The Remarkable Innate Resistance of Burkholderia bacteria to Cationic Antimicrobial Peptides: Insights into the Mechanism of AMP Resistance

Abstract

Gram-negative bacteria belonging to the genus Burkholderia are remarkably resistant to broad spectrum, cationic, antimicrobial peptides (AMPs). It has been proposed that this innate resistance is related to changes in the outer membrane lipopolysaccharide (OM LPS), including the constitutive, essential modification of outer membrane Lipid A phosphate groups with cationic 4-amino-4-deoxy-arabinose. This modification reduces the overall negative charge on the OM LPS which may change the OM structure and reduce the binding, accumulation, and permeation of cationic AMPs. Similarly, the Gram-negative pathogen Pseudomonas aeruginosa can quickly become resistant to many AMPs by multiple mechanisms, frequently including activation of the arn operon, which leads, transiently, to the same modification of Lipid A. We recently discovered a set of synthetically evolved AMPs that do not invoke any resistance in P. aeruginosa over multiple passages, and thus are apparently not inhibited by aminorabinosylation of Lipid A in P. aeruginosa. Here we test these resistance avoiding peptides, within a set of 18 potent AMPs, against Burkholderia thailandensis. We find that none of the AMPs tested have measurable activity against B thailandensis. Some were inactive at concentrations as high as 150 μM, despite all having sterilizing activity at ≤ 10 μM against a panel of common, human bacterial pathogens, including P. aeruginosa. We speculate that the constitutive modification of Lipid A in members of the Burkholderia genus is only part of a broader set of modifications that change the architecture of the OM to provide such remarkable levels of resistance to cationic AMPs.

INTRODUCTION

In the ongoing struggle against rising drug-resistant bacterial infections, it is imperative to identify novel antibiotic chemotypes that are less likely to induce resistance in order to stably advance infection prevention and treatment options. Resistance to conventional small molecule antibiotics can arise rapidly by selection due to their typical single site of action on a single biomolecule¹. Evolved resistance genes can readily be acquired by horizontal gene transfer^2–4. Furthermore, some means of resistance, such as efflux pump upregulation, can drive generic resistance to many small molecule antibiotics simultaneously. The existence of multiple active resistance pathways can lead to a state of pan drug resistance^{3, 5}.

Cationic antimicrobial peptides (AMPs) that target microbial membranes for disruption have long been considered promising alternatives to conventional antibiotics, and many are in preclinical development^6–13. There are good reasons to believe that they could be developed into effective antibiotics. In vitro, AMPs often have sterilizing microbicidal activity at low μM concentrations against many strains of bacteria^{8–10, 14–19} and thus are active in the same concentration range as conventional antibiotics against susceptible organisms. However, AMPs are generally also active against drug-resistant, multidrug resistant, and pan drug resistant organisms^{1, 20–37}. In fact, resistance to conventional antibiotics is broadly associated with collateral widespread susceptibility to AMPs³⁸.

AMPs have a unique mechanism of action, global membrane disruption³⁹, which effectively transforms an antibiotic barrier into a targeted site of action. AMPs do not target a single site, and thus, are not able to evoke resistance through the mutation of a single target gene^{40, 41}. Also, AMPs are not sensitive to drug-efflux pumps or other mechanisms of multi drug resistance. Nonetheless, resistance to some AMPs has been observed, and it has been experimentally selected^42–45. But avoidance or delay in resistance have also been observed^46–49. For example, in a direct comparison⁴⁶ using the Gram-negative pathogens Acinetobacter baumannii and P. aeruginosa, it was found to be much more difficult for bacteria to evolve resistance to some AMPs than to small-molecule antibiotics under the same conditions⁴⁶. Other groups, including us⁴⁹, have also reported the inability to select for resistance against some AMPs^{46, 47, 49}.

Multiple mechanisms of AMP resistance have been identified for Gram-negative bacteria^50–55, but the “resistome” is small^{40, 41}. The most commonly observed resistance mechanism is the modification of Lipid A phosphate groups with cationic moieties. Lipid A is a variable core lipid moiety that anchors the oligosaccharide component of LPS to the outer membrane (OM). Typical Lipid A molecules contain an anionic, di-phosphorylated disaccharide with 4-6 short/medium chain acyl groups. In response to exposure to sublethal concentrations of some antibiotics, Gram-negative bacteria can upregulate operons for the addition of cationic groups, including ethanolamine, glucosamine, and 4-amino-4-deoxy-arabinose (Ara4N) to the lipid A phosphates. In P. aeruginosa, for example, the addition of Ara4N to lipid A phosphates by the arn operon is a common mechanism of AMP resistance^{41, 51, 56, 57}.

The effect of Lipid A modification on AMP activity can be rationalized by the observation that some, and probably most, AMPs kill bacteria as a result of “self-promoted uptake” ⁵⁸ in which a saturating accumulation of peptide on the cell is required for lethal activity at the cytoplasmic membrane. Depending on the peptide and organism, between 1x10⁷ and 5x10⁸ peptide molecules must be bound to each cell for bactericidal activity^58–61. The polyanionic LPS layer of Gram-negative bacteria is likely saturated with peptide under these conditions, which has led researchers to conclude that massive accumulation on the outer membrane LPS, and perhaps OM destabilization, are necessary prerequisites for the permeabilizing activity of at least some AMPs at the inner membrane^{55, 58, 62}. It is thus reasonable that the most common mechanism of resistance to AMPs in Gram-negative bacteria is the addition of cationic groups to the anionic phosphates of the Lipid A moiety of OM LPS^{41, 50, 56, 63} as this modification will globally replace negative charges on LPS with positive charges, and also may change the LPS structure, reducing the OM binding potential for cationic AMPs.

However, not all AMPs are subject to the development of resistance in Gram-negative bacteria. Some AMPs are thus apparently not inhibited by the modification of Lipid A phosphates by Ara4N. For example, we recently described the synthetic molecular evolution of a family of potent, broad spectrum, hemocompatible antimicrobial peptides⁴⁹ that do not invoke any resistance in multiple passages against P. aeruginosa.. We assume that this means that aminorabinosylation of lipid A cannot confer resistance to these peptides. Under the same conditions, resistance developed rapidly to conventional antibiotics⁴⁹.

Here we extend our characterization of these potent, resistance-avoiding AMPs by testing their activity against Burkholderia thailandensis, which belongs to a genus of Gram-negative bacteria (class betaproteobacteria) that is biochemically similar to Pseudomonas (class gammaproteobacteria). B. thailandensis, like most Burkholderia species, produces Lipid A molecules that are constitutively aminorabinosylated, often on both Lipid A phosphate moieties⁵⁷. Here, we show that B. thailandensis is remarkably resistant to sterilization by cationic AMPs, even those against which P. aeruginosa is highly susceptible and cannot evolve resistance.

RESULTS

Synthetically evolved AMPs.

The unique resistance avoiding AMPs tested here are described elsewhere⁴⁹. They are the product of three generations of synthetic molecular evolution and rational design, undertaken to identify host-compatible, broad-spectrum antimicrobial peptides^{17, 18, 49, 64}. First, from a de novo library, we selected in parallel for members that either permeabilized bacteria-like lipid vesicles or that sterilized multiple species of bacteria simultaneously¹⁷. These peptides, exemplified by the 12-residue peptide ARVA (RRGWALRLVLAY), Fig. 1, have broad spectrum antimicrobial activity, but are potently inhibited by host cell binding, proteolysis, and serum protein binding^{61, 65}. Next, we used ARVA as a template to design a library from which we selected potent AMPs that are not inhibited by concentrated host cells⁴⁹. This screen gave rise to five “double broth sterilization”, or DBS, peptides which we have tested as both protease susceptible L-amino acid (L-DBS) peptides and as protease resistant D-amino acid (D-DBS) peptides, Fig. 1. We also tested a consensus sequence, D-CON (rrgwarrlafafgrr). Finally, a further round of rational optimization⁴⁹ gave rise to our most active AMP, called D-CONGA, which stands for D-amino acid, CONsensus sequence with Glycine Absent, Fig. 1. We note that all of these evolved peptides are C-terminal amides.

Figure 1.

Evolution of D-CONGA. A. The 1^st generation, de novo-designed library was previously screened for synthetic membrane permeabilization and for bacterial sterilization^17.18. B. The 2^nd generation library was designed from ARVA and screened for hemocompatible AMPs against Gram-negative pathogens⁴⁹. C. The 3^rd “generation” consisted of rational variants tested for broad spectrum activity against ESKAPE pathogens and low cytotoxicity. The 3^rd Gen. peptide D-CONGA is the best peptide identified⁴⁹.

As we previously showed, the D-DBS peptides, D-CON, and D-CONGA have excellent antimicrobial properties, suggesting that they could be developed into useful antibiotics, especially in the protection and treatment of wounds⁴⁹. They have potent (MIC ≤ 10 μM) sterilizing activity against all ESKAPE pathogens (E. coli, S. aureus, K. pneumoniae, A. baumanii, P. aeruginosa, and E. faecium) which include both Gram-negative and Gram-positive bacteria. The lead peptides also have potent activity against drug-resistant bacteria in vitro and in vivo, potent activity against biofilms, in vitro and in vivo, and retain activity in the presence of concentrated host cells, tissue, and serum proteins⁴⁹. Finally, they are also highly soluble, resistant to biodegradation, and have low cytotoxicity⁴⁹.

Resistance avoidance.

We previously tested some of these peptides for the ability to induce resistance in P. aeruginosa, which is inherently resistant, or can rapidly become resistant, to many conventional antibiotics and AMPs^{41, 63}. We tested whether P. aeruginosa could develop stable resistance to i) Four conventional antibiotics, ii) D-ARVA, the template peptide for the 2^nd generation library, iii) L-DBS1 and D-DBS1, a sequence selected from the second-generation library, and iv) D-CONGA the best lead peptide evolved during the recent work. As described elsewhere⁴⁹, P. aeruginosa PAO1 was grown in the presence of serially diluted antibiotic, and we selected the culture that grew at the highest concentration of antibiotic. This culture was propagated overnight in the absence of antibiotic to enable selection for stable resistance. The process was repeated for ten passages, which is equivalent to hundreds of generations. Against the four conventional antibiotics, resistance always increased (Fig. 2), and the bacteria evolved complete resistance (MIC ≥ 350 μM) to streptomycin and ceftazidime by 8 passages. Against the library template sequence D-ARVA, P. aeruginosa became resistant (MIC ≥ 67 μM) over four passages (Fig. 2). Slight changes in experimental protocols lead to resistance to D-ARVA in a single passage⁴⁹. However, against the lead 2^nd generation peptides L-DBS1, D-DBS1 and D-CONGA, we observed no measurable increase in resistance over ten passages. The activity of these peptides against P. aeruginosa is not sensitive to experimental details.

Figure 2.

Resistance avoidance in evolved antimicrobial peptides. P. aeruginosa was treated with serially diluted conventional antibiotics or AMPs⁴⁹. The culture that grew at the highest concentration was cultured overnight in the absence of antibiotic, and then the screening with antibiotic/AMP was repeated the next day. This was done for ten passages. A: The MIC values for each passage against four conventional antibiotics. B. The MIC values for each passage against four AMPs. C: Cross resistance of P. aeruginosa passaged against one antibiotic and one peptide and tested against a set of AMPs.

In Fig. 2C we show cross-resistance in P. aeruginosa after 10 passages against either the cephalosporin class antibiotic ceftazidime or the AMP D-CONGA. Passaging against D-CONGA caused no increase in resistance to D-CONGA, but it did cause resistance to D-ARVA, suggesting that the bacteria are responding to D-CONGA in a way that is ineffectual against D-CONGA, but enables resistance to D-ARVA. Even passaging against ceftazidime causes a partial increase in resistance to D-ARVA but did not increase resistance to D-CONGA. The fact that passaging against L-DBS1 did not evoke resistance to L-DBS1 or to L-CONGA indicates that protease secretion is not readily evoked as a resistance mechanism against these peptides in P. aeruginosa. Based on these data, we concluded that there is a unique, and important molecular aspect of the DBS peptides, D-CON, and D-CONGA that prevents or slows the development of resistance, compared to other AMPs. Some other AMPs have also been reported to have this property^46–49. The detailed molecular mechanism of such “resistance avoidance” is unknown, however we are confident that peptides like D-CONGA that do not readily invoke resistance in P. aeruginosa, are not inhibited by aminorabinosylation of Lipid A in that species.

Most Burkholderia species are innately and highly resistant to AMPs^{41, 54, 55, 66}. To test the correlation between resistance avoidance in P. aeruginosa and AMP resistance in Burkholderia, we measured minimum inhibitory concentrations (MIC) of D-ARVA, D-DBS peptides, D-CON, and D-CONGA against B. thailandensis. Activities are compared to Gram-negative P. aeruginosa and E. coli, and to Gram-positive S. aureus. For comparison, we also tested other AMPs from multiple sources, including indolicidin (a bovine neutrophil AMP in L- and D- forms), cecropin A (an insect AMP), melittin (a membrane lytic bee venom peptide⁶⁷), MSI78/pexiganan (a synthetic analog of frog skin AMPs), WLBU2 (a synthetic AMP), and ARNY, RNNY and NATT (16 residue AMPs related to ARVA¹⁷).

We used broth dilution assays in which ~2x10⁵ bacterial cells/ml and serially diluted peptide were mixed in nutrient broth, and the bacteria were allowed to grow overnight. We noted the lowest concentration without any growth and averaged multiple such experiments on the log scale. This assay is a stringent assay for sterilization, as overnight growth will generally allow any survivors to reach high cell density by the next day. Using susceptible bacteria, we have repeatedly shown, by plating on nutrient agar, that wells which remain clear after overnight growth do not contain any live bacteria. Thus, the MIC for an AMP is actually a minimum sterilizing concentration.

In Figure 3, we show the minimum inhibitory concentrations of these 18 different cationic antimicrobial peptides against B. thailandensis, and compare the results to MIC values against P. aeruginosa, E. coli, and S. aureus. The values shown are the average of 3-15 independent measurements made by broth dilution. Numbers shown in red boxes as “>conc” indicate that no sterilization was observed up to the stated maximum tested concentration. Numbers shown as “~conc” indicate that some, but not all, wells were sterilized only at the maximum concentration tested and that the MIC could be just above this maximum tested concentration.

Figure 3.

AMP activity against B. thailandensis, compared to E. coli. P. aeruginosa and S. aureus. Log averaged MIC values for 18 cationic antimicrobial peptides were measured as described⁴⁹. Serially diluted peptides were added to 2x10⁵ bacteria in growth media in 96 well plates and were incubated overnight at 37°C. After overnight incubation, optical density of each well was measured at 600 nm. Most measurements were equal either to transparent sterile media control wells or to opaque stationary phase growth control wells. In 3-15 repeat experiments the lowest sterilizing concentration was noted. The results were averaged in log space and the average was converted to concentration. Red cells with yellow text indicate little or no activity. Green cells with black text indicate measurable MIC value at ≤20 μM. nd=not determined.

Most of these peptides were studied at 20 to 50 μM concentration, but the most active one, D-CONGA, was studied up to 150 μM, as indicated in Figure 3. Remarkably, none of these potent, broad-spectrum AMPs have activity against B. thailandensis. In sharp contrast, most have sterilizing activity at less than 10 μM against most of the other microbes listed (green cells). The Gram-positive organism S. aureus is susceptible 14 of the 18 AMPs tested, which is a typical for this organism. With few exceptions, the D-DBS peptides D-CON and D-CONGA have sterilizing activity at ≤ 10 μM against all ESKAPE pathogens⁴⁹.

For a few of these peptides, we also performed colony forming unit (CFU) reduction assays to measure the reduction in viable bacteria, an assay that does not depend on sterilization. For D-ARVA and D-CONGA the reduction in viable B. thailandensis was less than 10-fold (<1 log) at peptide concentrations of 50-150 μM. This can be compared to sterilization of the 2x10⁵ susceptible P. aeruginosa/ml at 5-10 μM peptide in broth dilution experiments, demonstrating a reduction of viable P. aeruginosa by more than 4 logs at 10-20 fold lower peptide. These results show that B. thailandensis is at least 5 orders of magnitude more resistance to these evolved AMPs than P. aeruginosa.

DISCUSSION

Members of the genus Burkholderia have high inherent resistance to conventional antibiotics and antimicrobial peptides^{55, 68}, including resistance to the polymyxins^{41, 66, 69}, which are approved for use in humans. Resistance in Burkholderia is important because of the potential of some species to cause serious disease in humans. These pathogenic species include the Burkholderia cepacia complex which is prevalent in the lungs cystic fibrosis patients⁷⁰ and the biothreat organism Burkholderia pseudomallei, which causes meliodosis, a disease with high mortality in humans⁷¹. There are multiple mechanisms of resistance in Burkholderia⁵⁵, but a dominant one that is especially effective against AMPs, is the constitutive modification of the outer membrane LPS. For example, in Burkholderia one or both of the outer membrane Lipid A phosphate moieties are modified with cationic Ara4N moieties^{41, 57},^{57, 68, 69}. Some Burkholderia Lipid A variants also have one phosphate removed, with the remaining phosphate being aminoarabinosylated⁵⁷. The core polysaccharide of the Burkholderia LPS can also contain Ara4N mopieties⁶⁹ such that truncation of the LPS core polysaccharide has been shown to cause sensitivity to AMPs in Burkholderia⁷². In many other Gram-negative bacteria, stress operons can be triggered to carry out Lipid A modifications^{41, 51}. However, in Burkholderia these modifications are constitutive⁶⁸. Attempts to inactivate Ara4N modification pathways in Burkholderia⁶⁹ using conditional mutants showed that both the synthesis of Ara4N and the transfer of Ara4N to Lipid A are essential to viability.

One result of these modifications is a significant reduction in the net negative charge on the OM LPS, although structural changes are also possible. These changes likely inhibit the so called “self-directed uptake” of AMPs⁵⁸, a process that includes accumulation of cationic AMPs on the OM^{59, 61} via electrostatic interactions, including displacement of divalent cations⁷³, and subsequent disruption of the outer membrane prior to the lethal disruption of the inner membrane. As a result of the modifications to LPS, the AMP polymyxin⁷⁴ binds much more weakly to Burkholderia OM LPS than to P. aeruginosa LPS⁷⁴ and does not permeabilize the OM of Burkholderia. We and others have added to the understanding of this process by measuring the accumulation of AMPs on bacteria that is necessary for killing. The killing of Gram-negative bacteria by AMPs is a saturation-dependent event in which peptides bind massively to bacterial cells with a requirement that 1x10⁷ to 5x10⁸ peptides are bound to each cell for activity^{39, 59–61}. For example, about 3x10⁸ molecules of D-ARVA are required to kill one E. coli cell, despite an MIC of only 3 μM⁴⁹. Only a small portion of the total bound peptide actually interacts with the inner membrane⁷⁵, which has about 2x10⁷ lipid molecules³⁹, to cause lethal inner membrane permeabilization.

In this work we tested the hypothesis that cationic AMPs that cannot evoke the development of resistance in P. aeruginosa through Lipid A modification will also not be inhibited by the same OM LPS modifications that are constitutive in B. thailandensis. Our results do not support the hypothesis. Instead, we verify that B. thailandensis is remarkably resistant to cationic AMPs, and we show that this innate resistance includes resistance to AMPs that are apparently not inhibited by Lipid A modification by Ara4N in P. aeruginosa.

There are testable ideas that might explain the observation that some peptides cannot invoke resistance in P. aeruginosa but simultaneously have no activity against B. thailandensis. For example, the acute aminorabinosylation of Lipid A in P. aeruginosa may be associated with a secondary increase in inherent OM permeability, or some other reduction in overall fitness. Resistance-avoiding peptides may be those that can cross the leakier OM in P. aeruginosa to access the inner membrane without first massively accumulating on the OM. On the other hand, in Burkholderia aminorabinosylation of Lipid A is both constitutive and essential. Therefore, we speculate that members of the Burkholderia genus have an array of other OM and LPS modifications which act together to stabilize the structure of the LPS. Thus, both reduced charge and increased structure could function cooperatively to reduce binding and permeation of AMPs and other antibiotics across the OM.