Development of activity-based probes for caspase-6
In order to develop a probe to monitor caspase-6 activity, we used AB50 (Edgington et al., 2009
) as a starting scaffold and changed the peptide specificity region based on the reported caspase-6 substrate preferences ( and Supplemental Figure 1
). We chose the sequence Val-Glu-Ile-Asp (VEID) because it is the sequence recognized by caspase-6 on Lamin A/C, a substrate that is not effectively processed by the other executioner caspases (Rao et al., 1996
). Furthermore, many commercially available substrates and inhibitors for caspase-6 also make use of this sequence. We also designed a probe containing the sequence Ile-Val-Leu-Asp (IVLD) corresponding to the site on the huntingtin protein that has been shown to be cleaved by caspase-6 to generate neurotoxic fragments (Graham et al., 2010
). Finally, we used the optimal substrate sequence for caspase-6, Val-Glu-His-Asp (VEHD), reported from fluorogenic substrate screening assays (Thornberry et al., 1997
). We also screened purified, recombinant caspase-6 against our previously reported positional scanning library of AOMK inhibitors (Berger et al., 2006
). This screen indicated that caspase-6 strongly preferred threonine at both P2 and P4 positions, while caspase-3 did not (Supplemental Figure 2
). Therefore, we also synthesized a probe containing the sequence Thr-Glu-Thr-Asp (TETD). All of the probes were readily synthesized in Cy5-labeled form using a combination of solid phase and solution phase synthesis (Supplemental Scheme 1).
Direct comparison of LE22 and AB50 labeling in intact apoptotic cells
To test these new probes for caspase-6 labeling, we used an in vitro
model of apoptosis in which human COLO205 colorectal cancer cells were stimulated with a death receptor 5 agonist antibody (anti-DR5) to induce the extrinsic cell death pathway. Labeling of intact cells with each of the new probes showed similar patterns; however, the probe based on the lamin cleavage sequence, LE22, was by far the most potent (Supplemental Figure 1C
). We next compared the labeling of LE22 to our previously reported probe AB50 (). Most strikingly, we found that LE22 was more effective at labeling caspases than AB50. This may be due to a combination of increased cell permeability of the probe as well as increased potency for the executioner caspases. Although the labeling profile of LE22 was similar to AB50, it clearly labeled an 18 kDa protein and several higher molecular weight proteins between 30–36 kDa that were not labeled by AB50 (). Because LE22 was designed to target caspase-6, we hypothesized that these additional labeled proteins may be multiple forms of caspase-6. As an initial test of this hypothesis, we pre-treated cells with the AOMK inhibitor AB13 (Berger et al., 2006
) that we previously showed to be selective for mature forms of caspase-3 and -7 prior to labeling with LE22 (). Interestingly, inhibitor pretreatment completely blocked labeling of all of the AB50 labeled proteins but had no effect on the labeling of the LE22-specific 18kDa and 30–36kDa species.
To confirm the identity of the labeled caspases, we performed immunoprecipitation studies using antibodies raised against the cleaved forms of caspase-3, -6 and -7 (). As expected, we were able to precipitate cleaved forms of caspase-3 and -7 in cells labeled with AB50 or LE22. The caspase-6 antibody immunoprecipitated both the 18kDa protein as well as the higher molecular weight (30–36) kDa proteins that were only found in the LE22 labeled cells. These data suggest that LE22 is an overall more potent and sensitive probe than AB50 that labels caspase-3 and -6, and to a lesser extent caspase-7.
Cross-reactivity of caspase probes
In our previous studies, we found that the lysosomal cysteine protease, legumain is the major off-target of the AB50 probe (Edgington et al., 2009
). The active site of this enzyme has a similar overall fold to the caspases but is thought to predominantly cleave protein substrates after aspargine residues (Abe et al., 1993
). However, legumain is also able to bind aspartic acid-containing probes at the reduced pH of the lysosome (Kato et al., 2005
; Sexton et al., 2007
). To monitor selectivity, we used both AB50 and LE22 to label RAW cells, an immortalized mouse macrophage line that expresses high levels of active legumain and cathepsins (). As reported previously, we observed significant labeling of legumain by AB50. LE22, on the other hand, showed no labeling of legumain, even when used at high concentrations, indicating a markedly reduced cross-reactivity for this probe.
Cross-reactivity of probes towards other cysteine proteases
We also evaluated the specificity of both probes in vivo by treating wild type mice and analyzing kidney and liver extracts by fluorescence imaging and SDS-PAGE (). As a result of the high levels of expression of legumain in these organs, LE22 exhibits some cross-reactivity towards this protease, however the cross-reactivity was considerably less than AB50, consistent with the selectivity patterns observed for these probes in RAW cells ().
Application of LE22 to in vivo models of apoptosis
We next determined whether the increased potency of LE22 towards caspases -3, -6 and -7 relative to AB50 was also observed in in vivo models of cell death. We initially tested both probes in a mouse model in which dexamethasone (dex) was used to induce apoptosis in CD4+/CD8+ thymocytes (). For both probes, we observed a dex-dependent increase in both thymus fluorescence and caspase labeling as assessed by SDS-PAGE (). The dex-induced caspase labeling could also be blocked by pre-treating mice with the broad-spectrum caspase inhibitor, AB46. Immunoprecipitation experiments confirmed that AB50 labeled caspase-3 and a small amount of caspase-6, while LE22 labeled caspase-3 and -6 to a similar extent (). Caspase-7 activity was detected by both probes, but to a greater extent by AB50. In addition, in agreement with our results in the mouse RAW cells and normal mice, we only observed legumain labeling by AB50 and not by LE22.
Comparison of LE22 and AB50 in dexamethasone-induced thymocyte apoptosis
We also used LE22 and AB50 to label caspases in tumors induced to undergo apoptosis by chemotherapy. In this model, human COLO205 colorectal tumor cells were xenografted onto the backs of nude mice and later induced to undergo apoptosis by treatment with the anti-DR5 antibody. While both probes showed an anti-DR5-dependent increase in fluorescence within the tumor, LE22 treated tumors were brighter and showed better contrast over the non-treated controls (). After non-invasive imaging, we removed tumors and performed ex vivo imaging followed by SDS-PAGE analysis of tumor lysates (). Again we found that fluorescence increased in response to anti-DR-5 antibody treatment and signals from LE22-treated tumors were overall much brighter than those from AB50 treated tumors. Biochemical analysis of the tumor tissues verified that fluorescence intensity correlated with levels of caspase labeling. Again, LE22 showed stronger labeling of caspases than AB50 and also labeled less legumain.
Comparison of LE22 and AB50 in COLO205 tumors treated with anti-DR5 anti-body
We also wanted to determine if LE22 could be used ex vivo to monitor caspase-6 activation and maturation. We therefore labeled extracts from anti-DR5-treated tumors with AB50 and LE22 and analyzed labeling by SDS-PAGE. As we observed in intact COLO205 cells in vitro, there is a clear increase in labeling intensity of caspases for LE22 compared to AB50 (). We also used the caspase-3 and -7 specific inhibitor AB13, to block the activity of these two proteases. This treatment demonstrated that AB13 blocked labeling of all but three species by LE22 and completely blocked labeling of all species by AB50. Immunoprecipitation using caspase-3, -6 and -7 specific antibodies confirmed that the remaining proteins labeled by LE22 were in fact various forms of caspase-6 (). Overall, the labeling patterns were strikingly similar to what we observed in vivo, except that one of the pro-caspase-6 forms was obscured by legumain in the in vivo samples (legumain was not observed in lysates since labeling was carried out at neutral pH).
Monitoring caspase-6 activation with LE22
To determine whether differences in the sensitivity of AB50 and LE22 were due to permeability or uptake of the probes, we examined probe labeling in apoptotic lysates. As before, COLO205 cells were treated with anti-DR5 for four hours, followed by hypotonic lysis and subsequent probe labeling for 30 minutes (). Overall, we observed approximately 10-fold enhanced potency of LE22 compared to AB50. Interestingly, unlike in intact cells (), AB50 was able to label the mature form of caspase-6 in lysates. This result suggests that part of the reason for the lack of labeling of caspase-6 by AB50 may be due to reduced access of the probe to the intracellular caspase-6 pool. Cleaved caspase-6 has been shown to accumulate in the nucleus of COS cells upon staurosporine treatment (Warby et al., 2008
), suggesting that AB50 could have reduced nuclear access.
Direct comparison of LE22 and AB50 labeling in apoptotic lysates
In lysates, we saw more pronounced labeling of the larger forms of caspase-6, which were appropriately sized to be proforms that had not been cleaved between the large and small subunits (). We were initially surprised to see labeling of these forms, as activation of caspase-6 is thought to depend on removal of the prodomain and cleavage of the intersubunit linker. We also did not expect these forms to immunoprecipitate with an antibody that recognizes cleaved forms of caspase-6 (). This polyclonal antibody was raised against the c-terminus of the large subunit of caspase-6, and therefore should not detect full-length forms. The most plausible explanation for our results was that we were precipitating caspase-6 dimers in which a labeled, full-length uncleaved monomer was in a complex with a cleaved monomer. This half-cleaved complex would be similar to what has previously been reported for caspase-7 (Berger et al., 2006
; Denault et al., 2006
To test this hypothesis, we performed immunoprecipitation using probe-labeled apoptotic lysates that were denatured by boiling in SDS using an antibody that recognizes only cleaved caspase-6 (). Under denaturing conditions, the cleaved caspase-6 antibody precipitated negligible amounts of the full-length forms, suggesting that proforms of caspase-6 can be labeled by active site probes and that at least some fraction of these proforms can be isolated in complex with cleaved forms of caspase-6. To further confirm that the single chain (uncleaved) forms of caspase-6 possess catalytic activity, we pretreated lysates with two different inhibitors, AB46 and LE33, a version of LE22 in which the Cy-5 tag was replaced with biotin (). These results confirmed that active site directed inhibitors could block the labeling of the higher molecular weight proforms of caspase-6 by LE22. AB13, the caspase-3/-7 selective inhibitor, however, was unable to block labeling of those same pro-forms.
A conformational change is permissive for activation in the absence of cleavage
To gain more insight into the activation mechanism of caspase-6, we generated several cleavage site mutants in which the endogenous aspartic acid (`D') residues at three cleavage sites in the linker region were converted to alanine (`A') (Supplemental Figure 3
). Compared to wildtype, the reduction in activity of the D23A or D179A single mutants was minimal, and even the D23A/D179A double mutant showed only a two-fold decrease in Kcat
(Supplemental Table 1). The most dramatic reduction in activity, however, was observed for the D193A (~20-fold reduction in Kcat
). D23A or D179A mutations further reduced the activity of D193A. This confirms the earlier finding that caspase-6 can auto-process at D193, whereas D179 requires an additional enzyme for cleavage (Wang et al., 2010
). Cleaving the D193A mutants with 1% caspase-3 overnight increased the activity by about 40 fold. However, activity failed to reach levels expected for fully processed caspase-6, indicating that processing of the linker at D193 is absolutely essential to generate a fully active caspase-6 species. Interestingly, the amount of active enzyme in the preparations of non-cleaved caspase-6 as determined by titration with Z-VAD-FMK was only about 10–15% of what was expected based on absorption at 280 nm (A280
), while incubation with caspase-3 overnight led to activation of ~100% of the caspase-6 species in the preparations (not shown). Thus, in the total pool of non-cleaved caspase-6, only a fraction has activity, while the entire pool has the capacity to generate activity.
Since the non-cleavable caspase-6 mutants retained activity, albeit very little, we decided to utilize LE22 to visualize the active species of recombinant caspase-6 (). The full-length caspase-6 species containing D193A readily bound the probe (lanes 5, 7, and 8), albeit to a lesser degree than those species processed at D193 (lanes 1, 3, 4, and 6). Addition of 1% active caspase-3 to the D193A or D23A/D193A mutants dramatically increased probe labeling (lanes 9 and 10). The active site mutants of both caspase-6 and caspase-3 (lanes 2 and 12, respectively) were not labeled by the probe at all, indicating that probe binding is specific and depends on activity. In addition, several higher molecular weight species were also labeled by the probe. These are not likely to be contaminants carried over from E. coli, since they are not detected in the Cys mutant species and could represent aggregates of active caspase-6.
Caspase-6 undergoes a conformational change upon activation
Executioner caspases have been suggested to fluctuate between an `active' and an `inactive' conformation, with processing of the inter subunit linker stabilizing the `active' conformation (Fuentes-Prior and Salvesen, 2004
; Gray et al., 2010
). In theory, an inhibitor or active site probe such as LE22 could stabilize this active conformation. If this were the case, we would expect to see a steady increase in labeling over time when the uncleavable caspase-6 is labeled with the probe. However, we found that although the uncleaved species labels more slowly than the wild type (as expected from its decreased kcat
), maximum labeling of all species was achieved after ~90 minutes of incubation with an excess of probe (Supplemental Figure 4A
). This labeling was significantly lower (~7 fold) than the maximum labeling of WT caspase-6, even though equal amounts of total protein were used in the experiment (Supplemental Figure 4B
). This suggests that only a fraction of the uncleaved caspase-6 species is in an active conformation and, although this fraction labels more slowly than processed caspase-6, it does not increase over time.
The executioner caspases-3 and -7 have been described to form obligate dimers (Boatright et al., 2003
) and dimerization is an absolute requirement for caspase activation. We wondered whether the uncleavable caspase-6 mutants were primarily expressed as inactive monomers, while only the fraction that was labeled by LE22 formed active dimers. To investigate this hypothesis, we analyzed all caspase-6 species on a native pore limit gel after labeling with LE22 (). On such a gel, a native protein sample is resolved on a 4–20% polyacrylamide gradient. Migration of a protein or protein complex is limited by the pore size of the polyacrylamide, thus a protein complex will reach equilibrium higher up in the gel (at a larger pore-size) than the individual proteins would (Barrett et al., 1979
; Boatright et al., 2003
). To our surprise, the inactive, unlabeled species ran higher in the gel than the labeled species, which ran at approximately the same molecular weight as the caspase-3 dimer irrespective of linker or pro-domain processing. Probe labeled species resolved lower on the gel, suggesting that either native caspase-6 forms a higher oligomer or that inactive, single chain, caspase-6 is partially unfolded and therefore migrates slower through the gel. In support of the latter hypothesis, the crystal structure of ligand-free caspase-6 reveals a misalignment of active site residues, which is not observed for caspases-3 or -7 (Baumgartner et al., 2009
). When we compared the different species of single chain caspase-6 with WT on native gel either with or without LE22 labeling, we observed a significant shift upon probe binding (Supplemental Figure 4D
), suggesting that caspase-6 folds around its ligand, as has been suggested (Vaidya et al., 2011
To study the behavior of natural (endogenous) caspase-6 upon activation, we activated caspase-9 in cytosolic extracts from HEK293T cells by adding cytochrome c
and dATP, a commonly used model for caspase activation (; Stennicke et al., 1999
). As indicated by probe-labeling, both caspase-3 and caspase-6 are activated within 30 minutes and are fully processed within 60 minutes. At 30 minutes, a partially processed species of caspase-3 can also be observed that has activity, as indicated by probe binding, and is detected by the active caspase-3 antibody. We then probed the blots successively with antibodies against full length and cleaved caspase-3 and -6 respectively to distinguish between specific and non-specific antibody recognition. Although both caspase-6 and -3 are cleaved during activation, only caspase-6 undergoes a dramatic change in conformation as observed on the native gel (lower panels). Caspase-3 migrates at the same MW, whether or not it is cleaved/active, whereas active caspase-6 shifts to a lower MW relative to inactive caspase-6 upon activation. Furthermore, labeling of uncleaved caspase-6 can also be observed, both on the denatured and the native gel.
Altogether, our data suggest that a conformational change is permissive for initial caspase-6 activation, while cleavage, in particular auto-processing of D193, further enhances activity and stabilizes the active conformation.
Kinetic studies of caspase activation using activity based probes
We next used LE22 to examine the dynamics of executioner caspase activation in intact cells after induction of a death stimulus. For these studies, we treated COLO205 cells with anti-DR5 antibody over an eight hour time period and labeled with LE22 or AB50 in the final 30 minutes of the experiment (). We found that the p19 form of caspase-3 is active at early time points and matures over time, leading to increased activity of the p17 form. Interestingly, Caspase-6 activates more slowly than caspase-3, and shows the most activity at the late time points. In intact cells, the proforms of caspase-6 gradually show increased activity over time. In parallel, we also conducted a similar time-course in which lysates were used in place of intact cells (). In these samples we observed a sharp increase in activity of the proforms of caspase-6 immediately upon anti-DR5 stimulation. This activity remained relatively constant over time (). To obtain a clearer picture of caspase-6 activity we also used AB13 to block labeling of caspase-3 and -7 in the lysates before labeling with LE22 (). Using this method we were able to clearly identify all of the 6 major predicted forms of caspase-6 (see ). Additionally, we performed analogous studies in Jurkat cells using a variety of intrinsic and extrinsic death stimuli and observed similar trends in caspase-6 activation (Supplemental Figure 5
). The kinetics of activation were slower with anti-DR5, anti-Fas, or Staurosporine than with Etoposide. However, the pro-forms of caspase-6 increase in activity under all stimuli.
Monitoring maturation of executioner caspases during apoptosis
Finally, we performed a kinetic study in which AB13 was added at the time of anti-DR5 treatment to block caspase-3/-7 activity throughout the experiment (). If these executioner caspases were essential for caspase-6 activation in this model, we would expect to see dramatic changes in the kinetics of its maturation. However, the caspase-6 labeling pattern remained relatively unchanged. While auto-processing occurs at D193, full maturation to the p18 form of caspase-6 would not be expected to occur without cleavage at D179. This suggests that either caspase-6 is capable of autoactivation, or that other caspases are able to cleave at D179.