Crystal structure of a Trypanosoma brucei metacaspaseKaren McLuskeya,1, Jana Rudolfa, William R. Protoa, Neil W. Isaacsb, Graham H. Coombsc, Catherine X. Mossa,and Jeremy C. Mottrama,1

aWellcome Trust Centre for Molecular Parasitology, Institute of Infection, Immunity, and Inflammation, College of Medical, Veterinary, and Life Sciences,University of Glasgow, Glasgow G12 8TA, United Kingdom; bSchool of Chemistry, College of Science and Engineering, University of Glasgow, Glasgow G128QQ, United Kingdom; and cStrathclyde Institute of Pharmacy and Biomedical Sciences, University of Strathclyde, Glasgow G4 0RE, United Kingdom

Edited by Robert Huber, Max Planck Institute for Biochemistry, Planegg-Martinsried, Germany, and approved March 28, 2012 (received for review January23, 2012)

Metacaspases are distantly related caspase-family cysteine pepti-dases implicated in programmed cell death in plants and lowereukaryotes. They differ significantly from caspases because they arecalcium-activated, arginine-specific peptidases that do not requireprocessing or dimerization for activity. Toelucidate thebasis of thesedifferences and to determine the impact they might have on thecontrol of cell death pathways in lower eukaryotes, the previouslyundescribed crystal structure of ametacaspase, an inactivemutant ofmetacaspase 2 (MCA2) from Trypanosoma brucei, has been deter-mined to a resolution of 1.4 Å. The structure comprises a core caspasefold, but with an unusual eight-stranded β-sheet that stabilizes theprotein as a monomer. Essential aspartic acid residues, in the pre-dicted S1 binding pocket, delineate the arginine-specific substratespecificity. In addition, MCA2 possesses an unusual N terminus,which encircles the protein and traverses the catalytic dyad, withY31 acting as a gatekeeper residue. The calcium-binding site is de-fined by samarium coordinated by four aspartic acid residues,whereas calcium binding itself induces an allosteric conformationalchange that could stabilize the active site in a fashion analogous tosubunit processing in caspases. Collectively, these data give insightsinto the mechanistic basis of substrate specificity and mode of acti-vationofMCA2andprovide adetailed framework for understandingthe role ofmetacaspases in cell death pathways of lower eukaryotes.

apoptosis | clan CD | parasite | X-ray crystallography

Programmed cell death (PCD) is essential for animal develop-ment and the maintenance of adult tissues. PCD itself is a reg-

ulated process, and in animals, apoptosis is controlled through theaction of caspases, aspartic acid-specific cysteine peptidases (1).PCD is also essential for plant development (2), and multiplemarkers for apoptosis have been described in yeast and a broadrange of protozoan parasites (3), yet these organisms do not en-code caspases in their genomes; thus, alternative pathways musthave evolved for caspase-independent cell death to be regulated.In recent years, attention has focused on the metacaspases, whichare a highly conserved group of caspase-like cysteine peptidasesfound in plants, fungi, and protozoa but not in the metazoa (4). Inplants, metacaspases are essential for embryogenesis (5, 6) andhave been shown to act in antagonistic relationships, functioningas both positive and negative regulators of PCD (7). In yeast andsome protozoa, metacaspases have been implicated as mediatorsof cell death in response to oxidative stress and environmentalchange (3, 8–10).Various attempts have been made to draw parallels between

caspases and metacaspases in respect to structure, activation, andfunction (11). However, two striking differences between meta-caspases and caspases are their substrate specificities andrequirements for activation. Metacaspases have arginine/lysinespecificity, do not necessarily require processing or dimerizationfor activity, and are activated by calcium (5, 12–15). There are fewknown natural targets of the proteolytic activity of metacaspases.One, Tudor staphylococcal nuclease, a substrate for mcII-Pametacaspase in the Norway spruce, Picea abies, is also a substratefor human caspase-3, and this has been taken as indicating a levelof evolutionary conservation between PCD pathways (16). It isnow apparent, however, that metacaspases have a variety of cel-lular functions that are distinct from those of the caspases, which

may reflect the distinct differences in substrate specificity andactivation (11). For example, in yeast, the Yca1 metacaspase isrequired for clearance of cellular protein aggregates as well asregulation of the timing of the cell cycle, a function that has alsobeen identified in the parasitic protozoon Leishmania (9, 17–19).Metacaspases have been classified as members of the clan CD

(20) structural superfamily, which is represented by six peptidasefamilies: the caspases, legumains, separases, clostripains, gingi-pains, and the repeats in toxin (RTX) self-cleaving toxins (cysteineprotease domain), all of which are predicted to share a core cas-pase/hemoglobinase fold (CHF) containing the catalytic His/Cysdyad. The core of the CHF domain is a simple α/β-fold (21) de-fined by a four-stranded, parallel β-sheet and three conservedα-helices. Currently, there are crystal structures available for thecaspases (1) [including paracaspase (22)], gingipain (23), and anRTX toxin (24). Metacaspases are described as distant relatives ofthe caspases (4) and, despite low sequence identity with the cas-pases, have been grouped [along with the Arg-specific para-caspases (4, 22)] into a structural subfamily of the caspase-typepeptidases (family C14B). In the caspase structures, the funda-mental catalytic domain consists of a six-stranded β-sheet, whichcontains a large subunit (p20) and a small subunit (p10) (1) sep-arated by an intersubunit linker between strands 4 and 5. Activecaspases are functional dimers. The recently identified structure ofthe caspase domain of the mucosa-associated lymphoid tissuelymphoma translocation 1 (MALT1-C) paracaspase revealed anidentical topology (22). Caspases are activated by processing ofthe intersubunit linker or by dimerization (1, 25). Metacaspasesare divided into two structural types based on their primary aminoacid sequence (4, 11). Both types have been predicted to exhibita caspase-like domain composed of p20 and p10 subunits (26) butdiffer in that only type I has a predicted N-terminal prodomain,whereas type II contains an additional linker between the subunits.Plant genomes encode both types of metacaspases, whereas onlytype I has been reported in protozoa and fungi (11).The protozoan parasite Trypanosoma brucei is the causative

agent of human African trypanosomiasis, and its genome encodesfive metacaspases (MCA1–MCA5). MCA2, MCA3, and MCA5have the conserved catalytic residues (histidine/cysteine dyad) (27)and are collectively required for the bloodstream form of theparasite (28). MCA2, which is an active peptidase, is located inRAB11-positive endosomes and has roles in precytokinesis cellcycle control (28). T. brucei mutants lacking MCA2 have no ob-vious defect, indicating that functional redundancy exists betweenMCA2,MCA3, andMCA5.MCA4 is an inactive pseudopeptidase

Author contributions: K.M. and J.C.M. designed research; K.M., J.R., and C.X.M. performedresearch; W.R.P. contributed new reagents/analytic tools; K.M., J.R., N.W.I., G.H.C., C.X.M.,and J.C.M. analyzed data; and K.M., G.H.C., C.X.M., and J.C.M. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

Data deposition: The atomic coordinates and structure factors reported in this paper havebeen deposited in the Protein Data Bank in Europe, www.ebi.ac.uk/pdbe/ (PDB ID 4AF8,4AFP, 4AFR, and 4AFV).1To whom correspondence may be addressed. E-mail: karen.mcluskey@glasgow.ac.uk orjeremy.mottram@glasgow.ac.uk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1200885109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1200885109 PNAS | May 8, 2012 | vol. 109 | no. 19 | 7469–7474

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

8, 2

020

because of a cysteine-to-serine substitution in the active site, yet itfunctions as a membrane-linked virulence factor (29). MCA4 isalso bloodstream form-specific, released by the parasite and pro-cessed by MCA3 (29). This pseudopeptidase illustrates the di-versity of metacaspase function and the challenges faced inattempting to assign functional roles based on sequence alone.This study aimed to understand the structural basis for the dif-ferences between metacaspases and caspases, including the sub-strate recognition properties and mechanisms for activation of themetacaspases. To this end, we undertook structural and bio-chemical analysis of MCA2 from T. brucei, determining the pre-viously undescribed 3D structure of a metacaspase and providingincisive insights into how this family of enzymes functions.

Results and DiscussionMetacaspase Structure. Attempts to crystallize active MCA2(MCA2ac) were unsuccessful, most likely a result of the presenceof mixed degradation products caused by in vitro autoprocessingof the full-length recombinant enzyme (14). Therefore, the crystalstructures of two catalytically inactive forms ofMCA2 [MCAC213G

and MCAC213A (Table S1)] were determined to 1.4 and 1.6 Å,respectively (Tables S2 and S3). The MCAC213G and MCAC213A

structures are essentially identical (SI Methods). The highest res-olution structure of MCA2C213G is the most complete, and theterm MCA2 is used to describe the overall structure, unless oth-erwise stated. The MCA2C213G structure consists of residues 3–347, with the exception of two disordered loop regions consistingof residues 166–170 and 269–275 (denoted the 280-loop). MCA2is monomeric and comprises an eight-stranded β-sheet (β1–β8)consisting of six parallel and two antiparallel strands with2↑1↑3↑4↑7↑8↓5↑6↓ topology (30) (Figs. 1 and 2D). The first fourparallel strands lie in the same plane, exhibiting only a very slightright-handed twist, whereas the second half of the sheet is moretwisted, with the C terminus of β7 rotating out of the plane ofthe sheet and crossing under the N terminus of β8 (Fig. 1A). Theβ-sheet is surrounded by five α-helices (α1–α5) that lie approxi-mately parallel to it. Helices α1, α4, and α5 are found on one sideof the sheet, with α2 and α3 on the opposite side. At the C ter-minus of β3, a small section of antiparallel β-sheet (βA-βC) isfound with strands βB and βC, connected by a 10-residue linkercontaining a short section of α-helix (αA). This section of β-sheet ispositioned at the N terminus of α2 and α3, capping the helices andthe space between them (Fig. 1).The residues that constitute the catalytic dyad (H158 and

C213G/A) are found on loops situated at the C-terminal end ofstrands β3 and β4, respectively (Figs. 1 and 2D). The loop con-necting β4 and β5 comprising the catalytic cysteine (denoted thecatalytic loop) is 11 residues long (Figs. 1B and 2D) and showsa degree of flexibility (as defined by poorer electron density andhigher B-factors in this region). One of the most striking featuresof the MCA2 structure is the 70-residue N terminus preceding β1.This domain is an ordered loop containing only a small region of310-helix (α0′), two short β-strands (β0′ and β0), and a seven-res-idue α-helix (α0) (Figs. 1 and 3A). It encircles the main body of theenzyme, traversing the top in the region of the catalytic dyad andinteracting with itself as it crosses underneath (Fig. 3A). InMCA2C213A, there are 129 residues contributing to the interfacebetween the N-terminal domain and the main body of the protein,creating 32 hydrogen bonds and eight salt bridges. The interfacingresidues include the catalytic dyad of H158 and C213A, which are100% and 50% buried by the N terminus, respectively.

Comparison with Caspase-7. The overall fold of MCA2 is compa-rable to that of the mammalian caspases (SI Methods). It is moststructurally similar to caspase-7, with 65% of its secondarystructural elements identified in the caspase-7 structure. How-ever, the structure of MCA2 reveals some striking architecturaldifferences from the caspases, most notably in the length andarrangement of its β-sheet and in its extended N terminus (Fig.2). The β-sheet in MCA2 is two strands longer than in caspase-7,which exhibits a six-stranded β-sheet with 2↑1↑3↑4↑5↑6↓ topology.These differences result in the monomeric structure of MCA2

being much less compact than that of a caspase monomer(Fig. 2). MCA2 exists as an enzymatically active monomer (Fig.S1), and the loop regions on its surface create an unfavorableenvironment for the type of β-sheet/β-sheet dimerization found inthe caspases (Fig. 2). In addition, a structural overlay with thecaspase-7 dimer reveals that β5 from MCA2 overlays with β6 fromthe second caspase monomer (β6′), and thus occupies the spaceused in the caspase-dimer interface and renders it unavailable (Fig.2 B and D). Hence, the larger eight-stranded β-sheet in MCA2prevents it from forming a dimer along the same interface as thecaspases and stabilizes the protein as a monomer.Recombinant MCA2ac undergoes autocatalytic processing at

K55 in the N-terminal region and at K268 in the 280-loop (14)(Figs. 1B and 2D). However, cleavage does not generate separatesubunits because the C-terminal region is embedded in the β-sheetand the N-terminal region remains associated with the main bodyof the protein (Fig. S1). This processing has not been observedin vivo (28) and is not required for the enzyme to be proteolyticallyactive (14). Consequently, the autoproteolytic processing ofrecombinant MCA2ac in vitro may simply reflect the high con-centrations of the enzyme and the position of basic residues onsolvent-exposed loops.

Fig. 1. Structure and sequence of T. bruceiMCA2C213G. General loop regionsare shown in gray, the N-terminal region in cyan, α-helices in blue, andβ-strands in red, apart from βA-βC, which is shown in purple. (A) Tertiarystructure of the MCA2 monomer with the position of the catalytic dyad inyellow. (B) MCA2 sequence with the assigned secondary structure. The cata-lytic dyad is highlighted in red, and residues involved in the S1 pocket andsamarium binding are shown in orange and cyan, respectively. Known cleav-age sites (14) are highlighted (▼), and residues that are missing from theMCA2C213G electron density are shown in light gray and with a dashed line.

7470 | www.pnas.org/cgi/doi/10.1073/pnas.1200885109 McLuskey et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

8, 2

020

Effector caspases, such as caspase-7, are activated by cleavage inthe intersubunit linker generating the p10 and p20 subunits (Fig.2B). This cleavage allows part of the intersubunit linker from onecaspase monomer to interact with the neighboring monomer andhelps to order the active site (31). MCA2 is very unusual in thisrespect; the topology of the β-sheet is different, with β4 and β5 notbeing adjacent but, instead, separated by two antiparallel strands(β7 and β8) (Fig. 2D), and the catalytic loop that connects thesestrands does not harbor any basic residues as potential cleavagesites (Fig. 1B), and, indeed, none have been identified (14). Inaddition, this loop is much shorter than the intersubunit linker inthe caspases, suggesting that it does not have the same functionalrelevance. Thus, the eight-stranded β-sheet, the length and to-pology of which preclude caspase-like dimerization and explain itsmonomeric peptidase activity, clearly distinguishes MCA2 fromother caspase family members. This is further supported by thelack of an intersubunit linker preventing the formation of twodistinct subunits or activation by cleavage in this region. Conse-quently, the activation mechanism of MCA2 is quite distinct fromthat of the caspases.

N-Terminal Domain and Active Site. The catalytic dyad of MCA2 isburied beneath the N-terminal domain, which crosses the activesite in the direction of the N terminus to the C terminus (Fig. 3B).This is the same direction as the peptide inhibitors bound to the

other clan CD peptidases (23, 32, 33) and the predicted directionfor an MCA2 substrate. A 3D alignment of the structures ofMCA2 and caspase-7 revealed that the position of the catalyticdyad is conserved between the two enzymes (Fig. 2 B and D), andthis is also the case for clan CD members from other families:gingipain R (23) and the cysteine peptidase domain of the Vibriocholerae RTX (24). However, in all the structures apart fromMCA2, the dyad is found on the surface of the protein. In contrast,theN-terminal domain ofMCA2 appears to block substrate accessto the active site sterically, with the nature of the residues at thecrossover preventing intramolecular autocleavage.Occlusion of the active site by an N-terminal domain is also

observed in the proforms of the lysosomal cathepsins [clan CA,family C1 (20)], where proteolytic cleavage and removal of theprodomain are required for activation (34). However, in ca-thepsin B (35), the prodomain is only weakly associated with theenzyme, through five hydrogen bonds, and it crosses the activesite in the opposite orientation to that observed in MCA2.Structural elucidation of a caspase prodomain is yet to be ach-ieved, but it has been reported to be flexible, and hence disor-dered, in the crystal structures (36). In contrast, the N-terminaldomain of MCA2 was both visible and highly ordered in ourcrystal structures, with its position clearly showing that it wouldbe required to undergo a conformational shift to allow substratesaccess into the active site.Examining the structure of MCA2 without its N-terminal re-

gion reveals that the catalytic dyad forms part of a large acidicpocket (Fig. 3B), consistent with a binding site for the basic P1amino acid of an MCA2 substrate. The functional groups of C92,D95, S156, and D211 all point toward the surface of the S1pocket, and the bottom is lined with three well-ordered watermolecules (Fig. 3C), the locations of which were found to beconserved in all the MCA2 crystal structures. The N terminus ofMCA2 crosses the S1 pocket and the catalytic dyad at residues30–36, with Y31 forming hydrogen bonds to D95 and S156, alongwith a water molecule, which is, in turn, coordinated to a secondwater molecule and D211 (Fig. 3D). There are no other hydro-gen bonds between the S1 pocket and the N-terminal domain.Residues C92, D95, S156, and D211 are conserved within themetacaspase family, whereas Y31 is conserved in the N-terminaldomain of T. brucei metacaspases (Fig. S2). In addition to theirconservation, their location and orientation in the S1 pocketsuggest that these residues play a part in substrate recognition. Toinvestigate their significance, single alanine point mutations inMCA2acwereconstructedand theirproteolytic activitywas assessedusing a fluorogenic peptide substrate (Z-GGR-AMC; Fig. S3).D95 is positioned at the back of the S1 pocket in close proximity

to D211. Alanine mutations of these residues resulted in enzymesthat exhibited no autoprocessing in vitro or any noteworthyactivity toward Z-GGR-AMC compared with MCA2ac (0% forMCA2D95A, <5% for MCAD211A) (Fig. 3 E and F). Consequently,these aspartic acid residues are aptly positioned to recognize andcoordinate the basic side chain of an arginine or lysine residue andare involved in the enzyme’s substrate specificity in P1.MCA2C92Ashowed approximately half of the activity toward the substratecompared with MCA2ac, and the equivalent of C92 in Arabidopsisthaliana metacaspase 9 (C29) has similarly been shown to be im-portant for activity (37). Thus, C92 of MCA2 contributes to cor-rect formation of the substrate binding pocket. MCA2Y31A

exhibited a much higher degree of autoprocessing in vitro thanMCA2ac, or any of the other mutants (Fig. 3E), suggesting thatY31 controls access to the active site. MCA2Y31A had only 60% ofMCA2ac activity, probably attributable to the extensive autopro-cessing degrading the enzyme. In contrast MCA2S156A activity wasthreefold higher thanMCA2ac. We propose a model whereby Y31acts as a gatekeeper by using S156 as a latch; when the latch isreleased, substrate can access the active site. These data areconsistent with the N-terminal domain acting to regulate the ac-tivity of the enzyme by protecting the active site until it is activatedby calcium (see below).The participation of D95, D211, and Y31 in S1 binding of

MCA2 is further supported by analysis of substrate binding incaspase-7 andMALT1-C. In caspase-7, P1 substrate recognition is

Fig. 2. Comparison of MCA2 and caspase-7. Ribbon diagram of caspase-7(A) and metacaspase (C). General loops are colored gray/black, the mainβ-sheet is colored red, surrounding α-helices are shown in blue, and a smallsection of the β-sheet at the C terminus of β3 is shown in green. In MCA2,secondary structural elements found on the N-terminal domain are coloredcyan. (B) Topology diagram of caspase-7. The 341-loop and the position ofthe catalytic dyad are highlighted, and the β-strands are numbered from theN terminus. The intersubunit linker is drawn as a red line with the cleavagesite highlighted (▲). (D) Topology diagram of MCA2, with the catalytic loopshown in red. The 280-loop affected during the Ca2+ activation of MCA2 andknown autoprocessing sites (▲) (14) are shown. The CHF core secondarystructural elements are highlighted (*) in B and D.

McLuskey et al. PNAS | May 8, 2012 | vol. 109 | no. 19 | 7471

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

8, 2

020

dependent on the three highly conserved residues, R87, Q184, andR233 (1), and in MALT1-C, three negatively charged residues,D365, D462, and E500, were found to coordinate the P1Arg of theinhibitor (22). The structural overlay revealed that the secondarystructure of MCA2 is conserved with caspase-7 residues R87 andQ184, overlaying with residues L89 and D211, respectively. L89would not be important in recognizing a basic metacaspase sub-strate; however, D95, which sits nearby on the N-terminal end ofα1 and overlays with D365 inMALT1-C, was shown to be essentialforMCA2 activity. R233 of caspase-7, which lies on a loop (termedthe 341-loop; ref. 1, Fig. 2B) at the C terminus of β5 and undergoessignificant conformational changes during activation (33), has noordered equivalent in MCA2. In the structure of MALT1-C, thisloop contains the P1 binding residue E500. In theMCA2 structure,this corresponds to the disordered loop region at the C terminus ofβ7 (denoted the 280-loop; Figs. 1B and 2D), suggesting that thisloop could play a role in substrate binding of MCA2. In addition,Y31 from the MCA2 N-terminal domain occupies the same posi-tion as R233 in caspase-7 (and E500 in MALT1-C), implying thatY31 would be involved in binding a substrate at the P1 position.Despite the distinct structural and functional differences be-

tween the caspases and metacaspases, MCA2 uses residues atsimilar structural positions in its S1 binding pocket to other cas-pase family members to determine its substrate specificity. Al-though two of the three proposed MCA2 P1 binding residues donot share conserved secondary structure with the caspases, theposition of the residues in the S1 binding pocket remains pre-served. This suggests a common substrate recognition mechanismfor caspases, paracaspases, and metacaspases.

Calcium Binding/Activation. MCA2 and other metacaspases are de-pendent on calcium for activation (5, 12–15). Furthermore, thepeptide-based inhibitor Z-VRPR-FMK inhibits the autoprocessingactivity of MCA2ac, and calcium is required for inhibitor binding

(Fig. 4A and SI Methods). Varying concentrations of CaCl2 wereadded to crystallization trials in an attempt to identify calcium-binding sites in the protein. However, even concentrations as low as10 μM inhibited crystal formation. Lanthanides are often used todefine calcium-binding sites in proteins (38); consequently, samar-iumwas tested as a heavy atom derivative for the structure solution.A samarium-derivatized crystal (MCA2C213G-Sm) was obtainedthat revealed a single, fully occupied Sm3+-binding site on thesurface of the molecule (Fig. 4B). The samarium ion exhibits a co-ordination number of 7 from two water molecules and four highlyconserved aspartic acid residues: D173, D189, D190, and D220.These residues are well conserved in metacaspases from a diverserange of organisms (Fig. S2) but are not found in the caspases orparacaspases, which correlates with their lack of calcium-dependentactivation. The structures of MCA2C213G and MCA2C213G-Sm areessentially identical, with the exception that MCA2C213G-Sm lackselectron density for several additional residues in the 280-loop(266–279 compared with 269–275 for MCA2C213G; Table S3),showing an increased degree of flexibility in this region.To investigate the location of the calcium-binding site further,

a double mutant, MCA2D189A/D190A, was made. In the presence ofcalcium, this mutant had no detectable activity against Z-GGR-AMC (specific activity <0.14 mU/mg), nor did it autoprocess (Fig.4C). Furthermore, CD spectra of MCA2D189A/D190A confirmedthat the mutation had not adversely affected the overall structureof the protein (Fig. S4). This confirms the role of these asparticacid residues in the putative calcium-binding site and shows theyare essential forMCA2 activity. To gain additional information onhow calcium influences MCA2 activity, various CaCl2 soaks werecarried out on MCA2C213A crystals (MCA2C213A-Ca). Two data-sets were collected from each crystal soak at wavelengths of 0.97 Åand 1.7 Å, but Ca2+ could not be detected in any of these datasets(SI Methods). However, structures determined from calcium-soaked crystals revealed a large Ca2+-induced conformational

Fig. 3. N-terminal domain of MCA2 crosses over the S1 binding pocket. (A) N-terminal domain (blue) encircles the main body of the enzyme (gray). (B)Electrostatic surface potential of MCA2C213A (rotated ∼90° from A), calculated without the N-terminal domain, where blue and red denote positively andnegatively charged surface areas, respectively. The N-terminal domain is shown crossing the catalytic dyad as sticks with Y31 labeled. The water moleculesthat line the active site are shown as cyan spheres. (C) Close-up view of the S1 binding pocket. (D) Y31 interacts with residues in the S1 binding pocket. (E) SDS/PAGE analysis of in vitro autoprocessing exhibited by MCA2 and mutants purified by Ni2+-affinity chromatography. (F) Activity of MCA2ac and mutants wascompared by measuring the hydrolysis of the fluorogenic substrate Z-GGR-AMC (average of 3 experiments is shown ± SD).

7472 | www.pnas.org/cgi/doi/10.1073/pnas.1200885109 McLuskey et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

8, 2

020

change in the position of the 280-loop. This loop changes directionat the Cα position of A280, resulting in a 5.6-Å, 110° shift in thepositions of the Cα positions of G279 between MCA2C213A-CaandMCA2C213A (Fig. 4, inset). This movement also resulted in thedisorder of the first 15 residues of the N-terminal domain ofMCA2C213A-Ca, which does not appear to be functional but, in-stead, is a direct result of the 280-loop movement in the crystals.In both MCA2C213A and MCA2C213A-Ca, residues 280 and 281

form a section of β-sheet with N-terminal residues 32 and 31, re-spectively. In MCA2C213A, this loop then moves away from the N-terminal domain and crosses over a short loop between α5 and β8.InMCA2C213A-Ca, the 280-loop continues to run parallel, forminga section of β-sheet with the N-terminal domain (Fig. 4E). Incontrast to MCA2C213A, the 280-loop in MCA2C213A-Ca is in linewith the wall of the S1 binding pocket, sitting directly above the

catalytic dyad (Fig. 4E). Because the 280-loop is structurally ho-mologous to the 341-loop in the caspases (Fig. 2 B and D), itappears that in MCA2, the 280-loop moves to stabilize substratesin the active site in response to calcium binding. We also foundthat calcium binding alone does not induce a conformational shiftin the N-terminal region crossing the active site suggesting that thepresence of a substrate is also required to trigger the release of theN-terminal latch.

Conclusions. The 1.4-Å structure of T. brucei MCA2 has giveninsights into the substrate specificity and mode of activation ofMCA2. It has revealed that MCA2 possesses an unusual N-ter-minal domain, which likely serves to regulate substrate access tothe active site until the enzyme is activated by calcium in thepresence of a substrate. In MCA2, calcium is required for both

Fig. 4. Allosteric binding of calcium induces a conformational shift in MCA2C213A. (A) Inhibitor Z-VRPR-FMK binds to MCA2ac on the addition of Ca2+, causinga small size shift on SDS/PAGE, and protects MCA2ac from autoproteolysis in the presence of Ca2+. (B) Samarium ion (Sm3+) binds to the surface ofMCA2C213G viafour aspartic acid residues and twowater molecules (cyan spheres), defining the Ca2+-binding site. (C) SDS/PAGE analysis of in vitro autoprocessing exhibited byMCA2ac and MCA2D189A/D190A. (D) Composite diagram shows the structure of MCA2C213A-Ca (gray) and its 280-loop (+Ca2+, red) with the catalytic dyad high-lighted in orange and the N-terminal domain shown in green. Addition of the 280-loop from theMCA2C213A structure (APO, blue) and the bound Sm3+ ion fromtheMCA2C213G-Sm structure shows the 280-loop movement with respect to the position of the calcium-binding site. Missing residues from the 280-loop regionsare represented by dotted lines. (Inset) Structure of MCA2C213A-Ca shows an approximate 5.6-Å, 110° shift in the position of the MCA2C213A 280-loop. (E) In thestructure of MCA2C213A-Ca, the 280-loop (red) forms a section of the β-sheet with the N-terminal domain (green) that sits above the catalytic dyad (orange).

McLuskey et al. PNAS | May 8, 2012 | vol. 109 | no. 19 | 7473

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

8, 2

020

activation of the enzyme and binding of the peptide inhibitor Z-VRPR-FMK, and it was found to induce an allosteric conforma-tional change in the 280-loop, bringing it closer to the active site.This suggests that MCA2 requires this calcium-induced loopmovement to stabilize substrate binding, and hence to regulateactivity. Little is known about Ca2+ concentrations and/or Ca2+signaling in African trypanosomes, and nothing is known aboutCa2+ in the RAB11-positive recycling endosome compartment inwhich MCA2 primarily resides. However, endoplasmic reticulumstress-induced cell death in Leishmania is Ca2+-dependent (39);thus, it will be interesting to investigate whether there is any re-lationship between such cell death and the Ca2+ activation centralto metacaspase function.The structural similarity of MCA2 to the caspases suggests that

they evolved from a common ancestor. However, the length andconstruction of their β-sheets, and the differences identified in theintersubunit linker and domain structure, show that they are ar-chitecturally very distinct. Furthermore, the lack of structurallyconserved autocatalytic cleavage sites and dimerization interfacesin MCA2 is consistent with MCA2 having a mode of activationdifferent from that of the caspases. These findings, together withmetacaspase’s occluded active site, distinct substrate specificity,and dependence on calcium for activation (14), show that MCA2activity is regulated very differently from that of the caspases.Thus, caspases andmetacaspasesmay have evolved independentlyfrom an ancestral metacaspase-like peptidase, with each family ofenzymes evolving distinct activation mechanisms to regulate celldeath pathways. Overall, this work provides a context for the de-sign of metacaspase-specific inhibitors that can potentially be usedfor the development of novel antiparasite drugs (40).

MethodsThe construction of plasmids, purification of protein, and enzyme assays werecarried out as described previously (14) and in SI Methods. Initial crystals ofMCA2C213G or MCA2C213A were grown at 4 °C using vapor diffusion techni-ques in a 96-well sitting drop plate (Innovadyne) containing 500 nL of pro-tein and 500 nL of reservoir against a reservoir of 50 mM Hepes (pH 7.0),0.1% tryptone, and 20% (wt/vol) PEG3350 (PEG/Ion screen; Hampton). Thesecrystals were composed of multiple plates and were optimized in 24-wellsitting drop plates (Cryschem; Hampton) using microseeding techniques.Crystals were flash-frozen in liquid nitrogen using the reservoir solution plus20% (vol/vol) 2-methyl-2,4-pentanediol (MPD) as a cryoprotectant beforedata collection. A samarium-derivatized crystal (MCA2C213G-Sm) was pre-pared by soaking an MCA2C213G crystal in artificial mother liquor [AML; 50mM Hepes (pH 7.0), 0.1% tryptone, 20% PEG3350] containing 60 mM sa-marium acetate for 2 min. In addition, an MCA2C213A crystal soaked in 5 mMCaCl2 for 1 h (MCA2C213A-Ca) was prepared and cryoprotected in AML con-taining 20% MPD and 1 mM CaCl2. All native and multiwavelength anom-alous dispersion diffraction data were collected at Diamond Light Source(DLS; Didcot, United Kingdom) on beamlines I03 and/or I04 and processedwith Fast_dp at DLS. Details of data collection, structure determination, andstructural analyses are described in SI Methods.

ACKNOWLEDGMENTS. We thank E. Brown for help with cloning, A. Scottfor protein purification, T. Z. Murray for inhibitor assays, S. M. Kelly for CDanalysis, M. Gabrielsen for helpful discussion, and DLS for access (ProposalMx6683). This work was supported by Wellcome Trust Grant 091790 andMedical Research Council Grant 0700127. The Wellcome Trust Centre forMolecular Parasitology is supported by core funding from Wellcome TrustGrant 085349.

1. Fuentes-Prior P, Salvesen GS (2004) The protein structures that shape caspase activity,specificity, activation and inhibition. Biochem J 384:201–232.

2. Bozhkov PV, Filonova LH, Suarez MF (2005) Programmed cell death in plant em-bryogenesis. Curr Top Dev Biol 67:135–179.

3. Reece SE, Pollitt LC, Colegrave N, Gardner A (2011) The meaning of death: Evolutionand ecology of apoptosis in protozoan parasites. PLoS Pathog 7:e1002320.

4. Uren AG, et al. (2000) Identification of paracaspases and metacaspases: Two ancientfamilies of caspase-like proteins, one of which plays a key role in MALT lymphoma.Mol Cell 6:961–967.

5. Bozhkov PV, et al. (2005) Cysteine protease mcII-Pa executes programmed cell deathduring plant embryogenesis. Proc Natl Acad Sci USA 102:14463–14468.

6. Suarez MF, et al. (2004) Metacaspase-dependent programmed cell death is essentialfor plant embryogenesis. Curr Biol 14:R339–R340.

7. Coll NS, et al. (2010) Arabidopsis type I metacaspases control cell death. Science 330:1393–1397.

8. Herker E, et al. (2004) Chronological aging leads to apoptosis in yeast. J Cell Biol 164:501–507.

9. Zalila H, et al. (2011) Processing of metacaspase into a cytoplasmic catalytic domainmediating cell death in Leishmania major. Mol Microbiol 79:222–239.

10. Khan MA, Chock PB, Stadtman ER (2005) Knockout of caspase-like gene, YCA1, ab-rogates apoptosis and elevates oxidized proteins in Saccharomyces cerevisiae. ProcNatl Acad Sci USA 102:17326–17331.

11. Tsiatsiani L, et al. (2011) Metacaspases. Cell Death Differ 18:1279–1288.12. Vercammen D, et al. (2004) Type II metacaspases Atmc4 and Atmc9 of Arabidopsis

thaliana cleave substrates after arginine and lysine. J Biol Chem 279:45329–45336.

13. Watanabe N, Lam E (2005) Two Arabidopsis metacaspases AtMCP1b and AtMCP2bare arginine/lysine-specific cysteine proteases and activate apoptosis-like cell death inyeast. J Biol Chem 280:14691–14699.

14. Moss CX, Westrop GD, Juliano L, Coombs GH, Mottram JC (2007) Metacaspase 2 ofTrypanosoma brucei is a calcium-dependent cysteine peptidase active without pro-cessing. FEBS Lett 581:5635–5639.

15. Watanabe N, Lam E (2011) Calcium-dependent activation and autolysis of Arabidopsismetacaspase 2d. J Biol Chem 286:10027–10040.

16. Sundström JF, et al. (2009) Tudor staphylococcal nuclease is an evolutionarily con-served component of the programmed cell death degradome. Nat Cell Biol 11:1347–1354.

17. Lee RE, Puente LG, Kaern M, Megeney LA (2008) A non-death role of the yeastmetacaspase: Yca1p alters cell cycle dynamics. PLoS ONE 3:e2956.

18. Lee RE, Brunette S, Puente LG, Megeney LA (2010) Metacaspase Yca1 is required forclearance of insoluble protein aggregates. Proc Natl Acad Sci USA 107:13348–13353.

19. Ambit A, Fasel N, Coombs GH, Mottram JC (2008) An essential role for the Leishmaniamajor metacaspase in cell cycle progression. Cell Death Differ 15:113–122.

20. Rawlings ND, Barrett AJ, Bateman A (2010) MEROPS: The peptidase database. NucleicAcids Res 38(Database issue):D227–D233.

21. Aravind L, Koonin EV (2002) Classification of the caspase-hemoglobinase fold: De-tection of new families and implications for the origin of the eukaryotic separins.Proteins 46:355–367.

22. Yu JW, Jeffrey PD, Ha JY, Yang X, Shi Y (2011) Crystal structure of the mucosa-as-sociated lymphoid tissue lymphoma translocation 1 (MALT1) paracaspase region. ProcNatl Acad Sci USA 108:21004–21009.

23. Eichinger A, et al. (1999) Crystal structure of gingipain R: An Arg-specific bacterialcysteine proteinase with a caspase-like fold. EMBO J 18:5453–5462.

24. Lupardus PJ, Shen A, Bogyo M, Garcia KC (2008) Small molecule-induced allostericactivation of the Vibrio cholerae RTX cysteine protease domain. Science 322:265–268.

25. Boatright KM, Salvesen GS (2003) Mechanisms of caspase activation. Curr Opin CellBiol 15:725–731.

26. Vercammen D, Declercq W, Vandenabeele P, Van Breusegem F (2007) Are meta-caspases caspases? J Cell Biol 179:375–380.

27. Mottram JC, Helms MJ, Coombs GH, Sajid M (2003) Clan CD cysteine peptidases ofparasitic protozoa. Trends Parasitol 19(4):182–187.

28. Helms MJ, et al. (2006) Bloodstream form Trypanosoma brucei depend upon multiplemetacaspases associated with RAB11-positive endosomes. J Cell Sci 119:1105–1117.

29. Proto WR, et al. (2011) Trypanosoma brucei metacaspase 4 is a pseudopeptidase anda virulence factor. J Biol Chem 286:39914–39925.

30. Zhang C, Kim SH (2000) The anatomy of protein beta-sheet topology. J Mol Biol 299:1075–1089.

31. Shi YG (2004) Caspase activation: Revisiting the induced proximity model. Cell 117:855–858.

32. Shen A, et al. (2009) Mechanistic and structural insights into the proteolytic activationof Vibrio cholerae MARTX toxin. Nat Chem Biol 5:469–478.

33. Wei Y, et al. (2000) The structures of caspases-1, -3, -7 and -8 reveal the basis forsubstrate and inhibitor selectivity. Chem Biol 7:423–432.

34. Coulombe R, et al. (1996) Structure of human procathepsin L reveals the molecularbasis of inhibition by the prosegment. EMBO J 15:5492–5503.

35. Podobnik M, Kuhelj R, Turk V, Turk D (1997) Crystal structure of the wild-type humanprocathepsin B at 2.5 A resolution reveals the native active site of a papain-like cys-teine protease zymogen. J Mol Biol 271:774–788.

36. Vaidya S, Velázquez-Delgado EM, Abbruzzese G, Hardy JA (2011) Substrate-inducedconformational changes occur in all cleaved forms of caspase-6. J Mol Biol 406:75–91.

37. Belenghi B, et al. (2007) Metacaspase activity of Arabidopsis thaliana is regulated byS-nitrosylation of a critical cysteine residue. J Biol Chem 282:1352–1358.

38. Martin RB, Richardson FS (1979) Lanthanides as probes for calcium in biological sys-tems. Q Rev Biophys 12(2):181–209.

39. Dolai S, Pal S, Yadav RK, Adak S (2011) Endoplasmic reticulum stress-induced apo-ptosis in Leishmania through Ca2+-dependent and caspase-independent mechanism.J Biol Chem 286:13638–13646.

40. Berg M, et al. (2010) Design and evaluation of Trypanosoma brucei metacaspase in-hibitors. Bioorg Med Chem Lett 20:2001–2006.

7474 | www.pnas.org/cgi/doi/10.1073/pnas.1200885109 McLuskey et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

8, 2

020