I. Introduction

Top Previous Next References

Human subtlety will never devise an invention more beautiful, more simple, or more direct than does Naturebecause in her inventions, nothing is lackingand nothing is superfluous...
Leonardo da Vinci

Antimicrobial peptides represent ancient host defense effector molecules present in organisms across the evolutionary spectrum. Fundamental differences exist between microbial and mammalian cells that may represent targets for antimicrobial peptides. Among these, significant distinctions include membrane composition and architecture, energetics such as transmembrane potential and polarization, and structural features including sterols, lipopolysaccharide and peptidoglycan. Disparities such as these appear to translate to varying degrees of selective toxicity among distinct antimicrobial peptides, relating to peptide and target cell properties, as well as the biological settings in which the two interact.

Although hundreds of antimicrobial peptides have now been characterized as having widely diverse sequences, these peptides have been classified into relatively few conformational paradigms. Therefore, it may be argued that a high degree of degeneracy exists within the conformation code governing structure-activity relationships among antimicrobial peptides. Many of these molecules, within and beyond conformational classes, exhibit mechanisms of action that are highly complex and non-identical. Moreover, new evidence points to targets that lie interior to the cytoplasmic membrane as being important in antimicrobial mechanisms of these peptides. Thus, the assumption that antimicrobial peptides are uniform and indiscriminant membrane detergents is obsolete. Recognition of the sophisticated and thematic structure-activity relationships underlying distinct mechanisms of action among antimicrobial peptides will facilitate a more complete appreciation of their likely multiple roles in antimicrobial host defense.

Antimicrobial peptides have evolved as integral components of strategic and carefully regulated mechanisms of immunity to infection. However, microbial pathogens have not been passive to this evolutionary procession. Rather, prokaryotic and eukaryotic pathogens devote a considerable portion of their genomes to expressing complex and coordinately regulated countermeasures designed to subvert antimicrobial peptide targeting and mechanisms of action.

A clearer understanding of these parallel systems will advance two important, yet elusive goals. First, an awareness of the mechanisms employed by antimicrobial peptides will significantly improve our understanding of how these molecules act to defend against infection. Second, insights into these strategies will facilitate new opportunities and approaches to discover and develop pharmacologic agents that enhance or optimize immune mechanisms and suppress the ability of pathogens to subvert these mechanisms.

	II. Mechanisms of Antimicrobial Peptide Target Specificity and Selective Toxicity

Top Previous Next References

Polypeptides that exert antimicrobial activity have been isolated from essentially every tissue in which they have been sought. This intriguing observation has contributed to divergent interpretations regarding the potential functions of many of these peptides in antimicrobial host defense: peptides that may have little or no relevance in antimicrobial host defense can be demonstrated to inhibit or kill microorganisms in defined or austere conditions in vitroversuscomplementary peptides of varying structures, tissue sources, antimicrobial mechanisms, potencies, and/or spectra function in consort to provide optimal host defense against infection.

A pivotal consideration in this regard is the degree to which an antimicrobial peptide distinguishes between microbial and host cells in settings of potential toxicity. Evidence continues to mount in support of the concept that inherent structures or functions of microbial versus host cells contribute to selective antimicrobial discretion of some peptides. Alternatively, antimicrobial peptide access to potentially vulnerable host tissues may be limited by localization and/or or highly regulated expression. The following discussion highlights these themes as supported by recent studies.

1. Membrane Composition, Hydrophobicity, and Charge. The elementary component of essentially all biomembranes is the phospholipid bilayer. By definition, such bilayers are amphipathic, having both hydrophobic and hydrophilic domains. However, based on composition and influenced by cell energetics, biomembranes of prokaryotic versus eukaryotic cells differ significantly. For example, phosphatidylcholine (PC¹) and phosphatidylethanolamine (PE) normally have no net charge. Moreover, sphingomyelin (SM), a close analog of PC containing a palmitoyl residue, is also neutrally charged. In many membrane systems, the amounts of PC and SM are inversely related. Sterols such as cholesterol and ergesterol, found in eukaryotic but rarely in prokaryotic membranes, are also generally neutral. In contrast, hydroxylated phospholipids phosphatidylglycerol (PG), cardiolipin (CL; effectively a dimer of PG), and phosphatidylserine (PS), sustain a net negative charge. From these perspectives, it follows that the net charge of a biomembrane is based largely upon its phospholipid stoichiometry and architecture (Fig. 1). Cell membranes composed predominantly of PG, CL, or PS tend to be highly electronegative; such compositions are found in many bacterial pathogens. On the contrary, bilayers enriched in the zwitterionic phospholipids PE, PC, or SPcommonly found in mammalian cytoplasmic membranesare generally neutral in net charge. These characteristic membrane charge properties may also be compounded by differences in electrochemical gradients of prokaryotic versus eukaryotic cells (see below). Sterols within membranes may further differentiate mammalian and fungal cells from prokaryotes (Tytler et al., 1995; see below) as potential targets for antimicrobial peptides. Moreover, it is intriguing to note that peptides with primarily antifungal activity, such as many of those isolated from plants, tend to be relatively rich in polar neutral amino acids, suggesting a unique structure-activity relationship (Hancock and Chapple, 1999).

View larger version (30K): [in this window] [in a new window]   Fig. 1.   Comparative architecture of microbial and human cytoplasmic membranes. Cytoplasmic membranes of bacterial (E. coli, S. aureus, or B. subtilis) and fungal (C. albicans) pathogens are compared with that of the human erythrocyte in relative composition and distribution between inner and outer membrane leaflets. Membrane constituents ranging from anionic (left) to zwitterionic or neutral (right) are CL, PG, PE, PC, SM, and sterols (cholesterol or ergesterol, ST). Note the marked differences among microbial pathogens and human erythrocytes in phospholipid composition and asymmetry. These differences are believed to account for preferential antimicrobial peptide affinity for microbial versus host cells to the extent it exists for a given antimicrobial peptide. Key: open, E. coli; horizontal hatching, S. aureus; shaded, B. subtilis; checkered, C. albicans, solid, human erythrocyte.

2. Membrane Asymmetry. The compositional and architectural characteristics of prokaryotic and mammalian membranes are neither static nor symmetric. Distinctions between microbial and mammalian cells as targets for antimicrobial peptides include the configuration of phospholipid bilayer components. Recent evidence indicates that the distribution of phospholipids within cytoplasmic membranes is highly asymmetric. For example, only 2% of the total PE content in bovine erythrocytes is oriented toward the outer membrane leaflet (Florin-Christensen et al., 2001). Differences among asymmetric distribution, compositional stoichiometry, and saturation of phospholipid bilayers also significantly influence membrane phase transition and fluidity (Bayer et al., 2000; Verkleij and Post, 2000; McIntosh et al., 2001). These differences may extend to the inner and outer cytoplasmic membrane leaflets, or those of the outer membrane of Gram-negative bacteria or enveloped viral pathogens (which generally exhibit properties of their corresponding host cells). Accordingly, the charge and amphipathicity of the inner and outer membrane leaflets also vary considerably (Fig. 1). For example, in human erythrocytes, most glycosylated lipids (glycolipids), PC, PS, and SM are positioned on the exoplasmic membrane leaflet. Alternatively, when present, neutral or anionic phospholipids are typically localized on the cytoplasmic leaflet. Thus, differences in electronegativity resulting from leaflet asymmetry likely provide a further dimension influencing the relative affinity of antimicrobial peptides for biomembranes.

3. Microbial Ligands for Antimicrobial Peptides. The fact that D-and L-amino acid versions of antimicrobial peptides generally show little selectivity in binding suggests that stereospecific receptors are not present on target microbial cells (see Section III.). However, certain structures may be crucial for selective affinity of peptides for microbial pathogens. For example, Teuber and Bader (1976) demonstrated that radioactive mono-N-acetyl-¹⁴C or native polymyxin B absorbed to isolated cytoplasmic and outer membranes of Salmonella typhimurium within 60 s of exposure. Moreover, polymyxin B exhibited sigmoidal binding kinetics, suggesting saturation of cytoplasmic and outer membranes, with approximately 30 and 60 nmol of peptide bound per milligram of membrane, respectively. Importantly, based on the stoichiometry of LPS, PG, CL, and PE in the membranes, these investigators calculated that the theoretical binding capacities of polymyxin B were almost identical to the binding properties if LPS, PG, and CL were modeled to function as specific receptors for this peptide. This robust concordance between theoretical and experimental approximations of polymyxin B binding capacities, along with parallel binding and killing kinetics, argues that membrane anionic constituents themselves function as pseudoreceptors for this cationic peptide. Thus, electronegative ligands (e.g., PG, CL, LPS) likely provide impetus for the initial interaction between cationic peptides and certain pathogens (also see Sections III. and IV.).

4. Transmembrane Potential. Another fundamental difference between microbial and mammalian cells can be found in the charge separation between the extracellular and intracellular aspects of the cytoplasmic membrane. This electrochemical gradient, resulting from differing extents and rates of proton flux across the membrane, is termed the transmembrane potential (). The difference in between certain microorganisms and host cells may provide a means of selective targeting of microorganisms by cationic antimicrobial peptides. For example, normal mammalian cells exhibit a ranging from 90 to 110 mV. However, bacterial pathogens in logarithmic phase growth commonly exhibit of 130 to 150 mV. Such significant differences in membrane electrochemistry have been hypothesized as additional parameters guiding selective toxicity of antimicrobial peptides, through a mechanism that has been termed self-promoted uptake (Hancock, 1997 [also see Section III.]).

Tam et al. (2000) recently examined the influence of conformation on membranolytic selectivity of antimicrobial peptides. In these studies, antimicrobial activity and human cell cytotoxicity were assessed in conformationally restricted cyclic and noncyclic analogs of protegrin-1, an 18-amino acid cationic peptide exhibiting broad-spectrum antimicrobial activity. Antimicrobial assays in relatively low- and high-salt conditions revealed cyclic protegrins exert differential antimicrobial profiles against Gram-positive and Gram-negative bacteria, fungi, and human immunodeficiency virus-1. As compared to protegrin-1, the most constrained analog (a cyclic-tricystine protegrin termed ccPG-3) displayed a 10-fold decrease in hemolytic propensity to human cells and up to a 30-fold increased membranolytic selectivity against specific target pathogens. However, an analogous cyclic protegrin lacking a disulfide bond, or a cyclic mimic of protegrin-1 with one disulfide bond, exhibited antimicrobial and cytotoxic profiles equivalent to protegrin-1. Interestingly, circular dichroism showed that even cyclized protegrins stabilized by disulfide bonds display -strand structure in water/trifluoroethanol or phosphate-buffered environments. These findings suggest that conformational dynamics subsequent to initial binding contribute to antimicrobial peptide activity in selected membrane environments.

Related studies by Unger et al. (2001) provide additional insights into the structural basis for selective toxicity of antimicrobial peptides. These investigators examined the interaction of linear versus cyclic counterparts of melittin and magainin analogs (peptides displaying non-identical selective toxicity toward mammalian cells) with membrane models in vitro. As compared with linear versions, the cyclized peptides were less efficient in initial binding to phospholipid membranes. However, at normalized bound concentrations, linear and cyclic analogs retained equivalent potencies to induce membrane permeabilization. When bound to phospholipid membranes, these cyclized peptides reverted to ~75% of the helical structure of their linear analogs. Even more importantly, the cyclic melittin analog exhibited increased antibacterial activity, with reduced hemolytic propensity, whereas the cyclic magainin exhibited opposite biological functions. These observations were interpreted to suggest that conformation influences initial interactions of peptides with membranes, as well as ensuing disruptive actions on target membranes. Collectively, these findings emphasize the potential role for conformational dynamics subsequent to initial binding interactions in selective toxicity of antimicrobial peptides. In addition, the above studies lend insights into the potential for engineered conformational constraints to further dissociate antimicrobial activity from host cytotoxicity.

Recent studies by Oren and colleagues (1999) also shed light on the relationship between quaternary structure and selective toxicity among antimicrobial peptides. Human cathelicidin LL-37 is an antimicrobial peptide cytotoxic to both bacterial and mammalian cells. This peptide exists in equilibrium as monomers and oligomers in solution at low concentration but appears to undergo self-association within zwitterionic (mammalian-like) and electronegative (bacterial-like) artificial phospholipid membranes in vitro. Interestingly, in these models, LL-37 effected a detergent-like or carpet mechanism (see Section III.) in disrupting both membrane types, suggesting a structure-induced membrane perturbation in either setting. Supportive of this interpretation was the finding that the peptide conformed to a predominantly -helical configuration oriented parallel with the surface of zwitterionic membranes. Thus, a propensity to assume an invariable helical conformation and multimerize within lipid membranes of differing compositions may reduce the ability of antimicrobial peptides to exert selective toxicity against microorganisms versus host cells.

Experiments focusing on cationic antimicrobial peptides of varying structures and origins extend this theme of peptide interaction with model membranes of distinct phospholipid compositions (Zhang et al., 2001). In these studies, test peptides were uniformly cationic but varied in conformation, including -helical, -sheet, extended, and cyclic motifs. Regardless of conformation, all test peptides interacted with and penetrated into lipid monolayers composed of anionic PG, as measured by the release of preloaded calcein dye. In comparison, only -helical and extended peptides interacted with monolayers composed of more zwitterionic PC, albeit to a lesser extent than with the anionic lipids. Interestingly, a -sheet peptide induced rapid phospholipid translocation (movement of lipid from the inner facet to the outer facet of the membrane) at concentrations less than required for membrane permeabilization. Similarly, Kol et al. (2001) demonstrated that the ability of peptides of comparable conformation to induce phospholipid translocation was greater for those containing proportionately more lysine or histidine residues, compared with tryptophan. From these examples, it appears that antimicrobial peptides not only interact with biomembranes of specific composition and asymmetry but may also promote remodeling of these membrane properties within target cells.

Recently, Welling et al. (2001) tested the hypothesis that cationic antimicrobial peptides may discriminate between microbial cells and host tissues in vivo. Studies evaluated whether such peptides specifically accumulate in sites of infection, compared with sterile inflammatory lesions, due to preferential avidity for microorganisms. Peptide affinity and specificity for pathogens in vivo was assessed by intravenous injection of ^99mTc-labeled synthetic derivatives of human ubiquicidin or lactoferrin into animals experimentally infected with Staphylococcus aureus, Klebsiella pneumoniae, or C. albicans. As controls, sterile inflammatory sites were induced by the introduction of heat-killed microorganisms or purified LPS into thigh muscle. Labeled human defensin, human polyclonal IgG, and ciprofloxacin were examined as comparative agents. The ^99mTc-labeled peptides and defensin accumulated at a significantly higher rate and to a greater extent in bacteria- and C. albicans-infected lesions in mice and rabbits, compared with non-infected but inflamed tissues. These data were interpreted to indicate that the peptides distinguish between microorganisms and host tissues, and in doing so, accumulate at sites of infection in vivo.

In related studies, this same group examined the potential pharmacologic utility of antimicrobial peptides to localize to sites of infection (Welling et al., 2000). Biodistribution scintigraphy suggested that the ^99mTc-labeled peptides were rapidly removed from the circulation by renal excretion. However, despite this rapid clearance, the radiolabeled peptides efficiently discriminated between infected and non-infected tissue, with up to 5-fold increased binding to target versus nontarget tissues within 1 h in rabbits. Collectively, these results indicate that antimicrobial peptides rapidly localize and accumulate at sites of infection, likely due to preferential affinity for peptides to associate with target microorganism surfaces rather than non-infected tissues.

A similar, albeit more complex mechanism likely contributes to selective toxicity of antimicrobial peptides found in granules of phagocytic leukocytes. The fundamental antimicrobial functions of professional phagocytes include internalization of pathogens (phagocytosis), subjecting them to the harsh microenvironment of the phagolysosome. Neutrophils, monocytes, and macrophages of various mammalian species contain among the most potent antimicrobial peptides knowndefensins (see below). However, defensins may also exhibit among the least selective toxicity of any host defense peptides, often exerting membrane permeabilizing and other harmful effects on microorganisms and mammalian cells alike. Phagocytes normally interiorize and expose pathogens to lethal concentrations of these peptides within the maturing phagolysosome, rather than degranulating these potentially injurious components into the extracellular milieu. Within the restricted confines of the phagolysosome, defensins and other antimicrobial peptides are present in very high relative concentrations, where they may act harshly and synergistically with one another, along with oxidative killing mechanisms. In this way, defensins may be constrained to granules of mammalian phagocytes to minimize their potential for host cytotoxicity. Moreover, Shafer et al. (1986) and Yeaman (1997) have suggested that antimicrobial activities of defensins and platelet microbicidal proteins are potentiated in mildly acidic conditions, such as those found in the maturing phagolysosome.

Beyond the scope of this review, some antimicrobial peptides may also perform other important functions contributing significantly to antimicrobial host defense, including interfering with host cell receptor access to pathogens, recruitment of leukocytes to sites of infection, as well as potentiate their antimicrobial activities (Yeaman, 1997; Yeaman and Bayer, 1999; Cole et al., 2001; Tang et al., 2002). For example, Zhang et al. (2002) have recently found that CD8+ cytotoxic T lymphocytes elaborate -defensins 1, 2, and 3 in contributing to host defense against human immunodeficiency virus-1. Conceivably, these peptides act directly to alter or damage the human immunodeficiency virus virion, or indirectly by interfering with receptor targeting, eventual uncoating or replication, and/or enhanced intracellular destruction. Thus, the extracellular secretion of antimicrobial peptides at concentrations or in settings that do not result in host toxicity may play important roles in immunity. Through such strategies, the antimicrobial functions of peptides and phagocytes may be mutually amplified, while minimizing the potential for concomitant host cell toxicity.

	III. Mechanisms of Antimicrobial Peptide Action

Top Previous Next References

A striking feature among antimicrobial peptides as a group is their overall conservation of structure and charge themes across diverse phyla. Whether synthesized non-ribosomally with D- and L-amino acids, or from genetically encoded messenger RNA, antimicrobial peptides form amphipathic structures and are often cationic at physiological pH. As outlined above, amphipathicity and net charge are characteristics understandably conserved among many antimicrobial peptides. Furthermore, charge affinity is likely an important means conferring selectivity to antimicrobial peptides. In the context of these paradigms, the following discussion highlights current concepts relating to the molecular basis of antimicrobial peptide mechanisms of action.

An essential requirement for any antimicrobial host defense or therapeutic agent is that it has a selective toxicity for the microbial target relative to the host. Ideally, such compounds have affinity for one or more microbial determinants that are easily accessible, common to a broad spectrum of microbes, and relatively immutable. Nature has apparently yielded a class of molecules that meets these constraints in the evolution of antimicrobial peptides. Antimicrobial peptides initially target microbial cells, and thus fulfill criteria outlined above for identifying molecular determinants of pathogens that are accessible and broadly conserved. As a group, antimicrobial peptides have amphipathic features that mirror phospholipids, thus allowing them to interact with and exploit vulnerabilities inherent in essential microbial structures such as cell membranes. In the following section, several aspects of antimicrobial peptide structure relevant to antimicrobial activity and selective toxicity are considered thematically. Specifically, structural parameters such as conformation (), charge (Q), hydrophobicity (H), hydrophobic moment (M_H), amphipathicity (A), and polar angle (), are examined in some detail. It is important to note that these molecular determinants are interdependent, and therefore, modification of one parameter often leads to compensatory alterations in others. This holistic view of peptide structure-activity relationship relates to each of these key properties influencing mechanisms of action of antimicrobial peptides (Fig. 2). The following discussion is considered in this context.

View larger version (32K): [in this window] [in a new window]   Fig. 2.   Interrelationship among structural determinants in antimicrobial peptides. Fundamental composition and amino acid sequence influences not only the biochemical properties of the peptide [e.g., charge (Q), amphipathicity (A), and hydrophobicity (H)], but also govern their three-dimensional configuration [e.g., conformation (), polar angle (), and overall stereo geometry]. Therefore, changes in composition, sequence, and intramolecular bonds may profoundly effect the structure-activity relationships of antimicrobial peptides in solution, upon binding to target membranes, or as they may undergo conformational phase transition to activated states. Moreover, these features may be specific for distinct peptides as they interact with specific pathogens or in specific physiologic microenvironments. Therefore, optimal antimicrobial peptide efficacy lies in the relevant coordination of these relationships (shaded area) as they relate to microbial target versus host cells in a particular context of infection.

1. Conformation (). Although antimicrobial peptides differ widely in sequence and source, several themes in their three-dimensional topology appear predominant, and peptides have been categorized accordingly. The two largest groups are the -helical and -sheet peptides, whereas the majority of remaining peptides can be classified as those that are enriched in one or more amino acid residues [e.g., proline-arginine or tryptophan-rich (Hancock, 1997)]. Other classification schemes are based upon peptide source (e.g., neutrophils or other leukocytes), precursor (e.g., cathelicidin, derivatives of cathelin), extent of intramolecular bonds (e.g., cysteine array or cyclization in peptides), or other parameters.

2. Charge (Q). Many of the antimicrobial peptides characterized to date display a net positive charge, ranging from +2 to +9, and may contain highly defined cationic domain(s). Cationicity is undoubtedly important for the initial electrostatic attraction of antimicrobial peptides to negatively charged phospholipid membranes of bacteria and other microorganisms (see Fig. 2), and mutual electroaffinity likely confers selective antimicrobial targeting relative to host tissues. The fact that bacterial membranes are rich in the acidic phospholipids PG, PS, and CL confers their overall negative charge. Moreover, LPS and teichoic or teichuronic acids of Gram-negative and Gram-positive bacteria, impart additional negative charge to the surfaces of these respective organisms. Target cell is typically up to 50% greater in prokaryotes than in most mammalian cells. Thus, it has been proposed that such a chemiosmotic potential may act in an electrophoretic manner to concentrate positively charged peptides on microbial surfaces (also see Section II.).

3. Amphipathicity (A) and Hydrophobic Moment (M_H). Nearly all antimicrobial peptides form amphipathic structures upon interaction with target membranes. Amphipathicity can be achieved via a multitude of protein conformations; however, one of the simplest and perhaps most elegant is the amphipathic helix. The amphipathic -helix has a periodicity of three to four residues and is optimal for interaction with amphipathic biomembranes. While the extent of amphipathic helicity influences peptide activity against negatively charged membranes, it may have an even more pronounced effect in rendering peptides hemolytic against zwitterionic or neutral membranes. Thus, a high degree of helicity and/or amphipathicity yielding a segregated hydrophobic domain, is correlated with increased toxicity toward cells composed of neutral phospholipids (Dathe and Wieprecht, 1999).

1

1

5. Polar Angle (). Polar angle is a measurement of the relative proportion of polar versus nonpolar facets of a peptide conformed to an amphipathic helix. For example, in a hypothetical -helical peptide, in which one facet is exclusively composed of hydrophobic residues and the other solely composed of charged residues, the polar angle would be 180^o. A reduced segregation between these domains or an increased hydrophobic proportion of the helix would proportionately reduce the polar angle. In numerous studies of native and synthetic peptides, a smaller polar angle (and therefore a greater hydrophobic surface) is associated with increased capacity to permeabilize membranes (Dathe et al., 1997; Wieprecht et al., 1997; Uematsu and Matsuzaki, 2000). The polar angle has also been shown to correlate with the overall stability and half-life of peptide-induced membrane pores. In a recent study by Uematsu Matsuzaki (2000), the effects of polar angle on membrane permeabilization and pore formation were compared. Two model peptides with polar angles of 100^o and 180^o showed functional similarities with native -helical antimicrobial peptides in forming amphipathic helices, selective targeting of negatively charged membranes, and creating toroid or lipid-containing pores (see below). Results from these studies indicated that peptides with smaller polar angles induced greater membrane permeabilization, translocation, and pore formation rates (Uematsu and Matsuzaki, 2000). However, although the rate of pore formation was greater for peptides with smaller polar angles, the rate of pore collapse was higher. These results suggest that peptides with smaller polar angles achieve less stable pore structures compared with peptides having larger polar angles. Greater stability of pores formed by the latter peptides could result from larger charged surfaces, and/or more peptide molecules per channel. These concepts are consistent with those observed in native peptides, showing that peptide PGLa ( = 100^o) is more easily translocated than magainin 2 ( = 180^o; Matsuzaki, 1998). These results indicate that hydrophobic and hydrophilic stereogeometries in antimicrobial peptides play significant roles influencing the process and consequences of membrane interaction and disruption.

The existence of a broad diversity in antimicrobial peptide sequences and structures underscores the reality that no single antimicrobial peptide sequence has emerged as singularly effective against all pathogens in all settings. Moreover, Nature may have sustained such diversity as a strategy to prevent or delay evolution of microbial resistance to antimicrobial peptides. Nonetheless, a circumspect analysis of structural parameters associated with differential antimicrobial activity versus host cell toxicity among peptides reveals several themes. Conservation in secondary structure may be key to three-dimensional configurations facilitating antimicrobial activity of distinct peptides. Generally, extremes of certain features, such as charge, amphipathicity, hydrophobic moment, or polar angle may disfavor peptide antimicrobial activity and selective toxicity. A minimum threshold of charge, perhaps as low as +2, appears necessary for antimicrobial peptide selectivity toward microorganisms. This property is likely important for a number of reasons: 1) initial electrostatic attraction to negatively charged microbial membranes; 2) potential to displace membrane-associated cations; and 3) a strong trans-negative of many microorganisms may facilitate cationic peptide transitions in orientation on the membrane, entry into the polar membrane core, and/or translocate peptides from exoplasmic to cytoplasmic membrane facets. A moderate level of amphipathicity, independent of or in context of polarization of charge, appears to be more favorable in these respects. Segregation of charge and hydrophobicity paralleling the inherent amphipathicity of the target lipid bilayer may also promote peptide integration into and disruption of the microbial membrane. A third theme is that selectivity among membrane-lytic peptides may rely on moderate degrees of hydrophobicity, as excessive hydrophobicity may increase selectivity for zwitterionic membranes, increasing mammalian cytotoxicity. Thus, selective antimicrobial activity results from a delicate balance among three-dimensional hydrophobic and electrostatic interactions between an antimicrobial peptide and its target (Dathe et al., 1996, 1997).

1. Electrostatic Interactions. There is widespread acceptance that the initial mechanism by which antimicrobial peptides target microbes occurs via an electrostatic interaction. For example, cationic antimicrobial peptides and negatively charged lipid membranes of bacteria provide for a mutual and vigorous attraction. This supposition has been borne out by numerous studies in which a strong correlation between peptide charge and membrane binding activity has been demonstrated (Bessalle et al., 1992; Vaz Gomes et al., 1993; Matsuzaki et al., 1997; Dathe et al., 2001). This view is also supported by the conservation of positive charge within many antimicrobial peptides isolated from organisms across the evolutionary spectrum. The facts that electrostatic forces are active over relatively long molecular distances and that lysine and arginine interactions with phosphate groups in lipid bilayers are particularly strong (Mavri and Vogel, 1996) likely contributes to the initial attraction and membrane-targeting step of many antimicrobial peptides.

2. Receptor-Mediated Membrane Interactions. Early studies using all D-enantiomers of native and model peptides demonstrated equivalent antimicrobial activities of D- and L-isoforms. Thus, the prevailing dogma supported a non-receptor type interaction for antimicrobial peptides with most pathogen membranes (Bessalle et al., 1990; Wade et al., 1990). Since then, several studies suggest there may be important exceptions to this generalization. Perhaps the most well characterized example is that of nisin, a small, cyclic, non-ribosomally produced peptide that has been used in the food industry for several decades. Nisin exhibits antimicrobial activity in the nanomolar range and specifically binds to bacterial lipid II, a membrane bound component involved in peptidoglycan synthesis. When exposed to nisin, vesicles containing lipid II exhibit an ~1000-fold increase in fluorescein leakage compared with vesicles lacking lipid II (Breukink and Kruijff, 1999). It has been proposed that this specificity in nisin activity relates to a specific receptor-like interaction with lipid II and the proximity it confers to this peptide relative to the microorganism. Notably, lipid II is believed integral to peptidoglycan synthesis, and nisin is considerably more active against peptidoglycan-rich Gram-positive organisms than Gram-negative organisms (Breukink and Kruijff, 1999). Likewise, Brotz et al. (1998) have recently demonstrated that the lantibiotic mersacidin interferes with transglycosylation and peptidoglycan synthesis in Gram-positive bacteria by direct targeting of lipid II. In addition, tachyplesin has been demonstrated to have a specific affinity for LPS (Hirakura et al., 2002). Moreover, a number of studies have now shown non-equivalent activities for native all-L peptides, versus their all-D enantiomers (Fehlbaum et al., 1996; Vunnam et al., 1997). For example, in intriguing studies using PR-39, a proline- and arginine-rich peptide of porcine origin, the all-D enantiomer showed 1000-fold differences in species-specific activity against bacterial organisms (Vunnam et al., 1997). These studies suggest receptor-type interactions may be important for some peptides in targeting specific epitopes on the microbial surface.

Perhaps one of the more controversial issues within the field revolves around the fate of antimicrobial peptides following their initial interaction with biological membranes. The mechanism(s) by which peptides may permeabilize and traverse microbial membranes are not entirely clear and likely vary for different peptides. Uncertainty stems in part from technical difficulties associated with first-principle determinations or molecular modeling of peptide-lipid interactions. Attempts to crystallize antimicrobial peptides within a native lipid environment have been largely unsuccessful, and other methods of structure determination have various limitations. Conventional CD is an excellent tool for determining peptide secondary structures, such as -helices. However it necessitates the use of optically clear solutions and provides little information as to the relative size of conformer regions or their location (Blondelle et al., 1999; Sitaram and Nagaraj, 1999). Similarly, infrared spectroscopy (e.g., FTIR) is an important tool particularly well suited to study -sheet peptide conformations, but also has technical limitations. Fluorescence spectroscopy is convenient, and its high level of sensitivity allows for a small sample size. However, data are often highly dependent on the solvent or membrane mimetic system used. Nuclear magnetic resonance (NMR) studies also offer a powerful means to obtain structural information at the single residue or domain levels but can be limited by relatively slow rates of molecular reorientation. Recently developed methods include reverse phase-high pressure liquid chromatography-based and surface plasmon resonance; however, these techniques are limited to the extent they represent protein-whole microorganism interactions. Therefore, at the present time, the most comprehensive assessments of peptide-lipid structure often come from a combinatorial approach wherein a variety of methodologies are employed, and the results are considered collectively.

The following discussion considers events subsequent to peptide-target binding that may significantly influence peptide mechanisms of action and/or selective toxicity. Functional themes are reviewed using examples of prevailing models for these processes, which are proposed to occur via specific and nonspecific mechanisms. These data should be interpreted in the context of the specific biophysical methods employed; the particular conditions and assays utilized for assessing peptide antimicrobial activities are beyond the scope of this review. However, it should be understood that the potencies, spectra, and/or mechanisms of antimicrobial peptide action could be highly dependent upon conditions of testing. For example, media pH, osmotic and ionic strength, temperature, and viscosity (e.g., in peptide diffusion assays)individually and in combinationmay significantly influence peptide antimicrobial activities.

1. Threshold Concentration. At some point following initial membrane binding, peptides enter a second stage of membrane interaction, frequently referred to as the threshold concentration. In this phase, peptides begin to enter and traverse the lipid bilayer via a number of possible mechanisms, ultimately extending their antimicrobial action to targets interior to the cell membrane. Conceptually, the threshold concentration necessary to drive such events results from accumulation of peptides on the target surface. Parameters that likely influence this threshold include peptide concentration, propensity to self-assemble or multimerize, as well as phospholipid membrane composition, fluidity, and head group size (Yang et al., 2000). Additionally, it is important to note that individual peptide-membrane interactions can vary such that one type of peptide may act via multiple mechanisms dependent on conformation dynamics of the peptide or target membrane remodeling.

2. Conformational Phase Transition. A key event occurring after membrane binding is the process of peptide structural or conformational phase transition, most well documented for -helical antimicrobial peptides. Numerous studies using various biophysical methodologies show that many antimicrobial peptides are disordered in aqueous environments, exhibiting extended or random coil conformations in this setting (Bello et al., 1982; Dathe and Wieprecht, 1999). However, many such peptides rapidly assume highly structured amphipathic -helical conformation upon interaction with phospholipid bilayers or in membrane mimetic solvents. Interestingly, a number of peptides require a negatively charged bilayer to undergo this transition. For example, the frog skin peptide PGLa, disordered when exposed to membranes composed of the zwitterionic PC and SM membranes, adopts a helical structure in the presence of membranes composed of PG and PE (Latal et al., 1997). Similarly, magainins only undergo a helical transition when interacting with anionic membranes as demonstated by CD, (Matsuzaki et al., 1989, 1991), vibrational/Raman-FTIR (Williams et al., 1990; Hirsh et al., 1996) and NMR (Bechinger et al., 1993; Hirsh et al., 1996). Examination of cecropin analogs revealed that the extent of -helical conformation is proportionately dependent on the amount of negatively charged phospholipid within the model membrane (Wang et al., 1998). One mechanism by which such a relationship may promote peptide order relies on the inherent phospholipid packing within the bilayer. For example, interactions of the peptide with the phospholipid head groups may promote an optimal periodicity within the charged residues of the peptide, promoting folding of the -helix. As discussed above, this change in conformation would also likely alter peptide hydrophobic moment and polar angle. Another potentially important aspect of the conformational phase transition is that it may prevent indiscriminant membranolytic activity until the peptide identifies an appropriate target surface. Thus, a lack of bioactive structure at nontarget sites may be an important means by which antimicrobial peptides minimize host-cell toxicity.

4. The Barrel-Stave Mechanism. The term barrel-stave describes the overall topology of a membrane channel formed in this mechanism of membrane permeabilization. In this model, a variable number of channel-forming peptides are positioned in a "barrel-like" ring around an aqueous pore. The "stave" term refers to individual transmembrane spokes within this barrel, which may be composed of individual peptides or peptide complexes. In this mechanism, the hydrophobic surfaces of -helical or -sheet peptides face outward, toward the acyl chains of the membrane, whereas the hydrophilic surfaces form the pore lining (Ehrenstein and Lecar, 1977; Breukink and Kruijff, 1999). The initial step in barrel-stave pore formation involves peptide binding at the membrane surface, most likely as monomers. Upon binding, the peptide may undergo a conformational phase transition, forcing polar-phospholipid head groups aside to induce localized membrane thinning. At this point, the hydrophobic portion of the peptide is inserted into the membrane to an extent corresponding to the hydrophobicity of the membrane outer leaflet. Positioning of the positively charged amino acids near the phospholipid head groups facilitates this process. When bound peptide reaches a threshold concentration, peptide monomers self-aggregate and insert deeper into the hydrophobic membrane core. Aggregation allows for a minimal exposure of the peptide hydrophilic residues to the hydrophobic membrane interior, as the peptides adopt a transmembrane configuration. Continued accretion of peptide monomers results in further expansion of the membrane pore. Upon phospholipid translocation or relaxation of the pore, peptides are transported to the inner membrane leaflet aspect due to the concentration gradient of surface-bound peptide, as well as trans-negative . An example of such a mechanism of action has been proposed for alamethicin (Sansom, 1991; Beven et al., 1999; Yang et al., 2001). Alamethicin-induced membrane conductance has been measured to proceed as a pattern of multistate conductance levels. This finding suggests the existence of pores with openings of various diameters, corresponding to channels composed of four or more transmembrane-spanning peptides. However, there remain relatively few peptides for which there is compelling evidence of a barrel-stave mechanism, despite this model having been proposed more than a decade ago. More recent studies often support the toroid pore model (see below; Yang et al., 2000, 2001). These newer data may reflect refinements in methodology and offer a clearer understanding of biophysical properties of transmembrane pores or channels that may incorporate lipid and peptide moieties.

5. The Toroid Pore or Wormhole Mechanism. One of the most well characterized peptide-membrane interactions is that of the toroid pore. A primary difference between the toroid pore and barrel-stave models is that in the former, lipids are intercalated with peptide in the transmembrane channel. Therefore, this structure has been referred to as a supramolecular complex and represents a membrane-spanning pore lined with polar peptide surfaces as well as phospholipid head groups. The toroid pore model has been deduced principally from experiments using -helical peptides, including magainins and PGLa. In this model, peptides in the extracellular environment take on an -helical structure as they interact with the charged and hydrophobic bacterial membrane. Helices are initially oriented parallel to the membrane surface as confirmed by NMR, fluorescence quenching, and CD (Hara et al., 2001). The hydrophobic residues of the bound peptides displace the polar head groups, creating a breach in the hydrophobic region and inducing positive curvature strain in the membrane (Hara et al., 2001). The introduction of strain and thinning further destabilizes the membrane surface integrity, making it more vulnerable to ensuing peptide interactions. At a threshold peptide-to-lipid ratio (e.g., estimated to be 1:30 for magainin), peptides orient perpendicular to the membrane. At this point,