Activation of Caspase-12, an Endoplastic Reticulum (ER) Resident Caspase, through Tumor Necrosis Factor Receptor-associated Factor 2-dependent Mechanism in Response to the ER Stress*

When accumulation of a malfolded protein in the endoplastic reticulum (ER) is induced by various adverse conditions, such as hypoxia, glucose starvation, and perturbation of calcium homeostasis, cells respond to the stress by increasing transcription of genes encoding ER molecular chaperones, a process known as unfolded protein response. The signaling is initiated by IRE1s, ER stress sensors. Alternatively, excessive stress to the ER results in apoptosis. Caspase-12 is known to be essential for this ER stress-induced apoptosis. In this study, we analyzed the detailed regulatory mechanisms of IRE1s during ER stress. We identified c-Jun N-terminal inhibitory kinase (JIK) as a binding partner of IRE1α, and JIK was seen to modulate IRE1α-TRAF2 (tumor necrosis factor receptor-associated factor 2) complex formation and the resultant alteration to c-Jun N-terminal kinase signaling from IRE1s in response to ER stress. We also demonstrated that TRAF2 interacts with procaspase-12 and promotes the clustering of procaspase-12 and its activation by cleavage in response to ER stress. These results indicate that TRAF2 plays crucial roles not only in the signaling of the c-Jun N-terminal kinase pathway but also in activation of caspase-12 to transduce signals from IRE1s. Thus, we provide a missing link in the ER stress-induced apoptosis-signaling pathway, one which connects the stress sensor molecule IRE1 and the activation of caspase-12.

The endoplastic reticulum (ER) 1 is sensitive to alterations in homeostasis from a variety of different stimuli, such as glucose deprivation, perturbation of calcium homeostasis, and exposure to free radicals. Under such conditions, perturbation of protein folding and the accumulation of malfolded proteins in the ER induce ER stress (1). ER stress elicits two major cellular-protecting responses. One is the attenuation of protein synthesis, and the other is the up-regulation of genes encoding chaperones that facilitate the protein folding process in the ER known as the unfolded protein response (UPR). Both responses reduce the accumulation and aggregation of malfolded proteins in the compartments of the cells (1).
IRE1␣ (2) and IRE1␤ (3) are believed to be ER stress sensor proteins and play important roles in transducing the stress signals initiated by the accumulation of malfolded proteins from the ER to the cytoplasm and nucleus. IRE1s are known to participate in the UPR and control the expression of ER molecular chaperones. When cells are exposed to excess levels of stimuli causing ER stress, apoptotic signals are transduced from the ER and promote apoptotic cell death. It is reported that c-Jun N-terminal kinases (JNKs) are activated by the accumulation of malfolded proteins in the ER (4). JNKs constitute a family of signal transducers that are activated by a variety of exogenous stimuli, such as growth factor deprivation, Fas or Tumor necrosis factor ␣ (TNF ␣) treatments (5,6), anticancer drug treatments, and UV light irradiation. JNKs regulate gene expression through the phosphorylation and activation of transcription factors such as cJUN or the activator protein-1 family (7). The activation of JNK requires TNF receptor-associated factor 2 (TRAF2), a member of the TRAF family of proteins, which transduce signals from IREs. In addition, dominant-negative TRAF2, which is truncated in the N-terminal RING effector domain of TRAF2, inhibits the activation of JNK by signals from IRE1s (4).
During apoptosis induced by ER stress, caspase-12 is localized to the ER and is activated (8). The activation of caspase-12 is not mediated by other stimuli. Furthermore, it has also been reported that caspase-12-deficient mice are resistant to ER stress-induced apoptosis, but their cells are led to apoptosis in response to other stimuli (8). Although it has been shown that caspase-12 is activated during ER stress-induced apoptosis, the mechanisms of its activation by ER apoptotic signals are still unknown, and even less is known about how TRAF2 can transduce ER stress signals from IRE1s to its downstream signaling events. To address these issues in this study, we describe the identification and the characterization of interactions among the ER stress-associated molecules, and we present a picture of how they are coupled to the activation of this apoptotic signal cascade. CO 2 incubator and were used in all experiments. cDNA plasmids encoding human IRE1␣ and human IRE1␤ were kindly provided by Dr. R. J. Kaufman and Dr. D. Ron, respectively. Human JNK inhibitory kinase (JIK), human TRAF2, and mouse procaspase-12 expression vector were obtained by reverse transcriptional polymerase chain reaction in our laboratory. The procaspase-12 expression vector-fused HA or FLAG tag in its C-terminal were also engineered. Rabbit polyclonal anti-caspase-12 antibody was originally raised against recombinant polypeptide. The other antibodies were purchased from manufacturers, such as anti-FLAG antibody (Sigma), anti-HA epitope antibody, phosphospecific anti-JNK antibody, anti-JNK1 antibody, and anti-TRAF2 antibody (Santa Cruz Biotechnology, Santa Cruz, CA). In all of the experiments, typical results from at least three repeated experiments are shown.
Yeast Two-Hybrid System-To analyze the function of the IRE1 cytosolic domain in the apoptotic pathway, we employed a yeast twohybrid system to search for proteins that interact with IRE1. Yeast cells, strain Y190, were transformed with an expression vector encoding the GAL4 DNA-binding domain combined with the IRE1 kinase domain. Then the yeast-expressing bait protein was transformed with an expression vector containing the cDNA library fused to its DNA activation domain. We used Matchmaker cDNA library (mRNA sources: normal whole brain pooled from nine spontaneously aborted male/ female Caucasian fetuses aged 20 -25 weeks, 3.5 ϫ 10 6 independent transformants, CLONTECH). Screenings were performed according to the manufacturer protocol. As positive and negative controls, we used packaged transformants supplied with a screening system.
In Vitro Binding Assays-HEK293T cells were transfected with plasmids using LipofectAMINE 2000 transfection reagents (Life Technologies, Inc.) according to the manufacturer protocols. The total amount of transfected plasmid DNA was adjusted to the same levels within each individual experiment. Cells were harvested at 24 h after transfection and lysed with 0.2% Nonidet P-40 in Dulbecco's modified Eagle's medium-phosphate-buffered saline. Subsequently, 1 mg of soluble protein was incubated with 1 g/ml antibody for 2 h at 4°C, and proteins captured with this antibody were coprecipitated with G protein agarose (Life Technologies, Inc.). Immunoprecipitates or cell lysates were loaded onto appropriate SDS-polyacrylamide gels, electrophoresed, and immunoblotted with antibodies for detections.
JNK/Stress-activated Protein Kinase Assay-The activations of JNK in HEK293T cells transfected with mock vector or various expression vectors were examined at 3 h after treatment with or without stimulations. These experiments were performed using a stress-activated protein kinase/JNK assay kit (New England Biolabs). The detections of phosphorylated JNK under various conditions were also examined using cells transfected with JNK and various expression vectors. The immunoprecipitation was performed with a phosphospecific anti-JNK antibody followed by Western blotting with an anti-JNK1 antibody. Transfection, immunoprecipitation, and Western blotting were performed essentially as described above.
Metabolic-labeling Experiments-Metabolic labeling with 32 P i was performed as described in our previous paper (9). HEK293T cells in 6-well plates were incubated at 37°C for 3 h with or without ER stress after the addition of 32 P i . Radiolabeled lysates from each sample were immunoprecipitated with anti-FLAG antibody or anti-TRAF2 antibody. The immunoprecipitates were separated by 12% SDS-polyacrylamide gel electrophoresis and subsequently electrotransferred onto polyvinylidene difluoride filters (Millipore). The filters were then exposed to x-ray film for the detection of 32 P. To control for loading in this procedure, the filters were stained with the specific antibody after autoradiography.

RESULTS
Interactions among IRE1␣, TRAF2, and JIK-It is believed that IRE1s sense ER stress through their lumenal domains (10, Activations of JNK were also detected by immunoblotting analysis using the phosphospecific anti-JNK antibody. To control for loading, Western blotting analyses of lysates were performed with anti-JNK1 antibody (b). JIK promoted the activation of both JNK and c-Jun in response to ER stress, and the overexpression of mutant JIK suppressed these activations as much as did dominant-negative TRAF2. The relative intensities of protein bands were determined using the NIH Image software. 11). The cytosolic parts of IRE1s, the kinase domains and RNaseL domains, are reported to transduce ER stress signals to the downstream events to promote the transcription of ER molecular chaperones. These transduced signals are known to mediate the increase of gene expressions, such as ER molecular chaperones, prolyl peptidyl isomerases, and disulfide exchange proteins (12,13). In contrast, excess levels of stress in the ER result in apoptosis (14,15). We hypothesized that the cytosolic portion of ER stress sensor molecules can activate not only the UPR pathway but also the apoptotic-signaling pathway. To examine this hypothesis, the yeast two-hybrid systems were employed to search for genes that interact with the cytosolic kinase domain of IRE1␣. As a result of screening the human fetus brain cDNA library, JIK was identified as a possible binding partner of IRE1␣ (data not shown). It has been reported that JIK is a human STE20-related serine/threonine kinase and that JIK activity is decreased upon epidermal growth factor receptor activation, but it is not modulated by other exogenous stimuli, such as UV irradiation, TNF-␣, NaCl, H 2 O 2, and anisomycin treatments (16). However, there is no information about JIK in the ER stress signaling. Because JIK is implicated in the JNK signaling, we suspected that JIK also interacts with TRAF2 and forms complexes with both IRE1s and TRAF2.
To demonstrate the interactions of these molecules in mammalian cells, we performed coimmunoprecipitation experi-ments in HEK293T cells cotransfected with various combinations of expression plasmids for IRE1␣-FLAG, JIK-HA, TRAF2, and mock control. Immunoprecipitations of full-length JIK-HA with the anti-HA antibody and Western blot analyses with anti-FLAG or anti-TRAF2 antibodies revealed that JIK could be coimmunoprecipitated with both IRE1␣-FLAG and TRAF2 (Fig. 1, A and B). The results suggested that IRE1␣, TRAF2, and JIK form complexes and might influence the functions of one another. The same results were obtained when the order of precipitation was reversed (Fig. 1, A and B). During ER stress condition, it was reported that activated IRE1s could recruit TRAF2 to the ER (4). We also characterized the complex both in normal and the UPR-induced states. Treatments of 1.0 g/ml and 2.5 g/ml tunicamycin for 3 h resulted in no significant alteration of the bindings between JIK and IRE1␣ compared with that of no treated control. On the other hand, the complex formation between JIK and TRAF2 was facilitated by the same stimulations as described above (data not shown). These observations mean that TRAF2 is recruited to the JIK-IRE1␣ complex in response to the ER stress.
JIK Functions as a Regulator of the JNK-signaling Pathway during ER Stress-It is well known that ER stress activates JNKs (4). We speculated that JIK might play regulatory roles in JNK activation under ER stress conditions. Therefore, we measured the relative levels of JNK activity in cells that were treated with various manipulations by using an immune complex kinase activity assay. After treatments with 2.5 g/ml tunicamycin for 3 h, JNK was activated in 293T cells transfected with mock vectors. Overexpression of JIK resulted in the acceleration of JNK activation induced by treatment with tunicamycin. Alternatively, transfection of catalytically inactive mutant JIK(A181F183) (16) inhibited the activation of JNK by the same stress ( Fig. 2(a)). The cell lysates were also immunoprecipitated with phosphospecific anti-JNK antibody. Immunoprecipitated proteins were detected by immunoblotting analy-

FIG. 3. Implications of JIK in the ER stress-signaling pathway.
A, phosphorylation of TRAF2 increased by JIK coexpression. HEK293T cells were cotransfected with various plasmids as indicated, and metabolic-labeling experiments were performed as described under "Experimental Procedures" using anti-TRAF2 antibody. Increased phosphorylation of TRAF2 could be detected in the JIK-expressed cells compared with that in the untreated cells. The cell lysates were Western blotted with anti-TRAF2 antibody to verify similar levels of expression in all samples. B, the enhancement of the interactions between TRAF2 and IRE1␣ by JIK expression. HEK293T cells were transiently cotransfected with the indicated amounts (in g). We used antibodies in the indicated combinations for coprecipitation and detection. The expression of JIK strengthened the interactions between IRE1␣ and TRAF2.
FIG. 4. Altered interactions between TRAF2 and procaspase-12 under ER stress conditions. HEK293T cells were transfected with the indicated expression vectors and treated with or without tunicamycin (2.5 g/ml) for 3 h. Cell lysates were immunoprecipitated with anti-FLAG or anti-TRAF2 antibodies. These coprecipitated complexes were subjected to Western blotting using anti-TRAF2 or anticaspase-12 antibodies. ER stress signals reduced the interactions between TRAF2 and procaspase-12. Each lower panel shows Western blot to indicate the relative amount of protein expressions in each sample. sis using anti-JNK1 antibody. Transfection of wild JIK caused increases in the amounts of phospho-JNK, which is known as an active form of JNK during ER stress, but mutant JIK did not increase the amounts compared with the controls (Fig. 2(b)). The results were consistent with those of the JNK kinase activity assay as described above.
It has been reported that the cytoplasmic portion of IRE1s binds to TRAF2 that couples the ER stress sensor to JNK activation, and dominant-negative TRAF2 inhibits ER stressinduced c-Jun/JNK activation (4). As described above, we revealed that JIK activates the JNK pathway under the ER stress conditions and interacts with IRE1␣. Therefore, it is not inconceivable that JIK affects the function of TRAF2 through phosphorylation to regulate the signal transduction from IRE1s to JNKs during ER stress. To investigate whether JIK influences TRAF2, we examined the phosphorylation levels of TRAF2 with or without JIK coexpression by the metaboliclabeling method using 32 P i . As a result, transfection of wild JIK did not alter the expression levels of TRAF2, but the amounts of phosphorylated TRAF2 were significantly increased (Fig.  3A). At the same time in another coimmunoprecipitation experiment, we found that the overexpression of JIK also promoted interactions between TRAF2 and IRE1␣ (Fig. 3B). Taken together, these results suggested that JIK alters the status of TRAF2 phosphorylation and that the resultant-altered TRAF2 phosphorylation may change the interactions between IRE1␣ and TRAF2 themselves that might regulate the activity of JNK signaling under ER stress conditions.
Interactions between TRAF2 and Procaspase-12-It is still unknown whether the activation of JNK signaling by TRAF2 during ER stress is directly implicated in apoptosis. On the other hand, another group reported that ER resident caspase-12, which is one of the cystein-protease family, plays an essential role in ER stress-induced cell death (8). Procaspase-12 contains the caspase recruitment domain in its N-terminal region, which is known as a predicted prodomain (8). Because caspase recruitment domain is known as the domain that interacts with apoptosis-associated proteins, such as TRAF2, Apaf1, and other caspases (17), we tried to examine whether procaspase-12 is able to interact with TRAF2.
Total cell lysates from HEK293T cells transfected with pro-caspase-12 and/or TRAF2 expression vectors were immunoprecipitated using anti-TRAF2 antibody and were immunoblotted by anti-caspase-12 antibody. As shown in Fig. 4, procaspase-12 immunoreactive 60-kDa protein was detected only in TRAF2 and procaspase-12-cotransfected cells (Fig. 4, lane 3). No bands were detected in the cells transfected with either vector alone (Fig. 4, lanes 1 and 2). These results indicate that TRAF2 directly associates with procaspase-12 under normal conditions. On the other hand, treatments of TRAF2 and procaspase-12-cotransfected cells with 2.5 g/ml tunicamycin for 3 h inhibited the interactions between these molecules (Fig. 4, lanes  4 and 5). The same findings, such as the reduction of the coprecipitated procaspase-12 by TRAF2 during ER stress, were observed when IRE1␤ was overexpressed (Fig. 4, lanes 6 and  7). Reverse experiments also showed similar consistent results. We ascertained that the results documented above were not attributed to the reduction of protein expression by using direct Western blotting (Fig. 4, each lower panel). It is known that procaspase-12 is cleaved in response to ER stress (8). We examined whether TRAF2 overexpression affected the cleavage of procaspase-12. Our results showed that the coexpression of TRAF2 with procaspase-12, in fact, emphasized the cleavage of procaspase-12 ( Fig. 4(a)). This acceleration of the cleavage seems to be based on the activations of procaspase-12 as initiated by TRAF2.
Caspase-12 Homodimerization-Overexpression of procaspase-12 in cells encouraged the cleavage of caspase-12 itself (data not shown). We hypothesized that after procaspase-12 was recruited to the ER stress signal transducer in response to ER stress, this protease was activated through homodimerization and cleavage by some protein such as calpain (18). Initially, we tried to demonstrate procaspase-12 dimerization through coimmunoprecipitation experiments. Procaspase-12tagged HA epitope at its C-terminal was coexpressed in 293T cells with procaspase-12-tagged FLAG at its C-terminal. After immunoprecipitation of procaspase-12 with a FLAG antibody, coprecipitating HA-tagged procaspase-12 was detected by Western blotting with an anti-HA antibody. As we expected, procaspase-12-tagged HA was coprecipitated only from the lysates expressing both the HA-tagged procaspase-12 and the FLAG-tagged procaspase-12 (Fig. 5, lane 3). The same results

FIG. 5.
Dimerization of procaspase-12 and its cleavage. Homophilic associations of procaspase-12 were examined by the coprecipitation experiments. These oligomerizations or dimerizations of procaspase-12 were increased by various manipulations, such as treatments with tunicamycin for 3 h, IRE1␤, TRAF2, or dominant-negative TRAF2 coexpressions. To verify similar levels of procaspase-12 expression in all samples, the amounts of expression in the cell lysates were analyzed by immunoblotting with anti FLAG antibody (lower panel).
were obtained when the order of precipitation was reversed (data not shown). This procaspase-12 homodimerization was increased by both tunicamycin treatment and IRE1␤ expression (Fig. 5, lanes 4 -7). Moreover, the overexpressions of TRAF2 also reinforced this complex formation (Fig. 5, lanes 8  and 9). In addition, we found that TRAF2 increased the level of phosphorylated procaspase-12 in metabolic-labeling experiments (data not shown). These results revealed that TRAF2 plays a significant function in the oligomerization and phosphorylation of procaspase-12.
Associations of Mutant TRAF2 with Procaspase-12-TRAF2 can be divided into four subdomains. The N-terminal RING finger and the adjacent zinc finger motifs of TRAF2 are known to be required for nuclear factor B and JNK activation (19). The TRAFN, which forms ␣-helical coiled-coil structures, and the TRAFC subdomains independently interact with TRAF2 and TRAF2-associated proteins such as TNF-R2 and TRADD (20,21). It is reported that the binding domain of TRAF2 with IRE1s is the TRAFC domain (4). Thus, the distinct domains of TRAF2 are involved in the recruitment of signaling molecules and in the activation of downstream effectors.
The dominant-negative TRAF2 (RING finger deletion mutant, TRAF2-dR) was reported to be able to inhibit the activation of JNK signaling in response to ER stress (4). Consistent with this finding, the nuclear factor B activation ability of TRAF2 was also found to reside within its N-terminal half (19). In contrast, the self-association of procaspase-12 and its cleavage during ER stress was increased by TRAF2-dR in comparison with the mock control but not as much as by the full-length TRAF2 (Fig. 5). We also found that TRAF2 binds with procaspase-12dependent TRAFN domain by deletion mutant analysis (Fig. 6). The results suggested that the structure of TRAFN domain plays an important part for the ER stress signal transduction to caspase-12 activation. They also revealed that the ER stress signal transductions to activate the caspase-12 and to activate JNK pathway were mediated by distinct TRAF2 subdomains. However, against our expectations, none of the TRAF2 deletion mutants could get dominantnegative effects on the procaspase-12 activation (data not shown).

DISCUSSION
IRE1s are known as ER stress transducers transmitting ER stress signals from the ER to cytosol and to nuclei to fold malfolded proteins. In addition, IRE1s are also reported to activate the JNK-signaling pathway mediated by TRAF2 functions in response to the perturbation of protein folding in the ER (4). Also, JNK has been extensively demonstrated to be associated with apoptosis (22). Therefore, different types of signals are thought to be transduced from IRE1s, one a cell survival signal and another a death signal mediated by TRAF2.
In this study, we demonstrated that TRAF2 plays crucial roles not only in the signaling of the JNK pathway, which is controlled by JIK, but also in the activation of caspase-12 to transduce signals from IRE1s under ER stress conditions.
We identified JIK as a binding partner of IRE1␣ using a two-hybrid system. In mammalian cells, JIK had the potential to bind to both IRE1 and TRAF2, and the expression of JIK caused increases in the binding amounts of IRE1 and TRAF2 in a dose-dependent manner. These results suggested that JIK is implicated in the modification of IRE1s-TRAF2 complex formation to regulate the transduction of the signals that are initiated by perturbation of protein folding in the ER. This hypothesis is supported by the findings that overexpression of JIK increased JNK phosphorylation and JNK activities during ER stress and that the catalytically overexpressed inactive mutated JIK inhibited the activation of JNK in response to ER stress as much as the dominant-negative form of TRAF2. As phosphorylated TRAF2 was up-regulated in synchrony with the activation of JNK when JIK was overexpressed, this JIK action on JNK signaling was possibly because of increased phosphorylated forms of TRAF2. However, it is not known whether TRAF2 is phosphorylated directly or indirectly by JIK.
Previously, JIK was reported to inhibit the JNK activation induced by epidermal growth factor receptor (16). This finding is not consistent with our present data that JIK promoted the activation of the JNK-signaling pathway in response to ER stress. According to Tassi et al. (16), any stimuli except treatment with epidermal growth factor did not affect the JNK pathway, indicating that JIK may play diverse roles in modulating the signaling upstream of JNK pathway in response to various stimuli. However, at present the reasons for the discrepancy in the function of JIK are unclear. To clarify the mechanisms of the activation of JNK through JIK under ER stress conditions, further analyses are needed including identification of specific substrate(s) for JIK.
Caspase-12 is known to be essential for cell death induced by FIG. 6. Mapping of TRAF2 domains required for interaction with procaspase-12. A, expression vectors for various deletion mutants of TRAF2. These mutants were tagged with FLAG epitope at its N-terminal (TRAF2-dRZ, TRAF2-dRZC, and TRAF2-dRZN) or its Cterminal (TRAF2-dNC and TRAF2-dR). B, interactions between procaspase-12 and TRAF2 deletion mutants. These vectors were transiently cotransfected with the procaspase-12 expression plasmid. Coprecipitating FLAG-tagged mutant TRAF2s or procaspase-12 was detected by immunoblot analyses. Mutant TRAF2s containing the TRAFN domain bound with procaspase-12, but those lacking the TRAFN domain did not interact with procaspase-12 (panels (a) and (c)). Protein expressions in the cells were confirmed with Western blotting analyses (panels (b) and (d)). ER stress. Indeed, procaspase-12 is cleaved, and the activated forms are accumulated under ER stress conditions (8). However to date, the mechanisms of the activation of caspase-12 in response to ER stress have not been demonstrated. This study showed that TRAF2 plays an essential role in the activation of caspase-12. In unstressed cells, TRAF2 formed a stable complex with procaspase-12. The stimuli that induce ER stress led to the dissociation of procaspase-12 from TRAF2, and simultaneously dimerization (or oligomerization) of procaspase-12 was promoted. These findings raise the possible mechanisms that the dissociation of TRAF2 from caspase-12 is a trigger for the activation of caspase-12 during ER stress and that the resultant-free procaspase-12 is clustered to the ER.
Several recent studies demonstrate that procaspases, such as caspase-2, -8, and -10, can be activated through dimerization/oligomerization mediated through their prodomains. Specific adapter molecules are reported to be able to interact with these procaspases. For example, the prodomains of caspase-8 and -10 interact with the adapter molecule Fas-associated death domain protein (23)(24)(25). In a similar manner, caspase-2 is thought to bind the death receptor through the adapter RAIDD (24). It is suggested that the primary role of these adapter molecules may be to bring procaspase molecules into close proximity with each other to enable dimerization. The mechanism regarding the activation of procaspase-12 might be similar to these recent reports. Although we could not show the recruitment of procaspase-12 to IRE1s during ER stress in the present study, we cannot negate the possibility that procaspase-12 is recruited to IRE1s before oligomerization and that TRAF2 plays a role as an adapter molecule that recruit procaspase-12. Further studies are needed to elucidate the more detailed mechanisms responsible for the activation of caspase-12 focusing on the recruitment of procaspase-12 to IRE1s under ER stress conditions.
In conclusion, in this study we provide a missing link in the ER stress-induced apoptosis-signaling pathway, which connects between the stress sensor molecule IRE1 and the caspase-12. We demonstrate that TRAF2 is a key mediator that transduces the signals from the ER to cytosol during ER stress. Therefore, TRAF2 might become a target molecule with which we can try to control ER stress-induced apoptosis.