INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

The incidence of fungal infections has increased dramatically in the past 20 years because of the increase in the number of people whose immune systems are compromised by AIDS, aging, organ transplantation, or cancer therapy (3, 11). Accordingly, the increases in the rates of morbidity and mortality because of fungal infections have been regarded as a major problem (46). Most of the current antifungal drugs simply reduce the level of growth of fungi. Amphotericin B, called the drug of last choice for the treatment of most systemic mycoses, is a potent fungicidal agent, but it is very toxic to the kidney and the hematopoietic and central nervous systems (2, 30). The development of resistant fungal strains in response to the widespread use of current antifungal drugs will cause serious problems in the future (23). The recent emergence of fungal infections and resistant strains has stimulated the development of novel antifungal drugs (6, 24, 45).

In the past few years, membrane-active host defense molecules have been isolated from a variety of natural sources (4, 7, 12, 33). Interestingly, they are small peptides or proteins, some of which have been shown to have antibacterial and antifungal activities (4, 39, 47). Although native defense peptides themselves could not be used as therapeutic agents because of their low levels of activity and poor bioavailabilities, these peptides have received attention because of their low levels of toxicity against mammalian cells and the ideal mechanism of perturbing the membrane of the pathogen.

The development of novel antifungal agents from the peptides requires the design and synthesis of large numbers of individual peptides for activity optimization. This process is too time-consuming and limited by the difficulty of designing peptides with the desired structure in the lipid membrane, by nonlinear relationships between activity and structure (5, 13, 34), and by the lack of the detailed structural information concerning the synthesized peptide in the lipid membrane.

As an alternative method to overcoming these limitations, full peptide libraries have been developed (9, 10, 28, 38). However, the antimicrobial peptides identified by these libraries were less active than the currently available antimicrobial agents. Moreover, these peptides might be too short to act on the membrane of the target pathogen.

Since the decapeptide was reported to be the peptide with the minimal length necessary for the interactions of amphipathic -helical peptides with phosphatidycholine liposomes (36) and since the decapeptide derived from tenecin 1, an antimicrobial peptide from Tenebrio molitor, killed pathogens by changing the permeability of the lipid membrane of pathogens (32), combinatorial libraries composed of decapeptide mixtures were required to screen for peptides active against the membranes of pathogens. We expected that combinatorial libraries made up of a few amino acids instead of 20 natural amino acids must provide enough of a peptide mixture to screen for the activity-optimized peptide because the membrane of the pathogen must have less specificity than other biological targets such as enzymes, antibodies, and hormone receptors. Accordingly, we developed combinatorial libraries with decapeptide mixtures composed of seven amino acids (Lys, Leu, Val, Phe, Ser, Gln, and Pro) to screen for the peptide active against Candida albicans. We identified a novel decapeptide which had activity against bacteria as well as fungi but no hemolytic activity.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Synthesis of peptide libraries and individual peptides. Peptide libraries and individual peptides were synthesized on Rink amide methylbenzhydrylamine (MBHA) resin (PerSeptive Biosystem GmbH, Hamburg, Germany) by 9-fluorenylmethoxycarbonyl (Fmoc) chemistry (20, 21, 31). The amino acids used in the synthesis of peptide library were Leu, Lys, Phe, Pro, Ser, Gln, and Val (Calbiochem-Novabiochem Corp., La Jolla, Calif.). Side chain protection groups were as follows: for Lys, tert-butoxylcarbonyl; for Ser, tert-butyl; and for Gln, trityl. In this library, lysine was fixed at the C terminus, and the other nine positions consisted of seven amino acids at equimolar concentrations. A combination step including a division, coupling, and recombination process (28) was used to synthesize the peptide mixture. This process ensured the equimolarity of the peptides on the resin in the packets. Briefly, seven porous polypropylene packets, each containing 0.1 mmol of lysine-MBHA resin, were coupled to each protected Fmoc amino acid. All coupling reactions proceeded to completion (>99.5%), as assessed by the ninhydrin test of Kaiser et al. (29). The resulting resins from each packet were then combined and were thoroughly mixed. This resin mixture was separated into seven portions of equal weights, and these were placed into porous polypropylene packets, followed by removal of the Fmoc group. The resin in the packet was then reacted with solutions of the individual activated amino acid. The division, coupling, and recombination process described above was repeated eight more times, yielding a final mixture of 40,353,607 (7⁹) protected peptide resins. Cleavage of the peptide from the resin was achieved by treatment with a mixture of trifluoroacetic acid (TFA)-thioaninsole-ethane-dithiol-H₂O at a ratio of 80:5:2.5:5 (vol/vol) at room temperature for 12 h. After filtration of the resin and washing with TFA, a gentle stream of nitrogen was used to remove the excess TFA. The crude peptide was triturated with diethyl ether chilled at 20°C and was then centrifuged at 3,000 × g for 10 min. An individual peptide was synthesized by the solid-phase method on a 431A automatic peptide synthesizer (Applied Biosystems, Foster City, Calif.). The peptide was purified by high-performance liquid chromatography with a Waters Delta Pak C₁₈ column (25 by 100 mm; Waters, Milford, Mass.). Amino acid analysis, high-performance liquid chromatography (27), and electrospray mass spectrometry on a Platform II spectrometer (Fishons Instruments, Manchester, United Kingdom) were used to characterize the purified peptide.

Antifungal and antibacterial assays. In vitro antifungal assays were performed by the broth microdilution method by following the recommendation of the National Committee for Clinical Laboratory Standards (41). Sabouraud-2% dextrose broth (pH 5.6 at 25°C; Merck, Darmstadt, Germany) was used as the assay medium. C. albicans ATCC 36232 freshly grown on slopes of Sabouraud dextrose agar were suspended in physiological saline, and the cell concentration was adjusted to 10⁴ cells per 1 ml of 2× concentrated medium for use as the inoculum. Peptide solution was added to the wells of a 96-well plate (100 µl per well), and the wells were serially diluted twofold. The final concentrations of the peptide mixture ranged from 0.2 to 500 µg/ml. After inoculation (100 µl per well, 5 × 10³ cells per ml), the 96-well plate was incubated at 30°C for 48 h, and the absorbance was measured at 620 nm by using an enzyme-linked immunosorbent assay reader (SLT, Salzburg, Austria) to assess cell growth. The MIC was defined as the lowest concentration of the peptide that completely inhibited the growth of the test organism. To measure the minimal fungicidal concentration (MFC), 100 µl of the cell suspension was taken from each well and was washed three times with fresh Sabouraud broth. Then, each cell suspension was plated onto a Sabouraud dextrose agar plate and the plate was incubated at 30°C for 48 h. The MFC was determined by counting the numbers of colonies on the Sabouraud dextrose agar plate. An in vitro antibacterial assay was performed by the aforementioned method used for the antifungal assay, with the exception that the assay medium and the incubation temperature were different. In the antibacterial assay, antibiotic medium 3 (pH 7.0 at 25°C; Difco, Detroit, Mich.) was used, and the cells were incubated at 37°C for 24 h.

Hemolytic assay. The hemolytic assay method used in this study has been described elsewhere (15). Packed mouse erythrocytes were washed three times with buffer (150 mM KCl, 5 mM Tris-HCl [pH 7.4]), and then packed erythrocytes were suspended in 10 volumes of the same buffer (stock cell suspension). For antibiotic treatment, the cell stock suspension was diluted 25-fold with the same buffer and was preincubated in the water bath at 37°C for 15 min, and then the test sample was added. After incubation for 1 h, the sample was centrifuged at 4,000 × g for 5 min and the absorbance of the supernatant was determined at 540 nm. The hemolysis effected by 0.1% Triton X-100 was considered 100% hemolysis.

Preparation of liposomes. Liposomes were prepared by a freezing-thawing method. Lipid mixtures (Sigma, St. Louis, Mo.) were dissolved in chloroform and were dried with a stream of nitrogen gas to form a thin lipid film on the wall of a glass tube. The resulting thin film was hydrated in buffer (pH 7.4) that contained 12.5 mM aminonaphthalene-3,6,8-trisulfonic acid (ANTS), 45 mM N,N'-p-xylenebis(pyridinium bromide) (DPX; Molecular Probe Inc., Eugene, Oreg.), 68 mM NaCl, and 10 mM HEPES (Sigma), shaken for 30 min, and vortexed vigorously for 10 min. The resulting multilamellar vesicles were sonicated and shaken for 1 h at room temperature. The suspension was frozen-thawed for five cycles. The liposomes were separated from the unencapsulated material on Sephadex G-50 (Pharmacia, Upsala, Sweden) that was equilibrated with 10 mM HEPES buffer (pH 7.4) containing 150 mM NaCl and 1 mM ethylene glycol-bis(-aminoethyl ether)-N,N,N',N'-tetraacetic acid (Sigma).

Liposome leakage assay. The liposome leakage assay was based on the quenching of ANTS fluorescence by DPX (18, 43). The quenching of ANTS by DPX occurs by nonradiative energy transfer and depends on the average distance between two molecules. When the leakage occurred, ANTS and DPX were released from the liposomes and ANTS emitted fluorescence. Therefore, the leakage was measured directly by determining the relative change in fluorescence. The liposomes initially containing both ANTS (12.5 mM) and DPX (45 mM) emitted some fluorescence, which was set as the baseline, and the fluorescence from liposomes lysed with Triton X-100 was set as an indicator of 100% leakage. Fluorescence, excited at 360 nm and emitted at 535 nm, was measured with a Jasco J-777 spectrofluorometer (Jasco, Tokyo, Japan) (19).

CD measurement. Circular dichroism (CD) spectra were recorded on a Jasco J-715 spectropolarimeter (Jasco), using a quartz cell path length of 1 mm, at wavelengths ranging from 190 to 245 nm. A sample solution was prepared by mixing 50% (wt/vol) trifluoroethanol (TFE) and 20 mM sodium phosphate buffer at pH 7.4 or by the addition of 25 mM sodium dodecyl sulfate in 10 mM sodium phosphate buffer at pH 7.4. The peptide concentration was determined on the basis of amino acid analysis. The CD spectrum was recorded at room temperature and was obtained with a 0.5-nm bandwidth and a scan speed of 10 nm/min. Two scans were averaged to improve the signal-to-noise ratio. The CD data were analyzed for the percentage of the secondary structure of peptides by the method of Chen et al. (14).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Design of combinatorial peptide libraries composed of simplified amino acid sequences. Seven amino acids (Lys, Leu, Val, Pro, Ser, Phe, and Gln) instead of the 20 natural amino acids were selected for use in the synthesis of peptide mixtures by consideration of the characteristics of amino acids (for a review, see reference 34). Lys, which is one of the positively charged amino acids (Lys, Arg, and His), was chosen because of its high pK_a value and the easy deprotection of its side protection group in Fmoc chemistry. Leu and Val, which are in the class of aliphatic amino acids, were selected because of their secondary structure-forming propensities. Phe, which is in the class of aromatic amino acids (Phe, Tyr, and Trp), was selected because of the efficiency of its synthesis and deprotection. Ser was chosen for the hydroxyl amino acid, and Gln was chosen for the hydrophilic neutral amino acid. Pro, which is known as an  helix breaker, was also included. Cys was excluded because of its dimerization by disulfide bridge. Negatively charged amino acids (Glu, Asp) were excluded because the negative charge of the peptides interfere with the charge-charge interactions between the peptide and the negatively charged membrane. In our previous work, replacement of the amino acids of the antimicrobial decapeptide derived from tenecin 1 indicated that the presence of positively charged amino acid at the N terminus or the C terminus was critical for antifungal activity (26). Therefore, Lys was fixed at the C terminus for the efficient synthesis of the libraries.

Screening of the active peptide against C. albicans. To identify the most active amino acid sequence against C. albicans, each peptide mixture made up of nine libraries was prepared and assayed. As shown in Fig. 1, the most active peptide sequence had Lys at O₁ and O₂, which confirmed our previous result that a positively charged amino acid at the N or C terminus is critical for antifungal activity. The presence of Val at O₃ and O₄ resulted in the highest level of activity. The selectivity of the amino acid at O₅ was not great. When Val or Phe was located at O₅, the peptides had the same MICs; however, the 50% inhibitory concentration was lowest when the Phe was located at O₅ (data not shown). The presence of Lys at O₆ gave the highest level of antifungal activity. The MICs were the same when any aliphatic amino acid was located at O₇. However, Val was selected as the most active sequence at position O₇ when the 50% inhibitory concentration of peptides with various amino acids at O₇ were compared (data not shown). After complete screening, the most active decapeptide, KKVVFKVKFK-NH₂, named KSL, tested in this investigation was identified. The search for amino acid sequences similar to that of KSL by the BLAST program (http://www.ncbi.nih.gov/) showed that no similar amino acid sequence has been registered. As shown in Fig. 2, KSL did not show the perfect amphipathic -helical structure in a wheel diagram and had valine residues in the hydrophobic face.

View larger version (30K): [in this window] [in a new window]   FIG. 1.   Screening of libraries for activity that inhibits the growth of C. albicans. Each bar represents the MIC of the peptide mixture with a peptide defined at the O position with one of the seven amino acids used in the library; a, the initial position is defined (OXXXXXXXXK-NH2); b, the second position is defined (KOXXXXXXXK-NH2); c, the third position is defined (KKOXXXXXXK-NH2); d, the fourth position is defined (KKVOXXXXXK-NH2); e, the fifth position is defined (KKVVOXXXXK-NH2); f, the sixth position is defined (KKVVFOXXXK-NH2); g, the seventh position is defined (KKVVFKOXXK-NH2); h, the eighth position is defined (KKVVFKVOXK-NH2); i, the ninth position is defined (KKVVFKVKOK-NH2).

View larger version (22K): [in this window] [in a new window]   FIG. 2.   Helical wheel diagram of KSL.

Antimicrobial and hemolytic activity of KSL. As shown in Table 1, the peptide (KSL) with the most active sequence was synthesized, and its antibacterial and antifungal activities were measured. Although the size of KSL was much smaller than that of magainin II (47), the MIC of KSL for C. albicans was almost seven times lower than that of magainin II. Also, the antibacterial activity of KSL was more potent than that of magainin II against all test microorganisms, especially methicillin-resistant Staphylococcus aureus (MRSA). The MFC of KSL for C. albicans was 0.78 µg/ml, which indicated that KSL irreversibly inhibited the growth of C. albicans at the same concentration as the MIC (0.78 µg/ml). To lend credence to the activity in this assay, amphotericin B (15), fluconazole (22), and the antifungal hexapeptide identified by full peptide libraries (28) were assayed simultaneously and their activities were compared. Amphotericin B, which is known to be the most effective antifungal agent, had the same MFC and MIC (0.2 µg/ml) in this assay system. Fluconazole as a fungistatic agent did not show a clear cutoff value up to 50 µg/ml. The antifungal hexapeptide (28) had an MFC approximately 15 times higher (MFC, 12.5 µg/ml) than that of KSL (MFC, 0.78 µg/ml). To check the cytotoxicity against mammalian cells, KSL was added to mouse erythrocytes, and the level of hemolysis was measured. Figure 3 indicates the level of lysis of the mouse erythrocytes as a function of the concentrations of KSL, melittin, and amphotericin B. Amphotericin B and melittin (16) caused 100% lysis at concentrations greater than 10 µg/ml, while KSL did not show hemolytic activity at concentrations of up to 500 µg/ml. The concentration causing 50% hemolysis for KSL was approximately 600 times higher than the MIC for C. albicans (data not shown). This result indicates that KSL has a high degree of selectivity for fungi rather than mammalian cells.

View larger version (15K): [in this window] [in a new window]   FIG. 3.   Hemolytic activity of KSL. Erythrocytes were incubated in Tris buffer (150 mM KCl, 5 mM Tris-HCl [pH 7.4]) with various concentration of KSL for 1 h at 37°C. , KSL; , melittin; , amphotericin B.

Antimicrobial activities of individual peptides identified by combinatorial libraries. To confirm the results obtained with the combinatorial libraries, individual peptides were synthesized and assayed against bacteria as well as fungi. The sequences and characteristics of individual peptides are summarized in Table 2. Because the positive charge of the peptide was a critical factor for antimicrobial activity, hydrophobic amino acids were replaced only by other hydrophobic amino acids. All peptides had the same net positive charge, similar hydrophobicities, and similar hydrophobic moments. As shown in Table 3, KSL1 with valine at position 5 and KSL2 with phenylalanine at position 7 had the same MICs as KSL for C. albicans. The same MICs of KSL1 and KSL2 confirmed the results obtained with libraries in which the selectivity of hydrophobic amino acids at positions 5 and 7 for antifungal activity was not great. KSL3, KSL4, and KSL5, which were expected to be less potent than KSL on the basis of the results obtained with the libraries, in fact showed low levels of antifungal activity. In particular, KSL5, which contained Pro, which is known to be an  helix breaker, did not show activity at concentrations of up to 100 µg/ml. The activity of the individual peptide confirmed that KSL identified by the combinatorial library was the most active peptide against C. albicans. Interestingly, KSL2 showed the same activity as KSL against C. albicans, while it showed activity much better than that of KSL against MRSA and Mycobacterium smegmatis. This result indicated that the peptide identified by the libraries to be the most active against C. albicans could be different from that identified to be the most active against the other pathogens.

View this table: [in this window] [in a new window]   TABLE 2.   Sequences,  helicities, hydrophobicities, and hydrophobic moments of KSL and its analogsa

Assay of leakage of dyes from liposomes caused by KSL. To study the mechanisms of antifungal and antibacterial activity, we prepared liposomes with the phospholipid compositions of fungal and gram-positive bacterial membranes and measured the release of the dye from the liposomes induced by the peptide. Amphotericin B, which is known to be a membrane-active molecule (15), and fluconazole (22) were used as positive and negative controls, respectively. As shown in Fig. 4A, the increase in the concentration of KSL resulted in an increase in the level of leakage of dye from the liposomes whose compositions resembled that of the fungal membrane. The membrane-disrupting ability of KSL was much lower than that of amphotericin B, which was consistent with the relative potencies of the peptide and amphotericin B against C. albicans. Fluconazole, which is known to act on the enzyme in the cytoplasm as a primary target, did not cause leakage when it was used up to a concentration comparable to that of KSL. Figure 4B indicates the level of release of the dye from the liposome whose composition mimicked that of the membrane of gram-positive bacteria as a function of the concentration of KSL. In this experiment, magainin II, which is a membrane-active peptide (47), was used as a positive control. KSL and magainin II had similar leakage patterns and potencies as a function of the concentration. Thus, we suggest that the antifungal and antibacterial actions of the peptide are due to its interaction with and perturbation of the fungal and bacterial membranes. The hemolytic activities of KSL and melittin were measured by determining the level of release of dye from liposomes whose compositions resembled that of the membrane of human erythrocytes. As shown in Fig. 5, no significant leakage was observed that of with KSL at up to 25 µg/ml, while 3 µg of melittin, which is a cytotoxic peptide (16), per ml caused 100% leakage under the same assay conditions. From this study, we confirmed that KSL was highly selective between the fungal and erythrocyte membranes, which would prove to be a great advantage for pharmaceutical agents.

View larger version (15K): [in this window] [in a new window]   FIG. 4.   Leakage of the dye from liposomes. (A) The phospholipid composition of the liposomes resembled that of the fungal membrane (molar ratio of phosphatidycholine:phosphatidylethanolamine:phosphatidylserine:phophatidylinositol:cerebroside:cholesterol, 30:30:10:10:5:20). Antifungal compounds were incubated with liposomes for 1 h at 37°C. , amphotericin B; , KSL; , fluconazole. (B) The phospholipid compositions of the liposomes resembled that of the membrane of gram-positive bacteria (molar ratio of phosphatidyglycerol:cardiolipin, 3:1). The peptides were incubated with liposomes for 1 h at 37°C. , magainin II; , KSL.

View larger version (12K): [in this window] [in a new window]   FIG. 5.   Hemolytic activities measured by the leakage of the dye from the liposome by peptide. The phospholipid compositions of the liposomes resembled that of the human erythrocyte membrane (molar ratio of phosphatidylcholine:phosphatidylethanolamine:phosphatidylserine:sphingomyelin:cholesterol, 25:22:10:18:25). The peptides were incubated with the liposomes for 1 h at 37°C. , melittin; , KSL.

CD measurements. To investigate whether the antimicrobial activities of KSL and its analogs were correlated with the secondary structure, CD spectra were measured in various environments. According to the CD spectra, all peptides formed random coil structures in the phosphate buffer; however, all peptides except KSL5 formed a well-defined -helical structure in the presence of TFE (Table 2). KSL5, which had a proline in the middle of the amino acid sequence, had a random coil structure in the presence of 50% TFE. The helicity indicated that the activity of each peptide was not correlated with the -helical content. For example, the less active peptide KSL3 had 70%  helicity, while the more active peptides KSL and KSL1 had 56 and 42%  helicities, respectively.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

Antimicrobial peptides act on the lipid membrane of the pathogen, which has less specificity than other biological targets such as enzymes, antibodies, and hormone receptors. Therefore, we have designed novel peptide libraries consisting of simplified amino acid sequences to identify peptides with antifungal activity. Considering the relatively low level of specificity of the lipid membrane for the peptide, our peptide libraries can supply enough peptide mixtures with a variety of structures, hydrophobicities, and net charges. The results of recent studies in which the replacement of 20 natural amino acids in proteins by simplified amino acids did not strongly affect the structure and function support our suggestion (25, 40, 42).

Use of our novel peptide library technique has several advantages over use of the full peptide libraries and the traditional structure-activity relationship design method. The novel library developed in this study can identify the active amino sequence with less resin, labor, and cost than are required when full peptide libraries are used. This technique also overcomes the limitations of traditional structure-activity relationship design methods, such as the time-consuming optimization process, the nonlinear relationship between structure and activity, and a lack of information on the structure of the peptide in the lipid membrane. Furthermore, the results obtained in studies with these peptide libraries can provide useful information about the structure-activity relationship because the secondary structures of decapeptides can easily be characterized by CD spectroscopy.

The characterization of KSL, a peptide whose activity was optimized, provides useful information about the factor in the membrane-active peptide responsible antimicrobial activity. KSL has a unique amino acid sequence; five lysine residues (net positive charge, +6) at the hydrophilic face and three valine residues at the hydrophobic face in the -helical wheel diagram. Also, as shown in Fig. 2, KSL did not show the perfect amphipathic structure in the -helical wheel diagram, and the mean hydrophobic moment was calculated to be 0.22. This is different from the active peptides derived from structure-activity relationship studies (5, 13, 34). However, KSL showed more potent antifungal activity than the peptides that satisfied the perfect amphipathic structure criteria and contained a leucine residue in the hydrophobic face in the -helical wheel diagram. This result indicates the risk of using the traditional structure-activity relationship design method, in which more active peptides are designed by enhancing the  helicity only on the basis of the -helical wheel diagram.

Many structure-activity studies with antibacterial peptides indicate that an amphipathic structure and a net positive charge are fundamental factors for the activity (41-45). On the basis of the result obtained with the libraries, individual peptides were characterized to obtain an understanding of the structure-activity relationship. All individual peptides have a same net positive charge; however, activity is not correlated with the -helical content of the peptides calculated from the CD spectra measured in the presence of 50% TFE. An inactive analog with Pro in the middle of the amino acid sequence has a random conformation in the presence of 50% TFE, which is consistent with the general structure-activity relationship. Even though KSL adopts the -helical conformation in the presence of TFE, the most active peptide has Val residues instead of Leu residues in the hydrophobic face. This is a quite surprising result because Val has a low propensity for  helix formation, while Leu is one of the best -helix-forming residues. Also, KSL3 with Leu instead of Val in the hydrophobic face has a higher -helical content in the presence of TFE but has a lower level of activity than KSL. To obtain a more detailed structure, CD spectra of KSL and KSL3 were measured in the presence of 25 mM sodium dodecyl sulfate. These spectra reveal that both active peptides adopt a distorted -helical conformation and that the less active peptide KSL3 shows a decrease in the distorted -helical content (data not shown). There are two possible explanations for this; first, the real active secondary structure of KSL must be the distorted  helix because sodium dodecyl sulfate, with an anionic functional group in the end of an aliphatic lipid, rather than TFE must mimic the real biological membrane system. The other possible explanation is that there is some threshold of the secondary structure for antimicrobial activity. If the  helicity of the peptide is over this threshold, the structure is no longer a major factor in the activity. The former explanation seems to be more reasonable because the charge-charge interactions between the positively charged peptides with the negatively charged membrane is important for the binding and adoption of the structure of the positively charged peptide.

Interestingly, the peptides identified to have activity against C. albicans have high levels of activity against gram-positive and gram-negative bacteria. The different ratio and composition of the phospholipids between bacteria and fungi can explain this result. Bacterial membranes consisting of phosphatidylglycerol and cardiolipin have a more negative charge than fungal membranes, so positively charged peptides interact with the more negatively charged membranes of bacteria and enhance the permeability of the lipid membrane. This result supports the fact that regardless of the species of pathogens, the charge-charge interaction between positively charged antimicrobial peptides and negatively charged membranes is the common factor fundamental for the activities of the peptides. However, KSL and KSL2, which had the same activities against C. albicans, had different activities against MRSA and M. smegmatis, which strongly suggests that the lipid membrane of each microorganism has a sufficient specificity for differentiating the peptides consisting of simplified amino acid sequences. This fact indicates that our libraries can be applied to the screening of the membrane-active peptide against special pathogens such as MRSA.

The low level of hemolytic activity of the host defense peptides has been explained by several factors such as size, structure, and hydrophobicity. Magainin, cecropin, and other host defense peptides containing one or more amino acids known as -helix breakers have the potential to adopt a less perfect -helical structure. This imperfect amphipathic structure is regarded as a key factor for the differentiation of host cells from pathogenic cells (8). However, the low level of hemolytic activity of KSL without an -helix breaker supports the fact that the low level of hemolytic activity of this peptide must be due to its small size or its low level of hydrophobicity. The low level of hydrophobicity seems to be a key factor for the differentiation of mammalian cells from fungal cells because high-level hydrophobic interactions were reported to be necessary for the lysis of the erythrocyte membrane (35). Also, a very hydrophobic peptide, indolicidin, consisting of 13 amino acid residues, was reported to have hemolytic activity as well as antimicrobial activity (44).

Even though many host defense peptides have been isolated and their functions have been studied (10-13), most of them show antibacterial activity rather than antifungal activity. Some short host defense peptides such as indolicidin (1) and tachyplesin II (37) show antifungal activity as well as antibacterial activity. However, it is difficult to develop these native peptides into antifungal agents because of their cytotoxicity for the erythrocyte (13). The activity-optimized peptide identified in this study irreversibly inhibits the growth of C. albicans through its action on the lipid membrane and has a very potent and a broad range of activity against microorganisms but has no hemolytic activity. Considering the recent emergence of bacterial and fungal infections and resistant strains, it is possible that this peptide can be developed as a novel antimicrobial agent. Also, the characterization of KSL and its analogs will provide unique information about their structure-activity relationships.