LETTERS

Concerted multi-pronged attack by calpastatin toocclude the catalytic cleft of heterodimeric calpainsTudor Moldoveanu1, Kalle Gehring2 & Douglas R. Green1

The Ca21-dependent cysteine proteases, calpains, regulate cellmigration1, cell death2, insulin secretion3, synaptic function4 andmuscle homeostasis5. Their endogenous inhibitor, calpastatin,consists of four inhibitory repeats, each of which neutralizes anactivated calpain with exquisite specificity and potency6. Despitethe physiological importance of this interaction, the structuralbasis of calpain inhibition by calpastatin is unknown7. Here wereport the 3.0 A structure of Ca21-bound m-calpain in complexwith the first calpastatin repeat, both from rat, revealing themechanism of exclusive specificity. The structure highlights thecomplexity of calpain activation by Ca21, illustrating key residuesin a peripheral domain that serve to stabilize the protease core onCa21 binding. Fully activated calpain binds ten Ca21 atoms,resulting in several conformational changes allowing recognitionby calpastatin. Calpain inhibition is mediated by the intimatecontact with three critical regions of calpastatin. Two regions tar-get the penta-EF-hand domains of calpain and the third occupiesthe substrate-binding cleft, projecting a loop around the active sitethiol to evade proteolysis.

We determined the crystal structure of the complex betweenm-calpain and residues 134–219 of calpastatin inhibitory repeat 1(referred to as calpastatin; Fig. 1a, Supplementary Fig. 1 andSupplementary Table 1). The m-calpain heterodimer consists of an80 kilodalton (kDa) catalytic subunit and a 28 kDa regulatory sub-unit8. The large subunit contains the Ca21-dependent protease coredomain I–II (DI–II)7, DIII (which resembles C2 domains involved inmembrane targeting) and the Ca21-binding penta-EF-hand DIV toheterodimerize the homologous DVI of the regulatory subunit(Supplementary Fig. 1b)9. The catalytically inactive C105S 80 kDasubunit was used to overcome unwanted proteolysis observed withthe wild-type protein10. The regulatory subunit was substituted withthe 21 kDa DVI (Supplementary Fig. 1b)10. The 72 kDa calpastatininhibits heterodimeric calpains with nanomolar affinity, being com-posed of a non-inhibitory 12 kDa leader L domain and four 15-kDacalpain inhibitory repeats (Supplementary Fig. 1b)6. Each repeatcontains three regions (A–C) predicted to interact with calpain6.We optimized the calpain–calpastatin complex by truncating calpas-tatin from a longer construct (residues 119–238) on the basis ofresults from limited proteolysis and NMR spectroscopy, both ofwhich identified the amino- and carboxy-terminal mobile regionsthat impeded crystallization (Fig. 1b and Supplementary Fig. 2).

Calpastatin recognizes the Ca21-induced conformation of m-cal-pain but does not coordinate Ca21 in the complex (Fig. 1a). RegionsA and C fold as amphipathic helices when bound to the Ca21-induced hydrophobic pockets in the corresponding penta-EF-handdomains (Supplementary Fig. 3). Previously, homology modellingbased on the Ca21-dependent complex between DVI and region C ofcalpastatin predicted the mode of binding for region A11. Regions A

and B of calpastatin engage sites in DIV and DI–III, respectively.Region B associates with DI–III to obstruct the active site in theextended substrate-like orientation. Intrinsic disorder in the freestate enables calpastatin to adapt structurally on binding to thesubstrate-binding cleft of calpain, distantly resembling the inhibitoryconformation of the broad-specificity, structured protease inhibi-tors, the cystatins12 (Supplementary Fig. 4). Region B is anchoredon either side of the active site C105S and avoids proteolysis byforcing a kink (Gly 174-Ile-Lys-Glu-Gly 178) between the flankingresidues—Leu 173 at the substrate binding S2 subsite and Thr 179at S19 (Fig. 2a, b). The unprimed and primed substrate binding sub-sites are N- and C-terminal, respectively, from the scissile bond P1–P19. S and P distinguish protein and substrate subsites, respectively.N-terminal to the kink, residues Val 161-Leu 170 participate in hydro-gen bonds mediated by backbone atoms (Supplementary Table 2) andhydrophobic interactions with the DIII surface that juxtaposes theactive site (Fig. 2b). The remainder of region B binds to the DI–II(Fig. 2a). Ala 172 marks the S3 site whereas the Leu 173 interaction isof particular importance at the S2 pocket, the main specificity deter-minant of calpains13,14. The side chain of Leu 173 at S2 is superimpo-sable to Leu 2 of leupeptin in complex with the protease core DI–II ofm-calpain, mI–II (ref. 15; Supplementary Fig. 5). C-terminal to thekink, region B engages the S19-S29-S39 subsites with residues Thr 179-Ile-Pro 181. The conformation of Pro 181-Pro-Glu-Tyr 184 changesthe direction as the Glu 183-Lys 190 helix targets DI.

We tested the mechanism of inhibition by shortening the kinkthrough deletion of Lys 176, Glu 177 or both. In all instances, thelow nanomolar half-maximal inhibitory concentration (IC50) valuesfor m-calpain inhibition, derived from initial rate analysis, did notchange significantly compared to wild type (Supplementary Fig. 6a).However, all mutants succumbed to proteolysis within the kink,permitting catalytic cleft access and resulting in complete auto-proteolysis of the complex within hours (Fig. 2c). The 5-residue kinkis therefore essential to overcome proteolysis and its length hasindeed been conserved (Supplementary Fig. 2b)6. An 18-residue pep-tide B1, specific for DI–II (Ala 172-Glu 189), inhibited m-calpainwith an IC50 of 410 6 80 nM (mean 6 s.e.m., Supplementary Fig.6b). Others have shown that a 27-residue peptide, correspondingto Asp 163-Glu 189 of region B, inhibited m-calpain with an IC50 of,30 nM16. Residues Leu 173-Gly 174 and Thr 179-Ile-Pro 181 cor-responding to the substrate subsites P2-distorted P1 and P19-P29-P39, respectively, were identified as ‘hotspots’ that impaired inhibi-tion when replaced with Ala (ref. 17). Replacing the intact 27-residuepeptide with the corresponding N- and C-terminal peptides—Asp 163-Gly 174 and Lys 175-Ala 189 (human sequence)—abolishedcalpain inhibition17. Taken together, these data suggest that DIIIanchoring by region B enhances inhibitory activity .10-fold andsupport the structure-based occluding-loop mechanism. The

1Department of Immunology, St Jude Children’s Research Hospital, 332 N Lauderdale, Memphis, Tennessee 38105, USA. 2Department of Biochemistry, McGill University, 3655Promenade Sir William Osler, Montreal, Quebec H3G 1Y6, Canada.

Vol 456 | 20 November 2008 | doi:10.1038/nature07353

404

©2008 Macmillan Publishers Limited. All rights reserved

Ca21-dependent reversible interaction between calpastatin and cal-pain is biologically relevant and, in light of the length of the occludingloop, it indicates that the wild-type inhibitor, unlike the shorter loopmutants, may recycle once dissociated from the protease. This over-comes the need for new synthesis of calpastatin to maintain inhibi-tion under conditions of fluctuating Ca21 levels.

Only regions A, C and the DI–II-binding residues of B are con-served among calpastatins (Supplementary Fig. 2b). The divergentintervening sequences connecting regions A, B and C are devoid ofelectron density (red dots in Fig. 1a). NMR analysis of the complexbetween 15N-labelled calpastatin and unlabelled calpain (Fig. 1b andSupplementary Fig. 2a) corroborated that the intervening sequencesof calpastatin are intrinsically disordered (Supplementary Fig. 2b).Conversely, calpastatin regions A, B and C bind calpain, becomeordered and tumble slowly in solution as part of the 111 kDa

complex, and were undetectable by NMR. DIII contains hotspotsfor proteolysis in the free and Ca21-bound m-calpain18, and itsextensive interaction with calpastatin in the complex supports itsresistance to trypsin (Supplementary Fig. 2c). The modular organ-ization of calpastatin induced on binding to calpain emphasizes therole of the interspersed, flexible/disordered segments in the folding–binding transitions of the structured regions at distant sites incalpain. Similar disorder was confirmed by NMR for a complexbetween m-calpain and repeat 1 of calpastatin at 10 mM CaCl2(ref. 19). Calpastatin regions A and C interact with calpain whereasthe intervening regions are disordered19. Notably, calpastatin regionB corresponding to Lys 176-Leu 188, which in our complex targetsmainly DI, also interacts with calpain. At this CaCl2 level the proteasecore is not aligned for catalysis and, therefore, the N terminus of

DI

DII

DIII

DIVDVI

Region ARegion C

Inhibitoryregion B

A135

C146

P159E192

H209

T219

Heterodimerinterface

Stabilizedactive site rim

a

108

110

112

114

116

118

120

122

8.1 7.6 7.1 6.6

7.9 8.1 7.8 8.0

G131 G225 G196

T224

T149

S130D153

T195S222

I194

K151D152N150

K129

K226

T149 N150 K151 D152

1H (p.p.m.)

b

Occludedactive site cleft

15N

(p.p

.m.)

Figure 1 | Complex between Ca21-bound m-calpain and calpastatin.a, Overall structure shows regions A, B and C of calpastatin bound to DIV,DI–III and DVI of calpain, respectively. The intervening sequences ofcalpastatin are devoid of electron density (red dots). The central part of theinhibitory region B forms the occluding loop at the active site. The active sitein the protease core DI–II is stabilized by DIII. Calpain heterodimerization islargely defined at the DIV–DVI interface. Alternate conformations at theinterface between the DI–III core and the DIV–DVI heterodimer (blackdots) are possible30. b, 15N,1H-HSQC (heteronuclear single quantumcorrelation) spectrum of the complex between 13C,15N-labelled calpastatin(residues 128–226, Supplementary Fig. 2b) and unlabelled calpain identifiedflexible/disordered residues of calpastatin. Connected sample strips from theHNCA NMR experiment are inset.

S241

H169

G261

S105

W288Q290V259

K69

L165

D162

K161

A101L102

H262

DI

L170

A172 (P3)

L173 (P2)

G174 (distorted P1)

T179 (P1`)

E183

Y184

L188

K190

I180 (P2`)

P182

E189

α8α9

α10

T455

R457R461

T464F465

I466

D454R375

N376F489

N467 V161

D163M165

D166S167T168

Y169

L170

L162

P164

β5

P181 (P3′)

L187

DII D243 R337

DIII

|a

|b

1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6 M kDa

0.5× 5×Wild type

K176∆

75

25

15

5037

20

10

75

25

5037

20

10

1× 2×

15

174–178‘loop-out’

cM

Figure 2 | Binding and inhibition of calpain by region B of calpastatin.a, Detailed view of the interaction between calpastatin and the catalytic cleftin the protease core DI–II. At the subsite P1 calpastatin distorts from thesubstrate path and projects residues 174–178, which kink between the P2and P19 anchor sites. b, Detailed view of the interaction between calpastatinand a surface-accessible groove in DIII. Hydrogen bonds are represented asdashed lines. c, Autolysis of m-calpain (10 mM) in the presence of 1 mMCaCl2 and wild-type calpastatin or the K176D mutant was monitored at 0, 5,20, 80, 320 min, and 24 h (lanes 1–6, respectively) by SDS–polyacrylamidegel electrophoresis. Calpastatin:calpain (I:E) molar ratios and the size of themolecular markers (lane M) are indicated above and to the left of the panel,respectively. Unprocessed calpain subunits and calpastatin are identified(arrows). Autolysis is blocked by wild-type calpastatin above I:E of 2. TheK176D mutant fails to inhibit autolysis even at higher I:E, being digested inthe loop to produce Coomassie-stained fragments (asterisks).Fragmentation of the catalytic subunit is complete after ,1 h.

NATURE | Vol 456 | 20 November 2008 LETTERS

405

©2008 Macmillan Publishers Limited. All rights reserved

region B does not contact the unprimed side of the active site nor theDIII. The disorder in the calpain-bound calpastatin prompted us topredict minimal global conformational changes in calpain instigatedby calpastatin binding. This suggests that the rearrangement of calpaindomains is mainly induced by Ca21. Local calpastatin-induced con-formational changes in calpain are predicted for the gating loops of theprotease core, found in alternative conformations in the structure ofmI–II free or bound to inhibitors (Supplementary Figs 5 and 7)7,15.

The calpain–calpastatin structure offers an unprecedented oppor-tunity to study the Ca21-bound conformation of m-calpain (Fig. 3a).The structures of inactive calpain9,20–22, the Ca21-bound proteasecore7,23,24 and Ca21-bound and free DVI (refs 25, 26) have generatedvaluable, incomplete models for calpain activation8. The calpain–calpastatin structure (Fig. 1a) represents the Ca21-activated con-formation of m-calpain revealed by the realignment for catalysis ofthe protease core DI–II by two Ca21 atoms, as described for mI–II(ref. 7; Supplementary Fig. 7 and Supplementary Table 3). DI–II isintimately associated with DIII, which undergoes conformationalchanges to interact specifically with DI, serving to stabilize the prote-ase core and to maximize its catalytic activity. The EF-hand DIV andDVI bind four Ca21 atoms each27, mediate the heterodimer interfaceby pairing of EF-hand 5 as in the apo-calpain9 and show smallchanges between the Ca21-bound and free conformation(Supplementary Fig. 8 and Supplementary Table 3). Of significanceis the Ca21-induced displacement from DVI of the N-terminalanchor peptide9, which is unstructured in the complex.

To investigate the Ca21-induced conformational changes leadingto the heterodimer activation (Supplementary Video), we aligned theCa21-bound and free structure20 by overlapping DIII, which providesa central scaffold for the (re)arrangements of the vicinal domainsthrough protein–protein interactions (Fig. 3a). On binding Ca21,the upper DI–II lobe moves dorsal to frontal with respect to DIII,whereas the lower DIV–DVI lobe moves in the opposite direction.The tension on either side of the protease core, postulated in light of

the Ca21-free m-calpain structure9 and confirmed extensively bio-chemically8, is overcome by Ca21 binding (Fig. 3a). The discovery ofthe active conformation of the DI–III ensemble is significant as itidentifies the missing conserved features at the extensive DI–II–DIIIinterface (Fig. 3b).

We extrapolated the importance of this interface from structuralanalysis of the isolated protease core DI–II of m-calpain, mI–II.Owing to intrinsic instability of residues Gly 197-Gly 210 in DI, theunprimed side of the active site in mI–II collapses diminishing acti-vity .1,000-fold compared to full-length m-calpain24. In the Ca21-bound heterodimer, residues 197–210 are stabilized, through saltbridges and hydrogen bonds, by conserved basic residues in DIII(Fig. 3b and Supplementary Fig. 10). In particular, Arg 417 andArg 420 from the basic loop, which adopts a different conformationin Ca21-free calpain (Supplementary Fig. 9), and Arg 469 andArg 500 from the b-sandwich core of DIII may provide critical sup-port for the labile active site (Fig. 3b).

In limb-girdle muscular dystrophy (LGMD)-2A patients, p94 (cal-pain 3) point mutants in DIII result in the typical atrophic pheno-types associated with impaired p94 activity in limb-girdle and trunkmuscles5. The positions 490, 493, 541 and 572 in p94, correspondingto the conserved Arg residues (417, 420, 469 and 500, respectively) inDIII of m-calpain, are mutated to Trp, Gln or Pro in both familial andsporadic forms of the disease (http://www.dmd.nl)28. We probed theDI–II–DIII active interface by mutagenesis in m-calpain (Fig. 3c andSupplementary Fig. 11). The R417A and R420A substitutionsdecreased m-calpain activity to one-half and one-quarter, respect-ively, and doubled the Ca21 requirement in the latter. The R469Adecreased activity .60-fold and doubled the Ca21 requirement.Conservative Lys substitutions at 417, 420 or 469 did not rescuethe phenotypes of Ala mutations (Supplementary Fig. 11), collec-tively underscoring the importance of this interface for sustainingmaximal calpain activity and providing an explanation for the effectof disease-causing p94 substitutions in LGMD-2A patients.

Ca2+-freeacidic loop

DI

DII

DIII

DIV

DVIRelaxed504–530

linker

DII–DIIIlinker

Active recoiledbasic loop

α6α8α9 α10

β5

α5

α62

4

2

1

3

1

1

2

3

4

Basic α11

Basic α11

Ca2+-freeacidic loop

Extended504–530

linker

Inactive openbasic loop

DI

DII

DIII

DIV

DVI

AnchorT1

T2

T3

Y146

S196G197

D206E205

W106

E202

G209

R500E470

R469

R420

R417

H415

β6

β3

α9

α10

A194

G210

W144Recoiled

basic loop

Stabilized197–210AS rim

Ca2+-bound Ca2+-freea b

S105 C105H262

H262

N286N286

Alignedcatalyticallycompetent

AS

Catalyticallyincompetent

AS

c

Nor

mal

ized

m-c

alp

ain

activ

ity

1.0

0.5

0

100 1,000[CaCl2] (µM)

Figure 3 | Stabilization of active Ca21-bound calpain by DIII. a, The Ca21-bound and free calpain (PDB: 1KFU)20 were aligned structurally based onDIII. The Ca21-bound DI–II is freed from inactivating tension at theinterfaces between the anchor and DVI (T1), the acidic loop of DIII and thebasic helix of DII (T2), and the basic loop of DIII and DVI (T3)9,21. In Ca21-bound calpain, the DIII–DIV linker is relaxed by shedding region 505–513from the DIII core. DIII is Ca21-free and contacts DI–II and DIV (pinkareas) extensively, but not DVI. The basic region in DIII stabilizes the labile

active site (AS), and the DII–DIII linker contacts the basic helix in DII.Numbering of helices and strands is according to Supplementary Fig. 10.b, Detailed view of the interaction between the basic region in DIII and theactive site. c, Ala substitutions in DIII at Arg 417 (red), Arg 420 (light blue)and Arg 469 (dark blue) impaired m-calpain activity, normalized withrespect to the wild type (black); for the latter two it doubled the Ca21

requirement for half-maximal activity. Values are expressed as mean 6 s.d.

LETTERS NATURE | Vol 456 | 20 November 2008

406

©2008 Macmillan Publishers Limited. All rights reserved

Calpastatin inhibits m- and m-calpains6, which share the 28 kDaregulatory subunit predicted in both to bind calpastatin region Csimilarly (Supplementary Figs 3b and 12). Modelling of the complexbetween m-calpain and calpastatin illustrated that calpastatin regionsA and B bind conserved pockets of the 80 kDa catalytic subunitand probably produce a similar set of inhibitory interactions withm-calpain (Supplementary Fig. 12). We reported that mI–II is notinhibited by calpastatin repeat 1 (ref. 7). Here we show that trimmingdown calpastatin repeat 1 from either end to the 18-residue peptideB1 (Ala 172-Glu 189) increased inhibitory potency for mI–II. The 86-residue repeat 1 (134–219) reduced mI–II activity with an IC50 of154.7 6 16.3 mM, whereas the 57-residue regions BC (Asp 163-Thr 219) and AB (Ala 144-Lys 190, Fig. 1d), and the peptide B1 inhi-bited mI–II with IC50 values of 121.8 6 24.4, 80.5 6 10.2 and24.5 6 5.3 mM, respectively (mean 6 s.e.m., Supplementary Fig. 6b,c). In contrast, repeat 1 and fragments AB and BC exhibited lownanomolar IC50 values towards m-calpain (Supplementary Fig.6d). Our results confirm the specificity of calpastatin for the hetero-dimeric m- and m-calpains and suggest that calpains that lack thesmall subunit, DIV and/or DIII may not support the same set ofinhibitory interactions with calpastatin8.

The calpain and calpastatin proteins represent a major ubiquitouscellular proteolytic system, the imbalance of which has been impli-cated in necrosis associated with stroke and neuronal injury andperhaps Alzheimer’s disease, heart disease, cataract formation, type2 diabetes, cancer and LGMD-2A29. Our study shows the mechan-isms of activation by Ca21 and inhibition by calpastatin of m- andm-calpains (Fig. 4). Additional mechanisms of regulation for thecalpain–calpastatin system include phosphorylation, membrane tar-geting and differential localization (Supplementary Fig. 13). Thedetails of calpastatin specificity for the heterodimers may aid thedesign of new therapeutic agents.

METHODS SUMMARY

Repeat 1 of calpastatin and the m-calpain heterodimer, composed of an intact

catalytic subunit and lacking the Gly-rich DV of the regulatory subunit, were

expressed in Escherichia coli separately and were purified to homogeneity by fast-

performance liquid chromatography (FPLC). Excess radio-labelled (for NMR)

or unlabelled inhibitor was mixed with the protease in the presence of Ca21

followed by further FPLC purification of the complex. The complex was sub-

jected to limited proteolysis, NMR spectroscopy analysis, crystallization and

X-ray crystallography structure determination. On the basis of the NMR assign-

ments a shorter calpastatin fragment was generated and only its complex to

m-calpain produced crystals. X-ray diffraction data collection was performed

at three different US synchrotrons to improve the resolution. The crystal struc-

ture was obtained by molecular replacement using domains from previously

determined calpain structures as search models. The active conformation

of calpain and the mechanism of inhibition by calpastatin were probed by

site-directed mutagenesis using standard calpain activity and calpastatin inhibi-

tion assays on the basis of calpain autolysis and hydrolysis of fluorogenic

substrates. Truncated calpastatin constructs and synthetic peptides were tested

for inhibitory specificity against full-length and truncated calpain.

Full Methods and any associated references are available in the online version ofthe paper at www.nature.com/nature.

Received 6 March; accepted 15 August 2008.

1. Franco, S. J. & Huttenlocher, A. Regulating cell migration: calpains make the cut. J.Cell Sci. 118, 3829–3838 (2005).

2. Wang, K. K. Calpain and caspase: can you tell the difference? Trends Neurosci. 23,20–26 (2000).

3. Harris, F., Biswas, S., Singh, J., Dennison, S. & Phoenix, D. A. Calpains and theirmultiple roles in diabetes mellitus. Ann. NY Acad. Sci. 1084, 452–480 (2006).

4. Liu, J., Liu, M. C. & Wang, K. K. Calpain in the CNS: from synaptic function toneurotoxicity. Sci. Signal. 1, re1 (2008).

5. Kramerova, I., Beckmann, J. S. & Spencer, M. J. Molecular and cellular basis ofcalpainopathy (limb girdle muscular dystrophy type 2A). Biochim. Biophys. Acta1772, 128–144 (2007).

6. Wendt, A., Thompson, V. F. & Goll, D. E. Interaction of calpastatin with calpain: areview. Biol. Chem. 385, 465–472 (2004).

7. Moldoveanu, T. et al. A Ca21 switch aligns the active site of calpain. Cell 108,649–660 (2002).

8. Suzuki, K., Hata, S., Kawabata, Y. & Sorimachi, H. Structure, activation, andbiology of calpain. Diabetes 53, S12–S18 (2004).

9. Hosfield, C. M., Elce, J. S., Davies, P. L. & Jia, Z. Crystal structure of calpain revealsthe structural basis for Ca21-dependent protease activity and a novel mode ofenzyme activation. EMBO J. 18, 6880–6889 (1999).

10. Elce, J. S., Hegadorn, C., Gauthier, S., Vince, J. W. & Davies, P. L. Recombinantcalpain II: improved expression systems and production of a C105A active-sitemutant for crystallography. Protein Eng. 8, 843–848 (1995).

11. Todd, B. et al. A structural model for the inhibition of calpain by calpastatin: crystalstructures of the native domain VI of calpain and its complexes with calpastatinpeptide and a small molecule inhibitor. J. Mol. Biol. 328, 131–146 (2003).

12. Bode, W. & Huber, R. Structural basis of the endoproteinase–protein inhibitorinteraction. Biochim. Biophys. Acta 1477, 241–252 (2000).

13. Cuerrier, D., Moldoveanu, T. & Davies, P. L. Determination of peptide substratespecificity for m-calpain by a peptide library-based approach: the importance ofprimed side interactions. J. Biol. Chem. 280, 40632–40641 (2005).

14. Cuerrier, D. et al. Development of calpain-specific inactivators by screening ofpositional scanning epoxide libraries. J. Biol. Chem. 282, 9600–9611 (2007).

15. Moldoveanu, T., Campbell, R. L., Cuerrier, D. & Davies, P. L. Crystal structures ofcalpain–E64 and –leupeptin inhibitor complexes reveal mobile loops gating theactive site. J. Mol. Biol. 343, 1313–1326 (2004).

16. Betts, R., Weinsheimer, S., Blouse, G. E. & Anagli, J. Structural determinants of thecalpain inhibitory activity of calpastatin peptide B27-WT. J. Biol. Chem. 278,7800–7809 (2003).

17. Betts, R. & Anagli, J. The beta- and gamma-CH2 of B27-WT’s Leu11 and Ile18 sidechains play a direct role in calpain inhibition. Biochemistry 43, 2596–2604(2004).

18. Moldoveanu, T., Hosfield, C. M., Jia, Z., Elce, J. S. & Davies, P. L. Ca21-inducedstructural changes in rat m-calpain revealed by partial proteolysis. Biochim.Biophys. Acta 1545, 245–254 (2001).

19. Kiss, R. et al. Calcium-induced tripartite binding of intrinsically disorderedcalpastatin to its cognate enzyme, calpain. FEBS Lett. 582, 2149–2154 (2008).

DI

DII

DIII

DIV

DVI

Ca2+-freeacidicloop

Basicloopopen

LabileAS rim

Basicα11

34

1

2

12

12

3

4

DI

DII

DIII

DVI

DIV

DII–DIIIlinker

Inactive Active

DI

DII

DIII

DVI

DIV

Inhibited by calpastatin

Region A

Region C

StabilizedAS rim

‘Loop-and-block’AS cleft

Ca2+

4

1

2

12

2

34

3

1

Ca2+-freeacidicloop

Basicloop

recoiled

Extendedtransducer

Relaxedtransducer

Calpastatin IUP

Inhibitoryregion B

C B

A

Figure 4 | Calpain–calpastatin proteolytic system. A schematic diagramillustrating the Ca21-induced activation of calpain and its inhibition bycalpastatin. DIII has a fundamental role in relaying the Ca21-inducedstructural changes (red dotted arrows) from the peripheral domains to the

catalytically competent yet labile protease core. Concerted binding of theintrinsically unstructured protein (IUP) calpastatin to peripheral domainsand the active site of calpain results in low-nanomolar inhibition.

NATURE | Vol 456 | 20 November 2008 LETTERS

407

©2008 Macmillan Publishers Limited. All rights reserved

20. Reverter, D. et al. Flexibility analysis and structure comparison of two crystalforms of calcium-free human m-calpain. Biol. Chem. 383, 1415–1422 (2002).

21. Strobl, S. et al. The crystal structure of calcium-free human m-calpain suggests anelectrostatic switch mechanism for activation by calcium. Proc. Natl Acad. Sci. USA97, 588–592 (2000).

22. Pal, G. P., De Veyra, T., Elce, J. S. & Jia, Z. Crystal structure of a micro-like calpainreveals a partially activated conformation with low Ca21 requirement. Structure 11,1521–1526 (2003).

23. Davis, T. L. et al. The crystal structures of human calpains 1 and 9 imply diversemechanisms of action and auto-inhibition. J. Mol. Biol. 366, 216–229 (2007).

24. Moldoveanu, T., Hosfield, C. M., Lim, D., Jia, Z. & Davies, P. L. Calpain silencing bya reversible intrinsic mechanism. Nature Struct. Biol. 10, 371–378 (2003).

25. Blanchard, H. et al. Structure of a calpain Ca21-binding domain reveals a novel EF-hand and Ca21-induced conformational changes. Nature Struct. Biol. 4, 532–538(1997).

26. Lin, G. D. et al. Crystal structure of calcium bound domain VI of calpain at 1.9 Aresolution and its role in enzyme assembly, regulation, and inhibitor binding.Nature Struct. Biol. 4, 539–547 (1997).

27. Dutt, P., Arthur, J. S., Grochulski, P., Cygler, M. & Elce, J. S. Roles of individual EF-hands in the activation of m-calpain by calcium. Biochem. J. 348, 37–43 (2000).

28. Jia, Z. et al. Mutations in calpain 3 associated with limb girdle muscular dystrophy:analysis by molecular modeling and by mutation in m-calpain. Biophys. J. 80,2590–2596 (2001).

29. Saez, M. E., Ramirez-Lorca, R., Moron, F. J. & Ruiz, A. The therapeutic potential ofthe calpain family: new aspects. Drug Discov. Today 11, 917–923 (2006).

30. Hanna, R. A., Campbell, R. L. & Davies, P. L. Calcium-bound structure of calpainand its mechanism of inhibition by calpastatin. Nature doi:10.1038/nature07451(this issue).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements We thank P. L. Davies for facilitating the initial stages of thiswork and for sharing the coordinates for the calpastatin repeat 4–calpain complex,M. Osborne for help with the NMR analysis, B. Schulman for collecting the ALSsynchrotron data, the staff at Brookhaven National Light Source beam X29 and APSSERCAT for help with data collection, S. White and D. Miller for advice and sharingof synchrotron time, and C. Hosfield, A. Tocilj and R. Kriwacki for reviewing ourmanuscript. Funding was provided by the US National Institute of Health andNational Cancer Institute, the Canadian Institute of Health Research and theAmerican Lebanese Syrian Associated Charities.

Author Contributions T.M. performed the experimental work and data analysis.T.M., K.G. and D.R.G. were involved with the project planning and wrote the paper.

Author Information The sequences and three-dimensional coordinates for thecalcium-dependent complex between m-calpain and calpastatin have beendeposited in the PDB database under accession number 3DF0. Reprints andpermissions information is available at www.nature.com/reprints.Correspondence and requests for materials should be addressed to D.R.G.(douglas.green@stjude.org).

LETTERS NATURE | Vol 456 | 20 November 2008

408

©2008 Macmillan Publishers Limited. All rights reserved

METHODSCloning, mutagenesis, peptide synthesis, protein expression and purification.The rat calpastatin repeat 1 clone encoding residues Met 119-Ser 238

(Supplementary Fig. 2b, gi 13540322) was provided by S. Arthur as an

N-terminally His10-tagged construct in pET16b vector. Subsequent cloning in

the NcoI and XhoI sites of the kanamycin-resistant pET24d vector was per-

formed using this vector as PCR template and the following oligonucleotides

as primers, and engineering a stop codon to exclude the C-terminal His6 tag:

59med GCATGGCCATGGACAAGTCAGGCGTGAATGCTG, 39med GTGG-

TGCTCGAGTTACTTTCCAGTTGGAGAGCTACAG, 59sh GCATGGCCATG-GCTGCTTTGGATGACCTGATAG, 39sh GTGGTGCTCGAGTTAGGTGAAA-

TCAGATGACCAGGCA, 59BC GCATGGCCATGGACCCAATGGATTCTAC-

CTAC and 39BC GTGGTGCTCGAGTTAACAGGTGAAATCAGATGACA-

AGGC. We produced medium-sized (Met-Asp 128-Lys 226) and short (Met-

Ala 134-Thr 219) calpastatin repeat 1 constructs and peptide BC (Met-

Asp 163-Cys 220). In the short construct, we introduced a stop codon after

Lys 190 using the Quick Change protocol (Stratagene) and the forward primer

GAAACTTCTGGAGAAATAAGAAGCTATCACAGG (reverse not shown) to

generate peptide AB (Met-Asp 128-Lys 190). The E. coli BL21 DE3 strain was

used to express all five derivatives of calpastatin. In all calpastatin constructs the

initiating Met was removed during expression as indicated by intact mass deter-

mination by mass spectrometry. The m-calpain heterodimer, C105S m80 kDa/

21 kDa, which lacks the glycine-rich DV, and the protease core from m-calpain

(mI–II) were expressed in E. coli and purified as described previously7,10. Forward

mutagenesis primers for the 80 kDa subunit included R417A

CCAGAAGCATCGGGCGCGGCAGAGGAA, R420A GGCGGCGGCAGGCG-

AAGATGGGTGAG and R469A CCTTCATCAACCTCGCGGAGGTCCTCAAC,

and for calpastatin K176D GGCACTGGGTATAGAAGGGACTATTCC, E177DGCACTGGGTATAAAAGGGACTATTCCTC and K176/E177D GGCACTGGG-

TATAGGGACTATTCCTC. For 15N-labelling and 13C,15N-labelling, the medium-

sized and short calpastatins were expressed in M9 medium31 in the presence of 15N-

NH4Cl and 15N-NH4Cl plus 13C-glucose, respectively. All forms of calpastatin were

purified by boiling the cell lysate for 15 min, followed by Ni-NTA affinity chro-

matography (the original His-tagged construct), Q-Sepharose and C18 reversed-

phase HPLC or S200 gel filtration chromatography. All protein preparations were

exchanged into 20 mM HEPES (pH 7.5), 5 mM DTT storage buffer, were flash-

frozen in liquid nitrogen, and stored at 280 uC. Peptide B1 Ac-

ALGIKEGTIPPEYRKLLE-NH2 was synthesized and HPLC-purified at St Jude

Children’s Research Hospital Hartwell Center.

Calpain–calpastatin complex formation and purification. The m-calpain–cal-

pastatin complex was formed by slowly titrating 50 mM CaCl2 into a solution

(,20–30 ml) containing purified calpain (10–20 mg) and a 2–53 molar excess

of calpastatin in 50 mM HEPES buffer (pH 7.5) until the CaCl2 concentration

reached ,5–10 mM. The subsequent steps of purification were done in solutions

containing 5 mM CaCl2 to prevent dissociation of the complex. Excess untagged

calpastatin was removed by recovering the complex on Ni-NTA (Qiagen), whichbound to the column using the C-terminal His6-tag on the calpain large subunit.

Further purification involved Sephacryl S200 and Q-Sepharose chromato-

graphy. The complexes were stored in 20 mM HEPES (pH 7.5), 5 mM DTT,

10 mM CaCl2 at 280 uC.

Limited proteolysis and autolysis of the calpain–calpastatin complex. Limited

proteolysis18 or autolysis reactions were performed at 22 uC in a final volume of

less than 150ml, in 1 mM CaCl2 and 50 mM HEPES at pH 7.5. Purified 1 mg ml21

(9mM) m-calpain–calpastatin complex was proteolysed by 0.001 mg ml21

(0.04mM) trypsin (Sigma). For autolysis, 1 mg ml21 (10mM) m-calpain was

incubated at 1 mM CaCl2 with wild-type or mutant full-length calpastatin or

fragments AB and BC. At specific time intervals, aliquots were removed, the

reaction was stopped by the addition of 23 SDS gel sample buffer, and the

products were analysed by SDS PAGE.

Calpastatin inhibition assays. Hydrolysis assays by m-calpain (1 nM or 20 nM),

m-calpain mutants (1 nM) and mI–II (1.25–2.5mM) were performed in 100-ml

volumes containing 50 mM HEPES (pH 7.5), 200 mM NaCl, 1 mM DTT, various

concentrations of substrate (0–1.5 mM SLY-MCA (Sigma) or 0–50mg ml21

BODIPY-casein (Invitrogen)) and increasing inhibitor concentrations:

0–50 nM wild-type or mutant full-length calpastatin, fragments AB and BC,

0–20mM peptide B1 for m-calpain, and 0–750mM peptide B1 for mI–II.

Duplicates or triplicates were performed for each condition in 96-well plates

using a Molecular Devices microplate reader.

NMR assignments of calpastatin flexible regions in complex with calpain. The

complexes between 15N-labelled medium or short calpastatin and unlabelled

calpain was subjected to 15N,1H HSQC analysis in the storage buffer supplemen-

ted with 10% D2O at 25 uC. The spectra were collected on a Bruker Avance DRX

600 MHz spectrometer equipped with a triple-resonance CryoProbe. HNCA and

CBCA(CO)NH experiments with 15N/13C-labelled Asp 128-Lys 226 calpastatin

bound to unlabelled calpain were used to assign the mobile regions of the

inhibitor. Spectral processing was done using NMRpipe32 and spectral analysis

using nmrView33.

Calpain–calpastatin complex crystallization and structure determination.After extensive screening and expansions, the complex between m-calpain and

the NMR-trimmed Ala 134-Thr 219 calpastatin was crystallized in 4–9% PEG

3350, 5–10 mM CaCl2 and 50–100 mM NaOAc (pH 5.5) using the hanging-drop

method by mixing equal volumes of complex (10–15 mg ml21) and mother

liquor. Cryo conditions included mother liquor and up to 30% ethylene glycol.

The crystals were assigned to the tetragonal space group P42 with one molecule

per asymmetric unit and diffracted to 2.8–3.5 A at synchrotrons (Supplementary

Table 1). Multiple native data sets were collected at 5.0.1-3 beamlines of the

Advanced Light Source, X29 beamline at Brookhaven National Laboratories, andat SERCAT 22BM and 22ID beamlines at the Advanced Photon Source. The

structure was determined by molecular replacement using as search models the

Ca21-bound rat m-calpain protease core (1MDW), the DIII and DIV of Ca21-

free rat m-calpain (1DF0) and the Ca21-bound small subunit of rat m-calpain

(1DVI) and Phaser34 included in the CCP4 package35. Structural refinement was

processed using Refmac5 (ref. 36) and CNS37, with manual fitting performed

using Xfit38.

31. Sambrook, J., Fritsch, E. F. & Maniatis, T. Molecular Cloning: A Laboratory Manual(Cold Spring Harbor Laboratory Press, 1989).

32. Delaglio, F. et al. NMRPipe: a multidimensional spectral processing system basedon UNIX pipes. J. Biomol. NMR 6, 277–293 (1995).

33. Johnson, B. A. Using NMRView to visualize and analyze the NMR spectra ofmacromolecules. Methods Mol. Biol. 278, 313–352 (2004).

34. McCoy, A. J. et al. Phaser crystallographic software. J. Appl. Cryst. 40, 658–674(2007).

35. Collaborative Computational Project, Number 4. The CCP4 suite: Programs forProtein Crystallography. Acta Crystallogr. D 50, 760–763 (1994).

36. Winn, M. D., Isupov, M. N. & Murshudov, G. N. Use of TLS parameters to modelanisotropic displacements in macromolecular refinement. Acta Crystallogr. D 57,122–133 (2001).

37. Brunger, A. T. et al. Crystallography and NMR system: a new software system formacromolecular structure determination. Acta Crystallogr. D 54, 905–921 (1998).

38. McRee, D. E. A visual protein crystallographic software system for X11/XView. J.Mol. Graph. 10, 44–46 (1992).

doi:10.1038/nature07353

©2008 Macmillan Publishers Limited. All rights reserved