The Caspase-3 Precursor Has a Cytosolic and Mitochondrial Distribution: Implications for Apoptotic Signaling

The large subunit of caspase-3 was engineered as a MetSer29-Asp175 construct, expressed from an IPTG-inducible vector (pET11d; Novagen, Inc., Madison, WI) in Escherichia coli BL21(DE3)pLysS cells (Novagen, Inc.) and recovered by purification of inclusion bodies. The resulting recombinant protein (>98% pure as judged by SDS-PAGE) was denatured and resuspended in 6 M guanidine HCl, 25 mM Tris-HCl, pH 7.4, and used directly to immunize rabbits. Antibody MF393 was generated similarly and had identical characteristics (data not shown).

(a) Immunoblotting. Recombinant caspase large subunits (caspases 1–10, excluding caspase 6, and Ced-3) were generated by expression in bacteria as described previously (Rotonda et al., 1996). 10 ng of each purified caspase large subunit was immunoblotted with antiserum R280. The antibody recognized the large subunit of caspase-3 exclusively (Fig. 1,a). (b) Immunoprecipitation. Recombinant caspase-3 was biotinylated by incubation with 10 nM of the irreversible inhibitor biotin-DEVD-acyloxymethylketone (L772,094) in buffer containing 50 mM Hepes pH 7.0, 2 mM EDTA, 0.1% (wt/vol) 3-([3-cholamidopropyl]-dimethyllammonio)-1-propanesulfonate (CHAPS), 10% (wt/ vol) sucrose, and 5 mM DTT for 30 min at 37°C. Caspase-3 was then denatured by incubating either in 4 M guanidine-HCl/16.6 mM Tris-HCl, pH 7.4, at room temperature for 15 min, or in 0.7% (wt/vol) SDS in 67 mM Tris-HCl at 95°C for 5 min. 0.1-pmol amounts of the biotinylated native and denatured forms of caspase-3 were immunoprecipitated with R280 antiserum essentially as described (Taylor et al., 1995) (Fig. 1,b). Samples were visualized by immunoblotting with peroxidase-labeled streptavidin (Amersham Corp., Arlington Heights, IL) and enhanced chemiluminescence (ECL) (Amersham Corp.). R280 antiserum immunoprecipitated only the mature caspase-3 that had been denatured by boiling in SDS (Fig. 1 b). These results demonstrate that the R280 antiserum does not recognize the folded heterodimeric protease (although it can immunoprecipitate the precursor form of caspase-3 generated by coupled in vitro transcription/translation [data not shown]). (c) Competition Experiments. When decreasing amounts of antibody were used to immunoprecipitate a mixture of the precursor and mature forms of caspase-3 (generated by partially processing [³⁵S]methionine-labeled caspase-3 precursor with granzyme B), R280 exclusively immunoprecipitated the precursor form of caspase-3 (data not shown).

Purified p17 subunit of caspase-3 was spotted onto Immobilon (Millipore Corp., Bedford, MA) and used to affinity purify antibodies to caspase-3 from R280 antiserum as described (Olmsted, 1981). The amount of antibody was quantitated by comparing immunoblots of the affinity-purified antibody to dilutions of rabbit IgG standards. When these affinity-purified antibodies were used to immunoblot lysates of human umbilical vein endothelial cells (HUVECs) (Fig. 1 c), HeLa cells (data not shown), or keratinocytes (data not shown), they recognized only the caspase-3 precursor. A very minor band (<2% of the caspase-3 precursor intensity) of 120 kD was variably seen.

Human keratinocytes were obtained from neonatal foreskins and cultured in keratinocyte growth medium (KGM) (Clonetics Corp., San Diego, CA) as described (Casciola-Rosen et al., 1994a). Confluent secondary monolayers were used for experiments. HUVECs were obtained from Clonetics Corp., grown in endothelial cell growth medium (Clonetics Corp.), and used at the third or fourth passage. HeLa and WIF-B cells were cultured as described (Ihrke et al., 1993; Casciola-Rosen et al., 1994b). Apoptosis was induced by irradiation with UVB (Casciola-Rosen et al., 1994a), C2-ceramide (Haimowitz-Friedman et al., 1994), or staurosporine (Jacobson et al., 1996).

Immunofluorescence was performed on cells that had been grown on No. 1 glass coverslips.

Cells were fixed with 4% paraformaldehyde for 5 min at 4°C before being permeabilized with acetone (30 s at 4°C). To visualize the precursor form of caspase-3, cells were incubated for 30 min on ice with 10 μg/ml affinity-purified anti–caspase-3 antibody (R280) and then washed in PBS before incubating with fluorescein-conjugated goat anti–rabbit IgG (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) for 20 min at room temperature.

Mitochondria were labeled in intact cells with MitoTracker™ CMXRos-H₂ (Molecular Probes, Inc., Eugene, OR) according to the manufacturer's directions. Briefly, MitoTracker was diluted to a final concentration of 1 μM in serum-free medium (KGM for keratinocytes and DME for HUVECs). Coverslips containing viable keratinocytes and HUVECs were incubated in the appropriate MitoTracker-containing medium for 30–45 min in a 5% CO₂ humidified incubator. The cells were then fixed and permeabilized and either mounted immediately (for single-labeled mitochondria) or were double-labeled with anti– caspase-3 antibody R280 as described above. When specified, nuclei were visualized by labeling with 4′,6-diamidino-2-phenylindole (DAPI) (Molecular Probes, Inc.). In all cases, coverslips were mounted on glass slides with PermaFluor (Lipshaw, Pittsburgh, PA), and confocal microscopy was performed on a scanning confocal microscope system (model LSM 410; Carl Zeiss, Inc., Thornwood, NY).

Immuno-EM was performed essentially as described (Berryman et al., 1992). Briefly, WIF-B cells (Ihrke et al., 1993) were fixed at room temperature for 10 min (3% paraformaldehyde, 0.05% glutaraldehyde in Hank's salt solution) and scraped, and the cell pellet was fixed for an additional 45 min in this fixative. Cell pellets were washed with 3.5% sucrose in 0.1 M phosphate buffer, pH 7.4 (SPB), and postfixed with 0.25% tannic acid in SPB for 60 min at 0°C before quenching free aldehydes for 60 min with 50 mM ammonium chloride in SPB. After washing in sucrose-maleate buffer, pH 6.5, containing 4% sucrose, cells were stained en bloc with 2% uranyl acetate in sucrose-maleate buffer, pH 6.0, dehydrated, and embedded in LR gold resin at −20°C. Silver-gold sections were mounted on formvar-coated 200-mesh nickel grids, blocked in 1% BSA/TBST (10 mM Tris, pH 7.2, 500 mM NaCl, and 0.05% Tween 20), and stained for 2.5 h with affinity-purified R280 antibodies (15 μg/ml). After rinsing, grids were stained for 60 min with 12-nm colloidal gold-labeled donkey anti–rabbit IgG (Jackson ImmunoResearch Laboratories, Inc.; OD₅₂₀ = 0.05 [1:40 dilution]) followed by 2% aqueous osmium tetroxide (15 min) and then counterstained for 3 min with Reynold's lead citrate. Sections were viewed and photographed in an electron microscope (TEM type 1A; Carl Zeizz, Inc.), and prints were made to a final enlargement of 21,000 magnification. For quantitation, the area of mitochondria and extramitochondrial structures was measured, and the number of gold particles/cm² associated with mitochondria or other organelles was calculated (>200 gold particles in 10 separate prints were quantitated).

Mitochondria were isolated from human liver recovered at autopsy of a 17-yr-old male (cause of death: accidental) or rat liver as previously described (Nicholson and McMurray, 1984). All steps were performed at 4°C. The tissue was rinsed in SME buffer (0.25 M sucrose, 10 mM MOPS, 2 mM EDTA, pH 7.4), rid of blood vessels, and dispersed in SME buffer. The suspension was centrifuged twice at 480 g to sediment out nuclei and large cellular debris, and the mitochondria were recovered by centrifugation at 4,300 g for 10 min followed by a 2-min centrifugation at 10,000 g. After washing the mitochondrial pellet four times in SME buffer to minimize microsomal contamination, the mitochondria were resuspended in SME buffer to a protein concentration of 16 mg/ml and stored at −80°C.

To perform the experiments shown in Fig. 3, mitochondria (200 μg) were diluted in SME buffer with or without NP-40 (final concentration 1%). After a 1-h incubation at 37°C in the presence or absence of 0.2 μg granzyme B, the mitochondrial suspension was analyzed by SDS-PAGE and immunoblotted with the R280 anti–caspase-3 antibody essentially as described (Scoggan et al., 1996) except that the primary antibody was used at a dilution of 1:5,000. Identical results were obtained when mitochondria were disrupted by repeated freeze-thaw (data not shown). Progressive digitonin extraction and enzyme assays to measure adenylate kinase and fumarase were performed exactly as described on freshly isolated mitochondria (Nicholson et al., 1987).

Mitochondrial and cytosolic fractions were obtained by differential centrifugation of HeLa cell homogenates as described (Vander Heiden et al., 1997). Briefly, cells were washed twice with PBS, scraped, pelleted, and resuspended in 2.5 ml of ice-cold buffer A (250 mM sucrose, 20 mM Hepes, pH 7.4, 10 mM KCl, 1.5 mM MgCl₂, 1 mM EGTA, 1 mM EDTA, 1 mM DTT, and the protease inhibitors PMSF, leupeptin, chymostatin, pepstatin, and antipain). Cells were homogenized with a Balch/Rothman homogenizer (Balch and Rothman, 1985), with ball bearings that allowed calculated clearances of 0.0019 in or 0.0011 in. A postnuclear supernatant was prepared and centrifuged at 10,000 g for 25 min to obtain a pellet highly enriched in mitochondria. The pellet was resuspended in a volume of buffer A equal to that of the postmitochondrial supernatant. The postmitochondrial supernatant was further centrifuged at 100,000 g to obtain cytosol. Equivalent volume amounts of mitochondrial and cytosolic fractions were electrophoresed on 13.5% SDS–polyacrylamide gels, transferred to polyvinyl difluoride, and immunoblotted with R280 (1:5,000) or monoclonal antibodies to human cytochrome c (mAb 7H8.2C12 at 1.5 μg/ml; Pharmingen, San Diego, CA) and cytochrome oxidase (subunit IV) (mAb 20E8-C12 at 0.1 μg/ml; Molecular Probes, Eugene, OR). Relative distributions of proteins in the two fractions were assessed by densitometry as described below. While >95% of cytochrome c was associated with the mitochondrial fraction in cells homogenized with a ball bearing clearance of 0.0019 in, >95% of cytochrome c was found in the cytosol when cells were homogenized with a ball bearing clearance of 0.0011 in (Fig. 3,c and data not shown; seven separate experiments). In contrast, >95% of cytochrome oxidase was located in the mitochondrial fraction regardless of the ball bearing clearance (Fig. 3 c and data not shown).

Granzyme B was purified from granules isolated from cultured human natural killer leukemia YT cells essentially as described (Hanna et al., 1993; Quan et al., 1995). The granule contents were extracted with NaCl (the calcium was omitted), diluted, and then loaded onto a Mono-S cation-exchange column (0.5 × 0.5 cm; Pharmacia Biotech, Inc., Piscataway, NJ) that had been pre-equilibrated in 50 mM MES, pH 6.1, 25 mM NaCl. After washing with 12 vol of equilibration buffer, proteins were eluted with a linear gradient up to 1 M NaCl (in 50 mM MES, pH 6.1) over 40 column volumes. Granzyme B eluted at ∼0.6 M NaCl and was homogenous as judged by SDS-PAGE.

Control or apoptotic keratinocytes (incubated for the indicated times after UVB irradiation) were washed in Krebs Ringers buffer (KRB) containing 20 mM Hepes/NaOH, pH 7.4, 10 mM dextrose, 127 mM NaCl, 5.5 mM KCl, 1 mM CaCl₂, and 2 mM MgSO₄, and then harvested by scraping into KRB, followed by centrifugation at 300 g. The harvested cells were lysed in buffer containing 10 mM Hepes/KOH, pH 7.4, 2 mM EDTA, 1 mM DTT, 1% NP-40 (vol/vol), and the protease inhibitors PMSF, antipain, leupeptin, and pepstatin A to obtain whole cell lysates (protein concentration = 2.5 mg/ml). Aliquots of lysates were taken for protein assays, and DTT was then added to samples to obtain a final concentration of 5 mM DTT.

Whole cell lysates were incubated for 0 or 2 h at 37°C, as indicated in Fig. 7, before the addition of SDS electrophoresis sample buffer. Samples were electrophoresed on 10% SDS–polyacrylamide gels containing 0.087% bisacrylamide, and the proteins were transferred to nitrocellulose. Endogenous poly (ADP-ribose) polymerase (PARP) was detected by immunoblotting with a patient serum monospecific for PARP as described (Casciola-Rosen et al., 1995). The densities of bands on autoradiograms were determined on a PDI Discovery densitometry system (Protein Databases, Inc., Huntington Station, NY) with Quantity One software (Protein Databases, Inc.).

To measure Ac-DEVD-aminomethylcoumarin (Ac-DEVD-AMC) cleavage activity, lysates were added to reaction mixtures containing 10 or 100 μM Ac-DEVD-AMC, 100 mM Hepes, 10% sucrose, 0.1% CHAPS, 0.1 mM EDTA, and 1 mM DTT, pH 7.5, in a total volume of 100 μl. Production of AMC was monitored continuously at ambient temperature in an SLT Fluostar 96-well plate reader (Tecan U.S. Inc., Research Triangle Park, N.C.) using an excitation wavelength of 380 nm and an emission wavelength of 460 nm. To confirm that the observed activity was due to caspase-3 or a closely related homologue, each sample was tested for inhibition by N-ethylmaleimide, Ac-DEVD-CHO (K_i caspase-3 = 0.35 nM), and Ac-YVAD-CHO (K_i caspase-3 = 10,000 nM). Activities that were inhibited by N-ethylmaleimide (5 mM) and Ac-DEVD-CHO (500 nM) but not by Ac-YVAD-CHO (500 nM) were attributed to caspase-3 or a closely related homologue. Activity is expressed in U/mg, where a unit is defined as the amount of enzyme required to produce 1 pmol of AMC per min using 100 μM Ac-DEVD-AMC.

The subcellular location of the caspase-3 precursor in human foreskin keratinocytes was first determined with confocal fluorescence microscopy (Fig. 2). Using affinity-purified polyclonal antibody R280, which exclusively recognizes the caspase-3 precursor (see Materials and Methods), a diffuse cytosolic staining pattern with a superimposed punctate perinuclear component was obtained; low levels of diffuse nuclear staining were also observed (Fig. 2,b). The punctate component appeared similar to that seen for mitochondria, prompting us to double-stain keratinocytes with a mitochondrion-selective vital dye (MitoTracker) and R280. These studies confirmed that mitochondria were distributed in a punctate, perinuclear pattern in keratinocytes (Fig. 2,a), and that the punctate fraction of caspase-3 colocalized with mitochondria (Fig 2 c; overlapping red and green pixels seen as yellow).

The specificity of the caspase-3 precursor staining was confirmed in the following ways: (a) Competition experiments were performed by preincubating affinity-purified anti–caspase-3 (R280) antibody with a 100-fold excess of its p17 cognate antigen before use in immunofluorescence. In cells stained with these preincubated antibodies, the punctate pattern (mitochondria) was abolished, and the diffuse cytoplasmic staining was greatly diminished (data not shown). Since nuclear staining was minimally affected, this most likely represents nonspecific binding (data not shown). In contrast, when the anti–caspase-3 antibody was similarly preincubated with the p20 subunit of caspase-1, the punctate staining and diffuse cytoplasmic staining were not affected (data not shown). (b) No staining was observed when keratinocytes were incubated with secondary antibody alone, nor when an affinity-purified anti– caspase-1 antibody was used (data not shown). (c) When cells were single-stained with anti–caspase-3 antibodies, an identical punctate staining pattern was seen, confirming that this pattern was directly due to antibody binding, rather than bleedthrough between channels in the double-stained samples (data not shown). (d) Perinuclear punctate and diffuse cytoplasmic staining patterns were observed when cells were fixed with 2% paraformaldehyde/ 0.5% glutaraldehyde; the level of cytoplasmic staining was increased and nuclear staining was decreased using this fixation technique (data not shown). (e) An identical staining pattern was observed when a different rabbit anti– caspase-3 precursor serum (antibody MF393; raised against the unfolded p17 subunit of caspase-3) was used to stain keratinocytes (data not shown). Taken together, these data show that the caspase-3 precursor has a cytosolic and mitochondrial distribution in keratinocytes.

To examine whether this distribution was unique to keratinocytes, similar immunofluorescence experiments were performed using several other cell types. In HUVECs, mitochondria stained with a filamentous pattern that extended through the cytoplasm to the cell periphery (Fig. 2,d). The caspase-3 precursor shared an identical filamentous pattern but was also present diffusely in the cytoplasm (Fig. 2,e). Double staining confirmed colocalization of caspase-3 precursor and mitochondria (Fig. 2 f; overlapping red and green pixels seen as yellow). A similar mitochondrial and cytosolic distribution of the caspase-3 precursor was also observed in several other cells, including primary human myoblasts and myotubules, primary differentiated human neuronal cell cultures, MDCK cells, HeLa cells, and WIF-B cells (data not shown).

To confirm biochemically that the caspase-3 precursor is localized in mitochondria, intact mitochondria were prepared and immunoblotted with R280 antibodies. The 32-kD caspase-3 precursor alone was detected in these samples (Fig. 3,a, lane 1). To address whether caspase-3 was contained within these organelles or was associated with the outer surface, we incubated mitochondria in the presence of granzyme B, a cytolytic lymphocyte granule protease known to induce the processing and activation of caspase-3 (Darmon et al., 1995; Quan et al., 1996). In NP-40–permeabilized mitochondria (Fig. 3,a, lane 4), but not in mitochondria incubated in the absence of detergent (Fig. 3 a, lane 2), addition of granzyme B produced an 80% decrease in the amount of immunoblotted caspase-3 precursor, with concomitant appearance of the p17 subunit, as previously described (Martin et al., 1996; Quan et al., 1996). Similar results were obtained when repeated freeze-thawing was used to permeabilize mitochondria (data not shown). These biochemical data confirm the morphological finding that the caspase-3 precursor has a mitochondrial distribution and indicate that it is contained within the organelle.

To investigate the submitochondrial location of the caspase-3 precursor biochemically, purified mitochondria were treated with increasing concentrations of digitonin before recentrifugation. The relative amounts of the caspase-3 precursor, adenylate kinase (a mitochondrial intermembrane space marker enzyme), and fumarase (a mitochondrial matrix marker) were assayed in the resulting pellets and supernatants. The release of caspase-3 precursor from mitochondria at progressively higher digitonin concentrations coincided with adenylate kinase, with 50% of the total appearing in the supernatant at 0.15% (adenylate kinase) and 0.125% (caspase-3 precursor) digitonin (Fig. 3,b). In contrast, 50% fumarase release into the supernatant only occurred at 0.65% digitonin (Fig. 3 b).

Since intermembrane space contents frequently leak from mitochondria during subcellular fractionation procedures, establishing an accurate stoichiometry of distribution of caspase-3 precursor in mitochondrial versus postmitochondrial fractions required a fractionation procedure that maintained the outer mitochondrial membrane integrity (see Materials and Methods). Cytochrome c and cytochrome oxidase (subunit IV) were used as markers of the intermembrane space and inner membrane, respectively. Under homogenization conditions where >95% of cytochrome c was retained in mitochondria, the amount of caspase-3 precursor in mitochondria was ∼10% that found in cytosol (equal cell equivalents assayed; Fig. 3,c). In several separate experiments, caspase-3 immunoblotted as both the intact 32-kD form and a 29-kD form, most likely representing the protease without its prodomain (Fig. 3,c). In contrast, very low levels (<1%) of caspase-3 precursor were retained in mitochondria under conditions where >90% of cytochrome c was present in the cytosol (data not shown). Taken together, these data demonstrate that a significant proportion of the caspase-3 precursor is mitochondrial in distribution. Furthermore, since caspase-3 precursor is released from mitochondria at digitonin concentrations almost identical to those required to release adenylate kinase (Fig. 3,b) and retained in mitochondria only during homogenization conditions that maintain cytochrome c in the intermembrane space (Fig. 3,c), we conclude that caspase-3 is most likely located in the mitochondrial intermembrane space. These data are also consistent with the inaccessibility of the caspase-3 precursor to granzyme B in nonpermeabilized mitochondria (Fig. 3 a). It is also possible that the caspase-3 precursor is tightly associated with the outer surface of mitochondria.

To further confirm the ultrastructural localization of the mitochondrial caspase-3 precursor, ultrathin sections of WIF-B (Fig. 4) and HeLa cells (data not shown) were immunolabeled with affinity-purified R280 antibodies. Immunogold labeling was predominantly mitochondrial (63 ± 13%; n = 8 pictures, mean ± SD) and cytosolic (37 ± 12%; n = 8, mean ± SD); <5% of labeling was associated with other structures, including nucleus, endoplasmic reticulum, and plasma membrane (data not shown). The density of mitochondrial labeling (0.33 ± 0.09 gold particles/cm²; n = 246 mitochondria, mean ± SD) was approximately sixfold higher than that observed in all areas outside mitochondria (0.05 ± 0.02 gold particles/cm²). 30–50% of mitochondria were labeled with at least one gold particle; frequent examples of mitochondria labeled with multiple gold particles were found. It must be noted that significant extraction of cytosolic contents occurs during the fixation used to preserve optimal immunogenicity of the caspase-3 precursor for R280. It is likely that loss of cytosolic caspase-3 under these circumstances results in the different stoichiometries of caspase-3 distribution observed with biochemical and morphological methodologies. In all instances where mitochondrial labeling was observed, gold labeling was inside mitochondria, in close proximity to mitochondrial membranes. Particles were detected both immediately inside the outer membrane, most likely in the intermembrane space (Fig. 4, a and c) or well within the organelle, always near mitochondrial inner membrane cristae (Fig. 4, a, b, and d, and data not shown). These gold particles were found both alongside the membrane or within the cristal lumen (Fig. 4 c). Since the space within cristae is continuous with the intermembrane space, these data further support the confocal and biochemical data that localize the caspase-3 precursor to the mitochondrial intermembrane space.

Previous studies have described several novel features of mitochondria during apoptosis, including cytosolic leakage of cytochrome c (Liu et al., 1996; Kluck et al., 1997; Yang et al., 1997), and loss of mitochondrial transmembrane potential (for review see Kroemer et al., 1997). To determine the subcellular distribution of the caspase-3 precursor after induction of apoptosis, we performed immunofluorescence confocal microscopy on keratinocytes rendered apoptotic by irradiation with UVB. Nonirradiated cells showed the mitochondrial and cytosolic patterns of caspase-3 precursor staining described above (Fig. 5,b). In apoptotic cells, caspase-3 precursor staining was diminished, with no mitochondrial staining pattern visible (Fig. 5,e). In merged images, MitoTracker-stained mitochondria therefore appeared red (Fig. 5,f) rather than yellow (Fig. 5,c). A similar disappearance of mitochondrial caspase-3 precursor staining was observed in UVB-irradiated HUVECs and HeLa cells, as well as in keratinocytes and HeLa cells in which apoptosis was induced with C2-ceramide (data not shown). Interestingly, no enrichment of caspase-3 precursor staining was observed in apoptotic surface blebs (Fig. 5 e, arrowheads), although we have detected the mature active enzyme in these structures (data not shown).

We also addressed the effects of staurosporine treatment on the mitochondrial staining of the caspase-3 precursor in keratinocytes (Fig. 6) and HeLa cells (data not shown). This agent does not induce the characteristic surface blebbing and cytosolic shrinkage of apoptosis but does induce classic nuclear condensation and fragmentation (detected by DAPI staining in Fig. 6,a), caspase-mediated substrate cleavage, and internucleosomal DNA degradation (Jacobson et al., 1996; and data not shown). Since apoptosis occurs asynchronously in cell populations, adjacent nonapoptotic and apoptotic cells can be visualized simultaneously. In cells with the condensed, fragmented nucleus characteristic of apoptotic cells (Fig. 6,a, arrowheads), mitochondrial caspase-3 staining was absent (Fig. 6,b, arrowhead). In this field, the adjacent cell with a nonapoptotic chromatin pattern (Fig. 6,a, arrow) had normal mitochondrial staining of the caspase-3 precursor (Fig. 6 b, arrow). Interestingly, unlike cells in which apoptosis was induced by UV irradiation, mitochondria in staurosporine-treated apoptotic cells that lacked caspase-3 staining also failed to label with MitoTracker, indicating a markedly reduced mitochondrial transmembrane potential, possibly due to permeability transition.

Since antibody R280 recognizes only the caspase-3 precursor, decreased R280 staining in apoptotic cells is consistent with activation of caspase-3. To address biochemically whether the caspase-3 activity was present in cell populations containing morphologically apoptotic cells that lacked mitochondrial staining with antibody R280, we measured cleavage of an exogenous peptide substrate (Ac-DEVD-AMC) in keratinocyte or HeLa cell lysates known to support the activity of caspase-3 (Nicholson et al., 1995; Casciola-Rosen et al., 1996). In nonirradiated cells, background levels of Ac-DEVD-AMC cleavage activity were observed (14 U/mg); these did not increase up to 4 h after irradiation of parallel cultures (Fig. 7,a). Consistent with the presence of numerous apoptotic cells at time points >6 h after UVB irradiation, there was a marked increase in Ac-DEVD-AMC cleavage activity 8 h after irradiation (53 U/mg) (Fig. 7,a). The sensitivity of this activity to aldehyde inhibitors (Ac-DEVD-CHO but not Ac-YVAD-CHO) confirms that the activity is due to caspase-3 or a closely related homologue (data not shown). Similarly, cleavage of PARP, an endogenous macromolecular caspase-3 substrate, was only observed in cells incubated >6 h after UV irradiation (data not shown, and Fig. 7 b, lanes 1–6), providing additional biochemical evidence for the coincident presence at times >6 h after UVB irradiation of caspase-3 activity, morphologic apoptosis, and loss of mitochondrial caspase-3 precursor staining.

Activation of caspase-3 is a central event in the apoptotic process, upon which numerous different signaling pathways converge and through which multiple downstream substrates are cleaved. Recent data have strongly implicated the mitochondrion as a central integrating organelle in apoptotic signaling, which has the capacity to directly activate the execution pathways (see below). The studies reported here demonstrate that the caspase-3 precursor is present in both the cytosol and mitochondria in nonapoptotic cells. The existence of two spatially separate pools of the caspase-3 precursor predicts the existence of distinct activation pathways, regulated by different modulators of apoptosis. We propose that the mitochondrial caspase-3 precursor functions principally in Bcl-2–sensitive apoptotic pathways.

While it remains unclear how the caspase-3 precursor reaches its mitochondrial location, it is possible that it might either possess a cryptic targeting sequence (as does cytochrome c) or might interact with other molecules that target it to mitochondria. Identifying any interacting proteins and determining whether the mitochondrial and cytosolic caspase-3 precursors are equivalent in terms of activation mechanism remains a high priority.

Our studies demonstrate that the caspase-3 precursor localizes to mitochondria in healthy cells but is absent from this site in apoptotic cells. This loss of caspase-3 precursor staining in mitochondria is induced in a variety of cell types by diverse apoptotic stimuli including UV irradiation, staurosporine, or C2-ceramide. Caspase-3 activity is only detectable in cell populations containing morphologically apoptotic cells, indicating that the loss of mitochondrial caspase-3 staining and caspase-3 activation are temporally linked. The inability of R280 to recognize the mature form of caspase-3, as well as the asynchronous nature of apoptosis in cell populations, prevents us from distinguishing whether caspase-3 activation occurs in mitochondria, or whether the caspase-3 precursor and cytochrome c are redistributed from mitochondria before activation.

Interestingly, the mitochondrial pool of caspase-3 colocalizes with cytochrome c, which has also been shown to leak from mitochondria into the cytosol early during apoptosis (before alteration of mitochondrial transmembrane potential). Accumulating evidence suggests that the release of cytochrome c from mitochondria plays a critical role in initiating the apoptotic process: (a) The anti-apoptotic proteins Bcl-2 or Bcl-X_L (which localize to the outer mitochondrial membrane) prevent cytochrome c release from mitochondria in vitro and in vivo and prevent activation of caspase-3 (Kluck et al., 1997; Vander Heiden et al., 1997; Yang et al., 1997). (b) Cytochrome c, in conjunction with two protein cofactors (Apaf-1 and caspase-9/Apaf-3) and dATP/ATP, stimulate the conversion of the caspase-3 precursor to its active form in cell-free systems (Li et al., 1997; Zou et al., 1997). These studies strongly support the hypothesis that spatial separation of interacting proteins is key to modulation of the apoptotic process. This hypothesis predicts that cofactors that participate in cytochrome c–mediated processing of caspase-3 (e.g., Apaf-1 and caspase-9) are segregated from the caspase-3 precursor and cytochrome c in healthy cells, and that proapoptotic signals permit the efficient formation of an activation complex, with ensuing apoptosis.

The release of cytochrome c from mitochondria during apoptosis precedes mitochondrial membrane depolarization (Kluck et al., 1997; Yang et al., 1997), whereas release of AIF requires permeability transition (Susin et al., 1996). These differences might reflect the existence of multiple mechanisms through which mitochondria can influence cell death pathways. The observations that mitochondria in keratinocytes in which apoptosis is induced by UV irradiation still label with MitoTracker but fail to label when apoptosis is induced by staurosporine suggest that these different pathways might be operating to different extents in these settings. Whether active caspase-3 can affect mitochondrial membrane potential remains unknown but is of significant interest.

In conclusion, the coenrichment of caspase-3 at a site containing several molecules that regulate its activation suggests a mechanism whereby the apoptotic stimulus emanating from mitochondria is efficiently transduced. While all apoptotic stimuli appear to converge on a common set of “executioner” proteases, the upstream mechanisms that activate this final common pathway appear to vary for different stimuli. We propose that the mitochondrial subpopulation of caspase-3 precursor molecules is spatially and biochemically distinct and is coupled to a unique subset of apoptotic signaling pathways, which are Bcl-2 sensitive. These apoptotic signals are transduced by bringing together locally enriched but spatially separated molecules (e.g., cytochrome c, caspase-3 precursor, Apaf-1, caspase-9), thus permitting the formation of a multiprotein complex that catalyzes efficient activation of caspase-3. Alternatively, the mitochondrial caspase-3 precursor may have a unique upstream function in some forms of apoptosis, perhaps becoming activated in mitochondria before alteration of mitochondrial physiology. Determining the effects of Bcl-2 family members on the function of this mitochondrial caspase-3 pool is a major priority.

We thank Ann Hubbard for helpful discussions, Greg Martin for assistance with immuno-EM, and Chip Montrose (all three from Johns Hopkins University) for assistance with confocal microscopy.

This work was supported by National Institutes of Health grants 5T32-AI07247 and K12DK01298, a Career Development Award form the Dermatology Foundation/Lever Brothers Co. (LCR), AR44684 (LCR), the Peggy Meyerhoff-Pearlstone Foundation, the Howard Bank Memorial Fund, and an Arthritis Foundation (Maryland Chapter) Institutional Grant. A. Rosen is a Pew Scholar in the Biomedical Sciences.

AIF

apoptosis-inducing factor

CHAPS

3-([3-cholamidopropyl]-dimethyllammonio)-1-propanesulfonate

DAPI

4′,6-diamidino-2-phenylindole

DFF

DNA fragmentation factor

ECL

enhanced chemiluminescence

HUVEC

human umbilical vein endothelial cells

KGM

keratinocyte growth medium

KRB

Krebs' Ringers buffer

PARP

poly (ADP-ribose) polymerase