Caspase-2 at a glance

Caspase-2, formerly known as ICH-1 (Wang et al., 1994) or NEDD-2 (Kumar et al., 1994), is a highly conserved but functionally poorly defined member of the caspase family, a group of cystein-driven aspartate-directed endopeptidases that are involved in cell death, inflammation and differentiation (Li and Yuan, 2008). Caspase-2 contains an N-terminal caspase recruitment domain (CARD), followed by a large subunit containing the active site (p19) and a small subunit (p12). Thus, caspase-2 is most similar to caspase-9, the initiator of the intrinsic pathway of apoptosis (Li and Yuan, 2008). However, in contrast to ‘conventional’ initiator caspases, such as caspase-9 or the apical caspase of extrinsic apoptosis, caspase-8, caspase-2 does not process apoptosis effectors that need to be cleaved by initiators for their activation, such as caspase-3, -6 or -7 (Guo et al., 2002; Van de Craen et al., 1999), suggesting that caspase-2 has other substrates (see supplementary material Table S1 and poster). Functionally, caspase-2 has been implicated in the regulation of cell death that is induced by metabolic imbalance, DNA damage, endoplasmic reticulum (ER) stress, mitotic catastrophe and others (reviewed by Krumschnabel et al., 2009; Vakifahmetoglu-Norberg and Zhivotovsky, 2010). The precise function of caspase-2, that is whether it functions as an initiator or effector caspase, is still unknown, and studies on caspase-2 activation are confounded by the fact that caspase-2 is a substrate for caspase-8 as well as for caspase-3 (Paroni et al., 2001; Van de Craen et al., 1999).

Similar to other initiator caspases, caspase-2 is activated by proximity-induced oligomerization and trans-cleavage in vitro, and ectopic overexpression is sufficient for its activation in cells (Butt et al., 1998; Read et al., 2002). Endogenous caspase-2 becomes activated as part of different signaling pathways (see poster), some of which depend on the DNA damage-induced formation of the PIDDosome, a complex composed of the CARD- and death domain (DD)-containing adapter protein RAIDD (also known as CRADD) and the p53-inducible DD-containing protein PIDD (also known as LRDD) (Tinel and Tschopp, 2004). However, the significance of this platform in vivo has been challenged by genetic experiments that have demonstrated that PIDD and RAIDD are dispensable for caspase-2 activation (Kim et al., 2009; Manzl et al., 2009), suggesting that alternative modes of caspase-2 activation exist.

One such alternative mechanism for caspase-2 activation is possibly mediated by caspase-8 in the death-inducing signaling complex (DISC), which is formed upon CD95 (also known as FAS) receptor clustering (Lavrik et al., 2006; Olsson et al., 2009). The role of caspase-8-mediated caspase-2 activation is, however, controversial, because this event does not necessarily contribute to apoptosis downstream of CD95 (Lavrik et al., 2006). Hence, DISC-dependent caspase-2 activation might be a ‘bystander’ effect due to its incidental co-recruitment together with that of ‘non-canonical’ DISC components, such as RAIDD (Duan and Dixit, 1997) or receptor-interacting protein 1 (RIP1) (Ahmad et al., 1997).

Other alternative caspase-2 activation mechanisms have been described, including caspase-2 dimerization and subsequent auto-proteolysis, which is induced by K⁺ efflux in response to bacterial pore-forming toxins (Imre et al., 2012) or heat-shock-triggered protein aggregation (Tu et al., 2006). Furthermore, alternative mRNA splicing (Wang et al., 1994), post-translational modifications, including N-terminal acetylation (Yi et al., 2011) and phosphorylation, appear to impact on and fine-tune these activating events (see poster). The variety of activation mechanisms suggest that caspase-2 becomes activated under diverse conditions of cellular stress, as discussed below.

Caspase-2 has been implicated in the induction of cell death by pathogenic bacteria, such as Brucella, Staphylococcus aureus and Salmonella (Chen et al., 2011; Chen and He, 2009). Caspase-2-deficient murine macrophages are protected from Salmonella-induced cell death and show decreased processing of caspase-1, a major driver of inflammation upon infection (Jesenberger et al., 2000). Similarly, cells lacking caspase-2 are also protected from cell death induced by Staphylococcus aureus α-toxin, but the underlying mechanisms that are responsible for escape from cell death have not yet been defined (Imre et al., 2012).

In addition to bacterial infections, caspase-2 activation has also been described in the apoptotic response to viral infections by the single-stranded (ss) RNA Rhabdoviridae, such as Maraba virus or vesicular stomatitis virus (VSV). Both viruses are known to induce a strong ER or unfolded protein stress response (UPR), which can lead to apoptosis (Mahoney et al., 2011). Interestingly, by keeping protein levels of RAIDD low, the ER-stress response machinery appears to be able to delay caspase-2-dependent apoptosis upon infection (Mahoney et al., 2011). Under conditions of sustained ER stress, however, the ER stress response factor IRE1α (also known as ERN1), an ER transmembrane kinase-endoribonuclease (RNase), promotes the rapid degradation of microRNAs that target caspase-2 mRNA. This, in turn, causes a rapid induction of caspase-2 protein expression that might contribute to the induction of apoptosis (Upton et al., 2012). Mechanistically, apoptosis induction by caspase-2 has been linked to its cleavage of the pro-apoptotic BCL2 family protein BID into its active form tBID (Guo et al., 2002), and thymocytes or mouse embryonic fibroblasts (MEFs) from Bid^−/− mice have been reported to resist ER-stress-dependent apoptosis (Upton et al., 2008). However, as the microRNAs that are inactivated by IRE1α also repress multiple pro-apoptotic BCL2 family proteins that have been previously implicated in ER-stress induced apoptosis, such as BIM (miR-17) or PUMA (miR-125b) (Le et al., 2011; Xiao et al., 2008), it will be important to assess the functional significance of caspase-2 induction in response to ER stress.

Most reports implicate caspase-2 in controlling cell death in response to DNA damage (reviewed by Krumschnabel et al., 2009). But although it has been shown that caspase-2 and the PIDDosome, as well as BID, become activated in response to DNA damage, genetic analysis in mice failed to reveal an essential function for any of these genes in normal physiology or in response to DNA damage (Kaufmann et al., 2007; Manzl et al., 2009; our unpublished data). Although the possibility of compensatory events taking place in knockout mice cannot be entirely excluded, these data suggest that these genes act redundantly to contribute to the robustness of the DNA damage response.

Such a redundant mechanism has been identified in p53-mutated zebrafish, in which caspase-2 controls cell death in response to DNA damage downstream of the DNA damage response pathway. As in other organisms, inactivation of p53 function causes resistance to radiation in Danio rerio, which can be overcome by inhibition of the checkpoint kinase 1 (Chk1) (Sidi et al., 2008). The functional abrogation of Chk1 results in the failure of the G2-M checkpoint and premature entry into mitosis, leading to aberrant mitosis and apoptosis after DNA damage (see below). Under these conditions, cell death has been found to be independent of several well-known apoptosis regulators, including caspase-3, -8 and -9, and could not be inhibited by overexpression of BCL2, but requires caspase-2 and the PIDDosome components PIDD and RAIDD (Ando et al., 2012; Sidi et al., 2008). Activation of caspase-2 under these conditions (i.e. irradiation upon inhibition of Chk1 in the absence of functional p53) appears to be controlled by ataxia telangiectasia mutated (ATM) kinase, both in zebrafish and in human cancer cell lines (Ando et al., 2012; Sidi et al., 2008). ATM can directly phosphorylate PIDD, triggering the assembly of the PIDDosome and activation of caspase-2 (Ando et al., 2012). Although epistasis clearly posits that Chk1 acts as an inhibitor of the ATM-dependent activation of the PIDDosome, the mechanistic nature of this inhibition remains to be understood. Elucidating this mechanism is important to improve our understanding of the cellular responses upon DNA damage in the presence or absence of functional p53 and might help to predict whether the application of pharmacological Chk1 inhibitors will radiosensitize tumor cells with a given genetic makeup (reviewed by Garrett and Collins, 2011).

In addition to the proapoptotic protein BID, only a few other caspase-2 substrates with a link to cell death regulation have been identified so far (see supplementary material Table S1 and poster), including the p53-related transcription factor p63 (Jeon et al., 2012). In response to DNA damage, ΔNp63, a splice variant that inhibits the proapoptotic transcriptional activity of full length p63 (the TAp63 isoform) by dimerization in a dominant-negative manner, is modified by the ubiquitin-like protein insulin-stimulated gene 15 (ISG15), which is required for its subsequent cleavage and inactivation by caspase-2. This cleavage leads to the nuclear export of the inhibitory fragment, allowing TAp63 to induce transcription of relevant target genes like proapoptotic PUMA and NOXA (also known as BBC3 and PMAIP1, respectively) (Jeon et al., 2012). In a xenograft tumor model, no difference in growth was observed between cells expressing either wild-type ΔNp63, a version of TAp63 resistant to caspase-2 cleavage, or a version of TAp63 that is defective for ISG15 conjugation. Upon DNA damage, however, wild-type cells are more sensitive (Jeon et al., 2012), suggesting that TAp63 has an important role in the proapoptotic response to DNA damage, and that this requires caspase-2 (see poster).

Caspase-2 is the sole caspase that is able to translocate from the cytosol into the nucleus upon activation, corroborating the hypothesis that it might act as a nuclear protease to orchestrate the response to DNA damage (Baliga et al., 2003; Colussi et al., 1998). The role of nuclear localization is still debated (reviewed by Krumschnabel et al., 2009) and sophisticated imaging techniques have revealed that the bulk of caspase-2 activation in response to various stressors, including DNA damage (Bouchier-Hayes et al., 2009), is in the cytoplasm. Whereas the augmentation of p53 signaling that is mediated by caspase-2 appears to rely on the cleavage of the p53 inhibitor MDM2 in the cytoplasm (Oliver et al., 2011), as discussed below, processing of ΔNp63 conceivably occurs in the nucleus, suggesting that caspase-2 acts in the cytoplasm to control p53 and in the nucleus in the p63 pathway (see poster).

Cell death associated with an abnormal mitosis is often referred to as mitotic catastrophe. Confusingly, a consensus for the definition of ‘mitotic catastrophe’ is missing and the term carries various meanings (reviewed by Vitale et al., 2011). Caspase-2 has been linked to cell death that is induced by aberrant mitosis in cancer cells that transiently arrest in prometaphase after cell fusion (Castedo et al., 2004) and to cell death by abnormal mitosis that has been induced by microtubule-interfering agents, such as vincristine and Taxol (Ho et al., 2008), but the underlying mechanisms remain to be fully defined. Several lines of evidence support the notion that caspase-2 can trigger cell death that is induced by abnormal mitosis, but how this ‘abnormality’ is sensed and translated into caspase-2 activation is still unknown. One possible mechanism is the inhibition of caspase-2 auto-cleavage (by phosphorylation of the conserved residue Ser340 located within the linker region connecting its large and small subunit) by cyclin-dependent kinase 1 (CDK1) (Andersen et al., 2009). Mitotic arrest is likely to affect the propensity of caspase-2 to become activated: the degree of its phosphorylation might be altered by the balance of CDK1 and protein phosphatase 1 (PP1), which can dephosphorylate Ser340 (Andersen et al., 2009) to counteract the inhibitory phosphorylation by CDK1. Recently, caspase-2-deficient MEFs have been shown to accumulate micronuclei and hyperploidy faster than wild-type cells, indicating that they are genetically unstable (Dorstyn et al., 2012). Whether caspase-2 deficiency causes micronucleation because of its postulated role in the DNA damage checkpoint or because of its role in response to mitotic errors remains to be established.

Despite the evidence that caspase-2 controls cell death in response to DNA damage, caspase-2-deficient mice develop without gross phenotypes (O'Reilly et al., 2002). Mice lacking caspase-2 show signs of premature aging (Zhang et al., 2007) and a higher number of oocytes (Bergeron et al., 1998). The premature aging phenotype appears to be caused by increased oxidative stress in aged caspase-2-deficient mice (Shalini et al., 2012), which was found to correlate with decreased expression levels of the cysteine sulfinyl-reductases sestrin-2 and -3, reduced activity of the FOXO transcription factors FOXO1 and FOXO3a and/or increased activation of p53 and p21 (encoded by Cdkn1a) (Shalini et al., 2012). In contrast to aged cells, MEFs that have been derived from young Casp2^−/− mice show an impaired activation of p53 and p21, leading to evasion of senescence (Dorstyn et al., 2012). Therefore, the increased p53 activation noted in aged animals might be a consequence of changes that are induced by lack of caspase-2 in young mice that lead to accumulation of damage over time and subsequently increased p53 activity, which no longer is co-regulated by caspase-2 function. This phenomenon might be driven by the lack of an augmentative function of this protease on p53 stability (see below).

Caspase-2 is crucial for oocyte apoptosis in the mouse and in Xenopus laevis (Bergeron et al., 1998; Nutt et al., 2005). Work based on Xenopus oocytes and egg extracts has shown that pentose-phosphate pathway (PPP)-driven generation of NADPH appears to be critical for oocyte survival (Nutt et al., 2005). Engagement of the PPP by addition of glucose 6-phosphate could prevent oocyte death by inhibiting caspase-2 due to its phosphorylation at Ser135 (Ser164 in human). This event is catalyzed by calcium/calmodulin-dependent kinase II (CaMKII; encoded by CAMK2G) and leads to the subsequent binding of caspase-2 to 14-3-3ζ, which prevents caspase-2 activation (Nutt et al., 2005). Nutrient depletion promotes the release of caspase-2 from 14-3-3ζ, which, in turn, allows the dephosphorylation of Ser135 by PP1, relieving caspase-2 inhibition and causing cell death (Nutt et al., 2009). The ability of 14-3-3ζ to interact with and inhibit caspase-2 in oocytes and human cancer cell lines (Andersen et al., 2011) is regulated by acetylation. The control of the acetylation status of 14-3-3ζ is exerted partly through the activity of the deacetylase sirtuin 1 (SIRT1) that can respond to glucose 6-phosphate by an unknown mechanism (Andersen et al., 2011). Intriguingly, whereas the inhibition of SIRT1 in cancer cells is sufficient to drive caspase-2 release from 14-3-3ζ, it has been reported to not be sufficient to trigger cell death in the absence of another proapoptotic stimulus, for instance the microtubule poison Taxol (Andersen et al., 2011). How the metabolic control of caspase-2 intersects with other pathways that lead to caspase-2 activation is an important but still unresolved question.

On the basis of its reported proapoptotic properties, caspase-2 is considered to act as a tumor suppressor (Kumar et al., 1995). Consistently, reduced expression of caspase-2, which is encoded on human chromosome 7q in a region that is frequently lost in human tumors, has been reported in acute myeloid leukemia (AML) and acute juvenile lymphoblastic leukemia (ALL) patients that are refractive to therapy and have a poor prognosis (Holleman et al., 2005; Johansson et al., 1993; Mrózek, 2008). Reduced caspase-2 protein levels have also been reported in a meta-analysis of published data sets for Burkitt lymphoma (BL), mantle cell lymphoma (MCL), chronic lymphocytic leukemia (CLL) and hairy cell leukemia, as well as in solid tumors, such as hepatocellular carcinoma, gastric and ovarian cancer, as well as metastasizing tumors of the brain (Kumar, 2009; Ren et al., 2012).

Consistent with a tumor suppressive function of caspase-2, MEFs derived from Casp2^−/− mice that are transformed with the oncoproteins E1A and Ras show increased proliferation in vitro and are more tumorigenic when they are transplanted into nude mice (Ho et al., 2009). These cells also show an impaired arrest in response to irradiation, as well as decreased cell death (Ho et al., 2009). Failure to arrest the cell division cycle arrest in response to DNA damage (Ho et al., 2009) might lead to increased genomic instability and chromosomal aberrations, which can promote tumorigenesis. Consistent with this idea, it has been found that lymphomagenesis caused by overexpression of Myc, which also promotes DNA damage, is accelerated in Casp2^−/− mice (Ho et al., 2009; Manzl et al., 2012). Surprisingly, however, in other models of DNA-damage-induced tumorigenesis, for example in 3-methyl-cholantrene-driven fibrosarcomas or fractionated irradiation-driven thymic lymphomas, loss of caspase-2 does not accelerate tumorigenesis, which argues against a general role of caspase-2 in tumor suppression (Manzl et al., 2012). It remains to be determined whether caspase-2 can suppress other forms of oncogene-driven tumor formation, in addition to those mediated by Myc.

The tumor suppressive capacity of caspase-2 has also been linked to its role in regulating cell death upon oncogenic stress because the strong selective pressure to inactivate the p53 pathway in Myc-induced B cell lymphoma (Ho et al., 2009) is reduced in the absence of caspase-2 (Manzl et al., 2012). This can be explained by an intrinsic defect in p53 activation or stabilization in developing B cells in such mice (Dorstyn et al., 2012; Ho et al., 2009), and could be due to the loss of caspase-2-dependent cleavage of the E3 ligase MDM2, which converts it from an inhibitor into an activator of p53 (Oliver et al., 2011). Additionally, caspase-2 might also have a direct, but less-well understood, effect on the translation of Cdkn1a mRNA (which encodes p21) (Sohn et al., 2011).

In summary, the involvement of caspase-2 in cell death initiation appears to be restricted to a rather limited number of cellular conditions, such as p53 deficiency, aberrant mitosis or nutrient deprivation. Despite the high degree of conservation, caspase-2 is much less understood in terms of its regulation and function when compared with other caspases. It remains to be fully explored whether caspase-2 indeed acts simultaneously as an initiator and effector of apoptosis, an idea that is compatible with its proposed ancestral function. The basic cell death machinery that operated with only one caspase might have been replaced by more sophisticated mediators of cell death over the course of evolution, rendering caspase-2 largely redundant today. Of note, in some paradigms, we now believe caspase-2 does not exert a direct pro-death role, neither as initiator nor as an effector, but instead acts as ‘damage sensor’ mediating limited proteolysis for signaling, and that this reflects a more general phenomenon in caspase-2 biology. Caspase-2 might, therefore, be embedded in intricate cellular signaling pathways that allow cells to initiate apoptosis as the last resort in contrast to catastrophic signals, which require cellular suicide to be rapidly committed for the sake of the entire organism. Caspase-2 research will therefore remain challenging, but it is likely to unveil important cellular response pathways to different harmful processes, such as DNA damage, oxidative or metabolic stress and possibly others, that are yet to be linked to caspase-2 activity.