The oxidation of lipids has been shown to impact virtually all cellular processes. The paradigm has been that this involvement is due to interference with the functions of membrane-associated proteins. It is only recently that methodological advances in molecular-level detection and identification have begun to provide insights into oxidative lipid modification and its involvement in cell signaling as well as in major diseases and inflammation. Extensive evidence suggests a correlation between lipid peroxidation and degenerative neurological diseases such as Parkinson's and Alzheimer's, as well as type 2 diabetes and cancer. Despite the obvious relevance of understanding the molecular basis of the above ailments, the exact modes of action of oxidized lipids have remained elusive. In this minireview, we summarize recent findings on the biophysical characteristics of biomembranes following oxidative derivatization of their lipids, and how these altered properties are involved in both physiological processes and major pathological conditions. Lipid-bearing, oxidatively truncated and functionalized acyl chains are known to modify membrane bulk physical properties, such as thermal phase behavior, bilayer thickness, hydration and polarity profiles, as manifest in the altered structural dynamics of lipid bilayers, leading to augmented membrane permeability, fast lipid transbilayer diffusion (flip-flop), loss of lipid asymmetry (scrambling) and phase segregation (the formation of ‘rafts’). These changes, together with the generated reactive lipid derivatives, can be further expected to interfere with lipid–protein interactions, influencing metabolic pathways, causing inflammation, the execution phase in apoptosis and initiating pathological processes.
Oxidative stress is a major outcome of free radical-mediated injury, associated with apoptosis, inflammation and a number of pathological processes, such as age-related macular degeneration, Alzheimer's and Parkinson's diseases, prion disease, type 2 diabetes and cancer [1-5]. Reactive oxygen species (ROS) are formed during oxidative stress , and introduce a plethora of complex chemical modifications into all biomolecules, including proteins, nucleic acids and lipids . Changes in lipids, and phospholipids in particular, are of obvious importance when attempting to understand the consequences of oxidative stress on membrane-associated interactions and processes. In this context, phospholipids containing polyunsaturated fatty acyl chains are of specific interest because they yield a number of highly reactive derivatives upon oxidation. For example, mitochondria, the main organelles responsible for the generation of ROS [7, 8], harbor membranes containing major phospholipid constituents with polyunsaturated fatty acyl chains, such as tetra-linoleoyl-cardiolipin . This poses an immediate threat to the structural integrity of the mitochondrion, which is highly vulnerable to attack by ROS, for example, following an insult on its membrane structure by aggregated/misfolded protein oligomers composed of intracellular Alzheimer β-peptide (Aβ) .
Realization of the association between lipid peroxidation and the aforementioned pathological conditions prompted the first phase of studies on the effects of lipid peroxidation on the properties of model membranes. Most of these studies were performed using phospholipids isolated from cells and thus consisting of mixtures of lipids of varying chain lengths and degrees of acyl chain unsaturation. As a consequence, a complex mixture of oxidized lipids is formed upon peroxidation involving, in addition to the generation of oxidized lipid species, the formation of reactive small molecules such as malondialdehyde and 4-hydroxy-2-nonenal. Accordingly, although the results from this type of experiment represent to some extent the situation prevailing in real biomembranes exposed to oxidative stress, the exact molecular mechanisms involved are ambiguous because of poorly defined complex chemical structures and a lack of quantitative information. Nevertheless, a number of important results have been obtained and lipid oxidation has been shown to cause: (a) a loss of the membrane permeability barrier function ; (b) phospholipid phase transitions and phase separation ; (c) cross-linking of phospholipid head groups, such as ethanolamine ; and (d) augmented phospholipid flip-flop . The latter was explained by the involvement of nonbilayer phase structures . More recent studies have established altered conformational dynamics of the oxidatively modified acyl chains to be responsible .
This minireview deals mainly with effects caused by stable phospholipid oxidation products such as 1-palmitoyl-2-(9′-oxo-nonanoyl)-sn-glycero-3-phosphocholine (PoxnoPC) and 1-palmitoyl-2-azelaoyl-sn-glycero-3-phosphocholine (PazePC) (Fig. 1), for which biophysical data are now emerging and which have also been shown to exert activity in cellular studies [16-18]. We focus on the impact of oxidized phospholipids on the biophysical properties of lipid membranes and discuss molecular-level events, with the aim of identifying causative pathways that are manifest in physiological processes (most notably apoptosis) and lead to relevant pathological conditions. For more extensive discussions, the reader is referred to recent work from our laboratory [4, 19].
Membrane polarity and permeability
Oxidative derivatization of phospholipids, containing polyunsaturated fatty acyl chains, dramatically alters their amphiphilicity by incorporating polar moieties such as the highly reactive terminal aldehyde and carboxylic groups in the truncated sn-2 acyl chain. Accommodating these polar moieties within the low dielectric hydrocarbon phase would be energetically costly and, as a consequence, lipids such as those depicted in Fig. 1 will tend to adopt the extended conformation [20, 21]. This conformation is based on rotation of the C(1)–C(2) bond of the glycerol backbone, which results in flipping of the oxidized sn-2 chain, maximally by 180°, to extend into the aqueous phase. Recent monolayer measurements suggest that for a large fraction of the oxidized lipid the angle is ~ 90°, with the truncated chain parallel to the membrane plane (Fig. 2). In keeping with the dynamics of the bilayer lipids, the oxidatively modified chains fluctuate between aligned and extended conformations [15, 22, 23]. These fluctuations and altered biophysical properties were subsequently verified using computer simulations [24-26]. Phospholipids in the extended conformation nominated as ‘whiskers’ were also identified by NMR and were suggested as being adopted by phospholipids on the surface of cell membranes . One important and physiologically relevant consequence of phospholipid oxidative deterioration is that the bilayer polarity profile becomes modified so that it no longer forms a barrier to the penetration of water and accommodation of polar lipid head groups in the membrane interior [14, 15]. Fluorescence spectroscopic studies that utilized the dithionite assay [28, 29] to follow the transbilayer lipid distribution showed that the flip-flop rate of the nitro-2,1,3-benzoxadiaz-ol-4-yl-labeled phosphatidylserine analog in liposomes is greatly enhanced in the presence of oxidatively modified lipids . Thus, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) liposomes showed almost no flip-flop of the fluorescent phosphatidylserine (t½ > 2 weeks), in keeping with results from neutron scattering , however augmented flip-flop was seen (t½ ~ tens of minutes) for POPC liposomes containing up to 16 mol% of PazePC or PoxnoPC (Fig. 3). Interestingly, flip-flop rates did depend on membrane curvature and degree of oxidation. Faster transbilayer diffusion was measured in small unilamellar vesicles than large unilamellar vesicles, although PoxnoPC, having the less polar aldehyde moiety, resulted in slower diffusion in large unilamellar vesicles compared with that observed for PazePC. This is likely to reflect differences in conformational fluctuations of the oxanonanoyl and azeloyl chains. Apparently, the higher efficiency of PazePC in decreasing the energy barrier for flip-flop in large unilamellar vesicles, as characterized by tighter head group packing, is primarily associated with its augmented capability to adopt an extended conformation in order to decrease the free-energy penalty for accommodating the larger and partially charged terminal carboxylic group of azelaoyl chain into the hydrophobic bilayer interior.
Higher amounts of oxidized lipids further lead to a complete deterioration of the membrane permeability barrier. An increase in the PazePC or PoxnoPC content to > 20 mol% was shown to vastly augment dithionite permeability, which might indicate pore formation . This characteristic of oxidized lipids has been suggested to be involved in augmenting plasma membrane susceptibility to electropermeabilization, used to introduce plasmids and pharmaceutical agents into live cells . Taken together, these results demonstrate that lipid oxidation without involvement of proteins can significantly alter the integrity and permeability barrier properties of membranes.
Phase segregation and transition
Peroxidation alters the chemical structure and reactivity of lipids and also revamps their amphiphilic properties, which are crucial to their interfacial behavior, lipid–lipid interactions and self-assembly. Monolayer experiments have shown that oxidized phospholipid (oxPL) phases are characterized by a significant increase in average molecular area and decrease in surface potential, perhaps reflecting the extended conformation adopted by oxPLs [15, 22]. More specifically, PazePC occupies > 35% larger molecular area than its unoxidized analog, POPC, as seen in compression isotherms (Fig. 3A). The presence of a polar terminal group in the sn-2 acyl chain increases the water solubility and critical micelle concentration of oxPL . This property is also reflected in compression isotherms of PazePC as a sharp inflection followed by a plateau region at 38 mN·m−1, which is most probably caused by loss of the oxidized lipid from the monolayer into the aqueous phase (Fig. 4). Examination of monolayer electric properties revealed that the surface potential (ψ) of pure PazePC and PoxnoPC monolayers is generally lower than for unoxidized phospholipids, and is independent of the mean molecular area, indicating that the dipole moment vector is mostly oriented parallel to the monolayer surface [22, 33]. Of particular interest to this minireview are ternary lipid mixtures used to model so-called lipid ‘rafts’. These mixtures usually consist of a low main-transition temperature lipid (e.g. POPC), high transition temperature lipid (e.g. 1,2-dipalmitoyl-sn-glycero-3-phosphatidylcholine, sphingomyelin) and cholesterol, which segregate into a low transition temperature lipid-enriched liquid-disordered phase and a high transition temperature lipid and cholesterol-enriched liquid-ordered phase [34-36]. Monolayer compression isotherms have shown that coexistence of the liquid-disordered/liquid-ordered phase is surface pressure dependent and upon compression the components undergo a miscibility transition to produce one homogeneous phase  (Fig. 5A,B). In vesicles, this type of mixture forms micron-scale domains of immiscible liquid phases for a limited range of compositions. Intriguingly, oxidation has been found to stabilize phase segregation, preventing the miscibility transition in monolayers and vesicle systems [37-39]. Nanoscale analysis of mixed lipid monolayers revealed that oxygen-induced lipid oxidation can cause condensed domain-size evolution at surface pressures well above the expected miscibility transition . Similarly, the formation of large domains was observed in giant unilamellar vesicles upon their photo-oxidation and production by electroswelling . However, the chemically poorly defined oxidation techniques resulting in truly complex membrane assemblies impose ambiguity regarding the exact nature and extent of the chemical modification, making elucidation of the mechanisms underlying the observed phase segregation inaccessible. We recently studied the biophysical properties of the POPC/brain spingomyelin/cholesterol (1.5 : 1.5 : 1 molar ratio) monolayers upon introduction of a specific oxidatively modified phospholipid, PazePC . It was found that a gradual exchange of POPC to this oxidized analog postponed the miscibility transition in a concentration-dependent manner (Fig. 5C) and completely abolished it above a critical PazePC content (~ 25 mol% of the total phosphatidylcholine, corresponding to 9.4 mol% of the total lipid). The observed effect of oxidized lipid on miscibility transition might be assigned to enhanced hydrophobic mismatch, governing line tension at the boundary between liquid-ordered and liquid-disordered phases [43-45]. Accordingly, chemical modification caused by oxidative damage, more specifically oxidative truncation of the polyunsaturated sn-2 acyl chain in phosphatidylcholine, can be readily expected to promote thinning of the liquid-disordered phase in a concentration-dependent manner, as also evident from molecular dynamics simulations and X-ray diffraction studies [24, 31]. More specifically, molecular dynamics calculations suggest that incorporating oxidized lipids and peroxidation products of linoleic acid into phospholipid bilayers increases the area per molecule and decreases bilayer thickness. In keeping with the above, X-ray diffraction showed a marked reduction (from 36 to 32 Å) in the hydrocarbon core width of dilinoleoyl phosphocholine bilayers upon Fe2+/ascorbate-induced peroxidation of ~ 14.5% of the total lipid, and a decrease in overall membrane thickness, including surface hydration, from 48.7 to 44.6 Å . Similarly, a 4 Å reduction in hydrocarbon core width from 40 to 36 Å was observed in membranes reconstituted from bovine cardiac phosphatidylcholine, with only 1.1% of the available polyunsaturated fatty acyl chains being peroxidized. In addition, alterations in intermolecular packing following oxidative stress have been shown to facilitate interdigitation of terminal methyl segments, which favors a further decrease in membrane thickness. To this end, the available data suggest significant modification of lipid membrane systems upon exposure to oxidative stress, which is expected to have profound effects on the 2D and 3D organization of membrane constituents and their associated processes.
Biological and pathological concequences
Propagation of lipid peroxidation is a chain reaction, spreading rapidly to adjacent membranes and vicinal cells , and unlike most biochemical modifications of cellular constituents this process cannot be controlled by proteins . One exception to this appears to be the generation of oxidized phosphatidylserine, oxidized phosphatidylinositol and oxidized cardiolipin by cytochrome c , the reaction most likely being controlled by electrostatics between these acidic lipids and the cationic domains of cytochrome c. One of the consequences of the peroxidation of mitochondrial cardiolipin (cardiolipin → oxidized cardiolipin) is the release of cytochrome c from mitochondrial membranes . The latter has been assigned to oxidized cardiolipin being unable to maintain the cytochrome c–cardiolipin membrane interaction by the extended lipid anchorage , which provides hydrophobic interaction between membrane-associated cytochrome c and the lipid bilayer, without intercalation of the protein into the bilayer [52, 53]. Release of cytochrome c from mitochondria, in itself one of the hallmarks of apoptosis , has been shown to induce the apoptotic proteolytic caspase cascade [55, 56]. Oxidized mitochondrial membrane lipids have been implicated in mitochondrial permeability transition [57, 58], primarily associated with membrane depolarization and uncontrolled release of antioxidants (e.g. glutathione) and cytochrome c to the cytosol .
Another generic attribute of apoptosis is dissipation of the membrane lipid asymmetry and exposure of phosphatidylserine on the outer surface of the plasma membrane [60, 61]. Randomization of interleaflet lipid distribution is usually assigned to ATP-dependent, nonspecific ‘scramblases’ , however, positive identification of the scramblase and the triggering mechanisms remain perplexing [63, 64]. Notably, as outlined above, the presence of oxidized lipids has a profound effect on the transbilayer diffusion (flip-flop) of membrane lipids, accelerating this process to physiologically relevant rates . These findings are in agreement with the effects of PazePC on cultured cells and isolated mitochondria, similar to the 1-alkyl analog 1-hexadecyl-2-azelaoyl-sn-glycero-3-phosphocholine which is resistant to hydrolysis by phospholipase A1 . Addition of micromolar concentrations of these lipids to cultured cells causes apoptosis with rapid nonreceptor-dependent exposure of phosphatidylserine. Furthermore, these lipids become concentrated in the mitochondria and have been shown to depolarize isolated mitochondria and cause permeability transition assisted by the proapoptotic Bid and counteracted by the antiapoptotic Bcl-XL . These effects are reversible upon scavenging of these oxidized lipids by BSA. Similarly, investigation of proapoptotic protein Bax association with liposomes mimicking mitochondrial outer membrane revealed that the presence of PazePC-like oxPL facilitated the insertion of Bax into the membrane, distorting the bilayer's organization . Along these lines, Kagan and coworkers demonstrated an accumulation of oxidized phosphatidylserine in the plasma membrane of cultured cells and stimulated scrambling of nonoxidized phosphatidylserine and phosphatidylethanolamine . Because no inhibition of aminophospholipid translocase by oxidized phosphatidylserine was detected, the authors suggested that this lipid acts in a nonenzymatic fashion. Taken together, these findings indicate the potential physiological significance of oxidative stress in apoptosis and cancer, both of which involve a loss of membrane lipid asymmetry, with exposure of phosphatidylserine on the outer surface of the cell plasma membrane.
Several lines of research have concluded that lipid membranes induce the efficient formation of amyloid-type fibers by a number of cytotoxic proteins associated with major degenerative diseases [68, 69] and proteins involved in physiological processes such as apoptosis and innate immunity . Initially, augmented amyloid formation by membranes was associated with negatively charged lipids and their role was readily explained in terms of concentrating cationic peptides at the membrane surface, orienting the attached peptides and neutralization of their cationic charges, allowing for peptide–peptide approach and interactions in the membrane plane, the latter leading to efficient hydrogen bonding promoted by the low dielectric (hydrophobic) membrane environment [71, 72]. More recent studies have further shown that oxidized lipids promote amyloid formation, with major alterations in the aggregation/misfolding free-energy landscape . Thus, accelerated amyloid fibril formation was evident for Aβ peptide in the presence of oxidatively modified lipids and their soluble fragmentation products [73-76]. Several reports have outlined the potential for a complex interplay between cholesterol oxidation products and Aβ , and Schiff's base formation by Aβ and aldehyde-bearing lipid products has been shown to increase Aβ amyloidogenicity . It is becoming increasingly evident that lipid peroxidation is also significantly augmented in another neurodegenerative disorder, Parkinson's disease, associated with the formation of cytoplasmic protein aggregates, called Lewy bodies . Although the molecular mechanisms associated with neurodegeneration in Parkinson's disease remain poorly understood, α-synuclein, the major component of Lewy bodies [79, 80], has been suggested to have an affinity for oxPL-containing membranes. Findings by Zhu et al. suggest that inhibition of lipid oxidation by α-synuclein may represent a physiological function of the protein . Alternatively, association of α-synuclein with oxidized lipid metabolites has been proposed to cause mitochondrial dysfunction, in turn leading to dopaminergic neuron death and Parkinson's disease . Augmented oxidative stress is also evident in type 2 diabetes . Middle body obesity is a known risk factor for type 2 diabetes , with slow fractional lipid turnover of the abdominal fat necessarily increasing lipid peroxidation  and subsequent transfer and equilibrium of oxidized lipids through the plasma compartment to all cells . Accordingly, long-term accumulation and presence of oxidized lipids in the Langerhans β-cell outer surface would make these cells vulnerable to amyloidogenic attack by islet-associated polypeptide, present in high concentrations on the surface of the β cells following secretion of this peptide by these cells. Our own studies on islet-associated polypeptide revealed that although an acidic-phospholipid-containing bilayer could rapidly induce islet-associated polypeptide aggregation into mature amyloid fibrils , more complex processes were evident in bilayers containing the oxidized phospholipid PazePC . Specifically, our data suggest that oxPL significantly alter the free-energy landscape for the oligomerization of amyloid-forming proteins, opening new assembly pathways and prolonging the lifetimes of intermediate oligomer species existing during the assembly of peptide monomers into amyloid beta-sheet fibrils . This might underlie the cellular toxicity of amyloidogenic peptides and proteins under pathological conditions with associated oxidative stress .
Concluding remarks and perspectives
Conventionally, lipids involved in cellular signaling are assumed to act via specific interactions with proteins, such as platelet-activating factor, which represents an agonist for the typical seven-transmembrane helix G-protein-coupled platelet-activating factor receptor. Yet, very different membrane-mediated effects on proteins can also be depicted. As a simple example, reflecting the dependence of phospholipase A2 action on phospholipid lateral packing density allows osmotic stretching of lipid bilayers to change the lateral packing in a manner that makes membrane insertion of the phospholipase A2 possible, giving access for the phospholipid substrates into the active site of phospholipase A2. Accordingly, the physical state of the substrate converts to the action of an enzyme, a biochemical signal, readily demonstrating that phospholipase A2 is a mechanosensitive protein . Recognizing the importance of the membrane lipid physical (phase) state to the lateral organization of all membrane constitutes, it was postulated that lipids may control the physiological state of a membrane organelle by modifying its physical state . Along these lines, we previously showed that the formation of ceramide by sphingomyelinase results in an extensive phase separation, analogous to the so-called capping in cells by ligands such as concanavalin A [89, 90]. Ceramide formation was further shown to induce both 2D and 3D changes in membrane organization . Lipid oxidation and the introduction of oxidized lipids provide another, more dramatic, example of this principle. Notably, lipid oxidation manifests as loss of membrane phosphatidylserine asymmetry, previously assumed to require enzyme activity by ‘scramblase’. However, recent biophysical studies have demonstrate that the mere presence of oxPL in membranes is sufficient, alleviating the need for a scramblase enzyme.
Lipid peroxidation has been the subject of extensive studies for several decades, however the large diversity of oxidation-inducing factors (e.g. enzymes, ROS) and types of reactions have made elucidation of the specific roles of the end products a formidable challenge [47, 92]. Recent renewed interest in lipid peroxidation stems from findings of the involvement of oxPL in apoptosis, inflammation and several pathological conditions. Accordingly, inflammation and systemic oxidative stress being hallmarks of Alzheimer's disease, Parkinson's disease and cancer may shed light on the underlying molecular-level mechanisms of these and related diseases. For example, as pointed out recently , the apparent protective effects of Parkinson's disease and Alzheimer's disease against some cancers [94, 95] potentially evolve from augmented targeting of host defense proteins/peptides to cancer cells with oxPL-containing membranes. This also explains why administration of an antioxidant such as vitamin E can result in an increased incidence of cancer , with the antioxidant diminishing oxidative stress and thus reducing the exposure of oxPL on the surface of cancer cells, attenuating the efficiency of host defense proteins/peptides in eradicating these cells . In light of the above, lipid peroxidation can be considered as a signaling mechanism, initiated with oxidative lipid derivatization, which further translates into a change in membrane biophysical properties, followed by extensive modulation of lipid–protein interactions and finally alteration of biological processes on the whole-organism level.
Last, but not least, in this context, it is of interest that most anticancer drugs, such as adriamycin, cisplatin and taxol, as well as γ-radiation, all induce oxidative stress [97-100]. Accordingly, it seems feasible that their cytotoxic action involves the formation oxidized phospholipids, which then cause cell death. To this end, of all the structural components of a cell the only ones which are readily amenable to dietary modification are the lipid acyl chains. This scenario readily complies with the observed sensitization to anticancer drugs of cancer cells by dietary polyunsaturated fatty acyl  and also suggests that antioxidants such as vitamin E should be excluded from the diet of patients undergoing chemotherapy or γ-radiation therapy. Along the same lines, hyperbaric oxygen has been shown to enhance the efficiency of anticancer drugs . Taking into account the dramatic increase in the oxygenation of plasma, from 0.3 to 6% achieved by increasing the pressure from 1 to 3 atm, a drastic increase in the local concentration of ROS in tumors can be expected. The critical role of oxygen availability in cell killing  is in keeping with the key role of ROS and ROS-generated products in the execution phase of apoptosis.
This study was financed by the Finnish Academy ESF EUROMEMBRANE project oxPL, MEM/09/E006, BECS and Sigrid Juselius Foundation.