	>200	>16	20	>200	>16
15	>200	>16	21	>200	>16
16	>200	4(128)

^aFold reduction in colistin MIC indicated in parentheses.

Scheme 1. A. Structures of first set of indirubin derivatives. Scheme 1B synthetic route to indirubin derivatives. Conditions: (a) KI, NaOH, I₂, H₂O/MeOH; (b) CH₃COOAg, CH₃COOH, reflux; (c) Na₂CO₃, MeOH; (d) NH₂OH·HCl, pyridine.

The MIC of each derivative was first determined to quantify standalone toxicity (Table S1†), and all exhibited standalone MICs of greater than 200 μM. Each derivative was then assayed for colistin potentiation against KPB9 at a concentration of 5 μM (Table 1). Replacing the bromo group with either hydrogen (12), fluorine (13), methoxy (14), or methyl (15) led to a significant reduction in activity, with none of the derivatives reducing the colistin MIC to 16 μg mL⁻¹ (highest concentration tested). Replacement of the bromo substituent with a chloro group (16) returned a compound that was four-fold less active than the parent, reducing the colistin MIC 128-fold to 4 μg mL⁻¹. We additionally determined that replacement with a trifluoromethyl group (17) was also detrimental to activity and delivered a compound that effected only a 64-fold decrease in colistin MIC.

With only chloro and bromo substitutions maintaining colistin adjuvant activity, we next focused on studying the impact of moving either halogen from C6 to C5 or C7 (compounds 18–21). All four compounds did not reduce the colistin MIC to 16 μg mL⁻¹ (highest concentration tested), indicating that a C6-Br is the optimal substituent pattern from these analogs (Fig. 2).

Fig. 2. Structures of second set of indirubin derivatives.

With no improvement noted from modifying the isatin fragment, we next evaluated the impact that substituents on the indole moiety of compound 5 had upon colistin potentiation when screened at 5 μM. This set of analogues was synthesized following the route shown in Scheme 1, utilizing appropriately substituted indoles. Incorporation of a bromo substituent at the 5′- (22) or 6′- (23) position resulted in a moderate reduction in colistin potentiation (MICs of 4 and 8 μg mL⁻¹, respectively) in comparison to 5, while a bromo substituent at the 7′-position (24) resulted in comparable activity (1 μg mL⁻¹). Incorporation of a chloro substituent at either 5′ (25), 6′ (26), or 7′ (27) also delivered active adjuvants, suppressing the colistin MIC to 2 μg mL⁻¹, 0.5 μg mL⁻¹, and 1 μg mL⁻¹ respectively. Fluorine substitution was also tolerated at 5′ (28) and 7′ (29), with both compounds suppressing the colistin MIC to 2 μg mL⁻¹. When the fluoro group was moved to the 6′ position (30), activity was severely compromised, and this adjuvant did not reduce the colistin MIC to 16 μg mL⁻¹ (highest concentration tested).

Given the tolerance of 5′ and 7′ substitution, we expanded diversity at both positions by incorporating a phenyl, methyl, or trifluoromethyl group. Phenyl substitution at either the 5′ (31) or 7′ (32) abrogated activity, while loss of potentiation was observed with a 7′-methyl (33). Incorporation of a 5′-methyl (34), 5′-trifluoromethyl (35), or a 7′-trifluoromethyl (36) delivered active adjuvants, however all three compounds were less active than the parent compound 5 and returned colistin MICs of 4, 8, and 2 μg mL⁻¹ respectively.

With 6′-chlorine substitution delivering the most active adjuvant of the series, we next studied the impact on potentiation that addition of a second chloro substituent had upon activity. The 5′,6′-dichloro derivative (37) was >32-fold less active than 26, while the 5′,7′ analog (38) was similarly devoid of activity. Given this loss of activity, we then probed whether a 5′,7′-difluoro derivative (39) would show similar loss in activity. Interestingly, 39 was equipotent to parent compound 5. Finally, to further establish the necessity of the 6-Br, we synthesized a des-bromo analog of compound 27 (40, 6-H, 7′-Cl). Similar to what was observed for removal of the bromo substituent from compound 5 (compound 12), analog 40 was inactive (Table 2).

Second set of indirubin analogs colistin resistance suppression.

Compound	MIC (μM)	Colistin MICa (μg mL−1)	Compound	MIC (μM)	Colistin MIC (μg mL−1)
—	—	512		—	512
22	>200	4(128)	32	>200	>16
23	>200	8(64)	33	>200	>16
24	>200	1(512)	34	>200	4(128)
25	>200	2(256)	35	>200	8(64)
26	>200	0.5(1024)	36	>200	2(256)
27	>200	1(512)	37	>200	>16
28	>200	2(256)	38	>200	>16
29	>200	2(256)	39	>200	1(512)
30	>200	>16	40	>200	>16
31	>200	>16

^aFold reduction in colistin MIC indicated in parentheses.

Indirubins have been documented to have poor solubility, and previous work in the kinase field has established that incorporation of hydrophilic groups on the oxime oxygen can generate active kinase inhibitors that are more hydrophilic.²⁸ To test whether this was the case for colistin potentiation, we synthesized derivatives 41, 42, and 43 (Scheme 2), as similar substitutions have been shown to be tolerated for kinase inhibition.²⁸ All three compounds were devoid of colistin adjuvant activity.

Scheme 2. Synthetic route to compounds 41, 42 and 43. Conditions: (a) 4-(2-bromoethyl)-2,2-dimethyl-1,3-dioxolane, TEA, DMF; (b) HCl, MeOH; (c) 1,2-dibromoethane, TEA, DMF; (d) methylpiperazine, DMF; (e) piperazine, DMF.

With eight compounds (24–29, 36 and 39) exhibiting comparable activity to compound 5 at 5 μM, a dose–response study was conducted to determine if any of these next generation derivatives had superior activity at lower concentrations (Table 3). At 3 μM (1.1 μg mL⁻¹), compound 5 lost colistin potentiation activity against KPB9, while four derivatives 24, 26, 27 and 29 still reduced the colistin MICs to or below the breakpoint (2 μg mL⁻¹). At 1 μM, no compound suppressed the colistin MIC to the breakpoint, with 26 being the most active, suppressing the colistin MIC 32-fold to 16 μg mL⁻¹. Based upon these results, compounds 24, 26, 27 and 29 were selected for additional testing.

Dose response data for active analogs against KPB9.

Compound	Colistin MIC with 5 μM compound	Colistin MIC with 3 μM compound	Colistin MIC with 1 μM compound
—	512	512	512
5	1(512)	32(16)	>64a
24	1(512)	2(256)	32(16)
25	2(256)	16(32)	>64a
26	0.5(1024)	1/2(512/256)	16(32)
27	1(512)	2(256)	32(16)
28	2(256)	32(16)	>64a
29	2(256)	2(256)	>64a
36	2(256)	4(128)	64
39	1(512)	4(128)	>64

^aHighest concentration tested.

We first determined the adjuvant activity of these four compounds, along with compound 5, at 5 μM against three additional colistin resistant K. pneumoniae clinical isolates (KPC3, KPF9, and KPI2) as well as a K. pneumoniae strain harboring a plasmid containing the mcr-1 gene. All five compounds were highly active, returning colistin MICs of ≤8 μg mL⁻¹. Of these analogs, only compound 26 reached the colistin breakpoint levels of ≤2 μg mL⁻¹ at 5 μM (2 μg mL⁻¹) against all K. pneumoniae strains. We then tested these new derivatives for activity against colistin resistant strains of A. baumannii (AB3941), E. coli (EC25922^mcr-1), and P. aeruginosa (TRPA162). All five compounds showed no standalone microbicidal activity (Table S3†) and were screened for potentiation at 5 μM. None of the compounds suppressed the colistin MIC in either AB3941 or TRPA162 to 16 μg mL⁻¹ (highest concentration tested), while all five analogs reduced the colistin MIC in EC25922^mcr-1 to ≤2 μg mL⁻¹ (Table 4).

Potentiation of colistin by lead compounds 5, 24, 26, 27 and 30 against select members of a panel of colistin resistant strains.

Compound	KP C3	KPF9	KP I2	KPF2210291mcr-1	AB 3941	EC 25922mcr-1	TRPA 162
—	128	256	512	16	1024	16	1024
5	4	4	8	2	>16	2	>16
24	4	4	4	2/4	>16	1	>16
26	2/1	2	2	1	>16	2	>16
27	4	4	4	2	>16	2	>16
29	4	4	4/8	2	>16	2	>16

Time-kill curves were then constructed for the most active compound 26 in the presence of 0.5 μg mL⁻¹ (1× combination MIC), 2 μg mL⁻¹ (4× combination MIC), and 4 μg mL⁻¹ (8× combination MIC) colistin against KPB9 (Fig. 3). At 5 μM (2.0 μg mL⁻¹), 26 showed no standalone impact on bacterial growth, while a one log reduction in CFU mL⁻¹ was noted after 8 h when combined with 0.5 μg mL⁻¹ colistin. At 2 μg mL⁻¹ colistin in the presence of 5 μM 26, we observed a steep drop in initial CFUs after 2 h (two log reduction), then a slight rebound to level at a ca. one log reduction in CFUs out to the 24 h timepoint. Treatment with 5 μM 26 and 4 μg mL⁻¹ colistin suppressed the CFU count below the detection limit at all timepoints from 6–24 h.

Fig. 3. Time-kill curves for compound 26 and colistin against KPB9. Blue: control, red: 5 μM 26, grey: 5 μM 26 plus 0.5 μg mL⁻¹ colistin, yellow 5 μM 26 plus 2 μg mL⁻¹ colistin, green: 5 μM 26 plus 4 μg mL⁻¹ colistin. Dashed line indicates limit of detection.

Indirubins do not lyse red blood cells at their active adjuvant concentrations

In order to gain a preliminary insight into the eukaryotic toxicity of these inidrubin derivatives, we quantified the degree of hemolysis effected by the lead compound 26 upon defibrinated sheep blood. No lysis was observed for 26 at concentrations up to 400 μM (highest concentration tested) (Table S5†).

Lead compound 26 does not affect lipid A modification

As lipid A modification with either phosphoethanolamine or aminorhabinose has been established as the main mechanism for polymyxin resistance in most Gram-negative bacteria, we tested whether 26 reversed or inhibited this modification by analyzing the lipid A portion of LPS from KPB9. To this end, KPB9 was grown in either the absence or presence of 5 μM (2.0 μg mL⁻¹) 26 for eight hours, the resulting bacterial cells were then isolated, and lipid was A extracted. MALDI-TOF analysis of the lipid A from treated and untreated cells indicated no difference in the lipid A profile between the two samples. As the current MALDI-TOF method is not quantitative, one of two conclusions can be drawn. First, the mechanism of action is independent of lipid A modification. Second, achieving colistin susceptibility in colistin resistant bacteria does not require full reversal of lipid A modifications, and that slight reductions in lipid A modification underpinned by small molecule treatment, which would not be quantifiable with the current MALDI-TOF protocol, are enough to reverse colistin resistance. A quantitative method for quantifying lipid A modification is currently under active development and once rigorously established will be employed to differentiate these two mechanisms.

Conclusions

In conclusion, we have synthesized and tested a library of indirubin analogs for restoration of colistin susceptibility in colistin resistant bacteria. With no observed standalone toxicity (MICs >200 μM), eight compounds exhibit similar activity levels at 5 μM in comparison to the original lead 5. Unlike the lead, four of these derivatives retain activity at 3 μM. The most active compound, 26, lowered the colistin MIC in KPB9 to 0.5 μg mL⁻¹ at 5 μM (2.0 μg mL⁻¹) and to 2 μg mL⁻¹ at 3 μM (1.2 μg mL⁻¹). This lead compound achieved colistin breakpoint levels in additional strains of colistin-resistant K. pneumoniae as well as an E. coli strain harboring the mcr-1 gene. No hemolysis was noted for 26 up to 400 μM (80× its active concentration against all strains tested), while mechanism of action studies indicates that colistin potentiation is not dependent upon full reversal of lipid A modification. While indirubin derivatives have known liabilities in vivo, the ability to generate structurally similar analogs with diverse colistin potentiation activity will allow us to deliver a set of probes that will help quantify the full extent that lipid A modification must be reversed to achieve colistin sensitivity. Such studies are currently ongoing and will be reported in due course.

Conflicts of interest

The authors report no conflicts of interest.