Pyrrolomycins Are Potent Natural Protonophores

The escalating burden of antibiotic drug resistance necessitates research into novel classes of antibiotics and their mechanism of action. Pyrrolomycins are a family of potent natural product antibiotics with nanomolar activity against Gram-positive bacteria, yet with an elusive mechanism of action.

ABSTRACT

The escalating burden of antibiotic drug resistance necessitates research into novel classes of antibiotics and their mechanism of action. Pyrrolomycins are a family of potent natural product antibiotics with nanomolar activity against Gram-positive bacteria, yet with an elusive mechanism of action. In this work, we dissect the apparent Gram-positive specific activity of pyrrolomycins and show that Gram-negative bacteria are equally sensitive to pyrrolomycins when drug efflux transporters are removed and that albumin in medium plays a large role in pyrrolomycin activity. The selection of resistant mutants allowed for the characterization and validation of a number of mechanisms of resistance to pyrrolomycins in both Staphylococcus aureus and an Escherichia coli ΔtolC mutant, all of which appear to affect compound penetration rather than being target associated. Imaging of the impact of pyrrolomycin on the E. coli ΔtolC mutant using scanning electron microscopy showed blebbing of the bacterial cell wall often at the site of bacterial division. Using potentiometric probes and an electrophysiological technique with an artificial bilayer lipid membrane, it was demonstrated that pyrrolomycins C and D are very potent membrane-depolarizing agents, an order of magnitude more active than conventional carbonyl cyanide m-chlorophenylhydrazone (CCCP), specifically disturbing the proton gradient and uncoupling oxidative phosphorylation via protonophoric action. This work clearly unveils the until-now-elusive mechanism of action of pyrrolomycins and explains their antibiotic activity as well as mechanisms of innate and acquired drug resistance in bacteria.

INTRODUCTION

Recent reports have brought to light the increasingly devastating emergence of multidrug-resistant pathogenic bacteria (1). To combat this global health care challenge, it is essential that innovative approaches are used to identify novel classes of antibiotics that act through novel vulnerable targets and that are effective against drug-resistant isolates. To date, natural products and derivatives still account for 62% of all approved drugs and for about 80% of approved antibacterial drugs in the period between 1981 and 2014 (2). Nonetheless, due to various factors, including toxicity, poor pharmacokinetics, and lack of in vivo efficacy, a multitude of undeveloped natural products with potent in vitro antibiotic activity have not been further investigated. Understanding the mechanism of action of such compounds could help evaluate their drug-like potential and uncover novel bacterial targets that are chemically vulnerable. To this extent, in this study, we sought to discover the mechanism of antibiotic activity of pyrrolomycins.

Pyrrolomycins are a group of heavily halogenated small natural antibiotics produced by Streptomyces vitaminophilus (3), Streptomyces sp. strain UC 11065 (4), and Streptomyces fumanus (5). Pyrrolomycins carrying a carbonyl linker between the benzyl and pyrrole moiety are reported as potent anti-Gram-positive specific inhibitors (3, 6). Of the many natural pyrrolomycin analogues reported, the pentachlorinated pyrrolomycin D has been shown to be the most potent, with an antistaphylococcal MIC reported at around 1 ng/ml, and activity against biofilms (6), though the minimal bactericidal concentration (MBC) was reported in the low micrograms-per-milliliter range (7). Pyrrolomycin D cytotoxicity has been published to be moderate (6, 8), though its acute toxicity in mice is reported at 20 mg/kg of body weight (following intraperitoneal administration) (3).

The potent activity of pyrrolomycin D could help uncover an innate vulnerability of staphylococci and allow for further chemical optimization of this antibiotic scaffold. In this study, we present illuminating insight into the pyrrolomycin D spectrum of activity, describe diverse mechanisms of bacterial resistance to pyrrolomycins, and define the precise mechanism of action of these natural products.

RESULTS

Chemical synthesis of pyrrolomycins C, D, I, and J.

Pyrrolomycins C, D, I, and J (Fig. 1) were successfully produced by de novo chemical synthesis and their structures verified by nuclear magnetic resonance (NMR) (see supplemental Results) and high-resolution mass spectrometry. Each of the compounds was found to be more than 95% pure by high-performance liquid chromatography-UV (HPLC-UV) analysis.

FIG 1.

Chemical structures of pyrrolomycins C, D, I, and J.

In vitro profiling of pyrrolomycin activity.

The preferential antibacterial activities of pyrrolomycins C and D against Gram-positive bacteria versus Gram-negative bacteria were confirmed, as previously published (3) (Table 1). Antibiotic susceptibility testing in standard medium (cation-adjusted Mueller-Hinton broth [CAMHB]) confirmed that both Staphylococcus aureus and Streptococcus pneumoniae were very sensitive to pyrrolomycin D, slightly less sensitive to pyrrolomycin C, and more than 10-fold less susceptible to the methoxy analogues pyrrolomycins I and J (Table 1). The Gram-negative bacteria tested (all cultured in CAMHB) were all an order of magnitude less sensitive to pyrrolomycins than were the Gram-positive bacteria, though pyrrolomycins C and D remained more active than pyrrolomycins I and J (Table 1). Mycobacterium tuberculosis (grown in Middlebrook 7H9 supplemented with 10% oleic acid-albumin-dextrose-catalase [OADC], 0.2% glycerol, and 0.05% Tween 80) was weakly susceptible to pyrrolomycins, with similar high MICs for all analogues (Table 1).

TABLE 1.

MICs and cell line cytotoxicity of pyrrolomycins C, D, I, and J on a panel of bacteria in indicated growth medium^a

Strain	Culture mediumb	MIC95 (μg/ml) or IC95 (IC50) for:
		Pyr D	Pyr C	Pyr J	Pyr I	Rif	Cipr
S. aureus SH1000	CAMHB	0.025	0.1	1.5	3	0.05	1
S. pneumoniae	CAMHB	0.006	0.1	0.4	3	0.05	1
A. baumannii	CAMHB	3	6.25	>50	>50	>0.2	0.2
K. pneumoniae	CAMHB	5	27	>50	>50	>0.2	0.03
P. aeruginosa PAO1	CAMHB	12.5	33	>50	>50	>0.2	0.125
E. coli BW25113	CAMHB	3	6	>50	>50	>0.2	0.015
E. coli BW25113 ΔtolC	CAMHB	0.025	0.0125	1	1.5	>0.2	<0.015
E. coli BW25113 ΔacrA	CAMHB	3	/	/	/	/	<0.015
E. coli BW25113 ΔacrB	CAMHB	3	/	/	/	/	<0.015
M. tuberculosis H37Rv	7H9-GT-OADC	3	3	4	6	0.005	0.5
S. aureus SH1000	CAMHB + BSA	25	12.5	>50	>50	0.05	1
S. aureus SH1000	CAMHB + FCS	21	21	>50	>50	0.05	1
E. coli BW25113 ΔtolC	CAMHB + BSA	33	6	50	25	>0.2	0.015
E. coli BW25113 ΔtolC	CAMHB + FCS	12.5	1.5	25	17	>0.2	<0.015
M. tuberculosis H37Rv	7H9-GTy-ND	0.2	0.13	0.83	0.4	0.02	0.25
M. tuberculosis H37Rv	7H9-GTy-ND + BSA	3	1.5	25	25	0.1	0.5
HepG2 (IC95 [IC50])	DMEM + FCS	3 (1.1)	1.5 (0.62)	25 (8.3)	29 (8.6)
HEK-293	DMEM + FCS	3 (1.00)	3 (1.01)	25 (8.52)	25 (7.19)

^aData represent the average of at least 3 replicates. Rifampin (Rif) and ciprofloxacin (Cipr) were used as antimicrobial controls. Pyr, pyrrolomycin; IC₉₅, 95% inhibitory concentration; IC₅₀, 50% inhibitory concentration. Slash indicates that test was not performed.

^bCAMHB, cation-adjusted Mueller-Hinton broth; CAMHB + BSA, CAMHB supplemented with 0.5% (wt/vol) BSA fraction V; CAMHB + FCS, CAMHB supplemented with 10% (vol/vol) fetal calf serum; 7H9-GT-OADC, Middlebrook 7H9 medium supplemented with 0.2% glycerol, 0.05% Tween 80, and 10% OADC (final concentrations, 0.005% oleic acid, 0.5% albumin fraction V, 0.2% dextrose, 4 mg/liter catalase, and 0.95 g/liter NaCl); 7H9-GTy-ND, Middlebrook 7H9 medium supplemented with 0.2% glycerol, 0.025% tyloxapol, 0.2% dextrose, and 0.95 g/liter NaCl; 7H9-GTy-ND + BSA, Middlebrook 7H9 medium supplemented with 0.2% glycerol, 0.025% tyloxapol, 0.2% dextrose, 0.95 g/liter NaCl, and 0.5% albumin fraction V; DMEM + FCS, Dulbecco’s modified Eagle medium supplemented with 10% (vol/vol) fetal calf serum.

To assess whether Gram-negative bacteria were resistant to pyrrolomycins due to their efficient xenobiotic efflux systems, Escherichia coli strains lacking a component of the tripartite AcrAB-TolC efflux pump (E. coli ΔtolC, E. coli ΔacrA, and E. coli ΔacrB mutants) were tested for their susceptibility to pyrrolomycins. The data (Table 1) clearly showed that the E. coli ΔtolC mutant strain is highly sensitive to pyrrolomycins C and D, while surprisingly, the E. coli ΔacrA and E. coli ΔacrB mutant strains remained resistant. This suggests that wild-type E. coli resistance to pyrrolomycins is due to TolC-mediated efflux (AcrA/B independent) and not due to the absence of a pyrrolomycin intracellular target. The abundance of efficient xenobiotic efflux systems in the other Gram-negative bacteria may also account for their insensitivity to pyrrolomycins.

To investigate the observed insensitivity of M. tuberculosis to pyrrolomycins, experiments were designed to evaluate the role of growth medium components. Albumin (as bovine serum albumin [BSA]) is traditionally present in the culture medium of M. tuberculosis as part of the OADC or ADC supplements and functions to prevent toxicity associated with the Tween 80 detergent (9). However, albumin is known to bind drugs/compounds and alter their free fraction in medium, potentially affecting activity. Indeed, by using a medium supplemented with tyloxapol instead of Tween 80, it was clearly demonstrated that the removal of BSA rendered M. tuberculosis 10 times more susceptible to pyrrolomycins (Table 1). In a reverse experiment, the addition of BSA or fetal calf serum (FCS) to CAMHB reduced the activity of pyrrolomycin against S. aureus and the E. coli ΔtolC mutant by at least 2 orders of magnitude (Table 1). Overall, these data clearly suggest that pyrrolomycins are highly bound to albumin, and that differences in culture medium composition, particularly the presence of albumin, help explain the previously described low sensitivity of M. tuberculosis to pyrrolomycins.

The impact of medium supplementation with BSA or FCS (typically used in cell culture medium) also calls into question the validity of the previously reported high selectivity index (antibiotic activity versus cytotoxic activity) of pyrrolomycins (6, 8). Interestingly, the cytotoxicity of pyrrolomycins observed here against HepG2 and HEK293 human cells was in the same range as that observed for Gram-positive bacteria grown with 10% FCS supplementation (Table 1). These data illustrate that when correcting for differences in medium composition, there is no selectivity of pyrrolomycins C, D, I, or J for bacterial versus eukaryotic cells. This finding may explain the reported acute toxicity of pyrrolomycins in mice of 50 and 20 mg/kg for pyrrolomycins C and D, respectively (3).

Time-dependent killing and MBC.

Previous published data showed a large difference between the MIC and the minimal bactericidal concentration (MBC) for pyrrolomycins (7). This result was confirmed here, and the data show that the bactericidal activity is both concentration and time dependent (Fig. S1). Following a 24-h pyrrolomycin D exposure, a 2-log drop in CFU was achieved against S. aureus SH1000 and the E. coli ΔtolC mutant at 3 μg/ml and 0.75 μg/ml, respectively, compared to an MIC of 25 ng/ml for both strains. Analysis of the time-dependent bactericidal activity at 200 ng/ml (8 times MIC) showed a slow but gradual decline in CFU for both S. aureus SH1000 and the E. coli ΔtolC mutant (Fig. S1).

Deciphering genetic resistance to pyrrolomycin D.

(i) Pyrrolomycin D resistance in the E. coli ΔtolC mutant. To gain an insight into the mechanism of action of pyrrolomycin D, efforts were made to select resistant bacteria from the E. coli ΔtolC mutant and S. aureus streaked grown on solid medium containing pyrrolomycin D at concentrations above its MBC (1 μg/ml). Despite numerous attempts, and the addition of up to 6 × 10¹⁰ bacteria per plate (100 μl of optical density at 600 nm [OD₆₀₀] of 400), no resistant S. aureus clone could be isolated. For the E. coli ΔtolC mutant, when 2 × 10¹⁰ CFU were plated, two resistant colonies were isolated. Following culture of the resistant clones in compound-free medium, the resistance to pyrrolomycin D was confirmed (MIC, 2 μg/ml). Whole-genome sequencing and variant analysis of one of the resistant mutants (E. coli ΔtolC pyrr RC1) revealed no single nucleotide polymorphism, nor short insertion or deletion, but instead a large deletion between positions 4293549 and 4302404 on the genome reference (accession no. CP009273.1). This 8,855-bp deletion, confirmed by Sanger sequencing, encompasses nine genes transcribed in the order alsR, alsB, alsA, alsC, alsE, alsK, yjcS, ytcA, and mdtN (Fig. 2). The second pyrrolomycin-resistant clone that was isolated contained the exact same deletion. The role of this large deletion in pyrrolomycin resistance was confirmed using a genetic engineering approach on E. coli strain EP664 (BW25113 ΔtolC, where the kanamycin cassette was removed by FLP recombination target [FRT] excision). Replacement of the 8,855-bp region in EP664 with a kanamycin resistance cassette (resulting in E. coli strain EP673) was found not to alter its pyrrolomycin D sensitivity (Table 2). However, subsequent erasing of the kanamycin cassette by allelic exchange to generate the unmarked pyrr RC1 deletion in strain EP676 conferred resistance to pyrrolomycin D (Table 2). This work confirmed the causality of the 8,855-bp deletion in pyrrolomycin resistance and showed that resistance was likely due to the new gene arrangement rather than to the gene loss.

FIG 2.

Illustration of the genetic arrangement found in pyrrolomycin-resistant E. coli isolates, and engineered validation strains. From the parental E. coli ΔtolC::Km strain, a pyrrolomycin D-resistant isolate, E. coli ΔtolC pyrr RC1, was sequenced and shown to bear a 8,855-bp deletion between bases 4293549 and 4302404, deleting 7 whole genes and 2 partial genes, and forming a hybrid in-frame gene between the truncated alsR′ and ′mdtN genes. The disruption of the alsR gene (encoding a transcriptional repressor) likely leads to derepression of the als promoter P_als. Deletion mutants were generated via Red recombineering to validate that resistance was due to mdtOP expression via P_als derepression. To achieve this, the original E. coli ΔtolC::Km mutant strain was deleted of its Km cassette to obtain EP664 ΔtolC. The 8,855-bp als locus was then deleted and replaced with a Km cassette transcribed in the orientation opposite the als and mdt genes to yield EP673. EP673 was then cured of the Km cassette to generate EP676. Finally, to determine the involvement of mdtO and mdtP, these genes were deleted and replaced by a Km cassette to obtain EP680.

TABLE 2.

Resistance profiles of E. coli BW25113 ΔtolC strains to pyrrolomycins D and C

Strain	Characteristics	MIC95 (μg/ml) for:
		Pyr D	Pyr C	Km
ΔtolC mutant	ΔtolC::Km	0.025	0.125	>50
ΔtolC pyrr RC1 mutant	ΔtolC::Km, with selected pyrrolomycin resistance	3	6	>50
EP664	ΔtolC	0.025	0.125	4
EP673	ΔtolC ΔalsRBACEK-yjcS-ytcA-mdtN::Km	0.025	0.125	>50
EP676	ΔtolC ΔalsRBACEK-yjcS-ytcA-mdtN	3	6	4
EP680	ΔtolC ΔalsRBACEK-yjcS-ytcA-mdtNOP::Km	0.025	0.125	>50
EP676(pEP713)	Empty plasmid bearing PgapA (control)	0.5
EP676(pEP712)	PgapA-alsR plasmid (alsR overexpression)	0.062
EP664(pBAD30)	Empty plasmid without arabinose	0.062
EP664(pBAD30)	Empty plasmid with 10 mM arabinose	0.062
EP664(pEP714)	pEP714 is pBAD30::mdtP, without arabinose	0.062
EP664(pEP714)	pEP714 is pBAD30::mdtP, with 10 mM arabinose	1

The deletion in E. coli ΔtolC pyrr RC1 and EP676 places the mdtO and mdtP genes in close proximity to the als operon promoter (P_als). mdtO and mdtP encode resistance-nodulation-division (RND)-type multidrug efflux pump components homologous to AcrB (inner membrane transporter) and TolC (outer membrane channel), respectively. Thus, we reasoned that the P_als promoter activity may be high in EP676, as the local repressor gene alsR is truncated due to the deletion, and that mdtP overexpression could complement the loss of tolC. To verify these hypotheses, several genetic constructs were generated and evaluated. First, when the mdtOP genes were deleted in EP676, the resulting strain, EP680, turned pyrrolomycin sensitive, confirming the role of mdtOP in pyrrolomycin resistance (Table 2). Second, conditional expression of mdtP alone in EP664, a ΔtolC mutant strain bearing an intact als-mdt locus, led to a pyrrolomycin resistance phenotype, showing that the overproduction of the MdtP outer membrane channel was sufficient for pyrrolomycin resistance (Table 2). Finally, the overexpression of alsR from a plasmid rendered EP676 sensitive to pyrrolomycin D (Table 2), confirming the role of the als promoter and that of the AlsR repressor. Overall, these results demonstrate that the pyrrolomycin resistance phenotype selected is due to a genetic rearrangement that results in the als promoter-mediated upregulation of mdtP.

(ii) Selection of pyrrolomycin D-resistant clones from E. coli EP680. To isolate alternative pyrrolomycin-resistant E. coli mutants devoid of mutations associated with the MdtP channel, 1 × 10⁹ EP680 (ΔtolC Δals-mdtNOP) cells were deposited on plates containing low concentrations of pyrrolomycin D (50 ng/ml, twice the solid MIC). Seven colonies appeared after 48 h. All colonies were confirmed to present low-level resistance to pyrrolomycin D (Table 3). Whole-genome sequencing and variant analysis on two selected resistant clones (EP680-RC50.6 and EP680-RC50.11) revealed that both isolates carried a point mutations in slyA (encoding a nonessential transcriptional regulator of the MarR family), while the more resistant clone, EP680-RC50.6, carried an additional point mutation in ompA (outer membrane protein A) (Table 3). These mutations were confirmed by Sanger sequencing. The mutations in slyA were a premature stop codon at position 16 in one clone and a point mutation at the beginning of the helix-turn-helix (HTH) domain (Q49R) in the other clone, both likely generating a dysfunctional proteins. Pyrrolomycin susceptibility studies of the E. coli ΔtolC ΔslyA mutant (EP706) confirmed that slyA inactivation confers low-level pyrrolomycin resistance (Table 3). The role of SlyA in E. coli is still poorly known, but it is a MarR-type transcriptional regulator that targets numerous bacterial promoters (10). Though it is not clear how the inactivation of slyA causes low-level pyrrolomycin D resistance in E. coli, in Salmonella enterica, slyA (91% identical) has been shown to be involved in resistance to antimicrobial peptides (11). The mutation found in ompA (G63R) is located in the second transmembrane β-barrel of the protein. In contrast to this mutation, the deletion of ompA in the ΔtolC background rendered E. coli similarly sensitive (perhaps slightly more sensitive) to pyrrolomycin D, comparable to the susceptibility data reported for other antibiotics (12). While the role of OmpA in antibiotic susceptibility is still debated, it is believed that its role in susceptibility to non-beta-lactam antibiotics is by altering membrane integrity and permeability rather than antibiotic transport (12), with the identified mutations presumably resulting in a gain-of-function phenotype.

TABLE 3.

Pyrrolomycin D sensitivity of E. coli ΔtolC mutant strains on solid medium

E. coli strain	Characteristics	Solid pyrrolomycin D MIC (ng/ml)
EP680	ΔtolC ΔalsRBACEK-yjcS-ytcA-mdtNOP::Km	40
EP680-RC:50.6	slyA(Q49H) ompA(G63R)	80
EP680-RC:50.11	slyA(W16stop)	60
EP664	ΔtolC	40
EP705	EP664 ΔompA	30
EP706	EP664 ΔslyA	50

(iii) Selection of pyrrolomycin D-resistant clones from S. aureus SH1000. As no S. aureus resistant mutant could be isolated on high concentrations of pyrrolomycin D, selection was repeated at near MICs. Three resistant colonies could be isolated when 1.5 × 10¹⁰ CFU were plated at a concentration of 4 times the MIC (100 ng/ml). As with E. coli EP680 resistant mutants, these isolates appeared only after 48 h and were confirmed to have low-level resistance to pyrrolomycins D and C (Table 4). One of these resistant isolates (100-1) and the parental SH1000 S. aureus strain were subjected to whole-genome sequencing and variant analysis. The data showed a point mutation in the nonessential mprF gene (L78Stop) and a second one in the region upstream of the SAOHSC_01883 gene (conserved hypothetical), both confirmed by Sanger sequencing. Targeted sequencing of these regions in the two other resistant isolates showed that clone 100-2 carried the same mutations as clone 100-1, and that clone 100-3 carried a single-base insertion in mprF resulting in a premature stop codon. As different mutations in mprF were found in two resistant isolates, mutations in this gene are likely associated with pyrrolomycin resistance.

TABLE 4.

Sensitivity of S. aureus SH1000 and isolates resistant to pyrrolomycin D and daptomycin as measured on solid medium

S. aureus strain	SNP identified relative to SH1000a	Solid pyrrolomycin D MIC (ng/ml)	Daptomycin MIC95 (μg/ml)
SH1000	WT	20	1.25
100-1	mprF(L78Stop) (WGS) (S), 1797388(t→g) (WGS) (S)	35	0.078
100-2	mprF(L78Stop) (S), 1797388(t→g) (S)	30	0.078
100-3	mprF(Y419Stop) (S)	35	0.078

^aThe genetic variations identified by whole-genome sequencing (WGS) or targeted Sanger sequencing (S) are shown. WT, wild type.

mprF codes for a multiple peptide resistance factor (MprF) with phosphatidylglycerol lysyltransferase activity that modifies membrane phosphatidylglycerol (PG) to lysylphosphatidylglycerol (LPG). Increased mprF expression has been demonstrated to increase the cationic charge of the bacterial membrane and repel cationic molecules, such as daptomycin (13, 14), methicillin (15) (where mprF is named fmtC), and several other cationic antibacterial peptides (16). Bearing this in mind, the daptomycin susceptibility of the pyrrolomycin D-resistant isolates was determined, and all three were found to be 16-fold more sensitive than was the parental S. aureus strain (Table 4). This suggests that the absence of MrpF decreases the cationic charge of the membrane and favors daptomycin entry. As pyrrolomycin D is predicted to be anionic at physiological pH, decreasing the cationic charge of the membrane will likely disfavor its penetration and hence lead to low-level resistance.

Pyrrolomycin D is not an electrophile, and its activity is not impacted by antioxidants.

Pyrrolomycin D did not react with glutathione and hence is unlikely an electrophile that could target proteins nonspecifically (data not shown). Similarly, pyrrolomycin activity against S. aureus was not impacted by antioxidants, such as α-tocopherol or ascorbic acid (data not shown), suggesting that activity was not mediated through reactive radical species. To gain a first insight into whether pyrrolomycins can affect membrane potential, the impact of pH on pyrrolomycin activity was determined. The data clearly show that decreasing the medium pH improved the activity of pyrrolomycin D against both S. aureus and the E. coli ΔtolC mutant, similarly to the effect observed with the proton uncoupler carbonyl cyanide m-chlorophenylhydrazone (CCCP) (see Fig. S2 in the supplemental material).

Visualization of pyrrolomycin D activity by SEM.

Scanning electron microscopy (SEM) was used to determine potential changes in E. coli and S. aureus morphology following pyrrolomycin D exposure. Initial observations performed after 5 h exposure of 1 μg/ml pyrrolomycin D showed that most bacteria were no longer intact. To obtain images of cellular deformation prior to bacterial cell wall collapse, lower concentrations of pyrrolomycin D were used (100 ng/ml). With respect to S. aureus, no obvious consistent bacterial deformation could be observed following bacterial exposure. In contrast, blebbing or bulging of the E. coli ΔtolC mutant cell wall was clearly observed in the presence of pyrrolomycin D (Fig. 3). This deformation was largely localized at the division septum, sometimes seen at the old pole, and was reminiscent of images of E. coli following exposure to the antimicrobial peptide human defensin 5 (HD5) (17). Altogether, these data suggest that pyrrolomycins act by increasing the bacterial osmotic pressure that leads to local bacterial cell wall rupture and bleb generation, or by weakening the cell wall so that it is no longer able to contain the normal osmotic pressure.

FIG 3.

(A to D) SEM images of the E. coli ΔtolC mutant following a 5-h exposure to DMSO (A) or 100 ng/ml pyrrolomycin D (B to D, three representative SEM images). Cell wall blebs can be clearly observed on pyrrolomycin-exposed bacteria (indicated by white triangles).

Impact of pyrrolomycin D on S. aureus membrane potential.

To determine whether pyrrolomycins act as proton gradient-depolarizing agents, the potentiometric probe 3,3'-dipropylthiadicarbocyanine iodide [DiSC3(5)] was initially used. DiSC3(5) accumulates readily in bacteria with polarized membrane, and this strong accumulation causes self-quenching of its fluorescence signal. Upon depolarization of the membrane, DiSC3(5) is released, quenching is decreased, and fluorescence is increased. As previously reported (18), we found that in the absence of cells, the proton uncoupler CCCP, but also pyrrolomycins C and D, cause a concentration-dependent quenching of DiSC3(5) (quenching observed above 0.038 and 0.019 μg/ml, respectively). We confirmed that this quenching is not mediated by CCCP or pyrrolomycin absorbance of the DiSC3(5) excitation or emission wavelengths. To our knowledge, the mechanism of this interaction remains unknown. Due to the observed quenching, compound-dependent depolarization of bacterial membranes was evaluated at subquenching concentrations. The data showed that pyrrolomycin C was particularly potent at depolarizing the S. aureus membrane, with the methylated pyrrolomycin I being significantly less active. Pyrrolomycin D also caused depolarization at a very low concentration; however, the magnitude of the depolarization [DiSC₃(5) signal change] was inferior to that of pyrrolomycin C but comparable to that of CCCP (Fig. 4). Overall, these data suggest that pyrrolomycins C and D are more potent modulators of the membrane potential of S. aureus than CCCP.

FIG 4.

(A and B) S. aureus membrane potential measured using potentiometric dyes DiSC₃(5) (A) and DiOC₂(3) (B) in the presence of pyrrolomycin C (filled circle, solid line), D (filled square, solid line), I (filled inverted triangle, dotted line), J (filled triangle, dotted line), and CCCP (open circle, solid line). Asterisks indicate concentrations at which the molecules quench the fluorescence signal of DiSC₃(5).

To confirm that pyrrolomycins cause membrane depolarization in bacteria, an alternative potentiometric probe, DiOC₂(3), was used in combination with flow cytometry. This assay follows DiOC₂(3) changes in fluorescence from red (BL2) to green (BL1) following membrane depolarization (measured by the BL2/BL1 ratio; Fig. 4). In contrast to DiSC₃(5), we did not observe quenching of DiOC₂(3) upon the addition of pyrrolomycins or CCCP. In agreement with data obtained using DiSC₃(5), we found that pyrrolomycins C and D were nanomolar modulators of membrane potential, with pyrrolomycin C being 1 log more potent than pyrrolomycin D and 2 orders of magnitude more potent than CCCP. The methylated pyrrolomycins I and J were more than 10-fold less potent than their respective analogues pyrrolomycins C and D, but they were still able to depolarize the S. aureus membrane potential at higher concentrations (Fig. 4).

Impact on mitochondrial respiration.

Mitochondrial respiration (oxygen consumption) is a direct result of the electron transport chain maintaining a resting proton motive force (electrochemical gradient) that drives ATP synthesis through the ATP synthase. The addition of uncoupler depletes the proton motive force and the mitochondria (as well as bacteria) response by increasing proton efflux through an increase in the electron transport chain activity and, hence, respiration. In experiments measuring rat liver mitochondrial respiration, both pyrrolomycins C and D were potent stimulators of respiration, a feature commonly seen following the addition of protonophoric uncouplers (Fig. S3). This potent nanomolar activity on isolated mitochondria is likely associated with pyrrolomycin cytotoxicity to eukaryotic cells.

Protonophore activity of pyrrolomycins D and C on a planer lipid bilayer.

To examine the protonophoric activity of pyrrolomycins D and C in a pure lipid membrane system, we measured electric current across a planar lipid bilayer (BLM) under voltage-clamp conditions (Fig. 5A). As seen in Fig. 5B, the addition of nanomolar concentrations of pyrrolomycin D to bathing solutions at both sides of BLM induced a transmembrane electric current in a nanoampere range (voltage, 25 mV), and the value of the current increased with pyrrolomycin D concentration.

FIG 5.

(A). A schematic of the BLM (made from DPhPC) electrophysiological setup used to determine the protonophoric activity of pyrrolomycin. (B) Induced electrical current through planar BLM following the addition of indicated concentrations of pyrrolomycin D to both the cis and trans chambers. The solution was 50 mM Tris, 50 mM MES, and 10 mM KCl (pH 7.0). The BLM voltage was 25 mV. (C) Current-voltage (I-V) curves for 1 μM pyrrolomycin D under symmetrical (both cis and trans chambers at pH 7.0, black line, or both at pH 8.2, red line) and asymmetrical conditions (cis chamber at pH 7.0, trans chamber at pH 8.2, blue line). (D) Electrical current (in nanoamperes) induced through DPhPC membrane by 3 μM pyrrolomycin D, pyrrolomycin C, and CCCP (mean ± standard deviation [SD]; n = 4).

To determine the ion selectivity of pyrrolomycin D-mediated permeability, current-voltage (I-V) curves were measured under symmetrical (equal pH in both chambers) and asymmetrical (a different pH in both chambers; pH 1 = 7.0, pH 2 = 8.2) conditions (Fig. 5C). The value of zero-current potential (V_zero) under asymmetrical conditions was −57 mV, corresponding to the high proton selectivity of the conductance (V_theoretical = RT/F ΔpH = −71 mV at ΔpH = 1.2). As pyrrolomycin D, pyrrolomycin C demonstrated good proton selectivity, with a V_zero of −46 mV at ΔpH 1.2.

To place the protonophoric activity of pyrrolomycins in the context of the conventional protonophore CCCP, a comparison was performed at a 3 μM concentration of the compounds. Remarkably, pyrrolomycin D demonstrated much greater protonophoric activity (BLM current) than did the uncoupler CCCP (Fig. 5C). The activity of pyrrolomycin C was lower than that of pyrrolomycin D but still superior to that of CCCP (Fig. 5D). The larger protonophoric activity of pyrrolomycin D compared to pyrrolomycin C can be tentatively ascribed to better permeability of the compound with five chloride atoms (pyrrolomycin D) than with four chloride atoms (pyrrolomycin C). It has been shown earlier that chlorination increased the membrane permeability of hydrophobic anions (19, 20).

DISCUSSION

Thanks to their many functions, natural products are an incredibly rich source of antibiotic or antifungal molecules but also of potent anticancer agents and immunomodulatory molecules. Pyrrolomycins, produced by Streptomyces vitaminophilus and other bacteria, are described as some of the most potent natural antibiotics known, with activities in the low-nanomolar range against Gram-positive bacteria (3), as well as activity against biofilms (6). Other natural products share a similar chemical scaffold, such as pyoluleorin (21), TAN-876B (22), and marinopyrroles (23). Interestingly, while all these molecules are described as antibiotics, their mechanisms of actions are still elusive. By investigating the mechanism of action of pyrrolomycins, this work highlights the antibiotic drug potential of pyrrolomycins.

An important finding is that the antibiotic activity of pyrrolomycins is not specific to a select number of Gram-positive bacteria, as previously reported, but is instead much more general. We clearly demonstrated that E. coli, and probably other Gram-negative bacteria, are resistant to pyrrolomycins due to efflux of the xenobiotic through a TolC-dependent AcrAB-independent efflux system rather than to the absence of a target. In addition, our studies demonstrated that M. tuberculosis insensitivity to pyrrolomycins is due to the presence of BSA in the growth medium, which likely binds pyrrolomycins and sequestrates it away from the bacteria. The impact of both BSA and FCS in growth medium calls into consideration the previously reported high selectivity window of pyrrolomycins over eukaryotic cells in cytotoxicity assays, which contrasted with its reported acute toxicity in mice (3). These data show that one must be very careful when evaluating the spectrum of activity of antibacterial compounds, taking great care to account for differences caused by environmental factors such as growth medium.

Investigating the mechanism of resistance to a compound can often help clarify its mode of action. For most known antibiotics that target essential enzymes or the ribosome, there is an attainable frequency of resistance with mutations appearing in the target (rifamycins, beta-lactam antibiotics, fluoroquinolones, and aminoglycosides). However, when compounds act on nongenome-encoded targets, such as membrane lipids, sugars, or peptidoglycans, as is the case for vancomycin, teixobactin (24), or daptomycin (25), the frequency of resistance is often very low, with mutations associated with proteins involved in compound penetration or in the formation of nongenome-encoded targets. The difficulty of obtaining target-associated mutations is suggestive of a nonprotein, nonribosomal target. The few pyrrolomycin-resistant mutants that could be isolated in this work showed low-level resistance, and the mutation most probably affects compound penetration by affecting membrane charge. The only mutation found to cause higher-level resistance was identified in the E. coli ΔtolC mutant and was due to a large genomic deletion that placed the mdtOP efflux genes under the control of a now-active promoter (due to the loss of the AlsR transcriptional repressor). Genetic validation experiments clearly identified that overproduction of the MdtP outer membrane channel is able to phenocopy pyrrolomycin resistance through a mechanism that remains to be defined, but it likely equates to pyrrolomycin efflux from the bacteria, although it is AcrAB independent.

A combination of techniques (scanning electron microscopy, pH-mediated changes in MIC, and potentiometric dyes) were used to help narrow down the likely mode of action of pyrrolomycins to the perturbation of the bacterial membrane potential. Nevertheless, membrane potential can be perturbed by multiple mechanisms, direct or indirect (pore-forming toxins, the perturbation of ion channels, ion pumps, ionophores, etc.). To this extent, the BLM experiments were crucial to demonstrate that pyrrolomycins can specifically move protons across a BLM (containing no protein) in a dose-dependent manner. This clearly defined the mechanism of action of pyrrolomycins to be protonophores capable of depolarizing bacterial membranes and uncouple their oxidative phosphorylation. An efficient proton uncoupler needs to remain associated with the membrane and to have an ionizable group to allow for the transfer of protons from outside the membrane to the cytoplasm. Both pyrrolomycins C and D are very lipophilic (predicted logP values, 4.91 and 5.51, respectively, MarvinSketch 18.23.0; Chemaxon) and carry a pyrrole group and a phenyl hydroxyl group that are both ionizable groups. Analysis of pyrrolomycins I and J suggests that the phenyl hydroxyl group of pyrrolomycins C and D plays an important role in this proton gradient uncoupling. Predictions of the acid dissociation constant (pK_a, 5.1; Fig. S4) of the hydroxyl group of pyrrolomycins suggests that it would allow the shuttling of protons from outside to inside the bacteria and thus uncouple the proton motive force (as depicted in Fig. 6). The lipophilic nature of the pyrrolomycins likely favors their localization in the nonpolar membrane lipids, cycling from the outer leaflet to the inner leaflet to shuttle protons. To support this, the chlorine-rich FabI inhibitor triclosan has also been shown to act as a protonophore using its phenyl hydroxyl group to shuttle protons across the membrane (26). In line with these findings, the cytotoxicity observed with pyrrolomycins is likely due to the observed protonophore activity on mitochondrial membranes.

FIG 6.

Model of the proton gradient uncoupling by pyrrolomycins. Lipophilic pyrrolomycins likely accumulate in the nonpolar lipid membrane because of their physical chemical properties. When oriented to the less acidic internal side of the cytoplasmic membrane leaflet, the pyrrolomycin hydroxyl group will preferentially ionize, releasing a proton into the cytoplasm. When flipped to the outer leaflet of the membrane, the more acidic environment will favor the protonation of the hydroxyl group and, hence, the uptake of an extracytoplasmic proton, which in turn will be released in the cytosol. Overall, the cycling of this process will allow for the net movement of protons from outside the bacterial cell to the inside, uncoupling of the proton gradient, and disturbance of the membrane proton motive force.

A number of natural products have been described to exhibit protonophoric activity, including usnic acid (27), hypericin (28), clusianone (29), polyphenols (30), and others. Nonetheless, all these molecules are less potent than the conventional proton uncoupler CCCP. This work demonstrates that pyrrolomycins C and D are an order of magnitude more active than CCCP and, to our knowledge, the most potent known protonophores. It is plausible that natural products such as pyoluteorin, TAN-876B, and marinopyrroles, which have strong structural similarities to pyrrolomycins, may also act as protonophores.

This work illustrates the importance of evaluating the mechanism of action of antibiotic molecules before developing them as potential drugs. We show that pyrrolomycins are potent inhibitors of bacterial membrane potential through their protonophore activity that allows for the specific shuttling of protons across polarized lipid membrane bilayers. This potent activity, which is an order of magnitude higher than that of the benchmark CCCP, explains its broad spectrum of antibiotic activity and provides an explanation for their acute cytotoxic effects found in vitro and reported in vivo.

MATERIALS AND METHODS

Strains and media.

Staphylococcus aureus SH1000 was kindly provided by S. J. Foster (University of Sheffield). Escherichia coli BW25113 (CGSC 7636) and its derivatives E. coli ΔtolC (JW5503-1, ΔtolC732::kan, CGSC 11430), E. coli ΔacrA (JW0452-3, ΔacrA748::kan, CGSC 11843), and E. coli ΔacrB (JW0451-2, ΔacrB747::kan, CGSC 8609) were obtained from the Coli Genetic Stock Center (CGSC) and originated from the Keio Collection (31). Mycobacterium tuberculosis H37Rv was kindly provided by the Institut Pasteur in Paris, France. Acinetobacter baumannii LMG1025, Pseudomonas aeruginosa LMG 12228, Klebsiella pneumoniae LMG 2095, and Streptococcus pneumoniae LMG 14545 were obtained from the BCCM/LMG bacterial collection. HepG2 (HB-8065) and HEK-293T (CRL-3216TM) cell lines originated from the American Type Culture Collection (ATCC).

M. tuberculosis was conventionally grown in Middlebrook 7H9 supplemented with 10% OADC, 0.2% glycerol, and 0.05% Tween 80 or, to evaluate the impact of bovine serum albumin (BSA), in Middlebrook 7H9 supplemented with 0.2% glucose, 0.2% glycerol, 0.01% tyloxapol, 0.095% NaCl, and with or without 0.5% BSA. All other bacteria were grown in cation-adjusted Mueller-Hinton broth (CAMHB; Difco), alone or with supplementation of 10% fetal calf serum (FCS) or 0.5% BSA. Both HepG2 and HEK-293 cells were cultured and tested in Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% FCS.

Chemical synthesis of pyrrolomycins C, D, I, and J.

Pyrrolomycins C, D, I, and J were chemically synthesized as described previously (8, 32, 33) and are summarized in the supplemental material. The chemical identity and structures were confirmed by high-resolution mass spectrometry and ¹H-NMR in comparison to literature data.

Determination of MIC, MBC, and cytotoxicity.

The MICs for E. coli, S. aureus, S. pneumoniae, M. tuberculosis, A. baumannii, and K. pneumoniae were determined using the colorimetric resazurin microtiter assay (REMA). Briefly, compounds dissolved in dimethyl sulfoxide (DMSO) were distributed by an Echo liquid handler (Labcyte) into the wells of a 384-well plate (500 nl of compound/DMSO per well). Plates were subsequently incubated with a diluted bacterial culture (50 μl) and incubated (see Table S1 for conditions). Bacterial viability was then determined by the resazurin reduction assay, as measured by fluorescence (excitation [Ex], 530 nm; emission [Em], 590 nm). For P. aeruginosa, which does not allow the entry of resazurin, bacterial viability was determined by the bacterial optical density at 600 nm (OD₆₀₀). For all experiments, rifampin or ciprofloxacin was run in parallel as a control.

Solid MICs were determined by mixing autoclaved CAMHB with 1.5% agar with compound dilutions in DMSO in a 96-well plate. Bacteria were then added to the solidified medium (5 μl OD of 0.001), and wells were scored for growth after 24 and 48 h of incubation (37°C).

The minimal bactericidal concentration (MBC) was determined by measuring the impact of compounds on the CFU. Briefly, in a 96-well plate, 100 μl of bacterial suspension was incubated in the presence of compound. The number of viable bacteria was then determined at different time points by plating bacterial dilutions on LB agar, and the CFU were determined following overnight incubation at 37°C.

Cytotoxicity of pyrrolomycins.

For cytotoxicity assays, compounds were added in DMSO to a 384-well plate by Echo liquid handling. Fifty microliters of HepG2 or HEK-293 cells in DMEM (without phenol red) supplemented with 10% FCS was then seeded into each well (5,000 cells per well). Cells were incubated for 3 days (37°C, 5% CO₂), cellular viability was determined by the addition of resazurin (final concentration, 0.025%, 4 h), and fluorescence was determined (Ex, 530 nm; Em, 590 nm).

Isolation of pyrrolomycin D-resistant mutants, whole-genome sequencing, and variant analysis.

The selection of E. coli or S. aureus pyrrolomycin D-resistant clones was performed by plating 100 μl of a concentrated log-phase culture (concentrated to OD₆₀₀ of 1, 10, 100, and 500) onto CAMHB agar containing various concentrations of pyrrolomycin D. Plates were incubated for 48 h at 37°C to allow for the appearance of resistant colonies. Resistance of isolates was confirmed by growing colonies on CAMHB agar without pyrrolomycin D, followed by MIC analysis. Whole-genome sequencing and variant analysis of the resistant isolates and the parental wild-type were performed by Genoscreen (Lille, France). Identified genetic variations were confirmed by Sanger sequencing.

E. coli mutant construction.

To validate that the identified alsR-mdtN deletion in the E. coli ΔtolC mutant is the cause of resistance to pyrrolomycin D, this deletion was generated in the parental strain. For this, a number of strains were designed and are summarized in Fig. 2 and Table 2. The primers used are listed in Table S2.

To generate EP664, the BW25113 ΔtolC::Km Keio collection strain was transformed with pFLP2 encoding the FLP recombinase (34) to remove the kanamycin (Km) cassette. Ampicillin-resistant (Amp^r) transformants were screened for Km sensitivity (Km^s) to assess excision of the FRT-Km-FRT cassette (Table S2). One such Amp^r Km^s mutant was further streaked on an LB–5% sucrose plate to cure pFLP2 and yield the unmarked ΔtolC derivative, EP664.

EP673 was constructed from EP664(pEP1436) by red recombination at 30°C (35) using the PCR product generated with primers RH296 and RH297 and the pEP1446 template (Table S2). Correct integration of the Km1446 cassette into the als locus was assessed by PCR with primers RH278 and RH279, and pEP1436 was cured by streaking on an LB–5% sucrose plate at 37°C.

To construct EP676, arabinose-induced EP673(pEP1436) was electrotransformed with the hybridized 78-mer oligonucleotides RH298 and RH299 (Table S2). Clones were selected at 30°C on LB-Amp plates containing doxycycline (100 μg/ml) and then screened for Km sensitivity. Correct allelic exchange was confirmed by PCR with primers RH278 and RH279 and by sequencing of the product. A Km^s clone was then cured of pEP1436 by streaking on an LB–5% sucrose plate at 37°C.

To obtain EP680, arabinose-induced EP676(pEP1436) was electrotransformed with the PCR product generated with primers RH300 and RH301 on the pEP1446 template (Table S2). Integration of the Km1446 cassette at the mdtOP locus was checked by PCR with primers RH279 and K2F. A Km^r clone was then cured of pEP1436 by streaking on an LB–5% sucrose plate at 37°C.

To construct EP705 (ΔompA mutant) and EP706 (ΔslyA mutant), arabinose-induced EP664(pEP1436) was electrotransformed with the PCR product generated with primers RH732 and RH733, or RH728 and RH729 (Table S2) on the pEP1446 template. Integration of the Km1446 cassette in the ompA or slyA gene was checked by PCR with primer pairs RH734 and RH735 or RH730 and RH731, respectively. For each construct, a Km^r clone was then cured of pEP1436 by streaking on an LB–5% sucrose plate at 37°C.

Plasmid construction.

To generate pEP714, mdtP was first PCR amplified from the BW25113 genome with the RH723-RH724 primer pair and cloned into pCRbluntIITopo. The absence of mutation in the insert was assessed by sequencing. Then, mdtP was subcloned into pBAD30 as a SalI-HindIII DNA fragment.

To construct pEP712, alsR was first PCR amplified from the BW25113 genome with the RH726-RH727 primer pair and cloned into pCRbluntIITopo. The absence of mutation in the insert was assessed by sequencing. Then, alsR was subcloned as a SalI-XbaI DNA fragment into pFU95 digested with the same enzymes to yield pEP712, in which alsR is transcribed from the strong constitutive promoter PgapA. The control empty plasmid pEP713 corresponds to the intramolecular ligation of digested pFU95.

Sample preparation for electron microscopy.

Log-phase E. coli ΔtolC mutant and S. aureus SH1000 cultures grown in CAMHB were diluted to an OD₆₀₀ of 0.4, and 400 μl of culture was placed in a well of a 24-well plate containing a sterile glass coverslip. Cells were incubated for 5 h at 37°C with or without 100 ng/ml pyrrolomycin D. After incubation, 400 μl of 2.5% (wt/vol) glutaraldehyde (in phosphate-buffered saline [PBS]) was added to the wells for 3 min. All of the supernatant was subsequently aspirated and the cells covered with 400 μl of 2.5% (wt/vol) glutaraldehyde (overnight, 4°C). Bacteria were fixed by the addition of glutaraldehyde (2.5% [wt/vol]) glutaraldehyde in PBS at 4°C overnight.

Scanning electron microscopy.

After washing, the bacteria on the glass cover slide were treated with 1% osmium tetroxide in water (30 min in the dark). Bacteria were then dehydrated with increasing ethanol concentration baths up to 100% ethanol and dried with a critical point drier (Quorum Technologies K850, Elexience, France). The dry coverslips were then mounted on stubs and coated with 5 nm platinum (Quorum Technologies Q150T). Bacteria were observed by scanning electron microscopy using a secondary electron detector in a Merlin compact VP SEM (Zeiss, France) operated at 2 to 3 kV.

Membrane depolarization assay using DiSC₃(5).

Experiments using 3,3-dipropylthiacarbocyanine iodide DiSC₃(5) were conducted under conditions similar to those described previously (18). Briefly, log-phase bacteria were pelleted (12,000 × g, 5 min) and washed twice in buffer A (10 mM potassium phosphate buffer [pH 7] containing 250 mM sucrose) for S. aureus and in buffer B (5 mM potassium phosphate [pH 7.3] and 20 mM glucose) for the E. coli ΔtolC mutant. Bacteria were then resuspended (to an OD₆₀₀ of 0.085 for S. aureus and OD₆₀₀ of 0.4 for the E. coli ΔtolC mutant) in their respective buffers with 10 μM DiSC₃(5) and 100 mM KCl. One hundred microliters of cells was added to the wells of a 96-well plate and placed into a prewarmed (37°C) fluorescence plate reader (EnVision; Perkin Elmer). Bacterial accumulation of DiSC₃(5) was followed by measuring DiSC₃(5) fluorescence (excitation, 620 nm; emission, 685 nm) until it stabilized [fluorescence decreased due to fluorescence quenching upon DiSC₃(5) accumulation in the bacteria]. Upon fluorescence stabilization, different concentrations of pyrrolomycin or CCCP were added to the bacteria, and depolarization was followed by kinetic analysis of DiSC₃(5) fluorescence. Experiments were also performed in the absence of bacteria to determine quenching of DiSC₃(5) fluorescence by the compounds themselves. Bacterial membrane depolarization was determined as the compound-mediated increase in DiSC₃(5) fluorescence 10 min postexposure.

Membrane depolarization assay using DiOC₂(3).

To follow bacterial membrane depolarization by 3,3′-dihexyloxacarbocyanine iodide [DiOC₂(3)], S. aureus was grown in CAMHB to a log-phase culture (OD₆₀₀, ∼0.4). Bacteria were diluted 1:1,000 in PBS and 1 ml added to microtubes containing 10 μl of test compounds in DMSO at a 100× final concentration. Following a 15-min incubation at room temperature, 10 μl of 3 mM DiOC₂(3) in DMSO was added to the bacteria, mixed, and incubated for a further 15 min. The bacteria were subsequently analyzed by flow cytometry using a AttuneNxT cytometer (Thermo Fisher). For flow cytometry parameters, initially, the bacterial population was selected on forward scatter/side scatter (FSC/SSC) dot plot read in logarithmic scale, using PBS as a control for background noise. Optimal parameters of threshold were found to be FSC 2 × 1,000 and SSC 0.1 × 1,000. DiOC₂(3) fluorescence of the bacterial population was then analyzed in both the BL1 channel (excitation, 488 nm; emission, 530 nm) and BL2 channel (excitation, 488 nm; emission, 590 nm) for green and red fluorescence, respectively. Depolarization could be visualized as a decrease in BL1 fluorescence and an increase in BL2 fluorescence and was calculated using the BL2/BL1 ratio.

Liver mitochondrial respiration assay.

Rat liver mitochondria were isolated by differential centrifugation (28) in a medium containing 250 mM sucrose, 5 mM MOPS, and 1 mM EGTA (pH 7.4). The final wash was performed in the medium additionally containing bovine serum albumin (0.1 mg/ml). The protein concentration was determined using the Biuret method. Handling of animals and experimental procedures were conducted in accordance with the international guidelines for animal care and use and were approved by the institutional ethics committee of the A. N. Belozersky Institute of Physico-Chemical Biology at Moscow State University.

Respiration of isolated mitochondria was measured using a standard polarographic technique with a Clark-type oxygen electrode (Strathkelvin Instruments, UK) at 25°C using the 782 system software. The incubation medium contained 250 mM sucrose, 5 mM MOPS, and 1 mM EGTA (pH 7.4). The mitochondrial protein concentration was 0.8 mg/ml. Oxygen uptake is expressed as nmol/min·mg protein.

BLM experiments to define the protonophore activity of pyrrolomycins.

A bilayer lipid membrane (BLM) was formed by the brush (36) from a 2% decane solution of diphytanoylphosphatidylcholine (DPhPC) on a 0.6-mm aperture in a Teflon septum separating the experimental chamber into two compartments of equal 3-ml volumes. Electrical current across the BLM was measured under voltage-clamp conditions with two AgCl electrodes placed into the solutions on the two sides of the BLM via agar bridges, using a Keithley 428 amplifier (Cleveland, OH, USA). The voltage applied to the BLM was 50 mV.