Two phases of signalling between mitochondria during apoptosis leading to early depolarisation and delayed cytochrome c release

Mitochondria have multiple roles in cells, including participation in bioenergetics, signalling and apoptosis. Two key events associated with mitochondrial involvement in apoptosis are the mitochondrial permeability transition (MPT) and the release of signalling proteins such as cytochrome c (cyt c) from mitochondria to the cytosol (reviewed by Desagher and Martinou, 2000). MPT involves the opening of a channel across the inner mitochondrial membrane,which leads to depolarisation (dissipation of mitochondrial membrane potential, Δψ_m) concomitant with free passage of ions and other small molecules less than 1.5 kDa in size. Release of cyt c from the mitochondrial intermembrane space to the cytosol occurs through the outer membrane. By forming an apoptosome complex with Apaf-1, pro-caspase-9 and ATP or dATP, the released cyt c ultimately triggers activation of downstream postmitochondrial caspases (caspase-3 and caspase-7). Activation of downstream caspases by proteolytic cleavage results in the biochemical and morphological changes characteristic of apoptosis (reviewed by Kaufmann and Hengartner,2001).

An intriguing general question is whether communication between the multiple mitochondria within individual cells occurs to cooperatively enhance the signalling role of these organelles in apoptosis. Concepts of communication between mitochondria arose, in part, from their ability to modulate propagation and integrity of Ca²⁺ waves under physiological conditions (Jouaville et al., 1995). Such notions were refined to encompass mitochondrial Ca²⁺-dependent Ca²⁺ release mechanisms(Ichas et al., 1997) and mitochondrial ROS-induced ROS release waves(Zorov et al., 2000). In excitable cells such as myotubes, lateral signalling between mitochondria induced by exogenous apoptotic stimuli was shown to be propagated by cytosolic Ca²⁺ waves. These Ca²⁺ waves in myotubes also induced travelling mitochondrial waves that involved depolarisation by the mitochondrial permeability transition (MPT) and release of cyt c, leading to caspase activation and nuclear damage characteristic of apoptosis(Pacher and Hajnóczky,2001).

In non-excitable cells, coordination between subsets of mitochondria remains poorly explored in these aforementioned terms. Electrical coupling between mitochondria under non-pathological conditions has been shown in COS-7 cells (De Giorgi et al., 2000)and human fibroblasts (Amchenkova et al.,1988; Diaz et al.,2000). This coupling takes the form of Δψ_mflickering in individual mitochondria or of synchronous depolarisation in whole subsets of mitochondria, suggestive of intermitochondrial communication. We recently demonstrated intermitochondrial communication in non-excitable human osteosarcoma 143B TK^- cells during apoptosis, by applying a microscopic photosensitisation technique(Lum et al., 2002) with the mitochondria-specific dye Chloromethyl-X-Rosamine (CMXRos; MitoTracker Red), a potent photosensitiser (Minamikawa et al.,1999a). We have used the laser scanning beam of a confocal microscope to irradiate a subset of mitochondria in an individual cell. This resulted in complete depolarisation of non-irradiated mitochondria in the same cell (Lum et al., 2002). In the present study we have used this system to distinguish two temporally distinct phases of signalling between irradiated mitochondria and their non-irradiated counterparts, namely the initial rapid depolarisation of the non-irradiated organelles and their subsequent delayed release of cyt c. These findings extend the current knowledge of intermitochondrial communication and show that multiple modes of such interorganellar signalling occur in stressed cells undergoing apoptosis.

1,2-bis-(o-Aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid tetra-(acetoxymethyl) ester (BAPTA-AM) and all fluorescent dyes including rhodamine 123 (Rh123), CMXRos, tetramethylrhodamine methyl ester (TMRM), Fluo 3-AM, 2′,7′-dichlorofluorescein diacetate (H₂DCFDA),Annexin V conjugated to Alexa Fluor 488, and goat anti-mouse IgG conjugated to Alexa Fluor 488 were purchased from Molecular Probes. Staurosporine (STS),carbonyl cyanide m-chloro phenyl hydrazone (CCCP), cyclosporin A (CsA), and puromycin were purchased from Sigma-Aldrich. Monoclonal mouse anti-cyt c(6H2.B4) and anti-Bax IgGs were from Pharmingen. zVAD-fmk was purchased from Bachem (Switzerland). N-acetylcysteine (NAC) was from Sigma.

Human osteosarcoma 143B TK^– (hereinafter called 143B)cells were cultured as previously described(Minamikawa et al., 1999b). For generation of 143B cells stably overexpressing Bcl-2, cells were transfected with pEF Bcl-2 pGKpuro (gift from D. Huang, Walter and Eliza Hall Institute, Melbourne) using Lipofectamine (Invitrogen). Transfectants were selected by growth in puromycin (50 μg/ml). Cell lines were generated from single cells cloned using limiting dilution culture. Clones expressing high levels of Bcl-2 were identified by immunofluorescence staining of permeabilised, fixed cells with monoclonal anti-Bcl-2 antibody (Bcl-2-100;Sigma). One such clone was selected and used throughout this study. The cell line expressing GFP targeted to the mitochondrial matrix (MtGFP) was constructed by stably transfecting 143B cells with pCZ34(Zhang et al., 1998) as previously described (Lim et al.,2001).

143B cells stably expressing cyt c-GFP were generated by infection of cells with retroviral supernatant (kindly provided by J. Goldstein, La Jolla Institute for Allergy and Immunology, San Diego, CA) containing the vector pBabe-puro with the cloned insert being mouse cyt c tagged at its C-terminus with GFP. Cells were infected at ∼50% confluency and two rounds of infection at 37°C were carried out at a ratio of 1:2 (v/v) viral supernatant to cells. The duration of each round of infection was 6-8 hours. Cells were then left to grow in fresh media until confluent prior to subculturing into 96-well plates for obtaining single cell colonies by limiting dilution. Clones expressing cyt c-GFP were selected in puromycin (0.5μg/ml) and screened by confocal analysis of cyt c-GFP fluorescence and localisation to mitochondria. In each clone so retrieved, a mixed population was encountered that consisted of cells with varying levels of expression and cyt c-GFP localisation.

However, for the purposes of the microscopic photosensitisation studies in this work, only cells that show expression of cyt c-GFP fluorescence predominantly in mitochondria were irradiated.

Cells were seeded the day before irradiation into 35 mm dishes at a density of 6×10⁴ cells per dish, the bottom of which was fenestrated and sealed with a round grid-coverslip (CELLocate, Eppendorf, grid size 175μm). For photosensitisation, cells were loaded with CMXRos (200 nM, 15 minutes at 37°C) and washed with phenol-red-free RPMI 1640 medium (Gibco BRL Life Technologies). Cells were maintained at 37°C in a temperature-controlled chamber (Life Science Resources, Cambridge, UK) on the stage of an inverted Leica DMIRB confocal microscope (TCS-NT system). Single cells targeted for photoirradiation were initially observed under dim transmission illumination from a tungsten lamp and the cellular position on the grid coverslip noted. The target cell was then positioned at the centre of the visual field and an image of the whole cell was obtained at low laser intensity (zoom 1, 2-4 scans) to minimise phototoxicity. A portion of the cell(typically half of the cell) was photoirradiated at high laser intensity to induce damage on a subpopulation of mitochondria in a given cell. Typically,partial irradiation was conducted by continuous xy scanning using the 488 and 568 nm excitation from the Ar/Kr laser at 128 scans (1 second/scan), zoom 8,and imaged with a Leica PL APO 63×/NA 1.2 water immersion objective. Immediately after irradiation, the whole cell was imaged at low laser intensity (zoom 1, 2 scans) and the same cell monitored at appropriate time points following subsequent staining with other fluorescent indicators for cell death events. Unstained cells that were partially irradiated under these conditions were able to undergo cell division more than 8 hours after irradiation.

Fluorescent images were obtained under confocal microscopy conditions as above. Excitation wavelength/detection filter settings for each of the fluorescent indicators were as follows: CMXRos and propidium iodide (PI),568/665-nm longpass; GFP, Alexa Fluor-488, Fluo-3 and dichlorofluorescein(DCF), 488/530-nm bandpass filter. For simultaneous imaging of CMXRos with either Rh123, GFP, Alexa Fluor-488, Fluo-3 or DCF, laser excitation used was 488 and 568 nm. Image analysis and processing was performed with Image J(National Institutes of Health, USA).

Most experiments used Rh123 (10 μM) loaded into cells for 15 minutes at 37°C. In some experiments, TMRM (150 nM) was used instead of Rh123 and cells were incubated with TMRM for 20 minutes at 37°C subsequent to partial irradiation. Following loading with either Rh123 or TMRM, cells were washed and maintained in phenol-red-free RPMI 1640 medium (also containing 50 nM TMRM for cells stained with TMRM) for confocal imaging. In the case of Rh123 loading, cells were scored as manifesting high Δψ_mor loss of Δψ_m, respectively, in their non-irradiated mitochondria on the basis of Rh123 retention values determined previously(Lum et al., 2002) that were either above 60% or below 20% of controls.

To induce conversion of filamentous mitochondria into discrete punctate entities, cells were treated with CCCP (20 μM) for 60 minutes followed by washout of CCCP with fresh RPMI 1640 media. After incubation for 60 minutes at 37°C, cells had recovered their Δψ_m(Minamikawa et al., 1999b) and were then loaded with CMXRos prior to irradiation.

Release of cytochrome c into the cytosol and translocation of Bax to the mitochondria, respectively, were assessed by immunocytochemical staining of cyt c (Lum et al., 2002) and Bax (Lim et al., 2001). Cyt c release from mitochondria was indicated by weaker diffuse staining across whole cell sections that did not colocalise with mitochondria. Phosphatidylserine (PS) exposure on the surface of the plasma membrane was detected by an annexin V binding assay as described previously(Lum et al., 2002). Cells that showed annexin V binding typically displayed a fluorescent ring on the periphery of the cell. Uptake of propidium iodide (PI) was assessed by incubating cells as described (Lum et al.,2002); necrotic (but not apoptotic) cells showed staining of nuclei contingent on plasma membrane permeabilisation. Cells were imaged by confocal microscopy for these various parameters relating to activation of the apoptotic cell death pathway.

For detection of intracellular Ca²⁺ levels, cells were loaded with the membrane-permeant Fluo-3 AM (2.5 μM) and incubated for 10 minutes at 37°C followed by washout of excess dye with phenol red-free media. To detect intracellular ROS levels, cells were loaded with the membrane-permeant H₂DCFDA following partial irradiation. Staining of cells with H₂DCFDA (40 μM) was performed in serum-free media by incubation for 20 minutes at 37°C followed by washout of excess dye with phenol red-free media. H₂DCFDA is hydrolysed by intracellular esterases to form DCFH (the reduced form of DCF) upon cleavage of the acetate groups on H₂DCFDA (Sawada et al.,1996). ROS such as H₂O₂ and hydroxyl radical but not superoxide, readily oxidise DCFH to result in the fluorescent DCF(Vanden Hoek et al., 1997). The intracellular Ca²⁺ and ROS levels in each partially irradiated cell were quantified using Image J by measuring the mean Fluo-3 and DCF fluorescence, respectively, in each irradiated cell in pixel units expressed relative to fluorescence of non-irradiated cells in the same field.

The cellular responses to partial irradiation (including loss or retention of Δψ_m, cyt c release, cyt c-GFP release, PS exposure,and PI uptake) were scored as a percentage of the total cells analysed in each category. To assess whether the responses between two different treatments were significantly different (P≤0.05), the relevant proportions of cells from each treatment were comparatively analysed by a chi-squared contingency test using Microsoft Excel 2000 software. A student's unpaired t-test was used for assessing any significant difference in the mean fluorescence value of Fluo-3 and DCF, comparing two groups of cells subjected to different treatments.

We previously reported that signalling between mitochondria occurs during cell death induced by microscopic photosensitisation with CMXRos(Lum et al., 2002). Such signalling was indicated by the rapid loss of Δψ_m in non-irradiated mitochondria in response to partial irradiation on a subset of mitochondria in single cells loaded with CMXRos (the photosensitising agent)and Rh123 (the reporter of Δψ_m)(Fig. 1A, Fig. 2A). To account for the loss of Δψ_m in non-irradiated mitochondria following partial irradiation, one might consider that non-irradiated mitochondria are part of an extended filamentous network directly connected to mitochondria in the irradiated zone. This was considered unlikely at the outset, since we previously showed that partial irradiation of cells loaded with Rh123 alone led to substantial retention of Rh123 intensity in non-irradiated mitochondria(Lum et al., 2002). In this study we addressed this issue by dismembering the initially filamentous mitochondrial network in 143B cells using CCCP. This protonophore reversibly depolarises mitochondria in these cells(Minamikawa et al., 1999b),converting mitochondria to punctate morphology(Fig. 1B). CCCP-treated cells failed to maintain Δψ_m (all cells lostΔψ _m, Fig. 2A) but completely recovered Δψ_m after withdrawal of CCCP (100% cells show high Δψ_m, Fig. 2A). Partial irradiation of such cells after CCCP treatment and recovery did not affect the ability of non-irradiated mitochondria to undergo complete Δψ_m loss(only 11% cells retained high Δψ_m, Fig. 2A). This was comparable to that in cells partially irradiated but without prior treatment with CCCP(14% cells, Fig. 2A). Control cells that were neither irradiated nor treated with CCCP retained highΔψ _m (100% cells, Fig. 2A).

In cells not exposed to CCCP, if one filament was a single long mitochondrion which allows relay of energy (represented byΔψ _m) along the whole filament to facilitate intramitochondrial communication(Amchenkova et al., 1988), then one would expect a less efficient response to partial irradiation when the filaments are dispersed by CCCP into smaller separate entities. However, our data show that both the Δψ_m loss(Fig. 2A) and the death responses (Fig. 2B) in cells containing discrete punctate mitochondria were not significantly different(P>0.1) from those containing filamentous mitochondria. Moreover,filamentous mitochondria in CMXRos-loaded cells that were subjected to irradiation of 128 seconds themselves showed fragmentation into punctate structures by 60 seconds (Fig. 3A), hence preventing signal diffusion along the whole filament within the irradiated region. The high laser intensity utilised under such partial irradiation conditions did not result in the fragmentation of filamentous mitochondria in control cells(Fig. 3B) not loaded with CMXRos (but labelled with mitochondrial matrix-targeted GFP). Taken together,these data suggest that Δψ_m loss in non-irradiated mitochondria occurs via intermitochondrial and not intramitochondrial communication.

It has been proposed that caspases and mitochondria can engage in a self-amplification loop in which the release of mitochondrial apoptogenic proteins activates caspases that would in turn increase the mitochondrial membrane permeability (Marzo et al.,1998). Additional episodes of cyt c release may occur, which have been shown to be due to caspase-3-dependent cleavage of Bcl-2(Kirsch et al., 1999; Chen et al., 2000). To determine whether the signal originating from irradiated mitochondria is amplified by downstream caspases, we treated cells with a broad spectrum caspase inhibitor zVAD-fmk (100 μM) prior to partial irradiation. zVAD-fmk at this concentration was effective in preventing caspase-3 activation and cell killing following apoptotic induction by STS and MT-21 in 143B cells(data not shown). However, zVAD-fmk did not inhibit eitherΔψ _m loss in non-irradiated mitochondria or cyt c release(Table 1). This suggests that caspases are involved neither in depolarisation of non-irradiated mitochondria nor in cyt c release that is known to occur upstream of caspase-3 activation.

Our previous study using immunocytochemistry revealed the release of cyt c into the cytoplasm following partial irradiation(Lum et al., 2002). However,it was not possible to identify the source of such dispersed cyt c (either from irradiated or non-irradiated mitochondria) because cells are fixed at a certain point in time and the dynamics of cyt c release events could not be readily analysed. To resolve whether non-irradiated mitochondria are activated by irradiated mitochondria to release cyt c during the initiation of apoptosis, we employed 143B cells that stably express cyt c-GFP targeted to mitochondria. These cells behave similarly to the parent cell line in response to apoptotic inductions by STS (0.5 μM, 1 μM),H₂O₂ (1 μM, 2 μM), and MT-21 (50 μM). This was revealed by the lack of significant difference between the two cell lines in terms of cell viability and proportion of cells showing cyt c release following the various apoptotic inductions (data not shown). We also ascertained that in general the cyt c-GFP reporter follows bulk cyt c localisation by performing immunocytochemical staining with cyt c antibody on cyt c-GFP expressing cells subjected to the various apoptotic inducers above,including treatment with STS, H₂O₂ and MT-21 (data not shown). In the experiments below we use cyt c-GFP as a reporter of release of cyt c from non-irradiated mitochondria following partial irradiation. For technical reasons outlined below, the bulk cyt c release detected immunochemically and the cyt c-GFP-derived fluorescence signal are now dissociated.

Cells expressing cyt c-GFP, initially showing appropriate mitochondrial localisation of fluorescence, were subjected to partial irradiation after loading with CMXRos (Fig. 4A). This typically results in conversion of the bright punctate cyt c-GFP fluorescence in irradiated mitochondria into a diffuse, weak signal in the cytosol. This dispersal and loss of signal intensity results from a combination of the release of cyt c-GFP from irradiated mitochondria together with some extent of photobleaching of cyt c-GFP signal (green channel) in the irradiated mitochondria. The unbleached non-irradiated mitochondria, however,retain punctate mitochondrially localised cyt c-GFP fluorescence(Fig. 4A, 0 minutes). Therefore, it becomes possible to monitor subsequent cyt c-GFP release from the non-irradiated mitochondria. In the cell depicted in Fig. 4A, release is judged to be occurring about 25 minutes after partial irradiation and is complete after 30 minutes. Staining of non-irradiated mitochondria with CMXRos (red channel)remains clearly evident although a general reorganisation of mitochondria to cluster around the nucleus takes place in such apoptosing cells(Fig. 4A). Based on this principle, we detected cyt c-GFP release from non-irradiated mitochondria of 55% of cells tested within 2.5 hours subsequent to partial irradiation(quantified in more detail, below). Non-irradiated control cells in the same field do not release cyt c-GFP, as indicated by the retention of a substantial cyt c-GFP signal localised to mitochondria(Fig. 4B).

Monitoring of partially irradiated cells at every 30 minutes revealed that many of those cells releasing cyt c-GFP (∼30% of total cells) did so within 1 hour after irradiation (Fig. 5, filled bars). A smaller proportion of cells (∼10% of total)showed cyt c-GFP release in the intervals 1-1.5 hours or 2-2.5 hours after irradiation. A minor proportion of cells (∼3%) showed release at 1.5-2 hours. But almost half the cells had still to release cyt c-GFP 2.5 hours after irradiation (it was not practicable to observe cells systematically beyond 2.5 hours). These data clearly indicate that the onset of cyt c-GFP release in the cell population is delayed and heterogeneous. Furthermore in experiments where release of cyt c-GFP was monitored every 5-10 minutes in another series of cells (n=8) subjected to partial irradiation, we found that the cyt c-GFP release was completely executed within 5-10 minutes subsequent to the onset of release. This was judged by the disappearance of any residual punctate green fluorescence corresponding to the position of mitochondria marked by CMXRos in the red channel. The duration of this rapid release was independent of the time of onset of release (in this experiment at 15, 28, 30, 38, 38, 40, 40, 49 minutes post-irradiation), where observations ceased at 90 minutes.

To determine whether the cyt c-GFP release from the non-irradiated mitochondria is caspase-dependent, the cells were treated with zVAD-fmk (100μM) before partial irradiation. A similar distribution of cells showing cyt c-GFP release at the various time intervals was obtained either in the presence (Fig. 5, open bars) or absence of zVAD-fmk (filled bars). Moreover, the proportion of cells that did not release cyt c-GFP by 2.5 hours remained unchanged by zVAD-fmk treatment(Fig. 5). These data suggest that the release of cyt c-GFP from non-irradiated mitochondria is not mediated by a caspase-dependent feedback amplification loop.

It is instructive to compare the timing of cyt c-GFP release from non-irradiated mitochondria (delayed) with that of release of cyt c from irradiated mitochondria which is already recognised to be a very early event(within 30 minutes) for cells globally irradiated(Minamikawa et al., 1999a). Note that our partial irradiation technique emphasises fluorescence due to cyt c-GFP within the non-irradiated mitochondria, while immunocytochemical staining reveals the distribution of all cyt c within the cell. In our experiments, the extensive release of cyt c detected by immunocytochemistry occurs in the vast majority of cells by two hours after irradiation (88%, n=27) [(Lum et al.,2002) and other data not shown]. In principle, this release(measured in untransfected 143B cells) could have arisen from either irradiated or non-irradiated mitochondria. However, as we have shown above,using cells expressing cyt c-GFP, only about half (55%, n=33) have released cyt c-GFP from non-irradiated mitochondria by 2.5 hours. We interpret the difference in release between cyt c and cyt c-GFP to indicate that in most partially irradiated cells, the immunocytochemically detectable early cyt c release arises mainly from irradiated mitochondria. We also suggest that this early cyt c release from irradiated mitochondria plays a major role in activating the cell death pathway in response to CMXRos photosensitisation(cf. Minamikawa et al.,1999a).

We next examined the relationship between cyt c-GFP release andΔψ _m loss following partial irradiation. Individual cyt c-GFP expressing cells were monitored simultaneously for both cyt c-GFP release and depolarisation which was detected by TMRM loaded after partial irradiation with CMXRos. Note that TMRM and GFP fluorescence emission can be detected in separate red and green channels respectively (whereas Rh123 and GFP emissions overlap in the green channel). Even though the emission of TMRM fluorescence is detected in the same red channel as CMXRos and is thus recorded concurrently with the CMXRos fluorescence, it is still possible to monitor Δψ_m based on uptake of TMRM in the mitochondria. This is because CMXRos undergoes substantial photobleaching following irradiation and any mitochondria that retain high Δψ_m do show a restored red fluorescence upon subsequent loading of these partially-irradiated cells with TMRM. The cyt c-GFP expressing cells analysed in this manner showed that 88% of cells (n=16) underwent rapid depolarisation in non-irradiated mitochondria consistent with that observed in parent cells. Of these 16 cells, only 2 did not undergo such depolarisation,their non-irradiated mitochondria becoming relatively brightly labelled with TMRM (data not shown). In general, about half of cells which had undergone depolarisation (9 out of 16) showed cyt c-GFP release from non-irradiated mitochondria within 2 hours, the other 7 cells still retaining cyt c-GFP in mitochondria at that time (consistent with data in Fig. 5). These data indicate that cyt c-GFP release is not an obligatory immediate consequence of depolarisation in non-irradiated mitochondria.

A hallmark of the participation of mitochondria in apoptotic signalling is the regulation afforded by pro-apoptotic and pro-survival members of the Bcl-2 family (Desagher and Martinou,2000). Overexpression of Bcl-2 in 143B cells was able to protect against oxidative stress induced by hydrogen peroxide (data not shown), in agreement with its ability to reduce the formation of ROS and prevent lipid peroxidation mediated by oxygen radical(Degli-Esposti et al., 1999; Hockenbery et al., 1993). Following partial irradiation of such Bcl-2 overexpressing cells, bothΔψ _m dissipation in non-irradiated mitochondria and release of cyt c were minimised (Table 1). This finding is consistent with the anti-apoptotic property of Bcl-2 in stabilising the barrier function of mitochondrial membranes, for example preventing cyt c release and interfering with loss ofΔψ _m (Desagher and Martinou, 2000). Overexpressed Bcl-2 however did not markedly protect globally irradiated cells from Δψ_m loss. All of 12 cells analysed under our standard intensive irradiation conditions showed depolarisation in mitochondria throughout the cell, indicated by the failure of any mitochondria to stain with TMRM loaded after irradiation (although we have not systematically tried to find conditions of global irradiation that might discriminate between control 143B cells and those overexpressing Bcl-2).

By contrast, pro-apoptotic Bax is well known as a mediator of death signalling, following translocation of Bax to mitochondria, which results in cyt c release (Desagher and Martinou,2000). We therefore used immunocytochemistry to examine whether Bax is recruited to non-irradiated mitochondria in response to partial irradiation. CMXRos was used as a mitochondrial marker in this study. In contrast to the positive control treatment which showed mitochondrial localisation of Bax in the great majority of cells following apoptotic induction by STS, Bax was evidently not localised to the non-irradiated mitochondria in any but a small minority of cases following partial irradiation of CMXRos-loaded cells (Fig. 6). The loss of Δψ_m in non-irradiated mitochondria is therefore not contingent on recruitment of observable quantities of Bax from other cellular locations.

Finally, we tested whether CsA is able to block depolarisation of non-irradiated mitochondria. We found this not to be the case for CsA at the relatively high concentration of 20 μM(Table 1), nor at lower concentrations (data not shown). The significance of these findings in relation to a general model for the two phases of intermitochondrial signalling will be considered in the Discussion section.

We observed a significant increase in intracellular Ca²⁺ in partially irradiated cells, as detected by Fluo-3 fluorescence (3-6 fold greater than in non-irradiated cells in the same field(Fig. 7A). Partially irradiated cells typically show a brighter and diffuse Fluo-3 fluorescence in the entire cell (Fig. 7B), distinct from the weaker Fluo-3 signal in non-irradiated cells. The increased intracellular Ca²⁺ levels following partial irradiation were reduced by pre-treatment of cells with intracellular Ca²⁺ chelator BAPTA-AM(Fig. 7A). Such reduction in Ca²⁺ levels by BAPTA alleviated the loss ofΔψ _m in non-irradiated mitochondria (43% of cells, Table 1) and the occurrence of cyt c release (41% of cells, Table 1), suggesting that increases in intracellular Ca²⁺promote intermitochondrial signalling. Similarly, a significantly higher intracellular ROS level measured by DCF fluorescence (4-8 fold greater than in non-irradiated cells, Fig. 7C),was also observed as a diffuse and bright fluorescence in partially irradiated cells (Fig. 7D). Laser irradiation on its own did not induce a cellular response as demonstrated by minimal levels of either Fluo-3 fluorescence(Fig. 7A) or DCF fluorescence(Fig. 7B) following partial irradiation of control cells not loaded with CMXRos.

Pretreatment of cells with NAC, an antioxidant that is also a metabolic precursor to glutathione (GSH), also inhibited both Δψ_mloss in non-irradiated mitochondria and cyt c release to a similar extent as BAPTA-AM (Table 1). However,these latter effects of NAC could not be attributed to its function as an antioxidant since it was unable to reduce intracellular ROS levels(Fig. 7C); note that the DCF signal was inhibited by catalase, a direct H₂O₂scavenger (data not shown). It is possible that NAC could instead exert a protective effect via its ability to restore intracellular GSH levels that are depleted during apoptosis and thus prevent onset of MPT(Liu et al., 2001) and ameliorate cyt c release. Interestingly, BAPTA-AM substantially inhibited the increase in intracellular ROS levels (Fig. 7C) in partially irradiated cells loaded with CMXRos. This suggests that the ROS increase may be mediated by the increased Ca²⁺. Finally we showed that influx of extracellular Ca²⁺ provides a major component of the enhanced intracellular Ca²⁺ levels following photosensitisation of CMXRos-loaded mitochondria (noting the reduced Fluo-3 fluorescence in cells photosensitised in Ca²⁺-free medium; data not shown).

We have exploited the specific mitochondrial localisation of CMXRos as a photosensitiser to dissect intermitochondrial communication during apoptosis. The general mechanism of severe intracellular stress brought about by photoirradiation of photosensitisers is considered to emanate from a burst of ROS consequential to irradiation (Penning and Dubbelman, 1994). Mitochondrially localised photosensitisers other than CMXRos, such as verteporfin(Granville et al., 1998; Belzacq et al., 2001) and phthalocyanine Pc 4 (Varnes et al.,1999) have been shown, after global photoirradiation of cells, to lead to apoptosis via mitochondrial depolarisation, cyt c release, and activation of caspases. In the case of CMXRos in osteosarcoma cells, we are confident that a similar mechanism applies such that we have shown depolarisation, cyt c release, annexin V binding(Minamikawa et al., 1999a) and caspase-3 activation (T. Minamikawa, M.-G.L. and P.N., unpublished).

On the basis of data obtained here applying microscopic photosensitisation by partial irradiation, we present a schematic that outlines our present understanding of the intermitochondrial communication(Fig. 8). This scheme indicates how severe mitochondrial stress induced by CMXRos photoexcitation triggers biphasic responses in non-irradiated mitochondria leading to apoptosis. Thus,we have established that signalling between mitochondria during apoptosis occurs in two temporally independent phases, notably early depolarisation(induced within 10 minutes after irradiation) and followed by later cyt c release that occurs at a delayed but unpredictable time afterΔψ _m loss. Moreover, we suggest that Ca²⁺ is likely to mediate the signal between irradiated and non-irradiated mitochondria and that the elevated levels of intracellular Ca²⁺ may also facilitate increases in intracellular ROS. The photodynamic aspects of CMXRos photosensitisation themselves generate a burst of ROS, but we observed that overall cellular ROS levels are suppressed by the Ca²⁺chelator BAPTA. In other work, Ca²⁺ has been shown to increase ROS production by disrupting the lipid organisation of the inner mitochondrial membrane and, consequently the respiratory chain function(Grijalba et al., 1999). The ROS produced induces permeabilisation of the inner mitochondrial membrane in a non-specific manner. Moreover, mitochondrial ROS production during cell death has been shown to be regulated by extracellular Ca²⁺ flux(Tan et al., 1998), consistent with our observation that elevated intracellular Ca²⁺ in the irradiated cells has a large extracellular Ca²⁺ derivation.

We found that the early Δψ_m loss in non-irradiated mitochondria was not due to lumenal continuity of mitochondrial filaments in the 143B cells used in our system. The lumenal discontinuity of mitochondrial filaments has been reported in other cell types such as HeLa, Cos-7, cortical astrocytes, cortical neuron, and human umbilical venous epithelial cells. It is thought that such discontinuity facilitates the heterogeneous functioning of mitochondria in relation to Δψ_m, sequestration of Ca²⁺ and activation of MPT(Collins et al., 2002). Propagation of Δψ_m along the whole filament(Skulachev, 2001) in our case would be unlikely since mitochondria in the irradiated zone become physically separated from their non-irradiated counterparts by fragmentation into smaller structures (Fig. 3A). Such fragmentation of irradiated mitochondria favours the model where mitochondrial filaments represent multiple mitochondria connected by intermitochondrial junctions (Skulachev, 2001). Indeed, this has been shown in 143B cells where each of the small mitochondria is equipped with most or all of the metabolic and biosynthetic functions(Margineantu et al., 2002). Disintegration of such filaments into smaller punctate mitochondria has also been observed in cells treated with apoptotic inducers(Frank et al., 2001).

Pretreatment of cells with zVAD-fmk failed to abolish the early depolarisation in non-irradiated mitochondria(Table 1), indicating the lack of participation of caspases in amplification of signal from irradiated mitochondria. This supports a direct mode of signalling between mitochondria,rather than secondary responses mediated by downstream feedback amplification. A similar lack of involvement of caspases in intermitochondrial signalling in myotubes was also noted (Pacher and Hajnóczky, 2001). It is possible that some signal either acting upstream or on the mitochondrial level mediates the rapid depolarisation in non-irradiated mitochondria since the latter is inhibited by overexpressed Bcl-2. In this respect, we have ascertained that Bcl-2 impeded the non-irradiated mitochondria in responding to the signal from irradiated mitochondria, rather than acting on the irradiated mitochondria to prevent initiation of signal. This is based on our finding that global irradiation on Bcl-2 overexpressing cells rapidly depolarised all mitochondria.

To determine whether MPT is involved in mediating theΔψ _m loss in non-irradiated mitochondria following partial irradiation, we pretreated the cells with the MPT inhibitor CsA prior to irradiation. We found that depolarisation in non-irradiated mitochondria and cyt c release were not inhibited, perhaps suggesting that MPT has not occurred. However, selective inner membrane permeabilisation does occur after global irradiation of CMXRos-loaded cells(Minamikawa et al., 1999a),detected by the calcein-cobalt quenching procedure which was developed as a reliable way of assessing MPT opening in living cells(Petronilli et al., 1999). Yet in these globally irradiated cells, CsA did not block calcein release(Minamikawa et al., 1999a) in contrast to cells undergoing MPT induced by CCCP(Minamikawa et al., 1999b). Such insensitivity to CsA in partially irradiated cells (as well as in cells subjected to global irradiation) could be attributed to the exposure of mitochondria (possibly including non-irradiated mitochondria) to levels of Ca²⁺ and ROS severe enough to render the MPT pore inhibitory actions of CsA ineffective. In agreement with this proposition, Brustovetsky and Dubinsky showed that in brain mitochondria which have been severely depolarised by Ca²⁺ could not be repolarised by CsA(Brustovetsky and Dubinsky,2000). Moreover, ROS may increase sensitivity of MPT to Ca²⁺ through oxidation of thiol groups on the MPT pore, such that opening occurs at high Ca²⁺ levels in a CsA-insensitive manner(Halestrap et al., 1997). We therefore cannot eliminate the possibility that in the cases of both irradiated and non-irradiated mitochondria the primary event of depolarisation involves the opening of the MPT pore, but in a CsA-insensitive manner.

By employing a cell line that stably expresses cyt c-GFP, we found that cyt c-GFP released from non-irradiated mitochondria was delayed, occurring at a specifically unpredictable time but being rapidly completed (within 10 minutes) subsequent to the onset of release. The heterogeneous nature of the onset of cyt c-GFP release (Fig. 5) is in general, similar to that described by Goldstein et al.(Goldstein et al., 2000),although in the latter case cyt c-GFP release is delayed even more than under our partial irradiation conditions. Moreover, in this other study on apoptotic induction not involving photosensitisation, mitochondria maintainΔψ _m (or may even be hyperpolarised) prior to release of cyt c-GFP (Goldstein et al.,2000). But in the case of the partial irradiation studies we carried out, mitochondria in almost all cases are clearly depolarised long before cyt c-GFP release. This confirms that the release of cyt c from mitochondria is independent of the state of polarisation of mitochondria and cannot be obligatorily linked to MPT events(Lim et al., 2001), unlike in some apoptotic systems where MPT and cyt c release seem to be coincident and mutually interdependent (Kroemer et al.,1998).

We have found that cyt c-GFP release from non-irradiated mitochondria is not a secondary event due to downstream caspase mediated amplification loop(Fig. 5). This caspase-independent release of cyt c-GFP is also observed in excitable myotubes during intermitochondrial communication(Pacher and Hajnóczky,2001). Based on the data shown here, we suggest that Ca²⁺ is involved in what appears to be direct intermitochondrial communication.

In an attempt to uncover the events leading to cyt c release from the non-irradiated mitochondria, we looked for the participation of Bax. However,we found no evidence of mass recruitment of Bax to mitochondria. In our situation, we cannot exclude that Bax is recruited to non-irradiated mitochondria, in amounts below the level of detection, which may trigger cyt c release. The reduced Bax requirement may arise in circumstances where relatively little Bax is needed because of previous chronic bombardment of non-irradiated mitochondria with Ca²⁺ or other signals originating from irradiated mitochondria, such as ROS. By contrast, in cells undergoing STS-induced apoptosis the recruitment of Bax to mitochondria was the same irrespective of whether mitochondria were polarised or not (the latter induced by prior CCCP treatment) (Lim et al.,2001). Unlike in other systems, where it has been reported that MPT events serve as the signal for recruitment of Bax to mitochondria(Smaili et al., 2001; De Giorgi et al., 2002), the depolarisation of mitochondria resulting from CMXRos photosensitisation does not evidently lead to such Bax recruitment.

We thank D. C. Huang for kindly providing the Bcl-2 expression construct pEF Bcl-2 pGKpuro, N. Lazar-Adler for generating the cell lines stably expressing mitochondrially-targeted GFP and overexpressing Bcl-2, and J. C. Goldstein for providing the retroviral supernatant containing the cyt c-GFP vector. We also thank T. Minamikawa and I. S. Harper for technical advice and useful discussions.