Dynamics of mitochondrial structure during apoptosis and the enigma of Opa1

1. Introduction

Once cytochrome c is released from mitochondria and activates apoptosomes in the cytosol, death is the certain outcome for the host cell. What happens to the cell's mitochondria, on the other hand, is turning out to be a very complicated story. Adding to the complexity is that this organelle has three distinct compartments defined by two distinct membranes [1] (Fig. 1) that separate functionality. Co-incubation of isolated mitochondria with recombinant BH3-only proteins or peptides triggers the rapid and complete release of cytochrome c [2–4]. Other proteins such as Htra2/Omi, SMAC and Opa1 are released at about the same time. Currently, the mitochondrial community has a very limited understanding of how BH3-only proteins initiate this process.

It has been speculated that in order to release cytochrome c and other proteins, pores must be formed on the outer membrane of mitochondria [5, 6]. This pore is called the mitochondrial outer membrane pore (MOMP), its formation is triggered by tBid [7], and supposedly contains Bak or Bax proteins but VDAC channels are not involved [8]. It has also been speculated that cristae remodeling is a necessary part of cytochrome c release. There are a number of excellent reviews regarding the first speculation [9–11]. In this review, we will focus on the second speculation and address the question, “Is cristae remodeling necessary for the rapid and complete release of cytochrome c?”

2. The nature of cristae remodeling

Dramatic morphological changes of mitochondria are observed in cells undertaking apoptosis. Upon apoptotic stimulation in many situations, mitochondria change from a filamentous network into punctate fragments. Electron microscopic images revealed that the cristae, which are normally tubular or lamellar structures, become roundish and fragmented and then they disappear [12]. Blebbing of the matrix and extensive swelling of the mitochondria can occur, followed by the fragmentation of the mitochondria [2,5]. The release of cytochrome c from mitochondria activates apoptosome formation in the cytosol, and is a necessary step in mitochondrial-dependent apoptosis [13–16]. However, the question of whether large-scale morphological changes of mitochondria are necessary for the release of cytochrome c or the consequence of the released cytochrome c activating apoptosomes and altering mitochondria downstream has been a controversial issue. Proponents of the former model suggest that crista junctions, which are narrow tubular tunnels connecting cristae to the space between the outer and inner membranes, called the intermembrane space, must become wider to allow the passage of cytochrome c from the intracristal space to this intermembrane space. Once in the intermembrane space, cytochrome c can escape to the cytosol through the pores formed on the outer membrane. In this article, we present recent evidence with commonly used mechanisms to induce apoptosis favoring the latter model, being that more subtle forms of structural changes are taking place inside mitochondria. Among these changes are (1) narrowing of crista junction openings, and (2) disassembly of Opa1 protein complexes at the opening. We also discuss what this model suggests towards our understanding of mitochondria.

3. Timing of cytochrome c release

Doug Green and colleagues observed that when HeLa cells stably expressing GFP-Cyt C were stimulated with STS, Act D or UV to induce apoptosis, within each single cell the entire content of GFP-Cyt C was released from all the mitochondria within 10 min [17,18]. Furthermore, they found that transmembrane potentials of mitochondria were retained for at least 60 min from the time of cytochrome c release [19]. Several years later, Terry Frey et al. used correlated light microscopy and electron tomography to examine the 3D structures of mitochondria in cells that released GFP-Cyt C and found that in the presence of the pan-caspase inhibitor z-VAD, HeLa cells released the content of GFP-Cyt C from mitochondria with the same kinetics as in the absence of z-VAD [20]. Interestingly, little large-scale alterations of mitochondria took place in the presence of z-VAD. These recent observations suggest that large-scale morphological changes in mitochondria upon initiation of several forms of apoptosis are not necessary for the release of cytochrome c, but rather are the consequences of released cytochrome c activating cytosolic caspases. However, the warning was raised that immuno-fluorescent microscopy cannot distinguish different compartments of mitochondria. Thus, it was possible that GFP-Cyt C may be localized differently from endogenous cytochrome c, i.e., GFP-Cyt C could be residing only in locations adjacent to the outer membranes and thus remodeling of cristae was not required for release of GFP-Cyt C. The release of endogenous cytochrome c that resides in portions of cristae deep within mitochondria may still require cristae remodeling or other structural alterations. To address this question, researchers turned to in vitro experiments using isolated mitochondria incubated with the BH3-only proteins tBid, BimS and with BH3 peptides [4].

In in vitro experiments, isolated mouse liver mitochondria released the entire content of cytochrome c in 10–15 min in the presence of 10–20 nM tBid or BimS. However, the gross morphologies of treated and untreated mitochondria were virtually indistinguishable by transmission electron microscopy. Upon further analysis of mitochondria that released the entire content of cytochrome c using 3D tomography, it was found that the diameter of crista junctions was actually smaller. The release of cytochrome c was inhibited by the addition of excess anti-apoptotic proteins such as Bcl-XL and Mcl1 [4]. It has been estimated that over 85% of cytochrome c resides in cristae as opposed to the intermembrane space. The release of nearly 100% of cytochrome c must mean the release of most if not all the cytochrome c from within cristae within 10–15 min through narrowed crista junctions. Thus, essentially all cytochrome c can exit cristae and cross the outer membrane in as short as 10 min without large-scale structural alterations.

4. Molecular requirement for cytochrome c release at the mitochondrial outer membrane

How does cytochrome c traverse the mitochondrial outer membrane? The current model is that BH3-only pro-apoptotic proteins induce the formation of Bak and Bax oligomers, which then triggers pore formation through which cytochrome c and other intermembrane space proteins escape during apoptosis. Supporting this model is the observation that in the absence of Bak or Bax oligomerization, there is no cytochrome c release [21]. Examples are (1) in Bax/Bak double knockout cells tBid does not induce cytochrome c release from mitochondria [22], (2) addition of a recombinant mutant of tBid, G94E (tBid carrying a mutation in the BH3 domain) to mitochondria from wild-type mice does not induce Bak oligomerization, and does not induce apoptosis, (3) addition of excess anti-apoptotic proteins such as Bcl-XL or Mcl1 blocks tBid and BimS-induced Bak oligomerization and blocks apoptosis, and (4) chemical agents such as DCIC, MG132 and Ucf101 block Bak oligomerization and block apoptosis [4]. Experiments with mitochondrial outer membrane vesicles (OMV) loaded with GFP-dextran of various sizes up to several hundred kDa, also indicated the requirement of Bax or Bak for the tBid-induced dextran release from OMV [23].

It is not yet known what other proteins if any are involved in MOMP formation during apoptosis through which cytochrome c is released. It is understood, however, that the VDAC ion channel is not necessary for tBid-induced cytochrome c release [8]. The complication in determining the composition of the outer membrane pores rests with the fact that it is difficult to capture the images of cytochrome c being released through the pores [7].

5. Molecular events in the intermembrane space leading to the complete release of cytochrome c

The evidence we discussed so far points towards a simple model in which pro-apoptotic BH3-only proteins induce outer membrane pore formation through which the entire content of cytochrome c is released into the cytosol and large-scale alterations of mitochondrial morphologies are the consequences of cytochrome c activating apoptosomes in the cytosol. As we discuss in this section, however, subtle but significant discrepancies in this model were found. For example, it was found that the addition of MG132 blocked tBid-induced cytochrome c release by blocking Bak oligomerization [4]. When these mitochondria were analyzed by EM tomography, the average radius of crista junctions was 8.5 nm, about half the radius of crista junctions in untreated mitochondria (average 16 nm). In fact, the average radius was almost the same for crista junctions of mitochondria treated with tBid alone. Thus in the presence of MG132 and tBid, crista junctions narrowed while there were no pores on the outer membrane. When the accessibility of cytochrome c from the outer membrane was measured, it was found to be as accessible as cytochrome c in tBid-treated mitochondria. A search for a molecular marker for these subtly remodeled cristae found Opa1.

In isolated mitochondria, about 50% of Opa1 proteins are found in an oligomeric complex [4,24]. During tBid-treatment, oligomerized Opa1 complexes disappear rapidly, leaving only the monomeric form of Opa1 [4,24,25]. When isolated mitochondria were treated with MG132 and tBid, cytochrome c and Htra2/Omi as well as Opa1 were all retained, but there was no Opa1 complex. This led us to speculate that Opa1 complex assembly and disassembly regulate cytochrome c release from inside the cristae. One could imagine the Opa1 complex acting as a gate at crista junctions, retaining cytochrome c in the cristae under normal circumstances. During apoptosis, the Opa1 complex is disassembled by processes independent of caspases. When the gates collapse, they leave holes large enough to let through cytochrome c and many other proteins. But when there is no pore at the outer membrane, cytochrome c is still trapped inside the mitochondria. This model was tested by engineering a situation in which the outer membrane pores were formed but the Opa1 gates remained intact. To accomplish this, a point mutation in Opa1 was found which made the Opa1 protein resistant to disassembly. When this mutant protein was over-expressed in tissue culture cells and treated with apoptosis-inducing drugs, mitochondria released very little cytochrome c and became resistant to apoptosis. It was concluded that Opa1 controls passage of cytochrome c through crista junctions.

Thus how Opa1 oligomers are assembled and disassembled is a question central to understanding the molecular mechanism of tBid-induced cytochrome c release, and will be discussed later. But before we go on, we will address the differences between conflicting findings.

6. Two versions of the same incident — the Rashomon effect

Controversy is nothing new to Science. It often happens that competing camps with different models fight for a while only to find years later that no model was completely right. Further experimentation and/or analysis seems to always resolve these controversies. To resolve the different conclusions between Scorrano's group and our group, additional analysis is worthwhile because both groups used almost the same reagents [2,4,24]. Thus, we shall look more closely to the published articles and compared methods used and data derived from the experiments by Scorrano's group and ours.

Scorrano et al. used isolated mouse liver mitochondria as did we. However, they incubated the mitochondria with recombinant tBid protein in a buffer containing very little metal chelator (10 µM EGTA) compared to our buffer (1 mM EGTA and 1 mM EDTA) [2,4].¹ We both saw a rapid release of cytochrome c. Using a different buffer, however, they saw changes in mitochondrial morphology, such as extensive vesiculation of cristae followed by swelling of mitochondria, whereas we saw only a subtle structural change — the narrowing of crista junction diameter from 16 nm to 8.7 nm, which was only observable with EM tomography. Scorrano et al. classified the features of altered mitochondria into 4 classes [2]; Class I being untreated mitochondria, Class II showing extensively vesiculated crista structures, Class III mitochondria having asymmetric blebbing of the matrix, often with partially ruptured outer membranes. Swollen mitochondria with ruptured outer membranes, completely devoid of cristae structures are classified as Class IV. Within 2–5 min of incubation with tBid, Classes II–IV become more predominant. Using EM tomography, they too measured the diameter of crista junctions. In untreated mitochondria, the diameter of crista junctions was approximately 18.6 nm, similar to ours. However, in later phases, the average diameter increased to as much as 56.6 nm.

The morphologies of isolated mitochondria treated with tBid in low EGTA buffer are significantly different from the mitochondrial remodeling in HeLa cells undergoing etoposide-induced apoptosis described by Frey et al. [20], who categorized the progression of mitochondria during apoptosis in three stages. In the correlated light and EM studies by Sun et al. [20], mitochondria in stages 2 and 3 had already released the entire content of cytochrome c. Furthermore, in the presence of the pan-caspase inhibitor z-VAD, mitochondria in HeLa cells released the entire content of cytochrome c without ever reaching stages 2 and 3. Indeed, they found that the average diameter of crista junctions in etoposide+z-VAD treated HeLa cells was not different than before the treatment. Thus, it appears that extensive remodeling of cristae structures including the widening of the crista junctions is unnecessary for the complete release of cytochrome c.

In a more recent article, Scorrano et al. found that over-expression of Opa1 delays the onset of cytochrome c release triggered by various apoptosis-inducers in tissue culture cells [24]. Indeed, when mitochondria isolated from Opa1 over-expressing cells were treated with tBid, there were considerably fewer mitochondria in Classes II–IV, and the average diameter of crista junctions remained almost unchanged (16.1±2.1 nm for untreated cells, 15.2±2.3 nm for Opa1 over-expressing cells and 17.1±2.1 nm for cells over-expressing K301A mutant). When isolated mitochondria from wild-type cells and Opa1 over-expressing cells were treated with tBid, the crista junction diameter widened only in mitochondria from wild-type cells (45.4 ± 3.2 nm for wt mouse embryonic fibroblasts (MEFs) and 20.2 ± 3.1 nm for Opa1 over-expressing MEFs). They interpreted this result as evidence that Opa1 over-expression blocked the widening of the crista junctions, preventing cytochrome c release. Another possible interpretation would be that in the presence of an increased amount of Opa1 protein, fewer mitochondria had released cytochrome c and transitioned to Classes II–IV mitochondria. Thus, the majority of mitochondria might have remained in Class I with relatively narrow crista junctions. It is worth mentioning that over-expression of Opa1 by itself does not change the average diameter of crista junctions, suggesting that cytochrome c is not sequestered in the cristae by making crista junctions narrower. The conclusion is that widening of crista junctions is not necessary for the release of cytochrome c.

Even if mitochondrial cristae remodeling takes place only after cytochrome c release, it would be of some interest to know what causes these changes. We can only speculate at this point. The obvious candidates to bring about morphological changes would be caspases because in HeLa cells, the presence of the pan-caspase inhibitor z-VAD blocked these changes. It would be interesting to test what effect z-VAD would have on isolated mitochondria in low EGTA buffer when cytochrome c is released by adding recombinant tBid. Would remodeling of mitochondria take place? If z-VAD were to block the remodeling then maybe a mitochondrial z-VAD binding protein could be identified. There are actually candidate binding proteins — two caspases. Caspase 9 has been found inside mitochondria [26] and caspase 3 has been shown to act on complex I on the mitochondrial inner membrane [26a]. Since the presence of 1 mM EGTA and 1 mM EDTA block mitochondria crista remodeling, caspases could be the metalloproteases present in mitochondria that influence the remodeling. One other possibility is the release of calcium from the endoplasmic reticulum (ER) to trigger these changes. Since the ER is tethered to mitochondria by Mitofusion 2 [27], it is possible and also likely that the ER is present in many isolated mitochondria preparations. Perhaps tBid could induce calcium release from the ER causing mitochondria to swell, although a mechanism for this is unknown. However, if this were the mechanism, then cyclosporine A (CsA) would block the swelling, and possibly the release of cytochrome c. Interestingly, Scorrano et al. reported that CsA attenuated the release of cytochrome c from isolated mouse liver mitochondria, and also blocked the remodeling of cristae structures in mitochondria [2]. Furthermore, the remodeling did not require the BH3 domain of tBid, and was independent of the Bak protein [2]. On the other hand, we could not block tBid-induced cytochrome c by CsA, and the same G94E mutant of tBid used by Scorrano et al. could not induce the release of cytochrome c, whereas other BH3 only proteins such as BimS and BH3 peptides could [4]. It is perhaps worth noting that Frey et al. also found that pre-treatment of HeLa cells with CsA could not block the swelling of mitochondria induced by etoposide [20].

Disagreement about the effectiveness of CsA as an inhibitor of apoptosis is not just between three groups but is more widespread among the community. For example, using mouse liver mitochondria, it was suggested that CsA did not inhibit tBid-induced cytochrome c release [28], whereas another concluded the opposite [29]. These articles reported the use of isolated mouse liver mitochondria² with recombinant tBid and 1–3 µM of CsA. However, the buffers used for mitochondrial suspension were slightly different. Why such discrepancies among the scientific community is open for debate.

The anti-apoptotic effect of CsA on H₂O₂-treated cells is even more controversial. Some have claimed it to be inhibitory, whereas others have shown it to be ineffective [31,32]. The reason for the discrepancy may lie in the fact that H₂O₂ can initiate at least two different cell death pathways and which pathway is activated may depend on the cell type. The subject will be discussed later in this review. On the effectiveness of CsA related to cell death, however, there is at least one consensus and that is that CsA is effective in blocking nitrogen oxide-induced neural cells death [30].

7. The last step in the release of cytochrome c: how to break the cardiolipin–cytochrome c association?

BH3-only proteins induce outer membrane pore formation and open crista junctions by disassembling Opa1 oligomers, but many researchers now think that these events are not enough for the rapid and complete release of cytochrome c. In low salt buffer, they are enough to release intermembrane space proteins, such as short membrane-free Opa1 and Htar2/Omi [4], but not cytochrome c. Cardiolipin is a charged lipid almost exclusively found in the mitochondrial inner membrane where it constitutes about 20% of the total lipids [31]. There is an electrostatic force between cytochrome c and cardiolipin and unless this weak interaction is broken, it keeps cytochrome c inside mitochondria [32–34]. Indeed, when isolated mitochondria are incubated with recombinant tBid in buffer containing 10 mM KCl along with 300 M trehalose with or without metal chelators such as EGTA or EDTA, Opa1 oligomers are disassembled, opening crista junctions and releasing short-form Opa1, Htra2/Omi, and AK very rapidly [4,35]. However almost all cytochrome c remains within mitochondria. When the KCl concentration is raised to 60–80 mM, the complete release of cytochrome c is seen [4,36]. Intracellular concentration of K⁺ is estimated to be around 150 mM, more than enough to release cytochrome c from tBid-treated mitochondria. However, this concentration may include stored K⁺ in places such as mitochondria.

Supposing that we need to disrupt cardiolipin–cytochrome c interaction, one possibility is that calcium or other ions have entered the mitochondrial intracristal space, changing the ionic strength, and thus releasing cytochrome c from cardiolipin. K⁺ and Ca²⁺ are stored in the mitochondrial matrix [37]. However, since isolated mitochondria incubated with tBid do not release cytochrome c unless a large amount of KCl is added to the buffer, it seems reasonable to assume that the source of ion must come from outside the organelle. Another possibility is that cardiolipin is modified by reactive oxygen species (ROS) generated by altered respiratory chain reactions (both are reviewed in [38]). Here, we will briefly review the current knowledge regarding these two hypotheses.

8. ER-Ca²⁺-dependent cell death pathway

Unlike developmentally regulated programmed cell death that uses Bak/Bax-dependent apoptosis, external injuries to tissues such as wound healing or ischemia/reperfusion activate the ER-Ca²⁺-dependent cell death pathway.

To distinguish Bax/Bak-dependent intrinsic pathway of apoptosis from ER-Ca²⁺-dependent cell death pathway, Scorrano et al. used Bak/Bax double knockout cells (DKO cells) [22]. These DKO cells have little stores of Ca²⁺ in either ER or mitochondria. The sarco/endoplasmic reticulum Ca²⁺-ATPase (SERACA) is a Ca²⁺ transporter. Over-expression of SERCA enhanced Ca²⁺ uptake by the ER and restored Ca²⁺ levels to normal in DKO cells. This restored the sensitivity to ceramide, arachidonic acid and H₂O₂, but cells remained resistant to tBid-induced cell death [22]. Transfection of mitochondria-targeted Bax to DKO cells, on the other hand, restored the sensitivity of DKO cells to STS, etoposide and BFA-treatment, without restoring Ca²⁺ stores in these cells. As expected, DKO cells transfected with mitochondria-targeted Bax died an apoptotic death when tBid was introduced. However, these cells remained resistant to ceramide, arachidonic acid and H₂O₂ [22]. Thus, these sets of experiments showed that there exist two independent pathways of programmed cell death. In mouse fibroblasts, ceramide, arachidonic acid, and H₂O₂ use an ER-Ca²⁺-dependent programmed cell death which does not relay on MOMP to release cytochrome c whereas the expression of tBid triggers an ER-Ca²⁺ independent, Bax/Bak-dependent apoptosis, and STS and etoposide activate both pathways.

Lastly, agents such as H₂O₂ and ceramide induce Ca²⁺ release from the ER, which is taken up by mitochondria, and when the Ca²⁺ level in mitochondria reaches a threshold concentration, it activates the mitochondria permeability transition pores through which cytochrome c escapes, activating caspases in the cytosol.³ But this is very different from tBid-induced cytochrome c that does not require calcium influx. In tBid-induced apoptosis, cytochrome c escapes from mitochondria through Bak/Bax pores formed on the outer membrane. Interestingly, over-expression of Opa1 seems to delay cytochrome c release in both pathways [49]. It is not known how or even if cristae are remodeled during H₂O₂-induced or ceramide-induced cytochrome c release, but by the time cytochrome c is released, mitochondria are all likely to be swollen. Thus, remodeling of mitochondria under ER-Ca²⁺-induced cell death appears to be very different from tBid-induced cytochrome c, yet Opa1 plays a role in both of them. For this review, however, we will remain focused on tBid-induced cytochrome c release.

9. The role of Opa1 in maintaining cristae morphology

Mitochondria lacking Opa1 protein are largely devoid of cristae structures [66]. However, this finding is not unique. There are many more genes that affect the generation of cristae structures when their protein products are depleted [1]. One example gaining increasing interest is mitofilin [67]. The loss of specific isoforms of Opa1 can result in aberrant cristae morphogenesis and impaired cellular proliferation [68,69,66]. Furthermore, during siRNA treatment to inhibit Opa1 protein synthesis, the tubular network of mitochondria in cell lines becomes fragmented, indicating that Opa1 plays a role in mitochondrial fusion. Indeed, the yeast homologue of Opa1, Mgm1, is necessary for yeast mitochondrial inner membrane fusion [70]. However, we do not know exactly how Opa1 maintains cristae structures and how that is related to mitochondrial fusion activity.

Opa1 is a nuclear encoded gene and in humans, there are 8 splice variants of Opa1 [71,69]. In most tissue culture cells, at least two to three splice variants seem to be expressed. Each of these splice variants carry two to three cleavage sites by serine proteases. These sites were identified by Ishihara et al. [72] and designated MPP, S1 and S2. Once the product of a splice variant is imported into mitochondria, its mitochondrial localization signal peptide is cleaved at the MPP site, generating a mature membrane-bound Opa1 protein. Subsequently, second and possibly third cleavages can take place at either S1, S2 or both, producing shorter isoforms [72].

In most human and mouse cells, six isoforms of Opa1 are detectable by western blotting: two long membrane-bound isoforms and four short membrane-free isoforms. It is not known if each mitochondrion houses only long isoforms, only short isoforms or both. It is known, however, that the cells which had lost all the long isoforms of Opa1, such as found in patients suffering from mitochondrial myopathies or MEFs harboring an error-prone mitochondrial mtDNA polymerase α, contain fragmented mitochondrial networks and have reduced mitochondrial respiration rates [73]. Introducing various human Opa1 variants and their mutants into MEFs from Opa1 knockout mouse embryos, David Chan et al. demonstrated that at least one short isoform and one long membrane-bound isoform is necessary for the generation of fusion competent mitochondria [69]. Other approaches, such as over-expression of non-cleavable mutants or introduction of short isoforms, came to similar conclusions [68,71,74].

Because mitochondria need both long and short isoforms to support their fusion activity, knowing how conversion from long to short isoforms is regulated is of great interest. One could test if the same proteases cleave Opa1 during tBid-induced apoptosis. By knowing the identities of Opa1 processing enzymes, perhaps drugs could be developed to inhibit unwanted apoptosis by blocking Opa1 cleavage or do the opposite: prime tumor cells for apoptosis by activating Opa1 processing enzymes.

There are two conditions under which the complete conversion of all long isoforms of Opa1 to short isoforms takes place: one, during tBid-induced apoptosis and two, during the treatment of cells with the mitochondrial respiration inhibitor, CCCP. This latter process was inhibited by 1,10-phenanthroline, an inhibitor of metalloproteases, suggesting the involvement of m-AAA in the process [75]. Even though both processes convert Opa1 long isoforms to short isoforms, these two events differ in the way they alter cristae structures.

When isolated mitochondria are incubated with tBid, all the long isoforms of Opa1 are quickly converted to short isoforms and it is only the short isoforms that are released from mitochondria when Opa1 oligomers are disassembled [3]. However, the release of the short isoforms is not completed during the 30 min in which 100% of cytochrome c and Htra2/Omi molecules are both released. This might be because the membrane-free short isoforms are still attached to some other moieties in mitochondria, such as membrane-bound Parl. In tissue culture cells, the mitochondrial transmembrane potential is maintained for a few hours after the loss of cytochrome c [18,19]. On the other hand, CCCP-treatment of cells, as well as of isolated mitochondria, causes an immediate drop in the transmembrane potential and the long isoforms of Opa1 disappear from mitochondria in 10–30 min [75,66,69,76]. The mitochondrial network in these cultured cells becomes fragmented. The treatment does not cause immediate disassembly of Opa1 oligomers nor does it cause the loss of cristae structures. However, the long-term absence of Opa1 long isoforms seems to destabilize Opa1 complexes that eventually leads to the loss of cristae structures [68]. Thus, the changes to Opa1 brought on by two different treatments differ qualitatively — tBid-treatment causes immediate disassembly of Opa1 oligomers while CCCP-treatment cause only destabilization of Opa1 complexes.

10. Opa1 processing enzymes

Even though mitochondrial proteomes are slightly different from tissues to tissue, there are at most 20–30 serine proteases present in mitochondria. Three candidate serine proteases for processing mature membrane-bound Opa1 long isoforms are paraplegin, Yme1L and PARL. Mgm1 is cleaved by a rhomboid protease in mitochondria, Pcp1 [77]. When the mammalian homologue of Pcp1, PARL, is knocked out in mice, a wasting phenotype characterized by increased apoptosis was exhibited [49]. However, Parl^−/− cells displayed normal mitochondrial morphology, did not show defects in mitochondrial fusion/fission, and did not display primary respiratory defects. Nevertheless, Parl^−/− cells were more sensitive to apoptotic stimuli, such as etoposide, staurosporine and hydrogen peroxide. When exogenous Opa1 proteins were expressed in Parl^−/− cells, they became less sensitized to apoptotic stimuli. Thus, it is suggested that increased sensitivity to apoptotic stimuli in Parl^−/− cells may be due to altered Opa1 processing in these cells independent of mitochondrial fusion [24].

Yme1L (human Yme1-like) is an ATP-dependent membrane-inserted metalloprotease whose active site is in the intermembrane space thus classified as i-AAA [78]. David Chan's group showed that Yme1L cleaves Opa1 at S2 but not S1, They also discovered that the loss of transmembrane potential destabilized the long isoform of Opa1 and enhances the cleavage at S1 but not S2 [69]. Thus, Yme1L cannot be the sole protease that can generate short membrane-free Opa1 isoforms. van der Bliek's group showed that cells lacking either Parl or Yme1L responded to CCCP-treatment by cleaving the long isoforms of Opa1 [66]. Since Yme1L cannot cleave Opa1 at S1, their data imply that neither Parl nor Yme1L can cleave at S1. It was also found that Parl knockout did not affect Opa1 maturation nor did it affect induced proteolysis. Using a similar approach, Rojo's group also ruled out Parl and SPG7 (spastic paraplegia type 7) as serine proteases that generate short isoforms of Opa1 [76]. Reichert's group used yeast to reconstitute human Opa1 protein processing and found that human Parl expressed in yeast can cleave Mgm1 but not human Opa1 [79]. Ishihara et al. also expressed human Opa1 in yeast cells lacking various serine proteases and identified m-AAA as a candidate protease for Opa1 processing enzyme [72]. They also ruled out the involvement of paraplegin. In short, an S2 protease is known, but an S1 protease still has not been identified.

What could be the role of Parl? Mitochondria in Parl-depleted cells are more sensitive to apoptosis-inducing agents and their sensitivity is reversed by over-expression of Opa1. However, over-expression of Opa1 delays the induction of apoptosis in wild-type MEFs. Thus we do not know if Opa1 over-expression is restoring lost activity of Parl-processed Opa1 or simply protecting mitochondria by its abundant presence. Because Opa1 and Parl can co-precipitate each other [49], it is possible that the role of Parl is to stabilize Opa1 oligomers by providing an extra anchor. Stable Opa1 must be beneficial to the cell, making it more resistant to apoptosis. Not surprisingly, when wild-type Opa1 is over-expressed in MEFs, they become more resistant to apoptotic stimuli [24] while in Parl-depleted cells, Opa1 seems to make the cells less resistant [49,72].

The hypothesis that the role of Parl in cristae remodeling and apoptosis does not require its serine protease activity could be tested by expressing Parl point mutants defective in serine protease activity in Parl^−/− cells. If expression of such a mutant in Parl^−/− cells reverses the sensitivity, then Parl protease activity is not needed, and the role of Parl would be limited to its ability to associate with and stabilize Opa1 oligomers.

There is one more puzzle in this story. Prohibitins 1 and 2 are inner membrane proteins that form a large ring structure embedded in the inner membrane facing the intermembrane space, yet able to influence the matrix side. On the matrix side, the ring associates with an m-AAA protease complex [80]. It is known that knocking out one of the prohibitin genes destabilizes the protein product of the other, resulting in cells lacking both proteins. In prohibitin depleted cells, mitochondria are largely devoid of cristae much like Opa1 depleted cells, and both of the membrane-bound long isoforms of Opa1 are missing. The forced expression of a long isoform of Opa1 suppresses these defects. These data suggest that prohibitins regulate the serine protease(s) that cleaves membrane-bound long isoforms of Opa1.

11. Role of Opa1 in keeping cytochrome c inside crista

Using immuno-EM, Opa1 was found localized to the outer surface of the inner membrane seemingly throughout the intermembrane and intracristal spaces and not just at crista junctions [71,81]. Thus it seems that the Opa1 complex can be anywhere on the inner membrane.

Using yeast, Reichert et al. found that when protein synthesis and consequently mitochondrial import activities were arrested, Tim23, a part of the inner membrane portion of the import machinery that spans the mitochondrial inner membrane and outer membrane, was redistributed from predominantly inner boundary membrane localization to a more even distribution between inner boundary membranes and cristal membranes. They interpreted the results as follows: when mitochondrial protein import was arrested, it weakened the casual association between the N-terminus of Tim23 with the corresponding transporter complex of the outer membrane (a Tom), and since the C-terminus of Tim23 is embedded in the inner membrane, Tim23 diffused into the intracristal space. Their data suggest that there was no physical barrier blocking the lateral movement of inner membrane proteins across crista junctions [82].

If there is no barrier between inner boundary and cristal membranes, would there be lateral movement of cardiolipin across crista junctions as well? We do not know the distribution of cardiolipin since we do not yet have the means to separate inner boundary membranes from cristal membranes. But, since cardiolipin associating proteins such as cytochrome c and the oxidative phosphorylation (Oxphos) complexes are predominantly located in the cristal membranes, cardiolipin distribution might also be skewed towards cristal membranes. Reichert's group suggests that protein–protein association may also be a determinant factor for localization of cardiolipin [82]. Using their model, one could argue that once large complexes, such as the Oxphos complexes, are assembled, they would not easily be diffused and could even trap charged lipids like cardiolipin in the cristae. Furthermore, it is possible that the presence of Opa1 stabilizes Oxphos complexes and cytochrome c associating with them, thus keeping cytochrome c in the intracristal space. Indeed Frezza et al. showed that over-expression of wild-type Opa1 delayed the onset of H₂O₂-induced (ER-Ca²⁺-dependent) apoptosis [24].

What about soluble proteins and proteins that are loosely bound to membranes such as cytochrome c? Crista junctions in normal mitochondria can typically vary in mean diameter from 15 to 30 nm and vary much more in length. The mean diameter can be reduced to around 8 nm with tBid-treatment. Even at its narrowest, a hole of that size is big enough to let through 60–100 kDa proteins [83,84], hence it could let through cytochrome c molecules. At the same time, oligomerized Opa1 complexes may exceed several hundred kDa and they may be large enough to hinder free movement of soluble proteins across crista junctions.

12. Conclusion

“Happy families are all alike; every unhappy family is unhappy in its own way.” So begins Leo Tolstoy's novel, Anna Karenina. In many ways, studying apoptosis is like watching unhappy families. There are many ways leading to apoptotic cell death. Even if we restrict ourselves to the study of intrinsic forms of apoptosis, there are at least two distinct pathways: Bax/Bak-dependent pathway, which begins with the generation of BH3-only proteins that induce MOMP formation on mitochondria, and the ER-Ca²⁺-dependent pathway, which is triggered by calcium influx from the ER. This review focused on the molecular mechanism of tBid-induced cytochrome c release from mitochondria (Bax/Bak-dependent pathway) and internal changes that must take place before the release. Even though we briefly discussed the ER-Ca²⁺-dependent pathway for comparison purposes, we did not discuss how cristae are remodeled during Bax/Bak-independent ER-Ca²⁺-dependent cell death and the possible role of Opa1 in this pathway, nor did we examine the possible mechanisms of MOMP formation [7,9]. By narrowing the scope of this review, however, we were able to more completely cover what is known about the Bax/Bak-dependent pathway, identify key research issues and ongoing controversies, and to even propose experiments to help resolve these controversies. A model summarizing the cumulative knowledge of Opa1's role in tBid-induced apoptotic pathway is presented in Fig. 2. Even though tBid-induced cytochrome c takes place in the middle time frame of apoptosis, it does not happen in the way it was previously described: dying mitochondria first start to swell and then explode, spilling their entire contents throughout the cytosol. Instead, the evidence points to a complex interplay of events, causing subtle changes to mitochondrial structures. Three events necessary for the complete release of cytochrome c are the formation of mitochondrial outer membrane pores, the opening of crista junctions triggered by disassembly of Opa1 oligomers, and the dissolution of the cardiolipin–cytochrome c association. By examining the published evidence, one comes to understand the intricate timing of these events and the coordinated association of molecules that regulate this dynamic structural and functional correlation. Unresolved issues are (1) how Opa1 oligomers are assembled and disassembled and (2) how the electrostatic force between cardiolipin and cytochrome c is broken. The larger issue still to be addressed is how mitochondria cristae structure is generated and maintained. These are important questions because subtle defects in mitochondria are known to be the cause of premature aging and many age-related neurodegenerative diseases [85]. Furthermore, by better understanding this intricate organelle, we will learn how lipid and protein complexes can build and maintain very complex structures.