Conformational Changes of Calpain from Human Erythrocytes in the Presence of Ca2+ *

Small angle x-ray scattering has been used to monitor calpain structural transitions during the activation process triggered by Ca2+ binding. The scattering pattern of the unliganded enzyme in solution does not display any significant difference with that calculated from the crystal structure. The addition of Ca2+ promotes the formation of large aggregates, indicating the exposure of hydrophobic patches on the surface of the protease. In contrast, Ca2+ addition in the presence of the thiol proteinase inhibitor E64 or of the inhibitor leupeptin causes a small conformational change with no dissociation of the heterodimer. The resulting conformation appears to be slightly more extended than the unliganded form. From the comparison between ab initio models derived from our data with the crystal structure, the major observable conformational change appears to be localized at level of the L-subunit and in particular seems to confirm the mutual movement already observed by the crystallographic analysis of the dII (dIIb) and the dI (dIIa) domains creating a functional active site. This work not only provides another piece of supporting evidence for the calpain conformational change in the presence of Ca2+, but actually constitutes the first experimental observation of this change for intact heterodimeric calpain in solution.

Small angle x-ray scattering has been used to monitor calpain structural transitions during the activation process triggered by Ca 2؉ binding. The scattering pattern of the unliganded enzyme in solution does not display any significant difference with that calculated from the crystal structure. The addition of Ca 2؉ promotes the formation of large aggregates, indicating the exposure of hydrophobic patches on the surface of the protease. In contrast, Ca 2؉ addition in the presence of the thiol proteinase inhibitor E64 or of the inhibitor leupeptin causes a small conformational change with no dissociation of the heterodimer. The resulting conformation appears to be slightly more extended than the unliganded form. From the comparison between ab initio models derived from our data with the crystal structure, the major observable conformational change appears to be localized at level of the L-subunit and in particular seems to confirm the mutual movement already observed by the crystallographic analysis of the dII (dIIb) and the dI (dIIa) domains creating a functional active site. This work not only provides another piece of supporting evidence for the calpain conformational change in the presence of Ca 2؉ , but actually constitutes the first experimental observation of this change for intact heterodimeric calpain in solution.
Calpains (EC 3.4.22.17) are a family of cytosolic Ca 2ϩ -dependent cysteine endopeptidases widely distributed in all mammalian cells and comprising several genetically distinct isozymes and various homologues in different lower organisms (1,2). The physiological role of calpains is related to the transduction of extracellular signals mediated by changes in the permeability of membranes to Ca 2ϩ or by the mobilization of this ion from internal stores. These proteins are involved in physiological and pathophysiological conditions such as cell cycle regulation, apoptosis, cytoskeletal remodeling, Alzheimer's and Parkinson's diseases, and muscular dystrophies (1,(3)(4)(5)(6)(7)(8).
Most biochemical and structural studies concern the two ubiquitous enzymes namedand m-calpain exerting their in vitro catalytic activity at micromolar (10 -50 M) or millimolar (0.2-0.35 mM) Ca 2ϩ concentrations, respectively (1). The calpain from the Ca 2ϩ -activated proteolytic system of human erythrocytes is a -calpain localized in the cell soluble fraction that in vitro binds a maximum of eight equivalents of Ca 2ϩ with a K d of ϳ25 M (9). In vivo, -calpain seems to be activated at physiological Ca 2ϩ concentrations of 100 -300 nM, suggesting the involvement of other activators (10). These isoforms of calpain are heterodimers consisting of an 80 kDa catalytic large L-subunit and a 28 kDa small S-subunit, whose function probably consists in the regulation of the catalytic one (1). On the basis of amino acid homologies, the L-subunit has been described as comprising four domains, namely dI to dIV, while the S-subunit contains dV and dVI domains (11,12). Two subdomains dIIa and dIIb are distinguished within the dII domain, which shows some similarity to papain, while dIV and dVI are calmodulin-like domains containing EF-hands. A different domain classification, based on the crystal structure of rat calpain, has also been proposed (13). In the latter, the domain dI includes domain I and subdomain dIIa whereas dII corresponds to the sole subdomain dIIb of the previous nomenclature. Here, we use the second, structure-based, nomenclature accompanied by the first one in parentheses to facilitate reference to previous reports.
In the presence of Ca 2ϩ , calpain undergoes autoproteolysis (14). This reaction has been demonstrated to be inhibited by several inhibitors including leupeptin and trans-epoxysuccinyl-L-leucylamido-(4-guanidino)butane (E64). Previous studies indicate that the binding of both inhibitors requires the presence of Ca 2ϩ (9,(15)(16)(17). The kinetics of Ca 2ϩ binding and the onset of proteolytic activity in native calpain and its isolated subunits have been studied, and activity has been shown to appear only after a delay. This delay has been attributed to a Ca 2ϩ -induced conformational change making the active site of the enzyme accessible to the substrate (9).
It has been proposed that Ca 2ϩ causes the dissociation of the calpain heterodimer into the constitutive L-and S-subunits (16,18) thereby yielding the active form of the proteinase (19). However, whether the dissociation has to be considered as the early step in the calpain activation process is still under discussion. Indeed, Nakagawa et al. (20) suggested that the autolysis of the protein is necessary for the dissociation of m-calpain, whereas Pal et al. (21). observed calpain dissociation, although using the inactive C105S-80k/21k m-calpain form.
The crystal structures of Ca 2ϩ -free recombinant human and rat m-calpains have recently been determined at 2.3 and 2.6 Å resolution, respectively (12,13). In the absence of Ca 2ϩ , m-calpain displays a flat oval disk-like shape with the papainlike catalytic domains dI and dII (or dIIa and dIIb) located at one pole and the calmodulin-like domains dIV-dVI at the opposite pole. The analysis of the crystal structure reveals that some residues of the active site, which are crucial to the proteolytic activity of the protein, are not correctly positioned with respect to the substrate binding site in the unliganded state (12), suggesting the requirement for a structural rearrangement leading to the active form. The authors of the human m-calpain crystal structure postulated the existence of an electrostatic switch caused by the binding of Ca 2ϩ at the interface between dII (dIIb) and the negatively charged loop of dIII, a proposal supported by recent studies carried out on calpain mutants (22). More recently, Moldoveanu et al. (23) determined the crystal structure of a Ca 2ϩ -bound construct of -calpain comprising only the dI (dIIa) and dII (dIIb) domains. In this so-called minicalpain structure, Ca 2ϩ is seen bound to two non-EF-hand sites, one in each domain, thereby inducing a structural rearrangement of the active site leading to a correct positioning of the catalytic triad. Previous crystallographic studies on the Ca 2ϩ -free and Ca 2ϩ -bound isolated domain VI homodimer detected very small conformational changes at the level of the S-subunit upon Ca 2ϩ -binding (24 -26). However, direct experimental evidence of the overall structural modifications occurring in the complete calpain in solution during its activation is still lacking.
We have therefore undertaken a study of the effect of Ca 2ϩ and inhibitors of calpain on the enzyme conformation using small angle x-ray scattering (SAXS) 1 in solution. This method has been widely used to determine low resolution protein structures and conformational changes and is very sensitive to modifications of the state of aggregation of proteins in solution (27,28).
Synchrotron radiation x-ray small angle scattering measurements were carried out on calpain isolated from human erythrocytes. Different solutions of the proteinase were analyzed both in the absence and in the presence of Ca 2ϩ . We also investigated the calpain structure in the presence of two different inhibitors, the synthetic inhibitor trans-epoxysuccinyl-L-leucylamido (4-guanidino)butane (E64), which is known to react with the -SH group at the active site, and the neutral serine and thiol protease inhibitor leupeptin, again in the absence or presence of Ca 2ϩ . The shape of the unliganded molecule fits well with the envelope of the crystal structure. While Ca 2ϩ triggers the aggregation of calpain, the molecule remains heterodimeric in solution in the additional presence of inhibitors, allowing us to detect and characterize a conformational change, which appears to be compatible with the mechanism put forward on the basis of the crystal structures.

EXPERIMENTAL PROCEDURES
Purification of Calpain from Human Erythrocytes-Calpain was purified from human erythrocytes according to Michetti et al. (29) and dissolved in 50 mM sodium borate buffer, pH 7.5 containing 0.1 mM EDTA. Protein concentration was determined using the Bradford method (30). The effect of Ca 2ϩ was studied both in the presence and in the absence of 100 M E64 and leupeptin using the following five different Ca 2ϩ concentrations in excess with respect to the EDTA concentration (i.e. free Ca 2ϩ ): 10, 25, 100, 200, and 1000 M.
Scattering Experiments and Data Processing-SAXS measurements were performed using the synchrotron radiation beam line D24 at the DCI storage ring of LURE (Laboratoire pour l'Utilisation du Rayonnement É lectromagnétique, Orsay, France). The instrument, the data acquisition system (31), and the evacuated measuring cell (32) have already been described. To avoid any possible protein damage due to x-ray irradiation, the protein solution was slowly circulated in the measuring quartz capillary at the constant temperature of 20°C. In the course of several measuring sessions, samples from four different preparations were used. Freshly prepared samples of calpain were measured in the buffer used in the final step of purification. The monodispersity of each sample was checked immediately before SAXS measurements by SE-HPLC (Waters 486 System with a Shodex Protein KW-802.5 column). Eight frames of 100 s each were recorded using a position-sensitive proportional detector placed 1819 mm downstream from the sample so as to cover the range of momentum transfer q from 0.01 to 0.15 Å Ϫ1 (q ϭ 4sin/, where 2 is the scattering angle, is the radiation wavelength, ϭ 1.488 Å, absorption K-edge of Nickel). Frames were visually inspected to check for x-ray damage; none was found. Data were scaled to the transmitted intensity before computing the average and RMSD of each measurement and subtracting the scattering from the corresponding buffer.
The structural parameters characterizing calpain in solution were calculated using the following standard procedures. At very small angles, the intensity scattered by a particle can be approximated by a Gaussian curve (33) as in Equation 1, (Eq. 1) and the slope of a Guinier plot (ln[I(q)] versus q 2 ) readily yields the value of the radius of gyration R g of the particle and the value of the intensity at the origin I(0). This approximation is valid over a restricted q-range (typically R g qϽ1. 3). When put on an absolute scale, e.g. by means of a reference sample, I(0)/c (c, protein concentration of the sample) is proportional to the molecular mass m of the protein. Here, the calibration was performed using both the isolated subunit of the hemocyanin from Carcinus aestuarii (sedimentation coefficient 5 S, m ϭ 75 kDa) and its hexameric form (sedimentation coefficient 16 S, m ϭ 450 kDa) (34). The distance distribution function p(r) corresponds to the distribution of distances between any two volume elements within one particle. It has been determined using the indirect transform method as implemented in the program GNOM (35). This function provides an alternative estimate of the radius of gyration derived through the relationship shown in Equation 2.
Scattering intensities were computed from the atomic coordinates of the crystal structure of human m-calpain (1kfu.pdb) by using the program CRYSOL, which takes into account the hydration water by introducing a 3 Å thick border layer surrounding the molecule (36). The calculated scattering profile is fitted to the experimental pattern using only two adjustable parameters, the excluded volume of the particle V and the electron density in the border layer b , to minimize the discrepancy as shown in Equation 3, where N is the number of experimental points and I e (q i ) and (q i ) denote the experimental scattering curve and its RMSD, respectively. The ab initio shape determination was performed with the dummy atom model (DAM) method (37) using the program DAMMIN running on a Silicon Graphics O2 work station. A sphere of diameter D max is filled by closely packed small spheres (dummy atoms) of radius r 0 Ͻ ϽD max . The DAM structure is defined by a configuration vector X with N Ϸ (D max /r 0 ) 3 components. Using simulated annealing, the program searches for a configuration that fits the experimental data while a looseness penalty ensures the compactness and connectivity of the solution (37). No particular condition of oblateness of the particle shape was imposed as constraint in these calculations. The resulting shape together with the high resolution crystal structure were displayed and manipulated using the graphical software package ASSA (38). The superimposition of the pdb structures with the three-dimensional models of the protein obtained with the DAM method was performed using the program SUPCOMB (39). The program CRYSOL was used to calculate the SAXS pattern from the atomic coordinates of Ca 2ϩ -free human m-calpain (1kfu.pdb), which is shown in Fig. 2 together with the experimental scattering curve. The calculated curve, which corresponds to a dry excluded volume of 122 nm 3 and to an electron density contrast in the solvation shell ␦ b ϭ 44 e/nm 3 (i.e. density of the solvent in the shell 1.13 g/cm 3 ), neatly fits our experimental data ( ϭ 1.395). In parallel, an ab initio determination of the overall shape of the unliganded -calpain from the SAXS pattern was performed using the program DAMMIN (see "Experimental Procedures"). Ten independent calculations were performed and yielded very similar shapes. A typical result is presented in Fig. 3 superimposed on the crystal structure of the human m-calpain. The shape provides a satisfactory fit envelope to the crystal structure.

Structural Characterization of Unliganded Calpain in
The observations regarding the scattering patterns and the models indicate that the overall organization of both isoforms of the proteinase are very similar. This is not unexpected, considering that the high resolution structure of m-calpain from rat 2 (13) is very similar to its human counterpart 3 (12) with an RMSD over 800 C ␣ atoms of 1.55 Å. It also demon-strates that the overall shape of the protein does not undergo any significant distortion inside the crystal lattice, a non-trivial result in the case of a multidomain protein for which significant differences between the structure in solution and in the crystal may be observed (40).
Effect of Ca 2ϩ Addition--Calpain solutions were studied in the presence of variable amounts of Ca 2ϩ . At 10 and 25 M free Ca 2ϩ , the scattering curves were identical to that obtained in the absence of Ca 2ϩ (data not shown). In the presence of 100 M (and 200 M) free Ca 2ϩ , the SAXS pattern displays conspicuous differences from that of the unliganded protease (Fig. 4, dashed  line), most notably a marked upward curvature at small angles, which is an unambiguous indication of the formation of large soluble aggregates with a broad size distribution. The addition of 1 mM free Ca 2ϩ causes the formation of amorphous precipitates. Furthermore, preliminary experiments show that the aggregation process observed at 100 M Ca 2ϩ is essentially reversed upon addition of an excess of EDTA or in the presence of 100 mM NaCl, KCl, or NaSCN, confirming the observations of Pal et al. (

Conformational Change upon Ca 2ϩ Addition in the Presence
of Inhibitors-Although aggregation is thought to originate from a conformational change triggered by Ca 2ϩ , its effect dominates the scattering pattern at small angles to such an extent that this putative modification in shape cannot even be detected. To investigate whether the aggregation phenomenon can be related to the autoproteolysis process, we resorted to inhibitors of calpain. E64 is a specific inhibitor for cysteine proteases, which irreversibly binds to the SH group at the active site forming a thioether linkage. In the case of calpains, this inhibitor is effective only in the presence of Ca 2ϩ (9,17), as is the competitive inhibitor leupeptin (15).
In the presence of either E64 or leupeptin inhibitor at a concentration of 100 M, the addition of 100 or 200 M free Ca 2ϩ does not cause the previously observed strong increase in the intensity at small angles. This means that no high molecular mass aggregates are formed. The Guinier analysis of each scattering curve yields a slightly larger value of the radius of gyration (R g ϭ 38.5 Ϯ 0.7 Å) with respect to the value calculated for the protein in the absence of Ca 2ϩ . Furthermore, the comparison of the pair distribution functions of the protein under these two experimental conditions (Fig. 5) shows a shift to higher distances in the position of the maximum of the p(r) function (r max ) by about 4 Å, while the maximum dimension of the particle (D max ) increases from 120 to 130 Å in the presence of Ca 2ϩ . The value of I(0)/c (where c is the protein concentration), and therefore the molecular mass of the protein, remains constant, which rules out any significant dissociation of calpain into its subunits. The absence of large aggregates suggests that the observed small conformational change induced by Ca 2ϩ binding in the presence of inhibitors, does not lead to the exposure of hydrophobic surfaces. In contrast, the addition of 1 mM Ca 2ϩ leads to the formation of amorphous precipitates, as already observed in the absence of inhibitor. Thus, in the presence of inhibitor, the solution of calpain remains monodisperse upon the addition of 100 or 200 M free Ca 2ϩ ; the protein does not dissociate either but undergoes a transition to a more extended conformation. As a control that this change in the scattering pattern was not due to a spurious effect of the inhibitor, scattering curves were recorded with solutions of human erythrocyte calpain containing 100 M of each inhibitor but no free Ca 2ϩ . The two curves are virtually indistinguishable from that of the unliganded molecule (data not shown).
The calpain aggregation process together with subunit dissociation have been systematically studied with the inactive C105S-80k/21k m-calpain from rat using light scattering (21).
In particular, using both Tris and HEPES buffers at a pH value around neutrality and with a protein concentration of 0.7 M, the Ca 2ϩ concentration required for inactive C105S-80k/21k m-calpain to aggregate ranges from 0.8 to 2 mM. These values are well in excess of the free Ca 2ϩ concentration of only 100 M at which we reproducibly observed the formation of large soluble aggregates, but the concentration of -calpain used in our SAXS measurements was 14.6 M (i.e. more than 20-fold that used in light scattering measurements). Since the subunit dissociation is a process dependent on protein concentration, ionic strength and salt concentration, a direct and detailed comparison between our results and the studies of Pal et al. (21) is hindered by the differences between proteins (i.e. human -calpain versus rat m-calpain), by the large differences in the concentrations used, and by the different sensitivity of the various methods to the presence of small amounts of oligomers. In the conclusion to the study by Pal et al. (21), the subunit dissociation is described as a dynamic process triggered by Ca 2ϩ , and the aggregation is said to be most likely caused by the exposure of hydrophobic patches from the dimerization surfaces following dissociation.
Based on this view, a possible explanation for our observations suggests itself: the persistence of the heterodimeric form in the presence of Ca 2ϩ and inhibitors might reflect the influ- ence of inhibitors on the dynamic equilibrium of subunit association/dissociation triggered by Ca 2ϩ , which could be shifted toward association because of structural constraints imposed by inhibitor binding. However, while this proposal could apply in the case of the larger natural inhibitor calpastatin, which binds to both subunits and thereby stabilizes the heterodimer (21,41), it seems of little relevance in the case of the binding of the small inhibitors used in our work. Indeed, the crystal structures of complexes of papain with E64 (42) and leupeptin (43) show both inhibitors inserted in the active site, far from the interface between L-and S-subunits. Assuming that inhibitors bind in a similar way to calpain, their influence on the subunit association could only be indirect and most likely limited.
An alternative explanation for the persistence of the heterodimeric form in the presence of Ca 2ϩ and inhibitors is that under our experimental conditions (i.e. Ca 2ϩ /calpain molar ratio of ϳ7-14), Ca 2ϩ binding causes the small conformational change observed by SAXS with no detectable effect on the dynamic dissociation/association equilibrium and that the alteration of the latter can only occur at higher Ca 2ϩ concentration (precipitates are observed at 1 mM also with inhibitors) or following autoproteolysis in the absence of inhibitors. It seems plausible that the conformational change observed by SAXS in the -calpain from human erythrocytes in the presence of inhibitors and 100 M Ca 2ϩ also occurs in the case of the inactive C105S-80k/21k m-calpain but at a Ca 2ϩ concentration lower than that required for aggregation as determined by light scattering (21). This proposal could be established using a technique sensitive to conformation changes such as SAXS. Nevertheless, our results are already supported by the observation that the inactive mutant form of m-calpain C105S-m80k/30k does not dissociate in the presence of Ca 2ϩ as recently reported by Nakagawa et al. (20).
The results of ten independent ab initio analyses, carried out using the program DAMMIN on the SAXS data of the proteinase in the presence of 100 M E64 and 100 M free Ca 2ϩ , yield similar low resolution models. A typical model is reported in Fig. 6, model C. As already suggested by the observed increase of both the r max and the D max values from the p(r) analysis, the resulting shape appears to be more elongated than the model obtained for the native calpain (Fig. 6, model B). Most noticeable is the appearance of a protuberance at one end of the particle separated from the rest of the molecule by a slight constriction like a neck barely visible or absent on the models for the unliganded enzyme.
The crystal structure of the unliganded enzyme shows that the two domains dI and dII (or dIIa and dIIb) contain the catalytic triad responsible for the protease activity and are located at one end of the molecule (12,13). The active site is however split in two parts, one on each domain, and a mutual movement bringing the two domains together is required to activate the enzyme, an event triggered by Ca 2ϩ binding. Using a I-II minicalpain, which could be crystallized in the presence of Ca 2ϩ , a conformational change has been recently observed that brings the two domains dI (dIIa) and dII (dIIb) together and the active site residues in a suitable position for catalysis (23). When we consider this process in the context of the entire heterodimer, a more compact protuberance would be formed at the upper end of the proteinase, in agreement with our results. Moreover, at the opposite end of the protein our ab initio model of the Ca 2ϩ -bound calpain does not display significant differences with the model obtained for the unliganded form (Fig. 6,  C versus B and A). These results are in agreement with the crystallographic studies carried out on the Ca 2ϩ -free and Ca 2ϩbound isolated domain VI homodimer of the S-subunit that detected very localized conformational changes in this region of the proteinase (25), which would go unnoticed using SAXS (24,26).
In conclusion, our results constitute the first direct experimental evidence that Ca 2ϩ binding, in the presence of inhibitor, triggers a conformational transition of the undissociated heterodimeric form of calpain. Although at low resolution, our model is compatible with the already observed dI-dII (dIIa-dIIb) movement (23). Indeed, inhibitors which cannot bind in the absence of Ca 2ϩ do so irreversibly with high reactivity (E64) (9) or with high affinity (leupeptin, -calpain K i ϭ 3.2 ϫ10 Ϫ7 M, m-calpain K i ϭ 4.3 ϫ10 Ϫ7 M) (15) in the presence of Ca 2ϩ , which could imply that the active site undergoes a conformational change leading to the active conformation as already suggested by Crawford et al. (17). This result, together with previous studies discussed above, suggests that in the presence of Ca 2ϩ and inhibitors, human -calpain could also adopt an active conformation without subunit dissociation. Clearly, this proposal can only be validated by the high resolution structure of our complex. Precisely, the availability of monodisperse solutions of Ca 2ϩ -bound -calpain open the way to the crystallization and subsequent high resolution structural analysis of the intact heterodimeric calpain in the presence of Ca 2ϩ . The conformational change observed in the presence of Ca 2ϩ and inhibitors could be a likely requirement for the interaction with the natural inhibitor, calpastatin (14,44), to which we are currently extending our studies.