|Home | About | Journals | Submit | Contact Us | Français|
The complex process of apoptosis is orchestrated by caspases, a family of cysteine proteases with unique substrate specificities. Accumulating evidence suggests that cell death pathways are finely tuned by multiple signaling events, including direct phosphorylation of caspases, whereas kinases are often substrates of active caspases. Importantly, caspase-mediated cleavage of kinases can terminate prosurvival signaling or generate proapoptotic peptide fragments that help to execute the death program and facilitate packaging of the dying cells. Here, we review caspases as kinase substrates and kinases as caspase substrates and discuss how the balance between cell survival and cell death can be shifted through crosstalk between these two enzyme families.
Apoptosis is a form of programmed cell death characterized by distinct morphological changes that include cell rounding, membrane blebbing, cytoskeletal disassembly, chromatin condensation, and DNA fragmentation (Kroemer et al., 2008). These dramatic cellular alterations, which allow for packaging of the dying cell and its subsequent engulfment by neighboring cells or phagocytes, can be ascribed to the actions of caspases, a family of cysteinyl aspartate-directed proteases that cleave a wide range of cellular proteins (for a compiled list of published caspase substrates, refer to the CASBAH online database http://bioinf.gen.tcd.ie/casbah/) (Lüthi and Martin, 2007). Although the cleavage of many caspase substrates is required for the structural packaging of cellular contents during apoptosis, a subset of caspase substrates are signaling molecules whose cleavage alters their signaling properties to affect the internal environment of the dying cell. In turn, signaling molecules can modulate caspase function to positively or negatively alter the trajectory of the cell death program. Given the millions of reversible phosphorylation events necessary to maintain cellular homeostasis and to allow cells to adapt nimbly to changing internal and external environments, the bidirectional communication between caspases and the kinases/phosphatases that control the cellular phosphoproteome is of particular interest. This Review will consider the impact of caspase cleavage on kinase/phosphatase function, the ways in which phosphorylation can alter both caspases and their potential substrates, and the ways in which these classes of signaling molecules are linked to control cell death and survival.
Caspases are synthesized as inert zymogens whose activation is triggered by a diverse array of internal and external cues (reviewed in Li and Yuan, 2008). Upon receipt of apoptotic stimuli, cells activate initiator caspases (for example, caspase-2, -8, -9, and -10) that, in turn, proteolytically cleave and activate effector (also called executioner) caspases (for example, caspase-3, -6, and -7). Once active, effector caspases proteolytically cleave a range of substrates, leading to the dismantling of the dying cell (Fischer et al., 2003).
Procaspases contain an N-terminal prodomain, as well as sequences encoding the large (p20) and small (p10) subunits of the mature protease. The initiator caspases are characterized by long prodomains that serve as platforms for the recruitment of activating adaptor proteins. The prodomains of caspase-2 and -9 contain a caspase recruitment domain (CARD), whereas caspase-8 and -10 possess two tandem repeats of the death effector domain (DED). In either case, these domains interact homotypically with adaptors that promote caspase activation through a mechanism of induced proximity, wherein the close juxtaposition of two caspase molecules leads to the formation of an active caspase tetramer containing two small and two large subunits.
Activation of the initiator caspases may occur through either an extrinsic or an intrinsic pathway (reviewed in Danial and Korsmeyer, 2004). In the extrinsic pathway, engagement of cognate ligands with death receptors (for example, Fas) induces receptor trimerization and subsequent recruitment of death domain (DD)-containing adaptor proteins, such as Fas-associated death domain (FADD), to corresponding death domain motifs on a cytoplasmic region of the death receptors. The resulting death-inducing signaling complex (DISC), in turn, recruits, oligomerizes, and thereby activates zymogenic caspase-8 (or caspase-10) through homotypic interactions between the death effector domain within the caspase and the related death effector domain within the adaptor protein. Active caspase-8 can directly cleave and activate the effector caspases (for example, caspase-3) and/ or engage the intrinsic apoptotic pathway through cleavage of the Bcl-2 homology 3 (BH3)-only protein Bid. Cleaved Bid (tBid) translocates to the mitochondria, where it triggers activation of the intrinsic apoptotic pathway by promoting activation of the Bcl-2 proteins Bax and Bak, which induces mitochondrial outer membrane permeabilization (MOMP) and release of proapoptotic mitochondrial constituents into the cytoplasm (Figure 1).
The cellular decision to initiate apoptosis in the intrinsic pathway reflects a balance between proapoptotic and prosurvival Bcl-2 family proteins (reviewed in Chipuk and Green, 2008). Classified based on blocks of sequence homology, the proapoptotic Bcl-2 family members are subdivided into BH3-only proteins (for example, Bim, Bid, and Bad) and multidomain proteins (for example, Bax and Bak) that contain various blocks of homology. In response to apoptotic stimuli (such as DNA damage), BH3-only proteins are activated and, directly or indirectly, promote oligomerization of Bax/Bak, permeabilization of the mitochondrial outer membrane, and release of factors from the intermembrane space of the mitochondria (most notably cytochrome c). Prosurvival Bcl-2 family members (for example, Bcl-2, Bcl-XL, and Mcl-1) counteract this effect by sequestering proapoptotic family members. Once in the cytoplasm, cytochrome c promotes oligomerization of the caspase recruitment domain-containing adaptor protein Apaf-1. Heptameric Apaf-1 recruits the zymogenic form of caspase-9 through a CARD-CARD interaction and forms the apoptosome, which leads to dimerization-induced activation of caspase-9. Caspase-9 in turn cleaves effector caspases, leading directly to their activation (Figure 1).
The molecular mechanisms leading to activation of caspase-2 remain somewhat elusive. However, various lines of evidence have placed it upstream of the mitochondria in the intrinsic pathway, as it is known to cleave and activate the latent proapoptotic activity of Bid (Bonzon et al., 2006). A current model suggests that a p53-inducible adaptor protein, PIDD (p53-induced protein with a death domain), is induced in response to certain apoptotic stimuli, such as DNA damage. PIDD then engages a second adaptor protein, RAIDD (RIP-associated ICH-1/CED-3 homologous protein with a death domain), resulting in the recruitment of caspase-2 through a CARD-CARD interaction and the formation of the “PIDDosome” (a structure that is analogous to the apoptosome), where caspase-2 dimerizes and is activated (Figure 1). By virtue of being at the apex of an apoptotic pathway, caspase-2 is classified as an initiator caspase. Yet, it does not appear to directly activate any effector caspases. Moreover, caspase-2 seems to preferentially recognize five amino acid residues within substrates, rather than four residues, the typical recognition motif for other caspases. Only a few bona fide caspase-2 substrates, other than Bid, have been identified and validated (for review, see Krumschnabel et al., 2009).
Although the adaptor protein-mediated activation process described above provides a good measure of control over caspase activity within cells, it has become increasingly clear that these activities can be finely tuned through additional binding partners (for example, inhibitor of apoptosis [IAP] proteins) and posttranslational modifications (for example, nitrosylation, phosphorylation, and ubiquitination). With respect to phosphorylation, both the caspase activation process and intrinsic enzymatic activity are under the control of modifying kinases and phosphatases (Table 1). This permits cellular flexibility in setting a threshold for the induction of apoptosis in response to alterations in the cellular environment (for example, after growth factor stimulation or changes in cellular metabolism) through changes in the activity of pro- or antiapoptotic kinases and phosphatases. Phosphorylation may also control caspase activity indirectly by controlling other apoptotic modulators (including caspase binding partners). Here, we consider control of caspases that can be exerted via their direct phosphorylation (Table 1).
Inhibition of caspase-9 may result in the failure of apoptotic induction even when cytochrome c has been released from the mitochondria. Indeed, protection from apoptosis despite cytohrome c release has been described, although such protection has been attributed only in part to caspase-9 inactivation (reviewed in Schafer and Kornbluth, 2006). The first report of caspase phosphorylation demonstrated that human caspase-9 could be phosphorylated by the prosurvival kinase Akt at serine 196 (S196) and thereby inhibited (Cardone et al., 1998). The universality of this finding was subsequently questioned when it was noted that S196 is not conserved in all species (for example, in mouse and dog) and that mouse caspase-9 is not demonstrably phosphorylated by Akt (Fujita et al., 1999).
Perhaps the most extensively characterized and functionally validated phosphorylation site on caspase-9 lies within its prodomain at threonine 125 (T125). This site is phosphorylated by both Erk (extracellular signal-regulated kinase) and Cdk1 (cyclin-dependent kinase 1), and this phosphorylation suppresses caspase-9 activation (Allan et al., 2003; Allan and Clarke, 2007). Given that the Erk pathway is upregulated in a variety of cancers, this inhibitory phosphorylation of caspase-9 may contribute to apoptotic (and therefore chemotherapeutic) resistance. Suppression of caspase-9 by Cdk1-mediated phosphorylation of T125 has been shown to forestall cell death during mitosis, though the physiological necessity for such inhibition has not been adequately explained (for review, see Clarke and Allan, 2009). One interesting possibility is that the rounding of cell shape at mitosis results in a loss of contact with neighboring cells and would trigger detachment-induced cell death (anoikis) without the inhibition of caspase-9. Additional kinases—DYRK1A (dual-specificity tyrosine-(Y)-phosphorylation-regulated kinase A1) and p38 mitogen-activated protein kinase (p38 MAPK)—also phosphorylate T125 (Laguna et al., 2008; Seifert et al., 2008; Seifert and Clarke, 2009), and a recent study has shown that DYRK1A protects retinal cells from death during neuronal development (Laguna et al., 2008).
Given that T125 is a target of multiple kinases, it is important to determine how dephosphorylation of this site is regulated. With an interleukin-2 (IL-2)-dependent mouse T cell line, serine/threonine protein phosphatase 1α (PP1α) was shown to be the relevant phosphatase for dephosphorylation of T125 and activation of caspase-9 after IL-2 deprivation (Dessauge et al., 2006). It will be interesting to determine whether PP1α plays a role in the regulation of caspase-9 in other cell types, particularly in human tumor cells. Caspase-9 has also been described as a substrate of c-Abl, protein kinase Cζ (PKCζ), and casein kinase 2 (CK2), which phosphorylate it at tyrosine 153 (Y153), serine 144 (S144), and serine 348 (S348), respectively (although CK2 does not phosphorylate human caspase-9, which lacks a CK2 consensus site) (Brady et al., 2005; McDonnell et al., 2008; Raina et al., 2005). Phosphorylation of Y153 is induced upon DNA damage and reportedly enhances caspase-9 activity, whereas PKCζ phosphorylation is triggered in response to osmotic stress and suppresses caspase activity (Table 1).
The most evolutionarily conserved caspase, caspase-2, is a substrate of CK2, calcium/calmodulin-dependent protein kinase type II (CaMKII), and DNA-dependent protein kinase (DNA-PK). CK2 phosphorylates caspase-2 at serine 157 (S157) in the region connecting the caspase recruitment domain and the large subunit, blocking the dimerization and subsequent activation of caspase-2 (Shin et al., 2005). Interestingly, the loss of S157 phosphorylation results in the autoactivation of caspase-2 in a PIDDosome-independent manner, raising the possibility that caspase-2 activation is persistently suppressed by constitutively active CK2 (Shin et al., 2005). CaMKII-mediated phosphorylation of caspase-2 at serine 135 (S135) was identified using extracts from eggs of the frog Xenopus laevis, where this phosphorylation can suppress the binding of RAIDD to caspase-2 and subsequent caspase-2 activation (Nutt et al., 2005). As the egg extracts progressively deplete their nutrient reserves over time, phosphorylation of S135 is diminished and the extracts finally undergo “apoptosis,” as characterized by cytochrome c release from the mitochondria and caspase-3 activation. These biochemical events of apoptosis can be forestalled by either stimulation of the pentose phosphate pathway or addition of NADPH, a product of the pentose phosphatase pathway, to the extracts. This indicates that CaMKII-mediated phosphorylation of S135 is metabolically regulated (Nutt et al., 2005). Furthermore, a recent study has shown that phosphorylation at the S135 site is antagonized by protein phosphatase 1 (PP1) (Nutt et al., 2009). Of note, a region homologous to S135 and its flanking residues appears to be conserved in mammalian caspase-2, including that of humans. Thus, it is possible that phosphorylation of caspase-2 may be an evolutionarily conserved means to link the metabolic status of the cell to cell death pathways, although this remains to be validated. Interestingly, mouse caspase-2 appears to be critical for the death of oocytes, whose viability also likely depends on their own nutrient reserves. This raises the possibility that the overall regulatory mechanisms linking caspase-2 and metabolism through the action of modulatory kinases are also conserved (Bergeron et al., 1998).
Very recently, it was shown that in response to sublethal levels of DNA damage, caspase-2 can be directly activated by phosphorylation at serine 122 (S122; officially S139 in the most updated protein sequence of caspase-2) by DNA-PK and that this phosphorylation mediates cell-cycle arrest (via the DNA damage checkpoint) rather than apoptosis (Shi et al., 2009). Interestingly, procaspase-2 forms an intranuclear complex with PIDD through the catalytic subunit of DNA-PK (DNAPKcs) without the involvement of RAIDD, a characteristic of the PIDDosome (Shi et al., 2009). This finding was the first demonstration of a nonapoptotic role for caspase-2. Further identification of a nuclear substrate that is cleaved by caspase-2 to signal the DNA damage response will be of great interest.
Several studies have documented the regulation of caspase-8 activity by tyrosine phosphorylation, rather than serine/theonine phosphorylation. The Src family tyrosine kinases Src, Fyn, and Lyn phosphorylate tyrosine 380 (Y380) (for the A isoform of caspase-8, Y397 is phosphorylated; Jia et al., 2008) and suppress caspase-8 activation (Cursi et al., 2006; Senft et al., 2007). Phosphorylation of Y380 inhibits Fas-induced caspase-8 activation and mediates antiapoptotic signaling in EGF-stimulated cells or colon cancer cells (Cursi et al., 2006), providing a means by which these kinases that are widely implicated in carcinogenesis might block apoptosis via the extrinsic pathway. In addition to Y380, Lyn was also found to phosphorylate Y465, thus inhibiting caspase activity (Jia et al., 2008). Importantly, Src-homology domain 2 (SH2)-containing tyrosine phosphatase 1 (SHP1) binds to a consensus motif (tyrosine 310-glutamatic acid-isoleucine leucine; Y310EIL) located in the large subunit of caspase-8 upon phosphorylation of Y310 and dephosphorylates both of these tyrosine residues.
In addition to its apoptotic role, caspase-8 has also been implicated in regulation of cell migration and adhesion, as the loss of caspase-8 leads to reduced cellular motility (Helfer et al., 2006). Recent studies have suggested that Y380 phosphorylation targets caspase-8 to membrane ruffles where it functions as a scaffold in a manner that does not require enzymatic activity (Senft et al., 2007; Barbero et al., 2008). Given its ability to prevent apoptosis and support cell migration, Y380 phosphorylation may play an important role in modulating caspase-8 function in embryonic development and cancer progression.
Although control of the onset of apoptosis is often exerted at the level of the initiator caspases, a number of signaling pathways interrupt instead the transmission of the apoptotic signal through direct inhibition of effector caspases. For example, p38 MAPK phosphorylates caspase-3 and suppresses its activity (Alvarado-Kristensson et al., 2004). Of note, the phosphorylation site serine 150 (S150), located within the large subunit of caspase-3, is conserved in other initiator and effector caspases (caspase-1, -2, -4, -5, -7, -8, and -9); p38 MAPK is also known to phosphorylate caspase-8 at serine 364 (S364) (Alvarado-Kristensson et al., 2004). Interestingly, association of protein phosphatase 2A (PP2A) with caspase-3 has been shown to counteract the effect of p38 MAPK by dephosphorylating S150, suggesting that activation of PP2A may be a key point of caspase-3 regulation (Alvarado-Kristensson and Andersson, 2005). The means of targeting PP2A to this site is unknown. Caspase-3 is also a target of PKCδ. Phosphorylation of caspase-3 by this kinase has been shown to result in enhancement of its proteolytic activity, though the phosphorylation site has not been identified (Voss et al., 2005). Importantly, both the PP2A Aα subunit and PKCδ are known substrates of caspase-3. Caspase-3-mediated cleavage of either enzyme elevates their activity, providing the opportunity for positive feedback in the pathway to amplify caspase-3 activation and bolster transmission of the apoptotic signal (see below).
There are multiple examples in which kinases and phosphatases—such as PKCδ and PP2A—are activated by caspase cleavage to enhance the cell death process. Such activation can result from separation of catalytic and inhibitory domains by the cleavage event, subcellular relocalization of the cleaved products, and/or altered substrate preferences after proteolytic cleavage. Although simple in principle, for caspase cleavage to produce protein fragments that are functionally competent, these cleavage fragments must retain the ability to fold properly into an active conformation, be susceptible to any posttranslational modifications required for catalytic activity, and be refractory to further degradation (by the proteasome, for example). Surprisingly, based on a recent proteomic study, a substantial fraction (~35%) of random protein fragments generated by the proteolytic activity of caspases display marked stability (Dix et al., 2008). Indeed, there are instances where the cleavage products are more stable than the full-length proteins from which they were derived (for example, Cdc25A, see below). In addition, caspase-mediated cleavage frequently occurs within a linker region between two structural domains of a protein, thereby producing two intact (or nearly intact) domains (Dix et al., 2008). These findings are consistent with the idea that some of the cleaved products may be functionally stable. Here, we describe a number of notable examples of kinase or phosphatase regulation by caspase cleavage (Figure 2; for the full list of caspase substrates, see Tables S1-S3 available online).
Proteins of the protein kinase C (PKC) family are lipid-activated serine/threonine protein kinases that modulate multiple cellular processes governing cell proliferation, differentiation, and apoptosis (Griner and Kazanietz, 2007). Structurally, all PKC isoforms consist of an N-terminal regulatory domain and a C-terminal kinase domain that are connected by a hinge region susceptible to caspase-mediated proteolytic cleavage (Figure 2). PKC proteins reportedly cleaved by caspase-3 include the δ, ε, ζ, θ, η, and μ isoforms (see Table S1). Such proteolytic cleavage separates the kinase and autoinhibitory domains, potentially resulting in the generation of constitutively active kinase fragments.
Since the first observation of caspase-mediated activation of ubiquitously expressed PKCδ (Emoto et al., 1995), the roles and functions of this kinase have been the most extensively studied of all PKC isoforms. Caspase-3 proteolytically cleaves PKCδ at different sites in the mouse and human PKC isoform (DILD327↓N in mouse and DMQD329↓N in human, where ↓ indicates the cleavage site in the amino acid sequence). This cleavage generates an active 40 kDa kinase fragment in response to various apoptotic stimuli, including DNA-damaging agents (Emoto et al., 1995; Ghayur et al., 1996; DeVries et al., 2002). Human PKCδ has also been demonstrated to be a target of caspase-2, which has only a few documented substrates (Panaretakis et al., 2005). Notably, ectopic expression of the catalytic fragment of PKCδ is sufficient to cause apoptosis, whereas the expression of catalytically inactive or cleavage-resistant mutant forms of the protein markedly attenuate or delay cell death that is triggered by genotoxic agents. These observations support a role for proteolytically activated PKCδ in the dismantling of the cell during apoptotic execution (Ghayur et al., 1996; DeVries et al., 2002; D'Costa and Denning, 2005; DeVries-Seimon et al., 2007). Because PKCδ possesses membrane-targeting motifs at the N terminus and a nuclear localization signal (NLS) at the C terminus, the kinase fragment liberated by caspase cleavage translocates to the nucleus (DeVries et al., 2002), where it phosphorylates several proteins, including lamin B, DNA-PK, p53, p73β, and Rad9 (Griner and Kazanietz, 2007) (Figure 3). Although it is not known which of the nuclear substrates are the most significant PKCδ apoptotic effectors, it has been demonstrated that nuclear accumulation of PKCδ is critical for induction of apoptosis. Disruption of the NLS abolishes apoptosis that is triggered by overexpression of the C-terminal kinase fragment, whereas nuclear retention of full-length PKCδ can cause apoptosis (DeVries-Seimon et al., 2007). However, recent studies have also shown that both tyrosine phosphorylation (tyrosines 64 and 155) and the catalytic activity of PKCδ also regulate PKCδ nuclear targeting, indicating that caspase-mediated cleavage may not be sufficient to support nuclear accumulation of PKCδ (DeVries-Seimon et al., 2007; Humphries et al., 2008). The caspase-cleaved PKCδ fragment also translocates to the mitochondria, where it phosphorylates and targets the antiapoptotic Bcl-2 family protein Mcl-1 for proteolytic degradation (D'Costa and Denning, 2005; Sitailo et al., 2006) (Figure 3). Importantly, studies using PKCδ knockout mice demonstrated that the resistance of PKCδ−/− cells to a wide range of proapoptotic stimuli is partly due to the suppression of mitochondrial cytochrome c release. This raises the possibility that in addition to phosphorylation of nuclear substrates, the cleaved PKCδ may also directly modulate mitochondrial membrane permeability by phosphorylating another substrate such as Mcl-1 (Leitges et al., 2001; Humphries et al., 2006). Finally, it was reported that caspase-cleaved, but not full-length PKCδ, phosphorylates 14-3-3 proteins in a sphingosine-dependent manner, which may result in the dissociation of 14-3-3 from its targets such as FOXO (forkhead box o) proteins (see below). Because 14-3-3 binding can activate the PKCε isoform, phosphorylation of 14-3-3 by cleaved PKCδ may result in cross-talk between PKC isoforms, decreasing the susceptibility of PKCε to activation (Saurin et al., 2008). In contrast, phosphatidylserine, an activator of full-length PKCδ, inhibits the catalytic activity of cleaved PKCδ, suggesting that proteolytic cleavage may also affect susceptibility of PKCδ to specific modulators (Hamaguchi et al., 2003).
The serine/threonine kinase Rho associated kinase-1 (ROCK1) is a widely expressed effector of the Rho GTPases Rho, Rac, and Cdc42. ROCK1 regulates actin cytoskeleton reorganization, cell contractility, and motility through phosphorylation of various cytoskeletal proteins, including the myosin light chain (MLC). Caspase-mediated activation of ROCK1, but not its related isoform ROCK2, is required for the membrane blebbing that is typically observed in cells undergoing apoptosis (Coleman et al., 2001; Sebbagh et al., 2001). Accordingly, a pharmacological ROCK1 inhibitor or the expression of a dominant negative ROCK1 mutant protein prevents membrane bleb-bing (Coleman et al., 2001; Sebbagh et al., 2001). Caspase-3 cleaves (DETD1113↓G) and separates ROCK1 into an N-terminal kinase domain and a C-terminal autoinhibitory region (Figure 2). Thus, the proteolytic activity of caspase-3 generates a constitutively active kinase fragment that does not require Rho for its activation. During Fas-induced apoptosis, robust phosphorylation of MLC is observed independently of Rho activity and is correlated with ROCK1 cleavage. This phosphorylation is blocked by a ROCK1 inhibitor, suggesting that the aberrant kinase activity of the C-terminal ROCK1 fragment causes dysregulated cell contractility through hyperphosphorylation of MLC (Sebbagh et al., 2001). Interestingly, when pretreated with a ROCK1 inhibitor, Fas- (or TNFα-) stimulated cells still undergo normal apoptosis. However, this apoptotic cell death is accompanied by caspase activation and chromatin condensation but not by membrane blebbing (Coleman et al., 2001; Sebbagh et al., 2001), illustrating the linkage of a single event in apoptotic progression to a specific caspase cleavage product.
Mammalian homologs of the budding yeast Ste20 (Sterile 20) serine/threonine kinases can be divided into two subfamilies: p21-activated protein kinases (PAKs) and germinal center kinases (GCKs). Like ROCK1, PAK2 is widely expressed and is activated by the Rho GTPases Rac and Cdc42 to regulate cytoskeletal reorganization. Also similar to ROCK1, the PAK2 kinase has an N-terminal regulatory domain that contains a GTPase binding domain and a C-terminal catalytic domain (Figure 2). Two early studies independently demonstrated that PAK2, but not other PAK isoforms, undergoes apoptotic caspase-mediated proteolytic cleavage between the regulatory and catalytic domains (SHVD212↓G) to release an active catalytic fragment (Lee et al., 1997; Rudel and Bokoch, 1997). Ectopic expression or microinjection of the catalytic fragment of PAK2, but not full-length PAK2, triggers morphological manifestations of apoptosis (including cell rounding, membrane blebbing, and chromatin condensation) independently of mitochondrial cytochrome c release and caspase activation (Lee et al., 1997; Vilas et al., 2006). Conversely, expression of a catalytically inactive mutant of this fragment markedly delays the apoptotic morphological changes induced by Fas stimulation (Lee et al., 1997; Rudel and Bokoch, 1997). Together, these results strongly suggest that caspase-mediated activation of PAK2 regulates a signaling pathway that contributes to the dismantling of the cell.
Direct substrates of caspase-activated PAK2 remain to be identified. It should be noted that like ROCK1, which phosphorylates MLC at threonine 18 (T18) and serine 19 (S19), PAK2 can also phosphorylate MLC at S19 under nonapoptotic conditions, raising the possibility that caspase-activated PAK2 might trigger the same pathway as cleaved ROCK1 (Chew et al., 1998; Croft et al., 2005). That said, a constitutively active mutant version of the full-length PAK2 protein does not cause membrane blebbing in the presence of a caspase inhibitor, whereas a cleaved fragment of ROCK1 does (Coleman et al., 2001). Whether this represents different signaling properties of PAK2 and ROCK1 or distinct biological activities of cleaved and full-length PAK2 is unclear. Moreover, activation of the c-Jun N-terminal kinase (JNK) is also observed upon expression of the PAK2 catalytic fragment, suggesting that caspase-cleaved PAK2 may phosphorylate key targets other than MLC to promote apoptotic alterations in cell morphology (Vilas et al., 2006). Interestingly, full-length PAK2 is found in the cytoplasm, whereas the caspase-cleaved PAK2 kinase fragment translocates to the nucleus due to loss of an N-terminal nuclear export signal (NES) and the presence of a C-terminal NLS (Jakobi et al., 2003). This nuclear translocation may account, at least in part, for the chromatin condensation induced by the catalytic fragment of PAK2. In addition, another study has shown that caspase-cleaved PAK2 undergoes posttranslational myristoylation at the caspase-cleavage site, targeting the fragment to the plasma membrane and membrane ruffles (Vilas et al., 2006). As full-length PAK2 does not appear to promote apoptosis but rather supports cell survival in some settings (Jakobi et al., 2001), it would be interesting to determine how caspase-activated PAK2 and active full-length PAK2 trigger distinct signaling cascades.
The serine/threonine MST (mammalian Ste20-related) kinases are mammalian homologs of Ste20 kinases and the Drosophila kinase Hippo, and belong to the GCK subfamily (and thus, are distantly related to PAK2). MST kinases are ubiquitously expressed in a wide variety of tissues and are involved in the regulation of diverse cellular functions including morphogenesis, cell migration, proliferation, and apoptosis. In an early study, it was observed that the induction of apoptosis by various proapoptotic stimuli is accompanied by the activation of a 36 kDa kinase that phosphorylates myelin basic protein (MBP) in vitro but is distinct from caspase-cleaved PKCδ (Lu et al., 1996). This kinase was biochemically purified and identified as a catalytic fragment of MST1 (Lee et al., 1998). Later studies have demonstrated that MST2 and MST3 are also proteolytically activated by caspase-3 in response to a variety of proapoptotic stimuli (Table S1). In healthy cells, intrinsic MST kinase activity is regulated, at least in part, by dimerization and phosphorylation. However, caspase-mediated cleavage markedly increases activity of the MST kinase by ~10-fold and also causes translocation of the cleaved kinase fragment to the nucleus (Creasy et al., 1996; Lee et al., 2001). Like the other kinases described above, MST kinases consist of an N-terminal kinase domain and C-terminal regulatory domain. Though unique to the MST kinases, the regulatory domain contains both a dimerization motif and two NESs (Figure 2). Two cleavage sites in MST1 (DEMD326↓S, TMTD349↓G) and one cleavage site in MST2 (DELD322↓S) and MST3 (AETD313↓G) have been documented (Table S1). Importantly, all of the cleavage sites in MST kinases are located between the catalytic and regulatory domains (Figure 2). Thus, in a familiar paradigm, caspase-mediated cleavage results in removal of the regulatory domain and subsequent translocation of the constitutively active catalytic fragment to the nucleus (Lee et al., 2001) (Figure 3).
Proapoptotic functions of MST1 and MST2 have long been recognized, although their physiological cellular substrates and signaling cascades are only now being elucidated. Overexpression of MST1 or its N-terminal catalytic fragment alone induces cell rounding and chromatin condensation in many cell types. Overexpression of catalytically inactive mutants or RNA interference (RNAi)-mediated knockdown of MST1 inhibits the apoptotic morphology that is induced by prodeath stimuli, such as staurosporine or etoposide treatment (Ura et al., 2001; Wong et al., 2008). As with yeast Ste20, which activates the yeast MAPK cascade by phosphorylating yeast MAPK kinase kinase Ste11, MST1 and MST2 (but not MST3) are known to activate the JNK and p38 MAPK pathways (Graves et al., 1998; Song and Lee, 2008). Recently, it has been demonstrated that MST1-induced apoptosis can be suppressed by expression of a dominant-negative mutant version of JNK or by knockdown of either mitogen-activate protein kinase kinase 4 (MKK4) or MKK7, the two direct activators of JNK. This suggests that JNK activation may be required for the apoptotic effects of MST1 (Ura et al. 2001; 2007). Interestingly, several lines of evidence suggest that chromosome condensation and DNA fragmentation, the most prominent morphological manifestations of early and late apoptosis, respectively, are, at least in part, mediated by MST1 and JNK activation (Ura et al., 2007). In response to DNA damage, JNK phosphorylates histone H2AX at serine 139 (S139), thus promoting DNA fragmentation by caspase-activated DNase (CAD) (Lu et al., 2006). Other groups have also shown that chromosome condensation during early apoptosis is triggered by phosphorylation of histone H2B at serine 14 (S14) by caspase-activated MST1 (Cheung et al., 2003; Wong et al., 2008) (Figure 3). A recent study has demonstrated that phosphorylation of histone H2B by MST1 sequesters RCC1, a guanine nucleotide exchange factor for Ran, onto the chromosomes, resulting in a significant decrease in levels of nuclear RanGTP and consequently impairing nuclear import of proteins. Impaired import of nuclear factor-kappa B (NFκB), for example, would be expected to accelerate apoptosis (Wong et al., 2008).
The FOXO transcription factors induce expression of proapoptotic proteins such as Fas ligand and the BH3-only protein Bim. Akt phosphorylation of FOXO proteins promotes their binding to 14-3-3 proteins, thereby sequestering the transcription factors in the cytoplasm (for review, see Huang and Tindall, 2007). Importantly, this effect appears to be counteracted by MST kinases. Upon oxidative stress in primary neurons, MST1 phosphorylates FOXO3 at serine 207 (S207), leading to dissociation of the 14-3-3 protein and allowing FOXO3 to translocate to the nucleus (Lehtinen et al., 2006). Interestingly, recent evidence shows that MST and Akt also directly interact and suppress each other. Akt1 directly phosphorylates MST1 at threonine 387 (T387), preventing the caspase-mediated activation of MST1 (Jang et al., 2007). On the other hand, activation by caspase-mediated cleavage allows MST1 to impede Akt1 kinase activity (Cinar et al., 2007), suggesting that proapoptotic MST1 and prosurvival Akt1 create a negative feedback loop (Figure 3).
Hematopoietic progenitor kinase 1 (HPK1) is a hematopoietic-specific Ste20 homolog that participates in the proliferation and differentiation of hematopoietic cells. After growth factor stimulation or T cell or B cell receptor engagement, HPK1 turns on the JNK pathway through activation of MAPKKKs such as MEKK1 (Chen et al., 1999; Schulze-Luehrmann et al., 2002). It has been shown that, independently of JNK activation, HPK1 also stimulates NF-κB signaling through phosphorylation and activation of IκB kinases (IKKs) (Arnold et al., 2001). HPK1 is composed of an N-terminal kinase domain and a C-terminal regulatory domain connected via a proline-rich hinge region. As with MST kinases, cleavage by caspase-3 at the hinge region (DDVD385↓I) separates these domains of HPK1, resulting in kinase activation (Chen et al., 1999; Arnold et al., 2001) (Figure 2). Importantly, the cleaved catalytic fragment is still capable of activating the JNK pathway, but not NF-κB signaling (Chen et al., 1999; Arnold et al., 2001; Schulze-Luehrmann et al., 2002). Recent studies have demonstrated that caspase-mediated cleavage of HPK1 plays a role in the elimination of autoreactive T- and B-lymphocytes—a process termed activation-induced cell death (AICD)—by suppressing the NF-κB pathway; AICD-resistant cells retain intact HPK1 and are activated for NF-κB signaling (Brenner et al., 2005, 2007). Interestingly, activation of T and B cells causes an increase in caspase-3 activity that may result in HPK1 cleavage but not apoptosis (Brenner et al., 2007). Cleavage of HPK1 by an increase in nonapoptotic caspase-3 activity has also been reported during monocytic differentiation. In contrast to lymphoid cells, the resulting N-terminal cleaved product supports myeloid cell survival and differentiation through sustained activation of the JNK pathway (Arnold et al., 2007). These examples represent intriguing instances of kinase signaling that are induced by nonapoptotic caspase action. It will be of great interest to determine how basal caspase-3 activity increases within certain limits to avoid apoptotic induction and to identify the means by which HPK1 cleavage can induce different phenotypic outcomes in various cell types.
As mentioned above, apoptosis is frequently associated with JNK activation. However, the mechanism of JNK activation in response to apoptotic stimuli remains controversial. It appears that under certain circumstances, JNK is activated as a consequence of caspase activation, whereas in other instances, JNK is stimulated independently of caspase activity. In either case, JNK activation may amplify proapoptotic signaling by phosphorylating transcription factors (for example, c-Jun) and other signaling factors (for example, histone H2AX, 14-3-3, Bcl-2 family proteins) (for review, see Dhanasekaran and Reddy, 2008). PAK2, MST kinases, and HPK1 can activate the JNK pathway and, most importantly, can be also activated by caspases (this at least partially accounts for caspase-mediated JNK activation discussed above). MEKK1, one of the MAPKKKs, is also a direct substrate of caspase-3. MEKK1 cleavage by caspase-3 appears to be activated during anoikis, Fas stimulation, and DNA damage (Cardone et al., 1997; Deak et al., 1998; Widmann et al., 1998). MEKK1 is composed of a large N-terminal regulatory domain and a C-terminal kinase domain. Caspase-mediated cleavage (DTVD874↓G in mouse) liberates a constitutively active MEKK1 catalytic domain from a detergent-insoluble cellular compartment into the soluble cytoplasm (Deak et al., 1998). Overexpression of wild-type MEKK1 or the cleaved catalytic fragment promotes apoptotic DNA fragmentation, whereas a cleavage-deficient MEKK1 mutant fails to do so (Widmann et al., 1998). Moreover, overexpression of cleavage-resistant MEKK1 protects cells from anoikis-induced apoptosis, suggesting that caspase-mediated activation of MEKK1 promotes apoptosis (Cardone et al., 1997). Of note, although initial studies of caspase-mediated cleavage of MEKK1 were conducted exclusively with mouse MEKK1 protein, the putative cleavage site in MEKK1 is also conserved in the human protein (DTLD878↓G), and studies have shown that caspase-mediated activation of MEKK1 indeed occurs in human cells (Deak et al., 1998; Widmann et al., 1998). MEKK1 is also known to activate Erk and NF-κB, and it remains to be determined how caspase-cleaved and translocated MEKK1 compares to its full-length counterpart in phosphorylating downstream targets.
Receptor tyrosine kinases are a family of cell surface receptors composed of an extracellular ligand-binding domain, a transmembrane domain, and an intracellular tyrosine kinase domain. Upon engagement of their cognate ligands, the receptors dimerize and transmit an array of prosurvival signals through direct phosphorylation of cytoplasmic targets and through docking of various signaling proteins. To date, several receptor tyrosine kinases, including the rearranged during transfection (RET) receptor, MET kinase, anaplastic lymphoma kinase (ALK), epidermal growth factor receptor (EGFR), v-erb-b2 erythro-blastic leukemia viral oncogene homolog 2 (ErbB2; also known as Her2/neu), and potentially the platelet-derived growth factor receptor (PDGFR) have been reported as substrates of multiple caspases (see Table S1). Caspase-mediated cleavage of the receptor tyrosine kinases proteolytically removes the cytoplasmic region, thereby abrogating ligand-induced transmission of extracellular stimuli to downstream signaling molecules. However, the resulting tyrosine kinase cleavage product itself often has proapoptotic activity once liberated into the cytoplasm from the cell surface. Conversely, overexpression of cleavage-deficient mutant receptor tyrosine kinases can delay the onset of apoptosis.
The MET tyrosine kinase is a scatter factor/hepatocyte growth factor (SF/HGF) receptor that promotes cell motility, invasion, and angiogenesis (Bottaro et al., 1991). The caspase-cleaved (ESVD1000↓Y and DNID1374↓G in mouse) MET fragment contains its complete kinase domain, similar to the caspase-activated serine/threonine kinases described above (Foveau et al., 2007) (Figure 2). Although kinase activity is required for apoptotic induction by the MET fragment, it remains to be determined how the liberated kinase domain induces amplification of apoptotic signaling cascades (Tulasne et al., 2004). The extracellular domain of the MET kinase, anchored to the plasma membrane, is capable of binding HGF even after caspase-mediated removal of the cytoplasmic region, suggesting that it could potentially act as a “decoy receptor” (Deheuninck et al., 2008). In contrast to the MET kinase, the caspase cleavage products reported for other receptor tyrosine kinases do not contain intact kinase domains, suggesting that the proapoptotic effects of these fragments may be independent of tyrosine kinase activity. Rather, the cleavage products may recruit or sequester other apoptotic signaling proteins. ErbB2 may be one example. ErbB2 is cleaved by multiple caspases and its overexpression is associated with aggressive breast cancer. Recently, it was reported that ErbB2 contains a BH3-like sequence (1120LPSETD1125) at its cytoplasmic tail and that caspase cleavage products spanning this region translocate to the mitochondria to neutralize Bcl-XL and to promote mitochondrial cytochrome c release (Strohecker et al., 2008) (Figure 2). Mutation of the conserved leucine 1120 and/or aspartic acid 1125 to glutamic acid or the overexpression of caspase cleavage-deficient mutant proteins markedly attenuates apoptotic cell death in comparison to cells overexpressing wild-type ErbB2. This further supports a proapoptotic role for the ErbB2 fragment (Strohecker et al., 2008). Of note, this BH3-like sequence is also conserved in ErbB4, another member of the EGFR family that shares 58% overall similarity with ErbB2 (Naresh et al., 2006). It will be interesting to determine whether the BH3-like sequence and the caspase-mediated releasing mechanism are also shared with other growth factor receptors, especially PDGFR and EGFR, as these are also known caspase substrates (Table S1).
In addition to receptor tyrosine kinases, nonreceptor tyrosine kinases Etk, c-Abl, Fyn, and Lyn are subject to caspase-mediated cleavage in cells undergoing apoptosis (see Table S1). c-Abl is cleaved by caspases at multiple locations, leaving the kinase domain intact (Figure 2). Interestingly, a recent study demonstrated that cleavage of c-Abl at SLVD958↓A triggers nuclear translocation of a fragment containing the catalytic domain (Barilà et al., 2003) (Figure 3). It has been reported that nuclear localization of full-length c-Abl is critical for DNA damage-induced apoptosis, whereas its cytoplasmic retention can support cell survival, as in the case of the oncogenic Bcr-Abl kinase, a fusion protein of Bcr and c-Abl found in chronic myelogenous leukemia (Yoshida et al., 2005). In this regard, retinoblastoma protein (Rb), an inhibitor of nuclear c-Abl, is also a target of caspases (Jänicke et al., 1996), suggesting that caspase-mediated cleavage may robustly amplify the proapoptotic c-Abl pathway in the nucleus.
For the kinases described above, cleavage-induced activation enhances their (sometimes latent) proapoptotic activity, helping to efficiently execute the cell death program. Caspase-mediated cleavage may also functionally inactivate kinases, thereby terminating survival signals, as is true for the focal adhesion kinase (FAK) and Akt kinases.
FAK, a widely expressed nonreceptor tyrosine kinase that regulates integrin-mediated cell contact with the extracellular matrix through the assembly of focal adhesions, was one of the first kinases identified as a caspase substrate (Crouch et al., 1996; Wen et al., 1997). Inactivation of FAK by caspases (VSWD704↓S, DQTD772↓S) during the early stages of apoptosis results in the termination of survival signals from the extracellular matrix, thereby favoring apoptotic progression (Gervais et al., 1998b; Levkau et al., 1998a). Importantly, a caspase-generated C-terminal fragment of FAK consisting exclusively of the focal adhesion targeting (FAT) domain shares a high degree of homology with an endogenous FAK inhibitory protein FRNK (FAK-related nonkinase), suggesting that the C-terminal FAK fragment is capable of inhibiting FAK activity in a manner similar to FRNK (Gervais et al., 1998b) (Figure 2).
Similarly, the prosurvival kinase Akt is functionally inactivated by caspase cleavage in response to a variety of apoptotic stimuli, including anoikis (Bachelder et al., 2001) and growth factor withdrawal (Xu et al., 2002). Akt cleavage by caspase-3 at multiple sites (TVAD108↓G, EEMD119↓F, ECVD462↓S; also see Table S1 for sites identified only in vitro) leads to inactivation of the kinase by removing the membrane-targeting pleckstrin homology (PH) domain and/or the C-terminal hydrophobic motif, regions that are both critical for full activation of the kinase (Bachelder et al., 2001; Xu et al., 2002) (Figure 2). Overexpression of a Akt mutant protein unable to be cleaved by caspases (but not overexpression of the wild-type protein) markedly attenuates or delays apoptotic cell death, supporting a role for caspase-mediated inactivation of these prosurvival proteins (Bachelder et al., 2001; Xu et al., 2002).
Caspases also impede survival signaling by inactivating the NF-κB pathway. During apoptosis, several key components of the NF-κB pathway, including NF-κB itself, are targets of caspases whose action can thereby terminate the transmission of survival signals emanating from NF-κB (Levkau et al., 1999; Frelin et al., 2008). In addition, caspase-mediated cleavage activates the NF-κB inhibitor, IκBα, further extinguishing NF-κB signaling (Reuther and Baldwin, 1999). Activation of NF-κB signaling after stimulation by TNFα involves recruitment of the IκB kinase (IKK) complex to the TNF receptor through the action of receptor-interacting protein 1 (RIP1) kinase. Interestingly, RIP1 cleavage by caspase-8 (LQLD324↓C) not only attenuates the NF-κB pathway but also activates a TNF-induced proapoptotic signaling pathway. A C-terminal noncatalytic fragment of RIP1 containing a death domain promotes an interaction between TRADD (TNF receptor 1-associated death domain) and FADD, thereby activating the extrinsic apoptotic program (Lin et al., 1999; Kim et al., 2000) (Figure 2). Thus, RIP1 cleavage serves as a molecular toggle, shutting off survival signaling while simultaneously activating apoptosis.
During apoptosis, the protein Smac/DIABLO is released from the mitochondria to the cytoplasm, where it binds IAP proteins, including XIAP, cIAP1, and cIAP2 (Du et al., 2000). Once released, Smac not only antagonizes the ability of XIAP to inhibit caspases but also causes the autoubiquitination and degradation of cIAP1 and cIAP2 (Yang and Du, 2004; Varfolomeev et al., 2007; Vince et al., 2007). Interestingly, RIP1 kinase activity has been found to be required for TNFα-mediated apoptosis when induced by a Smac mimetic, a small molecule that is designed to mimic the action of Smac by binding to IAPs (Wang et al., 2008). As Smac mimetics trigger the degradation of cIAP1 and cIAP2 associated with the TNF receptor signaling complex, receptor stimulation promotes activation of caspase-8 through the RIP1/FADD/caspase-8 complex in a manner dependent on RIP1 kinase activity. Likewise, RIP1 kinase activity is also required for the induction of programmed necrosis occurring when activation of caspase-8 is inhibited after death receptor stimulation (Holler et al., 2000; Degterev et al., 2008). Most interestingly, antigen stimulation in the absence of caspase-8 results in the death of T cells by necrosis, which can be rescued by an inhibitor of RIP1 (Ch'en et al., 2008), suggesting that caspase-8-mediated cleavage of RIP1 inactivates its kinase activity in order to avert the initiation of the necrotic pathway. Recent studies have shown that RIP3, another member of the RIP kinase family, is also required for RIP1-dependent programmed necrosis (Cho et al., 2009; He et al., 2009). Like RIP1, RIP3 is also a substrate of caspase-8 (TEMD328↓G), although the physiological consequence of its cleavage remains to be fully investigated (Feng et al., 2007).
Regulators of cell-cycle progression, including proteins that govern DNA damage and spindle checkpoint pathways, are also subject to caspase-mediated cleavage during apoptosis. In earlier studies, it was shown that the DNA-PKcs undergo caspase-mediated cleavage and inactivation in response to various proapoptotic stimuli (Casciola-Rosen et al., 1995; Song et al., 1996). Several additional kinases/phosphatases have since been identified as caspase substrates (for the full list, see Table S2; for many other nonkinase/phosphatase modulators of the checkpoint/cell cycle that are targets of caspases, such as poly[ADP-ribose] polymerase, see Fischer et al., 2003).
In response to genotoxic or replication stress, checkpoint proteins are activated to halt cell-cycle progression to allow for DNA repair. However, if the damage is severe (or irreparable), apoptotic signaling pathways are engaged. Once this decision to die is made, there is no reason for the cell to expend energy on further DNA repair. Moreover, termination of repair processes may actually facilitate execution of apoptosis. Therefore, it is not surprising that several checkpoint and DNA repair kinases, including the pivotal checkpoint kinase Chk1, have been identified as caspase substrates during apoptosis (see Table S2). Chk1 is activated by the ataxia-telangiectasia mutated and Rad3-related (ATR) kinase after DNA damage or the stalling of replication. A recent study demonstrated that Chk1 cleavage by caspase-7 (SNLD299↓F, TCPD351↓H) generates an active N-terminal kinase domain separated from a C-terminal autoinhibitory domain and an NLS (Matsuura et al., 2008) (Figure 2). Interestingly, overexpression of the cleaved catalytic domain, but not a similar fragment lacking kinase activity, triggered apoptotic nuclear morphology (for example, DNA fragmentation) that is associated with phosphorylation of histone H2AX at S139. As this phosphorylation of H2AX is a hallmark of double-stranded DNA breaks, this raises the possibility that caspase-cleaved Chk1 amplifies apoptotic pathways by further promoting DNA damage (Matsuura et al., 2008). It will be interesting to determine how the dysregulated kinase activity of cleaved Chk1 triggers apoptotic chromatin modifications and whether caspase-activated Chk1 indeed plays a role in a physiological setting.
The serine/threonine kinase Cdk2 drives the transition from G1 to S phase by interacting with Cyclin E and Cyclin A. It has long been noted that apoptosis can be associated with an increase in cyclin-dependent kinase 2 (Cdk2) activity. In Xenopus embryos, caspase cleavage of Cyclin A2 was shown to generate a proteasome-resistant and kinase-active form of Cdk2/Cyclin A2 that participates in the apoptotic response to ionizing radiation (Finkielstein et al., 2002). Earlier studies demonstrated that the Cdk inhibitors p21Cip1/Waf1 and p27Kip1 undergo caspase-mediated cleavage and inactivation during apoptosis, consequently triggering activation of Cdk2 (Gervais et al., 1998a; Levkau et al., 1998b; Eymin et al., 1999). Another study reported that Wee1, which can promote inhibitory phosphorylation of Cdk2 (and Cdc2), is also subjected to caspase-dependent inactivation (Zhou et al., 1998). Furthermore, Cdc25A phosphatase, an activator of Cdk2, is proteolytically activated by caspase-3 during apoptosis (Mazars et al., 2008). Caspase-3 cleaves Cdc25A (DLLD223↓G) into an N-terminal regulatory domain and a C-terminal catalytic domain, resulting in enhanced phosphatase activity of the cleaved fragment (Figure 2). Moreover, because of the presence of a C-terminal NLS (and removal of an N-terminal NES) and the loss of targeting motifs for E3 ubiquitin ligase binding, the active catalytic fragment becomes highly stable and is retained in the nucleus to induce activation of Cdk2 (Mazars et al., 2008). These studies, which suggest that caspases can trigger the activation of Cdk2 through proteolytic activation of its activators and proteolytic inactivation of its inhibitors, support a model in which caspase cleavage of cell-cycle regulators may divert them from their duties in cell cycle control to promote apoptosis instead (Figure 4). In this regard, it is interesting that the Cdk2 activator Cyclin E, which is critical for driving the G1/S transition, is also a target of caspase-mediated cleavage in cultured hematopoietic tumor cells during DNA damage-induced apoptosis (Mazumder et al., 2002). Indeed, a caspase-cleaved fragment of Cyclin E (Cyclin E276–395), no longer capable of interacting with Cdk2, has been shown to bind to and sequester the Bax-binding protein Ku70, allowing the release of Bax from the inactive BaxKu70 complex and facilitating its activation (Mazumder et al., 2007).
The crosstalk between kinases and caspases extends beyond their abilities to act directly on one another. Kinases targeted to potential caspase substrates may also significantly alter the susceptibility of substrates to cleavage. For example, MST1 cleavage by caspases is inhibited by Akt phosphorylation at T387 (Jang et al., 2007). In addition, it has been shown that both CK1 and CK2 phosphorylate the BH3-protein Bid at threonine 58 (T58), serine 61 (S61), and serine 64 (S64) in the vicinity of a cleavage site (LQTD59↓G in the mouse Bid), protecting the protein from cleavage by caspase-8 (Desagher et al., 2001). It is possible that phosphorylation could also protect Bid from cleavage by caspase-2, though this has not yet been reported. Because Bid cleavage leads to activation of the intrinsic pathway of apoptosis, kinases that inhibit this proteolytic cleavage event might well be expected to raise the threshold for apoptosis, making it more difficult for apoptotic stimuli to trigger cell death (Figure 1). Similarly, it has been reported that phosphorylation of the tumor suppressor PTEN (phosphatase and tensin homolog deleted from chromosome 10) by CK2 appears to protect the phosphatase from the proteolytic activity of caspase-3 (Torres et al., 2003). Phospholipase C-γ1 (PLCγ1) is also targeted by caspase-3 and -7, but the cleavage is prevented by phosphorylation of tyrosine 771 (Y771) adjacent to the cleavage site (AEPD770↓Y) (Bae et al., 2000). IκBα is protected from caspase cleavage (DRHD31↓S) upon phosphorylation at serine 32 (S32) and serine 36 (S36) by the IKK complex (Barkett et al., 1997). Despite these examples, the prevalence of a mechanism for caspase substrate preservation by phosphorylation remains to be determined.
Phosphorylation can impede caspase-mediated substrate cleavage, but it is also possible that phosphorylation can enhance caspase-mediated cleavage of substrates to either inhibit or promote death, depending on the nature of the substrate. For example, cleavage of PKCδ by caspase-3 (DILD327↓N in mouse) is positively regulated by the Src kinase through phosphorylation at tyrosine 332 (Y332) (Lu et al., 2007). Likewise, it has been suggested that mixed lineage kinase-3 (MLK3) phosphorylation of the Golgi structural protein golgin-160 can enhance caspase cleavage at a site near the phosphorylation site (Cha et al., 2004). Phosphorylation of the Bcl-2 family member BimEL by JNK can also promote its cleavage by caspase-3 (ECD13↓R), though the cleavage may be secondary to the phosphorylation-dependent release of BimEL from microtubule sequestration (Chen and Zhou, 2004; Corazza et al., 2006). As databases of kinase substrates/phosphorylation sites and caspase substrates/cleavage sites continue to grow, it is likely that mining these databases for common targets will reveal even greater crosstalk between kinases and caspases in the control of proteolytic cleavage of substrates.
The actions of kinases, either pro- or antiapoptotic, are counterbalanced by phosphatases that can remove the phosphorylation from caspases or their substrates. Thus, it is perhaps not surprising that phosphatases too can be targets of caspase cleavage, though less is known about this mechanism of regulation (see Table S3). Protein phosphatase 2A (PP2A), a ubiquitously expressed serine/threonine phosphatase that is involved in numerous cellular signaling pathways, is one example of a phosphatase caspase substrate. The PP2A holoenzyme consists of three subunits: the A and B regulatory scaffold subunits and the C catalytic subunit. Subcellular localization and substrate specificity of the PP2A holoenzyme are determined in part by a large number of B subunits. Yeast two-hybrid screening of a human lymphocyte cDNA library revealed that the regulatory subunit Aα of PP2A is a substrate of caspase-3 (Santoro et al., 1998). Cleavage of the Aα subunit leads to enhancement of PP2A activity in vitro and in vivo, suggesting that PP2A may be proteolytically activated by caspase-3 during apoptosis (Santoro et al., 1998). Interestingly, the SET protein, a PP2A inhibitor, has also been recognized as a caspase substrate (Morita et al., 2000), raising the possibility that elevated PP2A activity during apoptosis may also be mediated by inactivation of its inhibitors (such as SET) by caspases. Similarly, PP1 inhibitor-3 (I-3), which specifically inhibits the phosphatase activity of PP1α and PP1γ1, is inactivated by caspase-3 during apoptosis, resulting in an increase in PP1α and PP1γ1 activity (Huang and Lee, 2008). In addition to controlling diverse signaling pathways, PP2A and PP1α may also remove inhibitory phosphorylations on caspase-3 and -9, respectively, and thus further contribute to apoptotic progression (Alvarado-Kristensson and Andersson, 2005; Dessauge et al., 2006).
Cross-regulation of kinases/phosphatases and caspases allows for fine tuning of the apoptotic threshold, as well as the opportunity to amplify apoptotic signals through the production of “strategic” cleavage products. In healthy cells, caspases are turned off by phosphorylation-mediated suppression in addition to other mechanisms. However, once cells receive a substantial stimulus that overcomes this suppression, caspases are turned on to cleave kinases and phosphatases, among other substrates. This not only helps to turn off survival signaling pathways but, in multiple cases, generates a number of proapoptotic protein fragments. Prosurvival kinases and phosphatases can be converted to fragments by proapoptotic activity, and some proapoptotic kinase/phosphatases are also activated by caspases. Consequently, once the balance shifts toward cell death, the caspase-kinase/phosphatase network amplifies apoptotic signals to efficiently dismantle the apoptotic cell. The potential crosstalk between caspases and kinases/phosphatases becomes even more elaborate when one considers the possibility that the kinases phosphorylating the caspases (for example, p38 MAPK) could potentially be activated by caspase-mediated cleavage of upstream components in their own regulatory pathways (for example, by the MAPKKKs or MAPKKKKs).
When caspase activation is blocked despite the presence of death stimuli, nonapoptotic cell death (for example, autophagy and necrosis) is often observed (Kroemer et al., 2008). Several lines of evidence suggest that caspase-mediated activation of kinases and phosphatases may also counteract autophagy and necrosis, directing a cell death signal toward the execution of apoptosis instead. For example, a recent study demonstrated that increased stability of the Cdk inhibitor p27Kip1 in quiescent cells promotes autophagy, whereas knockdown of p27Kip1 protein levels is sufficient to cause apoptosis in these cells (Liang et al., 2007). This suggests that if there is enough caspase activity to cleave p27Kip1 in the cells under nutrient-limiting conditions, the cells will choose apoptosis. Likewise, the RIP1 kinase has been put forth as a key protein in the induction of programmed necrosis. As described above, death receptor-induced necrosis requires RIP1 kinase activity, whereas kinase activity appears to be dispensable for activation of the NF-κB or apoptotic pathways (with the exception of Smac mimetic-induced apoptosis). Upon death receptor stimulation, cleavage of RIP1 by caspase-8 inactivates its kinase activity, forestalling programmed necrosis. The resulting noncatalytic C-terminal fragment of RIP1 favors the apoptotic pathway over the NF-κB survival pathway. Therefore, these findings place RIP1 at the molecular intersection of survival, apoptotic, and necrotic signaling pathways. Caspase-mediated activation of kinases and phosphatases also explains at least some of the morphological differences between apoptotic cell death and nonapoptotic cell death. Membrane blebbing and chromatin condensation, which can be triggered simply by caspase-mediated cleavage of substrates such as ROCK1 and MST1, respectively, are not observed in cells undergoing autophagy or necrosis.
Although a large number of caspase substrates, including the kinases and their regulators discussed here, have been identified, the temporal sequence of these cleavage events has not been finely resolved. Although some of the important signaling kinases described here are cleaved early in the apoptotic process before cell death is even morphologically evident (for example, HPK1, MEKK1, ErbB2, c-Abl, p21Cip1/Waf1, p27Kip1), others that are cleaved at later stages of apoptosis may merely represent final end-products of the long proteolytic cleavage cascade, with negligible effects on the progression of the apoptotic process (even if the putative cleavage fragments appear to promote apoptosis upon cellular overexpression). Conversely, it may be that there are functional reasons for late-stage proteolytic cleavage in the apoptotic process that we do not yet fully understand. In addition, with a few exceptions, the immediate substrates of caspase-activated kinase fragments still remain to be identified. Emerging proteomic technologies that enable identification and confirmation of caspase substrates (and their cleavage sites) (Dix et al., 2008; Mahrus et al., 2008), together with gene targeting and biochemical analyses, should help to unravel the molecular mechanisms underlying the coordination of complex cellular life and death decisions by the caspase-kinase/phosphatase networks.
The authors thank members of the Kornbluth lab for stimulating discussions and critical reading of the manuscript. This work is supported by NIH RO1s CA102707GM61919 (to S.K.) and by the Irvington Institute Fellowship of the Cancer Research Institute (to M.K.).
Supplemental Data include three tables and can be found with this article online at http://www.cell.com/supplemental/S0092-8674(09)01038-1.