Understanding the basis of antibiotic resistance: a platform for drug discovery

Characterizing the mechanisms of resistance

It was known that porin proteins in Escherichia coli play a role in antibiotic entry into the Gram-negative bacterial cell (Harder et al., 1981) and that the absence of some porin proteins is associated with low level multi-drug resistance (MDR; Mortimer & Piddock, 1993). Therefore, the outer-membrane protein (Omp) profiles of the Salmonella isolates from Patient B were determined (Fig. 1). Some isolates had wild-type profiles; however, five lacked OmpF and one lacked OmpF and OmpC (Piddock et al., 1993). Using methods developed in my laboratory (Mortimer & Piddock, 1991), the concentration of ciprofloxacin that accumulated inside the Salmonella isolates was determined (Fig. 2). There were varying levels accumulated by the different isolates, but these were not associated with any particular porin profile (Piddock et al., 1993).

In the absence of routine PCR and DNA sequencing to identify whether there was a mutation in a gene encoding the primary target protein of fluoroquinolone antibiotics, DNA gyrase, we took advantage of the knowledge that the resistant allele was recessive to the wild-type allele. A plasmid containing wild-type gyrA (Robillard, 1990) was introduced into the isolates and restoration of quinolone susceptibility was an indicator of a mutation in gyrA. Six of the isolates resistant to fluoroquinolones only became more susceptible to fluoroquinolones (Table 1). We cloned and sequenced gyrA from the NCTC type strain of S. Typhimurium (NCTC 74) and defined the quinolone resistance determining region of this gene (Griggs et al., 1996). With this information we sequenced gyrA of the clinical isolates and showed that in five of the isolates there was an unusual substitution in the quinolone resistance determining region of Gyr A (Ala119 to Glu) (Griggs et al., 1996). However, this did not fully explain the observed resistance. Likewise, introduction of a plasmid, pBP548, containing wild-type gyrB (Heisig, 1993) revealed that one isolate became more susceptible to fluoroquinolones. PCR and DNA sequencing revealed a Ser464 to Tyr substitution in GyrB (Gensberg et al., 1995). However, mutations in gyrA or gyrB also did not fully explain the MDR phenotype of several of the isolates.

As efflux had been shown to confer clinically relevant antibiotic resistance in Pseudomonas aeruginosa (Li et al., 1994), we explored whether efflux had a role in the antibiotic resistance of the Salmonella isolates. Northern blotting showed that four isolates had increased levels of acrB transcript (Piddock et al., 2000). We also carried out experiments to determine the concentrations of ciprofloxacin, chloramphenicol and tetracycline accumulated. In the presence of an efflux inhibitor, CCCP (which dissipates the proton motive force required by several efflux pumps, such as AcrB, to transport their substrates), the three drugs accumulated to similar concentrations to those seen in the pre-ciprofloxacin therapy isolate (Piddock et al., 2000). These data suggested that the MDR phenotype was delivered, at least in part, due to increased efflux mediated by overproduction of AcrB, a component of the AcrAB–TolC tripartite MDR efflux system.

By 2000, we had spent a long time researching the mechanism of antibiotic resistance in these isolates and obtained a considerable volume of data, but none fully explained the phenotypes. To this day, this is a problem with working with clinical isolates. Even with whole genome sequencing, which provides a huge amount of information, it is rare to observe a mutation in only one gene; therefore, biological experiments are essential to determine whether the observed mutation(s) gives the phenotype; detecting a single nucleotide polymorphism, deletion or an insertion provides hypotheses for the mechanism(s) delivering the phenotype. Therefore, I decided that it was necessary to investigate efflux in Salmonella and its contribution to antibiotic resistance, not in the clinical isolates, but in standard reference strains used by Salmonella researchers.

Multi-drug efflux: role in antibiotic resistance and virulence

Efflux is the pumping of a solute out of a bacterial cell. Efflux pumps are present in antibiotic-susceptible as well as -resistant bacteria and can be encoded on the genome or on a transmissible element (Piddock, 2006a). Expression of some efflux pumps can be induced by these substrates. Efflux pumps such as the Tet pump confer resistance to just tetracycline-type drugs, but other pumps, such as AcrB, will confer resistance to many different types of substrates. AcrB (and its homologues) is the transporter component of a tripartite system and resides in the inner membrane of Gram-negative bacteria. AcrA is a periplasmic adaptor protein which links AcrB to the outer membrane channel TolC through which substrates are transported to the outside of the cell (Blair & Piddock, 2009). Not only does S. Typhimurium act as a model organism for other Enterobacteriaceae, but also the AcrAB–TolC system acts as a paradigm for resistance nodulation division (RND) tripartite efflux pump systems.

Inactivation of any of the components of the AcrAB–TolC in S. Typhimurium confers multi-drug susceptibility (Table 2), showing that this system confers innate MDR. These efflux mutants also accumulate much higher intracellular concentrations of antibiotic and other substrates such as dyes (Blair et al., 2009; Webber et al., 2009); the same is true for other Gram-negative bacteria (see e.g. Li et al., 1995). The substrate profile of the AcrAB–TolC system is exceptionally wide, and it is difficult to find a molecule that is not a substrate of this system. Not only are antibiotics substrates, but Biolog Phenotype microarray data show that phenothiazines, anti-inflammatory agents, detergents and some biocides such as triclosan and cetrimide are also substrates (Bailey et al., 2008). Not only is AcrAB–TolC important in conferring innate MDR in many Enterobacteriaceae, but this system is also required for resistance conferred by several other mechanisms to be elaborated. Fluoroquinolone resistance is usually mediated by mutations in topoisomerase genes such as gyrA, but if acrB or tolC is genetically inactivated, the MIC changes are small and not clinically significant (Oethinger et al., 2000). The same is true for triclosan resistance via mutation in fabI (Webber et al., 2008). Furthermore, resistance mediated by some transmissible genes, e.g. florR (florfenicol resistance) or tetG (tetracycline resistance), is not expressed at clinically significant levels without an intact AcrAB–TolC MDR efflux pump (Baucheron et al., 2004; de Cristóbal et al., 2006). A functional AcrAB–TolC system is also required to select Salmonella resistant to substrates of this pump (Ricci & Piddock, 2009a); likewise the MexAB–OprM system is required in P. aeruginosa (Köhler et al., 1997).

Bile, a natural detergent, is also a substrate of AcrAB–TolC (Lacroix et al., 1996). As the primary reservoir of Salmonella entering the food chain and infecting humans is poultry, and fluoroquinolone-resistant bacteria were isolated from poultry (Wray et al., 1990), we next explored whether Salmonella survive in vivo because AcrAB–TolC confers bile resistance. In collaboration with colleagues at the Veterinary Laboratory Agency we carried out poultry experiments with the parental strain S. Typhimurium SL1344 in a 1 : 1 ratio with a mutant. In the absence of antibiotic the TolC mutant poorly colonized and did not persist (Buckley et al., 2006). The AcrB mutant transiently colonized, but by day 3 the numbers of bacteria were as low as for the TolC mutant. When complemented in trans, the 1 : 1 ratio was maintained. As the AcrB and TolC mutants had normal growth kinetics in laboratory medium (Webber et al., 2009), these experiments showed that the mutants were attenuated in vivo. In parallel, Nishino et al. (2006) showed that AcrAB and TolC were required for virulence in the mouse model of Salmonella infection.

To determine whether attenuation was due to bile susceptibility, experiments were carried out in two tissue culture infection models where there is no bile, the INT-407 intestinal cell system and RAW 264.7 macrophage system. The TolC mutant did not adhere to the tissue culture cells, and neither the AcrB nor the TolC mutant invaded the cells (Buckley et al., 2006). Subsequently, we showed the same was true when acrA was deleted (Blair et al., 2009). These experiments indicate that inactivation or deletion of any of the components of AcrAB–TolC attenuates the organism and that this is not due to bile susceptibility. There is now a considerable body of evidence to show that, irrespective of the bacterial species, inactivation of acrB or homologues attenuates the bacterium so that it is unable to infect its host, be it a plant, a bird or a mammal (Piddock, 2006b).

There are at least three hypotheses for the cause of attenuation associated with lack of AcrAB–TolC. Firstly, it was suggested that E. coli AcrB and TolC exported quorum sensing molecules (Yang et al., 2006). However, this was not reproduced by others in either E. coli or Salmonella (e.g. Webber et al., 2009). Secondly, loss of important proteins affects the production of known virulence effectors. Thirdly, AcrAB–TolC exports a virulence factor. There is evidence to support the second hypothesis. Whole transcriptome analysis with microarrays showed that inactivation of a single component conferred hundreds of gene expression changes (Webber et al., 2009). The greatest number of changes were seen when acrB was inactivated, and not tolC. This was surprising, because TolC is a very promiscuous protein and it was hypothesized that its loss would have the greatest impact. Amongst the differentially expressed genes were many that either regulate expression of acrAB, tolC or porin protein genes or are involved in virulence, including Salmonella pathogenicity island (SPI) genes, flagellar genes and those encoding proteins involved in anaerobic metabolism. Further experiments confirmed the transcriptional data; for instance the AcrB mutant was less motile and did not grow well under anaerobic conditions (Webber et al., 2009). Furthermore, many SPI loci genes had decreased expression, and Western blotting for the effector proteins revealed decreased production.

Regulation of production of multi-drug efflux pumps

In the isolates from Patient B, overexpression of efflux pump genes was associated with clinically relevant MDR due to increased efflux. Expression of genes is regulated within bacterial cells via several layers of control; there are the ‘master’ regulators such as HNS and CRP that have hundreds of genes within their regulon (e.g. Zheng et al., 2004) and there are specific regulators, such as MarA, which has about 60 genes within its regulon and so will target about 60 different promoters (Barbosa & Levy, 2000). There is also post-transcriptional control and post-translational control of genes and proteins, respectively (Browning & Busby, 2004).

The E. coli transcription factor MarA can activate expression of acrAB and tolC (Alekshun & Levy, 1999). If there are mutations preventing repression of marA by MarR, there is overproduction of MarA. Increased produc<1?show=[fo]?>tion of MarA (1) downregulates the expression of ompF (via production of MicF), thereby reducing antibiotic entry, and (2) upregulates the expression of acrAB and tolC, thereby increasing MDR efflux (Alekshun & Levy, 1999). In Salmonella, AcrR is a weak repressor of acrAB and inactivation of acrR had little effect upon antibiotic resistance (Table 3). Furthermore, increased expression of marA or soxS in laboratory mutants, clinical isolates or veterinary isolates of antibiotic-resistant Salmonella was not seen (Piddock et al., 2000; Cloeckaert & Chaslus-Dancla, 2001). However, overproduction of a homologue of MarA, RamA, was detected (Abouzeed et al., 2008; Ricci & Piddock, 2009b).

RamA is a homologue of MarA, SoxS and Rob and they are all members of the XylS–AraC family of transcriptional activators (van der Straaten et al., 2004). Interestingly, ramA is not found in E. coli or Shigella, but it is found in the majority of other Enterobacteriaceae. The ramA gene is co-located with ramR, which encodes a TetR-type repressor. RamA is overproduced when ramR or the target DNA sequence upstream of the ramA target is mutated, leading to lack of repression of ramA. The level of RamA influences the level of MDR (as determined by MIC of antibiotic) (Table 3) (Bailey et al., 2010). Cloning of ramA into pTRC conferred low level ramA expression in a ΔramA strain and under IPTG induction there is high level expression of ramA (Table 3). The level of RamA was also highly associated with the level of acrB expression. Data obtained with clinical and veterinary isolates also indicate that the RamA level can vary (Abouzeed et al., 2008; Ricci & Piddock, 2009b). Therefore, there is regulatory control impacting on how much RamA is produced and consequently the level of MDR.

Whole transcriptome analysis showed that inactivation or overexpression of ramA alters expression of over 200 genes, including acrAB, tolC and virulence genes (Bailey et al., 2010). Data obtained with Salmonella and other Enterobacteriaceae such as Klebsiella and Enterobacter spp. also indicate that RamA, not MarA, is the primary transcription factor regulating expression of acrAB and tolC (Schneiders et al., 2003; Chollet et al., 2004).

The RamR binding region between the ramR and ramA gene has been identified (Baucheron et al., 2012; Ricci et al., 2012). Furthermore, random mutagenesis of ramR in S. Typhimurium SL1344 ΔramR pMW82 ramA : : GFP and screening by flow cytometry led to the isolation of a super-repressor with the same phenotype as when ramA is deleted (Ricci et al., 2012). Expression of ramA is inducible (e.g. by chlorpromazine, an anti-psychotropic drug that is a substrate of AcrB) (Bailey et al., 2010; Lawler et al., 2013). Using chlorpromazine to induce ramA, the half-life of RamA was shown to be approximately 2 min (Ricci et al., 2014). Furthermore, the level of RamA is reset by the Lon protease, which proteolytically degrades RamA (Ricci et al., 2014). The regulation of expression of acrAB and tolC is complex (Fig. 3) and therefore it is no surprise that we could not easily decipher the underlying mechanisms of resistance in the isolates from Patient B with the tools we had in the 1990s.

Using basic research as a platform for drug discovery

In the last 10–15 years, there has been a huge amount of work in industry using genomic platforms to screen for new antibiotic targets or identify new antibacterial agents. Some inhibitors with excellent activity were identified in cell-free systems; unfortunately many had little or no activity. This is because either the molecules did not enter the bacterial cell or they were exported via MDR efflux systems such as AcrAB–TolC (Payne et al., 2007). Therefore, studies to understand resistance mechanisms directly inform those working on drug discovery, research and development programmes.

Since the early 2000s, proteins that comprise MDR efflux pumps have been a target for drug discovery, in the main for P. aeruginosa, seeking inhibitors to combine with fluoroquinolone drugs (Lomovskaya et al., 2001). Efflux is a good target because an efflux inhibitor will decrease the MICs of antibiotics that are already licensed. Therefore, there is no need to find new antibacterial drugs, as there are effective agents and biocides available for combination therapy. Furthermore, inhibition of efflux also inhibits the ability of the bacterium to infect its host, to form a biofilm (Baugh et al., 2012, 2014) and to evolve antibiotic resistance to substrates of the inhibited MDR efflux pump (Ricci & Piddock, 2009a).

There are drug discovery programmes seeking inhibitors of AcrB and/or TolC; however, another avenue of research is to prevent increased efflux, because many of the clinically useful drugs such as ciprofloxacin are effective so long as efflux pumps are not overproduced. This could be achieved by increasing RamR repression of ramA, which would give a similar phenotype to the super-repressor mutant we isolated (Ricci et al., 2012). My team has also described a bacteriophage that interacts with the TolC protein and which could be used to prevent colonizing of poultry by Salmonella (Ricci & Piddock, 2010). We have also investigated numerous different compounds and medicinal plants for efflux-inhibitory activity (Garvey et al., 2011). Transcription factors are also recognized discovery targets, and Levy and colleagues have identified inhibitors of E. coli MarA (e.g. White et al., 1997).

In summary, the work in my laboratory and others has shown that MDR efflux pumps, in particular AcrAB–TolC, confer both innate and attained resistance to clinically important antibiotics and other antibacterial molecules. However, this efflux pump is also required for Salmonella to infect its host. Therefore, the numerous properties afforded by AcrAB–TolC make both these proteins and those that regulate its production targets for antibacterial drug discovery.