Crystal Structure of Caspase-2, Apical Initiator of the Intrinsic Apoptotic Pathway*

The cell death protease caspase-2 has recently been recognized as the most apical caspase in the apoptotic cascade ignited during cell stress signaling. Cytotoxic stress, such as that caused by cancer therapies, leads to activation of caspase-2, which acts as a direct effector of the mitochondrion-dependent apoptotic pathway resulting in programmed cell death. Here we report the x-ray structure of caspase-2 in complex with the inhibitor acetyl-Leu-Asp-Glu-Ser-Asp-aldehyde at 1.65-Å resolution. Compared with other caspases, significant structural differences prevail in the active site region and the dimer interface. The structure reveals the hydrophobic properties of the S5 specificity pocket, which is unique to caspase-2, and provides the details of the inhibitor-protein interactions in subsites S1–S4. These features form the basis of caspase-2 specificity and allow the design of caspase-2-directed ligands for medical and analytical use. Another unique feature of caspase-2 is a disulfide bridge at the dimer interface, which covalently links the two monomers. Consistent with this finding, caspase-2 exists as a (p19/p12) 2 dimer in solution, even in the absence of substrates or inhibitors. The intersubunit disulfide bridge stabilizes the dimeric form of caspase-2, whereas all other long prodomain caspases exist as monomers in solution, and dimer formation is driven by ligand binding. Therefore, the central

Programmed cell death (apoptosis) is indispensable for all multicellular organisms in normal development and tissue homeostasis (1,2). Excessive apoptosis leads to disorders, including neurodegenerative diseases, whereas diseases such as cancer arise from insufficient apoptosis (reviewed in Refs. 3 and 4). The apoptotic process involves an intricate cascade of events, which is classified as occurring in two major pathways. The extrinsic pathway relies on a cell surface stimulus and is initiated by binding of death ligands to death receptors or by granzyme B. The intrinsic pathway occurs as a consequence of cellular stress and is mediated by the release of cytochrome c from the mitochondrion. Because the mitochondrion-dependent pathway is also employed as an amplification loop for the death receptor pathway, the terms "extrinsic" and "intrinsic" are used here strictly according to the origin of the death signals, indicating extracellular or intracellular initiation of apoptosis, respectively.
Caspase-2 was discovered as the first mammalian apoptotic caspase (5). Recent studies revealed that caspase-2 is engaged as an initiator in both the extrinsic and the intrinsic pathways of apoptosis. Additionally, caspase-2 serves in neuronal cells both as the default initiator and the default executioner caspase (reviewed in Ref. 6). The first function described for caspase-2 is that of an initiator of the extrinsic pathway, thereby mediating external death signals to the cellular apoptotic machinery. Upon binding of a death ligand to the death receptors, the adaptor molecule RAIDD (RIP-associated Ich-1/ CED-3 homologous protein with death domain, also known as CRADD) is recruited intracellularly, followed by binding of procaspase-2 via its N-terminal CARD 1 domain (7). In this complex, caspase-2 is activated via proteolytic cleavage, releasing the active protease to the cytoplasm. Unlike all other initiators, caspase-2 is completely inactive toward other caspase zymogens (8). Instead, the protease cleaves cytosolic Bid, which leads to an increase of the permeability of the outer mitochondrial membrane. As a consequence, cytochrome c is released to the cytoplasm and assembles with procaspase-9 and Apaf-1 to form the apoptosome, leading to caspase-9 activation and eventually to programmed cell death. More recently, caspase-2 was also found to be the apical initiator caspase of the intrinsic apoptotic pathway, which is activated during cell stress (8 -11). Both the extrinsic and the intrinsic pathways engage the same proteolytic cascade, but the mechanism of caspase-2 activation is different. Unlike in the extrinsic pathway, the molecular basis for the activation of caspase-2 as a consequence of cell stress remains obscure. The importance of caspase-2 in the intrinsic pathway, however, becomes increasingly evident: caspase-2 activation is crucial in chemotherapy-induced cell death, as shown for acute myeloid and acute lymphoblastic leukemia (12), caspase-2 is activated in response to the chemotherapeutic drug etoposide (10) and can mediate apoptosis in human breast cancer cell lines (13). Intriguingly, caspase-2 is most closely related to the Caenorhabditis elegans CED-3 of all mammalian caspases and is the most evolutionarily conserved caspase within the family (14). However, based on the lack of a dramatic phenotype in the caspase-2 null mouse (15), it is unlikely that the protease is the sole mediator of cell stress signals. Probably, cell stress-dependent and developmental * This work was supported by the Swiss National Science Foundation and the Baugartenstiftung (Zurich, Switzerland). 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.
The apoptosis can be ignited by caspase-2 and by other initiators, which act in a redundant manner (16).
All caspases show a high degree of specificity with an absolute requirement for cleavage after an aspartic acid residue and a recognition sequence of normally four consecutive amino acids N-terminal to the cleavage site. Using a combinatorial approach they have been classified into three distinct specificity groups: group I (caspases-1, -4, and -5) with preference for Trp-Glu-His-Asp, group II (caspases-2, -3, and -7) with a preference for Asp-Glu-X-Asp and group III (caspases-6, -8, -9, and -10) with a preference for (Leu/Val)-Glu-X-Asp (17). This characteristic specificity is crucial to the apoptotic process, because it involves cleavage of a particular group of proteins in an ordered manner rather than indiscriminate proteolysis (18). Given their pivotal role in the regulation of apoptosis, the caspases are important therapeutic targets, and it is evident that further investigation of caspase specificity and its relationship to structure is important for understanding cell death. A number of neurological diseases are connected to the caspase apoptotic pathway (19,20), and some small caspase inhibitors have been shown to block the effects of neuronal cell death (21)(22)(23)(24). A controlled activation of caspases, as desirable for cancer treatment, is more difficult to achieve. A possible strategy could involve the design of compounds that specifically promote dimerization of monomeric caspase zymogens.

EXPERIMENTAL PROCEDURES
Caspase-2 Production-Expression and purification were based on protocols previously described for a set of caspases (17,25,26), however, several modifications were introduced. The cloned genes for the large p19 (Met-Ala-Asn 150 -Asp 316 ) and the small p12 subunit (Met-Ala 331 -Thr 435 ) were inserted in the NcoI/BamHI sites of pET11d plasmids (Novagen). For separate expression of both subunits, Escherichia coli BL21-CodonPlus(DE3)-RIL cells (Stratagene), containing one of the two plasmids were grown to a density of A 600 ϭ 0.6 at 37°C in a 0.5-liter Luria-Bertani medium. Isopropyl-1-thio-␤-D-galactopyranoside (1 mM) was added, and the culture was shaken at 37°C for 2.5 h. Under these conditions, the subunits were localized in the inclusion body fraction where they constituted more than 90% of the total protein. Cells were harvested, washed in 10 mM Tris, pH 8.0, 1 mM EDTA, and broken using a French press. Inclusion bodies were sedimented, washed, and solubilized as described previously (17,26) and stored at Ϫ80°C. For refolding, equimolar amounts of each subunit were combined and rapidly diluted to a final concentration of ϳ100 g of subunit/ml. As refolding buffer, 200 ml of 100 mM Hepes, pH 7.5, 20% sucrose, and 10 mM DTT was used. Refolding was achieved overnight at 20°C under nitrogen. Precipitate was removed by centrifugation (5,000 ϫ g, 20 min), and the supernatant was concentrated 25-fold using an Amiconstirred cell (membrane: PLGC 10,000). To reduce the salt concentration, the protein solution was diluted 1:4 with 7.5% (w/v) sucrose, 4 mM DTT prior to chromatography. Active enzyme was separated from inactive subunits by anion-exchange chromatography using a MonoQ 5/5 column (Amersham Biosciences) equilibrated with 20 mM Hepes, pH 7.35, 10% (w/v) sucrose, 4 mM DTT. A further 4 mM DTT was added to all fractions immediately after chromatography. Fractions containing pure and active caspase-2 were pooled and concentrated by ultrafiltration to a final concentration of 1.2-2 mg/ml (yield, 5-8 mg/liter medium). The protease was subsequently inhibited with a 2-fold molar excess of Ac-LDESD-cho (Calbiochem).
Analytical Methods-Protein concentration was determined by absorbance spectroscopy at 280 nm using the theoretical extinction coefficient (⑀ 280 ϭ 17,930 M Ϫ1 cm Ϫ1 ) and with the Advanced Protein Assay from cytoskeleton. Amino acid analysis was performed using an amino acid analyzer (AminoQuant, Hewlett-Packard) according to the manufacturer's recommendations. All three methods yielded comparable results. SDS-PAGE was performed on 15% or 18% acrylamide gels in a Tris/glycine system. Analytical size exclusion chromatography was performed at 4°C using an Amersham Biosciences Superdex 200 HR10/30 column equilibrated with 20 mM Hepes, pH 7.5, 150 mM NaCl, 10% (w/v) sucrose. The enzyme (concentration: 8 -10 M; active and irreversibly inhibited with Ac-DEVD-cmk) was chromatographed in the same buffer at a flow rate of 0.5 ml/min. The elution volume was compared with a standard curve of the column calibrated with molecular weight markers and with active and inhibited caspase-8 using the same elution buffer described above.
Enzymatic Assay-The enzymatic activity of caspase-2 was determined using Ac-DEVD-amc and Ac-LDESD-amc in 100 mM Mes, pH 6.5, 10% polyethylene glycol 400, 0.1% Chaps, 10 mM DTT (27). Fluorescence was detected at ex ϭ 360 nm and em ϭ 465 nm for 20 min at room temperature using an HTS 7000 Plus Bio Assay Reader (PerkinElmer Life Sciences). For relative (V max /K m ) determinations, all substrate concentrations were first measured by amino acid analysis and then adjusted to a final concentration of 17.6 M; the active enzyme was diluted to a final concentration of 200 nM.
Mass Spectrometry-Prior to mass spectrometry, p19 of the uninhibited enzyme was separated from p12 by high-performance liquid chromatography using a C8 reverse phase column (Vydac), 0.05% trifluoroacetic acid in H 2 O as buffer A, 0.045% trifluoroacetic acid in 80% acetonitrile as buffer B, and a gradient from 20 to 100% B. The fractions were directly analyzed by mass spectrometry using a Bruker Biflex MALDI spectrometer and ␣-cyano-4-hydroxycinnamic acid as matrix.
To examine the reduced species, 50 mM Tris, pH 7.5, and 4 mM Tris(2carboxyethyl)phosphine hydrochloride was added prior to the analysis.
Crystallization and Data Collection-Crystals of the caspase-2/ LDESD complex were obtained by sitting drop vapor diffusion at 4°C by mixing 1.5 l of protein solution (in 20 mM Hepes, pH 7.5, 150 mM NaCl, 10% (w/v) sucrose, 8 mM DTT; protein concentration: 1.2-2.0 mg/ml) with 1 l of reservoir solution consisting of 25% (w/v) polyethylene glycol 6000, 100 mM Mops, pH 7.0. Crystals (ϳ600 ϫ 50 ϫ 50 m) grew within 1 week at 4°C. For data collection, crystals were frozen in the nitrogen stream after a short soak in 15% ethylene glycol, 30% polyethylene glycol 6000, 100 mM Mops, pH 7.0. X-ray data were collected from a single crystal at 90 K using monochromatic synchrotron radiation ( ϭ 0.9184 Å) and a MarCCD detector. This work was performed at the Swiss Light Source, Paul Scherrer Institut, Villigen, Switzerland. The crystals belong to space group P2 1 2 1 2 1 with unit cell dimensions of a ϭ 63.65, b ϭ 96.58, and c ϭ 97.55 Å. Data reduction was carried out with the programs DENZO and Scalepack (28). Due to anisotropic diffraction, the data were cut at 1.65 Å. The criterion for the high resolution limit was an I/(I) of Ͼ2.0 for the last resolution shell. Data statistics are given in Table I.
Structure Solution and Refinement-The structure was solved by molecular replacement with the program AMoRe (29) using a (p17/p12) 2 polyserine model of caspase-3 (1CP3) as template (30). Structure refinement was performed using the program CNS (31). No NCS restraints were used during the refinement of the structure. Electron density was visible for the majority of the amino acids. For dimer A, 8 N-terminal residues of p19 and 6 N-terminal residues of p12 and the C-terminal residue of both chains could not be modeled. For the p19 and p12 subunit of dimer B, 7 and 6 N-terminal residues could not be fitted, respectively.

RESULTS
Structure of Caspase-2-During activation, the single-chain procaspase-2 is proteolytically processed to active caspase-2 consisting of one p19 and one p12 subunit, hereafter referred to as a monomer. The structure of caspase-2 comprises two copies of the p19/p12 monomer that contain residues 8/9 -168 and 207-304/305, respectively (Fig. 1a). The root mean square deviation of the C␣ distance of superimposed monomers, including all 257 C␣ atoms is 0.30 Å. Therefore, both monomers can be considered to be identical. The core of the enzyme is formed by a 12-stranded ␤-sheet, which is surrounded by a total of 14 ␣-helices. The overall fold of caspase-2 is similar to that of other caspases for which structures have been determined previously (30,(32)(33)(34)(35)(36)(37)(38)(39). This is illustrated by a superposition of the C␣ chains with caspase-3, another representative of the specificity group II caspases, and caspase-9, the closest relative of caspase-2 based on sequence identity (Fig. 1b). Interestingly, caspase-2 shows an extraordinary long structured N terminus of the large p19 subunit. This tail structure contains a short helix and is so far unique to the CARD domain-containing caspases-2 and -1. In caspase-1, a surface disulfide bridge was observed to be stabilizing this extension, but its formation depends on the crystallization conditions. In caspase-2, there are also two cysteine residues at exactly the same position; however, they do not form a disulfide bridge, probably due to the reducing conditions maintained during purification and crystallization. As with other caspases, the dimer is mainly formed by interactions between residues of the central strands of the two p12 subunits (Fig. 1a). The additional interactions found in caspase-9, provided by a 7-residue loop inserted between helices 3 and 3a (36), are not present in caspase-2. In this region caspase-2 superimposes well with caspase-3 (Fig.  1b). Aside from the intersubunit disulfide, the absence of this FIG. 1. Structure of the caspase-2-Ac-LDESD-cho complex. a, ribbon diagram of the (p19/p12) 2 dimer. The p19 and the p12 chains are colored blue and gray, respectively. The Ac-LDESD-cho inhibitor is covalently bound to the activesite cysteine Cys 285 (caspase-1 numbering throughout) and is depicted as a stick model and colored according to atom type. The N and C termini of one monomer are indicated. The non-crystallographic 2-fold symmetry axis relating the two p19/p12 units is perpendicular to the plane of paper. The central disulfide bridge is located at the center of symmetry. Both the central disulfide bridge (involving Cys 390 and Cys 390# ) and the distal cysteines (Cys 329 ) are depicted as stick models. b, stereoview of caspase-2 (blue), caspase-3 (green), and caspase-9 (yellow). Superposition of the C␣ chains of one p19p12 monomer each of caspase-2, caspase-3 (1PAU, 37), and of the active monomer of caspase-9 (1JXQ, 36). The Ac-LDESD inhibitor of caspase-2 is shown in gray and is depicted as a stick model. All N and C termini are labeled and the 7-residue loop inserted at position 240 of caspase-9 is also indicated. c, stereo representation of the active site region: for the inhibitor a 2F o Ϫ F c electron density map, contoured at 1, was calculated (31) with the final refined phases. The p19 chain is shown in blue, the p12 chain in gray, and the inhibitor molecule, bound to the Cys 285 via a thioester bond, is shown as a stick model and colored according to atom type. The figure was generated using Setor (47).
insertion could contribute to the observed difference in dimer formation of caspase-2 compared with caspase-9.
The pentapeptidic inhibitor Ac-LDESD-cho is covalently bound to the active site cysteine and all atoms are well-resolved in the electron density (Fig. 1c). Because specificity of caspases is determined predominantly by three specificity loops, which show considerable variation both in length and amino acid composition, it is not surprising that significant structural differences prevail in the vicinity of the active site (Fig. 1b).
The Active Site and Inhibitor Binding-Both caspase-2 p19/ p12 monomers form an active site to which an inhibitor molecule is bound, this in contrast to its closest relative, caspase-9, which contains only one active site per (p19/p12) 2 dimer (36). The bound Ac-LDESD-cho inhibitor reveals the details of the pentapeptidic substrate recognition site of caspase-2, and its resemblance to the widely used tetrapeptidic Ac-DEVD inhibitor allows a direct comparison with other caspases bound to the latter compound.
In the S1 specificity pocket, aspartic acid is absolutely required for catalysis. Accordingly, interactions in S1 are strictly conserved among caspases. The aspartate side chain is buried within the deep positively charged S1 pocket and is stabilized by electrostatic interactions with two arginine residues (Arg 179 and Arg 341 ; caspase-1 numbering scheme throughout) and by a hydrogen bond with a conserved glutamine (Gln 283 ; Fig. 2). In addition, most caspase inhibitor complex structures show a hydrogen bond between the P1 backbone carbonyl and the backbone nitrogen of a highly conserved serine, which is replaced in caspase-2 by a glycine (Gly 238 ). Even though aldehyde inhibitors have been described as not exhibiting classic oxyanion hole binding to caspases (30), the C-terminal carbonyl group of Ac-LDESD-cho interacts directly with the amide proton of the strictly conserved Gly 238 in the oxyanion hole and with the imidazole moiety of the active site histidine (His 237 ).
The S2 pocket of the enzyme is not very selective and can accommodate a variety of side chains. Hence, S2 can be considered to be of minor importance for substrate recognition (17,40). In turn, all caspases show a distinct preference for glutamic acid in S3. As seen in other caspases, the glutamate side chain is stabilized by electrostatic interaction with the highly conserved Arg 341 . Both the P3 backbone nitrogen and the carbonyl oxygen form hydrogen bonds with the respective backbone carbonyl oxygen and nitrogen of the same conserved arginine involved in the surface charge interaction. A charge-charge interaction with a second arginine, Arg 177 in caspase-8 (39), is not present in caspase-2.
The S4 pocket has been described as the most selective specificity determinant, aside from the strict requirement for Asp in S1 (17). All group II caspases require a P4 Asp for efficient catalysis, whereas group I and III caspases tolerate a wide range of different residues with preference given to large aromatic/hydrophobic or large aliphatic side chains, respectively. Our structure provides the structural basis for the almost absolute requirement for a P4 aspartic acid in caspase-2. In the S4 binding pocket, which is tunnel-shaped and very narrow compared with the other group II caspases (Fig. 3), both carboxyl oxygens of the aspartate side chain form hydrogen bonds involving Tyr 382B , Asn 342 , and Trp 348 and, via an ordered H 2 O molecule, with Glu 380 and Arg 381 (Fig. 2). In caspase-3, Asn 342 binds the P4 Asp in a similar way. In caspase-7 only one of the two carboxyl oxygen atoms is involved in a hydrogen bonding network, because Asn 342 is substituted by a Ser. Taken together, the positions P1 and P4 are indispensable as anchor residues for the specificity group II caspases, whereas the chemical properties of amino acids tolerated in the other binding pockets may vary within certain limits.
The P5 Specificity Pocket-The dependence of cleavage efficiency on peptide length was examined by Talanian and coworkers (40). In this study, the DQQD sequence motif, internal cleavage site of caspase-2, was extended at both the N and C termini and subjected to peptide cleavage assays. The addition of a P5 residue increased the cleavage efficiency by more than one order of magnitude, whereas the addition of a P6 or P7 residue or of C-terminal extensions did not result in significant changes of V max /K m . We determined a 35-fold increased cleavage efficiency of the pentapeptide Ac-LDESD-amc compared with the tetrapeptide Ac-DEVD-amc. Because P2 is described to be of minor importance for substrate recognition and a Val/Ser substitution at this position was shown to have only little effect on catalysis (17,40) most of the observed increase in cleavage efficiency can be attributed to the P5 leucine. These results indicate that the additional enzyme substrate interaction in P5 plays a crucial role in caspase-2 but not in other investigated caspases. A surface comparison of the substrate recognition site with other caspases revealed that the caspase-2 binding groove is extended at the inhibitor's N-terminal region and forms an additional binding pocket, which allows the accommodation of a fifth residue and contributes to binding energy (Fig. 3). The hydrophobic S5 pocket is formed by Pro 385 and the phenyl moiety of Tyr 383 , resulting in the preference of caspase-2 for small hydrophobic P5 residues. In addition, the proper orientation of the P5 amino acid is secured by two hydrogen bonds involving both the amino and the carbonyl group of the N-terminal residue of the inhibitor and the backbone amide proton of Thr 343 of the enzyme plus the side-chain hydroxyl group of the same residue (Fig. 2). These observations, together with the fact that all atoms of the P5 residue are well ordered and clearly visible in the electron density, support the significance of the S5 binding site for substrate recognition in caspase-2.
A Central Disulfide Connects the Two p19/p12 Monomers-The two p19/p12 monomers are covalently linked by a central disulfide bridge, which is located at the 2-fold symmetry axis of the protease (Fig. 1a). This observation came as a surprise, because an equilibrium between the monomeric and the dimeric form was reported for all examined long prodomain caspases; an equilibrium that depends on the ligand binding status (36,(41)(42)(43). The presence of a covalent linkage between the two monomers in caspase-2, however, is not compatible with such a monomer-dimer equilibrium. As structure-based alignment indicates, this intersubunit disulfide bond can be formed exclusively by caspase-2, because no other caspase contains a cysteine at this particular location (data not shown). Furthermore, a total of four cysteine residues are properly aligned at the dimer interface; these cysteines are present in all caspase-2 sequences from different species available to date. The central cysteine pair, Cys 390 and Cys 390# (# referring to the second monomer), is engaged in the intermolecular disulfide bridge, whereas the two distal cysteines, Cys 329 and Cys 329# , do not form a disulfide with their respective neighbors, Cys 390 and Cys 390# , even though the geometry would allow this.
Further evidence for the existence of the central disulfide bridge comes from mass spectrometry. After separation of the large from the small subunit using reverse-phase chromatography, both subunits were analyzed using matrix-assisted laser desorption ionization time-of-flight mass spectrometry. In the case of p19, only the 19-kDa protein along with its multiple charged species was detected. With p12, in contrast, a dimer (mass: 23.83 kDa) was identified together with the double (11.91 kDa)-, triple (7.94 kDa)-, and quadruple (5.96 kDa)charged species. Aside from the observation of an obviously covalently linked p12 dimer, the existence of the triple-charged species would be inexplicable if there were no dimer. To check whether the linkage between the two p12 moieties is indeed a disulfide, the same sample was reduced under denaturing conditions and analyzed again. As expected, there was no p12dimer detectable any more, and no peak at 7.94 kDa for the triple-charged species was present either. Compared with the expected masses the averaged mass deviation was below 0.1% (data not shown).
The Oligomeric State of Caspase-2 in Vitro-A covalent link-age at the dimer-dimer interface must have consequences on the oligomeric state in solution, thus, active and irreversibly inhibited caspase-2 were analyzed by size-exclusion chromatography at comparable concentrations (Fig. 4). As the chromatograms show, caspase-2 also exists exclusively as a (p19/ p12) 2 dimer in solution, both when bound and unbound to an inhibitor. These findings are contrasted by caspases-9, -8, and -1, which all occur predominantly as monomers in the absence of an inhibitor. In these cases, dimer formation is either driven by ligand binding or by increasing the protein concentration to non-physiologically high values, such as that used in crystallization (36,(41)(42)(43).

DISCUSSION
Caspase-2, initially termed Ich-1 or Nedd-2, was the first mammalian cell death protease to be discovered (5). Several lines of evidence indicate that caspase-2 is positioned upstream of mitochondrial permeabilization and therefore represents the true apical initiator caspase in stress-induced apoptosis, although caspase-9 was originally believed to be the initiator caspase in cell stress signaling (6,8,9,44). These findings, combined with the key role of caspase-2 activation in human breast cancer and in response to different types of chemotherapies, first of all in acute myeloid and acute lymphoblastic leukemia, suggest that caspase-2 could serve as a promising drug target (10,12,13). Recently, a high molecular weight complex was described to which procaspase-2 is recruited in a way that is independent from cytochrome c and Apaf-1 (45); a complex possibly similar to the apoptosome in which procaspase-9 is activated. The individual components of this complex have yet to be identified and analyzed to delineate the molecular basis of caspase-2 activation in the intrinsic pathway.
Here we report the crystal structure of caspase-2, determined at 1.65-Å resolution (Fig. 1a). The caspase-2 (p19/p12) 2 dimer comprises two active sites, to each of which one inhibitor molecule is bound. This is in contrast to caspase-9, which has only one active site formed. In caspase-9, the two potential active sites are in contact with each other via a system of two Tyr and two Phe residues. Formation of one active site leads to a compensatory rotation of the side chains of Tyr 331 and Phe 390 . In a hypothetical dimer made from two active catalytic domains, the two reoriented Phe 390 side chains were found to create a steric clash. This eliminates the possibility of having two active monomers in a caspase-9 dimer (36). In caspase-2, Tyr 331 is also present but Phe 390 is substituted by a cysteine residue (Cys 390 ), rendering the two active sites independent from each other. Instead, Cys 390 forms an intersubunit disulfide with its symmetry related counterpart and is therefore probably of importance with respect to dimerization and activation, as discussed below in more detail.
The high resolution caspase-2 structure allows a detailed analysis of the interactions between the enzyme and the pen- tapeptidic substrate (Fig. 1c). The P1-P5 relative substrate specificities of caspase-2, as determined previously by kinetic studies (17,40), are attributed to the structural and amino acid composition differences in both the substrate binding groove and the loop regions around the active site. It became apparent that the observed P5 recognition originates from the closed shape of the S5 binding pocket, compared with the open and therefore more solvent-accessible corresponding regions of other caspases. The P5 binding pocket is formed by the side chains of Tyr 383 and Pro 385 , which provide the hydrophobic surface to interact with the P5 valine side chain of the inhibitor (Fig. 3). The hydrophobic interaction, together with an additional hydrogen bond with the P5 backbone of the inhibitor, ensures the proper positioning of the N-terminal residue of the ligand. Interestingly, the P4 aspartate, which functions in specificity group II caspases as an indispensable anchor residue, is located in a narrow tunnel-shaped binding pocket, whereas the caspases-3 and -7 contain more solvent-accessible S4 sites. Hence, we suppose that the increased catalytic efficiency upon addition of a fifth residue arises by part not from the P5 contribution directly but also from a stronger P4 binding as consequence from the P5 recognition and a resulting narrowing of the entire P4/P5 binding area. Taken together, these findings extend our understanding of the caspase-2 substrate specificity and open the way for the development of caspase-2-directed inhibitors of potential medical use as well as for specific caspase-2 substrates or inhibitors for analytical purposes.
An unexpected finding was the presence of an intersubunit disulfide bridge, which is located at the position of the 2-fold symmetry axis, involving Cys 390 from both monomers (Fig. 1a). The significance of the amino acid located at the center of symmetry with respect to dimer formation was analyzed previously by mutating residue 390 of caspases-8 and -9. The resulting interface mutants were significantly ineffective regarding dimer formation and consequently at inducing apoptosis, thereby supporting the critical role of residue 390 for dimerization (42). We believe that this intersubunit disulfide bond, which is unique to caspase-2, also exists in vivo at least transiently for the following reasons: the reducing environment of the cell was mimicked by the addition of DTT, which was present during the purification procedure and also in the crystallization buffer (4.4 mM). The fact, that the potential disulfide found stabilizing the N terminus of p19 in caspase-1 is not formed in caspase-2, also indicates a reducing environment. The central intersubunit disulfide, in contrast, is probably not affected by the reducing agent, because it is buried within the core of the enzyme. We observe this intersubunit disulfide in three structures arising from crystals grown under different conditions. In light of the importance of dimerization for the activation of caspases (42,43), this might suggest a novel mechanism of dimer stabilization of biological significance.
An examination of the oligomeric state of caspase-2 by size exclusion chromatography confirmed that the protease exists exclusively as a (p19/p12) 2 dimer not only in the crystal but also in solution, under reducing conditions and both in the presence and absence of a ligand. The existence of this central disulfide bridge was also proven by mass spectrometry: a dimer of p12 was detected, but not of p19, and as expected, addition of a reducing agent to the denatured p12 subunit resulted in complete separation of the two chains. In conclusion, caspase-2 revealed properties in solution that are clearly different from all other investigated long prodomain caspases, the death effector domain-containing caspase-8 (42,43) and the CARDcaspases-9 (36) and -1 (41), which are all predominantly monomers in the absence of a ligand and dimerize upon inhibitor binding.
These findings are complemented by the observation of two additional free cysteines properly aligned with the central disulfide (Fig. 1a). Given this striking alignment at this location, we suppose that this motif plays a role during the activation of caspase-2 and is more than just sheer coincidence. Possibly, the two Cys 329 -Cys 390 intrasubunit disulfides exist transiently before or while the two p19/p12 monomers associate to form the dimer.
A unified model for apical caspase activation has recently been proposed, whereby apical caspases are activated primar- FIG. 4. Examination of the oligomeric state of caspase-2 using size exclusion chromatography. The oligomeric state of caspase-2 was investigated by size exclusion chromatography using a Superdex-200 HR 10/30 column. The upper panels represent the situation in the absence, the lower panels in the presence of the irreversibly binding Ac-DEVD-cmk inhibitor. As expected, caspase-2 exists as a (p19/p12) 2 dimer in solution, irrespective of the ligand binding status, unlike caspase-8 (shown here; 43) or caspase-9 (36), in which dimer formation is driven by inhibitor binding. For caspase-2, the relative enzyme activity of the individual fractions is indicated by a histogram plot. Using a previously recorded standard curve, a molecular mass of 53 kDa was derived for the caspase-2 dimer. Caspase-8 was chromatographed at comparable protein concentration employing the same column but a different elution buffer. Using exactly the same buffer, an elution volume of 14.05 ml resulted for the caspase-8 dimer (M. Donepudi and M. G. Grutter, unpublished data), which is in good agreement with the corresponding value for the caspase-2 dimer (14.12 ml).
ily by dimerization of monomeric zymogens, and proteolysis represents a secondary event resulting in partial stabilization of the dimer (42). In addition, there are data suggesting that caspases-2 and -10 are also activated by oligomerization (46). In light of the oligomeric state of caspase-2 in solution, which is clearly different to all other long prodomain caspases investigated to date, a role of the intersubunit disulfide bridge formation during activation of caspase-2 seems plausible but remains to be proven in vivo. Caspase-2 certainly represents an interesting and exceptional case regarding dimer formation and suggests a way of dimer stabilization unseen so far in caspases.