Binding of protegrin-1 to Pseudomonas aeruginosa and Burkholderia cepacia

Background Pseudomonas aeruginosa and Burkholderia cepacia infections of cystic fibrosis patients' lungs are often resistant to conventional antibiotic therapy. Protegrins are antimicrobial peptides with potent activity against many bacteria, including P. aeruginosa. The present study evaluates the correlation between protegrin-1 (PG-1) sensitivity/resistance and protegrin binding in P. aeruginosa and B. cepacia. Methods The PG-1 sensitivity/resistance and PG-1 binding properties of P. aeruginosa and B. cepacia were assessed using radial diffusion assays, radioiodinated PG-1, and surface plasmon resonance (BiaCore). Results The six P. aeruginosa strains examined were very sensitive to PG-1, exhibiting minimal active concentrations from 0.0625–0.5 μg/ml in radial diffusion assays. In contrast, all five B. cepacia strains examined were greater than 10-fold to 100-fold more resistant, with minimal active concentrations ranging from 6–10 μg/ml. When incubated with a radioiodinated variant of PG-1, a sensitive P. aeruginosa strain bound considerably more protegrin molecules per cell than a resistant B. cepacia strain. Binding/diffusion and surface plasmon resonance assays revealed that isolated lipopolysaccharide (LPS) and lipid A from the sensitive P. aeruginosa strains bound PG-1 more effectively than LPS and lipid A from resistant B. cepacia strains. Conclusion These findings support the hypothesis that the relative resistance of B. cepacia to protegrin is due to a reduced number of PG-1 binding sites on the lipid A moiety of its LPS.


Introduction
Cystic fibrosis (CF) is a generalized exocrine disease resulting from the defective regulation of epithelial chloride ion transport within various organ systems, most importantly the lungs [1][2][3][4]. This dysfunction can arise from a variety of mutations within the cystic fibrosis transmembrane conductance regulator (CFTR) gene on chromosome 7, which encodes the transmembrane pump responsible for regulating transepithelial ion levels [1,2,4]. This defect is associated with an increased viscosity of the airway surface fluid, and according to some studies, an altered salt concentra-tion [5,6]. These and other pleiotropic effects of the CF gene abnormality increase the susceptibility of CF patients to bronchopulmonary bacterial infection by opportunistic pathogens.
Pseudomonas aeruginosa is the most common pathogen recovered from the lungs of CF patients [2,3]. This predisposition to P. aeruginosa infection has been attributed to several factors. For example, an increase in asialo-GM 1 glycoproteins on the apical surface of CF airway epithelial cells provides greater opportunity for adhesion by the type IV pili of P. aeruginosa [3,7,8]. Furthermore, the first extracellular domain of wild type CFTR has been found to act as a P. aeruginosa-specific receptor, which is involved in the clearance of P. aeruginosa from the airways via epithelial cell desquamation [9][10][11], a process greatly impaired in the CF lung. Other conditions that abet colonization of the CF lung include impairment of the "mucociliary escalator" and macrophage-mediated phagocytosis by the viscous airway secretions. If elevated salt concentrations exist within these secretions [5,6], they would inhibit α and β-defensins, endogenous antibacterial peptides produced by neutrophils and airway epithelia, respectively [2,3,[12][13][14][15]. This last point is still under review as recent evidence suggests that CF airway surface fluid may be isotonic not hypertonic [16].
In recent years Burkholderia cepacia has become the second leading cause of death in patients with chronic granulomatous disease (CGD). The phagocytes of such patients are deficient in their capacity to kill organisms by oxidative means because of a defective NADPH oxidase [17,18]. B. cepacia also poses a threat to patients with CF, either cocolonizing those already infected with P. aeruginosa or as the sole infecting agent [2,18]. Although rates of colonization can vary widely (between 5 and 70%), B. cepacia is isolated from sputum in only 10%-20% of patients with CF [2,3]. B. cepacia, a very diverse species, has been divided into several "classes" of bacteria, i.e., genomovars, collectively referred to as the B. cepacia complex. Genomovar III is the most common B. cepacia genomovar isolated from CF patients, accounting for approximately 80% of B. cepacia isolates in CF patients. Genomovars II and V have also been recovered from CF patients [18]. Of critical concern are B. cepacia's transmissibility from one patient to another and its propensity to give rise to the B. cepacia syndrome, which results in a rapid decline in pulmonary function [2,18]. The ability of B. cepacia as well as P. aeruginosa to cause chronic bronchopulmonary infections in CF patients is exacerbated by their intrinsic or acquired resistance to many conventional antibiotics. Peptide antibiotics, including protegrins, are currently under consideration as novel agents for treating pulmonary infections in CF patients.
This study tested the hypothesis that the relative sensitivity or resistance of P. aeruginosa and B. cepacia strains to PG-1 correlates to the extent of peptide binding. A previously described two-stage radial diffusion technique designed for peptide antibiotics was used to determine susceptibility. Binding of protegrins to intact P. aeruginosa and B. cepacia cells and to purified lipopolysaccharide (LPS) and lipid A prepared from these organisms was measured by three different techniques.

Bacterial strains and growth media
Two sets of genetically related P. aeruginosa strains were studied. Strain 144M, a serum-sensitive mucoid isolate from a CF patient, contains short O-side chain LPS, while its serum-resistant derivative, 144M(SR), which is also mucoid, has long O-side chain LPS [33]. Strain FRD-1 is a mucoid CF isolate [34] and FRD-2 is its spontaneous nonmucoid derivative [35] [36]. Log-phase cultures were grown at 37°C with shaking in trypticase soy broth (TSB) (Becton-Dickinson Microbiology, Cockeysville, MD).

Protegrin
PG-1 was prepared as previously described [25] and was approximately 96.5% pure. PG-1 was dissolved in and diluted with filter-sterilized acidified water (0.01% acetic acid) supplemented with 0.1% human serum albumin (Sigma, St. Louis, MO). The albumin in the PG-1 vehicle helped reduce nonspecific adsorption of protegrins when the stock solutions were serially diluted [22,37]. These PG-1 stock solutions were filter-sterilized prior to storage at either -20°C for short term (<3 months) or -80°C for long-term storage.

Radial diffusion assay
The two-stage radial diffusion assay protocol as described by Steinberg and Lehrer [37] and Lehrer et al. [38] was al-(page number not for citation purposes) tered by supplementing the underlay agar with 150 mM NaCl to ascertain PG-1's ability to kill P. aeruginosa and B. cepacia at an elevated salt concentration. These assays were interpreted as described below. The sample wells received 5 µl of PG-1 (various concentrations) or its vehicle. We analyzed the radial diffusion assay results in two ways. In one, we calculated the lethal concentration (C L ) as described by Hultmark et al. [39,40] from wells surrounded by net clear zones of 1 to 2.5 mm. In the other, we defined the minimal active concentration (MAC) as the lowest concentration of PG-1 that produced a measurable, totally clear, inhibition zone around the well.

Whole cell binding assay
Binding of PG-1 to intact cells was measured with a monoradioiodinated variant of PG-1. This protegrin variant differed from PG-1 in two ways: Phe7 replaced Tyr7 in the protegrin domain, and an N-terminal glycine-rich hexapeptide extension (GGGYGG) with a single tyrosine residue was present. In this report the modified peptide and its monoiodinated variant will be called "GGPG" and "I-GGPG", respectively. The full sequence of GGPG is GGGYGGRG-GRLCFCRRRFCVCVGR-CONH 2 . In preliminary experiments, we determined that PG-1, GGPG, and I-GGPG exhibited virtually identical antimicrobial potency and kinetics (data not shown).
GGPG was iodinated with NaI or Na 125 I (Pierce, Rockford, IL) by the iodobead method [41,42]. Unincorporated NaI or Na 125 I was removed by solid phase extraction on a Sep-Pak C-18 Plus Cartridge. I-GGPG was purified from uniodinated and di-iodinated GGPG by reverse phase-HPLC (Vydac, Hesperia, CA). Binding assays were done with stationary phase P. aeruginosa 9027 or B. cepacia 25416, both typically at 2.5 × 10 7 CFU/ml. The bacteria were incubated with 125 I-GGPG at 0° C for 60 min at pH 7.4 in a 10 mM Tris-acetate buffer containing 100 mM NaCl, 1% TSB, and 1% BSA. Samples were layered over 0.3 ml of a cushion composed of 5 parts of dibutyl phthalate and 3 parts of di-isodecyl phthalate (density 20 = 1.01) and centrifuged at approximately 14,000 × g. After removing the supernatant and phthalate oil, the base of the tube was severed with a razor blade and the stub, which contained the pellet, was recovered for radioactivity counting.

LPS and lipid A extraction and analysis
Bacterial LPS was isolated according to the protocol of Darveau and Hancock [43] with some modifications, as described below. Strains were grown overnight in TSB, collected via centrifugation (10,000 × g for 20 min at 0°C), and suspended in a pH 8.0 buffer (5.0 g of wet weight bacteria/30 ml of buffer) that contained 10 mM Tris-hydrochloride, 2 mM MgCl 2 , 100 µg/ml DNase I, and 25 µg/ml RNase I. The slurry was passed twice through a French Press set at 20,000 lbs/in 2 . DNase and RNase were again added, and the mixture was incubated at 37°C for 2 hours. Then, to each 15 ml of suspension, the following were added: 5.0 ml of 0.5 M tetrasodium EDTA in 10 mM Tris buffer, 2.5 ml of 20% SDS in 10 mM Tris buffer, and 2.5 ml of 10 mM Tris buffer. The preparation was mixed and centrifuged at 50,000 × g for 30 min at 20°C to remove the peptidoglycan. Protease (Sigma P5147) at 200 µg/ml was added to the supernatant and incubated at 37°C overnight with agitation. On the next day, 0.375 M MgCl 2 in 95% ethanol was added and the suspension was cooled to 0°C before centrifugation at 12,000 × g for 15 min at 0°C. The resulting pellet was resuspended in 2% SDS, 0.1 M tetrasodium EDTA in 10 mM Tris buffer and sonicated. This solution was incubated at 85°C for 30 min, and cooled to room temperature before protease (25 µg/ml) was added and incubated overnight at 37°C. After re-precipitation the following day with ethanol, the pellet was resuspended in 10 mM Tris buffer and centrifuged at 200,000 × g for 2 hours at 20°C in the presence of 25 mM MgCl 2 . The final LPS extract was suspended in 10 ml dH 2 O and dialyzed overnight against dH 2 O to remove any residual salt and SDS prior to being lyophilized. Contamination of the LPS samples by protein represented no more than 3.6% ± 0.21%, as estimated by measuring the dry weight and protein concentration with the Bio-Rad Protein Assay Kit. Nucleic acid contamination was considered to be minimal due to the repeated DNase and RNase treatments. LPS preparations were diluted using the PG-1 vehicle prior to assaying with PG-1.
Lipid A was extracted from isolated LPS preparations via acid hydrolysis [44,45]. Isolated LPS (25-30 mg) was suspended in 4 ml of a 1% acetic acid solution and heated in a water bath to 100°C under refluxing conditions for 3 hours. Once the lipid A was hydrolyzed, it precipitated out of solution and was extracted with five sequential 4 ml aliquots of chloroform. Lipid A extracts were collected and washed three times with 10 ml aliquots of dH 2 O before being lyophilized. Lipid A was dissolved in 100% dimethyl sulfoxide (DMSO), which was diluted to 2.5% with the PG-1 vehicle before application in the radial diffusion assay.
In order to measure the ability of LPS and lipid A to bind PG-1, we modified the two stage radial diffusion assay so that we could ascertain the amount of LPS or lipid A (in ng), that reduced the antimicrobial activity of 50 ng of PG-1 by 50 percent (called the effective concentration 50% or EC 50 ). Briefly, 2.5 µl of a 20 µg/ml PG-1 solution and 2.5 µl of LPS or lipid A (at various concentrations based on their dry weight) were combined, and added to the radial diffusion wells. After a 3-hour incubation, a nutrient rich overlay gel was poured and the plates were incubated overnight. The following day, zone diameters (including the wells) were measured in mm. Since the height (h) of the underlay gel was always 1.2 mm, the zone diameters were converted to volumes cleared of viable bacteria with the for-mula for the volume of a cylinder: V=π r 2 h. Since the 3 mm diameter sample well was devoid of both bacteria and peptide, its volume (8.5 µl) was subtracted from the gross clear zone volume (π r 2 h) to derive the net volume rendered free of bacteria. As will be shown below, this net volume was a linear function of the amount of PG-1 placed in the well, for amounts of protegrin between 0-50 ng/ well.

Surface plasmon resonance
Binding of LPS or lipid A to PG-1 was assessed with a Biacore 2000 system (BiaCore AB, Piscataway, NJ); a very similar approach was used to study LPS binding by peptide domains of horseshoe crab factor C [46]. This innovative technology uses surface plasmon resonance to study biomolecular interactions in real time. Briefly, the binding takes place on a small chip coated with a thin layer of metal. It is monitored by monochromatic, p-polarized light whose electric vector component is parallel to the plane of incidence. Under these conditions, the intensity of the reflected light is markedly reduced at a specific incident angle, producing a sharp "shadow". This phenomenon is called "surface plasmon resonance" (SPR) and the incident light angle at which the shadow is observed is the "SPR angle". Detection of the SPR angle depends on changes in the refractive index (RI) of the medium on the nonilluminated side of the chip. Since this RI depends on the mass concentration of macromolecules at this surface, monitoring the SPR angle provides a real-time measure of changes in the surface concentration. The SPR spectrum is a plot of the SPR angle against time, and displays the interaction's progress at the sensor surface. The SPR signal is expressed in resonance units (RU), and 1000 RU is equivalent to a change in surface protein concentration of about 1 ng/mm 2 or of about 6 mg/ml in the bulk protein concentration.
In our study, LPS and lipid A samples from P. aeruginosa 9027 and B. cepacia 25416 were immobilized on a hydrophobic HPA sensor chip (BR-1000-30, Biacore AB) that was found, in our preliminary experiments, to provide an optimal matrix for LPS and lipid A. The surface of HPA sensor chips is composed of long-chain alkanethiol molecules that form a flat, quasi-crystalline hydrophobic layer that facilitates hydrophobic adsorption of liposomes or micelles on a user-defined polar lipid monolayer. The chip was precleaned and conditioned by washing twice with 40 mM noctyl glucoside (Sigma Chemical Co.). To immobilize the LPS, 60 µl of a 1 mg/ml LPS preparation was dispersed in 100 mM NaCl containing 10 mM HEPES buffer (pH 7.4), and passed across the chip's surface for 30 min, at flow rate of 2 µl/min. Lipid A was dissolved and immobilized similarly, except that its concentration was reduced to 0.2 mg/ ml. Residual unbound LPS and lipid A were washed away with 0.1 M HCl and 0.1 M NaOH. BSA (0.1 mg/ml) was used to detect nonspecific binding on the surfaces coated by LPS or lipid A. PG-1 binding was measured by observ-ing the change in the SPR angle of the sample bound to the chip as 30 µl of PG-1 (20 µM, 10 µM, 6 µM, 3 µM, 2 µM, and 1 µM) flowed over the sample for 3 min at a rate of 10 µl/min. These PG-1 solutions were also prepared in 100 mM NaCl containing 10 mM HEPES buffer (pH 7.4). Each binding study was done in triplicate and the chip was regenerated between trials by washing it with 0.1 M HCl and 0.1 M NaOH.

Activity of PG-1 on P. aeruginosa and B. cepacia strains
The C L and MAC for PG-1 on P. aeruginosa and B. cepacia were determined in two-stage radial diffusion assays ( Table 1). The mean C L of PG-1 for the six P. aeruginosa strains was 0.30 µg/ml, and the mean C L for the five B. cepacia strains was 5.02 µg/ml, a 16.7-fold difference. The mean MACs for P. aeruginosa and B. cepacia were 0.17 µg/ml and 8.40 µg/ml, respectively. The C L and MAC values for P. aeruginosa and B. cepacia differed significantly (P exact = 0.004) when the data were analyzed by the Mann-Whitney test.
We also compared two sets of genetically related P. aeruginosa strains: 144M/144M(SR) and FRD-1/FRD-2. Whereas the LPS of strain 144M has a very short O-antigen side chain, LPS from strain 144M(SR) has a very long O-antigen side chain. However, despite these differences both strains had comparable C L and MAC values. Similarly, comparison of the alginate producer FRD-1 with FRD-2, its spontaneous nonmucoid derivative, revealed that both strains had very similar C L and MAC values.

Intact/whole cell binding
Incubation of 125 I-GGPG with intact P. aeruginosa 9027 (PG-1 sensitive) and B. cepacia 25416 (PG-1 resistant) cells demonstrated a marked difference in protegrin binding between the two strains ( Fig. 1). When the bacteria were exposed to 12.5 µg 125 I-GGPG /ml, approximately sixfold more protegrin molecules bound to P. aeruginosa than to B. cepacia, and at an 125 I-GGPG concentration of 25 µg/ml, 12 times as many protegrin molecules bound to P. aeruginosa. Another factor also merits mention in this regard. By light microscopy, the B. cepacia cells appeared larger than the P. aeruginosa cells. By electron microscopy, we found that the mean cross sectional diameters and lengths of B. cepacia 25416 cells were 0.755 µm × 2.55 µm. The corresponding values for P. aeruginosa 9027 cells were 0.615 µm × 2.35 µm. Assuming a cylindrical structure for both, we calculated that the surface area of the B. cepacia cells (6.00 µm 2 ) was approximately 33% larger on average than that of P. aeruginosa (4.54 µm 2 ).

Binding of PG-1 to LPS and lipid A
To compare binding of PG-1 to LPS and lipid A from P. aeruginosa 9027 and B. cepacia 25416, we devised a simple

Figure 1
Binding of 125 I-GGPG. Stationary phase bacteria, 2.5 × 10 7 CFU/ml, were incubated with the indicated concentrations of I 125 -GGPG for 60 min at 0 °C, and centrifuged through a phthalate ester "cushion" to separate free and bound peptides. The data points are mean ± SEM values from 5 to 9 replicates obtained in three separate experiments. Using the Mann-Whitney test, P was < 0.05 at values indicated by ** .

** <0.005
Mann-Whitney Sign-Rank Test ** Peptide concentration (µg/ml)  Fig. 2, for protegrin amounts between 0 and 50 ng, the volume of underlay gel that was rendered completely free of viable bacteria was linearly related to the amount of peptide introduced into the well. By mixing a constant amount of protegrin (within this linear range) with graded amounts of purified LPS and lipid A, we could readily determine how much of these ligands were needed to reduce the activity of 50 ng of PG-1 by 50% (the EC 50 values). The results of our assays with LPS and lipid A purified from the strains we used in the above 125 I-GGPG binding studies are shown in Fig. 3A To ascertain if these findings were representative, we prepared LPS and lipid A from another pair of bacteria (P. aer-uginosa FRD-2 and B. cepacia C4813) and repeated the studies. The EC 50 values for both LPS and lipid A obtained from P. aeruginosa FRD-2 were each 120 ng. The values for B. cepacia LPS and lipid A were estimated as approximately 2050 ng for both (Fig. 3B), a difference of 17.1-fold when compared to P. aeruginosa FRD-2 LPS and lipid A.

Surface plasmon resonance studies
We also used SPR spectroscopy to compare the binding of PG-1 to LPS and lipid A purified from P. aeruginosa 9027 and B. cepacia 25416. Injection of LPS or lipid A from both strains was stopped when the increase in RU of both reached 1000. Loosely bound molecules were washed away by increasing the flow rate to 100 µl/min, followed by two, one-minute short pulse injections of 100 mM NaOH. Injections of 0.1 mg/ml BSA prepared in the eluent buffer were passed across the LPS or lipid A coated surfaces for a five-minute period and showed an increase of around 100 RU, whereas the same amount of BSA passed Quantitative aspects of PG-1 mediated killing measured by two-stage radial diffusion assays. The left ordinate scale shows the net area of underlay gel totally cleared of viable bacteria using the equation (π d 2 -9)÷ 4, where d is the diameter of the cleared zone including the well diameter and 9 is subtracted to account for the area of the well, which is devoid of bacteria. The right ordinate scale shows the volume cleared (area cleared times the height of the underlay gel, 1.2 mm). All data points are means (n = 3). Note that Y is a linear function of X for these peptide concentrations. Area (A) and volume (V) are related by the equation: A (mm 2 ) = V (µl)/1.2 mm. across the uncoated, n-octyl glucoside washed surface showed an increase of roughly 1000 RU. This demonstrated that the surface of the chip was fully covered by LPS or lipid A. Since similar amounts of these ligands had bound to the chip, the response unit differences seen at concentrations between 1 and 20 µM indicate that approximately twice as much PG-1 bound P. aeruginosa lipid A as bound lipid A from the B. cepacia strain (Fig. 4). Binding differences were less marked when we studied LPS from these organisms (Fig. 5). Although up to one third more protegrin molecules bound to P. aeruginosa LPS than to B. cepacia LPS at protegrin concentrations of 1-6 µM, these differences disappeared when we tested higher PG-1 concentrations. At 10 µM PG-1, LPS from P. aeruginosa and B. cepacia bound PG-1 equally (Ϸ 750 response units). Equal binding was also seen at 20 µM PG-1 (Ϸ 1000 response units).

Discussion
In this study we compared the sensitivity of P. aeruginosa and B. cepacia strains to protegrin PG-1 with the ability of the intact bacteria and key components of their outer membranes (LPS and lipid A) to bind this peptide. The indices of susceptibility (C L and MAC values) obtained from our radial diffusion assays established that P. aeruginosa was considerably more sensitive to PG-1 than B. cepacia. We also found that at any given peptide concentration, many fewer protegrin molecules bound to intact B. cepacia and its lipid A than to the corresponding P. aeruginosa preparations. In model systems, the ability of protegrins to permeabilize membranes shows pronounced "concentrationgating", an indication that it is influenced by the density of bound protegrin molecules per unit area of membrane [47].
As prior quantitative studies of peptide binding are few in number, some readers may be surprised to note that Fig. 1 shows many millions of protegrin molecules are bound to an average bacterial cell, even at concentrations around the minimal active concentration. To acquire an equivalent density of surface-bound PG-1 molecules, B. cepacia needed exposure to much higher concentrations of PG-1 than P. aeruginosa. Thus, it is a reasonable inference that the relative resistance of B. cepacia to PG-1 is a consequence of Binding of PG-1 to LPS and lipid A. Binding/radial diffusion assays were conducted by mixing various amounts of LPS or lipid A with 50 ng of PG-1 in a final volume of 5 µl. In Panel A, the LPS and lipid A preparations were isolated from P. aeruginosa strain 9027 and B. cepacia strain 25416. In Panel B, the LPS and lipid A were isolated from P. aeruginosa FRD-2 and B. cepacia C4813. The mixtures of PG-1 with LPS or lipid A were introduced into the wells of radial diffusion assay plates seeded with P. aeruginosa FRD-1, a highly susceptible test organism. All data points are means (n = 3), and EC 50 values were extrapolated from these graphs. When these binding isotherms were subjected to linear regression analysis, the 95% confidence limits of the LPS and lipid A binding isotherms for P. aeruginosa and B. cepacia were non-overlapping, and the isotherms for the LPS preparations in Panel A overlapped only at the lowest ligand concentrations (data not shown).
(page number not for citation purposes) the smaller number of protegrin molecules bound per unit area of bacterial surface.
Earlier studies have shown that certain structural differences in the LPS of Gram-negative bacteria correlate with their susceptibility or resistance to antimicrobial (lipo)peptides [19,23,[48][49][50][51]. Particular attention has been given to alterations of the outer membrane that could modify binding to the cell surface. Given the notorious resistance of B. cepacia to conventional antibiotics [2], its relative resistance to PG-1 was not unexpected. What bacterial properties influenced sensitivity to protegrins? Certainly, the striking correlation with the extent of protegrin binding must be one of these. Our studies with genetically related P. aeruginosa strains did not implicate the length of the O-antigen side chain as a key factor. Furthermore, genetically related P. aeruginosa strains that differed in alginate production were equally sensitive to PG-1. Both results were somewhat surprising, since long O-side chain and alginate molecules should both bind PG-1. Perhaps limited alginate production under our experi- Why did fewer protegrin molecules bind to B. cepacia than to P. aeruginosa? Our radial diffusion/binding and SPR studies strongly implicated lipid A as a prime suspect in the case. The acyl chains of lipid A form an integral part of the outer membrane bilayer, and electrostatic or hydrophobic binding of protegrin to lipid A could easily disturb the organization and acyl-chain packing of the outer membrane, both vital in maintaining its integrity and barrier function. Accepting a direct proportionality between measured plasmon resonance (RU) and the extent of peptide bound to the sensor surface, the responses shown in Fig. 4 indicate that at saturation (20 µM), lipid A from B. cepacia bound almost exactly one-half as many protegrin molecules as lipid A from P. aeruginosa. How might this be rationalized?
Lipid A typically contains two phosphate groups, one on each glucosamine residue. In the lipid A of B. cepacia, one of these phosphates is modified by a phosphodiesterlinked 4-amino-4-deoxyarabinose residue [58,59]. It is noteworthy that a previous study reported that the phosphate content of B. cepacia LPS was only one-third of that of P. aeruginosa [60]. If most of the P. aeruginosa lipid A phosphates are unmodified, and if these phosphates are principal protegrin binding sites, then one would expect exactly the results shown in Fig. 4. This is also consistent with the suggestion, made in a recent report showing that cystic fibrosis isolates of P. aeruginosa had lipid A isoforms derivatized with aminoarabinose (or palmitate), that these structural modifications could enhance resistance to cationic antimicrobial peptides [61]. Although the acyl chains of lipid A might also provide alternative binding sites for protegrin, the orientation of lipid A on the HPA sensor chip ("butter-side down") probably removes them from consideration, at least under our study conditions.  binds about as much PG-1 as P. aeruginosa LPS (Fig. 5) especially at peptide concentrations of 10-20 µM, which allow binding to lower affinity sites. This interpretation helps explain why Fig. 3 suggests that B. cepacia LPS binds protegrin less effectively than the corresponding P. aeruginosa preparation, while the SPR results in Fig. 5 show that B. cepacia and P. aeruginosa LPS bound equivalent amounts of PG-1 at high (10-20 µM) protegrin concentrations. To resolve these seemingly discrepant results, it is important to recall that in SPR assays, the sensor chips were constantly bathed with a fixed protegrin concentration ranging from 1-20 µM and that in radial diffusion assays, the wells received a fixed initial amount of protegrin (50 ng). In the latter assays, the peptide was dispersed in a total volume of 10 µl, making its maximal initial concentration approximately 2.3µM -well below the 10-20 µM concentrations needed to saturate LPS. Moreover, in addition to binding the LPS or lipid A molecules we added to the well, some of the initially added protegrin could diffuse radially into the underlay gel. Both initial binding and radial diffusion will decrease the concentration of free (i.e., unbound) protegrin that remains in the well, and further skew its binding towards higher affinity binding sites on lipid A.
Although our data clearly establish the presence of a reduced number of binding sites for protegrin in lipid A from B. cepacia, as compared to lipid A from P. aeruginosa, they do not identify the cause of this phenomenon. These alterations could result from 4-aminoarabinose substitutions on the lipid A, from other changes that reduce the number of ion-binding sites, or from changes that affect hydrophobic interactions or membrane insertion. Additional studies that characterize B. cepacia LPS and lipid A can help pinpoint the exact structural differences that relate to the relative resistance of B. cepacia to protegrins and to endogenous antimicrobial peptides of humans.

Conclusion
Our studies demonstrate a correlation between protegrin sensitivity/resistance and protegrin binding in P. aeruginosa and B. cepacia, and support a hypothesis that the relative resistance of B. cepacia to protegrin is due principally to a reduced number of protegrin binding sites on the lipid A moiety of its LPS.