Characterization of pyridylpiperazine-based efflux pump inhibitors for Acinetobacter baumannii

Abstract Objectives In Acinetobacter baumannii, multidrug efflux pumps belonging to the resistance-nodulation-division (RND) superfamily result in decreased antibiotic susceptibility. Improving the activity of current antibiotics via efflux pump inhibitors (EPIs) represents an attractive alternative approach to control this bacterium. Pyridylpiperazines (PyrPips) are a new class of EPIs that can effectively inhibit the Escherichia coli RND efflux pump AcrAB-TolC and boost the activity of several antibiotics. Here we have evaluated and characterized whether the PyrPip chemical family is also able to boost antibiotic activity through inhibition of the RND efflux pumps in A. baumannii. Methods Comparative structural modelling and docking, structure-activity relationship studies alongside molecular genetic approaches were deployed to improve, characterize and validate PyrPips’ target. Results We showed that two enhanced PyrPip EPIs are capable of rescuing the activity of different classes of antibiotics in A. baumannii. By expressing A. baumannii main efflux pumps (AdeB, AdeG and AdeJ) individually in E. coli recombinant strains, we could gain further insights about the EPIs’ capacity to act upon each pump. Finally, we showed that PyrPip EPIs are mostly acting through AdeJ inhibition via interactions with two key charged residues, namely E959 and E963. Conclusions Our work demonstrates that PyrPip EPIs are capable of inhibiting RND efflux pumps of A. baumannii, and thus may present a promising chemical scaffold for further development.


Introduction
Acinetobacter baumannii is a critical emerging Gram-negative pathogen, now considered by the WHO and CDC as a health priority for research and development (R&D). 1 A. baumannii exploits diverse mechanisms to elude the action of antibiotics, including the use of efflux pumps to extrude a myriad of xenobiotics from within the bacteria. 2,3In A. baumannii, like in many other Gram-negative bacteria, efflux pumps belonging to the resistance-nodulation-division (RND) superfamily play a clinically relevant role in antibiotic resistance. 4Although at least nine have been described for A. baumannii, the AdeABC, AdeFGH and AdeIJK tripartite systems have been primarily linked to loss of antibiotic susceptibility. 4,5he fundamental role of RND efflux pumps in reducing antibiotic susceptibility in Gram-negative bacteria has prompted great interest in developing efflux pump inhibitors (EPIs) over the last 15 years.Primary examples of such discovered and characterized EPIs targeting Gram-negative RND pumps include the peptidomimetic PAβN, [6][7][8] pyranopyridines of the MBX series [9][10][11] and the pyridopyrimidine D13-9001. 12In addition to these EPIs, we recently described and characterized a new class of EPIs, known as pyridylpiperazines (PyrPips), 13,14 that boost antibiotic activity in Escherichia coli by binding a novel allosteric pocket in the transmembrane region of AcrB and inhibiting the activity of the AcrAB-TolC efflux pump.
A. baumannii is known to have low membrane permeability, driven by a combination of porins and efficient efflux pumps that hamper antibiotic activity. 15In A. baumannii, the main efflux pumps (i.e.AdeB, AdeG and AdeJ) have an overall low degree of amino acid sequence conservation when compared with AcrB of E. coli (49%, 40% and 57%, respectively); however, the residues around our previously described PyrPip inhibitor-binding site are highly conserved.In this work, we sought to investigate whether PyrPip could act as an EPI of A. baumannii RND pumps.Validation and characterization of PyrPips' target was achieved by deploying comparative structural modelling and docking, structure-activity relationship studies and molecular genetics approaches.We show that two enhanced EPIs are capable of rescuing the activity of different antibiotic classes in A. baumannii.We report that although our enhanced EPIs act primarily on AdeJ and to a lesser extent on AdeG and AdeB, further efforts are required to improve our knowledge behind the biology of A. baumannii efflux pumps, leading to more potent PyrPip EPIs to help combat antimicrobial resistance mediated by efflux pumps.

Chemical synthesis of pyridylpiperazines
The chemical synthesis can be found in the Supplementary Material (available as Supplementary data at JAC-AMR Online).

Bacterial strains and reagents
Bacterial strains, plasmids and primers used are listed in Tables S1 and S2.Bacteria were regularly cultured on LB broth (BD, DIFCO) or CAMHB at 37°C.When necessary, kanamycin or gentamicin were added to the cultures at a final concentration of 50 mg/L or 10 mg/L, respectively.Antibiotics and PAβN were purchased from Sigma Aldrich, Carbosynth Limited, Fisher Scientific and Euromedex.

MICs
The MICs of antibiotics against A. baumannii were determined according to the CLSI guidelines 16 with slight modifications.Briefly, overnight bacterial cultures were diluted to OD 600 = 0.001 in CAMHB.Where appropriate the bacterial culture was spiked with the EPIs of interest at indicated concentrations.These bacterial suspensions were transferred to a 96-well flat-bottomed microtitre plate (Falcon, France), and a gradient of antibiotics of interest was added by serial dilutions.Plates were incubated at 37 °C for up to 20 h, and bacterial viability was evaluated using the resazurin reduction assay and measured by fluorescence (POLARstar Omega, BMG Labtech: Ex: 530 nm, Em: 590 nm).MICs were defined as the concentration that prevented 90% of resazurin turnover compared with the non-treated bacteria.

Heterologous expression of the Ade efflux pumps in E. coli ΔacrAB
E. coli ΔacrAB, lacking its major RND efflux pump, was generated as previously described. 17To express the A. baumannii RND pumps from the E. coli ΔacrAB chromosome, each pump operon was first cloned in an oriR6K suicide plasmid bearing the E. coli acrR gene and PacrA promoter.The recombinant plasmids were constructed using an NEBuilder assembly kit and the Q5-HF polymerase (NEB, France).The acrR-PacrA DNA fragment was PCR amplified from E. coli ΔacrAB genomic DNA.The adeABC and adeIJK operons were PCR amplified from the genome of A. baumannii AB5075, and adeFGH from A. baumannii ATCC17978.Each assembly mix was transformed into E. coli UCC/pir+ (Lucigen, UK) by electroporation, and kanamycin-resistant clones were selected.Correct plasmid constructs were verified by Sanger sequencing.The generated plasmids were designated pJC1 (PacrA-adeIJK), pJC2 (PacrA-adeABC) and pJC3 (PacrA-adeFGH), and introduced into E. coli ΔacrAB by electroporation, selecting for kanamycin resistance.Correct plasmid integration at the acrR locus was assessed by PCR.To avoid plasmid excision in the absence of kanamycin, we then used Red-recombineering to exchange the oriR6K-KmR region with a gentamicin cassette and thus stabilize each chromosomal construct.

Generation of adeJ efflux pump deletions in A. baumannii
To delete adeJ in A. baumannii ATCC17978, an allelic exchange strategy based on the pMo130 suicide vector was used as described by Amin et al. 18 Briefly, 0.8 kb DNA fragments located up/downstream of the targeted gene were PCR amplified and cloned into pMo130 using an NEBbuilder assembly kit (NEB).Following confirmation of the correct plasmid assembly by Sanger sequencing, the suicide plasmid was transformed into A. baumannii ATCC17978 by electroporation.Single crossover events were selected on kanamycin; upon growth in LB + sucrose 15%, double-crossover kanamycin-sensitive strains were screened in which adeJ knockout was assessed by PCR.Subsequent WGS (microbesNG, Birmingham, UK) confirmed the anticipated genome sequence.

Generation of A. baumannii adeJ point mutants
Point mutations in adeJ were introduced into the A. baumannii ATCC17978 chromosome using the allelic exchange method described previously 18 and above.Briefly, a ∼1 kb DNA fragment centred on the adeJ targeted nucleotide was cloned into pCR-BluntII-TOPO (Fisher Scientific).The resulting plasmid was then used as a template for sitedirected mutagenesis with mutagenic oligonucleotide primers (Table S2) and Q5-polymerase (NEB).DNA fragments confirmed to carry the desired adeJ mutations were then subcloned into pMo130, and ATCC17978 was electrotransformed with the resulting plasmid.The correct allelic exchange was assessed by sequencing a PCR amplicon of the targeted region, followed by WGS (microbesNG, Birmingham, UK).

Quantitative RT-PCR
Bacterial cultures (OD 600 ∼0.05) were incubated at 37°C with shaking (180 rpm) until the OD 600 had reached 0.6-0.8.Bacterial RNA was then stabilized using RNAprotect Bacteria Reagent (Qiagen, France) according to the manufacturer's instructions, and then disrupted with 200 µL lysozyme (20 mg/mL in TE buffer (10mM TriS-HCl (pH 8.0), 1 mM EDTA), 10 min), followed by addition of 700 μL RLT buffer (Qiagen) containing 150 mM β-mercaptoethanol.Cells were further disrupted by beadbeating in lysing matrix B tubes (MP Biochemicals) using a FastPrep-24 (MP Biomedicals) and RNA was purified using a Qiagen RNeasy RNA purification kit according to the manufacturer's instructions.Contaminating genomic DNA was digested using TURBO DNA-free ™ Kit (Ambion), and DNA-free RNA was then reverse transcribed to cDNA using LunaSript ™ SuperMix kit (NEB) and random hexamers.Quantitative RT-PCR entailed using the KAPA SYBR FAST Mix (Sigma-Aldrich, France) and 200 nM of each primer (Table S2), and performed using a LightCycler ™ 480 (Roche).Samples were run as four biological replicates with three technical repeats each from a separate RNA purification batch of cells.Relative RNA concentrations were calculated using the 2^(−ΔCt) method. 19

A. baumannii efflux pump model building and molecular docking
In the absence of an Ade-PyrPip co-structure, potential interactions between the EPIs and the A. baumannii RND efflux pumps were investigated using structural modelling by satisfaction of spatial restraints 20 and molecular docking. 21The L-protomer of the BDM88855-bound AcrB (pdb: 7OUK) 13 was used as a homologue protein template (apo-state).AdeB, G and J were modelled onto this template using python scripts 22 for the MODELLER program, 23 giving five output models with the lowest DOPE and SCORE levels (both underlying the lowest energy barrier required for conformational stability).The top three models (based on scores) were subjected to geometric validation using MolProbity 24 and the best-fit model (versus AcrB L protomer) was used for docking studies with AutoDock using Vina 21

The PyrPip binding pocket is conserved in A. baumannii RND pumps
In A. baumannii, three major RND-type efflux pumps are described to impact antibiotic activity, namely AdeABC, AdeFGH and AdeIJK.The inner membrane components of these tripartite pumps share moderate overall amino acid similarities when compared with AcrB of E. coli; however, the region of the identified PyrPip binding pocket remains highly conserved (Figure 1).The E. coli AcrB-PyrPip co-structure showed that PyrPip binding was associated with significant changes to the TM region, 13 thus similar structural changes could be expected in the Ade proteins.Although the apo-structures of AdeB and AdeJ have been recently solved, 25,26 these structures could not be used for chemical docking as they do not account for the likely required TM movement.Therefore as an initial approach to evaluate whether AdeB, AdeG and AdeJ could accommodate PyrPip EPIs, structural models of each Ade pump were generated based on the published AcrB L-protomer-BDM88885 co-structure (apo structure). 13Molecular docking was then used to evaluate the binding poses of the BDM88855 EPI in each of these models.The three models confirmed a high level of three-dimensional structural identity with AcrB, with only a few amino acid substitutions (Figure 1) and no evidence of obvious steric hindrance or loss of essential binding interactions.Together, the analysis suggested that PyrPips could inhibit these pumps in A. baumannii.

Screening of PyrPips as EPIs in A. baumannii
Evaluation of the original PyrPip EPI hit (BDM73185) on A. baumannii ATCC17978 found it to be a mild booster (2-4-fold) of chloramphenicol, triclosan and linezolid antibiotic activity.Medicinal chemistry optimization of BDM73185 to more potent PyrPip analogues against E. coli AcrB 13 provided an extended PyrPip chemical library (>200 compounds) that was also evaluated in A. baumannii.This library was assessed in dose-response against A. baumannii in the presence of sub-MIC concentrations of efflux pump substrate antibiotics (chloramphenicol: 50 mg/L; fusidic acid: 50 mg/L; novobiocin: 5 mg/L; erythromycin: 2.5 mg/ L; azithromycin: 0.15 mg/L; and ciprofloxacin: 0.2 mg/L).Globally, this phenotypic screen clearly showed a disconnect between PyrPip potency observed in E. coli compared with that observed in A. baumannii.Overall, far fewer PyrPips were found to potentiate antibiotic activity, additionally encompassed by a lower potency.This finding suggested that despite no obvious difference in the PyrPip binding pocket of the Ade pumps, the structure-activity relationships governing the boosting of antibiotic activity were different to those established for E. coli AcrB.
Despite the overall inferior activity of PyrPips in A. baumannii, a total of 97 compounds were identified to potentiate the activity of at least one antibiotic by at least one dilution (2-fold), whereas 33 had activity to more than one antibiotic by at least two dilutions (4-fold).In general, we observed a high frequency of antibiotic boosting for fusidic acid, novobiocin and erythromycin, but to a lesser extent for azithromycin and ciprofloxacin.The chemical structure of active PyrPips in A. baumannii included four compounds with a quinoline core (compounds 1-4 in Table 1) and five with a pyridine core (compounds 6-10).The three most potent EPIs from the quinoline series were close analogues with meta-(compound 1), ortho-(compound 2) and para-(compound 3) benzylamine at position 6 of the quinoline ring (Table 1).A butanediamine substituent (compound 4) also allowed for boosting of antibiotic activity, though this compound was less potent.Regarding the pyridine series, three compounds (compounds 6-8) bearing an ester linker in position 5 of the pyridine ring were able to boost fusidic acid and novobiocin, though the most potent ester (compound 8) also showed intrinsic antibacterial activity of 62 µM.The two most potent boosters in this series were analogues bearing an ethynylphenyl moiety with either an ethylamine chain in the meta-(compound 9) or para-(compound 10) position.The initial chemical library did not contain the ortho-substituted ethylamine analogue, and so it was synthesized (compound 11), and found to be 3-to 6-fold more potent at boosting activity (EC 90 = 10 µM) than analogues 9 (EC 90 = 31 µM) and 10 (EC 90 = 62 µM).
A common characteristic among the most potent boosters, BDM91531 (compound 1) and BDM91892 (compound 11), is that they share a primary amine moiety.Replacing this group with another hydrogen bond-donating group, such as an alcohol, led to inactive compounds (compound 5, EC 90 ≥ 125 µM; compound 12, EC 90 > 62 µM), suggesting that the basic amine could interact with one of the acidic residues of the binding pocket via a salt bridge.In E. coli, both BDM91531 and BDM91892 were validated to inhibit AcrB (EC 90 = 0.24 and 1.5 µM, respectively) as PyrPip resistance mutations (S446P or A450P substitution in AcrB) 13 prevented antibiotic boosting activity of both compounds (EC 90 = 32 µM) (Table S3).

PyrPips inhibit A. baumannii RND pumps heterologously expressed in E. coli
Because A. baumannii expresses multiple RND efflux pumps, we studied the antibiotic boosting activity of PyrPips BDM91531 (compound 1) and BDM91892 (compound 11) on individual A. baumannii RND pumps using a heterologous expression system in E. coli lacking AcrAB.The adeABC, adeFGH or adeIJK operons were cloned into the E. coli ΔacrAB chromosome under the native promoter of acrA, to mediate a natural level of protein expression to PyrPip inhibition (pump overexpression could reduce EPI potency due to target overexpression).Quantitative PCR (qPCR) analyses confirmed that all three pumps were expressed at a comparable level in the recombinant strains (Table S4) although we recognize that this does not guarantee the correct and or expected assembly of the tripartite pumps in the E. coli membranes.Nonetheless, whereas expression of adeABC or adeFGH had no obvious impact on the growth rate, the adeIJK recombinant strain showed a moderate growth defect (Figure S1), similar to that previously reported. 5,27,28nitially, the impact of Ade efflux pump expression in E. coli ΔacrAB was determined by susceptibility to a selected panel of antibiotics (Table 2).Overall, heterologous expression of adeABC had limited impact on antibiotic susceptibility, which may be due to inefficiencies in AdeABC pump assembly in E. coli or to intrinsic differences in E. coli and A. baumannii cell wall architecture preventing proper functioning of this specific RND pump.This lack of AdeABC-related phenotype prevented the analysis of PyrPip inhibition on this RND pump.On the other hand, expression of adeFGH increased resistance to chloramphenicol and oxacillin (as well as the detergent SDS), whereas expression of adeIJK increased resistance to fusidic acid, novobiocin and erythromycin.These antibiotic profiles are largely in agreement with previous studies where these pumps were overexpressed, though the magnitude of resistance differs probably due to differences in expression level. 29,30ext, we challenged E. coli ΔacrAB and recombinant strains expressing A. baumannii RND efflux pumps in the presence of 40 µM BDM91531, BDM91892 or PAβN (Table 2).Results showed that all three molecules were able to increase antibiotic susceptibility.In particular, chloramphenicol and oxacillin resistancemediated by adeFGH expression-was affected.Moreover, resistance to fusidic acid, novobiocin and erythromycin-mediated by PyrPip efflux pump inhibitors for A. baumannii adeIJK expression-was also reduced.In all cases, PyrPips were also able to decrease SDS resistance mediated by adeFGH and adeIJK expression.As adeABC expression did not give a clear resistance phenotype, it was not possible to evaluate the efficacy of the EPIs on this pump.Together, these data strongly support that both BDM91531 and BDM91892 can inhibit the activity of AdeFGH and AdeIJK when expressed in E. coli.

Spectra of BDM91531 and BDM91892 activity on A. baumannii antibiotic susceptibility
To characterize the spectra of antibiotic boosting by BDM91531 and BDM91892 in A. baumannii beyond those used in the PyrPips' screening, we increased the antibiotics panel to 16 and included PAβN as an EPI control.Results largely reproduced the antibiotic boosting observed during the PyrPips' screening (i.e.chloramphenicol, novobiocin, erythromycin and fusidic acid), but also showed boosting of linezolid, aztreonam and trimethoprim (and piperacillin and oxacillin by BDM91531 only); however, no boosting was observed for levofloxacin, amikacin, gentamicin, cefepime and streptomycin by any of the inhibitors (Figure 2).When comparing these data with observations from the heterologous expression system in E. coli ΔacrAB, it suggests that the observed antibiotic boosting in A. baumannii by BDM91531, BDM91892 or PAβN is through AdeIJK inhibition.

Target confirmation of PyrPips in A. baumannii
Because A. baumannii encodes at least three RND pumps that are described to impact directly antibiotic susceptibility-though they are not equally expressed-we evaluated the relative expression of adeB, adeG and adeJ by qPCR.Data showed that A. baumannii primarily expresses adeIJK [average cycle threshold (Ct) = 19], followed by adeABC and adeFGH at much lower levels (average Ct = 24.35 and 26.40, respectively) (Table S4).Although transcriptional analysis does not fully reflect protein production levels, results are in line with previous data establishing AdeIJK as the predominant RND pump in A. baumannii strains. 5tructure-activity relationship data on PyrPips showed the primary amine moiety of BDM91531 and BDM91892 to be beneficial for potency.In fact, recent studies on E. coli AcrB showed that two acidic residues at the entrance of the PyrPip binding pocket in AcrB (E947 and D951) are crucial for PyrPip activity. 14Indeed, substitutions of any of these residues in AcrB led to a shift in potency for compounds BDM91531 and BDM91892, whereas PAβN activity was unaffected (Table S3). A. baumannii AdeJ also contains acidic residues at the equivalent positions (E959 and E963) (Figure 1).In the absence of an Ade-PyrPip co-structure or molecular dynamics data, in silico docking studies of BDM91531 and BDM91892 into the modelled AdeJ were deployed to evaluate PyrPip binding.The results showed that both EPIs could fit into the PyrPip binding pocket of AdeJ while making potential polar interactions between the PyrPip primary amine and the AdeJ acidic residues, E959 and E963.In addition to conservation of interactions with TM4 and TM5, a partial perpendicular π-stacking was also observed with AdeJ residue F960.Free energy (ΔG) calculations for each EPI suggested BDM91531 to have greater binding affinity to AdeJ compared with BDM91892 (DG • = −9.8 and −6.6 kcal/mol, respectively; Figure 2d).PyrPip efflux pump inhibitors for A. baumannii [exhaustiveness of 4096, grid dimensions 50 Å × 50 Å × 50Å spanning transmembrane domains (TMs) 4, 5, 10 and 11].Compounds were considered flexible during docking (number of rotatable bonds in the range 2 to 4).The top eight docking poses were kept to be used in structure-activity relationship (SAR) analysis.

Figure 1 .
Figure 1.The PyrPip binding pocket in A. baumannii AdeB, G and J compared with that of AcrB in E. coli.(a) protein sequence alignment of the PyrPip binding pocket region [transmembrane domains (TMs) 4, 5 and 10] in AcrB, AdeB, AdeG and AdeJ.Residues that interact with PyrPip in AcrB (in blue) are largely conserved (*).(b) In silico docking of BDM88855 into the PyrPip binding pocket modelled for AdeB, AdeG and AdeJ [modelling based on the Apo structure of the L-protomer of AcrB:BDM88855 (pdb: 7OUK)].The structural models showed no evidence of steric hindrance and highlighted the predicted conservation of PyrPip binding interactions with the Ade efflux pumps.Dotted lines: hydrophobic interactions; green lines: halogen interactions; blue lines: polar interactions.
were added at a final concentration of 40 µM.Bold numbers show when Ade pump expression increased antibiotic resistance at least 2-fold.Antibiotic susceptibility was measured in liquid culture by resazurin microtiter assay (REMA).Data depict the mean MIC of at least three separate biological replicates.91531 = BDM91531, 91892 = BDM91892.Jiménez-Castellanos et al.

Figure 2 .
Figure 2. (a) MIC (mg/L) for a panel of antibiotics in the presence or absence of 40 µM EPI (BDM91531, BDM91892 or PAβN) against A. baumannii (WT) and isogenic strain with AdeJ E959Q-E963Q substitution.Conditions where EPIs decreased antibiotic MIC are indicated in grey-shaded cells (≥2-fold).Conditions where EPIs were less efficient at antibiotic boosting in the adeJ mutant compared with in the WT are highlighted in blue.Values are the mean MICs of at least three biological replicates.(b and c) Structural modelling highlighting protein-ligand interactions of the AdeJ efflux pump with either BDM91531 or BDM91892.Dotted lines: hydrophobic interactions; green line: halogen interactions; blue lines: polar interactions.(d) Free energy values for ligand affinity.

Figure 3 .
Figure 3. Checkerboard assay showing the synergy between novobiocin activity and the EPIs: BDM91531, BDM91892 or PAβN anti-A.baumannii.Susceptibility studies were performed on A. baumannii ATCC17978 WT strain and isogenic strain with AdeJ E959Q-E963Q substitution.Values indicate the bacterial viability expressed as the mean percentage of resazurin reduction assay (yellow cells >10%, blue cells <10%).Data represent the median bacterial viability of at least three biological replicates.

Table 1 .
Structure and activity of the PyrPip EPIs with a quinoline core (compounds 1-5) or pyridine core (compounds 6-10), identified through screening to boost the activity of at least two antibiotics by 4-fold in A. baumannii a