Ion channel formation and membrane-linked pathologies of misfolded hydrophobic proteins: The role of dangerous unchaperoned molecules

Authors

  • Joseph I Kourie,

    1. Membrane Transport Group, Department of Chemistry, The Faculties, The Australian National University, Canberra, Australian Capital Territory, Australia
    Search for more papers by this author
  • Christine L Henry

    1. Membrane Transport Group, Department of Chemistry, The Faculties, The Australian National University, Canberra, Australian Capital Territory, Australia
    Search for more papers by this author

Dr Joseph I Kourie, Membrane Transport Group, Department of Chemistry, The Faculties, Building 33, Science Road, The Australian National University, Canberra, ACT 0200, Australia. Email: joseph.kourie@anu.edu.au

Summary

1. Protein–membrane interaction includes the interaction of proteins with intrinsic receptors and ion transport pathways and with membrane lipids. Several hypothetical interaction models have been reported for peptide-induced membrane destabilization, including hydrophobic clustering, electrostatic interaction, electrostatic followed by hydrophobic interaction, wedge × type incorporation and hydrophobic mismatch.

2. The present review focuses on the hypothesis of protein interaction with lipid membranes of those unchaperoned positively charged and misfolded proteins that have hydrophobic regions. We advance the hypothesis that protein misfolding that leads to the exposure of hydrophobic regions of proteins renders them potentially cytotoxic. Such proteins include prion, amyloid β protein (AβP), amylin, calcitonin, serum amyloid and C-type natriuretic peptides. These proteins have the ability to interact with lipid membranes, thereby inducing membrane damage and cell malfunction.

3. We propose that the most significant mechanism of membrane damage induced by hydrophobic misfolded proteins is mediated via the formation of ion channels. The hydrophobicity based toxicity of several proteins linked to neurodegenerative pathologies is similar to those observed for antibacterial toxins and viral proteins.

4. It is hypothesized that the membrane damage induced by amyloids, antibacterial toxins and viral proteins represents a common mechanism for cell malfunction, which underlies the associated pathologies and cytotoxicity of such proteins.

List of abbreviations:
AβP

Amyloid β protein

AD

Alzheimer's disease

ApoJ

Apolipoprotein J

CSF

Cerebrospinal fluid

DT

Diphtheria toxin

EqtII

Equinatoxin II

HA

Haemagglutinin

HIV

Human immunodeficiency virus

IFV2

Infectious flacherie virus

PC

Phosphatidylcholine

PrP[106–126]

Prion peptide fragment 106–126

sAβP

Secreted AβP

SDS

Sodium dodecyl sulphate

SI

Synapsin I

VacA

Vacuolating toxin of Helicobacter pylori

VP5*

Virion protein 5 fragment

VSV

Vesicular stomatitis virus

Introduction

It is increasingly recognized that membrane disturbance is a potent mechanism by which toxins, and also misfolded protein and peptide molecules, can induce cell membrane damage that leads to cell death.1–5 Several common pathologies are the result of a loss of membrane integrity, which can lead to imbalance of homeostasis and transmembrane electrochemical gradients and disruption of signalling systems.6–8 Many misfolded proteins that are produced during normal protein processing become capable of inducing cytotoxic effects via interaction with cytosolic membranes.9–12 Such cytotoxic protein molecules tend to contain significant numbers of exposed hydrophobic residues and are often classed as hydrophobic or amphipathic proteins.13 In addition to their hydrophobic nature, these peptides are often positively charged and this enables them to interact with negatively charged lipid membranes.13 Most cytotoxic peptides are positively charged as a result of lysine and arginine residues present in the sequence. The net charge of such molecules ranges from + 1 (e.g. RK-1) to + 8 (e.g. NK-lysin).14–16 Charge can vary with pH as a result of ionization of residues, such as histidines (e.g. the prion peptide fragment 106–126 (PrP [106–126])).17

It is hypothesized here that a common mechanism of action links misfolded native proteins, proteins secreted from pathological bacteria and from antimicrobial phagocytes and those present in venomous toxins, as well as fusion proteins present in viral envelopes. Cytotoxic peptides have been intensely studied because of their physiopathological significance, and for the relative simplicity of their structure–function relationships, due to their small size. Of particular interest are the autocytotoxic protein diseases of modified intracellular proteins that damage membranes of the producing cell. Much research has been done on these amyloidogenic proteins, including AβP, implicated in Alzheimer's disease (AD) and the modified scrapie prion protein conformer, PrPSc, which causes spongiform encephalopathies.18,19 These molecules are known to interact with cellular membranes and to damage their integrity. They form a range of ion channels in vitro.9–12,20–35 Our model agrees with those of others by explaining protein-induced membrane disturbance as involving both electrostatic and hydrophobic protein–membrane interactions. These interactions are the consequences of modifications in protein folding caused by genetic and/or cytosolic changes that lead to the exposure of the hydrophobic regions. These proteins exert their effects by: (i) inserting themselves into the membrane to form channels, as has been observed in vitro for AβP, PrP and many other hydrophobic and charged proteins;30,31 (ii) modifying intrinsic transport proteins, either directly or indirectly, by changing the membrane viscosity and lipid packing;6,7 and/or (iii) translocating themselves into the cell (e.g. viral proteins and toxins, such as α-sarcin, an antiribosomal protein).36–38 A broad characterization of the important structural and environmental factors influencing membrane interaction of cytotoxic peptides may reveal common mechanisms and motifs, allowing design of broad-based treatments and elucidating the general principles of protein–membrane interaction. The list of pathologies and diseases associated with hydrophobic protein-induced membrane damage is very long. More than 25 significant diseases have been proposed to be associated with amyloidogenic peptides alone.25

Malfunction of protein folding and exposure of hydrophobic regions

A central factor influencing the propensity of the peptide to interact with the lipid membrane is the presence, at the surface of the misfolded protein, of hydrophobic regions, which prefer the hydrophobic environment of the lipid membranes (Fig. 1). In addition, the presence of distinct positively charged regions in misfolded proteins enables better electrostatic interaction of these proteins with negatively charged biological membranes. Therefore, the mechanism of protein folding plays an important role in determining protein cytotoxicity. The correct folding is vital in conferring correct function and activity on proteins. Underlying autocytotoxicity involves changes in the intracellular environment that could induce alterations in protein folding and genetic mutations that could cause changes in the protein secondary structure and prevent proper folding. Protein processing in the cell is usually performed in the presence of chaperones and folding proteins, which act as guides for translocation of: (i) properly folded proteins to their correct sites; and (ii) misfolded proteins to degradation sites and/or internalized compartments (Fig. 1; see also Kopito and Sitia4 and Kourie and Henry5). Particularly in the case of autocytotoxic peptides (e.g. AβP and prions), which in their native form are not pathologic, it is hypothesized that a malfunction in the protein regulatory mechanism may be one of the factors responsible for cytotoxicity. The presence of protein refolding mechanisms for hiding the hydrophobic regions until they are needed may explain the key question of how peptides, such as antimicrobial molecules, retain their defence efficacy without harming the producing organism. Many proteins have been observed to undergo changes in secondary structure on interaction with membranes or due to other environmental factors. Both α-helix into β-sheet and β-sheet into α-helical transformations have been observed. In addition, membrane interaction may lead to an increase in secondary structure from random coil conformation.5,30

Figure 1.

Flow chart of the major steps involved in the mechanism of action of hydrophobic proteins and fragments (for detailed steps, see Kopito and Sitia,4 Kourie and Shorthouse13 and Kourie and Henry5).

Unfolding of the α-helix structure of a protein followed by refolding to β-sheet may be an important step in inducing a membrane-active protein structure for autocytotoxic peptides, such as prion and AβP. Normal cellular prion is stable and mostly composed of α-helical structure with very little β-structure. The form of prion responsible for neurodegenerative diseases contains an increased β-structure with a marked reduction in α-helix and an increase in β-structure.18,19 It is hypothesized here that this change in the structure could lead to changes in the prion translocation pathway to degradation sites and/or intracellular compartments and aggresomes.4,5 This results in the interaction of the strayed misfolded prion proteins with the endoplasmic reticulum and this is very likely followed by further interactions with the mitochondrial and nuclear membranes. Such a mechanism that involves exposure of the hydrophobic region and increased β-structure (two properties that are thought to enhance aggregation and are suitable for channel formation) explains prion's ability to destabilize and vacuolize cell membranes, as well as to form amyloid fibrils. A change in the intracellular pH is one of the important conditions that appears to play a role in protein misfolding. Morillas et al.39 show that native prion undergoes unfolding at low pH in the presence of urea, a denaturant. This unfolding is followed by concentration-dependent fibril formation and development of β-sheet secondary structure only in the presence of salts, such as NaCl. It is suggested that a partially folded monomeric intermediate may have β-sheet structure and be involved in the transition to oligomers; however, no monomeric β-sheet forms were observed. We hypothesize that such unfolding in the presence of lipid membranes may lead to insertion of monomers or oligomers into the membrane and that such a partial unfolding mechanism may be common to a wide range of peptides.

Conformational changes are induced in diphtheria toxin (DT) by the acidic endosomal environment on endocytosis. These changes lead to exposure of hydrophobic residues, thereby favouring membrane insertion and leading to membrane destabilization via pore formation.40,41 These studies indicate that endocytosis-induced structure and/or charge changes could represent a common step in hydrophobic protein activation. It is important to note that both β-sheet- and α-helix-based protein structural changes could confer cytotoxicity on refolded proteins (e.g. α-helix-based synapsin I,42 DT40,41 and natural α-sarcin37).

Considering the potential cytotoxicity of strayed hydrophobic proteins, the cell has developed mechanisms to neutralize such proteins. These mechanisms include degradation, compartmentalization, aggregation, plaques and fibrillization. The degradation of AβP, both in soluble form and as small fibrils, was examined in microglia by Chung et al.43 Fibrillar AβP is taken into the cells via receptor-mediated endocytosis at Scavenger Receptor A and degradation of some fraction of the peptide continues for 3 days. However, mostly undegraded aggregated AβP is released from the microglia during the next 9 days. Overall, only 15–20% is degraded; however, this is not due to cell saturation. It is hypothesized that the endosomes and lysosomes may provide an ideal acidic environment for continued aggregation and that microglia may actually promote fibrillization. An earlier study found that accumulation of AβP did take place within microglia, but found little release of the peptide.44 The authors of that study compared this engorgement to the process by which storage of cholesterol ester converts macrophages into foam cells. Soluble AβP is processed differently, being internalized not through receptors but via pinocytosis, processed through endosomes and lysosomes and released with very little degraded approximately 10 h later. The method by which soluble AβP avoids degradation is unknown; however, it may be that AβP monomers join aggregates within the cell and, therefore, reduce the surface area open to the degrading environment.43

Hydrophobic proteins associated with membrane damage

Antimicrobial proteins

Antimicrobial peptides are distributed widely in all higher animals, insects, arthropods, tunicates, bacteria and plants (for a review see Kourie and Shorthouse13). Antimicrobial peptides may be secreted at sites of inflammation or be continually produced by the organism and many of these peptides are thought to act via direct membrane interaction leading to disruption and lysis.13 Many of these peptides are broad spectrum, with some, such as magainin, acting on Gram-negative and Gram-positive bacteria, yeast and fungi and also showing haemolytic activity. Some fall into the general class of cytotoxic peptides, being short (≤ 40 amino acids) and linear. Examples are cecropin and magainin, from insects and amphibians, respectively. These are part of a large group of amphipathic α-helical peptides.45 This group of peptides, including alamethicin and gramicidin A, exhibits significant differences in membrane association behaviour. Cathelicidins are short mammalian antimicrobial peptides, which are produced as preproteins prior to secretion. They exhibit a variable content of α-helical, β-sheet and random coil secondary structure in different members of the group.46

The small proteins magainin and cecropin are membrane active. These peptides take an α-helical conformation in membranes. In solution, magainin exists as a random coil structure. It is suggested that this change in secondary structure in membrane environments is responsible for the development of antimicrobial activity.45 Magainin is an example of a structure with a hydrophobic N-terminal and a cationic C-terminal.47 Studies of the membrane-bound peptides indicate that the α-helices formed are amphipathic, with hydrophobic and hydrophilic side-chains on different faces of the helix. The helix appears to insert near the membrane surface and parallel to it, rather than by transmembrane penetration. This insertion would not directly achieve the observed alterations in transmembrane gradient, so it is possible that a small percentage of protein does span the membrane. However, further studies indicated that membrane spanning by this and similar peptides was not required for antimicrobial behaviour, indicating that some other mechanism than pore formation may also contribute to toxicity (see Tossi et al.47).

Although melittin is haemolytic, as well as exhibiting antibiotic activity, it shares many structural and pathological characteristics with the magainins and cecropins; indeed, hybrids, such as cecropin–magainin, are membrane active.45,47 At different concentrations and membrane compositions, melittin is found to cause various changes to the structure of the bilayer, including micellization, fusion of micelles into lamellar states and wedge insertion into the bilayer.45 VacA from Helicobacter pylori and Vibrio cholerae haemolysin induce cell vacuolation, which occurs in a similar manner to melittin-induced micellization and membrane fusion, acting as possible antibacterial mechanisms.48–51 This observation points to the similarities and significance of prion-induced vacuolation and AβP-induced endocytosis in neurodegenerative disease (e.g. prion disease and AD, respectively; see above). Differentiation of effects based on peptide concentration and membrane composition may well be a significant mechanism for confining toxicity to certain cells and membrane characteristics. For example, bacterial plasma membranes are rich in negative phospholipids compared with vertebrate cells, which appear neutral, and this will lead to different degrees of membrane association.45

Another antimicrobial peptide that affects membranes is tritrpticin, a cationic 13 amino acid peptide, which contains three Trp and four Arg residues. It is part of an antimicrobial peptide group known as cathelicidins, which are thought to cause disruption and lysis by binding to the microbial membranes. Schibli et al.46 proposed that membrane-associated tritrpticin formed an amphipathic structure. It is thought that the interaction of tritrpticin with membranes may be similar to that of magainin-2.46 It is hypothesized that both act by inducing positive curvature strain in membranes. Antimicrobial peptides that disturb cell membranes include Gramicidin A, a 15-residue antimicrobial peptide of Bacillus brevis, which also acts on human cells. This peptide is found to form an α-helix in the membrane and a transmembrane dimer acts as a cation pore. In addition, alamethicin, a fungal antimicrobial peptide, also exhibits transmembrane orientation and tends to form helical ‘barrels’, the number of helices in which depends on peptide concentration, membrane composition and other factors.45

After traversing the cell membrane, amphipathic α-helical peptides tend to interact electrostatically with the negatively charged membrane, prior to inserting the hydrophobic face of the helix into the membrane interior. The helices can associate to form raft-like aggregates. Further membrane disruption may be brought about by: (i) re-orientation to transmembrane ‘barrels’ or pores; (ii) thinning of the membrane itself; and (iii) the peptides acting in a detergent-like manner to cause disaggregation (the ‘carpet’ mechanism; see Tossi et al.,47 Shai and Oren52 and references within). All these mechanisms are believed to occur with different peptides and environments. None relies on membrane receptors, because all-d peptides exhibit the same activity, which indicates that hydrophobic interaction is important for antimicrobial cytotoxic action. The secondary events, such as ion gradient disruption, will also play a part in cell death.

Amyloid-forming proteins

This is a class of at least 16 cytotoxic peptides and proteins characterized by the formation of amyloid plaques and fibrils, which are ordered insoluble aggregates of one or several protein types. Such peptides are associated with various neurodegenerative and other pathologies, although it is not always certain that the fibrils are responsible for cytotoxic activity. Amyloidogenesis is often associated with channel interaction or membrane disruption. Membrane interaction may be required for the peptide structural changes that are necessary for aggregation. Apart from the designated class of amyloid peptides, many other peptides, including melittin and the defensins,53 aggregate in some conditions. This may be initiated by protein unfolding (e.g. α-sarcin) and/or lipid interaction.

The amyloids that have been investigated in great detail are prion and AβP. Normal cellular prion is a non-toxic protein, mostly composed of α-helical structures with very little β-structure. The isoform PrPSc is believed to be involved in causing several neurodegenerative diseases, such as Creutzfeldt–Jakob disease, kuru, scrapie and spongiform encephalopathies. The fragment 106–126 mimics some of the observed behaviour of the whole pathogenic protein and this sequence is believed to play a key role in the pathological effects of the protein. The sequence is part of, or close to, the small amount of β-sheet in PrPc and the PrP[106–126] fragment is found to increase in β-sheet content on binding to the membrane. It contains a hydrophobic core sequence, hypothesized to be involved in prion's neurodegenerative properties via membrane interaction and destabilization (see Kourie30). This hydrophobic sequence, which ranges from residue 113 to 120 and is alanine rich (AGAAAAGA), is involved in prion's ability to bind and destabilize lipid bilayers.

Amyloid β protein is involved in many neurodegenerative diseases, but it has been studied more specifically in AD. It is also present in small amounts under normal physiological conditions. It is thought that at least part of its cytotoxic behaviour involves forming Ca2+ channels within cell membranes, thus altering cell function and regulation. The AβP fragments, which can exist anywhere between 39 and 42 residues in length, are involved in forming these channels. These fragments have β-sheet structures in the N terminal and α-helices in the C-terminal of the peptide. It is in this α-helical C-terminal that a hydrophobic fragment, which is hypothesized to cause interaction and destabilization of the membrane, resides. This fragment, which extends from residues 29–35 (GAIIGLM), consists mainly of hydrophobic residues and is glycine rich. This region has been shown to modulate the secondary structure of AβP, because it has been shown that point mutations in this region lead to a decrease in peptide β-sheet content and also cytotoxic activity and aggregation.54,55

Another amyloid-forming peptide, namely calcitonin, which is found in the thyroid gland of mammals, is a hormone involved in the metabolism of calcium–phosphorus and in calcium homeostasis. Human calcitonin is a 32 amino acid peptide that consists of both α-helical and β-sheet secondary structures. Although primarily non-toxic and soluble in the body, human calcitonin forms insoluble fibrils in aqueous solution. Calcitonin fibril may be initiated by hydrophobic interaction of the hydrophobic region (residues 26–32) in the N-terminal, while the C-terminal is involved in the fibrillization of the peptide.56 The calcitonins are thought to interact either directly with membranes and/or at receptor sites.57,58 Structural studies of calcitonin in sodium dodecyl sulphate (SDS) micelles reveal an amphipathic α-helix in the central region 9–16 and a type I β-turn formed at residues 16–19 in the C-terminus, which are important in the structure and biological activities of this peptide. The amphipathic α-helix has been found to be important for the interaction of calcitonin with receptors.58,59 Motta et al.58 hypothesized that, following hydrophobic interaction by the amphipathic helices, the extended C-terminal of human calcitonin is involved in fibril formation by providing a template for α-helical rods. It was reported that β-sheet structure was also seen in the fibrils.57,58

Viral proteins

Several viruses also use protein hydrophobicity to destabilize cell nmembranes. A protein that exhibits amphipathic α-helical nature in membranes is Synapsin I (SI), a human immunodeficiency virus (HIV) coat protein associated with the vesicle fusion properties of the virus. Synapsin I is an amphipathic protein segment from the C-terminus of the HIV coat protein and, like melittin, has a net cationic charge and adopts an α-helical structure. However, SI is more hydrophilic, relying on electrostatic interaction with the negative lipid headgroups to initiate a process of embedding itself in a wedge-like interaction within the membrane and promoting membrane fusion and cell lysis. Synapsin I interacts only with negatively charged membranes.53 In particular, the conserved N-terminus has been shown to bind with high affinity to acidic phospholipids and both the N-terminus and three C-terminus regions are found to insert into the hydrophobic membrane core.42 According to Fujii,53 the greater hydrophilicity of some SI regions, in comparison with that of melittin, is responsible for the lack of interaction of SI with neutral membranes. In melittin, the distinct hydrophobic region, separated from the charged residues, enables strong enough hydrophobic interactions for insertion into neutral bilayers.

A proteolytic fragment of a rotavirus coat protein, virion protein 5 fragment (VP5*), also permeabilizes cell membranes and liposomes and causes membrane fusion. Its activity is temperature and dose dependent and occurs at neutral pH.60 The structure of VP5* contains a long single hydrophobic domain (residues 385–404 in the C-terminal region), enabling the formation of a pore for Ca2+ transport, which then modifies the virus, allowing translocation. Mutagenesis or truncation in this region led to loss of the ability of this fragment to cause membrane permeabilization.60 It is noted that the influenza M2 protein also forms voltage-gated channels. The SI-induced membrane fusion is suggested to occur when two protein molecules each bind a liposome and then dimerize.42 Such a mechanism, if confirmed, may prove common to a number of peptides involved in membrane fusion. This would also provide a further link between the membrane binding and aggregation capabilities of a peptide.

Mechanisms of hydrophobic protein interaction with membranes

Secondary structure, hydrophobicity and membrane interaction

Protein ‘misfolding’ is a key event in the development of cytotoxicity. This hypothesis led to investigations of whether particular protein structures are preferred for lipid interaction. There is clear evidence that increased content of both β-sheet and α-helix can confer cytotoxicity on a peptide. The ability to form ion channels, regardless of the secondary structure of the protein, is the significant factor in protein-induced membrane damage. The arrangement of amino acids and, in particular, the exposure of hydrophobic residues, is an important factor in the ability of a protein to interact with a lipid membrane, induce membrane fusion and destabilization. Fujii53 found that a scrambled SI protein was able to disrupt the membrane, probably via electrostatic interactions, but was unable to cause fusion or lysis of the membrane, demonstrating that the amino acid sequence and amphipathicity of the peptide have a role in its ability to cause membrane destabilization.

All-d, retro and retroenantio analogues of some amphipathic peptides, such as magainin and cecropin, show similar activity to their parent peptides.13 This indicates that it is amphipathicity and possibly the distribution of hydrophobic and charged residues, rather than specific sequences, which confer on the protein its membrane activity.47,61 However, specific single d-amino acid substitutions or mutations do significantly affect the structure and activity of peptides. Melittin is known to form pores in eukaryotic membranes, as opposed to the membrane disruption resulting from the toroidal pore or ‘carpet’ mechanism that has been invoked for some bacterial cells.52 Single d-substitution reduces peptide structuring, which reduces haemolytic activity, while having little effect on antimicrobial action, suggesting that peptide structure is more important for transmembrane pore formation.47,62 Similarly, ‘walking substitutions’ in melittin, in which a substitution or omission is inserted at each point in a peptide sequence, have shown that haemolytic and antimicrobial activity depend on hydrophobicity and the ability to form an amphipathic helix and correlate with an ability to interact with a hydrophobic medium.47,63

While it appears that no particular secondary structure confers an inherent advantage in membrane interaction, for any particular peptide it may be that one structure may allow better access for the hydrophobic residues to the membrane. This means that a change in secondary structure may lead to a change in, or initiation of, cytotoxicity of a particular peptide. This mechanism is particularly relevant to the investigation of autocytotoxic peptides, because the structural change or its trigger may be an ideal target for a therapeutic strategy. While changes in secondary structure are often important in membrane interactions, it is vital to recognize that often the alterations arise as a direct result of contact with the lipid environment. The hydrophobic lipid allows for lower energy exposure of hydrophobic residues, which necessitates a rearrangement of the whole protein. The change in structure seems often to be coincident with, rather than preceding, lipid interaction and insertion. This is not always the case. Prion protein is known to undergo structural changes when exposed to less acidic pH in solution64 and it is possible that this change or other environmental factors may drive the interaction with the membrane in some cases. Not all proteins undergo rearrangement of secondary structure. Diphtheria toxin inserts three α-helical segments into the cell membrane in order to allow translocation of its catalytic domain. These segments are observed before and after membrane interaction and it is hypothesized that the protein's conformational change is via a ‘molten globule’ intermediate, in which secondary structure is not altered but tertiary refolding does occur.40

The importance of secondary structure in the exposure of amino acids and hydrophobic residues was demonstrated by Gasset et al.,37 who analysed the structural requirements for α-sarcin to destabilize lipid bilayers. It was hard to explain the hydrophobic interactions of α-sarcin with a membrane, because it is highly polar while in its native configuration. Gasset et al.37 found that when α-sarcin unfolded, some β-sheet was retained and there was also more random coil. In the absence of lipid there was no aggregation by the unfolded protein, which suggests that the hydrophobic regions, including two Trp residues, were not exposed purely by unfolding. However, interaction with lipid vesicles did lead to a change in the structure of the protein, with an increase in α-helix and decrease in β-sheet content from the mostly unfolded variant. This indicates that association with lipid membranes may be required for a change in secondary structure, probably because of the exposure of residues to a different environment.37

Studies have determined that the normal cellular prion consists mainly of α-helices, whereas the neurotoxic prion protein segment, which corresponds to residues 106–126, consists predominantly of β-sheet structures. β-Sheet content increases at low pH or in the presence of lipids.64 It is thought that the formation of β-sheets in the PrP[106–126] is due to unfolding of the peptide in the presence of lipids and that this change in structure contributes to membrane destabilization and the toxic effects of PrP[106–126]. The link between structural alteration and membrane destabilization is confirmed by point mutations in the PrP hydrophobic region. Decrease in β-sheet content closely corresponded to decreased cytotoxicity. Jobling et al.64 substituted the hydrophobic residues alanine and valine with the hydrophilic residue serine in the PrP[106–126] hydrophobic core (residues 113–122). This substitution induced a reduction in the hydrophobicity of the region 113–122 and dramatically reduced the neurotoxicity of PrP[106–126]. They also found that, although at pH 5 both wild-type and mutant peptides exhibited random coil structure, the mutants that differed by more than one residue did not show the increase in β-sheet structure with pH observed for the wild-type PrP[106–126]. This suggests that the hydrophobic core of PrP[106–126] is important in maintaining the peptide's β-sheet content and plays a role in the peptide's neurotoxicity.

It has recently been postulated that the hydrophobic region of a protein may also play a role in modulating the secondary structure of the protein. This may explain the observed changes in peptide secondary structure on binding to the lipid membrane as the environment of the hydrophobic residues alters. Data so far produced confirm a possible link between changes in hydrophobic content and secondary structure. For example, it has been shown that mutations within the hydrophobic region of AβP (residues 25–35) reduce the β-sheet content and fibrillogenic properties of the peptide.65 Pike et al.55 showed that the hydrophobic region 29–35 of AβP modulated both secondary structure and neurotoxicity. It could be that it is not the entire hydrophobic region, but only certain residues, that are important in the maintenance of the preferred secondary structure of a protein. Identification of key residues would, as well as providing a basis for targeted therapeutic strategies for cytotoxic peptide pathologies, give information as to structural parallels between peptides and families of cytotoxic proteins. Studies have shown that substitution of residues, such as alanine and valine, has direct effects on peptide activity, for example in PrP[106–126].64 Similarly, Barrow and Zagorski66 found that valine was important in the β-sheet formation of AβP, because removal of four valines within the C-terminal reduced β-sheet formation.

Malovrh et al.67 examined the involvement of the five tryptophan residues, which are located at positions 45, 112, 116, 117 and 149, in maintaining the structure of the protein equinatoxin II (EqtII), a pore-forming toxin found in sea anemones. It has been hypothesized that the N-terminal of the protein is involved in interactions with lipid membranes and that the α-helical structure of the N-terminal is an important factor in the ability of EqtII to interact with lipid bilayer, although Anderluh et al.68,69 proposed that the α-helical structure is not a requirement for EqtII binding to lipid bilayers. Malovrh et al.67 described EqtII as being predominantly composed of β-structures, including β-sheets and β-turns, and possessing only a small amount of α-helix. Malovrh et al.67 used three mutations of EqtII, retaining different Trp residue/s, in order to study the functional effect of each Trp region on the secondary structure of the protein. Each mutation resulted in less haemolytic activity. The secondary structure of the mutation containing only the Trp116,117 residues was very similar to that of wild-type EqtII, whereas the mutant containing only the Trp149 residue was severely affected, with the amount of β-structures decreased and α-helix and random coil content increased compared with the wild-type EqtII.67 The mutant retaining only Trp45 also showed structural differences to the wild type. From these results, Malovrh et al.67 hypothesized that the substitution of Trp residues 112, 116 and 117 by phenylalanine promotes substantial change in the structural properties of EqtII, while substitution of the residues Trp45 and Trp149 does not. This suggests that the Trp112 Trp116 and Trp117 residues are involved in maintaining the structure of EqtII, which could be important in the ability of EqtII to bind to lipid bilayers. Malovrh et al.67 further demonstrated that the EqtII mutant containing only the Trp residues 112, 116 and 117 interacted and bound with lipid bilayers, whereas the EqtII mutants containing only the Trp residues 45 or 149 did not interact with the lipid bilayer. Given that the structure of the mutant EqtII containing residues 45 or 149 differed from the natural wild-type EqtII protein and that these mutants were also incapable of interacting with the lipid bilayer, it could be hypothesized that the secondary structure of the protein does influence its ability to interact with a membrane, or at least that these are related properties and that certain proteins have a secondary structure that is preferential for lipid bilayer interaction. The secondary structure of this protein may be very important, because the residues 112–117 are located in the hydrophobic interior of the protein, where they are protected from quenching, oxidation and other chemical alterations, which may decrease the ability of the protein to interact with lipid bilayers. This also provides evidence that the secondary structure plays a role in placing important amino acids and hydrophobic residues in positions that are favourable for their interaction in lipid membranes.

Overall, these studies indicate that the positioning of hydrophobic residues, the protein secondary structure and protein–lipid interaction are all linked. The existence of hydrophobic regions in a protein will lead to it adopting a structure that best stabilizes those residues. The peptide environment is likely to be altered by interaction with a lipid membrane, which may lead to a change in structure, just as an alteration in the cellular environment, for example, pH, may destabilize a particular secondary structure and lead to changes that may increase the access to and preference for the lipid membrane of the hydrophobic regions. In addition, certain secondary structures may favour the adoption of larger structures, such as oligomers or aggregates, which also have links to cytotoxicity in some cases. Finally, as just discussed, mutations in the hydrophobic regions of a peptide will have significant effects on the stability of the secondary structure and on the way in which the peptide interacts with the hydrophobic membrane. We hypothesize that all of these factors interact and affect each other's action and, in many cases, it is not possible to determine the single causative change, if any, that leads to changes in structure, cytotoxicity and peptide environment.

Energy calculations suggest that hydrophobic peptides, depending on their structure, differ in their ability to insert and form ion channels in membranes. As mentioned earlier, a very significant point is that free energy calculations on bilayer binding indicate that it is much more favourable (by 50–70 kcal/mol) to insert an α-helix into the bilayer than to insert a random coil, which later forms the more ordered structure.53 That conformational changes may take place outside the membrane environment is also indicated by the pH dependence of the membrane-disrupting and aggregation properties of many peptides. However, this does not negate the reasonable possibility that conformational change takes place on the surface of the bilayer, perhaps after electrostatic interactions or when the peptide is close to the hydrophobic inside of the membrane. Indeed, the many studies that show that the presence of vesicles does affect the secondary structure adopted by various peptides would seem to indicate that some interaction drives structural alterations just before insertion.

Amphipathic proteins

In order for a protein to destabilize a membrane, it must be able to adopt a three-dimensional structure that is either amphipathic or contains a region of hydrophobicity. It was once thought that an amphipathic or hydrophilic α-helical segment promoted membrane destabilization; however, it has been demonstrated that β-structures can also cause membrane destabilization through fusion of membranes. The type of membrane destabilization depends upon the balance of the electrostatic and hydrophobic properties and the amphipathicity of the protein. Melittin is largely hydrophobic, with a high hydrophobic moment and residues that form a hydrophobic dipole. This dipole is optimal for interaction with the lipid bilayer through hydrophobic forces leading to interactions of the charged amino acids with the lipids and destabilization of the membrane. Ohmori et al.70 used synthetic 18-residue peptides with a systematically varied hydrophobic–hydrophilic balance to investigate the effects of amphipathicity in membrane disruption and endocytosis. The peptides contained cationic lysine and hydrophobic leucine residues in ratios 13 : 5, 11 : 7, 9 : 9, 7 : 11 and 5 : 13. It was found that those peptides with 50% or more hydrophobic character were effective in forming stable aggregates of α-helical monomers and in transfecting the cell. It is suggested that the Leu13 peptide did not act solely via endocytosis, but also through stronger membrane-perturbation effects not observed in less hydrophobic peptides. Overall, the hydrophobic–hydrophilic balance does play an important role in determining the extent and mechanism of membrane disruption.70

Although some peptides may interact with membranes via purely electrostatic means, many are cationic and engage in electrostatic interaction with charged lipid headgroups before hydrophobic insertion. The degree of charge is an important parameter in membrane interaction. Increasing charge can, in some instances, increase peptide potency by increasing electrostatic attraction and/or by further stabilizing the secondary structure. An example is the antimicrobial peptide pardaxin (+2), which was more active in a +4 form. In contrast, if there is a concomitant decrease in hydrophobic residues, then structuring and activity may decrease. It has been found that the maximal activity for antimicrobial α-helical peptides occurs at around +4; however, this does not take account of targeting and self-protection.47 The haemolytic activity is directly correlated with the angle subtended by the hydrophobic sector in helical wheel diagrams of antimicrobial α-helical peptides. A five-residue sector is sufficient to confer antimicrobial, but not haemolytic, activity on a peptide.47,63

Models for membrane destabilization and insertion

The membrane changes effected by proteins can vary depending on the nature of the peptide and the lipid involved. Some peptides, particularly those characterized by electrostatic interaction, merely disrupt the membrane by binding to the charged headgroups. Such disturbance is likely to be reversible and short-term. Purely electrostatic interactions can cause aggregation of liposomes, but no mixing.53 However, most membrane-active proteins irreversibly affect membrane character by partial insertion into the hydrophobic interior of the bilayer. Most studies favour a model of electrostatic interaction between positively charged protein residues and negative or polar lipid headgroups, followed by insertion of some hydrophobic region of the protein into the membrane, possibly after a structural change.

Following protein insertion, the membrane can be affected in a number of ways. Apart from the general electrostatic disturbance, many proteins are responsible for fusion of vesicles. This fusion may be driven by dimerization of proteins in two separate membranes, as is proposed for the action of SI,42 or it may occur through the need to minimize exposure of the hydrophobic regions to the surrounding medium. In this case, the hydrophobic protein may act as a bridge, allowing mixing of the lipids in the membranes. Another result of protein insertion could be the loss of membrane integrity, through crowding the membrane, alterations in membrane thickness around the protein and/or through channel or pore formation. Mismatching of hydrophobic regions has also been postulated as a significant mechanism of membrane disruption.71

Electrostatic and hydrophobic interactions with membranes

There is much evidence to suggest that the interaction of many proteins with membranes occurs via a mechanism that involves two major steps: (i) electrostatic interaction of the charged residues with charged or polar lipid molecules; and (ii) insertion of hydrophobic regions into the hydrophobic interior. Electrostatic attraction yields a negative (exothermic) enthalpy contribution to free energy of the interaction, whereas the hydrophobic insertion contributes entropically.72 Membrane–protein interactions may be entirely enthalpy driven, as is the case for hydrophilic peptides, including somatostatin and magainin, or entirely entropy driven, as for the hydrophobic cyclosporine A, for which the interaction is endothermic. For some amphipathic peptides, both terminals may favour insertion, as for the actin-binding protein talin.72

As is demonstrated by the specificity of many cationic peptides for negative membranes, electrostatic interaction is often required for membrane disruption. It was found that SI only bound to the negatively charged phospholipids and did not bind to neutral or positively charged phospholipids.42 Even those proteins, such as melittin, that can interact with neutral membranes may tend to do so because they are more strongly cationic, which is hypothesized to allow them to approach polar regions, even within the overall neutral membrane environment.53 In many cases, electrostatic interaction is followed by a closer approach and insertion into the hydrophobic membrane core. Although this hydrophobic attraction is required for irreversible binding and for fusion and reorganization of the membranes,53 it is believed that some peptides exert their effects purely by electrostatic interaction, which leads to some membrane disturbance and may alter the membrane environment through neutralization of its negative charges. Fujii53 reports that peptides that are more amphipathic and/or hydrophilic, such as SI and defensins, destabilize membranes mainly via electrostatic interactions with the negative lipid headgroups, rather than through hydrophobic forces.

The membrane–protein interaction may be accompanied by a change in the secondary structure of the protein. This change will tend to expose hydrophobic residues, which will then stabilize by entering the interior of the membrane. Peptides will also alter in structure so as to group hydrophobic and charged residues in such a way that each region can be in its preferred environment. In a study of binding of the antimicrobial peptide tritrpticin to SDS micelles, Schibli et al.46 found that the structure changed, forming an amphipathic structure involving a two-turn fold, where the hydrophobic residues Trp and Phe cluster together into a hydrophobic patch. This patch, which is buried in the micelle, is believed to induce a positive-curvature strain in the bilayer, causing disruption of the membrane. The basic residues surrounding the hydrophobic patch also interact with the membrane electrostatically. Fujii53 discusses the prevalence among characterized proteins of amphipathic α-helices in promoting membrane fusion and also demonstrates the correlation between the fusion abilities of proteins and their hydrophobicity. A common structure adopted is that of an amphipathic α-helix in which hydrophobic residues are mainly along one external side of the spiral. The helix will then insert into the membrane parallel to and close to the surface. An example is the 17-residue H17 region of talin.73 Human calcitonin is thought to interact with the surface of membranes via an amphipathic α-helix of residues 9–16.58 Cheetham et al.42 showed that binding of SI to acidic phospholipid was initiated by electrostatic interaction with the surface of the membrane, followed by direct insertion of selected regions of the molecule into the hydrophobic core of the membrane. Two of the three conserved penetrating regions are in the form of amphipathic α-helices that could insert into the membrane. The third is proposed to undergo either a change to an amphipathic helical structure or to insert via a short hydrophobic loop.42 The hydrophobic interaction is stronger than the electrostatic effects and cannot be effectively site-specific phosphorylation.

α-Sarcin initially interacts with lipid bilayers via electrostatic interaction of this cationic protein with the negatively charged membrane. Gasset et al.37 hypothesized that the protein then unfolds to reveal β-sheets, which expose a hydrophobic domain by altering to an α-helix. This domain interacts with the hydrophobic core of the bilayer and causes aggregation and/or fusion of the lipid bilayer. Energy calculations revealed that it was more favourable for the protein to change structure before insertion than after entering the membrane with unaltered structure. Defensins are an example of a group of β-sheet structured peptides that interact both electrostatically and hydrophobically with lipid membranes.53 In amphipathic peptides, where hydrophobic interaction also occurs, the charged residues are likely to remain at the surface of the membrane among the lipid headgroups, although, in some cases, the peptide may undergo rearrangement to form aggregates, two-dimensional micelles or pores, which also place these residues in a favourable environment. More than one region of the protein could be involved in the interaction with membrane. Cheetham et al.42 found that the three regions of SI protein, which insert into the bilayer, can all orientate to form a single hydrophobic surface. This surface is opposite to the molecule's dimerization surface and it is proposed that dimerization of inserted molecules would provide a mechanism for the observed liposome aggregation. The re-orientation of hydrophobic and charged residues is essential. Schibli et al.46 proposed that membrane-associated tritrpticin formed an amphipathic structure, in which the hydrophobic residues Trp and Phe cluster together into a hydrophobic patch buried in the membrane core and the Arg residues are orientated on the opposite side of the peptide and are exposed to a hydrophilic environment.

Wedge-type incorporation of the protein into the membrane

This model describes a method of membrane disruption in which the hydrophobic region of the protein incorporates into the membrane in a wedge-like fashion, while the hydrophilic residues remain largely at the membrane surface. The model is used by Fujii53 to explain protein-induced membrane fusion of vesicles or liposomes. The model is proposed for amphipathic proteins and peptides, such as melittin, SI and defensins. The hydrophobic region of the peptide incorporates like a wedge into the lipid bilayer, while the charged or polar residues interact with the lipid headgroups at the surface of the membrane. Charged and polar residues also bring other liposomes into close proximity via electrostatic attraction and these can approach closely because the protein has disturbed the like-charged surfaces, as well as the hydration shell. Once together, the lipids are allowed to mix through the formation of a bridge, created by the hydrophobic portion of the peptides. Fujii53 also proposed that there are two possible peptide structures that can be formed via this process. In the first, the hydrophobic portion of the peptide remains embedded within the membrane, while the hydrophilic region is exposed to the extracellular environment and the peptide is orientated roughly parallel to the membrane surface. This type of insertion is confirmed by studies of viral proteins (e.g. infectious flacherie virus (IFV2)) showing a strong asymmetry in the distribution of hydrophobic glycine residues around the helix, suggesting that there is a segregation of residues between the sides of the helix.53 The second structure occurs with micellization of the peptide to form membrane-spanning pores or disc-like structures. The wedge-shaped proteins cause membrane destabilization by placing a strain on the lipid bilayer, which can create forces that interact with nearby proteins. The proteins can also create forces that cause interactions with nearby proteins through Gaussian curvature, van der Waals' attractions and hydrophobic and electrostatic interactions of the wedge-shaped protein with the lipid bilayer. Calcitonin, a peptide hormone with a role in calcium metabolism, has been hypothesized to interact with the surface of membranes via an amphipathic α-helix.13,53

Hydrophobic mismatch

Dumas et al.71 investigated the effects of hydrophobic mismatching of the lipid and protein on membrane organization. The correlation of hydrophobic lengths of lipid and protein is proposed to play a role in membrane organization and function. It was found that the length of the protein, compared with the length of the lipid's acyl chain, had an effect on the hydrophobic thickness of the lipids in contact with the protein, because the bilayer molecules stretch or bend to match the thickness of the incorporated protein. If the protein's hydrophobic segment is longer than the acyl chain length of the lipids, then local expansion of the bilayer will occur in a manner dependent on the hydrophobic length of the protein. If the protein's hydrophobic segment is shorter than the acyl chain length of the lipids, then the bilayer will undergo a corresponding contraction. The relationship between lipid/protein hydrophobic mismatch and thickness of bilayer was also studied by de Planque et al.73 A series of hydrophobic peptides of repeat sequence WALP, with varying sequence length, were incorporated into phosphatidylcholine (PC) lipid bilayers with varying acyl chain lengths. It was found that the lipid bilayers did change mean thickness to compensate for mismatching of hydrophobic lengths and that the greatest mismatch led to the greatest change in bilayer molecules. However, based on ‘first shell’ lipid maximal changes, the membrane response to mismatch was only partial.73 This may indicate that other factors, such as peptide tilt or structural alteration, also help to lower unfavourable mismatch. The difference in lipid response to a WALP peptide and α-helical gramicidin of similar length indicates that peptide surface characteristics, probably including the arrangement of Trp residues, also affect systemic response to mismatch.

Changes in the bilayer structure lead to an increase in free energy, which is at a minimum when the hydrophobic regions are approximately equal in length. The compression or expansion of the lipid bilayer, due to mismatched hydrophobic lengths, causes tension and elastic forces on the membrane, which lead to membrane destabilization. Such destabilization may lead to observed cytotoxic effects, such as membrane fusion or lysis, or membrane-mediated pore formation. Disturbance in the acyl chain order due to the mismatch in hydrophobic length of the protein is also expected to result in an overall shift in the mean phase transition temperature of the lipids.

A method commonly used for examining the effects of a peptide on the bilayer is determination of the phase-transition temperature. When the lipid has long-range two-dimensional order, but the hydrophobic tails are relatively disordered, it is in the ‘liquid-crystal’ phase. In contrast, the acyl chains of the molecules in the gel or lamellar phase are packed more tightly in an all-trans fashion.74 Different lipids exist in either phase under physiological conditions. The breadth, enthalpy and temperature of the phase transition are all affected by peptide insertion and mismatch. The alteration in the lipid phase-transition temperature between gel and liquid-crystalline phases may be a significant mechanism of peptide cytotoxicity. The consequent alteration of properties in the region surrounding the peptide would affect membrane microviscosity and the changes may alter both the inserting peptide structure and the structure and dynamics of nearby intrinsic proteins. The gel phase forms significantly thicker bilayers, which may alter the activity of channels.

Mismatch of hydrophobic length between the lipid and protein was also proposed to have an effect on the conformation, aggregation state, protein–protein interaction and activity of the protein. It was demonstrated that an increase or decrease in acyl chain length of the lipid, compared with the protein's hydrophobic length, led to a loss in the protein's activity.74 It is suggested that aggregation and conformation changes will occur to relieve hydrophobic mismatch.

As well as providing a possible mechanism of protein-induced membrane destabilization, the model of hydrophobic mismatch may also explain the preferential binding of proteins to different lipid membranes. Dumas et al.71 found that the stabilization of protein and lipid compositions could be achieved via a mechanism for sorting the proteins and lipids. Proteins were found to preferentially insert into bilayers of certain lipids in a way that corresponds to hydrophobic matching. This model may explain, in part, the difference in cytotoxic activity of proteins on different cell types.

Promoters/stimulators of hydrophobic protein interaction with membranes.

As shown above, changes in secondary structure are important in enabling a protein to interact with and destabilize lipid bilayers. Studies have shown that several factors (pH, lipid composition, temperature and salt content) induce changes in the protein structures that confer on these proteins the ability to expose their hydrophobic regions, needed for incorporation into the membrane and channel formation. Changes in pH can affect the folding and secondary structure of proteins and alter the net charge on a protein, which may affect its electrostatic interactions with the bilayer. For example, pH induces a net positive charge on amphipathic proteins, such as E. coliα-haemolysin.75 Changes in the intracellular pH environment induce different prion protein conformational transitions and susceptibility to protease digestion. These properties, which, in turn, are thought to be involved in the regulation of the hydrophobic regions of these proteins, affect the ability of these proteins to aggregate or interact with the lipid bilayer. It was found that pH was essential to the activation of viral fusion proteins (e.g. influenza virus haemagglutinin (HA) and the fusion factor of vesicular stomatitis (VSV) and rabies viruses).41,48,75,76 These viruses are imported into the cell via an endosomal pathway and the acidic environment of the endosome induces a conformational change, which exposes regions of higher hydrophobicity. These regions then insert into the endosome membrane before fusion with the viral envelope occurs.76 Studies on the fusion glycoprotein in VSV and rabies virus indicate strong pH sensitivity. A pH of 6.7 changes the interaction of the virions and increases hydrophobicity; however, further protonation at pH below 6.3 is necessary for fusion. The conformational change observed is reversible for the glycoprotein, but not for HA.

The composition of lipid membranes may be a key point of the observed differences in peptide–membrane interaction in different cell types; for example, some antimicrobial peptides, such as magainin, bind only with negative membranes, thus allowing them to target bacterial but not host organism cells. Obviously, charge is a key factor in lipid–protein interaction, with electrostatic interactions often responsible for the initial interaction before hydrophobic protein insertion. Lipid charge, headgroup, chain length and packing may all influence the preference. Dumas et al.71 found that peptides associated with lipids in a way corresponding to minimization of hydrophobic mismatch; however, the research of dePlanque et al.73 suggests that headgroup effects may, at times, dominate mismatch. Mukherjee and Maxfield74 report that some proteins partition to cholesterol/sphingolipid-rich domains (e.g. Src protein kinases) or glycosphingolipid receptors (e.g. Shiga toxin). An example of the use of membrane composition to determine cytotoxic activity is the antimicrobial peptide of sheep, namely SMAP-29. As well as being active against bacteria, this peptide also lyses human erythrocytes, but does not attack sheep erythrocytes, which have a different membrane composition.77

Important regions of hydrophobic proteins

In amphipathic, α-helical antimicrobial peptides, the N-terminal domain is always essential and often sufficient for antimicrobial activity.47 Tossi et al.47 performed statistical calculations on the N-terminal domain of 150 such peptides and their study revealed significant sequence homology in terms of the incidence of hydrophobic, charged, uncharged, polar and structural (Gly/Pro) residues. This incidence tends to place the hydrophobic residues on one face of the helix and the charged residues on the other. Amidation of the C-terminal, which contributes an extra stabilizing H-bond, is also common.

The study of Cheetham et al.42 using whole SI protein to determine the membrane-penetrating regions, shows that the three regions of amino acids 166–192, 233–258 and 278–327 in the highly conserved N-terminal C domain of SI, which are responsible for the hydrophobic interaction with and penetration of lipid bilayers, are surface exposed on the folded protein. Amino acid sequences 233–258 and part of 278–327 form amphipathic α-helices, while amino acid sequence 166–192 is hypothesized to either undergo conformational change to adopt an amphipathic α-helical structure upon binding with the lipid or to interact with the lipids via a hydrophobic loop.42

The key hydrophobic region of AβP extends from residues 29–35 (GAIIGLM). Similar regions have been observed in the cytotoxic peptide family and in toxins (see below). This region is thought to play an important role in determining the secondary structure, the neurotoxicity of the protein and ion channel formation. Consistent with this hypothesis is the observation that many of the cytotoxic properties of AβP are mimicked by the fragment AβP25-35, which contains this region. Membrane destabilization has been proposed as the mechanism by which this segment of AβP induces cytotoxicity.26

The toxic properties of prion have been similarly attributed, in part, to a hydrophobic segment, which consists of residues 113–122 (AGAAAAGAVV). The significance of this hydrophobic sequence in the toxic action of the prion has been demonstrated. For example, Jobling et al.64 investigated the properties of mutant PrP[106-126] peptides in which varying numbers of the hydrophobic residues were replaced by the hydrophilic residue serine. It was found that the substitutions decreased the formation of β-sheet structure and dramatically reduced the toxicity, aggregation and fibril formation of the protein. The mutant proteins were compared with the wild-type at pH 5–8. In the wild-type, the pH increase induced an increase in β-sheet structure, whereas this was not observed to the same extent in any substituted peptide. In particular, the two mutants in which the two valine residues were substituted showed almost no change in structure.

Fragment VP5* contains a hydrophobic domain in its C-terminal, which, it is hypothesized, is involved in permeabilizing cell membranes.60 When the GGA motif was removed from the C-terminal, membrane permeability was abolished. This suggests that this motif may be required for the peptide to be able to interact with lipid bilayers and cause membrane permeability and that deleting only a few residues may abolish peptide–membrane interaction.60

Important residues of hydrophobic peptides/proteins

Obviously, hydrophobic amino acids have been observed in the protein motifs involved in bonding with hydrophobic membranes. In particular, tryptophan, alanine, leucine, glycine and phenylalanine are commonly observed. Examples of proteins containing predominantly only one of these amino acids and of substitution of each type leading to attenuated membrane-binding and cytotoxic activity, demonstrate that no particular residue is required for membrane interaction but, rather, an accessible multiresidue region of predominantly hydrophobic groups would seem to make membrane interaction favourable. Tryptophan-rich hydrophobic proteins that cause membrane destabilization include tritrpticin, galanin, gramicidin, melittin, indolicidin and lactoferricin B, α-sarcin and EqtII (a eukaryotic cytolytic toxin).37,46,67,69 Glycine-based hydrophobic proteins that play an important role in membrane fusion and destabilization include prion, AβP, Tau, melittin and VP5* protein of rotavirus.57,58,60,78,79 Leucine, a good helix-capping residue, is often found at the N-terminal of α-helical antimicrobial peptides and plays an important role in stabilizing the membrane-bound protein structure.47 Alanine is another residue that is important in the hydrophobicity of a protein and has been implicated in the fusion and interaction of the protein with a lipid bilayer PrP, Tau and AβP.64,80,81 Combinations of these residues are commonly used to form synthetic hydrophobic peptides for bilayer interaction experiments and studies and these models mimic many of the observed membrane interactions of cytotoxic proteins, indicating that these residues do confer membrane binding and disruption properties on the proteins. For example, de Planque et al.73 used the sequence WALP, which is structurally related to gramicidin peptide, to form a hydrophobic peptide to study lipid/peptide hydrophobic mismatching. The synthetic peptides exhibited membrane insertion and disruption effects.

Mechanisms of blocking hydrophobic protein interaction with membranes

One possible mechanism of preventing initial interaction is to block the electrostatic interactions of unchaperoned proteins with membranes. Studies show that destabilization of the bilayer can be inhibited by the screening of positively charged amino acids from the negatively charged lipid headgroups, using isotonic concentrations of saline.53 In addition, more targeted agents could be used either to screen charge or, as in the case of pH changes, to alter the charge on the protein or the membrane. An alternative is to prevent the interaction of unchaperoned hydrophobic peptides with cellular membranes. This may be achieved by masking the hydrophobic regions with specific agents. Another approach involves diverting these peptides to interact with an alternative targeted protein or perhaps by using chaperones or gene-based mutation treatment to stabilize the structure of the protein, so that hydrophobic regions are folded properly and are not exposed. According to Calero et al.,82 apolipoprotein J (ApoJ) may be used as a defence against neuron damage caused by the aggregation and fibrillization of ΑβP in AD. Apolipoprotein J, also known as clusterin, is secreted by epithelial cells that line many organs, with the major site of secretion of ApoJ being in normal adult brains.83 Duguid et al.83 observed increased ApoJ secretion in neurons in several neurodegenerative diseases, such as AD, Downs' syndrome, scrapie and Pick's disease. The significance of ApoJ may also be indicated by the finding that its release in neurons is increased with ageing.84In vivo and in vitro experiments conducted by Calero et al.82 clearly indicate that secreted AβP (sAβP) is transported by lipoproteins in plasma and cerebrospinal fluid (CSF) and that ApoJ forms a stable complex with sAβP. By incubating highly fibrillogenic AβP with and without ApoJ, Calero et al.82 found that AβP aggregated and formed amyloid fibrils in the absence of ApoJ, but did not form fibrils in the presence of ApoJ. This suggests that the formation of a stable ApoJ/sAβP complex prevents aggregation and amyloid fibril formation by the toxic peptide.82In vitro experiments also indicated that ApoJ had a neuroprotective role in Alzheimer's disease, because it was able to fully protect cells against AβP toxicity. Calero et al.82 proposed that one of the mechanisms of neuroprotection by ApoJ could be due to the prevention of the partially folded proteins from interacting with and destabilizing cell membranes, by binding to the hydrophobic regions of the protein. These findings point to the possibility of developing similar complexing proteins to bind toxins in a way that lowers their cytotoxic activity.

Conclusions

Unchaperoned, intrinsic hydrophobic proteins are potentially pathogenic through causing autocytotoxicity. Living cells have used protein hydrophobicity for fusion of cellular membranes and manufacturing specialized transport pathways across membranes. Bacteria and viruses have also used this property, as a killing mechanism of the host cell. In turn, antibacterial organisms use this same property to kill bacterial cells. This mechanism of cytotoxicity is mediated by membrane damage that involves protein incorporation into the membrane and formation of non-specialized ion channels. The potency of the proteins acting via this mechanism of action is not dependent on the ability of these proteins to incorporate into the membrane and form a single type of ion channel but, rather, on the properties of a diverse population of non-intrinsic ion channels.

Because of the cytotoxic potential of hydrophobic proteins, cellular regulatory mechanisms have been evolved to allow the cell to cope with strayed, misfolded hydrophobic proteins. These mechanisms involve chaperoning of these potentially cytotoxic proteins for degradation, aggregation into aggresomes, fibrillization and internalization in cellular compartments to prevent them from interacting with membrane components. The underlying mechanism for the action of cytotoxic peptides is their properties of hydrophobicity, charged motifs and specific residues that confer on them an ability to interact with cellular membranes. The understanding of peptide–lipid interactions of strayed, hydrophobic proteins will enable the designing strategies for advanced drug delivery to overcome the deleterious effects. Specific blocking or screening of hydrophobic regions could remove the ability of such proteins to incorporate in the membrane and thus render them non-cytotoxic.

Acknowledgements

We thank Dr WL Armarego and Mr R McCart (The John Curtin School of Medical Research, The Australian National University, Canberra, ACT, Australia) for numerous discussions, suggestions and critical reading of the manuscript. The comments of Drs P Hardo (Benenden Hospital, Benenden-Cranbrook, UK) and M Bahadi-Hardo (William Harvey Hospital, Ashford, UK) on the manuscript are greatly appreciated. JIK is supported by the National Health and Medical Research Council of Australia (project grant numbers 970122 and 122808) and The Australian Research Council (grant numbers F99123, F0047 and F01008).

Ancillary