Correspondence: Hervé Le Moual, Department of Microbiology and Immunology, McGill University, 3775 University Street, Montreal, QC H3A 2B4, Canada. Tel.: +1 514 398 6235; fax. +1 514 398 7052; e-mail: email@example.com
Antimicrobial peptides (AMPs) are present in virtually all organisms and are an ancient and critical component of innate immunity. In mammals, AMPs are present in phagocytic cells, on body surfaces such as skin and mucosa, and in secretions and body fluids such as sweat, saliva, urine, and breast milk, consistent with their role as part of the first line of defense against a wide range of pathogenic microorganisms including bacteria, viruses, and fungi. AMPs are microbicidal and have also been shown to act as immunomodulators with chemoattractant and signaling activities. During the co-evolution of hosts and bacterial pathogens, bacteria have developed the ability to sense and initiate an adaptive response to AMPs to resist their bactericidal activity. Here, we review the various mechanisms used by Gram-negative bacteria to sense and resist AMP-mediated killing. These mechanisms play an important role in bacterial resistance to host-derived AMPs that are encountered during the course of infection. Bacterial resistance to AMPs should also be taken into consideration in the development and use of AMPs as anti-infective agents, for which there is currently a great deal of academic and commercial interest.
Mammalian antimicrobial peptides (AMPs) are diverse in sequence and are classified into families on the basis of their structures and functions (Hancock & Sahl, 2006). Two major families of AMPs in mammals are the defensins and the cathelicidins (Table 1). Defensins are cysteine-rich cationic peptides that form β-sheet structures and contain disulfide bonds. The position of the disulfide bonds is used to further classify defensins into subfamilies (α- and β-defensins in mice and humans). Of note, murine α-defensins are often designated as cryptdins (Eisenhauer et al., 1992). Cathelicidins are also positively charged, but do not have disulfide bonds. Rather, they form amphipathic α-helices with a positively charged face. There is only one cathelicidin member present in humans and mice, named LL-37 and murine cathelicidin-related antimicrobial peptide (mCRAMP), respectively. The positive charge of defensins and cathelicidins (and many other AMPs) promotes their preferential association with negatively charged bacterial membranes. Cationic AMPs interact with Gram-negative bacteria in a multistep process, first interacting with the lipopolysaccharide and then disrupting the outer membrane (OM) to gain access to the periplasmic space. Most AMPs appear to exert their bactericidal function by then disrupting the cytoplasmic membrane, although several recent studies suggest alternative targets such as lipid II and peptidoglycan synthesis (Brogden, 2005).
Also relevant to human health are bacterially derived AMPs such as polymyxin B, polymyxin E (also known as colistin), and bacitracin, which are used to treat Gram-negative infections, and nisin, which is used as a food preservative. Polymyxin E is used clinically to treat bacterial infections in cystic fibrosis patients and in multidrug-resistant infections. For example, most Escherichia coli and Klebsiella pneumoniae isolates containing the New Delhi metallo-β-lactamase 1 (NDM-1) were shown to be susceptible to polymyxin E (Kumarasamy et al., 2010). Finally, because of the problem of widespread emergence of drug-resistant bacteria and the dearth of new antibiotics in the drug-discovery pipeline, there is renewed interest in developing novel synthetic AMPs for use as anti-infective agents (Yeung et al., 2011). The present review focuses on the strategies developed by Gram-negative bacteria to sense AMPs and resist AMP-mediated killing. Resistance of Gram-positive bacteria to AMPs is as important but was reviewed elsewhere (Nizet, 2006; Koprivnjak & Peschel, 2011).
The importance of AMPs in host defense to Gram-negative infections
The importance of AMPs and bacterial resistance against AMPs in the outcome of Gram-negative bacterial infections in vivo is supported by both human and animal studies. A study of uropathogenic E. coli (UPEC) strains isolated from patients with pyelonephritis (severe ascending urinary tract infection) and children with uncomplicated lower urinary tract infections found that pyelonephritis-associated strains were more frequently resistant to LL-37 than strains isolated from children with uncomplicated infections (Chromek et al., 2006). Humans with genetic disorders leading to a lack of certain AMPs (e.g. specific granule deficiency and morbus Kostmann syndrome) suffer frequent and severe bacterial infections (Ganz et al., 1988; Putsep et al., 2002). However, these patients suffer from complex diseases with pleiotropic effects, thus making conclusions about causality difficult. Studies in genetically modified mice provide more direct evidence for the role of AMPs in Gram-negative bacterial infections, particularly in the case of Salmonella enterica serovar Typhimurium (S. Typhimurium). Transgenic mice expressing 8–10 copies of human defensin 5 (an α-defensin produced by Paneth cells) are protected from oral S. Typhimurium infection (Salzman et al., 2003a), whereas mice lacking MMP-7, a protease required for processing cryptdins, are susceptible to oral infection with S. Typhimurium (Wilson et al., 1999). Macrophages from wild-type mice are more effective at inhibiting S. Typhimurium replication than mCRAMP−/− macrophages (Rosenberger et al., 2004). Together, these experiments indicate that defensins and cathelicidins are important in the host defense against S. Typhimurium infection. Conversely, in a study of S. Typhimurium mutants selected for sensitivity to AMP-mediated killing, eleven out of twelve AMP-sensitive bacterial strains displayed decreased virulence in a mouse infection model, indicating that AMP resistance may be a critical co-requisite for bacterial virulence (Groisman et al., 1992).
Animal models have provided evidence for the role of AMPs in other Gram-negative bacterial infections as well. mCRAMP−/− mice are more susceptible to intestinal infection with Citrobacter rodentium (Iimura et al., 2005) and urinary tract infection with UPEC (Chromek et al., 2006). Newborn rats treated with a chemical that damages AMP-producing Paneth cells become more susceptible to infection with enteroinvasive E. coli (EIEC) (Sherman et al., 2005). Conversely, treatment of Shigella-infected rabbits with butyrate led to upregulation of cathelicidin and marked clinical improvement and survival rates (Raqib et al., 2006), and in a human xenograft model, LL-37 overexpression increased killing of Pseudomonas aeruginosa (Bals et al., 1999).
AMPs are important to control colonization by not only bacterial pathogens but also commensal bacteria. A recent study revealed that aberrant expression of Paneth cells α-defensins alters the composition of the intestinal microbiota without changing the total bacterial numbers (Salzman et al., 2010). This finding raises the possibility that differences in pathogen susceptibility described for animals with aberrant AMP expression or activity may, in part, be mediated indirectly by changes in the microbiota.
Bacterial adaptation to AMPs
To survive the bactericidal action of AMPs, bacteria must sense the presence of AMPs and adapt accordingly by precisely controlling the expression of genes involved in AMP resistance. In Enterobacteriaceae, genes controlling AMP resistance are usually under the control of the two-component signaling pathways PhoPQ and PmrAB and the RcsBCD phosphorelay system. In S. Typhimurium, PhoPQ controls PmrAB signaling by promoting the expression of the PmrD protein that binds to phosphorylated PmrA and prevents dephosphorylation, resulting in sustained activation of PmrA-regulated genes (Bijlsma & Groisman, 2003). There is compelling evidence that AMPs are sensed directly by the PhoQ sensor kinase. Following self-promoted uptake through the OM, α-helical AMPs such as LL-37 and C18G bind directly to an anionic region of the PhoQ periplasmic domain and activate the PhoPQ system, leading to expression of PhoP-activated genes (Bader et al., 2005). Structural data have shown that the anionic AMP-binding site of PhoQ faces the inner membrane, in a position favoring the interaction with the cationic face of AMPs lying on the membrane surface (Bader et al., 2005). In contrast to the PhoQ sensors from Enterobacteriaceae, the P. aeruginosa PhoQ protein lacks the AMP-binding domain and only responds to limiting concentrations of divalent cations (Prost et al., 2008). In agreement, a recent study suggested that ParS, which is part of the ParRS two-component system, might be the P. aeruginosa AMP sensor (Fernandez et al., 2010). Recently, various AMPs, including polymyxin B, were shown to activate the S. Typhimurium RcsBCD phosphorelay system through the OM lipoprotein RcsF (Farris et al., 2010). AMP-mediated disruption of OM integrity is likely sensed by the lipoprotein RcsF located in the inner leaflet of the OM leading to RcsBCD activation through a mechanism that remains unclear. The Rcs phosphorelay contributes to AMP resistance by promoting the expression of capsule genes and production of colanic acid, which is a precursor of 4-amino-4-deoxy-l-Arabinose (l-Ara4N), the sugar responsible for polymyxin B resistance upon addition to the 4′ phosphate of lipid A.
Proteolytic degradation of AMPs
Inactive AMP precursors are processed into active AMPs by host proteases. Active AMPs can be degraded into inactive fragments by bacterial proteases that are either secreted or localized at the OM. In a pioneer study, Schmidtchen et al. (2002) reported the P. aeruginosa elastase and a protease from Proteus mirabilis, both isolated from culture supernatants, inactivated LL-37. The P. mirabilis protease was later identified as the ZapA zinc-metalloprotease and confirmed to cleave human LL-37 and β-defensin 1, but not β-defensin 2 (Belas et al., 2004). Although these proteases usually have broad-spectrum activity against various proteins or peptides, strict substrate specificity can be observed. For example, the ZmpA and ZmpB zinc-metalloproteases from Burkholderia cenocepacia cleaved LL-37 and β-defensin 1, respectively (Kooi & Sokol, 2009). A number of proteases secreted by bacteria in the oral cavity have also been implicated in AMP resistance. For example, Porphyromonas gingivalis, which is the pathogen most associated with chronic periodontal disease, is highly proteolytic and secretes three proteases known as gingipains that belong to the cysteine family of proteases and cleave substrates after arginine and lysine residues. Degradation and inactivation of LL-37 and β-defensin 3 by gingipains was reported (Gutner et al., 2009; Maisetta et al., 2011).
Many Gram-negative pathogens, mainly of the Enterobacteriaceae family, rely on proteases found at the OM to inactivate AMPs. These proteases, exemplified by E. coli OmpT, belong to the omptin family (Hritonenko & Stathopoulos, 2007). Omptins share high amino acid sequence identity (45–80%) and adopt a conserved β-barrel fold with the active site facing the extracellular environment. Omptins possess a unique active site that combines elements of both serine and aspartate proteases, and interaction with lipopolysaccharide is critical for activity. Omptins impact bacterial virulence by degrading or processing a number of host proteins or peptides (Haiko et al., 2009). Escherichia coli K12 OmpT was reported to efficiently degrade the AMP protamine (Stumpe et al., 1998). Other studies have shown that S. Typhimurium PgtE and Yersinia pestis Pla cleave α-helical AMPs such as C18G and human LL-37 (Guina et al., 2000; Galvan et al., 2008). CroP, the omptin of the murine enteric pathogen C. rodentium, was shown to degrade α-helical AMPs, including mCRAMP (Le Sage et al., 2009) (Fig. 1a). CroP-mediated degradation of AMPs occurred before they reached the periplasmic space and triggered a PhoPQ-mediated adaptive response. OmpT of enterohemorrhagic E. coli (EHEC) was shown to inactivate human LL-37 by cleaving it twice at dibasic sites (Thomassin et al., 2012).
Shielding of the bacterial cell surface
Structures external to the bacterial cell envelope such as capsule polysaccharides (CPS), curli fimbriae, exopolysaccharides involved in biofilm formation, and the O-polysaccharide of lipopolysaccharide play a role in AMP resistance. They are proposed to act as a decoy by binding AMPs and reducing the amount of AMPs reaching the bacterial membrane (Fig. 1b). Campos et al. (2004) reported that a K. pneumoniae CPS mutant is more sensitive to AMPs than the wild-type strain with a concomitant increase in AMP-mediated OM disruption, indicating that CPS acts as a shield against AMPs. Consistent with the cationic nature of AMPs, another study reported that only anionic CPSs decreased the bactericidal activity of AMPs (Llobet et al., 2008). A similar protective role for CPS was observed in Neisseria meningitidis. An unencapsulated serogroup B strain of N. meningitidis was more susceptible to the bacterially derived AMP polymyxin B, α- and β-defensins as well as the cathelicidins LL-37 and mCRAMP (Spinosa et al., 2007). Interestingly, sublethal concentrations of AMPs upregulated the transcription of the capsule genes in N. meningitidis, suggesting that increased capsule synthesis is a bacterial adaptation downstream of AMP sensing (Spinosa et al., 2007; Jones et al., 2009).
Bacterial exopolysaccharides are the major constituent of the extracellular biofilm matrix (Sutherland, 2001). Exopolysaccharides are most often anionic polymers that are proposed to play a role in the resistance of bacterial biofilms to innate host defenses. For example, the β-d-manuronate and α-l-guluronate polymer alginate produced by P. aeruginosa was shown to promote the formation of interacting complexes with LL-37 (Herasimenka et al., 2005). Pseudomonas aeruginosa alginate and exopolysaccharides from other lung pathogens were reported to inhibit the bactericidal activity of LL-37, indicating that sequestration of LL-37 by exopolysaccharides lowers the concentration of AMP at its target site (Foschiatti et al., 2009).
Anchored to the bacterial OM, the O-polysaccharide of lipopolysaccharide has also been shown to contribute to AMP resistance in several Gram-negative pathogens. Loss of the Bordetella bronchiseptica O-polysaccharide, which is negatively charged because of the presence of uronic acid, rendered mutant strains highly susceptible to various AMPs (Banemann et al., 1998). As for CPS and exopolysaccharide, O-polysaccharide has been proposed to act as a protective shield preventing AMPs from interacting with the bacterial membrane. Similarly, S. Typhimurium mutants lacking the O-polysaccharide were more susceptible to polymyxin B (Nagy et al., 2006; Ilg et al., 2009). In contrast, loss of the B. cenocepacia O-polysaccharide did not result in higher sensitivity to polymyxin B (Loutet et al., 2006), suggesting some heterogeneity in shielding effects between bacterial species.
Polysaccharides appear to not be the only bacterial surface structures able to trap AMPs. In a recent study, curli fimbriae expressed by UPEC were shown to bind LL-37 and increase resistance to this AMP (Kai-Larsen et al., 2010). Binding of LL-37 to both monomeric and polymeric CsgA, the major curli subunit, might be due to the overall negative charge of CsgA at physiological pH.
Modification of the bacterial OM
In Gram-negative bacteria, the lipid A and core moieties of lipopolysaccharide can be covalently modified either within the OM or during lipopolysaccharide synthesis and transport to the OM. Lipopolysaccharide modifications are often regulated by environmental stimuli through two-component signaling systems. They promote virulence, modulate the TLR4-mediated inflammatory response, and confer resistance to AMPs (Miller et al., 2005). Lipopolysaccharide modifications, especially those of the lipid A moiety, were shown to largely impact bacterial resistance to AMPs by reinforcing the OM permeability barrier and neutralizing the negative charges of lipopolysaccharide thereby preventing AMP binding (Fig. 1c). Although lipopolysaccharide modifications have been most extensively studied in S. Typhimurium, their importance in conferring resistance to AMPs is also evident for many Gram-negative pathogens including Yersinia spp., E. coli, P. aeruginosa, and Neisseria spp. (Richards et al., 2010).
PagP is an OM enzyme that transfers a palmitoyl group from phospholipids to lipid A, resulting in a hepta-acylated lipid A. In S. Typhimurium, this modification was shown to reinforce the OM permeability barrier and increase resistance to the AMPs C18G and protegrin (Guo et al., 1998). Interestingly, PagP remains dormant in the OM, and it becomes activated upon OM disruption leading to perturbation in the lipid asymmetry (Jia et al., 2004). Disruption of the OM by self-uptake of AMPs is therefore likely to be one of the signals stimulating PagP activity.
Other lipopolysaccharide modifications occur at the periplasmic side of the OM prior to lipopolysaccharide transport to the OM. The arnBCADTEF operon (also known as pmrHFIJKLM operon) is responsible for the biosynthesis and transfer of L-Ara4N to the 4′phosphate of lipid A. The positive charge harbored by the l-Ara4N group at physiological pH neutralizes the negative charge of the 4′ phosphate and strongly contributes to polymyxin B resistance (Gunn, 2008). Interestingly, members of the Burkholderia cepacia complex that are inherently resistant to high concentrations of polymyxin B constitutively modify their lipopolysaccharide with l-Ara4N, and this modification is essential for cell viability (Loutet & Valvano, 2011). In contrast, Franscisella novicida uses a different strategy to resist polymyxin B; the lipid A phosphatase LpxF removes the phosphate group at the 4′-position of lipid A (Wang et al., 2007).
PmrC and CptA are phosphoethanolamine (pEtN) transferases that mediate the addition of pEtN to the 1 and 4′ phosphates of lipid A and to the phosphate of heptose 1 found in the lipopolysaccharide core, respectively (Gunn, 2008). Although these modifications have been shown to have a modest effect on S. Typhimurium resistance to polymyxin B, addition of pEtN to Neisseria gonorrhoeae and N. meningitidis lipid A greatly increased resistance to polymyxin B, LL-37, and protegrin (Tzeng et al., 2005; Lewis et al., 2009).
Pumping AMPs in or out of the cell
Bacterial transporters are divided into importers and exporters belonging to different families (Davidson et al., 2008). Members of the ABC transporter and the resistance-nodulation-division (RND) efflux pump families have been implicated in AMP resistance. ABC importers, which usually rely on a periplasmic-binding protein, are believed to import AMPs from the periplasm or the periplasm–inner membrane interface into the cytosol, where they are most likely proteolytically degraded and recycled as nutrients (Fig. 1d). In contrast, exporters of the RND family are thought to export AMPs from the intracellular environment into the extracellular environment. It appears most likely that RND pumps capture AMPs from the periplasm or from the periplasm–inner membrane interface, rather than from the cytoplasm. Export from the periplasm of various antibiotics that cannot cross the cytoplasmic membrane has been documented for RND pumps (Aires & Nikaido, 2005).
The evidence for involvement of ABC transporters in AMP resistance came from the generation of strains in which the transporter genes were deleted or inactivated. These strains were more susceptible to AMPs than the wild-type strains, as judged by performing survival assays or determining minimum inhibitory concentrations. Screening for S. Typhimurium mutants hyper-susceptible to the AMP protamine led to the identification of the sapABCDF operon coding for the Sap (Sensitive to antimicrobial peptides) ABC importer (Parra-Lopez et al., 1993). In addition to protamine susceptibility, the sap mutants exhibited hypersensitivity to the bee-derived AMP melittin and crude extracts from human neutrophil granules. Importantly, this study revealed that SapA, the transporter periplasmic-binding component, is required for AMP resistance, suggesting that AMPs directly bind to SapA and are subsequently transported by the Sap system into the cytoplasm (Parra-Lopez et al., 1993). The sap genes are also present in a number of other Gram-negative bacterial species. In Erwinia chrysanthemi, a phytopathogen that causes soft rot diseases in crops, a sap mutant strain was more sensitive than wild type to the plant AMPs α-thionin and snakin-1 (Lopez-Solanilla et al., 1998). In non-typeable Haemophilus influenzae (NTHI), a mutation in the sapA gene conferred increased sensitivity to killing by chinchilla β-defensin 1 (Mason et al., 2005). In a more recent study, Mason et al. (2011) reported that the Sap system is also required for heme-iron acquisition and that AMPs compete with heme for SapA binding. Importantly, direct evidence of Sap-mediated AMP import into the bacterial cytoplasm and subsequent proteolytic degradation was recently provided (Shelton et al., 2011). In Haemophilus ducreyi, the Sap transporter plays a role in resistance to LL-37 but not to human defensins (Mount et al., 2010). Interestingly, the Sap transporter of Vibrio fischeri did not confer resistance to any AMP tested, including LL-37 (Lupp et al., 2002). Thus, the Sap system does not appear to confer resistance to AMPs to all bacterial species expressing sap genes, and the specificity of the transporter depends on the ability of SapA to bind given AMPs. The yejABEF operon encodes for an ABC-type transport system that putatively imports peptides. Deletion of S. Typhimurium yejF, the ATPase component of the transporter, resulted in increased sensitivity to protamine, melittin, polymyxin B, and human β-defensins 1 and 2 (Eswarappa et al., 2008). Escherichia coli yejABEF has also been implicated in bacterial uptake of the bacteriocin microcin C (Novikova et al., 2007).
Efflux pumps of the RND family of transporters have been reported to export AMPs out of the cell. Loss of the N. gonorrheae MtrCDE efflux pump resulted in increased susceptibility of gonococci to LL-37 and the porcine AMP protegrin-1 (Shafer et al., 1998). Similarly, deletion of mtrC in H. ducreyi resulted in increased sensitivity to human LL-37 and β-defensins, but had little effect on α-defensin resistance (Rinker et al., 2011). The involvement of the AcrAB efflux pump in bacterial AMP resistance is more controversial. Deletion of the acrAB genes in K. pneumoniae decreased bacterial survival in the presence of polymyxin B, α- and β-defensins (Padilla et al., 2010). In contrast, deletion of the same genes in E. coli did not appear to affect survival in the presence of LL-37, α- and β-defensins (Rieg et al., 2009).
Downregulation of AMP expression
Another strategy that Gram-negative pathogens may employ to resist killing by AMPs is to actively suppress their expression by host cells (Fig. 1e). Shigella spp. inhibit the expression of LL-37 and some β-defensins in intestinal epithelial cells through a mechanism that requires a functional type III secretion system and the mxiE transcriptional regulator (Islam et al., 2001; Sperandio et al., 2008). This was associated with penetration of the bacteria deeper into the intestinal crypts. S. Typhimurium induce a decrease in cryptdin expression in mice that is dependent on PhoP and the type III secretion system-containing pathogenicity island SPI1 (Salzman et al., 2003b). Exotoxins of Vibrio cholerae and enterotoxigenic E. coli (ETEC) have also been reported to downregulate LL-37 and hBD1 expression by host cells (Chakraborty et al., 2008), and N. gonorrhoeae is reported to suppress AMP gene transcription, though unknown mechanisms (Bergman et al., 2005).
Gram-negative pathogens use multiple mechanisms to resist AMPs. The respective contribution of each resistance mechanism is unclear, and a large degree of heterogeneity between species in terms of the relative importance of each mechanism of resistance appears to exist. A better understanding of the mechanisms of bacterial resistance to AMPs is warranted, both to better understand the host/pathogen interaction and to facilitate efforts to exploit AMPs for therapeutic interventions. There are several non-mutually exclusive avenues that can be explored for clinical applications of AMPs. First, host AMPs could be administrated exogenously to kill bacteria and/or act synergistically with other antimicrobials. However, high doses of AMPs may have adverse effects because of their multiple biological activities in the host (Yeung et al., 2011). Second, synthetic AMPs (peptidomimetic oligomers) optimized for maximal bactericidal activity and devoid of adverse effects may be developed. These synthetic AMPs could easily be made resistant to both host and bacterial proteases by, for example, the incorporation of d-amino acids. An alternative approach is to target bacterial resistance to AMPs by developing compounds that target one or several of the resistance mechanisms described above. Such a strategy would disarm pathogens' ability to resist AMPs thereby promoting the bactericidal activity of endogenous AMPs. This type of “anti-virulence” approach is though to avoid the selective pressure leading to resistance, making it a potentially attractive alternative approach to conventional antibiotics.
This work was supported by the Canadian Institutes of Health Research (CIHR, MOP-15551) and the Natural Sciences and Engineering Research Council (NSERC, 217482). S.G. is supported by a Canada Research Chair. The authors thank J.L. Thomassin and J. Brannon for critical reading of the manuscript.