18-Crown-6-ether Utilizes Distinct Allosteric Interactions to Uncouple Transport by the Multidrug Efflux Pump EmrE

Abstract

The recent discovery that the model multidrug efflux pump from Escherichia coli, EmrE, can perform multiple types of transport suggests that this may be a compelling target for therapeutic intervention. Initial studies have identified several small-molecule substrates capable of inducing transporter-dependent susceptibility rather than the well-known antibiotic resistance phenotype. However, many questions regarding the underlying mechanism and regulation of this transporter still remain. Prior studies identified lysine 22 as well as threonine 56 as important residues for regulating the formation of an occluded state critical to the prevention of an uncoupled leak in the WT transporter. Here, we use NMR chemical shift perturbations and in vivo EC₅₀ assays to confirm that 18-crown-6-ether binds at lysine 22, while liposomal leak assays verify that this substrate triggers uncoupled proton leak. In addition to characterizing the mechanism of action of another susceptibility substrate for EmrE, the characterization of K22 mutants herein solidifies the importance of this residue, as well as the nearby residue T56, in the allosteric regulation of the C-terminal tail. With a high degree of familial conservation in addition to a suggested role in transporter evolution, mechanistic insight into the transport regulation of EmrE may be broadly applicable across small multidrug-resistant efflux pumps.

Introduction

To address the growing antibiotic resistance crisis, a better understanding of antibiotic resistance mechanisms and new strategies to combat resistance are required. One mechanism of bacterial resistance is through the active efflux of antibiotics by multidrug efflux pumps. These transporters are highly promiscuous, conferring resistance to broad classes of substrates. The small multidrug resistance (SMR) transporters are among the smallest, exhibit particularly promiscuous substrate profiles, and are found throughout bacteria and archaea. − Despite being one of the better studied transporter families, the mechanism by which SMRs facilitate multidrug efflux remains incompletely characterized. − The archetype of the SMR family from Escherichia coli (E. coli), EmrE, confers resistance to a broad array of toxic polyaromatic compounds by secondary active transport. By coupling the favorable import of proton down the established gradient of the proton motive force (PMF) to the unfavorable export of the drug, these transporters actively efflux antibiotic and antiseptic compounds against a concentration gradient. , More recently, unbiased screens using the biolog phenotypic microarray to identify potential substrates of EmrE and other SMR transporters identified small molecules for which expression of functional SMR transporters enhanced bacterial susceptibility rather than conferring resistance. , However, the mechanism by which SMR transporters switch between antiport and other modes of transport that underlies susceptibility remains largely understudied.

The mechanism of action for one substrate to which EmrE confers susceptibility has been studied in more detail. Liposomal transport assays showed that the small-molecule harmane triggers uncoupled proton leak through EmrE, which sufficiently disrupts the proton motive force in bacteria such that E. coli expressing functional EmrE have reduced growth and metabolic output. However, it has not yet been experimentally established whether all substrates to which EmrE confers susceptibility act in the same manner to trigger uncoupled proton flux. Proton-coupled symport or uniport of a toxic substrate would also be detrimental to the bacteria. Here, we investigate the mechanism by which EmrE confers susceptibility to the second highest ranked susceptibility hit identified in the prior biolog phenotypic microarray, 18-crown-6-ether (18C6E).

Despite its small size, several regions of EmrE are already known to be functionally important. Two glutamates (E14^A/E14^B) in the functional dimer are the only protonatable residues within the transmembrane region of EmrE. This residue is fully conserved across the SMR family and is required for EmrE transport in liposomes and the ability of EmrE to confer resistance or susceptibility in bacteria. ,, The C-terminus of EmrE is also sensitive to the identity of the bound substrate and important for coupling drug binding to proton release, suggesting that the C-terminus may act as a secondary gate to prevent proton release in the absence of a drug. Deletion of four residues from the C-terminal tail of EmrE (Δ107-EmrE) results in increased uncoupled proton leak compared to WT-EmrE in the absence of exogenous substrates. Molecular dynamics simulations suggest that the C-terminal tail interacts with residues K22, T56, and D84 to stabilize the closed structure. Upon addition of harmane, uncoupled proton flux through WT-EmrE increases to equal the level of proton leak through Δ107-EmrE. This further supported the role of harmane as an allosteric regulator of the secondary gate, with a likely binding site near D84. Prior studies have shown that crown ethers can interact with proteins through coordination of positively charged lysine side chains (K-crown), by stacking with aromatic and hydrophobic amino acids or carboxylic and guanidium groups (C-crown), or by forming mixed interactions by binding lysine while surrounded by hydrophobic groups (KC-crown). Therefore, we hypothesized that 18C6E may induce susceptibility by interacting with K22 to destabilize the occluded state and induce EmrE-specific susceptibility through allosteric interactions distinct from harmane.

In the presence of 18C6E, E. coli expressing WT-EmrE have reduced growth compared to E. coli expressing a nonfunctional point mutant, E14Q-EmrE. This shows that the phenotype is EmrE-specific, occurring at concentrations well below that at which it acts as an ionophore , and requires E14, the critical residue for transport of protons or substrates across the membrane. Using chemical shift perturbations of the dynamically impeded mutant S64V-EmrE, we validated lysine 22 (K22), the only lysine residue in EmrE, as a potential binding site for 18C6E. Mutation of K22 to alanine abolishes the susceptibility phenotype in growth assays, supporting the importance of this residue for the 18C6E interaction with EmrE. In vitro solid-supported membrane electrophysiology and pyranine fluorescence-based transport assays show that 18C6E triggers an uncoupled proton leak through EmrE similar to harmane, although it binds in a different location. Finally, we use NMR and fluorescence polarization assays to characterize the dynamics and affinity of K22A-EmrE and K22R-EmrE to confirm the importance of residue 22 in regulating the secondary gate of EmrE.

Results

NMR Chemical Shift Perturbations and In Vivo Growth Assays Support a Putative Allosteric Site

To identify residues that coordinate the binding of 18C6E, we first acquired ¹H–¹⁵N-HSQC NMR spectra of S64V-EmrE in the absence and presence of 18C6E. S64V-EmrE has a reduced rate of alternating access but normal affinity for small-molecule substrates. This mutant has well-resolved peaks that enable calculation of chemical shift perturbations (CSPs) for most residues in the dimer upon addition of the drug (Figure S1). These experiments were performed at pH 5.8 where solid-state NMR structures have suggested that S64V-EmrE is likely in the occluded state and assignments were performed. Only a few residues show large CSPs, including K22_B, T56_A, and several residues in the C-terminal tail of monomer A (Figure A,B). We previously identified a role for T56 and K22 in coordinating the C-terminal gate of EmrE through hydrogen bonding and salt bridge formation, respectively. Thus, these chemical shift perturbations are consistent with the hypothesis that 18C6E binds to the only lysine in EmrE, K22, to unlock the C-terminal gate. To further test this, we generated mutants of K22 and transformed MG1655 ΔemrE E. coli to conduct growth assays and determine the EC₅₀s.

1.

Chemical shift perturbations and EC₅₀ curves identify K22 as a potential allosteric site for 18-crown-6-ether. Changes in the chemical shifts of individual residues in the ¹H–¹⁵N-HSQC spectra of S64V-EmrE in the absence and presence of 18C6E plotted on the structure (A) or by residue number (B). The single lysine K22 as well as T56 and residues in the tail have CSPs above the 3 standard deviation line. Relative growth of MG1655 ΔemrE expressing WT-EmrE, E14Q-EmrE, and K22A-EmrE (C) or K22R-EmrE (D) was measured after 18 h at increasing doses of 18C6E. Much like with E14Q-EmrE, which cannot transport protons, cells expressing K22A-EmrE do not show a dose-dependent loss of cell viability in the μM range of 18C6E. Cells expressing K22R-EmrE show a minor increase in potency relative to those expressing WT-EmrE. All cells show a nonspecific loss of viability in the low mM range of 18C6E likely due to the previously reported ionophore activity.

Overnight growth curves of E. coli expressing K22A-EmrE and K22R-EmrE showed that E. coli expressing K22A-EmrE grew similarly to bacteria expressing transport-incompetent E14Q-EmrE both in the presence and absence of crown ether, with minimal impact of 18C6E on growth. In contrast, E. coli expressing K22R-EmrE grew similarly to bacteria expressing WT-EmrE under both conditions with a similar reduction in growth in the presence of 18C6E (Figure S2). This would suggest that 18C6E is capable of binding to either a positively charged lysine or arginine residue at this position, as hypothesized, but 18C6E interaction with this residue is abolished when it is replaced with the small, hydrophobic alanine. This was further supported by dose–response curves of bacteria expressing WT-, E14Q-, K22A-, and K22R-EmrE (Figure C,D). While all of the constructs tested had an EC₅₀ in the low millimolar range, likely corresponding to nonspecific ionophore activity of 18C6E, only cells expressing WT-EmrE and K22R-EmrE had biphasic curves with a second EC₅₀ in the micromolar range (WT_EC50 = 120 ± 3 μM, K22R_EC50 = 88 ± 1 μM), consistent with an EmrE-specific activity of 18CE6 at this lower concentration (Figure C,D). While Mueller–Hinton broth is a standard medium for antibiotic susceptibility testing, it does have variable ion concentrations and may contain multiple ions and amino acids that can interact with 18C6E, such as potassium and arginine. The precise value of the EC₅₀ is likely to vary with the media and pH and would potentially be lower if competing ions and amino acids were not present in the media. The key result from these assays is that 18C6E interacts with EmrE specifically at concentrations below those where 18C6E exhibits general ionophore activity and that K22 is a critical residue for this EmrE-mediated susceptibility to 18C6E.

18-Crown-6-ether Triggers Drug-Gated Proton Uniport

Based on our previous finding that EmrE enhances bacterial susceptibility to harmane because harmane binding triggers uncoupled proton leak through EmrE, we hypothesized that 18C6E also triggers uncoupled proton leak. To test this hypothesis, we utilized in vitro solid-supported membrane electrophysiology (SSME) assays. These assays were previously developed to determine the dominant transport mode and stoichiometry of loosely coupled secondary active transporters like EmrE. ,, Using the underlying principles of reversal potentials, applying four separate gradient conditions (Figure A) allows for determination of the dominant transport mode of a given substrate based on the pattern of net flux that matches the expected outcome for a given transport mode (Figure B). Before assessing the dominant transport mode of 18C6E, we first verified proteoliposome integrity by assessing transport of methyl-triphenylphosphonium bromide (MeTPP⁺), a previously characterized substrate that EmrE antiports with a 2H⁺/1MeTPP⁺ stoichiometry, to confirm that K22A-EmrE and K22R-EmrE are active and able to perform proton/substrate antiport. ,

2.

SSME confirms transporter activity and the mode of action of 18C6E. Four standard gradient conditions have previously been shown to be sufficient to determine the dominant transport mode of different substrates with SSME according to the free exchange model (A). If liposomes and/or transporters are present and protons can pass through the liposome or they leak in an uncoupled fashion through the transporter, then there will be a low basal signal in the presence of an inwardly directed proton gradient in gray. When drug is added to both sides, the C-terminal gate will be unlocked, and there will be an increase in transport through EmrE constructs in black. If an outwardly directed drug gradient is paired with the inwardly directed proton gradient, then this will be favorable for antiport and will result in an increase in signal for antiported substrates (red). Finally, if an inwardly directed drug gradient is paired with the inwardly directed proton gradient, then this will be favorable for symport and will result in an increase in signal for symported substrates (navy). These gradients generate predictable patterns for net transport based on the dominant transport mode in the presence of a given substrate (B). The canonical antiported substrate MeTPP follows the expected pattern for antiport, with both WT- and K22A-EmrE showing the typical reversal under the navy condition as the stronger drug gradient reverses the direction of proton transport (C). The characteristic indifference to the direction of the drug gradients with 18C6E suggests that simply having sufficient drug around is enough to facilitate uncoupled proton transport in the direction of the proton gradient (D). This is the expected outcome for a substrate that triggers drug-gated proton leak as a mode of action.

In the presence of a 2-fold, inwardly directed proton gradient (Figure A,C gray), WT-EmrE, K22A-EmrE, and K22R-EmrE each have a low level of uncoupled proton transport in the absence of any other substrate. Upon addition of equal concentrations of MeTPP⁺ to both sides of the membrane, the inward proton gradient drives two protons into the liposome for every one MeTPP⁺ molecule transported out, generating a net +1 charge influx per transport cycle (Figure A,C black). As the drug can unlock the secondary gate and the rates of alternating access and drug/proton release are faster than proton leak in the absence of substrate, this results in increased transported charge relative to the proton-only condition for all three constructs. , Imposing a 16-fold outward drug gradient in addition to the 2-fold inward proton gradient (Figure A,C red) produces a condition where both the proton and drug gradients favor antiport. Since MeTPP⁺ is antiported by EmrE, this causes a further increase in transported charge for WT-EmrE, K22A-EmrE, and K22R-EmrE (Figure A,C red). Finally, reversing the direction of the drug gradient, such that it is now competing with the proton gradient to determine the direction of transport (Figure A,C navy), results in reversal of the direction of transport for WT-EmrE and K22A-EmrE as the large drug gradient surpasses the reversal potential for the 2H⁺:1MeTPP⁺ stoichiometric antiport with a 2-fold inward proton gradient (Figure A,C navy). While K22R-EmrE does not quite reverse under the 16-fold inward MeTPP⁺, 2-fold outward H⁺ gradient condition, examination of the integrated currents reveals that net transport is initially negative before switching back to positive, suggesting that this mutant may have an altered ability to regulate coupled transport or form the occluded state (Figure S3).

To determine the dominant transport mode in the presence of 18C6E, we applied the same magnitude drug gradients used to assess MeTPP⁺ transport, but with a higher average concentration as the EC₅₀ values indicate that 18C6E is less potent (Figure D and Figure S4). As was the case with harmane, we see that 18CE6 is required for significant net charge movement, but the direction of transport is the same regardless of the orientation of the drug gradient, indicating that 18C6E-gated proton leak is the dominant transport mode. The magnitude and direction of the transported charge are the same for spontaneous proton leak (gray) and with an outward drug gradient (red) for each of the variants. There is increased charge transport for the equal-drug (black) and inward-drug gradient (blue), and these two conditions also have matching flux for each variant. The lower flux is observed when there is no (gray) or 5 μM (red) 18C6E on the outside of the liposome, while the higher flux is observed when there is 80 μM 18C6E (black, navy) on the outside of the liposome. This higher concentration is the same order of magnitude as the EmrE-specific EC₅₀ for WT- and K22R-EmrE in cells (Figure D), suggesting that 18CE6 affinity is in this concentration range and increased binding of 16CE6 on the higher proton concentration side more effectively opens the gate and increases proton flux. However, it is not clear why this concentration of 18C6E triggers increased flux through the K22A-EmrE variant, which did not show an EmrE-specific functional impact in bacteria at micromolar 18CE6 concentrations (Figure D).

To directly confirm that 18C6E triggers drug-gated proton uniport and test the behavior of the different K22 variants, we performed a liposomal proton leak assay using the pH-sensitive dye pyranine. Each EmrE variant was reconstituted into proteoliposomes with pyranine encapsulated in the liposome interior at pH 7. The liposomes were then diluted 100-fold into a pH 7.5 buffer. If protons leak out of the liposome down the proton concentration gradient, the internal pH will rise and pyranine fluorescence will increase (Figure A). We compared the fluorescence, normalized to time zero (just after dilution), of proteoliposomes with those of WT-EmrE, K22A-EmrE, K22R-EmrE, and E14Q-EmrE (Figure B). We have previously shown that even under a 10-fold pH gradient, proton permeation through liposomes containing nonfunctional E14Q-EmrE is very slow, while uncoupled proton leak through WT-EmrE is significantly faster (Figure S5A). To minimize uncoupled proton leak, we use an ∼3-fold gradient in proton concentration and include a 2 min baseline to allow for comparison of the slope of the fluorescence change due to leak before and after addition of the drug. We then added 5 mM 18C6E to the outside of the proteoliposomes, a concentration above the EmrE-specific EC₅₀ and below the nonspecific EC₅₀ corresponding to general membrane permeation/ionophore activity by 18C6E (Figure S5B). WT- and K22R-EmrE proteoliposomes show an increase in internal pH over time (Figure B navy and teal), consistent with an 18C6E-triggered proton leak, while E14Q- and K22A-EmrE proteoliposomes (Figure B gray and gold) do not. This is consistent with the in vivo data and the hypothesis that 18C6E binds to a positively charged residue (K22 or K22R) and opens the C-terminal gate to allow for proton flux. Since these assays are conducted on very different timescales (ms vs minutes), the rates of alternating access and drug affinities will have a much greater effect on the SSME assays (Figure A). Therefore, to further investigate how the 18C6E gate protons leak through EmrE and why K22A-EmrE shows different behavior in the two in vitro functional assays, we next sought to determine the impact of 18C6E on EmrE dynamics since gating and alternating access are critical to transport.

3.

Pyranine fluorescence indicates that 18C6E uncouples proton uniport by interacting with a positively charged reside 22. Drug-gated proton uniport occurs when substrates bind allosterically to disrupt interactions with the C-terminus to increase the relative rates of proton binding/release and alternating access to facilitate uncoupled proton leak. When the fluorescent dye pyranine is preloaded in liposomes with an interior pH of 7, dilution into a pH 7.5 buffer creates an outwardly directed proton gradient, which will increase the internal pH if protons leak out without concomitant substrate antiport, thus increasing the fluorescence of the pH-sensitive pyranine (A). After a baseline of 2 min to demonstrate that protons do not leak through the liposomes, addition of 18C6E to the outside of the liposomes induces drug-gated proton uniport in WT- (B) and K22R-EmrE (E) containing proteoliposomes but not K22A- (D) or control E14Q-EmrE (C) proteoliposomes, which cannot bind protons to leak.

The Role of K22 in Regulating Alternating Access

To better understand the role of K22 in EmrE gating and transport, we examined how the dynamics of K22A-EmrE and K22R-EmrE compare to those of WT-EmrE in the absence of a drug. Due to the asymmetry of EmrE homodimers, the residues in each subunit have distinct chemical shifts. When alternating access occurs, the subunit in conformation A switches to conformation B and vice versa, leading to A/B exchange that can be detected by NMR. This transition reflects the global conformational change of the transporter alternating access from open to one side of the membrane to open to the other side of the membrane (as denoted by the vertical arrows in Figure A). If this alternating access happens slowly on the NMR timescale (ms–s), then both sets of peaks corresponding to conformations A and B will be separately resolved and visible in the 2D spectrum (Figure F, top). However, if alternating access is fast relative to the NMR timescale (μs–ms), then only a single set of peaks will be visible at positions that correspond to the average chemical shift of the two exchanging conformations (Figure F bottom). Finally, if the rate of alternating access is intermediate relative to the NMR timescale (low milliseconds), then significant exchange broadening of the peaks will occur such that many of the peaks are broadened into the noise (Figure F, middle).

4.

¹H–¹⁵N TROSY-HSQC spectra allow for comparison of protein dynamics between WT-, K22A-, and K22R-EmrE in the presence and absence of substrates. NMR dynamics experiments were performed at pH 8 where WT-EmrE is in slow-intermediate exchange and has higher spectral quality due to the relatively slow rate of alternating access. Protein motion impacts the NMR spectrum (F), with distinct patterns observed depending on the rate of alternating access compared with the chemical shift difference between the two structural states in the absence of exchange. If the rate of alternating access is slow compared to the chemical shift difference between conformations A and B, then two sets of peaks will be visible, corresponding to the two distinct conformations (slow exchange regime, top). If the rate of alternating access is very fast, then a single set of peaks will be detected, with a chemical shift corresponding to the population weighted average between conformations A and B (fast exchange, bottom). Finally, if the rate of alternating access is comparable to the chemical shift difference, then substantial line broadening will occur, and much of the signal will fall below the noise level, causing peaks to disappear (intermediate exchange, middle). Compared to WT-EmrE, in the absence of a drug, K22R-EmrE has a slower rate of alternating access (A) while K22A-EmrE has a faster rate of alternating access (B). When K22A-EmrE is saturated with TPP⁺, there are major improvements in spectral quality and the slower dynamics resulting in peak splitting confirming the poor quality of the apo spectrum is due to dynamics (C). Both mutants have faster rates of alternating access upon addition of saturating 18C6E, while also displaying chemical shifts, suggesting that both can bind 18C6E at saturating concentrations (D, E).

K22R-EmrE displays slightly slower alternating access than WT-EmrE (Figure A), as evident by the slightly sharper (narrower) peaks as well as the increase in the number of resolved peaks. In contrast, K22A-EmrE appears to display a significantly faster rate of alternating access, as there are far fewer peaks and considerable line broadening in the spectrum (Figure B). To confirm that this poor spectral quality is due to dynamics and not instability or degradation of the sample, a portion of the drug-free K22A-EmrE was saturated with the substrate tetraphenylphosphonium bromide (TPP⁺). The addition of TPP⁺ dramatically reduces the rate of alternating access for WT-EmrE, and the same pattern is seen for K22A-EmrE (Figure C), with significant improvement in spectral quality and clear resolution of two sets of peaks indicative of slow exchange alternating access. This confirms that the poor spectral quality of the drug-free K22A-EmrE spectrum was indeed due to motion. These data support a role for K22 in regulating the rate of alternating access in EmrE, with the conservative K22R mutation retaining alternating access behavior similar to that of WT, while K22A-EmrE has much faster exchange.

We then compared the effect of saturating 18C6E on K22A-EmrE and K22R-EmrE (Figure D,E) to understand how 18C6E interacts with these variants and hopefully gain insight into why K22A-EmrE shows different functional behavior in the two proteoliposome transport assays (Figures and ). Both K22A-EmrE and K22R-EmrE display significant chemical shift perturbations upon addition of 18C6E, indicating binding of the ligand and changes in the overall structure or motion. K22R-EmrE shows severe line broadening, consistent with intermediate exchange or exchange of multiple different conformational states. K22A shows an increase in the number of visible peaks, suggesting a slowing of the dynamics, although the lines are still very broad. These data indicate that both constructs are able to bind to 18C6E, consistent with the SSME data, but 18C6E has profoundly different impacts on transporter dynamics in each variant. The high concentration of 18C6E in the NMR samples for saturation may also drive additional binding to other positively charged residues in EmrE. While K22 is the only lysine in EmrE, each EmrE monomer contains three arginine residues in the loops and tail that could be accessible to 18C6E, R29, R82, and R106, although nonspecific K-crown interactions of crown ether with arginine side chains or amine termini are typically only seen in short peptides.

Binding of 18C6E Affects Transporter Dynamics through Functionally Relevant and Nonspecific Interactions

To directly detect 18C6E binding to K22 and its effect on transporter motion, we generated a WT-EmrE sample with uniform ¹⁵N labeling and ²H¹³C¹⁵N-labeled lysine and performed an NMR titration with 18C6E. Since K22 is the only lysine in EmrE, this allowed us to specifically assess 18C6E binding at K22 while also monitoring global changes in the protein. The individual carbon atoms in the K22 side chain show chemical shift perturbations in the low millimolar range (Figure B). This is consistent with a high micromolar binding affinity for 18C6E, as suggested by 100 μM EC₅₀, given the high protein concentration in the NMR sample (0.6 mM). Much larger chemical shift changes occur above 50 mM 18C6E, suggesting that full saturation with 18C6E causes more profound changes in the transporter structure that likely do not reflect the physiologically relevant effect of the drug. The ϵ-amino group in the lysine side chain (Figure B) is only detectable in SOFAST ¹⁵N-HMQC spectra above 50 mM 18C6E, indicating that full saturation with 18C6E reduces the dynamics of the highly mobile K22 side chain (Figure C). However, the lipid peaks (*) in the ¹H,¹³C-HSQC NMR spectra do not show any chemical shift perturbation, providing direct evidence that even very high 18C6E concentrations do not significantly perturb the bicelle structure.

5.

NMR supports that 18C6E binding reduces dynamics through both specific and nonspecific binding. ¹³C-HSQC experiments (A) and SOFAST ¹⁵N-HMQC experiments (B) monitoring the side chain of lysine 22 display large chemical shifts and dynamic changes as 18C6E reaches saturation. TROSY-HSQC spectra of the ¹H¹⁵N¹³C-lysine-¹H¹⁵N-EmrE backbone suggest that alternating access dynamics initially decrease from fast-intermediate to intermediate when 18C6E is added up to 5 mM (C). Titrating additional 18C6E decreases alternating access dynamics further into slow exchange on the NMR timescale (D). These data suggest that binding occurs in two concentration regimes to induce dynamics changes through specific binding and, upon saturation, nonspecific binding.

At nearly physiological pH (pH 7.4), WT-EmrE is in fast-intermediate exchange in the absence of a substrate, with a single set of relatively broad peaks. As 18CE6 is titrated into the EmrE sample, the peaks shift and further broaden and then split into two distinct sets of peaks that sharpen as the concentration increases, reflecting a transition to slow exchange upon saturation with 18C6E (Figure F, top). The rate of alternating access of WT-EmrE varies over several orders or magnitude depending on which substrate is bound and shifts across the full dynamics range as a function of pH in the absence of a substrate. Thus, the change in dynamics with 18CE6 is within the range of that previously observed. Interestingly, this impact on transporter motion begins to occur at concentrations below 5 mM, indicating that relatively low concentrations of 18C6E are sufficient to open the gate and allow transport. This is consistent with the flux observed in the liposomal assays at micromolar concentrations (Figures and ) and the 100 μM EC₅₀ (Figure ). The high millimolar concentrations needed to induce global reduction in protein motion and detect the K22 side chain are above the concentration at which 18C6E acts as an ionophore and suggest binding of 18C6E to additional sites or effects of 18C6E on the bicelle structure occur at these concentrations that are not physiologically relevant.

Developing New Platforms for Assessing Affinity of Allosteric Substrates

To better quantify the affinity of 18C6E for multiple likely binding sites on EmrE, it was necessary to expand the available tools. The challenge lies in the wide range of affinities of different substrates for EmrE and for individual substrates binding at multiple locations, combined with the need to solubilize this integral membrane protein in a stable functional form. We therefore optimized reconstitution of EmrE into Saposin A (SapA) nanoparticles, which are highly stable and amenable to reconstitution with a variety of lipids and membrane protein. − We reconstituted EmrE with the variable EmrE:SapA:DMPC to optimize reconstitution efficiency (Figure A, left). The resulting particles were then subjected to additional size-exclusion chromatography to remove excess lipid and cargo-free nanoparticles, and the conditions for reconstitution B were found to produce the highest reconstitution efficiency (Figure A, right, Figure S7). To ensure that the functional dimer was reconstituted into the SapA nanoparticles, we recorded the intrinsic tryptophan fluorescence of EmrE in SapA nanoparticles in the absence and presence of 500 μM harmane (Figure B, Figure S7). SapA has only a single tryptophan, compared to eight tryptophans in the EmrE homodimer. Quenching of intrinsic tryptophan fluorescence has been widely used to study substrate binding to EmrE and is primarily due to quenching of W63 A/B in the primary binding site near E14. ,, A significant reduction in the intrinsic tryptophan fluorescence of EmrE in SapA nanoparticles occurs upon incubation with harmane, suggesting that functional EmrE homodimers are present and the majority of the tryptophan fluorescence arises from EmrE.

6.

Characterization of SapA-EmrE nanoparticles. Gel electrophoresis of various reconstitution ratios before and after size-exclusion chromatography shows isolation of pure Saposin:EmrE particles (A). Intrinsic tryptophan fluorescence of the sample with the highest reconstitution efficiency before and after addition of harmane suggests that a function dimer is present, and the majority of the fluorescence corresponds to EmrE as expected (B). Raw native mass spectra for Saposin A particles carrying the EmrE dimer were collected during a collision energy ramp from 0 to 200 V (C). The total normalized mass defect of Saposin A particles carrying the EmrE dimer collected during a similar collision energy ramp is shown as a function of mass (D). The stoichiometry of Saposin A is more polydisperse in this case and varies between 2–10 in filled particles and 2–4 in empty particles. Mass defect analysis of EmrE:Saposin A particles at the beginning of the ramp at 0 V shows similar polydisperse complexes (E). The transient nature of some of these complexes is revealed in (F) where the mass defects at higher collision voltages (150 V) show the loss of lipids, loss of Saposin A dimer units, and removal of the empty particles without the EmrE dimer.

Unlike the stoichiometry of membrane scaffold proteins in nanodiscs, the stoichiometry of Saposin A in lipid nanoparticles is more polydisperse and dependent on the membrane protein in the complex. , We first characterized empty particles with only lipid and SapA, which predominantly contained four SapA proteins (Figure S6). EmrE:SapA nanoparticles had a higher polydispersity in mass distribution and mass defects (Figure C,D). While all of the particles contained EmrE homodimers, the difference in mass can be attributed to the EmrE dimer displacing some lipids from the SapA nanoparticles. At the lowest collision settings (Figure E), we see prominent signals of 10-, 8-, 6-, 4- or 2-SapA nanoparticles, each of them carrying an EmrE dimer. We also see a small population of empty SapA nanoparticles carrying 2- or 4-SapA per particle. At the highest collision energy setting (Figure F), most of these particles have decayed, but some survive with their cargo, losing only lipids or SapA dimers. Thus, EmrE was incorporated as a dimer into the SapA nanoparticles, but the particle size and number of SapA per particle were variable. This heterogeneity in the number of SapA is consistent with previous findings; however, in light of the minimal contribution of Saposin A to the molar absorptivity of the sample and the relative prevalence and stability of the 6 SapA:2 EmrE discs, an average of 6 SapA per EmrE dimer was used to determine the concentration of the EmrE dimer for further studies.

Nanoparticles Allow for Higher Sensitivity in Characterizing Secondary Binders and Nonspecific Interactions

Quenching of intrinsic tryptophan fluorescence has been used extensively to characterize substrate binding to EmrE. The intrinsic tryptophan fluorescence observed for EmrE is primarily due to W63 in the hydrophobic core of the transporter near the primary binding site at E14. Substrate binding in this site causes significant quenching of this fluorescence, and dimerization is required for substrate binding with proper affinity. Utilizing the nanodiscs for intrinsic tryptophan fluorescence, the K _D values are 73 ± 15 μM for methyl viologen (Figure A) and 46 ± 7 μM for harmane (Figure B). The K _D for methyl viologen has not been reported previously, although this substrate is widely used for transport and resistance assays. − The K _D for harmane measured here is quite similar to the previously reported value of 19 ± 2 μM measured using EmrE in isotropic bicelles, confirming that EmrE is properly folded and dimerized in the nanoparticles. Unlike bicelles, the SapA nanoparticles are not in equilibrium with monomeric short-chain lipids and can be diluted to lower concentrations without jeopardizing their stability. This results in greater sensitivity in the fluorescence assay and the ability to use a 10-fold less EmrE protomer (Figure A–C). Harmane, a noncanonical substrate that triggers uncoupled proton leak, was recently crystallized in the primary binding pocket of EmrE and causes nearly complete quenching of W63 fluorescence (Figure B), indicating that this substrate can advance into the primary binding pocket even though antiport is not the dominant mode of transport. Upon addition of 18C6E, the intrinsic tryptophan fluorescence does decrease, but full quenching is not observed, and a stable baseline is not reached even at very high concentrations (10 mM, Figure C). This suggests that the K _D for the primary site is much lower or nonspecific and binding to K22 in the periphery away from W63 and E14 does not alter the fluorescence profile, as expected.

7.

Binding assays with EmrE:Saposin A nanoparticles confirm dimer activity and nanoparticle competency. Intrinsic tryptophan fluorescence assays with WT-EmrE nanoparticles produce similar magnitude K _D’s for a canonical resistance substrate, methyl viologen (A), and the previously characterized susceptibility substrate, harmane (B). Binding of 18C6E to K22R-EmrE nanoparticles does not fully saturate W63 up to the low mM range, further supporting allosteric binding as the phenotypically relevant interaction (C). Fluorescence polarization assays with synthesized BODIPY-18C6E show time-dependent differences in affinity and saturation with WT-EmrE (D) and K22R-EmrE (F) nanoparticles, while K22A-EmrE nanoparticles show only the less specific interaction (E).

To further probe the complexity of 18C6E binding to EmrE, we generated BODIPY-labeled 18C6E (benzo[18]­crown-6 BODIPY) as previously described. We then performed time course fluorescence polarization experiments, titrating WT-EmrE, K22A-EmrE, K22R-EmrE, or empty SapA nanodiscs into 50 nM BODIPY-16C6E by 5-fold serial dilutions. This assay allows us to probe interactions at very low concentrations, extending the range of binding affinity. Interestingly, interaction with K22A-EmrE is detected with a K _D of 154 ± 69 nM. This simple binding event is stable over 30 min (Figure D–F) and requires the presence of EmrE (Figure S8). In contrast, WT- (K _D = 97.4 ± 57 nM) and K22R-EmrE (K _D = 226 ± 106 nM) both show a slight left shift after 30 min and a further change in polarization suggesting that a secondary binding interaction may occur (Figure D,F). While these affinities are tighter than the concentration ranges over which 18C6E impacts EmrE function in vivo and in vitro, it highlights the need for a positive residue at position 22 to detect biphasic binding behavior.

The ability of 18C6E to bind to both lysine and arginine, as well as form more complex C-crown and KC-crown interactions with hydrophobic residues, as are found in the core of EmrE, makes it difficult to determine the absolute K _D of 18C6E binding at K22. Additionally, the need for millimolar concentrations of protein for NMR makes it difficult to assess the differences between binding of 18C6E that impacts the local chemical environment versus changing the dynamics of the protein. These changes in the global dynamics of EmrE upon binding of the substrate in the core are well-documented and have been recorded for substrates with affinities varying by several orders of magnitude. Additionally, competition between proton and drug for binding in the core renders the binding affinity pH-dependent. ,, As such, while delineating the functionally relevant interactions of 18C6E with EmrE across assays and mutants is necessarily complex, it is clear that there is a higher affinity interaction when EmrE is present, and this is of greater significance when a positive residue is present at position 22. These interactions of 18C6E at K/R22 coupled with the significant changes in the dynamics and transport behavior of K22 mutants overall support the critical role of K22 in regulating the secondary gate of EmrE.

Discussion

Transporters perform essential functions, moving small molecules and ions across the membrane barrier separating the cell from the external environment to import nutrients, export waste and other toxins, maintain cellular homeostasis, and communicate. Transporters facilitate this movement across the membrane through diverse mechanisms, including facilitated diffusion (uniport), ion-coupled active transport, or ATP-driven active transport. Traditionally, transporters have been classified based on their activity under the assumption that these protein machines have evolved to perform a specific biological function. As more extensive, detailed, and quantitative studies of transport have been completed, it has become apparent that some ion-coupled transporters can also perform uncoupled transport of substrate or coupling ion to the extent that the dominant transport activity changes depending on the substrate or environmental conditions. How a single transporter can switch between different transport modes is not fully understood. The quaternary ammonium cation (QAC) subfamily of small multidrug resistance (SMR) is particularly interesting because these promiscuous transporters switch between dominant transport modes that result in different biological outcomes: they confer resistance to some substrates while enhancing bacterial susceptibility to others.

As one of the smallest and best studied multidrug resistance transporters, EmrE is often regarded as a model for studying the minimal requirements of coupled transport. However, EmrE has proven to be surprisingly complex, revealing unexpected features of membrane protein topology and transport mechanism. It is becoming increasingly clear that a universal exchange model best describes EmrE activity. The free exchange model previously reported includes all of the states and transitions observed by NMR and can account for the ability of EmrE to confer resistance to some substrates and susceptibility to other substrates; − however, it also predicts a fast, uncoupled proton leak through the transporter that is only observable in SSME assays when a large number of liposomes are concentrated on the surface. , In light of our studies with the small molecule harmane and now 18C6E, the susceptibility phenotypes we have seen thus far can be better characterized as a drug-gated proton leak. This susceptibility is of particular interest as it provides a novel avenue of antibiotic development leveraging bacteria-specific mechanisms and can even synergize with existing antibiotic entities. However, in order for antibiotic and adjuvant development to directly target SMRs to gain traction, a better understanding of universal exchange transport and the features that regulate coupled versus uncoupled transport as the dominant transport mode are crucial.

Recent investigations of a C-terminal deletion mutant of EmrE, Δ107-EmrE, confirmed that this mutant rapidly leaks protons, agreeing with a secondary gating model where the C-terminal tail of EmrE prevents uncoupled proton leak in the WT transporter and small molecules can disrupt regulatory interactions with the tail to alter the mode of transport. MD simulations also identified a mechanism for this leak, revealing that truncation of the C-terminal tail facilitated formation of a water wire connecting the primary binding site for protons, E14, to bulk water on the open side allowing protons to hop along the wire and lowering the free energy barrier relative to WT-EmrE. These simulations also identified key interactions in the WT transporter, which coordinate the formation of an occluded state, which breaks the water wire. These interactions included a salt bridge between D84 and R106, as well as the carbonyl group of the C-terminus forming hydrogen bonds with T56 and occasionally a salt bridge with K22. Prior CSP analysis upon binding of harmane to E14Q-EmrE identified that relatively large shifts near D84 as well as R106 supporting secondary binding near this site can disrupt formation of the occluded state. In contrast, here, we see that with 18C6E binding, there are CSPs at K22, T56, and within the tail, suggesting an allosteric mechanism distinct from harmane (Figure A,B).

It is clear from both the pyranine fluorescence and SSME experiments that 18C6E can induce substrate-gated proton leak (Figures and ) and that this is distinct from the ionophore effect seen at high concentrations (Figure C,D). The dramatic changes in transporter dynamics (alternating access rates) upon either mutation of K22 or addition of 18C6E further signify the importance of the previously identified hydrogen bonding network between the C-terminal tail and K22/T56 as a secondary regulatory mechanism for EmrE (Figure ). The development of protocols for Saposin A nanoparticle reconstitution and applications with binding assays that allow for higher sensitivity and characterization of compounds with a wider range of affinities and chemical characteristics also expands upon the available tools to assess the substrate profile of EmrE. Due to the agnostic nature of Saposin A to the cargo protein of interest, the reconstitution and binding assays described here could greatly facilitate further development of high-throughput assays to identify novel chemical matter capable of inducing susceptibility phenotypes in SMRs and other transporters. Altogether, the work presented here greatly expands upon the understanding of the interactions governing transport regulation through EmrE and how some substrates can bind allosterically to disrupt these interactions. Additionally, we expand upon the current knowledge and available tools to assess the effects of noncanonical substrates across a wider range of affinities.

Conclusions

As an archetype for the small multidrug-resistant family of transporters, EmrE is among the better studied multidrug efflux pumps. Here, we show evidence that a highly conserved residue in the first loop, lysine 22, is critical for maintaining coupling in EmrE. Mutation of this site or binding of 18C6E disrupts interaction of this loop with the C-terminal tail, which is important for secondary gating. Through this mechanism, 18C6E binding induces uncoupled proton leak, which disrupts the proton motive force and enhances bacterial susceptibility to 18C6E. This is the second allosteric regulator of EmrE to induce uncoupled proton leak, and it does so through a unique set of interactions distinct from those of the previously identified molecule, harmane. However, both harmane and 18C6E disrupt the C-terminal tail, confirming the importance of the C-terminal tail in maintaining proton coupling in EmrE.

Materials and Methods

Plasmids and Strains

All in vivo experiments were performed in MG1655 ΔemrE E. coli cells (item number JW0531-2, E. coli Genetic Resource (CGSC), Yale) transformed with a low copy number plasmid under a pTrc promoter. In vivo experiments relied on constitutive expression of these plasmids, and expression levels were previously validated by Western blot analysis. Protein expression utilized BL21 (Gold) DE3 E. coli transformed with a pET15b plasmid containing the respective EmrE construct or SHuffle cells (New England Biolabs) transformed with pET15b containing Saposin A.

Microplate Growth Assays

Cells expressing plasmids of interest were grown in Mueller–Hinton broth (Sigma, 100 μg/mL ampicillin, pH 7.0) from a single colony to an OD of 0.2–0.4 at 37 °C. The cells were then diluted to a final OD of 0.01 in 96-well microplates containing a concentration range of 18-crown-6-ether. The plates were incubated and shaken in a microplate reader (BMG-Labtech or Tecan Spark) at 37 °C. OD₆₀₀ was measured every 5 min for 18 h. Experiments were performed with two biological replicates each containing three technical replicates. Error bars represent the standard deviation from the mean of the six total replicates.

EC₅₀ Value Determination

MG1655 ΔemrE E. coli cells expressing either WT-EmrE, E14Q-EmrE, K22A-EmrE, or K22R-EmrE were grown to an OD of 0.2–0.4 at 37 °C from a single colony. A concentration range of 18-crown-6-ether (0–50 mM) was serially diluted in microplates with a starting OD₆₀₀ of 0.01. Plates were then incubated with shaking for 18 h at 37 °C. OD₆₀₀ end points were taken using a BMG (Labtech) plate reader. Relative growth was calculated by dividing the measured OD₆₀₀ from a given concentration by the OD₆₀₀ for cells containing no drugs. Two biological replicates were performed in triplicate (n = 6) and fit to a biphasic dose–response curve using GraphPad (Prism).

EmrE Expression and Purification

BL21 gold (DE3) E. coli cells were transformed with pET15b-WT-EmrE, pET15b-K22A-EmrE, pET15b-K22R-EmrE, pET15b-S64V-EmrE, pET15b-W31Y-S64V-EmrE, pET15b-W45Y-S64V-EmrE, or pET15b-S64V-W76Y-EmrE plasmids and grown in M9-H₂O minimal media with 4 g/L glucose unless otherwise specified to an OD₆₀₀ of 0.9. The bacteria were flash-cooled and then induced with 0.33 M IPTG overnight at 17 °C. The E. coli cells were collected with centrifugation and lysed, the membrane fraction solubilized with decylmaltoside (DM) or dodecylmaltoside (DDM), and the proteins purified using nickel affinity chromatography followed by size-exclusion chromatography (SEC) on a Superdex 200 column as previously described. Protein concentrations were determined using absorbance at 280 nm with an extinction coefficient of 38,400 L/mol cm.

Chemical Shift Perturbation Mapping

¹⁵N²H-S64V-EmrE was expressed in M9-D₂O with 2.5 g of ¹²C-d ₇-glucose and back-exchanged during purification in Milli-Q buffers before reconstitution into isotropic bicelles (q = 0.33) at pH 5.8 (0.7–1 mM monomer concentration). The 18-crown-6-ether-bound EmrE sample was soaked in 18-crown-6-ether overnight with incubation at 45 °C. Excess solid was removed by centrifugation at 16,000g for 10 min. HN-Transverse relaxation optimized spectroscopy–heteronuclear single quantum correlation (HN-TROSY-HSQC) experiments were performed on a 900 MHz Bruker Avance spectrometer at 45 °C (d ₁ = 2 s). Spectra were processed and analyzed using NMRPipe and CCPnmr Analysis 3.0.4. Assignments were transferred from an S64V-EmrE backbone walk available on BMRB. CSPs were calculated according to Δω = ω_drug‑free – ω_drug‑bound.

Direct Binding by NMR Spectroscopy

¹³C¹⁵N²H-Lys-¹⁵N-EmrE was expressed in M9-H₂O with 30 mg/L ¹³C¹⁵N²H-Lys. ¹⁵N-K22A-EmrE and ¹⁵N-K22R-EmrE were expressed in M9-H₂O with 1 g of ¹⁵NH₄Cl. ¹⁹F-WT-EmrE, ¹⁹F-S64V-EmrE, ¹⁹F-W31Y-S64V-EmrE, ¹⁹F-W45Y-S64V-EmrE, and ¹⁹F-S64V-W76Y-EmrE were expressed in M9-H₂O with 60 mg/L 5-fluoroindole. All constructs were purified as above (0.7–1 mM final monomer concentration) and were reconstituted into isotropic bicelles (q = 0.33) at pH 8. The 18-crown-6-ether-bound K22A- and K22R-EmrE samples were soaked in drug overnight with incubation at 45 °C. Excess solid was removed by centrifugation at 16,000g for 10 min. HN-Transverse relaxation optimized spectroscopy–heteronuclear single quantum correlation (HN-TROSY-HSQC) experiments, HC-TROSY-HSQC experiments, and SOFAST HN-HMQC were performed on an 800 MHz Varian spectrometer or a 900 MHz Bruker Avance spectrometer at 45 °C (d ₁ = 2 s). ¹⁹F experiments were performed on a 600 MHz Bruker Avance spectrometer at 45 °C. Spectra were processed and analyzed using NMRPipe and CCPnmr Analysis 3.0.4.

Pyranine Fluorescence Assays

All data were acquired on a Tecan Spark instrument. The excitation wavelength was 465 nm (35 nm bandwidth), and the emission wavelength was 530 (25 nm bandwidth). The excitation spectrum maximum of pyranine shifts from 400 to 450 nm as pH increases, so with a constant 465 nm excitation wavelength, the observed fluorescence signal will increase as pH increases. The number of flashes was set to 30 to reduce the well to well measurement time. To minimize instrument integration time, replicates were allowed to equilibrate for the full 30 min, and an average of the Z-position and gain recorded by the instrument were used as manual input for the reported assays. Liposome stocks with an internal buffer concentration of 100 mM MOPS, 20 mM NaCl, and 1 mM pyranine, pH 7, were dispensed with 2 μL per well, and 197 μL of 100 mM MOPS and 20 mM NaCl (pH 7 or pH 7.5) buffer was rapidly dispensed into the well containing the liposomes recording commenced as soon as possible. After a 2 min equilibration to differentiate leak inherent to the constructs in the presence or absence of a gradient, 1 μL of 18-crown-6-ether was dispensed and mixed three times with a second, larger volume multichannel pipet. Reported data are average values of three replicate wells recorded for 15 min each to minimize well to well measuring times, with error bars representing the standard deviation of the mean.

Solid-Supported Membrane Electrophysiology Transport Assays

WT-, K22A-, and K22R-EmrE were expressed and purified, with the final SEC performed in assay buffer (50 mM MES, 50 mM MOPS, 50 mM bicine, 100 mM NaCl, 2 mM MgCl₂, and 40 mM DM, pH 7). All buffers were carefully adjusted to the desired pH exclusively with NaOH to ensure consistent Cl^– concentrations across the membrane for transport assays. Protein was reconstituted into 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) proteoliposomes at a lipid-to-protein ratio of 1:400 in pH 7 assay buffer. Briefly, 15 mg/mL stocks of POPC were diluted in assay buffer and incubated at 45 °C for 1 h. Lipids were bath sonicated for 1 min, and then, octyl glucoside (OG) was added to a final concentration of 0.5%. Lipids were sonicated for an additional 30 s and returned to 45 °C to incubate for 15 min. SEC fractions containing purified protein in DM were added to the lipid solution and incubated at RT for 25 min, and then, detergent was removed with biobeads (Biorad) as previously described. As a negative control, POPC lipids were put through a simulated reconstitution process without protein. Amberlite was removed from each sample via a gravity column, and uniform liposomes were obtained by extrusion through a 0.2 μM membrane using an Avanti miniextruder.

All SSME data were acquired by using a Nanion SURFE2R N1 instrument. Liposome aliquots were thawed, diluted 2-fold, and briefly sonicated. 10 μL of liposomes was used to prepare 3 mm sensors as previously described. Before experiments, sensor capacitance and conductance values were obtained to ensure sensor quality. Based on manufacturer recommendations, sensors used were limited to those with a capacitance of 15–40 nF and a conductance <10 nS. Sensors with capacitance ranges closer to ∼20 nF and a conductance of 1–2 nS were preferred for internal consistency.

For all experiments, buffers contained 50 mM MES, 50 mM MOPS, 50 mM bicine, 100 mM NaCl, and 2 mM MgCl₂ with an internal pH value of 7.0 and an external pH value of 6.7. For inward-facing drug gradients, the external drug concentration was 10 μM for MeTPP or 80 μM 18C6E and the internal drug concentration was 0.625 μM MeTPP or 5 μM 18C6E. For outward-facing drug gradients, the internal drug concentration was 10 μM MeTPP or 80 μM 18C6E and the external drug concentration was 0.625 μM MeTPP or 5 μM 18C6E. Both internal and external drug concentrations were 10 μM MeTPP or 80 μM 18C6E for the zero-gradient data. 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. Measurements were performed at a flow rate of 200 μL s^–1. 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. Reported data are average values of at least three sensors with error bars representing the standard deviation of the mean.

Saposin A Expression and Purification

Plasmid containing codon-optimized Saposin A was transformed into SHuffle (New England Biolabs, Inc.) Escherichia coli cells. Cells were grown at 37 °C in Terrific broth until the samples reached an OD600 of 0.4–0.8 at which point cells were induced with 1 mM isopropyl β-d-1-thiogalactopyranoside. Cells were grown overnight at 16 °C before being harvested by centrifugation at 5000 rpm for 10 min. For purification, cells were resuspended in Ni affinity wash buffer [25 mM Tris-HCl (pH 7.5) and 150 mM NaCl] and subjected to lysis by sonication. Lysates were centrifuged at 10,000 rpm for 10 min, and the supernatants were collected. Isolated supernatants were heated to 85 °C for 10 min prior to an additional centrifugation at 10,000 rpm for 45 min. Samples were then purified using nickel beads followed by removal of thrombin and size-exclusion chromatography (SEC) on a Superdex 200 column as previously described for EmrE. SEC buffers consisted of either 25 mM Tris-HCl (pH 7.5), 150 mM NaCl buffer for direct reconstitution (intrinsic tryptophan fluorescence), or 1× phosphate-buffered saline (PBS, 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, and 1.8 mM KH₂PO₄, pH 7.4).

Saposin A Nanodisc Reconstitution

1,2-Dimyristoyl-sn-glycero-3-phosphocholine was resuspended in 100 mM MOPS and 20 mM NaCl, pH 7, to 20 mg/mL and incubated at 45 °C for 1 h. Lipids were then sonicated for 1 min, and 0.5% w/v octyl glucoside was added. Lipid was then sonicated for an additional 30 s before incubating at 45 °C for an additional 15 min. DMPC was then added to purified EmrE in a 25:1 lipid:protein ratio with Saposin A at a 10:1 SapA:EmrE ratio (50 μM EmrE, 1.25 mM DMPC, and 500 μM SapA) and allowed to incubate at RT for 20 min. A 90-fold excess of biobeads was then added to remove detergent and incubated while rocking for 3 h. Biobeads were then drained by gravity columns, and excess Saposin A, lipid, and remaining PBS buffering components were removed by SEC in 100 mM MOPS and 20 mM NaCl, pH 7.

Intrinsic Tryptophan Fluorescence

Purified WT-, K22A-, and K22R-EmrE were reconstituted Saposin nanodisc buffer containing 100 mM MOPS at pH 7.0 and 20 mM NaCl. Harmane and methyl viologen were prepared from DMSO stocks at 50 mM and then serially diluted into black 96-well flat-bottom plates in assay buffer with 2% DMSO. WT-, K22A-, and K22R-EmrE nanodiscs were added to a final dimer concentration of 400 nM, and the plate was incubated at room temperature for 1 h. The final assay volume was 200 μL, and each concentration was present in triplicate. End point fluorescence was determined using a Tecan Spark, and data analysis was performed in Igor Pro v8. The excitation wavelength was 280 nm (15 nm bandwidth), the emission wavelength was 340 nm (20 nm bandwidth), and the measurement integration time was 40 μs with 50 flashes. The Z-position and gain were determined automatically by the Tecan instrument from the A1 position. Data were fit to a single binding isotherm detailed in the following equation:

$$F_{\mathrm{obs}}=F\mathrm{EH}[\mathrm{EH}]+F\mathrm{E}[\mathrm{E}]$$ 1

[EH] is calculated from the following equation:

$$[\mathrm{EH}]=\frac{(E_{\mathrm{T}}+H_{\mathrm{add}}+K_{\mathrm{d}})-\sqrt{(E_{\mathrm{T}}+H_{\mathrm{add}}+K_{\mathrm{d}}{)}^2-4E_{\mathrm{T}}{H}_{\mathrm{add}}}}{2}$$ 2

where ET is the total concentration of the EmrE functional dimer in the sample, H _add is the total added harmane in the sample, and K _d is the dissociation constant. The concentration of the unbound EmrE functional dimer ([E]) is given by the following equation:

$$[\mathrm{E}]=1-[\mathrm{EH}]$$ 3

Synthesis of 4′-Formylbenzo[18]­crown-6

Benzo­[18]­crown-6 (7.563 g, 24.23 mmol) and hexamethylenetetramine (3.6 g, 25.72 mmol) were added to a 150 mL round-bottom flask. Trifluoroacetic acid (18.3 mL, 238.98 mmol) was introduced, and the mixture was refluxed for 24 h under a nitrogen atmosphere. After completion, the resulting dark-red solution was cooled in an ice bath for 20 min.

Once cooled, ice and a few drops of hydrochloric acid were added to the reaction mixture. The ice bath was removed, and the solution was stirred at room temperature for 2.5 h. The reaction mixture was extracted with dichloromethane, and the organic phase was dried over anhydrous sodium sulfate. The solvent was evaporated under reduced pressure, yielding a crude product.

The crude material was purified by using silica gel column chromatography with dichloromethane/ethyl acetate (90:10, v/v) as an eluent. The solvent was removed from the collected fractions, yielding a light-yellow oil. The product was mixed with diethyl ether and refrigerated overnight, and the precipitate was collected by filtration to obtain 4′-formylbenzo[18]­crown-6 (4.131 g, 12.14 mmol, 47%) as a white solid that was utilized in the subsequent step without characterization.

Synthesis of Benzo[18]­crown-6 BODIPY

To a 500 mL round-bottom flask were added 4′-formylbenzo[18]­crown-6 (3.603 g, 10.58 mmol) and 2,4-dimethylpyrrole (1.86 mL, 18.06 mmol) along with dichloromethane (680 mL). The reaction mixture was maintained under a nitrogen atmosphere at room temperature and stirred for 15 min. Trifluoroacetic acid (164 μL, 2.14 mmol) was added, and the reaction was stirred for 2 h.

The reaction mixture was washed sequentially with 0.1 M NaOH (170 mL) and deionized water (170 mL). The organic phase was separated and dried over anhydrous sodium sulfate, followed by the removal of the solvent under reduced pressure. The resulting residue was transferred to a 250 mL round-bottom flask and dissolved in toluene (40 mL) under nitrogen at room temperature. After stirring for 15 min, p-chloranil (2.342 g, 9.52 mmol) was added followed by stirring for an additional 15 min. Subsequently, triethylamine (6.8 mL) was added followed by the dropwise addition of boron trifluoride diethyl etherate (6 mL) while stirring, and the reaction mixture was stirred for 2 h.

The reaction mixture was washed with deionized water (170 mL) and extracted with dichloromethane. The combined organic layers were dried over anhydrous sodium sulfate, and the solvent was removed under reduced pressure. The crude product was purified by silica gel column chromatography using a mixture of dichloromethane and ethyl acetate (95:5–0:100 v/v) as the eluent. The desired product, benzo[18]­crown-6 BODIPY, was obtained (500 mg, 0.896 mmol, 8%) as an orangish brown-colored compound, was collected by sequential rotary evaporation of several column fractions, and analyzed by ¹H NMR analysis.

Benzo­[18]­crown-6 BODIPY

10-(2,3,5,6,8,9,11,12,14,15-decahydrobenzo­[b]­[1,4,7,10,13,16]­Hexaoxacyclooctadecin-18-yl)-5,5-difluoro-1,3,7,9-tetramethyl-5H-4l4,5l4-dipyrrolo­[1,2-c:2′,1′-f]­[1,3,2]­diazaborinine: ¹H NMR (500 MHz, CDCl₃): δ 6.95 (d, J = 8.1 Hz, 1H), 6.83–6.77 (m, 2H), 5.97 (s, 2H), 4.23–4.19 (m, 2H), 4.13–4.08 (m, 2H), 4.00–3.95 (m, 2H), 3.93–3.89 (m, 2H), 3.83–3.67 (m, 12H), 2.54 (s, 6H), 1.47 (s, 6H). Compound procedures above were adapted from and ¹H NMR in agreement with a literature report (Figure S11).

Chemical Structure of Benzo[18]­crown-6 BODIPY

Structure was confirmed by ¹H NMR.

Fluorescence Polarization

Purified WT-, K22A-, and K22R-EmrE were reconstituted in Saposin nanodisc buffer containing 100 mM MOPS pH 7.0 and 20 mM NaCl. BODIPY-18C6E was prepared from DMSO stocks at 4.23 mM and diluted to 30 or 50 nM in assay buffer with 2% DMSO. WT-, K22A-, and K22R-EmrE nanodiscs were serially diluted in black 96-well flat-bottom half-area plates, and the final assay volume was 50 μL. Plates were recorded immediately after dilution for a 30 min to 8 h time course, and each concentration was present in triplicate. Fluorescence was determined using a TECAN Infinity instrument, and data analysis was performed in Prism10. The excitation wavelength was 485 nm (20 nm bandwidth), the emission wavelength was 535 nm (25 nm bandwidth), and the measurement integration time was 20 μs with 30 flashes. The Z-position and gain were determined automatically by the TECAN instrument from the maximum bound nanodisc concentration. The BODIPY-18C6E concentrations were selected for a stable difference of 100 mP between free and bound fluorophores over a multihour time course. Data were fit to a single binding isotherm.

All data sets are available on Mendeley Data at https://data.mendeley.com/preview/b3srh8vcy4?a=be2e7769-4ee6-4758-9dc3-3d64b31b9e69. DOI: 10.17632/b3srh8vcy4.1. EmrE: UniProt ID P23895.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.biochem.5c00348.

• Experimental data and descriptions of controls including additional NMR spectra and assignments, bacterial growth curves, description of SSME and pyranine assays, and nanoparticle reconstitution and characterization (PDF)

M.B. and K.A.H.-W. conceived this study. M.B., M.T., and V.K. designed the experiments. M.B., T.D., and K.M.H. performed in vitro biophysical experiments. M.J.R. performed the chemical synthesis. M.B., M.J.R., M.T., and V.K. performed NMR experiments. M.B., M.T., V.K., M.T.M., and K.A.H.-W. analyzed the results of experiments. M.B., T.D., M.T.M., and K.A.H.-W. wrote the manuscript. All authors participated in manuscript editing and revision. K.A.H.-W. and M.T.M. secured the funding for this work.

The authors declare no competing financial interest.