Conserved small protein associates with the multidrug efflux pump AcrB and differentially affects antibiotic resistance

AcrZ–SPA Copurifies with Components of the AcrAB–TolC Membrane Efflux Pump.

The 49-amino-acid AcrZ is highly conserved among enterobacteria and other Gram-negative species (Fig. 1). The protein is predicted to contain an N-terminal TM helix (11, 13), and previous subcellular localization experiments have shown that tagged derivatives partition with the membrane fraction and that the C terminus is in the cytoplasm (13). Our group has used the sequential peptide affinity (SPA) tag (14) to observe the accumulation of multiple sprotein fusions (11). Because AcrZ–SPA is easily detected (11) and a strain producing the tagged derivative is phenotypically wild type in a cell envelope stress assay (15) as well as in the antibiotic sensitivity assays described below, AcrZ–SPA was a good candidate for biochemical approaches to identify interacting protein partners. Chromosomally expressed AcrZ–SPA was purified from a crude lysate of exponentially growing cells on the basis of the 3xFLAG tag and the calmodulin- binding peptide that compose the SPA tag. Two high-molecular-weight proteins consistently observed to copurify with AcrZ–SPA (Fig. 2) were identified as AcrB (114 kDa) and AcrA (42 kDa) by mass spectrometry. This finding indicated AcrZ associates with the AcrAB–TolC efflux pump, an interaction supported by the observation that AcrZ–SPA is detected in both the inner and the outer membrane fractions in a wild-type strain but is found predominantly in the inner membrane fraction of an ∆acrB deletion strain (13).

AcrZ Associates with AcrAB–TolC via the Inner Membrane Protein AcrB.

We verified the interaction of AcrZ–SPA with the AcrAB–TolC complex by performing a reciprocal purification of AcrZ-SPA with AcrB-His₆ (a functional C-terminal hexahistidine-tagged fusion protein of AcrB). As shown in Fig. 3A, AcrZ-SPA is coeluted from the Ni²⁺-NTA column in the eluate fraction with AcrB–His₆ from the acrZ-SPA acrB-His₆ lysate, but not from the acrZ-SPA lysate. As a control, we carried out a similar purification with His₆-tagged PhoR, an unrelated multitransmembrane protein. This experiment is complicated by the fact that untagged AcrB was found to copurify with other hexahistidine-tagged proteins due to the histidine-rich cluster on its surface (16). Indeed, both AcrB and AcrZ-SPA were found to copurify with PhoR–His₆ from a phoR-His₆ strain carrying wild-type acrAB (Fig. S1A). However, AcrZ-SPA did not copurify with PhoR–His₆ from the corresponding strain lacking acrAB (Fig. S1A), providing further support that AcrZ–SPA interacts specifically with AcrB. The level of AcrZ–SPA was notably lower in the ∆acrAB mutant, suggesting that the binding with its target may be required for AcrZ stability.

Importantly, AcrB–His₆ does not interact promiscuously with random small membrane proteins, as neither SPA-tagged YbgT nor SPA-tagged YccB was retained on Ni²⁺-NTA resin in conjunction with AcrB–His₆ (Fig. S1B). We also found that AcrZ–SPA was retained by AcrB–His₆ on Ni²⁺-NTA resin in the absence of both AcrA and TolC (Fig. 3A), indicating that AcrZ can associate with AcrB in the absence of the two other members of the AcrAB–TolC complex. Together, these results indicate that AcrZ–SPA is retained on the column by virtue of its association with AcrB.

To examine whether untagged AcrZ interacts with the AcrB pump, we also performed limited tryptic digestion of AcrB–His₆ purified from wild-type or ∆acrZ mutant cells. For AcrB–His₆ purified from wild-type cells, one fragment of ∼40 kDa was visible after 30 min of digestion (Fig. 3C). In contrast, for AcrB–His₆ purified from cells lacking AcrZ, the 40-kDa fragment appeared at 10 min of digestion, and one additional fragment of ∼100 kDa was detected after 30 min of digestion (Fig. 3C). This result suggests that the accessibility of AcrB is altered by untagged AcrZ, likely via a direct interaction.

Two Dominant-Negative Mutations Map to the C-Terminal Half of AcrZ.

Dominant-negative mutants, which inhibit the function of the wild-type protein when the wild-type and mutant proteins are produced simultaneously, are useful in both biochemical and genetic approaches as they can be used to define the nature of a protein–protein interaction. We thus set out to identify dominant-negative mutations in acrZ. To this end, acrZ was amplified by error-prone PCR and cloned behind an arabinose-inducible promoter on pBAD24 (17). Wild-type cells transformed with mutagenized pBAD24-acrZ were first screened for growth under inducing conditions to ensure that the mutant versions of AcrZ were nontoxic to the cell.

As described below, ΔacrZ cells are sensitive to chloramphenicol. We exploited this phenotype in our screen for dominant-negative mutants by identifying cells that grew in the presence of chloramphenicol (at a concentration that is inhibitory to the growth of ΔacrZ cells) when expression of the plasmid-borne copy of acrZ was repressed (on plates containing glucose) but not when both the chromosomally encoded and plasmid-encoded copies of acrZ were expressed (on plates containing arabinose). Two robust dominant-negative mutations, which both mapped to the C-terminal half of AcrZ, were isolated. One missense mutation (G30R) altered a highly conserved glycine located at the C-terminal end of the α-helix to an arginine. The other mutation was a nonsense mutation that truncated the protein by four amino acids (G46Stop).

One possible explanation for the dominant-negative phenotype is that the mutations disrupt a productive interaction between wild-type AcrZ and AcrB. This could occur if the dominant-negative AcrZ mutants bind AcrB more tightly and occlude wild-type AcrZ. To test whether the SPA-tagged versions of either AcrZ^G30R or AcrZ^G46Stop are still capable of binding AcrB–His₆, we once again purified AcrB–His₆ on Ni²⁺-NTA from lysates of cells that produce either the wild-type or the mutant forms of AcrZ–SPA. Similar to wild-type AcrZ–SPA, AcrZ^G46Stop–SPA interacts with AcrB–His₆ (Fig. 3B), suggesting that this dominant-negative mutant acts by sterically occluding wild-type AcrZ from AcrB and preventing a productive interaction. On the other hand, AcrZ^G30R–SPA did not copurify with AcrB–His₆ (Fig. 3B). This observation suggests that AcrZ^G30R does not stably associate with AcrB. The mechanism by which this dominant-negative mutant blocks the activity of wild-type AcrZ is not known.

Suppressor Mutations Map to TM Helix 11 and the Cytoplasmic Domain of AcrB.

To further examine the binding of AcrZ to AcrB as well as screen for acrB mutations that could suppress the acrZ^G30R mutation, we tested for a direct interaction between AcrZ and AcrB using the bacterial two-hybrid system. Specifically, we assayed whether AcrZ and AcrB fused to the T18 and T25 fragments of Bordetella pertussis CyaA (18) could restore adenylate cyclase activity leading to increased β-galactosidase activity. Consistent with the results of the affinity purification assays, there was a strong interaction between wild-type AcrZ–T18 and T25–AcrB as evidenced by high levels of β-galactosidase activity compared with empty vector controls (Fig. 4A), whereas the G30R dominant-negative mutation in AcrZ–T18 abolished the interaction with T25–AcrB (Fig. 4A). The loss of β-galactosidase activity in the dominant-negative mutant cannot be attributed to a difference in protein amounts, as AcrZ–T18 and AcrZ^G30R–T18 accumulate to similar levels in the cell (Fig. S2).

To identify acrB mutations that could suppress the acrZ^G30R mutation, we randomly mutated the acrB-coding sequence in the T25 construct and screened for mutants that restored an interaction with AcrZ^G30R–T18 in the two-hybrid assay. Two missense mutations in T25–AcrB were isolated and identified as H526Y and L984P. As shown in Fig. 4, β-galactosidase activity in strains coexpressing AcrZ^G30R–T18 with T25–AcrB^H526Yor AcrZ^G30R–T18 with T25–AcrB^L984P was increased >10-fold compared with the AcrZ^G30R–T18 T25–AcrB strain to 23% (for H526Y) or 37% (for L984P) of wild-type levels. The H526Y suppressor mutation is in the cytoplasmic α-helix between TM6 and TM7, whereas the L984P suppressor mutation is in TM11 of AcrB (Fig. S3).

We also recombined the H526Y and L984P alleles into the chromosomal copy of acrB to test whether the mutant versions of AcrB could compensate for the chloramphenicol sensitivity introduced by pBAD24-encoded AcrZ^G30R. Consistent with the ability of the H526Y mutation to suppress the reduced binding of AcrB to AcrZ^G30R, we found that the suppressor mutation restored some resistance to chloramphenicol (Fig. 4B). Unfortunately, the strain expressing the L984P did not grow under the conditions of our assay and thus could not be tested.

acrZ Is Regulated by MarA, Rob, and SoxS.

acrZ shares strong synteny with modEF and modABC, two operons involved in molybdenum uptake and in regulating the synthesis of the molybdenum cofactor molybdopterin (Fig. 1A). In E. coli, molybdopterin is used (often in conjunction with Fe-S clusters) by ∼20 molybdo-enzymes to conduct a broad range of redox reactions, including the detoxification of compounds such as aromatic aldehydes, in both the cytoplasm and the periplasm (reviewed in ref. 19).

In LB medium, AcrZ–SPA is most abundant in exponentially growing cells, but it is also present in cells harvested from both stationary phase and overnight cultures (11). Given its synteny with modEF and modABC, we surmised acrZ might be regulated in accordance with molybdenum levels. However, neither excess extracellular molybdenum nor lack of the ModE transcription factor or the ModABC transporter affects AcrZ–SPA accumulation (Fig. S4A).

We also hypothesized that acrZ might be coregulated with the genes encoding its interacting partners, acrAB and tolC. This hypothesis is supported by the presence of a conserved class II mar/rob/sox box and σ⁷⁰ “−10” promoter region upstream of acrZ (Fig. 5A). MarA, Rob, and SoxS are highly homologous transcription factors that recognize the same consensus binding site (AYnGCACnnWnnRYYAAAY) (21) and are induced by noxious compounds such as antibiotics, detergents, and oxidizing agents. When SoxS was activated by methyl viologen (paraquat), AcrZ transcript accumulated as measured in a primer extension assay (Fig. 5B). This transcript initiates at a highly conserved adenine with appropriate spacing to the conserved “−10” promoter region. In addition, both AcrZ–SPA and AcrB protein levels were elevated when MarA, Rob, or SoxS were overproduced from plasmids (Fig. 5C).

ΔacrZ Mutants Are Sensitive to Many, but Not All, AcrB Substrates.

In accordance with its function as an efflux pump, cells deleted for any component of the AcrAB–TolC complex are extremely sensitive to numerous antibiotics (22). Consistent with both AcrZ association with AcrB and acrZ regulation by MarA, Rob, and SoxS, in a recent large-scale phenotypic study designed to identify functionally related genes (23), an ∆acrZ mutant showed a chemical sensitivity pattern similar to deletion mutants of acrA, acrB, rob, and yaiP (which encodes a protein of unknown function). The correlation coefficient is above 0.4 for all four pairs. However, the ∆acrZ mutant is sensitive only to 22 of the 50 stresses under which ΔacrB mutants are impaired for growth in these large-scale assays (23).

To directly test ∆acrZ sensitivity to a subset of these compounds, we carried out competition assays between the ∆acrZ mutant and wild-type cells as well as between ∆acrB and wild-type cells (Fig. 6). As previously reported, the ∆acrB mutant was hypersensitive to all of the compounds tested. In contrast, ΔacrZ cells were hypersensitive to some compounds such as puromycin, chloramphenicol, and tetracycline, but showed wild-type sensitivity to other AcrB substrates such as erythromycin, rifampicin, and fusidic acid. This result suggests that AcrZ is required for efficient efflux of a subgroup of AcrB substrates and is dispensable for others.

We also tested the antibiotic susceptibility of the wild-type and ∆acrZ mutant in monoculture by minimum inhibitory concentration (MIC) assays (Table 1) and on gradient plates (Fig. S5A). Consistent with the results from the competition assay, ∆acrZ showed increased sensitivity to puromycin, chloramphenicol, and tetracycline, but not to erythromycin, rifampicin, or fusidic acid. The gradient plate assay also showed that acrZ-SPA fusion is phenotypically wild type. In contrast to the deletion of acrZ, deletions of yajC, which encodes another protein shown to bind AcrB (16), and of mdtC, mdtF, and acrD, encoding three other multidrug transporters, do not result in increased chloramphenicol or tetracycline sensitivity (Fig. S5B). This result, together with aforementioned findings, suggests that AcrZ uniquely binds AcrB to enhance antibiotic resistance.

Strain Construction.

All strains are derivatives of the laboratory stock of E. coli K-12 MG1655 unless otherwise noted and are listed in Dataset S1. The plasmids and oligonucleotides used in this study are listed in Datasets S2 and S3, respectively. Chromosomal mutants were generated by using λ-Red–mediated recombineering (34, 35). Details about strain construction and genetic screens are provided in SI Materials and Methods.

Bacterial Growth.

Cells were propagated in standard fashion in liquid or on solid [1.5% (wt/vol) Bacto-agar] Luria–Bertani media (10 g of tryptone, 5 g of yeast extract, 10 g of NaCl per liter). Glucose and arabinose were used at concentrations of 0.2% (wt/vol). Antibiotics were used at the following concentrations for selection of marker genes: kanamycin, 30 μg/mL; chloramphenicol, 25 μg/mL; ampicillin, 100 μg/mL; and tetracycline, 12.5 μg/mL.

Competition Assays and MIC Assays.

Competition assays were performed as in ref. 15, and MIC assays were performed by serial agar dilutions as in ref. 22 with minor modifications. These antibiotic sensitivity assays are described in detail in SI Materials and Methods.

AcrZ–SPA and AcrB–His₆ Purification and Detection.

The SPA and His₆-tagged proteins were purified on the basis of their tags and the tagged proteins as well as untagged AcrB were detected by immunoblot analysis as described in detail in SI Materials and Methods.

Bacterial Two-Hybrid Assays.

Pairs of proteins to be tested were fused to the two catalytic domains T18 and T25 of adenylate cyclase based on their confirmed topologies. The resulting plasmids were cotransformed into a Δcya strain, and transformants were grown in LB medium supplemented with ampicillin, kanamycin, and 0.5 mM isopropyl-β-d-thiogalactopyranoside for 16 h at 30 °C. The overnight cultures (1 mL) were harvested and assayed for β-galactosidase activity as described (36).

Primer Extension Assays.

Cultures of wild-type MG1655 and ΔacrZ cells grown to mid-exponential phase (OD₆₀₀ ∼0.4–0.6) were split, and half of the culture was exposed to 250 μM paraquat for 10 min. Total RNA was extracted by hot acid phenol as described previously (37). Primer extension analysis was carried out using 5′-end, ³²P-labeled primer (1 pmol) and total RNA (5 μg) as described previously (38).