mitochondrial pathway for biosynthesis of lipid mediators

Abstract

The central role of mitochondria in metabolic pathways and in cell death mechanisms requires sophisticated signaling systems. Essential in this signaling process is an array of lipid mediators derived from polyunsaturated fatty acids. However, the molecular machinery for the production of oxygenated polyunsaturated fatty acids is localized in the cytosol and their biosynthesis has not been identified in mitochondria. Here we report that a range of diversified polyunsaturated molecular species derived from a mitochondria-specific phospholipid, cardiolipin, are oxidized by the intermembrane space hemoprotein, cytochrome c. We show that an assortment of oxygenated cardiolipin species undergoes phospholipase A₂-catalyzed hydrolysis thus generating multiple oxygenated fatty acids, including well known lipid mediators. This represents a new biosynthetic pathway for lipid mediators. We demonstrate that this pathway including oxidation of polyunsaturated cardiolipins and accumulation of their hydrolysis products – oxygenated linoleic, arachidonic acids and monolyso-cardiolipins – is activated in vivo after acute tissue injury.

Mitochondria of eukaryotic cells contain machinery capable of oxidizing substrates in a coupled enzymatic and electrochemical process that effectively generates energy in the form of ATP. In addition to the powerhouse function, mitochondria are now viewed as the major regulatory platform involved in numerous intra- and extracellular metabolic and physiological pathways – from synthesis of major intracellular biomolecules to assembly of inflammasomes, immune responses, generation of reactive oxygen species, dynamic regulation of their own organization via fission/fusion and mitophagy as well as control and execution of apoptotic and necrotic cell death¹. Although there is an intuitive necessity for the existence of specific signals for mitochondrial communications – the identification of these signals has not been fully achieved. An array of lipid mediators, with diversified and potent signaling effects on the normal homeostasis and responses to stress and disease, are generated through oxygenation of free polyunsaturated fatty acids (PUFA). Molecular machinery involved in the production of these bioactive oxygenated compounds has been mostly assigned to the cytosol². Surprisingly, mitochondria have not been identified as a site of lipid mediators biosynthesis³.

The rate limiting step in the production of lipid mediators is the availability of oxidizable PUFA⁴. Normally esterified into cellular (phospho)lipids, free PUFA are released by Ca²⁺-dependent phospholipases A₂ (PLA₂) and act as substrates for oxygenation reactions by several enzymes, including, cyclooxygenases (COX), lipoxygenases (LOX), cytochrome P450 isoforms, and peroxidases⁵. A plethora of lipid regulators including prostaglandins, prostacyclins, thromboxanes, resolvins, protectins, maresins, leukotrienes, lipoxines, lipoxenes, levulo-glandins, among others, with multitude of physiologic effects are formed^{6, 7}. Notably, the major precursor of lipid mediators, phosphatidylserine (PS)⁸, is lacking from mitochondria⁹. The majority of phospholipids constituting the inner and outer mitochondrial membranes (IMM and OMM) are manufactured outside of the organelle, with a notable exception of mitochondria-specific cardiolipin (CL). CL (1,3-bis(sn-3’-phosphatidyl)-sn-glycerol) is structurally unique as it contains two phosphatidyl groups linked to a glycerol backbone and four fatty acyl chains. With about 20 different, mostly PUFA residues available for esterification, the diversity of CLs and its oxygenated species may be very high thus making it an exceptionally good source of lipid mediators.

The final rate-limiting step in the production of CL takes place in the IMM to which the newly synthesized CL molecules are sequestered thus generating characteristic CL asymmetry¹⁰. Collapse of CL asymmetry and accumulation of its oxygenated products, have been identified as essential early steps of apoptosis culminating in the release of pro-apoptotic factors¹¹. CL oxygenation is catalyzed by CL/cytochrome c (cyt c) complexes that act as a potent CL-specific peroxidase¹¹ via an enzymatic mechanism similar to that of COX-2 catalyzed oxygenation of arachidonic acid (AA)¹². Given that cyt c-driven reactions yield a highly diversified set of oxidized CL (CL_ox) products, including different stereo-isomers with hydroperoxy-, hydroxy-, epoxy- and oxo-functionalities^{13, 14}, we hypothesized that mitochondrial CLs can be a source of bioactive lipid mediators. While lipid mediators, particularly eicosanoids, have been implicated in the regulation of intrinsic and extrinsic apoptotic pathways¹⁵, mitochondria have not been associated with their biosynthesis. Here we report that CL is oxidized to highly diversified polyunsaturated molecular species by cyt c thereby, representing a new biosynthetic pathway for lipid mediator production. A rich assortment of CL_ox species subsequently undergoes hydrolysis by Ca²⁺-independent iPLA₂γ thus generating multiple oxygenated FAs (FA_ox) that include well known lipid mediators as well as oxygenated species of lyso-CLs with yet to be determined biological functions.

RESULTS

Selective oxidation and hydrolysis of CL in mouse small intestine

To ascertain whether mitochondrial CL can be a source of lipid mediators we performed lipidomics analysis of two different tissues - small intestine and brain - after whole body irradiation (WBI) and controlled cortical impact (CCI), respectively. In small intestine (a radiosensitive tissue) of C57BL6 mice exposed to 10 Gy of WBI, LC/MS analysis revealed: i) a decrease of oxidizable polyunsaturated CL molecular species (Fig. 1a), ii) generation of CL_ox (Fig.1b) and iii) accumulation of mono-lyso-cardiolipins (mCLs) (Fig. 1c) and FA_ox (Fig. 1d). CL_ox were mainly represented by molecular species containing one, two and three additional oxygen atoms and their structures were confirmed by MS/MS analysis (Fig. 1b). While mCLs mainly included non-oxidized molecular species, both mono- and di-oxygenated species of mCL (mCL_ox), particularly with mono- and di-oxygenated LA, were also detected (Supplementary Table S1, Supplementary Fig. S1) whereby the mono-oxygenated mCL derivatives were predominant (Fig. 1c). Quantitatively, the amount of detectable CL_ox (0.12 ± 0.05 pmol/nmol of total phospholipids) was ~2.5 times lower than the level of mCL (0.30 ± 0.18 pmol/nmol of total phospholipids). This indicates that the hydrolysis reactions – possibly catalyzed by Ca²⁺-independent iPLA₂γ – were effective in converting CL_ox into mCLs. The major FA_ox species – mostly mono-oxygenated LA-derivatives - were 9-HODE and 13-HODE (Fig 1d). In addition, 9-KODE, 13-KODE, 9-HpODE, 13-HpODE, 12-HETE, 15-KETE, 15-HETE were also detectable albeit in significantly smaller amounts (Supplementary Fig. S2, Supplementary Table S2).

Figure 1. Exposure of mice to whole body irradiation (WBI) causes cardiolipin (CL) oxidation and accumulation of its hydrolysis products in small intestine.

a, Quantitation of polyunsaturated CL (left) and MS spectra of CL before and after WBI (10 Gy) (right). Each cluster included 4–5 signals with differing m/z. b, Typical MS spectrum of oxidized CL (CL_ox) obtained 10 hrs after WBI (left) and fragmentation pattern of molecular ion with m/z 1463.969 containing oxygenated linoleic acid (LA_ox) (m/z 295.227). c, Quantitation of total monolyso-CL (mCL) (left), MS spectra of mCL (middle) and oxygenated mCL (mCL_ox) (right). Both mono- and di-oxygenated molecular species of mCL_ox were detected 10 hrs after WBI, whereby the mono-oxygenated mCL species were predominant. d, Quantitation of oxygenated polyunsaturated fatty acids (PUFA_ox) (left) accumulated after WBI and MS/MS spectrum of molecular ion with m/z 295 (right). Data are presented as pmol/nmol of total phospholipids. Molecular species of PUFA_ox were represented by LA_ox with m/z 293 (13-KODE, 9-KODE); m/z 295 (13-HODE, 9-HODE, 9,10-EpOME, 12,13-EpOME); m/z 311 (9-HpODE, 8,13-DiHODE); m/z 325 (9,14-KHpODE, 9,14-HpKODE, 8,13-KHpODE, 8,13-HpKODE) and oxygenated arachidonic acid with m/z 319 (12-HETE) and m/z 317 (15-KETE). Data are means ± S.D., *p < 0.05 vs control, n=5.

To ascertain whether iPLA₂γ might be accountable for the hydrolysis of CL_ox leading to the generation of mCLs, we utilized 6E- (bromoethylene)tetrahydro-3R- (1-naphthalenyl)-2H-pyran-2-one, (R)-bromoenol lactone ((R)-BEL), an inhibitor of Ca²⁺-independent iPLA₂γ in vivo¹⁶. We confirmed that (R)-BEL did not inhibit peroxidase activity of cyt c/CL complexes (Supplementary Table S3). (R)-BEL blocked the irradiation-induced accumulation of mCL by 94.9±0.7%. Notably, irradiation induced accumulation of two major lyso-phosphatidylcholines (LPC) (16:0-LPC and 18:0-LPC) was not significantly affected by (R)-BEL (data not shown). The amount of detectable CL_ox in the small intestine of mice pretreated with (R)-BEL and exposed to WBI was estimated as 0.29 ± 0.15 pmol/nmol of total phospholipids, a ~2.5-fold increase vs. its content in the absence of the inhibitor. Thus iPLA₂γ has been mainly responsible for the hydrolysis of CL_ox and the production of mCL. Simultaneously, the total amount of released FA_ox was inhibited by 67.3±7.2%. In particular, 12-HETE and 15-HETE concentrations in irradiated intestinal samples from (R)-BEL treated animals were 45.7±16.0% and 76.3±28.0% of untreated irradiated intestinal samples. Expectedly, the decrease of oxidizable CL species was not significantly affected by (R)-BEL (data not shown). Overall, these data emphasize the major role of Ca²⁺-independent iPLA₂γ in the generation of mCLs and also reflect the likely involvement of other pathways – independent of iPLA₂γ – in the formation of FA_ox.

We next compared the effects of a cocktail of several reportedly effective inhibitors of COX-, LOX- and cytochrome P450-driven conventional biosynthetic mechanisms for lipid mediators species versus their generation via CL-dependent pathways. Specifically, we tested the effects of the inhibitors of: i) COX-1 and -2 - piroxicam (4-hydroxy-2-methyl-N-2-pyridinyl-2H-1, 2-benzothiazine-3-carboxamide-1,1-dioxide); ii) LOX - licofelone (6- (4-chlorophenyl)-2,3-dihydro-2,2-dimethyl-7-phenyl-1H-pyrrolizine-5-acetic acid); and cytochrome P450 epoxygenase/12-HETE activity (CYP2C) - MS-PPOH (N-(methylsulfonyl)-2- (2-propynyloxy)-benzene-hexanamide))^17–20 on the formation of FA_ox in the small intestine of irradiated mice. Because all three enzymes – COX, P450 and 5-LOX – are iron-containing proteins of which the former two are hemoproteins, we assessed the effects of inhibitors on the catalytic peroxidase properties of cyt c/CL complexes in vitro. Within the range of concentrations corresponding to the doses applied in our in vivo experiments, these inhibitors did not suppress the peroxidase activity (Supplementary Table. S3). We found that the inhibitor cocktail partially decreased the levels of mono-oxygenated AA and oxo-LA compared with the group of mice that received no drugs before irradiation (Supplementary Table S4). The production of di-oxygenated LA did not appear to be affected by the mixture of inhibitors. This suggests that generation of lipid mediators from CL/CL_ox is accomplished through the enzymatic reactions distinct from the known Ca²⁺-dependent mechanisms realized via the release of free FAs and their subsequent oxygenation.

Selective oxidation and hydrolysis of CL in rat brain

In rat brain, CLs were highly diversified and included mostly oxidizable polyunsaturated CLs (Fig. 2a). LC/MS analysis revealed that CCI caused depletion of oxidizable CLs containing LA, AA and docosahexaenoic (DHA) acids (Fig. 2a) and formation of CL_ox (Fig. 2b) and its hydrolysis products: mCL (Fig. 2c) and FA_ox (Fig. 2d). CL_ox were mainly represented by species containing one and two additional oxygen atoms (Fig.2b). MS/MS analysis identified mono-oxygenated species of CL_ox with either LA_ox or AA_ox (Fig. 2b). mCLs were represented by non-oxidized as well as oxidized mCL_ox species (Fig. 2c, Supplementary Table S1, and Supplementary Fig. S3). The CL_ox content was lower than the amount of mCL: 0.10 ± 0.02 and 0.26 ± 0.08 pmol/nmol of total phospholipids, respectively. In addition, the oxygenated species of non-esterified LA and AA were produced (Fig. 2d, Supplementary Fig. S4), whereby several lipid mediators were identified such as 9-KODE, 13-KODE, 9-HODE, 13-HODE, 9,12,13,-KEpOME, 9-HpODE, 13-HpODE, 12-HETE, 15-HETE (Supplementary Table S2). Markedly smaller amounts of oxygenated DHA were also detected (Fig. 2d). Quantitatively, the CCI-induced loss of CL species containing PUFA was in good correspondence with the accumulation of FA_ox (~5.1 and ~3.0 pmol/nmol of total phospholipids, respectively). This indicates that CL can be a source of CCI-induced production of FA_ox.

Figure 2. Exposure of rats to controlled cortical impact (CCI) causes cardiolipin (CL) oxidation and accumulation of its hydrolysis products in brain.

a, Quantitation of polyunsaturated CL (left) and MS spectra of CL in control and 3 hrs after CCI (right). The spectrum of brain CLs includes nine major clusters of signals with 8–9 components with differing m/z in each. b, Typical MS spectrum of oxidized CL (CL_ox) (left) and fragmentation pattern of CL_ox molecular ion with m/z 1491.984 (right) represented by two species containing either oxygenated linoleic acid (LA_ox) (m/z 295.227) or arachidonic acid (AA_ox) (m/z 319.227) and originated from m/z 1475.980. c, Quantitation of total monolyso-CL (mCL) (left) and MS spectra of mCL (middle) and oxygenated mCL (mCL_ox) (right) formed after CCI. Di-oxygenated species of mCL_ox were detected in injured brain. d, Quantitation of major polyunsaturated fatty acids (left) and MS/MS spectrum of molecular ion with m/z 319 (right). Data are presented as pmol/nmol of total phospholipids. Species of LA_ox with m/z 293 (13-KODE, 9-KODE); m/z 295 (13-HODE, 9-HODE, 9,10-EpOME, 12,13-EpOME); m/z 311 (9-HpODE, 8,13-diHODE), m/z 309 (8,13-HKODE, 9,14-KHODE); m/z 325 (9,14-KHpODE, 9,14-HpKODE, 8,13-KHpODE, 8,13-HpKODE) and AA_ox with m/z 319 (12-HETE,15-HETE) were identified. Small amounts of DHA_ox (m/z 343) were also detected. Data are means ± S.D., *p < 0.05 vs control, n=4.

To ascertain whether the CL/CL_ox-dependent pathway may be cell type specific, we compared the responses of two types of brain cells – cortical neurons and one of the major components of glia, astrocytes, to a standard treatment with H₂O₂ in vitro. We found that neurons (Supplementary Fig. S5, Supplementary Table S5) contained more oxidizable polyunsaturated CL species than astrocytes (Supplementary Fig. S6, Supplementary Table S5). Accordingly, the amounts and speciation of CL_ox after H₂O₂ exposure was also markedly richer in the former than in the latter (Supplementary Figs. S5 and S6). Similarly, the levels and diversification of mCLs was also significantly greater in neurons than in astrocytes (Supplementary Figs. S5 and S6).

Generation of oxidized TLCL and its hydrolysis products in mitochondria

To further characterize pathways of CL peroxidation and hydrolysis we utilized C57BL6 mouse heart and liver mitochondria in which CL was accountable for ~15% of total phospholipids (Fig. 3). MS analysis demonstrated that oxidizable LA residues were present in all molecular species of CL (Supplemental Fig. S7). After exposure to a pro-oxidant, t-BuOOH, isolated mitochondria revealed the same pattern of characteristic CL changes as in vivo: i) decreased amounts of oxidizable CLs (Fig. 3a) accumulation of non-oxidized mCL (Fig. 3b) and LA_ox, (containing one and two oxygens) (Fig. 3c).

Figure 3. Peroxidized cardiolipins (CL) undergo phospholipase A₂-catalyzed hydrolysis in mitochondria.

a, b, c, Oxidation of CL (a) and accumulation of monolyso-CL (mCL) (b) and oxygenated fatty acids (FA_ox) (c) in mouse heart mitochondria treated with t-BuOOH (150 µM). Oxygenated linoleic acid (m/z 295 and m/z 311) and oxygenated arachidonic acid (m/z 319 and 335) containing one and two oxygens were detected. Data are presented as pmols FA_ox per nmol of phospholipids. Data are means ± S.D., *p < 0.05 vs. control, n = 3. d,e,f,g,h, Effect of 6E-(bromoethylene)tetrahydro-3R-(1-naphthalenyl)-2H-pyran-2-one, (R)-bromoenol lactone) (R)-BEL, Ca²⁺-independent iPLA₂γ inhibitor, on hydrolysis of exogenous oxidized tetra-linoleyl CL (TLCL_ox) by mouse liver mitochondria. MS spectra of CL obtained from mouse liver mitochondria before and after addition of TLCL_ox (d). TLCL was oxidized by cyt c/H₂O₂ (see Supplementary Fig.S10). Content of TLCL_ox (e), FA_ox (f) and mCL (g) in mitochondria incubated with exogenous TLCL_ox in the presence and in the absence of (R)-BEL. MS spectrum of mCL from mitochondria incubated in the presence of TLCL_ox (h). Data are mean ± S.D., *p < 0.05 vs. non-treated mitochondria, n=3.

Identification of cytochrome c as the catalyst of CL oxidation and hydrolysis

To directly assess the involvement of cyt c in CL peroxidation and hydrolysis products as a source of lipid mediators, we compared the production of mCL and FA_ox in cyt c^+/+ and cyt c^−/− mouse embryonic cells (MECs) during apoptosis triggered by non-oxidant (actinomycin D, ActD) or oxidant (t-BuOOH) stimuli (Fig. 4a). No significant differences were found in the total content and molecular speciation of CLs (approximately 3% of total phospholipids; 31.1±4.9 and 33.7±7.1 pmols CL/nmol total phospholipids, respectively) in cyt c^+/+ and cyt c^−/− cells (Fig. 4b and Supplementary Fig. S8). Treatments with ActD or t-BuOOH caused a significant decrease of oxidizable polyunsaturated species of CL and accumulation of CL_ox species with one and two oxygens in cyt c^+/+ cells - but not in cyt c^−/− - cells (Fig. 4b and Supplementary Fig. S8). Quantitatively, the amounts of CL_ox produced in cyt c^+/+ cells challenged with either ActD or t-BuOOH were increased 2.6- and 3.0 -times compared to untreated controls. Analysis of hydrolysis products revealed the presence of both oxidized and non-oxidized mCL (Fig. 4c) as well as oxygenated LA (Fig. 4d). Mono- and di-oxygenated molecular species of mCL_ox were the predominant forms accumulating in cells upon the treatments (Fig.4c). Mono-oxygenated LA was the main product in ActD treated cyt c^+/+ cells and was represented by a mixture of 9-HODE and 13-HODE (Fig. 4d). No significant increase in the levels of CL_ox and CL hydrolysis products was detected in ActD and t-BuOOH treated cyt c^−/− cells (Fig. 4). Notably, ActD or t-BuOOH triggered apoptosis in cyt c^+/+ cells as evidenced by a marked increase of PS externalization (Fig. 4a), whereas these pro-apoptotic effects were not found in cyt c^−/− cells¹¹.

Figure 4. Execution of apoptosis is accompanied by cardiolipin (CL) oxidation and accumulation of its hydrolysis products in mouse embryonic cyt c^+/+ (but not cyt c^−/−) cells.

a, Assessments of a biomarker of apoptosis, externalized phosphatidylserine, in cyt c^+/+ and cyt c^−/− cells by annexin V binding (left). Flow cytometry data from a representative experiment after treatment of cells with ActD (100 ng/mL, 16 hrs): the number of annexin V/FITC-positive cells is shown in log scale (right). b) Consumption of CL in either ActD or t-BuOOH (200 µM, 16 hrs) treated cyt c^+/+ and cyt c^−/− cells (left) and MS spectra of CL (middle), oxygenated CL (CL_ox) (right) from cyt c^+/+ cells exposed to ActD. c, Accumulation of monolyso-CL (mCL) in either ActD or t-BuOOH treated cyt c^+/+ and cyt c^−/− cells (left) and MS spectra of mCL (middle), oxygenated mCL (mCL_ox) (right) from cyt c^+/+ ActD treated cells. d, Accumulation of oxygenated linoleic acid (LA_ox) in ActD and t-BuOOH treated cyt c^+/+ and cyt c^−/− cells (left) and MS/MS fragmentation pattern of LA_ox (m/z 295) (right) formed in ActD treated cyt c^+/+ cells. Mono-oxygenated LA_ox was predominantly accumulated and represented by both 9-HODE and 13-HODE. Data are mean ± S.D., *p < 0.01 vs control, n=3–4.

MECs express two isoforms of cyt c – somatic (s-cyt c) and testicular (t-cyt c)^21–23. Western blot analysis established that t-cyt c was accountable for ~45% of the total cyt c in MECs (Supplementary Fig. S9). In cyt c^−/− cells, the amount of t-cyt c was ~1.5 times lower than in cyt c^+/+ cells (0.83 ± 0.05 and 1.24 ± 0.08 ng/µg protein, respectively). As cyt c^−/− cells rely more on glycolysis rather than mitochondrial respiration for ATP generation²⁴ they maintain lower levels of mitochondria. Indeed, Western blots showed that the levels of several mitochondrial marker proteins (COX-IV, MnSOD, TIM23) were ~46%–83% lower in cyt c^−/− cells versus cyt c^+/+ cells (Supplementary Fig. S9). siRNA knock-down of the t-cyt c in cyt c^−/− cells to ~50% of its content (Supplementary Fig. S9) did not affect the already low level of CL oxidation after ActD treatment. Similar to cyt c^−/− cells the low level of mCL in cyt c^−/− with knock-down t-cyt c remained unchanged after ActD exposure. These data suggest that t-cyt c is not a significant contributor to the generation of CL-driven lipid mediators.

Hydrolysis of oxidized TLCL by mitochondrial Ca²⁺-independent PLA₂γ

Ca²⁺-independent iPLA₂γ has been suggested as a likely candidate-catalyst capable to hydrolyze oxidized phospholipids in mitochondria²⁵. To assess its role in the hydrolysis of CL_ox species, we bio-synthesized (using cyt c/H₂O₂ system) and purified (to the level of 99% using LC/MS) oxidized tetralinoleoyl-cardiolipin (TLCL_ox with 1–8 oxygens in four LA residues, see Supplementary Table S6 and Supplementary Fig. S10) and analyzed the products formed in the presence of a specific iPLA₂γ inhibitor, (R)-BEL, in rat liver mitochondria (Fig. 3d). Mitochondria effectively hydrolyzed TLCL_ox as evidenced by i) decrease of TLCL_ox content (Fig. 3e) and ii) accumulation of oxygenated LA (with one and two oxygens) (Fig. 3f) as well as non-oxidized mCL and mCL_ox (Fig. 3g,h). The hydrolysis of TLCL_ox was effectively blocked by (R)-BEL (Fig. 3).

Hydrolysis of oxidized TLCL by PAF acetylhydrolase

Mitochondria with externalized CL and/or CL_ox as well as cyt c can be released into circulation during injury/disease process²⁶. Thus circulating CL and CL_ox can be a source of lipid mediators as well. Several phospholipases, including lipoprotein-associated PLA₂ (LpPLA₂) or platelet activating factor (PAF) acetylhydrolase (PAF-AH), have been shown to selectively hydrolyze oxidatively modified phospholipids to liberate FA_ox²⁷. Recently, we have identified oxidized PS as a good substrate for Lp-PLA₂²⁸. Because TLCL is the most predominant species of CL in circulation, we tested the ability of PAF-AH to release oxygenated FAs from TLCL_ox produced in cyt c driven reaction (Supplementary Fig. S10). LC/MS analysis revealed accumulation of mCL containing both LA and LA_ox (Figs. 5a, 5b, 5c), whereby LA_ox was represented by molecular species with one and two oxygens (Fig. 5d). Markedly less effective hydrolysis was detected when non-oxidized TLCL was treated with PAF-AH (data not shown). Thus, PAF-AH predominantly hydrolyzes oxygenated molecular species of CL to generate lipid signaling molecules formed in cyt c catalyzed reaction.

Figure 5. Platelet activating factor-acetylhydrolase (PAF-AH) catalyzes hydrolysis of oxygenated tetra-linoleyl-CL (TLCL).

a, MS spectra of monolyso-CLs (mCLs) (left) and dilyso-CLs (dCLs) (right) formed during PAF-AH-driven hydrolysis of oxygenated TLCL (TLCL_ox). Arrows show oxygenated mCL (mCL_ox) and dCL (dCL_ox) species (and their m/z values) with indicated oxygen atoms added. b, c, Quantitative assessment of lyso-CLs (b), non-oxygenated and oxygenated linoleic acid (LA_ox) (c) formed during PAF-AH-driven hydrolysis of TLCL_ox. d, Identification and quantification of LA_ox liberated after hydrolysis of TLCL_ox by PAF-AH. Molecular ions of LA_ox with m/z 293 (13-KODE, 9-KODE), m/z 295 (13-HODE, 9-HODE, 9,10-EpOME, 12,13-EpOME), m/z 309 (8,13-HKODE, 9,14-KHODE), m/z 311 (9-HpODE, 13-HpODE and 8,13-DiHODE) and m/z 327 (9,12,13-HpEpOME) were detected. Data are means ± S.D., *p < 0.05 vs. LA, n=4.

Diversity of oxygenated FAs after hydrolysis of brain CL_ox generated by cyt c/H₂O₂

The high diversification and enrichment with PUFA of brain lipids makes neuro-CLs particularly interesting as a source of lipid mediators generated via CL oxidation by cyt c. Indeed, we identified 56 major molecular species of CL in lipid extracts from mouse brain (Supplementary Table S7) of which 55 were highly oxidizable polyunsaturated CL containing one to four PUFAs (Fig. 6a). To establish molecular identity and stereospecificity of peroxidizable PUFA in neuro-CLs we employed a mixture of PLA₁ with PLA₂ and established that the dominant PUFAs of CL were represented by AA and LA (Fig. 6a). DHA, EPA, eicosatrienoic and octadecatrienoic acids were detected as well but in relatively lower abundance. Cyt c/H₂O₂ caused significant decrease of PUFA-containing molecular species of CL (Fig. 6b and Fig. 6c). By employing PLA₁ plus PLA₂ as a tool to liberate FA from both sn-1 and sn-2 positions of CL_ox we identified by MS/MS analysis highly diversified molecular species of LA_ox and AA_ox (Fig. 6d) which included those detected in acutely injured tissues (brain after CCI and small intestine after WBI) (Supplementary Table S2) as well as in apoptotic cyt c^+/+ cells. In addition, oxidatively truncated FA molecular species were detected (Supplementary Table S2). Overall, 31 molecular species of FA_ox have been identified as the products of brain CL_ox hydrolysis.

Figure 6. Cytochrome c (cyt c)/H₂O₂ induces oxidation of polyunsaturated fatty acid (PUFA) residues of cardiolipin (CL) isolated from mouse brain.

a, Characterization of polyunsaturated fatty acids (PUFA) esterified to brain CL. Number of CL molecular species containing PUFA (left). PUFA distribution in brain CL species (right). b, c, MS spectra (b) and quantitative assessment of brain CL molecular species (c) obtained before (red) and after (blue) treatment with cyt c in the presence of H₂O₂. d, Quantitative assessment of PUFA of non-oxidized CL (red) and CL exposed to cyt c/H₂O₂ (blue). *p < 0.05 vs control (CL not treated with cyt c). Data are means ± S.D., n=3.

Discussion

Oxygenated free PUFAs – generated by a wide array of enzymes, including cyclooxygenases (COX-1 and COX-2), lipoxygenases (5-LOX, 12-LOX and 15-LOX) and cytochrome P450 isoforms [reviewed in^{5, 7}] - have numerous physiological roles. The potent biological effects of lipid regulators necessitate the maintenance of very low endogenous levels of free PUFA. As a result, their availability is the rate-limiting step in the generation of lipid mediators⁴. Release of PUFA precursors from their phospholipid storage sites is achieved via action of Ca²⁺-dependent phospholipases A₂ (PLA₂)^,30. We have reported a new Ca²⁺-independent pathway for selective generation of lipid mediators from a mitochondria-specific phospholipid, CL, whereby oxidation of FA residues is catalyzed by an intermembrane space protein, cyt c, directly in the esterified phospholipid. This is followed by hydrolysis of oxygenated CL species by two types of PLA₂ specific towards oxidatively modified phospholipids. This novel pathway disobeys the major dogmas of lipid mediator biochemistry with regards to: i) oxygenation of a phospholipid rather than a free FA as the reaction substrate, ii) mitochondrial – rather than cytosolic - localization- iii) an electron carrier cyt c – rather than an oxidase - as the reaction catalyst, iv) the final – rather than the initial – hydrolytic stage using oxygenated species of CL by PLA₂ to synthesize oxygenated PUFA, and v) Ca²⁺-independent - rather than Ca²⁺ dependent - nature of the pathway.

One can predict specific features of regulation of CL-dependent process that are markedly different from the “conventional” biosynthesis of oxygenated lipid mediators. The mitochondrial pathway is strictly dependent on the catalytic interactions between cyt c and CL. However, these two participants have limited access to each other. The former is confined to the intermembrane space while the latter is “cloaked” almost exclusively in the inner mitochondrial membrane³¹. Importantly, ~15% of cyt c is believed to be hydrophobically bound to the IMM likely via its binding to CL in the outer leaflet of the IMM and cannot be removed from mitochondria, as occurs to the majority of cyt c upon treatment with high ionic strength solutions³². The function of the tightly membrane-bound cyt c in normal mitochondria has not been identified. Given that CL-bound cyt c cannot act as electron acceptor from respiratory complex III³³, we speculate that catalysis of CL-dependent formation of lipid mediators may represent an unrecognized function of cyt c.

For the peroxidase activity of cyt c/CL complexes, the presence of sufficient amounts of oxidizing equivalents such as H₂O₂ or lipid hydroperoxides (“peroxide tone”^{34, 35}) is essential for triggering CL oxidation. Their supply may be driven by disrupted electron transport likely occurring as a consequence of cyt c binding with CL³⁴. The complex has a redox potential ~400 mV more negative than free cyt c – the effect that precludes cyt c’s function as an electron acceptor from mitochondrial complex III³³. As a result accumulating reduced intermediates serve to donate electrons to molecular oxygen yielding superoxide radicals. Spontaneous or catalyzed dismutation of the latter generates H₂O₂ that feeds the catalytic peroxidase cycle of cyt c/CL complexes. This dependency on H₂O₂ as the source of oxidizing equivalents disappears with the accumulation of FA-hydroperoxides or CL-hydroperoxides that can act as more efficient substrates for the peroxidase half-cycle of the process³⁵.

Cells can express, at different proportions, two isoforms of cyt c – s-cyt c and t-cyt c^21–24. We found that t-cyt c does not significantly contribute to the production of lipid mediators. To investigate if there are potential differences in s-cyt c versus t-cyt c participation in lipid oxidation, we compared the interaction of two isoforms of cyt c with CL using sequence analysis, molecular structural modelling, as well as molecular docking and computational simulations. The sequences of s- and t-cyt c and 3D-structures displayed high similarity (Supplementary Fig. S12). Further, ligand docking analysis revealed close similarity between the two isoforms in binding a typical peroxidation substrate, TLCL, as judged by closely matching predicted binding energy values (Supplementary Table S8). Finally, coarse-grained molecular dynamics simulations of cyt c interactions with lipid bilayers revealed equally avid binding to TLCL-containing membranes at early stages in the simulation and lack thereof with lipid bilayers devoid of TLCL (Supplementary Fig. S12). This suggests that t-cyt c, like s-cyt c, should be competent towards peroxidase functions, including CL peroxidation.

The failure of t-cyt c to operate as a catalyst of CL oxidation in MECs may mostly due to its presence as apo-protein whereby its biosynthesis into the catalytically active holo-enzyme may be the limiting factor. According to Kim et al., three times more t-cyt c was detected by radioimmunoassay, than by spectral assessments of the catalytically competent hemoprotein³⁶. Of note, the anti-t-cyt c antibody employed in our study does not distinguish between the apo- and holo-proteins. Differences in the localization of these two isoforms of cyt c in the intermembrane space and their differential interactions with CLs and other proteins may also be contributory to their dissimilar peroxidase activity towards CLs^{37, 38}. The suggested higher pro-apoptotic activity of t-cyt c vs s-cyt c may be realized through their differential interactions with pro-caspase complexes (Apaf) and the formation of apoptosomes³⁹.

A notable feature of cyt c-catalyzed oxygenation reactions described in this work is accumulation of not only of the expected hydroperoxy-CL and hydroxy-CL⁵ but also the presence of epoxy- and oxo-derivatives and truncated products of oxidative cleavage. Whether these derivatives are produced by enzymatic cyt c-catalyzed reaction or non-enzymatic process caused by degradation and release of cyt c’s heme – is unknown. Similar products require additional enzymatic activities (i.e., epoxidase, epoxide hydrolases) in the traditional pathways triggered by COX and LOX⁵. Our results with pharmacological inhibitors of two different pathways COX/LOX/P450 and (R)-BEL support a role of CL oxidation as a source of the generated lipid mediators. In line with our data, disturbed CL re-acylation in mice with knock-down tafazzin - modeling Barth syndrome - was accompanied by the altered pattern of HETE and oxidized LA and DHA metabolites⁴⁰.

The final step in biosynthesis of oxygenated free FAs requires hydrolysis of oxidatively modified CLs by PLA₂. Several representatives from different groups of secretory PLA₂ (sPLA₂) of a large superfamily of PLA₂ were shown to display hydrolyzing activity towards peroxidized phospholipids [reviewed in David et al.,²⁹]. Among those, the best studied is LpPLA₂ – a Ca²⁺-independent LpPLA₂ or type VIIA PLA₂ - which can be represented by intracellular and secreted forms³⁰ and is active towards oxygenated long-chain and truncated forms of PC⁴¹. Oxidized PS has been identified as a representative of anionic peroxidized phospholipids readily hydrolyzable by Lp-PLA₂²⁸.

CLs have been identified as signaling molecules in two major biological functions: mitophagy⁴² and apoptosis¹¹. In both cases, CL gets externalized to the mitochondrial surface suggesting that any injury to plasma membrane may be associated with the release of these “CL-decorated” mitochondria into extracellular environments. Assuming that apoptosis commonly transitions to necrosis, mitochondria may – with their externalized CL and CL_ox – act as damage associated molecular patterns (DAMPs)^26,43. Our studies reveal that not only CL and CL_ox but also CL_ox hydrolysis products – mCL and FFA_ox – may be released, along with mitochondria, from injured cells. Importantly, sufficient hydrophobicity of mCL retains its association with mitochondria. However, oxygenated mCLs may lose their association with the outer mitochondrial membrane and partition into the aqueous phase of extracellular compartments. Oxygenated FFAs also water-soluble, hence diffuse independently of mitochondrial surfaces. Release of CL⁴⁴ and cyt c⁴⁵ from injured host cells into the extracellular space, have been shown. Thus lipid mediators may also be generated extracellularly from CL with the contribution of cyt c and subsequent hydrolysis of CL_ox by Lp-PLA₂. Overall, CL-dependent lipid mediators may be represented by a diversified variety of membrane-associated and freely diffusible and circulating signaling molecules whose identification and quantitative analysis will represent an intriguing opportunity for the future studies.

Brain has an unprecedented diversification of CLs¹³. By using brain CLs, we showed that a significant variety of lipid mediators could be generated by cyt c-catalyzed oxygenation process. In fact, we found that all eight well known LA-based lipid mediators were generated by cyt c/H₂O₂ (Supplementary Table S2). In the AA series, we were able to identify nine species of lipid mediators (Supplementary Table S2). A relatively smaller number of oxygenated derivatives of docosapentaenoic acid and DHA were detected in spite of the significant proportion of these FA residues in brain CLs. Also, terminally hydroxylated metabolites of AA, including 17-HETE, 18-HETE, 19-HETE, and 20-HETE were not found, implying that terminal hydroxylation reactions are not catalyzed by cyt c/H₂O₂ and are likely formed enzymatically after AA liberation.

The ranking order of FA resiues oxidation in the model biochemical system - where all the CL substrates were equally available - was DHA>AA>>LA. The amounts of DHA, AA and LA decreased almost 20-, 5- and 0.3-fold, respectively. The energies for H-abstraction in α-position to bis-allylic double-bonds for different FA with multiple double bonds decrease in the order LA<AA<DHA. Accordingly, the propagation rate constants for FA oxidation determined for LA (one bis-allylic group), AA (3 bis-allylic groups), EPA (4 bis-allylic groups) and DHA (5 bis-allylic groups) increase at ratios of 1, 3.2, 4.0, and 5.4⁴⁶. This is consistent with the assumption that the H-abstraction is essential for oxidation of different polyunsaturated species of CLs by cyt c’s protein-immobilized Tyr radical: Tyr-O^• +CL-H-> Tyr-OH +CL^•. Thus, preferential oxidation of LA in brain CL’s in vivo does not correlate with the chemical reactivity suggesting that CL oxidation in mitochondria is not a simple reaction of CLs with cyt c. Availability of CLs for oxidation by cyt c in the inter-membrane space likely defines the specificity of the reaction towards different CL species. Our findings with brain CLs may provide a long awaited explanation for a well-known large variety of CL species in the brain and some other tissues (eg, small intestine, lung) – by implying that they are used as precursors for biosynthesis of diversified lipid mediators⁴⁷.

In conclusion, we describe a new pathway for generation of lipid mediators. This novel Ca²⁺-independent pathway localizes to mitochondria, involves cyt c-catalyzed oxygenation of CL esterified PUFA residues, is followed by hydrolysis of oxygenated CL species by two types of Ca²⁺-independent PLA₂ specific towards oxidatively modified phospholipids.

Methods

This section describes key experiments only; an extended experimental section is provided in the Supplementary Methods.

Controlled cortical impact to the left parietal cortex in 17 day old rats was performed as described previously⁴⁸. For all studies, a 6-mm metal pneumatically driven impactor tip was used. The velocity of the impact was 4.0 ± 0.2 m s⁻¹, with a penetration depth of 2.5 mm. Whole body irradiation: C57BL/6NHsd female mice were irradiated with dose of 10 Gy using a J. L. Shepherd Mark 1 Model 68 cesium irradiator at a dose rate of 80 cGy/min as described previously⁴⁹. 10 Gy of WBI is a LD100/30 for C56Bl6 mice, which means this dose causes death of all animals within 30 days after the exposure. Mice were euthanized 10 or 24 hrs later by CO₂ inhalation. All procedures were approved by the IACUC of University of Pittsburgh and performed according to the protocols established.

Analysis of CL and mCL molecular species was performed using a Dionex Ultimate™ 3000 HPLC system coupled on-line to a linear ion trap mass spectrometer (LXQ, ThermoFisher Scientific, San Jose, CA) using Luna 3 µm Silica (2) 100 Å column (Phenomenex, Torrance, CA) as described previously²⁸. To analyze oxygenated species, CL and mCL were isolated from total lipids by normal phase 2D-HPTLC using mobile phases as described by Rouser et al.,⁵⁰. To prevent lipid oxidation during separation, chromatography was performed under N₂ conditions on diethylenetriaminepentaacetic acid (DTPA) treated silica plates (5 × 5 cm, Whatman). CL and mCL were extracted from silica spots⁴⁹ and used for reverse phase ESI-LC/MS analysis. LC/MS analysis was performed using a Dionex UltimateTM 3000 RSLCnano system coupled online Q-Exactive hybrid quadrupole-orbitrap mass spectrometer (ThermoFisher Scientific, San Jose, CA) using a C₈ column (Luna 3 µm,100 Å, 150 × 2 mm, Phenomenex, Torrance, CA). For CL_ox an isocratic solvent system consisting of 2-propanol: water: triethylamine: acetic acid, 45:5:0.25:0.25, v/v was delivered at 150 µl/min for 20 minutes. For mCL_ox, a gradient of solvent A (acetonitrile: water: triethylamine: acetic acid, 45:5:0.25:0.25 v/v) and B (2-propanol: water: triethylamine: acetic acid, 45:5:0.25:0.25, v/v) was used at the flow rate of 150µl/min as follows: 0–10 min isocratic at 50% solvent B; 10–20 min linear gradient 50–100% solvent B; 20–32 min isocratic at 100% solvent B; 32–35 min linear gradient at 100–50% solvent B; 35–45 min isocratic at 50% solvent B. Spectra were acquired in negative ion mode using a spray voltage of 4.0 kV and a capillary temperature of 320 °C. Scans were acquired in data-dependent mode with an inclusion list for both CL/CL_ox, mCL/mCL_ox species, isolation width of 1.0 Da and a normalized collision energy of 24 in high energy collisional dissociation (HCD) mode. TMCL-(14:0)₄ (Avanti Polar Lipids Inc., Alabaster, AL) and mCL-(14:0)₃ were used as internal standards. The peaks with signal-to-noise (S/N) ratio of 3 and higher were taken into consideration.

Detection of oxidized FA. LA_ox was analyzed by LC/MS using a Dionex Ultimate™ 3000 HPLC system coupled on-line to LXQ linear ion trap mass spectrometer (ThermoFisher Scientific, San Jose, CA). A C₁₈ column (Luna, 3 µm, 150 × 2 mm, Penomenex, Torrance, CA) and gradient solvents (A: tetrahydrofuran/methanol/water/CH₃COOH, 25:30:50:0.1 (v/v/v/v) and B: methanol/water 90:10 (v/v)) containing 5 mM ammonium acetate were used. The column was eluted at a flow rate of 0.2 mL/min during first 3 min isocratically at 50% B, from 3 to 23 min with a linear gradient from 50% solvent B to 98% solvent B, then 23–40 min isocratically using 98% solvent B, 40–42 min with a linear gradient from 98% solvent B to 50% solvent B, 42–28 min isocratically using 50% solvent B for equilibration of the column. Spectra were acquired in negative ion mode using a spray voltage of 5.0 kV and a capillary temperature of 150 °C. Assay method for quantitative assessment of free AA_ox and DHA_ox was completed as previously described by Miller et al.⁵¹ Briefly, LC was performed using an acuity ultra-performance LC autosampler (Waters, Milford, MA). Separation of analytes was conducted on a UPLC BEH C₁₈, 1.7µm (2.1 × 100 mm) column. A gradient mobile phase of 0.005% acetic acid, 5% acetonitrile in deionized water and 0.005% acetic acid in acetonitrile was used with a run time of 6.4 minutes. Analysis was performed using a TSQ Quantum Ultra (ThermoFisher Scientific, San Jose, CA) triple quadrupole mass spectrometer with heated electrospray ionization in negative selective reaction monitoring mode. Analytical data were acquired and analyzed using Xcalibur software.

Abbreviations and Textual Footnotes

WBI

Whole body irradiation

CCI

controlled cortical impact

COX

cyclooxygenase

LOX

lipoxygenase

FA

fatty acids

FA_ox

oxygenated fatty acids

PUFA

polyunsaturated fatty acids

PUFA_ox

oxygenated PUFAs

LA

linoleic acid

LA_ox

oxygenated linoleic acid

AA

arachidonic acid

AA_ox

oxygenated arachidonic acid

EPA

eicosapentaenoic

DHA

docosahexaenoic acid

PS

phosphatidylserine

PE

phosphatidylethanolamine

PC

phosphatidylcholine

PI

phosphatidylinositol

PIP₂

phosphatidylinositol 4,5-bisphosphate

PLA₂

phospholipases A₂

iPLA₂γ

Ca²⁺-independent iPLA₂γ

LpPLA₂

Ca²⁺-independent lipoprotein-associated PLA₂

PAF-AH

Platelet activating factor-acetylhydrolase

CL

cardiolipin

TMCL

tetra-linoleoyl-cardiolipin

TLCL

tetra-mirystoyl-cardiolipin

TLCL_ox

oxygenated tetra-linoleoyl-cardiolipin

CL_ox

oxygenated cardiolipin species

mCL

mono-lyso-cardiolipin

mCL_ox

oxygenated mono-lyso-cardiolipin

cyt c

cytochrome c

s-cyt c

somatic isoform of cytochrome c

t-cyt c

testicular isoform of cytochrome c

IMM

inner mitochondrial membrane

OMM

outer mitochondrial membrane

2D-HPTLC

two-dimensional high performance thin layer chromatography

MS

mass spectrometry

LC/MS

liquid chromatography mass spectrometry

9-KODE

9-oxo-octadecadienoic acid

13-KODE

13-oxo-octadecadienoic acid

9-HODE

9-hydroxy-octadecadienoic acid

13-HODE

13-hydroxy-octadecadienoic acid

9,10-EpOME

9,10-epoxy-octadecanoic acid

12,13-EpOME

12,13-epoxy-octadecanoic acid

9,14-KHODE

9-oxo-14-hydroxy-octadecadienoic acid

8,13-HKODE

8-hydroxy-13-oxo-octadecadienoic acid

9,12,13-EpKOME

9-oxo-12,13-epoxy-octadecanoic acid

8,13-DiHODE

8,13-dihydroxy-octadecadienoic acid

8,13-KHpODE

8-oxo-13-hydroperoxy-octadecadienoic acid

8,13-HpKODE

8-hydroperoxy-13-oxo-9,11-octadecadienoic acid

9,14-KHpODE

9-oxo-14-hydroperoxy-octadecadienoic acid

9,14-HpKODE

9-hydroperoxy-14-oxo-10,12-octadecadienoic acid

9,12,13-HpEpOME

9-hydroperoxy-12,13-epoxy-octadecanoic acid

15-KETE

15-oxo-eicosatetraenoic acid

15-HETE

15-hydroxy-eicosatetraenoic acid

12-HETE

12-hydroxy- eicosatetraenoic acid

17-HETE

17-hydroxy- eicosatetraenoic acid

18-HETE

18-hydroxy- eicosatetraenoic acid

19-HETE

19-hydroxy- eicosatetraenoic acid

20-HETE)

20-hydroxy- eicosatetraenoic acid