Caspase-9 Can Be Activated without Proteolytic Processing*

The recombinant form of the proapoptotic caspase-9 purified following expression in Escherichia coliis processed at Asp315, but largely inactive; however, when added to cytosolic extracts of human 293 cells it is activated 2000-fold in the presence of cytochrome c and dATP. Thus, the characteristic activities of caspase-9 are context-dependent, and its activation may not recapitulate conventional caspase activation mechanisms. To explore this hypothesis we produced recombinant forms of procaspase-9 containing mutations that disabled one or both of the interdomain processing sites of the zymogen. These mutants were able to activate downstream caspases, but only in the presence of cytosolic factors. The mutant with both processing sites abolished had 10% of the activity of wild-type, and was able to support apoptosis, with equal vigor to wild-type, when transiently expressed in 293 cells. Thus caspase-9 has an unusually active zymogen that does not require proteolytic processing, but instead is dependent on cytosolic factors for expression of its activity.

The recombinant form of the proapoptotic caspase-9 purified following expression in Escherichia coli is processed at Asp 315 , but largely inactive; however, when added to cytosolic extracts of human 293 cells it is activated 2000-fold in the presence of cytochrome c and dATP. Thus, the characteristic activities of caspase-9 are context-dependent, and its activation may not recapitulate conventional caspase activation mechanisms. To explore this hypothesis we produced recombinant forms of procaspase-9 containing mutations that disabled one or both of the interdomain processing sites of the zymogen. These mutants were able to activate downstream caspases, but only in the presence of cytosolic factors. The mutant with both processing sites abolished had 10% of the activity of wild-type, and was able to support apoptosis, with equal vigor to wild-type, when transiently expressed in 293 cells. Thus caspase-9 has an unusually active zymogen that does not require proteolytic processing, but instead is dependent on cytosolic factors for expression of its activity.
Apoptosis, the ordered dismantling of animal cells that results in their death and removal from the organism, requires specific proteolysis of a subset of cellular proteins, mediated by caspases. Many apoptotic responses are initiated by activation of the apical caspases-8 or -9; the former by recruitment to ligated cell surface receptors belonging to tumor necrosis factor receptor-1 family (1,2), and the latter by recruitment to Apaf-1, in the presence of ATP or dATP, following delivery of cytochrome c (cyto-c) 1 from mitochondria (3). Activation of either of these two initiator caspases can lead to activation of the exe-cutioner caspase-3 (casp-3), by direct proteolysis of the casp-3 precursor (4,5). Indeed all caspase zymogens are thought to be activated by limited proteolysis within a linker segment which results in generation of the characteristic large and small subunits of the catalytically active enzyme. Thus it has been specifically demonstrated that casp-1, -3, -7, and -8 gain marked increases in activity after proteolytic processing (6 -9). This is typical of proteolytic enzymes from all families and catalytic classes. As pointed out by Neurath (10), the zymogens are stored in an inactive form to prevent adventitious proteolysis before it is needed at the required site of action. Frequently protease zymogens are activated by other proteases, in a cascade mechanism that results in either amplification or localization to specific sites. There are few exceptions to the rule that zymogen activation requires proteolysis, and these exceptional zymogens, such as tissue plasminogen activator (tPA), usually have specific allosteric modulators that allow them to become active at the required site, without proteolysis.
The initiator casp-8 has some of these exceptional properties. Casp-8 initiates apoptotic signaling induced by specific ligation of death receptors (reviewed in Refs. 11 and 12). Recruitment of procasp-8 to the cytosolic face of ligated Fas or tumor necrosis factor receptor-1 results in its autoactivation by a mechanism involving clustering of zymogen molecules (9,13). The zymogen possesses ϳ1% of the activity of the activated enzyme, and it is hypothesized that this small amount of activity of the zymogen is sufficient to drive cleavage of clustered casp-8 zymogen molecules to the active form. Thus casp-8 has a zymogenicity ratio of 100 (defined as the ratio of the activity of the enzyme to the zymogen). Nevertheless, proteolysis of the casp-8 zymogen is thought to be required for its activation (9,13).
In light of the casp-8 activation studies, it has been hypothesized that the initiator caspases, including caspases-8 and -9 (14) and Caenorhabditis elegans CED-3, share a common mechanism of activation (15). This would require clustering zymogens that possess a small amount of activity in their single chain forms, with production of the active protease dependent on proteolytic processing of the zymogen. In turn, this would mean that preventing proteolysis of the zymogen in the interdomain linker should inactivate apoptotic signal transmission. We tested this hypothesis for casp-9 by utilizing a cell-free system (16) that allows dissection of the requirements for zymogen activation.

EXPERIMENTAL PROCEDURES
Materials-Acetyl-DEVD-p-nitroanilide (Ac-DEVD-pNA) and carbobenzoxy-VAD-fluoromethyl ketone (Z-VAD-FMK) were from Bachem. Flourogenic 7-amino-4-trifluoromethyl coumarin (AFC) caspase substrates (Ac-DEVD-AFC and Ac-LEHD-AFC) were from Enzyme System Products. Dithiothreitol was from Diagnostic Chemicals Ltd. Sucrose was from Mallickrodt. All other chemicals were from Sigma. Rabbit anti-caspase-9 was kindly provided by Dr. Douglas R. Green, La Jolla Institute for Allergy and Immunology (8). Rabbit antisera to caspases-3 and -7 were from in-house collections (17). The casp-3 zymogen was expressed and purified from Escherichia coli as described previously (8). The protein concentration of purified zymogens was determined from the absorbance at 280 nm based on the molar absorption coefficients calculated from the Edelhoch relationship (18).
Methods-The specific mutants casp-9 (D315A), casp-9 (D330A), and casp-9 (D315A/D330A) were constructed using overlap polymerase chain reaction techniques. A CARD domain deleted form (⌬CARD casp-9) was constructed by truncating the cDNA and introducing an initiator codon substituting Val 139 . All constructs were expressed in E. coli and isolated by virtue of an engineered C-terminal His 6 purification * This work was supported by Danish Natural Science Foundation Grant 9600412 (to H. R. S.), Leukemia Society of America Grant 5483-98 (to Q. L. D.), National Institutes of Health Training Grant 5T32 GM07863 (to E. W. H.), and National Institutes of Health Grants NS37878 (to G. S. S.), CA72994 (to J. C. R.), and AG15402 (to G. S. S. and J. C. R.). 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 U.S.C. Section 1734 solely to indicate this fact.
Cytosolic extracts of 293 cells, and casp-9-depleted extracts were prepared as described (8). The extracts were stored in appropriate portions at Ϫ80°C. Activation of casp-9 in cytosolic extracts was initiated by addition of 1 mM dATP and 10 M cyto-c followed by incubation at 37°C for 30 min (3). These conditions routinely result in 100% activation of the endogenous executioner casp-3 as judged by comparison with recombinant, fully active casp-3. The caspase activity in 2.5 l of cytosolic extract was determined by measuring the cleavage of Ac-DEVD-AFC as described previously (8,20,21). Alternatively, the generation of activity can be followed by release of pNA from Ac-DEVD-pNA, reflecting the instantaneous generation of executioner caspases, which in 293 cells is predominantly due to casp-3 (8).
Cell death assays were performed as described previously (24). Briefly, 2.5 ϫ 10 5 293T cells were transfected with 1 g of the mutant or wild-type plasmids plus 0.1 g of cytomegalovirus-␤-galactosidase in six-well tissue culture dishes using calcium phosphate precipitation. Approximately 36 h after transfection, the cells were fixed with 0.5% glutaraldehyde and then stained with 5-bromo-4-chloro-3-indoxyl-␤-Dgalactopyranoside and examined by phase contrast microscopy. At least 300 ␤-galactosidase-positive cells were counted for each transfection and identified as apoptotic or nonapoptotic based on morphological alterations typical of adherent cells undergoing apoptosis, including becoming rounded, condensed, and detached from the dish (25). Percent apoptosis represents the mean value from two independent experiments (mean Ϯ S.D.).

RESULTS AND DISCUSSION
Purified Recombinant Casp-9 Is Maximally Active Only in the Presence of Cytosolic Components-Recombinant casp-9 was obtained in the fully processed form, consisting of two chains caused by autolytic cleavage at Asp 315 , as determined by N-terminal Edman degradation and MALDI-TOF analysis of the isolated material (Fig. 1, A and B). The CARD domaindeleted form of casp-9 was obtained in the same way, with the same internal cleavage site. Both purified proteins had very low activity when measured with Ac-LEHD-AFC and contained only 1-5% active sites as judged Z-VAD-FMK titration. Thus the recombinant material is largely inactive compared with other caspases such as 3, 6, 7, and 8, which normally demonstrate 50 -100% active sites following expression and purification by the same procedure (5). Moreover, activation of the natural substrate procasp-3 was extremely slow, leading us to speculate that the recombinant casp-9 forms purified from E. coli are not equivalent to those operating in vivo. To test this we immuno-depleted endogenous casp-9 from a cytosolic extract made from 293 cells, and reconstituted with the recombinant forms.
Quantitative immunoblot analyses revealed that the concentration of casp-9 in the extract was approximately 20 nM (not shown). Addition of 200 nM (10-fold excess of the endogenous concentration) of the interdomain cleaved forms of ⌬CARD or full-length casp-9 to the depleted extract resulted only in limited induction of DEVD-AFC cleaving activity (Fig. 1C). In this assay, the cleavage of DEVD-AFC measures the activity of protease(s) that are activated by casp-9 in cytosolic extracts and thereby acts as an indirect measure of the activity of casp-9 itself. The rate is greatly amplified upon addition of cyto-c/ dATP, which activates casp-9 in cytosolic extracts (3,26). While the activity of the ⌬CARD casp-9 decreases slightly upon addition of cyto-c/dATP, full-length casp-9 recapitulated the activation rate of the endogenous material, even when added at 20 nM, equivalent to the endogenous casp-9 concentration. This is equivalent to a 2000-fold amplification of activity relative to the action of fully processed casp-9 on purified procasp-3 and demonstrates that binding to a cytosolic factor, probably the reported casp-9 activator Apaf-1 (26), is essential for the activity of casp-9 and not just for its activation. This is consistent with the demonstration that cleavage of casp-9 in cytosolic extracts devoid of cyto-c failed to accelerate caspase activation (8).
Proteolytic Processing of Casp-9 Is Not Required for Activation-Based on the consensus view that procaspase activation requires proteolysis in the interdomain linker segment (reviewed in Ref. 4), we expected mutants that prevent processing would prevent activation. Moreover, since casp-9 is recruited to Apaf-1 for activation, the noncleavable mutants may displace endogenous casp-9 from its activation complex and thus prevent its activation. Thus we constructed the processing site mutants casp-9 (D315A), casp-9 (D330A), and casp-9 (D315A/ D330A), as well as a catalytically incompetent mutant casp-9 (C287A), see Fig. 1, A and B, for details, and tested them in the in vitro cytosolic reconstitution system. Substitution of endogenous casp-9 by any of the cleavage site mutants recapitulated caspase activation. The single mutants were only slightly less effective than endogenous casp-9, while the double mutant (D315A/D330A) was the least effective, but still reconstituted FIG. 1. Reconstitution of casp-9 in cytosolic extracts. A, recombinant proteins were analyzed for purity by SDS-PAGE. The wild-type, ⌬CARD, and D330A versions were obtained in the two-chain form due to autolytic processing at Asp 315 (white asterisk), whereas D315A and D315A/D330A were largely unprocessed, but contained 5.1% of material cleaved at Ser 307 (black asterisk). B, representation of the casp-9 precursor. The start codon for the ⌬CARD version replaces Val 139 . The position of the catalytic cysteine (Cys 287 ), the autolytic cleavage sites observed at Ser 307 and Asp 315 , and the Asp 330 site cleaved by casp-3 are denoted by arrows, and asterisks corresponding to the cleavage products identified in A. C, naive cytosolic extracts were immunodepleted of casp-9 and reconstituted with the indicated concentrations of casp-9 recombinants (nanomolar). Reconstitution of the depleted extracts recapitulated caspase activation, measured by DEVD-AFC substrate hydrolysis, only in the presence of cyto-c/dATP and only when casp-9 contained an intact CARD domain.
Interestingly, the linker site mutants contained a small amount of aberrantly processed material that was determined by N-terminal sequence analysis and MALDI-TOF to be due to cleavage between Ser 307 and Pro 308 (Fig. 1A). The amount of this material relative to the D315A/D330A single chain zymogen was determined to be 5.1 (Ϯ1.2)% based on quantitative image analysis of Coomassie-stained gels. The origin of this cleaved material is a mystery, but appears to require a competent catalytic site in casp-9, since it was absent from the C287A mutant. The presence of this aberrantly processed material raises two possibilities which pertain directly to whether proteolytic processing of casp-9 is required for activation. Is the small amount of aberrant material the activator, not the residual 95% of unprocessed material in the D315A/D330A cleavage site mutant? This is unlikely since an amount of wild-type processed casp-9 equal to the aberrantly processed material was only able to generate 9% maximal activity, as opposed to the 40% attained by the D315A/D330A cleavage site mutant ( Fig. 2A). Moreover, if casp-9 requires cleavage for activation, it should compete with unprocessed material (D315A/D330A) for the activator complex. Indeed, when we simulated this by adding 2 nM wild-type processed casp-9 to 18 nM catalytic mutant (C287A), activation was abolished, presumably because the catalytic mutant prevents recruitment of processed casp-9 to the activator complex. Therefore we are left with a second possibility: the unprocessed D315A/D330A mutant is directly responsible for generating activity. To test whether the aberrant material increased during the incubation, Western blot analysis was conducted on cytosolic extracts that had been activated by cyto-c/dATP (Fig. 2C). No increase in the amount of the aberrant material, quantitated by image analysis, could be observed. This rules out adventitious proteolytic activation of the D315A/D330A mutant during the course of the assay and confirms that casp-9 can become active without proteolytic processing.
The generation of DEVD-pNA cleavage as a function of time provides a measure of instantaneous casp-3 activation. Casp-9 mutants that can be cleaved at the Asp 315 autoprocessing site (wild-type casp-9 and D330A) demonstrate kinetics very similar to endogenous casp-9, but mutants that cannot be cleaved at Asp 315 (D315A and D315A/D330A) cause DEVD-pNA cleavage more slowly (Fig. 2B). However, in the D315A mutant there is a significant beneficial effect of being able to form a two chain enzyme by cleavage at Asp 330 . Thus, the rate of caspase activation increases significantly as an increasing amount of caspase activity is generated, presumably due to a feedback of casp-3 onto the Asp 330 site of casp-9 as proposed by Srinivasula et al. (14).
Because the activation of procasp-3 by casp-9 is a complex process, it is difficult to derive an exact quantitation of the activity of casp-9. However, the cumulative caspase activity caused by endogenous casp-9 is 2.5-fold that of the double mutant, and this occurs about four times faster for the endogenous casp-9 than for the double mutant (Fig. 2, A and B). The product of these two measures of activation indicates that the double mutant has at least 10% of the activity of endogenous casp-9 when both are allowed to activate under identical conditions in cyto-c-programmed cytosols. Thus, since the casp-9 D315A/D330A mutant can be considered a zymogen, the zymogenicity ratio, which describes the activity of an active enzyme relative to its latent precursor, is at most 10.
Cleavage Site Mutants Do Not Act as Dominant Negative Inhibitors of Endogenous Casp-9 -These results combine to suggest that the cleavage site mutants would not act as dominant negative inhibitors and that they indeed should support apoptosis. Only the catalytic mutant should act as a dominant negative. To test this hypothesis, the constructs were subcloned into pcDNA3 and transiently expressed in 293T cells. As demonstrated in Fig. 3, both D315A and D315A/D330A induced FIG. 2. Activation of casp-9 mutants. Purified casp-9 mutants were added to immunodepleted extracts at 20 nM, followed by addition of cyto-c/dATP. Caspase activity was measured either after 30 min at 37°C by using Ac-DEVD-AFC (A) or followed continuously at 37°C in the presence of 0.4 mM Ac-DEVD-pNA (B). In A, additional controls were performed with 1 nM processed casp-9 (designated 5% wt) and a mixture of 2 nM processed casp-9 with 18 nM catalytic mutant (designated 10% wt ϩ C287A). C, cytosolic extracts immunodepleted of endogenous casp-9, to which 20 nM recombinant proteins were added, were Western blotted with an anti-casp-9 antiserum before (Ϫ) or after addition of cyto-c/dATP for 30 min at 37°C (ϩ). Note that the portion of material cleaved at Ser 307 does not increase during activation by cytoc/dATP (black asterisks). The extra band seen following activation of the D315A extract is due to cleavage at Asp 330 by casp-3 (14). apoptosis to the level observed for the wild-type and D330A, whereas the catalytic mutant casp-9 (C287A) failed to induce more apoptosis than the empty vector control. Only the catalytic mutant acted to suppress casp-9 activation in vivo, presumably by competing with endogenous casp-9 for binding to the cytosolic activators.
Zymogenicity and Casp-9 -Several caspases undergo proteolytic processing when expressed in E. coli, probably due to intrinsic catalytic properties of the zymogens which, though low, are sufficient to allow autolysis under the high concentrations obtained in expression (4,5). However, unlike other caspases which demonstrate substantial activity when purified from E. coli, casp-9 appears to be largely inactive until it is delivered to the correct cellular environment. It is not clear whether the endogenous cytosolic factors (presumably Apaf-1 in the presence of cyto-c/dATP) bind casp-9 and then release it in an active conformation or whether casp-9 must remain attached to retain activity. Experiments with purified cytosolic factors should answer this question. Nevertheless, proteolysis of casp-9 does not seem to be required for its activation. Certainly, processing in the interdomain linker, characteristic for other caspases, enhances casp-9 activity, but only by 10-fold, thus the zymogenicity ratio of casp-9 is about 10. On the other hand, binding to cytosolic factors enhances activity 2000-fold to give a robust enzyme capable of activating available procasp-3 in a few minutes. In the natural cellular environment it is clear that casp-9 is processed during apoptosis, as demonstrated in (14) for example, and this can occur either due to an autoprocessing cleavage at Asp 315 or by trans-processing by casp-3 at Asp 330 . However, in light of our data it would now be necessary to distinguish between processing (which may be adventitious) and activation, which does not appear to derive its driving force simply from proteolytic processing.
These observations on the relation of proteolytic processing to zymogen activation are reminiscent of tPA, which demonstrates a very small zymogenicity ratio, 2-10 (27). However, upon binding to fibrin, the ability of tPA to activate its physiologic substrate plasminogen increases several thousandfold (28). Presumably enzymes such as tPA and casp-9 have abolished the requirement for proteolysis as a mechanism of substantially increasing their activities, because allosteric regulators substitute this function, fibrin for tPA and Apaf-1 for casp-9. In the case of tPA, specific side chain interactions, absent in other members of the chymotrypsin family, allow activity of the zymogen. However, in the absence of a molecular structure of the casp-9 zymogen, few clues are available to explain the high activity of the unprocessed protein. One clue is suggested by the structure of active casp-1 and casp-3, each composed of two catalytic units thought to arise from dimerization of monomeric zymogens (reviewed in Ref. 5). If activa-tion of zymogens of the initiator caspases-8 and -9 and CED3 operates by clustering, then the clustering phenomenon may be explained by adapter-driven homodimerization of monomers. In the case of casp-9, the putative dimerization would be envisaged to activate the protease, and from a mechanistic point of view, though not necessarily from a biologic one, proteolysis is an epiphenomenon. It will be interesting to determine whether similar processing-independent mechanisms for activation apply to the other initiator caspases.