An APAF-1·Cytochrome c Multimeric Complex Is a Functional Apoptosome That Activates Procaspase-9*

We report here the reconstitution of the de novo procaspase-9 activation pathway using highly purified cytochrome c, recombinant APAF-1, and recombinant procaspase-9. APAF-1 binds and hydrolyzes ATP or dATP to ADP or dADP, respectively. The hydrolysis of ATP/dATP and the binding of cytochromec promote APAF-1 oligomerization, forming a large multimeric APAF-1·cytochrome c complex. Such a complex can be isolated using gel filtration chromatography and is by itself sufficient to recruit and activate procaspase-9. The stoichiometric ratio of procaspase-9 to APAF-1 is approximately 1 to 1 in the complex. Once activated, caspase-9 disassociates from the complex and becomes available to cleave and activate downstream caspases such as caspase-3.

show excessive number of neurons in their brain, defects in facial features, and delayed recession of interdigital webbing, owing to a defect in apoptosis (15)(16). Cells derived from these animals show resistance to a variety of apoptotic stimuli such as chemotherapeutic agents, UV and ␥-irradiation, ceramide, and dexamethasone (15)(16).
Details of the biochemical mechanism of procaspase-9 activation by APAF-1 is sketchy. Previous studies have shown that APAF-1 forms a complex with caspase-9 in the presence of dATP and cytochrome c, two co-factors for APAF-1 function (12). APAF-1 protein truncated at the COOH-terminal WD-40 repeats is constitutively active in vitro, independent of cytochrome c and dATP (17)(18). In addition, truncated APAF-1 interacts with itself as detected by a co-immunoprecipitation and a yeast two-hybrid experiment, suggesting that oligomerization might be an important step for its function (17)(18). Similar experiments with CED-4 protein suggest that the oligomerization of CED-4 is also important for activating CED-3, the C. elegans caspase (19). However, there has been no direct evidence either for APAF-1 oligomerization or for the roles of dATP/ATP and cytochrome c in such a process.
In the current report, we reconstituted the procaspase-9 activation pathway using highly purified cytochrome c, recombinant APAF-1, and recombinant procaspase-9. By using such a system, we are able to analyze the role of dATP/ATP hydrolysis, cytochrome c binding to APAF-1, and APAF-1 oligomerization in the de novo activation of procaspase-9. The results demonstrate a multi-stage reaction with the formation of a multi-subunit APAF-1⅐cytochrome c complex as the key commitment step.

EXPERIMENTAL PROCEDURES
General Methods and Materials-We obtained dATP and other nucleotides and radioactive materials from Amersham Pharmacia Biotech, and molecular weight standards for SDS-PAGE and gel filtration chromatography were from Bio-Rad. General molecular biology methods were used as described in Sambrook et al. (20).

Construction of Alternatively Spliced Transcripts of APAF-1 in a Baculovirus Expression Vector and Site-directed Mutagenesis of APAF-1-
The [␣-32 P]dCTP-labeled 285-bp PCR product was used to screen the HeLa Exlox cDNA library as described (13). A 1.4-kilobase pair clone was characterized by DNA sequencing and assembled with APAF-1 cDNA using DNASTAR program. An additional 33 nucleotides encoding 11 more amino acids between the CARD and CED-4 homologous domain of APAF-1 was found. To confirm the existence of this alternatively spliced transcript, HeLa poly(A) ϩ mRNA was purified using a rapid mRNA purification kit (Amersham Pharmacia Biotech), and the first strand cDNA was carried out using a first strand cDNA synthesis kit (Amersham Pharmacia Biotech) with the specific primer (5ЈAA-CACTTCACTATCACTTCC3Ј), designed from WD-40 region of APAF-1 cDNA. An aliquot of 400 ng of this first strand cDNA mixture was amplified by two primers P1/P2 (5ЈTAATGATTCCTACGTATCATTC-TACAATGC3Ј and 5ЈGAATGATCTCTAACAGCTTC3Ј) (Fig. 1). The resulting 316-and 283-bp PCR products were subcloned into the PCR II vector using the TA cloning kit (Invitrogen) and sequenced.
A 3.63-kilobase pair cDNA encoding the full-length APAF-1 fused * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 1 The abbreviations used are: APAF-1, apoptotic protease activating factor-1; ATP, adenosine 5Ј-triphosphate; dATP, 2Ј-deoxyadenosine 5Јtriphosphate; PAGE, polyacrylamide gel electrophoresis; RT, reverse transcriptase; PCR, polymerase chain reaction; CARD, caspase recruitment domain; ATP␥S, adenosine 5Ј-O-(thiotriphosphate). with a 9-histidine tag at the COOH terminus was subcloned into NotI/ KpnI sites of the baculovirus expressing vector pFastBacI (Life Technologies, Inc.). To insert APAF-1L into the APAF-1-pFastBacI vector, a 425-bp 5Ј region of APAF-1 cDNA was removed by enzymatic digestion with NotI and BstEII from APAF-1-pFastBacI construct and replaced with a 458-bp DNA fragment excised out of the same enzyme digestion from the 5Ј part of APAF-1L including the 33 nucleotides.
To test if APAF-1 has another alternative transcript that encodes an additional WD-40 motif, the first strand cDNA was carried using a first strand cDNA synthesis kit (Amersham Pharmacia Biotech) with the specific primer (5ЈTTTCACTGTTTCCTGATGGC3Ј) designed from 3Ј region of APAF-1 cDNA. An aliquot of 200 ng of this first strand cDNA mixture was then PCR amplified by two primers P3/P4 (5ЈATGCGA-CATCAGCAAATGAG 3Ј and 5ЈACCTTTGAACGTGAGTCTG 3Ј). Both PCR products (187 and 316 bp) were subcloned into the PCRII vector using the TA cloning kit (Invitrogen) and the sequences confirmed by DNA sequencing. This additional WD-40 motif was subcloned in frame into the full-length APAF-1L by using PCR-SOEing method (21). The full-length APAF-1L-WD13 with 9-His tag was subcloned into pFastBac I vector at NotI and KpnI sites. The nucleotide-binding site mutations (G159E and K160T in Walker's A Box, D243A and D244A in Walker's B Box) were generated by a PCR-SOEing method. The mutants were confirmed by DNA sequencing.
Caspase-9 cleavage site (D315A) and active site (C287A) mutants were also generated by the PCR-SOEing method and confirmed by DNA sequencing. Wild type caspase-9 with or without a 9-histidine tag at COOH termini and both the cleavage site and active site mutant procaspase-9 fused with 9-His tag at the COOH termini were subcloned into pFastBacI vector at BamHI and EcoRI sites.
Production of Recombinant APAF-1 and Proaspase-9 Proteins in a Baculovirus Expression System-The expression plasmids were transformed into DH10Bac Escherichia coli cells (Life Technologies, Inc.). The recombinant viral DNA, bacmids, were purified according to the Bac-To-Bac Baculovirus Expression procedure (Life Technologies, Inc.) and confirmed by PCR amplification analysis. The DNA was then used to transfect the insect cells, Sf21 using CellFECTIN reagent (Life Technologies, Inc.). The cells were grown in IPL41 medium supplemented with 10% fetal calf serum, 2.6 g/liter tryptose phosphate, 4 g/liter yeastolate, and 0.1% Pluronic F-68 plus penicillin (100 units/ml), streptomycin (100 g/ml), and fungizone (0.25 g/ml). The expression of recombinant protein was analyzed by Western blot. The virus stocks were amplified to 100 ml and used to infect 1 liter of Sf21 cells at a density of 2 ϫ 10 6 cells/ml. The infected cells were harvested after 40 h for APAF-1 and 22 h for procaspase-9 by centrifugation and resuspended in 5 volumes of buffer A (20 mM Hepes-KOH, pH 7.5, 10 mM KCl, 1.5 mM MgCl 2 , 1 mM NaEDTA, 1 mM NaEGTA, 1 mM dithiothreitol, and 0.1 mM PMFS). The resuspended cells were lysed in buffer A by homogenization. After centrifugation at 10,000 ϫ g for 1 h, the supernatant was loaded onto a 3-ml nickel affinity column. The column was washed with 300 ml of buffer A containing 1 M NaCl and 20 mM imidazole for APAF-1 or buffer A containing 1 M NaCl and 15 mM imidazole for procaspase-9. After equilibrating the column with 20 ml of buffer A, the column was eluted with buffer A containing 250 mM imidazole. The eluted APAF-1 protein were stored in multiple aliquots at Ϫ80°C.
The eluted wild type procaspase-9 was further purified by diluting the nickel column eluate with 10 volumes of buffer A and loaded onto a Mono Q column (Amersham Pharmacia Biotech) equilibrated with buffer A containing 50 mM NaCl. The column was eluted with 20-ml linear gradient from 50 mM NaCl to 300 mM NaCl in buffer A. The procaspase-9 protein was eluted at 150 mM NaCl in buffer A. After dialyzing against buffer A, the protein was stored in multiple aliquots in Ϫ80°C.
The non-His-tagged procaspase-9 was expressed as described above and purified according to a procedure described in Ref. 12.
Western Blot Analysis-CED-4 homologous domain of APAF-1 was PCR-amplified and subcloned in frame into the NdeI-XhoI sites of the bacterial expression vector pET-15b (Novagen). The fusion protein was purified as described (13) and was used to immunize rabbits to generate a polyclonal antibody. Procaspase-9 (C287A) mutant was subcloned in frame into BamHI-XhoI sites of pET-15b vector. The fusion protein was generated as described previously (12), and a polyclonal anti-caspase-9 antibody was generated by immunizing rabbits with this protein. Immunoblot analysis was performed with a horseradish peroxidase-conjugated goat anti-rabbit (APAF-1, caspase-9) or goat anti-mouse (cytochrome c) immunoglobulin G using enhanced chemiluminescence Western blotting detection reagents (Amersham Pharmacia Biotech).
Assay for Caspase-9 and Caspase-3 Activation-Aliquots of 0.5 l (0.2 g) of His-tagged recombinant procaspase-9 and 5 l (0.8 g) of recom-binant APAF-1 were incubated in the presence or absence of 10 ng/ml cytochrome c and 0.1 mM dATP and 1 mM additional MgCl 2 at 30°C for 1 h in a final volume of 20 l of buffer A (13). After incubation, the samples were subjected to a 15% SDS-PAGE and transferred to a nitrocellulose filter, which was blotted with anti-caspase-9 antibody. Procaspase-3 was translated and purified as described (12). A 1-l aliquot of in vitro translated caspase-3 was incubated with the mixture of procaspase-9 activation reaction as described above. The samples were subjected to a 15% SDS-PAGE, and then the gel was transferred to a nitrocellulose filter, which was subsequently exposed to a phosphorimaging plate and visualized in a Fuji BAS-1500 PhosphorImager.
Assay for ATP Hydrolysis-Aliquots of 2 g of wild type or nucleotide-binding sites mutant APAF-1 were incubated with 100 nM ATP plus 2 Ci of [␣-32 P]ATP (Amersham Pharmacia Biotech) or 100 nM dATP plus 2 Ci of [␣-32 P]dATP (Amersham Pharmacia Biotech) and additional 0.5 mM MgCl 2 in the presence or absence of 0.4 g of cytochrome c at 30°C for 1 h in a final volume of 50 l of buffer A. After incubation, the reaction mixtures were incubated with nickel beads (Qiagen), and the bound APAF-1s were eluted as described in the legend of Fig. 2. The eluted samples were directly loaded on a TLC plate, and ATP/dATP hydrolysis was analysis as described in the legend of Fig. 2.
Analysis of APAF-1⅐Cytochrome c⅐Caspase-9 Complexes-HeLa cells were set at 2ϫ 10 6 per 150-mm dish in medium A (Dulbecco's modified Eagle's medium containing 100 units/ml penicillin and 100 g/ml streptomycin sulfate) supplemented with 10% (v/v) fetal calf serum and grown in a monolayer at 37°C in an atmosphere of 6 -7% CO 2 . After incubation for 24 h, two of the four dishes were added with staurosporine to a final concentration of 2 M. After treatment for 8 h, both treated and non-treated cells were harvested, and the S-100 fractions were prepared as described in Liu et al. (22). An aliquot of 50 l of the S-100 fraction (0.1 mg) was directly loaded onto a Smart System Superdex-200 gel filtration column (2 ml) (Amersham Pharmacia Biotech) equilibrated with buffer A and eluted with the same buffer. Fractions of 100 l were collected. Aliquots of 30 l of the fractions were directly subjected to 10% SDS-PAGE followed by Western blotting analysis using a rabbit antibody against APAF-1.

RESULTS
Multiple Spliced Forms of APAF-1-APAF-1 was originally described as a 130-kDa protein consisting of 1194 amino acids (13). During the course of cDNA cloning, several alternatively spliced forms were also detected. One alternative splicing event uses a different splicing donor site at exon 3 to insert an additional 11 amino acids in between the CARD and the CED-4 homologous domain of APAF-1 (Fig. 1A). This spliced form was designated as APAF-1L. Another alternative splicing event produces an additional exon between exon 17 and 18 encoding 43 amino acids, creating one more WD-40 repeat after the fifth WD-40 repeat. We designated this spliced form as APAF-1-WD13. To confirm that these alternatively spliced APAF-1s were indeed expressed in cells, we performed RT-PCR using mRNA isolated from HeLa cells as templates. After first strand cDNA synthesis, the templates were amplified by two sets of primers flanking these two alternative spliced regions (Fig.  1A). As shown in Fig. 1B, each set of primers yielded two PCR products with the predicted sizes corresponding to the two alternatively spliced messages. The identity of all four PCR products was subsequently confirmed by direct DNA sequencing. The WD13 spliced form of APAF-1 has been reported (GenBank™ accession number AB007873) and exists in the mouse version of APAF-1 cDNA (15).
All three spliced forms of APAF-1 were expressed in Sf21 insect cells using a baculovirus expression vector that fused APAF-1 to a 9-histidine tag at the COOH termini. Each recombinant protein was purified through a nickel affinity column and subjected to SDS-PAGE followed by Coomassie Blue staining. As shown in Fig. 2A, each recombinant protein has an estimated size between 130 and 140 kDa with APAF-1L-WD13 being slightly larger than the other two. The purified proteins were tested for procaspase-9 activation in the presence of dATP and cytochrome c. We found that among the proteins encoded by the three alternatively spliced mRNAs, APAF-1L-WD13 has the most stable cytochrome c and dATP-dependent caspase-3 activating activity (data not shown). Thus, we used this recom-binant protein in all of the following experiments and refer to it as APAF-1.

Reconstitution of Caspase-3 Activation with Recombinant
FIG. 1. Alternatively spliced transcripts of APAF-1. A, a diagram of APAF-1L-WD13. The CARD, CED-4 homologous region, Walker's A and B box consensus sequences for nucleotide-binding sites (31), and WD-40 repeats are indicated. The additional 11 amino acids in between the CARD and CED-4 homologous domain and an additional WD-40 repeat in between the fifth and sixth WD-40 repeat are indicated. P1/P2 and P3/P4 denote the PCR primers used for RT-PCR to detect the alternatively spliced APAF-1 transcripts. B, the RT-PCR products using mRNA from HeLa cells as templates and P1/P2 and P3/P4 as primers were analyzed by a 2% agarose gel electrophoresis and visualized by ethidium bromide staining.

FIG. 2. Purified recombinant APAF-1 and procaspase-9 have dATP and cytochrome c-dependent caspase-3 cleavage activity.
Recombinant APAF-1, APAF-1L, APAF-1L-WD13, and procaspase-9 were prepared as described under "Experimental Procedures." A, aliquots (2.5 g) of APAF-1, APAF-1L, and APAF-1L-WD13 were subjected to 10% SDS-PAGE, and the gel was subsequently stained with Coomassie Blue. B, aliquots (2 g) of wild type procasepase-9, C287A active site mutant, and D315A cleavage site mutant were subjected to 15% SDS-PAGE followed by staining with Coomassie Blue. C, aliquots of recombinant procaspase-9 (2 g) with wild type sequence (lanes 1-3), or cleavage site mutant (D315A, lanes 5 and 6), or active site mutant (C287A, lanes 4 and 6), and recombinant APAF-1L-WD13 (0.8 g) were incubated individually (lane 1) or together (lanes 2-6), in the presence (lanes 3-6) or absence (lanes 1 and 2) of 0.2 g of cytochrome c plus 100 M dATP at 30°C for 1 h in a final volume of 20 l of buffer A. The samples were subjected to 15% SDS-PAGE and transferred to a nitrocellulose filter. The filter was probed with a rabbit antibody against caspase-9. The antigen-antibody complexes were visualized by an ECL method as described under "Experimental Procedures." The filter was then exposed to a x-ray film for 10 s. D, aliquots of recombinant procaspase-9 (2 g) and recombinant APAF-1L-WD13 (0.8 g) were incubated individually (lanes 1 and 2) or together (lanes 3-7), in the presence (lanes 1 and 2, 4, 6, and 7) or absence (lanes 3 and 5) of 0.2 g of cytochrome c plus 100 M dATP (lanes 1 and 2 and 5 and 6) or 100 M dADP (lane 7) with 1 l of 35 S-labeled affinity purified procaspase-3 at 30°C for 1 h in a final volume of 20 l of buffer A. The samples were then subjected to 15% SDS-PAGE and transferred to a nitrocellulose filter, followed by exposing to a phosphorimaging plate for 12 h at room temperature.
APAF-1 and Procaspase-9 -To study the biochemical mechanism of caspase activation, we established the caspase-3 activation reaction using purified, recombinant APAF-1 and procaspase-9. It has been a technical challenge to generate recombinant procaspases in large enough quantities for biochemical analysis since overexpression of procaspases often results in their auto-activation. In order to overcome this difficulty, we infected SF21 cells with a baculoviral vector containing the cDNA of human caspase-9 or human caspase-9 fused with a 9-histidine tag. The cells were grown in the presence of 10% bovine calf serum and were harvested 24 h after infection before auto-processing occurs. The procaspase-9 was purified either by a nickel affinity column followed by a Mono Q column or by conventional chromatography described previously (12). As shown in Fig. 2B, purified wild type procaspase-9 (1st lane), or procaspase-9 with a cysteine to alanine substitution at its active site (2nd lane), or aspartic acid to alanine substitution at its cleavage site (3rd lane) migrated as a ϳ50-kDa polypeptide band on an SDS gel.
The recombinant APAF-1 and procasepase-9 were then tested for their respective functions by incubating them with or without dATP and cytochrome c. As shown in Fig. 2C, wild type procaspase-9 was processed to the 35-and 10-kDa active form only when incubated with APAF-1, dATP, and cytochrome c (lane 3). No processing was observed when cytochrome c or APAF-1 was omitted (lanes 1 and 2). In contrast, the cleavage site mutant D315A, or the active site mutant C287A, could not be processed even in the presence of APAF-1, cytochrome c, and dATP (lanes 4 and 5). This experiment indicated that the processing of procaspase-9 by APAF-1, dATP, and cytochrome c occurs after Asp-315 as previously demonstrated (17), and the activation is through auto-catalysis since the active site mutant failed to be processed. Interestingly, that mixing D315A and C287A mutants did not result in any cleavage, suggesting inter-molecular cleavage does not happen (lane 6).
To confirm that recombinant APAF-1-catalyzed procaspase-9 processing is functional, we incubated the procaspase-9 activation reaction with 35 S-labeled, affinity purified procaspase-3, a caspase downstream of caspase-9 (12,(22)(23). As shown in dATP/ATP Binding and Hydrolysis by APAF-1-APAF-1 contains in its CED-4 homologous domain consensus Walker A and B boxes for nucleotide binding (13). The nucleotide-binding sites are conserved between CED-4 and APAF-1, suggesting the importance of nucleotide binding in their function (13,(25)(26). To confirm that APAF-1 indeed binds ATP or dATP, purified APAF-1 protein was incubated with ␣-32 P-labeled ATP or dATP, and the nucleotide bound to APAF-1 was pelleted by nickel affinity resin and analyzed by thin layer chromatography (TLC). As shown in Fig. 3A, radiolabeled nucleotides were co-precipitated with APAF-1, indicating that APAF-1 indeed bound nucleotides. Interestingly, the majority of nucleotide bound to APAF-1 was ADP or dADP, rather than ATP, or dATP. This finding suggests that the nucleotides bound to APAF-1 are hydrolyzed. Incubation with cytochrome c did not affect dATP binding or hydrolysis (lanes 6 and 7). A mutant APAF-1 protein with four amino acids substitution in the conserved Walkers A and B boxes did not bind any nucleotide (lanes 4 and 6) even though the same amount of protein was precipitated (Fig. 3B).
The above data suggest that the binding and hydrolysis of ATP/dATP to ADP/dADP are important for APAF-1 function. To confirm this hypothesis, increasing amounts of a non-hydrolyzable ATP analog, ATP␥S, was incubated with APAF-1 and procaspase-9, in the presence of dATP (Fig. 3C, lanes 1-4) or ATP (Fig. 3C, lanes 5-8) in a procaspase-3 activation reaction. Consistent with the previous observation, 100 M dATP, or 1 mM ATP, is the optimal concentration in such a reaction. The presence of ATP␥S inhibited the caspase-3 activation reaction in a concentration-dependent fashion with 1 mM ATP␥S completely inhibiting the reaction. It is interesting to note that even at concentrations much lower than ATP or dATP present, ATP␥S shows noticeable inhibitory effects (Fig. 3C, lanes 2, 3,  6, and 7).  1 and 2), aliquots of [␣-32 P]ATP or [␣-32 P]dATP in 50 l of buffer A were incubated at 30°C for 1 h, followed by directly loading the TLC plate together with the above samples. B, aliquots of 5 l of resulting supernatants as in A were subjected to 10% SDS-PAGE followed by Western blotting analysis using a rabbit antibody against the CED-4 homologous domain of APAF-1 (1:2000). The antigen-antibody complexes were visualized by an ECL method as described under "Experimental Procedures." The film was exposed for 10 s. C, aliquots of 0.8 g of recombinant APAF-1L-WD13 were incubated with aliquots of 0.2 g of recombinant procaspase-9 purified as described under "Experimental Procedures," 1 l of in vitro translated 35 S-labeled caspase-3, 0.2 g of cytochrome c, 100 M dATP (lanes 1-4) or 1 mM ATP (lanes 5-8), and increasing amounts of ATP␥s in a final volume of 20 l in buffer A. After 1 h incubation at 30°C, the samples were subjected to 15% SDS-PAGE and transferred to a nitrocellulose filter followed by exposing to a phosphorimaging plate for 12 h at room temperature.
Interactions Between Caspase-9, APAF-1, and Cytochrome c-Previous studies have demonstrated that APAF-1 interacts with caspase-9 only in the presence of dATP/ATP and cytochrome c (12). To study further the interaction between these three proteins, we incubated purified APAF-1 (His-tagged), procaspase-9 (without tag), and cytochrome c in different combinations, and we precipitated APAF-1 with nickel resin. The supernatants and pellets were probed with antibodies against these three proteins. Shown in Fig. 4C, the same amounts of APAF-1 were precipitated by the nickel resin independent of cytochrome c, dATP, or procaspase-9. No APAF-1 was detected in the supernatants (data not shown). On the other hand, caspase-9 was pelleted only in the presence of APAF-1, dATP, and cytochrome c (Fig. 4B, lane 6). Interestingly, the majority of pelleted caspase-9 was already processed into the 35-kDa active form, indicating that procaspase-9 was rapidly autoactivated once bound to APAF-1. When dATP, APAF-1, or cytochrome c was omitted in the reaction, procaspase-9 stayed in the supernatant (Fig. 4A, lanes 5 and 7-9). Noticeably, some of the processed caspase-9 was also in the supernatant (Fig. 4A,  lane 6), suggesting that once processed, caspase-9 can be released from APAF-1. Consistent with the previous finding, cytochrome c was co-precipitated with APAF-1 in the absence and presence of dATP (13) (Fig. 4D). However, the amount of cytochrome c that was co-precipitated with APAF-1 was consistently increased in the presence of dATP, suggesting that dATP may stabilize the binding of cytochrome c to APAF-1.
Formation of APAF-1⅐Cytochrome c⅐Caspase-9 Complex in the Presence of dATP-The co-precipitation experiments with either nickel resin (Fig. 4) or antibody against caspase-9 (12) indicate that caspase-9 interacts with APAF-1 in the presence of cytochrome c and dATP. To characterize further the biochemical nature of this APAF-1⅐caspase-9 complex formed in the presence of cytochrome c and dATP, we separated these proteins and the complex on a gel filtration column. In the absence of dATP, the mixture of APAF-1, procaspase-9, and cytochrome c were separated by the sizing column (Fig. 5A). The protein peaks of APAF-1, procaspase-9, and cytochrome c were found at fractions 14, 16, and 21, respectively. Each protein migrates at the position corresponding to its mono- FIG. 4. Interactions between procaspase-9, APAF-1, and cytochrome c. Recombinant APAF-1 and procaspase-9 (not tagged) were purified as described under "Experimental Procedures." Aliquots of 4 g of APAF-1, 0.5 g of procaspase-9, and 0.8 g of cytochrome c were incubated individually (lanes 1 and 2) or in combinations of two (lanes 3 and 4 and 7-9) or all three (lanes 6), in the presence or absence of 100 M dATP at 30°C for 1 h in a final volume of 100 l of buffer A. After incubation, the samples were mixed with 400 l of buffer A, 60 l of nickel beads and incubated at 4°C for 3 h in a rotator. The mixtures were then pelleted by centrifugation, and the supernatants were collected and concentrated to about 40 l. The pellets were washed five times with buffer A and eluted with 250 mM imidazole in 80 l of buffer A. A, 20-l aliquots of the resulting supernatants were subjected to 15% SDS-PAGE and transferred to a nitrocellulose filters. The filters were probed with a rabbit antibody against caspase-9. The antigen-antibody complexes were visualized by an ECL method as described under "Experimental Procedures." The filter was then exposed to a x-ray film for 30 s. B, the resulting pellets were washed five times with buffer A and then eluted with 250 mM imidazole in 80 l of buffer A. Aliquots of 35 l of the eluates were subjected to 15% SDS-PAGE and transferred to a nitrocellulose filter. The filter was probed with a rabbit anti-caspase-9 antibody and then exposed to x-ray film for 1 min. C, the resulting eluates as in B were subjected to 10% SDS-PAGE followed by Western blotting analysis using a rabbit antibody against APAF-1 (1:2000), and the antigen-antibody complexes were visualized by an ECL method. The film was exposed for 5 s. D, the same filter as in B was erased and reprobed with a monoclonal anti-cytochrome c antibody followed by ECL detection and exposure to x-ray film for 1 min. 5. Formation of an APAF-1, caspase-9, and cytochrome c complex. APAF-1 was purified from HeLa-S100 as described previously (13). The histidine-tagged recombinant procaspase-9 was purified as described under "Experimental Procedures." A and B, an aliquot of 2.5 g of APAF-1 was incubated with 15 l (0.5 g) of recombinant procaspase-9 and 0.8 g of cytochrome c in the absence (A) or presence (B) of 100 M dATP in 30°C for 1 h in a final volume of 60 l in buffer A. After incubation, the reaction mixtures were fractionated by a Superdex-200 gel filtration column in a Smart System (Amersham Pharmacia Biotech) as described under "Experimental Procedures." The fractions of 100 l were collected, and aliquots of 30 l were subjected to 15% or 1% SDS-PAGE followed by Western blotting analysis using monoclonal antibody against cytochrome c, a rabbit polyclonal antibody against caspase-9 (1:2000), and a rabbit polyclonal antibody against APAF-1 as indicated. The antigen-antibody complexes were visualized by an ECL method as described under "Experimental Procedures." The filters was exposed to a x-ray film for 30 s for APAF-1 blot and 1 min for caspase-9 and cytochrome c blots. C, four dishes of HeLa cells grown in monolayer as described under "Experimental Procedures." Two dishes of cells were treated with staurosporine at a final concentration of 2 M for 8 h. S-100 fractions were prepared from both treated and nontreated cells as described under "Experimental Procedures" and fractionated by the Superdex-200 gel filtration column as described in A and B. Aliquots of 30-l column fractions were subjected to 10% SDS-PAGE followed by Western blotting analysis using a rabbit antibody against APAF-1 (1:2000). The filters was exposed to an x-ray film for 30 s. meric size compared with the molecular size standard. No stable association was observed between these proteins. All three proteins migrated at the identical positions if run individually (data not shown). In contrast, if the mixture was preincubated in the presence of dATP, all three proteins were observed at fraction 11, indicating that they now form a large complex (Fig. 5B). Caspase-9 was processed under this condition and showed two peaks in the column. The first peak was at fraction 11 correlating with the complex form, whereas the second peak was at fraction 16, indicating again that the processed caspase-9 was released from the complex.

FIG.
To confirm that this large complex (Ͼ1.3 million-dalton exclusion volume of this column) forms in vivo when cells undergo apoptosis, we prepared cell extracts from normally growing HeLa cells and cells that were undergoing apoptosis induced by staurosporine. The apoptosis in staurosporine-treated cells was confirmed by characteristic morphological changes observed under a microscope (data not shown). These extracts were analyzed by the same gel filtration column. Shown in Fig. 5C, APAF-1 in normal cell extract migrated at peak fraction 14, corresponding to its monomeric inactive form, whereas a significant amount of APAF-1 in apoptotic extracts had been shifted to fraction 11, corresponding to a large complex.
Both dATP and Cytochrome c Are Required to Form a Caspase-9-activating APAF-1⅐Cytochrome c Complex-To dissect further the individual steps leading to the formation of this multimeric protein complex, recombinant APAF-1 protein was analyzed on the same gel filtration column either alone (Fig.  6A), or preincubated with cytochrome c plus ATP␥S (Fig. 6B), or dATP (Fig. 6C), or dATP plus cytochrome c (Fig. 6D). Preincubation of APAF-1 with cytochrome c plus nonhydrolyzable ATP or dATP alone did not cause significant change in their column behavior. In addition, after preincubating APAF-1 with dATP, the reaction mixture was loaded on the same gel filtration column, and the resulting column fractions were assayed for APAF-1 activity in the absence or presence of dATP. The APAF-1 activity was only observed in the presence of dATP and the activity peak was observed at fraction 14, correlating with the monomeric form of APAF-1 (Fig. 6C, lower panel). The APAF-1 protein that migrated larger than fraction 14 was not active and therefore most likely represents nonspecific protein aggregates. Indeed, recombinant APAF-1 alone was found to have a tendency to self-aggregate into non-functional complexes (data not shown). These data suggest that incubation of APAF-1 with dATP, or cytochrome c, was not sufficient to drive APAF-1 into the functional multimeric complex nor forgo the requirement for dATP in activating procaspase-9 (Fig. 6C,  lower panel).
Strikingly, when APAF-1 was preincubated with dATP and cytochrome c, the peak of APAF-1 was shifted around fraction 11 and associated with cytochrome c (Fig. 6D). This APAF-FIG. 6. Both dATP and cytochrome c are required for APAF-1 multimerization. A, an aliquot of 4 g of recombinant APAF-1 was fractionated by a Superdex-200 gel filtration column in a Smart System (Amersham Pharmacia Biotech) as described under "Experimental Procedures." Aliquots of 10-l column fractions were subjected to 10% SDS-PAGE followed by Western blotting analysis using a rabbit antibody against APAF-1 (1:2000), and the antibody-antigen complexes were visualized by an ECL method as described under "Experimental Procedures." The film was exposed for 10 s. The positions of molecular weight standards (Bio-Rad) migrated in this column were marked at the top. B, an aliquot of 4 g of recombinant APAF-1 was incubated with 0.8 g of cytochrome c and 1 mM ATP␥s at 30°C for 1 h in a final volume of 60 l in buffer A. The reaction mixture was analyzed by the gel filtration column as in A. Aliquots of 10-l column fractions were subjected to 10% SDS-PAGE followed by Western blotting analysis using a rabbit antibody against APAF-1 and then followed by ECL detection and exposure to x-ray film for 10 s. C, an aliquot of 4 g of recombinant APAF-1 was incubated with 100 M dATP at 30°C for 1 h in a final volume of 60 l in buffer A. The reaction mixture was analyzed in the gel filtration column as in A. Aliquots of 10-l column fractions were subjected to 10% SDS-PAGE followed by Western blotting analysis as in A. The film was exposed for 10 s (upper panel). In the lower panel, aliquots of 25-l column fractions were incubated with 5 l (0.17 g) of recombinant procaspase-9, 1 l of in vitro translated 35 Slabeled procaspase-3, and 0.2 g of cytochrome c in the absence or presence of 100 M dATP as indicated at 30°C for 1 h in a final volume of 35 l of buffer A. The samples were then subjected to 15% SDS-PAGE and transferred to a nitrocellulose filter. The filter was exposed to a phosphorimaging plate for 14 h at room temperature. D, an aliquot of 4 g of recombinant APAF-1 was incubated with 0.8 g of cytochrome c and 100 M dATP in a final volume of 60 l in buffer A. After incubated in 30°C for 1 h, the sample was analyzed in the gel filtration column as in A. Aliquots of 10-l column fractions were subjected to 10% SDS-PAGE followed by Western blotting analysis using a rabbit antibody against APAF-1, or a monoclonal antibody against cytochrome c followed by ECL detection and exposure to x-ray film for 10 and 30 s, respectively (upper panel). In the lower panel, aliquots of 25-l column fractions were incubated with 5 l (0.17 g) of recombinant procaspase-9 and 1 l of in vitro translated 35 S-labeled procaspase-3 in the presence or absence of 100 M dATP and 0.2 g of cytochrome c at 30°C for 1 h in a final volume of 35 l of buffer A. The samples were then subjected to 15% SDS-PAGE and transferred to a nitrocellulose filter. The filter was exposed to a phosphorimaging plate for 14 h at room temperature. E, recombinant cleavage site mutant (D315A) procaspase-9 was purified as described under "Experimental Procedures." The APAF-1 and procaspase-9 (Asp-315) standards were quantified by amino acid analysis. An aliquot of 150 l (25 g) of recombinant APAF-1 was incubated with aliquots of 180 l (45 g) of recombinant procaspase-9 (D315A), 4.8 g of cytochrome c, and 100 M dATP in the final volume of 360 l in buffer A. After 1 h incubation in 30°C, the sample was then fractionated in the gel filtration column as in A. The fraction 10 and fraction 11 were concentrated to a final volume of 30 l and subjected to 10% SDS-PAGE. The gel was subsequently stained with Coomassie Blue and scanned in a densitometer.
1⅐cytochrome c complex was now fully functional in activating procaspase-3 when purified procaspase-9 was added (Fig. 6D,  lower panel). dATP and cytochrome c were no longer required for such a reaction.
APAF-1 and Procaspase-9 Are At a 1:1 Ratio in the Complex-To estimate the molar ratio of APAF-1 and procaspase-9 in the complex, we incubated APAF-1 with an excessive amount of procaspase-9 with the D315A mutation (ϳ5 to 1 molar ratio). This mutant cannot be processed within this complex and therefore cannot be released. The APAF-1⅐caspase-9 D315A complex was isolated from the fraction 10 -11 of the gel filtration column after preincubating with dATP and cytochrome c and subjected to SDS-PAGE together with the known amounts of APAF-1 and procaspase-9 followed by Coomassie Blue staining (Fig. 6E). No cytochrome c was detected by this staining method even though it can be detected by Western blot analysis (data not shown). The bands corresponding to APAF-1 and procaspase-9 were scanned by a densitometer. The results suggest that APAF-1 and procaspase-9 are present at an approximately 1:1 ratio in this complex. DISCUSSION The reconstitution of the procaspase-9 activation pathway with highly purified, recombinant APAF-1 and procaspase-9 described above revealed that caspase-9 activation was achieved by a three-step reaction as illustrated in Fig. 7. First, dATP/ATP binds to APAF-1 through its consensus nucleotide binding domain and is hydrolyzed to dADP or ADP, respectively; second, cytochrome c binds to APAF-1 and promotes the multimerization of APAF-1⅐cytochrome c complex when the dATP/ATP bound to APAF-1 is being hydrolyzed; third, once the multimeric complex is formed, procaspase-9 is recruited to the complex in a 1:1 molar ratio to APAF-1, and it becomes activated through auto-catalysis. The activated caspase-9 is then released from the complex, allowing it to cleave downstream caspases and new procaspase-9 to come into the complex to be processed.
Role of ATP/dATP Hydrolysis-The direct demonstration that APAF-1 binds to and hydrolyzes dATP/ATP confirms the notion that the evolutionarily conserved nucleotide-binding sites are important for its function. Mutations at the conserved nucleotide-binding site diminished the nucleotide binding capacity of APAF-1 and rendered the protein inactive (Fig. 3A, and data not shown). The inhibitory effect of a non-hydrolyzable ATP analog for both caspase activation (Fig. 3C) and ATP binding to APAF-1 (data not shown) indicates that both the binding of ATP/dATP to APAF-1 and their hydrolysis to ADP/ dADP are important for its function. Interestingly, ATP␥S showed an inhibitory effect even at concentrations much lower than the ATP present (Fig. 3C). One interpretation for this observation is that each subunit of APAF-1 must hydrolyze ATP/dATP to form the functional complex. The subunit that bound ATP␥S not only inactivates itself but also effectively inhibits other ATP-bound APAF-1 molecules that were going to form a complex with it. This observation supports the importance of APAF-1 multimerization.
The effects of binding and hydrolysis of ATP/dATP seem to be transient since preincubation of dATP with APAF-1 did not bypass the requirement of dATP (Fig. 6C, lower panel). Since intracellular ATP and dATP concentrations are around 10 mM and 10 M, respectively (27), it is conceivable that ATP/dATP is hydrolyzed by APAF-1 constantly. This hydrolysis, however, does not have any functional consequence since cytochrome c is usually sequestered away from APAF-1 in mitochondria.
Neer et al. (28) compared all the known WD-40 repeat-containing proteins and noted that most are regulatory, but none is an enzyme. ATP/dATP hydrolysis activity by APAF-1 proved it to be an exception.
Role of Cytochrome c in Promoting APAF-1 Multimerization-Holocytochrome c exists exclusively in the intermembrane space of mitochondria in living cells. The newly translated apocytochrome c in cytosol does not have apoptosis promoting activity (29). However, when cells undergo apoptosis in response to a variety of stimuli, cytochrome c is released from mitochondria to cytosol, where APAF-1 is located (30). No detectable amount of APAF-1 is found in the membrane fraction in both living or apoptotic cells (data not shown). Cytochrome c was found to interact with APAF-1 in the absence or presence of dATP as shown by co-immunoprecipitation experiment (13) (Fig. 4D). However, when subjected to gel filtration chromatography, the cytochrome c and APAF-1 complex was FIG. 7. Model of caspase-9 activation. See the text for details. APAF-1, cytochrome c, caspase-9, dATP, and dADP are indicated. only detected in the presence of dATP, indicating that the interaction between APAF-1 and cytochrome c in the absence of dATP is not stable and can be easily separated. The binding and hydrolysis of dATP did not seem to be influenced by the presence of cytochrome c, suggesting that the function of cytochrome c is not to regulate the dATP binding and hydrolysis by APAF-1 (Fig. 3A). In the presence of dATP, cytochrome c promotes the multimerization of APAF-1 from a monomeric form to a large complex that consists at least 8 subunits of APAF-1 (calculated based on its size Ͼ1.3 million daltons). Such a complex is fully functional in activating procaspase-9 and neither dATP nor cytochrome c is required any longer (Fig. 6D,  lower panel). These data are consistent with a model that cytochrome c bound to APAF-1 will stabilize the transient conformational change of APAF-1 resulting from dATP/ATP hydrolysis, allowing them to multimerize. The functions of both dATP and cytochrome c are probably accomplished once the multimerized APAF-1/cytochrome c is formed (Fig. 6D). However, because cytochrome c is still detected in the functional caspase-9-activating complex, it is still possible that cytochrome c plays an additional role in procaspase-9 activation such as stabilizing this complex.
Oligomerized APAF-1⅐Cytochrome c Complex: a Functional Apoptosome?-Several features of this stable APAF-1 and cytochrome c multimeric complex make it a candidate of apoptosome as follows: first, it is a big complex (Ͼ1.3 million daltons measured by gel filtration chromatography); second, the formation of this complex requires the hydrolysis of high energy bond of ATP, or dATP; and third, this complex is functional in term of activating procaspase-9. The formation of this complex will likely be the commitment step for cells to activate their caspases in response to stimuli that cause cytochrome c release. This complex is able to trigger many characteristic apoptotic features of dying cells through caspase activities. Future studies are needed to determine the exact molecular compositions of this complex.
Why multimeric complex? Since the full occupancy of procaspase-9 in the apoptosome results in approximately 1:1 molar ratio of APAF-1 to procaspase-9 (Fig. 6E), one obvious reason for apoptosome is to bring multiple procaspase-9 in close proximity, allowing them to cleave each other. Similar hypothesis have been proposed for CED-4 and APAF-1 WD-40 truncation (17)(18)(19). However, there is no evidence that inter-molecular cleavage indeed occurs, and simply mixing D315A and C287A mutants did not result in the cleavage of C287A even though both of them can be recruited onto the apoptosome ( Fig.  3C and data not shown). Another reason for the multimeric complex could be to increase the threshold of apoptosis to ensure nonspecific leakage of cytochrome c does not cause apoptotic response.