Plasmid transmitted small multidrug resistant (SMR) efflux pumps differ in gene regulation and enhance tolerance to quaternary ammonium compounds (QAC) when grown as biofilms

Small multidrug resistance (SMR) efflux pump genes are commonly identified from integrons carried by multidrug-resistant (MDR) plasmids. SMR pumps are annotated as ‘qac’ for their ability to confer resistance to quaternary ammonium compounds (QACs) but few qac are characterized to date. Hence, we have examined SMR sequence diversity, antimicrobial susceptibility, and gene expression from >500 sequenced proteobacterial plasmids. SMR sequence diversity from plasmid database surveys identified 20 unique SMR sequences annotated as qacE/EΔ1/F/G/H/I/L, or sugE. Phylogenetic analysis shows ‘Qac’ sequences are homologous to archetypical SMR member EmrE, and share a single sequence origin. In contrast, SugE sequences are homologous to archetypical member Gdx/SugE and likely originate from different species. SMR genes, qacE, qacEΔ1, qacF, qacG, qacH, and sugE(p), were over-expressed in Escherichia coli to determine their QAC antimicrobial susceptibility as planktonic, colony, and biofilms. SMRs (except qacEΔ1/sugE) expressed in biofilms significantly increased its QAC tolerance as compared to planktonic and colony growth. Analysis of upstream SMR nucleotide regions indicate sugE(p) genes are regulated by type II guanidinium riboswitches, whereas qacE and qacEΔ1 have a conserved class I integron Pq promoter, and qacF/G/H are regulated by integron Pc promoter in variable cassettes region. Beta-galactosidase assays were used to characterize growth conditions regulating Pq and Pc promoters and revealed that Pq and Pc have different expression profiles during heat, peroxide, and QAC exposure. Altogether, this study reveals that biofilm growth methods are optimal for SMR-mediated QAC susceptibility testing and suggests SMR gene regulation on plasmids is similar to chromosomally inherited SMR members.

Introduction phylogenetic relatedness to well characterized SMR subclass members EmrE and Gdx. A total of 20 unique SMR 117 protein sequences were identified from a bioinformatic survey of over 500 SMR encoding plasmids, where six 118 representative SMR sequences were selected (qacE, qacEΔ1, qacF, qacG, and qacH, and sugE(p)) for further 119 experimental AST analysis. Six SMR genes were cloned into the expression vector pMS119EH and transformed 120 into Escherichia coli K-12 strains BW25113 and KAM32 for AST with a library of 13 antimicrobials (9 QACs, 2 121 interchelating dyes, and 2 antibiotics). To determine if growth influenced SMR QAC tolerance phenotypes, 3 122 different AST methods were examined in this study: agar dilution, broth microdilution, and minimal biofilm 123 eradication devices (Innovotech MBEC assays; formerly known as the Calgary biofilm device) representing 124 colony, planktonic and biofilm culture growth physiologies, respectively. All three AST growth methods 125 reconfirmed the current hypothesis that qac and sugE genes confer distinct antimicrobial substrate selection 126 profiles based on their phylogenetic relationships to either the SMP or GDM subclass of the SMR family. A 127 comparison of AST growth methods tested our second study hypothesis, that SMR over-expression in bacteria (E. 128 coli) growing as colony or a biofilm exhibit greater and more significant QAC tolerance as compared to 129 planktonic growth methods. The third hypothesis we examined was that is plasmid/ integron encoded SMRs 6 related to the GDM subclass will be regulated by a Gdm + riboswitch as compared to SMP homologs by a discrete 131 plasmid encoded promoter. Previous studies indicated that qacEΔ1 genes rely on a 'Pc' or a 'Pq/ Pqac' promoter 132 region associated with class I integrons (15, 37) but is it unclear how other SMR genes are regulated. To address 133 the third hypothesis, we bioinformatically analyzed the 500 nt upstream region corresponding to each SMR 134 sequence we surveyed to identify if class I integron promoters 'Pc' and 'Pq/ Pqac/ P3' (as described by (15, 38)) 135 or Gdm + binding riboswitches (as reviewed by (12)) sequences were present on plasmids. Since very little is 136 known about Pq promoter induction, our final aim delved deeper into class I integron 'Pc' and 'Pq' promoter 137 regulation to determine if exposure to stress, specifically, sub-inhibitory QAC concentrations, heat (42 o C), and 138 oxidative (H 2 O 2 ) stress, had a similar influence on these promoters identified upstream from specific qac genes. 139 By cloning the Pc and Pq promoter regions into a β-galactosidase lacZ reporter plasmid we compared β-140 galactosidase activity using a 96-well microplate assay in E. coli promoter transformant that were grown under 141 various growth modes (planktonic and biofilm) and different stress conditions, highlighting Pq and Pc promoter 142 differences. Altogether this study, clarifies three major knowledge gaps related to plasmid/ integron transmitted 143 SMR antimicrobial resistance profiles and offers greater insight into how plasmid encoded SMR gene expression 144 during bacterial growth influences tolerance to a wide range of QACs. 145

Bioinformatic analysis of plasmid encoded SMR sequences indicate a relationship to GDM or SMP 147 subclass members. 148
The majority (87-92%) of the plasmids encoding for either a 'Qac' or 'SugE' sequence in our database 149 surveys were identified from γ-Proteobacteria, specifically the order Enterobacteriales, and to a lesser extent 150 sequence identified from orders Pseudomonadales, Vibrionales, Aeromonadales ( Fig. 1B; Table S1). Due to the 151 predominance of SMR genes isolated from Enterobacterial plasmids, no obvious trends were apparent to support 152 a link between specific environmental niches and SMR co-selection. Due to animal and human Enterobacteriales 153 contamination, these species can be detected from and thrive in a wide range of conditions, explaining their 154 detection in diverse environments. Plasmid sequencing biases favoring Enterobacterial clinical isolates as well as 155 food/ retail meat contaminants likely bias SMR predominance within Enterobacteriales (Fig. 1C). Therefore, it is 7 unclear if specific SMR members are preferentially enriched or selected by plasmids/ integrons in specific 157 environmental sources. Regardless, the diversity and environmental distribution of 'Qac' and 'Gdx/ SugE' from 158 our plasmid database sequence surveys remains consistent with SMR gene detection rates and trends reported by SMR protein sequences frequently annotated as 'Qac' formed three closely related subclades: i) 162 QacE/QacEΔ1 cluster, ii) QacG, and iii) the QacF/I/L/H clusters (Fig. 1A), where each Qac was distantly related 163 to E. coli EmrE (Fig. 1A). QacEΔ1 was the most frequently detected SMP member within diverse environmental 164 and clinical isolates, in agreement with previous SMR surveillance studies (31,43,46,47). This is not surprising, 165 considering QacEΔ1 has high conservation within the 3' conserved region of Class I integrons, which we 166 identified in 99.5% of plasmid sequences surveyed herein (Table S1, Table S2). Many Qac sequences we 167 examined were misannotated; QacEΔ1 was most frequently misannotated as 'EmrE', 'QacE', and 'Smr', while 168 'QacF' and 'QacH' sequences were misannotated as 'QacE', 'QacEΔ1', 'QacI', and 'QacL' (Table S1, Table S2-169 3) and highlights caution when determining SMR sequence association by annotation only. Multiple sequence 170 alignments of all Qac proteins showed high overall sequence identity (66-99% pairwise identities) to each other 171 but not to EmrE suggesting Qac sequences share a close and common sequence origin (Fig. S1). QacEΔ1 had the 172 lowest sequence similarity, due to the an in-frame replacement of the 4 th TM strand extending the original QacE 173 protein to 115 aa (Fig. S1) (28). QacE, QacF, QacG, and QacH proteins had the least amino acid variations (<15 174 residue differences total) suggesting they are all isoforms with low sequence similarity to EmrE (42-49% pairwise 175 sequence identity; Fig. S1). Most amino acid differences identified between different Qac sequences altered 176 hydrophobic residues located on the lipid-facing surfaces of each TM protein and away from the proposed H + and 177 drug binding active site residue 'E14' as shown in the monomeric QacE homology model generated from E.coli 178 EmrE X-ray diffraction crystal structure (3B5D.pdb (48); Fig. 1D). This suggests that most sequence variations 179 among QacE, QacF, QacG, and QacH isoforms are less likely to impact substrate recognition and substrate efflux. 180 Altogether, our analysis suggests that plasmid-transmitted 'qac' gene sequences in class I integrons have likely 181 originated from a single qac origin, and show close relationship to SMP members, in agreement with previous 182 SMR phylogenetic analyses (16,49). 183 Phylogenetic analysis of plasmid-encoded GDM members formed two distinct subclades, i) a γ-184 proteobacterial clade with homology to E. coli Gdx (NP_418572) and ii) an α-proteobacterial Rhizobiales 'SugE' 185 clade (Fig. 1A, S2). The most frequently identified GDM member was SugE(p) (YP_002302254), which had 83% 186 sequence identity to E. coli Gdx and was identified from γ-proteobacterial plasmids isolated from 187 Enterobacteriales, specifically E. coli and Salmonella enterica species (Table S1; TableS2). SugE(p) was 188 frequently detected on plasmids isolated from species found in contaminated foods and retail animal (primarily 189 poultry) sources (Table S1), suggesting the food industry may select for SugE(p) maintenance on plasmids from 190 these species. Interestingly, 7 GDM sequences we surveyed were 100% identical to chromosomally encoded E. 191 coli Gdx (Fig. 1B), indicating that the archetypical E. coli gdx selected for this study is also horizontally 192 transmissible on Enterobacterial plasmids. In contrast to plasmid encoded SMP members, all 7 GDM protein 193 pairwise sequence identities were much lower when compared to the archetypical E.coli Gdx (25-30%) sequence 194 ( Fig. 1E; Fig. S2). Altogether, this suggests that gdx/ sugE on plasmids are more likely to be acquired from 195 different sequence origins or species that acquired the plasmid. Alternatively, plasmid GDM sequences may have 196 lower selective pressure exerted to maintain their sequence identity on plasmids, however, horizontal gene 197 transfer makes determination of this difficult to discern using the current plasmid dataset. 198

AST of E. coli SMR gene transformants show significant QAC tolerance when grown as biofilms in a 199 wildtype efflux pump strain. 200
Six of the most frequently identified plasmid-encoded Qac and Gdx/ SugE(p) sequences we identified 201 from our proteobacterial plasmid surveys, highlighted in yellow in Figure 1A and listed in Table 1 (52) was selected as a multiple efflux pump gene (ΔacrB, ΔmdtK) deletion strain, which is frequently used for 210 efflux pump AST analyses (for an example refer to (53)). AST of SMR vector transformants BW25113 and 211 KAM32 strains respectively were examined with a library of 13 previous tested QACs and antibiotics selected on 212 the basis of their ability to confer antimicrobial tolerance in other SMR over-expression studies (as reviewed by 213 (4)) and results are summarized on Tables 2-3. 214 Nearly all BW25113 SMR transformants examined using broth or agar AST methods had mean MIC 215 values that were statistically insignificant (within a 2-fold change or identical) to the parental vector control 216 BW25113-pMS119EH, with the exception of the pEmrE transformant ( induction conditions (Fig. S3). Furthermore, the possibility that efflux pump toxicity caused by protein over-266 accumulation within E.coli membranes due to IPTG induction was also not observed in BW25113 or KAM32 267 transformants based on OD 600 nm LB-AMP growth curves (Fig. S4). Therefore, AST growth are an important 268 consideration for QAC tolerance determination, and broth microdilution AST methods appear to be least effective 269 as compared to agar spot plating techniques to determine MIC values for SMR efflux pump overexpression in a 270

KAM32 strain. 271
As compared to BW25113 transformant biofilms exposed to the same antimicrobials, KAM32 272 transformant biofilms resulted in fewer significant increased MBEC values as compared to parental vector 273 controls (Table 3). Interestingly, KAM32 transformant biofilm AST MBEC results identified some SMR genes 274 (qacEΔ1, qacF, sugE(c), sugE(p)) exposed to QACs (CET, BZK, DDAB, CTAB) had enhanced antimicrobial 275 susceptibility (<4-fold reduction in MBEC values) as compared to the vector control (Table 3). This indicates that 276 when dominant efflux pump acrB and mdtK pump are absent in E.coli, the overexpression of SMR efflux pumps 277 work against the cells making them more susceptible to the aforementioned QACs. To determine if SMR 278 overexpression was toxic for biofilm growth, thereby reducing overall biofilm formation in BW25113 and 279 KAM32 strains prior to antimicrobial exposure, the biofilm biomass accumulation on MBEC device pegged lids 280 after 24 hr biofilm growth was quantified using CV staining ( Fig. 2A-B). Both BW25113 and KAM32 281 transformants showed no significant reductions in CV stained biomass on the MBED device pegs when compared 282 to their respective parental vector controls ( Fig. 2A-B) indicating that SMR vectors did not significantly reduce 283 biofilm formation. Interestingly, some KAM32 transformants had significantly increased biomass formation (1.5-284 to 2-fold) for pQacE, pQacEΔ1, pQacG, pQacH, pSugE(p) transformants as compared to pMS119EH (Fig. 2B) suggesting these transformants may promote biofilm formation. Hence, SMR transformants did not significantly 286 impair biofilm biomass formation in either strain. To verify if the deletion of the dominant RND efflux pump 287 acrB was an influential factor in biofilm biomass formation, BW25113 biofilm formation was compared to 288 JW0451 (ΔacrB) and KAM32 was compared to its parental strain TG1 (Fig. 2C); no significant differences in 289 biomass were observed for any strain pair in this study. 290 Altogether, SMR gene expression in KAM32 increased the detection of significant MIC values by each 291 SMR transformant when grown as agar colonies. The antimicrobial tolerance of various SMR transformants 292 differed dramatically in KAM32 as compared to the wildtype BW25113, particularly when grown as a biofilm 293 emphasizing the importance of growth conditions, and efflux pump deletions when performing AST. Comparing 294 different E.coli strains also reveals that KAM32 can determine the broadest SMR substrates from AST profiles; 295 however, caution should be taken when comparing agar colony versus biofilm values, since SMR expression in 296 strains lacking acrB and mdtK increased susceptibility to some QACs when grown as biofilms. 297

Plasmid encoded SMR genes are regulated by either Type II Guanidinium riboswitches or class I integron 298 promoters based on phylogeny. 299
To determine how each SMR gene is regulated on their respective plasmids (and integrons), we examined 300 the 500 nucleotide (nt) upstream region of each SMR gene starting from the +1 start codon (Tables S2-S3). The 301 results demonstrated that all qacEΔ1 and qacE genes primarily possessed a 'Pq' promoter sequence (identical to 302 the sequence described by (15, 37)), -100 nt immediately upstream of +1 start codon (Fig. S5A,C). Remaining 303 qac genes, qacF, qacH, and qacG, had very low (<22%) or no sequence identity to either class I integron Pq 304 within the 500 nt upstream region ( Fig. S5B-C); the only exception was for 2 of the 91 qacH plasmids we 305 examined had >90% sequence identity to Pq promoter (Fig. S5B). Expanding the inspection of nearby open 306 reading frames on these plasmid maps encoding each qac gene showed that the majority (81-94%) of qacF, qacH, 307 and qacG sequences were located within the variable cassette region of class I integrons or nearby transposable 308 elements (Fig. S6). Hence, qacF, qacH, and qacG genes are most likely regulated by the class I integron intI1 Pc 309 promoter or to a lesser extent other promoter(s) associated with upstream transposable elements (Fig. S6). It is 310 important to note, that the majority of Pc promoter sequences (56-94%) we identified from nearly all 'qac' containing plasmids surveyed had >90% sequence identity to the so-called 'weak' version of the class I integron 312 promoter (Table S2; Fig. S8), similar to trends reported in a previous study (38). The remaining Pc sequences 313 were hybrids of the string and weak Pc promoters, as described by Guerin et al. 2011 (38), or were not identified 314 at all within the plasmids. Finally, a small proportion of all qacEΔ1gene encoding plasmids we surveyed (664 315 plasmids) had one or more additional SMR sequences located elsewhere on the plasmid, specifically, qacF (3/16; 316 18.8%), qacG (3/17; 17.6%), qacH (6/91; 6.6%) or sugE(p) (21/115; 18.3%) (Fig. S6). As a result, the expression 317 of additional SMR genes are likely regulated by different plasmidic/integron promoters or riboswitch elements 318 within the SMR gene. 319 Surveys of the 500 nt region upstream of sugE(p) encoding plasmids demonstrated high (>85%) sequence 320 identity to the E. coli gdx 5' UTR of type II Gdm + riboswitch for every sugE(p) regions we examined ( Fig. S7A-321 B). This strongly suggests that sugE(p) gene are controlled by Gdm + riboswitches supporting the third study 322

Pq promoters associated with SMR genes show higher expression in planktonic versus biofilm cultures. 329
Due to the high proportion of qacE and qacED1 associations with the class I integron Pq promoter and 330 qacF, qacG, qacH regulation by Pc 'weak' promoters, we synthesized and cloned the Pq (pPq) and Pc (pPc) 331 'weak' promoter regions (Fig. S8) into a lacZ reporter vector to characterize and compare promoter activity in E. 332 coli BW25113 using β-galactosidase 96 well assays (Table 1). β-galactosidase activity of pPq and pPc was 333 compared to control pMS119EH vector with (placZ; Ptac promoter positive control) and without lacZ gene 334 (pMS119EH; negative control) and induced with and without 0.05 mM IPTG (+/-IPTG) to match AST 335 experiments. β-galactosidase activity of planktonic (broth) transformant cultures grown in rich medium (LB-336 as compared to placZ+IPTG controls up to and including 11 hrs (Fig. 3A). At 24 hrs, all planktonic transformants 338 (excluding pMS119EH) demonstrated reduced levels of β-galactosidase activity similar to placZ-IPTG 339 transformants, which was anticipated as cultures reach stationary phase. At early to mid-log growth phases (3-7 hr 340 timepoints), planktonic pPc transformants had reduced (50-75%) β-galactosidase activity as compared to pPq, 341 suggesting Pq is a much stronger promoter than Pc under these conditions. In fact, pPq transformant β-  oxidative stress. QACs are predicted to enhance oxidative stress as part of their mechanism of action (6) and 358 unfortunately, we did not observe and significant differences response to sub-inhibitory QACs by pPq and pPc 359 transformants in LB ( Fig. 4D-E). This result may be due to the complexity of the rich LB medium which may 360 absorb QACs at sub-MIC concentrations; therefore, we repeated β-galactosidase assay in minimal medium (M9-361 AMP). Switching to M9 growth increased β-galactosidase activity differences between IPTG-induced and 362 showed statistically significant increases in expression when exposed to 0.03 mM H 2 O 2 as compared to no H 2 O 2 368 exposure. Unfortunately, exposure to 0.3 mM was toxic to all M9-AMP cultures so data could not be collected 369 (Fig. 5C). Only pPq transformants in M9-AMP demonstrated a small (10-14%) but significant increase β-370 galactosidase activity when exposed to QACs (CTAB and DDAB; Fig. 5D-F), suggesting that the Pq promoter is 371 more sensitive to QAC-induced damage by comparison to the Pc promoter grown planktonically. Therefore, Pq 372 and Pc promoters respond to various oxidative stressors during planktonic growth conditions, where the Pc 373 promoter is more responsive to H 2 O 2 -induced stress and to lesser extent heat-induced stress. 374

SugE/ Gdx sequence diversity is greater than Qac members and validates antimicrobial substrate selection 376 differences between SMR subclasses 377
Two major findings were apparent from the bioinformatic survey and phylogenetic analysis of SUG 378 protein sequences encoded on Gram-negative bacterial plasmids: 1) GDM members appear to distribute among sugE horizontal gene transfer to plasmids originate from different bacterial phyla/species or are subject to lower 382 selection pressure to maintain sugE sequences but in Enterobacteriales SugE(p) sequences appear to predominate 383 over others ( Fig. 1B-C). We speculate that GDM gene transfer between α-proteobacterial plasmids may biased in 384 favor of the species they co-habitat with, since SUG sequences we identified on α-proteobacterial plasmids failed 385 to identify gdx/ sugE homologs at significant BLASTn e-values (<1.0 x10 -3 ) in the respective genomes the 386 plasmid was isolated from. Why GDM sequence diversity on plasmids is so different remains is unclear. GDM 387 gene transfer and mobility on plasmids may be driven by selection pressure for Gdm + containing substrates, as 388 many Gdm + containing chemicals like QACs are used in high quantities (100 kilotons annually) in food, 389 agricultural, petrochemical, and clinical industries (57-59). GDM homologs have also demonstrated involvement 390 in tolerance to metal containing antimicrobials such as di-and tri-butyltin based on previous studies of 391 Aeromonas molluscorum sugE (60), suggesting some GDM members may recognize more than just the nitrogen 392 cation similar to the ability of EmrE to recognize compounds with phosphorous centers (eg. 393 tetraphenylphosphonium (as reviewed by (4)). Riboswitch type and activity in species that inherit the plasmid 394 may also be a factor driving sugE/ gdx gene mobility. Altogether, the close sequence association between qac 395 genes (particularly qacEΔ1) and sugE(p) make these SMR sequences useful genetic markers to monitor as proxies 396 for QAC tolerant phenotypes and QAC pollution monitoring. 397 In contrast to plasmid encoded GDM sequences, Qac SMP homologs showed very high sequence identity 398 to each other (Fig. 1D), but much lower identity to E. coli emrE ( Fig. 1A; Fig S2), suggesting a single origin for 399 plasmid qac sequences surveyed thus far. It remains unclear why qacEΔ1 predominates over all other SMR 400 sequences, including its original and functional progenitor qacE sequence ( Fig. 1B-C). We speculate that the 401 predominance of the less active QacEΔ1 may be more beneficial to overall cell fitness, when compared to fewer 402 'functional' Qac sequences we observed in the surveys i.e. QacF, QacG, and QacH, Evidence for this may be seen 403 in our planktonic E.coli SMR transformant over-expression data (Fig. S4) and from KAM32 Qac transformant 404 biofilm AST results (Table 3). Additional QAC mechanisms of tolerance may be larger influencing factor, such as 405 the involvement of other intrinsic and unrelated QAC selective efflux pumps like MdtM (61) or bacterial 406 adaptation to QACs resulting in porin, peptidoglycan and lipid biosynthesis alterations as noted in QAC 407 adaptation studies of E.coli (62-64). 408

AST growth conditions and other efflux pump losses impact QAC tolerance conferred by SMR members. 409
Overall, AST of 6 representative plasmid transmitted SMR transformants in E.coli K-12 strains resulted 410 in moderate (4-fold increase from parental vector control) increases in MIC values for various QACs (Tables 2-3).   (1)) that may be 436 more effective when expressed together during QAC exposure or during different growth physiologies as 437 suggested in a previous study (54). The results of this study also provide the first phenotypic AST characterization 438 of SMR genes qacF, qacG, and qacH which recognize and confer tolerance more QACs than QacE (Fig.1F). We 439 show that pQacEΔ1 transformants confer tolerance to CET and AC (Table 2-3), despite the disruption of the 4 th 440 transmembrane helix as noted in a previous study (28). These finding are helpful for ongoing surveillance studies 441 and pertinent for in silico predictions of QAC tolerance when monitoring specific qac members. Hence, SMR 442 genes remain a useful proxy for estimating QAC anthropogenic pollution and low-level antiseptic tolerance (13, 443 69). 444 Finally, AST results we obtained for KAM32 'Qac' transformants demonstrated the most significant 445 increase in MIC values to broadest range of QAC structures when grown as planktonic and agar colonies ( Table  446 3) but fewer QACs were consistently identified as having a significantly enhanced or reduced MIC or MBEC 447 values for all three AST growth conditions when compared (Fig. 1F). We suggest two factors that may have 448 influenced these discordant KAM32 AST results between growth methods. 1) Different antimicrobial diffusion 449 rates in liquid versus solid media may impact the antimicrobial agent concentrations exposed to cells during Since media compositions for planktonic and biofilm AST were identical, our results suggest the mode of growth 457 mode plays an influential role as well as the genetic background of the strains we tested (KAM32 versus 458

BW25113). 459
The second factor influencing KAM32 AST result discordance, may pertain to SMR proteins' ability to 460 function as a 'dual-topology' dimeric transporter in the plasma membranes of a bacterium (as reviewed by (48, 461 70)). A dual-topology dimer refers to the ability of a transporter to fold and insert in either insertion orientation in 462 a membrane, where the amine-and carboxyl-termini of the SMR protein can orient to face cytoplasm or the 463 periplasm but form an asymmetric functional dimer. We observed that some KAM32 SMR transformant biofilms 464 exposed to CTP and CET exhibited increased susceptibility (4-fold reduced MIC values) as compared to the 465 vector control (Table 3), suggesting some SMR efflux pumps acted as importers for specific antimicrobials rather than exporters of these drugs. Due to the dual-topology nature of SMR protein dimers, previous studies of E. coli 467 EmrE and SugE/ Gdx have shown an ability to act as antimicrobial importers depending on amino acid sequence 468 alterations and protein folding conditions (66, 71). It may be possible that the loss of both acrB and mdtK in 469 KAM32 grown as a biofilm creates membrane conditions favorable for SMR to function as an 'importer'. Further 470 examination into SMR folding in E.coli strains lacking efflux pumps acrB and mdtK grown as biofilms is needed 471 to clarify this interesting observation. 472

SMP expression is regulated by integron promoters while GDM members are regulated Type II Gdm+ 473 riboswitches 474
A recent study by Kermani et al. 2018 (2), stated that most SMR transmitted by plasmids relied on 475 plasmid promoters. Our findings indicate this statement only applies to 'qac' SMR subclass homologs, whereas 476 sugE(p) and perhaps other GDM members are regulated by a riboswitch. The findings for our survey show 477 Enterobacterales sugE(p) predominate over all other gdx/ sugE sequences, therefore, it was no surprise that the 478 type II Gdm+ riboswitch was detected similar to E.coli gdx riboswitch (Fig. S7). Translational control of plasmid 479 Additionally, our findings identified that only qacEΔ1 and qacE frequently associate with class I integron 484 Pq promoters, whereas qacF, qacG, qacH were most frequently regulated by the integrase intI1 Pc 'weak' 485 promoter region (Table S2, (Fig. S4). Fitness costs caused by potentially toxic gene over-491 expression from integrons may also explain why both the Pq and Pc integron promoters had β-galactosidase expression levels that were similar or lower than the 0.05 mM IPTG Ptac induction levels we selected as a 493 positive control for SMR gene over-expression (Fig. 3). We suggest that gradual toxic over-expression caused by 494 'leaky' Pq promoters may explain why qacEΔ1 predominates over qacE, as qacEΔ1 has some efflux activity. 495 Finally, gene over-expression fitness costs may explain why other functionally active qac gene isoforms are less 496 abundant and regulated by 'weak' Pc promoters. 497 Our characterization of both pPq and pPc promoters grown as either planktonic or biofilm cultures after 498 24 hrs in LB medium demonstrated similar β-galactosidase expression levels (Fig. 3A-B).

Examination of Pq and 499
Pc promoters in M9 demonstrated that Pq promoters had higher β-galactosidase activity than Pc promoters (2-4-500 fold Miller units increases) and Pq were inducible by peroxide and sub-inhibitory QAC exposure, whereas Pc 501 promoters showed slightly higher β-galactosidase activity when exposed to 42 o C heat stress (Fig. 4). The 502 conditions that induced Pq promoter in our β-galactosidase assays make sense when considering known QAC 503 mechanisms of action that generate oxidative damage to proteins and DNA (6). However, previous Pc promoter 504 studies indicate the promoter is regulated by the SOS response system, where UV light and DNA damage are 505 primary inducers of the response regulator LexA which regulates and binds to the Pc promoter (72). SOS 506 regulation may explain why we observed only modest increases in Miller units for peroxide and QAC exposure of 507 pPc transformants in M9 medium. The shorter length of the synthesized 'weak' version Pc promoter region we 508 synthesized for experimental analysis in this study may another influencing factor. Previous Pc promoter studies 509 typically examined DNA regions 2-4 times the size of our selected core Pc promoter region (37, 38, 73). We 510 avoided using longer Pc regions for analysis due to lower sequence identity between DNA regions surrounding 511 the conserved Pc region core in our plasmid sequence dataset. Future studies of Pq and Pc regions will ideally 512 focus on examining more diverse promoter regions beyond the core Pc region. 513 In conclusion we have shown that SMR gene distribution, sequence variation, and antimicrobial 514 susceptibility on plasmids match trends identified for chromosomally inherited SMR members as noted in recent 515 studies (2, 5). Our analysis support the hypotheses that SMR phylogeny and amino acid sequence can predict 516 antimicrobial susceptibility profiles and particularly when grown as a biofilm in strains with functional intrinsic 517 efflux pumps. Our analysis provides AST characterization of the most frequently detected SMR sequences transmitted on MDR plasmids and reveals SMR gene regulation appears to match chromosomally inherited SMR 519 gene regulation trends. Altogether, this information provides more context to ongoing antimicrobial resistance 520 genetic surveillance studies by providing QAC phenotypes to formerly uncharacterized genes and identifying 521 optimal growth conditions for AST. This study also provides useful insights into how AST growth methods 522 influence QAC antimicrobial phenotypes related to SMR over-expression and highlight a possible role for AcrB 523 and MdtK in regulating SMR pump activity. Future studies will ideally explore how SMR genes contribute to 524 multidrug resistant phenotypes with intrinsic efflux pump systems and other resistance genes carried on MDR 525 plasmids. 526

SMR sequence collection and multiple sequence alignments 528
Plasmid-borne SMR protein and nt sequence searches were conducted using NCBI 529 proteobacterial plasmids with identified SMR protein sequences were obtained, where only 533 plasmid 536 sequences had uniquely different accession sequence numbers. SMR protein sequences were further aligned using 537 the online server COBALT (77) (Table S1; Fig. S1), where it was determined that only 20 SMR protein 538 sequences had unique sequence identity (< 99%; Fig. S2). Unique SMR proteins were annotated as QacE, 539 QacEΔ1, QacF/I, QacG, QacH, and SugE using multiple sequence alignment software Jalview version (v) 2.10.5 540 (78) 541 The 500 nt upstream region of frequently identified SMR gene sequences were also examined (Tables S2-542 3). Plasmid sequence accession numbers were obtained from either UniProt or GenBank and aligned relative to 543 each SMR gene +1 reading frame start site (ATG/GTG) using EMBL-EBI Clustal Omega software (ClustalO; https://www.ebi.ac.uk/Tools/msa/clustalo/) linked to the Jalview software package (78). ClustalO nt alignments 545 were used to identify and compare sequence identity of Pq and Pc promoter regions as well as Types I-III Gdm + 546 riboswitch consensus sequence identity summarized in Fig. S8. If no identity/ consensus was determined within 547 the 500 nt upstream SMR sequence (as was the case for qacF, qacG, and QacH), manual SMR sequence locations 548 were determined to predict the promoter region on the plasmid. Plasmid sequences surveyed are listed in Table  549 S2-S3 and include Pq and Pc promoter detection by the BLAST+ software package (79) to perform BLASTn

Phylogenetic analysis of SMR proteins 554
Phylogenetic analysis of the 20 unique SMR protein sequences was performed using the online PhyML Maximum 555 Likelihood estimation methods (PhyML; (80-82); http://www.atgc-montpellier.fr/phyml/), with 100 bootstrap 556 replicates to determine branching confidence at specific nodes. During PhyML estimations, an approximate 557 Likelihood Ratio test and approximate Bayes estimation methods were used, both resulting in similar clusters; 558 aBayes analysis is shown in Fig. 1A. E. coli EmrE NP_415075 and SugE NP_418572 sequences represented 559 archetypical SMP and SUG subclass members respectively, and Archaeoglobus fulgidus QacE sequence 560 AAB89552 served as the outgroup for analyses (Fig. 1A). 561

Homology modelling 562
Homology models of QacE (NP_044260) and SugE(p) (YP_002302254) protein sequences were generated using 563 the online server I-TASSER (83). QacE and SugE(p) sequences showed the most significant homology based on 564 C-scores of 0.12 and 0.22 respectively to the E. coli EmrE X-ray crystal structure monomer A (3b5dA.pdb; (48)). 565 I-TASSER homology models were used to generate the protein ribbon cartoon images in Fig 1D-

Drugs, strains, plasmids, and cloning procedures used in this study 568
Antimicrobial compounds tested in this study were obtained from Tokyo Chemical Industry Co.  Plasmid-encoded SMR sequences representing each annotated subclade in Fig. 1A-C (ie. qacE, qacED1,  576 qacF, qacG, qacH and sugE (p)) as well as promoter region sequences, Pc and Pq, shown in Fig. 2A were 577 selected for further experimental analysis involving gene synthesis and cloning into pUC-57 by BioBasic Inc. 578 Gene Synthesis Services (Markham, Ontario, Canada; Table 1). SMR genes were individually subcloned into the 579 multiple cloning region 5' XbaI and 3' HindIII (or 3' PstI) restriction sites of the low-copy vector pMS119EH 580 (50) for IPTG inducible Ptac promoter expression using cloning procedures described by (85). A detailed 581 description of the cloning procedures to construct each lacZ reporter construct with upstream Ptac, Pc, and Pq 582 promoter regions is summarized in File S1 and involved a modified version of the pMS119EH vector. All vectors 583 were constructed and manipulated in E. coli strain DH5α and all vectors were sequence verified using Sanger 584 sequencing services from Eurofins Genomics (MWG Operon CA, USA). Forward and reverse sequencing primers 585 for pMS119EH, Ptac, and Trp repressor regions are listed in File S1. 586 Previously cloned SMR genes from E. coli emrE (86) and sugE (11), were also included in this analysis 587 in pMS119EH, as representative SMP and SUG family members for comparative analysis. Plasmids were 588 transformed into E. coli K-12 wildtype BW25113 strain (51) as well as an efflux pump deficient strain, KAM32 589 (acrAB, mdtK; (52); Table 1), for all AST described below. SMR protein accumulation from extracted cell 590 membranes of plasmid transformed BW25113 and KAM32 strains was performed and confirmed using 16% (T) 591 sodium dodecyl sulfate (SDS)-Tricine PAGE analysis using the E. coli cell membrane extraction procedure 592 described by (56); gel loading details are provided in Fig. S3. 593

Broth microdilution AST method 594
All AST analyses were performed using overnight (18 hr) cultures of each transformant grown in Luria Bertani 595 for AST at 37 o C with shaking at 155 rpm. Overnight cultures were standardized to an optical density at 600 nm 597 (OD 600nm ) of 1.0 unit prior to their final dilutions for each method. Broth microdilution AST of plasmid 598 transformed E. coli strains (BW25113 and KAM32) was performed as described previously (87). Briefly, 599 standardized overnight cultures were diluted 10 -3 into fresh LB-AMP medium containing log2 dilution gradients AMP broth (Fig. S4). Each microplate had wells containing uninoculated media and antimicrobial drug 604 concentration as baseline controls. Growth was based on OD 600nm measurements using a Multiskan Spectrum 605 Ultraviolet /Visible wavelength region microplate reader (ThermoScientific, Waltham, MA). Three biological 606 replicates of each strain were tested in duplicate (n=6) and statistically assessed using two-way student's T-test to 607 determine significant p-value differences as compared to the parental vector. MIC values for each transformant 608 were determined to be the lowest antimicrobial concentration that resulted in a lack of growth (OD 600nm value) that 609 was indistinguishable from uninoculated media containing drug control well. 610

Agar (spot) dilution AST method 611
Agar dilution AST of SMR plasmid transformants (BW25113 or KAM32) was performed as described previously 612 (87) on LB-AMP agar petri plates to maintain expression vectors. Overnight cultures were standardized as 613 described for broth microdilution AST and were diluted to 10 -3 for spot plating. A sterilized stainless steel 48-614 pinned replicator was used to spot 1 l of diluted culture onto LB-AMP agar plates, each plate containing a log2 615 serial dilution of the antimicrobial stock concentration and 0.05 mM IPTG. Agar dilution plates were incubated 616 overnight for 24 hrs and in some cases 48 hrs at 37 o C. Growth of each spot was scored by visual inspection, 617 where the mean MIC values for each transformant was determined to be the minimum drug concentration that 618 resulted in no visible colony growth. Table 2-3. A minimum of three biological replicates were measured for each 619 transformant and compared to the respective strain transformed with the parental pMS119EH vector (Tables 2-3). 620

β-galactosidase reporter assays of planktonic and biofilm cultures. 643
Planktonic LB-AMP broth cultures of plasmid-transformed E. coli BW25113 with placZ, pPc and pPq (listed in 644 Table 1) Figure S1. COBALT alignment of 533 SMR protein sequences detected on proteobacterial plasmids. Sequence 953 information corresponding to each COBALT sequence tag in the alignment 'lcl' is provided in Table S1. 954 955 Figure S2. The final multiple sequence alignment of the 20 unique Qac and SugE protein sequences. Boxed Qac 956 residues highlight amino acid residues that vary when compared between other Qac sequences. Green boxes 957 highlight conserved SugE residues. The alignment was generated in Jalview v2.10.5 [85]. 958 959 Figure S3. 16% (T) SDS-Tricine PAGE analysis of isolated total cell membranes isolated from A) BW25113 and 960 B) and KAM32 SMR transformants to determine SMR protein accumulation. Protein bands were imaged by UV 961 irradiation using 0.5% (v/v) trichloroethanol added to the gels during casting [63] and loading dye was run off the 962 bottom of the gel for 25 minutes to resolve bands shown in both panel images. Panel A shows total membrane 963 protein extracts isolated from BW25113 transformants over expressing each cloned SMR in the study. The total 964 membrane protein amounts loaded in each well was 25µg. Panel B shows total membrane protein extracts isolated 965 from KAM32 transformants overexpressing each cloned SMR in the study. The total membrane protein loaded in 966 each well of the gel in panel B was 25ug. In both panels, cultures grown for membrane extractions reached 0.5 967 units at OD 600nm and were induced to overexpress its respective SMR gene with 0.05 mM IPTG for 3 hrs at 37 o C 968 prior to harvesting cell pellets by centrifugation. Membrane isolations were performed as described by [63], and 969 total membrane proteins concentrations were determined by Modified Lowry Assay. All membrane extract 970 samples were standardized by diluting based on Lowry concentrations in nuclease free water before being mixed 971 into sample loading buffer (100 mM dithiothreitol, 150 mM Tris-HCl pH 7, 12% (w/v) SDS, 30% (w/v) glycerol, 972 0.05% Coomassie Brilliant Blue G250 dye) at a 2:3 sample: loading dye volume ratio prior to loading in wells of 973 gel. 974 975 Figure S4. Growth curves of BW25113 and KAM32 SMR transformants to determine optimal IPTG induction 976 levels. In all panels, final concentrations of IPTG from 0-500 mM were individually added to wells containing 977 standardized (OD 600nm 1.0) 10 -3 diluted cultures in fresh LB-AMP and grown for 24 hrs at 37 o C with shaking in 96 978 well microtiter plates. OD 600 nm values collected over 24 hrs measured in 30 min intervals is shown in each panel.