A mutation in the allosteric site of the caspase 3 dimer interface of Val266 to histidine abolishes activity of the enzyme, and models predict that the mutation mimics the action of small molecule allosteric inhibitors by preventing formation of the active site. Mutations were coupled to His266 at two sites in the interface, E124A and Y197C. We present results from X-ray crystallography, enzymatic activity and molecular dynamics simulations for seven proteins, consisting of single, double and triple mutants. The results demonstrate that considering allosteric inhibition of caspase 3 as a shift between discrete ‘off-state’ or ‘on-state’ conformations is insufficient. Although His266 is accommodated in the interface, the structural defects are propagated to the active site through a helix on the protein surface. A more comprehensive view of allosteric regulation of caspase 3 requires the representation of an ensemble of inactive states and shows that subtle structural changes lead to the population of the inactive ensemble.
A common allosteric mechanism for caspase inhibition involves short-range interactions between amino acids in the dimer interface and active site residues . Disrupting the interactions, either through mutagenesis [2,3] or through the binding of small molecules  destabilizes the active conformation and shifts the protein into the inactive or ‘closed’ state. Procaspase 3 is a stable but inactive homodimer, where the ‘monomer’ consists of a large and small subunit covalently connected by a linker of 17 amino acids. Upon cleavage of the IL (intersubunit linker) by initiator caspases, the polypeptide chain is released from its binding in the dimer interface cavity, allowing the substrate-binding loop (called L3) to move into the active site and form the base of the substrate-binding pocket . Loop movement leading to activation of cleaved caspases has been reviewed recently, and the reader is referred to the references therein . As a consequence of IL cleavage, Arg164, which resides adjacent to the catalytic Cys163, moves from a solvent-exposed position in the procaspase to a buried position in the dimer interface of the mature caspase, where it intercalates between Tyr197 and Pro201. The binding of inhibitors in the allosteric site of the interface reverses the conformational change by preventing insertion into the interface of a region of L3 known as the ‘elbow loop’, which contains Pro201 [1,7]. The stacking interactions between Tyr197, Arg164 and Pro201 are thought to stabilize the active conformation.
We showed that the allosteric site in the dimer interface is bifunctional and depends on the context of the protein. The same site that inactivates the mature caspase also can activate the procaspase [2,6]. For example, a mutation of Val266 to glutamate in the dimer interface results in a constitutively active procaspase 3. The procaspase is activated without chain cleavage because the mutation shifts the conformational ensemble to an active state. The constitutively active procaspase 3 rapidly kills cells, and importantly, the protein is inhibited poorly by XIAP (X-linked inhibitor of apoptosis), the endogenous caspase 3 inhibitor . A model of the activated procaspase 3 suggests that the mutation prevents binding of the IL in the dimer interface, essentially destabilizing the inactive conformer and favouring the active conformer . In a broader context, the results imply that introducing the activated procaspase 3 in a targeted manner may be an effective method for killing specific cells. Alternatively, small molecules also activate procaspase 3 allosterically through binding to the dimer interface  or other sites , presumably by a similar mechanism of releasing the IL from the interface.
Collectively, current results suggest that the ensemble of conformations is more complex than a simple two-state model, where the IL of the inactive procaspase 3 is cleaved by initiator caspases to yield the fully active mature caspase. Indeed, it has long been known that, in the absence of substrate, mature caspase 7 exists in the closed conformation. In this case, the cleaved IL remains bound in the dimer interface, and the protein rearranges to the active conformer in the presence of substrate [9,10].
In contrast with the V266E mutant of caspase 3, a mutation of Val266 to histidine was shown to inactivate the protein , and it was suggested that steric clashes in the interface prevent insertion of the elbow loop and Arg164, essentially mimicking the mechanism of the small molecule allosteric inhibitors [2,11]. In order to more fully understand the allosteric site of the caspase 3 dimer interface and the effect that amino acid mutations in the interface have on allosteric regulation, we determined the structure of the V266H variant by X-ray crystallography, and the results pointed to several residues that could potentially be involved in the mechanism of allosteric inhibition. We subsequently mutated those residues in the background of V266H and report the X-ray crystal structures of those proteins as well. Overall, results from crystallography, enzymology and molecular dynamics simulations show that allosteric regulation is more complex than a single short-range interaction pathway involving Arg164. The results demonstrate long-range connectivity between the interface allosteric site and the active site loops. The results also show that the conformational ensemble of inactive versus active states can be influenced by the interactions of several key residues, some of which are buried in the interface while others reside on the protein surface. Allosteric manipulation of the (pro)caspase conformational ensembles may provide new therapeutic interventions for either promoting or preventing apoptosis, but currently the ensembles are poorly understood.
MATERIALS AND METHODS
Cloning, expression and purification
Mutagenesis of caspase 3 to generate the single, double and triple mutants was performed as described in the Supplementary information available online at http://www.bioscirep.org/bsr/032/bsr0320401add.htm. Escherichia coli BL21(DE3) pLysS cells were transformed with each of the plasmids, and all proteins were expressed and purified following previously established protocols [2,12–14]. Caspase 3 (V266H) and caspase 3 (E124A and V266H) are enzymatically inactive (described below), and the protein purified as the procaspase form. The cleaved form was generated by treating with granzyme B (Calbiochem) in a buffer of 20 mM Hepes, 100 mM NaCl, 0.1% CHAPS, 10% sucrose and 10 mM DTT (dithiothreitol) (enzyme assay buffer) at 37°C for 2 h at a final caspase/granzyme B molar ratio of 90:1.
Crystallization and data collection
Proteins were dialysed in a buffer of 10 mM Tris/HCl, pH 8.5, and 1 mM DTT and concentrated to 10 mg/ml. Inhibitor, Ac-DEVD-CMK (acetyl-Asp-Glu-Val-Asp-chloromethyl ketone) reconstituted in DMSO, was then added at a 5:1 (w/w) inhibitor/peptide ratio. The protein was diluted to a concentration of 8 mg/ml by adding 10 mM Tris/HCl, pH 8.5, concentrated DTT, and concentrated NaN3 so that the final buffer consisted of 10 mM Tris/HCl, pH 8.5, 10 mM DTT and 3 mM NaN3. Crystals were obtained at 18°C by the hanging drop vapour diffusion method using 4 μl drops that contained equal volumes of protein and reservoir solutions over a 0.5 ml reservoir. The reservoir solutions for optimal crystal growth consisted of 100 mM sodium citrate, pH 5.0, 3 mM NaN3, 10 mM DTT and 10−16% (w/v) PEG [poly(ethylene glycol)] 6000. Crystals appeared within 3.5–6 weeks for all mutants and were briefly immersed in cryogenic solution containing 10% MPD (2-methylpentane-2,4-diol) and 90% reservoir solution. Data sets were collected at 100 K at the SER-CAT synchrotron beamline (Advance Photon Source, Argonne National Laboratory, Argonne, IL, U.S.A.) The X-rays had a wavelength of 1 Å (where 1 Å=0.1 nm), and 180° of data were collected for each protein at 1° intervals. Those mutants that crystallized with the symmetry of the orthorhombic space group I222 were phased with a previously published caspase 3 structure (PDB entry 2J30), as described . Caspase 3 mutants that crystallized with the symmetry of the monoclinic space group C2 were phased with another previously published structure (PDB entry 1NMS). A summary of the data collection and refinement statistics is shown in Supplementary Table S1 (at http://www.bioscirep.org/bsr/032/bsr0320401add.htm). The atomic co-ordinates and structure factors for caspase 3 (V266H), caspase 3 (Y197C), caspase 3 (E124A), caspase 3 (Y197C,V266H), caspase 3 (E124A,Y197C) caspase 3 (E124A,V266H) and caspase 3 (E124A,Y197C,V266H) have been deposited in the PDB under accession codes 4EHA, 4EHD, 4EHH, 4EHF, 4EHK, 4EHL and 4EHN respectively.
Enzyme activity assay
Initial velocity was measured in enzyme assay buffer at 25°C in the presence of varying concentrations of Ac-DEVD-AFC (acetyl-Asp-Glu-Val-Asp-7-amino-4-trifluoromethylcoumarin) substrate, as described previously [2,14–16]. The total reaction volume was 200 μl and the final enzyme concentration was 10 nM. Briefly, following addition of the substrate, the samples were λexcitation at 400 nm, and λemission was monitored at 505 nm for 60 s. The steady-state parameters, Km and kcat, were determined from plots of initial velocity versus substrate concentration.
Molecular dynamics simulations
Molecular dynamics simulations were performed with GROMACS 4.5 , using the Amber99 force field  and the TIP3P water model . All simulations started with structures obtained from X-ray crystallography as described above. For those proteins that crystallized with symmetry of the space group I222, with one monomer in the asymmetric unit, as was the case for wild-type caspase 3, dimers were generated by applying the appropriate 2-fold crystallographic symmetry. The proteins were solvated in a periodic box of 62 Å × 48 Å × 66 Å, with approximately 13500 water molecules. Sodium or chloride ions were added as required to neutralize the charge on the system. The system was first minimized using steepest descent, and then the waters were relaxed during a 20 ps MD simulation with positional restraints on the protein. Simulations of 50 ns were then run for each protein under constant pressure and temperature (300 K). A time step of 2 fs was used, and co-ordinates were saved every 5 ps. A cut-off radius of 9 Å was used for non-bonded interactions, and electrostatic interactions were calculated using the Particle Mesh Ewald method . In each simulation, the protein was equilibrated within 500 ps. Cavity volumes were estimated by multiplying the number of water molecules within 15 Å of the α carbon atom of residue 266 by the average volume of a single water molecule (~17 Å3), which was consistent throughout each simulation.
X-ray crystal structure of caspase 3 (V266H) reveals long-range interactions
In wild-type caspase 3, van der Waals interactions between Val266 and Tyr197 from each monomer provide close packing in the dimer interface. Our initial model of caspase 3 (V266H)  explained inhibition by a steric hindrance model, assuming that the His266–His266′ residues were in the preferred rotamer for histidine in which the side chains face each other edge-on across the interface (the prime (′) refers to residues in the second monomer). In this model, the ϵ2-nitrogen atom of each histidine ring faces Arg164, but the ring also partially occupies the space for Tyr197. It was thought that steric clashes between the His266 and Tyr197 side chains would prevent proper insertion of Arg164 into the dimer interface, thus destabilizing the active site. A very similar case is seen in caspase 9, where the presence of several bulky residues in the interface prohibits the formation of two productive active sites , so one active site forms, while the other remains disordered. Although the steric-hindrance model was feasible based on our current understanding of the caspase 3 allosteric site, the structure of caspase 3 (V266H) showed a very different story.
Wild-type caspase 3 crystallizes with the symmetry of the orthorhombic space group I222, containing one monomer per asymmetric unit, so the biological dimer of caspase 3 is generated through one of the two-fold crystallographic symmetry axis. Caspase 3 (V266H), however, crystallized with the symmetry of the monoclinic space group C2, which contains two monomers per asymmetric unit. From this information alone one can infer that the monomers are no longer symmetric in the V266H variant, and examination of the structure showed this to be true. The backbone atoms of both monomers superimpose reasonably well with the monomer of wild-type caspase 3 [RMSD (root mean square deviation) ~0.17 Å], but the dimer demonstrates a larger difference (RMSD=0.34 Å) because of differences in active site loop 1 and amino acid side chains in the interface of monomer B, so the description presented here focuses on monomer B. Surprisingly, the variant is able to accommodate the bulky histidine residues due to a 90° rotation about ϕ2 so that the histidine rings face each other across the interface (Figure 1A; see also Supplementary Figure 3B available online at http://www.bioscirep.org/bsr/032/bsr0320401add.htm). One observes several changes in the interface of the mutant when compared with that of wild-type caspase 3. The hydroxy group of Tyr197 is displaced ~0.8 Å away from the interface, and the movement is accompanied by an even larger shift in Tyr195, where the hydroxy group moves ~2.3 Å away from the interface (Figure 1A). In contrast, small changes are observed for the short-range allosteric residues Arg164 and Pro201, although we note that the proline residue contains a different ring pucker compared with that of wild-type. In addition, there are minor changes for Glu124, which resides above the interface and neutralizes the charge of Arg164. Overall, the structure of the V266H variant demonstrated that the inhibition could not be explained by disruption of the short-range allosteric network because the His266–His266′ side chains were accommodated in the interface, albeit with repositioning of neighbouring tyrosine side chains.
Amino acids in the dimer interface of caspase 3
Expanding away from the site of the mutation, one observes the consequences of the movements in the side chains of Tyr197 and Tyr195. In wild-type caspase 3, Tyr195 participates in a H-bonding network with Lys137, Thr140, Glu190 and two water molecules. The hydroxy group of Tyr195 forms through-water H-bonds with Glu190 and Thr140 (Figure 1B, upper panel). The Glu190 residue has been shown previously to be important for stabilizing the active site because it forms backbone H-bonds with Asp169′ across the interface and stabilizes active site loops 2 and 2′ downstream of the catalytic Cys163 in the so-called ‘loop bundle.’ (Figure 1C) [13,22]. In the V266H variant, by contrast, the Tyr195 hydroxy group moves to within 2.9 Å of the side chain of Thr140 and forms a direct H-bond (Figure 1B, lower panel, and Supplementary Figures S1A and S1B at http://www.bioscirep.org/bsr/032/bsr0320401add.htm). Although one water molecule in the Tyr195–Glu190 H-bond network is conserved in the mutant, a second water molecule replaces the hydroxy group from wild-type Tyr195 so that the interaction with Glu190 now occurs through two waters. In addition, a salt bridge observed in wild-type caspase 3 between Lys137 and Glu190 (Figure 1B, upper panel) is disrupted in the V266H mutant (Figure 1B, lower panel, and Supplementary Figures S1A and S1B). Although the side chain of Glu190 appears to be unaffected by the movement in Tyr195 and the loss of the salt bridge with Lys137, the structural data suggest that the conformational changes may be propagated across the dimer interface to affect the opposing active site by destabilizing the L2–L2′ loop bundle. The dynamics of this region are addressed more fully below.
The side chain of Thr140 is positioned near the water-filled interface cavity in the middle of helix 3, one of five surface helices (Figure 1C). In the mutant, the direct H-bond between Tyr195 and Thr140 distorts the N-terminus of helix 3 such that it moves by ~1 Å (Figure 2A), which may be the cause of the repositioning of Lys137. The movements are propagated further through two short β-strands on the protein surface, again by ~1–2 Å, so that the protein chain moves closer to the active site. Ultimately, Phe128, on one of the surface β-strands, clashes with Met61 (on active site L1). The movement results in a different rotamer for Met61 such that the side chain moves towards His121, the second catalytic residue, and affects its H-bonding with Thr62 in active site L1 (Figure 2A and Supplementary Figures S1C and S1D). The S1 binding pocket is narrower in the mutant because of partial occupation by the His121 side chain (Supplementary Figure S2 at http://www.bioscirep.org/bsr/032/bsr0320401add.htm). Overall, the structural data show that the mutation in the dimer interface at His266 is propagated through helix 3 to the surface of the protein in a way that results in a disordered L1 (Figures 2B and 2C) and a narrower S1-binding pocket.
Comparison of changes resulting from the Val266 to histidine mutation
Mutations coupled to V266H restore the enzyme activity
In order to further examine the changes in the dimer interface that result from the V266H mutation, we replaced Tyr197 with cysteine and Glu124 with alanine, both singly and in combination with the V266H mutation (Figure 3A). As noted above, Glu124 resides on a loop above the dimer interface, where the loop connects the two short β-strands that were observed to move in the V266H mutant. The side chain of Glu124 also neutralizes the positive charge of Arg164 (see Figure 1), so, in addition to the stacking interactions with Tyr197 and Pro201, the three residues (Glu124, Tyr197 and Pro201) are thought to stabilize Arg164 in the active conformation.
A double mutation of Y197C,V266H restores activity and structure similar to that of wild-type
Both single mutants, Y197C and E124A, exhibited about 3–4-fold lower activity than wild-type caspase 3 (Table 1 and Figure 3A), which was manifested primarily in lower kcat values and a ΔΔG° of ~0.6–1 kcal/mol. When combined, however, the double mutant, E124A,Y197C, had an ~100-fold lower activity, with an additional ΔΔG°~1.8–2.2 kcal/mol. The results indicated that while either single mutant was reasonably well tolerated, loss of both residues dramatically decreased activity.
|Protein .||Km (μM) .||kcat (s−1) .||kcat/Km (M−1·s−1) .|
|Y197C,V266H||12±4||0.02 ± (2×10−3)||1.67×103|
|E124A,Y197C||30±5||0.008 ± (3×10−4)||2.67×102|
|E124A,Y197C,V266H||22±3||0.02 ± (1×10−3)||9.10×102|
|Protein .||Km (μM) .||kcat (s−1) .||kcat/Km (M−1·s−1) .|
|Y197C,V266H||12±4||0.02 ± (2×10−3)||1.67×103|
|E124A,Y197C||30±5||0.008 ± (3×10−4)||2.67×102|
|E124A,Y197C,V266H||22±3||0.02 ± (1×10−3)||9.10×102|
When placed in context of the His266 variant, E124A had no effect on activity, that is, the E124A and V266H double mutant remained inactive. In contrast, the mutation of Tyr197 to cysteine in the context of V266H resulted in an increase in activity such that the activity of Y197C,V266H was only ~15-fold lower than that of wild-type caspase 3. There was no further increase in activity in the triple mutant compared with Y197C,V266H [~(0.9–1.7)×103 M−1 · s−1], although it is notable that the activity of the triple mutant is greater than that of the E124A,Y197C double mutant by approximately 3-fold.
X-ray crystal structures of variants coupled to V266H
The structures of the proteins shown in Figure 3(A) were determined by X-ray crystallography to resolutions >1.86 Å (Supplementary Table S1). Both single mutants, Y197C and E124A, crystallized in the orthorhombic space group I222, the same as wild-type caspase 3. The monomer of both proteins superimposed well with that of wild-type, with RMSD of 0.122 and 0.127 Å respectively. When the interfaces are compared with that of wild-type, one observes minor changes at the site of the mutation. The interface allosteric site of wild-type caspase 3 contains six conserved water molecules that form H-bonds across the dimer interface between Arg164, Tyr197, Arg164′, and Tyr197′  (Supplementary Figure S3A at http://www.bioscirep.org/bsr/032/bsr0320401add.htm). The six water molecules also are retained in the two mutants (Supplementary Figures S3C and S3D) as is the extensive H-bonding network in the interface. In the E124A variant, however, the side chain of Tyr195 is shifted 1.1 Å away from the interface so that it is 3.6 Å from Thr140 (Supplementary Figure S3D). In a broader view, there are no differences in helix 3, the two surface β-strands, or the active sites when compared with wild-type caspase 3 (Supplementary Figure S5 at http://www.bioscirep.org/bsr/032/bsr0320401add.htm).
The double mutant, E124A,V266H, crystallized in the monoclinic space group C2, the same as the single mutant V266H. Overall, the structure is very similar to caspase 3 (V266H), with movements along the protein surface between the interface and loop 1 (Supplementary Figure S6 at http://www.bioscirep.org/bsr/032/bsr0320401add.htm). The hydroxy group of Tyr195 is in a similar position as it is in the V266H variant, ~3 Å from Thr140, although the side chain of Lys137 is similar to that of wild-type and interacts with Glu190 (Supplementary Figure S6B). Surprisingly, the electron density supports two conformations for His266 in monomer B, where one conformation is similar to that observed in the V266H single mutant (Figure 1A) and the second conformation faces Tyr195 (Supplementary Figures S6A and S6C). In monomer A, the electron density supports only one conformation for His266, similar to that in the V266H variant.
The double mutant, E124A,Y197C, also crystallized in the monoclinic space group C2. As described above, this mutant demonstrated very low enzymatic activity, and the structure showed several features in common with the V266H variant (Supplementary Figures S7A and S7B at http://www.bioscirep.org/bsr/032/bsr0320401add.htm). When compared with wild-type caspase 3, the overall RMSD is 0.23 Å, and the side chain of Tyr195 is positioned ~3.2 Å from that of Thr140. Helix 3 is somewhat distorted, and the surface β-strands are moved towards the interface, with surface side chains nearly identical with those of the V266H variant and loop 1 partially disordered (Supplementary Figure S7C). In contrast with the V266H variant, however, the side chain of Met61 is similar to that of wild-type caspase 3, and the side chain of Lys137 also is positioned similarly to that of wild-type and interacts directly with Glu190. When the interface is compared with those of the single mutants, one observes that the network of water molecules is disrupted, where only the two water molecules that interact with Arg164 are retained (Supplementary Figure S4B at http://www.bioscirep.org/bsr/032/bsr0320401add.htm). Indeed, a general feature of all of the double mutants as well as the triple mutant is that the interfaces have fewer water molecules, and thus the water-mediated interactions among Arg164, Tyr197, Tyr195 and Glu124 are disrupted (Supplementary Figure S4). At present, it is not clear whether the loss of the water molecules affects the position of Tyr195.
When Tyr197 is replaced with cysteine in the background of V266H [caspase3 (Y197C,V266H)] or E124A,V266H [caspase 3 (E124A,Y197C,V266H)], the enzyme activity returns to levels approximately 10–20-fold lower than the single mutant Y197C or wild-type caspase 3 (Table 1). Both mutants crystallized in the orthorhombic space group I222, with one monomer per asymmetric unit. For the double mutant, Y197C,V266H, the histidine side chain rotates about the ϕ2 dihedral angle so that the ϵ2-nitrogen atom of each histidine ring faces Arg164, similar to our initial model. The nitrogen atoms in the imidazole ring and those in Arg164 and Arg164′ are all within 3 Å or less. Although Tyr195 does not occupy the same position as in wild-type caspase 3, it is further from Thr140, at 3.4 Å (Figure 3C), and Lys137 interacts with Glu190 similarly to wild-type caspase 3. Helix 3 is no longer distorted as in the single mutant V266H, and the positions of the short surface β-strands are similar to those of wild-type caspase 3. Similar results are observed for the triple mutant, and for both mutants the electron density shows loop 1 to be well-ordered (Supplementary Figure S8 at http://www.bioscirep.org/bsr/032/bsr0320401add.htm). In the triple mutant, as with monomer B in caspase 3 (E124A,V266H), the electron density supports two conformations for His266. One conformation is similar to that observed in the Y197C,V266H double mutant, where the ϵ2-nitrogen atom of each histidine ring faces Arg164. The second conformation is similar to that observed in the E124A,V266H double mutant where the imidazole ring faces the C-terminus of β-strand 8 near Tyr195.
Overall, the structural data for the double and triple variants show that the defects caused by His266 are corrected when Tyr197 is replaced with cysteine. Interestingly, the electron density supports two conformations for His266 when Glu124 also is mutated to alanine. In some cases, the position of Tyr195 is closer to that in wild-type caspase 3, although it shows much variability, even in mutants that are active. The position of Tyr195 in the mutants suggests a larger dynamic range for the Tyr195 side chain compared with other residues in the interface, and a survey of caspase 3 crystal structures confirms this conclusion. We examined 23 structures of caspase 3 from the PDB, all with resolution greater than 2.0 Å and in the I222 or P21212 space groups (one monomer per asymmetric unit) (Supplementary Table S2 at http://www.bioscirep.org/bsr/032/bsr0320401add.htm), and found little variation in Tyr197, Arg164, Glu124, or Pro201. The position of Tyr195, however, varied such that its distance from Thr140 ranged between 3.4 and 4.3 Å (Supplementary Figure S9 at http://www.bioscirep.org/bsr/032/bsr0320401add.htm).
Molecular dynamics simulations
In order to further examine the effects of the mutations, we performed molecular dynamics simulations on the proteins shown in Figure 3(A) for a total of 50 ns. Global structures at the end of the simulations were very similar to those of the crystal structures for all mutants, with RMSDs between 1 and 2 Å (results not shown). For wild-type caspase 3, the results show that for both monomers the N-terminus of loop 2′ is very flexible and does not remain bound near the active site. In addition, L1 also demonstrates a high amount of flexibility (Supplementary Figure S10 at http://www.bioscirep.org/bsr/032/bsr0320401add.htm). With regard to the amino acids described above, at the beginning of the simulation Lys137 is 2.7 Å from Glu190, but it quickly moves to an average distance of 5.6 Å and fluctuates between 2.5 and 11.6 Å (with similar values for monomer B), suggesting that the Lys137–Glu190 salt bridge is quite dynamic (see Figure 4B). In contrast, the carboxylate moiety of Glu124 from monomer A is positioned close to Arg164 (2.9 Å) at the beginning of the simulation, and the distance fluctuates only small amounts (2.9–3.3 Å) throughout the simulation. We note, however, that Glu124 from monomer B briefly moves out of the interface to a solvent-exposed position (10.3 Å from Arg164) before moving back into the interface (2.8 Å from Arg164) (Figure 4A). Thus, for the time of the simulation, Glu124 from monomer A remains in H-bonding distance with Arg164, whereas the interactions are somewhat more transient in monomer B. Finally, the distance between Tyr195 and Thr140 fluctuates between 2.5 and 5.5 Å over the 50 ns simulation for both monomers (Figure 4C), consistent with our survey of structures described above. The single mutants Y197C and E124A showed similar results as described for wild-type caspase 3 and are not discussed further (results not shown).
Distances were calculated for atom pairs over the course of the molecular dynamics simulation
In the single mutant, V266H, movements in both monomers occur in a similar manner and so are described as a single event unless otherwise noted. The simulations show that, although the global structure is similar after 50 ns to that of the starting structure (RMSD of 0.89 Å), there are several notable differences in the dimer interface. Within 5 ns of the start of the simulation, the side chain of His266 flips to the C-terminal side of β-strand 8, away from Arg164, where it remains throughout the simulation (Figure 5A). In the new position, His266 faces Tyr195 in a manner similar to that observed in the crystal structures of the double mutant E124A,V266H and the triple mutant. The movement of the His266 side chain results in an increase in the distance between Arg164 and His266 from 4.5 to ~9 Å, although the distance between His266 and Arg164′ across the interface decreases to between 5 and 6 Å (Figures 4E and 4F, and Supplementary Figures S1E and S1F). When the side chain of His266 flips to the C-terminal side of β-strand 8, Tyr195 moves closer to helix 3 and fluctuates between distances of 2.9 and 3.9 Å from Thr140. In addition, the side chain of Glu124 moves out of the interface to a solvent-exposed position (~8 Å from Arg164) (Figures 4A and 5A). Although the position is similar to that described above for monomer B of wild-type caspase 3, the side chain of Glu124 remains in the solvent-exposed position rather than moving back into the interface as observed for wild-type. Later in the simulation, a major rearrangement occurs in helix 3 of monomer B. The N-terminal half of the helix rotates so that the helix is closer to the interface (Figures 4D, 5B and 5C). This movement results in a ~90° rotation of Thr140 away from Tyr195 as well as movement of Lys137 away from Glu190. In this new position, Lys137 alternates between H-bonding with the carbonyl moiety of Pro201′ and the carboxylate of Glu124′, across the interface. In addition, Lys138 forms a new H-bond with Glu190 (Figure 5C), and the distance between Lys138 and Pro201′ decreases from 14 Å to less than 10 Å (Figure 4D). Overall, several new interactions occur due to the rotation of helix 3 and the change in Glu124 to a solvent-exposed position. In this conformation, the interface is narrower by ~4 Å, and the movement of helix 3 pushes the two short β-strands towards the active site (Figure 5B), resulting in the reduction in volume of the central cavity from ~1750 Å3 at 30 ns to ~950 Å3 at 50 ns. For wild-type caspase 3, the volume of the cavity remains constant at ~1900 Å3 throughout the simulation.
Molecular dynamics simulations show that helix 3 is destabilized in caspase 3 (V266H)
The overall dynamics of each mutant was analysed by comparing the calculated B-factors from each simulation. The dimer interface of wild-type caspase 3 remained largely rigid, with few fluctuations observed (Supplementary Figure S10A). The inactive V266H mutant, however, was much more dynamic over the course of the simulation, particularly with regard to the N-terminus of helix 3, as well as with numerous residues in the dimer interface (Supplementary Figure S10D). This increase in the dynamics of the dimer interface was also observed in the inactive mutants, E124A,V266H and E124A,Y197C (Supplementary Figures S10E and S10F respectively), while the active mutants were closer to wild-type (Supplementary Figures S10B and S10C). The E124A mutation resulted in an increase in motion at the site of the mutation, but this did not necessarily correlate with the loss of activity for the protein, as the E124A single mutant remains active.
In the case of Y197C,V266H, numerous similarities were observed when compared with the results for wild-type caspase 3. For example, Glu124 in both monomers remains generally within 3 Å of Arg164 (Supplementary Figure S11A at http://www.bioscirep.org/bsr/032/bsr0320401add.htm). In monomer A, the movements of Lys137 resemble those of wild-type (Supplementary Figure S11B), and although for monomer B the distance from Glu190 fluctuates more than that observed for wild-type, the average distance over the course of the simulation is still somewhat less than for the V266H mutant (7.8 Å compared with 8.5 Å) (Supplementary Figure S11B). In addition, Thr140 for both monomers remains within 5 Å of Tyr195, as seen in the simulations for both the wild-type and V266H proteins (Supplementary Figure S11C). In contrast with V266H, where both histidine side chains rotate to face Tyr195, in monomer A of the Y197C,V266H double mutant, His266 only momentarily rotates to the C-terminal side of β-strand 8, and within 5 ns moves to a more intermediate distance from Arg164 (average of 8.4 Å) when compared with the same interactions in the V266H single mutant (9.4 Å). This position allows for a ring-stacking interaction with Tyr195 (Supplementary Figure 12 at http://www.bioscirep.org/bsr/032/bsr0320401add.htm) and is maintained for the majority of the simulation (Supplementary Figures S11E and S11F). Finally, the movements in helix 3 described above for the V266H single mutant are observed in the double mutant, but only transiently at the beginning of the simulation. This change in helix 3 results in a decrease in the distance between the α carbons of residues Lys138 and Pro201′, from 13 to 10 Å, for less than 2 ns (Supplementary Figure S11D). In the V266H single mutant, this change has occurred by 40 ns, and is maintained for the duration of the simulation. Together, these results are consistent with the observation that the addition of the Y197C mutation was sufficient to restore activity when combined with the V266H mutation because the double mutant more closely resembles the wild-type protein.
In the E124A,Y197C double mutant, which was mostly inactive, numerous differences were observed when compared with wild-type. Although the Y197C,V266H mutant had a transient shift in helix 3, the E124A,Y197C double mutant had much more robust shift in the helix as compared with either V266H or Y197C,V266H. This feature can be observed by a decrease in distance between Lys138 and Pro201′ from 14 Å to ~10 Å, as well as a longer dwell time in this configuration (Supplementary Figure S11D). Over the course of the simulation, Lys137 was an average distance of 10 and 8 Å from Glu190 for monomers A and B respectively with very little time spent within hydrogen bonding distance, more closely resembling the results for V266H. In addition, after the first 10 ns of the simulation, Tyr195 from monomer A rotated away from Thr140 and utilized a different rotamer such that the side chain faced the cavity generated by the mutation of Tyr197 to cysteine (Supplementary Figure S12B), increasing the distance from Thr140 to ~14 Å (Supplementary Figure S11C). These movements were observed transiently in V266H (monomer B), but the separation was more prevalent for the E124A,Y197C double mutant. Interestingly, helix 3 was observed to move towards the interface, in a manner similar to that observed in the single mutant V266H, as shown by the distance between Lys138 and Pro201′ (Supplementary Figure S11D). Collectively, the results presented here show that the enzyme can retain activity when either Tyr197 or Glu124 is mutated, but not both. When mutated together, the E124A,Y197C double mutation results in a dynamic loop between the two surface β-strands, the ‘124-loop’ (Supplementary Figure S10F), destabilization of helix 3 (Supplementary Figure S11D), and an increase in conformational states for Tyr195 (Supplementary Figure 12B). Importantly, the collective data for this mutant show that there may be several ways to manipulate the allosteric site of the interface to populate the inactive states of caspase 3, apart from mutations at Val266.
Finally, when the three mutations are inserted together, the dynamic range of Y195 is more similar to that of wild-type. In this case, the presence of His266 appears to prevent the movement of Tyr195 towards the cavity generated by the Tyr197 to cysteine mutation (Supplementary Figures S11C and S12B). In addition, helix 3 does not appear to sample the alternate conformation, as shown by the constant distance of Lys138 and Pro201′ of between 13 and 14 Å (Supplementary Figure S11D). Together, these features may explain the increase in activity when compared with the E124A,Y197C double mutant.
In the present study, we examined the consequences of mutating Val266 to histidine in the allosteric site of caspase 3. This region of the protein was shown previously to be the site of binding for allosteric inhibitors of caspases 1, 3 and 7 [1,3,7], and a common allosteric mechanism was proposed in which the inhibitors prevent formation of the active conformation by blocking formation of the substrate-binding pocket (loop 3) . A consequence of this mechanism is that Arg164 on active site loop 2 remains exposed to solvent. This state has been referred to as the ‘off-state’ since it resembles the zymogen. The allosteric networks that affect the shift between ‘off-state’ and ‘on-state’ have been mutated to decipher the mechanism, so the roles of several key residues are reasonably well understood [3,23]. In the apo-caspase as well as the procaspase, the allosteric site appears to be used in a similar manner in that the IL binds in the site and prevents active site formation by blocking loop movement into the interface [9,10]. Removal of the IL from the allosteric site, either by mutation  or binding of substrate is sufficient to allow active site formation. So, a common theme between activating the procaspase versus inhibiting the mature caspase appears to involve stabilizing the active site loops through interactions in the allosteric site or reversing the process respectively.
We showed previously that mutation of Val266 to histidine inactivates the protein , and we suggested that steric clashes between His266 and Tyr197 could mimic the effects of the small molecule allosteric inhibitors by preventing insertion of the elbow region of active site loop 3 . The results presented here show that a more accurate representation of the allosteric mechanism includes a consideration of ensembles of active versus inactive conformations rather than discrete states that reflect the positions of the loops. The X-ray crystal structure of caspase 3 (V266H) showed that the histidine side chain was accommodated in the interface, but its presence affected the structure of the active site, resulting in a narrower S1 binding pocket and a disordered loop 1. The structural changes appeared to be propagated through helix 3 on the protein surface and through disruption of H-bonding interactions between Glu124 and Arg164. Molecular dynamics simulations of the V266H variant showed that His266 moves to the C-terminal side of β-strand 8 and displaces Tyr195 towards helix 3. Experimental evidence for this conformation of His266 was observed in the E124A,V266H double mutant as well as the E124A,Y197C,V266H triple mutant. Helix 3 is destabilized such that the N-terminal half of the helix rotates towards the dimer interface, and this conformation is stabilized by new interactions across the interface, particularly between Lys137 and Glu124′, and Lys138 and Glu190′. The conformational change does not occur in wild-type caspase 3, at least on the time scale of our experiments. In wild-type caspase 3, Glu124 remains inserted into the interface to interact with Arg164, and helix 3 does not sample the extended conformation.
Replacing Glu124 with alanine or Tyr197 with cysteine individually is not sufficient to shift the population of states to the inactive ensemble, although when combined, the double mutation of E124A,Y197C results in a largely inactive protein with features similar to the single V266H variant. By X-ray crystallography, we observe that helix 3 is distorted as is the active site, and MD simulations show that helix 3 populates the inactive conformation, where it is rotated towards the interface. In addition, when Tyr197 is mutated to cysteine, the side chain of Tyr195 rotates to fill the cavity generated by the smaller Cys197, but this occurs only if Glu124 also is replaced with alanine. A survey of caspase 3 structures shows several positions for the Tyr195 side chain, and this was confirmed in our MD simulations of the mutants. Although the structural data for V266H suggest that a H-bond between Tyr195 and Thr140 stabilizes the C-terminal end of helix 3, the importance of the interaction is not fully revealed in these studies.
In the context of V266H, the active ensemble is repopulated when Tyr197 also is replaced with cysteine. In the Y197C,V266H double mutant, Glu124 remains inserted in the interface and interacts with Arg164. The presence of His266 also likely prevents Tyr195 from rotating into the new cavity at Cys197. From the X-ray structural data, these features manifest as an ordered helix 3 and loop 1.
Previous structural data of allosterically inhibited caspases demonstrated large changes in the active site loops, particularly in loop 3, which forms the base of the substrate-binding pocket [1,3,7]. We show here that more subtle conformational changes also inactivate the protein. As we examined structures of proteins with inhibitor bound in the active site, we have compared active and inactive proteins in which the substrate-binding pocket is formed. In all of these structures, Arg164 as well as the elbow loop are inserted into the interface. From the structural and MD simulation data for the seven allosteric site mutants, we suggest the following depiction of caspase 3 as part of the inactive ensemble, where one or more of these changes result in inactivation. The side chain of Glu124 is exposed to solvent rather than interacting with Arg164. The side chain of Tyr195 moves to within H-bonding distance of Thr140, and the N-terminal region of helix 3 rotates towards the interface, allowing interactions across the interface among several charged residues, such as Glu124, Lys137, Lys138 and Glu190. As a result of these changes, the volume of the interface is decreased by ~800 Å3, the S1′ pocket is narrower, and the mobility of loop 1 is increased. The level of activity in each mutant examined here may reflect the extent to which this inactive ensemble is populated. The results also show that one of the changes is generally not sufficient to inactivate the protein, but rather several of the changes must occur together.
It is worth noting that while the ‘124-loop’ is present in all caspases, the specific interactions in the interface are not conserved. In caspase 9 for example, where only one of the two monomers is active, seven residues are inserted in the loop, and the additional amino acids extend across the interface to interact with the extended loop from the second monomer . The longer loops abolish the interface cavity in caspase 9, so presumably the interactions between the two loops stabilize the dimer, possibly at the risk of removing the allosteric control found in other caspases. For caspase 9 and other initiator caspases, the allosteric mechanism may be coupled to dimerization, as natural protein inhibitors bind to this site to form heterocomplexes [24,25]. Although the equivalent to Arg164 is conserved in caspase 1 (Arg286 in caspase 1 numbering), the equivalent of Glu124 is not utilized (Glu241 in caspase 1) because a glutamate on β-strand 8 forms a salt bridge with Arg286 [1,3].
As all caspases contain the catalytic histidine–cysteine dyad and because they require an aspartyl moiety in the P1 position, it has proven difficult to design tight-binding and specific caspase inhibitors. In contrast, small molecule inhibitors to the allosteric site of the interface show more promise because the interface differs for the caspase subfamilies. Current allosteric inhibitors of caspases 1, 3 and 7 stabilize the ‘off-state’ of the enzyme [1,3,7,23]. The results presented here suggest that in addition to the better-characterized ‘off-state’, the inactive ensemble also contains states in which the substrate-binding loop is formed. The design of allosteric modulators may take into account these more subtle changes in the allosteric site to shift the population to the inactive ensemble, such as stabilizing helix 3 in the extended conformation, for example. One would predict that such inhibitors would show high specificity since the allosteric site differs among caspases and may be of value for caspase-related therapeutics.
Jad Walters prepared the proteins, conducted enzyme activity measurements and solved structures of the proteins. Josh Schipper performed and analysed the molecular dynamics simulations. Paul Swartz solved the structures by X-ray crystallography and analysed the molecular dynamics simulations. Carla Mattos and Clay Clark designed the experiments and wrote the paper.
This work was supported by the National Institutes of Health [grant number GM065970 (to A.C.C.)]. Use of the Advanced Photon Source was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under contract number W-31-109-ENG-38.
Present address: Department of Chemistry and Chemical Biology, Northeastern University, 102 Hurtig Hall, 360 Huntington Ave, Boston, MA 02115, U.S.A.