A. baumannii ACICU (Iacono et al., 2008), A. baumannii AYE (Poirel et al., 2003), A. baumannii ATCC 19606 (Janssen et al., 1997), Escherichia coli 25922 and E. coli S17 λpir (Simon et al., 1983) strains were used for this study. E. coli ATCC 25922 was purchased from LGC Standards S.r.l., Italy). All strains were cultured under aerobic conditions at 37°C in Luria-Bertani (LB) broth/agar. LB broth, cation-adjusted Mueller-Hinton broth (CAMHB) and Tryptic soy broth (TSB) were used to perform growth curves, susceptibility tests and biofilm assays. The chemical reagents were chlorhexidine digluconate (CHX), carbonyl cyanide m-chlorophenylhydrazine (CCCP), triclosan (5-chloro-2-(2,4-dichlorophenoxy) phenol (TRI), the quaternary ammonium compounds benzalkonium chloride (alkylbenzyldimethylammonium chloride (BZK), dequalinium chloride (DQ), and cetrimide (alkyltrimethylammonium bromide (CT), piperine (1-piperoyliperidine, PIP) and resveratrol (3,5,4’-trihydroxy-trans-stilbene, RV). The antimicrobials and chemical reagents were purchased from Sigma-Aldrich (Sigma, Milan, Italy).

DNA and plasmid DNAs of A. baumannii ATCC 19606 and knockout mutants were extracted using the DNeasy Blood & Tissue Kit (Qiagen, Milan, Italy) and the Plasmid Mini/Midi Kits (Qiagen, Milan, Italy), respectively, according to the manufacturer’s instructions. A. baumannii ATCC 19606 was mutagenized as previously described (Amin et al., 2013; De Gregorio et al., 2015) with the following minor changes. The upstream and downstream fragments of target genes were amplified using the primers listed in Supplementary Table 1 and inserted in the TA Cloning pCR2.1 vector (Invitrogen); 100 μL of competent E. coli DH5α were transformed with TA-cloning vector. The upstream fragments were digested with NotI-BamHI and cloned into suicide vector pMo130-Tel^R, creating pMo130-TelR-Up. Next, the downstream fragments were digested with BamHI–SphI and inserted in pMo130-TelR-Up to obtain the plasmid pMo130-TelR-Up/Dw. The final plasmid was introduced into E. coli S17-1 λpir by CaCl₂ transformation and mobilized to the A. baumannii ATCC 19606 strain or single/double mutants via conjugation as described (Amin et al., 2013), to obtain single, double and triple mutants. Transconjugants were selected in LB agar containing 30 mg/L tellurite + 50 mg/L ampicillin and 50 mg/L kanamycin + 50 mg/L ampicillin, cultured in LB broth containing 14% sucrose. Serial dilutions were spread onto LB plates containing 14% sucrose. Colonies were screened for tellurite sensitivity to monitor excision of the suicide vector. The inactivation of adeB, adeJ, aceI and amvA genes were confirmed by PCR amplification using control primers (Supplementary Table 1).

A. baumannii ATCC 19606 was grown overnight at 37°C on LB broth, under shaking (200 rpm). The MIC and MBC of CHX was determined by a manual microdilution method according to the recommended procedures by the European Committee for Antimicrobial Susceptibility Testing (Eucast) of the European Society of Clinical Microbiology and Infectious Diseases (Escmid) (2000) and the Clinical and Laboratory Standards (CLSI, 2019). Susceptibility was assessed to MIC value < 4 mg/L as described (Rajamohan et al., 2010a). A. baumannii ATCC 19606 and deletion mutants were grown on CAMHB at 37°C for 24 h. Afterward, 50 μL of 1 × 10⁶ CFU/mL bacterial cells were added to each well of the microtiter plate containing 50 μL of the CAMHB with twice the final concentration of molecules studied. Then the plates were incubated at 37°C for 18–24 h. Non-treated bacteria were used as controls. All tests were performed in triplicate and repeated three times.

The tests were carried out using the checkerboard method according to the previously reported method (Hall et al., 1983). Serial dilutions of CHX (0.5–164 mg/L) were prepared and combined with serial dilutions of piperine (8–128 mg/L), resveratrol (32–128 mg/L), CCCP (0.5, 1, and 2 mg/L). Subsequently, 1 × 10⁶ CFU/mL of either A. baumannii ATCC 19606 or deletion mutants were added to each well of the microtiter plate. Then the plates were incubated at 37°C for 18–24 h. All experiments were repeated three times.

Biofilm formation was examined using a crystal violet (CV) staining assay according to the previously reported method (De Gregorio et al., 2020). Bacterial cell suspension was prepared at 0.5 McFarland standard and it was diluted 1:100 in TSB. Subsequently, 100 μL of 1 × 10⁶ cells/mL was transferred into a 96-well flat-bottomed polystyrene microtiter plate containing 100 μL of scalar doses of CHX (164–0.5 g mg/L) and incubated at 37°C for 24 h. Non-treated bacteria were incubated with 100 μL of broth and used as the control. The culture supernatant was gently discarded, the wells were washed twice with phosphate-buffered saline (PBS) 1 × pH 7.4 and the biofilms were stained with 200 μL of 0.1% crystal violet for 20 min. The wells were washed twice with PBS 1X, and dye was re-eluted with 100% ethanol. The absorbance was measured at 595 nm using a microplate reader (Bio-Rad Laboratories S.r.l.). The OD595/OD600 ratio was used to normalize the amount of biofilm formed to the total cell content.

A. baumannii ATCC 19606 cells were grown over night on LB broth at 37°C at 200 rpm. Subsequently, ATCC 19606 was diluted 1:100 in LB broth alone or LB broth with subMIC of CHX or RV or PIP or CHX plus RV or CHX plus PIP and grown at 37°C at 200 rpm for a further 3 h to reach the exponential phase (OD₆₀₀ = 0.5). Total RNA was isolated from three independent cultures according to the previously reported method (De Gregorio et al., 2018). The cDNAs were synthesized using QuantiTect Reverse Transcription Kit (Qiagen, Milan, Italy), according to the manufacturer’s protocol. Real-time RT-PCR assays were performed using SYBR Green master mix (Applied Biosystems) (Martinucci et al., 2016). The rpoB gene (the housekeeping gene) was used to normalize the expressions of target genes. The fold-change of the gene expression level was calculated using the 2^–ΔΔct method (Livak and Schmittgen, 2001). All experiments were performed three times in triplicate. The primers used in the qRT-PCR experiments were reported in Supplementary Table 2.

All statistical analyses were carried out using GraphPad Prism version 8.0 for Windows (GraphPad Software, San Diego, CA, United States). All experiments were performed at least three times and the results are shown as means ± SD. Differences between mean values were tested for significance using ANOVA. A P < 0.05 was considered to be statistically significant.

Comparison of cryo EM structures of AdeB (PDB code 7 kgd) and AdeJ (PDB code 7 m4q) were conducted using the DALI platform for pairwise alignment (Holm, 2020) and the software Coot (Emsley and Cowtan, 2004) and PyMol (Seeliger and de Groot, 2010).

A. baumannii ATCC 19606, AYE, ACICU strains having different antimicrobial susceptibility profiles and classified as susceptible, multidrug-resistant (MDR) and extensively drug-resistant (XDR) as described (Magiorakos et al., 2012), respectively, invariably showed both CHX MIC and MBC values of 32 mg/L and were considered tolerant to CHX (Table 1). Instead, E. coli ATCC 25922 showed a CHX MIC/MBC value of 2 mg/L and was considered susceptible (Table 1). A. baumannii ATCC 19606 was able to grow and retain viability in the presence of 4–16 mg/L subMIC concentrations of CHX, while A. baumannii ATCC 19606 growth was abolished at 32 mg/L CHX (Figure 1). Also, CHX subMIC concentrations of 8 and 16 mg/L decreased stationary phase cell density of A. baumannii ATCC 19606 by three and fourfold, respectively (Figure 1).

Because it has been demonstrated that CHX increased the expression of aceI efflux pump (EP) gene in A. baumannii ATCC 17978 (Hassan et al., 2013), we asked if CHX was able to regulate the expression of EPs genes in ATCC 19606. Preliminary data showed that basal level of expression of adeB, adeG, adeJ, belonging to RND superfamily, amvA and craA belonging to MFS superfamily, aceI, belonging to PACE superfamily, and abeS and abeM, belonging to the SMR superfamily were different in A. baumannii ATCC 19606. In particular, aceI, adeJ, adeB, and amvA were expressed at high levels, with expression levels normalized on rpoB of 0.49, 0.34, 0.25, and 0.28, respectively, while craA, abeS, and abeM at low levels (Supplementary Figure 1). As shown in Figure 2, CHX at subMIC concentrations (4 and 8 mg/L) increased the expression of adeB and adeJ EPs genes by 6x and 2x, respectively, while the expression of adeG EP gene and adeR and adeS regulatory genes were not affected. Moreover, subMIC concentrations of CHX increased the expression of aceI EP gene and amvA EP gene 5x by 4 mg/mL and 9x by 8 mg/mL, and 2x by 4 mg/mL, respectively (Figure 2). amvA EP gene expression was not induced in the presence of 8 mg/mL CHX. On the other hand, subMIC concentrations of CHX decreased the expression of craA EP gene 4x by 4 mg/L and 8x by 8 mg/L (Figure 2). The above data indicated that adeB, aceI and to lesser extent adeJ and amvA EP genes are activated by CHX in A. baumannii ATCC 19606.

To study the molecular mechanisms responsible for tolerance to CHX in A. baumannii, we analyzed the effect of inactivation of AdeB and AdeJ, AceI, and AmvA EPs, which are abundantly expressed and positively regulated by CHX in A. baumannii ATCC19606, on susceptibility to CHX. To this aim, CHX MIC and MBC were analyzed in A. baumannii ATCC 19606 marker-less mutants of adeB, adeJ, aceI and amvA EPs genes. As shown in Table 2, CHX MIC and MBC values were decreased by eight and twofold in ΔadeB and in ΔaceI mutant, respectively, compared with A. baumannii ATCC19606; in ΔamvA mutant CHX MIC was also decreased by twofold but CHX MBC was not affected. Instead, CHX MIC and MBC in ΔadeJ mutant were similar to A. baumannii ATCC19606 (Table 2). Furthermore, CHX MIC and MBC values were decreased by 16-fold in ΔadeB ΔaceI and ΔadeB ΔadeJ double mutants, eightfold in ΔamvA ΔadeB, and fourfold in ΔamvA ΔaceI double mutant, while CHX MIC was decreased by two fold, but CHX MBC not affected in ΔamvA ΔadeJ and ΔaceI ΔadeJ double mutants. Moreover, CHX MIC and MBC were decreased by 32-fold in ΔadeB ΔaceI ΔadeJ, 16-fold in ΔamvA ΔadeB ΔaceI and ΔamvA ΔadeB ΔadeJ, and twofold in ΔamvA ΔaceI ΔadeJ triple mutants (Table 2). CHX susceptibility with MIC and MBC values of 2–1 was recovered in ΔadeB ΔaceI and ΔadeB ΔadeJ double mutants, and ΔadeB ΔaceI ΔadeJ, ΔamvA ΔadeB ΔaceI, and ΔamvA ΔadeB ΔadeJ triple mutants (Table 2). The above data indicated that CHX MIC and MBC in A. baumannii ATCC 19606 were mainly sustained by the expression of adeB and that aceI, amvA and to a lesser extent adeJ played an additive effect.

To further study the role of EPs on CHX susceptibility in A. baumannii, we analyzed the effect the EP inhibitor CCCP in A. baumannii ATCC 19606 and EPs marker-less mutants. As shown in Table 3, CCCP reduced dose-dependently CHX MIC in A. baumannii ATCC 19606 and in ΔadeJ, ΔaceI, or ΔamvA single, double or triple mutants. CCCP reduced CHX MIC in ΔadeB, single, double or triple mutants but the effect was not dose-dependent. This indicates that inhibition of efflux pump activity restores susceptibility to CHX in A. baumannii ATCC 19606 and in ΔadeJ, ΔaceI, or ΔamvA, but not in ΔadeB mutants.

We next asked whether EPs knockout gene inactivation might affect the in vitro planktonic and sessile growth of A. baumannii ATCC 19606. A. baumannii ATCC 19606 and single, double or triple ΔadeJ, ΔaceI, ΔamvA, ΔadeB mutants showed similar sigmoid growth curves and no difference in growth rates, despite ΔamvA ΔadeB ΔaceI and ΔamvA ΔadeB ΔadeJ triple mutants showed a longer lag phase than A. baumannii ATCC 19606 and other deletion mutants (Supplementary Figure 2). We analyzed also biofilm growth of A. baumannii ATCC 19606 and single, double or triple EP mutants. As shown in Figure 3, biofilm formation of single, double or triple ΔadeJ, ΔaceI, ΔamvA, ΔadeB mutants grown in the absence or in the presence of 1/2 MIC CHX was decreased by 30–50% compared with A. baumannii ATCC 19606 parental cells. On the other hand, 1/2 MIC CHX decreased biofilm growth in ATCC 19606 parental, ΔadeJ, ΔaceI, ΔamvA, ΔadeB single mutants, ΔamvA ΔaceI, ΔamvA ΔadeB, ΔadeB ΔaceI, and ΔaceI ΔadeJ double mutants and ΔamvA ΔadeB ΔadeJ triple mutants, while induced biofilm growth in ΔadeB ΔadeJ or ΔamvA ΔadeJ double mutants, and ΔamvA ΔadeB ΔaceI, ΔadeB ΔaceI ΔadeJ, or ΔamvA ΔaceI ΔadeJ triple mutants (Figure 3).

The susceptibility to other biocides, which are used as antiseptics or disinfectants (McDonnell and Russell, 1999), was analyzed in A. baumannii ATCC 19606 wild type and EP deletion mutants. In accordance with previous findings (Chen et al., 2009), A. baumannii ATCC19606 and single EP deletion mutants showed TRI MIC and MBC of 0.06 and 0.125 mg/L, respectively, and were considered susceptible to TRI (Supplementary Table 3). On the contrary, A. baumannii ATCC19606 and single EP deletion mutants were tolerant to quaternary ammonium compounds DQ and CT, showing MIC and MBC values of 32–256 and 16–64 mg/L, respectively (Supplementary Table 3).

The mechanisms responsible for tolerance to BZK was studied in detail in A. baumannii ATCC 19606 parental strain and marker-less mutants of adeB, adeJ, aceI and amvA EPs genes. As shown in Table 4, BZK MIC and MBC values were decreased by four, two, and onefold in ΔadeB, ΔamvA, and ΔaceI mutants, respectively, compared with A. baumannii ATCC19606; BZK MIC and MBC were not affected in ΔadeJ mutant. Also, BZK MIC and MBC values were decreased by eightfold in ΔamvA ΔadeB, ΔadeB ΔaceI, and ΔadeB ΔadeJ double mutants, and twofold in ΔamvA ΔaceI, ΔamvA ΔadeJ, and ΔaceI ΔadeJ double mutants. Moreover, BZK MIC and MBC were decreased by 16-fold in ΔadeB ΔaceI ΔadeJ and ΔamvA ΔadeB ΔadeJ, eightfold in ΔamvA ΔadeB ΔaceI, fourfold in ΔamvA ΔaceI ΔadeJ triple mutants (Table 4). BZK susceptibility with MIC and MBC values of 2 was recovered in ΔadeB ΔaceI ΔadeJ and ΔamvA ΔadeB ΔadeJ triple mutants (Table 4). The above data indicated that BZK MIC and MBC in A. baumannii ATCC 19606 were mainly regulated by the functioning of adeB and to a lesser extent amvA, aceI, and adeJ EPs.

Overall, AdeB and AdeJ are two highly homologous proteins sharing a sequence identity of 49%. Both AdeB and AdeJ adopt a homotrimeric structure, with the typical RND-like fold (Su et al., 2019; Morgan et al., 2021; Zhang et al., 2021). Similar to AcrB of E. coli (seqid 50%), they are composed of a transmembrane domain formed by 12 transmembrane (TM) helices and a large periplasmic domain (Figure 4A). In this structural organization, the periplasmic domain harbors an entrance, a proximal and a distal substrate binding pockets (PBP and DBP, respectively). The PBP is separated from the DPB by a so-called “gate-loop” (or G-loop). Another conserved flexible loop (F-loop) connects the cleft entrance to the proximal drug-binding pocket. These loops are crucial to substrate discrimination in AcrB (Schuster et al., 2016). During substrate extrusion, AdeB and AcrB are thought to pass through a conformational change that forces the substrate to move from the PBP to the DBP for final extrusion (Schuster et al., 2016; Morgan et al., 2021).

A structural comparison of AdeB and AdeJ protomers was performed to analyze whether differences in the structural features of the two pumps may account for the major role observed for AdeB, compared to AdeJ, on CHX extrusion and susceptibility. As shown in Figure 4B, AdeB and AdeJ share a strictly conserved fold, with root mean square deviations (rmsd) ranging between 2.5 and 3.0 Å on the three chains. The analysis of the entrance binding sites of AdeB and AdeJ suggests different features that may explain a different involvement in CHX transport. Most relevant, the conserved W708 of AdeB is replaced by an arginine residue (R718) in AdeJ (Figure 4B). Other residues belonging to this cavity also differ. Specifically, V658, M706, I861 are replaced by L668, G714, and T831, respectively. These differences in the composition of the entrance site of AdeJ, compared to AdeB, make the pocked positively charged and not prone to bind the positively charged CHX. Significant differences are also observed in the F loops of the two pumps. In AdeB, the F-loop (661-PAIDELGT-668) resembles that of AcrB (669-PAIVELGT-676) of E. coli, in which residue I671 has been shown to be important for drug discrimination (Schuster et al., 2016). Differently, the F-loop of AdeJ does not contain this key isoleucine (669-PAMPELGV-676), which is thought to be part of a preferential small-drug entrance pathway. Additionally, a more negatively charged F-loop (due to the charge contribution of D664) in AdeB may also contribute to its stronger involvement in the transport of the positively charged CHX.

We next screened two natural compounds, RV and PIP, which have shown promising activity as EPs inhibitors (Sharma et al., 2010; Mirza et al., 2011; Singkham-In et al., 2020). We tested if these non-toxic compounds can decrease CHX MIC in A. baumannii ATCC 19606 and EPs gene knockout mutants and restore susceptibility to CHX. Both PIP and RV showed no antimicrobial activity against A. baumannii ATCC 19606 and ΔadeJ, ΔaceI, ΔamvA, ΔadeB mutants (MIC > 1,024 mg/L) (Tables 5, 6). We then determined the antimicrobial activity of PIP in combination with CHX by in vitro combination assay. As shown in Table 5, increasing doses of PIP up to 128 mg/L decreased CHX MIC and MBC by four fold in A. baumannii ATCC 19606 and by two to eightfold in ΔadeJ, ΔaceI, ΔamvA, ΔadeB mutants, being able to restore CHX susceptibility in single, double and triple mutants with inactivation of adeB gene. Furthermore, RV from 32 to 128 mg/L decreased dose-dependently CHX MIC and MBC and restored CHX susceptibility in A. baumannii ATCC 19606 and ΔadeJ, ΔaceI, ΔamvA, ΔadeB single, double and triple mutants. In particular, CHX susceptibility was restored by RV at 128 mg/L in A. baumannii ATCC 19606 and ΔaceI, ΔamvA, ΔadeB, or ΔadeJ single mutants, 64 mg/L in all double or EP triple mutants, 32 mg/L in double or triple EP mutants harboring deletion of adeB (Table 6).

To assess whether the effect of PIP and RV on CHX susceptibility was mediated by inhibition of EPs expression, we analyzed amvA, aceI, adeB, and adeJ expression in A. baumannii ATCC 19606 in the presence of 4 mg/L subMIC CHX in combination with 32 mg/L PIP or 32 mg/L RV. As shown in Figure 5A, PIP counteracted CHX-dependent increased expression of amvA, aceI, adeB, and adeJ, while it did not affect basal EP gene expression. On the other hand, resveratrol inhibited both basal and CHX-dependent increased expression of amvA, aceI, adeB, and adeJ, the highest effect found for adeB and amvA (Figure 5B). The above data suggested that different effects of PIP and RV on CHX MIC in A. baumannii ATCC 19606 were mediated by distinct regulation of amvA, aceI, adeB, and adeJ expression.

The effect of RV was also analyzed on BZK MIC and MBC in A. baumannii ATCC 19606 parental strain and EP deletion mutants. As shown in Table 7, RV from 32 mg/L to 128 mg/L decreased dose-dependently BZK MIC and MBC and restored BZK susceptibility in A. baumannii ATCC 19606 and single, double and triple EP deletion mutants. BZK susceptibility was restored by RV at 128 mg/L in A. baumannii ATCC 19606 and ΔaceI or ΔamvA, single mutants, 64 mg/L in ΔadeB, or ΔadeJ single mutants and in all double or EP triple mutants, 32 mg/L in ΔadeB single mutant and in all, but not ΔamvA ΔadeJ, double mutants (Table 7).

We analyzed also the effect of 2 mg/L BZK alone or in combination with 32 mg/L RV on amvA, aceI, adeB, and adeJ expression. As shown in Figure 6, two mg/L BZK alone inhibited in a non-significant way EP gene expression, and 2 mg/L BZK in combination with 32 mg/L RV significantly inhibited amvA, adeB, and adeJ expression by 10–15-fold and aceI expression by twofold respect to untreated cells. The above data indicated that the effect of RV on BZK susceptibility was mediated by inhibition of amvA, adeB, adeJ, and to a lesser extent aceI expression.