Modulation of the Hydrophobic Domain of Polymyxin B Nonapeptide: Effect on Outer-Membrane Permeabilization and Lipopolysaccharide Neutralization

Materials and Methods

Synthesis of PMBN Analogs.

All protected amino acids, coupling reagents, and polymers were obtained from Nova Biochemicals (Laufelfingen, Switzerland) and Bachem (Bubendorf, Switzerland). Synthesis-grade solvents were obtained from Labscan (Dublin, Ireland). Linear peptide chains were assembled by conventional solid phase synthesis, using an automated solid phase multiple peptide synthesizer (AMS-422; ABIMED, Langenfeld, Germany). 9-Fluorenylmethoxycarbonyl strategy was employed throughout the peptide-chain assembly (Atherton and Sheppard, 1989) followingthe company's protocol. Synthesis was initiated by using 9-fluorenylmethoxycarbonyl-Thr(tBu)-Wang resin (0.7 mmol/g) and performed on a 25-μmol scale. Side-chain amino protecting groups for 2,4-diaminobutyric acid were tert-utyloxycarbonyl and benzyloxycarbonyl (Cbz). Coupling was achieved using 4 equivalents of benzotriazole-1-yl-oxy-tris-pyrolidino-phosphonium hexafluorophosphate as a coupling agent in presence of 8 equivalents of 4-methylmorpholine, all dissolved in dimethylformamide (DMF). The fully protected peptide-bound resin was treated with piperidine (20% in DMF) for 20 min, and the free N-terminal amino moiety of Thr⁹reacted with 4 equivalents ofN-(benzyloxycarbonyloxy)succinimide and 4 equivalents ofN,N-diisopropylethylamine in DMF for 3 h. The fully protected peptide-bound resin was treated with trifluoroacetic acid (TFA)/water/triethylsilane (95:2.5:2.5; v/v/v) for 1 h at room temperature and the reaction mixture was filtered. The solution was cooled down to 4°C and the partially protected linear peptide was precipitated with ice-cold di-tert-utyl methyl ether/petroleum ether (30–40°C; 1:3, v/v) and centrifuged. The pellet was washed with the same mixture, dissolved in water/acetonitrile (2:3, v/v) and the solution was lyophilized. Cyclization was performed in DMF at peptide concentrations of 1 mM using benzotriazole-1-yl-oxy-tris-pyrolidino-phosphonium hexafluorophosphate/1-hydroxybenzotriazole/4-methyl morpholin (4:4:8 equivalents) as reagents for 2 h at room temperature (yield >95% according to analytical HPLC). The reaction mixture was concentrated in high vacuum and the cyclic peptidic product was precipitated by treatment with water. Final deprotection, i.e., removal of Cbz, was achieved by a mixture of TFA/bromotrimethylsilane/thioanisol/ethandithiol/m-cerasol (58:10:19:10:3, v/v/v/v/v) at 0°C for 1 h. The product was precipitated by the addition of cold tert-utylmethyl ether, centrifuged, washed with cold tert-utylmethyl ether, and lyophilized from water.

Reversed-Phase HPLC Purification and Analyses.

The crude synthetic peptides were purified with a prepacked LichroCart RP-18 column (250 × 10 mm; 7-μm bead size; E. Merck, Darmstadt, Germany) employing a binary gradient formed from 0.1% TFA in water (solution A) and 0.1% TFA in 75% acetonitrile in water (solution B). The column was eluted at t = 0 min, B = 0%, and at t = 48 min, B = 60%, using a flow rate of 5 ml/min. For purity evaluation, analytical reversed-phase HPLC was performed using a prepacked Lichrospher-100 RP-18 column (250 × 4 mm, 5-μm bead size; E. Merck) with the following binary gradient: at t = 0 min, B = 10%, and at t = 40 min, B = 60% at a flow rate of 0.8 ml/min.

HPLC separations and analyses were performed using a liquid chromatography system (SP8800; Spectra-Physics, Stahnsdorf, Germany) equipped with a variable-wavelength absorbance detector (ABI 757; Applied Biosystems Foster City, CA). The column effluents were monitored by UV absorbance at 220 nm. Purity of peptides was >98% (yields, 30–45%). The corresponding fractions were collected, lyophilized, and analyzed after exhaustive acid hydrolysis and precolumn reaction with 6-aminoquinolyl-N-hydroxysuccinimidyl carbamate to ascertain amino acid composition (2690 separations Module; Waters, Milford, MA). Mass spectra analysis was performed to determine molecular weights (VG-platform-II electrospray single quadropole mass spectrometer; Micro Mass, Manchester, UK).

PMBN.

PMBN was prepared by proteolysis of PMB (Sigma, St. Louis, MO) with papain (Sigma) as described elsewhere (Danner et al., 1989). The crude product was purified (>98%) by HPLC, then analyzed and characterized as described above.

Dansyl-PMBN.

Dansyl-PMBN was synthesized as described elsewhere (Tsubery et al., 2000a).

Calculation of Peptides Hydropathicity (GRAVY).

Peptides grand average hydropathicity (GRAVY) was calculated using the Web-available program developed by Kyte and Doolittle (1982)(http://www.expasy.ch/cgi-bin/protscale.pl). To calculate relative hydropathicity values of the peptides, the Dab residues, (i.e., the unnatural amino acid) were replaced by Lys residues and the peptides were considered linear. Each amino acid was assigned a hydropathy index, a value reflecting its relative hydrophilicity and hydrophobicity. The sum of the hydropathy indices of a given sequence divided by the number of residues in the sequence generates the GRAVY score. In our case, the difference is in only one amino acid; hence, the GRAVY score reflects the different hydropathy indices of the residues at positions 5 and 6.

Molecular Modeling.

The coordinates of PMB were obtained from Pristovsek and Kidric (1999). The model structure of LPS was constructed using the MSI package (MSI Inc., San Diego, CA). The complex PMB-lipid A was assembled in a way that the interactions were as described by Pristovsek and Kidric (1999). The model structure of the complex was energy minimized in vacuum. Initially, only the lipid A molecule was allowed to change, whereas the PMB molecule was fixed at the NMR structure. Subsequently, the restrictions were removed and both molecules were allowed to change. The complex PMBN-LPS was modeled by removal of residues −1 (6-methyl heptanoic acid) and 0 (Dab) from the above PMB-LPS complex and the energy was minimized again. The energy minimization was performed using the consistent-valence force field within the Discover module of the MSI package (MSI, Inc.). The convergence requirement was for the maximum derivative to be less than 0.001.

Determination of Minimal Inhibitory Concentration.

The employed clinical isolates of Escherichia coli, K. pneumoniae, and Pseudomonas aeruginosa were obtained as described elsewhere (Ofek et al., 1994). The Gram-negative bacteria were grown on nutrient agar plates (Difco Laboratories, Detroit, MI) and kept at 4°C. Lyophilized aliquots of peptides (2 mg, determined by weight and ascertained by amino acid composition analysis) were dissolved in sterile double distilled water and filtered using a 0.2-μm Acrodisc filter (Gelman Sciences, Ann Arbor, MI). An overnight culture in Isotonic Sensitest Broth (ISB; Oxoid, Basingstoke, Hampshire, England) was adjusted to 1 × 10⁵CFU/ml and inoculated onto microtiter plate wells, each containing 100 μl of a serial 2-fold dilution (1000–0.5 μg/ml) of the tested antibiotics/peptides in ISB. The MIC was defined as the lowest concentration at which no visible bacterial growth was detected after incubation for 20 h, at 37°C. Results are reported for three to four separate tests.

Outer Membrane Permeabilizing Activity.

Bacterial suspension (10 μl, 1 × 10⁵ CFU) was inoculated onto microtiter plate wells containing 100 μl of a serial 2-fold dilution (1000–0.5 μg/ml) of novobiocin (Sigma) in ISB. To each well, 10 μl of the test peptide was added to a final concentration of 50 μg/ml. The MIC was defined as the lowest concentration at which there was no visible bacterial growth after incubation for 20 h, at 37°C. Results are reported for three to four separate tests.

Dansyl-PMBN Binding and Displacement Assay.

The displacement assay was preformed as follows: 0.55 μM dansyl-PMBN was added to a quartz cuvette containing LPS solution (2 ml, 3 μg/ml, ∼2 × 10⁻⁷ M) in 5 mM HEPES, pH 7.2, and allowed to equilibrate at room temperature for 10 to 15 min. Subsequently, small portions (5–10 μl) of peptide solutions (1 × 10⁻⁶–1 × 10⁻³ M) were added. Inhibition of fluorescence was measured 5 min after each addition of peptides. Percent inhibition was plotted as a function of the added peptide concentration and IC₅₀ values were calculated from maximal specific displacement (I_max).

Circular Dichroism (CD) Studies.

CD spectra were recorded on an Aviv-202 circular dichroism spectrometer (Lakewood, NJ). Duplicate scans over a wavelength range of 190 to 250 nm were taken at a chart speed of 12 nm/min in a 0.1-cm path-length quartz cell at room temperature. Peptides were dissolved in 5 mM phosphate buffer, pH 7.2, at a final concentration of 0.2 mM. A baseline was recorded and subtracted after each spectrum. Ellipticity is reported as the mean residue ellipticity [Θ] in degrees cm²dmol⁻¹ × 10⁻³.

Inhibition of Cytokine Release.

Peptide solutions (1, 10, and 100 μM, final concentration) were incubated (10 min, 37°C) withE. coli-LPS (20 ng/ml, final concentration) in an assay medium (RPMI medium/10% newborn calf serum, 1 mM sodium pyruvate, 1% nonessential amino acids, and 9 μg/ml insulin) in polypropylene tubes. MONO-MAC-6 (MM6) cells (5 × 10⁵/tube) were added and tubes were incubated for 4 h for TNFα production and 18 h for IL-6 production. Cytokine levels were determined using matched antibody pairs according to the manufacturer's guide to custom enzyme-linked immunosorbent assay development protocol (Endogen, MA).

Results

Four PMBN analogs, [d-Trp⁵]PMBN, [d-Tyr⁵]PMBN, [Phe⁶]PMBN, and [Ala⁶]PMBN were synthesized using a combination of linear peptide synthesis and cyclization in solution. The peptides were purified to homogeneity (>98%) by HPLC and their correct amino acid composition and calculated molecular weights were ascertained by amino acid analysis and electrospray mass spectrometry, respectively (Table 1). d-Phe⁵ was replaced either with d-Trp or d-Tyr, whereas Leu⁶ was replaced with Phe or Ala (Fig. 1, Table1). Considering the hydrophobicity on the basis of relative retention time on a RP-18 column, [d-Trp⁵]PMBN and [Phe⁶]PMBN were equally hydrophobic to PMBN whereas [d-Tyr⁵]PMBN and [Ala⁶]PMBN were less hydrophobic (Table4). The hydropathicity scale (Kyte and Doolittle, 1982), however, indicated that [d-Trp⁵]PMBN was much less hydrophobic than PMBN (Table 4).

The peptides' structure was evaluated using CD measurements. [d-Tyr⁵]PMBN and [Ala⁶]PMBN (0.2 mM) in phosphate buffer exhibited a random structure similar to 0.2 mM PMBN. A minor difference in the CD pattern was observed at 218 to 230 nm. [d-Trp⁵]PMBN at the same concentration (0.2 mM) exhibited a maximal negative ellipticity at 200 nm and an additional maximal negative ellipticity at 220 nm. [Phe⁶]PMBN exhibited reduced ellipticity compared with PMBN and a maximal negative ellipticity at 297 nm (Fig.2).

Antimicrobial and OM Permeabilization Activities of PMBN Analogs.

Unlike PMBN, none of the analogs was active against P. aeruginosa (MIC >250 μg/ml, Table2). The peptides' (50 μg/ml) potency to increase the bacterial OM permeability toward novobiocin was evaluated. [Phe⁶]PMBN was as potent as PMBN, whereas [Ala⁶]PMBN was 4- to 8-fold less active than PMBN (Table 3). [d-Trp⁵]PMBN was 8-fold less potent in the OM permeabilization assay compared with PMBN. The activity of [d-Tyr⁵]PMBN, however, was very weak (Table 3).

LPS Binding.

The interaction of the PMBN analogs with the bacterial cell-free LPS was quantified using the dansyl-PMBN displacement assay. Table 4 shows that the peptide concentrations required for 50% displacement were 4 and 5 μM for [d-Trp⁵]PMBN and [Phe⁶]PMBN, respectively, similar to PMBN (IC₅₀ = 2.5 μM). However, [d-Tyr⁵]PMBN and [Ala⁶]PMBN exhibited significantly lower potency compared with PMBN (Table 4).

Neutralization of LPS Toxic Effects.

The ability of PMBN and its analogs to neutralize stimulatory effects of LPS on the human monocyte cell line MM6 was tested. Peptides were preincubated withE. coli- LPS and the mixture was allowed to interact with the MM6 cells. Levels of released TNFα and IL-6 were measured after 4 and 18 h, respectively, using enzyme-linked immunosorbent assay. As shown in Fig. 3, stimulation of MM6 with LPS (20 ng/ml) triggered the release of ∼40 and ∼70 ng/ml of IL-6 and TNFα, respectively. PMBN and [d-Trp⁵]PMBN were equally potent inhibitors of TNFα release (95–23%). A similar effect was observed for the inhibition of IL-6 release. PMBN and [d-Trp]PMBN inhibited IL-6 release at the range of 75–20% in a dose-dependent manner (1–100 μM). However, [d-Tyr⁵]PMBN was a much weaker inhibitor, causing no significant inhibition either in the TNFα or the IL-6 assay (20 and 10%, respectively). As shown, PMB exhibited maximal inhibition capacity even at 1 μM concentration.

Molecular Modeling of Peptide-LPS Complex.

The structure of PMB bound to LPS was determined by NMR spectroscopy (Pristovsek and Kidric, 1999). Based on the coordinates provided for PMB, a model corresponding to 1:1 PMB/lipid A complex was generated (Fig.4). The complex is characterized by electrostatic interactions between four of the five positive side chains of PMB (positions 0, 4, 7, and 8, Fig. 4) and two of the negative phosphate groups of the phosphorylated lipid A head-groups. The hydrophobic side chains at the N terminus and at positions 5 and 6 of PMB interact with the aliphatic chains of lipid A. The PMBN-LPS complex was generated upon removal of residues −1 (6-methyl heptanoic acid) and 0 (Dab) from the PMB molecule in the PMB-LPS complex and energy reminimization. The minimized structure suggests that the interaction between the positive Dab moiety at position 0 and the phosphate group in lipid A is replaced by an electrostatic interaction between Dab² of PMBN (numbering as for PMB) and the same phosphate group. The positive side chain of Dab² in PMB is exposed to the solvent in the PMB-LPS complex (Fig. 4). The hydrophobic segmentd-Phe⁵-Leu⁶still interacts with the aliphatic chains of lipid A.

Discussion

The interaction of PMB and PMBN with their bacterial target molecule, LPS, was extensively studied in recent years. The intermolecular peptide-lipid associations were attributed to the amphiphilic features of PMB, involving primarily the hydrophobic face of the peptide enforced into an appropriate alignment by the positively charged side chains of the 2,4-diaminobutyric acid residues (Srimal et al., 1996). Similar structural considerations were deduced and elaborated from two-dimensional NMR and molecular dynamics studies of PMBN and E. coli-LPS (Bhattacharjya et al., 1997;Pristovsek and Kidric, 1999). According to the latter study, PMB and PMBN share a type II′ β-turn structure for the free peptide and an envelope-like fold of the peptide ring for the LPS-bound peptide. The present structure-function study of PMBN focused on thed-Phe⁵-Leu⁶segment, the peptide ring hydrophobic domain of PMBN. As shown in our previous study, the substitution ofd-Phe⁵ withl-Phe (i.e., construction of [l-Phe⁵]PMBN) resulted in loss of activity (Tsubery et al., 2000a). This finding is in line with the notion that position i +1 of type II′ β-turn is generally occupied by a d-amino acid (or Gly) (Venhatachalam, 1968). Indeed, position 2 of all naturally occurring polymyxins is occupied by the d-form of Phe or Leu, whereas position 6 is occupied by the l-form of either Leu, Ile, or Thr. In the present study, PMBN was modified so that the configuration at positions 5 and 6 was retained but the hydrophobicity was changed.

According to the hydrophobicity scale the amino acids follow the order Phe > Leu > Trp > Ala > Tyr (Kyte and Doolittle, 1982; Engelman et al., 1986). The hydropathicity of the various newly synthesized PMBN analogs was evaluated experimentally using RP-HPLC as well as calculated based on the grand average of hydropathicity (GRAVY) program developed by Kyte and Doolittle (1982). According to RP-HPLC, the peptides hydrophobicity follow the order [d-Trp⁵]PMBN > PMBN = [Phe⁶]PMBN > [Ala⁶]PMBN > [d-Tyr⁵]PMBN. However, the GRAVY scale generated the following order: PMBN > [Phe⁶]PMBN > [Ala⁶]PMBN > [d-Trp⁵]PMBN > [d-Tyr⁵]PMBN (Table 4). The difference might perhaps be caused by the actual overall structures that the peptides assume in the aqueous-organic milieu (acetonitrile-water).

The present CD measurement show that [d-Trp⁵]PMBN has a relative minor negative shoulder at 218 to 240 nm compared with PMBN and exhibits a rather marked negative band centered at 222 nm. This additional maximal negative ellipticity, however, could be attributed to the contribution of the indole moiety of Trp (Atkinson and Pelton, 1992). The CD spectra of [Ala⁶]PMBN was identical to that of PMBN, whereas [Phe⁶]PMBN exhibited a CD pattern with reduced ellipticity. However, no major peptide structural changes were found among the four peptides and all displayed random coil structure.

Although most bacteria are resistant to the bactericidal activity of PMBN, P. aeruginosa is an exception. None of the newly synthesized PMBN analogs was able to inhibit the growth of P. aeruginosa. So far, any modification made in PMBN resulted in loss of direct bactericidal activity toward P. aeruginosa (Tsubery et al., 2000a,b). When the peptides were evaluated for their ability to permeate the bacterial OM, [d-Tyr⁵]PMBN was found to be inactive, whereas [d-Trp⁵]PMBN and [Ala⁶]PMBN exhibited reduced OM permeabilizing activity compared with PMBN. [Phe⁶]PMBN, however, was as active as PMBN. This loss and reduced activities of the respective peptides might perhaps be explained by their GRAVY (Kyte and Doolittle, 1982) (that is, the lower the hydropathicity, the lower the antibacterial activity). Indeed, [Phe⁶]PMBN has an OM permeabilization ability identical to that of PMBN and the values of their hydropathicity index are close. These observations are consistent with the finding that, as with other antimicrobial peptides, amphipathicity is an essential parameter for the activity of PMB and its analogs (Liao et al., 1995; Srimal et al., 1996; Bhattacharjya et al., 1997; Hancock, 1997; Pristovsek and Kidric, 1999).

However, the affinity of the peptides to cell-free LPS as evaluated by the Dansyl-PMBN displacement assay had greater correlation with the hydrophobicity scale drawn by RP-HPLC (Table 4). Thus, the interaction of the peptides with cell-free LPS is somehow different from their interaction with cell-bound LPS.

The interaction of LPS with immune cells via its receptor (CD 14) results in stimulatory effects leading, among others, to enhanced release of cytokines such as TNFα and IL-6 (Viriyakosol and Kirkland, 1995). The ability of PMBN and its analogs [d-Trp⁵]PMBN and [d-Tyr⁵]PMBN to neutralize the stimulatory effect of LPS was evaluated. This capacity relates to the binding of the peptides to the LPS molecule and impairing its association with cognate immune cell receptors. Thus, both PMBN and [d-Trp⁵]PMBN bound to LPS and prevented its interaction with the LPS-receptor on MM6 cells in a similar dose-dependent manner. [d-Tyr⁵]PMBN, however, was not able to inhibit the release of TNFα and IL-6. These results are in good correlation with the affinity of the peptides to cell-free LPS and to the hydrophobic scale drawn by RP-HPLC.

The structures of PMB and PMBN and its analogs bound to LPS were modeled based on coordinates provided for PMB (Pristovsek and Kidric, 1999) (Fig. 4). Elimination of the positively charged residue Dab⁰, which interacts with the negatively charged phosphate group in the PMB-LPS complex, is not likely to significantly weaken the LPS-peptide interaction. This is partly because of the conformational change of the side chain at position 2, allowing for the efficient interaction of the Dab² residue with the negatively charged phosphate group and partly because the exposed phosphate can interact with water (Fig. 4). Thus, modifications at position 2 may lead to the weakening of the interaction. This assumption is supported by our previous observation that substitution at position 2 with Lys reduces the peptide OM permeabilizing activity as well as its affinity to LPS (Tsubery et al., 2000a). Preservation of the cationic nature of the peptides is of great significance in relation to their ability to enhance penetration through biological membrane. However, it was shown that various linear cationic analogs of PMBN analogs are devoid of activity, suggesting that the cyclic structure is of major significance (Vaara, 1991).

The hydrophobic interaction of PMBN with the bacterial membrane is of great significance for its OM permeabilization activity. Indeed, the parent PMB molecule has two hydrophobic regions, the fatty acid moiety at the N terminus and thed-Phe⁵-Leu⁶segment in the peptide ring. The removal of the fatty tail abolished its direct antimicrobial activity. Modulation of the hydrophobic segment of PMBN reduced its OM permeabilization activity. The proximity between the aromatic ring of d-Phe⁵and the side chain of Leu⁶ promotes the formation of the β-turn in the peptide (Pristovsek and Kidric, 1999). Thus, substitution at these positions in PMBN with less hydrophobic residues may impair the stability of this β-turn. Such a structural change may affect the amphiphilic nature of the peptide and, in turn, weaken its LPS binding and activity.