CASH, a Novel Caspase Homologue with Death Effector Domains*

CASP-8 and CASP-10, members of a cysteine protease family that participates in apoptosis, interact with MORT1/FADD, an adapter protein in the CD120a (p55 tumor necrosis factor receptor), and CD95 (Fas/Apo-1) death-inducing signaling pathways, through a shared N-terminal sequence motif, the death effector domain. We report cloning of two splice variants of a novel protein, CASH, that contain two N-terminal death effector domains and can bind through them to each other, to MORT1/FADD, to CASP-8, and to CASP-10. The unique C-terminal part of the longer variant shows marked sequence homology to the caspase protease region yet lacks several of the conserved caspase active site residues, suggesting that it is devoid of cysteine protease activity. Overexpression of the short CASH splice variant strongly inhibited cytotoxicity induction by CD120a and CD95. Expression of the longer variant, while inhibiting cytotoxicity in HeLa cells, had a marked cytocidal effect in 293 cells that could be shown to involve its protease homology region. The findings suggest that CASH acts as an attenuator and/or initiator in CD95 and CD120a signaling for cell death.

The caspases, conserved cysteine proteases that cleave specific cellular proteins downstream of aspartate residues, play a critical role in all known programmed cell death processes (reviewed in Refs. 1 and 2). These proteases (also called the CED3/ICE proteases, after the first described members of the family) (3)(4)(5)(6) are produced as inactive precursors and become activated by proteolytic processing upon death induction. In addition to their homologous C-terminal region from which the mature proteases are derived, the precursor proteins contain unique N-terminal regions. Interactions of these "prodomains" with specific regulatory molecules allow differential activation of the various caspases by different death-inducing signals (7)(8)(9)(10)(11). Two recently described caspases with similar prodomains, CASP-8 (MACH/FLICE1/Mch5) (7,8,12) and CASP-10 (Mch4/FLICE2) (12,13), interact through their prodomains with MORT1/FADD (14,15), an adapter protein in the deathsignaling cascades activated by two closely related receptors of the TNF 1 /nerve growth factor family, CD120a (the p55 TNF receptor) and CD95 (Fas/Apo-1). This interaction, which is required for the signaling to death, involves a protein-binding motif called the "death effector domain" (DED) or the "MORT" motif, found in the N-terminal part of MORT1/FADD and in duplicate in the prodomains of CASP-8 and CASP-10.
Here we report cloning of a novel protein, CASH, that contains duplicated DED at its N terminus and binds through this region to MORT1/FADD. The C-terminal part of the protein shows marked sequence homology to the corresponding regions in CASP-8 and CASP-10 yet lacks several of the residues that are crucial for cysteine protease activity. Functional tests demonstrated an ability of the novel protein to trigger as well as to inhibit signaling for death.

EXPERIMENTAL PROCEDURES
Cloning of CASH␤ and Study of Its Binding Properties by Two-hybrid ␤-Galactosidase Expression Test-CASH␤ was cloned by two-hybrid screening (16) of a Gal4 activation domain-tagged human Jurkat T cell library (donated by J. H. Camonis, Curie Institute) for proteins that bind to CASP-10 using the HF7c yeast reporter strain (CLONTECH, Palo Alto, CA). Screening was performed in the absence of 3-aminotriazole according to the Matchmaker Two-Hybrid System Protocol (CLON-TECH). The binding properties of CASH␤, as well as CASH␣ were assessed in the yeast SFY526 reporter strain (CLONTECH) using the pGBT9-GAL4 DNA-binding domain and the pGAD1318 and pGADGH-GAL4 activation-domain vectors. Quantification of the binding in yeast by the ␤-galactosidase expression filter assay was performed as described (17).
Cloning of the Mouse CASH Splice Variants-An expressed sequence tag clone (GenBank accession number AA198928) was identified as the mouse homologue of part of the DED region in CASH. Based on this sequence we cloned the mouse CASH␣ and CASH␤ splice variants from mouse liver mRNA by reverse transcription-PCR. The reverse transcriptase reaction was performed with an oligo(dT) adapter primer (5Ј-GACTCGAGTCTAGAGTCGAC(T) 17 -3Ј) and the avian myeloblastosis virus reverse transcriptase (Promega) used according to the manufacturer's instructions. The first round of PCR was carried out with the Expand Long Template PCR System (Boehringer Mannheim) using the following sense and antisense primers: 5Ј-GGCTTCTCGTGGTTCCCA-GAGC-3Ј and 5Ј-GACTCGAGTCTAGAGTCGAC-3Ј (adapter) respectively. The second round was performed with Vent polymerase (NEB) using the nested sense primer 5Ј-TGCTCTTCCTGTGTAGAGATG-3Ј and adapter.
Northern Blot Analysis-A radiolabeled mRNA probe corresponding to the DED module region of CASH was prepared using the T7 RNA polymerase (Promega) and used for analysis of human multiple tissue blots (CLONTECH) according to the manufacturer's instructions.
Sequence Analyses-Sequence alignment and homology evaluation were performed using the GAP and PILEUP programs of the GCG package and by the CLUSTAL 1.5 software. Sequence data base search was performed using the BLAST program. As parameter of homology significance we used "smallest sum probability" (P(N)), i.e. the probability of observing by chance a score or a group of scores as high as the observed ones when performing a search of the same size. The consensus for the DED region sequence was deduced from the alignment of the DED modules in MORT/FADD (14,15), MACH/FLICE/Mch5 (7,8,12), * This work was supported in part by grants from Inter-Lab Ltd., Ness-Ziona, Israel, from Ares Trading S.A., Switzerland, and from the Israeli Ministry of Arts and Sciences. 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 nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM (19), and K13 (20), and the consensus for the protease-homology region was deduced from the alignment of CASP-1, 2, 3, 6, 7, 8, and 10 (21) and CED3 (3). Residues occurring in more than six of the aligned proteins were included in the consensus.
Expression Vectors-The CASH␣ deletion mutants and the CD120a/ CD95 chimera were produced by PCR and/or conventional cloning techniques. The CASH splice variants, the CD95 or CD120a signalingcascade proteins (all of human origin), and the baculovirus p35 protein were expressed in mammalian cells using the pcDNA3 expression vector (Invitrogen). ␤-Galactosidase was expressed using the pCMV-␤-gal vector (Promega).
Cells (5 ϫ 10 5 293 cells or 3 ϫ 10 5 HeLa cells per 6-cm dishes) were transiently transfected with the cDNAs of the indicated proteins together with the pCMV-␤-gal, using the calcium phosphate precipitation method. Each dish was transfected with 5 g of the pcDNA3 construct of interest or, when transfecting two different constructs, 2.5 g of each, and 1.5 g of ␤-galactosidase expression vector. Cells were rinsed 6 -10 h after transfection and then incubated for a further 14 h without additional treatment. Anti-CD95 monoclonal antibody (CH11 (Oncor, Gaithersburg, MD), 0.5 g/ml) and human recombinant TNF␣ (100 ng/ml) were applied to the cells together with cycloheximide (10 g/ml) and incubated for an additional 4 h. Cells were then stained with 5-bromo-4-chloro-3-indoxyl-␤-D-galactopyranoside (22) (4,5), and the consensus, deduced as described under "Experimental Procedures." CASP-1 and CASP-3 are shown without their prodomain regions. Amino acid residues are numbered to the right of each sequence. Dotted lines indicate gaps inserted in the sequence to allow optimal alignment. The DED modules are shaded. Amino acids that are identical in more than three of the proteins shown are boxed. Residues in the consensus sequence that are identical to those in hCASH␣ are shown in bold type. Within the region of protease homology, amino acids aligned with CASP-1 residues that were implicated in catalytic activity by x-ray crystallography are denoted as follows. The residues putatively involved in catalysis, corresponding to His 237 and Cys 285 in CASP-1, are darkly shaded and marked by closed circles below the alignment. The residues constituting the binding pocket for the carboxylate side chain of the P1 Asp, corresponding to Arg 179 , Gln 283 , Arg 341 , and Ser 347 in CASP-1, are less heavily shaded and marked by open circles. Known and suggested Asp-Xaa cleavage sites and the potential site of cleavage found at a similar location in CASH are shaded. Horizontal arrows indicate the N-and C-terminal ends of the small and large subunits of the CASP-1. The C termini of the proteins are denoted by asterisks. The significance of the homologies, shown, was confirmed as follows. (a) A search of the GenBank CDS translations and the Protein Data Bank, SwissProt and PIR data banks for sequence similarities to the putative DED and protease homology regions in CASH yielded multiple caspases and DED-containing proteins. Homology was highly significant with P(N) as low as 10 Ϫ15 and 10 Ϫ27 for some caspases and DED-containing proteins, respectively; no other protein had P(N) values lower than 10 Ϫ2 . (b) Of the residues in the consensus sequences defined for the DED and caspase motifs, more than 60 and 50%, respectively, were observed in the corresponding regions in CASH␣. (c) The CASH␣ C-terminal region showed identity of 23% to CASP1, 29% to CASP3, and 29% to CASP10 (identity higher than 25% is considered significant (32)).

RESULTS AND DISCUSSION
To search for the proteins that bind to CASP-10 (Mch4/ FLICE2), we performed two-hybrid screening of human Jurkat T cell cDNA library using CASP-10 as a bait. This screen yielded cDNA clones of MORT1/FADD, previously shown to bind to CASP-10 (13). It also yielded a partial clone of a novel cDNA, which like CASP-8 and CASP-10, contained two death effector (MORT) modules (7,8,12) just downstream of its N terminus (Fig. 1). Because of the similarity of the protein to the caspases (see below) it was dubbed CASH, for caspase homologue.
Northern blot analysis revealed that the molecule exists in at least three distinct transcript sizes, 1.5, 2.4, and 4.0 kilobase pairs (data not shown), whose proportions vary greatly among different tissues. To obtain the full-length cDNA of CASH, we screened human skin fibroblast cDNA library (CLONTECH) with a cDNA probe corresponding to the CASH sequence. We obtained two cDNA species, apparently corresponding to two splice variants of CASH. The proteins encoded by these two cDNAs shared the death effector domain-containing N-terminal region, but their C termini differed. One (CASH␤) had a short C terminus, corresponding to that of the originally cloned cDNA. The other (CASH␣) had a long C terminus.
The amino acid sequence in this longer C-terminal region showed rather high homology to those of the protease-precursor regions in CASP-8 and CASP-10 ( Fig. 1). However, it lacked several of the residues believed to be crucial for protease activity, suggesting that the protein is devoid of cysteine protease activity. Interestingly, CASH␣ contains a caspase-substrate sequence at the site corresponding to the proteolytic-processing site within the protease regions in CASP-8 and CASP-10 (shaded in Fig. 1). Preliminary data suggest that CASH␣ can indeed be cleaved at this site by CASP-8.
Based on the nucleotide sequence of an expressed sequence tag clone found to correspond to the mouse homologue of part of the DED region in CASH, we cloned the cDNAs of both the mouse CASH␣ and CASH␤ splice variants from mouse liver mRNA by reverse transcriptase-PCR. Sequence comparison revealed high conservation throughout the CASH␣ molecule (71% identity in DED region and 59% in protease homology region), suggesting that both the DED and protease homology regions in the protein contribute to its function (Fig. 1).
Two-hybrid testing of the interactive properties of CASH␣ and CASH␤ (Fig. 2) revealed that both variants interact with MORT1/FADD and CASP-8, most probably through their shared DED regions. Notably, although initially cloned by twohybrid screening for proteins that bind to CASP-10, CASH␤ was found in this test to bind rather weakly to CASP-10, and CASH␣ did not bind to it at all. The two CASH variants also self-associated and bound to each other, but did not bind RIP or TRADD (adapter proteins that, like MORT1/FADD, contain death domains but lack DEDs), nor did they bind to a number of irrelevant proteins used as specificity controls.
To examine the function of CASH, we expressed its two variants transiently in HeLa and 293-T cells and assessed the effects of the transfected proteins on the CD120a-induced signaling for cytotoxicity triggered by TNF or by overexpression of the receptor, as well as on the CD95-induced signaling for cytotoxicity triggered by antibody cross-linking of CD95 or by overexpression of a chimeric receptor comprised of the extracellular domain of CD120a and the intracellular domain of CD95 (Fig. 3). In both cell lines, expression of CASH␤ by itself had no effect on cell viability, but it strongly inhibited the induction of cell death by CD120a as well as by CD95. Expression of the CASH␣ variant affected the two cell lines very differently. In HeLa cells it inhibited the cytotoxicity of CD120a and CD95, similarly to CASH␤. In the 293-T cells, however, it resulted in marked cytotoxicity. Similar cytotoxicity was observed when the protein was expressed in 293-EBNA cells or 293 cells (not shown). This cytotoxic effect could be completely blocked by coexpression of p35, a baculovirus-derived caspase inhibitor (23,24).
To assess the contribution of the region of protease homology in CASH␣ to its cytocidal effect, we examined the functions of two mutants of the protein, CASH␣(1-385) and CASH␣(1-408), with C-terminal deletions at the region corresponding to that part of the protease domain from which the small subunit of the mature protease is derived. Both mutants were devoid of any cytotoxic effect. Moreover, like CASH␤ they protected the 293 cells from death induction by CD120a and CD95 (Fig. 3C).
The above findings indicate that CASH can interact with components of the signaling complexes of CD120a and CD95 and that it affects death induction in a way that may differ depending on the identity of the splice variant of CASH and on the cell type in which it is expressed.
The inhibition of cytotoxicity induction by CASH␤, and in the case of the HeLa cells also by CASH␣, is apparently mediated by the DED region in this protein. It probably reflects competition of the DED of CASH with the corresponding regions in CASP-8 and CASP-10 for binding to MORT1/FADD.
Less easy to explain is the way in which CASH␣ causes death of the 293 cells. The ability of the p35 protein to block this cytotoxic effect indicates that the cytotoxicity is mediated by the activity of caspases. Yet CASH␣, even though displaying marked sequence homology to the caspases is unlikely to have cysteine-protease activity because it lacks several of the conserved caspase active site residues. A more likely explanation is that it acts by activating other molecules that do have caspase activity.
An intriguing possibility is that CASH␣, though unable to act alone as a protease, can still constitute part of an active protease molecule. Crystallographic studies of CASP-1 and CASP-3 structure indicate that the small and large protease subunits in each processed enzyme are derived from distinct proenzyme molecules (25)(26)(27)(28). In view of the observed dependence of the CASH␣ cytotoxic activity on intactness of the region corresponding to the small protease subunit (Fig. 3C), it is tempting to speculate that this region in CASH␣ can associate with the large subunit region of certain caspase(s) in a way that results in reconstitution of an enzymatically active molecule. The resulting active heterotetramer should then be capable of activating other caspases, thus triggering cell death.
The discovery of CASP-8 and CASP-10 and of their association through MORT1/FADD with CD120a and CD95 (7,8,13) indicates a plausible mechanism for initiation of the deathinducing cascades by the two receptors. It does not, however, provide any clue to the cause of the marked variation in effectivity of death induction by these receptors, even among cells that express the receptors, their adapter proteins, and the caspases at similar levels. Nor does it explain the frequently observed differences in effectivity of death induction by the two receptors. CASH, through its effects on the signaling for death by CD120a and CD95, which vary depending on the identity of the CASH splice variant expressed and perhaps also on the identity of the specific caspases expressed in the cell, may well contribute to these variations in cytotoxicity induction. It should be stressed, however, that transfection studies such as those described here can provide only a partial view of the function of a protein. It is conceivable that at least part of the effects observed when expressing a protein in amounts far higher than its normal levels will turn out to be unrelated to its real function. Complementing these overexpression tests by assessing the effect of decreased expression of CASH, e.g. by deleting the CASH gene, should provide a more reliable notion of the physiological role of this protein.