|Home | About | Journals | Submit | Contact Us | Français|
Caspases, the cysteine proteases that execute apoptosis, are tightly regulated via phosphorylation by a series of kinases. Although all apoptotic caspases work in concert to promote apoptosis, different kinases regulate individual caspases. Several sites of caspase-7 phosphorylation have been reported, but without knowing the molecular details, it has been impossible to exploit or control these complex interactions, which normally prevent unwanted proliferation. During dysregulation, PAK2 kinase plays an alternative anti-apoptotic role, phosphorylating caspase-7 and promoting unfettered cell growth and chemotherapeutic resistance. PAK2 phosphorylates caspase-7 at two sites, inhibiting activity using two different molecular mechanisms, before and during apoptosis. Phosphorylation of caspase-7 S30 allosterically obstructs its interaction with caspase-9, preventing intersubunit linker processing, slowing or preventing caspase-7 activation. S239 phosphorylation renders active caspase-7 incapable of binding substrate, blocking later events in apoptosis. Each of these mechanisms is novel, representing new opportunities for synergistic control of caspases and their counterpart kinases.
Tens of billions of cells die each day by the controlled cell death pathways of apoptosis. Apoptotic programmed cell death is fundamental for all multicellular eukaryotes and is critical for organismal development, differentiation, eliminating damaged cells, and maintaining homeostasis. Dysregulation of apoptosis has been linked to diseases including cancer (Greenblatt et al., 1994; Pistritto et al., 2016; Plati et al., 2010; Vaux et al., 1988), autoimmune disorders (Worth et al., 2006), and neurodegeneration (Ghavami et al., 2014; Margolis et al., 1994). In particular, transformation to a cancerous phenotype occurs when cells establish barriers to apoptosis and develop resistance to signals designed to eliminate malignant abnormalities via apoptosis.
The apoptotic cascade is ultimately dependent on the activation of the caspase family of cysteine proteases. These enzymes dismantle the cell by cleaving a diverse set of protein substrates (Dix et al., 2008) with a specificity for acidic residues, typically aspartate, at P1 (Stennicke et al., 2000) although glutamate (Moretti et al., 2002) and phosphoserine can also be recognized (Seaman et al., 2016). Apoptotic caspases are divided into two classes: the initiator caspases (caspase-8, -9, and -10) and their substrates, the executioner caspases (caspase-3, -6, and -7), which cleave specific cellular targets to invoke cell death. Caspases are functional as homodimers, with each monomer made up of an N-terminal prodomain, one large, and one small subunit (Figure 1A). As a means of inherent regulation, executioner caspases are translated as inactive procaspase zymogens, with distinct cleavage events required for zymogen activation. Initiator caspases propagate the apoptotic signal by cleaving downstream executioner caspases at their intersubunit linker. Prior to cleavage, this linker rests across the dimer interface (Riedl et al., 2001). Cleavage of the intersubunit linker generates loops 2 and 2’ (L2 and L2’), which, in the active, substrate-bound conformation, lock together the substrate-binding groove on the opposite monomer (Figure 1B) (Witkowski and Hardy, 2009). This results in a dramatic conformational change in the dynamic active-site loop bundle from a zymogen state to an active state. Cleaved executioners can also experience a reversal to the zymogen-like state by small molecules (Hardy et al., 2004), which forces a loop rearrangement that traps L2’ over the dimer interface and results in the expulsion of loop 3 from the active site pocket. Following activation in the cell, each executioner caspase cleaves over a hundred substrates (Agard et al., 2012; Hill et al., 2016) to propagate the well-ordered termination of the dying cell.
Due to their cell death inducing potential, caspase activation and downstream activity is tightly regulated on multiple levels, with phosphorylation as one of the primary means to manage apoptotic caspase activity (Allan and Clarke, 2007; Alvarado-Kristensson et al., 2004; Andersen et al., 2009; Cardone et al., 1998; Li et al., 2011; McDonnell et al., 2008b; Suzuki et al., 2004). In fact, caspases and kinases co-regulate each other, resulting in an intricate interplay of dramatic post-translational modifications that ultimately impact cell death and survival (Dix et al., 2012; Kurokawa and Kornbluth, 2009; Zukowski and Litchfield, 2015). Together apoptotic caspases are recognized to initiate and execute cell death, but their individual roles continue to come to light. Even highly related caspases play non-overlapping biological roles and are independently regulated by inhibitor proteins, zinc, and posttranslational modifications (Dagbay et al., 2014). Interactions with kinases are a major contribution to the unique regulation of individual caspases. The delicate nature of this balance between caspases and kinases has been exploited by a number of cancers (Allan et al., 2003; Cursi et al., 2006), which have developed resistance to apoptosis, even after stimulation by chemotherapeutic agents (Li et al., 2011; Marlin et al., 2009).
Many cancers have capitalized on the vast signaling capacity of various kinases to manipulate cell processes (Dhillon et al., 2007). One particular example is the dysregulation of p21 activated kinase 2 (PAK2, γ-PAK) (Dummler et al., 2009; Mira et al., 2000). PAK2, the only ubiquitously expressed PAK, plays a role in a variety of biological pathways including cell motility, mitosis, survival, and apoptosis. Interestingly, PAK2 plays a dual role in the context of apoptosis: active full-length PAK2 stimulates cell survival (Jakobi et al., 2001; Marlin et al., 2009) while cleavage of the autoinhibitory domain results in propagation of the pro-apoptotic response (Lee et al., 1997; Rudel and Bokoch, 1997). This duality is dictated by the relationship of PAK2 with caspase cleavage (Marlin et al., 2011). PAK2 is cleaved by the executioner caspase-3 at D212, which separates the PAK2 kinase domain and autoinhibitory domains (Walter et al., 1998). The phosphorylated kinase domain, PAK2p34, is then translocated to the nucleus (Jakobi et al., 2003) where it phosphorylates a new set of substrates, which in turn promote programmed cell death. A number of breast cancer cell lines have shown hyperactivity of PAK2 (Mira et al., 2000), which tips the scale to favor full-length PAK2 and results in reduced levels of apoptosis. Full-length, active PAK2 stimulates cell survival through multiple mechanisms, including caspase phosphorylation (Li et al., 2011). This overactivity has further been linked to resistance to chemotherapeutic agents.
The complexity of the caspase-kinase crosstalk has been intensified with the discovery that a pathway involving PAK2 is able to phosphorylate the executioner caspase-7 at three distinct residues: S30, T173, and S239 (Li et al., 2011). Although PAK2 and caspase-7 co-localize and co-immunoprecipitate, whether this phosphorylation of caspase-7 is mediated by PAK2 directly or by another kinase in that pathway was not fully elucidated by prior work. Phosphorylation of all three residues S30, T173, and S239 is reported to silence caspase-7 enzymatic activity, but the contribution from each individual site has not been studied nor has the mechanism of inhibition been determined. As a result of a total loss in caspase-7 activity, phosphorylation by PAK2 leads to a reduction in apoptosis. In addition, knockdowns of PAK2 in various breast cancer cell lines lead to increased apoptosis levels when stimulated with chemotherapeutics including staurosporine or doxorubicin (Li et al., 2011) underscoring that PAK2 is critical in curtailing caspase-7 activity.
Here, we elucidate the molecular mechanism by which PAK2 phosphorylation inhibits caspase-7 function. Understanding the regulation of this executioner caspase has implications in cancer resistance and apoptotic activation. Our results show that PAK2 inhibits caspase-7 by two divergent mechanisms prior to and following caspase activation: initial phosphorylation allosterically slows activation by upstream initiator caspases by impeding cleavage at the intersubunit linker and a second phosphorylation site directly blocks substrate binding. Identifying and resolving the molecular mechanism behind both regulatory phosphorylation events provides valuable insight into therapeutically relevant means to control programmed cell death.
Phosphorylation of caspase-7 by a PAK2-dependent pathway inactivates caspase-7 and this appears to attenuate apoptosis in several breast cancers after stimulation by chemotherapeutic agents (Li et al., 2011). Caspase-7 phosphorylation by PAK2 occurs at three sites: S30, T173, and S239 (Figure 1A), but the impact of phosphorylation at these individual sites on caspase-7 function or their mechanisms of regulating caspase-7 have not been investigated. Prior to activation, caspase-7 exists as a zymogen. Proteolytic cleavage by upstream initiator caspases at the prodomain (D23) and the intersubunit linker (D198/D206) generates the active form of caspase-7 comprising large and small subunits of caspase-7 (Figure 1B). In the small subunit, S239 sits below the substrate-binding groove on loop 3, which is part of the active-site loop bundle. T173 is located at the bottom of the 160s helix, whereas S30, in the highly mobile N-terminal region, has not been resolved by any caspase-7 crystal structures.
To explore the impacts of phosphorylation, each of the three reported phosphorylation sites were replaced by glutamate to generate phosphomimetic variants, mirroring the negative charge and steric bulk of phospho-serine or -threonine. To ensure that activity was monitored in a cleaved (active) form, each phosphomimetic was expressed from a constitutively two-chain construct in which the large and small subunits are independently expressed (Witkowski and Hardy, 2011). Neither S30E nor T173E had a significant effect on caspase-7 kinetics (Figure 1C). However, S239E had a dramatic effect on caspase-7 activity, dropping the catalytic efficiency by nearly three orders of magnitude as compared to the wild-type enzyme.
The introduction of negative charges in combination with steric bulk are two attributes that enable phosphorylation to have such significant effects on protein function. We next aimed to determine whether size, charge, or both, were responsible for the S239E loss of activity. Substituting S239 with Gln, retains the size of the Glu but removes the contribution of a negative charge. Conversely, the S239D variant retains the negative charge but removes the bulk. Both S239Q and S239D resulted in a significant drop in catalytic efficiency, but neither as dramatic as S239E, suggesting that both size and charge play a large role in inhibiting caspase-7 activity at S239. Notably, all three mutations (S239E, S239D, and S239Q) had a significant effect on KM, highly suggesting that phosphorylation at S239 interferes directly in a steric manner with substrate binding. In addition, there was an effect on kcat for the S239E mutation, suggesting a disruption to the catalytic machinery. However, this kcat effect was much less pronounced with the Gln and Asp mutations, further suggesting that both size and charge together are responsible for inhibition.
To investigate PAK2 phosphorylation of the caspase-7 small subunit, which contains S239, we performed an in vitro phosphorylation assay. PAK2 T402E, a constitutively active, auto-activating variant was pre-activated with ATP then incubated with [γ-32P] ATP and wild-type caspase-7 or the S239A variant, which cannot be phosphorylated at residue 239 (Figure 2A). Based on the low activity of S239E, we anticipated phosphorylation on the small subunit. We observed active caspase-7 completely cleaved PAK2 at D212 (Figure S1) to generate the kinase and autoinhibitory domains. In addition, in both the wild-type caspase-7 and S239A reactions, a phosphorylated band similar in molecular weight to the small subunit of caspase-7 appeared. To determine whether this 14 kDa band was a cleavage product of PAK2 or the small subunit of caspase-7, we repeated the in vitro phosphorylation assay on catalytically inactive caspase-7 (C186A, Figure 2B), which was expressed using the constitutively two chain construct. PAK2 cannot be cleaved by inactive caspase-7 C186A. The lack of the 14 kDa band with inactive caspase-7 confirms that the band is indeed a cleavage product of PAK2. Contrary to our expectations, we did not observe small subunit phosphorylation under these conditions.
These results demonstrate the difficulty in dissecting the competitive interplay between caspase-7 and PAK2. If active caspase-7 predominates, it cleaves PAK2 before PAK2 can phosphorylate the caspase-7 small subunit, which inactivates caspase-7. We sought a reagent that could block caspase-7 activity thereby preventing cleavage of PAK2. Due to the location of S239, it was important that the reagent not fill the substrate-binding groove and thereby limit accessibility of S239. To meet these criteria we developed a cysteine protecting group (PG, L-cysteinyl-2-pyridyl disulfide, Figure S2A). PG (Figure S2B) modifies and inactivates the catalytic cysteine via reversible thiol chemistry (Ventura et al., 2015), but because of its small size, does not fill the full substrate-binding groove. In the presence of PG, PAK2 clearly phosphorylated the small subunit of wild-type caspase-7 at S239, since no phosphorylation of S239A was observed (Figure 2C). In addition, the PAK2 remains uncleaved and the large subunit of caspase-7 is phosphorylated at the same levels as with unprotected caspase-7 demonstrating that S239 can indeed be phosphorylated by PAK2.
In addition to silencing the caspase-7 activity, the PG plays an orthogonal role at a second site, altering the equilibrium ensemble of the active-site loops by binding to a known allosterically acting cysteine (C290). Upon activation by cleavage at the intersubunit linker, caspase-7 experiences a rearrangement of its loops, which sample multiple conformations. For example, caspase-7 bound to the small molecule DICA (at C290) forces loop 3 to adopt a more accessible conformation, with an increased exposure to solvent. We realized that binding of the PG at C290 should increase the accessibility of this region by forcing loop 3 into a more accessible conformation. To test this hypothesis, we mutated C290 to Ser, blocking the ability of the PG to bind at the allosteric Cys. We observed a decrease in phosphorylation on the small subunit of caspase-7 C290S compared to wild-type (Figure 2D). We did not observe a similar decrease when C186 was mutated to Ala (Figure S3) suggesting that PG binding to C290, but not C186, was responsible for loop movement and availability of S239 to phosphorylation by PAK2.
Thus, it appears that PG binding to C290 acts analogously to binding of other small molecules at C290, resulting in a structural change to loop 3 (Figure 2E). In unliganded caspase-7 (Figure 2E, blue) loop 3 is structured and protrudes inwards toward C290. When a small molecule, like DICA or PG binds at C290, loop 3 is forced upward in a loop accessible conformation that may resemble the zymogen full-length procaspase-7 (Figure 2E, green) or caspase-7 bound to DICA (Figure 2E, tan). This conformational shift increases the solvent accessibility of loop 3, including S239. The PG thus served two purposes (1) to silence caspase-7 activity allowing PAK2 to remain uncleaved so that phosphorylation could proceed and (2) to alter the equilibrium of the caspase-7 active-site loop bundle to favor the solvent accessible conformation of loop 3, which includes the phosphorylation site S239. It is clear that phosphorylation of S239 occurs exclusively when loop 3 is in an exposed, loop-accessible conformation. This suggests either that S239 is only available for phosphorylation under certain conditions, or that PAK2 can distinguish these two distinct conformations of caspase-7.
To determine the molecular mechanism of inhibition by phosphorylation at S239 we solved the crystal structure of the unliganded (apo) cleaved form of the caspase-7 S239E phosphomimetic at 2.2 Å resolution (Table S1). This caspase-7 phosphomimetic shares the same overall fold with all previously crystallized caspase-7 structures, including those unbound and bound to substrate. The heterotetrameric core consists of 12 β strands flanked by 10 α helices, and two symmetrical active sites made up a dynamic active-site loop bundle. Loop 3 sits just below the active site and contains R233, a critical residue for binding substrate (Figure 3A). In both the unliganded mature structure (3IBF) and the substrate-bound structure (1F1J), this arginine is positioned identically (Figure 3B). R233 acts as an anchor poised to make four pivotal hydrogen bonds with the preferred DEVD tetrapeptide substrate. R233 interacts with both the Glu in the P3 position of the peptide substrate as well as the essential Asp side chain in the P1 position. Meanwhile, two additional hydrogen bonds are made with the carbonyl and nitrogen stemming from the R233 backbone. This behavior of R233 is in contrast to its behavior in the S239E structure.
The structure of unliganded caspase-7 S239E suggests a two-factor mechanism by which phosphorylation of S239 inhibits caspase-7. First, the position of loop 3 has been perturbed by the phosphomimetic (Figure 3C). The bulk from the introduced Glu side-chain at position 239, which mimics the phosphoserine in length and negative charge, forces the entire loop 3 to shift outwards toward a solvent channel. This reorients R233 away from the active site pocket so it can no longer make critical bonds with the substrate (Figure 3C). In addition, the phosphomimetic introduces a negative charge in the precise region where an aspartate would bind in the S1 pocket. This inhibitory mechanism is consistent with the kinetic data (Figure 1C), which suggested both size (S239Q) and charge (S239D) worked in combination to inhibit caspase-7 activity after modification at S239. Thus, phosphorylation at S239 is effective at silencing activity of an already active caspase-7 once apoptosis has been initiated.
Phosphorylation of caspase-7 by PAK2 was also observed on the large subunit (Figure 2). This phosphorylation event was so fast that it occurred before the zero timepoint sample could be processed. In addition, the large subunit was phosphorylated regardless of the cleavage state of PAK2 (Figure 2A-C). Despite the overwhelmingly fast phosphorylation by PAK2 in the large subunit, no effect on the catalytic rate of cleaved caspase-7 was observed (Figure S4).
To determine if this phosphorylation occurred at the reported sites on the large subunit, S30 and T173 were individually replaced by alanine, a non-phosphorylatable residue. These substitutions were made in the background of caspase-7 catalytic site inactivation (C186A), which was used to prevent PAK2 from being cleaved. PAK2 did not phosphorylate S30A/C186A but did phosphorylate C186A and T173A/C186A. This was true for both full-length and cleaved caspase-7, indicating that S30 is the primary phosphorylation site on the caspase-7 large subunit (Figure 4A). The ratios of the band intensities for the 32P:commassie stained bands were similar for FL C186A (2.14) and cleaved C186A (2.17) as well as for FL T173A/C186A (2.17) and cleaved T173A/C186A (2.11) suggesting that phosphorylation at S30A occurs similarly on the zymogen and the cleaved forms of caspase-7.
Given that S239 phosphorylation directly impacts catalytic activity, it was surprising that the predominant site of phosphorylation, S30, had no direct effect on caspase-7 catalytic function. This suggested that S30 phosphorylation might play an orthogonal functional role. Executioner caspases including caspase-7 are dependent on activation by upstream initiator caspases including caspase-8 and -9, which cleave the intersubunit linker of full-length procaspase-7 at D198 or D206, generating active caspase-7. To determine if phosphorylation of the full-length caspase-7 impacted processing by initiator caspases, full-length caspase-7 C186A was phosphorylated by autoactivated PAK2 in the presence of [γ-32P] ATP or not phosphorylated in the no ATP condition. Phosphorylation was confirmed by autoradiography. Phosphorylated or unphosphorylated caspase-7 was then subject to cleavage by either caspase-9 or caspase-8. Unphosphorylated caspase-7 was cleaved by caspase-9 completely at the intersubunit linker after 3 hrs (Figure 4B). In contrast, phosphorylated caspase-7 was cleaved more slowly by caspase-9. Remaining full-length caspase-7 was quantified at each time point. We observed a significant decrease in the rate of cleavage when caspase-7 was phosphorylated (Figure 4C). Phosphorylated caspase-7 was also cleaved more slowly by caspase-8 (Figure 4D-E). For both initiator caspases, 30% less full-length caspase-7 was cleaved in the timecourse of the experiment. Notably, PAK2 has no effect on the function of either caspase-8 or caspase-9 under these assay conditions (Figure S5A-B) and does not phosphorylate caspase-9 at all (Figure S5C), indicating that the decrease in processing is the result of phosphorylation of caspase-7.
By substituting alanine, a non-phosphorylatable residue, at each of the reported PAK2 sites in the caspase-7 large subunit (T173 and S30), we interrogated the contribution of phosphorylation at these sites to cleavage by caspase-9 (Figure 4F). Caspase-7 T173A/C186A was incubated with PAK2 in the presence or absence of [γ-32P] ATP and subsequently treated with active caspase-9 (Figure 4F-G). The cleavage pattern was nearly identical to caspase-7 C186A (Figure 4B), suggesting phosphorylation at T173 plays no role in slowing cleavage by caspase-9 at the intersubunit linker. In contrast, the cleavage pattern for full-length caspase-7 S30A/C186A was identical in the presence or absence of phosphorylation (Figure 4F,H). This data strongly suggests that phosphorylation by PAK2 at S30 has considerable impact on the ability of caspase-9 to cleave caspase-7 at the intersubunit linker (residues 198-206).
S30 and the cleavage sites in the intersubunit linker (D198/D206) are distal in sequence, and due to the disorder in all structures of the caspase-7 prodomain, it is difficult to draw specific conclusions about the spatial proximity of these two regions. Nevertheless, phosphorylation at caspase-7 S30 clearly impacts the ability of caspase-9 to cleave caspase-7 at the intersubunit linker. Fluorescence polarization of FITC-labeled caspase-7 in the presence of a caspase-9, which forms constitutive dimer (Chao et al., 2005), showed a strong (1.8 μM) interaction (Figure 5A, S6A-C). In these experiments catalytically inactive caspase-7 and -9 were used to prevent cleavage if either caspase during the reaction. When FITC-caspase-7 C186A was incubated with autoactivated PAK2 to allow phosphorylation prior to the caspase-9 titration, the KD weakened substantially, to greater than 27 μM (Figure 5B). This suggests that phosphorylation at S30 blocks the caspase-7:caspase-9 interaction. Consistent with that interpretation, repeating the experiment with caspase-7 S30A/C186A, which is non-phosphorylatable at position 30, shows a similarly strong interaction (KD = 1.7 μM, Figure 5C) even in the presence of active PAK2. These data confirm the observation (Figure 4C and 4H) that phosphorylation at S30 has a profound effect on how the initiator caspase-9 interacts with the executioner caspase-7 at the initiation stage of apoptosis and may suggest that the N-terminal region of caspase-7 is critical for direct binding to caspase-9.
It is clear that in vitro, phosphorylation at S30 leads to slower cleavage of caspase-7 by caspase-9 at the intersubunit linker. To observe the impact of S30 phosphorylation in a biological context we used MCF7 cells, a widely used breast cancer model. MCF7 cells are intrinsically deficient for caspase-3, putting the burden of executioner apoptotic activity on caspase-7. In addition, MCF7 cells overexpress PAK2, which phosphorylates caspase-7 intracellularly. MCF7 cells were transiently transfected with either phosphorylatable or non-phosphorylatable (S30A) caspase-7. Both transfected variants carried the catalytically inactivating C186S substitution to prevent caspase-7 self-processing. As predicted, following transfection of MCF7 cells, non-phosphorylatable caspase-7 (S30A/C186S) was processed by added caspase-9 significantly faster than was phosphorylatable caspase-7 (C186S) (Figure 6 A-B).
After observing the inhibitory effects of phosphorylation of caspase-7 at S30 in the MCF7 cellular milieu, we hypothesized that we would observe a similar effect in growing cells. MCF7 cells transiently transfected with caspase-7 C186S or S30A/C186S were treated with both ABT-263, a Bcl-2 inhibitor, and A1210477, an Mcl2 antagonist, which together stimulate the intrinsic apoptotic pathway and activate caspase-9 (Leverson et al., 2015). Monitoring the appearance of the cleaved and activated caspase-7 (Figure 6C-D) revealed that caspase-7 S30A/C186S was processed from the full-length form to the cleaved form faster than the C186S variant. This suggests that, as was the case in vitro, when S30 is available for phosphorylation by PAK2, this site becomes phosphorylated, decreasing binding, recognition, and cleavage by caspase-9. Thus it becomes clear that PAK2 regulates caspase-7 activity by inhibiting active (cleaved) caspase-7 through phosphorylation at S239, as well as preventing or dramatically slowing caspase-7 activation by caspase-9 due to phosphorylation at S30.
The regulation of caspases is paramount for determining whether cells live or die. Global regulation of apoptotic caspases has proven to be heavily modulated by phosphorylation, with extensive phosphorylation on both initiator and executioner caspases typically leading to inactivation (for review see Kurokawa & Kornbluth 2009; López-Otín & Hunter 2010). Prior to this work, the molecular details of these inactivating events have been sparse. Emerging regulatory themes have demonstrated an intricate complexity, with caspases as kinase substrates and kinases as caspase substrates. Biases in this interplay that favor cell survival influence diseases of proliferation, like cancer. In the caspase-7:kinase co-regulation, PAK2 hyperactivity is observed to contribute to a cancerous phenotype, with enhanced cell survival and chemotherapeutic resistance (Li et al., 2011).
Our results mechanistically highlight the functional relevance of PAK2 directly phosphorylating the apoptotic target caspase-7 and demonstrate how PAK2 impacts caspase-7 activity independently prior to (S30 phosphorylation) or after (S239 phosphorylation) caspase-7 activation in apoptosis. The first mechanism slows initial activation by allosterically disrupting the ability of initiator caspases to activate full-length procaspase-7 prior to apoptosis, while the second mechanism involves direct blocking of substrate from binding to active caspase-7 during apoptosis (Figure 7).
Phosphorylation at S30 interrupts the binding interaction of caspase-7 with the initiator, caspase-9, and thereby slows caspase-7 cleavage at the distal intersubunit linker. Phosphorylation has been observed to directly block cleavage at the intersubunit linker of several caspases, including caspase-2 (Shin et al., 2005), caspase-3 (Duncan et al., 2011), caspase-8 (Cursi et al., 2006; Matthess et al., 2010), and murine caspase-9 (McDonnell et al., 2008a). A recent study identified a phosphorylation site distal from the intersubunit linker cleavage site that decreases caspase-8 processing through the sequential action of two kinases via an as yet unknown mechanism (Matthess et al., 2014). Phosphorylation of S30 of caspase-7 is the first report of direct allosteric regulation of zymogen activation in caspases, where phosphorylation at a distal site impacted intersubunit linker cleavage. In many previous cases the phosphorylation sites that impact the rate of activation were located either on the linker itself or directly adjacent in the immediate flanking sequence, suggesting that phosphorylation at those linker sites directly prevents binding of the activating caspase by introduction of a charged residue. Inhibition by phosphorylation at S30 is the first example revealing details of a one-step allosteric mechanism where a distant site is utilized to control caspase maturation from the zymogen to the active form. We know of no other proteases for which this mechanism of phosphorylation regulating zymogen activation by a single kinase has been reported, but one might speculate that zymogen activation of other caspases or other proteases may likewise be allosterically modulated by relevant kinases. This identification of a vulnerable allosteric site on caspase-7 opens the door for future modulation of caspase-7 activation by both caspase and kinase modulators.
S30 sits in the N-terminal region of caspase-7. Our data suggest that this region may be critical for directly mediating the interaction of caspase-7 with caspase-9, potentially through an exosite interaction, that leads to zymogen activation. This N-terminal region has never been structurally characterized as it has been disordered in the many structures of caspase-7. Nevertheless, this region has proven to be critical for the participation of caspase-7 in apoptosis using allosteric mechanisms. An exosite at a basic lysine patch (K38KKK) assists in the recognition of specific caspase-7 substrates (Boucher et al., 2012). While little is known about the region, it appears to be critical for caspase-7 specific activities. For example, the K38KKK is not present in caspase-3 and is one example of how caspase-7 is able to fulfill its non-redundant apoptotic roles during the demolition phase. The PAK2 motif containing S30 is not present in caspase-3. We have also observed that PAK2 does not phosphorylate caspase-3 in this N-terminal region in vitro (data not shown), which further suggests that this region of caspase-7 is a region that is critical for regulation of caspase-7 specifically.
While phosphorylation of caspase substrates has been observed to block their cleavage by caspases, our work is the first to show that phosphorylation of an active-site loop in a caspase directly blocks substrate binding, thus inactivating active caspase-7 and later during apoptosis. The crystal structure of caspase-7 phosphomimetic S239E suggests that steric bulk and charge from phosphorylation results in a misalignment of the active-site loop 3. This mechanism is similar to that observed for caspase-6 phosphorylation of S257 by Ark5 kinase, which leads to misalignment of the loops in the substrate binding groove and inactivation of caspase-6 (Velázquez-Delgado and Hardy, 2012). Many lines of evidence have suggested that active-site loop conformations are critical for caspase function. Cleavage of the intersubunit linkers liberates two loops which can reorganize to facilitate substrate binding. Mechanisms like those utilized by PAK2 and Ark5 take advantage of the requirement for precise positioning of the loops to enable substrate binding and cleavage. Small molecules and mutations that organize (Walters et al., 2009) or disorganize (Hardy and Wells, 2009; Hardy et al., 2004) the active site loops have proven to be effective activators or inhibitors. The discovery of the dimer-interface allosteric site in caspases led to speculation that a naturally-occurring regulatory small-molecule might bind and inactivate caspase-7. Our observation that S239 becomes accessible upon binding small molecules at this allosteric site (containing C290) may suggest that such a molecule would be synergistic with PAK2, enabling phosphorylation at S239, thereby inhibiting caspase-7 activity in a covalent, and perhaps more stable, manner. Future caspase therapeutics may fruitfully focus on this adaptable regulatory mechanism exploited by PAK2.
Together, phosphorylation of PAK2 at both S30 and S239 combine to exert powerful control over the activation and subsequent activity of caspase-7. The effects of phosphorylation are maximized by taking advantage of a dual mechanism: an initial phosphorylation at S30 is responsible for slowing the critical activation event of intersubunit linker cleavage followed by a second modification on S239 to disrupt the active site loops and directly block substrate binding (Figure 7). This phenomenon, wherein one kinase engages two different molecular mechanisms that each lead to the same overall cellular outcome, has been observed for other kinase-substrate pairs. Ribosomal protein S6 kinase β-1 (S6K1) phosphorylation of insulin receptor substrate 1 (IRS-1) at S1101 blocks its interaction with PI3K, leading to insulin resistance (Tremblay et al., 2007). When alternative IRS-1 sites, S307/S312 are phosphorylated by S6K1, it enhances proteasomal degradation, also leading to insulin resistance (Greene et al., 2003; Shah and Hunter, 2006). This pattern of regulation can also be observed for transcriptional regulation by the forkhead box proteins (FOXO). FOXO1 is phosphorylated by protein kinase B (PKB/Akt) at three sites, T24, S256, and S319, (Rena et al., 1999) which results in a loss of transcriptional activity. Similar to PAK2 regulation of caspase-7, PKB phosphorylation utilizes different mechanisms of inhibition to achieve the same overall biological goal. Phosphorylation at S256 limits FOXO1 DNA binding (Zhang et al., 2002). Meanwhile, phosphorylation at S319 creates a consensus sequence for CK1 to phosphorylate additional serine residues at positions 322 and 325 (Rena, 2002). This tight grouping of modified serines promotes the nuclear export of FOXO1, decreasing FOXO1 activity by a different mechanism. In these patterns of regulation, one kinase modifies a single substrate at multiple sites, which each exert mechanistically independent but biologically synergistic effects. This synergy in impact is perhaps expected, as uncorrelated outcomes would be non-productive. Thus, it is likely that full-length PAK2 would only play anti-apoptotic roles that have biological synergy to caspase-7 phosphorylation under related conditions.
In addition to S30 and S239, T173 has also been reported to be phosphorylated by PAK2 (Li et al., 2011). We have not observed phosphorylation of T173 under any conditions nor have we observed any functional effect of phosphomimetics at this residue. This could be because i) we have not explored the requisite conditions for phosphorylation, or ii) T173 is not a bonafide modification site but can only be recognized in a peptide, such as those used to identify this site (Li et al., 2011), not in an intact protein, or iii) T173 is a bystander residue, which has no functional effect upon phosphorylation. The concept of bystander residues is replete in the kinase literature. Our work on S30 demonstrates that even for functional phosphorylation events, it can be difficult to decipher the role phosphorylation at a given site might play. It is possible that we may not have found the right conditions for T173 phosphorylation by PAK2, but if T173 is a target of PAK2 under conditions where S30 and S239 are phosphorylated it should likewise have an anti-apoptotic effect. We may likewise predict that if phosphorylation of T173 by PAK2 has a different (e.g. anti-apoptotic) functional effect, we would expect to observe it only under very different cellular conditions. Additionally, it is possible that each of the three sites of phosphorylation are utilized under different biological conditions, providing nuanced control of caspase-7 activity by PAK2.
These findings illuminate the molecular details of a kinase inhibiting an apoptotic caspase. The implications extend to cancer resistance, suggesting a hyperactive PAK2 is able to limit apoptotic cell death and resist apoptotic stimulation by chemotherapeutic agents. In addition, our results pinpoint an allosteric region on a key apoptotic contributor, where alterations result in a slowed maturation and potential control over an executioner caspase. The significance of this regulation could extend beyond cancer resistance and apoptotic activation. Recent discoveries have implicated caspase-7 in inflammation (Erener et al., 2012; Lamkanfi et al., 2008), defense against L. pneumophila infection (Akhter et al., 2009), osteogenesis (Svandova et al., 2014), and tooth development (Matalova et al., 2013), all of which show caspase-7 has distinct roles from caspase-3. These non-apoptotic roles could share regulation by phosphorylation and the mechanistic details uncovered here.
Finally, our increased mechanistic understanding of details of this caspase:kinase interplay may guide precision therapeutic approaches that exploit synergies between caspase and kinase directed modulators. Likewise, identification of the S30 site as critical for interaction with caspase-9 may implicate this S30 region in playing a role in the recognition of other caspase-7 substrates. If this is the case, synthetic inhibitors that block the S30 region, mimicking S30 phosphorylation, may prevent cleavage of the subset of substrates that utilize the S30 region for binding. This is contrast to active-site directed inhibitors, which like S239 phosphorylation, block cleavage of all caspase-7 substrates. S30-mimicking inhibitors that offered precision control of substrate selection would be unprecedented in the realm of synthetic caspase regulation.
Caspase-9 exists primarily as a monomer in solution, as seen by SEC (Huber and Hardy, 2012). A constitutively dimeric caspase-9 has been reported, which decreases the distribution of caspase-9 monomers in solution and increases the more biologically relevant dimer (Chao et al., 2005). In this dimeric caspase-9, 5 residues at the dimer interface of caspase-9 were replaced with the residues from caspase-3. We followed this strategy, and replaced the codons for residues 402-406 to the caspase-3 equivalent C-I-V-S-M on the C-terminus of caspase-9 C287A using Phusion site directed mutagenesis (Thermo Scientific).
Purified recombinant PAK2 T402E was first allowed to auto-activate by incubation with 1 mM ATP in PAK2 activity buffer (50 mM Tris pH 7.5, 20 mM MgCl2, 0.1 mM EDTA, 2 mM DTT) for 1 hour at 30°C. This activated kinase was then inc ubated with caspase-7 variants at a 20:1 caspase:kinase ratio. Additionally, [γ-32P] ATP (10 μCi/μL) was added to radioactively label phosphorylation on the caspase subunits. Reactions were resolved by SDS-PAGE and visualized by autoradiography using a Typhoon FLA 7000 (GE Healthcare). Band quantification was carried out by ImageQuant software (GE Healthcare) using a set of [γ-32P] ATP standards.
PAK2 is also a substrate of caspase-7, and under certain conditions caspase-7 will cleave PAK2. In order to silence this enzymatic activity, but still run the in vitro phosphorylation assay, active forms of caspase-7 could be silenced by incubation with a protecting group (PG, L-cysteinyl-2-pyridyl disulfide) (Ventura et al., 2015) that targets free activated cysteines on the caspase. Prior to the in vitro phosphorylation assay, 20 μM caspase-7 was incubated with 2 mM PG for 1 hour in 20 mM Tris pH 7.5, 100 mM NaCl, and was buffer exchanged to reduce the concentration of reductant (DTT) remaining after purification. This inactive caspase-7 was then used in an identical fashion to the in vitro phosphorylation assay described above.
To monitor the impact of caspase-7 phosphorylation on cleavage of full-length caspase-7 by the initiators caspase-8 or -9, caspase-7 zymogen was first incubated with PAK2. In this experiment all of the caspase-7 variants had the catalytic cysteine mutated to alanine (C186A) in order to ensure that all observed cleavage was from the initiator caspases. Caspase-7 full-length C186A and phosphoknock-outs (S30A, T173A) were first incubated with pre-activated PAK2 (as described above) or with PAK2 in the absence of ATP, in PAK2 activity buffer for 3 hours at 30°C. Additionally, [γ-32P] ATP (10 μCi/μL) was added to radioactively label phosphorylation on the caspase subunits. The equivalent volume of water was added to each reaction in the “no ATP” controls. Treated caspase-7 was then diluted to 3 μM in caspase-9 activity assay buffer (100 mM MES pH 6.5, 20% PEG 400, and 5 mM DTT) or in caspase-8 activity assay buffer (10 mM PIPES pH 7.2, 0.1 M NaCl, 1mM EDTA, 10% sucrose, 0.05% CHAPS, and 5 mM DTT). The corresponding initiator caspase was added at a final concentration of 500 nM. The reaction was quenched at each time point by the addition of SDS sample buffer and boiled for 10 minutes. Cleavage was assessed by SDS PAGE analysis and band quantification was determined using Image Lab software (Bio-Rad).
Fluorescence anisotropy was monitored using a SpectraMax M5 plate reader (Molecular Devices) with a fixed excitation wavelength set to 485 nm and an emission wavelength set to 525 nm. Full-length caspase-7 C186A or full-length caspase-7 S30A/C186A was labeled with fluorescein isothiocyanate isomer 1 (FITC; Sigma) in 0.1 M sodium bicarbonate buffer at pH 9.0. After labeling, unreacted FITC was removed from labeled caspase-7 by buffer exchanging with 3K MWCO filters into 100 mM Tris pH 7.5 and 100 mM NaCl. A fixed concentration of 20 nM labeled protein was subject to a serial dilution of caspase-9 constitutive dimer C287A from 40 μM to 10 nM. All measurements were taken at 25°C aft er a one hour incubation.
MCF cells were first washed twice with ice-cold 1x PBS and were then lysed in 50 mM Tris pH 7.4, 150 mM NaCl, 1% Triton X-100, and protease inhibitors (1 mM 4-(2-aminoethyl)benzenesulfonyl fluoride, 0.8 μM aprotinin, 29 μM bestatin, 10 μM pepstatin, 1.3 mM EDTA) and phosphatase inhibitors (5 mM β-glycerophosphate, 20 mM sodium fluoride and 200 μM sodium orthovanadate) on ice for 30 minutes, then were centrifuged at 20,000 rcf for 20 minutes at 4°C. Quantification of total prot ein levels in the supernatant were determined using the bicinchoninic acid assay (Pierce). An equal amount of total protein was then subject to caspase-9 cleavage by the addition of 0.93 μg/μL of purified recombinant human caspase-9.
MCF7 cells were grown to ~90% confluency and then plated in six well plates with one million cells per well. Cells were transiently transfected with caspase-7 C186S DNA or caspase-7 S30A/C186S DNA using Lipofectamine 3000 (Invitrogen) according to the manufacturer instructions. Cells were then treated with DMSO, staurosporine (STS; 1 μM), or apoptosis inducing compounds (ABT-263 and A1210477; 10 μM) for three hours. After incubation with compounds, cells were trypsinized and washed twice with ice-cold 1x PBS buffer. Cells were then lysed as described previously and lysates were quantified using the bicinchoninic acid assay (Pierce). An equal amount of total protein was loaded into each well of a 12% acrylamide gel and cleaved caspase-7 was detected by western blot analysis. Protocol for Immunoblotting is in the supplementary information.
DNA expression constructs, antibodies, expression, purification, mutagenesis and activity assay protocols for caspase-7, -8, and -9 and PAK2, crystallization, structure determination, immunoblotting, cell culture and transfection can be found in the supplemental experimental procedures.
PAK2 is frequently hyperactive in cancer where it plays an anti-apoptotic role by phosphorylating and inhibiting caspase-7.
This work was supported by the National Institutes of Health (GM80532). We thank V. Stojanoff and E. Lazo for assistance with data collection and acknowledge NSLS X6A, which is funded by the National Institute of Health (GM-0080), and the National Synchrotron Light Source at Brookhaven National Laboratory, which is supported by the U.S. Department of Energy (DE-AC02-98CH10886). We thank S. Thayumanavan for providing support to K.R. for synthesis of the cysteinyl-2-pyridyl disulfide.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
S.J.E. initiated and performed all experimental aspects of the study, prepared all figures and was the principal author for the manuscript. K.R. synthesized the cysteinyl-2-pyridyl disulfide protecting group. J.A.H conceptualized the project, secured funding, directed the research project, wrote parts of the manuscript and edited the manuscript.
The accession number for the coordinates and structure factors reported in this paper is PDB: 5K20.
Supplemental Information includes Supplemental Experimental Procedures and six figures and can be found with this article online at molecular cell.com.