Potent and Selective Nonpeptide Inhibitors of Caspases 3 and 7 Inhibit Apoptosis and Maintain Cell Functionality*

Caspases have been strongly implicated to play an essential role in apoptosis. A critical question regarding the role(s) of these proteases is whether selective inhibition of an effector caspase(s) will prevent cell death. We have identified potent and selective non-peptide inhibitors of the effector caspases 3 and 7. The inhibition of apoptosis and maintenance of cell functionality with a caspase 3/7-selective inhibitor is demonstrated for the first time, and suggests that targeting these two caspases alone is sufficient for blocking apoptosis. Furthermore, an x-ray co-crystal structure of the complex between recombinant human caspase 3 and an isatin sulfonamide inhibitor has been solved to 2.8-Å resolution. In contrast to previously reported peptide-based caspase inhibitors, the isatin sulfonamides derive their selectivity for caspases 3 and 7 by interacting primarily with the S2 subsite, and do not bind in the caspase primary aspartic acid binding pocket (S1). These inhibitors blocked apoptosis in murine bone marrow neutrophils and human chondrocytes. Furthermore, in camptothecin-induced chondrocyte apoptosis, cell functionality as measured by type II collagen promoter activity is maintained, an activity considered essential for cartilage homeostasis. These data suggest that inhibiting chondrocyte cell death with a caspase 3/7-selective inhibitor may provide a novel therapeutic approach for the prevention and treatment of osteoarthritis, or other disease states characterized by excessive apoptosis.

Caspases have been strongly implicated to play an essential role in apoptosis. A critical question regarding the role(s) of these proteases is whether selective inhibition of an effector caspase(s) will prevent cell death. We have identified potent and selective non-peptide inhibitors of the effector caspases 3 and 7. The inhibition of apoptosis and maintenance of cell functionality with a caspase 3/7-selective inhibitor is demonstrated for the first time, and suggests that targeting these two caspases alone is sufficient for blocking apoptosis. Furthermore, an x-ray co-crystal structure of the complex between recombinant human caspase 3 and an isatin sulfonamide inhibitor has been solved to 2.8-Å resolution. In contrast to previously reported peptide-based caspase inhibitors, the isatin sulfonamides derive their selectivity for caspases 3 and 7 by interacting primarily with the S 2 subsite, and do not bind in the caspase primary aspartic acid binding pocket (S 1 ). These inhibitors blocked apoptosis in murine bone marrow neutrophils and human chondrocytes. Furthermore, in camptothecin-induced chondrocyte apoptosis, cell functionality as measured by type II collagen promoter activity is maintained, an activity considered essential for cartilage homeostasis. These data suggest that inhibiting chondrocyte cell death with a caspase 3/7-selective inhibitor may provide a novel therapeutic approach for the prevention and treatment of osteoarthritis, or other disease states characterized by excessive apoptosis.
Significant advances toward understanding the molecular mechanisms of apoptosis have been achieved in recent years, and the role of caspases as integral components of the signal transduction process has been intensely studied (1,2). Eleven human caspases have been identified thus far (3), and a critical question surrounding them is whether inhibition of downstream effector caspase(s) alone is sufficient for promoting cell survival.
Caspase-null mutants have been used to demonstrate the importance of individual family members in murine embryonic development, and cells derived from these caspase-deficient mice have allowed the association of individual caspases with specific morphological changes that occur during apoptosis (4 -8). However, because these animals either die in utero or possess short life spans, their utility in the development of relevant disease models is severely compromised.
The involvement of caspase(s) in apoptosis is more frequently characterized in vitro by measuring enzyme activities and evaluating the effects of inhibitors in cell lysates and tissue extracts. Cell-based studies have been performed using peptide inhibitors with limited cell permeabilities, irreversible peptide inhibitors, and prodrug peptide inhibitors (9 -11). The moderate caspase selectivities associated with these reagents have made it difficult to assess the importance of a specific caspase in apoptosis, and the interpretation of results is often ambiguous. Caspase 3 has been found to be activated in virtually every model of apoptosis (12). It belongs to a subfamily of effector caspases, which also includes caspases 6 and 7. These caspases are activated downstream of initiator caspases such as 8 and 10. Natural substrates of caspase 3 include many proteins involved in cell maintenance and/or repair (13). For example, oligonucleosome fragmentation (DNA laddering) is a characteristic feature of apoptosis and is mediated by caspase-activated deoxyribonuclease (CAD), 1 whose activation is effected by the caspase 3-mediated cleavage of the CAD inhibitor ICAD (14). The availability of a selective inhibitor of caspase 3 would allow the evaluation of the potential for inhibition of apoptosis at the level of an effector caspase.
Osteoarthritis (OA) is a degenerative joint disease histolog-ically characterized by the erosion of articular cartilage and is a potential candidate for an anti-apoptotic therapeutic agent.
Chondrocytes are the only cell type found in cartilage and are specific to this tissue. Elevated levels of apoptosis have been reported in superficial, mid-zone, and hypertrophic chondrocytes in OA cartilage (15,16). In addition, chondrocytes adjacent to OA lesions express high levels of bcl-2, a gene involved in the inhibition of apoptosis (17). Cartilage degradation appears to result from cleavage of matrix proteins by proteolytic enzymes derived from chondrocytes and/or synoviocytes. As OA progresses, the fibrillar collagen network is degraded and chondrocyte cell death increases (18 -21).
In several transgenic models, dysregulated apoptosis has been observed in chondrocytes associated with major structural and developmental abnormalities, thereby suggesting a critical role for the regulation of this process in articular cartilage (22)(23)(24). Increased numbers of apoptotic chondrocytes have also been observed in animal models with surgically induced OA (25). Therefore, the evidence suggesting a role for apoptosis in chondrocytes from normal cartilage is both extensive and convincing. Furthermore, during the development of OA, it has become evident that apoptotic events are dysregulated.
We have identified potent and selective inhibitors of effector caspases 3 and 7 and have evaluated their potential to inhibit apoptosis in two cell-based models. In addition, the caspase 3/7-selective inhibitors blocked apoptosis in an in vitro osteoarthritis model and demonstrated for the first time that caspase 3/7 activity alone is critical in chondrocyte apoptosis and that our inhibitors may lead ultimately to a novel therapeutic strategy. In addition, the compounds inhibited apoptosis in mouse bone marrow neutrophils, which have previously been shown to be dependent upon caspase 3 for apoptosis (5). Our results suggest that inhibition of caspase 3 and/or 7 is essential for the apoptosis of multiple cell types and that inhibition of this activity may represent a novel approach for the treatment of diseases characterized by excessive apoptosis, such as OA.

EXPERIMENTAL PROCEDURES
Inhibitor Synthesis-1-Methyl-5-nitroisatin is commercially available from Aldrich. Sodium 5-isatinsulfonate (Aldrich) was treated with phosphorus oxychloride in sulfalane at 80°C to yield 5-chlorosulfonylisatin (26). Treatment of 5-chlorosulfonylisatin in tetrahydrofuran with an equivalent of amine in the presence of one equivalent of diisopropylethylamine yielded the 5-dialkylaminosulfonylisatin. 5-Dialkylaminosulfonylisatin was alkylated by treatment with sodium hydride in dimethylformamide (27) and reacting the resulting salt with an alkyl halide at temperatures ranging from 25 to 80°C to give 1-alkyl-5alkylaminosulfonylisatin. All compounds were characterized by satisfactory proton NMR and mass spectrometry data.
Protein Supply and Enzyme Inhibition Assays-Caspase 9 was obtained from Chemicon International, Inc. (Temecula, CA).
Expression and Purification of Caspases 3, 4, and 8 -The DNA sequences encoding the caspase 3, 4, and 8 catalytic domains (no prodomain) were amplified by polymerase chain reaction from their respective full-length human caspase cDNAs using custom synthetic oligonucleotide primers and cloned into an Escherichia coli expression vector. The DNA sequences for an N-terminal hexa-His tag and factor Xa cleavage site were designed into the forward primers, and restriction sites were incorporated into the forward and reverse primers for cloning into the pet16b expression vector (Novagen), which was digested with restriction enzymes NcoI and XhoI (New England Biolabs). The amplified caspase DNA's cloned into pet16b were sequenced to verify correct DNA sequences. The clones were then transformed into E. coli expression host LW29(DE3), a derivative of BL21(DE3) containing a chromosomal LacIQ. 2 Single transformants were grown up overnight at 37°C in LB ampicillin (100 g/ml) in shake flasks. The overnight cultures were diluted 1/30 into fresh media and allowed to grow to an A 600 of 0.8 -1.0 before inducing with a final concentration of 1 mM isopropyl-␤-D-thiogalactopyranoside. The cultures were induced for 3-4 h and harvested by centrifugation. Cell pellets were frozen and stored at Ϫ70°C.
For purification, the frozen cells, ϳ300 g of E. coli cells harvested from a 10L bioreactor, were thawed and lysed in 1.5 liters of Buffer A (25 mM HEPES, pH 8.0, 0.1% CHAPS, 10% glycerol, and 10 mM ␤-mercaptoethanol) containing 0.5 M NaCl. After centrifugation, the supernatant was applied to a nickel-nitrilotriacetic acid column, which had previously been equilibrated with Buffer A. N-terminal His-tagged caspases were eluted with 300 mM imidazole in Buffer A. The eluate was dialyzed against Buffer A to remove the imidazole and salt and applied to a ToyoPearl DEAE-650M column. Caspases were eluted with a linear salt gradient, 0 -500 mM NaCl, in Buffer A. A typical yield was ϳ300 mg of each caspase. Purified proteins were characterized by N-terminal sequencing and matrix-assisted laser desorption/ionization mass spectrometry, which confirmed that each caspase was composed of equal molar of p20 and p10 subunits, and molecular weights were as expected from the cDNA sequence.
Expression and Purification of Caspases 1, 2, 6, and 7-For caspases 2, 6, and 7, the full-length DNA sequences containing an N-terminal hexa-His tag were cloned and expressed in E. coli as described above. After lysis of the cells, the presence of proteolytically processed enzyme was detected in the supernatant by Western blot with polyclonal antibody directed against the N-terminal peptide of each p20 subunit. The supernatant was applied to a ToyoPearl DEAE 650 M column and the active catalytic domain (p10/p20 complex) was eluted with a salt gradient as described above. Partially purified p10/p20 complex of each caspase was used for compound screening assays without further purification. For caspase 1, the N-terminal hexa-His-tagged proenzyme, which was expressed as an insoluble pellet, was solubilized in 8 M urea in buffer B (50 mM NaPO 4 , pH 7.5, 0.5 M NaCl) and applied to an nickel-nitrilotriacetic acid column. Procaspase 1 was refolded in the nickel-nitrilotriacetic acid column by washing the column with Buffer B and eluted with 300 mM imidazole in Buffer B. The proenzyme was desalted using a Sephadex G-25 column equilibrated with Buffer A and adjusted to 5 mM dithiothreitol. Activation of proenzyme to p10/p20 complex was achieved by concentrating to 4 mg/ml and incubating 2 days at 4°C. The activated protein mixture was further fractionated using Superdex 200. Fractions containing caspase 1 activity were pooled. SDS-polyacrylamide gel electrophoresis showed that the active enzyme was Ͼ90% pure.
Crystallography-Co-crystals of the complex between recombinant human caspase 3 and inhibitor 4 were grown from 4-l hanging drops formed by mixing equal volumes of protein (10 mg/ml in 20 mM HEPES, pH 7.0) and reservoir solution. The drops equilibrated at room temperature (21°C) through the vapor phase against 500 l of a reservoir solution that contained 15-18% polyethylene glycol 6000, 0.1 M sodium citrate, pH 5.9, 20 mM L-cysteine, and 5% glycerol. The crystals belong to the space group I222 with unit cell dimensions, a ϭ 67.2 Å, b ϭ 83.3 Å, c ϭ 96.0 Å. Diffraction data were collected at the X12B beamline of the National Synchrotron Light Source (Brookhaven National Laboratory, Upton, NY) from a single flash-frozen crystal with its largest dimension of less than Ϸ20 m. The structure of the caspase 3 ϩ inhibitor 4 complex was determined by difference Fourier using the coordinates of the isomorphous caspase 3 complex with Ac-DEVDdimethylbenzoate previously determined and refined to 2.0 Å (data not shown). Initially, the difference electron density maps showed a 3 peak in the active site. The interpretation of the map was aided by 2 SmithKline Beecham, unpublished data. assuming a proposed mode of inhibition in which the thiolate group of Cys 163 is covalently linked to C-3 of the inhibitor and the crystal structure of epi-5. The model was refined in X-PLOR (28,29), and O (30) was used for manipulation of the model. The average B-factor for protein atoms is 16.2 Å 2 and for non-protein atoms is 30.0 Å 2 . The model of the asymmetric unit includes residues 29 -173 and 185-277 of caspase 3, 45 water molecules, and the inhibitor molecule 4.
Neutrophil Apoptosis Assay-Mouse bone marrow cells were treated with 10 g/ml cycloheximide (CHX) at 2 ϫ 10 6 cells/ml in RPMI medium (Life Technologies, Inc.) containing 10% fetal bovine serum (FBS) (Hyclone Laboratories, Inc., Logan, UT), penicillin (10 units/ml), and streptomycin (50 ng/ml) (Life Technologies, Inc.) in 5-ml round-bottomed polypropylene tubes for 3 h at 37°C; inhibitors were added at the same time as CHX. Approximately half of the cells were used for evaluation of neutrophil viability by FACS analysis, and the remainder was used for a DNA ladder assay as described previously (8). To determine neutrophil viability following CHX treatment, cells were washed and stained with anti-Gr-1-fluorescein isothiocyanate (PharMingen, San Diego, CA), and propidium iodide and analyzed by flow cytometry. This procedure was used also to assess propidium iodide staining of chondrocytes. Samples were analyzed by FACS. For the DNA ladder assay, DNA released into the cytoplasm was extracted, run on a 1% agarose gel, and stained with ethidium bromide (8).
Bovine chondrocytes were isolated from the articular cartilage of histologically normal carpal metacarpal joints of calves (0 -3 months of age). Chondrocytes were isolated by sequential enzyme digestion as follows. Cartilage (2 joints) was incubated in 50 ml of 0.2% testicular hyaluronidase (type I-S from bovine testes) in DMEM without serum for 20 min at 37°C. The hyaluronidase was then removed, and the tissue was incubated in 50 ml of 0.25% Pronase E in DMEM without serum for 20 min at 37°C. The Pronase E was removed, and the cartilage was then incubated with 0.2% Clostridium histolyticum collagenase D (Roche Molecular Biochemicals) in DMEM plus 10% FBS overnight at 37°C. The digested tissue was filtered through a 100-m cell filter (Falcon), and the cells were pelleted by centrifugation (1200 ϫ g for 10 min). The primary bovine chondrocytes were grown routinely in Ham's F-12 medium supplemented as above plus antibiotic-antimycotic (Life Technologies, Inc.). All cells were cultured at 37°C in a humidified atmosphere of 95% air, 5% CO 2 . Upon reaching confluence, the cells were subcultured using trypsin-EDTA (Life Technologies, Inc.).
Chondrocyte Cell Death ELISA-Human chondrocyte cells (C20/A4) (31) or primary bovine chondrocytes were grown in 24-well plates at 20,000 cells/well overnight and then treated with either camptothecin alone (4 g/ml) (Biomol) or camptothecin (4 g/ml) and Z-VAD-FMK (50 M) or isatin sulfonamides for 24 h. Cell lysates were prepared by combining the cells from the monolayer with the cells, which detached during the treatment period. Cells floating in the media were pelleted by centrifugation for 5 min at 1,000 ϫ g, resuspended in the manufacturer's lysis buffer, and then added back to the monolayer, and the total cell population was brought to 500 l with lysis buffer. The samples were lysed for 30 min at 4°C and then centrifuged for 10 min at 14,000 ϫ g to clarify the lysates. Samples were evaluated for mono-and oligonucleosome DNA fragment formation in the cell death ELISA following the manufacturer's protocol (Roche Molecular Biochemicals).
Chondrocyte Promoter Reporter Assay-Human chondrocyte cells (C20/A4) (31) were stably transfected, and primary bovine chondrocytes were transiently transfected with a construct (pGL3-Basic-NEO r -COL2A1) in which the regulatory sequence (Ϫ577 to ϩ3426) (32) of the type II collagen (COL2A1) gene is driving a luciferase reporter gene. To prepare the expression vector, the neomycin-resistant gene (NEO r ) was inserted at the BamHI site of the pGL3-Basic vector (Promega). The 4.0-kb COL2A1 fragment was inserted at the XhoI/MluI sites and transfected into the C20/A4 cells and primary bovine chondrocytes using the calcium phosphate transfection method (Invitrogen). For stable cell line generation, cells were grown overnight in 10-cm dishes at 1 ϫ 10 6 cells/dish and transfected with 20 g of pGL3-Basic-NEO r -COL2A1. Single clones were selected by growth in Geneticin (G418) (Life Technologies, Inc.) at 400 g/ml for 2-3 weeks. Positive cells grown from single clones were maintained in media containing 200 g/ml G418. Transiently transfected primary bovine cells and stably transfected C20A/4 cells were seeded in 24-well plates at 25,000 cells/ well overnight and then treated with either camptothecin alone (4 g/ml) or camptothecin (4 g/ml) with Z-VAD-FMK (50 M) or isatin sulfonamides for 8 h. Cells were then washed with phosphate-buffered saline and lysed with 100 l/well of the manufacturer's lysis buffer (Promega) for 15 min while rocking at room temperature. Lysates were centrifuged for 30 s at 14,000 ϫ g, and the clear lysate was transferred to a new tube prior to reporter analysis. Samples (20 l) were transferred to a 96-well luminescence detection plate and reacted with 100 l of the luciferase assay reagent (Promega), which was injected by a Microlumat LB96P luminometer (Wallac) in order to measure luciferase activity.
Statistical Analysis-For each parameter, differences between groups were assessed by unpaired Student's t test using Excel (Microsoft Corp.). Differences with a value of p Ͻ 0.05 were considered significant.

RESULTS
High throughput screening for inhibitors of recombinant human caspase 3 resulted in the identification of 5-nitroisatin (1, Fig. 1), which inhibited the enzyme with a K i(app) of 0.5 M ( Table I). The potency of this compound and the non-peptidic nature of its structure made it an attractive starting point for a drug discovery effort.
Using the previously determined x-ray structure of caspase 3 (33), an initial binding hypothesis was developed. Our model was based on the assumption that a tetrahedral intermediate between the catalytic cysteine residue and the ketone carbonyl group of inhibitor 1 is formed (34). The mode of binding suggested that extension of the molecule from the 5-position would allow access to the S 2 -S 4 (35) regions of the active site. In addition, there are metabolic stability issues associated with nitroaromatics. Therefore, a chemical functionality that would allow for the facile incorporation of chemical diversity was sought as a replacement for the 5-nitro group (36). Evaluation of a series of groups resulted in the identification of isatin sulfonamide 2, which retained significant activity against caspase 3 (K i(app) ϭ 1.4 M). Relative to lead 1, isatin sulfonamide 2 possessed much improved selectivity for caspases 1, 3, and 7 (Table I). Preparation of an extensive series of compounds with variation of groups about the sulfonamide functionality and isatin nitrogen culminated in the identification of isatin sulfonamides 3 and 4, 60 and 15 nM inhibitors of caspase 3, respectively. In addition, these inhibitors exhibited 100-fold or greater selectivity for the highly homologous caspases 3 and 7 (13) versus all other family members except caspase 9. Against this caspase, the selectivity ranged from 10-to 50-fold.
The mechanism of inhibition versus caspase 3 was demonstrated to be reversible and competitive with respect to the substrate Ac-DEVD-AMC. In addition, no evidence for time dependence was observed. Reversibility was assessed by preincubating recombinant human caspase 3 with inhibitor 3. Samples of this mixture at several time points up to 2 h were diluted 40-fold into assay buffer, and enzyme activity was measured with the substrate Ac-DEVD-AMC. The observed inhibition at all time points did not vary significantly over time and was consistent with that expected for a freely and rapidly reversible inhibitor. Isatin sulfonamide 3 was competitive versus caspase 3 with a K i of 83 nM (data not shown).
The basis of this unique selectivity for caspases 3 and 7 was evaluated by examining the protein sequence alignments of the caspases against which the compounds were screened. Since these inhibitors were initially modeled to interact with the catalytic cysteine residue, we focused on those protein residues proximal to this region of the active site. This analysis revealed that the observed selectivity was likely due to three hydrophobic residues in the S 2 pocket (Tyr 204 , Trp 206 , and Phe 256 ) that are unique to caspases 3 and 7 (Table II). In addition, the sequence alignments of caspases 5, 10, and 13, which were not available for testing directly, suggest that the isatin sulfonamides may not inhibit these family members as potently as caspases 3 and 7. The residues corresponding to Tyr 204 and Trp 206 of caspase 3 are Val/Trp (caspases 5 and 13) and Tyr/Phe (caspase 10). The residue corresponding to Phe 256 of caspase 3 for each of these caspases cannot be determined because of variability in loop size in this portion of the protein.
A 2.8-Å resolution x-ray co-crystal structure of the complex between recombinant human caspase 3 and isatin sulfonamide 4 was obtained. Data collection and statistics are given in Table  III. The electron density map showed good supporting density for all atoms of the isatin framework and pyrrolidine ring but weak density for the phenoxymethyl side-chain ( Fig. 2A). Compared with the previously determined structure of caspase 3 (33), there are no significant differences in conformation of the amino acid side chains of the protein.
As predicted by our initial hypothesis, the x-ray co-crystal structure reveals that a tetrahedral intermediate is formed between the catalytic cysteine thiolate and the isatin ketone carbonyl group (Fig. 2B). The S 2 pocket is involved in extensive hydrophobic contacts with the pyrrolidine ring of the inhibitor and supports our initial proposal of the importance of this hydrophobic pocket for conferring specificity to the isatin sulfonamides. As predicted, inhibitor 1, which does not possess the pyrrolidine ring, exhibits only moderate selectivity between the caspases.
The S 1 subsite of the caspases confers high selectivity for the cleavage of substrates possessing a P 1 aspartic acid (37)(38)(39). In the x-ray co-crystal structure, this subsite is occupied only by a water molecule. Despite the absence of inhibitor binding in the S 1 subsite, a representative set of isatin sulfonamides (inhibitor 3 and two closely related analogs) exhibited little inhibition of other cysteine proteases such as cathepsins B, L, K, and S (IC 50 values Ͼ Ͼ5 M; data not shown) (40). In addition, these compounds exhibited little inhibition of human recombinant calpain I, a protease implicated in apoptosis (41) (Ͻ20% inhibition at 50 M) (data not shown).
During peptide substrate hydrolysis, the "oxyanion pocket" formed by the backbone amide NHs of Cys 163 and Gly 122 is involved in stabilization of the negatively charged oxygen atom of the tetrahedral intermediate. In the caspase 3/inhibitor 4 structure, the oxygen of the tetrahedral intermediate and the amide carbonyl oxygen are within hydrogen bonding distance of Cys 163 NH and Gly 122 NH, respectively. There is a hydrophobic and/or aromatic interaction between the edge of Tyr 204 with one face of the bicyclic isatin core, which likely contributes to the binding of the inhibitor. The phenoxymethyl side chain fits the data best when positioned in the shallow S 3 pocket lined by residues Ser 65 , Arg 207 , and Ser 209 . However, this density is weak (Fig. 2A), and a subsequent co-crystal structure with a closely related analog of inhibitor 4 shows strong electron den-   sity for this side chain on the surface of Phe 256 . 3 In addition, in the caspase 3/inhibitor 4 structure, Ser 58 of a crystallographic symmetry-related molecule occludes this region about Phe 256 , and likely influences the observed positioning of the inhibitor side chain.
We next evaluated the ability of isatin sulfonamide 4 to functionally inhibit apoptosis in a caspase 3-dependent cellbased model of apoptosis. Apoptosis as measured by dye exclusion in mouse bone marrow neutrophils derived from caspase 3 ex3Ϫ/Ϫ mutants had previously been shown to be dependent upon caspase 3 (5). We therefore treated wild-type neutrophils with cycloheximide to accelerate apoptosis, and cell viability in the presence and absence of inhibitor was measured after 3 h. A dose-dependent decrease in apoptosis as measured by double labeling for Gr-1 and propidium iodide exclusion following treatment with inhibitor 4 was observed. At 25 M, there was an 80% reduction in cell death (Fig. 3A). In addition, there was a complete abrogation of DNA laddering at 25 M (Fig. 3B). Since apoptosis is dependent upon caspase 3 in these cells (5), the difference between in vitro enzyme (IC 50 ϭ 30 nM) and cellular apoptosis (IC 50 Ϸ 10 M) inhibition potencies is suggestive of protein binding and/or limited cell penetration of the inhibitor.
To support this proposal, we demonstrated that the in vitro inhibition of recombinant human caspase 3 by isatin sulfonamide 4 in 30% Jurkat cell cytosol/buffer resulted in a 50-fold attenuation of inhibitor potency (data not shown). Full inhibitor activity was recovered upon dilution of the inhibitor/cytosol mixture with buffer (inhibitor potency is similar to that observed in 100% buffer), indicating that it is a reversible binding event between inhibitor and cytosolic constituents. The nature and/or target(s) of this interaction is unknown at this time. Despite the attenuation in inhibitor potency observed in neutrophils, these results clearly establish the utility of inhibitor 4 as a tool to study the importance of caspases 3 and 7 in cellular apoptosis.
The ability of these compounds to inhibit apoptosis was next examined in chondrocytes. Due to the scarcity of primary human chondrocytes, we used primary bovine chondrocytes and an immortalized human chondrocyte cell line C20/A4 (32) to investigate the effect of inhibitor 4 and a related compound with much lower activity, isatin sulfonamide 5, on chondrocyte apoptosis. The residues of caspase 3 with which the inhibitor interacts are identical between the bovine and human enzymes. Camptothecin-induced apoptosis, determined by cell death ELISA, was blocked completely by treatment with the non-selective, irreversible peptide inhibitor Z-VAD-FMK (10) at 50 M (Fig. 4, A and B). Treatment of C20/A4 cells and primary bovine chondrocytes with the caspase 3/7-selective inhibitor 4 also blocked cell death (IC 50 ϭ 6 M). Compound 4 was nontoxic at all concentrations tested as measured by trypan blue exclusion (results not shown). In addition, propidium iodide staining was also inhibited by isatin sulfonamide 4 (Fig. 5). The structurally related analog 5, which is a weak inhibitor of caspase 3 (Table I), failed to block cell death at concentrations up to 25 M (Fig. 4, A and B).
To examine the consequence(s) of inhibiting chondrocyte apoptosis, the activity of the transfected type II collagen promoter-luciferase reporter was investigated in stably transfected C20/A4 cells and transiently transfected primary bovine chon- drocytes. Following the induction of apoptosis, type II collagen promoter activity markedly decreased, suggesting a shutdown of chondrocyte matrix production (Fig. 6, A and B). The nonselective peptide inhibitor Z-VAD-FMK was able to significantly inhibit the loss of type II collagen promoter activity observed after induction of apoptosis. Similarly, the caspase 3/7-selective inhibitor 4 blocked the decrease of type II collagen promoter activity (Fig. 6, A and B). No effect was shown by compound 5, a significantly less active compound against the isolated enzymes. These data suggest that the caspase 3/7selective inhibitors block apoptosis in vitro yet maintain the transcriptional activity of the chondrocyte-specific type II collagen promoter. DISCUSSION The isatin sulfonamides are the first reported examples of potent and selective inhibitors of caspases 3 and 7 and repre-sent a novel strategy for the development of active agents which block apoptosis. In contrast to currently available agents, these compounds provide the opportunity to evaluate the inhibition of effector caspases in cell-based models of apoptosis. Previous inhibitor studies have suggested that binding in the S 1 , S 3 , and S 4 subsites is critical for potent and selective inhibition of caspases. Binding to the S 1 subsite confers selectivity for caspases, while binding to the S 3 and S 4 subsites is generally believed to be critical for selectivity between caspases (13). The x-ray crystal structure of the complex between inhibitor 4 and caspase 3 shows minimal inhibitor interactions with these subsites. Instead, selectivity is obtained via extensive hydrophobic contacts between the pyrrolidine ring of the inhibitor and residues Tyr 204 , Trp 206 , and Phe 256 of the S 2 hydrophobic pocket. Thus, the isatin sulfonamides are the first examples of potent caspase inhibitors that achieve their selectivity through interaction with the S 2 subsite.
Although our initial goal was to identify a selective inhibitor of caspase 3, the effort yielded inhibitors with dual selectivity for caspases 3 and 7. However, caspases 3 and 7 are frequently activated at similar time points during the apoptotic signal transduction cascade and may play non-redundant roles in the processing of downstream substrates. Since the isatin sulfonamides offer the ability to block both of these effector caspases, we believe these compounds may possess advantages over a caspase 3-selective inhibitor.
The selectivity of isatin sulfonamide 4 for caspases 3 and 7 permitted the demonstration that inhibition of these caspases results in the blocking of cell death in two apoptotic cell models, one of which has been shown to previously to be absolutely caspase 3-dependent (4,5). Inhibition of mouse bone marrow neutrophil apoptosis with caspase 3/7-selective inhibitors is in agreement with the previous report from caspase 3 ex3Ϫ/Ϫ mutant mice, which have been shown to be resistant to apoptotic insult (4,5). This result suggests that the mechanism by which these compounds exert their anti-apoptotic activity is by inhibition of caspase 3.
We next showed that chondrocyte apoptosis was also blocked by the caspase 3/7-selective inhibitors. The extracellular signals that stimulate chondrocyte apoptosis in vivo and the intracellular pathways triggered that ultimately result in DNA fragmentation and cell death are not well understood. Chondrocyte apoptosis can be triggered in vitro by multiple stimuli, which may reflect local physiological conditions during the development and/or progression of OA, e.g. enhanced local cytokine production (TNF-␣) and reduced growth factor concentrations (low serum) (42). Increased cytokine levels and reduced growth factor environment have been reported in aged/ diseased articular joints (42). Interleukin-1 and TNF-␣ are cytokines that have been proposed to play a role in both inflammatory and non-inflammatory joint diseases such as rheumatoid arthritis and OA and have both been reported to stimulate chondrocyte apoptosis (15,43). In addition, we have also shown that multiple stimuli, including TNF-␣ and growth factor withdrawal, induce caspase 3 activity in human and bovine chondrocytes (50).
Chondrocytes are responsible for the production and maintenance of the extracellular matrix, which functions to allow optimal fluid joint articulation. Therefore, inhibition of cell death alone may not be sufficient for a beneficial effect on the articular cartilage. Ideally, an anti-apoptotic agent would be required to prevent cell death yet maintain the chondrocyte as a matrix producing cell. Inhibition of apoptosis with the caspase 3/7 inhibitors does just that. In our studies, prevention of apoptosis by inhibition of caspase 3/7 activity resulted in maintenance of the transcriptional activity of the COL2A1 promoter. Maintenance of type II collagen promoter activity suggests that expression, of this major cartilage matrix protein, would be maintained in the presence of the inhibitors. Therefore, in a catabolic state an anti-apoptotic agent may allow the cells, which were destined to die, to survive and more importantly to continue to synthesize and deposit new matrix. A major issue in treating chronic degenerative diseases will be to prevent the unwanted death of cells while not allowing the proliferative diseases, such as cancer or rheumatoid arthritisassociated synovial hypertrophy (44,45), to flourish. Therefore, understanding the role of the specific apoptotic proteases and their aberrant function in these highly proliferative cells is an important challenge. If cancer cell survival is indeed enhanced by non-selective caspase inhibition, then therapeutic approaches will need to be selective for the target cell, so that general cellular hypertrophy is not stimulated. For inhibition of apoptosis to be therapeutically beneficial, rescued, "nondead" cells must function in a normal manner. A report by Davidson and Steller (46) described residual cell function in Drosophila retinal degeneration mutants, which have a condition similar to human retinitis pigmentosa. The cell survival protein p35 blocked apoptosis in the mutant photoreceptor cells, and the flies retained more visual function than mutant, untreated flies. This model suggests that late stage (mature) anti-apoptotic therapeutic strategies may be effective against chronic degenerative diseases.
Our data suggest that inhibiting chondrocyte cell death with caspase 3/7-selective inhibitors may provide a novel therapeu-FIG. 6. The maintenance of type II collagen promoter activity after apoptotic insult in the presence of the caspase inhibitors and Z-VAD-FMK. The C20/A4 chondrocyte cells stably transfected (A) and primary bovine chondrocytes transiently transfected (B) with a construct in which the regulatory sequences (Ϫ577 to ϩ3426) of the type II collagen (COL2A1) gene are driving luciferase reporter gene expression. Cells were grown in serum-containing medium as a control (control). Cells were treated with 4 g/ml CAM or 4 g/ml CAM plus varying concentrations of compound 4 or 5 for 24 h. Expression of the COL2A1 gene was measured by luciferase activity measurements using a luminometer. Results are expressed as relative light units (RLU). Results are mean Ϯ S.D. (n ϭ 3) (* ϭ p Ͻ 0.02, compared with CAM treatment alone). tic approach for the prevention and treatment of OA or other disease states characterized by excessive apoptosis that involves the caspase 3/7 pathway.