The adaptor protein CRK is a pro-apoptotic transducer of endoplasmic reticulum stress.

Excessive demands on the protein-folding capacity of the endoplasmic reticulum (ER) cause irremediable ER stress and contribute to cell loss in a number of cell degenerative diseases, including type 2 diabetes and neurodegeneration. The signals communicating catastrophic ER damage to the mitochondrial apoptotic machinery remain poorly understood. We used a biochemical approach to purify a cytosolic activity induced by ER stress that causes release of cytochrome c from isolated mitochondria. We discovered that the principal component of the purified pro-apoptotic activity is the proto-oncoprotein CRK (CT10-regulated kinase), an adaptor protein with no known catalytic activity. Crk(-/-) cells are strongly resistant to ER-stress-induced apoptosis. Moreover, CRK is cleaved in response to ER stress to generate an amino-terminal M(r)~14K fragment with greatly enhanced cytotoxic potential. We identified a putative BH3 (BCL2 homology 3) domain within this N-terminal CRK fragment, which sensitizes isolated mitochondria to cytochrome c release and when mutated significantly reduces the apoptotic activity of CRK in vivo. Together these results identify CRK as a pro-apoptotic protein that signals irremediable ER stress to the mitochondrial execution machinery.

Irremediable ER stress represents a form of intrinsic cell damage that culminates in activation of the BAX/BAK-dependent mitochondrial apoptotic pathway[3,4]. Homo-oligomerization of BAX and/or BAK consequently results in outer mitochondrial membrane permeabilization and release of pro-death mitochondrial proteins (e.g., cytochrome c) into the cytosol, causing activation of effector caspases[8-10]. For many forms of cell injury, including ER stress, we have a limited understanding of the cellular transducers that relay the information of upstream damage to BAX/BAK oligomerization at mitochondria. To date, the only known BAX and/or BAK activators are members of the pro-apoptotic BH3-only family, a diverse class of polypeptides containing a loosely conserved ~9-12 amino acid BH3 death domain[11,12,13]. The BH3-only proteins BID, BIM, NOXA and PUMA have been previously implicated in ER stress-induced death[14,15]; however, cells deficient in one or more of these proteins are not completely resistant to this form of apoptosis[16-18]. Therefore, it is highly probable that additional proteins that communicate ER stress to the mitochondrial apoptotic machinery remain to be discovered. To pursue this possibility, we took an unbiased biochemical approach to purify the major ER stress-induced cytosolic pro-apoptotic activity.

Given their strong resistance to ER stress-induced apoptosis [16,17], we reasoned the Bax-/-Bak-/- mouse embryonic fibroblasts (MEFs) present a unique and powerful tool to trap and identify pre-mitochondrial apoptotic signals activated by ER stress. Therefore, we challenged SV40-transformed Bax-/-Bak-/- MEFs with the pharmacological agent Brefeldin A (BFA), which blocks ER-golgi protein transport, to induce irremediable ER stress and initiate the pre-mitochondrial apoptotic program. We prepared cytosolic extracts (S100) from untreated or BFA-treated Bax-/-Bak-/- MEFs, then incubated isolated Jurkat mitochondria with these extracts, and measured the amount of mitochondrial cytochrome c released as a readout for pro-apoptotic activity. The S100 of BFA-treated Bax-/-Bak-/- cells triggers the release of ~90% of total intra-mitochondrial cytochrome c in a BAX/BAK-dependent manner[15], while the S100 fraction from untreated Bax-/-Bak-/- MEFs induces negligible (<5%) cytochrome c release (Fig. 1a). Thus, ER stress induces a cytosolic activity capable of releasing mitochondrial cytochrome c, which we have termed ytochrome eleasing ctivity (CcRA). We previously found that proteolytically active BID (tBID) is responsible for a portion of this CcRA (~30%), but that the majority of CcRA remains intact in BFA-treated Bid-/- S100[15]. Therefore, we designed and performed a biochemical purification strategy to isolate additional CcRA factors in the BFA-treated Bax-/-Bak-/- S100 (Fig. 1b).

Figure 1

Biochemical purification of ER stress apoptotic activity identifies proto-oncogene CT-10-regulated kinase (CRK)

(a) Induction of cytochrome c release from isolated Jurkat mitochondria by cytosolic extracts (S100) from untreated (UNT) and 24h Brefeldin A (BFA) 2.5 μg/ml treated Bax-/-Bak-/- MEFs. (b) FPLC purification scheme of cytochrome c releasing activity (CcRA) present in BFA Bax-/-Bak-/- S100. Active fractions from each purification step are indicated. (c) CcRA assay of the fractions from the final step of the purification (MonoQ ion exchange gradient). (d) Diagram of CRK isoforms, domains, and amino acid sequence.

The CcRA-containing fractions from the final step (MonoQ gradient) (Fig. 1c) of the purification scheme did not contain detectable amounts of any known BH3-only protein by immunoblotting (data not shown), and so were analyzed by MALDI mass spectrometry (Fig. 1c). Proto-oncogene CT10-regulated kinase (CRK) was the highest confidence protein identified by mass spectrometry in the active fractions, with approximately 25% of the total sequence represented in 6 tryptic peptides (Supplementary Table 1). C-crk encodes two splice isoforms, CRKI (28kD) and CRKII (38kD)[7], which have been previously recognized as adaptor components in multi-protein complexes involved in cell morphology, movement, proliferation, and differentiation[19]. CRKI and CRKII share a common Src homology 2 (SH2) domain and SH3 domain, while CRKII contains an additional C-terminal SH3 domain (Fig. 1d)[19]. As the peptides detected by mass spectrometry are common to both CRKI and CRKII, this information did not differentiate which isoform is present in the analyzed fractions (Supplementary Table 1).

To test if CRK plays a role in ER stress-induced apoptosis in cells, we challenged crk-/- and wild-type (WT) matched MEFs with BFA or tunicamycin (TUN), an ER stress agent that specifically inhibits N-linked glycosylation. As previously reported, these crk-/- MEFs are derived from genetically engineered embryos that fail to express either CRKI or CRKII[20] (Fig. 2a). Notably, S100 from BFA-treated crk-/- MEFs has significantly decreased CcRA compared to that from BFA-treated WT MEFs (Fig. 2b), indicating that CRK is required for the majority of the ER stress-induced apoptotic signal. In addition, crk-/- MEFs are strikingly resistant to ER stress-induced apoptosis, but as sensitive as WT MEFs to staurosporine (STS), a pan-kinase inhibitor known to activate the mitochondrial apoptotic pathway independently of ER stress, thereby confirming the pro-apoptotic role of CRK specifically in the ER stress-induced apoptotic pathway, rather than generally from other intrinsic stresses (Fig. 2c, d).

Figure 2

Crk-/- MEFs are significantly resistant to ER stress-induced apoptosis

(a, b) 18h BFA (2.5 μg/ml)-treated crk-/- MEF S100 contains significantly less CcRA in comparison to 18h BFA-treated (2.5 μg/ml) wild-type MEF S100. n=3, error bars = sd. (c) crk-/- MEFs are visually resistant (phase contrast) to ER stress-induced apoptosis (BFA 2.5μg/ml). Scale bar, 100μm. (d) crk-/- MEFs are strongly resistant to BFA and TUN-induced apoptosis, but equally sensitive to staurosporine (STS), in comparison to wild-type MEFs. n=3, error bars = sd.

To investigate which CRK isoform is necessary for ER stress-induced apoptosis, we reconstituted crk-/- MEFs by transient transfection with expression constructs encoding either crkI or crkII and evaluated their sensitivity to ER stress-induced apoptosis. Surprisingly, crkI and crkII are both able to independently rescue the sensitivity of crk-/- MEFs to ER stress-induced apoptosis, arguing that a sequence common to CRKI and CRKII is required for CRK's apoptotic activity (Fig. 3a,b). We next attempted to use retroviral vectors to establish stable cells lines of crk-/- MEFs expressing either crkI or crkII. While we had no difficulty recovering cells reconstituted with crkII, all attempts to establish stable crkI expression in crk-/- MEFs were unsuccessful, indicating that CRKI may be inherently cytotoxic. Stable reconstitution of crkII in crk-/- MEFs restores sensitivity to ER stress agents, but does not affect the response to STS (Fig. 3c-e), recapitulating the transient expression results (Fig. 3a). Moreover, stable overexpression of crkII and transient overexpression of crkI further sensitizes WT MEFs to ER stress (Fig. 3f-i). Together, these data strongly argue that CRK is a critical component of the ER stress-induced apoptotic pathway.

Figure 3

CRKI or CRKII restores sensitivity of crk-/- MEFs to ER stress-induced apoptosis

(a, b) Transient expression of CRKI or CRKII sensitizes crk-/- MEFs to BFA-induced apoptosis. n=3, error bars = sd. (c-e) Stable CRKII expression in crk-/- MEFs rescues sensitivity to 24h BFA- and 18h TUN-induced apoptosis, but does not change sensitivity to STS-induced apoptosis. n=3, error bars = sd. Scale bar, 100μm. (f, g) Stable overexpression of CRKII in wild-type MEFs further increases sensitivity to 18h BFA-induced apoptosis. n=3, error bars = sd. (h, i) Transient overexpression of CRKI sensitizes WT MEFs to 18h BFA (1.25 μg/ml)-induced apoptosis. n=3, error bars = sd.

To explore its potential role in ER stress, we monitored CRK mRNA and protein following BFA treatment (Supplementary Fig. S1). We detected no changes in CRKI or CRKII mRNA transcript levels upon ER stress (Supplementary Fig. S1). However, we found that ER stress causes depletion of full-length CRKII (Supplementary Fig. S2). To determine if CRKII is reduced in a specific subcellular compartment, we probed subcellular fractions from Bax-/-Bak-/- MEFs and found that CRKII is partially localized to the ER, mitochondria, and cytosol (Fig. 4a). Moreover, upon ER stress, 38kD ER- and cytosol- localized CRKII is reduced, while levels of the mitochondrion-localized full-length CRKII change very little (Fig. 4a). Interestingly, we found that upon ER stress, CRKII appears to be sequentially cleaved at least twice, resulting in several distinct fragments (Fig. 4a and Supplementary Fig. S3). Proteolytic cleavage is a post-translational modification recognized to activate other known pro-apoptotic proteins, such as BID, but had not been described previously for CRK[21-24].

Figure 4

CRK is proteolytically cleaved into an apoptotic signal upon irremediable ER stress

(a) Upon 24h BFA (2.5 μg/ml) treatment of Bax-/-Bak-/- MEFs, full-length CRKII is depleted in the cytosol and at the ER. CRKII-specific fragments (*) appear in the cytosol, ER, and mitochondria. (b) Transiently expressed CRKI is also cleaved upon 24h BFA (2.5 μg/ml) treatment in crk-/- MEFs. * = CRKI-specific fragment. (c) Loss of full-length, endogenous CRKI and CRKII observed upon 18h BFA treatment of WT and Bax-/-Bak-/- MEFs. (d) Upon ER stress, CRK is cleaved at D110. Mutation of this site (D110A) in CRKII prevents cleavage following 24h 2.5 μg/ml BFA treatment in stably reconstituted in crk-/- MEFs. (e, f) CrkII (D110A) is not able to rescue crk-/- MEF sensitivity to ER stress-induced apoptosis induced by 24h 2.5 μg/ml BFA, in contrast to crk-/- MEFs stably expressing wild-type (WT) CRKII. n=3, error bars = sd. (g) Diagram of CRK (1-110a.a.) cleavage fragment (NF110) produced upon ER stress. (h, i) Transient expression of NF110 induces apoptosis independent of ER stress. n=3, error bars = sd.

We tested if CRKI also undergoes cleavage events upon ER stress. Following ER stress, at least one CRKI-specific cleavage product is readily observed (Fig. 4b). In addition, we observe depletion of full-length endogenous CRKI and CRKII in both WT and Bax-/- MEFs upon BFA treatment (Fig. 4c), indicating that their cleavage occurs upstream of the mitochondrial apoptotic pathway. Furthermore, the loss of full-length CRKI and CRKII occurs only when the level of ER stress rises to cytotoxic levels, and correlates with the initiation of ER stress-induced apoptosis (Supplementary Fig. S4)[25-28].

In an attempt to determine the role of CRK cleavage for its apoptotic activity, we tested a small panel of protease inhibitors for their ability to block this event (data not shown). We found that the pan-cysteine protease inhibitor ZVAD-FMK prevents ER stress-induced loss of full-length CRKI and CRKII and protects Bax-/-Bak-/- MEFs from the cytopathic effects of ER stress (Supplementary Fig. S5). We previously identified caspase-2 as a pre-mitochondrial cysteine protease activated by ER stress[10]. However, caspase-2 is poorly inhibited by ZVAD-FMK and in vitro experiments with recombinant caspase-2 do not result in CRK cleavage (data not shown). These data suggest that a previously unidentified ER stress-activated cysteine protease is responsible for CRK cleavage.

To determine if the observed ER stress-induced cleavage of CRK is critical for its apoptotic activity, we individually mutated each potential cysteine protease cleavage site (aspartic acid) in the CRK sequence. When stably reconstituted into crk-/- MEFs, CRKII ΔD110A (crk-/- +crkII (ΔD110A)) was the only aspartic acid mutant unable to be cleaved in response to ER stress (Fig. 4d). Furthermore, non-cleavable CRKII ΔD110A is defective in restoring crk-/- MEF sensitivity to ER stress-induced apoptosis (Fig. 4e, f), arguing that this cleavage event is critical for its apoptotic activity. Cleavage at D110 is predicted to produce one fragment of approximately 25kD, which can be detected by a C-terminal-specific antibody (Fig 4d), and a second N-terminal fragment of ~14kD, which is undetectable using available antibodies. As both CRKI and CRKII restore crk-/- sensitivity to ER stress, it is likely the shared N-terminal fragment (~14kd) contains the critical domain for its apoptotic function. To test this prediction, we transiently expressed the N-terminally FLAG-tagged fragment (NF110) in the absence of ER stress (Fig. 4g) and measured apoptosis. As predicted, NF110 is able to potently induce cell death independently of ER stress (Fig. 4h, i). From these data, we conclude that CRK is cleaved upon ER stress at D110, to produce a pro-apoptotic fragment. Further studies will be necessary to identify the upstream protease and its connection to the unfolded protein response pathway.

To further investigate the mechanism by which CRK triggers cell death, we tested if CRK-induced apoptosis is a BAX/BAK-dependent process. Transient overexpression of CRKI or CRKII triggers apoptosis in WT MEFs, but not Bax-/-Bak-/- MEFs, confirming that both isoforms signal upstream of the BAX/BAK-dependent mitochondrial apoptotic pathway (Fig. 5a-d). Pro-apoptotic BH3-only proteins are the only known molecules capable of activating BAX and/or BAK either directly or by inhibiting anti-apoptotic BCL-2 family proteins[4,13]. Through sequence analysis, we identified a putative BH3-like domain within CRK that contains a number of conserved amino acids present in several recognized BH3-only proteins (Fig. 5e). This sequence is present in both crk splice forms and located within the common N-terminal 110 a.a. pro-apoptotic fragment (Fig. 1d, 5e). In support of our hypothesis that CRK contains a BH3-like domain, we determined that CRK is capable of binding a prototypical anti-apoptotic BCL-2 family protein (BCL-XL), a common feature of most BH3-only proteins. Following ER stress induction, transiently expressed FLAG-CRKII co-immunoprecipitates with BCL-XL on FLAG-specific agarose beads (Fig. 5f).

Figure 5

CRKII contains a putative BH3 domain and triggers BAX/BAK-dependent apoptosis

(a, b) CRKII and empty vector (pmx) were transiently overexpressed in WT and Bax-/-Bak-/- MEFs using retroviral infection. 24h post retroviral infection cells were treated with BFA (2.5 μg/ml) for an additional 24h and analyzed for Annexin-V expression by flow cytometry. n=3, error bars = sd. (c, d) CRKI and empty vector (pmx) were transiently overexpressed in WT and Bax-/-Bak-/- MEFs. 24h post transfection cells were treated an additional 18h with BFA (2.5 μg/ml) and analyzed for Annexin-V expression by flow cytometry. n=3, error bars = sd. (e) The sequences of the putative BH3-only domain of CRK and the “BH3 domain” point mutation D91A are aligned against BH3 domains of several known BH3-only proteins. (f) 293 cells were transiently transfected 24h with Flag-crkII or untagged crkII, then treated 14h BFA (1.25μg/ml). Lysates were incubated with FLAG-specific agarose beads. Beads were immunoblotted for endogenous BCL-XL. (g) Cytochrome c release from isolated Jurkat mitochondria incubated with decreasing doses of tBID and CRK BH3 domain peptide. n=3, error bars = sd. (h, i) Stable reconstitution of D91A crkII into crk-/- MEFs is significantly less effective at restoring ER stress-induced apoptosis (24h BFA treatment) in comparison to expression of wild-type crkII. n=3, error bars = sd.

To determine if this sequence has BH3-like pro-apoptotic activity, we treated isolated Jurkat mitochondria with a synthetic CRK BH3 domain (77-96 a.a.) and measured cytochrome c release. There are two classes of BH3-only domains, those that sensitize or activate BAX/BAK-dependent mitochondrial apoptosis[29]. BH3 domains that directly “activate” BAX and/or BAK at mitochondria, such as the BH3 domain of BID, are able to cause cytochrome c release from isolated mitochondria. In contrast, “sensitizing” BH3 domains, such as the BH3 domains of BAD and BIK, sensitize isolated mitochondria to release cytochrome c in the presence of a second “activating” BH3 domain. In vivo, sensitizer BH3-only proteins are thought to competitively bind anti-apoptotic proteins, releasing bound “activating” BH3-only proteins to induce mitochondrial permeability. While the CRK BH3 domain is unable to induce cytochrome c release alone, it significantly potentiates with low concentrations of truncated BID (tBID) to cause cytochrome c release (Fig. 5g).

To further examine if this putative BH3 domain is required for CRK's pro-apoptotic activity in cells, we mutated the highly conserved aspartic acid (D91A) in crkII and evaluated the ability of this mutant to rescue the crk-/- MEF phenotype. In comparison to wild-type crk, the BH3 mutant crk (D91A) has significantly decreased apoptotic activity, arguing that this region is critical for its pro-death signaling (Fig. 5h, i).

Together, these results identify CRK as a major pro-apoptotic signal required for the execution of ER stress-induced cell death. During ER stress, CRKI and CRKII are cleaved by a yet to be identified cysteine protease to generate an N-terminal fragment with potent apoptotic activity. Furthermore, CRK interacts with anti-apoptotic BCL-XL and its apoptotic activity is upstream of the BAX/BAK-dependent mitochondrial pathway. Both CRK isoforms contain a putative BH3 domain, which sensitizes isolated mitochondria to tBID-induced cytochrome c release and when mutated diminishes apoptotic activity in cells. These data argue that CRK is a previously unidentified BH3-only-like protein, which upon ER stress is proteolytically processed into a pro-death signal. Our findings suggest that CRK may be a valuable therapeutic target in diseases where ER stress-induced cell loss is implicated, including some forms of neurodegeneration and diabetes[1,2].

We have identified a previously unknown pro-apoptotic function common to both CRK isoforms. CRK was initially identified through its homology with transforming v-crk[. However, only CRKI has been shown to have transforming activity in some cell culture models, and it is upregulated in a number of human cancers[32,33]. Clues that CRK mediates apoptosis are present in other species. For example, the C. elegans Crk-homologue CED-2 regulates apoptotic engulfment[34,35]. Mammalian CRKII has been reported to induce death in some transformed cell types upon overexpression[36,37] and is required for apoptotic activity that can be detected in Xenopus egg extracts[38-40]. Our work is the first to connect CRK to apoptosis, specifically under ER stress, in mammalian cells.

In addition to the shared pro-apoptotic function we have discovered for CRKI and CRKII, there are notable and possibly functionally significant differences between the isoforms. Our inability to establish a cell line stably overexpressing CRKI, in contrast to multiple cell lines stably overexpressing CRKII, suggests that CRKI may be the more cytotoxic isoform. Indeed, the expression of endogenous CRKI is restricted to approximately 10% that of CRKII (Supplementary Fig. S1), perhaps to limit its toxicity. In support of this notion, we observe that CRKI is cleaved more efficiently and at earlier kinetics than CRKII in response to ER stress (Fig. 4c and Supplementary Fig. S4). These observations and our discovery that CRKI can be converted into a pro-apoptotic protein in response to ER stress raise the possibility that pharmacologic inducers of ER stress may have therapeutic efficacy in cancers where crkI is upregulated.