Dissociation and aggregation of calpain in the presence of calcium.

Calpain is a heterodimeric Ca(2+)-dependent cysteine protease consisting of a large (80 kDa) catalytic subunit and a small (28 kDa) regulatory subunit. The effects of Ca(2+) on the enzyme include activation, aggregation, and autolysis. They may also include subunit dissociation, which has been the subject of some debate. Using the inactive C105S-80k/21k form of calpain to eliminate autolysis, we have studied its disassociation and aggregation in the presence of Ca(2+) and the inhibition of its aggregation by means of crystallization, light scattering, and sedimentation. Aggregation, as assessed by light scattering, depended on the ionic strength and pH of the buffer, on the Ca(2+) concentration, and on the presence or absence of calpastatin. At low ionic strength, calpain aggregated rapidly in the presence of Ca(2+), but this was fully reversible by EDTA. With Ca(2+) in 0.2 m NaCl, no aggregation was visible but ultracentrifugation showed that a mixture of soluble high molecular weight complexes was present. Calpastatin prevented aggregation, leading instead to the formation of a calpastatin-calpain complex. Crystallization in the presence of Ca(2+) gave rise to crystals mixed with an amorphous precipitate. The crystals contained only the small subunit, thereby demonstrating subunit dissociation, and the precipitate was highly enriched in the large subunit. Reversible dissociation in the presence of Ca(2+) was also unequivocally demonstrated by the exchange of slightly different small subunits between mu-calpain and m-calpain. We conclude that subunit dissociation is a dynamic process and is not complete in most buffer conditions unless driven by factors such as crystal formation or autolysis of active enzymes. Exposure of the hydrophobic dimerization surface following subunit dissociation may be the main factor responsible for Ca(2+)-induced aggregation of calpain. It is likely that dissociation serves as an early step in calpain activation by releasing the constraints upon protease domain I.

The classicaland m-calpains are cytosolic Ca 2ϩ -dependent cysteine proteases that are ubiquitously expressed and differ in their sensitivity to Ca 2ϩ . They consist of an isoformspecific catalytic ϳ80-kDa subunit (from the genes capn1 and capn2, respectively) and a common regulatory ϳ28-kDa subunit (capn4). Several other calpain-related genes are now known, but within this report, the word calpain refers only to theand m-calpains. Whereas the exact physiological roles of calpains remain to be defined, many studies suggest that they have important cellular roles. They have been implicated in several important cellular functions, including signal transduction, apoptosis, cell cycle regulation, and cytoskeletal reorganization (1)(2)(3)(4). Unlike many other cysteine proteases, calpains tend to cleave substrates at interdomain boundaries, thereby serving to modulate the function of these substrates rather than simply digesting them (5). Several probable substrates have been identified both in vitro and in vivo including p53, protein kinase C, spectrin, Ca 2ϩ -ATPase, talin, and fibronectin (4 -8).
Bothand m-calpain are absolutely dependent on Ca 2ϩ for hydrolysis of their substrates (9 -11). The initial effects of Ca 2ϩ binding to calpain include a conformational change that is essential to assemble the active site and some limited autolysis of both subunits. Further results of Ca 2ϩ binding include aggregation of the whole enzyme or continued autolysis and degradation. The recent structure determination of m-calpain (12,13) has provided new insights into the structural basis of calpain activation by Ca 2ϩ . In the absence of Ca 2ϩ , the catalytic triad is not assembled so that the protease is inactive. Several structural features have been identified that maintain the active site in an inactive conformation. These involve on one side contacts between the large subunit N-terminal peptide of domain I and domain VI of the small subunit and on the other side contacts between domains II and III of the large subunit. Some of the later interactions have been clarified by mutational analysis (14), but the functional implications of the binding of the N-terminal peptide to domain VI in relation to enzyme activation, autolysis, dissociation, and aggregation are not well characterized.
Although there is a clear difference in the Ca 2ϩ requirement for activation (250 -350 M for m-calpain and 10 -50 M for -calpain), the basis for this difference remains elusive. However, the activation of both m-and -calpain appears to be very similar, and it was proposed that one function of Ca 2ϩ was to cause the dissociation of the calpain subunits, a dissociation that may be reversible before autolysis but is promoted and presumably irreversible following autolysis (15)(16)(17). In two other reports, however, subunit dissociation was not detected (18,19). The inconsistent results stemmed from the complication of Ca 2ϩ -induced aggregation of calpain, and the difficulty in designing unambiguous experiments. In the presence of heavy precipitation, further studies of the calpain activation process will become extremely difficult. Consequently, several important aspects of the process of calpain activation and subsequent aggregation remain poorly understood.
In our attempts to crystallize either wild-type or inactive m-calpain (C105S-m-80k/21k) in the presence of Ca 2ϩ , crystals were formed under certain conditions that contained only the small subunit together with an amorphous aggregate enriched in the large subunit. We therefore used light scattering and analytical ultracentrifugation to search for conditions in which Ca 2ϩ -induced aggregation and possibly dissociation could be avoided. Aggregation of calpain has frequently been noted, but the factors involved have not been systematically studied (15,16,19,20). Subunit dissociation was also studied by the above methods and also by an improved subunit exchange experiment. The results confirmed unequivocally the occurrence of subunit dissociation in the presence of Ca 2ϩ and provided some new insight into the nature of calpain aggregation.

MATERIALS AND METHODS
Protein Expression and Purification-Both m-calpain and -calpain consist of an 80-kDa large subunit, which has an ϳ62% sequence identity between the isoforms, and a 28-kDa small subunit, which is identical in the two isoforms. Upon Ca 2ϩ -induced activation, calpain undergoes autolysis in both subunits. The natural rat calpain small subunit contains 270 residues, of which the N-terminal glycine-rich region of ϳ83 residues (domain V) is unstructured in the crystal structure (13) and is rapidly removed by autolysis (20). In this work, our recombinant calpains contain either a 21-kDa small subunit of 184 residues (domain VI), corresponding closely to the natural small subunit autolysis product (20,21), or an N-terminal His-tag version of this 21-kDa subunit containing 205 residues. The absence of domain V and the presence or absence of the N-terminal His-tag do not affect the stability or Ca 2ϩ requirement of the enzyme. In this work, inactive recombinant calpains were used to eliminate the problem of autolysis. C105S-m-80k-CHis 6 /21k m-calpain and the 21-kDa small subunit were prepared as described previously (21) and stored in snap-frozen aliquots at a concentration of 10 -20 mg/ml.
A hybrid calpain large subunit was constructed by means of engineered BamHI and DraI restriction sites. The resultant cDNA codes for m-calpain residues 1-48, -calpain residues 59 -648, and m-calpain residues 637-714 (including the C-terminal His-tag). 1 This large subunit formed an active calpain entitled m-Bam--Dra-m-80k/21k and very similar to wild-type -calpain when coexpressed with the 21-kDa or NHis 10 -21-kDa small subunit. In its active form, m-Bam--Dra-m-80k/21k is almost indistinguishable from -calpain on column chromatography and casein zymography, and its Ca 2ϩ requirement of 120 M is close to that of -calpain (ϳ25 M) (m-calpain, 325 M). The same hybrid calpain large subunit was also prepared with the inactivating mutation C115S.
A recombinant form of rat calpastatin domain I with a C-terminal His-tag containing a total of 140 amino acid residues was also cloned, expressed, and purified as described previously (22).
Analysis of Crystal Content-Crystals were obtained within 3-7 days in several conditions together with substantial amounts of amorphous aggregate, and the two solids could not be readily separated. The whole droplet was transferred to an Eppendorf tube and centrifuged. The clear supernatant (mother liquor) was removed, and the solid phase was washed repeatedly with crystallization buffer. The solid phase was then treated briefly with 20 l of distilled water at room temperature, which instantly dissolved the crystals but did not significantly dissolve the amorphous precipitate. This residual precipitate was dissolved in SDS gel sample buffer, and the calpain subunit content of all the samples was analyzed by SDS gel electrophoresis.
Light Scattering-Light scattering by C105S-m-80k-CHis 6 /21k calpain was observed in a Perkin-Elmer LS50B spectrophotometer at room temperature. Both excitation and emission wavelengths were set at 320 nm, and the time-dependent change in scattering intensity was recorded. The solutions contained 70 -200 g/ml of calpain, 0 -0.2 M NaCl, 10 mM dithiothreitol, and 20 -330 mM buffer, in a total volume of 3.0 ml. The buffer system was varied to alter the pH, and the CaCl 2 concentration was varied from 0.1-50 mM.
Dynamic Light Scattering-The volume of 3.0 ml used for conventional light scattering required too much material for studies at a higher protein concentration. Therefore, dynamic light-scattering experiments were performed at 20°C using the DynaPro-MS/X instrument (Protein Solutions. Charlottesville, VA), which has an operating volume of 12 l. For these experiments, protein solutions in the range of 1-5 mg/ml in 50 mM Tris-HCl, pH 7.6, containing 50 mM CaCl 2 and 0.2-1.0 M NaCl were used. The monodispersity or polydispersity of the solutions was assessed, and the molecular weight(s) of the predominant species were calculated.
Analytical Ultracentrifugation-Sedimentation velocity experiments were carried out with m-calpain at 20°C in a Beckman XL-I analytical ultracentrifuge using absorbance optics following the procedure described by Laue and Stafford (23). Protein solutions were exhaustively dialyzed against 50 mM Tris-HCl, pH 7.6, containing 200 mM NaCl, 2 mM tri-(2-carboxyethyl)-phosphine, and either 5 mM EDTA or 5 mM CaCl 2 . Runs were performed at a protein concentration of 1.13 mg/ml both in the presence and absence of Ca 2ϩ at either 50,000 or 60,000 rpm for ϳ3 h, during which time a minimum of 50 scans was taken to monitor the sedimentation rate of the protein. For sedimentation equilibrium, experiments were performed at three different protein concentrations of 0.09 -0.3 mg/ml and at a minimum of two different speeds between 6000 -12,000 rpm. Each speed was maintained until there was no significant difference in r 2 /2 versus absorbance scans that were taken 2 h apart.
Subunit Exchange-To observe subunit interchange, the inactive mutants ofand m-type calpains, m-Bam-C115S--Dra-m-80k/ NHis 10 -21k and C105S-m-80k/21k, were dialyzed at 4°C in buffer A (50 mM Tris-HCl, pH 7.6, 2 mM EDTA, 10 mM 2-mercaptoethanol, 0.1% (v/v) Triton X-100). Equal amounts (0.75 mg each) of the two proteins were mixed and resolved by chromatography on a Bio-Rad UnoQ (1 ml) ion exchange column with a gradient of [NaCl] in buffer A. An identical mixture at 20°C was adjusted to 0.2 M NaCl, incubated with 5 mM net Ca 2ϩ for 30 min, and then quenched with excess EDTA. The mixture was dialyzed against buffer A overnight at 4°C and resolved on the UnoQ column as described above. The protein content of the fractions was analyzed by SDS gel electrophoresis, Coomassie Blue staining, and by immunoblotting.

RESULTS
Crystallization-Following the preincubation of C105S-m-80kCHis 6 /21k in the presence of 10 mM CaCl 2 , clusters of needle-shaped crystals appeared in droplets containing either 50 mM sodium acetate, pH 5.0 -5.5, or 50 mM MES, pH 6.5, in addition to 2 M ammonium sulfate and 5% isopropyl alcohol. Crystals of the same morphology appeared, although less reproducibly in 50 mM MES, 1.6 M ammonium sulfate, 10% isopropyl alcohol. In a different experiment, very thin hexagonal plate crystals were obtained from 100 mM MES, pH 6.5, 20% polyethylene glycol 5000 monomethyl ether, and 20% isopropyl alcohol. Under all these conditions, no crystals were obtained from calpain that had not been preincubated with CaCl 2 .
Crystals of either morphology diffracted x-rays to ϳ3.0 Å at a rotating anode x-ray source. However, further crystallographic characterization was not possible because of twin crystal formation. SDS gel electrophoresis showed that the crystals obtained from these trials contained only the 21-kDa small subunit. The mother liquor showed faint traces of the 21-kDa protein, and the amorphous precipitate contained the 80-kDa large subunit together with a very small amount of the 21-kDa protein (Fig. 1).
Aggregation and Calpastatin-Calpain Complex Formation-At low ionic strength, calpain aggregated on the addition of sufficient Ca 2ϩ , and this could be reversed rapidly and completely by excess EDTA and more slowly reversed by raising the NaCl concentration (Figs. 2, a and b). In buffers containing 70 g/ml calpain and 0.2 M NaCl, no detectable increase in light scattering was caused by the addition of Ca 2ϩ . Aggregation at low ionic strength was also prevented by the presence of calpastatin, and in this case, the small increase in light scattering was assumed to mark the onset of the calpastatin-calpain complex formation (Fig. 2c). The calpastatin used here was a recombinant form of rat calpastatin domain I containing a total of 140 residues (22). Table I lists the Ca 2ϩ requirements for these two different events and shows that aggregation of C105S-m-80k/21k calpain was dependent on the ionic strength and pH of the buffer as well as on the Ca 2ϩ concentration. The Ca 2ϩ concentration required in any given buffer for calpastatin-calpain complex formation was significantly higher than that required for aggregation (24).
Dynamic light-scattering experiments were conducted at higher calpain concentrations than the light-scattering work in order to approach conditions relevant to crystallization trials. They showed that a solution of 1 mg/ml of m-calpain in 50 mM Tris-HCl, pH 7.6, was monodisperse in the absence of Ca 2ϩ . The major species had a molecular mass of 95-105 kDa, and ϳ5% of the protein was present as high molecular mass species in the range of at least 10 3 kDa. At this low ionic strength, the solution became polydisperse upon the addition of Ca 2ϩ but remained monodisperse in 0.5 M NaCl with Ca 2ϩ . With 3-5 mg/ml calpain in 1 M NaCl in the presence of Ca 2ϩ , 50-70% of the scattering intensities were contributed by the monomer of ϳ100 kDa, and the remainder were provided by components of the order of 10 3 -10 5 kDa. The data did not indicate the presence of the homodimer of 21 kDa.
Ultracentrifugation Studies-Sedimentation studies also provided clear evidence of aggregation. Protein samples (1.13 mg/ml) gave rise to a sedimentation coefficient value (s 20,w 0 ) of 5.93 in the absence of Ca 2ϩ , whereas in the presence of Ca 2ϩ the coefficient value was 17.79. In the sedimentation equilibrium experiment, m-calpain (0.09 -0.3 mg/ml) in the absence of Ca 2ϩ showed a single molecular species of 100 kDa, but in the presence of Ca 2ϩ , multiple species were observed with molecular masses ranging from 276 to 600 kDa, suggesting the presence of a mixture of aggregates of 4 -8 calpain molecules or calpain large subunits. The data again did not provide evidence for the existence of the isolated small subunit. As a control, the isolated 21-kDa small subunit protein, which is known to exist as a homodimer (25), was analyzed at the same time under the same conditions. In the absence of Ca 2ϩ , it was monodisperse with an apparent molecular mass of 40 kDa. In the presence of Ca 2ϩ , it was polydisperse with a dominant species of an apparent molecular mass of 47.5 kDa.
Subunit Dissociation-The subunit exchange experiment depended on two factors, the separation by ion exchange chromatography of the -calpain-like m-Bam-C115S--Dra-m-80k/ NHis 10 -21k calpain from the m-type C105S-m-80k/21k calpain and the clear distinction on gel electrophoresis between the two different small subunits with and without the NHis 10 peptide. Buffers containing 0.2 M NaCl and 0.1% Triton X-100 were used for incubation with Ca 2ϩ to minimize the formation of insoluble aggregates.
The chromatograms (Fig. 3) show thatand m-calpain peaks, which were well separated before exposure to Ca 2ϩ , were recovered in the same positions after a 30-min exposure to Ca 2ϩ followed by quenching in EDTA and extensive dialysis. The peaks were less sharp following Ca 2ϩ treatment, and the yield of protein was ϳ60%, suggesting some degree of aggregation and some failure to reform heterodimers. The Coomassie Blue-stained gel (data not shown) and the immunoblot (Fig. 4) showed clearly that some NHis 10 -21k small subunit had been transferred from the -type calpain to the m-calpain, and conversely that some 21-kDa small subunit had been transferred from the m-calpain to the -type calpain. The conclusion is inescapable that in these inactive calpains, which cannot undergo autolysis, subunit interchange had occurred so that some large and small subunits must have dissociated in the presence of Ca 2ϩ and at least to some extent reversibly reassociated upon the removal of Ca 2ϩ . Densitometry was not performed, but the extent of exchange appears from the immunoblot to be of the order of 30%. The nature of this experiment with a column step of relatively low recovery does not permit precise quantification of the extent of subunit exchange. DISCUSSION The impetus for this work was the need to crystallize calpain in the presence of Ca 2ϩ . The structure of calpain in the absence of Ca 2ϩ (12,13) showed that the active site was not assembled to a catalytically active conformation, so that it was clearly of interest to solve the structure in the presence of Ca 2ϩ to understand the mechanism of calpain activation. Not unexpectedly, however, crystallization in the presence of Ca 2ϩ raised problems of aggregation and subunit dissociation. The suggestion that the calpain subunits dissociate in the presence of Ca 2ϩ has been the subject of some debate (15)(16)(17)(18)(19) arising not least from the difficulty of designing definitive experiments.
Here we provide evidence that clearly confirms the phenomenon of subunit dissociation first described by Yoshizawa et al. (15) in 1995. The results also suggest that in m-calpain and in the absence of autolysis, the aggregation/dissociation is a highly reversible process. The dissociation is normally not complete unless driven by other factors. Extrapolation of these clear in vitro results to the cell is necessarily speculative. It seems highly probable that the subunit dissociation is an essential aspect of calpain activation in vivo, but the ensuing aggregation observed in vitro is less likely to be relevant in vivo where calpain is diluted in a highly proteinaceous environment and where autolysis rapidly removes the activated calpain.
Calpain aggregation is clearly a function of protein concentration, ionic strength, and Ca 2ϩ concentration, but some new insights into the aggregation were obtained from light-scattering and ultracentrifugation experiments. First, within the experimental time frame (ϳ1 h) and in the absence of autolysis, aggregation is an equilibrium process, because it could be fully reversed by the addition of EDTA. Second, the inhibition of aggregation by higher salt concentration strongly suggests that FIG. 3. Fast Protein Liquid Chromatography separation of calpain isoforms on a 1 ml UnoQ (Bio-Rad) quaternary ion exchange column. The earlier peak contained m-Bam-C115S--DraI-m-80k/NHis 10 21k calpain, and the later peak contained C105S-m-80k/21k calpain. The column was run in 50 mM Tris-HCl, pH 7.6, 2 mM EDTA, 0.1% (v/v) Triton X-100, 10 mM 2-mercaptoethanol (buffer A) with a gradient of increasing NaCl concentration. a, a mixture of the two proteins resolved without previous exposure to Ca 2ϩ . b, a mixture of the two proteins resolved following exposure to 5 mM Ca 2ϩ for 30 min, quenching with excess EDTA, and dialysis overnight against buffer A. FIG. 4. Immunoblot analysis of the eluted peaks shown in Fig.  3. Fractions spanning the two main eluted peaks in each column were analyzed by means of immunoblotting. The left-hand section of the blot contains samples from the column (see Fig. 3a) previous to Ca 2ϩ treatment, and the right-hand section contains samples from the column (see Fig. 3b) following exposure to Ca 2ϩ . The upper portion of the blot was treated with a mixture of a polyclonal antibody to rat m-calpain large subunit and a monoclonal antibody to human -calpain large subunit, which cross-reacts with rat -calpain. The lower portion of the blot was treated with a polyclonal antibody to the rat calpain small subunit. Both blots were treated with appropriate second antibodies and then developed by enhanced chemiluminescence. It is important to note the appearance of the NHis 10 21k subunit in the m-calpain peak and the corresponding appearance of the 2kDa subunit in the -calpain peak only after Ca 2ϩ treatment as indicated by open arrows. hydrophobic interactions are involved. Although aggregation of calpain at Ͻ200 g/ml appeared to be suppressed by 0.2 M NaCl, both dynamic light scattering and ultracentrifugation run at Ն 1-mg/ml calpain showed that soluble high molecular weight oligomers were formed in the presence of Ca 2ϩ , which could only partly be suppressed by 0.5 or 1 M NaCl. Both of these methods failed to detect the presence of the isolated 21-kDa homodimer, which was anticipated as a result of subunit dissociation. Previous work on the 21-kDa subunit showed that it exists as a homodimer both in the presence and absence of Ca 2ϩ , and that the homodimer once formed is unlikely to dissociate in any of the conditions used here (25). The absence of free small subunits in the light-scattering and centrifugation experiments therefore suggests that subunit dissociation is far from complete in these conditions, and that the large soluble complexes are still composed predominantly of calpain (80 ϩ 21 kDa) heterodimers, which have undergone Ca 2ϩ -induced conformational changes leading to aggregation. It is possible to imagine a highly flexible intermediate form of the heterodimer in which the contact between the N-terminal peptide of the large subunit and domain VI in the small subunit has been lost, whereas the subunits remain at least partially attached through residual contacts between domains IV and VI. A third aspect of the aggregation studies was the inhibition of aggregation by a molar excess of a 140-residue domain of calpastatin. Separate sections of calpastatin are known to bind to domain IV in the large subunit and to domain VI of the small subunit (26). Therefore, it seems probable that calpastatin prevents calpain aggregation by binding to both subunits simultaneously and preventing subunit dissociation.
Whereas m-calpain dissociation appears to be incomplete in many buffer conditions, it is clear that calpain does indeed dissociate in the presence of Ca 2ϩ at pH 6.5 in the conditions prevailing in some crystallization droplets. In these precipitant conditions, the dissociated 21-kDa small subunit crystallized out, and its removal presumably promoted further subunit dissociation, leaving the large subunit to precipitate as an amorphous aggregate. Several crystallization conditions have also been found in which no visible aggregation of calpain occurs in the presence of Ca 2ϩ , but useful crystals, whether they are of the activated heterodimer or of the isolated large subunit, have not yet been obtained. Finally, the subunit exchange experiment also clearly demonstrated reversible subunit dissociation of calpain, because the interchange of small subunits between the m-and -type is only possible provided they dissociate. We had earlier failed to detect subunit dissociation by means of column chromatography of calpain in the presence of Ca 2ϩ and by a less definitive version of the subunit exchange experiment (19), but the present experiments provide unequivocal proof of subunit dissociation.
The importance of hydrophobic interactions and the effects of calpastatin support the idea that aggregation is a result of subunit dissociation. The crystal structure of calpain shows that the large and small subunits in the absence of Ca 2ϩ are bound mainly by interactions between the fifth EF-hand motif of each subunit (12,13) in a manner very similar to that shown for the 21-kDa homodimer (25,27). Dissociation must expose the complementary dimerization interface on both subunits, which is a large hydrophobic area of ϳ2780 Å 2 representing approximately one quarter of the surface area of domain IV and equally of domain VI (Fig. 5). Such an exposure is energetically unfavorable in aqueous solution and would be expected to promote either "correct" reassociation of the large and small subunits or random association leading to aggregation. The exposure of this surface would be even more disfavored in solutions of higher ionic strength, which explains the partial suppression of calpain aggregation in 0.2-1 M NaCl. Our modeling suggests that two large subunits could not dimerize at this site in domain IV in their correct relative orientation as observed in the domain VI homodimer because of steric interference by domains I-III. Thus, even partial exposure of this hydrophobic patch would promote the formation of randomly associated aggregates.
Based on the first x-ray structure of m-calpain (12), we proposed that the constraints imposed upon the protease domains I and II by the remainder of the molecule would act as a barrier in the activation of calpain. These constraints were provided on the one side by the small subunit domain VI, binding to the large subunit N-terminal peptide of domain I, and on the other side by a set of salt links between domains II and III. The release of the constraints would facilitate the assembly of the competent catalytic triad. We have reported experiments supporting the idea of the domain II/domain III interaction (14), and this work suggests that dissociation of the small subunit would release the constraint on domain I.
The evidence of aggregation, calpastatin complex formation, and subunit dissociation suggests that subunit dissociation is FIG. 5. Schematic representation to illustrate dissociation/aggregation pathway. The small subunit is presented as a stick model, and the large subunit is shown in the electrostatic surface representation. Red and blue colors highlight negative and positive charges, respectively. The green circle illustrates the central area of the exposed hydrophobic surface in large subunit that would initiate aggregation. The first step (a) of the pathway indicates subunit dissociation, which may be incomplete, and is completely reversible by EDTA. It is assumed that the existence of isolated monomeric dissociated subunits is transient and would be followed rapidly by step (b) in which large subunits and also Ca 2ϩ -bound heterodimers aggregate while isolated small subunits form a homodimer. The formation of the small subunit homodimer is thought to be essentially irreversible. The intermediate soluble aggregates will proceed irreversibly to step (c) to form insoluble aggregates. Upon Ca 2ϩ addition, there is also a possibility that the heterodimeric calpain may simultaneously aggregate (d) along with the association-induced aggregation (a-c). It appears likely that dissociation is physiologically important, because it serves as an early step in calpain activation by releasing the constraints upon protease domain I. The diagram was generated using GRASP (28). the principal cause of Ca 2ϩ -induced aggregation of calpain. Furthermore, we also show that both dissociation and aggregation are reversible. Our present understanding of these events is shown in Fig. 5. Ca 2ϩ -induced conformational rearrangement and partial dissociation leads to the formation of aggregates probably containing heterodimers as well as dissociated large subunits. This step is largely reversible, but the small subunit homodimer is probably no longer available for reassociation, and larger aggregates composed almost exclusively of large subunit will be formed almost irreversibly. In the case of active calpains, aggregation also occurs in vitro on initial exposure to Ca 2ϩ , but autolysis at the same time begins to degrade the large subunits to inactive fragments.