Billions of years of evolution created and optimized mechanisms for efficient translation of genomic information into thousands of finely tuned protein machines (1) and perfected functional interactions between the proteins through sophisticated multi-leveled signaling systems (2). Lipids of biological membranes constitute a critical part of this complicated signaling network (3). The metabolic coordination requires that the flow of signaling information proceeds with optimized levels of fidelity and speed. Conservative estimates indicate that the number of proteins in the human proteome is on the order of 10⁵-10⁶ (4, 5). Thus it is not surprising that the diversity of signaling lipids coordinating multiple protein-protein and lipid-protein interactions within and between subcellular organelles, cells, and tissues may be even greater resulting in the possible >10⁶ of individual molecular species in the lipidome. Engagement of membrane phospholipids (PLs) in the signaling process occurs via their biochemical modifications leading to the appearance of small amounts of “unusual” PL molecules such as their hydrolysis or peroxidation products (6). This review is focused on oxidatively modified (phospho)lipids as the signaling entities. Among them are well known lipid mediators represented by oxygenated free polyunsaturated fatty acids (PUFA) as well as oxygenated PUFA esterified into different classes of membrane phospholipids (PLs). The latter group will be the subject of the current review.

Cultural beliefs of successful societies have led to the common opinion that suicidal elimination is not the necessary way to resolve life conflicts that may encompass transient dark episodes within an otherwise bright present and even more wonderful future. This optimistic view has been expressed in many statements by politicians, writers and other artistic celebrities (including those by W. Churchill and E. Hemingway quoted above). On the molecular level, however, the ruthlessness of life/death elimination decisions is frequently a necessary attribute of the high fidelity of cell populations and their adaptive adjustments. The suicidal programs of cell death are genetically pre-determined and deciphering their specific mechanisms represents one of the emerging fields of cell biology. The lipid-derived signals may act within a given cell (autocrine signaling), affect cells within the surrounding neighborhood (paracrine signaling) or act on remote targets using the circulatory system for the transportation of death signals (endocrine signaling). Currently, more than a dozen regulated death programs have been identified in cells that accumulate excessive amounts of the (geno)toxic materials and hence are recognized by the surveillance machinery as irreparably damaged. It is believed, but not proven, that peroxidation of polyunsaturated lipids (PUFA-lipids) has been associated with the initiation and execution of many, if not all, of these programs (7). In spite of these general associations, neither the specific roles of peroxidized PLs in the fulfillment of the programs nor their chemical identity have been identified. Notable exceptions are apoptosis and ferroptosis, two programs for which the progress in redox lipidomics has resulted in the deciphering of death signals.

The structural core of biological membranes is formed by the bilayer of PLs—amphipathic molecules with long lipophilic hydrocarbon chains and water-soluble polar heads. The hydrophobic/hydrophilic balance of phospholipids dictates the organization of the bilayer in which lipophilic chains interact with each other while the polar head-groups are localized at interface with the aqueous phase. In PLs, the chains represent fatty acids covalently attached to two sites of the glycerol backbone whereas the third position is occupied by phospho-base that may be a non-charged zwitter-ion (when the negative charge of the phosphate is compensated by a positive charge) or carry a negative charge (when an extra negatively charged group is present). Fatty acyls of PLs may have no double bonds (saturated) or contain one (mono-unsaturated) or several methylene-interrupted double bonds [polyunsaturated PLs (PUFA-PL)]. While unsaturated PLs may be biosynthesized both in anaerobic and aerobic conditions, the huge diversity of PUFA-PLs is characteristic for different domains of aerobic life (8).

One of the most popular concepts explaining the presence of diversified PUFA-PLs in the lipid bilayer relates to their function as a regulator of membrane fluidity necessary for the rapid diffusion and conformational flexibility of membrane proteins (9). In spite of its attractive simplicity, this concept does not explain the huge molecular variety of PUFA-PLs. Indeed, contemporary lipidomics detects 10³–10⁴ individual molecular species of major classes of phospholipids in cells and tissues. This is a conservative estimate of the species with differing masses. Indeed, with the two acyls/PL molecule and a menu of >30 commonly found fatty acids the number of possible isomeric species of “two-legged” PLs should be close to 10³. However, for “three-legged” tri-glycerides this estimate would yield 10⁴ species and for “four-legged” mitochondrial cardiolipins (CL) – >10⁵ molecular species.

One of the prominent features of PUFA-PLs is their susceptibility to peroxidation via free radical mechanisms (10). These mechanisms may be comprised of enzymatic systems for the activation of oxygen and/or lipid substrates or occur non-enzymatically (see below). As the general schema of peroxidation includes abstraction of hydrogen from bis-allylic positions, PLs with multiple (four-six) double-bonds, are preferred substrates, particularly for non-enzymatic free radical reactions (10, 11). The primary product of the peroxidation process generates hydroperoxy-PLs (HOO-PLs). These products are not stable and readily undergo secondary decay reactions leading to a variety of electrophilic aldehydic-, keto-, hydroxy-derivatives as well as cyclic compounds (10). In terms of lipid diversification, this adds another order of magnitude to the possible number of individual phospholipids, thus bringing it to 10⁵ species. It should be noted, however, that the measurable amounts of peroxidized PLs in healthy cells and tissues is markedly lower than their non-oxidized parent PLs (12). This is partly due to the fact that only a fraction of PUFA-PLs are involved in peroxidation and also to the high reactivity of the secondary electrophilic decay products toward nucleophilic sites in proteins (13). As a result, the life-time of these products in “free” form may be relatively short as they form lipid-protein adducts. However, the levels of these products may increase many-fold in conditions associated with cell injury and death. These reactive secondary intermediates of PUFA-PL peroxidation represent the proximate entities affecting functions of numerous proteins (14). Given that the formation of these adducts is, in a way, a reaction of protein “lipidation” that may dramatically change the distribution and functional characteristics of the affected proteins, it has been hypothesized, although not proven, that electrophilic products of PL peroxidation and their adducts with specific proteins represent the proximate “gateways” of cell’s demise in regulated cell death programs.

Since the time of the first detailed description of regulated cell death almost five decades ago, about a dozen different programs have been identified (15). The best described programs include apoptosis, necroptosis, pyroptosis, ferroptosis, entotic cell death, netotic cell death, parthanatos, lysosome-dependent cell death, autophagy-dependent cell death, alkaliptosis, and oxeiptosis (15). The majority of them have been qualified as responses to different types of stresses causing irreversible changes not only to one particular cell but also representing a high-risk threat to the entire community of surrounding cells. Among the different causative factors, oxidative stress has been universally identified as one of the leading mechanisms engaged early at the initiation or later during the execution stages of the death programs (15) ( Table 1 ).

Given the vague definition of what exactly “oxidative stress” means, attempts have been made to connect the death programs with specific pathways and manifestations of the aberrant redox metabolism. Due to the high sensitivity of PUFA-PL to oxidative modifications, lipid peroxidation (LPO) has been considered as one of the common denominators of programmed cell death (57, 58). However, the specific role and mechanisms of LPO in the pathways leading to cell demise remain poorly defined. This is due, to a large extent, to difficulties in the analysis of highly diversified and very low abundance LPO products (12). These technological problems were resolved with the advent of high-resolution liquid-chromatography-mass spectrometry (LC-MS) based redox lipidomics (6, 12) with its capability to detect, identify and quantitatively characterize a variety of PL oxidation products. While the application of this technology may provide important information in any of the known death pathways, so far the significant results have been obtained mostly for two death programs—apoptosis and ferroptosis (16, 21, 34, 36).

Numerous studies of PL peroxidation in model chemical and biochemical systems using different initiating agents (e.g., azo-initiators of peroxyl radicals, Fe-ascorbate dependent generators of HO• radicals) demonstrated that the susceptibility to oxidative modification is largely defined by the number of double bonds in the fatty acid residues (59). As a result, PLs with hexa-, penta-, and tetra-enoyl residues are the predominant peroxidation substrates as compared to doubly or triply unsaturated PLs. Notably, the nature of the polar part of the PL molecules was not influential as a factor determining vulnerability to oxidation. In sharp contrast, redox lipidomics studies of programmed death associated peroxidation has determined that there is a high selectivity of the process toward specific classes of PLs. Execution of the intrinsic apoptotic program revealed a high selectivity toward peroxidation of mitochondrial CLs whereby the species with C18:2 represented the major substrates (16, 21). Notably, CL molecular species containing more PUFA residues remained non-oxidized. Moreover, in hetero-acylated CL species, oxidation of C18:2 residues occurred preferentially even when C22:5 and C22:6 residues remained non-oxidized within the same molecule ( Figure 1 ).

Ferroptosis-associated LPO was also highly selective toward a specific PL class—PUFA-containing phosphatidylethanolamines (PE) (34). Interestingly, two types of PE-molecular species with C20:4 and C22:4 displayed the highest sensitivity toward oxidative modification. In terms of positional specificity, the 15^th position in C20:4 and the 17^th position in C22:4 were the preferred oxidation sites. Importantly, these PE oxidation products exerted predictive features of ferroptosis biomarkers and displayed pro-ferroptotic activity upon co-incubations with target cells (34). A recent study demonstrated that not only di-acyl-phospholipids but also PUFA-plasmalogens (ether-phospholipids), synthesized in peroxisomes, underwent peroxidation in ferroptosis (60). Downregulation of ether phospholipids was associated with the increased resistance of cancer (carcinoma) cells to ferroptosis in vivo.

Selectivity and specificity of the PL peroxidation process in two different cases of apoptotic and ferroptotic (non-apoptotic) regulated cell death suggest a possible involvement of enzymatic catalytic mechanisms. Indeed, two different metalloproteins, a hemoprotein cytochrome c (cyt c) ( Figure 2A ) and a non-heme Fe-protein, 15-lipoxygenase (15LOX) ( Figure 2B ), have been identified as the highly likely enzymes initiating the peroxidation process in apoptosis and ferroptosis, respectively (16, 34). In both cases, the selectivity of the enzymatic peroxidation mechanisms is achieved due to the formation of lipid-protein or protein-protein complexes as described below.

PUFA residues of lipids are believed to be highly susceptible to oxidative modification by oxygen (10, 64). This process, LPO, proceeds via the formation of radical intermediates. The rate limiting stage is the initial formation of radicals that can further propagate the overall process. Therefore, the peroxidation rate is very low in the absence of catalysts. While there are many different radical initiators—physical factors like irradiation or chemical agents, like compounds spontaneously decomposing to form carbon-centered radicals—the most important biological peroxidation catalysts are transition metals, particularly iron (Fe) (65). Therefore, the levels of redox active free Fe-ions or “loosely bound” Fe in low molecular weight complexes are strictly controlled in cells and biological fluids ( Figure 3A ). Catalytic Fe of active enzymes is regulated by the protein structure. Fe required for these catalytic functions is delivered to the respective protein clients—Fe-sulfur proteins, hemoproteins, and non-heme Fe-proteins—by several specialized protein chaperons (66). Quantitatively, hemoproteins represent the most abundant endogenous source of Fe and catabolic degradation of these proteins accompanied by the release of ferrous ions (Fe²⁺) is operated by heme oxygenases (67). Specialized ferroxidases convert (Fe²⁺) to (Fe³⁺)—the form suitable for the intracellular iron storage by ferritin ( Figure 3 ).

Normally regulated Fe-catalyzed reactions are enzymatic. However, under special conditions such as Fe-overload, excessive catabolism of Fe-proteins or diseases, aberrant non-enzymatic Fe-driven lipid peroxidation by poorly controlled low molecular Fe-complexes may occur, becoming excessive or even overwhelming (68). These reactions commonly proceed via the production of reactive oxygen species (ROS), particularly HO• radicals, generated in Fenton/Haber-Weiss reactions or via the formation of lipid radicals generated during decomposition of organic hydro peroxides, including lipid hydroperoxides (69) ( Figure 3B ). As cell death programs are based on regulated mechanisms, the associated LPO is initiated by a selective and specific enzymatic process ( Figures 3C, D ). Interestingly, the enzymes involved in the production of lipid death signals via peroxidation mechanisms during apoptosis and ferroptosis—cyt c and 15LOX—usually are involved in different biological functions. For example, cyt c is shuttling electrons between mitochondrial respiratory complexes III and IV in the intermembrane space (70). 15LOX is a dioxygenase catalyzing the formation of oxygenated lipid mediators from free PUFA, particularly free arachidonic acid [AA or eicosatetraenoic (ETE) acid 20:4] (71). Upon the initiation of the cell death program, these enzymes change their properties/functions and switch their activity to the peroxidation of PLs. The transformation of enzymatic activity occurs due to protein interaction with other molecules—a mitochondria-specific PL molecule, CL, in the case of cyt c ( Figure 2C ), and the protein PEBP1 in the case of 15LOX ( Figure 2D ).

LPO—enzymatic or non-enzymatic—has a radical-mediated reaction in its nature. Radical intermediates are very short lived and cannot be directly detected by conventional high resolution analytical protocols such as LC-MS. The primary molecular products of LPO are hydroperoxides and they represent the first opportunity for the LC-MS based quantitative characterization. The analysis of lipid hydroperoxides has become a formidable task for several reasons. First, their chemical and metabolic instability and thermolabile nature most often results in the formation of secondary products, some of which are susceptible to degradation. Second, the sheer number of lipid signaling molecules is staggering which translates directly into a plethora of potential LPO products that, in many circumstances, occur at very low levels. Finally, while lipid hydroperoxides are the initial/primary products of lipid oxidation, secondary products such as aldehyde, ketone, hydroxy, and epoxide products add to the heterogeneity and complexity of the signaling language. Thus, this rich signaling language should include not only full-length peroxidation products, but also truncated PLs, cyclized PLs as well as fragments of oxidized fatty acyl chains resulting from secondary reactions of lipid hydroperoxides (34, 105).

Lipid hydroperoxides are present in very low steady-state concentrations and are unstable due to their cleavage to yield, dependent on the reducing or oxidizing environment, new alkoxyl- or peroxyl- radicals, which readily decompose into secondary products (34). Indeed, many studies have focused on small reactive lipid fragments, such as malondialdehyde, 4-hydroxynonenal, etc. that can act as secondary downstream reactive products in a variety of cell death mechanisms and can covalently modify proteins [3]. However, this will depend on whether the reactive group (a reactive aldehyde for example) resides with the “leaving” short lipid fragment or remains with the truncated parent PL. Either of these two groups of reactive products can potentially react with nucleophilic amino acids, such as histidine, cysteine, and lysine that reside in cellular proteins. It has been well established that the process of PL peroxidation generates electrophilic species that are able to modify proteins and change their structure, activity and function (106). As PEs are preferentially peroxidized during ferroptotic death, it is likely that the truncated reactive parent PL is necessary for driving the downstream effects of ferroptosis ( Figure 5 ). Indeed, small reactive lipid fragments can be formed from any polyunsaturated (phospho)lipid class, hence will not be specific for ferroptosis. By forming conjugates with a protein, lipidation by truncated reactive parent PLs will undoubtedly change the hydrophobic-hydrophilic balance, likely changing the distribution of the proteins into membranes (107–109) where they can form dreadful oligomeric pores.

Peroxidation reactions catalyzed by cyt c/CL complexes yield a highly diversified set of oxidized CL species with hydroperoxy-, hydroxy-, epoxy-, and oxo-functionalities (110). These CL peroxidation products are similar to the CLox signals detected in cells during execution of the intrinsic mitochondria-mediated apoptotic program triggered in cells by actinomycin D (16), staurosporine (17), and ionizing radiation (22, 111). Given that oxidizable PUFA-CLs are found exclusively in mitochondria, it is not surprising that oxidatively modified CL species, particularly mono-oxygenated C18:2-containing CLs, have been associated with the execution of the apoptotic death program (21). It should be noted, however, that a variety of CLox species, including hydroperoxy-, epoxy-, and oxo-CLs are detectable in cells and tissues in vitro and in vivo. For instance, CLox containing hydroxy-, epoxy-, oxo-, and hydroperoxy-functionalities have been detected in the small intestine of mice exposed to total body irradiation (110, 112, 113). Similarly, hydroxy- and hydroperoxy-CL species were detected in the brain of mice after traumatic injury (114). Notably, prevention of apoptosis by a mitochondrial electron acceptor, XJB-125, was associated with decreased levels of CLox in the peri-contusional zone of the traumatized brain.

It is possible that oxidatively truncated CLox species with high electrophilic potential can be formed in mitochondria. These products can readily attack nucleophilic sites in proteins and form protein adducts that will affect mitochondrial function and damage their integrity (115). Oxidatively truncated CL species can be formed in cyt c/CL catalyzed reactions in vitro ( Figure 6 ) and were detected in vivo in the ileum of mice exposed to total body irradiation ( Figure 7 ). While it is possible that truncated CLox products can modify a number of mitochondrial proteins, including those involved in the execution of mitochondria-dependent cell death programs (e.g., apoptosis), this work has not been accomplished and represents a goal for future investigations. The isolation and determination of the site of lipidation on proteins directly participating in the execution of cell death in apoptosis and ferroptosis is a formidable task. However, lipidated proteins, and more importantly lipidated peptides generated from protein digests, should impart a dominant hydrophobic characteristic whereby the peptides should be highly retained on reverse-phase solid supports. Adjusting and modifying gradients for extremely hydrophobic species and/or applying different chromatographic solid supports should aid in the separation and identification of lipidated species.

The oxidation of CL during apoptosis occurs very early during the time when its molecular reactions leading to the initiation of apoptosis remain located in a restricted space inside of mitochondria whereby its inhibition provides an effective target for preventing apoptosis. This suggestion was confirmed by the experimental results demonstrating, that genetic depletion of CL-synthase as well as deficiency of cyt c or mutation of its Tyr₆₇ residue result in an increased resistance of cells to apoptosis (20, 151–153). Several low molecular weight compounds preventing CL oxidation by inhibiting of peroxidase activity of cyt c have been designed and tested, including mitochondria-targeted electron scavengers and stable nitroxides radicals (116, 154–156).

Regulation of LPO which play crucial roles in the development of ferroptosis is a key approach in preventing ferroptotic cell death. One of the most important defenders is glutathione peroxidase 4 (Gpx4), converting toxic PL hydroperoxides to non-toxic alcohols by using GSH as the reducing substrate (142). Direct inactivation of GPX4 by small molecular weight compounds interacting with selenocysteine such as RSL3, ML162, ML210, FINO₂ as well as indirect inhibition by deprivation of GSH were effective in inducing ferroptosis (28, 126, 157) ( Table 1 ). One of the prerequisites for ferroptosis is the presence of ETE-PE, the substrate for the production of the death signal, HpETE-PE. Enzymes participating in the synthesis of ETE-PE such as ACSL4 responsible for the formation of CoA-derivatives of ETE and LPCAT3 catalyzing the esterification of CoA-ETE into lyso-PE, act upstream of GPX4 and are important ferroptosis regulators. Experimental data demonstrated a direct correlation between their expression and sensitivity to ferroptosis (34–36). GPX4 and ACSL4 double-KO cells have the ability to overcome the deadly effect of GPX4 deficiency and do not undergo ferroptosis (35).

Some cells contain additional enzymatic mechanisms such as inducible nitric oxide synthase (iNOS)/NO• which can interact with 15LOX and/or lipid radicals and neutralize them (37). The anti-ferroptotic effect of NO• involves the inhibition of 15-LOX-dependent oxidation of ETE-PE and neutralization of HpETE-PE as well as secondary lipid radicals formed during cleavage and oxidative truncation of this molecule, thus preventing their toxic effects. Increased expression of iNOS/NO• in M1 macrophages/microglia or addition of NO• donors to the M2 macrophages not expressing iNOS lead to their high resistance to ferroptosis. Due to its ability to neutralize the formation of HpETE-PE, iNOS can inhibit ferroptotic cell death acting upstream of GPX4.

Nrf2/NFE2L2 transcription factor is activated as a feedback loop to protect cells from LPO generated during ferroptosis and it is one of the key players in the protection of cells from ferroptosis. After dissociation from its complex with its negative regulator Keap1, Nrf2 translocates to the nucleus where it activates the transcription of target genes (158, 159). NRF2 inhibits ferroptosis through the regulation of hundreds of genes, including genes participating in the regulation of glutathione, GPX4 expression, iron, mitochondrial function and lipid metabolism (159, 160).

Contemporary understanding of the role, contribution and mechanisms of PL peroxidation as drivers of apoptotic and ferroptotic death programs, leads to the design of new classes of specific small molecule inhibitors with a potential for their application as therapeutic agents. To emphasize the importance of the structural and functional organization of the enzymatic complexes directly involved in the production of PL death signals, we will re-iterate their relevant and most important features in this section.

PLs are the major building blocks of the membrane bilayer and their structural role is indispensable and fundamental to compartmentalization and interfacing of thousands of processes in cells and their organelles. In this capacity, macroscopic characteristics of the PL assembles such as fluidity, flexibility, lateral and trans-membrane diffusion are essential for the function of membranes as interactive barriers. Interconnected with this is the signaling by membrane phospholipids that is associated with their post-synthetic modifications. Most commonly this occurs via oxidative transformations of polyunsaturated PL resulting in the production of a huge variety of new molecular species of lipids, integrated into a new concept of the epilipidome. The epilipidome includes not only oxidatively modified lipids but also a huge variety of their adducts with proteins thus merging with the epi-proteome, post-translationally modified proteins. Signaling by individual lipids and lipoprotein adducts comprising the epilipidome/epiproteome is broadly employed in almost each of the myriads of life-defining activities of cells and cell communities. This type of signaling is also essential for coordination of regulated cell death programs. Although these programs may be viewed as suicidal, their strict control and harmonized execution defines the borderline between health and disease. Therefore, deciphering the signaling phospholipid language of these programs is one of the central areas of research not only in cell biology but also in many fields of biomedicine.

While lipid peroxidation has been attributed to essentially all known death programs, its specific mechanisms and role have been clearly defined for only two of them—apoptosis and ferroptosis. Formation of free radical intermediates in the course of LPO occurs both during regulated enzymatic and random non-enzymatic processes. As a result of this, a broad spectrum of agents of different classes with high hydrogen/electron-donating capacities, particularly phenolic compounds and aromatic amines, may be effective in blocking the LPO component of death programs. This has created an optimistic view that many of these compounds may lead to “anti-suicidal” cell-based therapies. The indiscriminative nature of this strategy—likely affecting multiple vital biochemical reactions proceeding via free radical intermediates—may be associated with low effectiveness and serious side effects. Free radical scavengers/sacrificial antioxidants/chain-breaking antioxidants—have been designed and developed and ultimately tested in numerous clinical trials in more than three dozens of diseases. Disappointingly, the results of these were uniformly negative. This strategic mistake in searches for a “magic antioxidant bullet” should be avoided in the design of new anti-apoptotic and anti-ferroptotic therapeutic agents—the lesson has to be learnt. Future generations of small molecule regulators of regulated cell death program therapies must consider their selective and specific mechanisms, hence, to be precise and highly discriminative. In line with this, distinctive inhibitors/regulators have to be developed for controlling individual cell death programs—as has been discussed above.

VEK conceived the idea and wrote the manuscript. YYT, IIV, IHS, and VAT prepared the figures and wrote part of the manuscript. AAK, AAA, TSA FBC, AL, MWE, JSG, DHB, RKM, AKS, HB, and AAS wrote part of the manuscript. All authors contributed to the article and approved the submitted version.

This work was supported by NIH (HL114453, AI156924, AI156923, CA165065, CA243142, AI145406, NS076511, and NS061817) and grant No. 075-15-2020-926 World-Class Research Center (Ministry of Science and Higher Education of the Russian Federation). This project has been funded in whole or in part with federal funds from the National Cancer Institute, National Institutes of Health, under Contract No. HHSN261200800001E.

The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.

The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the National Institute for Occupational Safety and Health, Centers for Disease Control and Prevention.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.