Calpain Activation by Cooperative Ca2+ Binding at Two Non-EF-hand Sites*

The active site residues in calpain are mis-aligned in the apo, Ca2+-free form. Alignment for catalysis requires binding of Ca2+ to two non-EF-hand sites, one in each of the core domains I and II. Using domain swap constructs between the protease cores of the μ and m isoforms (which have different Ca2+ requirements) and structural and biochemical characterization of site-directed mutants, we have deduced the order of Ca2+ binding and the basis of the cooperativity between the two sites. Ca2+ binds first to the partially preformed site in domain I. Knockout of this site through D106A substitution eliminates binding to this domain as shown by the crystal structure of D106A μI-II. However, at elevated Ca2+ concentrations this mutant still forms the double salt bridge that links the two Ca2+ sites and becomes nearly as active as μI-II. Elimination of the bridge in E333A μI-II has a more drastic effect on enzyme action, especially at low Ca2+ concentrations. Domain II Ca2+ binding appears essential, because Ca2+-coordinating side-chain mutants E302R and D333A have severely impaired μI-II activation and activity. The introduction of mutations into the whole heterodimeric enzyme that eliminate the salt bridge or Ca2+ binding to domain II produce similar phenotypes, suggesting that the protease core Ca2+ switch is crucial and cannot be overridden by Ca2+ binding to other domains.

Calpains have been found in some prokaryotes (Porphyromonas gingivalis) and single cell eukaryotes (Saccharomyces cerevisiae) but are better characterized in multicellular organisms like C. elegans, Drosophila melanogaster, and humans (2). They have in common a cysteine protease core region composed of two structural domains, I-II (ϳ330 residues). Most calpains have additional C-terminal domains, the two most common being a C 2 -like domain and a Ca 2ϩ -binding penta EF-hand (PEF) 1 domain (18 -20). N-terminal combinations are also observed such as the zinc finger domains of SOL calpains (2). Several calpain isoforms (, m, and nCL-4) form heterodimers with the ubiquitously expressed calpain small subunit (ss1), through their homologous C-terminal PEF domains (2). This dimerization was illustrated in the crystal structures of the Ca 2ϩ -free m-calpain heterodimer (18,21). In addition to its C-terminal PEF domain, the small subunit contains an Nterminal glycine-rich domain V that is too mobile to see in the x-ray crystal structure. Other calpains have been shown to exist without the small subunit, including p94 (calpain 3), the calpain 3 lens-specific alternative transcript Lp82, and the Drosophila calpains A and B (2).
Although most calpains are complex multidomain proteins, the nCL-2Ј splice variant of nCL-2 (calpain 8) and the P. gingivalis calpain contain only the two protease core domains. Consistent with this observation, we have recently shown that the protease core from -calpain (referred to here as I-II or -minicalpain) is a Ca 2ϩ -dependent cysteine protease when expressed in isolation (22). The crystal structure of the Ca 2ϩbound I-II revealed two novel non-EF hand Ca 2ϩ binding sites, one in each domain, that are conserved from C. elegans to humans. Neither Ca 2ϩ ion is directly involved in catalysis, but rather Ca 2ϩ binding to these sites appears to drive the realignment of the active site residues to the positions seen in the related papain family of cysteine proteases. Based on comparison with the protease core in the inactive m-calpain heterodimer structure (18,21), major conformational changes must occur upon Ca 2ϩ binding. Each Ca 2ϩ site is postulated to assemble through the rearrangements of flexible loops, their loading permitting the two domains to switch into the active conformation in a concerted manner (Fig. 1). Although the x-ray structures reveal the starting and end points of the conformational change, they do not explain how the transition is effected. Here we elucidate the order of Ca 2ϩ binding to the core sites, the basis for cooperativity between them, and the overall mechanism for realignment of the active site residues. This activation mechanism is central and fundamental to the calpains.

EXPERIMENTAL PROCEDURES
Cloning, Expression, and Purification of /m-minicalpain Hybrids and I-II Mutants-The reciprocal /m minicalpain hybrid constructs, I-mII and mI-II, were generated by swapping I or mI coding segments into the corresponding region of the previously described pET24d mI-II or I-II constructs, both of which have a C-terminal His 6 tag (22,23). The AgeI unique restriction site, which occurs naturally in -calpain at the DI-DII boundary (-217 FTGG 220 ), was used as the swap site. An AgeI restriction site was engineered by Kunkel mutagenesis into the corresponding location within mI-II without alteration of its amino acid sequence using the primer ctcggcaatgccaccggtgaagtcttcaaag (24). Swapping NcoI-AgeI fragments produced the desired mI-II and I-mII hybrids. Primers used to generate amino acid replacements within pET24d-I-II are identified by their targeted amino acid substitution. In each case a diagnostic restriction site (underlined) was introduced. The primers were: D106A, NdeI agctccctggcatatggccgttcgggtgg; W298A, SacII ctccacttcacccgcggggttccgcat; E302R, ApaI gtcactccagggccctttccaccgcacttcacttcaccccag; D331A, NaeI gacctccagaattcgccggcttccatcttgacc; and E333A, NarI aaggacatccagaaggcgccgtcttccatct. Two mutations (a double and a single) within the pET24d-m80k construct were achieved using the primers E320A/D321A, SacII atccagaattctcccgcggcctgccgttctg; and R94G, KpnI ccttggcagatgtcggtaccggtagcacctcc. To perform protein crystallization in the presence of Ca 2ϩ without the risk of autolysis, the inactivating cysteine 115 to serine mutation was introduced in the D106A I-II construct using C115S, PvuII agccaggagccagctgtcccccagagc. Cloned and mutated inserts were sequenced (Cortec DNA Services Laboratory, Inc.) to confirm their identity and integrity. Expression in Escherichia coli, purification, and storage of the above constructs and their mutants followed the previously described protocols (22).
Differential Scanning Calorimetry-Correct folding of the minicalpain mutants was verified by differential scanning calorimetry (DSC) using a VP-DSC calorimeter (Microcal) (25). Temperature change was at the rate of 90°C/h, and scans were recorded from 20°C to 60°C. Prior to analysis, samples were dialyzed overnight in 4 liters of buffer containing 50 mM HEPES (pH 7.6), 200 mM NaCl, 1 mM EDTA, and 1 mM dithiothreitol. The dialysis buffer was filtered using a 0.22-m filter (Millipore) and used to obtain a baseline during several buffer-buffer scans. Protein-buffer scans were performed at a final protein concentration of 0.2 mg/ml. Using Origin 5.1 software, the data was corrected for the buffer transitions and modeled to one thermal transition to obtain the melting temperature.
Limited Proteolysis-Limited proteolysis of I-II and its mutants was performed at 22°C using L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated trypsin or 1-chloro-3-tosylamido-7-amino-2-heptanone-treated chymotrypsin (Sigma). The substrate concentration was 1 mg/ml in a reaction volume of 150 l or less. Trypsin or chymotrypsin was present at 0.1 mg/ml. To monitor the progress of proteolysis, aliquots at timed intervals were mixed with 5ϫ SDS sample buffer and were analyzed by SDS-PAGE and Coomassie Blue staining.
Activity and Intrinsic Tryptophan Fluorescence Assays-The steadystate kinetics of mutants and /m minicalpain hybrids were measured under similar conditions to those originally described for I-II using the peptide, succinyl-leucine-tyrosine-aminomethyl coumarin (SLY-MCA, Sigma), as substrate (22). These conditions were also maintained for the CaCl 2 activity titrations, which involved increasing CaCl 2 con-FIG. 1. Schematic outline of Ca 2؉ -induced conformational changes. The positions of key residues and peptide backbone before and after Ca 2ϩ -induced conformational changes are signified by dotted and solid lines, respectively. The movement of specific residues is indicated by red arrows. Domains I and II are represented by blue and cyan, respectively. A, calcium site 1 in domain I. B, the double salt bridge linking domains I and II. C, calcium site 2 in domain II. Because of the complexity of the structural changes whereby two loops move to form site 2, the before and after views are shown separately above and below the thick arrow. D, retraction of the tryptophan wedge from the active site cleft. centrations up to 175 mM for the minicalpains. Intrinsic tryptophan fluorescence (IWF) measurements and data analysis were performed as previously described (22).
Crystallization and Structure Determination-Crystallization of D106A/C115S I-II in the presence of Ca 2ϩ was done by the hanging drop, vapor diffusion method under slightly different conditions from those established for wild-type I-II, and produced crystals of a new space group. The well solution contained 1.5 M NaCl, 0.1 M MES (pH 6.5), and 10 mM CaCl 2 . The drop size was less than 5 l and contained an equal volume of well solution and protein solution (12.5 mg/ml). Crystals were cryo-protected by six serial soakings (each for up to 5 min) in stabilization solutions containing increasing 5-30% (v/v) glycerol. Diffraction data were collected at the Cornell High Energy Synchrotron Source beamline F1, which was equipped with a Quantum-4 charge-coupled device detector (Area Detector Systems Corp.) and a liquid nitrogen cryo unit (Oxford Cryosystems). The diffraction data were processed as previously described for I-II (22). The space group was P2 1 2 1 2 1 with one molecule per asymmetric unit ( Table I). The structure solution was determined by molecular replacement using AMoRe, and the structure of the Ca 2ϩ -bound I-II (PDB accession code 1KXR) as a model. Refinement and analysis of the structure were performed using the software described for I-II (22). The ribbon diagrams were generated using Molscript (26). The electron density diagram was made using Xfit, and all structure figures were rendered in RASTER3D (27,28).

RESULTS
Validation of Mutants-In this study, the Ca 2ϩ -dependent mechanism by which the active site of calpain is aligned for catalysis was investigated using domain swaps and site-directed mutagenesis (Fig. 2) (22). Where targeted amino acid changes interrupt or alter a specific step in the process, it is essential to establish that the change in activity is due to the loss of a functional group and not due to the mis-folding of the mini-calpain. In general, mis-folded proteins are either poorly produced in E. coli or accumulate in inclusion bodies. All mutants and swaps used here expressed well in E. coli, remained soluble, and gave final yields of Ͼ10 mg/4 liters of culture that are comparable to those obtained with I-II and mI-II. Another indication of correct folding is that the chromatography profiles of mutant and wild-type proteins were comparable (not shown). All the minicalpains produced discrete, sharp peaks that eluted at similar locations in the profiles.
To further validate the mutants, their thermal denaturation profiles were measured using DSC. Each minicalpain, mutant and wild type, showed a single thermal transition (Fig. 3A). The values ranged from 47°C (wild type) to 44°C (D106A). The latter mutant was the one for which an x-ray structure was determined, which revealed no significant differences in the protein fold. With the thermal transitions for the other mutants falling in between these flanking values, it can be safely assumed that none of them was structurally compromised. Limited proteolysis with chymotrypsin ( Fig. 3B) and trypsin (not shown) confirmed this. The digestion products were essentially the same for each minicalpain after 80 or 320 min, the only difference being a slightly faster rate of breakdown for some mutants like E302R and W298A. The latter and E333A were the subject of NMR analysis by comparison of its HSQC spectra with that of the wild-type I-II (not shown), which revealed no significant perturbations in the fold of the mutant.
Experimental Strategy-In conjunction with the domain swaps and site-directed mutagenesis, we have used two specific biochemical features of the calpain protease core to elucidate its mechanism of activation by Ca 2ϩ . One is the large change in IWF that accompanies the Ca 2ϩ -driven alignment of the active site for catalysis (22). We provide evidence below that this change is, as previously suspected, largely due to the movement of Trp 298 . The other is the Ca 2ϩ -dependent proteolytic activity of the core as measured by SLY-MCA hydrolysis. The gain in enzyme activity for I-II is biphasic. The first phase is sigmoidal. It reflects the cooperative binding of the two Ca 2ϩ and parallels the IWF change. The second phase is a gradual gain in activity at unphysiologically high Ca 2ϩ levels, which we attribute to general stabilization of the bilobal core. The core lacks the support of the other domains it would have in the whole enzyme. What readily distinguishes the m-and -minicalpains as enzymes is the greatly reduced activity of mI-II due to the collapse of a key helix in domain I leading to a rearrangement of hydrophobic core residues (23).
Minicalpain Domain Swaps Demonstrate the Order of Ca 2ϩ Binding-The increase in IWF during Ca 2ϩ titration was measured for four minicalpains, native I-II and mI-II, and the reciprocal domain swaps, mI-II and I-mII. All four gave characteristic sigmoidal profiles that fitted well to the Hill equation ( Fig. 4A and Table II). Ca 2ϩ binding was cooperative between the two domains as indicated both by the upward curvature of the Scatchard plots ( Fig. 4A, inset) and by the Hill coefficients of ϳ2. Ca 2ϩ concentrations for half-maximal change ([Ca 2ϩ ] 0.5 ) were ϳ27 and 190 M for I-II and mI-II, respectively ( Fig. 4A and Table II). These values are remarkably close to the Ca 2ϩ requirement for activation of the intact -(5-50 M) and m-calpains (200 -1000 M). The Ca 2ϩ requirements for the mI-II and I-mII swaps were 33 and 104 M, respectively ( Fig. 4A and Table II), intermediate between those of I-II and mI-II. Even more informative was the shape of the sigmoidal IWF profiles. The II-containing constructs, I-II and mI-II, produced very similar profiles that saturated at ϳ200 M, whereas the mII-containing constructs, mI-II and I-mII, both reached a plateau at ϳ1 mM. Thus, it appears domain II controls the end point of the titration and binds Ca 2ϩ second. The initial part of the transition was very similar for the II-containing constructs, I-II and mI-II, with both proteins beginning their change at Ͻ10 M Ca 2ϩ . We, therefore, reasoned that DI from either isoform has high Ca 2ϩ affinity and binds Ca 2ϩ first, with I being a slightly better binder than mI.
We further demonstrated the functionality and hybrid nature of the domain swaps by testing their enzyme activity. We have previously shown that I-II readily hydrolyzes the calpain substrate SLY-MCA (22). In contrast, mI-II fails to cleave SLY-MCA because of an intrinsic inactivation mechanism triggered by the collapse of a key ␣-helix at Gly 203 in domain I (23). This occurs in the Ca 2ϩ -loaded state and leads to the protrusion of Trp 106 into the active site cleft. Consistent with this, the where F o and F c are observed and calculated structure factor amplitudes, respectively. activity of the mI-containing swap, mI-II, was barely detectable with a k cat of 3.6 Ϯ 0.9 ϫ 10 Ϫ5 s Ϫ1 , and a K m of 0.46 Ϯ 0.01 mM ( Fig. 4B and Table II). In contrast, the I-containing swap, I-mII, where there is an alanine in place of Gly 203 , was quite active. Its k cat (2.3 Ϯ 0.3 ϫ 10 Ϫ4 s Ϫ1 ) was about half that of I-II, and its K m was slightly higher (0.403 Ϯ 0.009 mM versus ϳ0.3 mM). As seen with I-II, the activity of I-mII (but not that of mI-II) was augmented at higher ionic strength due to general stabilization of the core.
We have also modeled the swaps using the Ca 2ϩ -bound I-II and mI-II structures (22,23). The domain interface is highly conserved between the isoforms (not shown), and there are no obvious structural impediments to the hybrids undergoing activation by the same mechanism that brings the two domains together in I-II. Several inferences can be made from the domain swap analysis. Regardless of isoform, the domain I site has a very similar high affinity for Ca 2ϩ , being occupied prior to domain II. In contrast, the affinity of domain II is isoformspecific being the principal determinant of Ca 2ϩ requirement within the protease core.
Trp 298 Contributes Most to the Intrinsic Tryptophan Fluorescence Change of I-II-Ca 2ϩ loading results in a very substantial IWF increase (ϳ37% in I-II and ϳ28% in mI-II) that shows the cooperativity of the process and makes the conformational change easy to follow. The prime candidate for the major contributor to the IWF increase is Trp 298 (22). Its indole ring moves from the wedge-like position in between domains, where it is exposed to solvent on both sides, into a Ca 2ϩinduced hydrophobic pocket that fully buries one of its sides (Fig. 1). Because the Trp 298 residue remains surface-accessible, rather than contributing to a hydrophobic core, its mutagenesis to alanine is structurally feasible. Its position above the domain II catalytic triad residues (His 272 and Asn 296 ) matches that seen in other cysteine proteases, where it shields these residues from solvent exposure and thus facilitates catalysis. We made the W298A mutant in the inactive C115S I-II, because we predicted that activity would in any case be abolished by exposure of the key catalytic residues to solvent. Indeed, a W288Y mutation within the m-calpain heterodimer resulted in a significantly lower activity (Ͻ5% of wild type), despite this being a much more conservative amino acid substitution (29).
The correct folding of W298A was validated by differential scanning calorimetry (Fig. 3A), limited proteolysis (Fig. 3B), and two-dimensional NMR (heteronuclear single quantum coherence spectroscopy) using 15 N-labeled protein (not shown). In all of these tests the mutant was virtually indistinguishable from the wild type. However, this mutation drastically reduced the IWF increase upon Ca 2ϩ saturation from 37% (wild-type value) to 10% (Fig. 5A, gray trace). This increase was still sigmoidal and fitted the Hill equation with [Ca 2ϩ ] 0.5 ϳ 15 M, being Ͼ90% complete at ϳ40 M, the [Ca 2ϩ ] 0.5 value for I-II ( Fig. 3B and Table II). Thus, the W298A mutation did not affect the cooperativity between Ca 2ϩ binding sites as indicated by an unchanged Hill coefficient of 1.8 and an even more pronounced curvature of the Scatchard plot (not shown). Therefore, it appears that about 75% of the observed IWF change is contributed by Trp 298 during the advanced stages of the sigmoidal transition. The change in the environment of Trp 298 that causes the IWF increase coincides with formation of domain II Ca 2ϩ site, which implies that site II formation occurs during the late stage of the activation mechanism.
Ca 2ϩ Binding to Domain I Site Can Be Circumvented in Vitro-Asp 106 and Glu 185 provide full side-chain coordinations to the Ca 2ϩ site within domain I (Figs. 1 and 2B). Disruption of either residue has the potential to eliminate Ca 2ϩ binding. Of the two residues, Glu 185 moves least to accommodate Ca 2ϩ and might therefore be the first residue to sense Ca 2ϩ (22). However, there is an electrostatic stabilizing interaction formed between it and the microdipole of the central ␣-helix around which the domain I hydrophobic core is wrapped (Fig. 2B), suggesting that its replacement might negatively affect DI folding and stability. In contrast, Asp 106 is more peripheral, being part of the flexible loop (residues 99 -106) that closes in to provide the other protein coordinations to the Ca 2ϩ site. Because its side chain is not involved in major stabilizing interactions in the Ca 2ϩ -free structure and is more solventexposed than Glu 185 , it was chosen for mutation to A (D106A).
The IWF profile of D106A I-II during Ca 2ϩ titration was significantly different from that of wild-type I-II and produced only a 5% total change in IWF (Fig. 5 (A and B) and Table  II). The calcium requirement for this transition was much greater than for the wild type ([Ca 2ϩ ] 0.5 ϳ 145 M). The poor fit to the Hill equation, as indicated by a Hill coefficient of ϳ1.3 and a much flatter Scatchard profile, suggested that the domain I Ca 2ϩ binding site was abolished in this mutant (Fig. 5B) along with the natural cooperativity between the two sites.
We also followed the activity of these minicalpains during Ca 2ϩ titrations by monitoring the hydrolysis of SLY-MCA. The initial Ca 2ϩ -dependent sigmoidal increase in wild-type I-II activity (Fig. 4A, blue trace), which correlated with the increase in IWF and was complete at 200 M Ca 2ϩ , represents enzyme activation (22). This was followed by a nonspecific, gradual increase (hyperbolic) in activity brought about by divalent cations (and to a lesser extent by monovalent cations), which we have argued represents structural stabilization of the enzyme (22). (Within the heterodimeric calpains the C 2 -like domain III may fulfil this role of stabilizing the core.) At low Ca 2ϩ levels (Ͻ200 M), the D106A mutant of I-II gave a hyperbolic increase in activity instead of the wild-type sigmoidal change, suggesting that only one Ca 2ϩ site was functional (Fig. 6A, yellow trace). Despite this and the difference in IWF profiles, D106A was a surprisingly efficient enzyme. Indeed, the SLY-MCA hydrolysis kinetics of both wild type and D106A were comparable over the low mM Ca 2ϩ range (Fig. 6A).
D106A Crystal Structure Reveals an Empty DI Ca 2ϩ Site-To check for structural changes induced by the D106A mutation and to be sure that it completely abolished Ca 2ϩ binding at the DI site we solved the D106A/C115S I-II structure to 1.9-Å resolution. We established new conditions for the crystallization of I-II in the presence of Ca 2ϩ , which, in comparison to the original conditions (22), produced more stable crystals with higher symmetry that diffracted at higher resolution (Table I).
After the first round of refinement of the Ca 2ϩ -bound I-II model, which was used to obtain the molecular replacement solution for D106A I-II, the 2F o Ϫ F c map indicated that density corresponding to the Asp 106 side chain was absent beyond the C␤, as expected for the D106A substitution (not shown). In addition, the density corresponding to the Ca 2ϩ ion in domain I of I-II was not matched in the new map. Upon Ca 2ϩ removal from the model, the following round of refinement confirmed the absence of a Ca 2ϩ ion at this position, which was instead filled by a water molecule (wat1, Fig. 7, B and C). The second water molecule that coordinates the Ca 2ϩ in I-II (wat4, Fig. 7B) is absent in the D106A mutant structure, suggesting that Ca 2ϩ likely does not bind at this site even in the presence of 10 mM CaCl 2 used during crystallization. Higher resolution data (ϳ1.5 Å) might permit refinement of the occupancy at this site, which is further hampered because of the weak contribution of Ca 2ϩ to the electron density (20 electrons).
With the exception of these changes, the Ca 2ϩ -bound structure of D106A was essentially identical to I-II (Fig. 7A), even within the loop that contains D106A, meaning that this was in the "closed" Ca 2ϩ -coordinating position despite the absence of Ca 2ϩ in site 1 (Fig. 7B). The relative orientation of domains and realignment of active site residues was preserved. In addition, all critical interactions directly involved in the activation mechanism were maintained, including the Arg 104 -Glu 333 salt bridge, the occupied Ca 2ϩ site in DII, and the Ca 2ϩ -induced Trp 298 pocket (Fig. 7A). The D106A I-II structure shows that at unphysiologically high Ca 2ϩ levels used in vitro, Ca 2ϩ binding to domain I can be circumvented in the mutant. Activation of D106A I-II likely follows a slightly different molecular mechanism. We suggest that the loop that harbors Arg 104 could move close to the Glu 185 side chain, because it is free of electrostatic clashes that would otherwise be caused by the Asp 106 side chain in the absence of Ca 2ϩ . In this process Arg 104 would already be repositioned into the observed Ca 2ϩ -bound orientation, and would in turn facilitate the rotation and side-chain removal of Glu 333 from the domain II Ca 2ϩ site without having to overcome the energetic barrier imposed by forming the domain I Ca 2ϩ site (Fig. 1).
The Arg 104 -Glu 333 Salt Bridge Is More Important for Activa-tion than Ca 2ϩ Site 1-The double salt bridge, Arg 104 -Glu 333 , is the direct physical link between the Ca 2ϩ sites in domain I and II and, therefore, the underlying structural basis for cooperativity in the core (Fig. 1). To test this hypothesis, we abolished this interaction through E333A mutation (Fig. 2). This mutation radically changed the IWF profile during Ca 2ϩ titrations (Fig. 5). There was no increase in IWF until after 100 M CaCl 2 , which is consistent with the inability of Ca 2ϩ site 1 binding to influence the transition, site 1 being the first to bind Ca 2ϩ (Fig.  5A). The data from the beginning of the fluorescence increase (Fig. 5B, orange trace, [Ca 2ϩ ] 0.5 ϳ 300 M) did not fit the Hill equation, despite a Hill coefficient of 1.8. They generated the most severely flattened Scatchard plot of the series, indicating a major disruption in cooperativity (Fig. 5B, inset). Enzyme activity was also severely impaired at low Ca 2ϩ concentrations. It became detectable at 300 -400 M CaCl 2 (Fig. 6A), but the initial activation phase was stretched out to ϳ10 mM CaCl 2 (not shown) compared with ϳ200 M for I-II, and it saturated at a lower value (Fig. 6A).

Domain II Ca 2ϩ Binding Defines the Next
Step in the Activation Hierarchy of I-II-Having probed the domain I Ca 2ϩ site and the double salt bridge between domains I and II, we next looked at the domain II Ca 2ϩ site. A critical side-chain rearrangement must be made by residue Glu 302 to form this second Ca 2ϩ site (22). Glu 302 exchanges position with Val 301 to provide double coordinations to Ca 2ϩ , while creating room for a short anti-parallel ␤-sheet (Fig. 1). This sheet, along with the Val 301 side chain, lines the periphery of the Trp 298 -binding pocket together with the Ile 263 and Val 269 side chains. We considered that E302R might generate a Ca 2ϩ -independent enzyme, because the arginine side chain can be modeled with its guanidino group within the Ca 2ϩ binding position, where it could interact electrostatically with the rest of the coordination complex (Fig. 2C). This thought was partly driven by the threedimensional alignment model of the Ca 2ϩ -bound calpain 7 DI-II (not shown), which contains an Arg at this position and is believed to be Ca 2ϩ -independent (2). When Arg replaced Glu 302 , the mutation abolished enzyme activity even at elevated divalent cation levels. This was despite an associated 15% increase in IWF (Fig. 5A, green trace) ϳ 102 M, Hill coefficient ϳ 1.3, Fig. 5B). E302R presumably blocked formation of a functional domain II Ca 2ϩ site in I-II  and prevented the formation of the Trp 298 pocket, suggesting that proper site 2 assembly is essential for activation. We therefore made a less disruptive substitution of a Ca 2ϩcoordinating residue in DII, namely the D331A mutation. D331 provides one-side-chain coordination to Ca 2ϩ as well as a stabilizing coordination to the water molecule that in turn coordinates Ca 2ϩ (Fig. 2B) (22). This side chain is somewhat secondary to the activation mechanism, not being involved directly in critical inter-stage transitions such as Glu 333 or Glu 302 repositioning. Nevertheless, this mutation impaired the activation mechanism more so than the E333A salt bridge mutation as observed by a dramatic effect on its activity versus Ca 2ϩ titration during the activation phase (Fig. 6A, pink). There was only a trace amount of activity at sub mM Ca 2ϩ concentrations. The IWF change (totaling ϳ17%; Fig. 5A) did not correlate with the initial activation phase, being more hyperbolic than sigmoidal, and producing a poor fit to the Hill equation ([Ca 2ϩ ] 0.5 ϳ 100 M, Hill coefficient ϳ 1.3; Fig. 5B). The data indicate that both the structure and the activation mechanism are directly affected by the mutation. Furthermore, the linearity of the Scatchard plot obtained from the IWF data (Fig. 5B, inset) helped confirm that D331A abolished Ca 2ϩ binding in domain II.
Effects of Activation Mechanism Mutations on m-calpain Heterodimer-Based on these minicalpain mutation studies, the double salt bridge Arg 104 -Glu 333 formation and the proper assembly of the DII Ca 2ϩ site are the most critical stages of the activation mechanism, the latter being significantly more important. To validate this conclusion in the context of the whole enzyme, we chose to abolish these critical stages within the m-calpain heterodimer. R94G was used to mimic the effect of E333A in eliminating the salt bridge, and the E320A/D321A double mutation was used in the manner of D331A to retard assembly of the DII Ca 2ϩ site (note: m-calpain numbering lags by 10 from that of -calpain). Their activity profiles during Ca 2ϩ titrations were similar to those of the corresponding I-II mutants during the activation stage. R94G and E320A/D321A required ϳ1.5 mM and ϳ8 mM CaCl 2 for half-maximal activation, respectively, compared with ϳ350 M for the wild type (Fig. 6B). These extremely high Ca 2ϩ requirements were never achieved using any single site mutation within the DIV or DVI PEF domains (30). The most severe single EF-hand mutation only increased the CaCl 2 concentration for half-maximal activation to 0.83 mM. In addition, these enzymes were weaker, both being ϳ20% as active compared with wild type at 20 and 60 mM CaCl 2 , respectively (Table II). A lower activity was not observed with any of the PEF mutants or PEF DIV swaps, which essentially have wild-type activities but with a different Ca 2ϩ requirement (30,31). These results suggest that the activation mechanism observed within the isolated protease core can be confidently extended to the activation of the core within the context of the heterodimer. In this respect DII Ca 2ϩ site assembly is the ultimate rate-limiting event during the heterodimer activation, being strongly facilitated by the Arg 104 -Glu 333 cooperative interaction. DISCUSSION The presence of 10 -12 Ca 2ϩ binding sites throughout calpain complicates the issue of cooperativity in the whole enzyme, despite the structural information gathered thus far (22,30). In contrast, the protease core deals only with one pair of sites that act cooperatively to realign the active site cleft for catalysis. A comparison of the crystal structures of apo m-calpain and Ca 2ϩ -bound I-II reveals the net changes that are required for activation through the realignment of the active site (21,22). From these snapshots we previously surmised that the two non-EF-hand Ca 2ϩ sites were physically and mechanistically linked through the double salt bridge, but we had no biochemical evidence for this or for the order and progression of Ca 2ϩ binding (22). Elucidation of the activation mechanism is provided here through biochemical analysis of and mI-II domain swaps and I-II mutants of select residues involved at sequential steps of the mechanism.
Evidence for the order of Ca 2ϩ binding came initially from the domain swaps and their IWF change during Ca 2ϩ titration, which indicated that domain I binding determines the onset of the sigmoidal IWF profile, and domain II binding controls its saturation. We had previously shown that the 4-to 5-fold difference in Ca 2ϩ requirement of theand m-calpain isoforms is faithfully reflected in the Ca 2ϩ requirement for activation within the protease core, being complete at ϳ200 M Ca 2ϩ for I-II and at ϳ1 mM Ca 2ϩ for mI-II (23). We now show that saturation is largely determined by domain II, because both II-containing minicalpains complete their titration at ϳ200 M Ca 2ϩ whereas the mII-containing minicalpains do not reach saturation until ϳ1 mM Ca 2ϩ . Thus, domain II is involved in the later stage of the active site assembly, which includes the removal of Trp 298 from its wedge-like position in the active site cleft. This cannot occur until site 2 is formed. Although Trp 298 is one of 11 Trp residues in I-II, its mutation to Ala reduced the IWF change by ϳ75%. (The other 10 Trp residues are found in overlapping positions in the apo-m and holo-protease cores.) This mutation helped establish the order of Ca 2ϩ binding. W298A abolished the final stage of the sigmoidal IWF transition, emphasizing its involvement during the later steps of the activation mechanism. This Trp is present in all papainlike cysteine proteases, but only in calpains does it function as a removable wedge and undergo a radical change in environment with activation (32).
The fact that part of site 1 is already pre-assembled is suggestive of a role in the early stages of activation, but direct evidence that domain I binding determines the onset of the sigmoidal IWF profile came from the salt bridge knock-out mutation E333A. This mutation essentially uncoupled Ca 2ϩ binding to site 1 from the activation process. All cooperativity was lost, and the IWF transition did not begin until after 100 M Ca 2ϩ was reached. Clearly, what should have been an early boost from site 1 binding was unable to initiate the cooperative process, because the distal part of the salt bridge was missing and unable to transmit the Ca 2ϩ signal. E333A mutation also resulted in ϳ10-fold increase in the Ca 2ϩ requirement for activation, suggesting that domain II site formation is greatly facilitated by the cooperative link, which helps remove the Glu 333 side chain from the Ca 2ϩ ion position. Support for the cooperativity between the two protease-core Ca 2ϩ sites came from disruption of either end of the Arg 104 -Glu 333 double salt bridge, through E333A substitution in I-II and R94G in m-calpain. Sequence alignments revealed that about two thirds of the calpain isoforms have conserved the Arg-Glu salt bridge, further supporting the role for this ancestral cooperative link. In addition, other key residues (identified in Fig. 2B) are Ͼ95% conserved among all large subunit homologues, perhaps indicating the mechanism's importance in conveying the Ca 2ϩ signal to the active site within these isoforms, and suggesting an ancestral common origin for these calpains. The R94G mutation within the context of the heterodimeric m-calpain severely impaired activation and diminished activity (33). The equivalent mutation (R118G) occurs in the musclespecific calpain isoform, p94, in some patients with limb girdle muscular dystrophy 2A (13). Loss of function of this enzyme is responsible for the disease (35). R118G is one of many inactivating mutations catalogued in p94 but one of very few, which can now be explained mechanistically in light of our corresponding mutation in m-calpain. From this study we conclude that p94 R118G activity is impaired due to loss of site cooperativity within its protease core. A completely unexpected finding of the mutagenesis experiment was the resilience of the D106A mutation, which was designed to knock out the domain I Ca 2ϩ binding site. Although the crystal structure showed that Ca 2ϩ was missing from site 1, the movable loop containing D106A had in fact closed up to near the normal position but with water in place of Ca 2ϩ . Passive closure of this loop in the absence of Ca 2ϩ would of course permit the double salt bridge to form and facilitate Ca 2ϩ binding to site 2. Also, under the high Ca 2ϩ (10 mM) crystallization conditions, occupancy of site 2 might form the salt bridge from the reverse direction. Formation of the salt bridge would clearly facilitate the closing up of the movable site 1 loop toward Glu 185 , and, unlike Asp 106 , Ala 106 would not cause charge repulsion in the absence of Ca 2ϩ . However, in the wild type, domain I site confers cooperativity to the protease core and directly modulates the activation mechanism. We conclude that the cooperative link is more important than the domain I site, because it forms despite the D106A mutation and it more directly facilitates Ca 2ϩ binding to the rate determining domain II site.
It appears that mutations introduced along the mechanism from the domain I site, into the cooperative link, and further into the DII site have increasingly dramatic effects on activation, suggesting a similar hierarchy of significance for these stages. In this regard, the domain II site mutations within the m-calpain heterodimer, E320A/D321A, had more severe effects on activation and activity than mutations of any single EF-hand motif within the PEF domains IV and VI (30). Loss of Ca 2ϩ binding in domain II was more disruptive than losing the double salt bridge through R94G substitution. This result extends our model of heterodimer activation, which begins with Ca 2ϩ binding at both PEF domains and the concomitant release of the N-terminal large subunit anchor from the small subunit (22). These initial events release some of the tension on the protease core. The involvement of domain III in the protease core stabilization is postulated indirectly from the nonspecific stabilization of the core by divalent cations (23) and directly through mutations at the domain II-domain III interface (36). However, the putative Ca 2ϩ binding ability of this C 2 -like domain should not be excluded as a specific contributor to the Ca 2ϩ -induced release of tension on the core (37). When freed of tension, the protease core should realign through our proposed mechanism, which makes domain II occupancy the last and crucial rate-limiting step of heterodimer activation, and by analogy that of most calpain isoforms. Although additional work is needed to address the activation of the more complex calpains, our mechanism might completely describe the activation of some of the least complex ones, including nCL-2Ј (2) and the naturally occurring protease cores (akin to I-II) generated through calpain autolysis. In stark contrast to the initial concept that PEF domains were central to calpain activation by Ca 2ϩ (2,18,21), it appears that they are later additions to further control activation by Ca 2ϩ signaling. Calpain activation originates in the core and was the key step in evolving a Ca 2ϩ -dependent cysteine protease from a papain-like progenitor.