Substrate dependence of transport coupling and phenotype of a small multidrug resistance transporter in Pseudomonas aeruginosa

ABSTRACT

Small multidrug resistance (SMR) transporters are key players in the defense of multidrug-resistant pathogens to toxins and other homeostasis-perturbing compounds. However, recent evidence demonstrates that EmrE, an SMR from Escherichia coli and a model for understanding transport, can also induce susceptibility to some compounds by drug-gated proton leak. This runs down the ∆pH component of the proton-motive force (PMF), reducing the viability of the affected bacteria. Proton leak may provide an unexplored drug target distinct from the targets of most known antibiotics. Activating proton leak requires an SMR to be merely present, rather than be the primary resistance mechanism, and dissipates the energy source for many other efflux pumps. PAsmr, an EmrE homolog from Pseudomonas aeruginosa, transports many EmrE substrates in cells and purified systems. We hypothesized that PAsmr, like EmrE, may confer susceptibility to some compounds via drug-gated proton leak. Growth assays of E. coli expressing PAsmr displayed substrate-dependent resistance and susceptibility phenotypes, and in vitro solid-supported membrane electrophysiology experiments revealed that PAsmr performs both antiport and substrate-gated proton uniport, demonstrating the same functional promiscuity observed in EmrE. Growth assays of P. aeruginosa strain PA14 demonstrated that PAsmr contributes resistance to some antimicrobial compounds, but no growth defect is observed with susceptibility substrates, suggesting P. aeruginosa can compensate for the proton leak occurring through PAsmr. These phenotypic differences between P. aeruginosa and E. coli advance our understanding of the underlying resistance mechanisms in P. aeruginosa and prompt further investigation into the role that SMRs play in antibiotic resistance in pathogens.

IMPORTANCE

Small multidrug resistance (SMR) transporters are a class of efflux pumps found in many pathogens, although their contributions to antibiotic resistance are not fully understood. We hypothesize that these transporters may confer not only resistance but also susceptibility, by dissipating the proton-motive force. This means to use an SMR transporter as a target; it merely needs to be present (as opposed to being the primary resistance mechanism). Here, we test this hypothesis with an SMR transporter found in Pseudomonas aeruginosa and find that it can perform both antiport (conferring resistance) and substrate-gated proton leak. Proton leak is detrimental to growth in Escherichia coli but not P. aeruginosa, suggesting that P. aeruginosa responds differently to or can altogether prevent ∆pH dissipation.

INTRODUCTION

Multidrug-resistant Pseudomonas aeruginosa is a threat to public health, causing over 30,000 hospital-acquired infections per year in the United States and an estimated 85,000 deaths worldwide in 2019 (1, 2). Some strains appear resistant to all known antibiotics (3). Due to the high degree of antimicrobial resistance and the limited availability of effective therapeutic treatment options, P. aeruginosa is a compelling and clinically relevant system to investigate mechanisms of drug resistance. One such mechanism is secondary active transport of toxic compounds by efflux pumps. The small multidrug resistance (SMR) transporters are a family of efflux pumps that have been identified in many high-priority pathogens, but their contributions to antibiotic resistance are minimally understood (4–11). The SMR family of transporters may be broadly divided into four subfamilies, with the QAC subfamily representing the most promiscuous transporters, named for their ability to confer resistance to quaternary ammonium compounds (11).

Historically, SMR transporters were understood as antiporters, tightly coupling proton import to drug export (12, 13). However, in vitro biophysical investigation of the model QAC SMR transporter EmrE has demonstrated that EmrE is also theoretically capable of symport and drug or proton uniport, with transporter behavior determined by substrate identity and kinetic parameters (14–17). This differential transport behavior, termed the “free-exchange model,” will ultimately lead to the transporter conferring susceptibility to some chemicals instead of resistance (see Fig. 3A). Substrate-gated proton uniport has been experimentally demonstrated for EmrE in Escherichia coli, leading to the disruption of the ∆pH component of the proton-motive force (PMF), impacting cellular ATP production and other PMF-dependent processes (15). PMF disruption has been theorized to be a potential antibiotic and/or adjuvant strategy (18–22). Should other SMR transporters or even other families of efflux pumps confer opposing biological outcomes in a substrate-dependent manner, this may broaden our understanding of how to target pathogens, inducing susceptibility instead of merely disabling specific bacterial resistance processes.

PAsmr, also known as EmrE_Pae, is an SMR transporter expressed by P. aeruginosa, originally identified on the basis of sequence homology to EmrE and conferring resistance to ethidium bromide, methyl viologen, and other quaternary ammonium compounds (4, 7, 23, 24). smrPA (PAO1_4990, PA14_65990), the gene expressing PAsmr, has been identified in clinically isolated strains of multidrug-resistant P. aeruginosa (25). PAsmr has also been investigated as a potential target and model for efflux pump inhibiting peptides (23). We previously carried out a chemical sensitivity screen expressing PAsmr in E. coli, demonstrating that PAsmr displays functional promiscuity, the ability to confer either resistance or susceptibility depending on which substrate is present (24). This is similar to the behavior observed for EmrE (15).

Here, we further explore the ability of PAsmr to confer either resistance or susceptibility depending on substrate and hypothesize that these opposing biological outcomes are explained by different transport modes, based on our prior observations of EmrE (14, 15, 17). We use heterologous expression in E. coli to assess PAsmr activity and find that it can confer either resistance or susceptibility to different substrates. This confirms the phenotypes previously identified in an unbiased screen (24). We use in vitro assays to demonstrate that the resistance phenotypes are due to proton-coupled antiport, while substrate-triggered proton uniport is responsible for susceptibility phenotypes. Finally, we use a P. aeruginosa strain lacking PAsmr to look for phenotypic differences in the presence of both known and novel SMR substrates (Fig. 1) in the native organism, as well as how this phenotype corresponds to those observed upon heterologous expression in E. coli. The results highlight differences in the impact of proton-coupled transport and proton leak in E. coli and P. aeruginosa.

Fig 1.

Known and novel PAsmr substrates. Compounds used in this study.

MATERIALS AND METHODS

Growth assays in E. coli

MG1655 ΔemrE E. coli cells were transformed with either wild-type (WT) or non-functional (E14Q) PAsmr (PAO1-PA4990) cloned into the pWB plasmid, a low-copy, leaky-expression vector (24). Cells were grown in nutrient broth (Difco 234000), 100 µg/mL carbenicillin, from a single colony to an optical density (OD) of 0.2 at 37°C. Cultures were diluted to a final OD of 0.01 in microplates (Corning, REF: 351172) containing select concentration values of tested compounds. The plates were incubated in a microplate reader (BMG-Labtech or TECAN) at 37°C. OD₆₀₀ (or OD₇₀₀ for ethidium) was measured every 5 minutes for 18 hours. Experiments were performed in biological triplicate.

ATP quantitation in E. coli

ATP quantitation was carried out using the commercially available BacTiter-Glo Assay from Promega. Cells were grown as described in the growth assays until 18 hours (or 6 hours when specified) and then transferred to a black 96-well plate and mixed with the BacTiter-Glo Reagent. Plates were shaken for 5 minutes at room temperature (RT) and luminescence detected with a TECAN microplate reader. Experiments were performed in biological triplicate. Statistical significance was determined by Student’s t-test and based on P < 0.05.

PAsmr expression and purification

PAsmr was expressed as a maltose binding protein (MBP) fusion construct, using pet28-TEV-MBP, a gift from Zita Balklava & Thomas Wassmer (Fig. S5A) (26). Following TEV protease cleavage of the N-terminal MBP, the N-terminal 6xHis-tag may be used to purify PaSMR via affinity chromatography and then removed with thrombin leaving an extra N-terminal GSHGS. BL21(DE3) Gold cells transformed with pet28-MBP-PaSMR were grown in M9 minimal media (27), with protein expression induced by 0.33 mM isopropylthiogalactoside (IPTG) at an OD₆₀₀ of 0.9–1 at 17°C. Cells were harvested after 17 hours, and cell pellets were resuspended in lysis buffer (250 mM sucrose, 100 mM NaCl, 2.5 mM MgSO₄, 20 mM tris pH 7.5, 5 mM β-mercaptoethanol, 1 mg/mL lysozyme, DNAse, 1 µg/mL pepstatin, 10 µM leupeptin, and 100 µM PMSF) and lysed by sonication; 0.25 mg/mL TEV protease was added, and following incubation for 16–48 hours at RT or 48–72 hours at 4°C, the membrane fraction was separated by a high-speed spin (30,000 × g for 1), resuspended in the same buffer, and solubilized with 40 mM DM (decylmaltoside, Anatrace) at RT for 2 hours. After a second high-speed spin, 6×-His tagged PAsmr was purified via NiNTA affinity chromatography, thrombin cleavage, and size-exclusion chromatography as previously described for EmrE (Fig. S5B through D) (27). When more complete removal of MBP was required for downstream experiments (e.g., SSME), 50–100 mM maltose was included in all steps of purification through NiNTA column elution and an additional desalting step was added to the end of the purification. The purified sample was reasonably monodisperse (Fig. S5F), and the identity of PAsmr as the purified species was confirmed by whole-protein mass spectrometry analysis (Fig. S5E). We also confirmed that residual MBP was eliminated during the bicelle and liposome reconstitution process for in vitro experiments (Fig. S5G).

Solid-supported membrane electrophysiology

Purified WT PAsmr was reconstituted into 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) liposomes at a lipid-to-protein molar ratio of 250:1 in pH 7 assay buffer (100 mM potassium phosphate, 100 mM sodium acetate, 200 mM NaCl, and 4 mM MgCl₂). Liposome preparation was performed as described in reference (15). All SSME data were acquired using a Nanion SURFE2R N1 instrument as described in reference (28). Liposome aliquots were thawed, diluted twofold with assay buffer, and briefly sonicated. Ten microliters of liposomes was used to prepare 3-mm sensors as previously described (15). Before experiments, sensor capacitance and conductance values were obtained to ensure sensor quality. All experiments used assay buffer with internal pH values of 7.0 and external pH values of 6.7. For each drug, four conditions were tested: no drug, even drug, outward-facing drug gradient, and inward-facing drug gradient. Drug concentrations used are in Table S1. Sensors were rinsed with at least 500 µL of internal buffer before each measurement to set the internal buffer, pH, and drug concentrations as described in reference (28). Data acquisition occurred in three stages. First, sensors were perfused with an internal buffer, then transport was initiated by perfusion of the external buffer, and finally, perfusion of the internal buffer re-equilibrated the sensors. Signals were obtained by integrating the current during perfusion of the external buffer, with the final 100 ms of the initial internal buffer perfusion used as the baseline. Integrated currents were normalized by subtracting the integrated current for no drug/proton transport-only control. For the total transported charge, we reported the average values of three sensors, with error bars representing the standard deviation. Statistical significance was determined by Student’s t-test.

1D ligand-observed NMR

PAsmr was purified as described above. Binding of MeTPP⁺ and streptomycin was assessed by ¹H NMR. Samples contained 500 µM MeTPP⁺ or streptomycin with or without 25 µM PAsmr in DHPC/DMPC bicelles, in a buffer of 20 mM NaCl, 20 mM potassium phosphate, 8% D₂O, and 0.3 mM sodium trimethylsilylpropanesulfonate at pH 7. Samples containing protein also contained 0.05% sodium azide and 2 mM tris (2-carboxyethyl)phosphine. All spectra were acquired on a Bruker Avance III spectrometer operating at 600 MHz. 1D spectra were acquired using the Bruker pulse program zgesgp. WaterLOGSY experiments were recorded at specified temperatures using the Bruker pulse program ephogsygpno.2 with a mixing time of 1.5-s and 2-s recycle delay. All spectra were processed and visualized with Mnova.

Proton leak

PAsmr was reconstituted into 3:1 POPC:POPG liposomes at a lipid-to-protein molar ratio of 400:1. Proton leak was monitored by diluting PaSMR liposomes in inside buffer (50 mM sodium phosphate, 100 mM KCl, pH 7) into weakly buffered outside buffer (75 µM phenol red, 99 mM NaCl, and 1 mM KCl, pH 7) to 0.8 µM final PaSMR concentration. Valinomycin diluted in inside buffer was added to a final concentration of 1 µg/mL to create a negative inside membrane potential. FCCP diluted in inside buffer was added to a final concentration of 1 µg/mL as a positive control. At the end of the assay, HCl (50 nM final) was added to return to resting potential.

smrPA (PA14_65990) deletion in P. aeruginosa PA14

smrPA was deleted in PA14 using sacB counterselection (29, 30). Briefly, a plasmid vector (pJMP3262) containing an origin of replication that replicates in E. coli but not P. aeruginosa, a gentamicin resistance cassette, and the sacB gene was used as the backbone. The regions 1,000 bp upstream and downstream of smrPA were cloned into the plasmid vector via Gibson Assembly.

Following extraction of genomic DNA from PA14, the regions 1,000 bp upstream and downstream of smrPA were amplified via PCR using primers oJMP 1832 and 1833 for the upstream region and oJMP1834 and 1835 for the downstream region (Table 1). The resulting products were gel purified and cloned into pJMP 3262 via Gibson Assembly. The plasmid was then transformed into E. coli (sJMP 2630). The plasmid sequence was confirmed by Sanger sequencing using primers oJMP 1339, 1340, 1836, and 1837 (Table 1). The plasmid was then transformed into a conjugative E. coli strain (sJMP3257).

TABLE 1.

Primers used in PAsmr deletion

JMP primer	Sequence (5′−3′)
1832	CTAGAACTAGTGGATCCCCCCGGGCAAGGACAGGGAAAA
1833	TTCAGTGCCCATTGGTCATGGCGGCGTC
1834	CATGACCAATGGGCACTGAACGACCGTC
1835	ATATCGAATTCCTGCAGCCCCCTGTTCGGCGAGGAAGG
1881	GCTCGGTTTCCATGATCACC
1882	GGATGCCGTCTTCTTCCAGG
2374	CGCTTTTTTTACGTCTGCAGACTAGTTTGACAATTAATCATCCGGCTCGT ATAATGTGTGGAATTGTGAGCGGATAACAATTCTAGAATGGCCGCAGTTTCC
2376	GCTGCAAGTCCCATTAAGGAGGCGCGCCAACGGAAAGGGCGCAGTATA

The newly constructed plasmid donor (sJMP3257) and recipient (WT PA14 sJMP744) were grown overnight on plates and resuspended in 1 mL LB to OD₆₀₀ » 3. Three tubes were prepared, one for mating containing 800 µL LB, 100 µL of the recipient, and 100 µL of the donor strain, and two controls (900 µL LB and 100 µL of either donor or recipient strains alone). Cells were pelleted at 4000 × g for 3 minutes. Media were removed, and the pellet was resuspended in 25 µL of LB and spotted on a filter on a LB + diaminopimelic acid (DAP) plate. Following incubation for 3.5 hours at 37°C, the filter was placed in a 1.5-mL Eppendorf tube with 200 µL LB and vortexed for 30 s. The cells were further diluted in LB at 1:10, 1:100, 1:1,000, 1:10,000, and 1:100,000. Ninety microliters of each dilution was plated onto LB + gent (30 µg/mL) plates to select for P. aeruginosa transconjugants containing the plasmid integrated into the genome at the smrPA (PA14_65990) locus by single crossover. Donor and recipient controls were also plated on LB + gent (30 µg/mL) plates.

Several isolated colonies from the LB + gent (30 µg/mL) plates were grown in 10 mL LB overnight. The resulting overnight cultures were plated on LB + sucrose 15% plates and incubated for 48–72 hours at RT. Incubation at RT enhanced recovery of correct, double-crossover clones. To confirm the selection of ΔsmrPA, 40 isolated colonies from the LB + sucrose 15% plates were patched on LB + gent (30 µg/mL) + sucrose 15%, LB + gent (30 µg/mL), and LB plates. Any colonies growing on the plates containing gent (30 µg/mL) were discarded. To confirm the deletion of PAsmr, the gDNA was extracted from a few samples, and the primers oJMP1881 and oJMP 1882 were used to amplify the region including smrPA. Gel electrophoresis of the PCR product was used to confirm the deletion of the 333-bp gene product (Fig. 6A).

smrPA (PA14_65990) complementation in P. aeruginosa PA14

An IPTG-inducible promoter, Ptrc, was chromosomally inserted upstream of smrPA in the PA14 ΔsmrPA strain background to complement its function. The plasmid expression vector, pJMP 3389, which contains an apramycin (P. aeruginosa selection) and ampicillin resistance (E. coli selection) cassette, was used to create the smrPA-complemented strain. The PCR primers oJMP 2374 (containing P_trc) and oJMP 2376 (Table 1) were used to amplify PA14 genomic DNA and create the Ptrc-smrPA⁺ insertion. Following a gel purification of the PCR product and restriction digest of the mini-Tn7 plasmid pJMP3389, the Ptrc-smrPA⁺ insert was cloned into the digested plasmid vector via Gibson assembly.

The newly assembled plasmid vector was then transformed into E. coli (sJMP 2630) on LB + ampicillin (100 µg/mL) selective plates, and the resulting plasmid sequence was confirmed by whole plasmid sequencing (Nanopore sequencing from Plasmidsaurus). The plasmid was then transformed into a conjugative E. coli strain (sJMP3257). In a tri-parental mating to integrate Tn7 containing the P_trc-smrPA⁺ cassette into the P. aeruginosa ΔsmrPA genome, newly constructed plasmid donor (sJMP3257), recipient donor (PA14 ΔsmrPA), and transposase donor (sJMP 2953) were grown on agar plates overnight and resuspended in 1 mL of LB to OD₆₀₀ ≈ 15, OD₆₀₀ ≈ 5.5, and OD₆₀₀ ≈ 12, respectively. Four tubes, one for mating containing 700 μL LB, 100 μL of the recipient, 100 μL of the transposase donor, and 100 μL of the donor strain, and three controls (no recipient, no transposase donor, and no transposon donor controls) were prepared. The cells were pelleted, and the supernatant was removed. The resulting cell pellet was resuspended in 25 μL of LB and spotted on a filter on a LB + DAP plate. Following incubation for 3.5 hours, the filter was placed in a 1.5-mL Eppendorf tube with 200 μL LB and vortexed. The cells were further diluted in LB and plated on LB + apramycin (50 µg/mL) selective plates.

Growth assays in Pseudomonas aeruginosa

WT PA14 and PA14 ∆smrPA were grown in nutrient broth (Difco 234000) from single colonies to an OD of 1 at 37°C. Cultures were diluted to a final OD of 0.01 in microplates containing select concentration values of tested compounds. Plates were incubated in a microplate reader (TECAN) at 37°C. OD₆₀₀ (or OD₇₀₀ for ethidium) were measured every 15 minutes for 18 hours. Experiments were performed in at least biological triplicate. Data were analyzed by normalizing WT growth and standard deviation by growth of the knock-out strain.

RESULTS

E.coli expressing PAsmr display substrate-dependent growth effects

Previously, we performed an unbiased screen to identify potential substrates of PAsmr using heterologous expression of the transporter in Escherichia coli (24). The screen identified and subsequent experiments confirmed several novel substrates including the small molecules harmane and 18-crown-6 ether. The screen also identified the aminoglycoside streptomycin and the antibiotic chelerythrine chloride as potential substrates of PAsmr. To further assess the opposing biological outcomes seen in the prior study, ∆emre E. coli expressing WT PAsmr were grown in nutrient broth (NB), a low-ionic-strength medium. As a control, we used E. coli expressing E14Q PAsmr. E14Q is a known transport-dead variant of EmrE (31), and E14Q mutation has been shown to abolish the ability of PAsmr to confer resistance to toxic compounds when heterologously expressed in E. coli while maintaining normal growth when no drug is present (24). In that study, we showed that E14Q PAsmr is expressed to comparable levels as WT PAsmr in E. coli and there is minimal growth difference between E. coli expressing either variant of PAsmr in the absence of drug substrate. Therefore, the E14Q variant is a valuable negative control for phenotyping E. coli strains with heterologous expression of SMR transporters.

Expression of WT PAsmr conferred resistance to ethidium bromide (Fig. 2A; Fig. S1A and B); methyltriphenylphosphonium bromide (MeTPP⁺), which is closely related to known substrate TPP⁺ (Fig. 2B; Fig. S1C and D) (4); and chelerythrine chloride (Fig. 2C; Fig. S1E and F). PAsmr expressed in E. coli conferred susceptibility to harmane (Fig. 2D; Fig. S2A, B, S4A and C), 18-crown-6 ether, (Fig. 2E; Fig. S2C, D, S4B and D), and streptomycin (Fig. 2F; Fig. S2E and F). This phenotype may include reduced maximum OD or enhanced log phase depending on the growth conditions (Fig. 2D through F; Fig. S4). Several other putative hits identified in the prior screen showed no growth difference for E. coli expressing WT or E14Q PAsmr (Fig. S3), demonstrating that not all hits identified based on NADH production show a significant growth phenotype. Compared with previously tested growth in higher ionic strength media (24), growth in nutrient broth also resulted in earlier and more pronounced phenotypic differences between WT and E14Q PAsmr (Fig. 2; Fig. S1 to S4). PAsmr phenotypes for harmane and 18-crown-6 ether in low-ionic strength media are consistent but more pronounced than those in Muller-Hinton Broth (24), suggesting that a decreased ability to compensate for ∆pH dissipation may occur in low-ionic strength media. Overall, for compounds tested previously in growth assays, phenotypic results were consistent with growth in E. coli in higher ionic strength media (24) and confirmed that PAsmr confers resistance to the known EmrE substrate, MeTPP⁺. The susceptibility to streptomycin in E. coli shown here is the first demonstration of PAsmr conferring susceptibility to an aminoglycoside antibiotic.

Fig 2.

PAsmr expression in E. coli shows a substrate-dependent phenotype. (A–C) E. coli expressing WT PAsmr shows improved growth relative to the E14Q variant in the presence of ethidium, MeTPP⁺, and chelerythrine, suggesting that PAsmr confers resistance to these compounds. (D and E) E14Q-PAsmr-E. coli shows improved growth relative to WT PAsmr in the presence of harmane, 18-crown-6 ether, and streptomycin, suggesting that functional PAsmr confers susceptibility to these compounds in E. coli. (G) ATP measurements of WT and E14Q PAsmr with harmane, 18-crown-6 ether, and streptomycin at 18 hours and streptomycin at 6 hours. P values were determined by Student’s t-test.

ATP changes due to susceptibility are consistent with proton leak

While the functional promiscuity of PAsmr has been observed once previously (24), the molecular basis of susceptibility and the mechanism by which expression of functional PAsmr activity impairs growth of E. coli have not been established. We hypothesized that this susceptibility is due to substrate-grated proton leak through PAsmr, and this hypothesis is supported by the observation that some phenotypes are enhanced by a low-ionic strength medium. To further test this hypothesis, we also measured ATP levels of E. coli expressing PAsmr or non-functional E14Q PAsmr in the presence of the susceptibility substrates. If the PMF is dissipated due to proton leak, this should result in decreased levels of ATP production and increased levels of ATP consumption to restore PMF homeostasis (32).

ATP levels of E. coli expressing WT PAsmr and E14Q PAsmr were measured after 18 hours of growth. With no drug, E14Q PAsmr had slightly increased ATP levels, which may be explained by slight proton leak through PAsmr or some basal transport by PAsmr that is detrimental to cell growth (Fig. 2G). While the native role of Qac SMR transporters is not fully known, tolerance to pH and osmotic stress have been suggested as native functions and may explain this slight growth difference (33, 34). Addition of harmane and 18-crown-6 ether resulted in significantly decreased ATP compared with the control, consistent with the proton leak hypothesis and the observed growth phenotypes (Fig. 2G). In the presence of streptomycin, ATP levels of cells expressing WT PAsmr were increased, even though the carrying capacity at endpoint was only slightly different from that of the transport-dead control (Fig. 2F). To understand this result further, we measured ATP in the presence of streptomycin after 6 hours of growth, when the greatest difference in the growth phenotype was observed. After 6 hours, cells expressing E14Q PAsmr had significantly greater ATP levels than cells expressing WT PAsmr, consistent with the susceptibility phenotype. Aminoglycosides have a complex and not fully understood mechanism of action. Inhibition of translation increases ATP levels by reducing ATP used for protein synthesis with time-dependent changes in ATP levels noted for E. coli exposed to aminoglycosides (35). In addition, changes in transmembrane voltage are implicated in aminoglycoside uptake and/or bactericidal activity (36, 37), adding complexity to how PAsmr-mediated proton leak and streptomycin activity may combine to influence the proton-motive force and ATP levels.

PAsmr substrate phenotype in E. coli corresponds to transport mode

Next, we investigated how PAsmr can confer opposing phenotypes depending upon substrate. We hypothesized that PAsmr may confer resistance via antiport but susceptibility via drug-gated proton uniport based on our prior studies of EmrE. To test if PAsmr performs only antiport or may perform drug-gated proton leak or other modes of transport (Fig. 3A), we tested four substrates in a purified system using solid-supported membrane electrophysiology (SSME) (28). PAsmr was expressed and purified as a MBP fusion construct (Fig. S5), and following MBP cleavage and size-exclusion chromatography, fractions with minimal MBP were reconstituted into liposomes, which were adsorbed onto a gold electrode sensor. We note that PAsmr is an antiparallel homodimer with one subunit inserted in each orientation to form the functional dimer, removing any concern about the orientation of the protein when reconstituted into the liposome. In SSME, signal is detected as net charge moves in or out of the liposomes when different gradient conditions are applied. A twofold outward pH gradient is established across the liposome in each assay, and baseline charge transport (which is minimal; see Fig. S6 to S9) is determined. With the same twofold pH gradient, the drug gradient is varied: equal drug on both sides, greater drug concentration outside, or greater drug concentration inside (Fig. 3B). Net charge transport in response to establishment of these gradients distinguishes between antiport, symport, and drug or proton uniport (Fig. 3C). Integrated charge traces (Fig. 3D through G; Fig. S8 and S9) are individually normalized by subtracting baseline transport with no drug present.

Fig 3.

PAsmr performs both antiport and proton uniport depending on substrate. (A) Based on the free-exchange model of EmrE, four transport modes are theoretically possible for PAsmr (16). (B) Solid-supported membrane electrophysiology was performed using three different drug gradient conditions plus no-drug condition, all with a consistent twofold proton gradient. (C) Resulting ion transport for each condition allows discrimination between four different transport modes. (D) PAsmr transports MeTPP⁺ and (E) chelerythrine via antiport. (F) Harmane and (G) 18-crown-6 ether trigger increased proton transport regardless of drug gradient. P values were calculated by Student’s t-test for each comparison between even drug condition and gradient conditions. Representative traces are a single replicate where all four conditions are run on a single sensor and normalized to the no drug present (twofold proton gradient only) condition. The results from three different independent sensors (replicates) are then averaged for the bar graphs and statistical tests. Additional replicates and raw traces are available in Fig. S6 and S7.

In all cases, empty liposomes were used as a control (Fig. S8 and S9) and did not show significant differences between the four tested conditions, indicating proper reconstitution of liposomes and minimal leak through the liposomes. For all compounds, addition of drug in equal concentration inside and outside of the liposome resulted in transported charge (Fig. 3D through G; Fig. S6 to S9). For MeTPP⁺ and chelerythrine chloride, addition of an outwardly directed 32× drug gradient resulted in the largest net charge transport, and an inwardly directed 32× gradient resulted in reversal of charge and significant deviation from the “even drug” condition, a hallmark of coupled antiport (28) (Fig. 3D and E; Fig. S6 and S8). This is consistent with the expected pattern and reversal observed for antiport (Fig. 3C) and confirms that PAsmr performs proton-coupled antiport of chelerythrine and MeTPP⁺.

Harmane and 18-crown-6 ether, however, induced net positive transport under almost all conditions and net flux does not reverse under any drug gradient conditions (Fig. 3F and G; Fig. S7 and S9). This is consistent with harmane and 18-crown-6 ether not being transported, but instead triggering proton uniport down the twofold proton gradient (Fig. 3C). For the SSME experiments, we use substrate concentrations in a similar range to the concentration for which we observe phenotypes in bacteria. Unfortunately, in the case of streptomycin, this concentration range is below the threshold where we can detect signal with SSME. We therefore used NMR to confirm that streptomycin binds to PAsmr.

Ligand-detected NMR confirms streptomycin interaction

WaterLOGSY is a highly sensitive one-dimensional NMR technique that detects substrate binding under conditions where dissociation is fast (typically micromolar to millimolar affinity). In this experiment, water is excited and magnetization is observed on the free ligand. If magnetization transfer occurs from water directly to ligand, it will be of opposite sign compared with when magnetization transfer occurs from water to protein to ligand. This is because the sign of the NOE (operative mechanism of magnetization transfer) is highly dependent on the rotational correlation time of the molecule and has an opposite sign for small molecules versus macromolecules. Bound ligand behaves like a macromolecule for the purposes of NOE-based magnetization transfer, and NOE transfer is much more efficient in this bound state than for direct water to ligand transfer due to the relative dynamics of the components. Signal is detected on the free ligand, since the fast tumbling of the free ligand results in much greater sensitivity and resolution. As a consequence of these factors, conditions must be chosen where ligand is in exchange between bound (magnetization transfer) and free (detection) on a relatively rapid timescale in order for the experiment to work. Under these conditions and using the usual conventions for data processing and analysis, 1H spectra will show positive signal for free substrate (or substrate alone with no protein present), while substrate or substrate regions bound to protein will show a negative signal (38). To confirm the WaterLOGSY effect with PAsmr, we first tested binding of MeTPP⁺ at 45°C and found that MeTPP⁺ alone showed positive WaterLOGSY signal, but upon addition of PAsmr and lipids, MeTPP⁺ and lipids both showed negative signal consistent with binding to PAsmr (Fig. S10). Therefore, we used WaterLOGSY to confirm that streptomycin was interacting specifically with PAsmr. At 20°C, the streptomycin signal in the WaterLOGSY experiment is positive when mixed with lipids alone. Upon addition of PAsmr, several streptomycin peaks became negative, consistent with specific moieties of streptomycin interacting with PAsmr (Fig. 4). At lower temperatures, the full streptomycin spectrum became negative (Fig. S11). The different temperatures used for MeTPP⁺ and streptomycin WaterLOGSY experiments reflect the difference in affinity of these two ligands and the requirement to have an off-rate on a suitable timescale for WaterLOGSY detection. These NMR data confirm that PAsmr interacts specifically with streptomycin.

Fig 4.

Streptomycin has selective interactions with PAsmr. Streptomycin has characteristic chemical shifts (black) and a positive WaterLOGSY signal with lipid alone (orange). Addition of PAsmr results in a negative WaterLOGSY signal at some but not all characteristic shifts (asterisk), confirming specific interaction of PAsmr and streptomycin. Additional spectra are shown in Fig. S10 and S11 and available in the deposited data.

Substrate is required to trigger proton leak

There was slight proton leak observed through PAsmr liposomes with a proton gradient but no drug present (Fig. 3; Fig. S8 and S9), so to confirm that the increase in proton uniport observed in the SSME experiments was induced by drug and not simply due to potential inherent proton leak through PAsmr, we performed an additional voltage-driven proton leak assay as previously performed for EmrE (17). PAsmr was reconstituted in tight 3:1 POPC:POPG liposomes with a high internal potassium concentration at pH 7. These liposomes were then resuspended in a weakly buffered solution with a low potassium concentration and phenol red, which has pH-sensitive absorbance at 559 nm (Fig. 5A). Addition of valinomycin creates a negative inside membrane potential as potassium moves down its concentration gradient. If PAsmr leaks protons, this negative inside membrane potential would drive proton leak through this pathway, causing the external pH to rise (lower proton concentration) leading to a change in Abs₅₅₉ of phenol red. This did not occur (Fig. 5B). The external pH shifted only after addition of a protonophore (FCCP) that allows direct proton transport across the membrane. These data indicate that PAsmr does not leak protons under a voltage gradient and substrate is required to trigger proton uniport or leak (Fig. 5B; Fig. S12).

Fig 5.

PAsmr does not leak proton at neutral pH. (A) PAsmr was reconstituted into liposomes and resuspended with a 100-fold potassium gradient in neutral pH and phenol red. Addition of valinomycin dissipates the potassium gradient and establishes a negative inside membrane potential, favoring proton transport into the liposome if it is possible. (B) Phenol red absorbance at 559 nm reveals no change upon valinomycin addition (top), indicating that PAsmr does not leak proton under these conditions. Protonophore addition increases phenol red absorbance, and HCl addition decreases absorbance, confirming establishment of a membrane potential. Protonophore addition first also does not result in an increase of absorbance, indicating that both membrane potential and protonophore are necessary for proton transport through liposomes (bottom). Representative traces are shown, and additional replicates are provided in Fig. S12.

smrPA deletion phenotype in P. aeruginosa PA14 is substrate dependent

Finally, to test the function of PAsmr and impact of proton leak in the native organism, we created a deletion strain in PA14, a clinically derived virulent strain of Pseudomonas aeruginosa. We deleted the gene encoding PAsmr, smrPA, from the PA14 genome using sacB counterselection to generate an unmarked strain (Fig. 6A). We then compared growth of WT PA14 and the PA 14 ΔsmrPA in nutrient broth. There was no significant growth defect in the ΔsmrPA strain compared with the wild type when cultured in NB alone (Fig. 6B). We then tested several known and recently discovered PAsmr substrates in PA14 and the ∆smrPA strain. Li et al. previously demonstrated that deletion of smrPA (then known as EmrE_Pae) decreased resistance to ethidium bromide, acriflavine, and several aminoglycoside antibiotics in PAO1, a less virulent strain of P. aeruginosa (7). Deletion of smrPA did not result in growth differences in the presence of methyl viologen or acriflavine near their minimum inhibitory concentrations, consistent with the results of Li et al. (7). Deletion of smrPA from PA14 substantially increased susceptibility to ethidium bromide at the tested concentration (Fig. 6D), consistent with the published results (7). To confirm that smrPA expression was responsible for this phenotype, we reintroduced smrPA⁺ + the knockout strain under control of an IPTG-inducible promotor, P_trc. In the presence of 1 mM IPTG, resistance to ethidium was restored (Fig. 6C). The ∆smrPA strain also showed increased susceptibility to chelerythrine chloride (Fig. 6E), consistent with the resistance conferred by PAsmr in E. coli (Fig. 2C). This confirms that chelerythrine chloride is a substrate of PAsmr and the transport activity contributes measurably to resistance to this compound.

Fig 6.

P. aeruginosa ∆smrPA phenotype is substrate dependent. (A) Confirmation of the smrPA deletion in PA14 by PCR with flanking primers. (B) ∆smrPA does not impact growth in nutrient broth alone. (C) ∆smrPA with smrPA induced under a different promotor can confer resistance to ethidium bromide. (D) ∆smrPA is more susceptible to ethidium bromide. (E) ∆smrPA is more susceptible to chelerythrine chloride. (F, G) ∆smrPA has a smaller and more varied impact on growth in the presence of streptomycin.

18-crown-6 ether, harmane, and streptomycin trigger proton leak through PAsmr that is detrimental to bacterial growth in E. coli. However, 18-crown6 ether was not bactericidal to PA14 or PA14 ∆smrPA even at high concentrations (10 mM) and no difference in growth was observed between the two strains (Fig. S13F). Upon testing harmane and streptomycin, we observed small and somewhat variable differences between PA14 and PA14 ∆smrPA (Fig. 6F and G; Fig. S13A through D). For each of these substrates, the experiment was performed in triplicate on 3 different days (nine replicates total). For harmane, no significant difference was observed in two out three trials (Fig. S13A and B), and a resistance phenotype was observed in the third trial only after a very long lag in growth (Fig. S13C). The ∆smrPA strain showed increased susceptibility to harmane at 2 mM but not at lower concentrations (Fig. S13A through C). Streptomycin showed a small resistance phenotype in two out of three trials (Fig. 6F; Fig. S13D) and no significant phenotype in the third trial (Fig. 6G). We also tested tobramycin, a clinically relevant aminoglycoside, for comparison with streptomycin and saw no significant phenotype in PA14 (Fig. S13E). The small and less consistent phenotypes observed for harmane and 18-crown-6 ether in P. aeruginosa suggest that substrate-triggered proton leak through PAsmr has less of an impact on this organism. Thus, P. aeruginosa responds differently and may be better able to compensate for proton leak than E. coli.

DISCUSSION

Our model of SMR transporter function predicts that these efflux pumps act as double-edge swords for bacteria by conferring resistance to some compounds and susceptibility to others. This investigation serves as a starting point for the further characterization of PAsmr as well as the molecular determinants of substrate profile and phenotype for SMR transporters more broadly. While PAsmr was identified based on sequence homology to EmrE, there are limited data on its function and mechanism. The data presented here confirm that PAsmr transport activity confers resistance to the previously identified substrate, ethidium, and to chelerythrine chloride, which was previously not known to be a substrate of PAsmr. Currently known PAsmr substrates, phenotypes, and transport modes are summarized in Table 2.

TABLE 2.

Known PAsmr substrates and phenotypes

Substrate/compound	Interaction
Tetraphenylphosphonium	Inhibits methyl viologen transport in vitro (4)
Benzalkonium	Inhibits methyl viologen transport in vitro (4); PAsmr inhibition re-sensitizes PAsmr-E. coli (23)
Acriflavine	Inhibits methyl viologen transport in vitro (4); smrPA deletion from PAO1 decreases MIC (7)
Ethidium	Inhibits methyl viologen transport in vitro (4); transports/effluxes (23); smrPA deletion from PAO1 decreases MIC (7); smrPA deletion from PA14 decreases resistance; WT expression in E. coli confers resistance; transported in vitro via antiport
Methyl viologen	Transported in vitro (4), weak resistance phenotype (24)
Cetylpyridinium	PAsmr inhibition re-sensitizes PAsmr-E. coli (23), WT expression in E. coli confers resistance (24)
Cetyltrimethylammonium	PAsmr inhibition re-sensitizes PAsmr-E. coli (23)
Gentamicin	smrPA deletion from PAO1 decreases MIC (7)
Kanamycin	smrPA deletion from PAO1 decreases MIC (7)
Neomycin	smrPA deletion from PAO1 decreases MIC (7)
Nalidixic acid	smrPA deletion from PAO1 decreases MIC (7)
Chelerythrine	smrPA deletion from PA14 decreases resistance; WT expression in E. coli confers resistance; transported in vitro via antiport
Harmane	WT expression in E. coli confers susceptibility; triggers proton uniport
18-crown-6 ether	WT expression in E. coli confers susceptibility; triggers proton uniport
Streptomycin	WT expression in E. coli confers susceptibility; binds
Methyltriphenylphosphonium	WT expression in E. coli confers resistance; transported in vitro via antiport

Recently, we discovered that EmrE can confer resistance to some substrate and susceptibility to others by performing different types of coupled and uncoupled transport depending on the identity of the substrate (15). Proton-coupled antiport of toxic compounds leads to active efflux and confers resistance to those substrates, the well-known function of the SMR transporter family (Fig. 3A). Harmane triggers uncoupled proton uniport through EmrE, dissipating the proton gradient and negatively impacting ATP levels and growth in E. coli. The solid-supported membrane electrophysiology data in Fig. 3 show that both harmane and 18-crown-6 ether trigger proton leak through PAsmr, the first direct confirmation that drug-gated proton uniport occurs in an SMR transporter other than EmrE. When PAsmr is heterologously expressed in E. coli, this transport activity is similarly detrimental to bacterial growth and ATP levels—E. coli expressing non-functional transporter have higher ATP levels than cells expressing functional transporter in the presence of harmane and 18-crown-6 ether, and functional transporter also impairs growth.

Our observation that the aminoglycoside antibiotic streptomycin interacts with PAsmr supports the need for greater understanding of SMR transporters in clinically relevant organisms and raises further questions about the impact of proton leak. 18-crown-6 ether and harmane trigger proton leak, but the mechanism by which PAsmr induces susceptibility to streptomycin in E. coli remains unconfirmed, although streptomycin does bind to PAsmr (Fig. 4). While 18-crown-6 ether and harmane trigger a consistent decrease in ATP production, E. coli expressing PAsmr were able to eventually recover growth in the presence of streptomycin, leading to increased ATP production compared with control by the time they reached stationary phase (Fig. 2G). This may be explained by PMF-mediated uptake of aminoglycosides or the previous observation that streptomycin itself causes time-dependent changes in ATP levels due to its well-known function suppressing protein synthesis (35). Thus, an alternative possibility is that streptomycin-triggered proton leak through PAsmr reduces the ATP level initially, and then, this is overcome by the reduced use of ATP in protein production upon streptomycin targeting of the ribosome.

When moving to the native organism, P. aeruginosa, we observed significant differences in the consistency of PAsmr to confer resistance or susceptibility across different organisms. In the case of proton-coupled antiport-mediated resistance to toxic compounds, the function of PAsmr was consistent in E. coli and P. aeruginosa. While Li et al. demonstrated that smrPA deletion results in a reduced minimum inhibitory concentration to some substrates in PAO1, only traditional transport substrates and aminoglycosides were tested, and full growth curves were not examined (7). We demonstrated that deletion of smrPA from PA14, a more virulent model strain, resulted in the loss of resistance to ethidium bromide compared with the WT, which is surprising given the numerous additional efflux mechanisms present in this strain (39, 40).

However, the difference in growth between WT PA14 and PA14 ∆smrPA in the presence of substrates that trigger proton leak was more subtle and inconsistent. As PAsmr did confer resistance to ethidium and chelerythrine, the more subtle phenotype with these substrates is not due to the lack of PAsmr activity. Thus, the proton leak through PAsmr triggered by harmane and 18-crown-6 ether does not have a strong impact on growth of P. aeruginosa. This highlights important nuances in how proton leak impacts bacterial growth in different organisms and prompts further investigation about the greater ability of P. aeruginosa to compensate for or prevent loss of ∆pH due to transporter-mediated proton leak compared with E. coli. Using PAsmr to trigger proton leak provides a pathway to disrupt ∆pH specifically without more general membrane disruption, in contrast to small molecule PMF disruptors that act via the membrane, such as nigericin or CCCP. Thus, triggering proton leak through PAsmr provides a new route to probe the mechanisms by which P. aeruginosa responds to specific PMF disruption. Although the growth phenotypes were subtle and somewhat variable, the ability to induce proton leak through PAsmr may still provide a useful route to develop compounds that can synergize with other drugs that are dependent on the PMF or impacted by the metabolic response of P. aeruginosa to ∆pH dissipation. Understanding the response to PMF perturbation may reveal additional new targets for P. aeruginosa membrane energetics or compensatory responses to membrane perturbation. Regulation of efflux pumps is a key resistance mechanism in P. aeruginosa, and understanding the expression of transporters, including SMR transporters, is also an important future direction. Ultimately, understanding the organismal response to diverse transport behavior in complex resistance networks will aid future antibiotic development efforts.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/jb.00151-24.

Supplemental figures and table. jb.00151-24-s0001.pdf.

Fig. S1-13; Table S1.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.