An unusual class of anthracyclines potentiate Gram-positive antibiotics in intrinsically resistant Gram-negative bacteria

Abstract

Objectives

An orthogonal approach taken towards novel antibacterial drug discovery involves the identification of small molecules that potentiate or enhance the activity of existing antibacterial agents. This study aimed to identify natural-product rifampicin adjuvants in the intrinsically resistant organism Escherichia coli.

Methods

E. coli BW25113 was screened against 1120 actinomycete fermentation extracts in the presence of subinhibitory (2 mg/L) concentrations of rifampicin. The active molecule exhibiting the greatest rifampicin potentiation was isolated using activity-guided methods and identified using mass and NMR spectroscopy. Susceptibility testing and biochemical assays were used to determine the mechanism of antibiotic potentiation.

Results

The anthracycline Antibiotic 301A₁ was isolated from the fermentation broth of a strain of Streptomyces (WAC450); the molecule was shown to be highly synergistic with rifampicin (fractional inhibitory concentration index = 0.156) and moderately synergistic with linezolid (FIC index = 0.25) in both E. coli and Acinetobacter baumannii. Activity was associated with inhibition of efflux and the synergistic phenotype was lost when tested against E. coli harbouring mutations within the rpoB gene. Structure–activity relationship studies revealed that other anthracyclines do not synergize with rifampicin and removal of the sugar moiety of Antibiotic 301A₁ abolishes activity.

Conclusions

Screening only a subsection of our natural product library identified a small-molecule antibiotic adjuvant capable of sensitizing Gram-negative bacteria to antibiotics to which they are ordinarily intrinsically resistant. This result demonstrates the great potential of this approach in expanding antibiotic effectiveness in the face of the growing challenge of resistance in Gram-negatives.

Introduction

Antibiotic resistance, in particular multidrug-resistant (MDR) strains of bacteria, represents a threat to modern medicine and the successful treatment of bacterial infections. The CDC in the USA has recently released a report highlighting the imminent threat that these resistant organisms represent.¹ A combination of improved infection control and an arsenal of new drugs targeting methicillin-resistant Staphylococcus aureus (MRSA)² has led to a reduction in the occurrence of MRSA bacteraemias in both UK and US hospitals.^3–6 However, this feat has been somewhat overshadowed by the spread of antimicrobial drug resistance amongst Gram-negative human pathogens,⁷ a phenomenon that dramatically diminishes therapeutic options.^2,8,9 In addition to ‘acquired’ resistance in Gram-negative pathogenic bacteria, the wealth of Gram-positive antibiotics at our disposal are redundant due to the relative intrinsic resistance of these organisms.¹⁰ The ‘intrinsic resistome’ of bacteria has been shown to be a complex network of various elements that contribute to this phenotype.¹⁰ Membrane-spanning pumps that actively efflux drugs and the presence of the outer membrane (OM), which effectively hinders the entry of certain molecules, appear to be the main contributors to the intrinsic resistome of Gram-negative pathogenic bacteria.¹⁰ Indeed, in the past decade only two antibacterial agents possessing Gram-negative activity have been launched: tigecycline¹¹ (an anti-MRSA agent exhibiting enhanced in vitro Gram-negative activity in comparison with other tetracyclines) and doripenem¹² (a carbapenem showing similarities to meropenem). There are many factors contributing to the slow progress in discovery and development of anti-Gram-negative agents; at the forefront is the difficulty associated with identifying molecules that can penetrate the OM and are recalcitrant to efflux.

The protection bestowed by these intrinsic resistance elements becomes apparent when they are compromised or deleted, resulting in sensitization to a variety of antibacterial agents.^13,14 This observation reveals an alternative approach to traditional drug discovery; if we could target these intrinsic resistance mechanisms with small molecules, we could extend the activity spectrum of numerous existing clinically used antibacterial agents, as a combination therapy.¹⁰ More than 50 years ago, cyclic cationic lipopeptides such as polymyxin B (PMB) and colistin (polymyxin E) were discovered.¹⁵ These molecules can be active alone or in combination with hydrophobic antibiotics, permeabilizing the OM and thus allowing entry of otherwise excluded molecules into the cell.¹⁵ Despite associated toxicities of the polymyxins, colistin has been re-introduced for the treatment of MDR strains of Acinetobacter baumannii, Pseudomonas aeruginosa and Klebsiella pneumoniae when no other therapeutic options are available.^16,17 Furthermore, inhibition of the tripartite antibiotic efflux assemblies that span the bacterial membrane offers an appealing approach, since one efflux pump is capable of mediating resistance to a number of structurally diverse drugs.^18,19 However, despite the identification of molecules capable of inhibiting antibiotic efflux in different bacterial species, the introduction of an efflux pump inhibitor into the clinic continues to be awaited.^2,8,9 Further to inhibition of the more traditional components of intrinsic resistance (the OM and efflux), recent studies have shown that the intrinsic resistome is more complex than originally foreseen, with a number of additional genes contributing to relative intrinsic resistance levels.^14,20,21 These additional, less obvious contributors extend the spectrum of targets for the discovery and development of small molecules inhibiting elements of the intrinsic resistome. Indeed, we have proven success in the discovery of antibiotic adjuvants,²² and more recently we have identified small molecules that potentiate the Gram-positive antibiotic novobiocin against E. coli.²³ This large-scale chemical screen (n = 30 000 compounds) identified molecules whose adjuvant activity was related to changes in cell shape and permeability.²³ Our previous findings highlight the potential of combination screening for the identification of small molecules targeting the intrinsic antibiotic resistome, offering a novel and innovative approach to traditional antibiotic discovery.

A large majority of antibiotics and antifungals on the market today were discovered in the ‘Golden Age’ of antibiotic discovery (1940s to 1960s), largely through the isolation and screening of actinomycete extracts. The success of Actinomyces as producers of a wide array of secondary metabolites with novel therapeutic properties has led to renewed interest in natural product discovery. Based on this success, we have recently compiled a collection of >9000 actinomycete isolates [the Wright Actinomycete Collection (WAC)].²⁴ In this study, we screened a subsection of this collection in the presence of subinhibitory concentrations of the Gram-positive antibiotic rifampicin to identify successful antibiotic–adjuvant combinations.

Materials and methods

Bacterial strains, growth conditions and reagents

Bacterial strains used in this study are listed in Table S1 (available as Supplementary data at JAC Online). All strains were grown in Mueller–Hinton II broth (cation adjusted) (Becton, Dickinson and Co., Franklin Lakes, NJ), at 37°C with aeration for 18 h. Antibiotics used were rifampicin, novobiocin, erythromycin, vancomycin, linezolid, PMB and polymyxin B nonapeptide (PMBN) (Sigma, Mississauga, ON). Anthracyclines purchased were doxorubicin (AK Scientific, Union City, CA) and daunorubicin (Erfa Canada, Montréal, QC). Susceptibility testing was performed using the microdilution broth method, according to CLSI guidelines (M7-A9).²⁵ The MIC was defined as the lowest concentration that yielded no visible growth at an optical density of 600 nm (OD₆₀₀).

Screening the natural product library (NPL)

The NPL is composed of microbial fermentation extracts derived from strains within the WAC.²⁴ E. coli BW25113 was used to screen the NPL in the absence of and in combination with rifampicin at a concentration of 2 mg/L (0.25 MIC). The robustness of the screen was measured by calculating the Z′ score using the equation below (where σ_p and σ_n are the standard deviations of the positive and negative controls, respectively, and μ_p and μ_n are the means of the positive and negative controls, respectively):²⁶

Screening was performed in duplicate in 384-well flat-bottom plates with a final volume of 30 μL. A Beckman Biomek FX liquid handler (Beckman Coulter Inc., Fullerton, CA) was used to dispense 1.5 μL of fermentation extracts from the NPL. This was followed by addition of 28.5 μL of diluted cell culture (prepared according to CLSI susceptibility testing guidelines)²⁵ containing rifampicin and the same for cell culture containing no antibiotic. Positive growth controls (DMSO only) and negative growth controls (sterile medium) were included in rows 1, 2, 23 and 24 of each plate and the plates were incubated with aeration at 37°C. The OD₆₀₀ was measured after 18 h, growth controls were used to generate percentage growth data, and replicates were plotted against each other. To identify extracts that only reduced the growth of E. coli in combination with rifampicin, the OD₆₀₀ reading from the screen in the absence of rifampicin was subtracted from the OD₆₀₀ reading in the presence of rifampicin. Resulting hits with an overall negative growth percentage represented successful extract and rifampicin combinations.

Re-culturing of the WAC strains and preparation of extracts confirmed hits; in 96-well plates, 5 μL of each extract was combined with diluted cell culture in the absence and presence of 2 mg/L rifampicin to a final volume of 100 μL, plates were incubated at 37°C with aeration for 18 h and the OD₆₀₀ was measured.

Characterization of WAC450 and purification of antibiotic adjuvant natural products

Genomic DNA for WAC450 was prepared by standard methods.²⁷ For 16S rRNA analysis, PCR was performed using primers F27 [5′-AGAGTTTGATC(A/C)TGGCTCAG-3′] and R1492 [5′-TACGG(C/T)TACCTTGTTACGACTT-3′].²⁸ Amplicons were sequenced at the MOBIX Central Facility (McMaster University, Hamilton, Canada).

For large-scale purification of antibiotic adjuvants, WAC450 was grown in 6 L of Bennett's medium for 6 days at 30°C with aeration. The cell culture was filtered and the filtrate incubated with 2% (w/v) HP-20 (Diaion) resin for 1 h. The resin was loaded onto a 1 L column and molecules were eluted by gravity with H₂O (2 L), 20% MeOH (1 L), 40% MeOH (1 L), 60% MeOH (1 L), 80% MeOH (1 L) and 100% MeOH (1 L), yielding six fractions. Fractions were assessed for bioactivity using rifampicin potentiation as a guide throughout; activity was observed in fractions that were deep red in colour, which were eluted in the 80–100% MeOH fractions. These fractions were pooled and concentrated under reduced pressure with 2% (w/v) silica gel 60 and loaded onto a 40 g RediSep Rf silica FLASH chromatography column (Teledyne Isco, Lincoln, NE). Secondary metabolites were semi-purified using a CombiFlash Rf (Teledyne) chromatography system and eluted with 100% chloroform followed by a linear gradient to 100% chloroform:methanol:acetic acid:H₂O, 80 : 20:14:6 (v/v/v/v). Fractions were concentrated under reduced pressure and assayed for bioactivity. Active fractions were further purified by passage through a Sephadex LH-20 column (100 mL), with 100% methanol. Purity of fractions was assessed by reverse-phase HPLC coupled with mass spectrometry detection (LC/MS).

LC/ESI/MS data were obtained by using an Agilent 1100 Series LC system (Agilent Technologies Canada, Inc.) and a QTRAP LC/MS/MS System (Applied Biosystems). Reverse-phase HPLC was performed using a C18 column (SunFire C18 5 μm, 4.6 × 50 mm, Waters) with an Agilent 1100 LC binary pump at a flow rate of 1 mL/min, under the following conditions: isocratic 5% solvent B (0.05% formic acid in acetonitrile) and 95% solvent A (0.05% formic acid in water) for 1 min, followed by a linear gradient to 97% B over 10 min.

Compound structures were confirmed by 1D and 2D NMR experiments on a Bruker AVIII 700 MHz instrument equipped with a cryoprobe in an appropriate deuterated solvent. Chemical shifts are reported in ppm relative to tetramethyl silane using the residual solvent signal as an internal signal. High-resolution mass spectra (HRMS) were obtained using an XL Orbitrap Hybrid mass spectrometer (Thermo-Fisher, Bremen, Germany) equipped with electrospray interface operated in positive ion mode.

Antibiotic 301A₁ adjuvant mechanism-of-action studies

The fluorescent probe 3′-dipropylthiacarbocyanine (DiSC₃)²⁹ was used to measure the transmembrane potential in E. coli BW25113. Cells were prepared by addition of a 0.001% inoculum to 50 mL of Luria–Bertani (LB) broth; cultures were incubated at 37°C for 4 h with aeration. The assay was performed as previously described,²² with PMB (20 mg/L) as a positive control. The cells were incubated with DiSC₃ for 2 min and the dye was allowed to stabilize, compounds were injected at 200 s and the change in fluorescence was monitored for a further 200 s. For Antibiotic 301A₁, a concentration of 32 mg/L was used.

To assess the effect of Antibiotic 301A₁ on efflux in E. coli BW25113, the Nile Red efflux assay³⁰ was used. The assay was performed as described by Bohnert et al.,³⁰ with some minor modifications. Nile Red and carbonyl cyanide m-chlorophenylhydrazone (CCCP) were purchased from Sigma. The efflux-deficient ΔtolC E. coli BW25113 strain³¹ was used as a negative control, and DMSO concentrations were kept constant in all assays. A final volume of 3 mL of diluted cells was used in the cuvette for optimal fluorescence traces. Following stabilization of the dye, rapid energization was triggered by the addition of 150 μL of 1 M glucose and changes in fluorescence were measured in a spectrofluorimeter (Photon Technology International) with a slit width of 10 nm and excitation and emission wavelengths of 552 and 636 nm, respectively. The mixture was stirred throughout with a metal stirrer. Antibiotic 301A₁ was used at concentrations of 25, 50 and 100 μM.

Generation of rifampicin-resistant E. coli BW25113 mutants

Rifampicin-resistant mutants were generated by plating of ∼10⁹ E. coli BW25113 cells from saturated cultures onto LB agar containing 125 mg/L and 250 mg/L rifampicin. Colonies were selected after 18 h of incubation at 37°C. PCR was used to amplify the rpoB gene using the following primers: rpoB I forward, 5′-GACAGATGGGTCGACTTGTCAGCG-3′; rpoB I reverse, 5′-AGGTGGTCGATATCATCGACT-3′; rpoB II forward, 5′-TCGAAGGTTCCGGTATCCTGAGC-3′; and rpoB II reverse, 5′-GGATACATCTCGTCTTCGTTAAC-3′.³² Amplicons were sequenced at the MOBIX Central Facility (McMaster University, Hamilton, Canada).

Cytotoxicity studies

Human Embryonic Kidney 293 (HEK293) cells were used for cytotoxicity assays. The cell line was cultured in minimal essential medium (MEM) supplemented with 10% fetal bovine serum (Life Technologies Inc., Burlington, ON) in a controlled environment (37°C, 5% CO₂). The cells were seeded at ∼10⁴ cells/mL and grown to confluence. The medium was removed and immediately replaced with fresh medium containing serial dilutions (125–0.05 mg/L) of test compound and incubated for 18 h. The live cell count was carried out in a Burker chamber using microscopy. Cell viability was assessed by observation of morphology (‘rounding up’) and Trypan Blue staining using microscopy.

Results

Intrinsic resistance of E. coli to rifampicin

Intrinsic resistance to different classes of Gram-positive antibiotics in Gram-negative bacteria is often the result of a synergistic combination involving both active efflux and the OM.¹⁰ In the case of rifampicin, reduced permeability resulting from the presence of the OM represents the most significant contributor to the intrinsic resistance of E. coli to rifampicin.^13,15,33 Active efflux of rifampicin by the AcrAB pump has also been implicated in the intrinsic resistance of E. coli to rifampicin, albeit to a much lower level in comparison with the OM.^34,35 All screening and susceptibility testing was performed in the Keio collection parental strain,³¹ E. coli BW25113, and with single-gene knockouts from this collection.³¹ By deletion of the gene encoding BamB (ΔbamB), which renders E. coli hyper-permeable to many small molecules,³¹ we confirmed that E. coli BW25113 is sensitized to rifampicin if the OM is compromised (Table 1). However, susceptibility levels were not affected when the tripartite efflux protein TolC was absent (Table 1). Since active efflux has previously been implicated in the intrinsic resistance of E. coli to rifampicin, we also investigated the effect of deletion of the genes encoding AcrA (ΔacrA) and AcrB (ΔacrB). Consistent with previous studies,^34,35 we found a 2-fold decrease (from 8 to 4 mg/L) of the MIC in the ΔacrB deletion strain. However, susceptibility changes were unaffected if the periplasmic protein AcrA was lacking (data not shown).

Table 1.

Susceptibility (mg/L) of E. coli BW25113, derivatives^a of this strain and Bacillus subtilis to rifampicin and members of the anthracyclinone family of molecules

Antibiotic/compound	MW (Da)	E. coli BW2511334	E. coli BW25113 ΔtolC34	E. coli BW25113 ΔbamB34	E. coli BW25113 ΔtolC ΔbamB	B. subtilis 168
Rifampicin	822.94	8	8	<0.125	<0.125	0.0125
Doxorubicin	579.98	>256	4	64	4	4
Daunorubicin	527.52	>256	4	64	4	4
Antibiotic 301A2	670.71	>256	8	128	16	16
Antibiotic 301A1	684.73	>256	8	128	16	8
Avidinorubicin	1214	256	64	128	128	16
Unnamed	1244	256	32	32	32	8
Antibiotic 301B	1261.29	64/128	ND	ND	ND	ND

ND, not determined.

^aAll strains originating from the Keio collection were verified by PCR.

Screening the NPL

We screened a subsection of our NPL (1120 extracts), derived from 560 strains of actinomycetes. Screening was performed in duplicate, at 0.25 MIC (2 mg/L rifampicin) based on our previous experience in combination screens.²³ The Z′ score³⁶ was calculated to be >0.8, consistent with a robust screening methodology. In addition to screening in the presence of rifampicin, E. coli BW25113 was screened in the absence of target antibiotic as a negative control to discount extracts that possessed antibacterial properties themselves, rather than acting as an adjuvant. Growth controls were included for each plate, allowing percentage growth to be calculated; each replicate was subsequently plotted against each other (Figure 1a). Following normalization of the data to the controls, extracts were identified that potentiated the activity of rifampicin (Figure 1b). In line with our previous work,²³ hits were defined as combinations of extract and antibiotic that inhibited bacterial growth to <60%; overall, four hits were identified, resulting in a hit rate of 0.36%. Small-scale fermentation of strains identified as producers of antibiotic adjuvants confirmed the reproducibility of secondary metabolite production and rifampicin potentiation. We decided to isolate and identify the active compound from the extract (WAC450) exhibiting the most prominent rifampicin potentiation in the screen (Figure 1b).

Figure 1.

Replica plots of NPL screening results. (a) Screening of E. coli BW25113 in the presence of 2 mg/L rifampicin against 1120 microbial fermentation extracts derived from 560 strains of actinomycetes within the WAC. (b) Subtraction of screening results from E. coli BW25113 against the same extracts, in the absence of antibiotic, from (a), to discount extracts that possess antibacterial properties themselves. The fermentation extract from WAC450, exhibiting the most prominent rifampicin potentiation, is circled.

Isolation and identification of rifampicin adjuvant natural products from WAC450

WAC450 was originally isolated from a soil sample collected from Gillam, Manitoba, Canada. The active compound from WAC450 was isolated using rifampicin potentiation as a guide during purification. Initially, conditioned medium, cell pellet and solid-medium fermentation extracts were assessed for adjuvant activity. All conditions exhibited adjuvant activity; however, the culture broth possessed the most potent activity, followed by solid medium and low activity in the pellet extract. Analysis of 16S rDNA determined WAC450 to be related to Streptomyces lysosuperificus (98% identity); fermentation products were deep red in colour and during activity-guided purification rifampin-potentiation activity was associated with the red fractions. Identity of the WAC450 red active compound was determined by LC/MS. It was observed that the prominent active peak possessed UV/visible spectroscopic properties indicative of an anthracycline, as seen in our previous studies of the anti-cancer drug doxorubicin.³⁷ A search for known compounds within the absorbance range of 475–485 nm, and the 685.26 m/z ion detected in the positive ion mode, identified Antibiotic 301A₁ (Figure 2), an anthracycline antibiotic with a mass of 684.23 g/mol, as a likely candidate. Interestingly, a number of additional ions were also detected in the positive ion mode (671.10, 1215.84, 1231.74, 1246.80 and 1262.48 m/z) within the culture broth and pellet methanolic extracts of this strain, which also had UV/visible spectroscopic properties indicative of the anthracyclines. Indeed, all compounds in the mixture had the same aglycone scaffold, but are decorated with different complex amino and/or nitro sugars (Figure 2). This family of antibiotics belong to the anthracyclinone family,^38–42 possessing an aglycone structurally related to nogalamycin (Figure 2),^43,44 with a positively charged bicyclo amino sugar but lacking a methyl ester on one end of the aglycone. The structure of our purified molecule was determined by high-resolution MS and multidimensional NMR and found to be consistent with previous reports (Table S2, available as Supplementary data at JAC Online).^39,41

Figure 2.

Chemical structures of anthracyclinone and anthracycline molecules identified or referred to in this study.

Antibiotic 301A₁ potentiates rifampicin activity against E. coli

Members of this anthracyclinone family have documented antibacterial activity; however, this activity is limited to Gram-positive organisms.⁴⁵ Susceptibility testing of purified compounds identified within the WAC450 culture broth (Figure 2) revealed that the molecules with higher molecular weight (MW) (>700 Da), possessing additional sugar moieties to that of Antibiotic 301A₁, had marginally higher anti-Gram-negative activity than lower-MW molecules (Antibiotic 301A₁ and Antibiotic 301A₂) (Table 1). Furthermore, susceptibility testing with an E. coli strain deficient in efflux (ΔtolC), the hyper-permeable strain (ΔbamB) and a strain harbouring both mutations revealed that a combination of both efflux and the OM contributes to the intrinsic resistance of E. coli to Antibiotic 301A₁, with a greater emphasis on efflux (Table 1). Due to the increased anti-Gram-negative activity of the higher-MW molecules, we focused our attention on the smaller-MW molecules, Antibiotics 301A₁ and 301A₂, which have minimal antibacterial activity alone, but synergize in combination with rifampicin. Furthermore, during purification of these smaller-MW molecules, we obtained a significantly higher yield of pure Antibiotic 301A₁ than Antibiotic 301A₂. For convenience, further mechanistic studies of these adjuvant molecules focused on Antibiotic 301A₁.

To determine whether Antibiotic 301A₁ synergizes in E. coli BW25113 or has an additive effect in combination with rifampicin, the microdilution chequerboard method was used to determine the FIC index (FICI).⁴⁶ An FICI of ≤0.5 is indicative of a synergistic combination of molecules; indices of 1 and ≥4 are considered to indicate no interaction and antagonism respectively.⁴⁷ Antibiotic 301A₁ strongly synergized with rifampicin (FICI = 0.156). EUCAST defines rifampicin resistance in the intrinsically susceptible organism S. aureus as an MIC of ≥0.5 mg/L.⁴⁸ The MIC of rifampicin for E. coli BW25113 was reduced to 0.25 mg/L in combination with 32 mg/L Antibiotic 301A₁ (Figure 3a).

Figure 3.

Antibiotic 301A₁ synergizes with Gram-positive antibiotics in E. coli BW25113. Chequerboard analysis showing combinatorial synergistic effect of Antibiotic 301A₁ with rifampicin (a), erythromycin (c) and linezolid (d). No interaction is evident for novobiocin (b).

Evaluating the antibiotic and bacterial adjuvant spectrum of Antibiotic 301A₁

To determine whether Antibiotic 301A₁ synergizes with other antibiotics that also have poor activity against Gram-negative bacteria, the FICI was calculated for the molecule in combination with novobiocin, erythromycin and linezolid (Figure 3b–d). Growth of E. coli was unaffected by the combination of Antibiotic 301A₁ with novobiocin (FICI = 0.625). However, there was synergistic inhibition of growth when combined with the synthetic antibiotic linezolid (FICI = 0.25) or erythromycin (FICI = 0.156).

To assess the spectrum of adjuvant activity in other Gram-negative bacteria, we tested synergy against a panel of Gram-negative pathogenic bacteria (P. aeruginosa PAO1, A. baumannii BM4587 and K. pneumoniae H0142423). The susceptibilities of these organisms to linezolid and rifampicin were first determined; all three strains were intrinsically resistant to both antibiotics, with a much higher level of resistance observed in the K. pneumoniae strain (Table 2). To ascertain whether the OM contributes to the intrinsic rifampicin resistance in these organisms, susceptibility testing was performed in the presence of the OM-permeabilizing compound PMBN.⁴⁹ Following addition of PMBN (30 mg/L) the MICs of rifampicin for P. aeruginosa and A. baumannii were reduced to <0.5 mg/L and the MIC was significantly reduced to 64 mg/L for K. pneumoniae (Table 2). However, PMBN had little effect on the MICs of linezolid for both A. baumannii and K. pneumoniae, but did reduce the MIC of linezolid for P. aeruginosa (Table 2). We then used an efflux-deficient strain of P. aeruginosa K2732 (ΔmexB, ΔmexX, ΔmexCD-oprJΔmexEF-oprN; MexAB-OprM⁻ MexCD-OprJ⁻, MexXY⁻, MexEF-OprN⁻)⁵⁰ and determined the susceptibility of this strain in the presence of linezolid. The MIC was reduced from >512 mg/L in wild-type P. aeruginosa to 32 mg/L in the efflux-deficient strain (Table 2), confirming that efflux plays a significant role in the intrinsic resistance of P. aeruginosa to linezolid. Interestingly, since the MIC of linezolid for wild-type P. aeruginosa was reduced in the presence of PMBN (Table 2), we also tested susceptibility of the efflux-deficient P. aeruginosa strain in the presence of PMBN. The MIC was reduced to 2 mg/L, indicating that, in contrast to E. coli, both efflux and the OM play a role in the intrinsic resistance of P. aeruginosa to linezolid. Following confirmation that these organisms are susceptible to rifampicin and linezolid if their intrinsic resistance elements are compromised, we determined the synergy of Antibiotic 301A₁ in these Gram-negative bacteria. No synergy was observed at 0.25 MIC for rifampicin and linezolid in P. aeruginosa, efflux-deficient P. aeruginosa or K. pneumoniae. However, synergy was observed in A. baumannii at 0.25 MIC for both drugs. Notably, the MICs of linezolid and rifampicin for A. baumannii were comparable to the susceptibility levels observed in E. coli (Tables 1 and 2).

Table 2.

Susceptibility of Gram-negative pathogenic bacteria to rifampicin and linezolid

Strain	Rifampicin MIC (mg/L)	Rifampicin MIC (mg/L) + 30 mg/L PMBN	Linezolid MIC (mg/L)	Linezolid MIC (mg/L) + 30 mg/L PMBN
P. aeruginosa PAO1	32	<0.5	>512	32
P. aeruginosa K2732	32	<0.5	32	2
A. baumannii BM4587	8	<0.5	256	256
K. pneumoniae H0142423	>512	64	>512	512

Antibiotic 301A₁ structure–activity relationship

To establish whether the sugars that decorate the aglycone of the anthracyclinone family are critical for rifampicin adjuvant activity, decilorubicin was degraded by the known method of strong acid hydrolysis,³⁹ producing decilorene (Figure 2). Using LC/MS to analyse the degradation product, a 492.72 m/z ion was detected in the positive ion mode, consistent with the anticipated mass of decilorene (491.15 g/mol).³⁹ Susceptibility testing using the chequerboard method revealed that decilorene (32 mg/L) did not reduce the growth of E. coli in combination with rifampicin (data not shown). This finding indicates that the adjuvant activity of Antibiotic 301A₁ is compromised by removal of the sugars decorating the aglycone. We hypothesized that the ability of the molecule to cross the OM of E. coli may have been altered by removal of the sugars. To establish whether this factor was responsible for the reduced activity of decilorene, we investigated the adjuvant activity of decilorene in the hyper-permeable E. coli strain (ΔbamB).³¹ However, since permeabilizing E. coli increased the susceptibility of this organism to rifampicin (Table 1), we assessed adjuvant activity in combination with linezolid, as Antibiotic 301A₁ also synergized with this Gram-positive antibiotic (Figure 4). In contrast to rifampicin, the intrinsic resistance of E. coli to linezolid appears to be solely due to active efflux of the drug from the cell^13,51 and permeabilization of the OM does not affect the linezolid MIC.¹³ However, even in the hyper-permeable strain of E. coli, decilorene had no effect on the MIC of linezolid, indicating that reduced permeability is not the underlying factor resulting in reduced adjuvant activity.

Figure 4.

Fluorescence-based adjuvant mechanism-of-action studies. (a) Antibiotic 301A₁ does not affect membrane potential in E. coli. Following self-quenching and stabilizing of DiSC₃ dye, compounds were injected at 200 s. With the addition of the permeabilizing antibiotic polymyxin B (20 mg/L; red), an increase in fluorescence is observed. Rifampicin (2 mg/L; green), linezolid (128 mg/L; purple) and Antibiotic 301A₁ (32 mg/L; blue) did not result in an increase in fluorescence. (b) Efflux assay. Nile Red efflux following injection of 50 mM glucose at 100 s in wild-type E. coli BW25113 (blue) and efflux-deficient (ΔtolC) E. coli BW25113 (red). (c) Concentration-dependent (100 μM, red; 50 μM, purple; 25 μM, green; and 0 μM, blue) inhibition of efflux by Antibiotic 301A₁.

We sought to investigate the structure–activity relationship of Antibiotic 301A₁ further by ascertaining whether rifampicin synergy is specific to Antibiotic 301A₁ and members of this anthracyclinone family, or whether synergy is applicable to molecules with similar chemical structures. Daunorubicin and doxorubicin are natural-product anthracycline antibiotics (Figure 2) structurally related to Antibiotics 301A₁ and 301A₂. These molecules all share a common planar aromatic chromophore; however, Antibiotic 301A₁ also possesses a bicyclo amino sugar at one end of the aglycone and lacks a methyl ester at the other (Figure 2). Furthermore, the amino sugar attached to daunorubicin and doxorubicin (daunosamine), is a dimethylamine sugar in Antibiotic 301A₁. Susceptibility testing established that, as in the case of Antibiotic 301A₁, E. coli is intrinsically resistant to daunorubicin and doxorubicin primarily due to efflux (Table 1). Susceptibility testing of E. coli using the chequerboard method revealed that even at concentrations up to 512 mg/L of the two anthracyclines, the MICs of rifampicin and linezolid remained unchanged.

Comparison of cytotoxicity with the anticancer drug doxorubicin

Given that the structurally similar molecule doxorubicin is currently used clinically in cancer chemotherapy, we sought to investigate the cytotoxicity of our compound Antibiotic 301A₁. Live counts of HEK293 cells incubated in the presence of serial dilutions of doxorubicin, Antibiotic 301A₁ and avidinorubicin revealed that doxorubicin was 5-fold (MIC = 1.2 mg/L) more toxic than both Antibiotic 301A₁ and avidinorubicin (50 mg/L).

Investigating the molecular basis of Antibiotic 301A₁ adjuvant activity

Since the OM reduces uptake of rifampicin in E. coli, we hypothesized that the adjuvant activity of Antibiotic 301A₁ may simply be the effect of increased permeability, and thus enhanced uptake of rifampicin into the cell. To assess the influence of Antibiotic 301A₁ on the E. coli OM, we investigated membrane depolarization by Antibiotic 301A₁ using the lipophilic potentiometric dye DiSC₃. In order for bacteria to successfully generate energy via the electron transport chain, it is crucial that the transmembrane proton motive force (PMF) is established; the main component of the PMF is an electrical potential gradient (ΔΨ). DiSC₃ is sensitive to changes in the ΔΨ of the cytoplasmic membrane;²⁹ it is taken up by the cells according to the magnitude of the ΔΨ and is concentrated in the cytoplasm, where it self-quenches its own fluorescence. If a compound disrupts the cytoplasmic membrane and thus depolarizes the ΔΨ, the dye is released, resulting in an increase in fluorescence. The permeabilizing antibiotic PMB⁴⁹ was used as a control; following addition of this drug there was a marked increase in fluorescence, as anticipated (Figure 4a). In contrast, the presence of Antibiotic 301A₁ had no effect on the fluorescence of DiSC₃.

An additional approach taken to investigate the impact of Antibiotic 301A₁ on membrane permeability involved testing of the susceptibility of E. coli to vancomycin in the absence and presence of Antibiotic 301A₁. The Gram-positive antibiotic vancomycin is normally too large to cross the OM of Gram-negative bacteria; however, in the presence of molecules that change or disrupt the OM, the MIC is reduced.^52,53 In the presence of vancomycin alone, E. coli BW25113 had an MIC of 256 mg/L; this MIC remained unchanged in the presence of 32 and 64 mg/L of Antibiotic 301A₁. Overall, our findings suggest that Antibiotic 301A₁ does not affect the integrity or the ΔΨ of the E. coli outer and cytoplasmic membranes.

Since Antibiotic 301A₁ does not seem to elicit its adjuvant activity via non-specific membrane disruption, we sought to confirm that the phenotype associated with combination of the two molecules is directly connected to rifampicin successfully inhibiting its target, RNA polymerase (RNAP).^54,55 In E. coli, the mutation frequency of rifampicin resistance is high and mutants (rifampicin-resistant mutants) are readily selected; mutations are located within the rpoB gene, affecting the β-subunit of RNAP.⁵⁶ We generated rifampicin-resistant mutants by growing E. coli BW25113 in the presence of high concentrations of rifampicin (125 and 250 mg/L). Sequencing of rifampicin-resistant mutants established that the mutations have been previously characterized^57,58 and are located within the rpoB gene; we selected three different mutants (I₅₇₂F, D₁₅₆N and H₅₂₆Y) to further investigate the adjuvant activity of Antibiotic 301A₁. In rifampicin-resistant mutants, rifampicin was no longer capable of inhibiting RNAP. If the growth of E. coli is still inhibited in the presence of subinhibitory concentrations of rifampicin and Antibiotic 301A₁, then the phenotype must be associated with a target other than RNAP, in the case of rifampicin potentiation. Growth of the rifampicin-resistant mutants was not affected by the presence of rifampicin (concentrations ranging from 2 to 250 mg/L) and Antibiotic 301A₁, indicating that the phenotype is directly associated with binding of rifampicin to RNAP.

In addition to rifampicin potentiation, we observed that Antibiotic 301A₁ also enhanced the activity of linezolid in E. coli (Figure 3). Since active efflux significantly decreases the activity of linezolid in E. coli,⁵¹ we sought to investigate possible inhibition of efflux by Antibiotic 301A₁ as the underlying mechanism of linezolid potentiation. To assess efflux inhibition, we used E. coli cells preloaded with Nile Red.³⁰ Use of this dye is considered to give a more true representation of the rate of efflux, since dyes such as ethidium and doxorubicin become concentrated in the cytosol due to binding of proteins and nucleic acid, resulting in notably slower efflux.³⁰ Thus, the Nile Red efflux assay is regarded as a real-time measure for studies of efflux competition in E. coli. Energy-depleted E. coli cells were preloaded with Nile Red in combination with low concentrations of the proton conductor CCCP. Following removal of CCCP, efflux was initiated by the addition of glucose, resulting in extrusion of the dye and a decrease in fluorescence (Figure 4b). The efflux-deficient ΔtolC E. coli strain was used as a control; following energization of the preloaded cells with glucose, no decrease in fluorescence was observed (Figure 4b). After incubation of the preloaded cells with Antibiotic 301A₁, we observed concentration-dependent competition of dye efflux (Figure 4c). This finding suggests that Antibiotic 301A₁ may synergize with linezolid in E. coli through competition with the drug for efflux, thus decreasing the rate of linezolid extrusion.

To further investigate this observation, we utilized strains of A. baumannii over-expressing efflux pumps. In A. baumannii, increased expression of resistance–nodulation–division (RND) systems is a major cause of MDR.⁵⁹ We tested a series of A. baumannii strains over-expressing different RND systems [BM4688 (AdeABC),⁶⁰ BM4666 (AdeIJK)^60,61 and BM4690 (AdeFGH)].⁶² Since we ascertained that Antibiotic 301A₁ synergizes with linezolid in this organism, we hypothesized that the phenotype would be reduced or lost if active efflux was increased due to over-expression of RND systems. Susceptibility testing of the various A. baumannii strains to linezolid resulted in a 2-fold increase (256–512 mg/L) in the MIC of linezolid for strains over-expressing RND efflux systems. Antibiotic 301A₁ in combination with 0.25 of the linezolid MIC resulted in a reduction in growth in strains over-expressing AdeABC and AdeFGH, suggesting that these RND systems are not involved in the adjuvant mechanism of action. However, synergy was lost in the strain over-expressing the RND system AdeIJK, indicating that this RND system is connected to the phenotype. These results confirm that efflux is connected to the linezolid adjuvant mechanism of action of Antibiotic 301A₁.

Discussion

In order to control and diminish the serious health threat associated with antibiotic resistance,¹ it is critical that new and innovative antibacterial therapies are developed. An orthogonal approach involves the discovery of small molecules that enhance or potentiate the activity of existing antibacterial agents; we have proven success in this line of attack and have identified a number of successful combinations.^22,23,63 The emergence of antibiotic resistance is an ever-present threat due to the ability of bacteria to acquire resistance to an antibiotic they were previously susceptible to; this occurrence dramatically reduces treatment options in the clinic. However, an additional contributing factor is the innate or intrinsic ability of bacteria to resist the inhibitory effect of certain antibacterial agents. Indeed, this phenomenon renders the majority of Gram-negative bacteria impervious to a plethora of antibiotics that are used clinically to treat infections caused by Gram-positive pathogenic bacteria. Since soil-dwelling Actinomyces have provided us with an arsenal of antibiotics and antifungals with which to fight infectious disease, we sought to investigate whether we could identify secondary metabolites that potentiate the activity of existing antibiotics, which are otherwise ineffective in the model Gram-negative organism E. coli.

Screening of fermentation extracts derived from 560 strains of Actinomyces resulted in the identification of a family of unusual anthracyclines that enhance the activity of rifampicin in E. coli. Interestingly, these molecules have documented activity in Gram-positive organisms⁴⁵ but are largely ineffective against E. coli due to active efflux (Table 1). The higher-MW molecules (avidinorubicin and Antibiotic 301B) had marginally higher activity against E. coli alone and our mechanistic studies involved the lower-MW molecules. Previous studies involving these compounds have focused entirely on the higher-MW molecules and therapeutic applications involving their anti-tumour⁴⁵ and platelet aggregation inhibitory activity.⁴¹ To our knowledge, no studies have involved the lower-MW molecules (Antibiotics 301A₁ and 301A₂), and in the last 20 years no studies have been documented on any members of this family of molecules. However, a number of studies have investigated the structurally similar antibiotic anti-tumour drug nogalamycin (Figure 2), which inhibits DNA-directed RNA synthesis in vitro.⁶⁴ Indeed, the shared positively charged bicyclo amino sugar of both Antibiotic 301A₁ and nogalamycin is thought to underlie the unusual nucleic acid intercalating properties of the drug.⁶⁵ Structure–activity relationship studies of Antibiotic 301A₁ revealed that other anthracyclines, such as doxorubicin and daunorubicin, do not synergize with rifampicin or linezolid in E. coli. Furthermore, removal of the sugar moiety attached to one end of the aglycone abolishes the adjuvant activity of Antibiotic 301A₁, and this loss of activity does not appear to be associated with reduced permeability to the drug. Cytotoxicity studies revealed that Antibiotic 301A₁ and members of this family are significantly less toxic than the anticancer drug doxorubicin. These findings indicate that the adjuvant activity of Antibiotic 301A₁ is not a generic phenomenon applicable to all anthracyclines. Furthermore, FICIs of Antibiotic 301A₁ with other Gram-positive antibiotics revealed that the drug also synergizes with the synthetic antibiotic linezolid and that these synergistic combinations also reduce the growth of the important human pathogen A. baumannii.

We sought to investigate the mechanism underlying rifampicin and linezolid potentiation by Antibiotic 301A₁. Since the activity of rifampicin is reduced in E. coli due to the relative impermeability of the OM, we considered membrane disruption a likely candidate. However, our findings suggest that this molecule does not greatly compromise the membrane. Furthermore, linezolid is known to be actively effluxed in E. coli and the OM does not appear to impact susceptibility levels in E. coli. Efflux assays revealed that Antibiotic 301A₁ competitively inhibits extrusion of Nile Red, suggesting that the molecule also competes with linezolid, reducing the rate of efflux of the drug and increasing intracellular levels of the antibiotic. To further investigate this phenomenon we utilized a series of A. baumannii strains over-expressing RND efflux pumps (AdeABC,⁶⁰ AdeIJK^60,61 and AdeFGH).⁶² Interestingly, we only observed loss of the linezolid–Antibiotic 301A₁ synergistic phenotype in the strain over-expressing AdeIJK. This finding reinforces the observation that Antibiotic 301A₁ competes with linezolid for efflux by AcrAB–TolC in E. coli. The AdeIJK RND efflux system is found in all strains of A. baumannii⁵⁹ and is responsible for the intrinsic resistance, but not acquired resistance, of this organism to numerous antibiotics.⁶⁶ The inner membrane transporter AdeJ belongs to the AcrB protein family and shows 57% identity with AcrB in E. coli,⁶⁶ an observation that may explain why we see the linezolid–Antibiotic 301A₁ synergy in both E. coli and A. baumannii.

Since rifampicin is a poor substrate for the AcrAB–TolC efflux system, we acknowledge that competition for efflux by Antibiotic 301A₁ may not be the only underlying mechanism resulting in rifampicin potentiation. However, the structure of AcrB bound to rifampicin has been reported, revealing that the protein has two binding pockets (proximal and distal).⁶⁷ Molecules with a higher MW are found within the proximal binding pocket and are thought to travel from there to the distal pocket; in contrast, smaller-MW molecules are thought to pass through the proximal binding pocket without taking part in specific interactions and then bind to the distal pocket.⁶⁷ It is possible that rifampicin binds to the proximal pocket of AcrB and occludes access of the transporter to other lower-MW molecules. If this were the case, binding of rifampicin would inhibit efflux of lower-MW molecules such as linezolid and Antibiotic 301A₁, resulting in synergy. This phenomenon may explain the synergistic relationship between rifampicin and Antibiotic 301A₁. To investigate this hypothesis, we performed chequerboard analysis with rifampicin and linezolid. The FICI indicated that the combination is not synergistic and that linezolid efflux is not inhibited. However, this does not rule out the possibility that rifampicin occludes the efflux of Antibiotic 301A₁. We have also ascertained that the synergistic phenotype is connected with rifampicin inhibiting its target, RNAP. Since members of the anthracyclinone family are known to intercalate nucleic acid and thus inhibit RNA synthesis, it is possible that synergy is also the result of two distinct drugs inhibiting RNA synthesis.

This study has identified a natural-product small molecule capable of sensitizing E. coli to antibiotics to which it is ordinarily impervious. Characterization of the mechanisms underlying potentiation revealed apparently distinct factors contributing to the antibiotic adjuvant spectrum. Antibiotic 301A₁ therefore potentiates antibiotics in different bacterial genera by a complex mechanism. We note that, like other well-studied antibiotic adjuvants such as phenylalanine arginyl β-naphthylamide (PAβN),^52,68 Antibiotic 301A₁ is a diamine with exocyclic amines predicted to be protonated at physiological pH. As is the case for PAβN, these charges may contribute to the complexity of the mode of action. Only a fraction (20%) of our NPL was screened in this study, highlighting the potential of future studies to identify additional antibiotic adjuvants within this collection.

Funding

This study was supported by grants from the Canadian Institutes of Health Research (MT-13536), the Natural Sciences and Engineering Research Council Grant (237480), a Canada Research Chair in Antibiotic Biochemistry (to G. D. W.) and the BSAC (GA2013_027R).

Transparency declarations

None to declare.

Supplementary data

Tables S1 and S2 are available as Supplementary data at JAC Online (http://jac.oxfordjournals.org/).