Block access time

In general, mature AMP sequences consist of between 10 to 60 residues, in which cationic (i.e. lysine and arginine) and hydrophobic (e.g. leucine and isoleucine) amino acids are especially prominent (Zasloff, 2002; Brogden, 2005). While this leads to most AMPs having a positive charge of +2 or greater (Wang et al., 2009), the primary sequences of AMPs are varied, and do not contain conserved motifs, even between closely related species (Peschel & Sahl, 2006). As organisms produce more than one AMP species in certain tissues (Zasloff, 2002), it is thought that such variation is tolerated due to functional redundancy, thus leading to a greater evolutionary flexibility in the classical “arms race” with pathogens (Hancock

& Sahl, 2006). What seems to matter more than primary AMP sequence is overall secondary structure. Three main structural classes are apparent: "-helical, with an amphipathic nature (i.e. hydrophobic and cationic side chains segregated to distinct faces of the helix); !-sheet, with a similar amphipathic nature but with multiple disulphide bonds acting as stabilisers; and an extended conformation, lacking clear secondary structure in which certain residues are favoured (e.g. tryptophan [39%] in indolicidin, proline [49%] and arginine [26%] in PR-39) (Gallo & Huttner, 1998;

Brogden, 2005). Figure 3.1 gives examples of peptides from each of these classes.

In addition, cyclic peptides in a head-to-tail circular configuration, such as the

#-defensin RTD-1, have also been described (Tang et al., 1999).

Figure 3.1: Examples of various AMP structures. Common structural classes include A,

!-helical; B, "-sheet; and C, extended. Cationic residues are shown in blue, while anionic are shown in red. Selected peptide details (amino acids [aa], origin, Protein Data Base entry [PDB ID]) are as follows: magainin 2 (23 aa, African clawed frog, PDB ID 2MAG), LL-37 (37 aa, human, PDB ID 2K6O), lactoferricin (25 aa, bovine, PDB ID 1LFC), protegrin 1 (18 aa, porcine, PDB ID 1PG1), "-defensin-3 (45 aa, human, PDB ID 1KJ5), tritrpticin (13 aa, synthetic, PDB ID 1D6X), and indolicidin (13 aa, bovine, PDB ID 1G89). Figure modified from Nguyen et al. (2011), PDB accessible at http://www.pdb.org (Berman et al., 2000).

What unites many AMP molecules is their amphiphilicity when bound to lipid membranes. Positively charged and hydrophobic residues cluster in distinct regions of an AMP, be it on opposite sides of an "-helix or faces of a !-sheet (Toke, 2005).

Of particular relevance to this work is the structure of cathelicidins, a class of mammalian AMPs (12 to 39 aa) so named because of the conserved pre-pro regions that they share prior to the cleavage event that gives rise to the active peptide (Gennaro & Zanetti, 2000; Ramanathan et al., 2002). Expressed in a variety of epithelial and lymphatic tissues (Dorschner et al., 2001; Nizet et al., 2001), a subgroup of these peptides are linear and unstructured in solution, but circular dichroism and nuclear magnetic resonance spectroscopy studies (Yu et al., 2002;

Park et al., 2003) have revealed that they form an "-helix when associated with model phospholipid membranes (e.g. the human LL-37, Figure 3.1) (Toke, 2005).

3.1.2.2 Functions

Traditional antibiotics often have well-defined targets: for example, the !-lactams bind bacterial cell wall transpeptidases, while macrolides bind bacterial 50S ribosome subunits to inhibit protein synthesis (Devasahayam et al., 2010). AMPs, on the other hand, predominantly act on less specific targets, such as biological membranes (van't Hof et al., 2001). Such a lytic mode of action has been elucidated by several in vitro studies (Saiman et al., 2001; Park et al., 2001; Fantner et al., 2010), including examining the release of the contents of mimetic membranes and microbes themselves, as well as electron microscopy (Toke, 2005). The cytoplasmic membranes from both Gram-negative and Gram-positive bacteria contain anionic head groups in their phospholipid make up, resulting in an overall negative charge (Zasloff, 2002). In addition, Gram-positives possess a number of anionic teichoic and lipoteichoic acid moieties in their cell wall, while Gram-negatives exhibit anionic lipopolysaccharides (LPS) on their outer membrane (Hale

& Hancock, 2007; van't Hof et al., 2001). Both these features contribute to the electrostatic attraction of positively-charged AMPs, with LPS being especially important with regards to Gram-negatives. These negative moieties provide a passage to the inner cell membrane via a “self-mediated” uptake mechanism, in which the competitive displacement of LPS-associated divalent cations (e.g. Ca²⁺, Mg²⁺) allows outer membrane permeabilisation and subsequent AMP translocation to occur (Sawyer et al., 1988).

Once associated with a bacterial membrane, unstructured AMPs typically undergo a conformational change such as the coil-helix transition exhibited by LL-37 mentioned previously (Gennaro & Zanetti, 2000). For "-helical cathelicidins, this conformation is energetically favourable with regards to membrane insertion due to a reduced energy cost imparted by intramolecular NH-CO hydrogen bonding along the peptide backbone (Dathe & Wieprecht, 1999). The realignment of residues into an amphipathic configuration allows the hydrophobic face to insert into the membrane between the phospholipid head groups, while shielding the polar peptide

backbone from the lipid membrane interior (Oren & Shai, 1998; Brogden, 2005).

Cationic AMP insertion events are further facilitated by the internally negative membrane potential of a bacterium (Toke, 2005). When a critical threshold peptide concentration is reached (which may be near complete saturation of the membrane [Melo et al., 2009]), the integrity of the bilayer is compromised. Several mechanisms for this have been proposed, none of which are mutually exclusive (see Figure 3.2) (Hale & Hancock, 2007). Bacterial death results within minutes due to the loss of membrane polarisation and overall integrity (Brogden, 2005). In silico molecular dynamics simulations have been performed to try to further understand such mechanisms of action, and may give additional insight as modelling times progress from nanosecond bursts to microseconds (Bond & Khalid, 2010).

Figure 3.2: Cartoon of proposed mechanisms of AMP-mediated membrane destabilisation. Once a critical AMP threshold concentration has been reached, membrane lysis can occur via A, barrel-stave formation by multiple AMP molecules, where multiple hydrophilic faces associate to form a transmembrane pore; B, a “carpet”

mechanism akin to the effect of a detergent; C, toroidal pore formation, where AMPs insert between phospholipid head groups and promote membrane curvature much like the inside of a doughnut; or D, disordered toroidal pore formation (less structured with respect to a conventional toroidal pore). Figure modified from Melo et al. (2009).

In contrast to bacterial membranes, host membranes such as those of mammalian cells are more neutral in charge, containing zwitterionic outer leaflets that attract

and bind cationic AMPs poorly (Oren & Shai, 1998). In addition, their membranes are typically less fluid than bacterial membranes due to the “stiffening” properties conferred by cholesterol, further reducing the susceptibility to AMP-mediated lysis (Toke, 2005). It must be emphasised, however, that mammalian cells are not impervious to lytic activity – AMPs frequently exhibit undesirable haemolytic properties. For example, melittin, a 26 aa "-helical peptide from the honeybee, exhibits equal bactericidal and haemolytic activity at its minimum inhibitory concentration (MIC) due to possession of a hydrophobic N-terminal domain coupled with a cationic C-terminal domain (van't Hof et al., 2001). Mutational studies with different AMPs give conflicting evidence as to which property is most important (Shin et al., 2000; Travis et al., 2000; Yang et al., 2003; Hilpert et al., 2006; Chen et al., 2007; Jiang et al., 2008), but adjusting the charge, the proportion of hydrophobic/cationic residues, and the residue distribution over a peptide can abrogate haemolytic activity (Dathe & Wieprecht, 1999). Overall, haemolysis seems to be context dependent for each peptide, although high absolute hydrophobicity leads to greater haemolytic activity due to an increased affinity for zwitterionic membranes (van't Hof et al., 2001; Frecer et al., 2004; Toke, 2005; Matsuzaki, 2009). It appears that there is a subtle balance between amino acid composition, charge, lipophilicity, amphipathicity and structure.

AMPs have also been shown to have alternative (or concurrent) targets to the cell membrane (Brogden, 2005). Clues to this came from the observation that, when used at their MIC, some AMPs were able to kill a target without leading to obvious membrane disruption. For example, analogues of pleurocidin, a 25 aa "-helical peptide from the winter flounder, only lead to membrane depolarisation at concentrations ten-fold higher than its MIC (Patrzykat et al., 2002). It makes sense that intracellular-acting AMPs still exhibit lytic activity at higher concentrations, as they are required to cross the cell membrane to access potential targets. Because of their cationic properties, such targets are often anionic in nature (Hale &

Hancock, 2007). Examples of AMPs with intracellular targets include PR-39 (39 aa, porcine origin, rich in proline and arginine, extended structure), which has been shown to inhibit protein synthesis in E. coli (Boman et al., 1993); indolicidin (mentioned previously), which caused filamentation of E. coli cells by altering

cytoplasmic membrane septum formation (Subbalakshmi & Sitaram, 1998); and buforin II (21 aa, toad origin, "-helix structure), which is able to accumulate in the E. coli cytoplasm and bind DNA (Park et al., 1998).

Another alternative function of some AMPs is an ability to modulate a host’s immune response (e.g. LL-37; see Section 1.1.3.1) (Hancock & Sahl, 2006). As mentioned above, cationic AMPs bind LPS moieties during their self-mediated uptake in Gram-negative bacteria. Bacterial lysis results in liberation of these endotoxins, which provoke a strong systemic immune response in humans that may lead to septic shock (Zanetti, 2005). Among other AMPs, LL-37 has been shown to bind free LPS, and thus limit its ability to stimulate the production of pro-inflammatory cytokines such as tumour necrosis factor-" (Bowdish et al., 2005b).

Furthermore, the presence of serum seems to abrogate the membrane lytic activity of LL-37 (Gennaro & Zanetti, 2000; van't Hof et al., 2001). It has therefore been hypothesised that, although LL-37 shows good antimicrobial activity in vitro, its primary function in vivo may be as an immunomodulator (Bowdish et al., 2005b). In addition to endotoxin binding, LL-37 and other cationic AMPs (such as defensins) may also act as chemokines to attract macrophages; translocate into lymphocytes (presumably via their innate membrane penetration properties) to induce expression of anti-inflammatory genes; and promote wound healing in general (Bowdish et al., 2005a). Because of the above, AMPs that possess an immunomodulatory function in addition to antimicrobial activity have been re-christened “host defence peptides”

in order to reflect this increased functional workload (Bowdish et al., 2005a).

There are concerns that the use of AMPs in a clinical setting may lead to the eventual rise of resistance, which may have the undesirable consequence of also promoting resistance to endogenously produced AMPs (Perron et al., 2006).

However, because the primary mode of action of many AMPs is the general disruption of the cell membrane, it has been posited that resistance to such a broad mechanism of action is unlikely to arise in vivo, i.e. a fundamental membrane redesign would be required (Zasloff, 2002). Some specific resistance mechanisms do exist, however, such as the use of D-alanine amino groups by Staphylococcus aureus to reduce the negative charge of surface anionic teichoic acid molecules

(Peschel et al., 1999), and the use of outer membrane proteins as potential target decoys (e.g. adhesin A in the Gram-negative bacterium Yersinia enterocolitica [Visser et al., 1996]). The Gram-negative bacterium Burkholderia cepacia, an opportunistic pathogen involved in cystic fibrosis, is also extremely resistant to AMPs due to lipid A modifications (Mahenthiralingam et al., 2005). Other resistance strategies include protease production (particularly effective against linear AMPs), the use of capsule polysaccharides to prevent AMP access to the membrane, and increasing the rigidity of the membrane to make it more difficult for AMP insertion (Peschel & Sahl, 2006). Despite this, AMPs have remained effective antimicrobials over millions of years in many species, probably due to a multitude of targets and the production of several different peptides in a single organism (Zasloff, 2002).

Consistent with this, it takes many passages of a microbe at sub-lethal AMP concentrations to induce resistance (Gennaro & Zanetti, 2000; Perron et al., 2006):

cross-resistance to other AMPs is limited (Samuelsen et al., 2005), and animal models lacking certain endogenous AMPs have shown little increase in susceptibility to infection (Hancock & Sahl, 2006).

In summary, the large number of different activities that AMPs can possess has lead to the idea that AMPs may be “dirty drugs”, i.e. hit multiple targets with varying affinities, and such a multifaceted approach could lead to new microbial treatments (Peschel & Sahl, 2006).