A Complex between Atg7 and Caspase-9

Background: Several cross-talk mechanisms between autophagy and apoptosis have been identified, where the same protein plays a central role in the opposing processes. Results: This study identified a novel cross-talk mechanism involving an Atg7·caspase-9 complex. Conclusion: Atg7 and caspase-9 mutually regulate each other's activity. Significance: The Atg7·caspase-9 complex may determine the balance between autophagy and apoptosis in response to stress. Several cross-talk mechanisms between autophagy and apoptosis have been identified, in which certain co-regulators are shared, allowing the same protein to participate in these opposing processes. Our studies suggest that caspase-9 is a novel co-regulator of apoptosis and autophagy and that its caspase catalytic activity is dispensable for its autophagic role. We provide evidence that caspase-9 facilitates the early events leading to autophagosome formation; that it forms a complex with Atg7; that Atg7 is not a direct substrate for caspase-9 proteolytic activity; and that, depending on the cellular context, Atg7 represses the apoptotic capability of caspase-9, whereas the latter enhances the Atg7-mediated formation of light chain 3-II. The repression of caspase-9 apoptotic activity is mediated by its direct interaction with Atg7, and it is not related to the autophagic function of Atg7. We propose that the Atg7·caspase-9 complex performs a dual function of linking caspase-9 to the autophagic process while keeping in check its apoptotic activity.

Several cross-talk mechanisms between autophagy and apoptosis have been identified, in which certain co-regulators are shared, allowing the same protein to participate in these opposing processes. Our studies suggest that caspase-9 is a novel co-regulator of apoptosis and autophagy and that its caspase catalytic activity is dispensable for its autophagic role. We provide evidence that caspase-9 facilitates the early events leading to autophagosome formation; that it forms a complex with Atg7; that Atg7 is not a direct substrate for caspase-9 proteolytic activity; and that, depending on the cellular context, Atg7 represses the apoptotic capability of caspase-9, whereas the latter enhances the Atg7-mediated formation of light chain 3-II. The repression of caspase-9 apoptotic activity is mediated by its direct interaction with Atg7, and it is not related to the autophagic function of Atg7. We propose that the Atg7⅐caspase-9 complex performs a dual function of linking caspase-9 to the autophagic process while keeping in check its apoptotic activity.
Autophagy is a multistep homeostatic and stress-induced process that mediates the degradation of cytoplasmic components through a lysosomal pathway (1). Autophagy involves the coordinated function of multiple Atg proteins that choreograph the various phases of the degradation process from the generation of the autophagosome to its fusion with an endocytic organelle and/or lysosome. Two ubiquitin-like reactions are required for autophagosome formation: (i) the ubiquitination-like conjugation of activated Atg12 to Atg5 and (ii) the post-translational lipidation of Atg8/LC3 3 with phosphatidyle-thanolamine (2). Atg7 is an E1-like enzyme that is required for both reactions, utilizing Atg10 as the E2 enzyme for the Atg12⅐Atg5 conjugation and Atg3 as the E2 enzyme for the lipidation of LC3 (3)(4)(5). The preformed Atg12⅐Atg5 conjugate is also required for the formation of lipidated LC3. Its phosphatidylethanolamine lipid moiety mediates membrane tethering and hemifusion of LC3; therefore, it serves as a reliable marker for a nascent autophagosome (6).
Autophagy is implicated in multiple human diseases that can be characterized by an imbalance in protein, organelle, or cellular homeostasis (7). In cancer, autophagy is involved in tumor biogenesis as well as in its response to therapy (8,9). Although autophagy may function under certain circumstances as a Type II mechanism of cell death, there is increasing evidence that in advanced cancer, it improves the fitness of the metabolically stressed cancer cells. This survival function is mediated via the elimination of damaged organelles and cytoplasmic stress by-products, the recycling of basic building blocks, and the utilization of recycled metabolites as an internal source of energy (10,11). As a survival mechanism, autophagy is expected to exhibit complete polarity to apoptosis. Such polarity has been detected in tumor cells, where either autophagy or a programmed cell death mechanism prevails (12). Reported cross-regulation mechanisms between autophagy and apoptosis hinge on shared proteins with distinct functions in the opposing processes (13). Such co-regulators include Bcl-2, which interacts with proapoptotic Bcl-2 family members as well as with Beclin-1 and Atg12 (14,15); Atg5, an essential autophagy player (16) that promotes mitochondrial apoptosis following its cleavage by calpains (17); FLIP, which inhibits caspase-8 activity, but also binds to Atg3, which blocks its conjugation to LC3 (18); and Atg12, which is essential for autophagy and also binds to Bcl-2 or Mcl-1 to promote apoptosis (15). In addition, caspases were found to cleave and inactivate Beclin-1 and Atg4D (19 -22), whereas autophagy degrades active caspase-8 (23).
In the current study, we present evidence for a novel crosstalk mechanism between apoptosis and autophagy: autophagic involvement of caspase-9 (C9), which is independent of its caspase activity. Potential involvement of catalytically active caspases in autophagic degradation has been reported in Drosophila (24,25). Although the critical function of caspases in apoptosis is firmly established, it has become evident that certain caspases mediate other cellular processes distinct from apoptotic cell death (26 -29). Thus, non-apoptotic functions have been determined for caspase-1, -3, -7, -8, -11, and -14, which involve their catalytic activity (30 -38). Two recent studies identified non-apoptotic activities for C9 that are dependent on its catalytic activity, including myotube formation in skeletal muscle (39) and the involvement of C9 enzymatic activity in lysosomal function (40). Our findings suggest that the Atg7⅐C9 complex can distinctly impact autophagy and apoptosis and may serve as a therapeutic target for the concomitant removal of apoptosis repression and the institution of autophagy inhibition.
Cell Lines-The tumor cell lines, including Hct116, HeLa, MB-MDA-231, and RKO, were obtained from ATCC. VP-16resistant MDA-MB-231 cells were selected by multiple cycles of etoposide treatments. Tumor cell lines and WT or engineered MEF were grown in Dulbecco's modified Eagle's medium containing 10% fetal calf serum, 20 mM L-glutamine, and 100 units/ml each of penicillin and streptomycin.
Transfection-Transfections were carried out with Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions using non-linearized plasmids for transient expression and linearized plasmids for stable expression. To generate caspase-9 KO MEF that stably express Tet-inducible C325A caspase-9 mutant, the cells were first stably transfected with linearized pcDNA6/TR that encodes the Tet repressor. Blasticidin selected clones with verified Tet repressor expression were further stably transfected with the linearized Tetinducible plasmids pcDNA4/TO, pcDNA4/TO/lacZ (Invitrogen), or pcDNA4/TO/C325A caspase-9. All procedures involving linearization, transfection, and stable clone selection were carried out as described previously (41)(42)(43).
RNAi-Beclin-1, caspase-9, and Vps34 specific siRNAs were obtained from Dharmacon as the siGENOME SMARTpool, which consists of four nonoverlapping siRNAs. The matching negative control from Dharmacon was the siGENOME Nontargeting siRNA Pool 2. Additionally, three distinct caspase-9 siRNAs were obtained from Invitrogen as the Stealth Select RNAi CASP9 set along with their matching negative controls, the Stealth RNAi negative controls LO and Hi GC. All RNAi results obtained by the Dharmacon siGENOME SMARTpool were confirmed by at least an additional distinct caspase-9specific siRNA. Hct116, MDA-MB231, RKO, MEF WT, and MEF caspase-9 KO cells (2.5 ϫ 10 5 ) were plated in a 6-well plate and, following 16 h (at 20% confluence), were transfected with 25-100 nM siRNA in Opti-MEM medium (Invitrogen) without fetal calf serum using oligofectamine reagent (Invitrogen) according to the manufacturer's transfection protocol. After 4 h, fetal calf serum was added to a final concentration of 10%. Mouse caspase-9 shRNA-expressing pLKO.1 plasmids (set of five non-overlapping target sequences) were obtained from Open Biosystems (catalog no. RMM4534-NM_015733). The non-mammalian shRNA-expressing negative control pLKO.1 plasmid was obtained from Sigma-Aldrich (catalog no. SHC002).
Molecular Cloning of Tet-inducible Mouse C325A Caspase-9 Mutant Expression Plasmid-Total RNA was isolated from WT MEF cells using RNA STAT-60 reagent (Tel-Test "B", Inc.). Reverse transcription of the MEF total RNA was carried out using SuperScript II RNase H Ϫ reverse transcriptase (Invitrogen), using an oligo(dT) [12][13][14][15][16][17][18] primer. PCR was performed with the Expand High Fidelity PCR system (Roche Applied Science) following the manufacturer's protocol. An amplicon containing the catalytic mutant (C325A) mouse caspase-9 open reading frame (ORF), which extends 6 nucleotides (nt) into the 5Ј-untranslated region (UTR) and 39 nt into the 3Ј-UTR was generated in two steps using the following four primers by overlap extension using the PCR method (44): 5Ј-CGGGGTAC-CCTCGCCATGGACGAGGCG-3Ј (forward; outside) and 5Ј-ACCACCCGCGGCCTGGATGAAGAAGAGCTT-3Ј (reverse; inside C325A underlined) and 5Ј-TTCATCCAGGCC-GCGGGTGGTGAGCAGAAAGACCAT-3Ј (forward; inside-C325A underlined) and 5Ј-CCGGAATTCCTTCGGAGAGA-ATAATGAGGC-3Ј (reverse; outside). The putative mutant caspase-9 amplicon was size-selected on a 1% agarose gel and extracted using the QIAquick gel extraction kit (Qiagen). It was then subjected to a second round of PCR with the forward and reverse outside primers and was size-selected and extracted as above. The purified mutant amplicon was digested with the restriction endonucleases KpnI and EcoRI and ligated (T4 ligase; New England Biolabs) into the Tet-inducible vector pcDNA4/TO (Invitrogen). Following transformation of DH5␣ cells (Invitrogen), randomly selected putative mouse C325A caspase-9 clones were sequenced by the University of Pittsburgh Genomics and Proteomics Core Laboratories.
Generation of Tet-inducible N-terminal 3ϫ FLAG Mouse C325A Caspase-9 Mutant Expression Plasmid-We carried out PCR using the recombinant C325A mouse caspase-9 plasmid (above) as a template with the following primers: 5Ј-CGGGG-TACCGCCATGGACTACAAAGACCATGACGGTGATTA-TAAAGATCATGACATCGATTACAAGGATGACGATGA-CAAGATGGACGAGGCGGACCGG-3Ј (forward; 3ϫ FLAG tag underlined) and the reverse outside primer described above for mouse C325A caspase-9. All other procedures for the processing of the N-terminal 3ϫ FLAG mouse C325A amplicon, its ligation to pcDNA4/TO, and sequence verification were as above.
Molecular Cloning of Mouse N-terminal V5 Atg7-Total RNA was isolated from WT MEF cells and reverse transcribed as described above for mouse C325A caspase-9. PCR was carried out as above using the following primer pair: 5Ј-CGCGGA-TCCGCCATGAGAGGGCCCTTCGAAGGTAAGCCTATC-CCTAACCCTCTCCTCGGTCTCGATTCTACGCGTACCG-GTATGGGGGACCCTGGACTG-3Ј (forward; V5 tag underlined) and 5Ј-ACGCGTCGACTTGCTTGCTTCAGAC-AGTCTC-3Ј (reverse). The putative V5-tagged Atg7 amplicon was size-selected on an agarose gel and extracted as described for caspase-9 above. It was subjected to a second round of PCR as above. The purified V5-Atg7 amplicon was digested with restriction enzymes BamHI and SalI and ligated to pcDNA4/TO, which had been digested with BamHI and XhoI. Following transformation of DH5␣ cells, randomly selected clones were sequenced by the University of Pittsburgh Genomics and Proteomics Core laboratories.
Molecular Cloning of Human ATG7 Expression Plasmid-Total RNA was isolated from MCF-7 breast cancer cells, and reverse transcription and PCR were all carried out as described above for the mouse C325A caspase-9 mutant. The following primers were utilized to generate a full-length ATG7 ORF that extends 6 nt into both the 5Ј-and 3Ј-UTRs: 5Ј-CGGGGTAC-CGAAATAATGGCGGCAGCTACG-3Ј (forward) and 5Ј-CCG-GAATTCGCCATCTCAGATGGTCTCATC-3Ј (reverse). The putative full-length amplicon was processed as described for C325A caspase-9, including digestion by KpnI and EcoRI. It was ligated to the mammalian expression vector pCR3.1 that had also been digested with KpnI and EcoRI. Transformation and sequence verification were performed as described above.
In Vitro Transcription-Translation-Atg7 and caspase-9 cDNAs were expressed in the TNT T7 transcription-translation reticulocyte lysate system (Promega). Each coupled transcription-translation reaction contained 1 g of plasmid DNA in a final volume of 50 l in a reticulocyte lysate reaction mixture according to the manufacturer's instructions. After incubation at 30°C for 90 min, the reaction products were immediately used or stored at Ϫ78°C.
Immunoblotting and Immunoprecipitation-Cell lysates were prepared with 1% Nonidet P-40, 20 mM Tris-HCl, pH 7.4, 137 mM NaCl, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, and 10 g/ml aprotinin. The immunoblotting and immunoprecipitation procedures were described in our previous publications (41,42,45,46). All immunoprecipitations were controlled by a sham procedure with nonspecific matching immunoglobulins. Quantification of scanned protein bands was performed by the US-SCAN-IT Gel software.
Cell Microscopy and Image Acquisition-Confocal images were obtained with an Olympus Fluoview 1000 confocal microscope and the companion software FV10-ASW1.6. Images were acquired with the use of the same setting at a resolution of 1020 ϫ 1024 pixels. Morphometric measurements were performed using Metamorph (Universal Imaging) on at least 50 cells/condition. Physiologically expressed endogenous LC3 puncta were monitored by two measures: (i) counting of dot number/cell and (ii) determination of cumulative dot area/cell area. Cumulative dot area and cell area were determined by Metamorph on images where the set threshold eliminated the detection of low or non-puncta staining.
For electron microscopy, cells were fixed with 2% paraformaldehyde and 2% glutaraldehyde in 0.1 M phosphate buffer (pH 7.0), followed by 1% OsO 4 . After dehydration, thin sections were stained with uranyl acetate and lead citrate for observation under a JEM 1011CX electron microscope (JEOL, Peabody, MA). At least 50 cells/treatment were utilized for quantification of all digitally acquired microscopic images.
Statistical Analysis-All immunoblot analyses are representative of at least three experiments. Images are representative of ϳ50 cells from at least three repeats. Quantification by Metamorph was performed on at least 50 cells/treatment, and the results were confirmed in three independent experiments. Statistical analysis was performed by GraphPad Prism V software, utilizing the nonparametric tests Mann-Whitney U (MWU) and Wilcoxon signed-rank, as appropriate.

Involvement of Non-apoptotic C9 in Autophagy-Recent
studies have demonstrated the induction of cytoprotective autophagy by chemotherapeutic agents, including TRAIL, in apoptosis-resistant tumor cells (46 -49). Apoptotic sensitization of such tumor cells could be achieved by pharmacological inhibition of autophagy or through RNAi of Atg genes, including Beclin-1, Vps34, and UVRAG (46,48). Upon inhibition, 30 -50% of the treated cells turn Annexin V-positive, and all such treated cells lose their clonogenic capability, as determined by a long term clonogenicity assay (46). In our efforts to elucidate the mechanism(s) of cell death induced by autophagy inhibition, we knocked down C9 with the expectation that it would inhibit the apoptotic shift. Surprisingly, C9 knockdown (KD) did not inhibit the execution of the autophagy-to-apopto-  sis shift but rather enhanced it significantly (data not shown). Utilizing tumor cell lines with significant TRAIL resistance (VP16-selected MDA-MB-231, RKO, and Bax Ϫ/Ϫ Hct116), we determined that comparable TRAIL sensitization was mediated by Beclin-1, Vps34, or C9 RNAi (Fig. 1, A-C). Likewise, C9 KD shifted to cell death the starvation response of Bax Ϫ/Ϫ Hct116 cells (Fig. 1D). To ascertain that these observations were not related to off-target effects, we performed the experiments with multiple distinct C9 siRNAs that demonstrated effective sensitization of Bax Ϫ/Ϫ Hct116 cells to TRAIL (Fig. 1,  E and F). The conclusion from this initial series of experiments was that C9 performs a non-apoptotic function, which in the context of cellular adaptation to stress supersedes its apoptotic activity.
The impact of C9 on autophagy was assessed by modifications in its expression level in cells undergoing autophagy in response to rapamycin or starvation, classical inducers of autophagic flux. The modifications in C9 expression levels included the utilization of C9 KO MEF from two sources (50,51), C9 KD, and transient and stable transfection of C9 KO MEF with the murine catalytically inactive C325A C9 mutant. Distinct and large endogenous LC3 puncta were detected in WT MEF but markedly less so in C9 KO MEF under basal conditions or following treatments by rapamycin or starvation (Fig. 2,   A and B). Further, LC3 decorated vesicles were detected only in WT MEF treated with rapamycin or exposed to starvation and not in similarly treated C9 KO MEF. The expression of C9 also impacted the ratio of LC3-II/LC3-I in either control or rapamycin-treated MEF; whereas rapamycin treatment of WT MEF increased this ratio by enhanced generation of LC3-II, there was no change in the already reduced level of LC3-II formation in C9 KO MEF (Fig. 2C). Likewise, whereas rapamycin enhanced p62 degradation in WT MEF, it did not impact the p62 expression level in C9 KO MEF. Although both LC3-I and LC3-II were present in C9 KO MEF, the difference in the LC3-II/I ratio between the two cell lines suggests that C9 enhances the mechanism responsible for the conversion of LC3-I to LC3-II. This interpretation is supported by the impact of C9 KD on the autophagic response of Bax Ϫ/Ϫ Hct116 to starvation or rapamycin (Fig. 2, D-J). In particular, the LC3-I to LC3-II conversion was more efficient in rapamycin-treated cells in combination with non-target siRNA control than with C9 siRNA (Fig.  2J). Likewise, the p62 degradation levels mediated by rapamycin were reduced in the same cells in which C9 KD reduced the conversion of LC3-I to LC3-II. The impact of C9 on the autophagosome formation process was corroborated by the reduced expression of LC3 puncta in C9 KD Hct116 cells subjected to starvation (Fig. 2, D-F). The level of LC3 puncta under basal conditions was also impacted by the expression level of C9, as demonstrated by C9 KD in WT MEF (Fig. 2, G-I). Conversely, transfection of C9 KO MEF with the C325A C9 mutant, but not vector control, increased the basal expression of LC3 puncta (Fig. 2, K-M).
Because protein degradation is mediated by both proteasomal and lysosomal mechanisms, characterization of an autophagic effector(s) may be best conducted in cells with inhibited proteasomal activity. In support of such an approach, it has been documented that proteasomal inhibition induces an autophagic response as a compensatory mechanism for protein degradation (52,53). In line with the compensatory notion, we observed a marked difference in the response of WT and C9 KO MEF to the proteasomal inhibitor MG132. A significant reduction in LC3 puncta formation in response to MG132 was detected in the absence of C9 (Fig. 3, A-D). The WT MEF LC3 puncta were highly co-localized with ubiquitin-decorated proteins that accumulated in response to MG132, but such co-localization was not observed in C9 KO MEF harboring a reduced number of LC3 puncta. The reduced formation of autophagosomes as detected by LC3 puncta or LC3-II/LC3-I ratio was corrected by transfection of the C9 KO MEF with the C325A C9 mutant (Fig. 3E). Thus, the ratio of LC3-II/LC3-I was increased 7-fold upon transient transfection of C325A C9 mutant as compared with MG132-treated vector control cells. This change in LC3-II/LC3-I ratio was accompanied by a reduction in the expression level of p62 as compared with vector-transfected C9 KO MEF. Because the C9 expression level in our transient transfection is not controllable, we generated C9 KO MEF that stably express Tet-inducible C325A C9. Low level constitutive expression of C9 (due to leakiness) was detected in the absence of Tet (Fig. 3F). Thus, we utilized both non-treated and Tettreated C9 KO MEF that were stably transfected with Tet-inducible C325A C9 as cells expressing different incremental levels of C9. Stable expression of C325A C9 increased the LC3 expression signal above vector-transfected cells in both control and MG132-treated MEF (Fig. 3, G and H). Increased formation of autophagosomes was also documented by transmission electron microscopy in C9 KO MEF reconstituted with C325A C9 relative to those reconstituted with vector under basal conditions (Fig. 4, A and B) or in response to rapamycin (Fig. 4, C and  D). Cumulatively, these findings suggest that non-apoptotic C9 enhances the autophagic response mediated by archetypical inducers of autophagy, including starvation, rapamycin, and proteasomal inhibition. The findings further suggest that C9 significantly affects the autophagosome formation process (as assessed by autophagosome morphology and LC3 puncta) and appears to specifically impact the LC3-I to LC3-II conversion process.
Interaction of C9 with Atg7-To expand our investigation of the newly identified autophagic function of C9, we screened C9 immunoprecipitated complexes for an array of autophagic proteins. We identified Atg7 as a C9-interacting protein. Thus, a complex between endogenous Atg7 and endogenous C9 was confirmed by a two-way immunoprecipitation in Hct116, HeLa, and two non-small cell lung cancer (A549 and H460) cell lines (Fig. 5, A-C). Two-way co-immunoprecipitation of Atg7 and C9 was also confirmed for C9 KO MEF that were tran-siently transfected with 3ϫ FLAG-C325A C9 and V5-Atg7 (Fig.  5D). The specificity of the interaction was confirmed by a lack of C9 co-immunoprecipitation in Atg7 KO MEF (Fig. 5E). Further, a direct interaction between Atg7 and C9 was confirmed by co-immunoprecipitation of in vitro translated Atg7 and C9 proteins (Fig. 5F).
We then assessed the impacts of starvation and apoptosis on the Atg7⅐C9 complex formation. As determined by either Atg7 immunoprecipitation (Fig. 5G) or C9 immunoprecipitation (Fig. 5H), the Atg7⅐C9 complex formation was enhanced by starvation and decreased by staurosporin-mediated apoptosis. The disruption of the Atg7⅐C9 complex in response to staurosporin may suggest that the apoptotic activity of C9 favors its dissociation from Atg7.
To start mapping the Atg7 binding site within C9, we generated, in addition to the 3ϫ FLAG-tagged full-length C325A C9 construct, three deletion mutants (⌬CARD, ⌬large subunit (⌬LS), and ⌬linker domain and small subunit (⌬LD ϩ ⌬SS)) (Fig. 6A). FLAG-tagged full-length C325A C9 and the indicated deletion mutants were transiently transfected into C9 KO MEF, which were subjected to immunoprecipitation of endogenous Atg7. The C9 large subunit was identified as the core of the binding area because Atg7 co-immunoprecipitated the fulllength C9, ⌬CARD C9, and ⌬LD ϩ ⌬SS but not the ⌬LS (Fig.  6B). Subsequently, we generated two additional C9 mutants, each with a partial deletion of the LS: 3ϫ FLAG-C9⌬Val 177 -Arg 263 (⌬A) and 3ϫ FLAG-C9⌬Ala 264 -Asp 353 (⌬B). Each of these mutants was transfected with V5-Atg7 into C9 KO MEF and assessed for Atg7 binding. The 3ϫ FLAG-C9⌬B mutant failed to co-precipitate with V5-Atg7 (Fig. 6, C and D), identifying the region that stretches from Ala 264 to Asp 353 within C9 LS as required for the interaction of C9 with Atg7.
To determine if the C9 interaction site with Atg7 is also required for C9 autophagic function, we assessed the LC3-II levels in C9 KO MEF transfected with either full-length C325A C9 or C9⌬Ala 264 -Asp 353 (C9⌬B) and treated with MG132 (Fig.  6E). MG132 treatment enhanced the LC3-II level in C9-transfected cells but not in those transfected with either vector or with the C9⌬B deletion mutant. These findings suggest that the C9 site, which is required for interaction with Atg7, is also required for its impact on LC3-II formation.
Mutual Impacts of Atg7 and C9 on Each Other-The possibility that Atg7 is a substrate for C9 proteolytic activity or C9-activated caspase-3/7 was excluded in tumor cell lines treated by staurosporin or etoposide. Thus, neither a change in the expression level of endogenous Atg7 nor the appearance of any Atg7 cleavage products was detected in staurosporin-or VP16-treated cells confirmed for the processing and activation of C9 (Fig. 7, A, C, and E). In line with the validation that Atg7 does not serve as an in vivo substrate of C9, the interaction of Atg7 with C9 appears to have an inhibitory effect on the processing of the C9 prodomain. Thus, overexpression of Atg7 repressed the processing of C9 under apoptotic conditions mediated by staurosporin in Hct116 and HeLa cells (Fig. 7, A, C, and E). In contrast to Atg7, transfection of either Atg5 or Beclin-1 (Fig. 7, B and D) did not inhibit the staurosporin-mediated processing of C9 or other downstream caspases. To investigate if the Atg7-mediated inhibition was dependent on the autophagic activity of Atg7, we transfected Atg7 into Atg16 KO MEF and assessed its impact on staurosporin-mediated processing of the C9 prodomain (Fig. 7F). The staurosporinmediated processing of C9 was inhibited in Atg7-transfected, but not in vector-transfected Atg16 KO MEF. Because Atg16 is an essential autophagic effector and the Atg16 KO MEF are autophagy-deficient (54), the C9-targeted inhibitory activity of Atg7 appears to be independent of the autophagic function of Atg7. Although C9 processing was markedly inhibited in the presence of Atg7 (Fig. 7, A, C, and E), the processing of caspase-7, -8, or -3 or poly(ADP ribose) polymerase was only partially inhibited, suggesting that the apoptotic activity of staurosporin was not completely dependent on C9. Indeed, transfection of Atg7 KO MEF with Atg7, but not with vector control, significantly inhibited, but only partially (percentage of inhibition, 32.3 Ϯ 3.3% (mean Ϯ S.E.)), the staurosporin-mediated cell death of these cells (Fig. 8A). The observed partial inhibi-   tion may relate to the ability of staurosporin, a broad spectrum kinase inhibitor, to initiate multiple cell death pathways in which the activation of caspase-7 and -8 may be both upstream and downstream of C9. In support of this interpretation, benzyloxycarbonyl-VAD-fluoromethyl ketone, a pan-caspase inhibitor, arrested the cell death mediated by staurosporin treatment (24 h), whereas benzyloxycarbonyl-LEHD-fluoromethyl ketone, a C9-specific inhibitor, demonstrated only partial inhibition at a level of ϳ25%. Reconstitution of Atg7 in Atg7 KO MEF also reduced the catalytic activity of C9, as assessed by the processing of a specific C9 fluorescence substrate in the response of these cells to starvation or MG132 (Fig. 8B). To investigate the capability of Atg7 to sequester C9, we transfected C9 KO MEF with V5-Atg7 or vector control, immunoprecipitated the transfected Atg7, and added active recombinant C9 to the pellet. The washed pellet and its supernatant (sup) were assessed by immunoblotting for the presence of Atg7 and C9, and the supernatant was further assessed for C9 enzymatic activity (Fig. 8C). An increased level of recombinant C9 was present in the Atg7 immunoprecipitate (1:1.7 ratio of vector/Atg7 transfection), and the C9 activity in the matching supernatant was correspondingly decreased (1:0.6 ratio). These experiments suggest that the Atg7⅐C9 complex could function by sequestering C9 from its activation site. The ability of Atg7 to inhibit C9 enzymatic activity was confirmed in a cell-free system, where active recombinant C9 was mixed with increasing doses of purified GST-Atg7. Each of the utilized doses of GST-Atg7 significantly inhibited recombinant C9 protease activity, as compared with the GST control (Fig. 8, D and E).
To investigate the impact of C9 on the Atg7 autophagic function, we transfected Atg7 KO MEF with WT Atg7 concomitant with a C9 shRNA construct(s) (Fig. 8F). The transfection of Atg7, but not the vector control (lanes 1, 2, 7, and 8) repaired the cell's capability of producing the Atg12⅐Atg5 complex and of converting LC3-I to LC3-II. However, in the presence of each of four independent C9 shRNAs or their mix (lanes 4, 6, 10, 12, and 14), LC3-II formation was reduced, whereas the conjugation capability required for Atg12⅐Atg5 complex formation remained unchanged. In agreement with the LC3-II formation levels, degradation of p62 was reconstituted by Atg7 transfection and reversed to a reduced level by the C9 KD. These findings suggest that the autophagic function of C9 is mediated at least partly by its interaction with Atg7, it is localized to the molecular phase of LC3-II formation, and it significantly impacts the degradation of p62.

DISCUSSION
Our studies revealed an unexpected non-caspase function for C9 that is localized to the early phase of autophagosome formation and is mediated via its interaction with Atg7. Depending on the cellular context, Atg7 and C9 mutually impact each other; whereas Atg7 represses the apoptotic activity of C9, the latter facilitates the Atg7-dependent formation of autophagosomal LC3-II. Induced autophagy is associated with increased formation of the Atg7⅐C9 complex, which facilitates LC3-II formation while keeping C9 apoptotic activity in check.
Reduced expression of the Atg7⅐C9 complex is associated with an apoptotic response, where C9 is relieved from sequestration and repression by Atg7.
The survival activity of C9 was initially discovered by an RNAi approach that was expected to block a shift from autophagy to apoptosis upon autophagy inhibition but was surprisingly found to exacerbate the ensuing apoptotic process. Whereas KD of caspase-8 inhibited apoptosis and elucidated its major role in the apoptotic process induced by inhibition of autophagy, C9 KD intensified the apoptotic response, suggesting the supersession of C9 survival activity over its apoptotic activity in the context of an adaptive response to cellular stress.
Our findings suggest that the processing of C9 is effectively inhibited in cells that overexpress Atg7 and that this inhibition is not dependent on the cytoprotective effect of autophagy because it occurs in Atg16 KO MEF, which are deficient in an essential autophagic protein. The Atg7 interaction with the C9 large subunit is consistent with the ability of Atg7 to interact with both pro-C9 and with (CARD-deficient) recombinant C9. Pro-C9 monomers are mainly cytosolic, and their interaction with Atg7 may sequester them from the mitochondria and apoptosome. Because C9 processing occurs mainly on the apoptosome, cytosolic sequestration of C9 by Atg7 may indirectly inhibit C9 activation and processing. Additionally, Atg7 may interact with apoptosome-associated C9, inhibiting its activation and/or processing. Unlike other caspases, whose activation requires prodomain processing (i.e. intrachain cleavage), the activation of C9 has been considered independent of its auto- . a.u., arbitrary units. C, recombinant C9 is sequestered by its interaction with Atg7. C9 KO MEF were transfected with V5-Atg7 or vector control and subjected to V5 IP. Recombinant active C9 was then added to the protein A-Atg7 beads. Following washing, the pellet was assessed by immunoblotting, and the protein bands were quantified by densitometric measurements. The wash supernatant was assessed for C9 activity by the C9 Glo assay. D and E, the protease activity of recombinant C9 is inhibited by recombinant Atg7. Recombinant C9 was incubated alone or with wheat germ in vitro translated GST-Atg7 or the GST-control protein (1-5 g/ml, 1 h, 37°C). C9 enzymatic activity was determined by the C9 Glo assay. The data are means Ϯ S.D. of triplicate determinations in one experiment of three performed with similar results; *, p Ͻ 0.05 (Wilcoxon signed rank test). E, immunoblotting of the proteins utilized in D. The doses of the loaded proteins are equivalent to the lowest and medium doses utilized in D. F, C9 KD reduces the Atg7 conjugation activity required for LC3-II formation and p62 degradation. Atg7 KO MEF were transfected with WT V5-Atg7 or matching vector control (24 h) and subsequently treated with non-target (NT) control shRNA (lanes 1, 2, 7, and 8) and four independent C9 shRNAs (lanes 3-6 and 9 -12) or the shRNA mix (lanes 13 and 14) for 48 h. The cell lysates were assessed by immunoblotting. Lanes 1-6 and 7-14 represent two independent experiments. LC3-II was formed only in the presence of Atg7, and its level was reduced by the C9 shRNAs. Likewise, p62 degradation that was observed in Atg7-transfected cells (lanes 2 and 8) was blocked by the distinct C9 shRNAs (lanes 4, 6, 10, 12, and 14). Atg12⅐Atg5 complex was also produced only in the presence of Atg7, but its production level was not impacted by C9 KD. Densitometric quantification of LC3-II (in relative units) is indicated. Also, the ratios of p62 in Atg7 transfectants relative to p62 in their matching vector controls (lanes 2 versus 1, 4 versus 3, 6 versus 5, 8 versus 7, 10 versus 9, 12 versus  11, and 14 versus 13) are indicated. catalytic processing (55,56). Although this issue has remained controversial, autoprocessing of C9 is either required for its activation (57) or for the molecular timer-mediated recruitment of pro-C9 to the apoptosome (55,58). Thus, the Atg7mediated inhibition of C9 processing could potentially inhibit the activation of C9 at the level of the apoptosome or interfere with its recruitment to the apoptosome. Additionally, apoptosome-independent activity of C9 has been described (59), and thus, Atg7 may be involved in the regulation of this cell death mechanism. These potential mechanisms of Atg7-mediated inhibition of C9 are now being intensely investigated.
Currently, it is not clear what determines the participation of C9 in apoptosis or autophagy. It is possible that a significant release of mitochondrial apoptogenic proteins would shift the balance toward apoptosome activation, whereas insufficient mitochondrial involvement would allow the opposing regulation of C9 by Atg7. Our findings suggest that the Atg7⅐C9 complex is involved in the determination of the balance between apoptosis and autophagy. Such toggling activity could potentially serve as a therapeutic target for the concomitant enhancement of apoptosis and the inhibition of autophagy. Although reciprocal regulation between Atg7 and C9 is a likely mechanism for the potential determination between participation of C9 in either autophagy or apoptosis, additional molecular scenarios should be considered. Because specific C9 phosphorylation sites are associated with inactivation of its enzymatic function, a phosphorylation mechanism may serve as a potential determinant in the involvement of C9 in either apoptosis or autophagy. Additionally, association of C9 with XIAP or Bruce or its potential ubiquitination by these E3 ligases (60, 61) may force C9 into an autophagic rather than apoptotic mode. Further studies should determine the exact molecular mechanisms that tie C9 to autophagosome formation and to the ultimate determination between its opposing functions within either apoptosis or autophagy.