Functional promiscuity of small multidrug resistance transporters from Staphylococcus aureus, Pseudomonas aeruginosa, and Francisella tularensis

Small multidrug resistance transporters efflux toxic compounds from bacteria and are a minimal system to understand multidrug transport. Most previous studies have focused on EmrE, the model SMR from Escherichia coli, finding that EmrE has a broader substrate profile than previously thought and that EmrE may perform multiple types of transport, resulting in substrate‐dependent resistance or susceptibility. Here, we performed a broad screen to identify potential substrates of three other SMRs: PAsmr from Pseudomonas aeruginosa; FTsmr from Francisella tularensis; and SAsmr from Staphylococcus aureus. This screen tested metabolic differences in E. coli expressing each transporter versus an inactive mutant, for a clean comparison of sequence and substrate‐specific differences in transporter function, and identified many substrates for each transporter. In general, resistance compounds were charged, and susceptibility substrates were uncharged, but hydrophobicity was not correlated with phenotype. Two resistance hits and two susceptibility hits were validated via growth assays and IC50 calculations. Susceptibility is proposed to occur via substrate‐gated proton leak, and the addition of bicarbonate antagonizes the susceptibility phenotype, consistent with this hypothesis.

. Several studies have implicated SMR transporters in a variety of bacterial processes in addition to toxin efflux, including biofilm production (Andremont et al., 2010;Bay et al., 2017;Willsey et al., 2018), osmotic stress regulation (Bay & Turner, 2012), and toxic metabolite transport (Higashi et al., 2008;Higgins et al., 2019;Kermani et al., 2018).In addition, they are widespread in environmental and clinical isolates (Bjorland et al., 2001;Furi et al., 2013;Ignak et al., 2017;Katongole et al., 2020;Kiddee et al., 2013;Kondori & Mansury, 2023;McNeil et al., 2016McNeil et al., , 2023;;Suma et al., 2023;Zahedani et al., 2021), but our understanding of the functional role of the SMR transporters found in these isolates is limited.Recently, our lab discovered that the model SMR transporter from Escherichia coli, EmrE, confers susceptibility to some substrates rather than resistance.Here we assess whether this functional promiscuity is unique to EmrE or extends across other members of the small multidrug efflux subfamily of drug and metabolite transporters.To that end, we tested the ability of three different SMR transporters to confer either resistance or susceptibility to different compounds when heterologously expressed in E. coli, under conditions where our prior studies have established the relationship between the phenotype observed in bacteria and the biochemical function of the transporter in vitro.This enables a direct comparison of the transport behavior and functional promiscuity of EmrE and the three additional SMR homologs.
EmrE from E. coli was the first SMR transporter to be functionally characterized both in vitro and in vivo and is still the most extensively studied SMR homolog (Burata et al., 2022;Yerushalmi et al., 1995).EmrE functions as an antiparallel homodimer and is a highly promiscuous transporter that exports quaternary ammonium compounds, polyaromatic cations, dyes, and planar toxins across the inner membrane of E. coli (Bay et al., 2017;Morrison et al., 2012Morrison et al., , 2015;;Saleh et al., 2018;Yerushalmi et al., 1995).The original mechanistic model of coupled 2 H + : 1 toxin antiport, coupling efflux of substrate to inward movement of protons down the proton-motive force (PMF), was based on early experimental data showing electrogenic transport of a + 1 toxin and electroneutral transport of a + 2 toxin (Robinson et al., 2017;Rotem & Schuldiner, 2004;Schuldiner, 2009;Yerushalmi et al., 1995;Yerushalmi & Schuldiner, 2000a).Proton and drug binding both occur at two central glutamate residues, E14 on each monomer, and mutation of this glutamate abolishes transport (Muth & Schuldiner, 2000).However, more recent mechanistic studies have demonstrated that EmrE is not a strictly coupled proton/ drug antiporter and may be able to perform substrate-gated proton uniport, and possibly other modes of transport, in addition to antiport (Robinson et al., 2017;Spreacker et al., 2022Spreacker et al., , 2023)).This new mechanistic model of EmrE transport has major biological implications since any of the alternative transport modes should lead to enhanced susceptibility, rather than resistance to the small molecule substrate (Spreacker et al., 2022).This is fundamentally different than mere polyspecificity (Figure 1a), where one protein can transport a variety of different substrates, but all substrates are moved in the same direction and with the same stoichiometry to the coupling ion (proton in the case of the SMR transporters).Mechanistic promiscuity (Figure 1b) as observed in the SMR transporters and a few other transporter families (including Nramp, LacY, and YiiP) is distinct, since different substrates are transported with different stoichiometry with respect to the coupling ion, including in both coupled and uncoupled modes (Bazzone et al., 2016(Bazzone et al., , 2017;;Bozzi et al., 2019;Hussein et al., 2023).Recently, we confirmed this functional promiscuity experimentally, showing that the small molecule harmane induces uncoupled proton leak through EmrE, resulting in a growth defect and reduced NADH production in E. coli expressing functional transporter (Spreacker et al., 2022).Thus, EmrE can perform at least two types of transport: proton-coupled substrate antiport and substrate-gated proton leak.Proton-coupled antiport actively effluxes toxic substrates, enhancing bacterial growth and survival, and is the basis for the well-established ability of EmrE to confer resistance to a broad class of small molecule substrates (Yerushalmi et al., 1995).Substrate-triggered proton leak dissipates the transmembrane pH gradient, presumably triggering additional metabolic changes that confer susceptibility rather than resistance to some compounds based on the type of transport activity triggered by the individual substrate (Spreacker et al., 2022).
Other members of the SMR family have been identified through sequence analysis (Kermani et al., 2018;Lolkema et al., 1998;Nelson et al., 2017) and have been divided into four subtypes based on sequence, substrates, and regulatory elements as recently reviewed in (Burata et al., 2022).Since experimental assessment of transport or antimicrobial resistance phenotypes has only been performed for a relatively small number of SMR homologs and a limited number of substrates, this classification is ongoing (Bay & Turner, 2012, 2016;Brill et al., 2012;Kermani et al., 2018;Lytvynenko et al., 2016;Mitchell et al., 2019;Nasie et al., 2012;Saleh et al., 2018).The Qac subtype (drug and antiseptic exporters, named for their ability to efflux quaternary ammonium compounds), which includes EmrE, is the most promiscuous subtype with broad substrate polyspecificity.This family also includes the genes referred to as qacE, qacG, emrE, and smr and classification is based on conformity to the consensus sequence, that of emrE (Burata et al., 2022).However, it is not known whether the mechanistic promiscuity observed for EmrE applies to other members of the Qac subfamily.Most EmrE homologs have only been identified via sequence similarity with minimal biochemical characterization, leaving their transport capabilities and substrate profiles not well defined.Understanding what defines a substrate for the Qac subfamily more broadly, and for individual transporters specifically, will help determine the molecular interactions responsible for this mechanistic promiscuity that can ultimately lead to enhanced antimicrobial susceptibility.
Prior functional screens of EmrE have focused on quaternary ammonium compounds, dyes, and other antimicrobial compounds that are common substrates of many different multidrug (MDR) efflux pumps, such as methyl viologen and ethidium bromide (Banigan et al., 2015;Bay & Turner, 2012;Brill et al., 2015;Saleh et al., 2018).
No other members of the Qac subfamily have been screened against a chemically diverse panel of small molecules to understand their substrate profiles.To better understand the potential mechanistic promiscuity and substrate profiles of other members of the Qac subfamily, here we investigate the function of three other SMRs when heterologously expressed in E. coli: SAsmr (QacC) from

Staphylococcus aureus, PAsmr from Pseudomonas aeruginosa and
FTsmr from Francisella tularensis.These three homologs were chosen from SMR transporters found in pathogens of high interest for their ability to cause disease in humans or animals and were selected to encompass the diversity of the Qac subfamily.Two homologs were chosen from the relatively small subset for which there is some prior biochemical characterization (PAsmr, SAsmr), while one has not been biochemically characterized (FTsmr).Two are chromosomally encoded like EmrE (PAsmr, FTsmr), while one is found on a plasmid (SAsmr).Pairwise sequence identity between all four homologs (EmrE, SAsmr, PAsmr, FTsmr) ranges from 34% to 46%.Within the QAC transporters of the SMR family (PFAM 00893), the four homologs tested here sample different portions of the main cluster (Figure S1).
PAsmr from P. aeruginosa was found to confer resistance to dyes and the aminoglycoside antibiotics gentamycin, kanamycin, amikacin, and neomycin in P. aeruginosa (Li et al., 2003).It is thus one of the few SMR transporters shown to confer resistance to antibiotics used against its host organism.PAsmr has also been shown to transport methyl viologen and tetraphenylphosphonium (TPP + ) in vitro (Ninio et al., 2001).SAsmr (QacC) from S. aureus has been shown to confer resistance to ethidium bromide in both Staphylococcus and E. coli (Abd El-Aziz et al., 2021;da Silva Abreu et al., 2021;Fuentes et al., 2005;Lee et al., 2020;Tang et al., 2020) and has previously been expressed and purified at the scale needed for biophysical characterization (Poget et al., 2007(Poget et al., , 2010)).FTsmr from F. tularensis, the causative agent of tularemia, is almost completely uncharacterized, but an smr sequence has been identified as part of an arsenic resistance locus in Francisella species (Kassinger & van Hoek, 2021;Snowden & Simonsen, 2022).
We hypothesize that PAsmr, SAsmr, and FTsmr are capable of the functional promiscuity previously reported for EmrE and described above (Robinson et al., 2017;Spreacker et al., 2022).If true, heterologous expression of any of these SMR transporters in E. coli will have a positive impact on NADH production and bacterial growth in the presence of some small molecule substrates (resistance) and will have a negative impact on NADH production and bacterial growth in the presence of other small molecule substrates (susceptibility).To test this hypothesis, we performed a F I G U R E 1 Functional promiscuity and sequence similarity of four SMRs.(a) While SMRs may perform polyspecificity, or recognition of multiple substrates toward the same outcome, we hypothesize that they also perform functional promiscuity, or substrate-dependent biological outcomes.(b) Cartoons showing aromatic (Trp, Phe, Tyr) and charged (Asp, Glu, Lys, Arg, His) amino acid residues and their approximate locations on a monomer of each SMR homolog.Residues conserved across all four SMRs are shaded in gray, and those charged or aromatic residues that are not consistent across all four are colored according to their type.As these are the residues most likely to interact with substrates, analysis of these differences provides clues may explain differing phenotypes.(c) Alignment of EmrE, SAsmr, PAsmr, and FTsmr shows 30%-50% sequence identity for SAsmr, PAsmr, and FTsmr compared to EmrE, with highly conserved residues highlighted in yellow.

| RE SULTS
To compare the substrate specificity profile and functional promiscuity of SAsmr, PAsmr, and FTsmr with EmrE, each transporter was heterologously expressed in MG1655 ΔemrE E. coli.Mutation of the critical glutamate residue to glutamine (E14Q-EmrE, E14Q-PAsmr, E13Q-SAsmr, E13Q-FTsmr) blocks transport and is a wellestablished non-functional mutant of SMR transporters (Yerushalmi & Schuldiner, 2000b).Both WT and non-functional mutants of each SMR homolog express equally well and are inserted into the membrane with similar efficiency in E. coli (Figure S2g), and there is minimal difference in the growth of E. coli expressing each transporter in media alone (Figure S2d-f).In addition, growth assays show that there is no difference in growth between E. coli expressing WTor E14Q-EmrE versus E. coli transformed with empty plasmid in media alone (Figure S2a) or in the presence of harmane (susceptibility substrate) or the presence of ethidium (resistance substrate) (Figure S2b,c).This shows that the expression of an SMR transporter under these conditions does not have a measurable fitness cost.

| SMR transporters have broad substrate profiles
To broaden our understanding of the substrate specificity profiles for these transporters we performed Biolog Functional Phenotyping Microarrays using the chemical sensitivity panel (Shea et al., 2012).
The goal of this screen was to assess whether the SMR transporters only provide resistance to toxic compounds or can confer both resistance and susceptibility in a substrate-specific manner, as an indication of their ability to perform different types of uncoupled and/ or coupled transport.This Biolog phenotypic microarray is a panel of ten 96-well microplates with 240 different compounds and includes many drug-like compounds of interest for understanding SMR activity.Many known SMR substrates are polyaromatic cations, and this commercially available assay is designed to avoid interference from the natural fluorescence of such compounds, enabling a single consistent assay format for characterization across several hundred potential substrates selected without bias toward common MDR efflux pump substrates.Our recent Biolog phenotypic microarray screen of E. coli revealed that EmrE confers resistance or susceptibility depending on the identity of the substrate, an unprecedented result (Spreacker et al., 2022).Here, we repeated this screen on SAsmr, PAsmr, and FTsmr expressed in MG1655 ∆emrE E. coli to understand if this functional promiscuity applies to the QAC subfamily at large.
Some data for EmrE are included to facilitate comparison with the additional SMR transporters.
The Biolog screen was run in parallel with E. coli expressing either wildtype or non-functional (E13Q-or E14Q-point mutants that eliminate the primary binding site) SMR transporter and measured NADH production over 24 h using a colorimetric indicator.
While empty plasmid is a more common negative control, it is not uncommon for membrane protein expression to impact bacterial metabolism (Gubellini et al., 2011;Wagner et al., 2007).We therefore used cells expressing the non-functional point mutant as our negative control.Phenotypic differences between cells expressing functional transporter and cells expressing non-functional transporter unambiguously identify SMR substrates since the only difference is transport activity.Thus, this approach eliminates false hits due to variations in compound solubility, effects on media, or protein expression.The area under the curve represents the total metabolic activity and the difference between cells expressing functional and non-functional transporter (∆AUC) reflects the impact of transporter activity on NADH production.
If NADH production is greater when wildtype transporter is expressed, it indicates that transport activity is beneficial to cell growth and metabolism, which we denote as "conferring resistance" to that compound.If NADH production is greater when the non-functional point mutant of the same transporter is expressed, it indicates that SMR activity is detrimental, and the SMR transporter "confers susceptibility" to that compound (see Methods for selection criteria).
The entire assay comparing the impact of functional vs nonfunctional transporter for each SMR homolog was run in biological triplicate.For each screen, ∆AUC was calculated for individual wells and scored as positive (resistance) or negative (susceptibility) hits if the difference was significantly (>3 S.D.) above or below the 10% trimmed mean, as described in the methods.Each set of Biolog plates contains four wells, two wells at a lower concentration and two wells at a higher concentration, for each compound.Since the entire assay was run in triplicate for each transporter, this results in a total of 12 wells for a single compound, or a maximum possible score of ±12.To account for the non-zero rate of false-positives or false-negatives in scoring individual wells, as well as the potential that the transporter could confer resistance or susceptibility at one compound concentration but not the second concentration in the Biolog plates, we selected any compound that scored ≥ +5 as a resistance hit or ≤ −5 as a susceptibility hit.All the known EmrE resistance substrates (methyl viologen and acriflavine) that are present in the Biolog compound set were identified as resistance hits using this scoring system, confirming the validity of this scoring method.
The screen identified compounds in both categories, resistance, and susceptibility, for all four of the SMR homologs and generated a large but manageable list of potential hits (Figures 2 and S3; Table 1).Hits were clustered via agglomerate hierarchical clustering (Figure 2).
We chose four compounds with large resistance or susceptibility hit scores for at least one SMR transporter for further phenotypic analysis: methyl viologen (MV), chelerythrine chloride (CC), harmane, and 18-crown-6 ether (18c6e) (Spreacker et al., 2022).Methyl viologen is a canonical EmrE substrate widely used for assessing SMR activity (Morimyo et al., 1992;Ninio et al., 2001), but is only a resistance hit for three of the four transporters.Chelerythrine is an antibacterial natural product (He et al., 2018) that was selected for additional studies because it had a high positive hit score (resistance) for all four transporters.Harmane was chosen because it had a large negative hit score (susceptibility) for EmrE and was the only susceptibility hit previously characterized (Spreacker et al., 2022).
18-crown-6 ether was a relatively weakly scored susceptibility hit but was chosen because it clustered with the most highly scored susceptibility hits (Figure 2).Harmane and 18-crown-6 ether are also each susceptibility hits for two of the four transporters, EmrE and FTsmr for harmane, EmrE and SAsmr for 18-crown-6 ether, providing a comparison of specificity between susceptibility hits.These four compounds vary greatly in structure (Figure 3).Other hits with high scores (positive or negative), especially susceptibility hits oxytetracycline and hexachlorophene, were not included in this set because of solubility issues, but are an important direction for future study.The full results (Figures 2 and 3; Tables 1 and S1) echo what we observe for these four compounds: all four SMR homologs confer resistance to some compounds and susceptibility to others, as we previously observed for EmrE (Spreacker et al., 2022).While there is often consistency in the phenotype across different homologs for each substrate, there is also some variation in which compounds are substrates for each transporter.This variation in substrate specificity is expected given the sequence variation among the homologs (Figures 1 and S1), particularly in the TM1-3 region known to be important for determining the specificity profile of the SMR transporters for drug resistance (Bay & Turner, 2012;Brill et al., 2015;Saleh et al., 2018).
Previous studies of EmrE substrate specificity suggested that substrate charge and hydrophobicity were key parameters affecting the affinity and transport rate of EmrE substrates (Bay & Turner, 2012;Morrison & Henzler-Wildman, 2014;Rotem & Schuldiner, 2004).These studies focused on known compound classes to which EmrE confers resistance (polyaromatic cations, quaternary ammonium compounds, etc.).Using our larger and more chemically diverse Biolog dataset, we compared the hydrophobicity (cLogP) and predicted charge for all the hits for each SMR (Figure 4).
Hydrophobicity does not appear to correlate with score, with cLogP values varying greatly within both classes of resistance and susceptibility hits.However, the charge does show some association with resistance versus susceptibility classification.Susceptibility hits were almost universally uncharged at neutral pH.In contrast, the majority of resistance hits were positively charged, as expected based on previous literature (Rotem & Schuldiner, 2004).A few resistance hits are likely to be negatively charged, and these will need to be studied further.This trend confirms the previous characterization that SMR transporters confer resistance to hydrophobic cations, and previous screens focusing on this compound class may have missed uncharged substrates also recognized by SMR transporters.

| Growth assays and dose-response curves confirm the metabolic results of the Biolog microarray
We in the membrane at comparable levels (Figure S2), this differential reflects variation in the ability of the different SMR homologs to transport and thus confer resistance to these compounds (Spreacker et al., 2022).In general, these two assays measuring growth and metabolic activity are consistent for each of the SMR homologs, as previously observed for EmrE (Spreacker et al., 2022).While the Biolog screen measures metabolic output and these growth assays measure OD 600 , there are parallels between the hit score from the Biolog screen and the strength of the resistance phenotype seen in the growth assays.For instance, PAsmr had a low score for methyl viologen (Figure S3c), and the growth difference between E. coli expressing WT or the nonfunctional mutant E14Q is barely significant (Figure 5b).Similarly, FTsmr demonstrates the smallest growth differential in the presence of chelerythrine chloride (Figure 5f) and has the lowest Biolog score for that compound (Figure S3b).This relationship is more tenuous for susceptibility substrates.For instance, 18-crown-6 ether, which was not a hit for PAsmr (Figure S3c), does lead to susceptibility in the growth assays (Figure 6e).This suggests that susceptibility substrates may have a different interaction with the transporters than resistance hits.chloride (Figure S4a-d).These results act as a benchmark for the ability of other SMRs to complement EmrE in E. coli.
Functional SAsmr showed significant resistance to chelerythrine chloride (Figure S5d), but minimal resistance to ethidium bromide, methyl viologen, and cetylpyridinium chloride (Figure S5a-c), highlighting that the impact on growth and metabolism occurs in a narrow concentration window.This result is consistent with prior reports that SAsmr (qacC) conferred resistance to cetylpyridinium chloride and similar compounds in S. aureus (Littlejohn et al., 1992).
PAsmr is known to confer resistance to dyes and aminoglycosides in P. aeruginosa (Li et al., 2003) along with the MexAB-OprM efflux system.Similarly, dose response curves and IC 50 values reported here demonstrate that PAsmr confers at least 2-fold resistance to ethidium bromide (Figure S6a).Further, similar resistance to chelerythrine and cetylpyridinium is seen in E. coli expressing functional PAsmr (Figure S6c,d), but little resistance to methyl viologen (Figure S6b).This is consistent with its known resistance to cetylpyridinium in E. coli and identifies chelerythrine as a novel substrate for PAsmr (Mitchell et al., 2019).
There is little information available on the function of FTsmr in vivo, but we see significant, consistent resistance to methyl viologen and cetylpyridinium in E. coli expressing functional FTsmr (Figure S7b).IC 50 values increase by under 2-fold for E. coli expressing WT-FTsmr in the presence of ethidium and chelerythrine.While this increase is not significant according to our thresholds, it demonstrates that even FTsmr can complement EmrE in E. coli.
The more novel result, that SMR homologs can confer susceptibility to some substrates rather than resistance, was also further assessed by measuring dose-response curves for E. coli expressing each SMR in the presence of harmane or 18-crown-6 ether (Tables 3,  3, Figure S4g).Although harmane had a greater, more consistent susceptibility phenotype in the Biolog assay (Figure 3), only EmrE showed a consistent reduction in harmane IC 50 values with functional transport.In contrast, 18-crown-6 ether was a weaker, less-consistent hit in the Biolog assay but there is a significant decrease in IC 50 values for this substrate with functional EmrE (Figure S4g), SAsmr (Figure S5g) and PAsmr (Figure S6g).While there is less consistency between assay formats for the susceptibility hits, these results still demonstrate that EmrE and its homologs from other SMRs can confer susceptibility to some substrates in vivo.The inconsistency between this assay, the growth assays, and the Biolog assay may also be explained by differential effects of susceptibility substrates on growth and metabolism as discussed in the next section.Further study of the mechanisms of susceptibility for different substrates and transporters will help elucidate this relationship further, as will much broader screening to identify more substrates.
These results show some variation in the specific substrate profile for each SMR homolog, as expected based on sequence variation in regions of the transporters known to be important for substrate binding and specificity.Although the results of the growth and metabolic assays characterizing the ability of SMR transporters to confer susceptibility or resistance to different compounds in E. coli are not perfectly consistent, these orthogonal assays confirm that the functional promiscuity previously described for EmrE (Spreacker et al., 2022) does extend to at least three other SMR homologs within the same subfamily.

| Bicarbonate eradicates the phenotype of harmane and 18-crown-6-ether for all tested SMR transporters
Functional promiscuity of EmrE leads to susceptibility to harmane because harmane binds to the transporter and triggers uncoupled proton flux (Spreacker et al., 2022).In other words, EmrE acts as a harmane-gated proton uniporter, but it functions as a protoncoupled antiporter of ethidium and other resistance substrates.
Uncontrolled proton leak through EmrE dissipates the ∆pH component of the proton motive force, leading to the specific growth and metabolic defects detected in the assays in E. coli.This is consistent with the time-dependent bacteriostatic effects observed in the growth curves for harmane-induced susceptibility with functional SMR transporter, while methyl viologen has a more immediate and complete suppression of bacterial growth unless a functional transporter is present to confer resistance.In addition, E. coli compensate for ∆pH dissipation by enhancing ∆ψ (membrane potential) to attempt to maintain a stable proton motive force (Bakker & Mangericht, 1981), which may explain the more variable phenotype observed in the different assays for the susceptibility hits.
We hypothesize that ∆pH dissipation is also one mechanism by which the other SMR transporters confer susceptibility.Therefore, F I G U R E 4 Substrate charge, not hydrophobicity, appears to be a determining factor in resistance versus susceptibility.The cLogP and formal charge of Biolog hits from EmrE (a), SAsmr (b), PAsmr (c), and FTsmr (d) reveal patterns that could differentiate resistance substrates from susceptibility substrates.The cLogP values and formal charges were taken from PubChem.There appears to be no influence of hydrophobicity on whether a substrate is a resistance or susceptibility hit of these SMR transporters.However, formal charge varies between the two groups of hits.Resistance hits (score ≥5) have a charge, whether it is positive or negative, but susceptibility hits (score ≤ −5) are almost all neutral.we repeated the dose-response curves for the susceptibility hits (harmane and 18-crown-6-ether) in the presence of 25 mM sodium bicarbonate.Sodium bicarbonate can diffuse directly through the membrane and dissipates ∆pH independently of any membrane protein or transporter.This will suppress the susceptibility phenotype if it is due to a proton leak since there will no longer be any ∆pH to dissipate.This is exactly what was previously observed for EmrE in the presence of harmane (Spreacker et al., 2022), and is replicated here (Table 3, Figure S4f,h).For the other three transporters, the addition of bicarbonate increases the IC 50 value of harmane equally for both functional and non-functional transporter, and any observed susceptibility phenotypes are abolished with the addition of bicarbonate.
18-crown-6 ether showed a more consistent susceptibility phenotype in E. coli expressing the various SMR homologs.In the presence of 25 mM sodium bicarbonate, the IC 50 values for 18-crown-6-ether increase by 5-fold or more (Table 3) showing a significant suppression of the impact of SMR activity.These results are consistent with substrate-gated proton leak and ∆pH dissipation as the mechanism of harmane-and 18-crown-6-ether-induced susceptibility.

| DISCUSS ION
The increasing rates of antibiotic resistance across bacterial species from E. coli (Ramstad et al., 2021), Staphylococcus (Lee et al., 2020;Littlejohn et al., 1992), P. aeruginosa (Heir et al., 1999;Li et al., 2003), and F. tularensis (Kassinger & van Hoek, 2021), among others, highlights the need for better understanding of the mechanisms underlying antibiotic resistance and the identi- While the resistance phenotype is present in all conditions, the strength of this resistance is different between each SMR homolog.Similar assays for EmrE are published in (Spreacker et al., 2022).Error bars represent standard deviation.et al., 2020;Robinson et al., 2017;Spreacker et al., 2022;Thomas et al., 2018).The Qac subfamily of SMRs has been investigated as a minimal model system to study mechanisms of multidrug efflux due to their small size and broad substrate profiles.However, their native function is not known and it is not clear how they contribute to antimicrobial resistance across the wide range of bacteria in which they are found.Here we compared the activity of several different members of the Qac subfamily and showed that the mechanistic promiscuity observed for EmrE is a more general property observed in other members of this subfamily.
The alternative transport activity of EmrE other Qac transporters is of interest because substrate-gated proton leak dissipates the ∆pH component of the proton motive force through a proteinmediated mechanism.Should other members of this subfamily and potentially other transporters entirely be capable of functional promiscuity, not just polyspecificity, this may ultimately provide novel methods of targeting multidrug resistance, such as PMF dissipation or antimicrobial influx (Figure 1a).While understanding these transporters in their native organisms is ultimately key to therapeutic development, the benefit of expressing the three SMR homologs initially in E. coli is that it allows direct comparison of substrate specificity profiles and the ability of each transporter to substitute for EmrE without the confounding factor of variation in inherent resistance/susceptibility of each bacterial species.Additionally, while it is unlikely that SMRs are typically the primary resistance mechanism in their native organism, it is striking that our results indicated a significant enough resistance phenotype due to the function of these SMR transporters that it can be detected even in the presence of major efflux pumps such as AcrAB-TolC.While the Biolog assay is a limited screen and more data is needed to determine general structure-activity relationships for substrates, it provides a good starting point to demonstrate functional promiscuity and challenge the assumptions that all SMR substrates are charged, polyaromatic or quaternary ammonium compounds.
F I G U R E 6 SMRs confer susceptibility to harmane and 18-crown-6 ether.Average (n = 3) growth curves for functional (black) and nonfunctional (red) EmrE, SAsmr, PAsmr, and FTsmr expressed in MG1655 ∆emrE E. coli cells are displayed for harmane (a-c) and 18-crown-6 ether (d-f).The corresponding concentrations of each drug and the specific SMR are listed in each graph.Bacteriostatic phenotypes like those seen previously for EmrE occur a few hours after the assay is initiated, suggesting a metabolic link to the phenotype.Similar assays for EmrE are published in (Spreacker et al., 2022).Error bars represent standard deviation.

mM 18c6e
To gain insight into the different specificity profiles between SMR transporters, we evaluated differences in charged and aromatic residues, given the importance of these residue types in multidrug recognition across MDR transporters and transcription factors.The loops and tail regions of EmrE have been implicated previously as potential additional binding sites (Banigan et al., 2015;Glaubitz et al., 2000;Saleh et al., 2018;Thomas et al., 2018), and recently harmane was shown to interact with the C-terminal tail, TM2, and the TM3-4 loop (Spreacker et al., 2022).While EmrE and PAsmr have similar amino acid sequences in these regions, SAsmr and FTsmr have additional charged and aromatic residues in TM2, and SAsmr lacks an aspartate in the TM3-4 loop (Figure 1c).FTsmr and SAsmr displayed minimal differences in susceptibility to harmane, suggesting some of these differences may be crucial for harmane's susceptibility phenotype and proton uniport (Spreacker et al., 2022).In addition, PAsmr displays minimal resistance to methyl viologen, consistent with earlier reports (Li et al., 2003).PAsmr has fewer aromatic residues than the other three SMRs, which may reduce ability to interact with the aromatic methyl viologen.Differences even in the abilities of different SMRs to confer resistance suggests that there may be even greater functional differences within the subtype that challenge the assumption that Qacs primarily transport quaternary ammonium compounds, and there may be unrealized native functions or other resistance contributions.While evaluation of these differences is preliminary, it suggests future hypotheses and regions of interest in determining substrate-phenotype relationships and may someday allow prediction of SMR phenotypes based on sequence.
Finally, this work suggests a possible long-term future direction of development of chemical adjuvants to include in existing antibiotic treatments.Inhibition of SMRs has been studied previously as a new strategy to combat resistant bacteria (Mitchell et al., 2019;Ovchinnikov et al., 2018), but SMR family members are rarely the dominant efflux pump contributing to antibiotic resistance in bacteria.
Thus, simple inhibition is not likely to have a major clinical impact.We do not propose to inhibit SMR-mediated efflux, but rather shift these transporters from proton-coupled-substrate-antiport to substratetriggered-proton-leak, effectively using small molecules to drive an entirely different transport function.It is unlikely that ∆pH dissipation will ever provide the bactericidal efficacy desired for an antibiotic, but it has potential as antibiotic adjuvant for two reasons.First, a potential adjuvant may synergize with existing antibiotics whose activity is influenced by the proton motive force.Second, the major multidrug efflux pumps in bacteria are proton-coupled antiporters, and dissipation of ∆pH to the extent that it impacts bacterial growth and metabolism, as demonstrated here in E. coli, will perturb the energy source for other proton-coupled pumps.Usage of antibiotic adjuvants is increasing as standard antibiotics are losing their efficacy.To date, the only FDAapproved adjuvants are beta-lactamase inhibitors, but other classes are in development (Liu et al., 2019).Some of these classes target efflux pumps for inhibition or to act as routes of concentrative uptake of antibiotics in bacteria.This work suggests that the functional pro-

| Sequence-similarity network
A sequence-similarity network was generated for the PFAM Family PF00893 (Multi_Drug_Res) using the EFI-EST webserver, as previously described by Kermani et al (Kermani et al., 2020;Zallot et al., 2019).An alignment score of 20 was used, and the 50% identity group analyzed further in Cytoscape using the prefuse forcedirected layout (Shannon et al., 2003).

| Expression validation
C-terminal 6xHis tags were added to the SMR genes and E13/14Q mutants in the pWB primers via Gibson assembly (for expression validation only).MG1655∆emre E. coli were transformed with the plasmids and allowed to grow in Mueller-Hinton Broth for 20 h.
Pellets were harvested at 4000 g for 30 min.SDS-PAGE sample preparation was carried out by adding lysis buffer (250 mM sucrose, 100 mM NaCl, 2.5 mM MgSO 4 , 20 mM tris pH 7.5, 5 mM βmercaptoethanol, 1 mg/mL lysozyme, DNAse, 1 μg/mL pepstatin, 10 μM leupeptin, and 100 μM PMSF), and lysing by sonication.The membrane fraction was separated by a high-speed spin, resuspended in the same buffer, and solubilized with 40 mM DM (decylmaltoside, Anatrace) at RT for 1 h.Samples were then heated at 37°C for 30 min, spun for 30 min to remove debris and genomic DNA, and mixed with SDS loading dye and an additional 2% SDS.
Samples were run on a 12.5% Bis-Tris gradient gel and transferred to nitrocellulose membrane for 1 h at 100 V.The membrane was blocked with Qiagen Blocking Reagent (Qiagen Penta-His HRP Conjugate Kit) for 1 h, probed with 1:500 Penta-His HRP antibody in blocking solution for 1 h, and washed with ECL reagent and visualized.

| Biolog Phenotype MicroArrays
monitored the growth of MG1655 ΔemrE E. coli cells expressing each functional or non-functional SMR in the presence of methyl viologen (Figure 5a-c), chelerythrine chloride (Figure 5d-f), harmane (Figure 6a-c), and 18-crown-6 ether (Figure 6d-f) to validate hits from the Biolog screen (Figure 3) in an orthogonal assay.In the presence of methyl viologen and chelerythrine chloride (Figure 5), MG1655 ΔemrE E. coli cells expressing functional SMR proteins (black) have increased growth compared to cells expressing their non-functional counterparts (red).Even though this resistance phenotype remains the same, the strength of the growth differential differs between the different SMR homologs.Since we are expressing each of the transporters in the same strain of E. coli and they express and insert Dose-response curves of E. coli expressing EmrE, SAsmr, PAsmr, or FTsmr with each of these four compounds plus ethidium (a known resistance substrate of EmrE and common MDR efflux pump substrate) and cetylpyridinium (Biolog resistance hit and antiseptic) quantitatively establish the functional behavior of these transporters(Table 2,.IC 50 values calculated from these curves show significant (at least 2-fold) resistance conferred by functional EmrE in the presence of ethidium, methyl viologen, chelerythrine chloride, and cetylpyridinium F I G U R E 2 Hierarchical clustering of phenotypic microarray hits.Compounds in the phenotypic microarray screen were scored according to their impact on bacterial metabolism, with scores of ±5 or greater magnitude considered hits.Hits were hierarchically clustered using Seaborn.Red indicates compounds to which the transporter confers resistance and blue indicates compounds to which the transporter confers susceptibility, with color intensity based on the magnitude of the score.
Summary of Biolog hit scores for specific compounds studied in this work.Biolog screen: n = 3.

F
The functional promiscuity of SMR transporters extends beyond EmrE.Heat maps displaying the average ∆AUC (area under the curve) from three replicates of the Biolog assay for EmrE, SAsmr, PAsmr, and FTsmr.Data is shown for methyl viologen, chelerythrine chloride, harmane, and 18-crown-6 ether, with low and high concentration wells as indicated.Red indicates resistance and blue indicates susceptibility, with color intensity based on the strength of the phenotype., Figures S4e,g, S5e,g, S6e,g and S7e,g).Dose response curves for EmrE with harmane (FigureS4e) or 18-crown-6 ether demonstrate a significant decrease in IC 50 values when E. coli express functional EmrE (Table fication of novel targets for antibiotic development.We recently discovered that the SMR transporter EmrE can perform substrategated proton uniport (leak) in addition to its well-known proton/ substrate antiport activity.This flips the paradigm of transporters as performing a single specific type of transport (uniport, symport, or antiport) with a single biological outcome to an expanded model where transporters may perform different types of coupled or uncoupled transport with different impacts in bacteria.These results demonstrate the importance of considering not just substrate specificity, but mechanistic promiscuity when assessing Qac transporter function (Glaubitz et al., 2000; Hussey F I G U R E 5 SMR transporters confer resistance to methyl viologen and chelerythrine chloride.Average (n = 3) growth curves for functional (black) and non-functional (red) EmrE, SAsmr, PAsmr, and FTsmr expressed in MG1655 ∆emrE E. coli cells are displayed for methyl viologen (a-c) and chelerythrine chloride (d-f).The corresponding concentrations of each drug and the specific SMR are listed in each graph.
miscuity of the SMR transporter family may inform discovery of alternative targets for development of antibiotic adjuvants with a novel mechanism of action.TA B L E 2 IC 50 values of compounds to which EmrE and its homologs confer resistance.indicate over 2-fold changes between WT and nonfunctional protein and/or a difference in in vitro phenotype near the calculated IC 50 value.

SMR % Sequence Identity SAsmr 100 34 41 35 FTsmr 34 100 44 38 EmrE 41 44 100 45 PAsmr 3 5 39 46 100 Blue shading represents approximate TM helix locations based on EmrE topology
ginosa, S. aureus, or F. tularensis to the various compounds in the Biolog assay.We found compounds to which each SMR conferred either resistance or susceptibility, experimentally demonstrating that PAsmr, SAsmr, and FTsmr are all capable of functional promiscuity that induces susceptibility in E. coli as well as broad substrate polyspecificity when expressed heterologously.Substrates to which the SMR transporters conferred resistance in E. coli (referred to hereafter as "resistance substrates") were almost universally charged, while substrates to which the SMR transporters conferred susceptibility in E. coli (referred to hereafter as "susceptibility substrates") were primarily uncharged.Four compounds were validated with growth assays, and resistance and susceptibility in E. coli were assessed via IC50s.Finally, the addition of bicarbonate in the presence of the susceptibility substrates resulted in equal growth of E. coli expressing WT or non-functional SMR, thereby abolishing the susceptibility phenotype.This matches the previously characterized behavior of EmrE with substrates shown in vitro to trigger proton leak, thus supporting the hypothesis that PAsmr, FTsmr, and SAsmr can perform substrate-induced proton leak like EmrE.
IC 50 values of compounds to which EmrE and its homologs confer susceptibility.IC50 underlined indicate over 2-fold changes between WT and non-functional protein and/or a difference in phenotype near the calculated IC MG1655 ∆emrE E. coli cells containing either WT-or E13/14Q-SMR (non-functional) constructs from the bacterial strains used in this TA B L E 3