Antimicrobial resistance, especially multidrug resistance in bacterial pathogens, is among the top 10 global threats to humanity [1
]. Among the large array of different defense mechanisms adapted by bacteria, the overexpression of efflux pumps has a significant role in conferring multidrug resistance. AcrAB-TolC is one of the most extensively studied efflux pump systems in Gram-negative bacteria, playing a crucial role in the multidrug resistance in bacteria such as Eshcherichia coli
]. AcrAB-TolC is a member of the resistance nodulation division (RND) superfamily. AcrAB-TolC efflux pump confers resistance to a broad spectrum of antimicrobial compounds including β-lactams, tetracycline, novobiocin, and fluroquinolones [5
]. This tripartite efflux transporter consists of three major protein components [7
], an outer membrane channel TolC, a periplasmic adaptor protein (PAP) AcrA, and an inner membrane proton-driven antiporter AcrB [9
]. TolC forms a channel that spans the outer membrane and acts as the exit pathway of substrates translocated from the inner membrane and the periplasmic space. AcrA has function in stabilizing the connection between the two membrane components TolC and AcrB [13
]. The RND transporter protein AcrB is responsible for substrate recognition and energy transduction. Upon binding of a substrate, AcrB uses the energy from the proton flow down its concentration gradient through a proton translocation pathway in the transmembrane domain to drive the conformational change necessary to move the substrate upward toward the exit tunnel [4
]. TolC is shared by several efflux systems, hence E. coli
strains deficient in TolC are more sensitive to a wider variety of chemicals (e. g. detergents, drugs, bile salts, and organic solvents) [16
With the dedication of many research groups, the structure and mechanism of drug efflux by the RND pumps have been brought to light. The first crystal structure of the pump component was determined for TolC by Koronakis et al. [18
] in 2000. TolC is a trimer with an overall length of 140 Å with 40 Å in the β-barrel domain mainly composed of β strands, and 100 Å in the periplasmic domain mainly composed of α-helices. The periplasmic end of the TolC tunnel is sealed at the resting state, which likely opens by an allosteric protein–protein interaction mechanism [19
]. In 2002, Murakami et al. first reported the crystal structure of AcrB, followed by the proposal of the functional rotation mechanism [20
]. Later in 2006, Mikolosko and coworkers determined the crystal structure of AcrA. In contrast to the trimeric TolC and AcrB, AcrA forms a hexamer in the pump assembly [23
]. The assembled pump structure was first proposed as the “deep interpenetration model”, which shows that AcrB and TolC have direct interactions with AcrA wrapped around on the outside to strengthen the interaction [24
]. More recently, Wang et al. proposed a new model based on Cryo-EM studies, known as the “tip-to-tip model”. In this model, AcrA hairpins form a barrel-like conformation, contacting TolC in a tip-to-tip arrangement [25
]. The recent determination of the complex structure first by cryo-EM, then by X-ray crystallography, confirmed the tip-to-tip model [7
]. Energy does not seem to be required to assemble the AcrAB-TolC complex and AcrAB could interact with the TolC channel to form a AcrAB-TolC complex even in the absence of known substrate [31
]. The dynamic process that leads to the formation of the complex is still elusive.
The dominant negative effect describes the phenomena in which an excess of a functionless mutant of a protein in the presence of its wild type counterpart, reduces the observed activity due to competition from the mutant for interaction with functional partners of the protein of interest. In the AcrAB-TolC complex, over-expression of functionless AcrB or AcrA mutant in wild type E. coli strains is expected to drastically reduce the assembly of functional efflux complex, and thus reduce the efflux activity and increase the sensitivity to substrate compounds. However, we tested the overexpression of several functionally defective mutants in the wild type E. coli strain, but did not observe the expected level of reduction. We speculate that the assembly of the AcrAB-TolC is a precisely controlled process involving delicate proof-reading procedures.
The dominant negative effect describes the situation in which the phenotype is dominated by the negative impact of the functionless mutant. The observation of the dominant negative effect has been used in many studies to investigate the mechanism of protein–protein interaction, including the identification of protein–protein interactions interface [46
], determination of enzymatic activity related to oligomerization [47
], and the effect of mutations in genetic disorders [48
In the process of AcrAB-TolC assembly, there are many steps where the incorporation of a functionless AcrA or AcrB mutant would negatively impact the efflux activity. First, all three proteins in the system are oligomers. AcrB and TolC are obligate timers, while AcrA is believed to exist as a dimer or trimer in the free form and assembles into a hexamer in the pump complex [13
]. While AcrA and AcrB are believed to form a complex in the absence of substrate and efflux, TolC assembles with AcrAB during active efflux. When a functionless AcrA mutant is expressed in excess in a wild type E. coli
cell containing genomic AcrA, we expect them to compete with their genomic counterpart to engage genomic AcrB, forming non-functional interactions to reduce the overall efflux activity. Similarly, we expect functionless AcrB mutant expressed from a plasmid to compete with genomic AcrB for genomic AcrA. In addition, we expect the competition for genomic TolC will further enhance the dominant negative effect. For this competition to occur, we chose mutants that are defective due to mechanisms not directly related to the interaction between AcrA and AcrB. The structure of the AcrAB-TolC complex and location of mutants mentioned in this study are shown in Figure 6
. We determined the expression level of the mutants relative to their genomic counterpart. Using serial dilution and quantitative Western blot analysis, we found that the expression levels of the AcrB mutants were 10–20 folds of the level of the genomic AcrB, and the AcrA mutants were 2 to 20 folds of the level of the genomic AcrA. With this high level of excess, we expect to observe a strong dominant-negative effect if the mutants were actively involved in the pump assembly, competing for binding partners. We constructed five groups of AcrA and AcrB mutants, defective in different aspects. We observed that the effect of certain mutations was not always the same for different substrates. For example, T978A mutation in AcrB is detrimental to all substrates tested except for R6G, while F178A mutation in AcrB drastically reduced the MIC for ERY, but not as much for other substrates tested. This difference in mutation effects has been observed in many studies characterizing AcrA and AcrB mutants, for example, [23
]. We speculate that this difference could be due to differences in the binding and interaction of specific substrates with the pump complex. The substrates vary drastically in their size, shape, structure, and charged state. As a result, the subgroup of residues that they interact with on their way to be transported are not likely to completely overlap. Therefore, point mutations introduced in AcrA and AcrB could have different impacts on specific substrates.
Mutants defective in the proton translocation pathway still form trimers [34
] and interact properly with AcrA (Figure 1
c). Then, was why no dominant negative effect observed? One possibility is that the genomic AcrA and AcrB are transcribed together, sharing the same mRNA. Hence, the newly produced AcrA and AcrB could be clustered as well. As a result, the genomic AcrA and AcrB form a AcrAB complex as soon as they are translated and inserted into the membrane (AcrB) or secreted into the periplasm with lipid anchoring (AcrA). Since the local concentration of the genomic proteins are high, they associate with each other with a much higher chance than associate with a plasmid-encoded partner. Another requirement for the observed activity is that the AcrAB complex, once formed, should be resistant to dissociation. Otherwise, the high concentration of AcrB mutant in the cell membrane would be effective in competing with genomic AcrB to form a nonfunctional AcrAB complex.
We further examined the impact of over-expressing both AcrA and AcrB, in which the entire sequence coding AcrA and AcrB were inserted in a plasmid. As a result, we expect them to form AcrA-AcrBD408A complex to compete with plasmid encoded AcrAB for genomic TolC. Yet, while the expression level of the complex was ~20 fold higher than that of the genomic AcrAB, no significant reduction of MIC was observed. It appears that TolC could differentiate the two complexes, even though the only difference between them is the single residue mutation down in the transmembrane domain of AcrB.
The second group of AcrB mutants examined are defective in substrate binding. These mutations are not expected to affect AcrB structure [42
], nor do they impact the interaction between AcrA and AcrB (Figure 2
b). However, we did not observe a significant reduction of MIC.
We also examined the effect of expressing AcrA mutants on the efflux activity. The first group of residues we tested are ones that impact AcrA interaction with TolC, as they are involved in the tip-to-tip interaction with TolC [44
]. We confirmed that a representative mutant in this group, R128D, still binds to AcrB (Figure 3
b). Similarly, we expect the expression of these mutants will lead to competition with genomic AcrA for genomic AcrB, and the formation of a nonfunctional complex. However, no reduction of MIC was observed. To further probe the interaction of AcrA and AcrB, we created a strain of E. coli
that is genetically modified to replace the signal peptide of AcrA with the signal peptide of OmpA (BW25113spmut
). The resultant AcrA can still be secreted into the periplasm and is fully functional, but the lipid anchoring is lacking [13
]. With this free-floating AcrA, we expect the interaction between AcrA and AcrB to be weakened, which may increase the competitiveness of the plasmid encoded AcrA, which is lipid anchored. Yet, we did not observe a reduction of MIC when the functionless mutants were expressed in strain BW25113spmut
The next group of AcrA mutants contains changes at the inter-subunit interface of AcrA to disrupt formation of the functional hexametric ring. While the A155D single mutation was enough to completely abolish activity, expression of the protein in BW25113 did not lead to a reduction of MIC. Finally, we examine a group of AcrA that forms disulfide bond locked conformation that is functionally incompatible. Again, similar as other groups, over-expression of these mutants did not have a significant impact on MIC.
In conclusion, we examined the effect of plasmid-encoded AcrA and AcrB mutants in wild type E. coli
cells, to probe the potential disruption of normal AcrA-AcrB-TolC assembly in the presence of excess mutants of AcrA or AcrB. To our surprise, none of the five groups of mutants showed the so-called “dominant negative” effect. This observation indicates that the RND pump assembly process in Gram-negative bacteria is a precisely controlled process that prevents the formation of functionless complex. An alternate explanation is the possibility that efflux depends on only a very small population of AcrAB-TolC pumps active at a given moment, as the population of active AcrAB-TolC pumps was not detectable in situ in E. coli
]. If the majority of AcrAB and TolC in the cells are idle, then the effect of over-expressing functionless mutant would be greatly limited. In addition, our results suggest that dissociation kinetics of the AcrAB complex is very slow. Once formed, the complex remains bound and does not dissociate easily.