Interchangeability of periplasmic adaptor proteins AcrA and AcrE in forming functional efflux pumps with AcrD in Salmonella enterica serovar Typhimurium

Abstract Background Resistance-nodulation-division (RND) efflux pumps are important mediators of antibiotic resistance. RND pumps, including the principal multidrug efflux pump AcrAB-TolC in Salmonella, are tripartite systems with an inner membrane RND transporter, a periplasmic adaptor protein (PAP) and an outer membrane factor (OMF). We previously identified the residues required for binding between the PAP AcrA and the RND transporter AcrB and have demonstrated that PAPs can function with non-cognate transporters. AcrE and AcrD/AcrF are homologues of AcrA and AcrB, respectively. Here, we show that AcrE can interact with AcrD, which does not possess its own PAP, and establish that the residues previously identified in AcrB binding are also involved in AcrD binding. Methods The acrD and acrE genes were expressed in a strain lacking acrABDEF (Δ3RND). PAP residues involved in promiscuous interactions were predicted based on previously defined PAP-RND interactions and corresponding mutations generated in acrA and acrE. Antimicrobial susceptibility of the mutant strains was determined. Results Co-expression of acrD and acrE significantly decreased susceptibility of the Δ3RND strain to AcrD substrates, showing that AcrE can form a functional complex with AcrD. The substrate profile of Salmonella AcrD differed from that of Escherichia coli AcrD. Mutations targeting the previously defined PAP-RND interaction sites in AcrA/AcrE impaired efflux of AcrD-dependent substrates. Conclusions These data indicate that AcrE forms an efflux-competent pump with AcrD and thus presents an alternative PAP for this pump. Mutagenesis of the conserved RND binding sites validates the interchangeability of AcrA and AcrE, highlighting them as potential drug targets for efflux inhibition.


Introduction
MDR efflux pumps play a major role in antibiotic resistance of bacteria by reducing the intracellular concentration of drugs. 1,2 The resistance-nodulation-division (RND) family of efflux pumps confer clinically relevant antibiotic resistance on Gram-negative bacteria, such as Salmonella enterica. [3][4][5][6] Tripartite RND pumps span the double membrane of Gram-negative bacteria and consist of an inner membrane RND transporter, a periplasmic adaptor protein (PAP) and an outer membrane factor (OMF). [7][8][9] The majority of RND pumps exhibit a broad substrate profile, which includes multiple classes of antibiotics, biocides, detergents, dyes and metals. [10][11][12] S. enterica has five RND pumps: AcrAB, AcrD, AcrEF, MdtABC and MdsABC. 5 The AcrAB pump is constitutively expressed in S. enterica and displays a wide substrate profile, consisting of multiple classes of antibiotics, bile salts, detergents and dyes. 5 The AcrEF system has a similar substrate profile to AcrAB but is not constitutively expressed. 5,13 In S. enterica, AcrB is 80% identical to AcrF, whereas AcrD is 64% and 65% identical to AcrB and AcrF, respectively. 14,15 This sequence divergence is reflected in the substrate profile of AcrD, which is narrower than that of AcrB and AcrF. In Escherichia coli, AcrD has been shown to export aminoglycosides and anionic b-lactams. [16][17][18] Homology modelling of E. coli AcrD, combined with molecular dynamic simulations, has also suggested that the different substrate specificities between AcrB and AcrD stem from the corresponding differences in the physicochemical and topological properties of their binding pockets. 19,20 Until now, this view of AcrD substrate selectivity has been assumed to also apply to the AcrD pump in S. enterica.
The RND transporter genes are usually co-located with their cognate PAP in a single operon. In S. enterica there are four RNDassociated PAPs: AcrA, AcrE, MdtA and MdsA. 5 Based on sequence analyses and structural alignments, AcrA and AcrE are the most closely related, with an amino acid sequence identity of 69.3% (Figure 1a). The predicted secondary structure of AcrA and AcrE is nearly identical to that of the experimentally determined structure of AcrA from E. coli (Figure 1b). 21 Both MdtA and MdsA are more sequentially divergent, with MdsA sharing less than 30% identity with AcrA and AcrE, which is predicted to translate into significant differences in the 3D structure. 5 Although AcrA is the cognate PAP for AcrB, the RND pump AcrD was shown to depend on AcrA to form a functional tripartite efflux system since it lacks an associated PAP-encoding gene. 22 Indeed, AcrA has been reported to also function with AcrF in E. coli 23 and recently, AcrE has been demonstrated to function with AcrB in S. enterica serovar Typhimurium. 21 The major RND transporterbinding residues of AcrA have been highlighted by cryogenic electron microscopy structural studies 8,9 and validated by mutagenesis. 21 Our comparative analysis of Salmonella PAPs demonstrated that these critical residues fall within a discrete number of linear sequence sites, which we termed RND-binding boxes. 21 These are shared between AcrA and AcrE, potentially explaining their interchangeability 21 but are not shared between MdtA and MdsA, explaining their inability to function with non-cognate RND transporters. 21 Although AcrA and AcrE have been shown to be largely interchangeable, the ability of AcrE to function with AcrD remains unknown.
Here, we have investigated the substrate specificity of S. Typhimurium SL1344 AcrD. We furthermore explored whether the interoperability of AcrA and AcrE extends to the RND transporter AcrD and whether this interaction is driven by the same residues that have been shown to be important for other PAP and RND combinations.

Bacterial strains
All strains used in this study are listed in Table 1. The S. Typhimurium strains were derived from the WT strain SL1344, a pathogenic strain first isolated from an experimentally infected calf. 24 All strains were grown in LB broth at 37 C with aeration.

Construction of gene deletion mutants
The DacrB mutant strain was constructed previously. 25 All other mutant strains were constructed using the k red recombinase system described previously, antibiotic markers were removed, and the process repeated to make double-and triple-knockout S. Typhimurium SL1344 strains (Table  1). 26 All the primers used for generating gene knockouts and cloning are listed in Table S1, available as Supplementary data at JAC Online.

Plasmid construction
All plasmids used in this study are listed in Table 1. The acrD and acrA genes were amplified from S. Typhimurium SL1344 by PCR and cloned into pHSG398 and pACYC177 plasmids, respectively, as described previously. 22 Expression of the acrE gene is repressed by H-NS. 13 Therefore, to clone acrE into pACYC177 and obtain sufficient expression, a forward primer was designed containing the trc promoter and the acrE ribosomal binding site (Table S1). The synthetic trc promoter is derived from the E. coli trp and lacUV5 promoters, driving a high level of transcription. 27 The acrE gene was amplified from S. Typhimurium SL1344 genomic DNA by PCR using the acrE cloning F and R primers (Table S1), which introduced ScaI and BamHI sites, respectively. The PCR fragment contained the trc promoter and a region 14 bp upstream to 2 bp downstream of acrE. This fragment was digested with ScaI and BamHI and cloned into the corresponding sites of pACYC177, where an ampicillin resistance gene was located. The resulting plasmid pACYC177 acrE solely possessed a kanamycin resistance marker. The control pACYC177Kan R plasmid was constructed as described previously. 22

Construction of mutant pacrA and pacrE plasmids
The acrA and acrE point mutants were generated using the GeneArt s R Gene Synthesis Service (Invitrogen, Germany) and subsequently cloned into the pACYC177 plasmid using the Subcloning Service (Invitrogen). All plasmids were sequenced to check for the presence of the desired point mutations and to ensure there were no unwanted secondary mutations.

Determination of antimicrobial susceptibility
The MICs of various antimicrobials were determined using the agar dilution method according to CLSI guidance. 28

AcrD of S. Typhimurium SL1344 does not transport aminoglycosides
Despite being isolated several decades ago, 29 the substrate specificity of AcrD remains relatively poorly characterized experimentally. Therefore, an additional rationale of this study was to investigate the substrate specificity of S. Typhimurium SL1344 AcrD, especially in the context of PAP-RND interactions, which may provide modulatory effects on the specificity of the pump. Previously, it has been reported that E. coli AcrD exports aminoglycosides. 17,18 However, there is a lack of experimental evidence in Salmonella and most of the features of Salmonella AcrD are inferred, based on close sequence similarity to E. coli AcrD (97.4%; Figure S1). While some previous work has addressed this, aminoglycosides have not been specifically investigated. 5,15 Therefore, we investigated the substrate range of AcrD in S. Typhimurium SL1344.
The DacrD SL1344 strain did not exhibit any significant increase in susceptibility to any of the antimicrobials tested, as previously reported. 15 This is likely because expression of acrD is generally low in laboratory conditions and, for many compounds, any effect would be masked by the presence of AcrB. 15 Therefore, pacrD was transformed into the DacrB strain, and the effect of acrD overexpression on antimicrobial susceptibility of the resulting transformant was determined. The DacrB/pacrD strain displayed significantly increased MIC values of reported AcrD substrates aztreonam, carbenicillin, cloxacillin, fusidic acid, nafcillin, novobiocin and oxacillin (Table 2), consistent with previous studies, 22 suggesting that protein is functionally expressed and incorporated into the membrane. Surprisingly, the introduction of acrD into the DacrB (DacrB/pacrD) strain did not result in a significant increase in MICs of the aminoglycosides kanamycin, gentamicin, spectinomycin or streptomycin ( Table 2), implying that AcrD is not measurably contributing to aminoglycoside efflux. This is in contrast to the reported role of AcrD in the aminoglycoside resistance of E. coli, AcrD can function with AcrA or AcrE JAC Figure 1. (a) A pairwise sequence alignment of AcrA and AcrE of S. enterica highlighting their predicted close structural homology. The top secondary structure is derived from the previously reported homology model of AcrE, 21 while the bottom secondary structure corresponds to the experimental AcrA structure from E. coli (PDB ID 5O66; chain G), which has no sequence gaps with the AcrA of S. enterica. (b) Mapping the sequence differences between the Salmonella AcrE and AcrA onto the homology model of the AcrE. 21 The non-conserved substitutions are shown in sidechain and semi-transparent sphere representation. The mapping demonstrates that the bulk of the discrepancies, which may be expected to account for the functional differences between the PAPs, map to their b-barrel and membrane-proximal domains. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.
wherein deletion of acrD was shown to decrease aminoglycoside MICs 2-to 8-fold 17 and expression of acrD from a plasmid in an acrB::aph DacrD strain increased aminoglycoside MICs 2-fold. 30 In agreement with our findings, the AcrD efflux pump of the Gram-negative plant pathogen Erwinia amylovora has also been reported to not play a role in aminoglycoside resistance. 31 A possible explanation for the differences in the substrate profiles of AcrD between E. coli and S. Typhimurium could be the observed discrepancy between the residues in their respective access and deep binding pockets ( Figure S1). Due to the lack of experimental AcrD structure, the functional significance of the residues of the respective drug-binding pockets of AcrD is inferred from their positional homology with corresponding AcrB residues, structures of which have been experimentally defined for both E. coli 32-34 and more recently for Salmonella. 35 Specifically, the presence of a serine in the deep binding pocket of S. Typhimurium AcrD at position 610, which in E. coli AcrD is occupied by an alanine, could possibly impact the previously described lipophilic character of the drug-binding cavity. 20 There are also two additional discrepancies, which could be seen as non-conservative substitutions, namely that of E. coli AcrD isoleucine to phenylalanine at position 633 (I633F) in S. Typhimurium, and leucine to glutamine at position 565 (L565Q), both of which are likely to cause steric hinderance and impact the electrostatics of the access binding pocket, respectively. 19,20 These subtle differences may account for the notable differences in substrate recognition by AcrD between the two species. AcrD can function with AcrA or AcrE

AcrE forms a functional PAP-RND pair with AcrD
AcrD has been previously shown to depend on AcrA to function as an efflux system. 22 Therefore, owing to the high similarity of the predicted RND-binding sites between the PAPs AcrA and AcrE, 21,36 we hypothesized that AcrE should also function with AcrD. To test this, we deleted the acrAB, acrD and acrEF genes in S. Typhimurium SL1344 to create a strain without active RND-dependent efflux, as indicated by significantly increased susceptibility to AcrB, AcrF and AcrD substrates ( Table 2 and Table S2). The MdtABC and MdsABC systems are much less similar to the three AcrB/AcrD/AcrF-based systems and play a minor role in resistance. Consistent with this, they are not expressed under standard laboratory conditions 5 and, furthermore, their inactivation did not have any additive effect on antimicrobial susceptibility. 5,21 Hence, these systems were not inactivated. Firstly, we validated the previously reported AcrA dependency of AcrD in S. Typhimurium SL1344. 22 The pacrA and pacrD plasmids were co-transformed into the D3RND strain and the antimicrobial susceptibility of the resulting transformant was determined. We found that co-expression of acrA and acrD in the D3RND strain significantly decreased susceptibility to known AcrD substrates aztreonam, carbenicillin, cloxacillin, fusidic acid, nafcillin, novobiocin, oxacillin and ticarcillin (Table 2).
Secondly, to determine whether AcrE and AcrD form a functional complex together, pacrD and pacrE were co-transformed into the D3RND strain and the susceptibility to validated AcrD substrates was tested. Co-expression of acrE and acrD in the D3RND strain significantly increased the MICs of aztreonam, carbenicillin, cloxacillin, fusidic acid, nafcillin, novobiocin, oxacillin and ticarcillin (Table 2). There was no difference in MIC values between co-expressing acrD with acrA or acrE, which demonstrates the full interchangeability of the two PAPs (Table  2). Furthermore, co-expression of either acrE and acrD or acrA and acrD in the D3RND strain did not increase MIC values of the tested AcrB substrates (i.e. acriflavine, crystal violet, ethidium bromide, erythromycin, methylene blue, rhodamine 6G and tetracycline), 5 clearly showing AcrD-mediated efflux (Table S2). Overexpression of either acrD or acrE alone in the D3RND strain did not significantly increase MIC values of the AcrD substrates tested ( Table 2), signifying that AcrE requires the presence of AcrD to form a functional, efflux-competent complex. Alav et al.
Our data suggest interchangeability between AcrA and AcrE in S. Typhimurium SL1344. One possible explanation for the interoperability between AcrA and AcrE is that the latter may function as a backup PAP for when AcrA function is impaired or lost. This idea is supported by evidence from studies demonstrating that in S. Typhimurium, in the absence of acrA, it was possible to select for acrE overexpression. 21,37 Another study demonstrated that in the absence of acrA and acrE, it is possible to restore the phenotypic defect in active efflux by complementing with either acrA or acrE. 36 Disruption of the RND-binding residues in AcrA or AcrE impairs AcrD-mediated efflux of substrate drugs AcrE has been shown to complement AcrA and is capable of interacting with AcrB. 38 Recently, we showed that this promiscuity between Salmonella AcrA and AcrE stems from the highly conserved RND-binding sites (termed RND-binding boxes) between these two PAPs. 21 Specifically, within the Salmonella AcrA, we identified several residues mapping to the b-barrel and membrane proximal domains that were important for AcrB binding. 21 There, the disruption of the F292 or G363 residues in AcrA produced the most pronounced phenotypic effect, resulting in severely abrogated active efflux and significantly increased susceptibility to AcrB substrates. 21 Therefore, to investigate whether these residues are also important for binding of the newly determined cognate PAPs to AcrD, the point mutations corresponding to F292G or G363F were constructed in both pacrA and pacrE (F291G and G362F, respectively) and co-transformed with pacrD into the D3RND strain. Based on structural analysis, we chose F292G and G363F as target mutations due to their radical change of respective sidechain properties.
Consistent with the data obtained in co-expression with AcrB, 21 the disruption of F292 or G363 in AcrA resulted in impaired AcrDmediated efflux of AcrD substrates, confirming that the same residues required for binding of AcrA to AcrB are also required for its binding to AcrD (Table 2). These point mutations do not impact the protein levels and folding, as previously demonstrated. 21 To determine whether the corresponding residues in AcrE are also important for AcrD binding, F291 and G362 were mutated ( Figure S2). As expected, the F291G or G362F point mutations in AcrE also impaired AcrD-mediated efflux in the D3RND strain (Table 2). These data suggest that the PAP-RND binding sites previously identified based upon AcrA-AcrB interaction are indeed both sequentially and functionally conserved between AcrA and AcrE and account for the productive recognition and formation of functional tripartite pumps.

Conclusions
Here, we report that the PAP AcrE can form a functional complex with the RND transporter AcrD, further validating the interchangeability between the homologous PAPs AcrA and AcrE. Furthermore, this interchangeability is likely to be due to the highly conserved and specific RND-binding sites between these two PAPs. Our report highlights that the redundancy between these two PAPs must be taken into account when targeting them for efflux inhibition. Therefore, the residues we identified here could inform future design of effective efflux inhibitors targeting PAPs or tripartite complex assemblies.