The N- and C-terminal autolytic fragments of CAPN3/p94/calpain-3 restore proteolytic activity by intermolecular complementation

Significance

CAPN3/p94/calpain-3, a calpain protease family member, has uniquely rapid and exhaustive autolytic activity. Here, we demonstrate that active protease core domain of CAPN3 is reconstituted by de novo intermolecular complementation (iMOC) between its two autolytic fragments. We also show that iMOC between one of the autolytic fragments and an unautolyzed full-length CAPN3 effectively generates autolytic activity. So far, CAPN3 is the only example of calpain that exhibits activity mediated by intermolecular complementation of its autolytic products. Mutations in the human CAPN3 gene cause limb-girdle muscular dystrophy type 2A (LGMD2A or calpainopathy). Understanding the genotype–phenotype links in the pathology of LGMD2A will be accelerated by incorporating iMOC-CAPN3 as one of active CAPN3 entities.

Abstract

CAPN3/p94/calpain-3, a calpain protease family member predominantly expressed in skeletal muscle, possesses unusually rapid and exhaustive autolytic activity. Mutations in the human CAPN3 gene impairing its protease functions cause limb-girdle muscular dystrophy type 2A (LGMD2A); yet, the connection between CAPN3’s autolytic activity and the enzyme’s function in vivo remain unclear. Here, we demonstrated that CAPN3 protease activity was reconstituted by intermolecular complementation (iMOC) between its two autolytic fragments. Furthermore, the activity of full-length CAPN3 active-site mutants was surprisingly rescued through iMOC with autolytic fragments containing WT amino acid sequences. These results provide evidence that WT CAPN3 can be formed by the iMOC of two different complementary CAPN3 mutants. The finding of iMOC-mediated restoration of calpain activity indicates a novel mechanism for the genotype–phenotype links in LGMD2A.

CAPN3 (also known as p94 or calpain-3) belongs to the calpain super family (EC3.4.22.18, clan CA, family C2) (1, 2). The consensus protease core sequence referred to as “CysPc” (cd00044 in the conserved domain database of the National Center for Biotechnology Information, NCBI) is shared by gene products belonging to the calpain super family. Because CAPN3 has calpain-type beta-sandwich (CBSW) and penta-EF-hand [PEF(L)] domains, it is classified as a “classical” calpain (3–5). Structural studies have revealed that calpain’s Ca²⁺-requiring protease activity is mediated by Ca²⁺ binding to the CysPc domain (6–10). This property has been difficult to capture biochemically for some calpains using conventional assay methods.

CAPN3 undergoes an apparently Ca²⁺-independent and exhaustive autolysis in several ex vivo protein-expressing systems, which has marked this calpain species as an intriguing yet intractable subject in the research field (11). CAPN3, on the other hand, can be stably expressed in skeletal muscle cells where endogenous CAPN3 is predominantly found (12). The autolytic activity of CAPN3 is dependent upon the presence of the CAPN3-specific insertion sequences, IS1 and IS2 (Fig. 1A) (11). The mechanism(s) by which these insertions transform CAPN3, which is otherwise quite similar to those of CAPN1 and CAPN2 with the exception of the interactions with the calpain small subunit (CAPNS1), into being highly autolytic, is not yet clear. Although it is known that CAPN3 avoids “random” autolysis in skeletal muscle by associating with specific binding proteins such as titin and PLEIAD/SIMC1, the general regulation of CAPN3 autolysis in vivo is still poorly understood (13–16).

Fig. 1.

Schematic representation of the CAPN3 constructs used in this study. (A) Diagram at Top illustrates the structure of CAPN3. Mutant constructs were designed to inactivate the protease activity and/or delete CAPN3-specific domain structures. PC1 and PC2, protease core domains 1 and 2; CBSW, calpain-type β-sandwich; PEF(L), penta-EF-hand in calpain large subunit; NS/IS1/IS2, CAPN3-specific insertion sequences; C129/H334/N358, active-site amino acid residues Cys, His, and Asn. e.g., “C > S” and “HN > AA” indicate amino acid substitutions C129 to Ser and H334/N358 to Ala, respectively. (B) Background of the hybrid hypothesis. The hybrid hypothesis is based on the previous finding that the compound heterozygosity of LGMD2A-associated mutations, G222R and R748Q, is associated with a less severe disease phenotype than that associated with homozygosity of the individual mutants (i and ii) (34). The two mutations attenuate and accelerate CAPN3 autolysis, respectively (iii and iv). Comparison of autolytic fragments expected in each genotype proposes hybrid intermolecular complementation (bidirectional arrows) as a means of restoration of WT activity (v). *Autolytic activity of CAPN3:R748Q also targets CAPN3:G222R.

After CAPN3 was identified as the gene responsible for limb-girdle muscular dystrophy type 2A (LGMD2A), various aspects of CAPN3’s biological function were intensively studied (17). Studies using transgenic knockout (KO) (18–23), as well as knockin (KI) mice (24, 25) demonstrated that defective CAPN3 protease activity is the primary cause for the disease. In vitro analyses of the CAPN3 missense mutations found in patients with LGMD2A suggested that both the loss of apparent Ca²⁺-independent autolyzing activity and, in contrast, the excess expression of autolyzing activity are associated with the improper function of the CAPN3 mutants (26). The biological significance of the peculiar autolytic activity associated with CAPN3 is yet to be defined.

It was reported that autolysis of CAPN3 in the IS1 region alters the accessibility of substrates as well as inhibitors to its protease core, indicating that autolysis is positively involved in the activation of CAPN3 (27). In addition, studies, including ours, suggested that the autolyzed fragments of CAPN3 remain associated as an active protease (27–29). Such an interaction was, however, considered temporary. Also, there is a notion that down-regulation of CAPN3 is an important outcome of autolysis based on the fact that the overexpression of wild-type (WT) CAPN3 in mouse skeletal muscle was not found to be harmful, whereas the overexpression of CAPN3:ΔIS1, whose autolytic but not proteolytic activity is attenuated (30), causes the development of a degenerating skeletal muscle phenotype (31). Thus, CAPN3’s extremely strong propensity to autolyze may contribute to the regulation of its protease activity in both positive and negative directions.

Possible interactions between two CAPN3 molecules have also been investigated. Studies focusing on homodimerization of CAPN3 have revealed the unique feature in the C-terminal PEF(L) domain (32, 33). With respect to the pathological mechanisms of LGMD2A, previously, we found that patients with LGMD2A carrying compound heterozygous mutations (two different missense mutations in CAPN3), may in some instances exhibit more benign symptoms than patients who carry any of the corresponding homozygous mutations (34). One of the mutations is located in the N-terminal protease core domain 1 (PC1), and reduces the autolytic activity, whereas the other is in the C-terminal PEF(L) domain, and accelerates autolysis. This discovery prompted us to consider that the coexpression of two distinct CAPN3 mutants may lead to functional complementation. Among several molecular mechanisms hypothesized, our “hybrid hypothesis” proposes that the limited autolysis of each CAPN3 mutant generates two autolytic fragments (Fig. 1B) that reconstitute the WT CAPN3 activity.

In the present study, we show that the protease activity of CAPN3 can be reconstituted by de novo intermolecular complementation (iMOC) between two autolytic fragments, suggesting that WT CAPN3 activity can be restored in individuals carrying compound heterozygous mutations if they are located in complementary positions. Surprisingly, the autolytic activity of inactive CAPN3 mutants such as CAPN3:C129S (FL:SHN) (Fig. 1A) was rescued by autolytic fragment-mediated iMOC. These results suggest that the protease core domain of CAPN3 can be derived as a complex of two peptide fragments, and that autolysis is, at least in part, mediated by chimeric iMOC between full-length (FL) and autolyzed fragments of CAPN3. These findings provide novel insights into the unique properties of CAPN3.

Results

CAPN3 Autolytic Fragments Reconstitute an Active Protease Complex Through iMOC.

As an early event of autolysis, CAPN3 is proteolyzed within or close to the IS1 region, generating N- and C-terminal fragments containing the active-site amino acid residues (aars), Cys129, and His334 and Asn358, respectively (Fig. 1A, N_31K and C_58K). “N_xxK” represents an N-terminal fragment containing PC1 and part of PC2, exhibiting a relative molecular mass (M_r) of xx kDa (“Nterm” inclusively refers to these fragments). “C_xxK” represents a C-terminal fragment encompassing part of PC2 and the rest of the C-terminal domain, exhibiting a M_r of xx kDa (“Cterm” inclusively refers to these fragments). To explore the possibility that two different CAPN3 mutants compensate for each other and reconstitute the active enzyme following autolysis, two nonoverlapping autolytic fragments of CAPN3 represented as the constructs N_31K (amino acid region 1–274) and C_58K (amino acid region 323–821) (Fig. 1A) were coexpressed to determine whether they had the propensity to generate protease activity similar to that of intact CAPN3. To detect proteolytic activity, C-terminal binding protein 1 (CTBP1), which functions as a substrate for CAPN3 (16) was used (Fig. 2A).

Fig. 2.

CAPN3 protease activity is reconstituted by iMOC between autolytic fragments. (A) Two constructs of CTBP1 were used to generate substrates for monitoring the protease activity of CAPN3. Bidirectional arrows indicate the fragments detected by antibodies. Bidirectional arrows (i) and (ii) show segments detected by anti-MYC before and after proteolysis; (iii) and (iv) show segments detected by anti-FLAG before and after proteolysis. (B) MYC-CTBP1 was coexpressed with various CAPN3 constructs in COS7 cells. A proteolytic fragment of CTBP1 (A) was detected when CAPN3:WT was expressed [lane 2, anti-MYC (ii)]. The combination of N_31K and C_58K also caused CTBP1 proteolysis (lane 3). Comparison of two theoretically inactive combinations, N_31K and C_58K:HN > AA (lane 4), and N_31K:C > S and C_58K (lane 5), indicated that endogenous COS7 protein(s) may compensate for the function of N_31K in complementing C_58K (*), but not vice versa. Decreased amount of N_31K and C_58K in lane 3 suggests that iMOC between the fragments leads to their proteolysis. Arrow and open arrowhead indicate the N-terminal (N_31K) and the C-terminal (C_58K) fragments of CAPN3. (C) The N-terminal region of CAPN2, CAPN2:N_27K, corresponding to N_30K, was able to substitute for N_31K in iMOC with C_58K. CTBP1 proteolysis was evaluated using the MYC-CTBP1-EGFP-FLAG construct (lane 4, anti-GFP). In the same sample, decreased amounts of CAPN2:N_27K and C_58K were also observed. (D) The combined expression of N_30K and C_58K:ΔIS2 led to the iMOC-induced proteolysis of MYC-CTBP1 (lane 5 vs. lanes 3 and 4). CAPN2:C_53K was unable to substitute for C_58K:ΔIS2 (lane 7). (E) Deletion of the C-terminal 7 aars from N_30K, in N_29K, abrogated its ability to complement C_58K, as assessed by MYC-CTBP1 proteolysis (lane 3). (F) Proteolytic activity generated by the various combinations of N- and C-terminal fragments from CAPN3 and CAPN2 are summarized. The NS, IS1, and IS2 domains were not required for the autolytic fragment-mediated iMOC. The N- but not the C-terminal fragment derived from CAPN2 was able to replace the corresponding region of CAPN3. n.d., not done. Open arrowhead indicates the C_58K fragment generated by autolysis. Thick arrows indicate the position of the expressed N-terminal fragment. Arrows in i–iv indicate the CTBP1 fragments shown in A.

When MYC-epitope tagged-CTBP1 (Fig. 2A, Upper Scheme) was coexpressed with CAPN3:WT in COS7 cells, a MYC-labeled proteolytic fragment was generated (Fig. 2B, lane 2, anti-MYC) (16). Similarly, the coexpression of N_31K and C_58K also resulted in the proteolysis of MYC-CTBP1 (Fig. 2B, lane 3). However, when the two catalytic amino acid residues in C_58K, His and Asn, were replaced by Ala, CTBP1’s proteolysis was suppressed (Fig. 2B, lane 4). These results showed that proteolytic activity was generated through cooperation of the catalytic amino acid residues residing in two independently synthesized protein fragments, and suggest that autolytic fragments from different CAPN3 molecules can reconstitute protease activity through iMOC. Further autolysis by reconstituted CAPN3 activity was also observed (Fig. 2B, anti-C3IS2, lane 3).

The N-Terminal Region of CAPN2 Can Undergo iMOC with the C-Terminal Autolyzed Fragment of CAPN3.

To determine whether the observed iMOC is CAPN3 specific, the CAPN2:N_27K fragment was used in place of CAPN3 N_31K. CAPN2:N_27K corresponds to CAPN3 N_30K, which is 7 aars shorter than N_31K and lacks the IS1 sequence (Fig. 1A). Surprisingly, CAPN2:N_27K was able to substitute for N_31K, resulting in CTBP1 proteolysis (Fig. 2C, lane 4 vs. lanes 2 and 3). We further verified that the reconstituted proteolytic activity was dependent on the active-site Cys105 residue of CAPN2 (lane 5) as well as that CAPN2:N_27K interacted with C_58K (Fig. S1A). Because CAPN2 lacks CAPN3-specific N-terminal insertion sequence (NS), this result indicates that both NS and IS1 in CAPN3 are dispensable for the iMOC mediated by autolytic fragments, and that the N-terminal domains from different calpain species can undergo iMOC.

The Specificity of CAPN3 iMOC Resides in Its C-Terminal Autolytic Fragment but Does Not Require IS2.

When N_30K was coexpressed with C_58K:ΔIS2 together with MYC-CTBP1 in COS7 cells, CTBP1 proteolysis was observed (Fig. 2D, anti-MYC; lane 5 vs. lanes 3 and 4). Therefore, in addition to IS1, IS2 was also dispensable for the iMOC-mediated protease activity of CAPN3. On the other hand, CAPN2:C_53K, which corresponds to CAPN3 C_58K:ΔIS2, did not undergo detectable iMOC with N_30K, as evaluated by CTBP1 proteolysis (Fig. 2D, lanes 6 and 7). These results indicate that the iMOC-mediated protease activity of CAPN3 was dependent on sequences in its C-terminal region, but that IS2 was not required. Although the C-terminal region corresponding to CAPN3 C_58K is less conserved than the N-terminal region, it is still highly similar to the corresponding regions in CAPN1 and -2 (Fig. S2B). We could not identify any specific C-terminal sequences that were unique to CAPN3, indicating that the C-terminal specificity that contributes to CAPN3-mediated iMOC may be attributed to the secondary structure of this region.

β-Strand Structures in PC2 Contribute to the Formation of iMOC-CAPN3.

To gain further insight into the molecular mechanism of iMOC, we referred to the structure of the Ca²⁺-bound form of rat CAPN2/CAPNS1. In PC2, an antiparallel β-sheet was formed by the insertion of β-strand 6 (β6) between β1 and β2 (Figs. S2 and S3) (6). In a previous study, it was suggested that the corresponding structure in PC2 of CAPN3 supports the association of autolytic fragments after autolysis in IS1 (27). Thus, we examined the contribution of the β-strands in PC2 to iMOC-mediated CAPN3 protease activity using N_29K, which lacks these structures (Fig. 1A and Fig. S3). Because N_29K coexpressed with C_58K did not reconstitute CTBP1 proteolysis (Fig. 2E, lane 3) or coimmunoprecipitate with C_58K (Fig. S1B), we concluded that the β2 strands in the PC2 domain of CAPN3 were required for the interaction of the autolyzed fragments.

iMOC-Generated CAPN3 Exhibits Substantial Autolysis.

To compare cellular targets of WT and iMOC-mediated CAPN3 protease activity, proteomic differences in COS7 cells expressing either form of CAPN3 were analyzed using reagents for isobaric tagging to determine the relative and absolute amounts quantitatively (iTRAQ). This experiment is anticipated to find possible substrates differentially proteolyzed by two forms of CAPN3. As a protease-inactive control for FL-CAPN3, FL:SHN has been sufficient; however, for the iMOC-CAPN3, all three catalytic amino acids were replaced with Ala to minimize nonspecific activity.

Under both conditions, among the proteins identified with more than five peptides, CAPN3 itself and coexpressed CTBP1 were significantly down-regulated (Fig. 3). In contrast, calpastatin (CAST), a previously established substrate for CAPN3, was efficiently more down-regulated only in cells expressing FL-CAPN3 (1.61 vs. 1.14). In COS7 cells, however, it was suggested that complementation generates activity at an efficiency of less than 5% of FL-CAPN3 activation (Fig. S4). Therefore, the possibility that CAST is a differential substrate candidate for FL- and iMOC-CAPN3, at least in vitro, was explored in a different approach, which is described below.

Fig. 3.

Autolysis is the major proteolytic event in COS7 cells expressing iMOC-CAPN3. Proteomic analyses of COS7 cells expressing CAPN3 were performed on two different experimental sets. In each set, protein expression was compared between cells expressing either WT or inactive mutant (Mut) CAPN3. As a protease-inactive control for FL-CAPN3, FL:SHN has been sufficient; however, for the iMOC-CAPN3, all three catalytic amino acids were replaced with Ala to minimize nonspecific activity. The experimental procedure is also schematically shown in Fig. S6. pep no., number of peptides with a confidence level >95%.

CAPN3 Fragments Separately Translated in Vitro Undergo iMOC and Proteolyze CAST in a Na⁺-Dependent Manner.

To further determine whether CAST is a differential substrate for FL- and iMOC-CAPN3, an in vitro translation (IVT) system using wheat germ extract, in which cotranslation guarantees 100% coexpression of proteins, was used to produce the autolytic fragments of CAPN3. When the complementary fragments of CAPN3 were cotranslated, the interaction between N_31K or N_31K:C > S and C_58K, was detected. The resulting autolytic activity was found to be dependent on the active-site Cys aar, confirming the results of the COS7 cell experiment (Fig. 4A). In addition, it was demonstrated that CAST is more efficiently proteolyzed by FL-CAPN3 than iMOC-CAPN3 (Fig. S5 B and C).

Fig. 4.

In vitro translated CAPN3 fragments form active protease complexes de novo. (A) iMOC mediated by in vitro translated fragments of CAPN3. Cotranslation of the CAPN3 fragments N_31K and C_58K produced autolytic activity (lane 3 vs. lanes 1, 2, and 4; bracket and gray arrowhead). Both N_31K and N_31K:C > S were coimmunoprecipitated with the C_58K fragment using the anti-C3IS2 antibody (lanes 7 and 8, anti-C3NS). Autolytic fragments of N_31K were also detected in the immunoprecipitant (lane 7, bracket). *Nonspecific signal. (B) Posttranslational formation of iMOC-CAPN3. Separately translated products were mixed, and the proteolytic activity was evaluated using Cy3-labeled cast-d1 as a substrate. (C) Proteolysis of Cy3-cast-d1 was detected in the FL (i) or iMOC (ii), but not in the iMOC_{C > S} reaction (iii) (Cy3-image, lanes 2 and 5 vs. lane 8). In both i and ii, cast-d1 proteolysis and CAPN3 autolysis was increased in the presence of Na⁺ (lanes 3 and 6 vs. lane 8). A total of 5 mM EDTA/K inhibits translation; thus, translation during the iMOC incubation was considered to be negligible. *Carry over from the adjacent lane (lane 7).

The proteolysis of CAST was also assayed using a Cy3-labeled domain 1 fragment of CAST (Fig. 4B, Cy3-cast-d1). Because the cotranslation of iMOC-CAPN3 resulted in significant autolysis (Fig. 4A, lane 3), separately translated CAPN3 fragments were mixed immediately before their incubation with Cy3-cast-d1 (Fig. 4B). In the presence of 5 mM EDTA/K, which was sufficient to inhibit the translation reaction (Fig. S5D), cast-d1’s proteolysis was detected in the FL and iMOC samples, but not in the iMOC_{C > S} sample (Fig. 4C, lanes 2 and 5 vs. lane 8). Consistent with the fact that CAPN3 is a Na⁺-dependent protease (35), these activities were accelerated by the presence of 0.1 M NaCl (Fig. 4C, lanes 3 and 6 vs. lane 9). Under these conditions, the autolysis of CAPN3 also proceeded in both the FL and WT iMOC samples (Fig. 4C, lanes 2, 3, 5, and 6), but not in the iMOC_{C > S} samples (Fig. 4C, lanes 8 and 9). These results demonstrated that iMOC-CAPN3 retained the characteristics of FL-CAPN3, including the proteolysis of CAST, autolytic activity, and the propensity for increased proteolysis in the presence of Na⁺.

N- and C-Terminal Autolytic Fragments of CAPN3 Are Not Targeted to Myofibrils.

In transient expression experiments using primary skeletal myotube cultures, CAPN3 is found to be predominantly localized to the sarcomeric M-line region (12). To examine the effect of autolysis on cellular localization, the N- and C-terminal autolyzed fragments were expressed as GFP fusion and FLG-tagged proteins, respectively, in primary myotubes. Remarkably, the results indicated that neither of the fragments retained the M-line localization characteristic of FL-CAPN3 (Fig. 5 A and B, iv vs. ii), suggesting that upon autolysis, CAPN3 may localize to the cytosol for functions distinct from that of FL-CAPN3 by undergoing iMOC.

Fig. 5.

Altered subcellular localization of N_31K and C_58K in myotubes compared with FL-CAPN3. N_31K and C_58K:HN > AA expressed in primary chick skeletal myotubes showed a diffuse localization, distinct from the striated M-line localization of CAPN3 FL:SHN (A and B, iv vs. ii). Arrows indicate Z-lines labeled by α-actinin or the N terminus of titin/connectin (Z1-Z2). Arrowheads indicate the striated localization of FL:SHN, visualized by anti-GFP or anti-FLAG antibodies. (Scale bars, 10 μm.)

iMOC Mediated by Mutant FL-CAPN3 and Wild-Type Autolytic Fragments Rescues CAPN3’s Autolytic Activity.

During the course of these studies, to our surprise, a protease-inactive mutant of CAPN3, FL:AHN, was found to undergo proteolysis when coexpressed with N_31K, but not N_31K:C > S (Fig. 6A, lane 2 vs. lanes 1 and 3, gray arrowhead). However, the coexpression of N_31K with FL:AAA, another protease-inactive mutant in which all three catalytic amino acid residues were replaced with Ala, did not lead to the generation of proteolytic fragments (lane 5). These results demonstrate that the autolytic activity of CAPN3 was restored by iMOC between C129 in N_31K and H334/N358 in the FL molecule.

Fig. 6.

The N-terminal autolytic fragment of CAPN3 restores the autolytic activity of inactive mutant CAPN3 through iMOC. (A) When expressed in COS7 cells, FL:AHN, a catalytically inactive mutant, failed to undergo autolysis (lane 1). Coexpression of N_31K with FL:AHN generated a fragment with a similar mobility to the C_58K fragment and trace amounts of N-terminal fragment (lane 2, gray arrow). These fragments were not detected when N_31K was replaced by N_31K:C > S (lane 3) or when FL:AHN was replaced by FL:AAA (lanes 4–6), a catalytically inactive mutant in which all three active site residues were replaced by Ala. (B) Deletion of the IS1 region from N_31K in N_30K did not affect iMOC-mediated autolysis of FL:AHN (lane 3 vs. lane 4, gray arrowhead). However, the autolytic reaction was compromised by Ca²⁺ binding-site mutations in the PC1 domain of N_31K (lane 5 vs. lane 6). (C) The N-terminal region of CAPN2, CAPN2:N_27K, could substitute for N_30K or N_31K in generating the C_58K-like fragment from FL:SHN (lane 4, gray arrowhead). The active-site Cys105 in CAPN2 was required for the reaction (lane 5). (D) The C_58K fragment generated from FL:AHN by CAPN3:WT was reduced in size by introducing a C-terminal truncation, FL:AHN:ΔC27 (lane 1 vs. lane 5). A similar change in product size was observed after the coexpression of FL:AHN:ΔC27 with N_30K (lane 2 vs. lane 6). Arrowhead and open arrowhead indicate the full-length (FL) product of the protease-inactive CAPN3 mutants and the C_58K fragment, respectively. Open and closed arrows indicate the positions of the 27-aar truncated C_58K generated from FL:AHN:ΔC27 and the expressed N-terminal fragment, respectively.

In the iMOC-mediated FL autolytic reaction observed above, N_30K was competent to substitute for N_31K (Fig. 6B, lanes 3 and 4), as in the case of iMOC mediated by two autolytic fragments. Furthermore, CAPN2:N_27K could also restore CAPN3 FL:SHN autolytic activity, in a CAPN2 Cys105-dependent manner (Fig. 6C, lane 4 vs. lane 5). Notably, the mutations D120A/E199A present in N_31K;DE > AA abrogated the iMOC-mediated autolysis (Fig. 6B, lane 6). These amino acid residues correspond to the Ca²⁺-binding site in PC1 (CBS1) in CAPN1 and CAPN2 and were shown to be involved in the Na⁺ responsiveness of CAPN3 (35).

The proteolytic product generated from FL:AHN (or FL:SHN) following iMOC with N-terminal competent fragments and detected by anti-C3IS2 antibodies, appeared to be identical to the C_58K fragment (Fig. 6 A–D, gray arrowhead, anti-C3IS2). When a C-terminal 27-aar truncation was introduced into FL:AHN (FL:AHN:ΔC27; Fig. 1A), the proteolytic fragment generated from this mutant following coexpression with either CAPN3:WT or N_30K exhibited a reduction in molecular weight by ∼27 aars (Fig. 6D, lanes 1 and 2 vs. lanes 5 and 6, open arrow). Thus, the 58-kDa product generated from FL:AHN by iMOC-mediated autolysis likely represented the C-terminal fragment, identical to C_58K produced by the canonical autolysis of CAPN3:WT. The N-terminal proteolytic products generated from FL:AHN had sizes close to N_31K: these fragments are supposed to be competent for iMOC if not for the C129A mutation (Fig. 6B, lanes 3 and 4, bracket).

Notably, the lack of autolytic activity associated with FL:CAA was rescued by C_58K (Fig. S1C, lane 10), but not by C_58K:HN > AA (lane 11). These results indicate that the coexpression of a CAPN3 mutant and the appropriate N- or C-terminal autolytic fragment (for example, FL-CAPN3 with a mutation at the N terminus + N-terminal fragment without a mutation), results in a protease activity similar to that of iMOC-CAPN3.

IS1 and IS2 in the FL-CAPN3 Are Required for iMOC-Mediated FL Autolysis.

Next, the role of the CAPN3-specific sequences (NS, IS1, and IS2) in FL-CAPN3–mediated iMOC was assessed using a series of deletion mutants in which the active-site Cys was replaced by Ser (for mutants, see Fig. 1A). The mutant FL:SHN:ΔNS, which was recognized as a substrate for CAPN3:WT, was found to undergo iMOC with N_30K or N_31K, resulting in the generation of autolytic fragments (Fig. 7A, lanes 3 and 5 vs. lane 4), indicating that the NS in FL-CAPN3 is dispensable for iMOC with the N-terminal autolytic fragments of CAPN3. On the other hand, coexpression of the IS1 deletion mutant FL:SHNΔ:IS1 with either N_30K or N_31K did not result in production of the proteolytic C_58K fragment (Fig. 7B, lanes 3–5). This result indicates that, like the canonical autolysis of FL-CAPN3, the IS1 sequence is required for CAPN3 to undergo iMOC-mediated autolysis.

Fig. 7.

The CAPN3-specific insertion sequences IS1 and IS2 are required for the iMOC-mediated autolysis of FL-CAPN3. (A) The protease-inactive deletion mutant FL:SHN:ΔNS was coexpressed with CAPN3:WT or the N-terminal fragments in COS7 cells. In addition to CAPN3:WT, both N_30K and N_31K were able to generate the C_58K autolytic fragment from FL:SHN:ΔNS (lanes 3 and 5 vs. lane 2, open arrow). The autolytic fragment was not generated when N_31K was replaced with N_31K:C > S (lane 4). (B) Coexpression of FL:SHN:ΔIS1 with N_31K or N_30K did not result in the generation of the C_58K autolytic fragment (lanes 3 and 5 vs. lane 6). (C) Coexpression of FL:SHN:ΔIS2 with CAPN3:WT generated a slightly smaller fragment than C_58K (lane 2, lower open arrowhead, C_58K:ΔIS2), which was not generated when N_31K or N_30K was substituted for CAPN3:WT (lanes 3 and 4). Similarly, CTBP1’s proteolysis was undetectable when N_31K or N_30K was coexpressed with FL:SHN:ΔIS2 [anti-MYC; lanes 3 and 4 vs. lane 2 (B)]. (D) The effects of deleting NS, IS1, or IS2 on iMOC-mediated autolysis are summarized. The lack of either IS1 or IS2 abrogated iMOC-mediated autolysis of FL-CAPN3. Note that the ability of CAPN3 to serve as a substrate for CAPN3:WT did not depend on IS2.

FL:SHN:ΔIS2 was capable of serving as a substrate for CAPN3:WT, as the coexpression of these two CAPN3 forms resulted in the generation of a proteolytic fragment corresponding to the IS2-deleted form of C_58K (C_58K:ΔIS2) (Fig. 7C, lane 2). In contrast, C_58K:ΔIS2 was not observed when FL:SHN:ΔIS2 was coexpressed with N_31K or N_30K (Fig. 7C, lanes 3 and 4), indicating a failure of these fragments to undergo iMOC leading to autolysis. In addition, FL:SHN:ΔIS2 was unable to cooperate with N_31K or N_30K to proteolyze CTBP1, unlike N_30K+C_58K:ΔIS2 (Fig. 7C, lanes 3 and 4 vs. Fig. 2C, anti-MYC, lane 5). Taken together, these results indicate that both IS1 and IS2 in the FL-CAPN3 are required for the iMOC-mediated autolysis of inactive FL-CAPN3 (Fig. 7D).

Discussion

“Complementation” refers to a phenomenon in which functional activity is restored through the noncovalent interaction of two (or more) different proteins, as in the case of α-complementation in Escherichia coli β-galactosidase, in which the α-fragment and ω-fragment of this enzyme restore enzymatic activity when expressed in trans (36, 37). In addition, it has been demonstrated that physiological functions of some proteins including membrane receptors and enzymes are mediated by intermolecular complementation (38–41). In the present study, we showed that autolyzed fragments of CAPN3, neither of which has protease activity when expressed alone, complement each other to reconstitute protease activity in a manner similar to α-complementation. Our data also strongly support and provide a molecular mechanism for the hybrid hypothesis (34), a model based on the finding that patients with LGMD2A exhibiting compound heterozygous mutations in different domains of CAPN3 may show reduced pathology compared with patients with single homozygous mutations (Fig. 1B). In this model, two complementary fragments generated from different CAPN3 molecules form an active protease complex by sharing active-site amino acid residues in each fragment. One of the remarkable aspects of CAPN3 is that the complementary complex can be formed spontaneously from separately generated fragments de novo. To our knowledge, such a property has only been reported for pPR, a protease in cytomegalovirus (41), which is described below.

Intriguingly, the initial autolytic cleavage of CAPN3 requires the CAPN3-specific sequences IS1 and IS2 (11, 42), whereas these sites are dispensable for the iMOC of autolyzed fragments. In addition, as CAPN2:N_27K can replace CAPN3 N_30K in the fragment-based iMOC, it is tempting to predict that many, if not all of the CAPNs could potentially undergo iMOC (Fig. 2C and Fig. S3). In fact, CAPN1 and CAPN9 have 3D structures (43) similar to that of CAPN2 in the region of PC2 containing β1, α1, and β2, which are thought to be crucial for iMOC. From physiological aspect, however, such iMOC compatibility may not be relevant. As of now, it is assumed that most CAPNs do not spontaneously undergo autolysis to generate the fragments corresponding to N_31K and C_58K of CAPN3. In addition, the calpain species that never coexist with CAPN3 in skeletal muscle have no chance to be involved in complementation with CAPN3 fragments.

It should also be noted that there is no evidence that ubiquitous calpains expressed in skeletal muscle complement CAPN3’s function. This could be due to the fact that the physiologically essential characteristic of CAPN3, such as its sensitivity to Ca²⁺ or Na⁺ as well as cellular localization, is not complemented by the fragment from other calpain species.

Consistent with the above notion, in contrast to CAPN2:N_27K, CAPN2:C_53K was not functionally equivalent to the corresponding CAPN3 fragment, C_58K. Thus, this region of CAPN3 appears to encode the functional specificity for iMOC (Fig. S2B). The specific feature(s) in the C-terminal region of CAPN3 that distinguish it from that of CAPN2 is currently unclear. Previous findings indicate that both CAPN1 and CAPN2 can form a heterodimer with the calpain small subunit (CAPNS1) via the PEF(L) domain. However, a similar heterodimer of CAPN3 with CAPNS1 has not been detected, which may also suggest that the carboxyl-terminal region of CAPN3 and other CAPNs plays different roles in these related proteases. In line with such difference, the EF5-hand of CAPN3 is unique in that it is a binding site for Ca²⁺ as well as an interacting site for homodimer formation of the PEF(L) domain (33). It was also shown that a CAPN3 mutant completely lacking IS2 and PEF(L) (CAPN3:S581X), expresses protease activity comparable to that of FL-CAPN3 (35). Further studies will be required to elucidate the precise role of the carboxyl-terminal region in regulating CAPN3’s autolysis and the function of iMOC.

The rescue of inactive FL-CAPN3 involves iMOC mediated by either the N- or C-terminal autolytic fragment (Fig. 8A and Fig. S1C). The N-terminal fragment-mediated iMOC provided the active-site Cys, whereas both the NS and IS1 domains were dispensable. On the other hand, the presence of both IS2 and IS1 were required in the FL-CAPN3 mutant for the iMOC-mediated restoration of activity. The effect of the ΔIS1 mutation was expected, as it compromises the structure of the CAPN3 autolytic sites. Because IS2 plays a role in the Ca²⁺-independent activity of CAPN3 (11, 35), its deletion may change the structure of PC2. However, we demonstrated that the IS2 in C_58K was not required for the formation of the active CysPc through iMOC with the FL:CAA mutant (Fig. 8B and Fig. S1C). Therefore, IS2 was required in FL-CAPN3, but not in C_58K. Although the stoichiometry of the N- or C-terminal fragment and FL-CAPN3 in the iMOC complex is not clear, the CysPc domain in CAPN3 somehow circumvents steric hindrance by the extra PC1 or PC2 domains, probably via flexibility introduced by IS1.

Fig. 8.

iMOC contributes to the previously unidentified regulation of CAPN3. (A) The FL-CAPN3 is an active protease (diagram at Top). Through autolysis, it is converted into two fragments, designated Nterm and Cterm, which flank the IS1 region (i–iii). The iMOC-mediated reconstitution of protease activity requires the coexpression of complementary Nterm and Cterm fragments. The N-terminal fragment of CAPN3 can be replaced by the corresponding region in CAPN2 (i). Although the C-terminal IS2 region is not required for iMOC-mediated autolysis, the C-terminal region of CAPN2 is unable to substitute for CAPN3 Cterm (ii and iii). Shaded box indicates that the construct is derived from CAPN2. (B) iMOC between Nterm and inactive FL-CAPN3 mutants restores autolytic activity (i) and requires the presence of both IS1 and IS2 (ii–iv). The Cterm, C_58K, also undergoes iMOC with FL-CAPN3 mutants, restoring autolytic activity (v), but does not require IS2 in the Cterm (vi). *The results are shown in Fig. S1C. (C) Revised model of the CAPN3 autolytic cycle. In this model, autolysis is viewed as both a positive and negative regulator of CAPN3 function. The preautolytic FL-CAPN3 is regulated by previously identified factors (i). Initially, autolysis leads to nicking in IS1 (ii), resulting in the formation of nFL-CAPN3, allowing it to function as a protease (iii). The N- and C-terminal fragments separate (iv) and undergo iMOC (v), possibly involving additional regulatory mechanisms. After iMOC, the reconstituted CAPN3 may be alternatively regulated and take on additional functions (vi). At the same time, each autolytic fragment is capable of accelerating the autolysis of FL-CAPN3 by chimeric iMOC (iMOC2-N and iMOC2-C) (vii). The initial autolytic reaction is also accelerated by iMOC-CAPN3 (vii). The second round of autolysis is probably linked to the efficient down-regulation of CAPN3 (viii).

The homodimer formation of CAPN3 may have a certain role in iMOC (34). In the modeled structure of the FL-CAPN3 homodimer, protease core domains of two monomers are located at the outside of the complex, far from the interface (32). Therefore, it is not necessarily a monomer form of CAPN3 that performs iMOC. For example, iMOC may be more efficient for the autolytic fragments generated from different CAPN3 LGMD2A mutants in the same dimer. Alternatively, a dimer of FL-CAPN3 and/or C_58K may have a structural advantage to perform iMOC with N_31K. The molecular mechanism involved in CAPN3 iMOC will be an interesting area of investigation for structural biologists.

The critical difference between CAPN3 iMOC and the complementation of β-galactosidase is that CAPN3 innately generates the complementary fragments as natural products of autolysis. Such a mechanism coupling autolysis and complementation for the restoration of protease activity has been well studied for a capsid scaffold protein that functions as a capsid maturational protease (pPR, pUL80a) in cytomegalovirus (CMV) (41). During the course of viral capsid maturation, pPR functions as a critical protease in three forms: pPR, assemblin, and the reconstituted form of assemblin autolyzed products. Although the different forms of pPR may share the same substrate specificity, changes in its molecular size and conformation are necessary for its roles in maintaining efficient viral reproduction and for the removal of scaffold proteins from the capsid before incorporation of the viral DNA (44, 45). Interestingly, the fragments artificially generated from the assemblin homolog in HSV type I has the potential to undergo iMOC, but, such fragments are not naturally generated by autolysis (46, 47); it is attributed to the fact that HSV assemblin does not have an autolytic sequence corresponding to the one found in CMV assemblin (48). It has also been suggested that the difference in the functionality of two closely related assemblin homologs may reflect certain physical constraints, such as the size of the capsid and viral genome in each species.

FL-CAPN3 was previously shown to be most stably expressed in the myofibril-bound fraction of skeletal muscle cells (15). FL-CAPN3 is hypothesized to be highly regulated, requiring induction by certain biological conditions. The iMOC-induced activation of CAPN3 revealed in the present study suggests an alternative pathway for expressing CAPN3 activity (Fig. 8). In analogy to pPR, we hypothesize that the iMOC form may have greater mobility than FL-CAPN3 within skeletal muscle cells, which are densely packed with myofibril components (49). The observation that transiently expressed autolyzed fragments were found to be diffusely distributed in skeletal muscle cells instead of localizing to the sarcomeric M-line region (Fig. 5) supports this hypothesis in part. Reassembly of two fragments, however, is required to form an active enzyme, which may cause further kinetic and spatiotemporal constraint for function. In this regard, the overall kinetics of CAPN3 should be further investigated together with its interacting molecules through the process of autolysis to iMOC formation.

Our proteomic analysis showed that both FL- and iMOC-CAPN3 exhibited substantial autolytic activity. This property of iMOC-CAPN3 is consistent with the finding that the autolysis of CAPN3 was maintained by its autolytic products and may explain the exhaustiveness of CAPN3’s autolysis. We also found that CAST is more efficiently proteolyzed by FL-CAPN3 than by iMOC-CAPN3 in ectopic protein expression systems. Although it is still an open question whether there are physiologically significant differences in substrate specificity between FL-CAPN3 and iMOC-CAPN3, they have different sensitivity to inhibitors (Fig. S6A) (27). Considering that FL-CAPN3’s autolysis is very rapid, CAPN3 nicked by autolysis (nFL-CAPN3) (Fig. 8C) must be somewhat stable, requiring time to separate into fragments (27–29). It is currently unknown whether the preautolytic FL-CAPN3 proteolyzes other substrates before it is converted to nFL-CAPN3, or whether FL- and nFL-CAPN3 have distinct substrate preferences.

The present study revealed that autolyzed fragments of CAPN3 generate de novo protease activity via iMOC. This finding provides a possible molecular mechanism for the functional compensation between two distinct CAPN3 molecules (Fig. 1B). Therefore, at least three active molecular states of CAPN3 can be distinguished as shown in Fig. 8C; FL-CAPN3, nFL-CAPN3, and iMOC-CAPN3. Further studies should enhance our understanding of the mechanism by which CAPN3 exerts its protease activity and extend our understanding of the genotype–phenotype relationship in LGMD2A, as well as the molecular mechanisms underlying the pathology of this disease.

Materials and Methods

cDNA Constructs.

For protein expression in mammalian cells, cDNAs were subcloned into the pSRD expression vector, as described previously (16). The pEGFP-C1 and pcDNA3.1 (Invitrogen) expression vectors were used for protein expression in chick skeletal myotubes (16). For in vitro transcription and translation, the human CAPN3 cDNAs were subcloned into the pEU-E01 vector (CellFree Sciences) (50). Where indicated, the proteins were expressed as MYC-, FLAG-, or His-tagged forms. Enzymes used for manipulating recombinant DNA were purchased from Takara Bio, Iwai Chemicals, New England Biolabs, and Agilent Technologies. The mutations described here were introduced by long PCR using PfuTurbo DNA polymerase, as described previously (26). The sequences of all of the constructs were verified.

Cell Culture and Protein Expression.

Proteins were expressed in COS7 cells as previously described (16). COS7 cells were grown in DMEM (Sigma-Aldrich) supplemented with 10% (vol/vol) FBS that was heat inactivated at 56 °C for 30 min before use. The cells metabolically labeled using SILAC Protein Quantitation kit (Pierce) were also used for protein expression. Plasmid prepared at a concentration of 1 μg/μL was transfected into cells using TransIT-LT1 (Mirus Bio), according to the manufacturer’s instructions. For each transfection, 1.2–1.5 × 10⁵ cells were seeded in a 3.5-cm dish 12–16 h before transfection, and 1 μg of plasmid per construct was used as the standard condition. The cells were harvested 48–60 h after transfection using ice-cold PBS, resuspended in sonication buffer [20 mM Tris⋅Cl, pH 7.5, 1 mM EDTA-K, pH 8.0, 1 mM DTT, 2 mM 4-(2-aminoethyl) benzenesulfonyl fluoride hydrochloride (AEBSF)], and lysed by sonication. For proteomic analysis, the cell lysate was prepared in 20 mM triethylammonium bicarbonate containing 20 μM ALLNal (Sigma-Aldrich) and 20 μM calpeptin (Calbiochem) and subjected to sonication and centrifugation at 18,000 × g for 20 min at 4 °C (51). Protein concentration was quantified using the DC protein assay (Bio-Rad) or Direct Detect (Merck).

In Vitro Transcription and Translation.

In vitro transcription and translation were performed using the wheat-germ cell-free expression system according to the manufacturer’s instructions (CellFree Sciences) (50). The mRNAs for each construct were separately transcribed. The translation reaction was carried out at 16 °C for 18–20 h without shaking. For the coexpression of CAPN3 fragments, an equal volume of transcription product for each construct was mixed before the translation reaction. The volume of the translation reaction was adjusted to 30–100 μL, maintaining the ratio of constituents as indicated in the instructions. Where indicated, the translation reaction was carried out in the presence of protease inhibitors.

Western Blot Analysis.

Proteins were separated by SDS/PAGE and transferred onto PVDF membranes (Millipore). The membranes were probed with appropriate primary antibodies and horseradish peroxidase-coupled secondary antibodies (Nichirei) followed by visualization using a POD immunostaining kit (Wako) or the Clarity Western ECL Substrate (Bio-Rad). Chemiluminescence signals were detected by the ImageQuant LAS 4000 imaging system (GE Healthcare). Scanned images were processed for presentation using Adobe Photoshop CS6 (Adobe Systems).

Immunoprecipitation.

In vitro translation reaction mixtures were diluted in a 10- to 15-fold excess volume of lysis buffer (50 mm Tris⋅HCl, pH 7.5, 150 mm CsCl, 1 mm EDTA-K, pH 8.0, 1% Triton X-100, 1 mm AEBSF, 0.1 mm leupeptin, 10 μg/mL aprotinin, and 10 mM iodoacetamide) (52), and incubated on ice for 30 min with occasional mixing. The supernatant was collected after centrifugation at 20,630 × g for 5 min at 4 °C. Preabsorption of nonspecific interactions was performed using Protein-G Sepharose (GE Healthcare). Subsequently, the supernatant was incubated with goat anti-C3IS2 IgG (29) for 8–12 h, followed by the precipitation of immunocomplexes with Protein-G Sepharose according to the manufacturer’s instructions. Precipitated immunocomplexes were washed twice with lysis buffer and twice with a wash buffer that was modified from its previous description (50 mm Tris⋅HCl, pH 7.5, 150 mm CsCl). Immunoprecipitated proteins were denatured in SDS/PAGE sample buffer and subjected to SDS/PAGE. Immunoprecipitation using anti-FLAG M2 affinity gel (Sigma-Aldrich) was performed as described (52).The same protocol was used for the immunoprecipitation of proteins expressed in COS7 cells; cells collected from one dish were lysed in 1 mL of lysis buffer.

Immunofluorescence Analysis.

The transfection of primary chick skeletal myotubes and immunofluorescence microscopy were performed as described previously (53). Briefly, transfection was performed using Effectene (Qiagen). Six days later the transfected myotubes were incubated in relaxing buffer (150 mM KCl, 5 mM MgCl₂, 10 mM 3-(n-morpholino) propanesulfonic acid, pH 7.4, 1 mM EGTA, and 4 mM ATP) for 15 min, then fixed with 2% (wt/vol) formaldehyde in relaxing buffer for 15 min. The primary and the secondary antibodies were used at concentrations described below. Coverslips were mounted onto slides with Aqua Poly/Mount (Polysciences) and analyzed on an Axiovert microscope (Zeiss, Oberkochen) using 63× (NA 1.4) or 100× (NA 1.3) objectives. The images were processed for presentation using Adobe Photoshop CS6 (Adobe Systems).

Antibodies.

The antibodies used in this study included anti-FLAG mouse monoclonal antibody (1:1,000; M2, Stratagene), anti-MYC mouse monoclonal antibody (1:1,000; 9E10, Developmental Studies Hybridoma Bank), anti-GFP rabbit polyclonal antibody (1:2,000; Abcam, and 1:1,000; mFX75, Wako), anti-CAPN3 IS2 goat polyclonal antibody (1:1,000, Cosmo Bio (designated as anti-C3IS2), anti-CAPN3 rabbit polyclonal antibodies (1:1,000; RP4-Calpain3, Triple Point Biologics (designated as anti-C3NS), and 1:400; anti-CAPN3-1, Myomedix (designated as anti-CAPN3), anti-His mouse monoclonal antibody (1:1,000; His·Tag, Novagen), antisarcomeric–α-actinin mouse monoclonal antibody (1;15000; EA53, Sigma-Aldrich), and antititin Z1-Z2 rabbit polyclonal antibody (1:100) (53). Mouse monoclonal pan calpain antibody (anti-CAPNs) was kindly provided by Seiichi Kawashima (Tokyo Metropolitan Institute of Medical Science, Tokyo).

Calpastatin Proteolytic Assay.

All procedures were performed under light-protected conditions. Cast-d1 (Takara Bio) was labeled using Cy3 monoreactive dye (GE Healthcare) according to the manufacturer’s instructions with some modifications. The labeling reaction was stopped by adding lysine at a final concentration of 20 mM, and labeled protein was purified using a PD SpinTrap G25 column (GE Healthcare). Thirty nanograms (2.1 pmol) of labeled cast-d1 was incubated with 4 μL of in vitro translation (IVT) reaction mixture in a total volume of 10 μL at 16 °C for 26 h. The final composition of the incubation solution was 20 mM Tris⋅HCl, pH 7.5, 1 mM EDTA/K, 1 mM DTT, 2 mM AEBSF, and either 5 mM EDTA/K or 0.1 M NaCl. The reaction was stopped by the addition of SDS sample buffer, and the sample was subjected to SDS/PAGE. Fluorescent signals were detected using Typhoon 9400 laser scanner (GE Healthcare), followed by image processing using Adobe Photoshop CS6 (Adobe Systems). The same gel was used for Western blot analysis.

Proteomic Analysis.

Two sets of experiments were performed (Fig. S6A). The cell lysate was prepared as described in Cell Culture and Protein Expression. For each sample group, 20 μg of protein from each of two independent transfection procedures were combined, and a total of 40 μg was processed according to the instructions in the iTRAQ 4-Plex Reagent kit (AB SCIEX). Ten micrograms each of endoproteinase Lys-C (Wako) and trypsin (Promega) were used to digest 40 μg of protein. The digests from each sample were divided into two equal portions for labeling with two different iTRAQ tags (Fig. S6A). After the labeling reaction, the peptides were mixed and then separated into three fractions, by eluting with 87.5 mM, 175 mM, and 350 mM KCl, using the ICAT Cation Exchange Buffer Pack (AB SCIEX). Each fraction was desalted using a Sep-Pak Light C18 cartridge (Waters) and concentrated by evaporation. MS/MS spectral data were acquired using the TripleTOF 5600⁺ system (AB SCIEX) equipped with a Nano LC Ultra 1D Plus system (Eksigent).

Protein Identification and Quantification.

Proteins were identified and quantified using Protein Pilot (ver. 4.5, AB SCIEX). The Paragon algorithm in the Protein Pilot software was used as a search program with the following settings: trypsin digestion, methyl methanethiosulfonate modification of cysteines, iTRAQ 4-plex labeled peptides, instrument 5600, no special factors, default iTRAQ isotope correction settings, quantification, bias correction, background correction, biological modifications, amino acid substitutions, and thorough ID parameters selected. To improve the identification of protein from COS7 cells, which are established from monkey, the human database was manually supplemented with the sequences for other primates (UniProt database released in September 2012). This procedure increased the number of identified proteins by 2–3%, however, approximately 65% of identified proteins were found only in a human database (Fig. S6B). Inspection of identified proteins suggests that the sequence conservation between human and monkey is high enough for the search program with the settings we used (54). Therefore, it was concluded that the use of a human database for the present study is reasonable and necessary. In the two sets of experiments, 2,769 [(I) FL-CAPN3] and 3,169 [(II) iMOC-CAPN3] proteins were identified with a global false discovery rate (FDR) of <5% (Fig. S6B).

SILAC-Based Quantitation.

The immunoprecipitated proteins from metabolically labeled COS7 cells were mixed 1:1 and digested using endoproteinase Lys-C (Wako). After concentration and desalting, peptides were analyzed by the TripleTOF 5600⁺ system (AB SCIEX). Peptides and proteins were identified using Protein Pilot (ver. 4.5, AB SCIEX). Among the peptides with a confidence level of >95%, consistently modified peptides that are found in both light- and heavy-labeled forms are selected for quantitation. SILAC ratios (light:heavy) were calculated for the peak intensity of selected peptides using PeakView (ver. 2, AB SCIEX).

Sequence and Structure Analysis.

Protein sequence data were retrieved from the UniProtKB or NCBI database. The sequences were aligned using MAFFT, and the result was processed using Genetyx (ver. 11) (Genetyx). To investigate the structure of the reconstituted protease domain of CAPN3, the crystal structure of the protease domain of rat CAPN2 in the presence of Ca²⁺ (Protein Data Bank 1MDW) (6) was used as a template. Using the MolFeat ver. 5 (FiatLux), the coordination of the two autolytic fragments of CAPN3 was modeled. The processed image was exported for further modification using Adobe Photoshop CS6 (Adobe Systems).