Main

Since the discovery that the major effectors of programmed cell death are members of the caspase family of proteases, significant efforts have been directed towards understanding the multifarious roles that members of this evolutionary-conserved family of proteases play in apoptosis. Although the mechanisms of activation of the worm, fly and mammalian caspases are now relatively well understood,1, 2 much less is known about how these proteases contrive to kill cells. Although hundreds of mammalian caspase substrates have now been identified – and therein lies the root of the problem – we still know surprisingly little about how caspases coordinate the terminal events in apoptosis that result in death of the cell.

Discriminating Innocent Bystanders from Legitimate Targets

For this overview, we have searched the literature for caspase substrates and have compiled a somewhat daunting list of 390 proteins that have been reported to be cleaved by caspases, either in vitro, during apoptosis or inflammation (see Supplementary Table 1). Owing to this profusion of caspase substrates, the major challenge now is to identify the subset of these proteins that have functional significance for the process of apoptosis and/or inflammation. Although it is within the bounds of possibility that all proteins that are cleaved by caspases during apoptosis contribute to the demise of the cell on some level, it seems highly unlikely that all caspase substrates play equally significant roles in this process. On the contrary, it seems rather more likely that many proteins are simply caught up in the proteolytic maelstrom that envelops a cell during apoptosis and become cleaved by caspases– by accident rather than design – during the terminal phase of cell death. Thus, the nub of the problem is to discriminate what we have previously termed the ‘innocent bystanders’ from the ‘legitimate targets’ of caspase-mediated attack.3

Unfortunately, there is no easy way to identify functionally important caspase substrates from the proteolytic noise that almost inevitably accompanies a process where several proteases become activated within a short time frame.4, 5, 6 Painstaking analysis of individual caspase substrates – involving the mapping of precise caspase cleavage sites, the creation and expression of non-cleavable mutants and/or cleaved forms of the substrate, the generation of gene knockouts or non-cleavable knock-ins – is required to generate a definitive link between proteolysis of an individual substrate and a specific ‘apoptotic parameter’. For the vast majority of substrates that have been described to date (see Supplementary Table 1), such analyses have not been carried out. Rather, the consequences of proteolysis of individual substrates has either been inferred or in some cases, the cleaved form of the substrate has been overexpressed and the functional effects of this have been evaluated in otherwise healthy cells. In many cases, we also await confirmation that putative caspase substrates are actually cleaved by caspases at physiological concentrations of enzyme. For example, the recently identified cytokine, IL-33, has been reported to be a substrate for caspase-1.7 However, the latter proposal is based solely upon the observation that this cytokine is cleaved in vitro by recombinant caspase-1 at arbitrary concentrations of this enzyme.7

One way of approaching the question of whether a caspase substrate is potentially more relevant than others is to ask whether this protein is also cleaved by caspases in other species. However, this approach has not been adopted in many cases and has led to generalizations concerning the significance of some substrates for apoptosis that are unlikely to be borne out upon more detailed analysis. Although it is certainly possible that some caspase substrates may be cleaved in a species-specific manner, it seems implausible that the same functional endpoints are achieved through proteolysis of a completely different array of proteins in different species. This view is buttressed by observations that many of the same morphological endpoints are seen in apoptotic cells from organisms as divergent as flies and man.8 Thus, it is probable that a conserved cohort of caspase substrates are cleaved by worm, fly and mammalian caspases, and that these represent the legitimate targets that ensure controlled cell destruction, as well as the morphological and functional hallmarks of apoptosis.

Exceptions to the above are likely to be caspase substrates that play roles in signal transduction cascades that operate only in certain organisms. For example, nematodes do not appear to possess death receptors; therefore, signalling intermediates in death receptor pathways, such as BID and RiP-1, which are important caspase substrates in mammals are not present in this organism. Similarly, it is also possible that tissue-specific proteins may be targeted by caspases to regulate some unique aspect of apoptosis in such tissues.

Caspases are Required for the Apoptotic Phenotype

There has been an ongoing debate as to whether caspases are essential for apoptosis of mammalian cells or whether other proteases can step into the breach to achieve the same end points when caspase activity is stymied.9, 10, 11, 12 Some investigators hold the view that essentially similar cell death endpoints can be achieved regardless of whether caspase activity is muzzled. Under conditions where caspase activity is experimentally disabled – either through genetic inactivation of specific caspases or caspase-activating proteins such as Apaf-1 or through exposure of cells to poly-caspase inhibitors, such as Z-VAD-fmk – cell death still occurs in response to many pro-apoptotic triggers and this is taken as evidence that caspases are not required for apoptosis. However, in the vast majority (if not all) of these cases, the morphological end points seen when caspases are prevented from participating are not what are commonly regarded as hallmarks of apoptotic cell death.11, 12 Rather, caspase inhibition typically converts the phenotype of the dying cell from apoptosis into necrosis.11, 12

So why does caspase inhibition/inactivation fail to protect cells from stimuli that provoke apoptosis? In mammals, this is owing to events upstream of caspase activation that provoke the release of mitochondrial intermembrane space proteins such as cytochrome c. Although this does not apply to all pro-apoptotic stimuli (engagement of death receptors for example), it applys to stimuli where BH3-only proteins play instrumental roles in promoting mitochondrial permeabilisation and assembly of the Apaf-1/caspase-9 apoptosome downstream.13 In such cases, owing to the rapid decline in mitochondrial function, most cells are irreversibly committed to die irrespective of whether caspases are activated downstream. However, in the absence of caspase activation, the phenotype of the resulting death is necrotic rather than apoptotic. The exception seems to be certain post-mitotic cells such as neurons and cardiomyocytes, as these cells can recover from mitochondrial permeabilisation as long as caspase activation is prevented downstream.14, 15

In the case of the nematode worm, Caenorhabditis elegans, it is clear that the CED-3 caspase is essential for all developmental-related programmed cell deaths in this organism.16, 17 Strong loss-of-function mutations in CED-3 block all programmed cell deaths in the worm, thereby illustrating that a default cell death programme is not present in this organism. Similar observations have also been made with respect to mutations in the CED-3 – activating adaptor molecule, CED-4.16

In the fruitfly, it is also the case that loss-of-function mutations associated with the CED-4 homologue, ARK (dApaf-1/HAC-1)18 or its caspase-binding partner, DRONC, lead to extensive disruption of the normal patterns of cell death that are seen during development.19, 20, 21 Furthermore, despite intensive genetic screening using both worm and fly, no other proteases have emerged as playing influential roles in programmed cell death in these organisms.

In mammals, there is also good genetic evidence to argue that caspases and their activators play central roles in apoptosis. To date, knockouts have been generated for the apoptosis-associated caspases, caspase-2, caspase-3, caspase-6, caspase-7, caspase-8 and caspase-9, as well as the caspase-9 adaptor protein Apaf-1.22, 23, 24 In several of these cases, programmed cell death is severely disrupted and embryonic lethality is observed. However, as discussed above, mutation or inactivation of caspases rarely confers clonogenic rescue upon a cell owing to the fact that one of the most commonly utilised routes to apoptosis in mammals, the mitochondrial pathway, involves permeabilization of mitochondria before caspase activation.13 Because of this, blocking caspase activation downstream of mitochondrial outer membrane permeabilization is typically futile as cells die as a result of mitochondrial dysfunction. However, death without caspase participation rarely, if ever, manifests the normal spectrum of changes that define apoptosis.25 Nonetheless, some investigators have labeled this type of cell death ‘atypical apoptosis’ or ‘caspase-independent apoptosis’ or ‘aponecrosis’ which can give rise to misunderstandings concerning whether caspases are really required for apoptosis.

Global Perspectives on Caspase-Mediated Substrate Proteolysis

Given the embarassment of riches presented by approximately 400 mammalian caspase substrates that have now been described (see Supplemental Table 1), it is unfortunately not yet possible to paint a clear picture of how caspases achieve the sterotypical changes that are seen during apoptosis. Indeed because most of the known caspase substrates have been identified serendipituously, it is very likely that hundreds of additional caspase substrates are likely to emerge upon systematic proteomics-based analyses of the proteomes of apoptotic cells.

Although we have tabulated essentially all of the caspase substrates that have been reported in the literature to date, it is important to point out that many of these substrates are simply reported to be cleaved during apoptosis and data pertaining to their functional significance have yet to emerge. Moreover, in most cases the caspase cleavage sites within these proteins have not been mapped, nor has it been tested whether these proteins are also targeted for proteolysis by caspases in other mammalian species, such as the mouse. Therefore, the precise role of the majority of caspase substrates in apoptosis remains speculative, at best, and many are almost certainly insignificant for the completion of the process. However, some caspase substrates have been definitively linked to specific cell death end points as we will discuss below.

The CASBAH: The CAspase Substrate DataBAse Homepage

Because of the expanding list of caspase substrates, we have created an online database containing all of the reported mammalian caspase substrates (http://www.casbah.ie). This database is fully searchable and contains linkouts to the UniProt (the Universal Protein Resource) database, which contains extensive additional information on each protein. In addition, where the caspase cleavage site(s) with a substrate have been mapped, this information is included within the CASBAH entry, along with relevant references to the primary papers that first reported a particular caspase substrate. We hope that this will be a useful resource for investigators working in this area.

Relative contributions of Caspases to Cell Demolition

The majority of studies indicate that caspases are certainly not equal in terms of their ability to cleave cellular substrates. The apoptosis-associated initiator caspases (caspase -8, -9 and -10 and possibly also caspase-2) appear to be highly specific proteases that cleave few substrates apart from their own precursors and other caspases downstream. Apart from other caspases, the only well-established substrates for the initiator caspases, which have been identified to date, is the BH3-only protein BID and RIP kinase. BID is cleaved by caspases -8 and -10 upon stimulation of death receptors and plays an important role in the propagation of pro-apoptotic signals through promoting mitochondrial outer membrane permeabilization and cytochrome c release.26, 27 RIP is recruited to the tumour necrosis factor and Fas receptor complexes and controls nuclear factor-kappa B (NFκB) activation via recruitment of the regulatory subunit of the IκB kinases complex.28, 29 However, RIP can be inactivated through proteolysis by caspase-8 and this facilitates propagation of the death signal as opposed to NFκB-mediated survival signals.28, 29

There is little evidence that the initiator caspases can contribute to the generalised proteolysis seen during apoptosis. However, it is useful to point out that high (i.e. nonphysiological) concentrations of recombinant caspases will cleave practically any caspase substrate in vitro. Clearly, this does not offer proof that this protease is responsible for this event in vivo, although it is often implied that this is the case. So, which proteases are responsible for most of the substrate proteolysis seen during the demolition phase of apoptosis? Evidence from caspase-deficient cells and cell lines, as well as immunodepleted cell-free extracts, point towards caspase-3 as the major effector caspase.25, 30, 31 Loss of caspase-3 abolishes or dramatically delays the kinetics of the majority of substrate proteolysis seen during apoptosis. Thus, the vast majority of substrates listed in Supplementary Table 1 are most likely preferentially cleaved by caspase-3 during apoptosis. Caspase-7 possesses similar specificity to caspase-3 towards synthetic tetrapeptide substrates and these enzymes are sometimes regarded as functionally redundant.32, 33 However, activity towards synthetic tetrapeptide-based substrates does not give information concerning interactions with full-length substrate proteins, particularly at surfaces outside of the catalytic pocket of the enzyme. Therefore, it is certainly plausible that caspases -3 and -7 exhibit distinct activities towards full length substrate proteins. Indeed, gene knockout and immunodepletion experiments argue that caspase-7 cannot readily substitute for the absence of caspase-3, although this may also relate to the relative expression levels of either caspase within cells.4, 31, 34, 35, 36, 37

Few cellular substrates for caspase-6 have been identified thus far. Although several substrates for caspase-6 have been reported, the majority of these claims rest upon data generated in vitro using high concentrations of recombinant caspases. The most well-established substrate for caspase-6 appears to be the nuclear lamins A and C,31, 38, 39 whereas lamin B is cleaved by caspase-3.31 Caspase-6 is also responsible for the proteolytic maturation of caspases -2 and -8 within the caspase cascade that is initiated downstream of assembly of the Apaf-1 apoptosome.4, 40, 41 However, loss of caspase-6 appears to have no overt effect on developmental apoptosis.22, 36 Thus, the contribution of caspase-6 to the demolition phase of apoptosis remains enigmatic.

How do Caspases Coordinate Apoptosis?

Of the hundreds of caspase substrates that have been reported to date, a small subset of these have been convincingly linked with specific alterations to the cellular architecture that are generally accepted to be hallmarks of apoptosis. The list of these ‘intended victims’ is surprisingly short at present.

It is worth reminding ourselves that apoptosis is a morphologically defined mode of cell death that is associated with several distinct features.42, 43 Of course, there are likely to be other modes of programmed cell death that display different characteristics, but these modes of cell death are, almost by definition, not apoptotic in nature. On the basis of the original studies of Kerr et al.,42 apoptotic cells typically display pronounced compaction of chromatin and, in the majority of cases, the nucleus undergoes progressive fragmentation and dispersal throughout the cell body. The latter events are highly distinctive and can be completely blocked by inhibition of caspase activity.44, 45 Studies on the serine/threonine kinase rho-associated kinase I (ROCK I) have strongly implicated this kinase in apoptosis-associated nuclear fragmentation (Figure 1).46, 47 Caspase-mediated proteolysis of ROCK I leads to loss of the C-terminal autoinhibitory region of this molecule and produces a constitutively active kinase which is capable of contributing to myosin light chain phosphorylation and other events leading to increased actin–myosin contractility (Figure 1). This increased contractility of the actin cytoskeleton, in tandem with caspase-mediated proteolysis of the nuclear lamins that contribute to weakening of the nuclear envelope,48 has been linked with the dismembering of the nucleus into several fragments as typically occurs during apoptosis.46, 47

Figure 1
figure 1

Caspase-dependent proteolysis of ROCK I contributes to plasma membrane blebbing and nuclear fragmentation. Caspase-dependent cleavage of ROCK I removes the regulatory C terminus of the molecule, resulting in a constitutively active kinase. Active ROCK I promotes phosphorylation of myosin light chain kinase which, in turn, regulates myosin-dependent actin rearrangements. Rearrangement of actin microfilaments can contribute to plasma membrane blebbing and, in tandem with caspase-dependent cleavage of nuclear lamins, nuclear fragmentation

Full size image

Cells undergoing apoptosis also retract from neighbouring cells, losing contact with other cells and the substratum and usually undergo extensive plasma membrane blebbing which can also lead to collapse of the cell into numerous small fragments, termed apoptotic bodies.42 However, the production of apoptotic bodies seems to be very dependent on cell type, as some cell types fragment to a greater or lesser degree than others. Once again, neutralisation of caspase activity is a very efficient means of blocking apoptosis-associated cellular retraction, plasma membrane blebbing and apoptotic body production.30, 49 Although it remains unclear precisely how caspases coordinate cellular retraction, this very likely relates to the ability of caspases to cleave several cytoskeletal proteins such as actin, tubulin, fodrin and vimentin. Many other proteins with structural roles within the cell are also cleaved during apoptosis (see Table 1) and it is likely that proteolysis of such proteins contributes to the dramatic morphological rearrangements that take place during this mode of cell death.

Table 1 Caspase substrates: structural proteins
Full size table

ROCK I has also been heavily implicated in the production of plasma membrane blebs during apoptosis, again through reorganisation of the actin–myosin cytoskeleton owing to the caspase-dependent loss of ROCK I-auto-inhibition (Figure 1).46, 47, 49, 50 In agreement with this, depolymerisation of the actin cytoskeleton using fungal toxins efficiently abolishes the production of apoptotic bodies.51

Although not a morphological feature of apoptosis, one of the earliest biochemical signatures of this process to be identified was the internucleosomal scission of DNA into multiples of approximately 200 bp,43 an event that is also under the control of caspases. Caspase inhibitors very efficiently block this extensive DNA hydrolysis, apparently through delaying the caspase-3 -mediated inactivation of inhibitor caspase-activated DNase (ICAD)/DNA fragmentation factor (DFF)45, the inhibitory subunit of the caspase-activated DNase (CAD)/DFF40 nuclease.52, 53 Multiple transcription factors are also cleaved by caspases and a range of translation initiation factors and ribosomal proteins are also affected (see Table 2). Predictably, this results in transcriptional and translation shut down relatively early in the process. Similar to DNA hydrolysis during apoptosis, it is not clear why the transcriptional and translational machineries are targeted by caspases. A plausible explanation is that this may guard against the possibility that these machineries could be used to replicate viruses that might have provoked apoptosis in the first place. Similarly, apoptosis-associated DNA hydrolysis may also safeguard against the persistence of viral DNA. Alternatively, the latter event may simply render chromatin more manageable for subsequent removal by phagocytes.

Table 2 Caspase substrates: regulators of transcription/translation
Full size table

Numerous kinases and other signaling intermediaries have also been reported to be cleaved by caspases during apoptosis (see Table 3). In many cases, this results in the production of constitutively active forms of such kinases but, with the exception of ROCK I as discussed above, it is likely that many of these events are not especially relevant for apoptosis.

Table 3 Caspase substrates: kinases and signalling intermediaries
Full size table

A major goal of programmed cell death is to facilitate the swift clearance of dying cells while avoiding release of cellular contents. A number of alterations to the composition of the membranes of apoptotic cells have been documented,54 perhaps the most well known of which is the externalisation of phosphatidylserine on the outer leaflet of the plasma membrane.55 Although this event is also blocked by inhibition of caspase activity,56 it remains unclear how phosphatidylserine externalisation or other apoptosis-associated membrane events, is regulated by caspases. In addition to becoming licensed for removal by phagocytes, apoptotic cells may also actively attract the attentions of such cells by secreting molecules with chemotactic properties. This may occur through the caspase-dependent release of a chemoattractant lipid, lysophosphatidylcholine (LPC), possibly mediated by cleavage and activation of calcium independent phospholipase A (iPLA) by caspase-3 (Figure 2).57

Figure 2
figure 2

Caspases promote the release of chemotactic factors via iPLA2 Activated effector caspases cleave the calcium independent phospholipase A2 resulting in a constitutively active C-terminal fragment. Active iPLA2 hydrolyses membrane bound phosphatidylcholine to bioactive LPC and arachidonic acid. LPC is released from the cell to the extracellular space and acts as a chemoattractant for phagocytic cells, which engulf and remove the apoptotic cell

Full size image

Concluding Remarks

The knowledge that there are in excess of 400 proteins that become targeted for proteolysis by caspases during the terminal phase of apoptosis suggests that we are still some way from fully understanding how these proteases coordinate this cell death paradigm. It seems rather unlikely that failure to cleave individual substrates is likely to have any impact on the kinetics of cell death in the great majority of cases. Rather, failure to cleave particular substrates may well alter certain aspects of the apoptotic phenotype; a good example of which is the pronounced delay in the kinetics of DNA fragmentation upon disablement of the ICAD/CAD system. However, discovery of especially influential caspase substrates appears to be the exception rather than the rule, with the majority of substrates that have been reported to date playing relatively minor roles in the process. Owing to the sheer number of proteins that are cleaved by caspases during the throes of death, it has thusfar proved difficult to identify a subset of proteins that have particular relevance for the signature events that define apoptosis. Nonetheless, a blueprint for how caspases kill is slowly emerging from the dense thicket of caspase substrates that have been identified to date.