The Building Blocks of Antimicrobial Resistance in Pseudomonas aeruginosa: Implications for Current Resistance-Breaking Therapies

1P. aeruginosa is a ubiquitous environmental bacterium capable of establishing opportunistic infections in both plants and animals, including humans. It is the primary cause of Gram-negative nosocomial infections and of lung infections in people with cystic fibrosis (CF), chronic obstructive pulmonary disease (COPD) and non-CF bronchiectasis. P. aeruginosa has a large genome of 5.5-7 million base pairs, with remarkable plasticity. Its ability to adapt to a range of environmental niches and its high nutritional versatility stems from this genome plasticity. In addition, a large variety of intrinsic and acquired resistance mechanisms exist within the P. aeruginosa population. Understanding these mechanisms of resistance and their interplay can help develop targeted therapies for combatting MDR P. aeruginosa infections.

2Antibiotic resistance mechanisms can be broadly divided into three categories; intrinsic, acquired and adaptive. Intrinsic resistance mechanisms are those genetically encoded in the core genome of the organism, whereas adaptive resistance mechanisms are those induced by environmental stimuli, and acquired resistance arises from gain of resistance genes from other organisms or as a consequence of selection of advantageous mutations.

3Intrinsic resistance mechanisms of P. aeruginosa include its low outer membrane permeability (12- to 100- fold lower than that of Escherichia coli), the presence of antibiotic efflux pumps and β-lactamases, such as OXA-50 and AmpC. Acquired resistance mechanisms from horizontal gene transfer include acquisition of transferable aminoglycoside modifying enzymes and β-lactamases, while acquired resistance as a result of de novo mutational events often takes the form of overexpression of efflux pumps and β-lactamases, along with decreased expression or modification of target sites and porins. Acquired resistance through adaptive mutations is common in CF isolates due to the prolonged periods of selective pressure during extended antibiotic treatment of chronic infection and the actions of host immune effectors. Adaptive resistance mechanisms are those that are induced through external stimuli, such as stress factors and the presence of certain antibiotics. This is different from acquired mutational resistance as adaptive resistance is transient and unstable. Adaptive resistance mechanisms are not permanent, unlike mutational events, and become inactive upon removal of the stress factor. Adaptive resistance often involves regulatory pathways and leads to altered gene expression, changes in protein production or target alteration. For example, two-component signaling systems (TCSS), MexXY induction and biofilm formation.

4In this review, we will discuss the different resistance mechanisms P. aeruginosa employs to evade antibiotic action. Intrinsic resistance mechanisms, such as low-outer membrane permeability, efflux, lipopolysaccharide (LPS) modification and the bacterial enzyme AmpC, will be described, with a focus on how these contribute to high-level resistance via interplay with acquired mutations and adaptive mechanisms. In addition, we will outline the roles of bacterial enzymes and commonly acquired resistance mechanisms, as well as phenotypic contributions to resistance, such as motility and biofilm formation. Novel therapeutic efforts that directly target these resistance mechanisms will be reviewed and future priorities discussed.

5Antibiotics must cross the bacterial membrane in order to act on intracellular targets ( Figure 1 ). Accumulation of antibiotics in the cell is intrinsically diminished in P. aeruginosa due to its low outer membrane permeability. The relative dearth of porins within the outer membrane (OM) of P. aeruginosa decreases the rate at which antibiotics can penetrate the cell. Porins not only function in transporting nutrients and other molecules across the membrane, but they also play a role in signaling, adhesion and stability of the membrane. P. aeruginosa has 26 different types of porin, of which OprF is the most abundant in P. aeruginosa and has been implicated in a variety of functions.

6OprF is present in two forms, the highly abundant two-domain closed channel and the single domain open channel. The open channel conformation occurs in <5% of the OprF protein population. The two-domain closed channel conformation is thought to interact with peptidoglycan and therefore to play a structural role by stabilizing the OM. The single domain open channel spontaneously transitions between different sub-conformations that are weakly conductive and exists for only a short time in its fully open state. The low occurrence of the open channel and its weak conductive state aid in the low permeability of P. aeruginosa. Clinical OprF deficient mutants of P. aeruginosa with increased antibiotic resistance have been observed. However, several studies have shown that complete loss of OprF alone does not seem to be a major route to antibiotic resistance amongst clinical isolates. The direct role of OprF in antibiotic resistance is somewhat ambiguous, despite being implicated in quorum sensing and biofilm formation. Interestingly, OprF is detectable in CF patient sputum samples and is upregulated 40-fold in anaerobic conditions, in comparison to aerobic conditions. It has been found to be necessary for growth in the anaerobic environment of CF sputum. Overexpression of OprF was also observed in a P. aeruginosa artificial sputum biofilm model of the CF lung and was shown to be necessary for the tight microcolony formation that occurs during the early stages of biofilm formation in the CF lung. Therefore, although the direct role of OprF in antimicrobial resistance may be ambiguous, there is evidence of a wider role in the establishment of other P. aeruginosa resistance mechanisms during lung infections.

7OprD is the second major porin protein in P. aeruginosa and one of the most well-characterized, due to its involvement in the entry of carbapenem antibiotics, mainly imipenem and meropenem. Carbapenem resistance has been linked to mutations causing downregulation of oprD expression and to OprD amino acid substitution mutations found in hypermutator isolates.

8OprD has been found in high abundance in biofilm outer membrane vesicles (OMVs). OMVs bud from the outer-membrane and are filled with periplasmic components. They have several functions that serve to enhance bacterial survival, such as modulating host immunity, delivering virulence factors, acquiring nutrients and contributing to structural support in biofilms. OprD-rich OMVs are thought to absorb carbapenems through OprD, leading to accumulation in the OMV and simultaneously lowering the carbapenem concentration in the bacterial cell. Similarly, P. aeruginosa may package OprD into OMVs to decrease cellular levels of this porin and thereby reduce the passage of carbapenems across the membrane and into the cell.

9Chronic P. aeruginosa infection in CF can lead to reduced oxygen tension, due to nearby polymorphonuclear leukocyte activity, causing some bacteria to grow anaerobically. Increased antibiotic tolerance is observed in P. aeruginosa that grow in well established, anoxic biofilms. Several studies have found strongly decreased expression of oprD in 48-hour and 96-hour anoxic biofilms, which may partly explain the enhanced resistance to carbapenems in biofilm. This is in contrast to oprF, which is upregulated under anoxic environments such as the CF lung. Increased expression of oprF would not substantially increase membrane permeability, as the closed channel is the most common form of OprF. The increase in closed channel OprF porins along with the decrease in oprD expression may work in conjunction to further lower outer membrane permeability in P. aeruginosa and protect the cell from antibiotic permeation, although this has not been experimentally determined. Alternatively, the benefit derived from increased expression of oprF in the CF lung may relate to its role in QS signaling and/or as a host immune system sensor.

10Khuata et al. found that sialylation occurs in multiple porin proteins in the presence of normal human serum. OprD was identified as a sialoglycoprotein and sialylation of OprD resulted in a decreased interaction with β-lactam antibiotics in comparison to non-sialylated OprD. It was also found that piperacillin and ceftazidime were able to penetrate non-sialylated OprD much more readily than the sialylated form. Accordingly, in the presence of normal human serum, increased survival in the presence of these antibiotics is observed, as a result of OprD sialylation. These observations highlight the effect that mode of growth and host factors can have on the susceptibility of P. aeruginosa to antibiotics.

11OprH is a 21.6 kDa porin protein and is the smallest porin found in P. aeruginosa. OprH interacts with LPS through electrostatic interactions and is implicated in polymyxin resistance through the low Mg2+ concentration sensing PhoPQ and PmrAB TCSS. The genes for the PhoPQ TCSS form an operon with oprH. However, studies with oprH knockout P. aeruginosa have demonstrated that the porin is not essential for polymyxin resistance. Instead, OprH is thought to play a supplementary role in polymyxin resistance through stabilizing the OM during Mg2+ starvation. OprH has also been linked with gentamicin resistance in small colony variants of P. aeruginosa, although it is not clear whether upregulation of the operon alone was responsible for the resistance phenotype, as several other genes were also upregulated.

12Bacterial efflux pumps are categorized into five families, the two ancient super families that are made up of the adenosine triphosphate binding cassette (ABC) and the major facilitator super family (MFS), and three smaller families that consist of the multidrug and toxic compound extrusion family (MATE), the small MDR family (SMR) and the resistance nodulation-cell division family (RND). RND efflux pumps are the main multidrug efflux systems responsible for antibiotic extrusion and resistance in P. aeruginosa. P. aeruginosa possesses at least 12 RND efflux pumps, of which four are of clinical relevance and are the main contributors to antibiotic resistance through efflux, namely, MexAB-OprM, MexCD-OprJ, MexEF-OprN and MexXY-OprM ( Figure 2 ). All RND efflux pumps cross the cytoplasmic membrane, the periplasm and the OM. The RND efflux pump complex is comprised of the secondary active efflux pump protein embedded in the inner membrane, the outer membrane factor protein, and a membrane fusion protein that extends through the periplasm and connects the two. Efflux pumps are powered by proton motor force and are able to export drugs from both the periplasm and the cytosol into the environment. Efflux and the low-permeability OM work in a cooperative manner, whereby those antibiotics that manage to permeate the OM are quickly extruded by efflux pumps, back into the environment.

13Efflux pumps not only function in antimicrobial efflux, but are also recruited in response to cellular stress. Stress response signals including host factors, detergents and other endogenous inducers of the bacterial stress response may function as pressures for the selection of antimicrobial resistant efflux-expressing mutants, even in the absence of antimicrobials. For example, inflammation in the CF lung means P. aeruginosa is continuously exposed to reactive oxygen species (ROS), which is likely to drive the high prevalence of MexAB-OprM and MexXY-OprM over-expressing mutants Additionally, in vitro studies have shown that ROS exposure selects for MexXY-OprM expressing, aminoglycoside resistant mutants.

14The MexAB-OprM efflux pump is constitutively active in P. aeruginosa wild-type strains and, along with other RND efflux systems, causes low level resistance but can also significantly contribute to the development of multi-drug resistance. MexAB-OprM extrudes carbapenems, chloramphenicol, fluoroquinolones, lincomycin, macrolides, novobiocin, penicillins, tetracyclines and all β-lactams (except imipenem and biapenem), as well as the antiseptic triclosan and the surfactant sodium dodecyl sulfate. MexAB-OprM null mutants are markedly susceptible to all of the aforementioned antimicrobials, whilst mutants that overproduce MexAB-OprM show high levels of resistance. MexAB-OprM consists of an inner membrane protein, MexA, and a fusion protein, MexB, that links MexA to the outer membrane protein OprM. The efflux pump proteins are encoded in an operon, the expression of which is regulated by MexR. MexR is encoded upstream of the mexAB-oprM operon and acts as a transcriptional repressor by binding to the mexA-mexR intergenic region near the promotors of mexR and a secondary promotor for mexA. Loss of a functioning MexR causes hyperexpression of mexAB-oprM. Mutations in mexR found in clinical P. aeruginosa isolates can lead to the expression of MexR protein that is no longer able to dimerize or bind to DNA to repress mexAB-oprM expression. In addition, mutations leading to premature termination of the MexR peptide resulting in complete absence of the production of a functioning MexR have been identified in clinical isolates.

15MexCD-OprJ is not normally expressed in wild-type strains and is functional in nfxB mutants only, it is therefore not classified as contributing to the natural resistance of P. aeruginosa. Antibiotic substrates of the MexCD-OprJ efflux system are chloramphenicol, erythromycin, fluoroquinolones and tetracycline. The NfxB protein negatively regulates MexCD-OprJ and different mutations within the nfxB gene lead to varying levels of MexCD-OprJ expression, leading to clinical isolates with different levels of resistance to antimicrobials.

16The MexEF-OprN efflux system extrudes chloramphenicol, fluoroquinolones, tetracycline, trimethoprim and imipenem. The mexEF-oprN operon is regulated by the transcriptional activator MexT. MexT positively regulates mexEF-oprN expression, in contrast to most other RND-efflux systems, which are negatively regulated by a repressor protein. However, often wild-type strains of P. aeruginosa carry an inert mexT gene encoding inactive MexT, causing the suppression of mexEF-oprN expression. Besides acting as a positive regulator for the expression of the mexEF-oprN efflux system when MexT is active, it also acts as a repressor of oprD transcription. Active MexT leads to a reduction of the OprD porin in the bacterial membrane, causing increased resistance to carbapenem antibiotics. The multidrug resistant nfxC-type mutant overexpresses the mexEF-oprN operon and is resistant to chloramphenicol, fluoroquinolones, tetracycline, trimethoprim and imipenem.

17The MexXY-OprM efflux system contributes to P. aeruginosa intrinsic resistance to aminoglycosides, tetracycline, erythromycin and cefepime. MexXY-OprM is regulated by the MexZ repressor protein and mutations in or around mexZ lead to overproduction of MexXY-OprM. The mexZ gene is one of the most commonly mutated in isolates from people with CF, leading to loss of MexXY-OprM repression and thereby increasing its expression. Mutations in mexZ that arise during long-term infection in CF can result in resistance to tobramycin, a first-line antibiotic in CF clinics. One study looking at levels of mexY expression and mexZ mutation found that both were present at a significantly higher frequency in adults with chronic infection than in children with chronic or new infections and that mexZ mutation strongly correlated with aminoglycoside resistance. This suggest that mexZ mutation undergoes positive selection during long term chronic infection.

18LPS is a central component of the Gram negative bacterial membrane. It serves as the outer leaflet of the OM and is essential for host-pathogen interactions, virulence and antimicrobial resistance. P. aeruginosa LPS is composed of three domains: the core oligosaccharide, the lipid A portion and the O-antigen region. Modification of LPS is a resistance mechanism that protects the bacterial cell against cationic antimicrobial peptides (CAPs) and polymyxin antibiotics. LPS modification in response to a low Mg2+ environment, or to exposure to CAPs and polymyxins, is achieved by the addition of a 4-amino-4-deoxy-L-arabinose (L-Ara4N) moiety to the phosphate unit on the lipid A group of the LPS. Modification of the lipid A arabinose subunit leads to a lower net negative charge of the bacterial membrane, which leads to a decrease in electrostatic interactions with positively charged antimicrobials. The decrease in electrostatic interaction between the antibiotic and the membrane prevents the drug from binding the LPS which would normally destabilize and disrupt the membrane. LPS modification in P. aeruginosa is regulated by several two-component signaling systems (TCSS): PmrAB, PhoPQ, CprRS and ParRS. These TCSS regulate the transcription of the arnBCADTEF operon, which encodes the genes responsible for the addition of L-Ara4N to the lipid A group, in response to a variety of environmental signals. PmrAB and PhoPQ regulate the activation of arnBCADTEF under Mg2+ deficient growth conditions. ParRS and CprRS both play an active part in adaptive resistance against CAPs and can stimulate arnBCADTEF upregulation, independently of each other, in the presence of these peptides. Mutations in these regulatory TCSS that lead to the constitutive activity of the arn operon confer colistin resistance to P. aeruginosa. One study found that mutations in the pmrB histidine kinase gene of the PmrAB TCSS commonly arise during the development of high level polymyxin resistance in experimental evolution experiments performed with CF isolates. Conversely, a missense mutation in pmrB in isolates taken from a murine model of P. aeruginosa chronic respiratory infection has been found to be associated with increased susceptibility to a wide range of antibiotic classes but also with resistance to the host antimicrobial enzyme lysozyme. High level resistance to polymyxin antibiotics is forged through epistatic interactions between multiple regulatory genes and can occur via several pathways, but a functional arn operon and the aminoarabinosylation of the lipid A are requirements for polymyxin and colistin resistance in P. aeruginosa.

19β-lactam antibiotics interfere with cell wall synthesis and disrupt peptidoglycan recycling by inhibiting penicillin binding proteins (PBP), enzymes responsible for synthesis and maintenance of peptidoglycan. β-lactamases are enzymes that interfere with this process by hydrolyzing the β-lactam ring that is present in all β-lactam antibiotics. β-lactamases are classified according to different systems, the Ambler classification groups enzymes according to amino acid sequence and the Bush-Jacoby-Medeiros system classifies enzymes according to function and phenotype.

20The β-lactamase AmpC is an Ambler class C antibiotic inactivating enzyme. It contributes to the natural resistance of P. aeruginosa and is constitutively produced at basal levels. The chromosomal blaAmpCgene is drug inducible with aminopenicillins and cephalosporins, whereupon AmpC inactivates the β-lactam antibiotic through the process of hydrolysis. AmpC induction occurs through two different mechanisms: perturbation of the peptidoglycan recycling pathway and through loss of PBP4. Peptidoglycan is composed of cross-linked, alternating N-acetylglucosamine (GlcNAc) and N-acetylmuramic acid (MurNAc) residues. Peptidoglycan recycling requires the collaboration of PBPs, endopeptidases, carboxypeptidases, N-acetylmuramoyl-l-alanine amidases and lytic transglycosylases (LTs). LT’s cleave the β-1,4 glycosidic linkage between GlcNAc and MurNAc residues during growth and division, to generate new material to be inserted into the peptidoglycan layer. This cleavage causes the release of anhydromuropeptides (anhMPs). The anhMPs are transported to the cytoplasm through an inner membrane permease AmpG, where they are processed for recycling back into the peptidoglycan biosynthesis pathway by the β-N-acetylglucosaminidase NagZ and the N-acetylmuramoyl-l-alanine amidase AmpD. Disruption of anhMP processing leads to their accumulation in the cytoplasm where they can bind to AmpR, the transcriptional regulator of AmpC. Two anhMPs, N-acetylglucosamine-1,6-anhydro-N-acetylmuramyl pentapeptide and 1,6-anhydro-N-acetylmuramyl pentapeptide, have been identified as signaling molecules that activate β-lactamase expression in P. aeruginosa.

21Often β-lactamase resistance in clinical P. aeruginosa strains is caused by inactivation of the PBP4 encoding gene dacB, causing ampC overexpression and activation of the CreBC (BlrAB) TCSS. PBP4 functions to create an anhydromuropeptide (anhMP) that controls AmpR. Loss of PBP4 leads to the disappearance of this specific anhMP from the peptidoglycan recycling pathway, which changes AmpR regulation, thereby inducing ampC expression. Upregulation of ampC is commonly observed amongst P. aeruginosa isolates from people with CF, it is selected for by clinical use of β-lactam antibiotics and the presence of hypermutable strains in the CF lung can hasten the emergence of this mutational overexpression.

22Metallo β-lactamases (MBLs) belong to Ambler classification group B, they contain zinc in the active site, as opposed to a serine based hydrolytic system. The most clinically relevant MBLs in P. aeruginosa are the Verona integron-encoded MBLs (VIM) and active-on-imipenem type MBLs (IMP). They are often located on integrons along with other resistance genes, which allows these genes to integrate into the chromosome or a plasmid and disseminate through bacterial populations. The presence of MBL causes high-level resistance against carbapenem antibiotics.

23Extended spectrum β-lactamases (ESBL) hydrolyze oxyimino-aminothiazolyl cephalosporins (cefotaxime, cefuroxime, cefepime, ceftriaxone and ceftazidine) as well as penicillins and other cephalosporins, excluding cephamycins. The most common ESBL in P. aeruginosa are the Ambler class A SHV-, VEB-, PER- and GES-type enzymes and the OXA-type enzymes from Ambler class D. The presence of ESBLs confers a higher level of resistance to extended spectrum cephalosporins than AmpC, impermeability and efflux pump hyper-expression combined. However, ESBL enzymes can be inhibited by clavulanic acid, sulbactam and tazobactam. ESBL producing P. aeruginosa have been frequently isolated, although their detection is complicated by the activity of chromosomally encoded β-lactamases in P. aeruginosa.

24Aminoglycoside modifying enzymes (AME) are capable of catalyzing the modification of the -OH or -NH2 groups of the 2-deoxystreptamine nucleus or the sugar moieties of aminoglycoside antibiotics. There are three types of AMEs; acetyltransferases (AACs), nucleotidyltransferases (ANTs) and phosphorlyl transferases (APHs). All of these decrease the binding affinity of the antibiotic to its target by transferring a functional group onto the aminoglycoside. All three types of AMEs have been observed in P. aeruginosa and are common determinants of aminoglycoside resistance in non-CF isolates. AMEs are not, however, the most common aminoglycoside resistance mechanism in people with CF. Instead, impermeability and efflux are the more commonly observed events. However, studies have observed a variable prevalence of these enzymes in between 5.6% and 52% of isolates.

2516S rRNA methylases are another group of modifying enzymes carried by P. aeruginosa. These enzymes are carried on plasmids and are thought to have come from an aminoglycoside producing organism that utilizes these enzymes to protect its own ribosomes from aminoglycoside action. 16S rRNA methylases cause the methylation of a specific nucleotide of the 16s rRNA A-site, to protect the ribosome from aminoglycoside activity. Strains carrying 16s rRNA methylase genes show a high level of resistance to most clinically relevant aminoglycoside antibiotics. The prevalence of these enzymes is not common in CF and other respiratory associated infection isolates However, rmtD carrying P. aeruginosa have been reported in Africa, Asia, Europe and South America and are, along with ESBLs, often associated with a pan-resistant genotype.

26DNA gyrase and topoisomerase IV are both involved in DNA replication and management of chromosome integrity. DNA gyrase functions to cleave double stranded DNA, introduce negative supercoiling and re-ligate the strand in an ATP-dependent reaction. Topoisomerase IV functions by decatenating and relaxing positively supercoiled DNA. Fluoroquinolones like ciprofloxacin act on DNA gyrase and topoisomerase IV to prevent the re-ligation step, leaving the DNA cleaved and thereby acting as a bacteriostatic antibiotic, blocking DNA replication.

27Resistance against ciprofloxacin occurs most commonly through mutations in the DNA gyrase subunits GyrA and GyrB or the topoisomerase subunits ParC/ParD but also through the overexpression of efflux. Mutations in gyrA and gyrB are commonly identified in CF associated infections and are likely to arise during prolonged antibiotic treatment with ciprofloxacin. The most frequently observed mutations in both CF and non-CF isolates, as well as in in vitro experiments, are the substitution of an amino acid at position 83 in GyrA from threonine to isoleucine and an amino acid substitution of serine to leucine at position 87 in ParC. The second most common mutation is GyrA position 87, where tyrosine is substituted by asparagine, glycine or tyrosine. These positions of GyrA are important for the binding of ciprofloxacin to DNA gyrase and the amino acid substitutions greatly lower the affinity of this binding. Resistance to ciprofloxacin is higher when both mutations in gyrA and parC are present, in comparison to only gyrA mutation. However, it is thought that gyrA mutation is necessary for high level resistance against ciprofloxacin, due to the strong affinity of the native form of the protein for this antimicrobial. When only gyrB, parC or parD mutations are present or when efflux is the main mechanism of ciprofloxacin resistance, only low level resistance is observed. A combination of efflux, gyrAB and parCD mutation leads to a higher level of resistance and again highlights that resistance in P. aeruginosa is complex, often involving multiple mechanisms.

28Mutations arising in the fusA1 gene lead to amino acid substitutions in a critical part of the translational machinery, elongation factor G (EF-G). Mutations within different domains of EF-G have been identified in clinical strains and are linked to decreased aminoglycoside susceptibility. Mutations in fusA1 have been reported in both CF and non-CF isolates globally. The mutation is often paired with upregulation of efflux, mainly the MexXY-OprM efflux pump, leading to strong aminoglycoside resistance. However, on its own, fusA1 mutation can cause strains to be 4- to 8- fold more resistant to aminoglycosides. This was shown in vitro, where three fusA1 mutants in domains II, III and IV were created. The molecular mechanism linking mutations within EF-G to increased aminoglycoside resistance remains unclear. Aminoglycosides do not bind to or target EF-G, thus resistance to aminoglycosides must be an indirect effect of EF-G alteration.

29Bacterial motility is a vital part of P. aeruginosa pathogenicity and plays a role in host colonization and establishment of infection. P. aeruginosa employs several different forms of motility that are environmentally dependent and often require the use of type IV pili, flagella or rhamnolipid surfactants. RNA-seq and transposon mutant studies have shown that both swarming and surfing phenotypes are associated with significant changes in gene expression, leading to an increase in antibiotic resistance. Swarming motility was linked to the altered regulation of 1,581 genes, including 104 regulatory genes, including transcription factors, TCSS and sigma factors. Surfing motility was linked to altered regulation of up to 2,078 genes and out of these, 31 resistome genes were identified which were known to cause an increase in antibiotic susceptibility when mutated.

30Both swarming and surfing play a role in adaptive resistance. They are environmentally dependent and associated with conferring a high level of antibiotic resistance in comparison to alternative motility phenotypes. Swarming cells have been shown to be significantly more resistant to aminoglycosides, β-lactams, chloramphenicol, ciprofloxacin, tetracycline, ethoprim, erythromycin and azithromycin. Surfing cells however, showed significantly higher resistance to polymyxins, aminoglycosides, fluoroquinolones, tetracycline, chloramphenicol, trimethoprim and several β-lactams, but not to macrolides. On semi-solid surfaces with low nitrogen concentrations, P. aeruginosa is known to utilize swarming motility. The conditions under which P. aeruginosa exhibits swarming motility appear to mimic those of the lung, where semi-solid mucous overlays epithelial cells. Likewise, the CF and bronchiectasis lungs contain high levels of mucin, the component necessary for surfing motility. Although P. aeruginosa motility is often lost in long term chronic illness such as in CF, a plethora of genetic heterogeneity exists in the bacterial populations residing in the CF lung. Furthermore, swarming and surfing motility may occur during early colonization and adaptation to the CF lung.

31The biofilm mode of growth is a major impediment in the struggle to eradicate P. aeruginosa infection from the CF lung, due to the increased ability of bacteria in biofilms to withstand antibiotic treatment. Biofilms are composed of bacteria surrounded by extracellular polymeric substances like exopolysaccharides, extracellular DNA and polypeptides. This can lead to an increase in tolerance to antimicrobial agents of 100-1000 times, compared to planktonic cells. Contributing features include quorum sensing, decreased ability to penetrate biofilm, presence of oxygen gradients, altered metabolism and slow bacterial growth rate. Biofilms are common in chronic P. aeruginosa infection of the CF lung but have also been shown to be of significance in patients with COPD, bronchiectasis and chronic wounds. Airway mucins are found in abundance in people with CF, COPD and bronchiectasis. In vitro studies have shown that the presence of airway mucins are fundamental for the development of biofilm structures with enhanced tolerance to antimicrobials. Further, mucin may serve as a suitable attachment surface for P. aeruginosa biofilm formation.

32Iglesias et al. investigated antibiotic pharmacodynamics within biofilm structures in the context of CF. Bacterial counts, metabolic activity and biomass of PAO1 biofilms grown in artificial sputum media (ASM) or trypticase soy-based medium were compared. They found that bacteria in ASM reached the same CFU and metabolic activity as biofilms formed in trypticase soy-based medium, although ASM biofilms grew slower and had a marginally higher biomass. When both biofilms were subjected to antibiotic treatment, ASM grown biofilms were substantially more resistant to tobramycin and colistin, but not ciprofloxacin and β-lactams. These studies indicate that environmental factors play an important role in determining biofilm structure and antimicrobial resistance. For comprehensive reviews on P. aeruginosa biofilms refer to: Olsen, 2015 and Maurice, Bedi and Sadikot, 2018.

33Multidrug resistance is often the result of the acquisition of external resistance genes and/or through mutational resistance. Combining resistance genes encoding mechanisms previously described here, along with intrinsic resistance, can lead the development of MDR and XDR P. aeruginosa. MDR and XDR P. aeruginosa arise due to the compilation of resistance mechanisms, amassing building blocks into a barrier of resistance strategies that protect the bacteria from antimicrobial assault. Intrinsic resistance forms the base layer of this barrier, with additional acquired and adaptive resistance mechanisms forming further layers, collectively constructing a robust barrier against antimicrobials ( Figure 3 ). Antibiotic resistance genes can quickly and easily be disseminated through P. aeruginosa populations en masse, via megaplasmids (>420kb). Megaplasmids contain dynamic accessory genomes where frequent recombination and duplication events take place, leading to diverse and adaptive multidrug resistance traits. Accordingly, resistance to antimicrobial agents in P. aeruginosa clinical isolates is highly complex, with frequent interplay between intrinsic, adaptive and acquired resistance mechanisms. AmpC, low outer membrane permeability and efflux systems often work together in resistance to carbapenems, chloramphenicol, fluoroquinolones, macrolides, penicillins, tetracyclines and β-lactams and resistance may be enhanced through the accumulation of mutations leading to up- or down- regulation of each of these systems. The outer membrane porin OprH works in conjunction with the two-component signaling systems PhoPQ and PmrAB in modifying the bacterial LPS to regulate protection to polymyxin antibiotics. Protection against polymyxin antibiotics is additionally mediated via TCSS ParRS and/or CprRS in the presence of cationic antimicrobial peptides. Aminoglycoside resistance is achieved through mechanisms such as aminoglycoside modifying enzymes, fusA1 mutation, 16S rRNA methylation, along with MexXY-OprM upregulation. Likewise, gyrA and parC mutation leading to altered DNA gyrase and topoisomerase IV leads to fluoroquinolone resistance which can be heightened through the cooperation of MexAB-OprM and/or MexCD-OprJ, as well as MexEF-OprN. In addition, the presence of β-lactamases such as ESBLs and MBLs may further enhance carbapenem, cephalosporin and penicillin resistance. A perfect storm of highly resistant P. aeruginosa and a “dry pipeline” of traditional antimicrobials are driving innovation in novel therapeutic approaches that directly target resistance mechanisms. Counteracting these mechanisms could prolong the life of existing antimicrobials.

34Increasing OM permeability to hydrophobic and amphiphilic compounds challenges the issue of intrinsic low-outer membrane permeability. For example, polymyxin B nonapeptide causes a 2- to 40-fold increase in susceptibility to ciprofloxacin, norfloxacin, and ofloxacin and 80- to 200-fold increase in susceptibility to nalidixic acid. Nalidixic acid is the precursor to ciprofloxacin, norfloxacin, and ofloxacin, and is not ordinarily more efficacious than the optimized antibiotics currently used in clinical practice. However, when the membrane becomes more permeable through polymyxin B nonapeptide treatment, nalidixic acid becomes a more powerful anti-pseudomonal. Polymyxin B nonapeptide is highly toxic and has therefore never been considered for clinical application. However, more than 30 years after the first report of OM sensitizers, three have been approved for clinical studies; the anti-protozoal pentamidine and the polymyxin B analogues SPR206 and SPR741. None of these OM sensitizers, with the exception of SPR206, have been proven to have promising anti-pseudomonal activity. SPR206 is active against P. aeruginosa with a similar potency to polymyxin B and is currently undergoing phase-one clinical trial. OM sensitizers appear to be a promising approach to resistance that can by-pass intrinsic, acquired and spontaneous resistance. However, further investigation into additional non-toxic compounds that act on the P. aeruginosa OM is essential, especially for the treatment of polymyxin resistant isolates.

35The most widely researched P. aeruginosa efflux-pump inhibiter (EPI) is Phe-Arg-β-naphthylamide (PAβN), a broad-spectrum peptidomimetic compound capable of interfering with all four clinically relevant P. aeruginosa RND efflux pumps. PAβN potentiates chloramphenicol, fluoroquinolones, macrolides, ketolides, oxazolidinones and rifampicin but not aminoglycosides or β-lactams. The proposed mechanism of action for PAβN is that it functions as a substrate of the Mex-series efflux pumps and outcompetes the antibiotic for extrusion, preventing the antibiotic from leaving the cell. However, PAβN and derivatives of this compounds are not yet approved, as adverse toxicology and pharmacokinetic profiles were identified during phase 1 clinical trials. Instead, PAβN is solely utilized for research on EPI’s and AMR in vitro, rather than as a therapeutic, and may be used to validate the discovery of future EPI’s.

36The pyridopyrimidine derivative D13-9001 is active against the MexAB-OprM efflux pump. D13-9001 has shown promising in vitro and in vivo activity, as well as high solubility and low-toxicity profiles. D13-9001 obstructs normal functioning of MexAB-OprM in two ways. Firstly, it prevents conformational changes by binding tightly to the hydrophobic trap. Secondly, it prevents substrate binding to MexB, through the interaction of the D13-9001 hydrophilic component and the substrate binding channel of MexB. EPIs will need to be broad-spectrum if they are to be used as an adjuvant to antibiotics that are substrates of several efflux-pumps. The specificity of D13-9001 would limit its usage to co-administration with antibiotics extruded exclusively by MexAB-OprM. However, a study by Ranjitkar et al. found that there are several mechanisms of resistance to D13-9001 potentiator activity in P. aeruginosa, when the agent is used together with carbenicillin, an antibiotic that is substrate specific to MexAB-OprM. Loss of potentiating activity of D13-9001 occurred rapidly due to a F628L substitution in mexB, which is known to play an important role in inhibitor binding.

37Polyamines are aliphatic carbon chains containing several amino groups and are essential organic polycations present in every form of life. Polyamines are implicated in cell maintenance and viability and in the functioning of a wide array of organ systems, including, the nervous and immune systems.

38Fleeman et al. identified a polyamine scaffold as a strong efflux pump inhibitor with no direct antimicrobial activity. Five lead agents were found to potentiate aztreonam, chloramphenicol and tetracycline by causing a 5- to 8-fold decrease in the MIC90. In addition, the polyamine derivatives did not disrupt the bacterial membrane, unlike other polyamines, which can lead to the identification of false positives for EPIs. Moreover, polyamines did not display toxicity to mammalian cell lines and did not inhibit calcium channel activity in human kidney cells.

39Phage therapy, the use of bacteriophages to infect and lyse bacterial cells, has been widely discussed. Traditional phage therapy involves the administration of one, or a mixture, of phages that will invade the bacterial cell and clear infection. A different approach to phage therapy has been proposed, whereby phages would be used to steer antibiotic resistance evolution, selecting for phage resistance and antibiotic susceptibility. For example, the lytic Myoviridae bacteriophage, OMKO1, utilizes OprM of the multidrug efflux systems MexAB and MexXY as a receptor-binding site. Selection for resistance to OMKO1 bacteriophage attack creates an evolutionary trade-off in MDR P. aeruginosa, by changing the efflux pump mechanism, leading to an increased sensitivity to ciprofloxacin, tetracycline, ceftazidime and erythromycin, four drugs from different antibiotic classes. Phage steering can be achieved when the binding receptor for the bacteriophage is implicated in both antibiotic resistance and phage resistance. The advantage of this approach lies in the two distinct, and opposing, mechanisms leading to bacterial eradication.

40The prototypical example of successful anti-resistance therapeutics are the β-lactamase inhibitors. β-lactamase inhibitors such as clavulanic acid, sulbactam and tazobactam are widely used to combat resistance mediated by β-lactamases. However, the majority of clinically used β-lactamase inhibitors have a limited spectrum and mainly target Ambler class A β-lactamases, excluding KPC-type β-lactamase. Progress has been made in the development of novel β-lactamase inhibitors with a wider spectrum of activity. Three novel β-lactamase inhibitors, avibactam, vaborbactam and relebactam, function against Ambler class A, C and D β-lactamases. However, only avibactam and relebactam are efficacious against P. aeruginosa infection.

41Lamut et al. designed 4,5,6,7-tetrahydrobenzo[d]thiazole-based DNA gyrase B inhibitors and incorporated these inhibitors with siderophore mimics. The siderophore mimic served as an inducer for increased uptake of the gyrase B inhibitors into the bacterial cytoplasm. Out of the ten gyrase B inhibitors tested against P. aeruginosa, four were able to inhibit ≥50% but only under iron-supplemented conditions, which is not reflective of the host environment during infection. Several more attempts have been made at designing broad-spectrum anti-bacterial and anti-biofilm therapies targeting DNA gyrase or topoisomerase but none have shown good activity for P. aeruginosa.

42Murepavadin is a novel, non-lytic, species specific, outer-membrane protein targeting antibiotic for the treatment of P. aeruginosa infections, including those caused by MDR strains. Murepavadin is derived from the β-hairpin host defense molecule protegrin 1 (PG-1) and optimized to counteract unfavorable absorption, distribution, metabolism, excretion and toxicity (ADMET) properties normally associated with PG-1. It is a macrocycle compound consisting of PG-1 loop sequences linked to a D-proline-L-proline sequence, the latter of which is important for its stability and subsequent strong antibacterial potential. Murepavadin functions through binding to the LPS transport protein D (LptD), an OMP necessary for LPS biogenesis in Gram-negative bacteria. The interaction between murepavadin and LptD causes inhibition of LPS transport, which leads to alterations of the LPS on the bacterial OM and eventually, cell death.

43Murepavadin derivatives have been screened for activity against Gram-negative ESKAPE pathogens, including P. aeruginosa. These derivatives were initially shown to be effective, although the majority of leads were found to have high MIC values in the presence of 50% human serum and showed signs of lytic activity in human red blood cells. Therefore, compounds were generated consisting of β-hairpin macrocycles linked to the peptide macrocycle of polymyxin B. One of these compounds, compound 3, showed strong antimicrobial activity (MIC 0.5-2 μg/mL for P. aeruginosa isolates), low toxicity to mammalian cells, low plasma protein binding, good human plasma stability and no lytic activity towards human red blood cells. This compound was shown to perturb and permealise the bacterial membrane through interacting with the β-barrel domain of BamA in E. coli ATCC 25922. BamA is part of the β-barrel assembly machinery BAM complex, which serves to fold and insert outer membrane proteins in the OM. The binding interaction between BamA and compound 3 locks BamA in its closed state through changing the conformational composite in the β-barrel lateral gate between open and closed states. It is not known what causes compound 3 to permeabilize the membrane. It may inhibit the folding activity of the BAM complex, leading to incorrectly folded proteins being misplaced in the inner membrane. Alternatively, BamA may only serve as an extra binding site for compound 3, thereby evading the LPS-modification resistance mechanism of Gram-negative pathogens.

44Quorum sensing regulates a wide range of genes involved in virulence and bacterial adaptation. For instance, QS is required for the surfing and swarming motility phenotypes associated with increased resistance to antimicrobials. The surfing phenotype is regulated via three QS systems in P. aeruginosa; Las, Rhl and Pqs. In addition, QS has been found to influence tolerance to antibiotics in P. aeruginosa biofilms. QS provides structural rigidity through the regulation of Pel polysaccharides and eDNA release necessary for the extracellular polysaccharide matrix. In addition, the production of rhamnolipids, surfactants important for the establishment and maintenance of biofilms, is controlled under QS. Therefore, QS has been recognized as a significant potential target for developing anti-resistance therapies. Strategies to combat antimicrobial resistance by targeting adaptive resistance mechanisms have significant potential for reversing antibiotic resistance in P. aeruginosa. Adaptive resistance is often mediated through complex global regulatory systems, such as the QS system, and regulate an extensive set of genes involved in resistance. Targeting these regulatory systems may prevent the activation of expression of these resistance genes that would normally be expressed under the environmental conditions of infection.

45Ajoene is a natural sulphur-containing compound extracted from garlic that has been shown to modulate biofilm formation by inhibiting QS-induced production of virulence factors. It targets the Gac/Rsm component of QS, leading to a decreased expression of rsmZ and rsmY small regulatory RNAs. rsmZ and rsmY bind the global regulatory protein RsmA, and unbound RsmA represses the translation of genes by preventing ribosome binding to the Shine-Dalgarno site. Several genes involved in QS are under RsmA regulation, and low expression of rsmZ and rsmY in the presence of ajoene promotes RsmA mediated repression of these target genes. However, the therapeutic applicability of ajoene is limited due to availability, instability, hydrophobicity and relatively high MIC values. Efforts are being undertaken to overcome these issues through modification, the use novel delivery systems and a targeted route of administration and through the development of synthetic ajoene analogues.

46Another novel QS-inhibitor derived from a natural source is the plant flavonoid naringenin. Naringenin diminishes the production of QS-regulated virulence factors in P. aeruginosa by binding directly to LasR, thereby competing with the activator of LasR, N-(3-oxo-dodecanoyl)-l-homoserine lactone (HSL).

47It is ineffective at outcompeting HSL when the activator is already bound to LasR. Thus, the QS-inhibitor will only sufficiently interfere with the QS response when administered during early exponential growth, when naringenin can compete with unbound HSL for LasR binding. Naringenin is only suitable for combatting P. aeruginosa populations at low cellular densities, which often does not represent the clinical infection scenario. The full potential of QS-inhibition will only be realized if an inhibitor is developed that is capable of targeting P. aeruginosa QS signaling regardless of bacterial density and/or QS status.

48Bacterial biofilms pose a physical barrier for drug penetration, which is one of the reasons that bacteria in a biofilm mode of growth are more resistant to antimicrobials. This phenomenon may be subverted with the use of nanocarriers that encapsulate antimicrobials and facilitate drug diffusion through the bacterial biofilm. In addition, nanocarriers can also protect drugs from degradation, ensure controlled drug release, and cause increased uptake by the drug target, leading to an overall higher efficiency of encapsulated drugs. Drug delivery methods can be diverse in chemical structure and nature ( Table 1 ). Most published studies concur that encapsulated antibiotics are more effective at preventing or eradicating biofilm formation than their free drug counterpart, although several concluded that the activity of encapsulated and free antibiotic was equal.

49Another promising antibiotic adjuvant targeting biofilms is the non-bactericidal, inhaled adjuvant, nitric oxide (NO). Exposure of P. aeruginosa biofilms to low-dose NO has been shown to cause dispersal of biofilms, rendering the infection susceptible to subsequent antibiotic treatment. NO functions by increasing bacterial phosphodiesterase activity which, in turn, leads to a reduction in the vital secondary signaling messenger, cyclic di-GMP. Cyclic di-GMP is vital for intracellular regulation of biofilm formation. Howlin et al. carried out in vitro biofilm studies using CF sputum clinical samples. Biofilms were treated with NO only, tobramycin only, tobramycin and ceftazidime, NO + tobramycin and NO + tobramycin and ceftazidime. Biofilms treated with NO showed a relative decrease in biofilm biomass and surface bound thickness in comparison to the untreated control. Further, treatment of biofilms with NO + tobramycin and NO + tobramycin and ceftazidime led to complete eradication of biofilm biomass and surface bound thickness. In comparison, biofilms treated with tobramycin or tobramycin and ceftazidime led to a 243% and 155% increase in biofilm biomass and a 199% and 174% increase in biofilm thickness, respectively, when compared to the untreated biofilm control. There is some evidence for the safety of NO administration in CF patients in vivo and NO is currently undergoing clinical trials to measure clinical efficacy.

50Poly-acetyl-arginyl-glucosamine (PAAG), also called SNSP113, is a novel inhaled adjuvant therapy currently undergoing phase one clinical trials. PAAG is a polycationic glycoprotein that functions by permeabilizing the bacterial membrane and is active against methicillin resistant Staphylococcus aureus (MRSA), Burkholderia spp., Mycobacterium spp. and E. coli. PAAG has been shown to effectively disperse Burkholderia cepacia complex biofilm structures extracted from the CF lung. In P. aeruginosa, PAAG is has been shown to effectively eradicate persister cells, which is important for the prevention of recurrent P. aeruginosa infections and subsequent exacerbations in people with CF. In addition to serving as an effective antibiotic adjuvant, PAAG also reduces inflammation and promotes viscoelasticity and mucociliary clearance, making it a suitable drug candidate to improve the quality of life for patients with a variety of mucus diseases.

51The global overuse and misuse of antibiotics during the last 80 years has led to a profound increase in antimicrobial resistance. Between 2000 and 2010, global antibiotic consumption increased by nearly 70% and antibiotic resistant infections have accordingly become more pervasive, according to global epidemiological antimicrobial resistance surveillance networks. AMR is a complex, One Health issue, involving human, animal and environmental factors. The solution to AMR is therefore also likely to be a complex one, involving multiple strategies; maintaining AMR surveillance, containing AMR transmission, reducing selection pressure, developing novel antimicrobials or reverting antibiotic resistant microbes back to the susceptible phenotype with the use of antibiotic adjuvants. Although progress in the development of naturally derived and peptide-based antimicrobials has been made. The conservation of existing antibiotics through careful stewardship is paramount to help mitigate the gap between the demand for new drugs and the diminishing supply pipeline. Antibiotic adjuvants will also play an important role in extending the shelf life of our existing antimicrobial therapeutic agents.

52Adjuvant strategies targeting resistance mechanisms in P. aeruginosa could rejuvenate traditional antibiotic therapy by potentiating drug activity as well as slowing the development of antibiotic resistance. As described in this review, antimicrobial resistance in P. aeruginosa is regulated through a complex interplay of mechanisms. Resistance encoded in the core genome of P. aeruginosa, such as low outer membrane permeability, Mex-type efflux pumps and AmpC β-lactamase amount to the basal level of resistance against antimicrobials. This intrinsic resistance is present in the all P. aeruginosa strains and serves as a foundational level, which can be expanded upon. This expansion can be induced by environmental influences, such as host factors and signalling molecules, that switch on adaptive resistance mechanisms. Acquired resistance mechanisms, such as antibiotic target modifications generated via mutation, and antibiotic modifying enzymes or resistance plasmids, acquired by gene transfer, may serve as additional building blocks to expand the arsenal of resistance mechanisms a particular strain might carry. Several novel therapeutic strategies, targeting one or more of these mechanisms, have been described in this review. In light of recent findings, OM perturbants capable of sensitizing the Gram-negative bacterial membrane to previously non-active antibiotics seem an opportune strategy to combat resistance. OM perturbants can by-pass intrinsic as well as acquired and spontaneous resistance mechanisms, making them highly promising drug candidates, for which the development of resistance would be unlikely. However, efforts to finding perturbants suitable for targeting the P. aeruginosa membrane must be increased.

53A second promising strategy is phage steering, which uses the natural predators of bacteria and the forces of evolutionary pressure to our advantage. Counterbalancing antibiotic resistance with phage susceptibility creates a double edged sword to circumvent key AMR mechanisms. In addition to these strategies, adjuvants targeting adaptive resistance mechanisms are worthy of consideration, due to the potential to disrupt multiple bacterial resistance and virulence processes with agents targeting a single regulator. Targeting global regulatory systems that would normally control the expression of resistance genes under infection conditions will prevent the activation of those genes with potential knock-on advantages in inhibition of virulence mechanisms. QS and two-component signaling systems are particularly attractive targets from this perspective, as are the regulators of biofilm formation.

54There are several challenges in developing resistance-breaking therapy for P. aeruginosa infection. Firstly, due to its comparatively large genome and highly adaptive nature, a plethora of regulatory systems, as well as limited drug penetration and active efflux, many antibiotic adjuvants designed for Gram-negative pathogens do not show efficacy against P. aeruginosa. Secondly, toxicity has been proven to be the major hurdle for adjuvants designed against P. aeruginosa, leading to many being abandoned at early phases of development. Drug safety assessment is a long, expensive, but crucial process and toxicity is most likely where drug targets share structural similarity with human proteins. In this respect, bacterial signaling systems are good candidates, as prokaryotic and eukaryotic signaling systems are highly divergent, with eukaryotes lacking TCSS or phosphorelay systems.

55As with all newly developed drugs designed to be used as combination therapy, care must be taken in determining the correct dosing and investigating clear synergy profiles. Drug levels necessary for synergy in vitro may not be achievable in vivo. Synergy in vivo may be affected by failure to obtain desired levels of drugs in the target tissue, drug metabolism or plasma protein binding. In addition, it is of paramount importance to evaluate drugs in relevant models that reflect the environmental conditions of infection. This will increase the predictive power of preclinical testing, reducing the costly progression of unpromising agents to clinical trials. The lack of well-validated in vivo models for testing CF anti-infective therapeutics limits the speed of development of new drugs. Murine models using cystic fibrosis transmembrane conductance regulator (CFTR) knockout animals, or transgenics in which the severe gut phenotype associated with loss of CFTR has been corrected are available, and have proved useful, but do not develop the characteristic features of acute and chronic P. aeruginosa infection seen in those with CF. Progress on the use of ferret and porcine infection models with mutated CFTR has been made, although these are limited by the availability of suitable immunobiology reagents. Intranasal administration of P. aeruginosa into the healthy murine lung often leads to either rapid clearance or sepsis. To create a model of persistent infection, it is usually necessary to immobilize P. aeruginosa. This can be achieved by encapsulating the bacteria into agar or alginate beads, where the bacteria are protected from clearance by immune effectors and where their mode of living more closely mimics bacterial biofilms present in chronic infection. However, this bead model is technically demanding and requires surgical transtracheal instillation of the bead suspension, leading to additional complications and mortality not representative of bacterial infections in CF. Alternatively, long term lung infection can be achieved using P. aeruginosa isolates from CF, some of which naturally establish chronic infection in mice, without the need for implantation into beads. This model has the advantages of using a natural infection route and having no requirement for surgical intervention, and offers opportunity to study lung infection over prolonged periods. However, the density of infection achieved in the lung is low, making some analyses challenging.

56Questions remain regarding the commitment of governments and pharmaceutical manufacturers to ongoing investment in antibacterial drug development, particularly as financial and research priorities are reshuffled by the ongoing SARS-CoV2 crisis. Despite the understandable current emphasis on anti-viral agents and vaccines, it is important that we do not lose ground in the fight against AMR. Indeed, emerging evidence suggests that antibiotic use has increased dramatically in the COVID-19 era. On top of this, increased usage of sanitizers and disinfectants globally may induce the development of cross-resistance to antibiotics. The ESKAPE pathogens, for which new medicines are urgently needed, continue to cause serious community-acquired and nosocomial infections, and if investment into research and drug development for these bacterial pathogens is diminished, it will exacerbate the global health and economic costs associated with the ongoing pandemic.

57FL wrote the manuscript with supervision and input from all others. All authors contributed to the article and approved the submitted version.

58FL is supported by a PhD studentship from the Rosetrees Trust (M750). DN is supported by a Sir Henry Dale Fellowship funded by the Wellcome Trust and the Royal Society (Grant number 204457/Z/16/Z).

59The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.