Research article

Mitochondrial targeting of the electrophilic lipid 15-deoxy-Δ12,14prostaglandin J2 increases apoptotic efficacy via redox cell signalling mechanisms

Anne R. Diers, Ashlee N. Higdon, Karina C. Ricart, Michelle S. Johnson, Anupam Agarwal, Balaraman Kalyanaraman, Aimee Landar, Victor M. Darley-Usmar


Prototypical electrophiles such as the lipid 15-deoxy-Δ12,14-prostaglandin J2 (15d-PGJ2) are well recognized for their therapeutic potential. Electrophiles modify signalling proteins in both the cytosol and mitochondrion, which results in diverse cellular responses, including cytoprotective effects and, at high doses, cell death. These findings led us to the hypothesis that targeting electrophiles to specific compartments in the cell could fine-tune their biological effects. To examine this, we synthesized a novel mitochondrially targeted analogue of 15d-PGJ2 (mito-15d-PGJ2) and tested its effects on redox cell signalling. Mito-15d-PGJ2 caused profound defects in mitochondrial bioenergetics and mitochondrial membrane depolarization when compared with 15d-PGJ2. We also found that mito-15d-PGJ2 modified different members of the electrophile-responsive proteome, was more potent at initiating intrinsic apoptotic cell death and was less effective than 15d-PGJ2 at up-regulating the expression of HO-1 (haem oxygenase-1) and glutathione. These results demonstrate the feasibility of modulating the biological effects of electrophiles by targeting the pharmacophore to mitochondria.

  • apoptosis
  • 15-deoxy-Δ12,14-prostaglandin J2
  • haem(heme)
  • mitochondrion
  • redox signalling
  • thiol


Electrophilic compounds have attracted great interest as potential therapeutic agents in several fields, including cancer, cardiovascular disease and neurodegeneration [14]. Their primary mechanism of action occurs through their ability to covalently modify key proteins in redox cell signalling pathways [57]. Electrophiles can be derived from several sources including the diet, environment or endogenously through enzymatic or non-specific lipid peroxidation processes [4,8]. Endogenously generated electrophiles can be formed through enzymatic pathways, most notably via cyclo-oxygenase. Arachidonic acid oxidation results in the formation of a family of lipid prostaglandins, some of which are electrophilic in nature [9,10]. These electrophilic cyclo-oxygenase metabolites have been shown to possess anticancer, anti-inflammatory and cardioprotective properties [1113].

Indeed, one of the best characterized electrophilic lipids generated from cyclo-oxygenase is 15d-PGJ2 (15-deoxy-Δ12,14-prostaglandin J2). It is particularly interesting because of its high degree of specificity for modification of protein cysteine residues (thiols) in signalling proteins [14]. For example, 15d-PGJ2 has been shown to increase expression of phase II detoxification enzymes and cytoprotective intracellular antioxidants such as HO-1 (haem oxygenase-1) and the enzymes responsible for GSH synthesis [1416]. Although 15d-PGJ2 is also known as a PPARγ (peroxisome-proliferator-activated receptor γ) agonist, the induction of HO-1 and GSH requires the presence of the electrophilic reactive centre in the lipid and is not recapitulated by other non-reactive PPARγ agonists [15]. We have termed the highly selective family of proteins that form covalent adducts with electrophilic lipids the ‘electrophile-responsive proteome’ [17]. This is a discreet proteome given that of the several hundred proteins in a cell which have potentially reactive thiols only approx. 10% form covalent protein adducts with 15d-PGJ2.

In a therapeutic context, 15d-PGJ2 has been proposed as both a potential anticancer and anti-inflammatory agent because of its ability to regulate redox-sensitive aspects of angiogenesis, growth arrest, expression of inflammatory cytokines and cell death through the covalent modification of proteins [1,3]. For example, it has been shown to inhibit angiogenesis through the suppression of inflammatory enzymes and cytokines; this occurs through the direct modification of key components of the NF-κB (nuclear factor κ-light-chain-enhancer of the activated B-cell) signalling pathway. Specifically, the potent anti-inflammatory effects of 15d-PGJ2 are attributed to the covalent modification the p50 subunit of NF-κB by 15d-PGJ2, which results in the inhibition of its DNA-binding activity [18]. It was also found that 15d-PGJ2 modifies two cysteine residues within the C-terminal zinc finger of ER (oestrogen receptor)-α, and this modification inhibits its DNA-binding activity, decreases target gene expression and inhibits proliferation [19]. The cell-death effects of 15d-PGJ2 have been reported to occur through a pleiotropic mechanism involving activation of PPARγ and through interactions with mitochondrial proteins which lead to the activation of apoptosis [12,20]. It is well recognized that these redox signalling pathways may be important therapeutic targets; however, the major limitation has been the inability to selectively target the specific redox signalling pathways that regulate the beneficial effects of electrophiles.

Evidence clearly supports the fact that the protein targets of 15d-PGJ2 in the electrophile-responsive proteome are found in different compartments in the cell, including the cytosol and mitochondria, and this is consistent with reports that 15d-PGJ2 acts by pleiotropic mechanisms [18,21]. In work pioneered by Murphy and colleagues, a novel strategy for intracellular targeting of compounds to the mitochondrion has been developed. The conjugation of a delocalized, lipophilic cation to a compound of interest directs its accumulation within the mitochondrion and is sustained by the mitochondrial membrane potential [22,23]. This strategy has been employed to direct therapeutic antioxidants to the mitochondrion [2426]. On the basis of these previous studies, suggesting that the electrophilic modification of cytosolic targets are protective, whereas modification of mitochondrial proteins may regulate apoptosis, we hypothesized that targeting an electrophile to the mitochondria would enhance the apoptotic effects of this compound while limiting the induction of intracellular antioxidants.

To test this hypothesis, we synthesized a novel mitochondrially targeted analogue of 15d-PGJ2 (mito-15d-PGJ2) and examined its effects on apoptotic cell death and induction of intracellular antioxidants. We found that mito-15d-PGJ2 was more potent at initiating intrinsic apoptosis than 15d-PGJ2. Additionally, mito-15d-PGJ2 was less effective at up-regulating Keap1 (Kelch-like enoyl-CoA hydratase-associated protein 1)-dependent antioxidant expression (HO-1 and GSH) and caused profound defects in mitochondrial bioenergetics and mitochondrial membrane depolarization when compared with 15d-PGJ2. A mitochondrially targeted analogue of the non-electrophilic lipid prostaglandin E2, mito-PGE2, and the targeting moiety TPMP (methyltriphenylphosphonium) did not recapitulate these effects. Taken together, these results demonstrate for the first time the feasibility of activating specific redox signalling pathways with an electrophile by selectively targeting specific compartments within the cell.



All chemicals were of analytical grade and purchased from Sigma–Aldrich unless otherwise specified. 15d-PGJ2 and prostaglandin E2 were purchased from Cayman Chemicals. EZ-link 5-(biotinamido)pentylamine was purchased from Pierce Chemicals and was used for the synthesis of bt-15d-PGJ2 (biotinylated 15d-PGJ2) as described previously [6]. TPMP bromide was used as a mitochondrial-targeting control. 2-Amino-ethyl-TPMP bromide was provided by Dr B. Kalyanaraman.

Cell culture

MDA-MB231 human mammary adenocarcinoma cells were cultured in RPMI 1640 medium (Cellgro) containing 10% (v/v) FBS (fetal bovine serum). MCF10A-immortalized human mammary epithelial cells were cultured in MEGM (mammary epithelial cell growth medium) supplemented with MEGM SingleQuots (Lonza). MEFs (murine embryonic fibroblasts), derived from wild-type and Nrf2 (nuclear factor-erythroid 2-related factor 2)-null mice as described previously [27], were cultured in DMEM (Dulbecco's modified Eagle's medium) containing 10% (v/v) FBS and supplemented with 50 μM 2-mercaptoethanol. All experiments using MDA-MB231 cells were performed in RPMI 1640 medium with 0.5% FBS at approx. 50% confluence. Cells were cultured in six-well cluster plates except when used for bioenergetic measurements when specialized Seahorse Bioscience culture plates were used. For lipid exposures in six-well cluster plates 1 ml of medium was used (a 10 μM addition of lipid was then equivalent to approx. 160 pmols lipid per μg of protein). For lipid exposures in Seahorse Bioscience culture plates, 250 μl of medium was used (a 10 μM addition of lipid gave a similar concentration of approx. 125 pmols lipid per cell). Experiments utilizing MEFs and MCF10A cells were conducted at confluence in DMEM containing 0.5% FBS or MEGM respectively, with supplementation as described above.

Synthesis of mito-15d-PGJ2 and mito-PGE2

In order to target the lipids to the mitochondria, we utilized the lipophilic cation TPP+ (triphenylphosphonium) [28]. The procedure is a modification of a method reported previously [6,29] and is described in detail in the Supplementary Experimental section (available at HPLC, UV–visible spectral analysis, and mass spectrometric confirmation of synthesized lipids are shown in Supplementary Figures S1–S3 (available at Structures for the parent compound and synthesized lipids are also shown in Figure 1.

Figure 1 Structures of lipids used in the present study

Structures of 15d-PGJ2 and the mitochondrially targeted lipid derivatives mito-15d-PGJ2 and mito-PGE2. Electrophilic carbon centres are denoted with (*).

Assessment of cell viability

After lipid exposure, aliquots of the medium from the treatment wells were taken, and cells were harvested by scraping in PBS lysis buffer containing 0.1% Triton X-100. After centrifuging cell lysates (14000 g for 10 min) to remove debris, LDH (lactate dehydrogenase) activity in the supernatant and aliquots of cell medium was measured spectrophotometrically by monitoring the oxidation of 0.3 mM NADH at 340 nm. Apoptosis and necrosis were measured, after the indicated treatment had been performed for 16 h, by flow cytometric analysis using an Annexin V FITC apoptosis detection kit (Calbiochem) using a BD Biosciences LSR II flow cytometer. Briefly, gating parameters were set using PI (propidium iodide) only, Annexin V only and no staining controls, and 10000 events were collected for each experimental sample. Cells staining positively for both PI and Annexin V were considered late apoptotic. Cells staining positively for PI only or Annexin V only were scored as necrotic or early apoptotic respectively. Cells which were negative for both PI and Annexin V staining were scored as viable cells.

Glutathione measurement

The total glutathione (GSH and GSSG) level was determined in cell lysates, treated as indicated, using the recycling assay described previously [30]. Briefly, after treatment, cells were lysed in 10 μM DTPA (diethylenetriaminepenta-acetic acid) and 0.1% Triton X-100 in PBS, pH 7.4. Total glutathione in lysates was determined spectrophotometrically by monitoring the reduction of 5,5′-dithio-bis(2-nitro-benzoic acid) at 412 nm. Values were normalized to protein content assayed by the Bradford method (using a Bio-Rad Laboratories protein assay kit).

Western blot analysis

Cell lysate proteins were separated by SDS/PAGE (10% or 12.5% gels) and transferred on to nitrocellulose membranes (Bio-Rad Laboratories). Protein amounts were quantified by the Bradford method (using a Bio-Rad Laboratories protein assay kit), and equivalent amounts of protein were loaded. Uniform protein loading was verified by Ponceau S staining of the membranes, which showed no significant differences in protein levels on the blots among samples. The membranes were blocked with 5% (w/v) dried non-fat milk/TBS-T (Tris-buffered saline with 0.05% Tween 20) solution for 1 h at room temperature (24 °C), and then incubated with primary antibody overnight at 4 °C or for 3 h at room temperature. The antibody dilutions used are as follows: anti-HO-1 antibody (Stressgen), 1:1000; anti-caspase 9 antibody (Cell Signaling Technology), 1:1000; anti-VDAC (voltage-dependent anion channel) antibody (Mitosciences), 1:3000, anti-β-actin antibody (Cell Signaling Technology), 1:1000; and anti-COX (cytochrome c oxidase) subunit I antibody (Mitosciences), 1:1000. After washing with TBS-T, membranes were incubated with an HRP (horseradish peroxidase)-conjugated secondary antibody (GE Healthcare). Membranes were then developed using SuperSignal West Dura chemiluminescence substrate (Pierce Chemicals) and imaged using a charged-coupled device camera imaging system (AlphaInnotech).

Measurement of mitochondrial function

To measure mitochondrial function in intact MDA-MB231 cells, a Seahorse Bioscience XF24 Extracellular Flux Analyzer was used [31]. The optimal seeding density of MDA-MB231 cells needed to obtain a measurable OCR (oxygen consumption rate) and ECAR (extracellular acidification rate) was established, and both the ECAR and OCR show a proportional response with respect to the cell number (results not shown). For subsequent experiments a seeding density of 40000 cells per well was selected to allow both potential increases and the inhibition of OCR and ECAR to be assessed. We noted that high levels of FCCP [carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone] inhibited mitochondrial respiration presumably due to the loss of the ability to accumulate respiratory substrates. Accordingly, oligomycin, FCCP and antimycin A concentrations were optimized in order to elicit maximal effects in preliminary experiments (results not shown). The mitochondrial function assay employed in the present study uses sequential injections of oligomycin, FCCP and antimycin A to define a number of mitochondrial parameters such as basal OCR, ATP-linked OCR, proton leak, maximal respiratory capacity, reserve respiratory capacity and non-mitochondrial oxygen consumption. It is important to note that in order to calculate these parameters, we have assumed that the oligomycin-insensitive OCR is attributable to proton leak; however, oligomycin has been shown to hyperpolarize the mitochondrial membrane [32], and therefore the resulting OCR is probably an upper estimate of the contribution from proton leak. The mitochondrial inner membrane electrochemical potential (Δψ) was assessed using JC-1 (5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazolylcarbocyanine iodide; Invitrogen). Briefly, cells grown in 12-well culture plates were treated as described in the Figure legends, and 7.4 μM JC-1 was added directly to the MDA-MB231 cells in culture medium and incubated for 30 min. Next, cells were washed with PBS, and the red and green fluorescence was measured using a fluorescence plate reader. Results are expressed as the ratio of red to green fluorescence.

Protein adduct detection and mitochondrial fractionation

Following experimental incubations, MDA-MB231 lysates were analysed by two-dimensional IEF (isoelectric focussing) SDS/PAGE. Total cell lysates (50 μg) were separated in the first dimension on a pH 3–pH 10 gradient (IPG strips; Bio-Rad Laboratories) followed by resolution by SDS/PAGE (12.5% gels), and proteins were transferred on to nitrocellulose membranes as described previously [15]. After blocking for 1 h, the membranes were probed using an anti-TPP+ antibody (at a dilution of 1:10000; a gift from Professor M.P. Murphy, Medical Research Council Mitochondrial Biology Unit, Cambridge, U.K.) and with either an HRP-conjugated secondary antibody or with streptavidin conjugated to HRP (GE Healthcare) for detection of mito-15d-PGJ2 and bt-15d-PGJ2 adducts respectively. Membranes were developed as described above, and total protein was detected using Deep Purple stain (GE Healthcare). Blots were analysed using PDQuest software (Bio-Rad Laboratories). For mitochondrial fractionation, after treatment, MDA-MB231 cells were harvested by scraping into ice-cold HMIM (heart mitochondria isolation medium) buffer (20 mM Tris/HCl, pH 7.35, 300 mM sucrose and 2 mM EGTA at 4 °C) containing protease inhibitor cocktail (Roche) and 1 mM PMSF as described in detail in the Supplementary Experimental section. Protein adducts were detected using reagents described above. Membranes were developed using ECL+ (enhanced chemiluminescence; GE Healthcare) and chemifluorescence was imaged using a Typhoon fluorescence detector (GE Healthcare).

Statistical analysis

Results are reported as means±S.E.M. for n≥3, as indicated in the Figure legends. Statistical significance was evaluated by one-way ANOVA among the groups using GraphPad Prism 4. The minimum level of significance was set at P<0.05. Tukey's multiple comparison test was used for post-hoc analysis of significance between groups.


Cytotoxicity of 15d-PGJ2 and mito-15d-PGJ2

It is established that lipid electrophiles at low concentrations induce phase II cytoprotective enzymes, but at higher concentrations initiate cell death [16]. For example, the parent compound in these studies, 15d-PGJ2, has been shown to induce apoptotic cell death in a number of cancer cell lines at doses ranging from 5 to 50 μM [3335]. To assess cytotoxicity of mito-15d-PGJ2 in human mammary cells, non-tumorigenic MCF10A mammary epithelial cells and MDA-MB231 mammary adenocarcinoma cells were exposed to 3–30 μM mito-15d-PGJ2 for 16 h and then cell death was assessed by using LDH release as an indicator of late apoptotic or necrotic cell death. Treatment of both cell lines resulted in significant cell death at concentrations at or above 10 μM (Figure 2A). Additional examination of the cell-death effects of mito-15d-PGJ2 relative to 15d-PGJ2 were conducted in MDA-MB231 cells. We next determined the relative cytotoxicity of mito-15d-PGJ2 by treating MDA-MB231 cells with increasing concentrations (0.3–30 μM) of either mito-15d-PGJ2 or 15d-PGJ2 for 16 h. Cytotoxicity was assessed using FACS analysis, after staining with PI and Annexin V. Mito-15d-PGJ2 treatment resulted in a greater decrease in viable cells (those that are not positive for either Annexin V or PI) at lower concentrations when compared with 15d-PGJ2 (Figure 2B). Hardly any early and late apoptotic cells were detectable after exposure to 15d-PGJ2, whereas this represented the majority of the cell population after exposure to mito-15d-PGJ2 (Figures 2C and 2D). In contrast, 15d-PGJ2 induced predominantly necrotic cell death at 30 μM, whereas this represented only 3–10% of the cells treated with mito-15d-PGJ2 across the dose range (Figure 2E).

Figure 2 Effect of 15d-PGJ2 and mitochondrially targeted compounds on cell viability and the induction of apoptosis

(A) Viability of MCF10A and MDA-MB231 cells exposed to increasing concentrations (3–30 μM) of mito-15d-PGJ2 for 16 h was determined by LDH release. (BE) Apoptosis and necrosis after exposure to 15d-PGJ2 or mito-15d-PGJ2 (0.3–30 μM for 16 h) were determined by flow cytometry using PI and Annexin V staining. The percentage of cells which are (B) viable (PI negative and Annexin V negative), (C) early apoptotic (PI negative and Annexin V positive), (D) late apoptotic (PI positive and Annexin V positive) and (E) necrotic (PI positive and Annexin V negative) are shown. (F) Caspase 9 activation in MDA-MB231 cells treated with 10 μM 15d-PGJ2, mito-15d-PGJ2, mito-PGE2 or TPMP was assessed using Western blot analysis of pro-caspase 9 and the levels were quantified. Staurosporine (1 μM) was used as a positive control. Equivalent amounts of protein were loaded for each sample and equal protein loading was confirmed after transfer on to nitrocellulose membranes by using Ponceau S staining. A representative Western blot (upper panel) is shown. Results in the lower panel represent means±S.E.M., n=3. *P<0.05 compared with EtOH (ethanol vehicle control). **P<0.01 compared with EtOH. #P<0.05 compared with 15d-PGJ2.

As a further confirmation that mito-15d-PGJ2 causes apoptotic cell death, MDA-MB231 cells were treated with 10 μM 15d-PGJ2 or mito-15d-PGJ2 for 4 h, and then cell lysates were probed for pro-caspase 9, a protein which is cleaved upon the activation of the intrinsic (mitochondrial) apoptotic signalling. Treatment with 15d-PGJ2 resulted in a partial loss, or activation, of pro-caspase 9, whereas mito-15d-PGJ2 caused complete loss of this apoptotic mediator (Figure 2F). Mito-PGE2 and TPMP were used as controls for the effect of the TPP+ moiety and electrophilicity and had no effect on the levels of pro-caspase 9. Staurosporine (1 μM) was used as a positive control for caspase activation. Taken together these results show that mito-15d-PGJ2 induces cell death primarily through apoptosis, and more robustly than 15d-PGJ2, which is consistent with the hypothesis that mitochondrial targeting then biases the mechanism of cell death towards apoptosis over necrosis.

Induction of the intracellular antioxidants HO-1 and GSH by 15d-PGJ2 and mito-15d-PGJ2

It has been shown that the modification of critical thiols in the cytosolic protein Keap1 by 15d-PGJ2 results in the translocation of the transcription factor Nrf2 to the nucleus and subsequent transcription of genes under the control of the EpRE (electrophile-response element) [15,36]. Such genes include HO-1 and subunits of GCL (glutamyl cysteine ligase), which controls the production of GSH [37]. Accordingly, we next investigated the effect of 15d-PGJ2 and mito-15d-PGJ2 on the induction of HO-1 and GSH. MDA-MB231 cells were treated with increasing concentrations (0.3–10 μM) of 15d-PGJ2 or mito-15d-PGJ2 for 4 h and cell lysates were analysed for HO-1 protein levels. 15d-PGJ2 caused a dose-dependent increase in HO-1 protein from 0.3–3 μM and a slight decrease at higher concentrations (Figure 3A), similar to the biphasic dose–response we have described previously [16]. However, mito-15d-PGJ2 induced HO-1 protein expression to a much lesser extent. TPMP (10 μM) was used as a control for the effect of the TPP+ moiety and had no effect on HO-1 levels.

Figure 3 Differential induction of HO-1 and GSH by 15d-PGJ2 and mito-15d-PGJ2

MDA-MB231 cells were treated with increasing concentrations (0.3–10 μM) of 15d-PGJ2 and mito-15d-PGJ2. (A) Cells were treated for 4 h then HO-1 protein levels were determined using Western blot analysis and the levels were quantified (lower panel). Equivalent amounts of protein were loaded for each sample and equal protein loading was confirmed after transfer on to nitrocellulose membranes by using Ponceau S staining. A representative Western blot is shown (upper panel). (B) Cells were treated for 4 h, then compound-containing medium was removed and RPMI 1640 medium containing 0.5% FBS was added for an additional 16 h. Cell lysates were then analysed for total glutathione content and values were normalized to the total protein level. Results represent means±S.E.M., n=3. **P<0.01 compared with ethanol vehicle control. #P<0.05 compared with 15d-PGJ2.

As an additional readout of the effect of the electrophiles on the Keap1/Nrf2 system, GSH was also assessed by treating MDA-MB231 cells with increasing concentrations (0.3–10 μM) of 15d-PGJ2 or mito-15d-PGJ2 for 4 h. Compound-containing medium was then removed to attenuate cytotoxicity, and the cells were incubated for a further 16 h to allow time for the de novo synthesis of GSH. Similar to the result with HO-1, 15d-PGJ2 treatment resulted in a significant increase in total GSH levels; however, mito-15d-PGJ2 was unable to induce GSH (Figure 3B).

Nrf2 dependence of the regulation of antioxidant expression and cell death by 15d-PGJ2

We have reported previously that the induction of HO-1 by 15d-PGJ2 can be inhibited by modification of mitochondrial protein thiols [16]. To test the possibility that mito-15d-PGJ2 is functioning through a similar mechanism, cells were pretreated with mito-15d-PGJ2 for 4 h then 15d-PGJ2 was added to stimulate HO-1 and GSH levels. Mito-15d-PGJ2 does not impair the 15d-PGJ2-stimulated induction of HO-1, indicating that is not inhibiting the mitochondrial component required for HO-1 synthesis (Figure 4A). Interestingly, in the combined treatment group, mito-15d-PGJ2 pretreatment enhances HO-1 induction by 15d-PGJ2. These results are consistent with previous studies showing that induction of HO-1 occurs by a number of mechanisms, some of which appear to be independent of Keap1/Nrf2 [38]. In contrast, the induction of GSH appears to require Nrf2 and, consistent with this finding, pretreatment with mito-15d-PGJ2 does not impair 15d-PGJ2-dependent induction of GSH (Figure 4B).

Figure 4 Nrf2 dependence of the regulation of antioxidant expression and cell death

(A) MDA-MB231 cells were pretreated with 1 μM mito-15d-PGJ2 for 4 h, then 1 μM 15d-PGJ2 was added for an additional 4 h. HO-1 protein levels were determined using Western blot analysis and the levels quantified. Equivalent amounts of protein were loaded for each sample and equal protein loading was confirmed after transfer on to the nitrocellulose membranes by using Ponceau S staining. A representative Western blot is shown (upper panel). (B) Cells were treated as described in (A), then the compound-containing medium was removed and RPMI 1640 medium containing 0.5% FBS was added for an additional 16 h. Cell lysates were then analysed for total glutathione content and values were normalized to the protein level. (C) Viability of MDA-MB231 cells co-incubated with 10 μM mito-15d-PGJ2 and 3 μM 15d-PGJ2 for 16 h was determined by the level of LDH release. (D) Nrf2 wild-type and Nrf2-null MEFs were treated with 15d-PGJ2, mito-15d-PGJ2 or mito-PGE2 (3 and 10 μM) for 16 h then LDH release was measured. Results represent means±S.E.M, n=3–6. *P<0.05 compared with ethanol vehicle control. **P<0.01 compared with ethanol vehicle control. #P<0.05 compared with 15d-PGJ2.

It has been established that activation of the Keap1/Nrf2 system by low levels of electrophilic stress results in increased resistance to cytotoxic levels of reactive oxygen and nitrogen species. It is then possible that the failure of mito-15d-PGJ2 to activate the EpRE may contribute to its cytotoxicity. To test this, two different experimental strategies were employed. First, MDA-MB231 cells were co-incubated with a cell-death-inducing concentration of mito-15d-PGJ2 (10 μM) and a concentration of 15d-PGJ2 which increases antioxidant expression (3 μM). After a 16 h treatment, cell death was assessed by LDH release and was unchanged by co-treatment with 15d-PGJ2 (Figure 4C). This is consistent with the early induction of the apoptotic process before significant induction of phase II cytoprotective enzymes can occur. Secondly, wild-type and Nrf2-null MEFs were treated with 3 or 10 μM 15d-PGJ2 or mito-15d-PGJ2 and the level of cell death was assessed. If the cytotoxicity was attenuated by activation of the Keap1/Nrf2 system, we would have expected the Nrf2-null cells to be sensitized to the electrophiles; however, we found that cytotoxicity of 15d-PGJ2 and mito-15d-PGJ2 was independent of Nrf2 levels in the cell (Figure 4D).

Protein modification by bt-15d-PGJ2 and mito-15d-PGJ2 are different

The results thus far demonstrating the different biological properties of 15d-PGJ2 and mito-15d-PGJ2 suggest that the proteins modified by the electrophilic lipids will also differ. To test this, MDA-MB231 cells were incubated with 10 μM bt-15d-PGJ2 or mito-15d-PGJ2 for 1 h, and the protein lysates separated by two-dimensional IEF SDS/PAGE followed by Western blotting and detection for protein–lipid adducts using the anti-TPP+ antibody or streptavidin conjugated to HRP. The patterns of protein stains for all treatments showed no significant differences (results not shown). Control cells not treated with electrophilic lipid showed no major positive reactions with the exception of one spot on the biotin blot (Figures 5A and 5B), which we ascribe to an endogenous biotin-containing carboxylase [6]. Figure 5 shows detection of approx. 63 proteins that are positive for biotin (Figure 5C) and approx. 32 proteins that are positive for TPP (Figure 5D). The images were selected to minimize differences in potential sensitivity between the two detection techniques by selecting images with the same intensity for a common spot (indicated by an arrow on Figures 5C and 5D). The overlay of these two images was achieved using the PDQuest proteomics software (Figure 5E) and shows red spots that are predominantly reactive with bt-15d-PGJ2 and green spots that are predominantly reactive with mito-15d-PGJ2, so that the proteins that react with both lipids are yellow. In Figure 5(F), these results are summarized and indicate that ten proteins are common targets of both electrophilic lipids, with 22 targets unique to mito-15d-PGJ2 and 53 targets unique to 15d-PGJ2. It was not possible to unequivocally overlay the protein adduct maps with the protein stain to identify these targets as many of these proteins are present in the cell at low abundance [6].

Figure 5 Visualization of bt-15d-PGJ2- and mito-15d-PGJ2-modified proteins by two-dimensional IEF SDS/PAGE

Cell lysates from MDA-MB231 cells were separated by two-dimensional IEF SDS/PAGE. Ethanol-vehicle-control-treated (1 h) cell lysates were probed using (A) streptavidin–HRP and (B) anti-TPP+ antibody with an HRP-conjugated secondary antibody to determine background reactivity. Cells were treated (10 μM for 1 h) with (C) bt-15d-PGJ2 or (D) mito-15d-PGJ2 and protein adducts were detected using streptavidin–HRP and anti-TPP+ antibody respectively. Arrows indicate common spots with the same intensity that were used to select images. (E) A merged image of the bt-15d-PGJ2 (red) and mito-15d-PGJ2 (green) adduct blots. (F) Analysis of spot quantity was determined and the number of common and unique spots from bt-15d-PGJ2 and mito-15d-PGJ2 membranes is shown schematically.

15d-PGJ2 and mito-15d-PGJ2 localize to mitochondria and affect mitochondrial membrane potential

To assess the relative localization and formation of protein adducts by the lipid electrophiles, mitochondrial fractions were isolated from cells treated with bt-15d-PGJ2 or mito-15d-PGJ2. Validation of the fractionation protocol used demonstrates substantial enrichment of the mitochondrial proteins in the mitochondrial fraction compared with the cytosolic fraction (see Supplementary Figure S4 available at

Protein adduct formation was assessed in fractionated samples from bt-15d-PGJ2- and mito-15d-PGJ2-treated cells using one-dimensional SDS/PAGE. Consistent with our previous results [21], bt-15d-PGJ2 localizes to the mitochondria and forms stable protein adducts. Mito-15d-PGJ2 protein adducts are also enriched in the mitochondrial fraction (Figure 6A). Mitochondrial purity was also confirmed by the VDAC enrichment in the mitochondrial fraction. Protein adducts in each fraction were quantified and expressed as the fold increase in protein adducts in the mitochondrial fraction compared with the homogenate (Figure 6B). These results demonstrate that there is a significant increase in mito-15d-PGJ2–protein adducts in the mitochondria compared with the homogenate. There was also a trend towards an increase in mitochondrial protein adducts with bt-15d-PGJ2 treatment; however, this result was not significant.

Figure 6 Localization of 15d-PGJ2 and mito-15d-PGJ2 to the mitochondrion and effects on mitochondrial membrane potential

(A) MDA-MB231 cells were treated with 10 μM bt-15d-PGJ2 (bt-15d) or mito-15d-PGJ2 (Mito-15d) for 1 h then mitochondrial and cytosolic fractions were prepared and resolved using SDS/PAGE (10% gels). Adduct formation was detected using an anti-TPP+ antibody (for mito-15d-PGJ2) or streptavidin–HRP (for bt-15d-PGJ2). ECL+ coupled to HRP was used to detect chemifluorescence protein adducts. VDAC protein levels were also determined using Western blot analysis (lower panels) to confirm mitochondrial preparation purity. (B) Protein adducts were quantified and expressed as the fold-enrichment over the homogenate. EtOH, ethanol vehicle control. (C) MDA-MB231 cells were exposed to increasing concentrations (1–10 μM) of 15d-PGJ2, mito-15d-PGJ2 or TPMP for 4 h, then the mitochondrial membrane potential was assessed using JC-1. Results are represented as the ratio of red/green fluorescence and are means±S.E.M, n=3–6. *P<0.05 compared with EtOH. **P<0.01 compared with EtOH. #P<0.05 compared with 15d-PGJ2.

We next investigated whether the modification of mitochondrial proteins by lipid electrophiles altered mitochondrial function by determining the effect of 15d-PGJ2 and mito-15d-PGJ2 on mitochondrial membrane potential (Δψ). MDA-MB231 cells were treated for 4 h with increasing concentrations (1–10 μM) of 15d-PGJ2 or mito-15d-PGJ2 then JC-1 fluorescence was assessed. Treatment with 15d-PGJ2 resulted in a dose-dependent decrease in the mitochondrial membrane potential. However, mito-15d-PGJ2 caused a more significant depolarization of the mitochondrial membrane (Figure 6C). TPMP was used as a control for the effect of the TPP+ moiety and did not cause mitochondrial membrane depolarization.

Measurement of mitochondrial function after exposure to 15d-PGJ2 and mito-15d-PGJ2

To assess cellular bioenergetics in intact MDA-MB231 cells, the Seahorse Bioscience XF24 Extracellular Flux Analyzer was used to determine the rates of change in both pH and oxygen in the medium surrounding the cells [31,39]. We have utilized a mitochondrial function protocol (schematically represented in Figure 7A). This protocol is a modification of one described previously [40]. Initially, basal OCR is measured and, after 20 min, the inhibitor of the mitochondrial ATP synthase, oligomycin (0.3 μg/ml), is added. This causes a decrease in OCR which we ascribe to the mitochondrial activity utilized for ATP synthesis. The remaining OCR can be ascribed to all processes which allow ion movement across the mitochondrial inner membrane, and we collectively term this proton leak. Next, the maximal mitochondrial function that can be sustained in the cells with endogenous substrates is assessed by the addition of the proton ionophore FCCP (1 μM). As expected, this stimulates oxygen consumption, as it is no longer constrained by the proton gradient in the mitochondria. The difference between the basal OCR and this maximal rate is then termed the reserve capacity of the mitochondrial function in the cells under these conditions. Finally, the complex III inhibitor, antimycin A, is added to assess the oxygen consumption of COX, the terminal member of the electron transport chain. The remaining low level of OCR can be ascribed to partial reduction of oxygen and is therefore ascribed to ROS (reactive oxygen species) formation.

Figure 7 Effect of 15d-PGJ2, mito-15d-PGJ2 and mito-PGE2 on basal and maximal respiration capacity

MDA-MB231 cells were treated for 30 min with 10 μM 15d-PGJ2, mito-15d-PGJ2 or mito-PGE2 in specialized 24-well Seahorse Biosciences V7 microplates, then bioenergetic function was assessed using the Seahorse Biosciences XF24 Analyzer. (A) A schematic diagram demonstrating the use of specific inhibitors to determine the sites of cellular oxygen consumption is shown. (B) Basal OCR and (C) maximal OCR for the treatments are shown. EtOH, ethanol vehicle control. (D) The proportion of maximal respiration utilized for ATP-linked respiration (ATP-linked), proton leak (Proton), and ROS formation (ROS), and the reserve respiratory capacity (Reserve) was calculated after measurement of the OCR using the Seahorse Biosciences XF24 Analyzer after sequential injections of oligomycin (0.3 μg/ml), FCCP (1 μM) and antimycin A (10 μM) from MDA-MB231 cells treated as described. (E) Basal ECAR was also measured concomitantly to the experiment in (D). The Seahorse Biosciences XF24 Analyzer protocol included 2 min of mixing, 2 min of waiting and 3 min of measurement times for each measurement. (F) Caspase 9 activation was assessed by Western blot analysis of pro-caspase 9 in samples harvested from a parallel plate, treated as described. A representative Western blot is shown. Results represent means±S.E.M, n=5. **P<0.01 compared with EtOH. #P<0.05 compared with 15d-PGJ2.

Using this protocol, mitochondrial function was measured in confluent MDA-MB231 cells which were first exposed to 10 μM 15d-PGJ2, mito-15d-PGJ2 or mito-PGE2 for 30 min. The lipids were then removed and the mitochondrial parameters measured. There was no significant effect of 15d-PGJ2 or mito-PGE2 on basal or maximal respiration; however, profound decreases in basal and maximal OCR were seen with the mito-15d-PGJ2 treatment (Figures 7B and 7C). The analysis of the different components of OCR, as described above, was also performed and the results are shown in Figure 7(D). It is clear that 15d-PGJ2 has little effect on mitochondrial function, whereas mito-15d-PGJ2 suppresses all bioenergetic parameters. In addition to the measurement of OCR, the ECAR, a surrogate marker of glycolysis, was also determined (Figure 7E). Under unstimulated conditions, ECAR represents predominantly lactate production from glycolysis. Treatment with either 15d-PGJ2 or mito-PGE2 resulted in no change in ECAR; however, after a 30 min exposure to mito-15d-PGJ2, ECAR is significantly increased. It is important to note that at this time point apoptosis has not yet been activated (Figure 7F) thus these mitochondrial bioenergetic defects precede apoptosis. Taken together these results show that mito-15d-PGJ2 modifies mitochondrial function, whereas 15d-PGJ2 has a minimal effect on any of the bioenergetic parameters assessed. Moreover, glycolytic flux, as indicated by an elevated ECAR, is increased with mito-15d-PGJ2, which may represent a compensatory mechanism to overcome lost mitochondrial function.


The control of redox signalling in cells is now emerging as a key area for our understanding of both the aetiology of disease and the development of novel therapeutics. It is recognized that redox signalling offers potential therapeutic targets; however, the major limitation has been the inability to selectively activate or inhibit a specific pathway [41]. To our knowledge, this is the first demonstration of selective targeting of a redox signalling pathway by modifying an electrophilic lipid to target the mitochondrion. We selected the compound 15d-PGJ2 for these studies because of its recognized potential as a therapeutic agent and its well-established ability to modify proteins and activate cell signalling in both the cytosol and mitochondrion. We have demonstrated that by conjugating the delocalized cation TPP+ to 15d-PGJ2 we are able to enhance its cytotoxicity by promoting apoptosis over necrosis when compared with the parent compound (Figure 2). It is likely that the mechanism through which this occurs is the covalent modification of proteins as the non-electrophilic lipid analogue to mito-15d-PGJ2, mito-PGE2, has no effect on cell death or caspase 9 activation. There are multiple potential mechanisms by which mito-15d-PGJ2 could enhance apoptosis in cells. It is unlikely that mito-15d-PGJ2 activates PPARγ-dependent cell death as PPARγ is localized primarily in the cytosol and blocking the carboxy functional group on the lipid leads to loss of its PPARγ-dependent properties [42,43].

As mito-15d-PGJ2 is added to the outside to the cell it will first accumulate in the cytosol, on the basis of the membrane potential across the plasma membrane, and only then will it cross the inner mitochondrial membrane and accumulate in the mitochondrion [22,44]. Consistent with this model, we found that the protein adducts presented a distinct pattern when separated by two-dimensional IEF SDS/PAGE. As expected, the electrophile-responsive proteome for mito-15d-PGJ2 shares some common targets with 15d-PGJ2, but a far greater number of proteins (53 for bt-15d-PGJ2 and 22 for mito-15d-PGJ2) are distinct (Figure 5). The fact that mito-15d-PGJ2 has substantially fewer protein targets is consistent with our hypothesis that the sub-proteomes modified by reactive electrophiles are cell-domain-specific. There are, however, some limitations to this analysis of the electrophilic proteome. For example, not all the modified proteins are capable of entering the second dimension of the two-dimensional IEF SDS/PAGE due to precipitation in the isoelectric strip (results not shown). Nevertheless, it is clear from the proteins that are represented that the patterns are distinct. Moreover, as expected, modified proteins were present in both the cytosolic and mitochondrial fractions of cells exposed to both mito-15d-PGJ2 and the untargeted analogue (Figure 6). Fractionation of the cells after exposure to the electrophiles resulted in a greater enrichment of adducted proteins, relative to the unfractionated sample, for mito-15d-PGJ2 compared with 15d-PGJ2. In further support of a selective impact on mitochondrial function by the addition of the TPP+ group to 15d-PGJ2, we found that mito-15d-PGJ2 caused a profound mitochondrial defect in oxidative phosphorylation at concentrations where 15d-PGJ2 had no discernable effect (Figure 7).

There are several potential mechanisms for the increased toxicity of mito-15d-PGJ2. First, mito-15d-PGJ2 may more extensively modify mitochondrial targets of the parental compound 15d-PGJ2. We have reported previously that 15d-PGJ2 modifies components of the mitochondrial permeability transition pore, including ANT (adenine nucleotide translocator), and promotes permeability transition in endothelial cells [20]. There is also the potential that mito-15d-PGJ2 enhances cell death by limiting the cytoprotective response of cells to the electrophilic stress they experience. Nrf2-dependent gene transcription results in an increase in the overall antioxidant capacity of cells and has recently been demonstrated to also initiate anti-apoptotic responses [45]. A last potential mechanism for increased cytotoxicity of mito-15d-PGJ2 maybe through the modification of new mitochondrial protein targets as a result of the targeting of 15d-PGJ2. Our results suggest that the loss of Nrf2-dependent gene transcription does not explain the enhanced cytotoxicity of mito-15d-PGJ2 (Figure 4). Co-incubation with an Nrf2-activating concentration of 15d-PGJ2 does not protect against mito-15d-PGJ2-induced cell death. Moreover, wild-type and Nrf2-null MEFs have similar cytotoxicity profiles, pointing to the fact that Nrf2 is not a key mediator of the cell death phenotype on acute exposure to an electrophile. However, it is important to note that induction of antioxidant defences with low non-toxic concentrations of electrophiles has been shown to be essential for the protection against subsequent oxidative stress [46].

The fact that the loss of Nrf2 signalling does not appear to be a mediator of the enhanced cell-death effects of mito-15d-PGJ2 does not preclude the role of mitochondrial thiol modification as a mechanism. We have demonstrated that both mito-15d-PGJ2 and 15d-PGJ2 localize to the mitochondrion and, in two-dimensional protein adduct formation analyses, we have identified a number of common protein targets for the two compounds, which are probably the cell-death targets (Figures 5 and 6). Interestingly, we also show that mito-15d-PGJ2 has profound effects on mitochondrial bioenergetics that are not observed with untargeted 15d-PGJ2 (Figure 7). These results suggest a potential role for both the extent of modification of common targets and a gain-of-function activity of mito-15d-PGJ2, through the modification of unique mitochondrial protein targets, that results in impaired mitochondrial respiration and may contribute to cell death. Consistent with this idea, our results also demonstrate that the concentrations at which mitochondrial function is impaired are also those which are required to initiate apoptosis.

We reasoned that if mito-15d-PGJ2 was accumulated in the mitochondrion it would be less effective at activating cytosolic signalling pathways. To test this we chose to measure the level of initiation of the EpRE-dependent genes, such as HO-1 or the proteins controlling GSH synthesis, by the electrophilic lipids [37]. We found that the ability of mito-15d-PGJ2 to up-regulate HO-1 is greatly attenuated and for GSH that it is essentially abolished when compared with 15d-PGJ2 (Figure 3).

In the present study, we have demonstrated the feasibility of targeting specific cell responses by targeting a subcellular proteome (the mitochondrion). We successfully targeted the cellular response of cell death and selected against another (antioxidant up-regulation). It is important to note that the addition of the TPP+ moiety to 15d-PGJ2 may have an impact on the electrophilic-responsive proteome through mechanisms distinct from mitochondrial targeting. For example, it probably changes the lipophilicity of the parent electrophile and may also provide additional steric factors which may change the reactivity to target proteins. Taken together these results indicate that both the reactivity of the thiol proteome and the physical chemical properties of the electrophile will determine the specific electrophile-responsive sub-proteomes which are modified and thus the biological responses. This has important implications for both the understanding of the basic mechanisms through which electrophiles mediate redox cell signalling and the potential to refine the protein targets of electrophiles through intracellular targeting strategies.


Anne Diers, Ashlee Higdon, Karina Ricart, and Michelle Johnson performed experiments. Anne Diers, Ashlee Higdon, Michelle Johnson, Aimee Landar and Victor Darley-Usmar designed experiments and analysed results. Anupam Agarwal provided MEFs. B. Kalyanaraman provided reagents for synthesis of mito-15d-PGJ2. All authors contributed to the preparation of the manuscript.


This work was supported by the National Institutes of Health [grant numbers ES10167; DK 75865 (both to V.D.U.); DK 079337 (to A.A.); T32 HL007918 (to A.R.D.)]; and by the American Heart Association through the Scientist Development programme [grant number 0635361N (to A.L.)] and a Predoctoral Fellowship [grant number 0815177E (to A.N.H.)].


We are grateful to Dr J. Chan (Department of Pathology, School of Medicine, University of California-Irvine, Irvine CA, U.S.A.) for the Nrf2 wild-type and knockout MEFs and to Professor M. P. Murphy (Medical Research Council Mitochondrial Biology Unit, Cambridge, U.K.) for the anti-TPP+ antibody.

Abbreviations: 15d-PGJ2, 15-deoxy-Δ12,14-prostaglandin J2; bt-15d-PGJ2, biotinylated 15d-PGJ2; COX, cytochrome c oxidase; DMEM, Dulbecco's modified Eagle's medium; ECAR, extracellular acidification rate; EpRE, electrophile-response element; FBS, fetal bovine serum; FCCP, carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone; HO-1, haem oxygenase-1; HRP, horseradish peroxidase; IEF, isoelectric focusing; JC-1, 5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazolylcarbocyanine iodide; Keap1, Kelch-like enoyl-CoA hydratase-associated protein 1; LDH, lactate dehydrogenase; MEF, murine embryonic fibroblast; MEGM, mammary epithelial cell growth medium; mito-15d-PGJ2, mitochondrially targeted 15-d-PGJ2; mito-PGE2, mitochondrially targeted prostaglandin E2; NF-κB, nuclear factor κ-light-chain-enhancer of the activated B-cell; Nrf2, nuclear factor-erythroid 2-related factor 2; OCR, oxygen consumption rate; PI, propidium iodide; PPARγ, peroxisome-proliferator-activated receptor γ; ROS, reactive oxygen species; TBS-T, Tris-buffered saline with 0.05% Tween 20; TPMP, methyltriphenylphosphonium; TPP+, triphenylphosphonium; VDAC, voltage-dependent anion channel


View Abstract