Analysis of limited proteolytic activity of calpain-7 using non-physiological substrates in mammalian cells



M. Maki, Department of Applied Molecular Biosciences, Graduate School of Bioagricultural Sciences, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8601, Japan

Fax: +81 52 789 5542

Tel: +81 52 789 4088



Calpain-7 is a mammalian ortholog of a fungal non-classical calpain named PalB, which is an intracellular cysteine protease and functions in fungal alkaline adaptation in association with the endosomal sorting complex required for transport (ESCRT) system. Despite our previous finding [Osako Y et al. (2010) FEBS J 277, 4412–4426] of autolytic activity, neither physiological nor non-physiological substrates of calpain-7 have yet been identified, and experimentally useful substrates that show robust evidence of intermolecular proteolytic activity of calpain-7 are required. In this study, we found limited proteolysis of C-terminally truncated ALG-2-interacting protein X (ALIX; (ALIXΔC), but not full-length ALIX, when the mutant was co-over-expressed with calpain-7 in HEK293T cells and analyzed by western blotting. The extent of ALIXΔC cleavage by calpain-7 was enhanced by co-expression with several ESCRT proteins. We investigated whether fusion of casein, a commonly used substrate for a variety of proteases including calpains, to the Bro1 domain confers the ability to serve as a substrate of calpain-7, but no specific cleavage was observed. However, when domain 1 of calpastatin, an endogenous inhibitory protein of ubiquitous classical calpains, was fused with the Bro1 domain, the fusion protein was cleaved at the C-terminal border of subdomain B (an inhibitory center for calpains) of calpastatin. These results demonstrate for the first time that calpain-7 has limited proteolytic activity and substrate preference. Moreover, the proteolytic assay system developed enabled us to perform mutational analysis of calpain-7, which revealed the importance of not only the N-terminal microtubule-interacting and trafficking (MIT) domains but also the C-terminal C2 domain-like domains for proteolytic activity.


ALG-2-interacting protein X


ALIX-interacting motif


calpastatin domain 1


charged multivesicular body protein


endosomal sorting complex required for transport


monoclonal antibody


monomeric green fluorescent protein


MIT-interacting motif


microtubule-interacting and trafficking


multivesicular body




proline-rich region


tumor susceptibility gene


vacuolar protein sorting


Calpains, intracellular Ca2+-dependent cysteine proteases, are divided into two sub-groups: classical calpains and non-classical calpains [1, 2]. All classical calpains have a catalytic cysteine protease domain, a C2-domain-like domain and a penta-EF-hand (PEF) domain. Non-classical calpains have a catalytic cysteine protease domain, a C2L domain and an additional unique domain, but lack the PEF domain. For example, calpain-7 has a catalytic cysteine protease domain, two C2L domains and a tandem repeat of microtubule-interacting and transport (MIT) domains. Based on their expression patterns, calpains may be classified into tissue-specific or ubiquitous calpains. Tissue-specific calpains have specific roles in tissues or organs; for example, calpain-3 has a role in skeletal muscle and calpain-8 and calpain-9 have roles in the gastrointestinal tract [3-5]. Ubiquitous classical calpains, such as calpain-1 (μ-calpain) and calpain-2 (m-calpain), are thought to be involved in fundamental cellular functions and to regulate a variety of cellular events, including signal transduction, apoptosis and the cell cycle, as well as membrane trafficking, with limited proteolysis of their substrates [2, 6]. As calpain-7 falls into the ubiquitous category, it may function in a common pathway in cells [7]. Although activation mechanisms for these ubiquitous classical calpains and inhibition by the endogenous inhibitor calpastatin have been extensively studied by biochemical analyses, including analysis of 3D structures, knowledge of such mechanisms for non-classical calpains is limited [8-12].

Yeast genetics and comparative analyses in mammalian cells have revealed a variety of ESCRT (endosomal sorting complex required for transport) and associated proteins that contribute to formation of the multivesicular body (MVB) and sorting of ubiquitinated membrane-bound cargo proteins destined for vacuolar (lysosomal) degradation [13]. The ESCRT system also controls retroviral budding, the final step in cytokinesis and autophagy in mammalian cells [14, 15]. ESCRT machineries are composed of four complexes named ESCRT-0, -I, -II and –III, as well as an accessory vacuolar protein sorting (VPS) 4 lysosomal trafficking regulator-interacting protein (LIP) 5 complex that dissociates ESCRT proteins from endosomal membranes [16, 17]. VPS4 contains an MIT domain that interacts with MIT-interacting motif (MIM)-containing proteins of ESCRT-III such as CHMPs (charged multivesicular body proteins) 2, -3, -4, -6, and its ATPase activity is partially regulated by this interaction [18, 19]. We previously demonstrated that the tandem repeat of MIT domains of calpain-7 interacts with CHMP1A, -1B, -2A, 4B and 4C and increased sodium tolerance (IST) 1 [20, 21]. ALIX (ALG-2-interacting protein X, now known as an ESCRT accessory protein) has three distinct domains that interact with various proteins related to membrane trafficking [a Bro1 domain that interacts with CHMP4s, a V domain that interacts with viral late domains and syntenin, and a Pro-rich region (PRR) that interacts with tumor susceptibility gene 101 (TSG101), ALG-2, endophilins and CIN85] [22-25]. ALIX is thought to be recruited to endosomal membranes after conformational changes by binding to TSG101 and CHMP4s [26].

Orthologs of calpain-7 exist from yeast to mammals, and calpain-7 is the most evolutionarily conserved member of the calpain family [27, 28]. Genetic studies have revealed that fungal and yeast orthologs of calpain-7 (PalB and Rim13, respectively) have critical roles in alkaline adaptation by processing a transcription factor (PacC and Rim101, respectively) [27-30]. The calpain-7 orthologs and substrates are thought to be recruited to endosomal membranes by interacting with ESCRT-III proteins and ALIX homologs, respectively. However, neither potential physiological substrates, including mammalian homologs of PacC/Rim101, nor any non-physiological substrates have been found yet for calpain-7. We previously demonstrated that calpain-7 has autolytic activity that is enhanced by formation of a ternary complex with IST1 and CHMP1B in HEK293T cells [31]. Although autolysis of calpain-7 was suggested to occur by both inter-molecular and intra-molecular reactions, experimentally useful substrates that show robust evidence of inter-molecular proteolytic activity of calpain-7 are required. Here we show that C-terminally truncated ALIX containing the Bro1 domain and a part of the V domain serves as a substrate of calpain-7. Moreover, we designed a non-physiological substrate by fusing calpastatin domain 1 with the Bro1 domain, and identified a cleavage site. However, this strategy was not applicable to casein, a commonly used protease substrate. These findings strongly suggest that calpain-7 has limited proteolytic activity and substrate preference. The developed proteolytic assay system was useful to perform structure–function analysis of calpain-7 in mammalian cells.


Calpain-7-dependent cleavage of C-terminally truncated ALIX

To determine whether autolytic activity of calpain-7 is differently modulated by closed and open forms of ALIX [26], monomeric GFP (mGFP)-fused wild-type (WT) calpain-7 (mGFP–calpain-7) or mGFP–calpain-7C290S (a mutant in which the putative catalytic Cys, Cys290, is replaced by Ser) was transiently expressed with FLAG-tagged full-length ALIX (closed conformation) or its open-conformation deletion mutants ΔPRR, ΔC and NT (schematically represented in Fig. 1A) in HEK293T cells. An autolytic fragment corresponding to 30 kDa (designated 30K) was detected by western blotting using monoclonal antibody (mAb) against GFP when the WT mGFP–calpain-7, but not the C290S mutant, was expressed (Fig. 1B). The intensities of 30K bands were not obviously different between over-expression of full-length and deletion mutants of FLAG–ALIX (Fig. 1B, upper panel) or between FLAG–ALIX and control (FLAG vector, data not shown), suggesting that over-expression of ALIX had little effect on autolysis of calpain-7 itself. Unexpectedly, however, FLAG–ALIXΔC generated a proteolytic fragment (designated ΔC-short) that was not detected by co-expression with mGFP–calpain-7C290S (Fig. 1B, lower panel). Compared with the migration rate of the FLAG–ALIXNT mutant, the cleavage site is predicted to reside approximately at amino acid residue number 500 in the V domain based on the estimated molecular mass of ΔC-short (approximately 50 kDa). Although FLAG–ALIXΔC cleavage by calpain-7 is a non-physiological proteolytic event, the results obtained imply that ALIX-interacting proteins may also be substrates of calpain-7 in mammalian cells. The FLAG–ALIXΔC protein identified as a proteolytic substrate for calpain-7 was used as a model substrate to investigate the factors that modulate the proteolytic activities of calpain-7.

Figure 1.

Cleavage of C-terminally truncated ALIX by calpain-7. (A) Schematic representations of ALIX. ALIX has a Bro1 domain, a V domain and a proline-rich region (PRR). Deletion mutants used in the following experiments are indicated. The numbers below the bars indicate positions in amino acid residues. Full, full-length. (B) Expression plasmid for mGFP–calpain-7 wild-type (WT) or mGFP–calpain-7C290S (C290S) was used for co-transfection with vectors for FLAG-tagged ALIX or its deletion mutants into HEK293T cells as indicated above the panels. Total cell lysates were subjected to SDS/PAGE and western blotting with mAb GFP (upper panel) and mAb against FLAG (lower panel). mGFP–calpain-7 bands of full-length (FL) and 30 kDa (30K) are indicated by an arrow and an open arrowhead, respectively (upper panel). The limited proteolytic band specifically observed in WT is indicated by an arrow (lower panel), and is named ΔC-short.

Effects of ESCRT proteins on calpain-7 autolysis and ALIXΔC cleavage

We examined which ESCRT proteins enhance the calpain-7 autolysis and ALIXΔC cleavage. mGFP–calpain-7 and FLAG–ALIXΔC were co-expressed with each ESCRT component, i.e. FLAG-tagged signal-transducing adaptor molecule (STAM) 2 and hepatocyte growth factor receptor substrate (HRS) (Fig. 2A) for ESCRT-0, Myc-tagged VPS28, -37A, -37B, -37C and -37D, MVB12A and -12B and TSG101 (Fig. 2B) for ESCRT-I, FLAG-tagged EAP45, -30 and -25 (Fig. 2C) for ESCRT-II, and FLAG-tagged CHMP1A, -1B, -2A, -2B, -3, -4A, -4B, -4C, -5 and -6 in (Fig. 2D) and Strep-tagged CHMP7 and IST1 (Fig. 2E) for ESCRT-III. Figure 3A shows the rates of calpain-7 autolysis (upper panel) and ALIXΔC cleavage (lower panel) quantified by densitometric analysis of signal intensities of western blotting bands obtained in three independent transfection assays. Among these ESCRT proteins, TSG101, CHMP1A, -1B, -2A, -4A, -4B, -4C, -5 and -7 enhanced the autolytic activity of mGFP–calpain-7, and STAM2, HRS, TSG101, CHMP1A, -4A, -4B, -4C and -5 enhanced cleavage of FLAG–ALIXΔC. Expression of a C-terminally truncated mutant of CHMP1B (FLAG–CHMP1BΔα6) enhanced both calpain-7 autolysis and ALIXΔC cleavage (Fig. S1).

Figure 2.

Effects of co-over-expression of ESCRT proteins on the autolysis of calpain-7 and cleavage of ALIXΔC. (A–E) Co-expression effects of each component of ESCRT-0 (A), ESCRT-I (B), ESCRT-II (C) and ESCRT-III (D, E) on autolytic activity of mGFP–calpain-7, and calpain-7-dependent cleavage of FLAG–ALIXΔC were investigated as shown in Fig. 1. Ctrl, empty vector. The asterisks in (A) indicate FLAG-tagged STAM2 and HRS, the asterisk in (D) indicates a band attributable to CHMP1A, probably a dimer, and the asterisk in (E) indicates a band of endogenous cellular biotinylated protein that co-migrated with Strep-tagged CHMP7 in SDS/PAGE.

Figure 3.

Relationship between the rate of autolytic activity and ALIXΔC cleavage. (A) Rates of autolytic activities and FLAG–ALIXΔC cleavage influenced by co-over-expressed ESCRT proteins are shown. Cells transfected with an empty vector were used as a control (Ctrl) in each experiment. Values are expressed as means of more than three independent experiments ± SEM. (B) The rates of mGFP–calpain-7 autolysis and FLAG–ALIXΔC cleavage shown in (A) are plotted on the axis and axis, respectively. The overall correlation coefficient between the rate of autolytic activity and ALIXΔC cleavage was 0.64. Factors are classified into four groups (i–iv) based on the degree of enhancement of calpain-7 autolysis and ALIXΔC cleavage.

Relationships between calpain-7 autolysis and ALIXΔC cleavage

The degree of effects on autolysis of calpain-7 and on limited proteolysis of ALIXΔC differed among ESCRT proteins. To assess the correlation of each effect, we plotted the relative rates of proteolytic enhancement in respective transfection experiments by normalization against the rates in transfection experiments using the control vector (Fig. 3B). Although the overall correlation was found to below (= 0.64), it was possible to categorize the factors into four groups: (a) factors that have a greater contribution to ALIXΔC cleavage (STAM2 and HRS), (b) factors that have a greater contribution to autolytic activity (CHMP1B, -2A, -5 and -7), (c) factors that affect both autolysis and ALIXΔC cleavage at similar levels (TSG101, CHMP1A, -1BΔα6, -4A, -4B and -4C), and (d) factors that have no effect (VPS28, -37A, -37B, -37C and -37D, MVB12A and -12B, EAP45, -30 and -20, and CHMP2B, -3 and -6).

Importance of the CHMP4–ALIX interaction for ALIXΔC cleavage

Although interaction between calpain-7 and ALIX was not detectable by co-immunoprecipitation assays (data not shown), ALIXΔC cleavage by calpain-7 was presumed to take place in the ESCRT system because both ALIX and calpain-7 bind CHMP4 proteins [20,22]. To understand the activation mechanism of calpain-7 and ALIXΔC cleavage, it is necessary to take two factors into account: activation of calpain-7 and recruitment of ALIXΔC. As CHMP4s appear to be the most effective activator of calpain-7 (Fig. 2) and bind to ALIX, we first investigated whether the ALIX–CHMP4 interaction is important for calpain-7-dependent cleavage of ALIXΔC. We performed limited proteolysis assays similar to those for which results are shown in Fig. 2 except using a Bro1 domain mutant of FLAG–ALIXΔC (FLAG–ALIXI212DΔC) whose binding ability to CHMP4 is diminished by substituting Ile212 with Asp [23]. Although co-expression of all CHMP4 isoforms promoted calpain-7 autolysis regardless of the Bro1 domain mutation, the extent of FLAG–ALIXI212DΔC cleavage significantly decreased compared with FLAG-ALIXΔC (Fig. 4). To exclude the possibility that the mutation in the Bro1 domain of ALIX affected other factors independently of the ALIX–CHMP4 interaction, we constructed an expression plasmid for FLAG–CHMP4B1-206, which lacks the ALIX-interacting motif (ALXIM), and one for FLAG–CHMP4B1-188, which that lacks ALXIM and the MIT-interacting motif 2 (MIM2) (Fig. 5A,B). Although these mutants affected calpain-7 autolysis to some degree (Fig. 5B, upper panel), they did not enhance FLAG–ALIXΔC cleavage as much as wild-type FLAG–CHMP4B did (Fig. 5B,C, lower panels). In agreement with the results shown in Fig. 4, cleavage of FLAG–ALIXI212DΔC was not significantly promoted by co-expression with WT FLAG–CHMP4B and its mutants. These results indicate that binding of CHMP4B to the Bro1 domain is important for calpain-7 to cleave FLAG–ALIXΔC. Among the CHMP4 isoforms, enhancement of calpain-7 autolysis by CHMP4B and CHMP4C was greater than that by CHMP4A (Fig. 4A, upper panel). These results are consistent with the fact that calpain-7 binds to CHMP4B and CHMP4C but not to CHMP4A by co-immunoprecipitation assays (Fig. S2).

Figure 4.

Decrease in cleavage of ALIXΔC by mutation in CHMP4 binding site. (A) Either FLAG–ALIXΔC or FLAG–ALIXΔCI212D was co-expressed with mGFP–calpain-7 and FLAG–CHMP4A, -B or -C, and autolytic activities and cleavage of ALIXΔC were investigated as shown in Fig. 2. Ctrl, empty vector. (B) Rates of autolytic activities and cleavage of FLAG–ALIXΔC in (A) were calculated as described in the legend to Fig. 3A. Values are expressed as means of three independent experiments ± SEM.

Figure 5.

Decrease in cleavage of ALIXΔC by mutation in ALIX-binding site in CHMP4B. (A) Schematic representation of CHMP4B. The Snf7 domain is followed by two interacting motifs: the MIT-integrating motif (MIM2) and the ALIX-interacting motif (ALXIM). The numbers below the bar indicate positions in amino acid residues. (B) Either FLAG–ALIXΔC or FLAG–ALIXΔCI212D was co-expressed with mGFP–calpain-7 and FLAG–CHMP4B, CHMP4B1-206 or CHMP4B1-188, and autolytic activities and cleavage of ALIXΔC were investigated as shown in Fig. 2. FL, full-length; Ctrl, empty vector. (C) Rates of autolytic activities and cleavage of FLAG–ALIXΔC in (A) were calculated as described in the legend to Fig. 3A. Values are expressed as means of three independent experiments ± SEM.

Designing a non-physiological substrate of calpain-7

Assuming that calpain-7 acts in the ESCRT system and ALIX recruits its substrates as in yeast and fungi, i.e. proteolysis of transcription factors PacC/Rim101 of the ALIX homolog PalA/Rim20 by the calpain-7 orthologs PalB/Rim13, physiological substrates of calpain-7 may be cleaved when substrates are placed in the proximity of calpain-7 in the ESCRT system [32]. As no such calpain-7 substrates have yet been identified, we attempted to design a non-physiological substrate of calpain-7 by fusing the Bro1 domain with the two well-known naturally unfolded proteins (also termed intrinsically unstructured proteins): casein, a commonly used protease substrate, and calpastatin, an endogenous inhibitor of calpain-1 and calpain-2 [33]. We fused either bovine αs1-casein or human calpastatin domain 1 (castD1) to the C-terminus of the FLAG-tagged Bro1 domain of ALIX (Fig. 6A), and then FLAG–Bro–casein and FLAG–Bro–castD1 were co-expressed with mGFP–calpain-7 in HEK293T cells. As shown in Fig. 6B, when wild-type mGFP–calpain-7 was expressed simultaneously with a potential substrate, a specific proteolyzed fragment was not detected for FLAG–Bro–casein (lower left panel) but was detected for FLAG–Bro–castD1 (lower right panel, closed arrowhead), and the fragment was not detected by co-expression with the C290S mutant of mGFP–calpain-7 used as a negative control. The amount of cleaved fragment of FLAG–BroI212D–castD1, which has a mutation at the CHMP4-binding site in the Bro1 domain, was lower than that of FLAG–Bro–castD1 (Fig. 6B, lower left panel, lanes 1 and 3). As casein is a good substrate for a variety of proteases, including calpain-1, -2 and -3, the resistance of casein to calpain-7 is rather surprising. On the other hand, the limited proteolysis of calpastatin by calpain-7 is not a unique phenomenon among calpains as calpastatin has been known to be cleaved by calpain-3 at multiple sites [34].

Figure 6.

Cleavage of ALIXBro1-fused calpastatin by calpain-7. (A) Schematic representations of ALIXBro1-fused proteins and calpain-7 mutants used in this study. The numbers below the bars indicate positions in amino acid residues, excluding the FLAG tag and a linker region. (B) FLAG–ALIXBro1–casein (FLAG–Bro–casein), FLAG–ALIXBro1I212D–casein (FLAG–BroI212D–casein), FLAG–ALIXBro1–calpastatin domain 1 (FLAG–Bro–castD1) or FLAG–BroI212D–castD1 were co-expressed with mGFP–calpain-7 or mGFP–calpain-7C290S in HEK293T cells as indicated above the panels. The total cell lysates were subjected to SDS/PAGE and western blotting. Proteolytic bands of FLAG–Bro–castD1 are indicated by a closed arrowhead (lower panel). Asterisks indicate proteolysed bands that were produced by cellular proteases other than calpain-7. (C) Deletion mutants of mGFP–calpain-7 were co-expressed with FLAG–Bro–castD1, and cleavage of these proteins was analyzed as in (B). The asterisk indicates a band corresponding to a fragment observed in both WT and the C290S mutant of calpain-7.

Next, we investigated structure–function relationships of mGFP–calpain-7 using three deletion mutants in the N- or C-terminal region (Fig. 6A) of the enzyme and FLAG–Bro–castD1 as a substrate for co-transfection experiments, in which autolytic fragments were detected by mAb against GFP (Fig. 6C, upper panel) and limited proteolytic products were detected by mAb against FLAG (Fig. 6C, lower panel). The amounts of the 30K autolytic fragments were smaller in the ΔMIT mutant than in the wild-type (WT), and no autolytic fragments were detected in the ΔNt and ΔC2L1/2 mutants. The 35 kDa fragment (Fig. 6C, asterisk) in the ΔC2L1/2 mutant may correspond to the faint 35 kDa band that appeared for the active-site C290S mutant (Fig. 6B, upper panels), suggesting a proteolytic fragment generated by other cellular proteases in the transfected cells (Fig. 6C, upper panel). Production of the limited proteolytic fragment from FLAG-Bro-castD1 (Fig. 6C, lower panel, arrowhead) was similarly affected by deletion of domains in calpain-7, as in the case of production of the 30K autolytic fragment. Although the ΔMIT mutant (deletion of MIT domains) still possesses a weak proteolytic activity, further deletion of amino acids from 160 to 211 (the region additionally lacking in the ΔNt mutant) caused complete loss of the activity, suggesting that the linker region between MIT domains and the protease domain has an important role. Moreover, loss of proteolytic activity in the mutant with deletion of the C2L domains indicates an essential role of these domains for the activities. A similar structure–function relationship was observed when using FLAG–ALIXΔC as a substrate (Fig. S3). It remains to be established whether there is a physical link between the C2L domains and the ESCRT system, or whether these domains are important for proper folding of the protease domain to confer proteolytic activity.

Determination of a cleavage site in calpastatin domain 1

As Bro2castD1 was found to be a substrate of calpasin-7, we decided to determine the cleavage site by analyzing the N-terminal sequence of a limited proteolytic fragment that was derived from the C-terminal region. FLAG-tagged Bro–castD1 was fused with mGFP C-terminally (FLAG–Bro–castD1–mGFP, Fig. 7A) to enable purification of the C-terminal fragment using GFP–TrapA resin. First, we checked the cleavage efficiency of the newly designed substrate with Strep-tagged calpain-7 by western blotting (Fig. 7B). To our disappointment, the cleavage efficiency was very poor and a cleavage fragment specific to wild-type Strep-tagged calpain-7 (odd lanes, not generated by the CS mutant) was barely detectable when the western membranes were probed with anti-FLAG mAb (top panel) and anti-GFP mAb (2nd row panel). As FLAG–Bro–castD1 also gave a very faint band (indicated by a closed arrowhead) under the conditions used, the poor efficiency of the proteolysis appears not to be due to fusion of mGFP to the C-terminal side of Bro-castD1 in the new construct, but rather due to using Strep-tagged calpain-7 instead of mGFP–calpain-7 as an enzyme source for the co-transfection assays. Fusing a large molecule (mGFP) to the N-terminal MIT domains of calpain-7 may partly mimic binding of the MIT domains to ESCRT-III proteins to activate the enzyme. To circumvent the poor proteolytic efficiency for isolation of the C-terminal fragment, we co-expressed Myc-tagged TSG101, which was found to be efficient in enhancing autolysis of calpain-7 and cleavage of FLAG–ALIXΔC (Fig. 2B), and this strategy was effective for increasing the yield of the limited proteolytic fragment detected by anti-FLAG mAb (Fig. 7B, top panel, lanes 2 and 6). The C-terminal fragment was also detected by western blotting using anti-mGFP mAb (2nd row panel, lane 2, open arrowhead). After a large-scale transfection experiment followed by GFP–TrapA purification (Fig. S4), the bands of interest were excised from a Coomassie Brilliant Blue-stained poly (vinylidene difluoride) membrane and subjected to peptide sequencing by Edman degradation. The determined N-terminal sequence was KEGIT, corresponding to amino acid residues 212–216 of castD1 (Fig. 7C).

Figure 7.

Determination of the cleavage site in calpastatin domain 1. (A) Schematic representation of calpastatin domain 1 (castD1) fused N-terminally to ALIXBro1 and C-terminally to mGFP for determination of the cleavage site. The numbers below the bar indicate positions in amino acid residues, excluding the FLAG tag and a linker region. (B) FLAG–Bro–castD1–mGFP or FLAG–Bro–castD1 were expressed in combination with Strep-tagged calpain-7 or Strep-tagged calpain-7C290S and Myc-tagged TSG101 as described for HEK293T cells. N- and C-terminal proteolytic fragments of FLAG–Bro–castD1–mGFP produced specifically by WT Strep-tagged calpain-7 were detected by western blotting using anti-FLAG mAb (top panel) and anti-GFP mAb (2nd row panel), and are indicated by closed and open arrowheads, respectively. The asterisk indicates non-specific signals observed in western blotting using anti-GFP and anti-Myc mAb. Western blotting was also performed using anti-calpain-7 polyclonal antibody (3rd row panel) and anti-Myc mAb (bottom panel). (C) Cleavage site for calpain-7 in calpastatin domain 1 determined by N-terminal amino acid sequencing. The numbers indicate amino acid residue numbers for calpastatin (DDBJ/GenBank/EMBL accession number D16217). (D) FLAG–ALIXBro1, FLAG–Bro–castD1151-195, FLAG–Bro–castD1151-211 or FLAG–Bro–castD1 were co-expressed with mGFP–calpain-7 as well as with Myc-tagged TSG101. A cleaved fragment of FLAG–Bro–castD1 showed the same mobility as FLAG–Bro–castD1151-211.

Calpastatin has four repetitive inhibitory domains (domains 1–4), and each domain has three highly conserved regions (sub-domains A, B and C), each of which binds to specific domains of the conventional calpains: sub-domain A binds the large-subunit PEF domain, sub-domain B binds the protease domain, and sub-domain C binds the small-subunit PEF domain. Crystal structural analyses of complexes between rat (Rattus norvegicus) calpain-2 and calpastatin domains 1 and 4 revealed that the sequence corresponding to KREV(196–199) in human calpastatin sub-domain B of domain 1 loops out from the catalytic center and inhibits calpain activities [11, 12]. The calpain-7 cleavage site is located between ELLAK(207–211) and KEGIT(212–216) at the C-terminal border of sub-domain B, and electron densities of the corresponding amino acids in the crystal structures are very low. Compared with statistical analyses of the amino acid frequency near the cleavage site (P4–P7′) of calpain-1 and calpain-2 relative to the general frequency of amino acids in total proteins [35], the identified cleavage site of calpain-7 (P4-LLAK-P1 and P1′-KEGITGP-P7′) has a similar feature only at the P1 position, which has a higher frequency in Lys. A program for calpain cleavage site prediction using multiple kernel learning has been developed ( [36], and was used to predict cleavage sites in castD1. However, the results did not agree with the identified cleavage site of calpain-7. These facts suggest that the substrate specificity of calpain-7 is different from those of calpain-1 and calpain-2. Skeletal muscle-specific calpain-3, categorized as a classical calpain, is not inhibited by calpastatin but cleaves the inhibitory protein at multiple sites including the N-terminal side of the KREV(196–199) sequence. As calpain-7 lacks PEF domains and its protease domain is distantly related to those of classical calpains [27, 28], attraction of calpastatin to its protease domain may not be similar to that of classical calpains. We cannot exclude the possibility that calpain-7 also cleaves sites on the N-terminal side other than the identified cleavage site in castD1. However, this possibility is unlikely due to the fact that FLAG–Bro–castD1 truncation mutants containing stop codons at the positions corresponding to Lys196 and Lys212 migrated faster and similarly to migration of the N-terminal fragment detected by anti-FLAG mAb (Fig. 7D).


Although physiological substrates of calpain-7 have not yet been identified, by using FLAG–ALIXΔC as a model substrate, we found that limited proteolytic activity of calpain-7 depends on the ESCRT system in a way similar to that of fungal and yeast proteolysis by PalB/Rim13. However, some differences were observed. In yeast, ESCRT-II components and Vps20 (an ortholog of CHMP6) are essential for Rim101 processing by Rim13 [37, 38], but ESCRT-II components and CHMP6 showed little enhancement of FLAG–ALIXΔC processing (Figs 2 and 3). On the other hand, while Vps27 (and ortholog of HRS) is dispensable in yeast, HRS as well as STAM2 (ESCRT-0 components) enhanced FLAG–ALIXΔC processing but not autolysis. It is possible that the discrepancy is due to the differences in methods employed: gene disruption in yeast versus over-expression in the present study. Indeed, over-expression of IST1 had little effect on autolysis of mGFP–calpain-7 (Fig. 2), but knockdown of IST1 decreased the efficiency of autolysis (Fig. S5). According to a structural model, Snf7/CHMP4 is a major ESCRT-III component and forms polymers that are seeded with Vps20/CHMP6 (interacting with Vps25/EAP20, an ESCRT-II component) and capped with Vps2/CHMP2 and Vps24/CHMP3 [16]. Both Vps2 and Vps24 have been shown to be dispensable for Rim101 processing in yeast [37, 38], but interaction of Vps24 with PalB is essential for PacC processing in fungi [39]. Thus, although direct interactions have not been demonstrated, the variation or lack of the MIT domain in Rim13 may allow the protease to interact with ESCRT components in a different manner to that for Aspergillus PalB and human calpain-7 (i.e. Rim13 interacts with Snf7/CHMP4, PalB interacts with Vps24/CHMP3, and calpain-7 interacts with CHMP1s, CHMP4s and IST1), and the protease activation mechanisms may have evolved differently. The C-terminal region of Rim13 has recently been shown to be essential for Rim101 processing and punctate localization of Rim13–GFP in yeast cells under alkaline conditions [40].

Recent reports indicate that the V domain of ALIX binds to K63-linked di- or polyubiquitin chains with an affinity higher than to K48-linked or linear chains [41, 42]. To investigate whether there is a relationship between limited proteolysis of ALIX by calpain-7 and the ubiquitin binding capacity, we constructed expression vectors for the same alanine-substituted ubiquitin-binding deficient mutants of ALIX (QEE_3A, Q435/E439/E442; ERE_3A, 453E/456R/460E; 6A, QEE_3A/ERE_3A), the V domain of which has been reported to lose binding ability to K63-linked polyubiquitin chains [41], and performed calpain-7-dependent proteolysis assays (Fig S6). The mutations did not cause the full-length FLAG–ALIX protein to become susceptible to mGFP–calpain-7, and did not increase or decrease the amounts of the limited proteolytic fragment of FLAG–ALIXΔC (ΔC-short). Thus, ubiquitin binding to the V domain of ALIX does not appear to regulate the susceptibility of ALIX to calpain-7, at least under the conditions used in this study. Although both FLAG–ALIXΔPRR and FLAG–ALIXΔC mutants are thought to have an open conformation, only FLAG–ALIXΔC was cleaved by mGFP–calpain-7. The V domain (360–702) has a V-shape structure formed by two arms comprising a three-stranded coiled coil (six long and five short α-helices) [24]. The deletion after 660 amino acids may disrupt the second three-stranded coiled coil (coil 1: α4, amino acids 418–427; α5, amino acids 430–471; coil 2: α6, amino acids 481–484; α7, amino acids 486–516; α8, amino acids 518–521; coil 3: α13, amino acids 645–699), and the resulting unstructured region may be susceptible to cleavage by calpain-7 (see Fig. S7 for a structural model). We also attempted to determine the cleavage site of FLAG–ALIXΔC–mGFP. However, the cleavage efficiency was too low for further analysis (data not shown). Fusion of GFP to the truncated V domain may alter its conformation and hence its ability to resist protease attack.

Based on recent knowledge of the ESCRT system and new findings obtained in this study, we propose a model for activation of calpain-7 in mammalian cells (Fig. 8). Both the enzyme (calpain-7) and hypothetical substrates bound to the V domain of ALIX are recruited to the ESCRT scaffold formed on the endosomal membrane via direct or indirect interaction with CHMP4s through each specific domain and interacting motifs (the MIT domain interacts with MIM, and the Bro1 domain interacts with ALXIM). A CHMP1/IST1 complex may also recruit calpain-7 to the membrane [31], but association with other ESCRT components remains to be established. Among the ESCRT-I components, only TSG101 enhanced autolytic and limited proteolytic activities (Figs 2 and 3). This is consistent with the results of a previous study suggesting that binding of TSG101 to the PSAP motif in ALIX PRR disrupts intramolecular interactions between ALIX PRR and the Bro1 domain, resulting in exposure of a patch in the Bro1 domain, which increases affinity for CHMP4s, and causing localization of ALIX to membranes coated with an ESCRT-III lattice [26]. CHMP5 is known to be absent from the ESCRT-III complex. Interestingly, it enhanced autolysis of mGFP–calpain-7 most efficiently among ESCRT components, and enhanced cleavage of FLAG–ALIXΔC moderately (Figs 2 and 3). While ALIX binds only CHMP4s among the ESCRT-III proteins, Brox (a farnesylated Bro1 domain protein with a CAAX motif) binds CHMP4s and CHMP5 [43, 44]. Thus, calpain-7 may be recruited to endosomal membranes by alternative ESCRT-involving pathways in addition to the common ESCRT systems found in yeast and fungi. The finding of catalytic activity of calpain-7 toward Bro1 domain-fused proteins has encouraged us to pursue further studies to identify physiological substrates in mammalian cells in the future.

Figure 8.

Hypothesis for ESCRT-dependent calpain-7 proteolysis on endosomal membranes. Calpain-7 and ALIX are recruited to the scaffold of ESCRTs formed on endosomal membranes via the MIT domain in calpain-7 and via the Bro1 domain and PRR in ALIX, respectively, by interacting with ESCRT components as indicated by double-headed arrows. Bro1 domain-fused calpastatin becomes a non-physiological substrate, as shown in this study. Unidentified calpain-7 substrates may be recruited to endosomal membranes by interaction with ALIX, probably via the V domain, in analogy with the fungal and yeast PalB/Rim13 proteolytic system. ESCRT-0 proteins are also involved in recruitment and activation of calpain-7 via unknown mechanisms.

Experimental procedures

Antibodies and reagents

The following mouse monoclonal antibodies were used: anti-FLAG (clone M2; Sigma, St Louis, MO, USA), anti-GFP (clone B-2; Santa Cruz Biotechnology, Santa Cruz, CA, USA) and anti-Myc (clone PL14; MBL, Nagoya, Japan). Anti-GFP antiserum (catalog number A6455) was obtained from Invitrogen/Molecular Probes (Carlsbad, CA, USA). Horseradish peroxidase-conjugated Strep-Tactin was obtained from IBA GmbH (catalog number 2-1502-001, Göttingen, Germany). A rabbit polyclonal antibody against recombinant human calpain-7 was raised in rabbits as described previously [20].

Construction of plasmids

Construction of the plasmids expressing calpain-7, calpain-7C290S, ALIX, ALIX mutants, ESCRT-I, -II and -III and its accessory proteins was performed as described previously [22, 45, 46]. Plasmids expressing FLAG-tagged ESCRT-0 proteins and Myc-tagged ESCRT-I proteins were kindly provided by M. Komada (Department of Biological Sciences, Tokyo Institute of Technology, Japan) and W.I. Sundquist (Department of Biochemistry, University of Utah, Salt Lake City, UT), respectively.

Mammalian expression plasmids for 3xFLAG-ALIXΔCI212D, pCMV3xFLAG-CHMP4B1-188 and pCMV3xFLAG-CHMP4B1-206 were obtained by PCR-based site-directed mutagenesis using primer pairs ALIXΔCI212D forward (5′-GAAAGATGCCgaCATA-GCTAAATTAGC-3′) and reverse (5′-GCTAATTTA-GCTATGtcGGCATCTTTC-3′), CHMP4B1-188 forward (5′-TCAGTGGACCCGAAACAtaaCCTCTACCAAATGTTCC-3′) and reverse (5′-GGAACATTTG-GTAGAGGttaTGTTTCGGGTCCACTGA-3′), CHMP4B1-206 forward (5′-CAAAACCCGCCAAGA-AGtAAGAAGAGGAGGACG-3′) and reverse (5′-CGTCCTCCTCTTCTTaCTTCTTGGCGGGTTTTG-3′) (substituted nucleotides are shown in lower case) and the template plasmids pCMV3xFLAG-ALIXΔC and pCMV3xFLAG-CHMP4B, respectively. To construct pFLAG-Bro-castD1(151–270) and pFLAG-Bro-casein, we first mutated pFLAG-ALIX to add an SalI site downstream of the Bro1 domain-encoding cDNA sequence by site-directed mutagenesis using primers 5′-CTGTTTGAGAAGATGGTcgaCGTGTCAGTACAGCAG-3′ (forward) and 5′-CTGCTGTACTGAC-ACGtcgACCATCTTCTCAAACAG-3′ (reverse) (substituted nucleotides shown in lower case), and then removed a fragment after digestion with SalI at the newly created SalI site and the SalI site in the multiple cloning site, to create pFLAG-ALIXBro1. The cDNAs encoding calpastatin D1 (castD1, amino acids 151–270) and casein were amplified using a full-length human calpastatin cDNA [47, 48] and pBR322-casein (a kind gift from M. Nagao, School of Agriculture, Kyoto University, Japan) [49] as templates and primer pairs castD1 forward (5′-TGAGAAGATGGTCGACTCAGGCATGGATGCTGCTTTG-3′) and reverse (5′-CCCCCTCGAGGTCGACTCAACTTCTGACTGTCCCTGCTG-3′), and casein forward (5′-TGAGAAGATGGTCGACAGGC-CCAAACATCCTATCAAG-3′) and reverse (5′-CCC-CCTCGAGGTCGACTCACCACAGTGGCATAGT-AG-3′), and inserted into SalI-digested pFLAG-ALIXBro1 using an In-Fusion® Advantage PCR cloning kit (Clontech, Palo Alto, CA, USA). For mammalian expression of FLAG–Bro–castD1(151–248)–mGFP, a DNA fragment of mGFP was amplified using primers 5′-TAAGTCGACGATCCAAGCGGAGGCTCT-3′ (forward) and 5′-TAAGTCGACTTACTTGTACAGC-TCGTCCA-3′ (reverse), and inserted into SalI-digested pFLAG-ALIXBro1, and EcoRV was introduced after the Bro1 domain of ALIX by PCR-based mutagenesis using primers 5′-CTGTTTGAGAAGATGGatatCGATCCAAGCGGAGGC-3′ (forward) and 5′-GCCT-CCGCTTGGATCGatatCCATCTTCTCAAACAG-3′ (reverse) (substituted nucleotides shown in lower case). A cDNA fragment corresponding to part of castD1 (amino acids 151–248) was amplified using primers 5′-TTTGAGAAGATGGATATCTCAGGCATGGATG-CTGCTTTG-3′ (forward) and 5′-TCCGCTTGGATC-GATATCTCCAGCAGCTGTAGGCGAC-3′ (reverse), and inserted into the EcoRV-digested plasmid as described above.

pFLAG-Bro-castD1151-195 and pFLAG-Bro-castD1151-211 were constructed by site-directed mutagenesis using pFLAG-Bro-castD1 as a template and primer pairs Bro-castD1151-195 forward (5′-CTACATAGAG-GAATTGGGTtAAAGAGAAGTCACAATTC-3′) and reverse (5′-GAATTGTGACTTCTCTT-TaACCCAATTCCTCTATGTAG-3′), and Bro-castD1151-211 forward (5′-AGGGAACTATTGGCTA-AAtAGGAAGGGATCACAGG-3′) and reverse (5′-CCTGTGATCCCTTCCTaTTTAGCCAATAGTTCC-CT-3′) (substituted nucleotides shown in lower case).

Cell culture and DNA transfection

HEK293T cells were cultured in Dulbecco's modified Eagle's medium (Nissui, Tokyo, Japan) supplemented with 4 mm glutamine, 5% fetal bovine serum, 100 U·mL−1 penicillin and 100 μg·mL−1 streptomycin at 37 °C under humidified air containing 5% CO2. Cells were transfected with plasmid DNAs by the conventional calcium phosphate precipitation method.

Calpain proteolysis assay

HEK293T cells were transfected with expression vectors for mGFP-fused calpain-7 and FLAG-tagged ALIX or its mutants. At 24 h after transfection, the cells were washed with NaCl/Pi, and harvested cells were lysed using Laemmli sample buffer (62.5 mm Tris/HCl, pH 6.8, 10% glycerol, 2% SDS, 5% β-mercaptoethanol, 0.002% bromophenol blue). To detect proteolysed fragments of mGFP–calpain-7 and FLAG–ALIXΔC, SDS/PAGE was performed on 12.5% and 10% Laemmli gels, respectively, followed by western blot analyses. Signals for immunoreactive bands were detected using an LAS-3000mini lumino-image analyzer (Fujifilm, Tokyo, Japan) with Super Signal West Pico chemiluminescent substrate (Thermo Fisher Scientific Inc., Rockford, IL, USA) and quantified using Multi Gauge version 3.X (Fujifilm).

Purification of GFP-tagged fragments for N-terminal amino acid sequencing

One day after HEK293T cells had been seeded into 90 mm dishes, the cells were transfected using expression vectors for C-terminally mGFP-fused protein with Strep-tagged calpain-7 and Myc-tagged TSG101. After 48 h, the cells were washed with NaCl/Pi, and the harvested cells lysed in buffer A (10 mm HEPES/NaOH, pH 7.4, 142.5 mm KCl, 0.2% Nonidet P-40) supplemented with protease inhibitors (0.2 mm phenylmethanesulfonyl fluoride, 0.2 mm 4-(2-Aminoethyl)-benzenesulfonyl fluoride hydrochloride (AEBSF, Nacalai Tesque, Kyoto, Japan), 3 μg·mL−1 leupeptin, 1 μm L-trans-Epoxysuccinyl-leucylamido(4-guanidino)butane (E-64, Peptide Institute, Osaka, Japan), 1 μm pepstatin). The lysates were incubated with GFP–TrapA beads (ChromoTek, Martinsried, Germany) for 1 h at 4 °C, and then the beads were pelleted by brief centrifugation (100 g, 1 min) at 4 °C and washed three times with buffer A. The bound proteins were resolved by SDS/PAGE and then transferred to a poly(vinylidene difluoride) membrane and subjected to staining with Coomassie Brilliant Blue R-250 (Nacalai Tesque, Kyoto, Japan). The band of interest was cut out, and N-terminal sequencing analysis was performed by APRO Science Co. Ltd (Tokushima, Japan).


We thank all members of the Laboratory of Molecular and Cellular Regulation for valuable suggestions and discussion. This work was supported by a Grant-in-Aid for Scientific Research on Priority Areas to M.M. and a Grant-in-Aid for Japan Society for the Promotion of Science Fellows to Y.M.