Efflux pump inhibitor CCCP to rescue colistin susceptibility in mcr-1 plasmid-mediated colistin-resistant strains and Gram-negative bacteria

Abstract

Objectives

Efflux in bacteria is a ubiquitous mechanism associated with resistance to antimicrobials agents. Efflux pump inhibitors (EPIs) have been developed to inhibit efflux mechanisms and could be a good alternative to reverse colistin resistance, but only CCCP has shown good activity. The aim of our study was to identify CCCP activity in a collection of 93 Gram-negative bacteria with known and unknown colistin resistance mechanisms including isolates with mcr-1 plasmid-mediated colistin resistance.

Methods

Colistin MIC was evaluated with and without CCCP and the fold decrease of colistin MIC was calculated for each strain. In order to evaluate the effect of this combination, a time–kill study was performed on five strains carrying different colistin resistance mechanisms.

Results

Overall, CCCP was able to reverse colistin resistance for all strains tested. The effect of CCCP was significantly greater on intrinsically colistin-resistant bacteria (i.e. Proteus spp., Serratia marcescens, Morganella morganii and Providencia spp.) than on other Enterobacteriaceae (P < 0.0001). The same was true for bacteria with a heteroresistance mechanism compared to bacteria with other colistin resistance mechanisms (P < 0.0001). A time–kill study showed the combination was bacteriostatic on strains tested.

Conclusions

These results suggest an efflux mechanism, especially on intrinsically resistant bacteria and Enterobacter spp., but further analysis is needed to identify the molecular support of this mechanism. EPIs could be an alternative for restoring colistin activity in Gram-negative bacteria. Further work is necessary to identify new EPIs that could be used in humans.

Introduction

Efflux in bacteria is a ubiquitous mechanism to combat antibiotic therapy.¹ Multidrug pumps are often non-specific, leading to cross-resistance with several antimicrobial compounds, and can interact synergistically with other resistance mechanisms in order to increase the resistance level. Colistin resistance mechanisms are complex and involve many genes that have not yet been fully identified.² A well-known, identified mechanism involves reducing the negative charge of the outer membrane by adding positive charges such as sugars (phosphoethanolamine, aminoarabinose) to lipid A of the lipopolysaccharide, resulting in the decreased electrostatic attraction of polymyxins.³ Several mutations in genes such as the two-component systems pmrA/pmrB and phoP/phoQ, the negative regulator of PhoP/PhoQ, mgrB and the aminoarabinose biosynthesis operon arnBCADTEF have been involved in colistin resistance.³ Recently, a transferable colistin-resistance gene mcr-1, which codes for a phosphoethanolamine transferase, has been discovered and appears to have already spread worldwide.² Colistin resistance in Gram-negative bacteria has emerged over the last few years with its increasing use in human medicine, but also its wide use in agriculture, especially in animal breeding.^4,5 In some bacteria, efflux pump mechanisms have been reported to play a role in colistin resistance. Many efflux pumps have been identified as reducing colistin susceptibility, such as KpnEF and the AcrAB–TolC complex.^2,3,6 These mechanisms often lead to resistance to other antimicrobial compounds such as aminoglycosides and fluoroquinolones. Various efflux pump inhibitors (EPIs) have been tested on Gram-negative bacteria such as CCCP, 2,4-dinitrophenol (DNP), PABN, reserpine, omeprazole and verapamil.¹ Only CCCP and DNP, which inhibit the energy of efflux pumps, showed a significant decrease in colistin resistance.⁷ The effect of EPIs such as CCCP in the colistin resistance reversion of Klebsiella pneumoniae, Pseudomonas aeruginosa, Acinetobacter baumannii and Stenotrophomonas maltophilia⁷ supported the hypothesis that this mechanism plays a role in colistin resistance. More recently, colistin resistance reversion by CCCP has also been demonstrated in the naturally colistin-resistant species Serratia marcescens.⁸ In our laboratory, we isolated a collection of well-defined colistin-resistant strains including naturally resistant Gram-negative bacteria and acquired-resistance bacteria (mcr-1, mgrB, pmrAB, etc.). The mechanisms of resistance for some of these strains are, however, still unknown. The objective of our study was to evaluate the effect of CCCP on this group in order to evaluate whether the mechanism of resistance can influence the response to efflux pump inhibition and to evaluate its effect on mcr-1 plasmid-mediated colistin-resistant strains.

Materials and methods

Bacterial strains and colistin resistance mechanisms

Strains used in this study are listed in Table 1. Ninety-two strains isolated from various human and pig samples were included in this study. Samples came from France, Laos, Thailand, Nigeria, Algeria and Saudi Arabia (Table 1). Some of these strains have already been published (Table 1). Mechanisms of resistance to colistin have been studied previously with the analysis of phoP, phoQ, pmrA, pmrB, mgrB and mcr-1 performed by real-time PCR, standard PCR and/or sequencing.^9–11 Two reference strains of Escherichia coli, one susceptible to colistin (ATCC 25922) and one mcr-1-positive strain (NCTC 13846), were used as controls. The mcr-1-positive strain has been included in the statistical analysis.

Table 1.

Summary of strains used in this study.

Strain (reference)	Name	Source/ country	Mechanism of resistance to colistin	Microdilution assay (mg/L)			Conclusion effect	Disc diffusion assay (mm)
				MIC of colistin	MIC of colistin + CCCP	MIC fold change		MH agar	MH agar + CCCP	difference
E. coli	ATCC 25922	–	colistin susceptible	2	0.5	4	–	19	27	8
E. coli	NCTC 13846	human/England	mcr-1	4	1	4	reverse	16	19	3
E. asburiaea	TH66	human/Thailand	unknown	32	0.5	64	reverse	6	22	16
E. asburiae6	LH74	human/Laos	unknown	>256	<0.25	>1024	reverse	6	27	21
E. cloacae6	NH52	human/Nigeria	unknown	>256	<0.25	>1024	reverse	18	22	4
E. cloacaea	NH132	human/Nigeria	unknown	>256	0.25	>1024	reverse	17	22	5
E. cloacaea	NH131	human/Nigeria	unknown	>256	0.25	>1024	reverse	17	25	8
E. cloacae23	SB1	human/France	mcr-1	8	0.5	16	reverse	17	21	4
E. coli10,24	P17	pig/Laos	mcr-1	4	0.25	16	reverse	12	21	9
E. coli10,24	LH30	human/Laos	mcr-1	4	0.25	16	reverse	15	25	10
E. coli25	6R	human/Saudi Arabia	mcr-1	8	<0.25	>32	reverse	16	21	5
E. coli10,24	P6	pig/Laos	mcr-1	8	0.5	16	reverse	15	22	7
E. coli25	44A	human/Saudi Arabia	mcr-1	8	0.5	>16	reverse	12	20	8
E. coli26	SE65	human/Algeria	mcr-1	8	<0.25	>32	reverse	12	21	9
E. coli10,24	LH57	human/Laos	mcr-1	8	0.5	16	reverse	15	20	5
E. coli10	TH176	human/Thailand	unknown	8	0.5	16	reverse	17	26	9
E. coli10,24	LH257	human/Laos	mcr-1	16	0.5	32	reverse	17	21	4
E. coli25	134R	human/Saudi Arabia	mcr-1	16	0.5	32	reverse	15	19	4
E. coli25	96R	human/Saudi Arabia	mcr-1	16	0.5	32	reverse	15	19	4
E. coli25	143R	human/Saudi Arabia	mcr-1	16	<0.25	>64	reverse	16	21	5
E. coli10	235	chicken/Algeria	mcr-1	16	<0.25	>64	reverse	15	21	6
E. coli10,24	P10	pig/Laos	mcr-1	8	0.5	16	reverse	15	21	6
E. coli25	117R	human/Saudi Arabia	mcr-1	16	<0.25	>64	reverse	12	19	7
E. coli10,24	LH1	human/Laos	mcr-1	16	<0.25	>64	reverse	15	22	7
E. coli25	95R	human/Saudi Arabia	mcr-1	32	0.5	64	reverse	15	20	5
E. coli25	85R	human/Saudi Arabia	mcr-1	32	0.5	64	reverse	14	20	6
E. coli10	TH99	human/Thailand	mcr-1	64	<0.25	>256	reverse	16	21	5
E. colib	EC1CR	human/France	unknown	64	0.5	128	reverse	16	32	16
E. coli10,24	LH121	human/Laos	mcr-1	16	0.5	32	reverse	17	23	6
E. coli25	1R 2014	human/Saudi Arabia	mcr-1	4	<0.25	>16	reverse	15	23	8
E. colib	EC2CR	human/France	unknown	4	0.5	8	reverse	17	22	5
E. colia	TH66	human/Thailand	unknown	4	<0.25	>16	reverse	16	25	9
K. oxytoca9	FHA124	human/France	unknown	128	1	128	reverse	18	27	9
K.pneumoniae9	LH12	human/Laos	mgrB stop	16	<0.25	>64	reverse	8	20	12
K. pneumoniae25	119R	human/Saudi Arabia	mcr-1	32	0.5	64	reverse	17	21	4
K. pneumoniae9	FHA60	human/France	mcr-1	32	0.5	64	reverse	16	23	7
K. pneumoniae9	LH92	human/Laos	mcr-1	32	0.5	64	reverse	11	18	7
K. pneumoniae9,27	LH17	human/Laos	mcr-1 + PmrB T157P	32	0.5	64	reverse	15	22	7
K. pneumoniae9	TH34	human/Thailand	PmrB S205P	32	0.5	64	reverse	10	20	10
K. pneumoniae9	LH375	human/Laos	unknown	32	0.5	64	reverse	15	22	7
K. pneumoniae9	TH114	human/Thailand	unknown	32	0.5	64	reverse	8	22	14
K. pneumoniae9,27	LH 131	human/Laos	mcr-1 + mgrB stop	64	0.5	128	reverse	14	20	6
K. pneumoniae9	TH176	human/Thailand	unknown	64	0.25	128	reverse	15	21	6
K. pneumoniae9	TH166	human/Thailand	unknown	64	1	64	reverse	16	22	6
K. pneumoniae9	TH164	human/Thailand	unknown	64	1	64	reverse	16	20	4
K. pneumoniae9	FHM77	human/France	unknown	64	0.5	128	reverse	17	22	5
K. pneumoniae9	FHM169	human/France	mgrB stop	64	1	64	reverse	16	23	7
K. pneumoniae9	LH102	human/Laos	unknown	128	1	128	reverse	16	21	5
K. pneumoniae9,27	LH61	human/Laos	mcr-1 + mgrB subs	>256	<0.25	>1024	reverse	16	20	4
K. pneumoniaeb	KP7CR	human/France	unknown	4	1	4	reverse	18	22	4
K. pneumoniaeb	KP1CR	human/France	unknown	16	1	16	reverse	18	20	2
K. pneumoniaeb	KP3CR	human/France	unknown	16	1	16	reverse	17	19	2
K. pneumoniae9	TH205	human/Thailand	unknown	32	1	32	reverse	17	20	3
K. pneumoniaeb	KP4CR	human/France	unknown	8	1	8	reverse	17	21	4
K. pneumoniaeb	KP8CR	human/France	mgrB del	32	1	32	reverse	18	22	4
K. pneumoniaeb	KP9CR	human/France	unknown	>256	1	>256	reverse	16	20	4
K. pneumoniae9	TH68	human/Thailand	unknown	16	1	16	reverse	17	22	5
K. pneumoniae9	TH21	human/Thailand	unknown	64	1	64	reverse	17	23	6
K. pneumoniae9	LH94	human/Laos	unknown	16	0.5	32	reverse	15	22	7
K. pneumoniaeb	KP5CR	human/France	unknown	32	1	32	reverse	18	25	7
K. pneumoniae9	LH140	human/Laos	unknown	4	0.5	8	reverse	12	22	10
K. pneumoniaeb	KP2CR	human/France	unknown	4	0.5	8	reverse	14	25	11
K. pneumoniae9	FHA105	human/France	mgrB ND	32	0.5	64	reverse	15	27	12
K. pneumoniae9	FHM120b	human/France	mgrB ND	64	1	64	reverse	16	24	8
K. pneumoniaeb	KP6CR	human/France	unknown	8	1	8	reverse	12	31	19
M. morganiic	KON	human/France	natural	>256	0.5	>512	reverse	6	NG	>30
M. morganiia	FM102	human/France	natural	>256	1	>256	reverse	6	34	28
M. morganiia	FHA60	human/France	natural	>256	1	>256	reverse	6	32	26
P. aeruginosaa	FHM-PACOLR1	human/France	unknown	>256	0.5	>512	reverse	25	30	5
P. aeruginosac	AMO	human/France	unknown	>256	0.5	512	reverse	23	23	0
P. aeruginosac	GON	human/France	unknown	>256	<0.25	>1024	reverse	23	32	9
Providencia alcalifaciensa	TH44	human/Thailand	natural	>256	1	>256	reverse	6	28	22
P. alcalifaciensa	TH66	human/Thailand	natural	>256	<0.25	>1024	reverse	6	28	22
Proteus mirabilisa	FH112	human/France	natural	>256	0.5	>512	reverse	6	17	11
P. mirabilisa	TH41	human/Thailand	natural	>256	<0.25	>1024	reverse	6	22	16
P.vulgaris29	P97	human/Algeria	natural	>256	<0.25	>1024	reverse	6	38	32
Providencia rettgeria	HI734	human/France	natural	>256	1	>256	reverse	6	21	15
P. rettgeria	TH66	human/Thailand	natural	>256	<0.25	>1024	reverse	6	30	24
Salmonella enterica30	100RC3	human/Saudi Arabia	unknown	16	0.5	32	reverse	17	37	20
S. enterica30	65RC	human/Saudi Arabia	unknown	16	1	16	reverse	17	34	17
S. marcescens28	1237	human/Algeria	natural	>256	<0.25	>1024	reverse	256d	0.064	4000
S. marcescens28	Oxa-48	human/France	natural	>256	<0.25	>1024	reverse	256d	0.094	2723
S. marcescens28	2186	human/Algeria	natural	>256	0.5	>512	reverse	256d	0.094	2723
S. marcescens28	1332	human/Algeria	natural	>256	0.5	>512	reverse	256d	0.094	2723
S. marcescens28	1262	human/Algeria	natural	>256	<0.25	>1024	reverse	256d	0.125	2048
S. marcescens28	50	human/Algeria	natural	>256	<0.25	>1024	reverse	256d	0.125	2048
S. marcescens28	1036	human/Algeria	natural	>256	<0.25	>1024	reverse	256d	0.125	2048
S. marcescens28	122	human/Algeria	natural	>256	1	>256	reverse	256d	0.125	2048
S. marcescens28	200	human/Algeria	natural	>256	1	>256	reverse	256d	0.125	2048
S. marcescens28	1211	human/Algeria	natural	>256	0.5	>512	reverse	256d	0.125	2048
S. marcescens28	1072	human/Algeria	natural	>256	0.5	512	reverse	256d	0.125	2048
S. marcescens28	237	human/Algeria	natural	>256	0.5	>512	reverse	256d	0.19	1347
S. marcescens28	567	human/Algeria	natural	>256	1	>256	reverse	256d	0.19	1347
S. marcescens28	2701	human/Algeria	natural	>256	1	>256	reverse	256d	0.19	1347

ND, not detected; NG, no growth; del, deletion.

a

A. O. Olaitan and J. M. Rolain, unpublished data.

b

S. A. Baron, N. Cassir and J. M. Rolain, unpublished data.

c

S. A. Baron and J. M. Rolain, unpublished data.

d

Performed using the Etest method (mg/L).

Effect of CCCP on colistin MIC using the broth microdilution method

Exploration of the effect of the oxidative phosphorylation uncoupler CCCP on colistin MIC was performed as follows: MIC determination of colistin was performed according to EUCAST recommendations using the broth microdilution method. Colistin sulphate salt (MP-Biomedicals, LLC, Illkirch Graffenstaden, France) concentrations ranged from 0.25 to 256 mg/L. Two tests were performed in parallel: one without adding CCCP and one adding CCCP to each CAMHB well. A stock solution of CCCP was prepared at 5 mg/mL in DMSO. The final concentration of CCCP in the CAMHB was 10 mg/L with a DMSO concentration of 0.2%.⁷ A growth control well containing 10 mg/L CCCP in CAMHB was also added for each strain, in order to check the absence of effect of CCCP alone on these strains. Visualization of bacterial growth was done by adding iodonitrotetrazolium to wells.

The resulting MIC fold changes after the addition of CCCP was calculated as the ratio of the CCCP-free antibiotic’s MIC level to that of the CCCP-added antibiotic. As previously described,⁸ the positive criterion for the presence of efflux pumps in isolates was an ≥8-fold decrease in colistin MIC after adding CCCP. The mean fold change was calculated by species and by colistin resistance mechanism identified as previously described:⁸

$$\left[1/\text{total\hspace{0.17em}sample\hspace{0.17em}size}\left(n\right)\right]\times \Sigma \left(\text{MIC\hspace{0.17em}fold\hspace{0.17em}change\hspace{0.17em}}\times \text{\hspace{0.17em}frequency\hspace{0.17em}of\hspace{0.17em}fold\hspace{0.17em}change}\right)$$

where the ‘frequency of fold change’ is the number of times a particular MIC fold change was recorded for that species. The symbol of superiority for MIC fold changes was considered as an equal for the mean fold change analysis. Statistical analyses were performed using non-parametric one-way analysis of variance (ANOVA) (GraphPad Software Inc., La Jolla, CA, USA). A P value <0.05 was taken into account for the MIC fold change.

The resulting MIC after adding CCCP was compared to the EUCAST cut-off (established at 2 mg/L) and the effect was considered reversed if the strain became susceptible to colistin again.

Effect of CCCP on colistin MIC on Mueller–Hinton (MH) agar plates

Colistin susceptibility testing was performed according to EUCAST recommendations using the Etest (bioMérieux, Marcy-l’Étoile, France) or the disc diffusion method on MH agar with commercial antibiotic discs (bioMérieux). Initially, Etests were performed on S. marcescens in order to better visualize the ‘cocarde’ effect,¹² which corresponds to an absence of inhibition next to the disc surrounded by a ring of inhibition close to the edge of the bacterial growth unaffected by the antibiotic. The withdrawal from the market of Etests forced us to use the disc method for the other species (http://ansm.sante.fr/var/ansm_site/storage/original/application/7e5c29808ae0bc4d37761e98cc8eeb06.pdf). Colistin MIC was determined on two types of media: MH agar with DMSO, and MH agar with 10 mg/L CCCP (Sigma–Aldrich, Saint-Quentin-Fallavier, France) dissolved in DMSO. When a strain was totally inhibited by colistin + CCCP, a plate containing 10 mg/L CCCP only was used to confirm the absence of effect of the EPI alone.

The significance of the change in diameter of colistin inhibition by the disc diffusion method with and without CCCP was evaluated by a Student’s t-test in a matched series. Results were considered as statistically significant when they had a P value <0.05. S. marcescens results were considered separately as the MIC was determined on an agar plate for this species and a fold change was calculated as described above.

Time–kill study

A time–kill study was performed on one strain of each colistin resistance mechanism group, namely, K. pneumoniae FHM169 (mgrB stop), E. coli 44A (mcr-1 positive), Proteus vulgaris P97 (intrinsically resistant to colistin), Enterobacter asburiae LH74 and K. pneumoniae KP4CR (unknown resistance). A fresh culture of bacteria was inoculated in the three following conditions: CAMHB + DMSO, MH broth + colistin (2 mg/L) and CAMHB + colistin (2 mg/L) + CCCP (10 mg/L) and incubated for 24 h at 37°C with shaking. At times 0 h, 30 min, 1 h, 2 h, 4 h, 8 h and 24 h, several dilutions of each culture were spread on Trypticase soya agar and incubated for 24 h before colony counting.

RNA expression

In order to evaluate the impact of CCCP on mcr-1 gene transcription, we quantified mcr-1 RNA in two mcr-1-positive E. coli strains, 44A and P10. We inoculated fresh colonies into three different LB broth cultures, containing DMSO, 2 mg/L colistin + DMSO or 2 mg/L colistin + 10 mg/L CCCP, that were incubated at 37°C with shaking for 4 h. One millilitre of this culture, calibrated to an OD of 0.19 (corresponding to ∼1 × 10⁸ cfu/mL), was extracted using the TRIzol^® Max^TM Bacterial RNA isolation Kit (Thermo Fisher, Waltham, MA, USA). Briefly, after centrifugation, the pellet was resuspended with 700 μL of TRIzol and 50 μL of 4-bromoanisole (BAN). After 15 min of centrifugation at 4°C at 12 000 g, the aqueous phase containing the RNA was suspended in a new Eppendorf tube and precipitated using 500 μL of isopropanol. The RNA was then washed with 75% ethanol and suspended in 30 μL of water after centrifugation. This RNA was purified from the remaining DNA using the DNA-free^TM kit procedure (Thermo Fisher) following the recommendations of the manufacturer. Finally, 20 μg of the RNA was used to quantify gene expression using the SuperScript^TM III Platinum One-Step qRT-PCR Kit (Invitrogen, Carlsbad, CA, USA) following the manufacturer’s recommendations. Each quantification was performed in duplicate. The ΔΔCt method was used to quantify gene expression. The dxS housekeeping gene was used as calibrator using the following primers: F-GCTTCACAATGCCTTTGCCA, R-TATAACGATGGCCCGTCAGC and the following probe: 6 FAM-CGTCAGTTCCACGCCGACCG whereas the mcr-1 primers and probe were those previously described.¹¹

Results

We tested the effect of CCCP on 91 Enterobacteriaceae including 32 K. pneumoniae, 26 E. coli, 6 Enterobacter spp., 2 Salmonella enterica, 1 Klebsiella oxytoca, 14 S. marcescens, 3 Morganella morganii, 3 Proteus spp., 4 Providencia spp. and 3 P. aeruginosa. All strains, except for the colistin-susceptible reference strain ATCC 25922, were found to be resistant to colistin, with an MIC between 4 and >256 mg/L (Table 1). Mechanisms of resistance to colistin had already been identified in some of the strains studied (Table 1).

All isolates used in this study grew in the presence of CCCP at the concentration of 10 mg/L in DMSO without any colistin, in broth medium and on agar plates, confirming that this substance alone has no effect at this concentration on these Gram-negative bacteria (Figure 1).⁷

Figure 1.

Agar method [left, colistin inhibition diameter on an MH + DMSO plate (photograph not to scale) and on an MH + 10 mg/L CCCP plate (photograph not to scale); right, colistin MIC determined using the Etest method with and without 10 mg/L CCCP]. Liquid method (colistin MIC determined using the microdilution method with and without 10 mg/L CCCP). NC, negative control. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Of the 93 strains tested, 90 strains had an MIC fold change ≥8, confirming the presence of the efflux pump mechanism of resistance. Only two strains of K. pneumoniae showed an MIC fold change <8, but their colistin MICs were low (2 and 4 mg/L, respectively) and a reversal effect was still observed after adding colistin (0.5 and 1 mg/L, respectively). Likewise, the addition of CCCP led to reversed colistin resistance (MIC <2 mg/L) in all the strains studied. The mean fold change determined by species varied from 24 for S. enterica to 1024 for colistin-heteroresistant Enterobacter (Table 2). We noticed that the effect of CCCP on intrinsically resistant bacteria was much greater than on other Enterobacteriaceae (P < 0.0001).

Table 2.

Summary of mean fold change in colistin MIC after adding CCCP, by species and by colistin resistance mechanism

Species	Colistin resistance mechanism	Mean fold change
M. morganii	natural	341.3
Proteus spp.	natural	853.3
Providencia spp.	natural	640.0
S. marcescens	natural	621.7
Enterobacteriaceae intrinsically resistant to colistin	natural	618.7
E. coli	mcr-1	46.6
	unknown	36.8
Enterobacter spp.	unknown	64
	heteroresistance	1024
	mcr-1	16.0
Klebsiella spp.	mcr-1	64.0
	other mutation	181.3
	unknown	60.4
S. enterica	unknown	24.0
Other Enterobacteriaceae	—	132.5
Total Enterobacteriaceae	—	260.7
P. aeruginosa	unknown	682.7
Total Gram-negative bacteria	—	274.2

We then decided to compare the impact of CCCP according to the mechanism of resistance identified. Bacteria were separated between ‘intrinsic’ or ‘natural’ resistance, ‘heteroresistance’, which describes a phenomenon where subpopulations of seemingly isogenic bacteria exhibit a range of susceptibilities to a particular antibiotic,¹³ ‘mcr-1 plasmid-mediated’ resistance, ‘other’ resistance when the resistance was caused by a known chromosomic mutation and ‘unknown’ when the resistance mechanism was still unknown. By comparing the resistance mechanism (Table 3), we observed that the efflux mechanism was significantly greater in bacteria naturally resistant to colistin and in bacteria with a heteroresistance mechanism than in bacteria with other resistance mechanisms (P < 0.0001) (Figure 2a). On the other hand, there were no significant differences between ‘intrinsic’ and ‘heteroresistance’ groups or between ‘mcr-1’, ‘other mutation’ or ‘unknown’ groups (Figure 2a).

Table 3.

Summary of mean fold change in colistin MIC after adding CCCP, by colistin resistance mechanism in Enterobacteriaceae

Resistance mechanism	Number of isolates	Mean fold change
mcr-1	24	51.8
Other mutation	9	181.3
Heteroresistance	18	711.1
Naturala	10	614.4
Unknown	32	112.9

a

Excluding S. marcescens.

Figure 2.

(a) Mean fold change of colistin MIC by species and mechanism of resistance. (b) Colistin inhibition diameter on MH agar plate without and with 10 mg/L CCCP. EIRC, Enterobacteriaceae intrinsically resistant to colistin.

The effect of CCCP on agar plates showed significant CCCP activity on colistin resistance. The diameter of inhibition of colistin (mm) increased from 14.1 ± 4.4 mm without CCCP to 23.5 ± 5.3 mm with CCCP (P < 0.0001) (Figure 2b). For S. marcescens strains, the fold change on agar plates with Etest MIC determination varied from 1347-fold to 4000-fold (P < 0.0001). Strains were all resistant to colistin with MIC >256 mg/L before adding CCCP, whereas colistin MIC values after adding CCCP varied from 0.064 to 0.19 mg/L (mean 0.128 mg/L). However, no correlation between the disc diffusion assay and the microdilution method was observed. This can be explained by the fact that the disc diffusion method is not a relevant method to determine colistin susceptibility as the colistin does not diffuse correctly in an agar plate.¹⁴

Analyses of the time–kill study showed that the association of colistin + CCCP was bacteriostatic on the five strains tested (Figure 3). No difference was observed between the strains with different resistance mechanisms. This is concordant with a previous time–kill study done on K. pneumoniae⁷ that showed an inhibiting effect on bacterial growth, but no killing effect.

Figure 3.

Time–kill study of colistin (2 mg/L) + CCCP (10 mg/L) on five colistin-resistant strains with different mechanisms of resistance to colistin.

Finally, the effect of the association of colistin + CCCP on mcr-1 gene expression was performed on two mcr-1-positive E. coli. The addition of 2 mg/L colistin led to a decrease of 1.65-fold for strain 44A and to an increase of 0.84-fold for strain P10 (Figure 4). However, the combination of 2 mg/L colistin + 10 mg/L CCCP decreased mcr-1 expression 40.5-fold and 26.9-fold, respectively, for strain 44A and strain P10.

Figure 4.

mcr-1 gene expression in E. coli 44A and P10 in the absence of antibiotic, after 2 mg/L colistin and after 2 mg/L colistin + 10 mg/L CCCP.

Discussion

The role of efflux in colistin resistance is still unknown in enterobacteria. In our study, we showed a reversal effect on colistin resistance by CCCP on all the strains studied. This effect was more significant in ‘intrinsically resistant’ and ‘heteroresistant’ enterobacteria than in other Enterobacteriaceae, suggesting that the mechanism blocked by CCCP is critical for colistin resistance in these species. Although the main mechanism described for these species is the modification of lipid A by amino sugars, susceptibility to EPIs was significantly greater in this group than in the ‘other mutation’ and ‘mcr-1’ groups for which colistin resistance is also mediated by lipid A modification. Interestingly, the time–kill study showed no differences between the different strains carrying different colistin resistance mechanisms. The association of colistin + CCCP was bacteriostatic on all strains tested, supporting previous results observed on a colistin-resistant K. pneumoniae strain.⁷ Many mechanisms have been identified so far, with the most recent, mcr-1, carried by a transferable plasmid, mostly in E. coli and K. pneumoniae.² The transcriptomic analysis of the mcr-1 gene in two mcr-1-positive E. coli isolates showed that the association of colistin and CCCP inhibits the transcription of the mcr-1 gene. This could likely be explained either by inhibition of efflux (CCCP increases the colistin concentration, which potentiates its effect and inhibits the transcription of the mcr-1 gene) or by an unknown action of CCCP.

Efflux has been reported to play a role in colistin resistance, such as with AcrAB–TolC in E. coli¹⁵ and MexXY–OprM in P. aeruginosa.¹⁶ EPIs have been used to evaluate the effect of efflux pump up-regulation in resistance to colistin. Effects of EPIs on colistin resistance vary according to the type of EPI. For example, PABN and 1-(1-naphtylmethyl)-piperazine (NMP) are EPIs believed to act in competition with drugs in the pocket of the action site of AcrB¹⁷ and they failed to recover colistin susceptibility in colistin-resistant Gram-negative bacteria.^7,18 This inefficiency has also been reported for other well-known EPIs such as verapamil, omeprazole and reserpine.⁷ On the other hand, CCCP and DNP, which are non-specific EPIs, showed good activity in restoring colistin susceptibility.^7,8,18 Therefore, it is difficult to identify the efflux pump responsible for the resistance because they act on the energy source of efflux pumps, the proton motive force, and can modulate other protein activity in the membrane.¹ Ni et al.⁷ have also suggested that the CCCP effect observed on colistin activity may be due to regeneration of negative charges on the cell membrane. In another study, Park and Ko¹⁸ hypothesized the decrease of ATP production caused by CCCP action could be responsible for increased colistin activity in these cells. Our study suggests that CCCP probably has an effect as an EPI. Moreover, the role of soxRS, a modulator of the efflux pump AcrAB-TolC, has recently been reported to be responsible for colistin resistance in Enterobacter spp.⁶ TolC is an outer membrane protein that can interact with many inner membrane proteins, especially membrane fusion proteins, in order to create a channel for various protein extrusions.¹⁹ Usually, bacteria use at least one efflux pump system in the inner membrane such as AcrAB in E. coli that, in association with TolC, enables antibiotic extrusion. Some studies showed that some drugs were dependent on TolC without being dependent on AcrAB, suggesting that TolC can be associated with other inner membrane proteins to evacuate drugs.¹⁹ But until now, none of the studied efflux pumps could prove this hypothesis. In our study, the action of CCCP was particularly important for the heteroresistant strains. In these strains, PABN, which is known to permeabilize the outer membrane of Gram-negative bacteria,²⁰ showed an effect on colistin resistance in Enterobacter cloacae and in Enterobacter aerogenes, suggesting that AcrAB–TolC is up-regulated in these strains, but no effect was observed in S. marcescens strains (data not shown). Despite its resistance being similar to the heteroresistance of Enterobacter spp., some other genes seem to be involved in colistin resistance. In a study on S. marcescens, it has been shown that CCCP increases the accumulation of ciprofloxacin in a WT strain, but not in an sdeB-deficient mutant. The resistance–nodulation–division tripartite efflux pump SdeAB–HasF is homologous to AcrAB–TolC and seems to be the major efflux pump system in Serratia.²¹ This pump seems to extrude a wide variety of components, is dependent on proton motive force and may be a good candidate for a colistin-resistance target gene. Further studies are needed on efflux pump inhibition in colistin resistance to better understand the pathways of this complex mechanism, particularly by describing efflux pump genes and their expression in susceptible/resistant/heteroresistant mutant strains. Our study further suggests that several EPIs could be developed in the future to tackle the issue of emerging colistin resistance in Gram-negative bacteria as demonstrated in the past with β-lactamase inhibitors and resistance to β-lactams. Some studies are currently trying to develop associations of EPIs with other molecules²² to fight bacterial infections.

Funding

This work was supported by the French Government under the ‘Investissements d’avenir’ program managed by the Agence Nationale de la Recherche (reference: Méditerranée Infection 10-IAHU-03).

Transparency declarations

None to declare.