Proximal Binding Pocket Arg717 Substitutions in Escherichia coli AcrB Cause Clinically Relevant Divergencies in Resistance Profiles

ABSTRACT Recent mutations in RND efflux pumps in clinical strains have further increased multidrug resistance. We show that R717L and R717Q substitutions (found in azithromycin-resistant Salmonella enterica spp.) in the Escherichia coli efflux pump AcrB dramatically increase macrolide, as well as fluoroquinolone, resistance. On the other hand, cells became more susceptible to novobiocin and β-lactam cloxacillin. We urge the control of, and adjustments to, treatments with antibiotics and the need for novel antibiotics and efflux pump inhibitors.

MIC results of 20 compounds. Compared to wild-type AcrB, R717Q/L give divergent resistance spectra. MICs generally increased for macrolides and fluoroquinolones but decreased for novobiocin (NOV) and b-lactam cloxacillin (CLX). For clarithromycin (CLR) and erythromycin (ERY), the MICs of R717Q/L were .256 and 256 mg/mL, respectively, compared to 128 mg/mL for the wild type. The MIC for AZM increased significantly by 4-and 8-fold for R717Q and R717L (64 and 128 mg/mL, compared to 16 mg/mL for wildtype AcrB). Mutant cells grew as full and healthy colonies on the agar plates at their highest viable macrolide concentrations. Ofloxacin (OFX), levofloxacin (LVX), and moxifloxacin (MXF) MICs increased 2-fold. For R717Q, the monobactam aztreonam (ATM) MIC increased by 4-fold. R717L increased the ethidium bromide (EtBr) MIC 2-fold. Interestingly, R717Q and R717L both caused 2-fold NOV and CLX MIC decreases.
To compare the R717Q/L mutations, we introduced other clinical substitutions in AcrB-Ec, found in N. gonorrhoeae MtrD and S. Typhimurium AcrB. In MtrD, K823E/N were observed in macrolide-resistant strains (22)(23)(24)(25). In AcrB, G288D conferred increased fluoroquinolone resistance in a clinical strain (26). Therefore, we introduced the substitutions G288D, E826K, and E826N in AcrB-Ec. Lys823 in MtrD corresponds to Glu826 in AcrB-Ec Two drug molecules are present in the proximal binding pocket (rifampicin and erythromycin). The Arg717 residue and the R717Q/L substitutions are located at the entrance of the pocket. (c) Close-up inwards view of the proximal binding pocket and Arg717 area. Rifampicin can be seen behind the Arg717 residue and the R717Q/L substitutions. CH1-4, Channel 1-4. Colors: red box, substitution area; yellow spheres, rifampicin; orange spheres, erythromycin; green sticks, Arg717; orange sticks, R717Q; red sticks, R717L. PDB accession codes: 4DX5 (for the main structure and the mutagenesis); 3AOC (for the erythromycin coordinates); 3AOD (for the rifampicin and the wild-type Arg717 side chain coordinates).
Solid plate MICs (Table 1) showed an up to 8-fold increase in macrolide MICs for R717Q/L mutants. Furthermore, clinically relevant G288D generally caused decreased MICs. The reversed substitution E826K decreased multiple drugs' MICs. Thus, the 8-fold AZM, the 2-fold ERY, and the $4-fold CLR MIC increases show a significant macrolide export gain-of-function for AcrB-Ec by the R717Q/L substitutions. Therefore, we checked the growth of R717Q/L cells in macrolide, CLX, and NOV supplemented liquid medium. Fig. 2 (macrolides) shows that wild-type cell growth is significantly inhibited at 256 mg/mL and completely inhibited at 512 mg/mL ERY (top lane). Contrarily, R717Q/L mutants grow fully at 256 mg/mL and significantly grow even under 512 mg/ mL. Similarly, for CLR (middle lane), wild-type cell growth is inhibited at 256 mg/mL, while mutant cells can grow fully. At 512 mg/mL, mutant cells still slightly grow. For AZM (bottom lane), wild-type cell growth was already partly inhibited at 16 mg/mL and fully at 32 mg/mL. Contrarily, mutants could fully grow, even in 64 mg/mL (even in 128 mg/mL, the mutants seem very slightly viable). These results corroborate the significant resistance increase caused by R717Q/L for macrolides, especially AZM. Under all conditions, R717L seemed somewhat more viable than R717Q. Fig. 3 shows inhibited cell growth for R717Q/L, compared to the wild type, in CLX and NOV supplemented medium, corroborating the plate MIC results (Table 1).
To further check the clinical significance, we performed Kirby-Bauer disk diffusion susceptibility tests. Table 1 shows R717Q/L also increased fluoroquinolone resistance, besides increased macrolide resistance. S. Typhi strains from Pakistan and Nepal show that R717Q/L cause reduced fluoroquinolone susceptibility; however, they note a double mutation in gyrA contributes to this phenotype (15,16). Nonetheless, we wanted to further investigate the consequences of R717Q/L in AcrB-Ec on fluoroquinolone and macrolide resistance. Fig. 4 shows the disk diffusion susceptibility results for macrolides, fluoroquinolones, NOV, and minocycline (MIN). When AcrB-Ec is expressed, the inhibition zones for all compounds decrease (Fig. 4, WT versus KO). R717Q/L mutations significantly decrease the inhibition zones further for all macrolides and fluoroquinolones (visible by the naked eye in Fig. 4a, quantified in Fig. 4b). R717L has slightly more impact on the inhibition zone decreases than R717Q (Fig. 4b). Interestingly, R717L also decreases the inhibition zone for MIN. Similar to Table 1 and Fig. 3, R717Q/L increased NOV susceptibility ( Fig. 4a and b).
We showed that R717Q/L substitutions in AcrB-Ec confer significant increased macrolides resistance, with an up to 8-fold increase in MICs. These findings corroborate the phenotypes of AZM-resistant S. Typhi and Paratyphi A strains harboring the R717Q/L mutations (19)(20)(21). Interestingly, R717Q/L caused a 2-fold MIC decrease for CLX and NOV. This could (partly) explain the susceptibility of AZM-resistant S. Typhi R717Q strains from Pakistan, which were still susceptible to third-generation cephalosporins   (16). According to the EUCAST database, the theoretical AZM epidemiological cut-off value for E. coli is 16 mg/mL (34). Our wild-type E. coli MG1655 strain has an AZM plate MIC of 16 mg/mL, and R717Q/L mutants 64 and 128 mg/mL, respectively (Table 1). Mutants also have an AZM liquid MIC of 128 mg/mL (Fig. 2). These substitutions caused a hydrophilicity decrease at the entrance of the PBP (Fig. 1, 5, S5). Gln is polar while Leu is hydrophobic; still, both substitutions cause macrolide MIC increases. Macrolides are hydrophobic molecules; thus, the hydrophilicity decrease (and the removal of a positive charge) partly explains the increased resistance. Additionally, the side chains are significantly shorter for R717Q/L, enlarging the PBP entrance (Fig. 1c, Fig. 5), further explaining the enhanced efflux of bulky macrolides (13). We showed that bulky Trp substitutions at the PBP entrance decreased ERY MICs significantly (14), corroborating  the impact of space-limiting substitutions on macrolide-export. The increased space and the increased hydrophobicity can explain why R717L seems slightly more active in exporting macrolides than R717Q (Table 1, Fig. 2). Furthermore, the inhibition zone for R717L was also significantly smaller than for R717Q for fluoroquinolones and MIN (Fig. 4). Basically no difference was found between the mutants on the growth under CLX, while R717L seemed slightly more viable than R717Q under NOV (Fig. 3).
As explained, besides increased macrolide and fluoroquinolone resistance, we observed decreased MICs for CLX and NOV. Therefore, it may be clinically interesting to combine multiple antibiotics to treat typhoid and paratyphoid fever to mitigate resistance and enhance treatment. For example, a combination of b-lactams and AZM may enhance the treatment of Salmonella infections. Additionally, we observed this significant gain-of-function in E. coli AcrB for the first time, showing that these mutations in other pathogenic bacteria may significantly affect clinical treatment options. These results further imply the importance of adjusted antibiotics treatments, and the need for novel antibiotics and efflux pump inhibitors.
E. coli MG1655 (35) was used to create DacrB (NKE96) (36) by gene deletion (37). Bacterial strains were grown at 37°C in Luria-Bertani broth (38). His-tagged AcrB was expressed from pBAD33acrBhis, and point mutations were introduced by PCR (GenScript). Susceptibility testing in liquid and on solid media was determined by adding the toxic compounds by serial dilutions. Cell cultures (supplemented with 10 mM arabinose) were incubated until OD 600 nm 0.6 and diluted to 0.05. For liquid growth curves, OD 600 nm readings were performed. For LB agar experiments, cells were stamped on agar plates and incubated overnight. Disk diffusion susceptibility was performed according to (39,40), with slight modifications. In short: Mueller-Hinton agar plates were supplemented with resazurin (41). Overnight cultures were diluted and grown until OD 600 nm 0.6, and then diluted to OD 600 nm 0.1 with PBS buffer. Cells were streaked, and antibiotic discs were added. Plates were left overnight at 37°C (42).
Data availability. Data are available in the published article itself, and the supporting figures and tables are available as supplemental material. Other data that support the findings of this study are available from the corresponding authors upon request.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 4.4 MB.