Target (MexB)- and Efflux-Based Mechanisms Decreasing the Effectiveness of the Efflux Pump Inhibitor D13-9001 in Pseudomonas aeruginosa PAO1: Uncovering a New Role for MexMN-OprM in Efflux of β-Lactams and a Novel Regulatory Circuit (MmnRS) Controlling MexMN Expression

Efflux pumps contribute to antibiotic resistance in Gram-negative pathogens. Correspondingly, efflux pump inhibitors (EPIs) may reverse this resistance.

ABSTRACT

Efflux pumps contribute to antibiotic resistance in Gram-negative pathogens. Correspondingly, efflux pump inhibitors (EPIs) may reverse this resistance. D13-9001 specifically inhibits MexAB-OprM in Pseudomonas aeruginosa. Mutants with decreased susceptibility to MexAB-OprM inhibition by D13-9001 were identified, and these fell into two categories: those with alterations in the target MexB (F628L and ΔV177) and those with an alteration in a putative sensor kinase of unknown function, PA1438 (L172P). The alterations in MexB were consistent with reported structural studies of the D13-9001 interaction with MexB. The PA1438_L172P alteration mediated a >150-fold upregulation of MexMN pump gene expression and a >50-fold upregulation of PA1438 and the neighboring response regulator gene, PA1437. We propose that these be renamed mmnR and mmnS for MexMN regulator and MexMN sensor, respectively. MexMN was shown to partner with the outer membrane channel protein OprM and to pump several β-lactams, monobactams, and tazobactam. Upregulated MexMN functionally replaced MexAB-OprM to efflux these compounds but was insusceptible to inhibition by D13-9001. MmnS_L172P also mediated a decrease in susceptibility to imipenem and biapenem that was independent of MexMN-OprM. Expression of oprD, encoding the uptake channel for these compounds, was downregulated, suggesting that this channel is also part of the MmnSR regulon. Transcriptome sequencing (RNA-seq) of cells encoding MmnS_L172P revealed, among other things, an interrelationship between the regulation of mexMN and genes involved in heavy metal resistance.

INTRODUCTION

The Gram-negative bacterial outer membrane (OM) establishes a permeability barrier that limits the influx of toxic compounds (1). Limited influx often conspires with active efflux, particularly by resistance-nodulation-cell division (RND) family pumps, to prevent sufficient intracellular access/accumulation of antibacterial compounds necessary for potent whole-cell antibacterial activity (2). Genetic inactivation of efflux pumps can render Pseudomonas aeruginosa susceptible to various antibacterial compounds that otherwise lack clinically useful potency. There has been substantial interest in the identification of efflux pump inhibitors (EPIs) for use in combination regimens which, in theory, could potentiate the activity of a broad range of pump substrate antibiotics (3). The genome of P. aeruginosa encodes at least 12 members of the RND pump family (4), but only a subset has so far been strongly associated with either intrinsic or mutationally mediated resistance to antibiotics. These include MexAB-OprM, which has a broad range of antibiotic substrates, is constitutively expressed, and is therefore an important contributor to intrinsic resistance in P. aeruginosa as well as to increased resistance due to regulatory mutations that increase pump expression (5–12). Additional pumps known to impact susceptibility to antibiotics are MexXY (OprM), which is inducible by antibiotics that inhibit protein synthesis (13–16), and MexCD-OprJ and MexEF-OprN, which are typically silent but which can be strongly upregulated in so-called nfxB (17–19) and nfxC (20–22) mutants, respectively. Regulatory mutations (e.g., in the repressor gene mexZ) or mutations that interfere with protein synthesis can also lead to the constitutive expression of MexXY and corresponding reductions in susceptibility to antibiotics (13, 16, 23–28). Overexpression of these and other pumps has been found in clinical isolates, presumably selected by therapeutic exposure to antibiotics (2). Given the multiplicity of pumps and their potential overlapping antibiotic pump substrate profiles, the approach of efflux pump inhibition to potentiate antibiotic activity in P. aeruginosa may be complex, and the specificity of an EPI may need to be broad if it is partnered with an antibiotic that is recognized by many pumps. Alternatively, more pump-specific EPIs could potentially find utility if a partner antibiotic had a narrow spectrum of clinically meaningful pump recognition. Examples of EPIs relevant to P. aeruginosa include phenylalanylarginine-β-naphthylamide (PaβN) and related molecules (29, 30), which are proposed to be broad-spectrum (pump) EPIs, and D13-9001, which is more specific to the MexAB-OprM pump in P. aeruginosa (31, 32). In the case of D13-9001, one notion would be to partner this EPI with an antibiotic, such as aztreonam (ATM) (31), whose extrusion from cells is mediated primarily by MexAB-OprM. The antibacterial activity of most β-lactams and monobactams can be decreased substantially by expression of β-lactamases, which could make the benefits of efflux inhibition less promising. However, some β-lactams, such as temocillin, are noted for their stability to a range of β-lactamases, and the emergence of newer more stable monobactams (33) suggests that efflux inhibition could find utility in this context. Despite the strong interest in EPIs over the years as an approach to antibiotic potentiation, there is a paucity of information on mechanisms of resistance to these molecules. To further characterize D13-9001 and explore this more specific EPI approach, we employed combination resistance selection experiments using D13-9001 (Fig. 1) paired with the MexAB-OprM-specific antibiotic carbenicillin (CAR) to identify mechanisms that would reduce susceptibility to this class of EPI. This revealed target-based (MexB) and non-target-based (MexMN pump upregulation) mechanisms of resistance, providing interesting perspectives on the specific EPI approach as well as the potential clinical relevance of MexMN.

FIG 1.

Structure of the narrow-spectrum EPI D13-9001, described in reference 31.

RESULTS AND DISCUSSION

D13-9001 in combination with the MexAB-OprM pump substrate antibiotic carbenicillin selects for alterations in the target protein MexB.

To identify mechanisms that can reduce susceptibility to D13-9001, we employed a combination selection experiment whereby the MIC of the MexAB-OprM pump substrate antibiotic CAR was established in the presence of a fixed level of the MexAB-OprM pump inhibitor D13-9001 (8 μg/ml, as described in Materials and Methods). This fixed level of D13-9001 increased the activity of CAR from 64 or 128 μg/ml to 4 μg/ml, and since D13-9001 does not have appreciable intrinsic antibacterial activity, this presumably reflected inhibition of MexAB-OprM. Using this MIC plate setup, serial passage experiments (tracking the apparent decrease in susceptibility to CAR) were then conducted to enrich for mutants that were less susceptible to the combination. We employed a strain of P. aeruginosa mutated in ampC (NB52019-CDK0002; Table 1) for these experiments to eliminate selection of AmpC β-lactamase hyperexpression. Decreased susceptibility to the combination occurred fairly rapidly under these conditions (4 passages to growth at a CAR MIC of >256 μg/ml). Mutants that had no shift in susceptibility to CAR alone (indicative of specific resistance to D13-9001) were isolated from the final passaging wells. The specific loss of CAR potentiating the activity of D13-9001 was then confirmed by checkerboard analysis. Several of these mutants were screened for mutations in the mexB structural gene, and one isolate, NB52019-CDK0026, harbored a mutation encoding a F628L substitution. There was a ≥32-fold shift in the potentiating activity of D13-9001 against this mutant compared to that against its parent strain (Table 2). This suggested that the change in MexB was responsible for reducing susceptibility to the potentiating activity of D13-9001, consistent with reported costructures showing the importance of this amino acid for interaction with D13-9001 (32) (Fig. 2; see below). Similar shifts in D13-9001 potentiating activity were seen for the monobactams ATM and carumonam (CMN) (Table 2). To directly confirm the role of MexB with the F628L substitution (MexB_F628L) in reducing susceptibility to efflux inhibition by D13-9001, the MexB wild type (MexB_WT) or MexB_F628L was expressed from a plasmid in P. aeruginosa strain NB52245 (ΔmexB ΔmexXY ΔmexCD-oprJ) (Table 1). Either protein restored the efflux of the structurally unrelated pump substrate compounds chloramphenicol (CAM), levofloxacin (LVX), and CHIR-090 (34) and the novel LpxC inhibitor compound 1 (35), indicating that both were functional for efflux (see Table S2 in the supplemental material). As expected, the activity of CAM was potentiated by D13-9001 in NB52245 (pAK1900-MexB_WT), whereas NB52245 (pAK1900-MexB_F628L) was much less susceptible to potentiation by D13-9001. In the presence of CAM at 8 μg/ml, the potentiating activity of D13-9001 shifted >32-fold, from 2 μg/ml for MexB_WT to >64 μg/ml for MexB_F628L (Fig. S1; compare Fig. S1B and C). Furthermore, a shift in susceptibility to efflux inhibition by D13-9001 was also seen using the alamarBlue efflux assay (Fig. 3A to C). We next examined the frequency of spontaneous single-step mutations conferring resistance. For this, agar was prepared with a combination of D13-9001 at a fixed concentration of 64 μg/ml (not antibacterial) and CAR at 8 μg/ml (8× agar MIC of CAR in the presence of D13-9001), and NB52019-CDK0002 was plated for selection of mutants. Mutants would have to be circa 8-fold less susceptible to efflux inhibition by D13-9001 in order to grow. The frequency of spontaneous mutants under this condition was approximately 1.25 × 10⁻⁹. One mutant selected from this experiment, NB52019-CDK0029, also had a mutation in mexB encoding F628L (also with a ≥32-fold decrease in susceptibility; Table 2). F628 plays an important role in D13-9001 inhibitor binding to MexB and AcrB (Escherichia coli), according to site-directed mutation studies (32), and selection of the F628L variant directly using D13-9001 in combination with CAR further supports that observation and validates our selection process. A second mutant, NB52019-CDK0028, was >32-fold less susceptible to efflux inhibition by D13-9001 and had a mutation in mexB encoding MexB_ΔV177 (Table 2). When expressed from plasmid pAK1900-MexB_ΔV177 in the P. aeruginosa NB52244 (ΔmexB ΔmexXY ΔmexCD-oprJ) strain background, MexB_ΔV177 decreased susceptibility to CAM and LVX to levels close to those of wild-type MexB but was not effective in decreasing susceptibility to CHIR-090 or compound 1 (Table S2). Therefore, MexB_ΔV177 is functional for efflux but appears to be compromised in its ability to pump certain compound scaffolds. It was unlikely that MexB_ΔV177 would be defective for efflux of β-lactams, since our mutant selection procedures used this class of inhibitor, but we confirmed this using the novel monobactam LYS228, which is a MexAB-OprM substrate (33) but is stable to the β-lactamase expressed from pAK1900. MexB_F628L and MexB_ΔV177 reduced susceptibility to LYS228 similarly to wild-type MexB (Table S2). Lastly, potentiation of CAM by D13-9001 was decreased >32-fold in cells expressing MexB_ΔV177 (Fig. S1), confirming that MexB_ΔV177 is the specific determinant of decreased susceptibility to D13-9001 in NB52019-CDK0028.

TABLE 1.

Bacterial strains and plasmids

Strain or plasmid	Relevant characteristics	Source or reference
Strains
P. aeruginosa
NB52019	P. aeruginosa PAO1 K767	68
NB52245	P. aeruginosa K2896 ΔmexB ΔmexXY ΔmexCD-oprJ	58
NB52040	P. aeruginosa Z61; ampC encodes premature stop Q294*	ATCC 35151
NB52019-CDK0002	NB52019 ampC inactivated	This study
NB52019-CDK0005	NB52019 with engineered PA1438 encoding the L172P substitution	This study
NB52019-CDK0006	NB52019-CDK0005 ΔmexB	This study
NB52019-CDK0007	NB52019-CDK0005 ΔmexN	This study
NB52019-CDK0008	NB52019-CDK0005 ΔmexB mexN	This study
NB52019-CDK0032	NB52019-CDK0005 ΔoprM	This study
NB52019-CDK0009	NB52019-CDK0002 PA1438 encoding the L172P substitution selected via passaging	This study
NB52019-CDK0026	NB52019-CDK0002 mexB encoding the F628L substitution selected via passaging	This study
NB52019-CDK0028	NB52019-CDK0002 mexB encoding a V177 deletion selected via plating	This study
NB52019-CDK0029	NB52019-CDK0002 mexB with F628L selected via plating	This study
E. coli SM10	thi-1 thr leu tonA lacY supE recA RP4-2-Tc::Mu Kmr, mobilizer strain	69
Plasmids
pAK1900	E. coli-P. aeruginosa shuttle cloning vector, Plac upstream of the multiple-cloning site; Apr Cbr	70
pAK1900-MexBWT	pAK1900 harboring a gene encoding wild-type MexB	This study
pAK1900-MexBF628L	pAK1900 harboring a gene encoding MexBF628L	This study
pAK1900-MexBΔV177	pAK1900 harboring a gene encoding MexBΔV177	This study
pEX18Ap	Gene replacement vector, mob + sacB, Apr	71
pSR3	pEX18Ap:PA1438	This study
pSR4	pEX18Ap:ΔmexN	This study
pSR5	pEX18Ap:ΔmexB	This study
pSR6	pEX18Ap:ampC(Q294*)	This study
pXS1	pEX18TcGm:ΔoprM	This study

TABLE 2.

Shift in susceptibility to β-lactam potentiation activity (efflux inhibition) of D13-9001 for P. aeruginosa PAO1 mutants

Strain	Relevant characteristics	MPC4 (MPC8)c of D13-9001 (μg/ml)
		CAR	ATM	CMN
NB52019-CDK0002	PAO1 ΔampC	≤1 (4)	2 (4)	≤1-2 (4–32)
NB52019-CDK0026a	MexBF628L ΔampC	64 (>64)	64 (>64)	>64 (>64)
NB52019-CDK0029a	MexBF628L ΔampC	32 (64)	>64 (>64)	>64 (>64)
NB52019-CDK0028a	MexBΔV177 ΔampC	>64 (>64)	>64 (>64)	>64 (>64)
NB52019-CDK0009a	PA1438L172P ΔampC	8 (>64)	>64 (>64)	>64 (>64)
NB52019	Wild-type PAO1	2 (4)	2 (8)	4 (32)
NB52019-CDK0005b	PA1438L172P	16-64 (>64)	>64 (>64)	>64 (>64)
NB52019-CDK0006b	PA1438L172P ΔmexB	>64 (>64)	>64 (>64)	>64 (>64)
NB52019-CDK0007b	PA1438L172P ΔmexN	2 (4)	2 (4)	2 (8)

^aSelected mutants.

^bMutation engineered into PAO1 (NB52019).

^cMPC₄ and MPC₈, the minimal D13-9001 concentrations that potentiate the activity of the partner antibiotic 4-fold and 8-fold, respectively.

FIG 2.

Overlay of crystal structure of MexB (PDB accession number 3W9J; in green) on the homology model of MexB_ΔV177 (in yellow). Residues F178 and F628, which pi stack with the inhibitor, are labeled. F178 in the modified sequence has an altered geometry.

FIG 3.

Inhibition of efflux by D13-9001 in P. aeruginosa strains measured by a whole-cell alamarBlue efflux assay. (A to C) Reduced inhibition of MexB_F628L by D13-9001. Pump-deficient P. aeruginosa strain NB52445 complemented by pAK1900-MexB_WT (red lines) or pAK1900-MexB_F628L (blue lines) was monitored over time for the reduction of resazurin to resorufin in the presence of 2 (A), 4 (B), or 64 (C) μg/ml of D13-9001. Efflux by MexB_WT was inhibited at all three concentrations, indicated by the rapid rate of signal generation (red lines). MexB_F628L was not inhibited significantly at 2 or 4 μg/ml D13-9001 but was inhibited at the higher level of 64 μg/ml (blue lines). (D) Reduced inhibition of efflux by D13-9001 at 64 μg/ml in cells upregulated for MexMN expression. Efflux was strongly inhibited in wild-type strain PAO1 (NB52019; red squares) but not in the selected MexMN-upregulated strain NB52019-CDK0009 (blue diamonds) or the engineered MexMN-upregulated strain NB52019-CDK0005 (green triangles). As expected, deletion of mexB (which would be fully inhibited at 64 μg/ml D13-9001) in the MexMN-upregulated mutant did not alter efflux (NB52019-CDK0006; purple multiplication signs), whereas deletion of mexN strongly reduced efflux (NB52019-CDK0007; blue crosses). Therefore, MexMN efflux is not inhibited even at 64 μg/ml of D13-9001. RFU, relative fluorescence units.

Based on the minimized structure of this MexB_ΔV177 sequence (Fig. 2, yellow), a loss in effectiveness of D13-9001 could be attributed to the altered geometry of F178 caused by movement of the entire flexible region resulting from the residue shift. Truncation of this region (Q176 to Y182) would reduce the ability of F178 to effectively pi stack with the ligand (36) relative to wild-type MexB (Fig. 2, green), as the two planes of the phenylalanine side chain and bicyclic aromatic ring of the ligand are modeled to be no longer parallel. The interactions of the pyridopyrimidine of D13-9001 with F178 and of the thiazole ring of D13-9001 with F628 were described in the original cocrystal structure (32), and here we show the importance of V177, specifically, the role of orienting the key aromatic side chain, to stabilizing the ligand in the MexB structure.

Mutations in the putative sensor kinase gene PA1438 reduce susceptibility to β-lactam potentiation by D13-9001.

Certain mutants isolated from our combination selection experiment lacked mutations in mexB. Genome sequencing of one such mutant, NB52019-CDK0009, revealed a point mutation in the gene PA1438 (4) encoding a L172P substitution. PA1438 is annotated as a probable two-component sensor kinase (with PA1437 as a response regulator) (4). There was a ≥8-fold decrease in the D13-9001 potentiation of CAR and circa 32-fold decreases in the potentiation of ATM and CMN (Table 2). To verify that this mutation was responsible for decreasing susceptibility to the potentiating activity of D13-9001, the mutation was engineered into the NB52019 (PAO1) strain background, which recapitulated the shift in D13-9001 susceptibility seen for the original selected mutant (compare the shifts for strains NB52019-CDK0005 [engineered] and NB52019-CDK0009 [selected]; Table 2).

PA1438 regulates expression of the MexMN efflux pump in P. aeruginosa.

Located immediately adjacent to PA1437-PA1438 on the PAO1 genome (Fig. 4A) were two genes (PA1435 and PA1436) encoding putative RND efflux family membrane fusion protein and inner membrane pump components, respectively (4). These were previously determined to encode an efflux pump, designated MexMN, by cloning into an antibiotic-susceptible E. coli strain, where they reduced susceptibility to phenicols (37). Consistent with the genomic proximity of mexMN to PA1437-PA1438, mexN transcripts were elevated circa 250-fold in PA1438 mutant strain NB52019-CDK0009 (selected mutant) and 150-fold in the engineered mutant, NB52019-CDK0005 (Fig. 4B). No significant upregulation of transcript levels for mexA, mexC, mexE, or mexY was observed, although a small decrease in mexA and mexC transcript levels (approximately 2-fold) was observed for NB52019-CDK0005 (data not shown). Therefore, the PA1437-PA1438 two-component regulator pair controls expression of the adjacent mexMN pump genes either directly or indirectly. Furthermore, the PA1437 and PA1438 genes themselves were upregulated >50-fold, consistent with autoregulation (Fig. 4). To our knowledge, this is the first report of a specific regulatory role for this putative two-component regulatory system. Based on the new role identified here, we propose to name PA1437 and PA1438 mmnR and mmnS, for MexMN regulator (response regulator) and MexMN sensor, respectively.

FIG 4.

(A) Genetic organization of the PA1437-PA1438 two-component regulator genes in proximity to the mexMN efflux pump locus. The proposed names for PA1437-PA1438 are indicated in parentheses. (B) Changes in transcript abundance mediated by PA1438_L172P measured by RT-qPCR. Note that the oprD transcripts remained approximately 2.6-fold downregulated if mexN was deleted from strain NB52019-CDK0005.

MexMN effluxes β-lactams and reduces the β-lactam potentiation activity of D13-9001.

The strong upregulation of mexMN expression engendered by MmnS_L172P suggested that efflux by MexMN reduced the β-lactam potentiation activity of D13-9001. Therefore, we deleted mexN from the genome of strain NB52019-CDK0005, which has MexMN upregulated due to MmnS_L172P. Supporting a direct role for MexMN efflux in mediating the shift in susceptibility to D13-9001, the potentiation of CAR activity (presumably reflecting inhibition of MexAB-OprM) was restored upon loss of mexN (strain NB52019-CDK0007, Table 2). In contrast, deletion of mexB had no significant effect (NB52019-CDK0006; Table 2). This suggested that MexMN might efflux D13-9001 directly or, alternatively, that MexMN could extrude β-lactams but was insusceptible to inhibition by D13-9001. Therefore, we examined the β-lactam recognition profile of MexMN using the MexMN-overexpressing mutant strain NB52019-CDK0005 from which mexB (NB52019-CDK0006), mexN (NB52019-CDK0007), or both (NB52019-CDK0008) had been deleted (Table 3). As expected, deletion of mexB in the NB52019 (PAO1) strain background (where MexMN is not expressed) increased susceptibility to several β-lactams, including ATM, BAL30072, CAR, CMN, cefepime, methicillin, moxalactam (MOX), sulbenicillin (SUL), ticarcillin (TIC), and temocillin (TMC) (NB52021; Table 3), consistent with many β-lactams being substrates for MexAB-OprM (38). In contrast, deletion of mexB in MexMN-overexpressing strain NB52019-CDK0005 had much less of an impact (e.g., for CAR) or no impact (e.g., for ATM and CMN) (NB52019-CDK0006; Table 3), consistent with MexMN functionally replacing MexAB in the export of these drugs. Furthermore, deletion of mexN, in addition to mexB (NB52019-CDK0008), clearly increased susceptibility to a range of β-lactams (e.g., ATM, BAL30072, biapenem, CAR, CMN, MOX, SUL, TIC, TMC) (NB52019-CDK0008; Table 3). Therefore, MexMN joins the list of RND family pumps capable of extruding β-lactams and monobactams. Importantly, D13-9001 was not able to potentiate the activity of CAR, ATM, or CMN in the MexMN-upregulated strain from which mexB was deleted, where efflux was presumably solely or largely due to MexMN (NB52019-CDK0006; Table 2), indicating that MexMN is likely refractory to inhibition by D13-9001. This is in keeping with the reported high specificity of D13-9001 for the MexAB-OprM pump in P. aeruginosa (31, 32). Consistent with recognition of β-lactams by MexMN, the β-lactam-based β-lactamase inhibitor (BLI) tazobactam (TAZ) was a substrate of MexMN, as measured by shifts in its intrinsic antibacterial activity (Table 3). This suggests that MexMN upregulation could reduce the effectiveness of BLIs. In contrast, MexMN did not appear to have a significant impact on the intrinsic activity of the non-β-lactam (diazabicyclo-[3.2.1]-octane [DBO])-based BLI NXL-105 (Table 3). A range of additional non-β-lactam antibiotics was also tested in this panel, and none were clearly identified to be MexMN substrates. This is not definitive, since some compounds (e.g., tetracycline) are also effluxed by other pumps (e.g., MexXY), and this could mask the detection of efflux by MexMN using these strain backgrounds (Table 3). However, novobiocin and nalidixic acid are known substrates of MexAB-OprM, and MexMN appeared to contribute slightly or not at all to their efflux (Table 3). Consistent with this, D13-9001 potentiating activity for these antibiotics was not affected by the upregulation of MexMN in NB52019-CDK0005 or NB52019-CDK0009 (data not shown). This implies that MexMN reduces susceptibility to the D13-9001 potentiation activity only for MexMN substrate antibiotics, in turn suggesting that the reduced potentiation by D13-9001 is not due to its own efflux by MexMN. Taken together, these data indicate that the mutation in PA1438 (MmnS) reduces susceptibility to the β-lactam-potentiating activity of D13-9001 by strongly upregulating the D13-9001-insusceptible efflux pump MexMN, which substitutes for MexAB-OprM in extruding β-lactams, such as CAR or ATM, in P. aeruginosa. Furthermore, the alamarBlue efflux assay indicated that this dye is also a substrate of MexMN (Fig. 3D). The efflux of resazurin in the MexMN-upregulated strain NB52019-CDK0005 was not inhibited by D13-9001, regardless of whether MexAB was functional (Fig. 3D). When mexN was deleted, leaving only MexAB-OprM (NB52019-CDK0007; Fig. 3D), inhibition of efflux by D13-9001 was restored.

TABLE 3.

Efflux of antibacterial compounds by the MexMN efflux pump of P. aeruginosa

Compounda	MIC (μg/ml)
	NB52019 PAO1	NB52021 ΔmexB	NB52019-CDK0005 PA1438 (L172P)	NB52019-CDK0006 PA1438 (L172P) ΔmexB	NB52019-CDK0007 PA1438 (L172P) ΔmexN	NB52019-CDK0008 PA1438 (L172P) ΔmexB ΔmexN
ATM	4	0.25	8	8	4	0.25
BAL30072	2	0.25	4	4	1	0.25
Biapenem	0.5	0.5	8	8	8	4
CAR	64	1	64–128	32	64	1
CMN	4	0.25	16	16	4	0.25
FEP	2	1	2	1	2	1
CAZ	2	0.5–1	4	4	1	0.5
CFT	0.5	0.5	1	1	0.5	0.5
IPM	1	1	8	8	8	8
MEM	1	0.25	4	2	4	1
MET	>256	64	>256	128	>256	64
NXL-105	0.25	0.25	0.25	0.25	0.25	0.25
MOX	32	8	64	32	32	8
PIP	4	0.5	4	1	4	0.5
SUL	>256	64	>256	>256	>256	64
TAZ	>1024	128	>1,024	1,024	>1,024	128
TIC	16	0.5	32	8	16	0.5
TMC	512	≤2	512	128	512	≤2
AZM	128	256	128	128	128	128
CHL	32	8	32	8	32	8
CIP	0.25	0.125	0.25	0.125	0.25	0.125
NOV	1024	128	1,024	256	2,048	64
NAL	128	16	64	16	128	16
GEN	1	1	0.5	0.5	0.5	0.5
LVX	0.5	0.25	0.25	0.25	0.5	0.125
TET	32	32	32	32	64	32
TOB	0.5	0.5	0.25	0.25	0.25	0.25

^aATM, aztreonam; AZM, azithromycin; CAR, carbenicillin; CHL, chloramphenicol; CIP, ciprofloxacin; CMN, carumonam; FEP, cefepime; CAZ, ceftazidime; CFT, ceftolozane; GEN, gentamicin; IPM, imipenem; LVX, levofloxacin; MEM, meropenem; MET, methicillin; MOX, moxalactam; NAL, nalidixic acid; NOV, novobiocin; PIP, piperacillin; SUL, sulbenicillin; TET, tetracycline; TIC, ticarcillin; TMC, temocillin; TOB, tobramycin.

OprM is an outer membrane channel for MexMN.

The genes encoding MexM and MexN are not colocalized with a gene encoding an outer membrane channel component on the PAO1 genome (4). Previous studies examining efflux by MexMN expressed in E. coli indicated that the OprM component of MexAB-OprM can function as an outer membrane channel for MexMN in that heterologous background (37). Having a mutant in which MexMN is strongly upregulated allowed us to determine directly if OprM functions as an outer membrane channel for MexMN in P. aeruginosa. As mentioned above, deletion of mexB in the MexMN-overexpressing mutant did not substantially alter susceptibility to many β-lactams, whereas deletion of both mexB and mexN did. Therefore, if OprM, which functions with MexAB, also functioned with MexMN, deletion of oprM should similarly inactivate both pumps and decrease susceptibility. Deletion of oprM in the MexMN-upregulated mutant NB52019-CDK0005 strongly increased susceptibility to CAR, ATM, and CMN (strain NB52019-CDK0032; Table 4), confirming that OprM functions as an outer membrane channel for MexMN in P. aeruginosa PAO1.

TABLE 4.

MexMN is functionally linked to the OprM outer membrane channel

Strain	MIC (μg/ml)
	CAR	ATM	CMN	LVXa
NB52019-CDK0005 PA1438L172P	64	16	16	0.5
NB52019-CDK0006 PA1428L172P ΔmexB	32	8	16	0.25
NB52019-CDK0032 PA1438L172P ΔoprM	2	0.5	0.5	≤0.031

^aNote that LVX is also pumped by MexXY-OprM. Therefore, susceptibility to this antibiotic will also increase more when oprM is deleted than when mexB is deleted.

MmnRS influences expression of the porin channel gene oprD.

Upregulation of MexMN expression in P. aeruginosa mutant strain NB52019-CDK0005 decreased susceptibility to biapenem and imipenem (IPM) 8- to 16-fold (Table 3). This suggested that these antibiotics are substrates of MexMN; however, inactivation of mexN or mexB and mexN did not restore susceptibility (Table 3). IPM gains access to the periplasm of P. aeruginosa cells in part through the OprD outer membrane porin channel (2). Correspondingly, mutations affecting OprD function or expression are known to increase resistance. There was an approximately 3-fold reduction in oprD transcript levels in MmnS_L172P-expressing mutant strains NB52019-CDK0005 and NB52019-CDK0009 (Fig. 4), suggesting that reduced OprD levels contributed to biapenem and IPM resistance in these mutants. Expression of oprD continued to be downregulated if mexN was inactivated (data not shown), consistent with the observation that inactivation of mexN did not restore susceptibility to biapenem or IPM. This also indicates that oprD downregulation is not a downstream effect of MexMN pump overexpression, in turn implying that oprD expression is regulated either directly or indirectly via the MmnSR two-component system.

Preliminary identification of the MmnRS regulon.

Regulatory networks controlling expression of RND efflux pumps can be complex but often involve repressors, such as MexR (for MexAB-OprM) and NfxB (for MexCD-OprJ), or positive activators, such as MexT (for MexEF-OprN), with two-component regulation of efflux perhaps being less common. Examples include BaeRS, the cell wall stress regulatory pair that controls a regulon including mdtABC and acrD efflux genes in E. coli (39–41); EvgAS, which regulates acrAB, tolC, emrKY, mdtEF, and mdfA in E. coli (42, 43); CpxRA, which has recently been shown to regulate mexAB-oprM in P. aeruginosa (44) and several efflux pumps in Klebsiella pneumoniae (45, 46); and ParRS, which controls expression of the MexXY efflux pump and porin OprD in P. aeruginosa (47). It is not known what signals MmnRS naturally responds to or what the natural physiological function of MexMN-OprM might be; however, mexMN and mmnR-mmnS (PA1437-PA1438) were upregulated approximately 4- to 8-fold in response to copper shock in P. aeruginosa and mexMN was upregulated 12- to 23-fold in copper-adapted cultures (48). Deletion of mexM caused a modest increase in susceptibility to growth inhibition by copper, suggesting a role in survival under copper stress. To explore this further, changes in transcript abundance mediated by MmnS_L172P were determined using transcriptome sequencing (RNA-seq) (Table S3). Confirming our initial reverse transcription (RT)-quantitative PCR (qPCR) results, both mexMN and mmnRS expression was significantly upregulated (>300-fold and >30-fold, respectively) and oprD expression was downregulated (1.8-fold), as was that of several other porin genes. Expression of many genes, including several hypothetical genes, was changed (Table S3); however, some themes emerged; another RND family efflux pump, the copper-inducible divalent metal cation (zinc, cadmium, and cobalt) efflux pump czcABC, and its cognate two-component regulatory pair, czcRS (49), were significantly upregulated. The copper resistance genes ptrA and pcoAB and the copper-dependent nitrous oxide reductase (nos) locus (50) were upregulated, as were the hasIS heme-responsive sigma and antisigma factors along with the associated hasAp and hasR heme acquisition and transport genes (51, 52). Intriguingly, two hypothetical genes upstream of hasI (PA3413 and PA34140) were very highly upregulated, suggesting a potential link to hasI. Genes involved in virulence were also upregulated, including the CFTR inhibitory factor (cif) gene (72).

As mentioned above, oprD was downregulated 1.8-fold along with the oprD family porin opdC, both of which were downregulated in copper-adapted cultures (48). Intriguingly, we also observed the strong upregulation of another oprD family transporter, opdT, which was previously reported to be upregulated in copper-adapted cultures. Exposure of P. aeruginosa to copper downregulated expression of oprD via the regulator CzcR and/or CopR, increasing imipenem resistance (49, 53). Although the downregulation of 1.8- to 2.7-fold for oprD resulting from MmnS_L172P was modest, similar reductions were associated with a significant loss of detectable OprD by Western blotting (49, 53). Our results suggest that MmnR may directly act in regulating oprD expression; however, it was shown that overexpression of CzcR repressed oprD, and therefore, MmnS_L172P may mediate reduced oprD expression through induction of czcR. The coregulation of mexMN or czcABC and oprD is reminiscent of the P. aeruginosa pump MexEF-OprN, where downregulation of oprD expression can occur concomitantly with pump upregulation in so-called nfxC mutants (2). Moreover, the interrelationship of mexMN and oprD regulation and the demonstration that the substrate specificity of MexMN includes β-lactams may be consistent with reports that mexMN also appears to reside within the broad regulon of ampR, which includes the P. aeruginosa chromosomal ampC β-lactamase gene, mexEF-oprN, and oprD, among many others (54). The potential complex interrelationships among these regulatory circuits and any role(s) in resistance and, possibly, virulence warrant further exploration.

Concluding remarks.

Here we identified two mechanisms by which P. aeruginosa can become resistant to a highly specific inhibitor of the MexAB-OprM efflux pump D13-9001. The first is alteration of the protein target of the inhibitor, MexB. The second addresses the notion that the long-term effectiveness of pump-specific inhibitors is predicated on partner antibiotics being specifically extruded by the same pump. We found that the least-studied RND pump in P. aeruginosa, MexMN, when upregulated can efflux β-lactams in place of the inhibited MexAB-OprM. In MexMN-upregulated mutants, the specificity of D13-9001 becomes a liability since MexMN is not susceptible to inhibition. Addition of MexMN (OprM) to the list of RND pumps capable of extruding β-lactams may have implications vis-à-vis the relatively high percentage (15 to 29%) of P. aeruginosa isolates from patients with cystic fibrosis (CF) that are susceptible to ticarcillin (TIC) and temocillin (TMC) due to defects in MexAB-OprM efflux (55–57). TMC has been proposed to be a potential therapeutic option to treat this subpopulation in patients with CF (56) since TMC exhibits good stability to β-lactamases and is not significantly impacted by defects in the porin OprD (56). However, TIC and TMC are both pump substrates of MexMN, and therefore, treatment of MexAB-OprM pump-defective strains from patients with CF with TMC or by pairing TMC with a MexAB-specific EPI for broader coverage could be degraded over time by selection of MexMN upregulation. The ability of upregulated MexMN to reduce susceptibility to D13-9001 illuminates some of the potential complexities of trying to match specific EPIs with appropriate partner antibiotics, especially in pathogens such as P. aeruginosa that have many less well-characterized RND pumps, although it remains to be seen if this phenomenon would occur during the clinical use of such combinations.

MATERIALS AND METHODS

Bacterial strains, plasmids, and growth conditions.

The bacterial strains and plasmids used in this study are shown in Table 1. Bacteria were grown in lysogeny broth (1% NaCl, 0.5% yeast extract, 1% tryptone) or cation-adjusted Mueller-Hinton broth (CAMHB) or agar. When necessary, the growth media for the P. aeruginosa strains were supplemented with 150 μg/ml carbenicillin (CAR) (30 μg/ml for E. coli), 25 μg/ml triclosan (Irgasan), and 8% sucrose. Compound 1 (-) was prepared at Novartis (35). D13-9001 was prepared following published procedures (31). BAL30072 and NXL-105 were prepared according to the procedures described in patents WO 2008/116813 (73) and US8063219 (74). CHIR-090 is previously described (34). All other antibiotics were obtained from commercial sources.

Selection of single-step spontaneous mutants with decreased susceptibility to the antibiotic-potentiating activity of D13-9001 using compound combinations.

Single-step selection of mutants with decreased susceptibility to the D13-9001–CAR combinations was performed by growing P. aeruginosa strain NB52019-CDK0002 to an optical density at 600 nm (OD₆₀₀) of approximately 0.6 (mid-log phase) in CAMHB. The culture was pelleted by centrifugation and suspended in fresh medium, and aliquots were plated on CAMHB agar containing both D13-9001 and CAR to select for resistant isolates. The agar contained a combination of D13-9001 (not antibacterial) at 64 μg/ml and CAR at 8 μg/ml (8× agar MIC of CAR in the presence of D13-9001). Under this plating condition, mutants would have to be circa 8-fold less susceptible to inhibition of efflux by D13-9001 in order to grow. Serial dilutions were also plated on Mueller-Hinton broth (MHB) agar without compound for determination of the number of CFU. Mutant frequencies were calculated as the number of CFU on drug-containing plates divided by the number of CFU plated.

Selection of mutants with decreased susceptibility to the antibiotic-potentiating activity of D13-9001 by serial passage in compound combinations.

To select mutants with decreased susceptibility to the carbenicillin-potentiating activity of the MexAB-OprM-specific EPI D13-9001, the concentration of EPI D13-9001 was fixed at 8 μg/ml in MHB, and the MIC for the MexAB-OprM substrate antibiotic carbenicillin was established in this medium (MIC = 4 μg/ml). Then, a serial passage in this medium containing the fixed concentration of D13-9001 was conducted to enrich for mutants with decreased susceptibility to the compound combination. For the serial passages, two different levels of inoculum were used; in one study, 50 μl of a 1:50 dilution of bacteria prepared from passage 0 (from the growth well below the MIC well that had growth similar to that in the no-compound control well) was used as the inoculum for subsequent passages, and a parallel passaging was done using 2.9 μl of a 1:100 dilution of cells. After four passages in either case, the MIC of CAR had increased to >256 μg/ml and cultures from the wells with CAR at 256 μg/ml were streaked onto solid medium to isolate resistant colonies. These were screened by checkerboard analysis with D13-9001 and CAR to identify mutants that were less susceptible to CAR in the presence of D13-9001 but had no decrease in susceptibility to CAR alone, indicating a mechanism specific for D13-9001.

DNA protocols.

The oligonucleotides used in this study are listed in Table S1 in the supplemental material and were purchased from Integrated DNA Technologies (IDT; Coralville, IA). P. aeruginosa genomic DNA was isolated using a DNeasy blood and tissue kit (Qiagen, Hilden, Germany). PCR was carried out using Phusion high-fidelity master mix (Thermo Fisher, Waltham, MA). PCR fragments were purified from agarose gels using a QIAquick gel extraction kit (Qiagen, Hilden, Germany). DNA sequencing was performed by Quintara Biosciences (South San Francisco, CA). Plasmids pAK1900-MexB_WT, pAK1900-MexB_F628L, and pAK1900-MexB_ΔV177 were constructed as follows: a 3,195-bp segment of the mexB gene was PCR amplified from strain NB52019-CDK0002 (with wild-type mexB), NB52019-CDK0026 (with mexB encoding F628L), and NB52019-CDK0028 (with mexB encoding ΔV177) genomic DNA with primers SR44 and SR45. The PCR products were cut with HindIII and SalI FastDigest restriction enzymes (Thermo Fisher, Waltham, MA) and cloned into plasmid pAK1900 (Table 1), which was cut with the same enzymes, in the same orientation as the lac promoter of pAK1900. The sequences of the inserts were confirmed using primers SR44 and SR45. Plasmids pAK1900-MexB_WT, pAK1900-MexB_F628L, and pAK1900-MexB_ΔV177 were then transformed into P. aeruginosa NB52245, which lacks MexAB-OprM, MexCD-OprJ, and MexXY-OprM function (58), using a previously described method (59). Whole-genome sequencing of the P. aeruginosa mutants and identification of single nucleotide polymorphisms were conducted as previously described (60). Detailed descriptions of the construction of P. aeruginosa mutants used in this study are described in the Methods in the supplemental material.

RT-qPCR.

The oligonucleotide primers and probes used for gene expression studies (Table S1) were purchased from Integrated DNA Technologies (IDT; San Diego, CA) and were designed using the IDT PrimerQuest design tool. RNA isolation and RT-qPCR were performed as described previously (61) with the following modifications. RT-qPCR was performed using a qScript XLT one-step RT-qPCR ToughMix kit (Quanta Biosciences, Beverly, MA) on a Bio-Rad CFX96 real-time detection system. The 25 μl RT-qPCR reaction sample consisted of 5 μl of a serial dilution of RNA template (concentration range, 2 ng/ml to 200 μg/μl) and 20 μl of a RT mix containing RT-qPCR ToughMix, primers (450 nM), and probe (150 nM), as recommended by the manufacturer. Each sample was run in duplicate. The quantification cycle (C_q) values were determined using Bio-Rad software (v3.1). Target gene (mexN, mexA, mexC, oprD) expression was quantified by normalization to rrsE expression, as previously described (61).

RNA-seq.

Confluent P. aeruginosa bacterial growth was harvested from a Mueller-Hinton agar plate, diluted to an OD₆₀₀ of 0.05 in cation-adjusted Mueller-Hinton broth, and incubated at 37°C with shaking until the OD₆₀₀ was ∼0.5 to 0.6. Ten milliliters of culture was harvested from three individual flasks (biological triplicates) for each strain. RNA isolation and RNA and DNA quantity and quality assessments were performed as previously described (62). rRNA was depleted using an Illumina Ribo-Zero rRNA removal kit (Bacteria; Illumina, San Diego, CA). RNA-seq libraries were constructed with a ScriptSeq RNA-seq library preparation kit (Illumina), pooled, and size selected (375 to 475 bp) with a Pippin Prep system (Sage Science, Beverley, MA). Libraries were sequenced on an Illumina MiSeq with 150 cycles kit (2 × 75 paired-end reads, single index) following the manufacturer’s instructions. FASTA sequence and GFF annotation files for Pseudomonas aeruginosa PAO1 (GenBank accession number NC_002516.2) were used to build a transcriptome index for Salmon (v0.9.1) software (63). Transcript abundances were then quantified with Salmon software, which was run in quasimapping mode with the gcBias flag. tximport (v1.4.0) tool (64) to construct gene-level count estimates. Standard differential expression analysis was performed with the DESeq2 (v1.16.1) package (65). The work flow followed was that described in the DESeq2 vignette (http://bioconductor.org/packages/devel/bioc/vignettes/DESeq2/inst/doc/DESeq2.html) in Bioconductor software (66). In particular, a false discovery rate (FDR) of 0.05 and a low-count threshold of 10 were chosen; the latter led to the exclusion of 103 genes from further analysis. Three hundred seventy genes showed statistically significant differential expression at the chosen FDR.

Structural analysis.

To generate a hypothesis of the structural effect of the MexB V177 deletion, a crystal structure containing [[2-({[((3R)-1-{8-{[(4-tert-Butyl-1,3-thiazol-2-yl)amino]carbonyl}-4-oxo-3-[(E)-2-(1H-tetrazol-5-yl)vinyl]-4H-pyrido[1,2-a]pyrimidin-2-yl}piperidin-3-yl)oxy]carbonyl}amino)ethyl](dimethyl)ammonio]acetate (ABI-PP) in MexB (PDB accession number 3W9J) (32) was prepared with the Molecular Operating Environment (MOE; v2013.08) system (Chemical Computing Group Inc., Montreal, QC, Canada, 2017) by deleting chains not associated with ligand P9D in the B chain and adding protons to all remaining residues. To see the crude effect of deletion of V177, it was deleted and an amide bond between Gln176 and Phe178 was created. This was followed by minimizing the backbone and side chains of residues Q176 to Y182 using the AMBER10 force field (67).

Data availability.

The RNA-seq data set has been deposited in NCBI GEO under accession number GSE123403.