Antimicrobial Activities and Structures of Two Linear Cationic Peptide Families with Various Amphipathic β-Sheet and α-Helical Potentials

Abstract

Many naturally occurring antimicrobial peptides comprise cationic linear sequences with the potential to adopt an amphipathic α-helical conformation. We designed a linear 18-residue peptide that adopted an amphipathic β-sheet structure when it was bound to lipids. In comparison to a 21-residue amphipathic α-helical peptide of equal charge and hydrophobicity, this peptide possessed more similar antimicrobial activity and greater selectivity in binding to and inducing leakage in vesicles composed of bacterial membrane lipids than vesicles composed of mammalian membrane lipids (J. Blazyk, R. Weigand, J. Klein, J. Hammer, R. M. Epand, R. F. Epand, W. L. Maloy, and U. P. Kari, J. Biol. Chem. 276:27899-27906, 2001). Here, we compare two systematically designed families of linear cationic peptides to evaluate the importance of amphipathicity for determination of antimicrobial activity. Each peptide contains six lysine residues and is amidated at the carboxyl terminus. The first family consists of five peptides with various capacities to form amphipathic β-sheet structures. The second family consists of six peptides with various potentials to form amphipathic α helices. Only those peptides that can form a highly amphipathic structure (either a β sheet or an α helix) possessed significant antimicrobial activities. Striking differences in the abilities to bind to and induce leakage in membranes and lipid vesicles were observed for the two families. Overall, the amphipathic β-sheet peptides are less lytic than their amphipathic α-helical counterparts, particularly toward membranes containing phosphatidylcholine, a lipid commonly found in mammalian plasma membranes. Thus, it appears that antimicrobial peptides that can form an amphipathic β-sheet conformation may offer a selective advantage in targeting bacterial cells.

Many host defense peptides function by virtue of their ability to induce leakage in the plasma membranes of target microorganisms without causing extensive damage to host membranes (23). One class of these cationic antimicrobial peptides consists of linear molecules with the potential to adopt an amphipathic α-helical conformation. While these peptides lack a secondary structure in solution, interactions with lipid bilayers in membranes induce an amphipathic α-helical structure that is believed to be a prerequisite for antimicrobial activity. Disruption of membrane integrity could result from either the formation of pore-like structures (13, 24) or a more general disruption (26). Magainin 2 has been shown to forms channels consisting of toroidal pores (36), while cecropin carpets the membrane surface and destabilizes lipid bilayer (12).

A second structural class includes peptides such as mammalian defensins (22) and protegrins (31) with several intramolecular disulfide bonds that stabilize a conformation containing amphipathic β sheets. This class also is believed to exert its antimicrobial effects primarily through membrane disruption. For example, protegrin-1, like magainin 2, forms toroidal pores (35). Two other less common classes includes extended peptides that are rich in one or two amino acids, such as indolicidin (29), and loop peptides, such as thanatin (10). These classes contain well over 500 naturally occurring cationic antimicrobial peptides, and their structural and functional properties have recently been reviewed (2, 27).

Much effort has been devoted to the design of analogs of the naturally occurring peptides that possess enhanced antimicrobial activities coupled with low toxicities toward host cells. Hemolysis of red blood cells has been used most often as a measure of the peptides' effects on mammalian cell integrity. For instance, replacement of the glutamate at position 19 in magainin 2 with alanine increased both antimicrobial and hemolytic activities (5, 14). An increase in the cationic character of magainins from +5 to +7 was shown to enhance hemolytic activity as well (6).

We previously studied a more potent derivative of PGLa (15, 32) consisting of a 21-residue amidated peptide containing three heptameric repeats, (KIAGKIA)₃-NH₂. This peptide has the ability to form a highly amphipathic α-helical conformation, with all six of its lysine residues clustered on one side of the helical face. We compared the structural and functional properties of this peptide to those of another peptide with the same overall hydrophobicity and charge that is unique and different from the classes described above. This 18-residue linear amidated peptide with three hexameric KIGAKI repeats, (KIGAKI)₃-NH₂, has no propensity to form an amphipathic α helix but can form a highly amphipathic β-sheet structure. The antimicrobial potency of (KIGAKI)₃-NH₂ was equivalent to that of (KIAGKIA)3-NH₂. Significantly, (KIGAKI)₃-NH₂ appeared to be much more selective in inducing leakage in and binding to membranes containing phosphatidylethanolamine (PE), a neutral phospholipid found in many bacterial plasma membranes, than membranes containing phosphatidylchoine (PC), a major neutral lipid in mammalian plasma membranes, suggesting that this linear amphipathic β-sheet peptide may have an advantage in discriminating between bacterial and mammalian cells (1).

(KIGAKI)₃-NH₂ represents a new class that is unique from the β-sheet antimicrobial peptides described above since it lacks disulfide bonds or other covalent bonding constraints to stabilize secondary structure. The results presented here extend our prior studies on (KIGAKI)₃-NH₂ to test systematically (i) the importance of amphipathic structure (either as an α helix or a β sheet) in determining antimicrobial activity, (ii) whether our results with (KIGAKI)₃-NH₂ can be extended to other linear amphipathic β-sheet peptides, and (iii) if so, whether these peptides also share similar selectivity advantages over α-helical antimicrobial peptides. In order to test these hypotheses, we synthesized two families of peptides with the same overall charge (+7) and mean hydrophobicity (−0.08) using a consensus hydrophobicity scale (8). Each peptide contains six lysine residues and is amidated at the carboxyl terminus. The two families contain peptides with a wide range of amphipathic characters (i.e., either no amphipathic potential, the potential to form an amphipathic α helix, or the potential to form an amphipathic β sheet) (Tables 1 and 2).

TABLE 1.

Amino acid sequences of Hex and Hept peptide families

TABLE 2.

Mean hydrophobic moments and antimicrobial and hemolytic activities

Peptidea	μH		MIC (μg/ml)d			% Hemolysis (500 μg/ml)
	α helixb	β sheetc	E. coli	S. aureus	P. aeruginosa
1,2-Hex	0.00	0.02	128	>256	128	0
1,3-Hex	0.00	0.60	8	16	16	1
1,4-Hex	0.00	0.18	32	>256	64	0
1,5-Hex	0.00	0.63	8	8	8	0
1,6-Hex	0.00	0.02	>256	>256	>256	0
1,2-Hept	0.30	0.01	8	16	16	13
1,3-Hept	0.14	0.15	32	256	>256	0
1,4-Hept	0.40	0.02	8	8	32	16
1,5-Hept	0.38	0.14	8	8	16	17
1,6-Hept	0.05	0.02	>256	>256	>256	0
1,7-Hept	0.25	0.11	8	16	16	8

^aSee Table 1 for peptide sequences.

^bμ_H was calculated by using a consensus hydrophobicity scale and a side chain offset angle of 100°, which assumes that the peptides adopt an α-helical conformation (6).

^cμ_H was calculated by using a consensus hydrophobicity scale and a side chain offset angle of 180°, which assumes that the peptides adopt a β-sheet conformation (6).

^dFor MICs, differences of greater than a factor of 2 are considered significant.

The data show that antimicrobial potency is associated with an amphipathic structure as either an α helix or a β sheet. Furthermore, the active amphipathic β-sheet peptides are less hemolytic and appear to be more selective than the amphipathic α-helical peptides. These results support our earlier findings and suggest that antimicrobial peptides with an amphipathic β-sheet potential may be sufficiently potent and selective for clinical development.

MATERIALS AND METHODS

Peptide design and synthesis.

All peptides were synthesized by using 9-fluorenylmethoxy carbonyl chemistry on an Advanced Chem Tech model 90 peptide synthesizer. The crude peptides were purified by reverse-phase high-pressure liquid chromatography (HPLC). Purity was checked by reverse-phase HPLC, capillary electrophoresis, and electrospray mass spectrometry. The Hex family comprises five 18-residue peptides made up of a trimeric hexamer repeat in which there are two lysines. One is fixed at position 1; and the other is located at either position 2, 3, 4, 5, or 6. The other amino acids in the hexamer repeat (I, G, A, and I) remain constant. The Hex family repeats consist of the following: 1,2-Hex, KKIGAI; 1,3-Hex, KIKGAI; 1,4-Hex, KIGKAI; 1,5-Hex, KIGAKI; and 1,6-Hex, KIGAIK. Similarly, the Hept family comprises six 21-residue peptides made up of a trimeric heptamer repeat in which there are two lysines. One is fixed at position 1; and the other is located at either position 2, 3, 4, 5, 6, or 7. The other amino acids in the heptamer repeat (L, A, G, L, and A) remain constant. The Hept family repeats consist of the following: 1,2-Hept, KKLAGLA; 1,3-Hept, KLKAGLA; 1,4-Hept, KLAKGLA; 1,5-Hept, KLAGKLA; 1,6-Hept, KLAGLKA; and 1,7-Hept, KLAGLAK. All peptides contain a single tryptophan residue at either position 8, 9, or 10 that replaces an isoleucine or leucine. The complete peptide sequences are listed in Table 1.

The amphipathicity of each peptide was estimated by calculating the mean hydrophobic moment (μ_H), assuming that the peptide adopts either an α-helical or a β-sheet conformation (8), as shown in Table 2. For the Hex family, none of the peptides are amphipathic as α helices. 1,3-Hex and 1,5-Hex can form highly amphipathic β-sheet structures, each with a μ_H of 0.6, while 1,4-Hex is slightly amphipathic as a β-sheet (μ_H = 0.2). 1,2-Hex and 1,6-Hex are not amphipathic either as an α helix or β sheet. For the Hept family, 1,2-Hept, 1,4-Hept, 1,5-Hept, and 1,7-Hept have the potential to form amphipathic α-helical conformation (μ_H = 0.3-0.4). 1,3-Hept and 1,6-Hept are not strongly amphipathic either as an α helix or β sheet.

Materials.

1-Palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE), and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoglycerol (POPG) were used as supplied from Avanti Polar Lipids, Inc. Calcein, trifluoroethanol, and buffer materials were from Sigma Chemical Co. The phosphorus content in lipid stock solutions was determined by a spectrophotometric analysis (4).

Antimicrobial and hemolytic assays.

Antimicrobial susceptibility testing with Staphylococcus aureus (ATCC 29213), Escherichia coli (ATCC 25922), Escherichia coli ML-35, and Pseudomonas aeruginosa (ATCC 27853) was performed by a modification of the CLSI (formerly National Committee for Clinical Laboratory Standards) microdilution broth assay (1). Mueller-Hinton broth (BBL) was used to dilute the peptide stock solution and to dilute the bacterial inoculum. The inoculum was prepared from mid-logarithmic-phase cultures. Microtiter plate wells received aliquots of 100 μl each of the inoculum and the peptide dilution. The final concentration of peptide solution ranged from 0.25 to 256 μg/ml in twofold dilutions. The final concentration of bacteria in the wells was 5 × 10⁵ CFU/ml. Peptides were tested in duplicate. In addition to the test peptide, three standard peptides and a nontreated growth control were included to validate the assay. The microtiter plates were incubated overnight at 37°C, and the absorbance was measured at 600 nm. The MIC was defined as the lowest concentration of peptide that completely inhibited the growth of the organism. Hemolysis at peptide concentrations of 500 μg/ml was determined by using a 5% suspension of freshly drawn human erythrocytes which had been washed twice in phosphate-buffered saline. After incubation at 37°C for 30 min, the suspension was centrifuged at 10,000 × g for 10 min and the absorbance at 400 nm was measured. Complete hemolysis was determined by adding 0.2% Triton X-100 in place of the peptide (1).

CD spectroscopy.

Circular dichroism (CD) spectra were measured in a 1-mm quartz cuvette by using a Jasco J-715 spectropolarimeter. Spectra were recorded from 250 to 190 nm at a sensitivity of 100 millidegrees, a resolution of 0.5 nm, a response of 8 s, a bandwidth of 1.0 nm, and a scan speed of 50 nm/min, with a single accumulation. The buffer contained 5 mM potassium phosphate, pH 7.0. The peptide concentration was 20 μM. Large unilammelar vesicles (LUVs) were prepared from aqueous dispersions of phospholipids at a concentration of ∼1 mg/ml in phosphate buffer. Following five freeze-thaw cycles, the mixture was extruded 10 times through a 0.1-μm-pore-size polycarbonate membrane in an Avanti miniextruder apparatus, resulting in ∼100-nm-diameter LUVs.

Peptide-induced leakage from calcein-loaded LUVs.

The ability of the peptides to release calcein (M_r = 623) from LUVs of various lipid compositions was compared. LUVs were prepared as described above, except that the buffer consisted of 50 mM HEPES, 100 mM NaCl, 0.3 mM EDTA, and 80 mM calcein, pH 7.4. Calcein-loaded vesicles were separated from free calcein by size-exclusion chromatography with a Sephadex G-50 column and calcein-free buffer. Calcein leakage was monitored with a Varian (Walnut Creek, CA) Cary Eclipse spectrofluorometer by measuring the time-dependent increase in the fluorescence of calcein (excitation, 490 nm; emission, 520 nm). Assays were performed by measuring eight wells simultaneously in a 96-well plate. A 180-μl aliquot of the LUV suspension was added to each well by using a multichannel pipette, followed by the addition of a 20-μl aliquot of peptide solutions at various concentrations to give the desired lipid-to-peptide ratio of 2 to 256. The final concentration of lipid in the assay was 10 μM. A negative control omitted peptide. Complete leakage was determined by the addition of 20 μl of 10% Triton X-100 in place of the peptide. Each value represents at least six separate measurements obtained with at least two different LUV preparations. We determined that the percentage of calcein leakage at 3 min following the addition of the peptide represents at least 90% of the maximal leakage. For all peptides tested, maximal leakage was observed after 10 to 15 min. Each value represents duplicate measurements obtained with three different LUV preparations, with error bars calculated as the standard deviation of the six datum points for each value.

Peptide binding to LUVs measured by tryptophan fluorescence.

Peptide interactions with LUVs were measured with a Varian Cary Eclipse spectrofluorometer equipped with a manual polarizer accessory by the method developed by Ladokhin et al. (18). Fluorescence emission spectra were collected from 290 to 500 nm at 1-nm increments by using an excitation wavelength of 280 nm at a signal-to-noise ratio of 500 and a quartz cuvette (2 by 10 mm) at 25°C. The excitation and the emission slit widths were 10 nm and 5 nm, respectively. LUVs were prepared as described above by using a buffer containing 50 mM HEPES, 100 mM NaCl, and 0.3 mM EDTA, pH 7.4. The peptide and lipid concentrations were 10 μM and 500 μM, respectively, for a lipid-to-peptide ratio of 50:1.

All spectra were collected with emission and excitation polarizers oriented at 90° and 0° relative to the vertical, respectively. The peptide interactions with the LUVs were assessed by determination of the tryptophan emission intensity at 330 nm. Measured intensity values at 330 nm were corrected for the light-scattering effects of the LUVs at each lipid concentration by using the ratio of intensities (Ivalues) at 330 nm of 10 μM tryptophan (Trp) in the absence and the presence ofLUV, as shown in the following equation: I_corrected = I_measured · (I_{Trp + buffer}/ I_{Trp + LUVs}). Each value represents triplicate measurements for two different LUV preparations, with error bars calculated as the standard deviation of the six datum points for each value.

Peptide-induced leakage of ONPG into E. coli ML-35 cells.

Diffusion of extracellular o-nitrophenyl-β-d-galactopyranoside (ONPG) into the cytoplasm of E. coli ML-35 cells was monitored by measuring the production of o-nitrophenol (11, 21, 30). E. coli ML-35 constitutively expresses cytoplasmic β-galactosidase and is lactose permease deficient, which thus prevents the uptake of ONPG. Peptide-induced permeabilization of the plasma membrane is measured by the rate of production of o-nitrophenol, which absorbs strongly at 405 nm, following cleavage from galactose by β-galactosidase. A 100-μl aliquot of logarithmic-phase E. coli ML-35 (10⁸ CFU/ml) was added to 700 μl of 10 mM sodium phosphate, 100 mM NaCl, and 1.5 mM ONPG, pH 7.5, in a 37°C preheated cuvette. A 200-μl aliquot containing the desired concentration of peptide was added at zero time. An equivalent volume of 0.5% trifluoroacetic acid replaced the peptide solution in the negative control. The rate of o-nitrophenol production was monitored at 405 nm for 15 min. Complete permeabilization (100% control) was measured by using sonicated bacteria. None of the peptides had any direct effect upon β-galactosidase activity. In all cases, the reaction rate was linear over the 15-min time course. The percent maximal rate of ONPG cleavage was calculated from the ratio of the peptide-induced reaction rate to that of the 100% control. Each value represents duplicate measurements obtained with two different E. coli ML-35 cultures, with error bars calculated as the standard deviation of the four datum points for each value.

RESULTS

Comparison of antimicrobial and hemolytic activities of the peptides.

The antimicrobial and hemolytic activities of the Hex and Hept peptides are shown in Table 2. For the Hex family, only the 1,3- and 1,5-peptides showed appreciable antimicrobial activities against the three test organisms. All of the Hex peptides were nonhemolytic at 500 μg/ml. For the Hept family, the 1,2-, 1,4-, 1,5-, and 1,7-peptides possessed relatively low MICs. These peptides, however, were more hemolytic (8 to 17% hemolysis at 500 μg/ml) than the active Hex peptides. Only 1,3-Hept and 1,6-Hept were nonhemolytic, but both peptides showed negligible antimicrobial activities.

Peptide-induced leakage from calcein-loaded LUVs and E. coli ML-35 cells.

A comparison of the Hex family of peptides for their abilities to release calcein from LUVs with various lipid compositions is shown in Fig. 1. None of these peptides was effective at promoting leakage in LUVs composed of neutral POPC, while all of the peptides were able to induce complete leakage from anionic LUVs composed of POPG at lipid-to-peptide ratios of 4 to 8. In LUVs composed of a 4:1 mixture of neutral (either POPC or POPE) and acidic (POPG) lipids, none of the Hex peptides induced more than 30% calcein release. Significantly, however, the two Hex peptides with good antimicrobial activities, 1,3-Hex and 1,5-Hex, caused much greater leakage in POPE-POPG LUVs than in POPC-POPG LUVs. The three other members of the Hex family with little propensity to form amphipathic β-sheet structure (1,2-Hex, 1,4-Hex, and 1,6-Hex) were all less effective at promoting calcein release from POPE-POPG LUVs than from POPC-POPG LUVs.

FIG. 1.

Percent release of calcein from LUVs composed of POPC (open bars), POPG (solid bars), 4:1 (mol/mol) POPC-POPG (fine hatched bars), or 4:1 (mol/mol) POPE-POPG (coarse hatched bars) 3 min following the addition of Hex peptides.

In contrast, the Hept family showed strikingly different behaviors in the calcein leakage experiments. As shown in Fig. 2, the four Hept peptides with good antimicrobial activities, 1,2-Hept, 1,4-Hept, 1,5-Hept, and 1,7-Hept, induced considerable leakage in neutral POPC LUVs, even at relatively high lipid-to-peptide ratios (64 to 128). The two inactive peptides, 1,3-Hept and 1,6-Hept, were not very effective at causing calcein release from POPC LUVs. As with the Hex peptides, all of the Hept peptides were able to promote the complete release of calcein from acidic POPG LUVs at lipid-to-peptide ratios below 16. The active Hept peptides, however, were very effective at inducing leakage from POPC-POPG LUVs, in comparison to the effectiveness of the active Hex peptides, which induced negligible release. In all cases for the Hept peptides, much greater potency was observed for calcein release from POPC-POPG LUVs than from POPE-POPG LUVs. Interestingly, of the four active Hept peptides, all except 1,7-Hept were able to cause at least 20% calcein release from POPE-POPG LUVs; no release was detected for 1,7-Hept.

FIG. 2.

Percent release of calcein from LUVs composed of POPC (open bars), POPG (solid bars), 4:1 (mol/mol) POPC-POPG (fine hatched bars), or 4:1 (mol/mol) POPE-POPG (coarse hatched bars) 3 min following the addition of Hept peptides.

The Hept peptides were much more effective at inducing leakage in E. coli ML-35 cells than the Hex peptides, as shown in Table 3. Of the four Hept peptides with an MIC of 2 μg/ml, 1,2-Hept, 1,4-Hept, and 1,5-Hept were particularly potent, inducing 60 to 80% leakage over 30 min. In contrast, 1,5-Hex, which also possessed an MIC of 2 μg/ml, was able to induce only 16% leakage from E. coli ML-35 cells over this time period.

TABLE 3.

Effects of peptides on permeability of E. coli ML-35 membranes

Peptide	MIC (μg/ml)a	% Leakageb
1,2-Hex	64	4 ± 5
1,3-Hex	4	30 ± 1
1,4-Hex	32	4 ± 6
1,5-Hex	2	16 ± 5
1,6-Hex	128	8 ± 4
1,2-Hept	2	64 ± 1
1,3-Hept	32	5 ± 3
1,4-Hept	2	79 ± 12
1,5-Hept	2	70 ± 16
1,6-Hept	128	3 ± 2
1,7-Hept	2	31 ± 3

^aMIC for E. coli ML-35.

^bPercent leakage reflects the ratio of peptide-induced entry of exogenous o-nitrophenyl-β-d-galactopyranoside, where it is cleaved by cytoplasmic β-galactosidase, in intact versus sonicated E. coli ML-35 cells at 30 min following the addition of peptide. In each case, the peptide concentration was equal to the MIC.

Secondary structures of the peptides.

The conformations of the peptides were assessed by CD spectroscopy. The CD spectra of all of the Hex and Hept peptides were characteristic of a random structure (with a minimum below 200 nm) in buffer (data not shown). Figure 3 shows a comparison of the secondary structures of the peptides in the presence of LUVs containing POPG, 4:1 POPC-POPG, or 4:1 POPE-POPG at a lipid-to-peptide ratio of 20. For the Hex peptides with POPG LUVs (Fig. 3A), only 1,5-Hex and 1,3-Hex (which both possess significant antimicrobial activities) showed a maximum near 200 nm and a minimum just below 220 nm, suggesting a β-sheet structure, with the former possessing a greater degree of a β-sheet structure than the latter. The remaining Hex peptides appeared to be unstructured. In the presence of POPG LUVs, the CD spectra of all of the Hept peptides, in contrast, were characteristic of an α-helical structure, with a maximum near 190 nm and minima near 208 and 222 nm (Fig. 3D). In LUVs composed of a mixture of neutral and acidic lipids, the conformations of the Hex and Hept peptides differed significantly. With 4:1 POPC-POPG LUVs, none of the Hex peptides exhibited structure (Fig. 3B), while all of the Hept peptides adopted an α-helical conformation equal to or greater than that observed with POPG LUVs. In contrast, with 4:1 POPE-POPG LUVs, the two active Hex peptides (1,3-Hex and 1,5-Hex) showed some β-sheet structure (Fig. 3C), but not as much as they did in the presence of POPG LUVs. All of the Hept peptides with the exception of 1,6-Hept were slightly α-helical in the presence of 4:1 POPE-POPG LUVs, but much less so than with 4:1 POPC-POPG LUVs. These differences in the two mixed lipid systems, where the neutral lipid is composed of POPC or POPE, could arise from at least two factors: (i) differences in the ratio of free to bound peptides or (ii) differences in the conformation of bound peptides.

FIG. 3.

CD spectra of peptides in the presence of POPG LUVs (A and D), 4:1 (mol/mol) POPC-POPG LUVs (B and E), or 4:1 (mol/mol) POPE-POPG LUVs (C and F) at a lipid-to-peptide ratio of 20. (A to C) Hex peptides 1,2-Hex (solid gray), 1,3-Hex (solid black), 1,4-Hex (dashed gray), 1,5-Hex (dashed black), and 1,6-Hex (dash-dotted gray); (D to F) Hept peptides 1,2-Hept (solid black), 1,3-Hept (solid gray), 1,4-Hept (dashed black), 1,5-Hept (dash-dot black), 1,6-Hept (dashed gray), and 1,7-Hept (dash-double dot black). The black and gray lines correspond to peptides with high and low antimicrobial activities, respectively.

Fluorescence emission of tryptophan.

Since each peptide contains a single tryptophan residue, its fluorescence emission was used to assess the degree of binding through changes in the local polarity of the tryptophan side chain. In aqueous solution, the emission spectra for all peptides were nearly identical, with a maximum near 355 nm. The influence of LUVs upon fluorescence emission properties was assessed by measuring changes in intensity at 330 nm. The lipid-peptide association generally results in both a blue shift and an increase in intensity in the emission spectrum. Since the change in the emission maximum does not correlate linearly with the proportion of peptide bound, we used the method of Ladokhin et al. (18) as a linear estimate of peptide binding.

Changes in tryptophan emission intensity at 330 nm (corrected for scattering effects as described above) were measured at a lipid-to-peptide ratio of 50 by using LUVs composed of either POPC, POPG, 4:1 POPC-POPG, or 4:1 POPE-POPG, as shown in Fig. 4. All of the Hex peptides (Fig. 4A) showed little interaction with POPC and a much stronger interaction with POPG (relative intensity increase, 2.5 to 3). None of the Hex peptides showed a significant interaction with either 4:1 POPC-POPG or 4:1 POPE-POPG LUVs, with relative intensity increases comparable to those observed with POPC LUVs. Figure 4B shows that the Hept peptides generally exhibited higher relative intensities in all lipid systems. In particular, 1,4-Hept and 1,5-Hept showed greater interactions with POPC LUVs, and all of the Hept peptides showed stronger interactions with POPG LUVs (relative intensities, 4 to 5) than the Hex peptides. Most striking, however, were the results obtained with the two mixed lipid systems. For the four active Hept peptides (the 1,2-, 1,4-, 1,5-, and 1,7-Hept peptides) the degree of interaction with 4:1 POPC-POPG LUVs was nearly as great as that observed with POPG LUVs. Only the two inactive Hept peptides (1,3- and 1,6-Hept) showed significantly less interaction. In contrast, all of the Hept peptides interacted much less with 4:1 POPE-POPG LUVs.

FIG. 4.

Relative intensity changes in the tryptophan emission at 330 nm of the (A) Hex and (B) Hept peptides in the presence of POPC LUVs (open bars), POPG LUVs (solid bars), 4:1 (mol/mol) POPC-POPG LUVs (fine hatched bars), or 4:1 (mol/mol) POPE-POPG LUVs (coarse hatched bars). The peptide and lipid concentrations were 10 μM and 500 μM, respectively, for a lipid-to-peptide ratio of 50. The intensities were corrected for scattering effects, as described in the Materials and Methods.

DISCUSSION

Most naturally occurring antimicrobial peptides are cationic and amphipathic in nature. Some cytotoxic peptides, like melittin, are segregated into polar and nonpolar regions by virtue of their primary structures. The overwhelming majority of antimicrobial peptides, however, become amphipathic only by adopting an appropriate secondary structure (i.e., either an α helix or a β sheet). Examples of the former class include magainins (37) and cecropins (20), while defensins (28) and tachyplesin (25) are representative of the latter class (23). In our previous work (1, 16), we found that, unlike the native β-sheet antimicrobial peptides, in which the secondary structure is stabilized by disulfide linkages, linear peptides could be induced to form β sheets upon binding to lipid bilayers or membranes by optimizing the primary sequence for this secondary structure.

We showed that (KIGAKI)₃-NH₂ not only forms a β-sheet structure but also appears to be more selective in binding to and disrupting lipid vesicles resembling bacterial membranes (POPE and POPG mixtures) than those resembling mammalian membranes (POPC) (1). In an effort to determine whether these properties were unique to (KIGAKI)₃-NH₂ or general to this novel class of antimicrobial peptides, we compared two different families of peptides that were identical in charge and overall hydrophobicity.

The properties of membrane-binding amphipathic peptides have been studied for more than 20 years and were described in the seminal works of Kaiser and Kézdy (17) and DeGrado and colleagues (19). The relationship between amphipathic character and activity in α-helical peptides is well established. A strong correlation between hydrophobic moment and helicity with membrane binding and perturbation was shown for these peptides (7, 33, 34); however, since no systematic comparison of the importance of amphipathicity in α-helical versus β-sheet peptides has been demonstrated to date, we examined how differences in amphipathic character influenced the structural and functional properties of the peptides that adopted these secondary structures.

The data presented here confirm that the ability to form a highly amphipathic structure (either an α helix or a β sheet) is essential for the antimicrobial potencies of these peptides. In the Hex family, where three of the five peptides had the potential to form an amphipathic β-sheet structure and no possibility of forming an amphipathic α-helical structure, there was a strong correlation between amphipathicity and antimicrobial activity. 1,3-Hex and 1,5-Hex, which both possess μ_H values >0.6 as β sheets, were the most potent (i.e., lowest MICs) against all organisms tested (Table 2). Four of the six Hept peptides with μ_H values ≥0.25 as α helices were significantly more antimicrobially active than the other two members of the Hept family (Table 2). Therefore, since all of these peptides are identical in net charge and overall hydrophobicity, these properties alone are not sufficient to endow antimicrobial activity.

There are, however, noteworthy differences between the active Hex and Hept peptides. The Hept peptides were more hemolytic than the Hex peptides at 500 μg/ml (Table 2). In addition, the Hept peptides are generally much better at inducing leakage in E. coli ML-35 cells than the Hex peptides (Table 3). 1,5-Hex was not nearly as effective at promoting leakage in these bacteria as 1,4-Hept and 1,5-Hept, even though the MICs for all of these peptides were 2 μg/ml. Because 1,5-Hex possessed such limited lytic and membrane-binding abilities, we concluded that the antimicrobial activity of this peptide may involve mechanisms beyond simply increasing plasma membrane permeability (16). If the peptide can penetrate into the cytoplasm, 1,5-Hex may interact with intracellular components, as is the case with some other antimicrobial peptides (3).

None of the Hex peptides could promote the release of calcein from POPC LUVs, while all of the antimicrobially active Hept peptides were lytic even at high lipid-to-peptide ratios. Both 1,3-Hex and 1,5-Hex, the only peptides that can form a highly amphipathic β-sheet structure, were more effective at inducing leakage in POPE-POPG LUVs (Fig. 1) than in POPC-POPG LUVs, in marked contrast to the Hept peptides, particularly those with α-helical μ_H values ≥0.25 (Fig. 2). These lytic preferences are confirmed by the structural data, which showed the peptide conformation in the presence of LUV (Fig. 3) and peptide-LUV interactions probed by tryptophan fluorescence (Fig. 4).

The sole structural difference between POPC and POPE is the additional three methyl groups present in choline compared with the structure of ethanolamine. Therefore, the volume of the polar head group of POPC is greater than that of POPE, which results in different packing properties of these lipids in bilayers and membranes. Unsaturated PC lipids form stable bilayers in water, while unsaturated PE lipids usually form hexagonal phases in water (9). The major neutral lipid in many bacterial plasma membranes, such as E. coli, is PE rather than PC, which is abundant in mammalian plasma membranes. Significant differences in the ability to bind to and permeabilize membranes containing PC and those containing PE as the neutral lipid were observed for the amphipathic β-sheet and α-helical peptides. The most likely explanation for this behavior is that the β-sheet peptides bind more easily and penetrate more deeply into bilayers containing the smaller-volume PE as the neutral lipid than those containing PC. The fact that the α-helical peptides bind more efficiently and have a greater destabilizing effect upon membranes containing PC is consistent with their enhanced hemolytic activities.

In view of their low hemolytic activities, lack of interaction with PC-containing membranes, and high antimicrobial activities, linear cationic peptides with high amphipathic β-sheet potentials may be promising candidates for antimicrobial agents, especially for topical applications, with good selectivity between bacterial and mammalian cells. We are extending our investigation of this class of peptides by examining shorter (≤11 residues) and even more amphipathic sequences that possess greater antimicrobial potencies than the 18-residue peptides described here.