The SNIPer: Design, Optimization, and Characterization in Mammalian Cells
Our initial split-TEV construct was based on a previous design developed for protein-complementation assays (Wehr et al., 2006
. The full-length TEV variant S219V was split into two component fragments: N-TEV (residues 1-118) and C-TEV (residues 119-242). This design inactivates TEV protease by separating the catalytic triad residues Asp44 and His81 from the Cys151 nucleophile. N-TEV was fused to the C-terminus of FKBP with a 10-amino acid (GGGGS)2×
linker between the two domains, and C-TEV was fused to the C-terminus of Frb (the rapamycin-binding domain of mTOR kinase). Therefore, protease activity is conditionally restored by the addition of rapamycin, which promotes split-TEV fragment association and ligand-dependent protein folding.
Engineering the SNIPer for Conditional Proteolysis
We assayed for reconstituted protease activity in 293T cells by co-transfecting both N-and C-TEV halves and a FRET-based ECFP-YPET reporter that is cleaved by TEV. In the presence of rapamycin (Rap), the split-TEV construct extending to residue 242 cleaved the protease reporter as compared to an empty vector control ()
. However, significant processing also occurred in the absence of Rap, suggesting the two halves can associate and reconstitute activity in the absence of a small molecule dimerizer. Another group recently implemented the published N-TEV and C-TEV design in mammalian cells and similarly observed leaky TEV protease activity in the absence of Rap (Williams et al., 2009
). They reduced Rap-independent activation by lowering the amount of split-TEV DNA that was transiently transfected or by targeting one fragment to the cell membrane via a genetically encoded myristoylation tag. However, our goal was to engineer cell death pathways using stable cell lines, which requires a split-TEV design with no background activity. We hypothesized that the C-terminus of the TEV was causing spontaneous refolding of the fragments in the absence of Rap (Figure S1A
for a model)
. A systematic set of deletion mutants was created from the C-TEV119-242
fragment to generate four truncated C-TEV constructs. Co-expression of FKBP-N-TEV and Frb-C-TEV-219 yielded inducible proteolysis upon addition of Rap with reduced leaky proteolytic activity in mock-treated cells as confirmed by the FRET assay ()
and immunoblot analysis (Figure S1B
, top panel)
Further optimization was required due to detectable TEV activity for FKBP-N-TEV and Frb-C-TEV-219 in the absence of Rap (
, bottom panel)
. We tested the opposite FKBP and Frb protein fusions to N-TEV and to the enhanced C-TEV-219 fragment, and we observed that Frb-N-TEV(1-118) and FKBP-C-TEV(119-219) maintained robust TEV activity in the presence of Rap with little background. As an initial test of split-TEV stability, we found that proteolysis of the FRET reporter was not enhanced after treatment with Rap and 500nM PS-341 (Bortezomib), a clinically available proteasome inhibitor (Figure S1C)
, indicating that a proteasome inhibition was not sufficient to promote increased split-TEV activity in cells. Each split-TEV protein fragment was also detected by immunoblotting after blocking protein translation with cyclohexamide for 8hr, suggesting both were stable in cells (Figure S1D)
. Both removal of the C-terminal end of the C-TEV fragment and fusion pair optimization to FKBP and Frb were necessary to generate a rapamycin-inducible split-TEV protease for mammalian cells. We refer to this optimized construct pair as the SNIPer (Single Nick in Proteome) since there are no known natural TEV cleavage sites in the human proteome except for those we introduce.
We observed rapid activation of the SNIPer in individual cells (
, left panel)
using the TevS-FRET reporter in transiently transfected HEK293 cells. Cells were monitored over a 3 hr time frame over which a linear increase in TEV activity was measured by single-cell fluorescent microscopy. Approximately 50% of the ECFP-YPETENLYFQ|S
substrate was cleaved by 100 ± 20 min. Importantly, activation of the SNIPer at the single-cell level was more synchronous than endogenous caspase pathways during experiments in which SNIPer-expressing cells were directly compared to HEK293 cells transfected with ECFP-YPETDEVD|R
, a reporter for caspase-3 and -7 activity (Albeck et al., 2008a
) (, right panel
), and treated with 1uM staurosporine (STS). We conclude that SNIPer activation kinetics matches or exceeds endogenous DEVDase activation kinetics during apoptosis, which prompted its further use to probe executioner caspase signaling in cells.
Engineering Orthogonal Procaspase-3 and -7 Alleles for Selective Activation by the SNIPer
The executioner caspases are translated as inactive zymogens that are cleaved by initiator caspases, by other executioner caspases in trans,
or by auto-activation. Two sites of processing exist in procaspase-3 and -7: one at the junction of the prodomain and large subunit (site-1) and the second between the large and small subunits in the intersubunit linker (site-2) (Earnshaw et al., 1999
). To date, there are no reported crystal structures of procaspase-3 or procaspase-7 with defined electron density in the intersubunit linker or the prodomain. However, previous studies have shown that mutation of several residues within the “loop bundle” of the intersubunit linker of caspase-3 influences the activity of the protease (Feeney et al., 2006
). Therefore, we inserted a TEV recognition sequence (TevS) into the P1 position at site-2 on procaspase-3, replacing the P1 Asp of the endogenous caspase site, P5-IETD-P1′ with the preferred P6 Glu for the Tev cleavage site (TevS), P7-ENLFYQ-P1′. This resulted in a chimeric substrate at site-2 (IETENLYFQ) that maintained the integrity of the caspase cleavage site from P4-P2 while eliminating the aspartic acid critical for cutting by the caspases themselves (
, top panel)
. In principle, this approach would render the sites cleavable by the SNIPer only. Since caspases and TEV protease share similar preference for the small P1′ residues (Gly, Ser and Ala), the wild-type peptide sequence after the scissile bond was also maintained. Once the TevS is cleaved, a small peptide ‘scar’ is left behind with an ENLYFQ sequence on the C-terminus.
Design and In Vitro Characterization of Orthogonal Procaspase-3 and -7 Alleles
With this insertion strategy we generated four alleles in both procaspase-3 and -7: one where TevS was inserted between the pro-domain and the large subunit (TevS-1), one where the TevS was inserted between the large and small domain (TevS-2), one where two insertions were made (TevS-1,2) and one that contained the TevS inserted between the large and small subunits but that also contained a mutation of Asp to Ala mutation at site-1 to prevent prodomain cleavage (D2A-TevS-2) (
, bottom panel)
. To begin to test the impact of these mutations on processing and activity, a His-tagged version of the Casp3-D2A-TevS-2 protein was over-expressed in E. coli
and purified by Ni-NTA affinity chromatography. Casp3-D2A-TevS-2 was co-incubated with recombinant TEV protease, and the processed samples were analyzed by SDS-PAGE to confirm that the engineered caspase construct was indeed cleaved at the expected single site (Figure S2A)
. Furthermore, the measured catalytic efficiency (kcat
) for the engineered caspase was in good agreement with the value for wild-type active caspase-3 produced in our lab (Figure S2B)
To further probe the functional consequence of processing at site-1 and site-2 in vitro
, we constructed a matrix of the three procaspase-3 and -7 alleles harboring TevS sites that replace one or both endogenous caspase processing. Granzyme-B (GrB), a serine protease secreted by cytotoxic T-lymphocytes, is capable of cleaving procaspase-3 at site-2 (Casciola-Rosen et al., 2007
), but it does not process Casp3-TevS-2 since that cleavage site is mutated. Incubation of Casp3-WT and Casp3-TevS alleles with active GrB or Tev confirmed Casp3-TevS-2 is an orthogonal caspase; it is activated by TEV protease but not by GrB ()
Each procaspase-3 and -7 allele was expressed as a proenzyme in E. coli
, treated with GrB or TevS for 1 hr, and relative caspase activity was measured with Ac-DEVD-AFC ()
. Immunoblots confirmed the predicted cleavage patterns for TEV protease or GrB for procaspase-3 allele series (Figure S2C)
. Cleavage by GrB at IETD175
at site-2 was sufficient to activate WT and TevS-1 procaspases variants, both of which are WT for site-2. TEV processing at TevS-2 was sufficient to activate both caspase-3 and -7-TevS-2, regardless if the prodomain is cleavable. TEV protease processing at TevS-1 was not sufficient to induce activity for either isoform. Together, these results confirm that processing at site-2 is both necessary and sufficient for activation in vitro
by either TEV protease or GrB. Processing of the prodomain is neither sufficient for activation nor necessary for full levels of activation in conjunction with processing at the inter-subunit linker.
Selective Activation of Orthogonal Procaspase-3 and -7 Alleles in Stable Human Cell Lines
The matrix of caspase-TevS alleles was stably engineered as single-copy integrants into HEK293 cells as described in the Supplementary Experimental Procedures
in order to minimize the over-expression phenotypes associated with transient transfection. HEK293 SNIPer cell lines expressing Casp3-WT, Casp3-TevS-1 or Casp3-TevS-2 were induced for 6 hr with 10nM Rap and assayed for DEVDase activity. Addition of Rap induced a 8-fold increase in caspase-3 activity in the Casp3-TevS-2 line as compared to the same cells treated with DMSO only (
, left top and bottom panels)
. After 8 hr, the Casp3-TevS-2 cells showed robust GFP-Annexin-V staining (Ernst, 1998
), a hallmark of apoptosis involved in the recruitment of macrophages for engulfment and clearance of apoptotic cells. By 16 hr, only 25% of the cells remained viable. In contrast, no increase in DEVDase activity, Annexin-V staining or cell death was observed with the Casp3-WT or Casp3-TevS-1 lines treated with Rap. Similar functional results were obtained for the Casp7-TevS-2 line, which showed a 6-fold induction of DEVDase activity by 6 hr, an increase Annexin-V staining by 8 hr, and a large decrease in cellular viability by 16 hr ()
. Overall, the relative phenotypic consequences of the WT, TevS-1 and TevS-2 lines were similar for procaspase-3 and procaspase-7, although total DEVDase activity and Annexin-V staining were less for activated Casp7-TevS-2 as compared to the corresponding procaspase-3 allele. These data establish that SNIPer-mediated cleavage and activation of either procaspase-3 or -7 at site-2 alone
is sufficient to induce robust caspase activation and rapid apoptosis in human cells.
Phenotypic Analysis of SNIPer-Mediated Activation of Procaspase-3 and -7 in Cells
To further understand the functional consequences of prodomain cleavage in cells, we compared the D2A-TevS-2 and TevS-(1,2) alleles (). In parallel to the in vitro results, prodomain removal is neither necessary nor sufficient for caspase activation and apoptosis in cells. A comparison of the lines expressing TevS-2, D2A-TevS-2 and TevS-(1,2)-Casp7 suggested that prodomain removal results in modestly lower DEVDase activity at 6 hr (~40%) and reduced Annexin-V staining.
It has been proposed that the executioner caspases may amplify apoptotic pathways via cleavage of Bid to generate tBid, which induces mitochondrial outer membrane permeability (MOMP), cytochrome-C release and activation of caspase-9 (Green, 2005
) (Figure S3A)
. To examine if executioner caspase activation can ‘feedback’ and result in the activation of initiator caspases, we tested whether SNIPer-mediated activation of the Casp-7-TevS-2 allele was sufficient to promote MOMP. We observed several biochemical hallmarks of MOMP after 4 hr of Rap induction, including Bid cleavage and cleavage of procaspase-9 (Figure S3B)
. As expected, extensive PARP processing was also detected. These results further confirm the existence of positive feedback signaling between the executioner caspases and upstream apoptotic pathways.
Activated Caspase-3 and -7 are Proteasome Substrates and are Stabilized by a Proteasome Inhibitors
Processing of engineered procaspase alleles by the SNIPer was examined in cells by immunoblotting with an anti-FLAG antibody that detects the C-terminal His/FLAG epitope tag on both Casp3-TevS-2 and Casp7-TevS-2 constructs. The anti-FLAG antibody detects the full-length proenzyme (32 kDa), a processed form cleaved only at site-1 (29 kDa), and the form cleaved at site-2 to generate the small subunit (12 kDa). We also used a polyclonal caspase-3 antibody that recognizes both caspase-3 subunits. After the addition of Rap to the Casp3-TevS-2 cell line for 6hr, which correlates with the induction of DEVDase activity and apoptosis, the small subunit (~12kDa) of cleaved caspase-3 could not be readily detected by immunoblot Similar results were observed for Casp7-Tevs-2 ().
Caspase-3 and -7 Regulation by the Proteasome
Recent studies have suggested that caspase-3 is ubiquitinated and degraded by the proteasome (Suzuki et al, 2001
; Choi et al. 2009
). To address the role of the Ubiquitin-Proteasome System (UPS) in mediating turnover of SNIPer-activated caspases, we added one of two different proteasome inhibitors, MG-132 or PS-341, in combination with Rap, which stabilized the cleaved caspase subunits for both caspase-3 and caspase-7, respectively ()
. Furthermore, Rap and PS-341 treatment promoted the accumulation of polyubiquitinated caspase-3 detected by immunoblotting as compared to Rap alone ()
. To evaluate the effects of the proteasome inhibition on the induction of caspase activity, a time course of Rap-induced processing of the procaspase-3 and -7 allele series was performed ±5uM MG-132. Addition of MG-132 resulted in a 6- to 8-fold enhancement in Rap-induced DEVDase activity for those caspase-3 and -7 alleles that were active in cells (TevS-2, D2A-TevS-2 and TevS-1,2 variants) (
; Figure S4A)
An increase in caspase activity induced by MG-132 directly correlated to the severity and kinetics of the apoptotic phenotype ()
. For example, at 4h, MG-132 increased Annexin-V staining to 60% of cells compared to 25% with Rap alone. By 8 hr, almost 50% of the cells were double positive for Annexin-V and propidium iodide, a marker for the loss of cell membrane integrity. Thus, selective activation of either caspase-3 or -7 in combination with MG-132 robustly increased the level of DEVDase activity, accelerated apoptotic hallmarks such as Annexin-V staining and, by 8 hr, resulted in a large fraction of late-stage apoptotic cells. A minor population of cells were PI positive only, excluding necrotic cell death as the primary response to proteasome inhibitors. The addition of both Rap and MG-132 for as long as 8 hr to Casp3-WT or Casp7-WT control lines did not induce a increase in caspase activity (
), which further established a direct relationship between the UPS and the activated forms
of capase-3 and -7 generated by the SNIPer. We also tested the possibility that proteasome inhibitors stabilized the SNIPer itself. However, HEK293 cells stably expressing both SNIPer and the TevS-FRET reporter showed no detectable proteasome inhibitor-dependent enhancement of TEV activity by the FRET assay, which was further confirmed by immunoblotting to assess FRET reporter cleavage (Figure S1C)
We observed similar levels of Casp7-TevS-2 activation in experiments in which MG-132 or the more potent and selective PS-341 was directly compared (Figure S4B)
. To determine if proteasome inhibition promoted executioner caspases in a second cell line, we co-expressed the SNIPer in HeLa cells with retroviruses expressing the Casp3-WT and Casp3-TevS-2 alleles, and we successfully activated the SNIPer and Casp3-TevS-2 in a Rap-dependent fashion (Figure S4C)
. Moreover, the level of caspase-3 induction was potently enhanced by MG-132.
Activation of Caspase-6 is not Proapoptotic except in Conjunction with Proteasome Inhibition
Procaspase-6 activation was tested in vitro
and in cells with the SNIPer. Although procaspase-6 is defined as an executioner caspase and is closely related to caspase-3 and -7 by sequence homology, several distinct features have been observed for this isoform. First, procaspase-6, unlike procaspase-3 and -7, has been reported to self-activate when over-expressed in 293T cells (Klaiman et al., 2009
). There is an additional loop in the intersubunit linker that bisects a region known as the “safety catch” at the homologous site in procaspase-3. Second, procaspase-6 prefers Val or Thr at P4 unlike caspase-3/-7, which prefer acidic residues at this position (Thornberry et al., 1997
). Procaspase-6, therefore, resembles the initiator caspases by a tendency to self-activate and by similar substrate preferences at P4.
To probe caspase-6 signaling in apoptosis, we applied the TevS insertion strategy to engineer a series of orthogonal procaspase-6 alleles. In addition to site-1 at the junction of the prodomain and the large subunit, there are two reported cleavage sites in the intersubunit linker of procaspase-6 (): DVVD|N, which we refer to as site-2a, and TEVD|A, or site-2b. Although TEV protease can process sites with a P1′ Asn with reduced efficiency, an Ala was substituted at P1′ to enhance proteolysis at Site 2a. Thus, three TevS alleles were engineered by insertion at each of the three reported sites (site-1, site-2a with a P1′ Ala, and site 2b).
Proteasome Inhibition Converts Procaspase-6 to an Apoptotic Executioner Caspase Isoform
Each construct was expressed in E. coli and processed in vitro with recombinant TEV protease or recombinant active caspase-3, which is capable of activating WT procaspase-6. A fluorogenic Ac-VEID-AFC peptide substrate, the preferred substrate for caspase-6, was used in these experiments (). TEV protease could process and activate the TevS-2 variants (2a and 2b), whereas recombinant caspase-3 activated Casp6-WT, and, unexpectedly, Casp6-TevS-2b. Because this allele is WT at site-2a, and caspase-3 was not able to activate the TevS variants at site-2a, we infer that site-2a is a functional caspase-3 processing site (). TEV protease did activate Casp6-TevS-2a with an Asn at P1′ (data not shown), but total VEIDase activity was less than that of the Ala mutant.
Casp6-TevS variants were transduced into HEK293 cells with the SNIPer, and cell-based assays were employed to measure caspase-6 activity in cells. Casp6-TevS-2a was nearly fully processed by the SNIPer by 6 hr (Figure S5A)
, but induced VEIDase activity was only 2.4-fold above the control (
, left panels).
Cellular viability was not significantly affected by caspase-6 activation at 16 or 24 hours post-Rap for the Casp6-TevS-2a line as compared to Casp6-WT or to an empty vector control ()
, and we did not observe a significant apoptotic response by Annexin-V staining (
, right panels).
We hypothesized that the UPS may negatively regulate caspase-6 activity as well. Cells harboring the SNIPer and the Casp6-TevS-2a allele were treated with Rap±MG-132. Remarkably, VEIDase activity with Rap and MG-132 was elevated 14-fold at 4hr compared to Rap alone or to Casp6-WT or Casp6-TevS-1 control lines (
, left panels).
Amplification of VEIDase activity was accompanied by a dramatic increase (50%) in Annexin-V positive cells (
, right panels)
. To further probe the mechanism by which MG-132 promotes caspase-6 activity and apoptosis, we immunoblotted for caspase-6 before and after SNIPer-mediated cleavage. Surprisingly, MG-132 stabilized the unprocessed Casp6-TevS-2a zymogen (Figure S5A
, lanes 1 and 2)
. This result was independent of SNIPer-induced cleavage and in stark contrast to the mode of caspase-3 and -7 regulation by proteasome. When Casp6-TevS-2a cells were treated with Rap±MG132 (Figure S5A
, lanes 3 and 4),
the small subunit was also stabilized with MG-132. Similar results were obtained with PS-341. We next tested if caspase-6 activation induced MOMP, and after 8hr of sustained Casp6-TevS-2a activation with Rap only, we observed no evidence for MOMP. However, addition of PS-341 and Rap caused a large increase in Bid cleavage, procaspase-9 and PARP (Figure S5B)
. These results suggest that both procaspase-6 and activated caspase-6 are targets for negative regulation by the UPS, and caspase-6 is only proapoptotic in the presence of proteasome inhibitors. The severity of the apoptotic phenotype directly correlates with increased VEIDase activity in cells, stabilization of both the pro- and active forms (unlike casase-3 and -7), and recruitment of the endogenous apoptotic machinery via MOMP.
Caspases Attack the 26S Proteasome, and Caspase Activation Synergizes with Proteasome Inhibition
Recently gel-based proteomic data from several other laboratories have identified 10 proteins from the 26S proteasome, mostly in the 19S cap, that are cleaved during apoptosis (Sun et al., 2004
; Adrain et al., 2004
; Van Damme et al., 2005
; Thiede et al., 2005
Jang et al., 2007
). We have described a new proteomics strategy for the global identification of protease substrates in mammalian cells (Mahrus et al., 2008
). This method utilizes an engineered protein ligase called subtiligase to biotin-tag and enrich for nascent N-termini produced during proteolysis, which are then sequenced by LC/MS/MS. A distinct advantage to this technology is that the exact site of proteolysis is identified. Our group has identified >1000 caspase substrates in several cell lines induced by a variety of apoptotic drug stimuli (S. Mahrus, H. Nguyen, and J. Wells, manuscript in preparation). Within this data set, we found 12 caspase substrates in the 26S proteasome and defined their cleavage sites from MS/MS data (Table S1)
. Ten of these are novel. Surprisingly, the substrate list included two additional α-subunit sites and three β-subunit sites of the 20S catalytic core particle. One of the three β-subunit cleavage sites identified by our degradomics experiments, PSB7_HUMAN or β2, is a bona fide
proteasome catalytic subunit, providing trypsin-like hydrolase activity to the complex. The site of caspase proteolysis in the β2 subunit is only 10 residues from the catalytic N-terminal nucleophile and was likely missed in previous gel-based studies. These mass spectrometry data do not address the extent of cutting for each of these subunits by caspases, nor their functional impact. Nonetheless, the aggregate of our data plus previous publications show that 20 of 33 of the subunits in the 26S proteasome can be cleaved in a cellular context by caspases ()
. Further studies will be required to assess the functional impact and timing of these individual cuts on proteasome function.
Reciprocal Negative Regulation between the 26S Proteasome and Executioner Caspases
We then assayed for proteasome function during apoptosis in our system. As caspase activity increased upon activation of the Casp3-TevS-2 allele by the SNIPer, proteasome activity decreased coincidently (). These data lead us to hypothesize that there is a negative feedback loop between the executioner caspases and the proteasome (). This model predicts that caspase activation should synergize with proteasome inhibition. To test this prediction, we modulated the activation of Casp3-TevS-2 by the SNIPer by titrating Rap (EC50 of ~4 nM for SNIPer activation of the caspases) in combination with titration of MG-132 to achieve a range of proteasome inhibition (IC50 of 0.8 uM). We observed dramatic synergy between caspase-3 activation and inhibition of the proteasome (). For example, at the EC50 for SNIPer activation by Rap (~2 nM), the addition of saturating MG-132 (5uM) yielded a 7.7-fold induction of caspase activity as compared to 1.6-fold induction with Rap only. Similarly, at the EC50 for MG-132 (0.8 uM), addition of saturating amounts of Rap (8nM) induced a 8.9-fold increase in caspase activity as compared to 2.8-fold with Rap alone.