Identification and Characterization of Ich-3, a Member of the Interleukin-1β Converting Enzyme (ICE)/Ced-3 Family and an Upstream Regulator of ICE

We report here the isolation and characterization of a new member of the ice / ced-3 family of cell death genes, named ich-3 . The predicted amino acid sequence of Ich-3 protein shares 54% identity with murine interleukin-1 (cid:98) converting enzyme (ICE). Overexpression of ich-3 in Rat-1 and HeLa cells induces apoptosis, which can be inhibited by CrmA and Bcl-2. The mRNA and proteins of ich-3 are dramatically induced in vivo upon stimulation with lipopolysaccharide, an inducer of septic shock. The ich-3 gene product can be cleaved by cytotoxic T cells granule serine protease granzyme B, suggesting that Ich-3 may mediate apoptosis induced by granzyme B. Ich-3 does not process proIL-1 (cid:98) directly but does promote proIL-1 (cid:98) processing by ICE. These results suggest that Ich-3 may play a very important role in apoptosis and inflammatory responses and may be an upstream regulator of ICE. Generation of Ich-3 Antibodies and Western Blotting Analysis— A 15-amino acid peptide (HTEFKHLSLRYGAKFD)8-multiple antigen peptide-linked within the p20 region of Ich-3 was used for the genera- tion of polycolonal antibodies. The peptide and rabbit polycolonal antibodies against p20 region of Ich-3 was made by Research Genetics (Huntsville, AL) and purified using 4% N -hydroxysuccinimidyl chloro-formate-activated cross-linked beaded agarose from Sigma (H8635) ac- cording to the manufacturer’s protocol. For Western blotting, 10 (cid:109) g of purified bacterial Ich-3 fusion protein was subjected to SDS-polyacryl- amide gel electrophoresis on a 15% polyacrylamide gel. Proteins were then transferred onto Immobilon-P membrane (Millipore, Bedford, MA) and incubated with 1 (cid:109) g/ml rabbit anti-Ich-3 p20 peptide antibody for 2 h at room temperature. After three times of washing, the membrane was incubated with horseradish peroxidase-linked anti-rabbit Ig antibody for 45 min at room temperature (Amersham Corp.). After washing three times, antibodies bound to the membrane were revealed with the ECL Western blotting reagent (Amersham Corp.). The anti-Ich-3 monoclonal antibody was isolated from rats immu-nized with bacterial expressed Ich-3 protein using conventional proto- cols. The specificity of this monoclonal antibody was proved by Western blotting using proteins from the ich-3 (cid:50) / (cid:50) mice tissues. 4 To determine induction of Ich-3 by LPS, proteins were isolated from tissues of 7–10- week-old mice before or 4 and 20 h after LPS injection (40 mg/kg). 60 (cid:109) g of proteins were loaded in each lane on a 12% polyacrylamide gel for SDS-polyacrylamide gel electrophoresis. After transferring the proteins onto Immobilon-P membrane, Western blotting was carried out as by using rat anti-mouse Ich-3 monocolonal antibodies. In Vitro Cleavage Assay— For in vitro cleavage of Ich-3 by GraB, 10 mg of purified His-tagged Ich-3 protein was incubated with 20 ng of GraB in the presence of 50 m M Tris-HCl, pH 7.5, 0.5 m M EDTA, 0.5 m M sucrose, and 10 m M dithiothreitol in a total volume of 10 (cid:109) l. The mixture was incubated at 30 °C for 1 h, and the cleavage was detected by Western blotting with a peptide antibody against the p20 portion of Ich-3.

Interleukin-1␤ converting enzyme (ICE) 1 family is a growing family of cysteine proteases involved in cytokine maturation and apoptosis (1). ICE is a cytoplasmic cysteine protease responsible for proteolytically processing pro-interleukin-1␤ (31 kDa) into active form (17 kDa) (2). The amino acid sequence of ICE shares 29% identity with Caenorhabditis elegans cell death gene product Ced-3 (3). Expression of ice in a number of mammalian cell lines induces apoptosis (4,5). Microinjection of an expression vector of crmA, a cowpox virus gene encoding a serpin that is a specific inhibitor of ICE, prevents death of neurons of dorsal root ganglia and ciliary ganglia induced by trophic factor deprivation (6,7). Expression of crmA can also suppress apoptosis induced by TNF␣ and Fas (8 -11). These experiments suggest that the members of the ICE family play important roles in controlling mammalian apoptosis.
Cytotoxic T lymphocytes (CTL) are important players in host cell-mediated immunity (12). Granzyme B (GraB) is a serine protease that plays a major role in apoptosis induced by CTLs because mice that are deficient for GraB generated by gene targeting technique are severely defective in CTL-induced apoptosis (13). GraB can induce apoptosis of many if not all cell types in the presence of pore forming protein perforin (14,15). A recent report showed that CPP32, a member of the ICE family, is activated by cytotoxic T-cell-derived GraB, suggesting that CPP32 is important for CTL killing (16). CPP32, however, cannot be the only ICE family activated by CTL because CrmA is a very poor inhibitor of CPP32 (17). Tewari et al. (18) showed that expression of crmA completely blocks the Ca 2ϩ -independent component of CTL killing (i.e. Fas-mediated); if CPP32 were the only ICE family member responsible for CTL cytotoxicity, expression of crmA should not suppress CTL killing. We predict that there are additional members of the ICE family that play an important role in CTL-induced apoptosis. The amino acid sequence of GraB is not homologous with ICE; however, GraB and ICE share many enzymatic similarities. Like ICE, GraB requires Asp at P1 position for cleavage. Inhibitors of ICE or the ICE family, CrmA, ICH-1 S , and a mutant ICE are effective inhibitors of GraB/perforin-induced apoptosis. 2 Embryonic fibroblasts that are deficient in ICE from iceϪ/Ϫ mice are resistant to GraB/perforin-induced apoptosis, 2 suggesting that ICE is critical for GraB/perforin-induced apoptosis in at least certain cell types. ICE itself cannot be directly cleaved by GraB (20), and thus, although ICE is required for GraB/perforin-induced apoptosis in certain cells, GraB does not activate ICE directly. One possibility is that GraB activates another ICE family member that may then directly or indirectly activate ICE, and the activator of ICE can be inhibited by CrmA.
The mammalian ICE family now includes ICE, Nedd-2/ ICH-1, CPP32/YAMA, MCH-2, TX/ICH-2/ICErelII, and ICE re -lIII (2,5,(21)(22)(23)(24)(25)(26). Overexpression of nedd-2/ich-1 L induces cell death very effectively (5,21). Expression of CPP32/YAMA in full-length cDNA induces apoptosis of insect Sf9 cells but not * This work was supported in part by grants from National Institute of Aging and from Bristol Myer-Sqibb (to J. Y.) and in part by grants from the Medical Research Council of Canada and National Cancer Institute of Canada (to A. H. G.). 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 that of mammalian cells (22). 3 Recombinant CPP32/YAMA is inactive, and cleavage of CPP32/YAMA by ICE in vitro activates the precursor (24), suggesting that in vivo CPP32/YAMA may be activated by another protease to induce apoptosis. Expression of MCH2␣ also induces apoptosis of insect Sf9 cells but not that of mammalian cells (27). 3 Thus, the members of the ICE family can be classified into two classes: those that when overexpressed in mammalian cells can induce apoptosis (e.g. ice and ich-1) and those that when overexpressed in mammalian cells cannot induce apoptosis (e.g. CPP32 and Mch-2). These experimental evidence suggest that in vivo members of the ICE family may be arranged in proteases cascades, and certain members of the ICE family may activate other members of the ICE family.
We report here the isolation and characterization of a new member of the ICE family named ich-3. The predicted amino acid sequence of Ich-3 exhibits 46% identity with murine ICE, 45% identity with human ICE, 60 and 54% identities with human ICE-like proteases TX (TX, ICE rel -II, and ICH-2 are the same protein) and ICE rel -III, respectively. It shares 26 -32% of sequence identity with Ced-3, human ICH-1 L , and CPP32/ YAMA. Overexpression of ich-3 in Rat-1 and HeLa cells induces apoptosis, which can be inhibited by CrmA and Bcl-2. Expression of ich-3 is dramatically elevated in vivo after stimulation of LPS, an endotoxin secreted by Gram-negative bacteria that induces sepsis. In addition, Ich-3 can be cleaved by granule serine protease granzyme B in vitro. Ich-3 does not process proIL-1␤ directly but promotes processing of proIL-1␤ by ICE. Our results suggest that ich-3 may play an important role in apoptosis and inflammatory responses and may be an upstream regulator of ICE.

MATERIALS AND METHODS
Cloning and Construction of Plasmids-Mouse thymus cDNA library (Stratagene) containing 10 6 plaque-forming units was screened by human ice cDNA as a probe. Hybridization was performed at 40°C overnight in 5 ϫ SSPE, 20% formamide, 10 ϫ Denhardt's solution, 1% SDS. Filters were washed in 1 ϫ SSPE, 0.5% SDS twice at room temperature and then twice at 42°C. Two ich-3 cDNA clones were originally obtained and subcloned into pBluescript II (named m29 and mNO). Additional ich-3 clones were also obtained from the same cDNA library by direct screening with a ich-3 probe and were subcloned in pBluescript II (named BSNO1, BSNO3, BSNO9, and BSNO12). These clones contain inserts with overlapping segments of the ich-3 gene. mNO contained the longest insert (2 kilobase pairs) including the ATG initiation codon; however, this insert was longer than the size of ich-3 mRNA determined by Northern blot. mNO contains an unexpected duplication of ich-3 3Ј cDNA sequence at its 5Ј end. To confirm the 35-base pair upstream sequence from the ATG codon is derived from ich-3 mRNA, we performed the reverse transcriptase PCR analysis by using the primer set mNOF (5Ј-CTTCACAGTGCGAAAGAAC) and m29P2 (5Ј-GGTCCA-CACTGAAGAATGTCTGGAGAAGCATTTCA). An amplified fragment of expected size was obtained, indicating that we have cloned entire coding region of ich-3 (data not shown).
To construct mutant ich-3 gene in which the coding region for the active cysteine residue is changed into a glycine residue, two primers containing the mutation were synthesized: NO2GA (GTGCAGGCCG-GCAGAGGTGGG) and NO2GB (CCCACCTCTGCCGGCCTGCAC). The mutant construct p␤actS6Z was made from two rounds of PCR using two pairs of primers. First round of PCR was to generate the mutant cDNA as two half cDNA fragments. The 5Ј fragment from the N terminus to the mutation site was made using ich-3 cDNA as a template and M34 (CCCTCGAGCGGCCGCCATGGCTGAAAACAAACACCC) and NO2GB as primers. The 3Ј fragment from mutated site to the C terminus was generated using ich-3 cDNA as template and NO2GA and mNOR (AAGTCGACTTGCCAGGAAAGAGGTAGAAATATC) as primers. The PCR was performed in the condition: 1 ϫ Vent DNA polymerase buffer (Biolabs), 0.3 mM dNTPs, 0.5 M each primers, and 1 unit of Vent DNA polymerase (BioLabs) in a total volume of 25 l. DNA was denatured at 94°C for 1 min, annealed at 60°C for 1 min, and elongated at 72°C for 1 min with 28 cycles. In the second round PCR, the mixture of 5Ј fragment and 3Ј fragment was used as template and M34 and mNOR Fusion as primers. The product of second PCR was a complete ich-3 cDNA with a mutation that changed the active cysteine to a glycine. The condition of the PCR was the same as in the first round. The PCR product was inserted into EcoRV site of pBluescript II and sequenced to insure that no additional mutation was introduced during the PCR reactions. The expression construct of the mutant ich-3 (p␤actS6Z) was constructed in the similar way as the original wild type construct.
Northern Blot Analysis and RT-PCR-Total RNA from different tissues of mouse were isolated by TRIzol total RNA Isolation (Life Technologies, Inc.). For isolation of total RNA from mice stimulated by LPS, 7-10-week-old mice were injected with LPS at a dose of 40 mg/kg of body weight, and 5 h after LPS injection, tissues were isolated for RNA preparation. The 32 P-labeled probe was the 3Ј fragment of ich-3 generated by PCR using two primers NO2GA and mNoR. Hybridization was performed overnight at 62°C in 1% BSA, 1 mM EDTA, 0.5 M sodium phosphate, pH 7.2, and 17.5% SDS. The blots were washed in 40 mM sodium phosphate, pH 7.2, 1 mM EDTA, 1% SDS and 70 mM NaCl at 65°C and autoradiographed. For RT-PCR, first strand cDNA was synthesized by use of total RNA and random priming with Moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc.) as described previously (5). The primers used for PCR to amplify ich-3 were mNOF (5Ј-CTTCACAGTGCGAAAGAACT-3Ј) and m29P2 (5Ј-GGTC-CACACTGAAGAATGTCTGGAGAAGCATTTCA-3Ј). The PCR was performed in the condition: 1 ϫ Taq polymerase buffer (Promega), 0.3 mM dNTPs, 2.5 mM MgCl 2 , 0.5 M each primer, 1 unit of Taq DNA polymerase (Promega) in a total volume of 25 l. DNA was denatured at 94°C for 1 min, annealed at 55°C for 1 min, and elongated at 72°C for 1 min with 30 cycles. The PCR of ␤ actin was performed under same condition.
Cell Culture and Transfection Studies-Rat-1 cells, HeLa cells, and COS cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. For transfection, cells were seeded at a density of about 2.5 ϫ 10 5 in each of the 6-well dishes. 1 g of expression construct and 3 l of lipofectamine reagent were used according to the protocol from Life Technologies, Inc. The expression of lacZ fusion genes in cell cultures was detected by X-Gal staining as described previously (4). For cotransfections using more than one construct, CaCl 2 transfection method was used. Briefly, for each 6-well plate dish, 1-5 g of DNA was mixed with 108 l of water and 15.5 l of 2 M CaCl 2 . Then the DNA-CaCl 2 mixture was added slowly into 125 l of 2 ϫ HBS (280 mM NaCl, 10 mM KCl, 1.5 mM Na 2 HPO 4 , 12 mM dextrose, and 50 mM HEPES, pH 7.2). After incubation at room temperature for 20 -30 min, the DNA-CaCl 2 mixture was added into the dish and incubated at 37°C for 3-5 h. The cells were shocked by 15% glycerol in 1 ϫ HBS for 1 min and grown in complete medium until harvesting.
Assay of proIL-1␤ Secretion-An EcoRI fragment of mouse proIL-1␤ cDNA was cloned into pcDNA3 and placed under control of CMV promoter. The construct was named as pCMVS11. To test whether Ich-3 can process proIL-1␤, pCMVS11 was cotransfected with ich-3-LacZ fusion construct pCMVM26Z into COS cells. proIL-1␤ (pCMVS11) was also cotransfected with ice-lacZ fusion construct and with vector (p␤actGal). Vector DNA was added to each transfection to equalize the total amount of transfected DNA. 24 h after transfection, supernatant was collected and stored at Ϫ80°C or used immediately for ELISA according to the manufacturer's protocol (Genzyme, Cambridge, MA). In some experiments the cells were stained by X-Gal as described previously (4) to test the efficiency of the transfection.
Expression and Purification of Ich-3 from Escherichia coli-An EcoRI fragment from the cDNA clone BSNO12 encoding the full-length ich-3 was subcloned into EcoRI site of pTrcHis (Invitrogen). The construct was named as pTrcHisS9. In this construct, the N-terminal of ich-3 was fused to a polyhistidine (six histidines) coding region, and the fusion gene was placed under the control of the inducible trpB and lacUV5 hybrid promoter. The production of the fusion protein was induced by 0.3 mM isopropyl-1-thio-␤-D-galactopyranoside, and the protein was purified by His⅐Bind Buffer kit from Novagen according to the protocol. The protein was stored at Ϫ20°C in 10% glycerol. 3 E. S. Alnemri, personal communication.

Generation of Ich-3 Antibodies and Western Blotting Analysis-A
15-amino acid peptide (HTEFKHLSLRYGAKFD)8-multiple antigen peptide-linked within the p20 region of Ich-3 was used for the generation of polycolonal antibodies. The peptide and rabbit polycolonal antibodies against p20 region of Ich-3 was made by Research Genetics (Huntsville, AL) and purified using 4% N-hydroxysuccinimidyl chloroformate-activated cross-linked beaded agarose from Sigma (H8635) according to the manufacturer's protocol. For Western blotting, 10 g of purified bacterial Ich-3 fusion protein was subjected to SDS-polyacrylamide gel electrophoresis on a 15% polyacrylamide gel. Proteins were then transferred onto Immobilon-P membrane (Millipore, Bedford, MA) and incubated with 1 g/ml rabbit anti-Ich-3 p20 peptide antibody for 2 h at room temperature. After three times of washing, the membrane was incubated with horseradish peroxidase-linked anti-rabbit Ig antibody for 45 min at room temperature (Amersham Corp.). After washing three times, antibodies bound to the membrane were revealed with the ECL Western blotting reagent (Amersham Corp.).
The anti-Ich-3 monoclonal antibody was isolated from rats immunized with bacterial expressed Ich-3 protein using conventional protocols. The specificity of this monoclonal antibody was proved by Western blotting using proteins from the ich-3Ϫ/Ϫ mice tissues. 4 To determine induction of Ich-3 by LPS, proteins were isolated from tissues of 7-10week-old mice before or 4 and 20 h after LPS injection (40 mg/kg). 60 g of proteins were loaded in each lane on a 12% polyacrylamide gel for SDS-polyacrylamide gel electrophoresis. After transferring the proteins onto Immobilon-P membrane, Western blotting was carried out as described above by using rat anti-mouse Ich-3 monocolonal antibodies.
In Vitro Cleavage Assay-For in vitro cleavage of Ich-3 by GraB, 10 mg of purified His-tagged Ich-3 protein was incubated with 20 ng of GraB in the presence of 50 mM Tris-HCl, pH 7.5, 0.5 mM EDTA, 0.5 mM sucrose, and 10 mM dithiothreitol in a total volume of 10 l. The mixture was incubated at 30°C for 1 h, and the cleavage was detected by Western blotting with a peptide antibody against the p20 portion of Ich-3.

Ich-3 Is a Member of the ICE Family-
To identify additional members of the ICE family, we screened a mouse thymus cDNA library at low stringency condition using human ice cDNA as a probe. Two positive clones were identified that encode a protein similar but not identical to murine ICE. The protein encoded by these two clones was named Ich-3. A full-length cDNA of ich-3 was isolated through additional cDNA library screening (see "Materials and Methods").
The cDNA sequence of ich-3 ( Fig. 1) contains an open reading frame of 373 amino acids. The first ATG translational start codon is at the nucleotide 35-37. An opal stop codon is at the nucleotides 1154 -1156. There is a canonical poly(A) signal (AATAAA) at its 3Ј noncoding region. The predicted molecular mass of Ich-3 is 42 kDa. The amino acid sequence of Ich-3 is most homologous to human TX (60% identity), which is also named as ICErelII and ICH2 (25,26). Ich-3 shares 46, 45, and 54% of identities with murine ICE, human ICE, and human ICErelIII, respectively (Table I). Ich-3 is less homologous to C. elegans Ced-3, human ICH-1 L , human CPP32, and human MCH2 with 26, 30, 32, and 24% identities, respectively. Like all the other members of the ICE family, Ich-3 also lacks an extended serine-rich region that is present in Ced-3 (3). The majority of sequence heterogeneity occurs in the prepeptide region, whereas those areas within and around the conserved pentapeptide QACRG, the active site for the ICE family, is highly homologous. These results indicate that Ich-3 protein is a member of the ICE family.
ICE is synthesized as a p45 precursor form that is cleaved during activation into the p20 and p10 subunits (2). The cleavage is dependent on aspartic acid residue in the P1 position. Examining the residues involved in the maturation of the ICE precursor (2,28,29), the residue of Ich-3 corresponding to the ICE residue involved in processing p10 N terminus is conserved, whereas the residues corresponding to the processing site for the N terminus and C terminus of p20 are not conserved in Ich-3 (Fig. 2). Two nearby Asp residues (Asp 59 and Asp 80 ) in the Ich-3 sequence may serve as potential processing sites for   Ich-3  60 46  45  54  26  30  32  24  TX/ICE relIII /ICH-2  49  53  73  27  28  32  23  mIce  62  49  28  26  32  28  hIce  50  29  27  30  27  ICErel-III  25  25  30  23  Ced-3  28  34  33  According to the x-ray crystal structural analysis of ICE (28,29), His 237 , Gly 238 , and Cys 285 of ICE are involved in catalysis, and all three are conserved in Ich-3 (His 206 , Gly 207 , and Cys 254 ) ( Fig. 2A). The residues that are part of the P1 Asp binding pocket in ICE (Arg 179 , Gln 283 , Arg 341 , and Ser 347 ) are also conserved in Ich-3 (Arg 148 , Gln 252 , Arg 310 , and Ser 316 ) ( Fig. 2A). However, the residues in ICE that make up the groove for binding P2-P4 residues of the substrate (ICE Val 338 , Trp 340 , His 342 , Pro 343 , Arg 383 , and Gln 385 ) are quite different in Ich-3, and the corresponding amino acids are Leu 307 , Tyr 309 , Asp 311 , Lys 312 , and His 352 with only one Gln 352 conserved. These comparisons predict that Ich-3 is also a cysteine protease with preference for Asp at P1 position, but it may recognize a slightly different set of substrates than those cleaved by ICE.
ich-3 Is Expressed in Many Tissues and Is Induced by LPS-To study the expression pattern of ich-3, total RNA was isolated from different tissues of mice. Northern blots were probed with an ich-3 cDNA probe from the active site to 3Ј end Dotted lines are spaces in the sequence to allow optimal alignment. The catalytic Gly 238 , Cys 285 , and His 237 residues are marked by asterisks above the residues as indicated by x-ray crystallography analysis (28,29). The residues whose amino acid side chains form the P1 pocket are indicated by a caret above the residue. And those for binding P2-P4 residues are indicated by a {. Known and predicted Asp-Xaa cleavage sites that result in the p20/p10 subunits are indicated by arrows below the residue. The potential processing residues are underlined. Residues conserved in more than three ICE members are highlighted. The numbers at the end of each lane are the numbers of amino acid of the protein. B, the structure motifs of hICE, mICE, Ich-3, and hTX. The predicted Asp residues of Ich-3 cleavage sites are indicated. The position of the absolutely conserved pentapeptide sequence QACRG, which includes the catalytic cysteine residue, is indicated above the bars. The black bars and hatched bars represent p20 and p10 domains, respectively. in high stringency condition containing 17.5% SDS under which only sequences with more than 95% identity would hybridize (see "Material and Methods"). The ich-3 probe hybridizes to a single band of 1.4 kilobases mRNA, which is very similar to the sizes of ice mRNA and the reported size of TX/ICErelII/ICH2 (Fig. 3). The expression pattern of ich-3 is very similar to that of ice, except that of brain where ice expression can be easily detected. ich-3 expression can be detected in adult heart, lung, thymus, spleen, and kidney. Higher levels of ich-3 is found in adult thymus and spleen, which is similar to the distribution of TX/ICH-2/ICE rel II mRNA (23,25,26). Only RT-PCR can detect low levels of ich-3 expression in brain (Fig. 3B).
The expression levels of ich-3 in wild type mice in control conditions are very low. Endotoxins (LPS) are strong inducers of proIL-1␤ synthesis and mature IL-1␤ secretion. To examine if the expression of ich-3 can be induced by LPS, we prepared RNA from mice either before or 5 h after injection of a lethal dose LPS (40 mg/kg of body weight). Northern blot analysis showed that ich-3 RNA expression is dramatically induced at least 30-fold after LPS stimulation in thymus, lung, spleen, and kidney but not in brain where ich-3 expression is low in control mice (Fig. 3). ice transcription is not induced in spleen, kidney, lung, heart, and brain after LPS stimulation. ice mRNA is only induced significantly in thymus (Fig. 3). We also examined the levels of Ich-3 protein before and after LPS stimulation. Proteins were isolated from tissues of 7-10-week-old mice before and 4 and 20 h after LPS stimulation (40 mg/kg) and analyzed on Western blot using a monoclonal antibody that recognizes Ich-3 specifically. Ich-3 protein level is very low in Western blot of tissues isolated from control mice. In LPS stimulated mice, Ich-3 is detected as two proteins of 43 and 38 kDa. 43 kDa is very close to the predicted protein size of full-length ich-3 cDNA that we have. We know that these proteins are from ich-3 locus because the ich-3Ϫ/Ϫ mice that we have generated using gene targeting technique are specifically missing these two proteins. 4 LPS stimulation results in at least 20 -30-fold increase in levels of Ich-3 proteins. In spleen two additional bands of 30 and 26 kDa were detected that may be cleavage products of 43 or 38 kDa. Elevated levels of Ich-3 proteins are found at both 4 and 20 h after LPS stimulation, suggesting that LPS induces an immidiate and sustained increase in levels of Ich-3 proteins. In contrast, Western blot analysis of the same tissue samples using a polyclonal anti-ICE antibody detected no difference in expression before and after LPS stimulation (data not shown). These results suggest that ich-3 may be an important regulator of endotoxic shock in mice.
Overexpression of ich-3 Induces Apoptosis-To examine whether expression of ich-3 may be able to induce apoptosis, we used the same transient expression system used for ice and ich-1 (4,5). The mouse ich-3 cDNA was fused with the E. coli lacZ gene and placed the fused gene under the control of either chicken ␤-actin promoter (p␤actM24Z) or CMV promoter (pCMVM26Z). A mutant ich-3 was generated by site-directed mutagenesis in which the Cys residue in the conserved pentapeptide QACRG domain was converted to a Gly residue. This mutant was also fused to the lacZ gene and placed under the control of chicken ␤-actin promoter and named p␤actS6Z. These expression constructs were transfected into different culture cells, and their ability to induce apoptosis was tested by counting round dead blue cells to flat blue cell after X-Gal staining. As show in Fig. 4 and Table II, induction of Rat-1 cell apoptosis by ich-3 is as efficient as ice, both at about 97%. The percentage of cell death induced by ich-3-lacZ under the control of chicken ␤ actin promoter (p␤actM24Z) is similar to that of CMV promoter (pCMV26Z). ich-3 is less effective in inducing HeLa cell apoptosis (43%) than that of ice (94%). Because Rat-1 cells are not transformed, whereas HeLa cells are of tumor origin, this result suggests that ich-3-induced apoptosis may be more sensitive to apoptosis suppressors than that of ice. Consistent with this hypothesis, bcl-2 is somewhat more effective in suppressing ich-3-induced cell death than that of ice (Table II).
The cowpox virus gene crmA encodes a serpin that is a specific inhibitor of ICE (30). CrmA is much more effective in inhibiting ICE-induced apoptosis than ICH-1 L -induced apoptosis (5). CrmA is 10 4 -fold more potent in inhibiting ICE than CPP32 (17). These results suggest that CrmA can discriminate among different members of the ICE family. Because expres- FIG. 3. Expression of ich-3. A, induction of Ich-3 mRNA expression by LPS. Total RNAs were isolated from tissues of 7-10-week-old mice with or without LPS injection (40 mg/kg). The ϩ and Ϫ signs represent with or without LPS injection. 5 g of total RNA from each tissue was loaded per lane. Total RNAs isolated from thymus, spleen, kidney, lung, and brain, respectively. The amount of RNA was adjusted by ␤-actin blotting. B, expression pattern of ich-3 in different tissues from wild type mice. 1 g of total RNA was used for RT-PCR. The amount of PCR product was adjusted by ␤-actin. C, induction of Ich-3 protein expression by LPS. Proteins were isolated from tissues of 7-10-week-old mice before and after LPS injection. The numbers (0, 4, and 20) represent the hours after LPS injection. The amount of protein loaded from each sample was 60 g/lane except thymus, which was 20 g. The Western blot was probed with a rat anti-Ich-3 monoclonal antibody. sion of crmA can suppress trophic factor deprivation-induced neuronal cell death (6), CTL, Fas, and tumor necrosis factor ␣-induced apoptosis (8,9,11,18,31), it became critical to examine whether cell death induced by a particular ICE family member can be suppressed by CrmA. Expression constructs of ice and ich-3 were transiently transfected into Rat-1 cells stably expressing crmA (4), and the percentage of round dead blue cells among total blue cells was counted. As showed in Table II, ich-3 induced only 55% cell death in rat-1/crmA cells compared with 97% cell death in Rat-1 cells. Similar inhibition of cell death was observed in ice-induced cell death, which is reduced from 97 to 57%. Such experiments showed that CrmA is as effective in suppressing ich-3-induced cell death as that of ice.
Ich-3 Can Be Cleaved by Granzyme B in Vitro-Recent studies suggest that ice may be involved in GraB/perforin-mediated CTL-induced apoptosis. 2 CTLs induce apoptosis via granzymes in the presence of the pore forming protein perforin (14,15). It has been shown that ICE cannot be cleaved directly by GraB; nevertheless, ICE is important for GraB-induced apoptosis in at least certain cell types. 2 Other ICE family members may be processed by GraB, which in turn may directly or indirectly activate ICE. To examine whether GraB can cleave Ich-3, we expressed a His-tagged Ich-3 protein in E. coli. His-tagged Ich-3 protein purified from bacteria was mixed with or without active GraB and incubated at 30°C for 1 h. The cleavage products were identified by Western blot with a peptide antibodies against the p20 or a monoclonal antibody against p10 portion of Ich-3. As shown in Fig. 5 (right panel), the full-length Ich-3 band disappeared after incubation with GraB; a new 20-kDa band appeared that is detected by an anti-Ich-3 p20 antibody and a new 10-kDa band appeared that is recognized by a monoclonal antibody against p10 of Ich-3. The Ich-3 protein purified from bacteria is processed into p30 (perhaps by autocleavage) but not p20 and p10, whereas GraB can cleave Ich-3 into p20 and p10. Fragments around 30 kDa are the predicted sizes of the cleavages at Asp 59 and Asp 80 . An additional cleavage at Asp 281 will generate a 20-and a 10-kDa subunit. To confirm that p10 and p20 are generated from predicted p30 region, we expressed a T7-tagged p30 ich-3 in E. coli. Cleavage of this T7-tagged p30 generated predicted p20 and p10 subunits recognized by p20-and p10-specific antibody (Fig. 5, left panel). The cleavage of Ich-3 by GraB suggests a possible role played by Ich-3 in granzyme B/perforin-induced apoptosis.
Ich-3 Does Not Process proIL-1␤ Directly but Can Potentiate ICE for Cleavage of proIL-1␤-Mice with a homozygously disrupted ice gene are severely defective in generating mature IL-1␤ (19); hence, ICE plays a critical role in processing pro-IL-1␤ to mature IL-1␤. Because both mature IL-1␤ and ich-3 mRNA can be dramatically induced by LPS in vivo, we hypothesize that Ich-3 may directly or indirectly contribute to proIL-1␤ processing. A transient transfection assay combined with ELISA was used to test the ability of Ich-3 in cleaving proIL-1␤. A mouse proIL-1␤ expression construct pCMVS11 was cotransfected into COS cells together with either ice (p␤actM10Z) or ich-3 (pCMVM26Z) expression constructs. 24 h after transfection, secretion of mature IL-1␤ into the culture medium was quantified by an ELISA assay (Genzyme, Cambridge, MA). As show in Fig. 6, cotransfection of ice with proIL-1␤ resulted in a significant amount of secretion of mature IL-1␤. The amount of mature IL-1␤ generated by ICE ranging from 70 to 600 pg/ml is proportional to the amount of proIL-1␤ and ice used in the transfection. In contrast, when ich-3 was cotransfected with proIL-1␤, no significant secretion of mature IL-1␤ was observed, indicating that Ich-3 could not process pro-IL-1␤ by itself. Cotransfection of expression vectors of both ice and ich-3 with that of mouse proIL-1␤ into COS cells resulted in 50% increase in the amount of mature IL-1␤ secretion than ice alone and thus, Ich-3 can promote processing of proIL-1␤ by ICE. There is no increase of mature IL-1␤ production when ice was cotransfected with vector (p␤actGal) or mutant ich-3 (p␤actS6Z) (Fig. 6), suggesting that Ich-3 enzyme activity is required for promoting ICE function in generating mature IL-1␤ in vivo. DISCUSSION We described here the molecular cloning and characterization of murine ich-3, a new member of the ice/ced-3 family. The predicted Ich-3 protein is 373 amino acids long and contains the 100% conserved ICE family signature peptide QACRG. Five additional members of the ICE family have been identified (5,25,26). These ICE homologs can be classified into two different groups by their sequence homology: one group (ICE) is more homologous to ICE than to Ced-3 (TX/ICErelII/ICH2 and ICErelIII) and the other (Ced-3) is more homologous to Ced-3 FIG. 5. Cleavage of Ich-3 protein by granzyme B. 10 g of Histagged Ich-3 protein purified from E. coli was incubated with 20 ng of GraB in the presence of 10 mM dithiothreitol at 30°C for 1 h. The result was detected by Western blotting with a peptide antibody (Ab) against the p20 portion of Ich-3 and a monoclonal antibody against p10 portion of Ich-3.

TABLE II Overexpression of Ich-3 in tissue culture cells induces apoptosis
The constructs ␤-galactosidase control (pact␤GalЈ vector), ice-lacZ (Ice-lacZ fusion under ␤-actin promoter control ϭ p␤actM10Z), ich-3-lacZ (ich-3-lacZ fusion under cytomegalovirus promoter control ϭ pCMVM26Z), and mutant ich-3 (p␤actS6Z, mutant ich-3 under ␤-actin promoter control) were transiently transfected into different cell lines. 24 h (40 h in COS cells) after transfection, the cells were fixed and stained for X-Gal. The data (means Ϯ S.E.) are the percentages of round blue cells among total number of blue cells counted. The numbers in the parentheses are the number of blue cells counted. The data were from at least three independent experiments. ND, not determined.

Ich-3, an Upstream Regulator of ICE
than to ICE or equally homologous to ICE and Ced-3 (ICH-1, CPP32/YAMA, and MCH-2). Murine Ich-3 is more homologous to ICE than to Ced-3 and therefore belongs to the ICE group. The amino acid sequence of Ich-3 is 60% identical to TX, which is close to the identity shared by murine and human ICE (62% identity). The expression pattern of ich-3 is also similar to TX; both are expressed in many tissues except the brain. It is possible that Ich-3 is the murine equivalent of human TX. We cannot conclude at the moment, however, that ich-3 is in fact murine version of human TX, because TX has been shown to be able to cleave pro-ICE (23), whereas we have not found that Ich-3 can cleave pro-ICE in a similar assay (see below). The functions of ich-3 shown here are in general novel, even if Ich-3 and TX are comparable proteins. Expression of ich-3 mRNA is low in normal healthy tissues. The levels of Ich-3 proteins are generally undetectable on Western blots of tissues from healthy mice. LPS stimulation dramatically induces ich-3 mRNA and proteins, which persists at least 20 h after LPS stimulation. In contrast, expression of ice is not elevated in most tissues after LPS stimulation with the exception of thymus where its level is moderately elevated. Ich-3 protein is undetectable in normal conditions in mice. Upon stimulation by LPS, two proteins of 43 and 38 kDa are detected. Both proteins are products of ich-3 gene because a null mutation in ich-3 locus eliminates both proteins. 4 43 kDa is very close to the predicted protein size (42 kDa) generated from full-length ich-3 cDNA. The 38-kDa protein may be an alternatively spliced product of ich-3. These results suggest that Ich-3 may play a very important role in inflammatory responses. Consistent with its role in inflammatory responses, mice with a homozygous null mutation in ich-3 gene are resistant to LPS-induced septic shock. 4 Ich-3 proteins, however, are not likely to be directly involved in processing of pro-IL-1␤ for the following two reasons. First, there is no in vivo evidence of existence of another protease playing a significant role in pro-IL1␤ processing because ICE knock-out mice are at least 90% defective in processing pro-IL-1␤ (10,19). Second, expression of ich-3 in COS cells does not lead to pro-IL-1␤ processing di-rectly; rather it promotes processing of pro-IL-1␤ by ICE. This result suggests that Ich-3 is an upstream regulator of ICE. It is not clear, however, how Ich-3 activates ICE. The simplest possibility that Ich-3 directly cleaves ICE to activate, but it may not be true because we have consistently failed to observe cleavage of pro-ICE by Ich-3 either in enzymatic assay using GraB activated Ich-3 or in cells by double transfection. We hypothesize that there may be one or more intermediate step between Ich-3 and ICE. Expression of ich-3 in COS cells activates this intermediate step(s) which in turn activates ICE. This may also explain why we only observe 50% increase in mature IL-1␤ production when we co-express both ice and ich-3: because there is an intermediate step(s) that is in limited quantity in COS cells. This intermediate step may be another member of the ICE family. Alternatively, Ich-3 may activate ICE indirectly by inactivating an ICE inhibitor.
A question was raised whether the role of ICE is primarily in inflammation or apoptosis (19). It is clear now that ICE has functions in both processes because iceϪ/Ϫ cells are defective in both production of mature IL-1␤ and Fas-and GraB-induced apoptosis (10,19). 2 The same question can be asked for ich-3: expressing ich-3 can induce apoptosis, which indicates that ich-3 has the ability to induce apoptosis, which does not prove that it has a role in inducing cell death in vivo. ich-3Ϫ/Ϫ thymocytes are partially resistant to Fas-induced apoptosis, and ich-3Ϫ/Ϫ EF cells are resistant to GraB-induced apoptosis. These in vivo data are consistent with in vitro data presented here, which all suggest that Ich-3 is an upstream regulator of ICE.
Like iceϪ/Ϫ mice, a lethal dose of LPS fails to induce production of IL1 in the sera of ich-3Ϫ/Ϫ mice. The critical difference, however, is that ich-3-deficient macrophages and monocytes in vitro can produce mature IL-1␤ as well as wild type cells when stimulated with LPS and ATP (for macrophages) or LPS alone (for monocytes); thus, ich-3 mutant cells still have the normal ICE function, whereas ice-deficient macrophages and monocytes do not produce mature IL-1 when stimulated in vitro (19). These results suggest that Ich-3 may be an upstream regulator of ICE in vivo. When mice are stimulated with LPS, Ich-3 may be induced first and activated, which in turn indirectly activates ICE. This hypothesis is entirely consistent with the data presented here; Ich-3 does not process proIL-1␤ directly but does promote proIL-1␤ processing when ICE is present. The requirement for Ich-3 is bypassed in vitro, however, when cells are stimulated with a strong signal.