TNFα-induced lysosomal membrane permeability is downstream of MOMP and triggered by caspase-mediated NDUFS1 cleavage and ROS formation Free

TNFα is a pleiotropic inflammatory cytokine, which was discovered decades ago (Carswell et al., 1975; Pennica et al., 1984). Since then, it has been shown to exert many functions in various physiological and pathological processes. In particular, its implication in the survival and death of normal and cancerous cells has been widely investigated. For most mammalian cells, TNFα is non-cytotoxic, but triggers, after ligation with TNF-receptor 1 at the membrane, a TRADD–TRAF2–RIP1–c-IAP1/2 complex (complex I)-dependent signaling pathway that ends up in the activation of the transcription factor NF-κB (Walczak, 2011; Micheau and Tschopp, 2003; Andrew, 2004; Wang et al., 2008). NF-κB in turn induces genes for cell survival (such as those encoding Bcl-x_L, c-IAPs, c-FLIP, etc.) and pro-inflammatory mediators (TNFα itself, COX-2, iNOS, etc.). Upon c-IAPs inhibition by so called Smac mimetics, low expression of the caspase-8 inhibitor c-FLIP, or protein synthesis inhibition by cycloheximide (CHX), complex I dissociates from the membrane and forms in the cytosol a second complex (complex II) which now includes caspase-8 and its activator FADD to cleave the BH3-only protein Bid to tBid (Micheau and Tschopp, 2003; Andrew, 2004; Wang et al., 2008). tBid activates mitochondrial outer membrane permeabilization (MOMP) via Bax/Bak (i.e. Bax and Bak) activation leading to the release of cytochrome c, the formation of the Apaf-1–caspase-9 apoptosome and the processing and activation of effector caspases-3/7 (i.e. caspase-3 and -7), the main executioners of apoptotic programmed cell death (Kroemer et al., 2007). Recently, another type of TNFα-induced cell death, called programmed necrosis or necroptosis has been discovered when caspases are inhibited by the pan-caspase inhibitor Z-VAD-fmk. This pathway includes RIP1 and other novel signaling components (RIP3, MLK, etc.) whose functions are only now being unveiled (Degterev and Yuan, 2008, He et al., 2009; Zhang et al., 2009).

A characteristic feature of TNFα-induced apoptotic and necroptotic cell death is the production of reactive oxygen species (ROS) such as superoxide anions (O₂⁻·) and hydrogen peroxide. The intracellular source of these ROS, their link to cell death signaling components and their exact role in the effector phase of apoptosis or necroptosis have been largely debated (Van Herreweghe et al., 2010; Taujimoto et al., 1986; Meier et al., 1989; Yazdanpanah et al., 2009; Kim et al., 2010; Woo et al., 2000). While in macrophages and neutrophils TNFα triggers ROS production as a quick oxidative burst required for an early innate immune response, in fibroblasts and other nonphagocytic cells this production occurs continuously over a time period of hours (Van Herreweghe et al., 2010; Taujimoto et al., 1986; Meier et al., 1989; Yazdanpanah et al., 2009). This suggests that phagocytic and nonphagocytic cells may exploit distinct molecular machineries for ROS generation. In innate immune cells TNFα is known to stimulate the activity of the plasma membrane-residing Nox2/gp91 phox subunit of NADPH oxidase which produces superoxides within seconds in order to kill invasive microorganisms (Yazdanpanah et al., 2009; Lambeth, 2004). Interestingly, after treating with TNFα, mouse fibroblasts can also produce ROS via NADPH oxidase. In this case, ROS generation depends on a signaling complex composed of TRADD, RIP1, Nox1 and the small GTPase Rac1 (Lin et al., 2004). Other reports further implicate calcium mobilization, activation of the Rac-cPLA2-LTB₄ cascade and activation of acid sphingomyelinases in this process (Taujimoto et al., 1986; Yazdanpanah et al., 2009; Woo et al., 2000; Shoji et al., 1995). Due to the role of RIP1, it was suggested that this pathway of ROS production is crucial for necrotic cell death. However, the other important component RIP3, which forms a pro-necrotic complex with RIP1 after phosphorylation (Cho et al., 2009), did not activate NADPH oxidase but enzymes controlling glycolytic flux and glutaminolysis, thereby causing ROS formation via mitochondrial complex I (Zhang et la., 2009; Vandenabeele et al., 2010). Hence for necroptosis, and probably also for apoptosis, ROS may be rather produced by mitochondria, preferentially at the ubisemiquinone site (Schulze et al., 1993). To trace mitochondrial ROS, the Murphy group succeeded to generate synthetic compounds, called MitoSOX™ and MitoQ, which after removal of the targeting sequence by mitochondrial esterases, are stably inserted into the inner mitochondrial membrane and act as ROS detection probe and antioxidant, respectively (James et al., 2005; Smith et al., 2011).

Mitochondrial ROS production has been closely linked to a fall in the mitochondrial membrane potential. Although this process is seen in all apoptotic cells, including when cells die in response to TNFα, it is not an early event of cell death signaling but rather a consequence of Bax/Bak-mediated MOMP. Ricci et al. convincingly showed that caspase-3, which is activated after MOMP, cytochrome c release and apoptosome formation, feeds back on complex I of the respiratory chain by cleaving the p75 NDUFS1 subunit of this complex and thereby provoking a fall in the mitochondrial membrane potential and ROS production (Ricci et al., 2004). However, this feedback loop has not yet been investigated for TNFα. By contrast, a recent study using HEK 293, HeLa and cervix carcinoma cells reported that TNFα-induced mitochondrial ROS production was mediated by the recruitment of the survival factor Bcl-x_L to a mitochondrial protein called Romo1, which then resulted in a reduction of the mitochondrial membrane potential (Kim et al., 2010). Other studies claimed that MOMP itself triggered ROS production (Kirkland et al., 2002; Kirkland and Franklin, 2003; Kirkland and Franklin, 2007; Jiang et al., 2008). Finally, complex III of the respiratory chain was found to be involved in mitochondrial ROS formation induced by Bak in vascular endothelial cells (Childs et al., 2008). Thus, it is still unclear at which step in the apoptotic pathway ROS are produced and how they impact on the cell death machinery.

One widely discussed mode of action of ROS in the cell death pathway is their destabilizing effect on the lysosomal membrane leading to lysosomal membrane permeability (LMP) of various extent. Certain ROS species such as hydrogen peroxides can easily penetrate the organelle and be converted into highly reactive hydroxyl radicals due to the iron content of lysosomes (Fenton reaction). These radicals are known to effectively damage membranes by lipid peroxidation. While a full lysis of lysosomes has been associated with necrosis, a partial, more selective lysosomal membrane perforation is supposed to assist apoptosis by releasing cathepsins (Boya and Kroemer, 2008). An oxidative burst elicited by interferon-γ was also reported to induce LMP and cathepsin release (Kowanko and Ferrante, 1987). Cathepsins are cysteine, serine or aspartyl proteases, which, once in the cytosol can impinge on the apoptotic machinery by either directly processing and activating caspases or cleaving Bid to tBid (Boya and Kroemer, 2008; Blomgran et al., 2007; Guicciardi et al., 2000; Guicciardi et al., 2005; Droga-Mazovec et al., 2008). The problem with this pathway is that it was previously claimed to be an early event of apoptosis occurring before MOMP and effector caspase-3/7 activation and prior to ROS production (Werneburg et al., 2002; Werneburg et al., 2004; Gyrd-Hansen et al., 2006). We, however, recently published that for a variety of apoptotic stimuli, LMP is not an early requiring, but a late amplifying event of apoptosis induction because it strictly depends on Bax/Bak (i.e. MOMP), apoptosome formation and caspase-3/7 activation (Oberle et al., 2010). The link from caspase-3/7 to LMP has however not been sorted out yet. Moreover, our study did not include TNFα, which has so far been the golden standard stimulus for implying LMP as an early apoptotic trigger (Guicciardi et al., 2000; Guicciardi et al., 2005; Droga-Mazovec et al., 2008; Werneburg et al., 2002; Werneburg et al., 2004).

Here we show that also for TNFα, LMP is a late amplifying process downstream of Bax/Bak, the apoptosome and caspase-3/7. In addition, we uncovered the connection between caspase-3/7 activation and LMP. As previously shown by Ricci et al., 2004, caspase-3 cleaves the p75 subunit of complex I and thereby produces ROS, in particular O₂⁻·, which then damage the lysosomal membrane leading to LMP.

TNFα is only able to trigger apoptosis of mammalian cells when the expression of NF-κB induced survival genes is blocked by transcriptional or translational inhibition (Walczak, 2011; Micheau and Tschopp, 2003; Andrew, 2004; Wang et al., 2008). We therefore used in our study a combination of TNFα and CHX (designated TNFα/CHX, T/C in figures) to treat mouse embryo fibroblasts (MEFs). We first monitored the origin and amount of ROS generated by this treatment. For that purpose, we used dihydroethidium (DHE) and 2′,7′-dichlorodihydrofluorescein (DCDHF) to estimate O₂⁻· and H₂O₂/nitric oxide (NO) by FACS analysis and fluorescence microscopy detection, respectively. After exposure to 10 ng/ml TNFα and 1 µg/ml CHX for 4 h MEFs were strongly stained with DHE (Fig. 1A,B), but not with DCDHF (data not shown) indicating that this treatment majorly induced the production of O₂⁻·. To detect the source of O₂⁻· production, we incubated the cells with the mitochondrial red fluorescent ROS sensor MitoSOX™. As shown in Fig. 1A,B, TNFα/CHX-induced O₂⁻· were effectively produced in mitochondria, and this effect was strongly inhibited by a pretreatment with the mitochondria-specific antioxidant MitoQ. This compound mimics the endogenous mitochondrial antioxidant coenzyme Q10 (James et al., 2005; Smith et al., 2011) thereby markedly scavenging the ROS produced at the ubisemiquinone site. As a positive control for ROS production at the same site, we used rotenone (Fig. 1A,C). Interestingly, the amount of ROS/O₂⁻· produced by the treatment of MEFs with 0.5 µM rotenone was approximately the same (46% by ScanR cytometry, Fig. 1C) as that generated by TNFα/CHX for the same time period (50%, FACS analysis, Fig. 1B). Again the rotenone-induced ROS production was completely ablated by MitoQ (Fig. 1A) confirming that the ROS induced by TNFα/CHX is of mitochondrial origin. It should be noted that DHE can be intracellularly oxidized to ethidium (ETH) by ROS and then stains DNA. Indeed, both DHE and MitoSOX™ (which is a DHE analog) show a clear DNA and other intracellular (mitochondrial) staining after cellular treatment with TNFα/CHX or rotenone (Fig. 1A).

To elucidate if ROS production induced by TNFα/CHX is an early event occurring before or simultaneously with MOMP or rather a late event downstream of MOMP, we determined the amount of O₂⁻· generation by DHE FACS analysis in WT, Bax/Bak^−/− [double knockout (KO) in both Bax and Bak], Apaf-1^−/−, caspase-9^−/− and caspase-3/7^−/− (double KO in both caspase-3 and -7) MEFs. As shown in Fig. 2A, the ROS production observed after 4 h of TNFα/CHX treatment (see Fig. 1B) was completely ablated in all the KO cell lines, although Apaf-1^−/− MEFs retained a slightly higher ROS level above background (which was however not significant). These data clearly show that ROS is generated downstream of MOMP and caspase-3/7 in response to TNFα/CHX. In order to exclude possible respiratory chain deficiencies in Bax/Bak^−/− MEFs, we measured oxygen consumption of intact cells. As shown in supplementary material Fig. S1A,C, Bax/Bak^−/− cells did not show any respiration deficit as they even consumed 55±17% more oxygen than WT cells. After treating with TNFα/CHX, WT cells expectedly respired less but the Bax/Bak^−/− counterparts still consumed oxygen at a higher rate (supplementary material Fig. S1A,C). This is consistent with the finding that the Bax/Bak-deficient cells were protected from TNFα/CHX-induced apoptosis (Fig. 6A) and cytochrome c release (MOMP) (Kroemer et al., 2007) and did not produce any ROS (Fig. 2A). In the presence of the complex I inhibitor rotenone, MEFs did hardly respire indicating that we indeed measured the respiration of mitochondria (supplementary material Fig. S1A).

After knowing that ROS production was downstream of MOMP we wanted to know how it was caused. The most likely mediator was caspase-3 because it was previously shown to feed back on mitochondria by cleaving the p75 NDUFS1 subunit of complex I, thereby triggering a fall in the mitochondrial membrane potential and ROS production (Ricci et al., 2004). We first treated WT MEFs with the general caspase inhibitor Z-VAD-fmk and found that TNFα/CHX-induced mitochondrial O₂⁻· production measured by DHE and MitoSOX™ fluorescence was entirely inhibited (Fig. 1A). We then used HeLa cells expressing either WT p75 or a mutant version that cannot be cleaved by caspase-3 any more (p75^D255A). As shown in Fig. 2B, while p75^WT HeLa cells produced significant amounts of O₂⁻· in response to TNFα/CHX, this was not the case with the caspase non-cleavable p75 mutant cells. The p75^WT, but not the mutant protein, was indeed cleaved concomitantly with the release of cytochrome c from mitochondria into the cytosol whereas another subunit of complex I, p17 NDUFB6 remained intact (Fig. 2C, left panel). Importantly, the kinetics of cytochrome c release was only slightly delayed in p75^D255A as compared to p75^WT cells indicating that caspase-3-mediated cleavage of p75 did not majorly affect MOMP, but a process downstream of it (ROS production) (Fig. 2C, left panel). To determine whether this caspase-3-mediated feedback effect for ROS production was specific for TNFα/CHX, we performed the same DHE FACS analysis with p75^WT and p75^D255A HeLa cells treated with the genotoxic agent etoposide or the ER stress stimulus tunicamycin, which are also known to induce MOMP and ROS production (Kroemer et al., 2007). Indeed, both stimuli triggered ROS production but no difference was detected between p75^WT and p75^D255A mutant cells (Fig. 2B). Moreover, the p75 protein was not proteolyzed despite effective cytochrome c release (Fig. 2C, right panel). These data indicate that TNFα/CHX specifically provokes ROS generation via the caspase-3-mediated cleavage of p75 of complex I which supports the finding above that ROS production originates from a mitochondrial source.

Like other apoptotic stimuli TNFα is able to induce LMP. However based on previous studies, it has been postulated that LMP is an early event occurring before MOMP, i.e. right after TNF-R1 stimulation (Boya and Kroemer, 2008; Blomgran et al., 2007; Guicciardi et al., 2000; Guicciardi et al., 2005; Droga-Mazovec et al., 2008; Werneburg et al., 2002; Werneburg et al., 2004; Gyrd-Hansen et al., 2006). Surprisingly, the acid test, namely the role of Bax/Bak and the apoptosome in TNFα-induced LMP has not been performed yet. We therefore looked at TNFα/CHX-triggered LMP in WT, Bax/Bak^−/−, Apaf-1^−/−, caspase-9^−/− and caspase-3/7^−/− MEFs by two means, the decrease of lysosomal red fluorescent acridine orange (AO) staining and the release of cathepsin B and -L (cathepsin B/L, CTB/L) activity and/or protein into the cytosol. As shown in Fig. 3A,B, the TNFα/CHX-induced diminishment of AO staining and the concomitant appearance of cathepsin B/L activity (Z-FRase) in the cytosol were both entirely prevented in all KO MEFs indicating that LMP triggered by TNFα/CHX clearly occurred downstream of MOMP. To identify if mitochondrial ROS production via caspase-3-mediated p75 cleavage was the mediator of this LMP, we quantified diminished AO staining and cathepsin B protein release in WT MEFs co-treated with TNFα/CHX and either MitoQ or the general antioxidant N-acetylcysteine (NAC) and in TNFα/CHX-treated p75^WT and p75^D255A HeLa cells. For comparison, we again exposed MEFs to etoposide and tunicamycin. While LMP was markedly (although not completely) inhibited by both MitoQ and NAC when MEFs were treated with TNFα/CHX, much less LMP inhibition was detected by MitoQ in the case of etoposide and tunicamycin (Fig. 3C,D). As discussed below, the partial dependence of LMP on ROS in the case of these two drugs could be due to ROS production independent of p75 cleavage. Moreover, the release of cathepsin B induced by TNFα/CHX was significantly diminished by MitoQ (Fig. 3E). In addition, while only 49% of p75^WT HeLa cells retained lysosomal AO staining (Fig. 3D), and cathepsin B largely appeared in the cytosol (Fig. 3F) after 4 h of TNFα/CHX treatment, 82.5% of the p75^D225A mutant cells exhibited AO staining (Fig. 3D) and no cytosolic cathepsin B release was observed in these cells (Fig. 3F). Consistent with these data, a cellular treatment with rotenone, which provokes mitochondrial ROS production (see Fig. 1C), also effectively triggered LMP (Fig. 5A). Thus, our results show for the first time that TNFα/CHX-induced LMP is mediated via caspase-3-triggered p75 cleavage and mitochondrial ROS production.

To validate this mechanism, we needed to show that ROS production was prior to LMP. For that purpose, we performed a kinetic FACS analysis of TNFα/CHX-treated HeLa cells, using AO (Fig. 4A) and DHE (Fig. 4B) stainings for LMP and ROS measurements, respectively. Since AO exhibits both green and red fluorescence and DHE stains red, the data could not be depicted on a single FACS dot-plot. But by comparing the extent of LMP and ROS at each hour post-treatment we could show that ROS production peaked at 2 h (Fig. 4B, black bars) when only 20% of the cells had undergone LMP (Fig. 4A). This confirms that mitochondrial ROS production in response to TNFα/CHX occurs before the induction of LMP.

To get an idea if it is principally possible that ROS can provoke LMP in vitro, we incubated a post-mitochondrial fraction containing lysosomes with a buffer containing no antioxidants (oxidizing) or reducing agents for up to 60 min and then measured the amount of cathepsin B/L activity (Z-FRase) remaining in the lysosomes. As shown in Fig. 5B, left panel, the lysosomal Z-FRase activity gradually decreased with time of incubation, and this was markedly inhibited by the addition of NAC or glutathione (GSH) to the buffer system (Fig. 5B, right panel). Simultaneously, the AO staining, which was diminished in these lysosomes after 60 min buffer incubation also recovered by the presence of the two antioxidants (Fig. 5B, right panel). We could therefore demonstrate that ROS is able to effectively mediate LMP in vitro.

We finally needed to show that the caspase-3-mediated p75 cleavage and mitochondrial O₂⁻· production had any physiological significance for TNFα/CHX-induced apoptosis, at least as an amplification mechanism. As reported before, this apoptosis was entirely prevented in Bax/Bak^−/−, Apaf-1^−/−, caspase-9^−/− and caspase-3/7^−/− MEFs (Fig. 6A) confirming that in these cells TNFα/CHX used the intrinsic mitochondrial pathway for apoptosis induction. The mitochondria-specific antioxidant MitoQ (Fig. 6B) or the presence of the caspase-3 non-cleavable p75^D225A mutant (Fig. 6C) also partially protected the cells from TNFα/CHX-induced apoptosis. By contrast NAC was unable to interfere with this type of apoptosis (Fig. 6B) despite the fact that it blocked LMP (Fig. 3D), most likely because this antioxidant was slightly cytotoxic on its own (data not shown). For unknown reasons untreated p75^D225A cells respired less effectively than their WT counterparts (supplementary material Fig. S1B,C). Like WT MEFs, p75^WT cells consumed less oxygen upon TNFα/CHX treatment. p75^D225A cells showed a slightly improved respiration under these conditions confirming that their respiratory chain was less impaired and they produced less ROS than p75^WT cells (Fig. 2B).

However, the p75^D225A cells were not entirely protected against TNFα/CHX-induced apoptosis since the p75/ROS pathway is not expected to be an initiating, requiring event leading to MOMP but a downstream amplifying process to further enhance apoptosis through LMP, cathepsin release and additional caspase activation. Indeed, the released cathespins seemed to contribute to this amplification loop as both cathepsin B/L^−/− MEFs as well as MEFs pre-treated with the cysteine protease inhibitor E64d showed a similar diminishment of TNFα/CHX-induced apoptosis as p75^D225A HeLa cells (Fig. 6C).

TNFα, the founding member of the TNF-superfamily, mediates both necroptotic and apoptotic cell death (Walczak, 2011; Micheau and Tschopp, 2003; Andrew, 2004; Wang et al., 2008; Kroemer et al., 2007; Degterev and Yuan, 2008, He et al., 2009; Zhang et al., 2009). Although many investigations have been conducted over the last few years the detailed mechanisms of these death signaling pathways are still not completely known. This particularly concerns our understanding about where ROS are produced intracellularly, how ROS and LMP are mechanistically linked and how both processes contribute to TNFα-induced apoptosis (Boya and Kroemer, 2008; Shen and Pervaiz, 2006; Zdolsek et al., 1993; Brunk et al., 1995). Our study here bridges these missing gaps. We identified that TNFα/CHX-induced LMP is not an apoptosis initiating, but amplifying process downstream of MOMP, and it is triggered by mitochondrial ROS generated as the result of a caspase-3-mediated cleavage of the p75 NDUFS1 subunit of complex I. A similar finding was obtained in two different cellular systems, MEFs and HeLa cells, although molecular tools for manipulating p75 levels were not available for MEFs, and HeLa cells were less sensitive to TNFα/CHX-induced apoptosis due to the expression of HPV E6 and E7 genes (Garnett et al., 2006; Filippova et al., 2002).

The mitochondrial production of ROS by TNFα has been shown in previous reports, and it was also suggested that these ROS somehow contribute to the apoptotic or necrotic process (Kim et al., 2010; Lin et al., 2004; Shoji et la., 1995; Schulze et la., 1993; Shen and Pervaiz, 2006). However, since other intracellular sources of ROS exist, such as the plasma membrane residing NADPH oxidase complex (Yazdanpanah et al., 2009; Woo et al., 2000; Lambeth, 2004), we had to make sure that the majority of ROS implicated in TNFα-induced death signaling is of mitochondrial origin. Indeed, the mitochondria specific antioxidant MitoQ, but not the general antioxidant NAC, partially protected fibroblasts from TNFα/CHX-induced apoptosis. Moreover, DHE and MitoSOX™ stainings were completely abrogated by MitoQ and in Bax/Bak^−/− MEFs indicating that the major source of ROS production induced by TNFα/CHX is indeed the mitochondria. We further showed that rotenone which interferes with complex I, triggered similar ROS production as TNFα/CHX, and this was also blocked by MitoQ. Finally, the fact that HeLa cells expressing the caspase non-cleavable p75^D255A mutant did not produce any ROS in response to TNFα/CHX confirms that complex I is one major source of this ROS generation. It has been reported that ROS can be a direct consequence of Bax/Bak-mediated MOMP (Kirkland et al., 2002; Kirkland and Franklin, 2003; Kirkland and Franklin, 2007; Jiang et al., 2008, Childs et al., 2008), because of the loss of cytochrome c from the respiratory chain complex III/IV and the fall in the mitochondrial membrane potential. This might be the case for apoptotic stimuli such as etoposide or tunicamycin, which did not provoke p75 cleavage despite effective cytochrome c release, and did not depend on p75 for ROS production and apoptosis. As with TNFα/CHX, the kinetics of cytochrome c release induced by these two drugs were also similar between p75^WT and p75^D255A cells. However, the situation is clearly different for TNFα/CHX. Although it similarly triggers cytochrome c release and ROS production as other apoptotic stimuli, the latter event is entirely blocked in Apaf-1^−/−, caspase-9^−/− or caspase-3/7^−/− MEFs. This reinforces the notion that in response to TNFα/CHX ROS is generated downstream of MOMP, e.g. by a caspase-3-mediated cleavage of p75, as has previously been shown by Ricci et al., 2004. In their report they also clearly revealed that the fall in the mitochondrial membrane potential occurs after cytochrome c release in a caspase-3/p75 cleavage dependent manner (Ricci et al., 2004). More studies are required to determine if other apoptotic stimuli (for example other members of the TNF-superfamily such as FasL or TRAIL) use the same mechanism to amplify apoptosis via LMP.

Although the major ROS species, which we measured in our system were O₂⁻· (as evidenced by DHE or MitoSOX staining), we cannot exclude that they were further converted into hydrogen peroxide (H₂O₂) by mitochondrial superoxide dismutases. In fact, in contrast to O₂⁻·, H₂O₂ is more cell permeable and may be the ROS species that leave mitochondria and enter lysosomes where they react with iron to produce the highly reactive hydroxyl radicals (^•OH) (Fenton reaction) which then induce LMP via lipid peroxidation. Indeed the role of intralysosomal iron in oxidant-induced cell death has been previously reported (Yu et al., 2003). Moreover, we could show here that incubating a post-mitochondrial cellular fraction containing lysosomes with an oxidizing buffer triggered LMP in vitro over time. The reason why we did not detect major H₂O₂ production by DCDHF staining might have been due to the use of CHX, which was required for TNFα to effectively induce apoptosis (Walczak, 2011; Micheau and Tschopp, 2003; Andrew, 2004; Wang et al., 2008), but which also is known to downregulate the mitochondrial superoxide dismutase and ferritin heavy chain (Sasazuki et al., 2004; Pham et al., 2004). This may explain why ROS levels reached such high levels within a short time period (2 h) of the TNFα/CHX treatment (Fig. 4B). These levels were already at the maximum before a significant LMP was observed (Fig. 4A) reinforcing that notion that LMP is caused by ROS.

The surprising finding of our study was that TNFα-induced LMP turned out to be a process downstream of MOMP. Many previous studies proposed an early induction of LMP by this cytokine, for example via the activation of acid or neutral (FAN) sphingomyelinases, cathepsin B or caspase-8 (Boya and Kroemer, 2008; Blomgran et al., 2007; Guicciardi et al., 2000; Guicciardi et al., 2005; Droga-Mazovec et al., 2008; Werneburg et al., 2002; Werneburg et al., 2004; Gyrd-Hansen et al., 2006; Ullio et al., 2012). How the lysosomal membrane is damaged by these molecules has however not been unveiled so far, although a direct perforation by Bax or Bak (Werneburg et al., 2007, Castino et al., 2009; Kagedal et al., 2005; Feldstein et al., 2006) or the implication of ROS (without showing from where they in fact originated) were proposed (Zdolsek et al., 1993; Brunk et al., 1995; Roberg and Öllinger, 1998; Brunk and Svensson, 1999; Terman et al., 2006). LMP results in the release of cathepsins, which then either cleave Bid to tBid, process caspases or degrade Bcl-2-like survival factors to induce MOMP (Boya and Kroemer, 2008; Blomgran et al., 2007; Guicciardi et al., 2000; Guicciardi et al., 2005; Droga-Mazovec et al., 2008). These signaling pathways, and therefore LMP, were supposed to be initiating, requiring events for TNFα-induced apoptosis. Astonishingly, none of the previous studies verified if LMP was dependent on MOMP, i.e. if it was blocked in Bax/Bak^−/− cells. Our study now reveals for the first time that TNFα/CHX-induced LMP, as determined by a diminished red fluorescent staining of the lysosomotropic dye AO and the release of cathepsin B/L activity and cathepsin B protein into the cytosol, is entirely dependent on Bax/Bak, the apoptosome and caspase-3/7. Since ROS production is also dependent on this pathway, and that LMP is inhibited by MitoQ treatment or in p75^D225A HeLa cells, the mechanistic link between TNFα and LMP is mitochondrially generated ROS. Thus, rather than being an initiating process, LMP occurs as an amplification loop downstream of MOMP. Such a mechanism explains why inhibiting ROS generation or performing the experiments in p75^D225A HeLa cells did not entirely eliminate TNFα/CHX-induced apoptosis, because cell death could still proceed at a low level without the amplification loop. This adds TNFα to the growing list of apoptotic stimuli, which we have previously shown to trigger LMP downstream of MOMP (Oberle et al., 2010). We therefore propose that in most, if not all cases, LMP and the release of cathepsins are not required for apoptosis induction but rather amplify the process to make sure that cell death occurs effectively, rapidly and irreversibly.

3T9 MEFs from wild type (WT) C57BL/6 mice or mice deficient of Apaf-1, caspase-9 or both Bax and Bak (Bax/Bak^−/−) were obtained from Andreas Strasser, WEHI, Melbourne. SV40 T antigen transformed MEFs deficient of both caspase-3 and -7 (caspase-3/7^−/−) were kindly provided by Richard Flavell, Yale, USA. 3T9 transformed MEFs deficient for both cathepsin B and L (CTB/L^−/−) are described in Oberle et al., 2010. HeLa cells stably expressing the wild-type (WT) p75 NDUFS1 subunit of respiratory chain complex I or its caspase-3 non-cleavable mutant p75^D255A are described in Ricci et al., 2004. All cells were cultured in DMEM supplemented with 10% FCS, 100 units/ml penicillin and 100 µg/ml streptomycin at 37°C/5% CO₂.

For TNFα/CHX-induced cell death, fibroblasts were preincubated or not with 5 µM MitoQ or 30 µM E64d (Biomol, Germany) for 24 h and then treated with 10 ng/ml of TNFα in combination with 1 µg/ml CHX for 4 h. For etoposide or tunicamycin-induced cell death, fibroblasts were treated with 100 µM etoposide or 1 µg/ml tunicamycin with or without 5 µM MitoQ for 24 h. Apoptotic cells were determined by flow cytometry after staining with His–GFP–Annexin-V/propidium iodide (PI) as described in Egger et al., 2003.

Cellular fractionation was carried out using the Cell Fractionation Kit Standard from Abcam which retains mitochondria intact. Briefly, cells were collected, washed once with buffer A and centrifuged. The pellet was resuspended in buffer A to 6.6×10⁶ cells/ml, an equal volume of buffer B was added and then the cells were incubated for 7 min on a rotator at room temperature and centrifuged sequentially at 5000 g and 10,000 g at 4°C for 1 min each. The supernatant was collected as cytosolic fraction. The cytosol-depleted pellet was resuspended in buffer A and mixed with buffer C containing detergent and then incubated for 10 min on a rotator at room temperature. After two sequential centrifugations as above the supernatant was collected as mitochondrial fraction. To get the lysosomes the cytosolic fraction was further centrifuged at 100,000 g, 1 h. The pellet represented the lysosomal fraction.

Cathepsin B (CTB) and cathepsin L (CTL) specific activities of cytosolic and heavy membrane fractions were measured in a buffer containing 200 mM NaOAc, 1 mM EDTA, 0.1% Brij 35, 0.1% Triton X-100 at pH 6.7. Briefly, 5 µl of cytosolic or heavy membrane fraction (25 µg protein) was mixed with 90 µl assay buffer and 5 µl of the cathepsin B substrate Z-FR-AMC (25 µM) in each well of a black Nunc 96 MicroWell™ plate (Nunc, Denmark) and incubated at 37°C for 1 h before measuring fluorescence at excitation 355 nm and emission 460 nm in a Fluoroskan Ascent (Thermo Labsystem).

Cells were exposed or not to apoptotic stimuli and LMP was detected by flow cytometry (FL1:FL3) after staining with 5 µM AO in 1×Hank's balanced salt solution (HBSS) with Ca²⁺/Mg²⁺ (0.137 M NaCl, 5.4 mM KCl, 0.25 mM Na₂HPO₄, 0.44 mM KH₂PO₄, 1.3 mM CaCl₂, 1.0 mM MgSO₄, 4.2 mM NaHCO₃) for 10 min at 37°C, or alternatively by ScanR cytometry (Olympus) after in situ staining and mounting. To measure LMP induced by ROS in vitro, a post-mitochondrial fraction was extracted from MEFs loaded with 5 µM AO, washed twice and then incubated with or without 1 mM N-acetylcysteine (NAC) or 1 mM glutathione (GSH) at 37°C for up to 1 h. Afterwards the post-mitochondrial fraction was centrifuged as described above to separate cytosol from membranes, and AO fluorescence was measured at excitation 485 nm and emission 538 nm. In addition, the distribution of cathepsin B/L activities between the membrane and cytosolic fraction was determined by the Z-FR-AMC assay as described above.

After treatment with various pro-apoptotic stimuli in the presence or absence of 5 µM MitoQ or 100 µM of the general caspase inhibitor Z-VAD.fmk (Promega) the MEFs were incubated for 10 min in 1×HBSS with Ca²⁺/Mg²⁺ buffer containing 2 µM dihydroethidium (DHE) or 1 µM 2′,7′-dichlorofluorescein (CM-H₂DCFDA) or 1 µM MitoSOX™ red mitochondrial superoxide indicator (Invitrogen). After two washes the cells were either subjected to FACS analysis (for DHE and CM-H₂DCFDA) or fluorescence microscopy (Zeiss) or ScanR fluorescence cytometry (Olympus) (for MitoSOX™). As positive control for ROS production, cells were incubated with 0.5 µM rotenone for 2 or 4 h.

The experiment was carried out using high-resolution respirometry (Oxygraph 2k Oroboros Instruments) according to the manufacturer's instruction. Oxygen consumption was measured in intact cells pretreated or not with TNFα/CHX for 2 h. Chamber A and B were filled with 2 ml of complete DMEM and compensated for at least 5 min to stabilize temperature and oxygen. 2×10⁶ in 50 µl DMEM were injected into each chamber and oxygen consumption was measured as described in Kuznetsov et al., 2008. As a control for mitochondria-specific respiration the cells were treated with 0.5 µM rotenone for 2 h prior to harvesting.

Cytosolic or heavy membrane fractions (containing all other organelles except the nuclei) were collected as described above and the membrane fraction was lysed in RIPA buffer (50 mM Tris-HCl, pH 7.4, 1% Triton X100, 1% n-octylglucoside, 0.1% SDS, 1 mM EDTA, 150 mM NaCl) containing proteinase inhibitors (1×protease inhibitors cocktail, Roche) for 30 min on ice. 20 µg of cytosolic or membrane proteins were subjected to SDS-PAGE followed by western blotting to nitrocellulose (NC) membranes according to standard protocol. Immunodetection was performed by incubating the NC membranes with the first antibody in PBS containing 0.05% Tween-20 for 4 h followed by a horse radish peroxidase (HRP)-conjugated secondary antibody for 1 h. The following first and secondary antibodies were used: goat polyclonal anti-cathepsin B (R&D systems), rabbit polyclonal anti-cytochrome c and anti-COX IV (3E11) (Cell Signalling), mouse polyclonal anti-NDUFS1 (ab52690), rabbit polyclonal anti-NDUFB6 (ab103531), rat monoclonal anti-Lamp-1 (1D4B) (Abcam), mouse monoclonal anti-β-actin (Pierce), and HRP-conjugated donkey anti-goat, donkey anti-rabbit and goat anti-mouse antibodies (Dianova). Signals were visualized by enhanced chemiluminescence (ECL) (Pierce).

All quantitative data are shown as means±standard deviation (s.d.) and the statistical significance was determined by two-tailed Student's t-test.

We thank Michael P. Murphy, MRC, Cambridge, United Kingdom for MitoQ, Andreas Strasser, WEHI, Melbourne, Australia for the Bax/Bak^−/−, Apaf-1^−/− and caspase-9^−/− 3T9 MEFs and Richard Flavell, University of Yale, CT, USA, for the SV40 transformed caspase-3/7^−/− cells. We are grateful to Robert Pick, Walter Brendel Center, Munich, Germany for the technical assistance with the ScanR and Thomas Reinheckel, Freiburg, for helping with the cathepsin assays. The authors declare that they have no conflict of interest.