Keywords

The outer membrane of the Gram-negative cell acts as a permeability barrier that protects the cell from various noxious substances, including numerous antimicrobials [1]. Polymyxins exert their antimicrobial action via direct interaction with the lipid A component of the lipopolysaccharide (LPS) which leads to a disruption of this critical barrier function. Accordingly, understanding the mechanism of polymyxin antibacterial activity requires a brief review of the architecture of LPS and the outer membrane. The complex asymmetrical structure of the outer membrane comprises an inner phospholipid leaflet, as well as an outer leaflet that predominantly contains LPS, proteins and phospholipids [1]. LPS is composed of three domains, a conserved inner core 2-keto-3-deoxyoctonoic acid (Kdo) bound to lipid A and a variable O-antigen composed of repeating units of various polysaccharides [1,2,3,4]. The structure of lipid A consists of aβ-10-6-linked D-glucosamine (GlcN) disaccharide that is phosphorylated at the 1- and 4′-positions and decorated by a variable number of saturated hydrocarbon chains, generally C10-C14 in length [2, 4]. Lipid A is intercalated within the outer leaflet, where in the saturated lipid A hydrocarbon chains are tightly packed together within the membrane through van der Waals forces, thereby acting as an anchor for the entire LPS structure [1, 5, 6]. Divalent cations (Mg2+ and Ca2+) associate with the lipid A phosphoresters and function to bridge adjacent LPS molecules [1, 5, 6]. The barrier function of the outer membrane is further accentuated by a highly repulsive anionic charge conveyed by lipid A phosphorester moieties, as well as phosphate and carboxylate functionalities decorating the core and O-antigen sugars [1, 5, 6].

Amphipathicity of the polymyxins is critical for their outer membrane permeabilizing action. The conserved elements in the chemical structure of polymyxins that contribute to this amphipathicity includes the two hydrophobic domains (the N-terminal fatty acyl chain and the hydrophobic position 6–7 segment) separated by segments of polar (Thr) and cationic Dab side chains. The elucidation of the three-dimensional NMR solution state structure of polymyxin B in complex with LPS revealed the polymyxin B molecule is folded such that the polar and hydrophobic domains form two distinct faces, thereby conferring structural amphipathicity (Fig. 4.1) [7, 8].

Fig. 4.1
figure 1

Left panel. NMR-based model of polymyxin B bound to the lipid A component of E. coli LPS [7, 8]. Right panel. Chemical structures of polymyxin B, colistin and nonapeptide. Polymyxin residues: Thr: threonine; Leu: leucine; Phe: phenylalanine; Dab: α,γ-diaminobutyric acid

Full size image

4.1 Primary Mode of Action of Polymyxins

Polymyxins are believed to exert their primary antimicrobial mode of action by permeabilizing the outer membrane via a direct interaction with LPS . Polymyxins zone-into their primary cellular target, LPS through the initial electrostatic interaction of the cationic L-α,γ-diaminobutyric acid (Dab) side-chains with the phosphate groups of the lipid A component of LPS, displacing divalent cations (Ca2+ and Mg2+) that bridge adjacent LPS molecules (Fig. 4.2, Stage 1) [1, 5, 6, 9,10,11]. This initial electrostatic interaction allows the N-terminal fatty acyl chain and hydrophobic position 6–7 motif (Polymyxin B: D-Phe6-L-Leu7 and colistin: D-Leu6-L-Leu7) of the polymyxin molecule to insert into the fatty acyl chain layer of the lipid A molecules. The insertion of the hydrophobic domains of polymyxins possibly act to weaken the packing of adjacent lipid A fatty acyl chains causing expansion of the outer membrane monolayer. The fact that polymyxin B nonapeptide (derived by proteolytic cleavage of the fatty acyl-Dab1 from the N-terminus of the polymyxin) is devoid of antibacterial activity highlights the importance of the hydrophobic interactions for the mechanism of polymyxin action [12]. Subsequently, the polymyxin molecule inserts and disrupts the physical integrity of the phospholipid bilayer of the inner membrane leaflet via membrane thinning by straddling the interface of the hydrophilic head groups and fatty acyl chains or transient poration [6, 7, 10, 13, 14]. This ‘self-promoted’ uptake mechanism is believed to produce disruption of the outer membrane structures, which leads to bacterial cell death (Fig. 4.2, Stage 2) [5, 6, 9,10,11]. It has also been proposed that the MOA of polymyxins involves producing contacts between the periplasmic leaflets of the inner and outer membranes that promotes phospholipid exchange between the inner and outer membrane leaflets. This in turn would result in the loss of phospholipid compositional specificity, potentially to an osmotic imbalance that contributes to lytic cell death [15, 16]. This postulate is based on evidence that polymyxin B when bound to anionic phospholipid vesicles is capable of forming vesicle-to-vesicle contacts [13, 15,16,17]. The ability of polymyxins to disrupt the inner membrane structure is coincident with their inhibition of the inner membrane respiratory enzyme the alternative type 2 nicotinamide adenine dinucleoti dedehydrogenase (NDH-2) in Mycobacterium smegmatis and in a number of pathogenic Gram-negative bacteria [18, 19], which intuitively leads us to the next section that covers secondary modes of action of polymyxin lipopeptides.

Fig. 4.2
figure 2

Schematic diagram depicting the putative mode of action of polymyxins. Stage 1. Electrostatic attraction between the positively charges polymyxin molecule and the negatively charges bacterial outer membrane surface Fig. 4.2 (continued) leads to the displacement of divalent cations that help stabilize the outer membrane structure by bridging adjacent LPS molecules. Stage 2. The positively charged polymyxins displace divalent cations that bridge adjacent LPS molecules. The hydrophobic insertion destabilizes the outer membrane. Stage 3. The polymyxin molecule penetrates into the inner membrane and inhibits the respiratory enzyme NDH-2

Full size image

4.2 Secondary Mode of Action of Polymyxins

Although cationic peptides such as the polymyxins are traditionally thought of as outer membrane-active agents [20], the bacterial outer membrane is not necessarily the sole target for their mode of action [21,22,23]. Secondary targets involved in the bactericidal activity of polymyxins remain poorly characterized. Based on available evidence, one possible secondary mode of action of polymyxin B and colistin in Gram-negative bacteria involves the inhibition of bacterial respiration [24, 25].

Instead of the multi-subunit complex I found in mammalian cells, protozoa, bacteria and plants possess a single sub-unit non-proton pumping, rotenone insensitive alternative type II NADH-menaquinone oxidoreductase (NDH-2) [26,27,28,29]. The NDH-2 enzyme contains a single non-covalently bound flavin adenine dinucleotide (FAD) cofactor and catalyzes the oxidation of NADH with menaquinone [28,29,30,31]. In general, the bacterial respiratory chain consists of three complexes with quinones and reduced nicotinamide adenine dinucleotide (NADH) acting as the carriers that shuttle electrons and protons between large protein complexes [32,33,34,35,36]. The exact organization of enzymes varies among different bacteria [32,33,34]. In Complex 1, three inner membrane respiratory enzymes of the NADH oxidase family have been identified: proton-translocating NADH-quinone (Q) oxidoreductase (NDH-1), NADH-Q oxidoreductase which lacks an energy-coupling site (NDH-2) and the sodium-translocating NADH-Q oxidoreductase [32,33,34, 36].

We have shown that polymyxin B, B1, B2 and colistin can inhibit NDH-2 activity in the inner membranes of three different Gram-negative bacterial species (Escherichia coli, Klebsiella pneumoniae, Acinetobacter baumannii) in a concentration-dependent manner (Fig. 4.2, Stage 3). The mechanism of NDH-2 inhibition by polymyxin B was investigated in detail with E. coli inner membrane preparations and conformed to a mixed inhibition model with respect to ubiquinone-1 and a non-competitive inhibition model with respect to NADH. The structure of the polymyxins (cyclic peptides) being distinct from those of the NDH-2 substrates NADH and Q1 is supportive of the inhibition kinetic data, in that they are unlikely to compete for the same sites on the enzyme. Our kinetic data are in line with the reported data for Gluconobacter oxydans which showed that the inhibition by gramicidin S and scopafungin was non-competitive with respect to NADH [37]. Scopafungin, which like polymyxin B and colistin possesses a cyclic ring and a long acyl chain in its structure, displayed a mixed inhibition mode with respect to ubiquinone, whereas gramicidin S was a competitive inhibitor [37].

The IC50 values for the inhibition by polymyxin B and colistin of NDH-2 activity in inner membrane of the three different Gram-negative bacterial species were in most part comparable, indicating that inter-species differences in NDH-2 do not impact the inhibitory activity of the polymyxins. Polymyxin B was a better inhibitor compared to colistin, which is in line with reported results with the Gram-positive M. smegmatis NDH-2 [37]. Although polymyxin B and colistin display high IC50 values (polymyxin B IC50 = 50 μM; colistin IC50 = 251 μM) for NDH-2 inhibition, under in vivo conditions there remains the possibility that very high local concentrations of the antibiotic can accumulate at the site of infection that fall within these IC50 value ranges. Coincidently, we have garnered in vitro evidence that suggests that polymyxins can accumulate in the inner membrane of Gram-negative bacteria [38]. Therefore, the high IC50 values do not dismiss the possibility that NDH-2 represents one of the secondary pathways that is targeted once the polymyxin penetrates the outer membrane. Notably also colistin inhibited NADH-quinone oxidoreductase activity in the polymyxin-susceptible strain of K. pneumoniae with a comparable IC50 to that of the polymyxin-resistant strain, suggesting polymyxin resistance in these strains is not at the level of the inner membrane respiratory enzymes. Our previous study had indicated that the resistant derivative of K. pneumoniae exhibited less negative charge than the wild type that lead to failure of polymyxin interaction at the outer membrane [39]. The NDH-2 activity was not inhibited by CMS, polymyxin B nonapeptide and colistin nonapeptide. The loss of inhibitory activity seen with the polymyxin nonapeptide and CMS suggests that the N-terminal fatty acyl chain and the positive charges of the polymyxin molecule are critical for NDH-2 inhibitory activity [40].

The fact that NDH-2 enzymes are not found in mammalian mitochondria and are mainly expressed by protozoa, bacteria and plants makes them very attractive drug targets [41]. Our data suggest that one of the secondary target sites of polymyxins is the type II NADH-quinone oxidoreductase respiratory enzyme that forms an integral part of the bacterial electron transport pathway. In view of the dry antibiotic pipe-line, together with the increasing incidence of multidrug resistant in Gram-negative bacteria , NDH-2 represents an important target that can be exploited for the development of new antibiotics against these problematic pathogens. Notably, energy metabolism and NDH-2 in particular, is emerging as an important drug target in Mycobacterium tuberculosis and Plasmodium falciparum [28, 29, 31, 42,43,44,45].

A recent preliminary biochemical study reported that rapid killing of A. baumannii by polymyxins is mediated by a hydroxyl radical death pathway, which although under explored, potentially represents another secondary mode of action whereby polymyxin kill Gram-negative bacterial cells [46]. Coincidently, it has been proposed that most antibiotics cause bacterial cell death via a common mechanism whereby they disrupt bacterial metabolism leading to the generation of reactive oxygen species (ROS) that eventually kills the bacterial cell [47].

4.3 Stress Responses to Polymyxin Treatment in Gram-negative Bacteria

The antibacterial activity of bactericidal antibiotic is not solely governed by its mode of action and its ability to interact with targets [47,48,49,50,51]. There are a number of bacterial response factors associated with exposure to sub-lethal concentrations of polymyxins which will be reviewed in this section. These factors include activation of adaptive resistance mechanisms, the stimulation of protective changes to cell physiology, and even induction of resistance mutations.

The protective stress response, also named adaptive resistance, is an auto-regulated antibiotic induced phenomenon and reversal to the sensitive phenotype in the absence of inducer [52, 53]. Extensive studies in recent years have provided significant insight into the outer membrane remodelling mechanisms responsible for adaptive resistance to polymyxins, perhaps best studied in Salmonella [54,55,56,57].

In response to polymyxin exposure, the outer membrane undergoes extensive remodelling of structural alterations contributing to adaptive resistance to polymyxins which is triggered by two-component systems (Fig. 4.3) [52, 54, 58]. In Salmonella, the response to polymyxins is mediated by the PhoPQ two-component system, where polymyxins interact with and activate the sensor PhoQ by displacing divalent cations from their metal binding sites in the sensor domain [59] and then activates the PhoP response regulator to up-regulate a variety of target genes and ultimately promote adaptation to the stress. Specific changes in OM regulated by activated PhoP include: increasing O-antigen chain length, acylating, inhibiting deacylating, hydroxylating lipid A , derivitizing lipid A and LPS core phosphates with cationic groups, palmitoylating OM phosphatidylglycerols, and increasing the level of OM cardiolipins [54, 56]. Therefore, upon PhoPQ activation an extensive alteration of LPS, GPLs, and proteins elaborates the OM barrier more impermeable to polymyxins , thereby promoting bacterial survival (Fig. 4.3) [54, 56].

Fig. 4.3
figure 3

(a) Glycerophospholipid and lipid A modifications that ensue with Gram-negative membrane remodelling associated with the development of polymyxin resistance. (b) Polymyxin-induced outer membrane remodelling, intracellular biochemical perturbations and resistance pathways. The initial outer membrane disorganisation caused by polymyxin exposure is followed by intracellular redox perturbations of NDH-2 activity. These events are accompanied by activation of repair pathways and outer membrane remodelling. The PhoP/PhoQ two component regulatory system activates the lipid A deacylase PagL and PmrD which in turn activates PmrA. PmrA activates the expression of arnBCADTEF which are a collective of enzymes modifying lipid A with cationic groups (e.g. 4-amino-4-deoxy-L-arabinose or phosphoethanolamine) to repel the polymyxin. PmrA also activates PmrR responsible for the repression of LpxT (phosphorylation of lipid A) and LpxR (deacylation of lipid A) genes

Full size image

4.4 Regulation of LPS Remodelling

Upon sensing polymyxins , PhoPQ increases transcription of pmrD and the pmrCAB operon to activate the response regulator PmrA (Fig. 4.3) [60,61,62]. The activated PmrA induces expression of a short membrane peptide, PmrR, which binds to and inhibits the lipid A phosphorylase LpxT, an enzyme responsible for increasing the negative charge of the outer membrane leaflet [56, 63]. The activated PmrA induces transcription of genes encoding enzymes that covalently modify lipid A and core sugar phosphates with positively charged aminoarabinose (L-Ara4N) and phosphoethanolamine (pEtN) [54]. The initial step for L-Ara4N moiety modification begins with the oxidation of UDP-glucose in the cytosol by the PhoP and PmrA-regulated UDP-glucose dehydrogenase (PagA, Ugd, or pmrE) [54, 64]. The remaining steps responsible for L-Ara4N moiety modification are encoded within an operon pmrHFIJKLM (also known as arnBCADTEF), which is activated by PhoPQ through PmrA [54, 64]. When the biosynthesis proceeds to complete UDP formylated-L-Ara4N synthesis, the formylated-L-Ara4N residue is transfer from UDP to undecaprenyl phosphate carrier lipid on the inner leaflet of the inner membrane and then deformylated to form undecaprenyl phosphate-L-Ara4N [54, 64]. Next, the undecaprenyl phosphate-L-Ara4N is flipped into the outer leaflet of the IM where the membrane protein ArnT (also known as PmrK) transfers the L-Ara4N moiety to nascent lipid A phosphates [54, 64]. Finally, O-antigen is loaded onto the core structure and then the assembled LPS molecules are moved from the inner membrane through the periplasm to the OM of cell surface by the lipopolysaccharide transport (Lpt) proteins complex acquired driving powers from cytoplasmic ATP hydrolysis [54, 65]. The PmrAB-controlled pEtN transferases encoded by eptA (also known as lptA or pmrC) and cptA (or eptB) also contribute to polymyxins resistance via their modification of lipid A and LPS core respectively, with positively charged pEtN [54, 56]. Decreased negative charge conferred by cationic groups L-Ara4N and pEtN on lipid A molecules diminishing binding sites plays a significant role in polymyxins resistance while modification of cationic groups on LPS core plays modest effect on resistance [54, 56, 66]. Though varies in details across Gram-negative species, the positively charged modification of lipid A mediated by two-component systems is the most commonly seen mechanism of polymyxins resistance [52, 56, 67, 68].

An abundance of uniquely hydroxylated myristoyl groups has been detected in lipid A structures under PhoPQ activation in S. typhimurium [69]. The modification of hydroxylation of myristoyl groups is catalyzed by the inner membrane dioxygenase, LpxO, which act as part of a coordinated stress response [54, 56, 70].

PhoPQ activation regulates the composition proportions of O-antigen repeats through the action of the PmrA-regulated inner membrane protein complex Wzzst and WzzfepE [71,72,73]. Wzzst selectively determines the formation of ‘long’ (L) O-chain lengths around 16–35 repeat units, while WzzfepE is responsible for the ‘very long’ (VL) lengths with over than one hundred subunits. Activation of pmrA increased the fraction of such L-type and VL-type O-antigen in LPS molecules [71,72,73]. Raising the proportions of L-type and VL-type O-antigen leads to heightened resistance to serum [71]; moreover, VL-type enhanced two-fold increase in polymyxin B resistance [73]. Thus, it is plausible that such types of O-antigen somehow promote barrier function and contribute to polymyxins resistance [54].

Three outer membrane enzymes PagP (or CrcA), PagL and LpxR with active-site exposed to the outer leaflet of OM are related to acylation or deacylation in barrier remodelling [54, 56]. In response to PhoPQ activation, the transcription of pagP is stimulated and thus upregulated the encoded proteins that modify lipid A with palmitate [74]. Palmitoyl group transferring from phospholipid (GPL) donors to lipid A occurs on the extracellular active-site of PagP and results in hepta-acylated lipid A species [54, 56, 64, 74]. The other two outer membrane enzymes, PagL and LpxR, acting as deacylation of lipid A are subjected to post-translational inhibition by L-Ara4N modified lipid A as their active sites are found on the extracellular surface and in close proximity to lipid A [56, 64]. Increased palmitoylation and inhibited deacylation enhance the hydrophobicity of the OM and prevent penetration of the amphipathic polymyxin molecules [54, 56].

4.5 Regulation of Glycerophospholipids (GPL) Remodelling

Recent research indicates that glycerophospholipids (GPL) are also regulated components of the OM barrier in Gram-negative bacteria [55, 75]. Penetration of polymyxin molecules can cause LPS layer to become displaced and shed from the outer leaflet and activated the PhoPQ system [59, 68, 75]. To maintain bilayer barrier integrity, GPL from the inner leaflet migrate into the outer leaflet of the OM to replace the breached areas of LPS with GPL as a consequence of locally weakened barrier which are only detectable in stressed cells [54, 55, 75, 76]. The outer membrane protein PagP with dual substrate specificity activated by the PhoPQ system can function as a membrane-intrinsic probe to restoration of the permeability barrier [55, 64]. In addition to lipid A , PagP transfers palmitoyl groups from GPL to the polar head group of phosphatidylglycerol (PG) that have flipped onto the surface of OM forming palmitoyl-PG [54, 55]. Therefore, once LPS layer disrupted and PhoPQ system activated, GPL may be increasingly translocated to the outer leaflet and further acylated by pagP to enhance barrier hydrophobicity [54, 55, 64]. Also, the research work detected modest yet significantly increases in cardiolipin (CL) amount of the OM on activation of PhoPQ, which was speculated to form functional micro-domains that promote OM lipid re-modelling [54, 55]. The phospholipid (PL) bilayer structure of OM is more permeable than the asymmetrical LPS-PL bilayer barrier, so the OM phospholipase A (OMPLA) and inter-membrane transport system Mla pathway are functioned to prevent surface exposure of PLs and maintain lipid asymmetry of the OM if necessary [75, 77]. Thus, coordinate regulation of LPS and GPL forms a remodelled OM barrier critical for bacterial protective responses and survival to polymyxins.

4.6 Other Responses to Polymyxin Treatment in Gram-negative Bacteria

In Salmonella, antimicrobial peptides and cations occupy the overlapping binding site of the sensor PhoQ [59]. Divalent cations are bound to the acidic surface region of PhoQ sensor under normal conditions, while the displacement of these cations by antimicrobial peptides results in a conformational change that activates PhoQ and triggers the hierarchical regulation [59]. To date, various two-component systems have been reported to associate with the adaption to sub-inhibitory concentrations of antimicrobial peptides in Pseudomonas aeruginosa , including the widespread PhoPQ and PmrAB systems, and the ParRS, ColRS and CprRS systems [52, 58, 78, 79]. In contrast to Salmonella, P. aeruginosa senses divalent cations and cationic peptides via different mechanisms. Divalent cations are detected by PhoQ and PmrB but not peptides [58, 79]. Whereas ParS and CprS can detect cationic antimicrobial peptides regardless of Mg2+ concentrations which are independent two-component systems that might recognize different properties of peptides or the different effects of peptides on cell at specific concentrations. As for polymyxins, both participated at all concentrations, with a greater involvement of ParRS which is likely to be the key component [78, 79]. The occurrence of at least two direct polymyxins response systems and three associated response regulatory systems in P. aeruginosa highlights the complexity of the adaptive resistant pathways in this organism.

Several tripartite efflux systems play considerable roles in the intrinsic and acquired resistance in P. aeruginosa. Each system consists of three proteins with presumed functions: a cytoplasmic membrane component of the resistance-nodulation-division family acting as a transporter, an outer membrane component forming channels, and a membrane fusion protein linking the two membranes [80, 81]. At the gene level, a constitutively expressed operon, mexAB-oprM, coding for an efflux system (MexAB-OprM) which contributes intrinsic resistance in P. aeruginosa produced at a basal level in wild-type bacteria [80]; however, it has been observed that the MexAB-OprM efflux system is overexpressed in the metabolically active subpopulations of P. aeruginosa biofilm , conferring an unspecific adaptive resistant phenotype to polymyxins [52, 82]. Another outer membrane efflux pump system MexXY lacking OM protein in its own operon utilises OprM of the MexAB-OprM system and forms MexXY-OprM system that has been shown to provide natural to aminoglycosides and various unrelated antibiotics [80, 81]. It has been demonstrated that polymyxins can promote expression of the mexXY operons besides pmrAB and arnBCDTEF-ugd, and to coordinately downregulate the oprD gene that promoting β-Lactams resistance through the activation of the two-component systems ParRS [78]. These researches indicate that polymyxins are able to induce multidrug adaptive resistance of cells through the activation of distinct mechanisms (efflux, porin loss, and LPS modification) in P. aeruginosa [78, 82].

Microarray and high-throughput RNA-seq analysis revealed a global change pattern of gene expression leading to adaptive responses. When P. aeruginosa cells are exposed to sub-inhibitory concentrations of colistin, approximately 0.5% of 5500 genes showed significantly changed expression levels in the colistin-treated sample. Among them, 13 were upregulated and 17 were downregulated. The upregulated genes are involved in quorum sensing (QS) and biofilm formation besides well-known LPS modification, while the downregulated genes are involved in motility (swarming and swimming motility) and osmotolerance [83]. Upon exposure to a much higher concentration of polymyxin (10 × MIC) treatment, a wider profile of global changes was identified from protective responses of Yersinia pestis to survive the stressful environments. A total of 291 genes were differentially expressed and 158 of them were induced. Among the 158 upregulated genes, 22 were regulatory genes including 8 two-component systems that globally or locally governing a wide set of stress-protective functions, 19 genes were involved in remodelling of cell envelope encoding membrane components or polysaccharide surface structures, 4 operons including 9 genes were essential for dissimilation of sn-glycerol 3-phosphate which is a direct precursor for phospholipid biosynthesis [84], 10 were heat shock proteins that play important roles in preventing aggregation of proteins and repairing misfolded or damaged proteins caused by environmental stresses, 5 were related to drug resistance (3 of tellurium resistance and 2 of multidrug transport system), 11 genes organized into four operons were components of siderophore-based iron acquisition systems which known as high-pathogenicity island shared by three pathogenic yersiniae [85]. Our recent high-throughput RNA-seq study of the transcriptomic response of Acinetobacter baumannii to colistin under conditions that able to kill partial cells revealed hundreds of genes differentially expressed including those of two-component systems, glycerophospholipid metabolism, lipopolysaccharide biosynthesis, biofilm synthesis, drug resistant proteins, heat shock proteins as discovered from other Gram-negative strains (data unpublished). Moreover, genes involved in nucleotide excision repair and peroxisome were also significantly induced by killing concentration of colistin. Two genes encoding the peroxisome superoxide dismutase SOD1 and catalase KatE (belonging to the antioxidant system) were upregulated after colistin treatment.

4.7 Mutations and Death

Recently, it has been demonstrated that a number of bactericidal antibiotic classes trigger the endogenous production of lethal active forms of hydroxyl radicals in bacteria through the Fenton reaction [46, 47, 51, 86] which depends on the availability of hydrogen peroxide, an iron species and reducing equivalents [86,87,88]. Hydrogen peroxide is generated within cells as a by-product of oxidative metabolism when molecular oxygen accidentally acquires electrons from the reduced cofactors of flavoproteins [50, 88] and bactericidal antibiotics elevated the generation of deleterious reactive species [48]. Iron is necessary for bacteria to survive and acquired from environment by biosynthetic iron chelators known as siderophores [86]. Hydrogen peroxide is capable of interacting with intracellular ferrous form iron unincorporated or associated with biological molecules including iron-sulphur-dependent dehydratases and mononuclear iron proteins, oxidizing the iron and forming hydroxyl radicals in the process [50, 86]. This is a cyclical process in vivo since intracellular reductants can reduce the oxidized iron back to ferrous form [86]. The event of iron disintegration from proteins eliminates a variety of enzyme activity and the hydroxyl radical is an extremely powerful oxidant that reacts with virtually all organic molecules including giving a wide variety of DNA damage [50]. Ultimately, the oxidative damage of hydroxyl radical to DNA, lipids, and proteins eventually reaches levels that cannot be controlled and thus contribute to cell death [47, 48].

It has been demonstrated that the rapid killing of Gram-negative species A. baumannii, Escherichia coli and Francisella novicida by polymyxins is in part mediated by a hydroxyl radical death pathway [46]. In addition, this mechanism of killing occurs in polymyxin-susceptible A. baumannii isolates including multidrug-resistant clinical isolates but this response is not induced in a polymyxin-resistant isolate [46]. The mechanism by which polymyxin treatment induces the production of hydroxyl radicals in Gram-negative bacteria is not clear yet. Polymyxin-induced hydroxyl radical death does not occur in polymyxin-resistance isolates of A. baumannii but in susceptible isolates [46], and polymyxin resistance in A. baumannii is commonly due to blocking the entries through remodelling of the outer membrane [67, 68, 89]. Therefore, it seems that the event of the hydroxyl radical death pathway induced by polymyxins happened after the initial drug-target interactions on the outer membrane. The secondary MOA of polymyxins that they can inhibit NDH-2 activity in the inner membranes of detected Gram-negative species [90] may result in accumulation of nicotinamide adenine dinucleotide (NADH) which can be utilized as a source of reducing equivalents that contribute to Fenton reaction (Fig. 4.3b, bottom left part) [88]. In vitro the NADH-Fe (III)-EDTA-H2O2 system drives an ongoing Fenton reaction and DNA ensue breaks in such system while NAD+ is ineffective. Furthermore, the rate of DNA nicking corresponds to the rate of NADH oxidation [88].

Exposure of Gram-negative species to low concentrations of hydrogen peroxide resulted in DNA damage that causes mutagenesis and kills the cell [87, 88]. It has been proposed that antibiotic-induced ROS may provide a mechanism of acquiring beneficial mutations when stresses are small but induce lethality when stresses are large [48]. A recent study revealed that treatments with sub-lethal levels of bactericidal antibiotics resulting in up to 10 times increases in the mutation rate relative to an untreated control, a strong correlation between ROS (reactive oxygen species) formation and fold change in mutation rate, and heterogeneous increases in the minimum inhibitory concentration for a range of antibiotics irrespective of the drug target [49]. Colistin was bactericidal in a concentration-dependent manner [91]. Although colistin resistance is rare, colistin-resistant A. baumannii, P. aeruginosa and K. pneumoniae isolates have been reported world-wide. The emergence of colistin resistance or heteroresistance after colistin treatment can be easily selected in vitro and in vivo with mutations of key genes for protective responses such as the initial two-component regulatory systems [91,92,93]. However, a recent study of colistin mutant prevention concentrations (MPCs) for A. baumannii, P. aeruginosa and K. pneumoniae were shown to be very high (>64 μg/mL) [93], which recommended polymyxin combination therapy to prevent the emergency of resistant mutants and the risk of toxicity at high concentrations [93].

Accordingly, bacterial cells contain scavenging enzymes to prevent the accumulation of reactive oxygen species. Two well-known antioxidant system enzymes, catalase and SOD, can transform active radicals into oxygen and water [72, 73]. Our unpublished work of transcriptomic response of A. baumannii to colistin supported such protective response as 6 genes encoding peroxidases (including catalase and SOD) were significantly unregulated. Moreover, nucleotide excision repair genes for DNA damage repairing in our unpublished data, heat shock proteins [85] for repairing mis-folded or damaged proteins, iron acquisition systems [85] for recovering the function of iron disintegrated proteins were up-regulated significantly which might partially arise from the response to hydroxyl radical damage.

In summary, polymyxins are capable of inducing cell damage and death by interfering with their primary targets which trigger stress responses that induce outer membrane remodelling preventing polymyxins from entering. While downstream secondary target-polymyxin interactions might trigger redox-related physiological alterations that result in the formation of toxic reactive oxygen species as well as stimulating repairing responses which further contribute to cellular damage, mutations and death [48]. The formidable challenge of sub-lethal polymyxin treatment leading to multidrug resistance rather than completely killing the bacteria required us to expand our understanding of the polymyxin stress responses in Gram-negative bacteria on a detailed system-wide level. Such understanding helps in providing a foundation for finding key protective responses molecules. Those molecules can be utilised as therapy targets in combination with polymyxin-treatments to improve current therapeutic options and avoid resistant mutations.

4.8 Imaging Polymyxin Penetration and Localization in the Gram-negative Bacterial Outer Membrane

The unavailability of valid imaging probes with native activities is a significant barrier to examine the intra-cellular localization of polymyxins in Gram-negative bacterial cells [94]. Most reports of polymyxin probes either employed inactive nonapeptide derivatives or dansylated polymyxin B, which was derived by non-specifically reacting polymyxin B with dansyl-chloride [95,96,97,98]. We have previously highlighted the deficiencies of directly amine-coupling dansyl groups onto the Dab side chains in semi-synthetic preparations of dansyl-polymyxin B [99]. Analysis of these semi-synthetic dansyl-polymyxin B preparations revealed the existence of mono-, di-, tri-, and tetra-dansyl substituted species [99]. Furthermore, as polymyxin B is comprised of two major components (B1 and B2), the potential for either of these components to be substituted at any of the five Dab side chains with up to four dansyl molecules results in a highly variable mixture of dansylated derivatives [99]. Commercial preparations of 4-bora-3a,4a-diaza-s-indacene (BODIPY) labelled polymyxin B displayed a markedly reduced antibacterial activities compared to polymyxin B [100]. Moreover, our group previously reported a fully synthetic [dansyl-Lys]1polymyxin B3 probe that was devoid of antibacterial activity [99]. These findings are consistent with our understanding of polymyxin SAR wherein the dansyl modification of the Dab side chains inactivates antibacterial activity [99]. Clearly, there is very little value in using these semi-synthetic preparations as imaging probes since they lack native antibacterial activityand pharmacological properties of the parent compound, polymyxin B.

We recently reported the regio-selective modification of the polymyxin B core scaffold at the N-terminus with the dansyl fluorophore to generate an active probe (probe (1)) that mimics polymyxin B pharmacologically (Fig. 4.4, bottom left panel) [38]. The design and synthesis of a dansyl molecular probe through the regio-selective modification of the polymyxin B core structure was undertaken with the aforementioned lipid A interaction principles in mind; in order to mimic the polymyxin B structure as closely as possible and to maintain its native antibacterial activity. The dansyl group was utilized for the fluorescent probe as it has suitable spectral properties and its relative small size would reduce the chance of steric effects. The regio-selective incorporation of the dansyl group into the hydrophobic N-terminal centre of the polymyxin B core scaffold is prudent as it has a minimal impact on the native antibacterial activity of the polymyxin B scaffold. Therefore, the strategy we employed was to replace the N-terminal fatty acyl group of polymyxin B with the amino acid L-octylglycine, where the eight-carbon fatty acyl chain emulated the N-terminal fatty acyl chain of polymyxin B, whilst the N α-amino group would provide a convenient point of attachment for the dansyl group (eliminating the need for additional orthogonal protection during the synthesis).

Fig. 4.4
figure 4

Bottom left panel. Chemical structure of the probe (1). Top left panel. NMR-based model of the probe (1)-Kdo2 lipid A complex. Right panel. Laser scanning confocal microscopy image of K. pneumoniae ATCC 13883 cells treated with probe (1)

Full size image

The antimicrobial activity of probe (1) was screened against a panel of ATCC and recent clinical isolates of polymyxin-susceptible and -resistant strains of P. aeruginosa, A. baumannii and K. pneumoniae. The probe showed antibacterial activity against polymyxin-susceptible strains (MIC 4–16 mg/L), compared to colistin (MIC 0.125–2 mg/L) and polymyxin B (MIC <0.125–2 mg/L). Probe (1) also displayed activity against polymyxin-resistant P. aeruginosa and A. baumannii strains (MIC 4–8 mg/L); colistin and polymyxin B (MIC >32 mg/L). It is understood that Gram-negative pathogens resist the action of polymyxins by introducing cationic modifications onto the phosphate groups on the lipid A component of LPS [101,102,103,104,105,106]. The most common mechanism involves esterification of lipid A phosphates with aminoarabinose, or ethanolamine [101,102,103,104,105,106]. The molecular tailoring serves to reduce the net negative charge of the outer membrane surface, thereby repelling the electrostatic attraction with positively charged polymyxin molecules [107]. The NMR-based molecular model of the probe (1)-Kdo2 Lipid A complex implies that the combination of the L-octylglycine and the dansyl substituent at the N-terminus provides additional hydrophobic interactive forces that compensate for the electrostatic repulsion of the aminoarabinose phosphate modifications (Fig. 4.4, top left panel). The molecular model indicates that electrostatic interactions with the 1-phosphorester group on lipid A are not hampered by the dansyl group. The model further suggests that the hydrophobic dansyl group interacts with the apolar environment formed by the fatty acyl chains of lipid A. TEM imaging of K. pneumoniae ATCC 13883 cells treated with probe (1) at 0.5 × MIC revealed the formation of numerous protrusions or blebs extending from the outer membrane of the cells that possibly represent outer membrane fragments. A similar blebbing effect was observed with Gram-negative bacterial cells treated with polymyxin B and colistin [108].

Time-lapse laser scanning confocal microscopy imaging of the penetration of probe (1) into K. pneumoniae ATCC 13883 cells revealed that the probe initially accumulates in the outer membrane and subsequently penetrates into the inner membrane and finally becomes homogenously distributed into the cytoplasm (Fig. 4.4, right panel). Intriguingly, confocal imaging and spectrophotometric lysis assay experiments with spheroplasts isolated from K. pneumoniae ATCC 13883 revealed that probe (1) also accumulated within and disrupted the inner membrane structure. Coincidently, our group has recently show that polymyxin B and colistin inhibit the NDH-2 oxidoreductase inner membrane respiratory enzyme, which also may contribute towards their bactericidal effect (cf. preceding discussions; Fig. 4.1) [19]. Furthermore, the imaging experiments revealed that at sub-MIC concentrations, probe (1) tends to accumulate on the surface of the bacterial cell and partly penetrates into the outer membrane. Whereas at <MIC concentrations, probe (1) accumulated on the surface of the bacterial cell and entered into the cytoplasm. Fundamentally, the localization studies from the time-lapse laser scanning confocal microscopy imaging results helps validate the mechanistic model of polymyxin action (Fig. 4.1). Based on the imaging data, it is evident that polymyxins initially accumulate in the outer membrane, followed by a gradual penetration into the inner membrane and finally enter the cytoplasm in K. pneumoniae. These findings are consistent with the secondary mode of action of polymyxin B which involves an inhibitory activity against the inner membrane NDH-2 enzyme [19]. It will be most interesting to elucidate the intracellular bacterial cell target(s) for the polymyxins, which to date remains uncharacterized.

The commercial availability of this probe would greatly facilitate molecular imaging studies on both the mode of action and pharmacokinetics ; and contribute towards the development of a new generation of polymyxin lipopeptides with superior activity against polymyxin-resistant Gram-negative ‘superbugs’.

In summary, polymyxins are bioactive natural products with beneficial pharmacological activities and their antimicrobial properties have been investigated for decades. Albeit, their precise mode of action remains unknown and much progress has been made towards understanding their bacterial killing effect. The compendium of data has highlighted that their bacterial killing mechanism is more complex than simply the long-standing notion that they impart a membrane disorganising effect. Secondary, pathways such as the bacterial redox chain have been implicated in their killing effect. Clearly, much work needs to be done to comprehensively elucidate the precise mode-of-action of these valuable lipopeptide antibiotics.