Structural Basis of Proline-Proline Peptide Bond Specificity of the Metalloprotease Zmp1 Implicated in Motility of Clostridium D

Article

Structural Basis of Proline-Proline Peptide Bond Speciﬁcity of the Metalloprotease Zmp1 Implicated in Motility of Clostridium difﬁcile

Graphical Abstract Authors Magdalena Schacherl, Christian Pichlo, Ines Neundorf, Ulrich Baumann

Correspondence [email protected] (M.S.), [email protected] (U.B.)

In Brief Zmp1 is a proline-proline speciﬁc metalloprotease that displays an aliphatic-aromatic network of amino acid side chains and a unique loop to shape its active site for substrate binding. Schacherl et al. describe unbound and peptide-bound structures of Zmp1 that explain the high speciﬁcity for proline- proline peptide bonds.

Highlights Accession Numbers d The structure of Clostridium difﬁcile Zmp1 is similar to 5A0P anthrax lethal factor 5A0S 5A0R d The S-loop of Zmp1 undergoes large motions upon substrate 5A0X binding d An aliphatic-aromatic network and the diverting loop contribute to Pro-Pro speciﬁcity d An active site mutant E143A possesses remarkable residual activity

Structural Basis of Proline-Proline Peptide Bond Speciﬁcity of the Metalloprotease Zmp1 Implicated in Motility of Clostridium difﬁcile

Magdalena Schacherl,1,* Christian Pichlo,1 Ines Neundorf,1 and Ulrich Baumann1,* 1Institute of Biochemistry, University of Cologne, 50674 Cologne, Germany *Correspondence: [email protected] (M.S.), [email protected] (U.B.) http://dx.doi.org/10.1016/j.str.2015.06.018

SUMMARY (Brazier, 1998). C. difficile-associated disease (CDAD) is a severe and often fatal colon inflammation with symptoms ranging from Clostridium difficile is a pathogenic bacterium diarrhea (George et al., 1977) to pseudomembranous colitis causing gastrointestinal diseases from mild diarrhea (George et al., 1978) and toxic megacolon (Bartlett, 2006). The to toxic megacolon. In common with other patho- risk group for this disease are hospitalized patients older than genic bacteria, C. difficile secretes proteins involved 65 years after antimicrobial treatment, e.g. with clindamycin in adhesion, colonization, and dissemination. The (Bartlett et al., 1978; Thomas et al., 2003). Owing to its high recently identified Zmp1 is an extracellular metallo- recurrence and high mortality rate, C. difficile infections have become a major problem in hospitals and nursing homes (Makris protease showing a unique specificity for Pro-Pro and Gelone, 2007). The emergence of the ‘‘epidemic’’ or ‘‘hyper- peptide bonds. The endogenous substrates of virulent’’ strains BI/NAP1/027 in North America and Europe has Zmp1 are two surface proteins implicated in adhe- caused a drastic increase in new infections and fatality rates in sion of C. difficile to surface proteins of human cells. the last 15 years (O’Connor et al., 2009). Thus, Zmp1 is believed to be involved in the regula- Zmp1 is secreted by all C. difficile strains, including the virulent tion of the adhesion-motility balance of C. difficile. and multidrug-resistant NAP1 ribotype 027 strain R20291 (Ca- Here, we report crystal structures of Zmp1 from fardi et al., 2013). It is implicated in the regulation of C. difficile C. difficile in its unbound and peptide-bound forms. motility (Figure 1) through cleavage of the putative C. difficile The structure analysis revealed a fold similar to Bacil- adhesins CD2831 and CD3246 (Hensbergen et al., 2014), which lus anthracis lethal factor. Crystal structures in the makes it an interesting target to study. It was shown for other open and closed conformation of the S-loop shed human pathogens that proteases involved in adhesion, colonization, and dissemination are essential for infections and light on the mode of binding of the substrate, and recurrence, as for example reported for the regulation of reveal important residues for substrate recognition B. anthracis adhesion to human endothelial cells via the cleavage and the strict specificity of Zmp1 for Pro-Pro peptide of the major surface layer protein BslA by the secreted protease bonds. InhA (Tonry et al., 2012). Similarly, Bordetella pertussis SphB1 protease maturates the major adhesin filamentous hemaggluti- nin at the cell surface (Coutte et al., 2003b), and the knockout INTRODUCTION of SphB1 strongly affects the ability of B. pertussis to colonize the mouse respiratory tract in vivo (Coutte et al., 2003a). Taking Zinc-dependent endoproteases are involved in many essential into account these observations, Zmp1 could be a potential drug biological processes such as protein degradation, proenzyme target for CDAD treatment of new, but also of recurrent, activation, and amino acid supply (Neurath and Walsh, 1976). C. difficile infections. A majority of these enzymes belong to the MEROPS clan MA While no three-dimensional structural information is available (Rawlings et al., 2014) and exhibit the conserved sequence motif for Zmp1 or close orthologs, there is a weak similarity (less HEXXH, where the two histidines ligate the catalytic zinc ion, and than 20% sequence identity) to B. anthracis LF, which is the glutamic acid acts as catalytic base activating a zinc-bound composed of four domains (Pannifer et al., 2001), namely water molecule for nucleophilic attack on the scissile peptide the N-terminal protective-antigen-binding domain (PA-binding bond (Matthews, 1988). A subgroup of the MA clan is the M34 domain) I, domains II and III responsible for substrate binding, family of metalloendopeptidases (Rawlings et al., 2014). Among and the metalloprotease domain IV, which is distantly related its members are the lethal factor (LF) from Bacillus anthracis and to thermolysin (Matthews, 1988). the recently discovered Zmp1 from Clostridium difficile (Cafardi In LF the third zinc ligand is a glutamic acid C-terminal of the et al., 2013; Hensbergen et al., 2014). two histidines, and a tyrosine residue that is assumed to stabilize In the past decade, C. difficile has become one of the major the tetrahedral intermediate of the hydrolysis reaction is spatially causes of nosocomial antibiotic-associated diarrhea infections located near the zinc ion (Pannifer et al., 2001; Tonello et al., (Bouza, 2012). It is transmitted through the spore form of this 2004). LF is part of the tripartite anthrax toxin of B. anthracis Gram-positive anaerobic bacterium via the fecal-oral route (Smith et al., 1956; Beall et al., 1962). Its main substrates are

1632 Structure 23, 1632–1642, September 1, 2015 ª2015 Elsevier Ltd All rights reserved Figure 1. Model of Proposed Role of Zmp1 and Sequence Conservation in Zmp1 Substrates (A) Model of the possible role of Zmp1 in the regulation of motility in C. difficile infections. Secreted Zmp1, the endogenous C. difficile substrate adhesins CD2831 and CD3246, as well as the receptor molecules on the enterocyte surface are depicted. (B) Prevalent amino acids around the scissile bond (positions P4–P40) in the endogenous C. difficile substrates CD2831 and CD3246 in WebLogo format (Crooks et al., 2004). (C) Sequences found in the human protein substrates IgA1/2, Hsp90b, and fibrinogen, as in (A). (D) Amino acids at the P3 to P30 positions previously identified in a peptidic screen. The inverted triangles indicate the protease cleavage sites. mitogen-activated protein kinase kinases (MAPKKs) (Duesbery tween two prolines, we have determined crystal structures of the et al., 1998; Pellizzari et al., 1999; Vitale et al., 2000). The Zmp1 wild-type enzyme in the unbound state, as well as of conserved amino acids around the scissile bond in MAPKKs inactive mutants in complex with substrate- and product-like are (P1)[Pro/Arg/Lys/Gln/Gly];[Ile/Ala/Leu/Phe](P10), where ; peptides. These structures reveal a fold similar to anthrax LF cat- symbolizes the scissile bond and P1 and P10 denote the neigh- alytic domain where a double-kinked conformation of the sub- boring residues (Schechter and Berger, 1967). In a peptide strate is enforced by a unique loop of Zmp1 as well as by an screen the recognition motif was determined as (P1)Pro;[Tyr/ active-site cleft that is shaped by an alternating aliphatic-aro- Leu/Ile/Met/Phe/Val](P10)(Turk et al., 2004). matic side-chain stack. The shape and chemical nature of the C. difficile Zmp1, on the other hand, shows a very unique S sites nicely explain the substrate specificity of Zmp1. The specificity for Pro-Pro peptide bonds that is different from LF wild-type protein is crystallized in its fully active form, as demon- (Hensbergen et al., 2014). Its preferred recognition sequence strated by proteolytic activity assays based on fluorescence de- in proteinaceous substrates is (P4)Val-[Val/Leu/Ile]-Asn-Pro; quenching of synthetic peptides and on cleavage of two human Pro-Val-Pro-Pro(P40) with a stringent specificity for Asn in the substrates. P2, and Pro in the P1, P10, and P30 positions. Screening of a pep- This structural information can further be utilized for the devel- tide library (Hensbergen et al., 2014) resulted in a broader selec- opment of specific inhibitors that interfere with C. difficile tivity, accepting, for example, also less efficiently Ala in the P1 or motility, which may support other therapeutic strategies. the P10 position (Figures 1C and 1D). While there are numerous peptidases known to cleave Pro-X RESULTS and to a lesser extent X-Pro peptide bonds with X being any amino acid beside Pro (Cunningham and O’Connor, 1997), Proteolytic Activity of Zmp1 Constructs there are only few examples of Pro-Pro-bond cleavage. A survey Fibronectin and fibrinogen were examined as proteinaceous of the MEROPS database (Rawlings et al., 2014) resulted in substrates in an activity assay at 37C with subsequent SDS- 1,413 hits for Pro-X and 1,247 for X-Pro, but only 40 for Pro- PAGE analysis (Figures 2A and 2B). Fibronectin seems to be a Pro peptide bond cleavage events; of the latter, 14 are attributed rather poor substrate for Zmp1, as even after 24 hr and a prote- to Zmp1. ase to substrate molar ratio of 5:1, no complete cleavage was To understand this striking sequence specificity and how achieved (Figure 2A). Fibrinogen (b chain), on the other hand, Zmp1 cleaves the proteolytically less sensitive peptide bond be- was almost completely digested by Zmp1 after 1 hr in a protease

Structure 23, 1632–1642, September 1, 2015 ª2015 Elsevier Ltd All rights reserved 1633 ure 2C). Compared to with wild-type Zmp1 (set to 100% cleavage after 1 hr at 37C), the single mutant Y178F displays 41% activity and E143A 18% activity (Figure 2C). The double mutant E143A/Y178F does not possess detectable proteolytic activity in this assay. This high residual activity of both single mutants is surprising. Usually, mutation of the glutamic acid of the HEXXH motif leads to virtually complete inactivation, as shown for numerous proteases of MEROPS clan MA, such as LF (Tonello et al., 2004), astacin (Yiallouros et al., 2000), and aminopeptidase A(Vazeux et al., 1996). Mutation of the transition state-stabilizing tyrosine also completely abolishes proteolytic activity in the case of LF and reduces it to a few percent in astacin.

Overall Structure of Zmp1 The data collection and reﬁnement statistics for all reported structures in this manuscript are given in Table 1. The overall structure of Zmp1 (Figure 3) consists of an upper N-terminal domain (NTD) and a lower helical C-terminal domain (CTD) with respect to the active-site helix a4, which harbors two of the three zinc-binding residues His142 and His146 and the catalytic base Glu143 in the typical metalloprotease motif (H142E143TAH146). Glu189 contributes to the second Zn2+ coordination sphere, with its side chain hydrogen bonding to the imidazole ring of His142. Similarly for His146, there is a water- mediated interaction with Asp149. The last ﬁve residues of helix a4 (Ile148-Val152) are folded as a p helix (in i+5 conformation), a feature occasionally found at the end of a helices. The NTD is composed of a twisted four-stranded b sheet, three a helices

(a1–a3) backing the b sheet, and three short 310 helices (h1– h3) connecting the strands b2 and b3. Furthermore, it accommodates the edge strand b3, the bulge edge segment, and the S-loop involved in substrate recognition and binding (nomencla- ture similar to that used by Tallant et al., 2010). The CTD features

four a helices (a5–a8) and one 310 helix (h4). The third zinc ligand Glu185 is located on helix a6, which connects the NTD and the CTD. This feature ascribes Zmp1 to the clan of gluzincins (Hooper, 1994). Opposite of the S-loop, the CTD harbors a further determinant of substrate specificity, namely the S10- Figure 2. Proteolytic Activity of Zmp1 Variants wall-forming segment. It carries a tyrosine residue (Tyr178) that (A) Activity of Zmp1 against human fibronectin. Fibronectin (1 mM) was incu- probably acts as an electrophile by stabilizing the tetrahedral bated with Zmp1 (5 mM) for 1, 2, 3, and 24 hr and analyzed via SDS-PAGE. The transition state, analogous to astacins (Grams et al., 1996; Yial- asterisks indicate cleavage products. (B) Activity of Zmp1 against human fibrinogen. Fibrinogen (1 mM) was incu- louros et al., 2000), serralysins (Hege and Baumann, 2001), and bated with Zmp1 (1 mM or 0.05 mM) for 1, 2, 3, and 24 hr and analyzed via SDS- thermolysin (Matthews, 1988). The phenolic OH group accepts a PAGE. The asterisk indicates the cleavage product (fibrinogen b chain). hydrogen bond from the indole ring of Trp103, which is located at (C) Activity of Zmp1 variants wild-type (wt), E143A, Y178F, and the double the upper rim of the active-site cleft. mutant E143A/Y178F in a fluorescence dequenching-based assay using the peptide Dabcyl-KEVNPPVPDE-Edans after 1 hr. The activity of the wild-type S-loop Flexibility enzyme was set to 100%. Error bars represent SD (n = 3). Two alternative conformations of the S-loop were observed in the Zmp1 crystal structures reported here. In the unbound form to substrate molar ratio of 1:1, and after 24 hr in a protease to (Zmp1 wild-typeunbound and Zmp1 E143Aunbound), one of the substrate molar ratio of 1:20 (Figure 2B). two molecules in the asymmetric unit shows an open conforma- To study the importance of the residues Glu143 and Tyr178, tion (Figure 4A), while the other and all four independent copies of which were assumed to be involved in the proteolytic mecha- the peptide complex structures reveal a closed conformation nism, different variants of Zmp1 were subjected to a de- (Figure 4B). In the latter, the S-loop completely covers the primed quenching-based fluorogenic activity assay. The internally side of the substrate-binding groove (Figure 3). Substantial quenched fluorescence resonance energy transfer (FRET)-pep- movements of the backbone but also of several side chains are tide Dabcyl-KEVNPPVPDE-Edans (50 mM) was incubated with necessary to perform this conformational change (Figure 4C), 1 mM of each Zmp1 variant: wild-type, the single mutants with the largest being performed by Lys101 shifting the Ca E143A and Y178F, and the double mutant E143A/Y178F (Fig- atom by 3 A˚ and the Nz atom by 10 A˚ . Analysis with the DynDom

1634 Structure 23, 1632–1642, September 1, 2015 ª2015 Elsevier Ltd All rights reserved Table 1. Data Collection and Reﬁnement Statistics Structure/Dataset Peak Zn-SAD Zmp1unbound Zmp1 E143Aunbound Zmp1 E143Aproduct Zmp1 E143A/Y178Fsubstrate Data Collection

Space group P212121 P212121 P212121 P21 P21 a, b, c (A˚ ) 43.17, 71.68, 117.70 43.17, 71.77, 117.80 43.50, 72.39, 118.53 37.06, 42.96, 119.21 37.44, 42.94, 123.24b a, b, g () 90, 90, 90 90, 90, 90 90, 90, 90 90, 92.54, 90 90, 96.1, 90 Wavelength (A˚ ) 1.28254 1.00 1.00 1.00 1.00 Resolutiona (A˚ ) 45.5–1.67 (1.72–1.67) 45.5–1.40 (1.49–1.40) 45.9–2.56 (2.71–2.56) 40.4–1.25 (1.33–1.25) 40.9–1.70 (1.80–1.70) No. of observations 779,717 (34,025) 310,074 (45,851) 78,821 (10,167) 404,889 (56,107) 254,035 (31,806) No. of unique reﬂections 80,960 (5,597) 71,874 (11,172) 12,587 (1,825) 102,218 (15,653) 43,083 (6,069) Multiplicity 9.6 (6.1) 4.3 (4.1) 6.3 (5.6) 4.0 (3.6) 6.0 (5.2) Completeness (%) 99.2 (93.0) 98.2 (95.6) 98.5 (91.8) 98.6 (93.7) 97.8 (87.8) b Rmerge (%) 6.8 (56.5) 5.2 (52.6) 17.7 (81.2) 4.4 (45.3) 12.3 (41.1) b

Structure Rmeas (%) 7.1 (61.7) 5.9 (60.2) 19.3 (89.3) 5.0 (52.6) 13.4 (45.2) 21.86 (2.81) 15.66 (2.48) 9.74 (2.42) 15.82 (2.67) 9.58 (2.78) c CC1/2 (%) 100 (85.1) 99.9 (85.5) 99.6 (86.2) 99.9 (85.1) 99.3 (86.3)

23 Reﬁnement

6214,Spebr1 2015 1, September 1632–1642, , d Rwork /Rfree (%) 15.60/18.17 19.95/25.97 14.15/16.20 17.96/20.82 No. of non-H protein atoms 3,109 3,054 3,149 3,147 No. of water molecules 532 107 454 428 No. of ions/heavy atoms 2 Zn2+ 2Zn2+ 2Zn2+ 2Zn2+ No. of other molecules – – 4 Glycerol – No. of TLS groups/chain 9 3 8 8 Root-mean-square deviations Bond lengths (A˚ ) 0.006 0.005 0.015 0.006 Bond angles () 1.022 0.629 1.489 0.932 Average B factor (A˚ 2) ª

05Esve t l ihsreserved rights All Ltd Elsevier 2015 All protein atoms 16.12 31.64 17.19 18.47 Water 29.74 27.69 30.44 31.11 Other atoms 10.12 30.32 11.06 67.76 Ramachandran plote (%) Most favored 98.72 96.37 99.0 98.73 Additionally allowed 1.28 3.37 1.0 1.27 Disallowed 0 0.26 0 0 PDB entry 5A0P 5A0S 5A0R 5A0X SAD, single-wavelength anomalous dispersion; CC, correlation coefficient. aHighest-resolution shell in parentheses. b Rsym = SSjI(hkl; j) j/SS with I(hkl; j) being the jth measurement of the intensity of the unique reflection (hkl), and the mean over all symmetry-related measurements; Rmeas = SS[n(hkl)/(n(hkl 1)]1/2 jI(hkl; j) j/SS (Diederichs and Karplus, 1997). 1635 c CC1/2 from Diederichs and Karplus, 2013. dRandom 5% of working set of reflections (Brunger, 1997). eMolProbity (Chen et al., 2010). protease domain (Figure 5). Together with domains II and III it builds up a groove that recognizes and binds substrates (Panni- fer et al., 2001; Maize et al., 2014). Domain III was recently shown to be responsive to ligand binding and to change its conformation depending on the bound ligand/substrate (Pannifer et al., 2001; Maize et al., 2014). We performed a structural alignment of Zmp1 with B. anthracis LF (Figure 5). The major part of Zmp1 aligns to LF domain IV with a root-mean-square deviation of 2.48 A˚ over 157 aligned residues. Interestingly, the S-loop spatially superposes with a part of LF domain III. Analogous to the interaction of LF domains III and IV (Pannifer et al., 2001), the S-loop shares a hydrophobic interface with the part of the NTD covered by the S-loop. From these findings and the afore- mentioned large movements, we conclude that the S-loop exhibits a function similar to that of LF domain III, and thus contributes to sequence specificity and substrate binding.

Substrate Recognition In a ﬁrst attempt to structurally characterize substrate binding to Zmp1, co-crystallization trials were set up using the substrate

octapeptide Ac-EVNPPVPD-NH2 and the Zmp1 mutant Zmp1- E143A, where the catalytic base of the HEXXH motif is mutated

to Ala. After a few days a new crystal form in space group P21 appeared. Surprisingly, only the unprimed part Ac-EVNP- COOH of the peptide was discernible in the electron density with a very clear defined carboxy terminus. A glycerol molecule was found in the primed part of the active-site cleft (Figure 6A). Thus, despite the mutation of the catalytic base Glu143, some residual proteolytic activity is retained, as also demonstrated in activity tests using an internally quenched fluorogenic substrate Figure 3. Structure of Zmp1 Overall structure of Zmp1 in cartoon representation. The N-terminal domain peptide (see above). (NTD) is colored slate blue, the active-site helix a4 orange, and the C-terminal To obtain crystals of Zmp1 with the full-length substrate domain (CTD) green. The edge strand b3 (cyan), the diverting loop (light green), peptide, comprising primed and unprimed sites, two changes 0 the S-loop (yellow), the bulge edge segment (red), and the S1 -wall-forming were introduced based on the observations of the product segment (purple) are labeled. The residues involved in zinc ion binding, the peptide complex. The peptide was shortened at the C ter- catalytic base, and the catalytic tyrosine are shown as sticks, and the zinc ion minus by one residue, resulting in the heptapeptide Ac- as a sphere. EVNPPVP-NH2, to make it fit into the crystal packing of the monoclinic space group. To abolish residual activity, the double server (Lee et al., 2003) yields a 98% closure of the S-loop with a mutant E143A/Y178F was prepared, which indeed proved to be rotation angle of 19.4 around the bending residues Ser83-Leu86 proteolytically inactive. The two complex structures are identical and Pro114-Gly115. Those residues lie in the loops between with respect to the interactions made by the unprimed peptide strand b2 and helix h1 and in the bulge edge segment right before residues. the edge strand b3, respectively. In the flexible part (Thr87- In the substrate peptide, all X-Pro peptide bonds are in the Val113) three interactions mediate S-loop closure in the unbound trans conformation, while potential cis conformers are probably crystal structure (Figure 4C, wheat). In the closed state, the unable to bind in a productive manner. The peptide binds in an Lys101 side chain is involved in hydrogen bonds with the side antiparallel fashion to the edge strand of the protease, but the chains of Glu184 and Glu185 and a water molecule. In addition, approximate angle between the peptide backbone and the pro- the phenolic OH group of Tyr182 OH interacts with Gly102 (dis- tease b-sheet plane is about 30, so only two main-chain tance of 2.83 A˚ ) and the side chain of Trp103 is engaged in a hydrogen bonds are formed via the P4 and P1 main-chain hydrogen bond with Tyr178 OH (2.51 A˚ ). In the open conforma- carbonyl oxygen atoms. The Pro at the P10 position introduces tion (yellow) all these interactions are broken, as the distance of a sharp bend, turning the course of the backbone away from the involved residues from each other increases. In the peptide the edge strand. This is very different from LF-peptide com- complexes, closure is probably also triggered by the interaction plexes (PDB: 1PWV; Turk et al., 2004) where the P10 tyrosine of Lys101 and the P2 Asn residue of the substrate (see below). has backbone dihedral angles (F, J)of(169, 66), while in Zmp1 these angles are (75, 177) and the restricted conforma- The S-loop Exhibits a Function Analogous to B. anthracis tion of the proline F angle does not allow a more b-strand-like LF Domain III conformation. LF from B. anthracis is a 90-kDa polypeptide composed of four Interestingly, the primed site of the peptide is bound in a ‘‘dou- major domains (I–IV), of which domain IV is the catalytic metallo- ble-kinked’’ conformation, with an angle of 122 (along the

1636 Structure 23, 1632–1642, September 1, 2015 ª2015 Elsevier Ltd All rights reserved Figure 4. Conformational Space of the S-loop (A) Surface representation of the open conformation of the S-loop. Important residues are shown as sticks in both (A) and (B). (B) Surface representation of the closed conformation of the S-loop. (C) Movement of important residues in S-loop opening and closure (indicated by double-sided arrows). K101 performs the largest movement from the closed conformation (wheat) to the open conformation (yellow), followed by E104 and other residues in the S-loop. Three residues mediate the closure: K101 interacts with E184 and E185, G102 with Y182, and W103 with Y178 (indicated by dashed lines). neighboring Ca atoms Asn3*-Pro4* and Pro4*-Pro5*), when dues in the S-loop and the bulge edge segment is further considering the P10 residue Pro5* (Ca) as the hinge and 109 at substantiated by the fact that they are conserved throughout the P20 residue Val6* (Ca) as the second hinge (Figure 6B). The the species (Figure S2). Leu95, Pro100, Trp110, Val113, and first kink is necessary to overcome the steric hindrance pro- Pro114 are strictly conserved, and the amino acid at position duced by the long loop composed of residues Ser128–Asp135, 103 is an aromatic residue, either a tryptophan or a tyrosine which diverts a 180 exit from the substrate-binding site and residue. forces the peptide to adopt this specific conformation. We In the product peptide complex structure with Zmp1-E143A, named this loop the ‘‘diverting loop’’ (Figures 3, 5, and 7). This the C terminus of the proline P1 residue (Pro4*) coordinates loop connects the b4 strand and the active-site helix. In anthrax the catalytic zinc ion via one oxygen atom of its carboxylate LF this loop is shortened by four residues and the active-site group, completing the tetrahedral zinc coordination together helix is elongated by three residues at its N terminus (Figure 7). with His142 (Nε2), His146 (Nε2), and Glu185 (Oε1). Therefore the loop is more flat and 6 A˚ further away from the cat- The proline at P10 position (Pro5*) is perfectly embedded in the alytic zinc ion than in Zmp1, thus allowing a straighter course of S10 subsite (Figure 6B). Its side chain interacts with His142, the substrate backbone (Figure 7). His134, Ala136, and Leu139. Its main-chain carbonyl oxygen ac- There are various interactions of the side chains of the sub- cepts a hydrogen bond from the Trp103 side chain. The back- strate, explaining nicely the observed substrate specificity (Fig- bone carbonyl oxygen of the P20 residue (Val6*) makes a ure 1B). The ammonium group of Lys101 interacts with the P2 hydrogen bond to the backbone NH group of Asp135. The S20 residue Asn3* (Od1) accompanied by a complete closure of the subsite is composed of the side chains of Asp135, Ala136,

S-loop. The NH2 group of the Asn3* side chain forms a hydrogen Asn175, Tyr178, and Leu179 (Figure 6B). Asp135 is located at bond with the main-chain carbonyl oxygen atom of Gly117 that is the bottom of the S20 subsite and restricts the size of the pocket. located on the edge strand. These two interactions explain the The proline at P30 completes the unique aliphatic-aromatic observed preference for Asn at the P2 position. network (Tyr94–Trp103) described above, by aligning as the In addition to hydrogen bonds, the peptide is bound by hy- last residue at the rim of the network (Figure 6B). This produces drophobic/van der Waals interactions (Figures 6 and S1), which the second kink in the substrate between the axes Pro5*-Val6* are also determinants of substrate speciﬁcity. The top of the and Val6*-Pro7* with an angle of about 109. The pyrrolidine unprimed part of the substrate-binding cleft up to the P10 posi- ring of proline is in perfect parallel orientation to the indole ring tion is built by an unusual network of alternating aliphatic and of Trp103 and in a 90 angle to the benzene ring of Phe178 (in aromatic residues that are aligned in a beads-on-a-string like wild-type Zmp1 to the phenol ring of Tyr178). It interacts with fashion, starting from Tyr94 located on helix h2 through the side chain of Trp103, Thr106, Phe178 (Tyr178), and a water Leu95, Trp110, Pro100, and ending at Trp103, all located on molecule that is bound to the side chain of Asn175. The back- the S-loop and the bulge edge segment. The latter three bone oxygen is bound to two water molecules coordinated by amino acids stack against each other in an almost parallel Asn175. The amide that remains attached at the C terminus of orientation of their side chains. Pro100 and Val113 together the peptide after cleavage from the Rink amide resin during with Trp110 and Trp103 encircle the peptide P1 proline residue solid-phase synthesis points toward the solvent and indicates and form a hydrophobic pocket that snugly accommodates the the direction in which the C-terminal residues (P40 and further ﬁve-membered ring. The importance of the hydrophobic resi- residues) would be located.

Structure 23, 1632–1642, September 1, 2015 ª2015 Elsevier Ltd All rights reserved 1637 Figure 5. Zmp1 Is Structurally Similar to Anthrax Lethal Factor Metalloprotease Domain IV Superposition of Zmp1 with B. anthracis LF in stereo view. Zmp1 (red) superposes well with LF metalloprotease domain IV (cyan), and the Zmp1 S-loop aligns partially with LF substrate-binding domain III (orange).

clashes with Glu189 at the bottom of the S20 site and with Asn175 and Leu179 flanking it to the solvent. Proline at P30 is strictly conserved in the endogenous substrates (Figure 1B). Its side chain stacks against the indole ring of Trp103 of the aliphatic-aromatic stacking network (Figure 6B). DISCUSSION Proline is unique among the proteinogenic amino acids owing to its secondary amino group functionality whereby the aliphatic On the basis of the crystal structures reported here, we are now side chain is bonded to the nitrogen. This restricts its conforma- able to link the knowledge about sequence specificity described tional flexibility to a backbone F angle of about 75. Further- earlier (Cafardi et al., 2013; Hensbergen et al., 2014) to structural more, its a-amino group in a peptide bond lacks the polar amide features of Zmp1. The residue at position P4 forms two back- hydrogen and cannot act as H-bond donor, and the X-Pro pep- bone interactions to Ser119 (Figures 6A and S1). The side chain tide link is more prone than other peptide bonds for the of the glutamate at P4 interacts directly with N3* of the peptide cis conformation. The imino-peptide bond synthesis is more itself and with His150 (Figures 6A and S1), but also makes difficult than for other amino acids: Ribosomes show a much water-mediated interactions with Asp149 and Asp155 side- slower incorporation rate (Pavlov et al., 2009), and in bacteria chain carboxyl groups. Owing to the many different possible in- the special elongation factor eF-P is needed to prevent ribosome teractions, the S4 pocket seems to be suitable for many different stalling at proline-rich sequences (Doerfel et al., 2013; Ude et al., side chains, thus high sequence variability can be expected at 2013). position P4. The S3 pocket is built up and restricted in size by Likewise, enzyme-catalyzed hydrolysis of proline-containing the three hydrophobic residues Tyr94, Leu95, and Trp110; it is peptide bonds also appears to be difficult, and a rather limited backed up by Leu116 located on the edge strand b3(Figures 6 set of proteases is known to be able to cleave proline-adjacent and S1). With this configuration a preference for non-polar peptide bonds. Because most proteolytic enzymes bind their aliphatic residues at P3 can be explained (Figure 1B). The substrates as an additional b strand in an extended conformation C. difficile substrates CD2831 and CD3246, which possess mul- at the edge of a b sheet (Madala et al., 2010; Gomis-Ruth et al., tiple cleavage sites, show only asparagine at position P2 (Fig- 2012), proline-containing substrates are obviously much less ure 1B). In the substrate complex crystal structure, this side suited for this binding mode since it requires a F angle of chain is stabilized by five interactions (Figures 6A and S1). about 139. In addition, owing to the higher basicity of the Because of its size and charge distribution, asparagine is the a-amino group (or imino group) of proline, it is a worse leaving only amino acid that can form all five interactions. The interaction group than other residues. with Lys101 is the one that drives S-loop closure and probably As mentioned earlier, only 40 genuine cleavage events of Pro- locks the peptide into the right conformation to position the Pro peptide bonds are documented in the MEROPS database, P1/P10 prolines for efficient cleavage. The amino acids Pro100, thus corroborating the particular resilience of this bond to Trp103, Trp110, Val113, His134, and Leu139 (Figure 6) deter- enzyme-catalyzed hydrolysis (Vanhoof et al., 1995). mine the size and hydrophobicity of the S1/S10 pocket. This C. difficile Zmp1 is the first protease that has a described explains why only proline and alanine residues are tolerated at specificity for such Pro-Pro peptide bonds. The substrate-bind- the substrate positions P1 and P10. Alanine can adopt many ing cleft of Zmp1 is shaped to handle non-extended, kinked pep- more main-chain conformations than proline, which makes tide conformations, partly owing to the unique diverting loop that such a peptide more flexible and causes a higher entropy loss is absent in LF (Figure 7) and other metalloproteases such as at the substrate-binding pocket. This explains why alanine- thermolysin. Another striking feature of Zmp1 that might be containing peptides at the P1 or P10 positions are less efficiently related to its ability to cope with Pro-Pro peptide bonds is the cleaved (Hensbergen et al., 2014). In addition, alanine makes surprisingly high residual proteolytic activity found in the fewer interactions with the surrounding aromatic and aliphatic active-site single mutants E143A and Y178F (Figure 2C). While residues. At position P20 aliphatic residues such as valine, isoleu- analogous mutations in LF (Tonello et al., 2004) and astacin cine, and alanine are found. While valine fits very well, the alanine (Yiallouros et al., 2000) almost completely inactivate the respec- side chain does not fill the S20 pocket, and larger amino acids tive enzymes, Zmp1 seems to overcome this handicap to a such as leucine or the aromatic residues would produce steric certain extent, at least in an in vitro situation with a peptidic

1638 Structure 23, 1632–1642, September 1, 2015 ª2015 Elsevier Ltd All rights reserved Figure 6. Product and Substrate Complex of Zmp1 (A) Unprimed substrate-binding site of Zmp1 occupied by the N-terminal product (Ac-EVNP- OH) after cleavage of the substrate peptide. The protein is in cartoon representation (cyan); the substrate-binding subsites S4–S1 are marked by half-circles. The peptide (green) and the edge strand b3 are shown as sticks, and dashed lines indicate the most important interactions.

(B) Substrate peptide Ac-EVNPPVP-NH2 (orange) bound to the Zmp1 substrate-binding site. Dashed lines indicate the most important interactions between the peptide and the primed substrate- binding site of Zmp1. A dashed ruler highlights the kink in the peptide with a 122 angle at Pro5* and a 109 angle at Val6*. See also Figures S1 and S2. substrate under the conditions described. It appears that the Expression of Zmp1 Constructs zinc-coordinating water molecule is reactive enough even The resulting vectors pET28a-Zmp1 wild-type, pET28a-Zmp1(E143A), without an assisting catalytic base (glutamic acid). Furthermore, pET28a-Zmp1(Y178F), and pET28a-Zmp1(E143A/Y178F) were transformed into competent BL21 Star (DE3) cells (Invitrogen) by heat shock, plated on the addition of water to the carbonyl carbon of the scissile pep- LB agar plus 50 mg/ml kanamycin plates, and incubated overnight at 37 C. tide bond may not be the rate-determining step, as has been Precultures grown overnight from a single colony at 37C were used to inocu- shown for the HIV-protease-assisted Phe-Pro peptide bond late 1 l of expression culture (LB medium plus 50 mg/ml kanamycin) to an 0 hydrolysis, where proton transfer to the P1 nitrogen is the optical density at 600 nm (OD600) of 0.1. Gene expression was induced with slowest step (Okimoto et al., 1999). Some residual activity of 0.5 mM isopropyl-b-D-1-thiogalactopyranoside at an OD600 of 0.6, and cell the analogous E/A mutant enabling activation of profragilysin growth was continued for another 4 hr at 37 C. Cells were harvested by centrifugation at 7,000 3 g,4C for 20 min. Cell pellets were washed with Tris- has been described, but without further quantification (Goulas buffered saline (TBS) (20 mM Tris [pH 7.5], 200 mM NaCl), pelleted again, et al., 2011). and stored at 80C until use. Eventually, the surprising Pro-Pro specificity may also have its origin in the unique aromatic-aliphatic side-chain stack at the Purification of Zmp1 Constructs upper rim of the active-site cleft, together with the diverting A cell pellet from 1 l of culture was solubilized in TBS with 10 mg/ml DNaseI loop and the mobile S-loop. These elements uniquely shape (AppliChem). Cells were lysed on ice/water by sonication for 15 min at the active-site cleft. 30% amplitude (Vibra-Cell VCX500, Sonics), cell debris was pelleted at 10,000 3 g, for 10 min at 4C, and the supernatant was cleared by ultracentri- fugation at 165,000 3 g at 4C for 30 min. The supernatant was adjusted with EXPERIMENTAL PROCEDURES 1 M imidazole (pH 7.5) to a final concentration of 10 mM imidazole and loaded onto 2-ml NiNTA Superflow resin (Qiagen). Protein was eluted (TBS, 250 mM Cloning of Wild-Type Zmp1 imidazole) after two wash steps with 10 mM and 30 mM imidazole. Protein The truncated (amino acids 27–220, lacking the native signal peptide) version concentration was determined at 280 nm using the molar extinction coefficient of the gene zmp1 (CD2830) from C. difficile strain 630, codon-optimized for of 25,900 M1 cm1, and 2 U of thrombin (Sigma-Aldrich) per milligram of pro- Escherichia coli (Geneart, Thermo Fisher Scientific), was PCR amplified tein were added, followed by dialysis overnight at 4C against a 50-fold volume 0 using the oligonucleotides Zmp1-fw: 5 -GGAATTCCATATGGATAGCACCAC of TBS. The protein was applied onto the same NiNTA Superflow column equil- 0 0 0 CATTC-3 and Zmp1-bw: 5 -CCGCTCGAGTTATTTTGCCAGATTCTG-3 , and ibrated with TBS supplemented with 10 mM imidazole, and the flow-through was subcloned into vector pET28a (Merck-Millipore) using the restriction sites was collected. The protein was concentrated and applied onto a HiLoad NdeI and XhoI. This construct yields a protein with a thrombin-cleavable N-ter- Superdex 200 16/600 column (GE Healthcare) equilibrated with TBS. Zmp1 minal 6x-His tag. constructs eluted as single peaks corresponding to a size of 20 kDa. The yield for all constructs was 50 mg of protein per liter of culture. Mutagenesis of Zmp1 Active-site mutants were generated using the protocol published earlier Crystallization (Zheng et al., 2004). For the E143A mutant the PCR was performed using All crystals were grown by the sitting-drop vapor diffusion technique. For the pET28a-Zmp1 as template and the oligonucleotides Zmp1-E143A-fw: orthorhombic unbound form, wild-type protein (12 mg/ml) was mixed with pre- 50-GAACTGCATGCAACCGCACATGCCATTG-30 and Zmp1-E143A-bw: cipitant in a ratio of 2:1 to form 3-ml drops that were then equilibrated at 20C 50-GCGGTTGCATGCAGTTCCAGATTAATG-30. For the constructs pET28a- against 1.8–2.55 M ammonium phosphate and 0.1 M Tris (pH 7.5–9.0). Highly Zmp1(Y178F) and pET28a-Zmp1(E143A/Y178F), the vectors pET28a-Zmp1 intergrown crystals appeared after 2 days. Seeds were prepared (Beads-for- wild-type and pET28a-Zmp1(E143A), respectively, were used as templates Seeds, Jena Biosciences) and diluted 1:1,000 in the original reservoir solution. and the following oligonucleotides: Zmp1-Y178F-fw: 50-GTTAATTTTCT 1 ml of seeds was added to the wells containing 1.8–2.25 M ammonium phos- GGGTGTTTATCCG-30 and Zmp1-Y178F-bw: 50-CCCAGAAAATTAACATT phate (pH 7.5–9.0) to yield well-ordered crystals in all conditions after another GCCCAGG-30. A reaction with 18 cycles was performed with an annealing 2 days. Data were collected from a crystal grown in 2.1 M ammonium phos- temperature of 60C for 1 min and an elongation step at 68C for 18 min. phate, 0.1 M Tris (pH 8.0) that was cryoprotected with 20% glycerol. The reactions (50 ml) were purified using a PCR cleanup kit (Sigma-Aldrich) Orthorhombic crystals of the Zmp1 E143A variant were grown by mixing and digested with 1 U DpnI for 1 hr at 37C. 4 ml of the reactions was trans- protein (11 mg/ml) in a 1:2 ratio with precipitant (2.1 M D/L-malic acid [pH formed into competent E. coli TOP10 cells (Invitrogen), plated on LB agar 7.0]) to form 300-nl drops at 20C. Needle clusters appeared after 2 weeks. plus 50 mg/ml kanamycin plates, and incubated overnight at 37C. The pre- Single crystals were released using two acupuncture needles as tools, and pared vectors were sequenced to identify positive clones. flash-frozen in liquid nitrogen without further cryoprotection.

Activity Assay Using Fibrinogen and Fibronectin Long-term reactions of 5 mM Zmp1 with 1 mM ﬁbronectin from human plasma (Sigma-Aldrich) and reactions of 0.05 and 1 mM Zmp1 with 1 mM ﬁbrinogen type I from human plasma (Sigma-Aldrich) were set up at 37C in PBS. The sample volume was set to 100 ml and samples of 20 ml were taken after 1, 2, 3, and 24 hr before analysis on 10% SDS-PAGE gels.

Solid-Phase Peptide Synthesis

The substrate peptides Ac-EVNPPVPD-CONH2 and Ac-EVNPPVP-CONH2 were synthesized by automated multiple solid-phase peptide synthesis (SyroI; MultiSynTech) on the 4-(20,40-dimethoxyphenyl-Fmoc-aminomethyl)-phenoxy (Rink amide) resin (Novabiochem; 30 mg, resin loading 0.5 mmol/g) using the ﬂuorenylmethoxycarbonyl/tert-butyl-strategy utilizing a double coupling pro- 0 Figure 7. The Diverting Loop Is a Unique Feature of Zmp1 cedure and in situ activation with Oxyma Pure-N,N -diisopropyl-carbodiimide (Oxyma-DIC) (Novabiochem and IRIS Biotech). N-Terminal acetylation was Close-up stereo view of Zmp1 (orange) and anthrax LF (yellow) along the carried out using 10 equivalents (eq.) of acetic anhydride (Ac O; Sigma- substrate-binding groove. The substrate-heptapeptide in Zmp1 and the sub- 2 Aldrich) activated with 10 eq. of N,N–diisopropylethylamine (DIPEA; Fluka) in strate-undecapeptide in LF are shown as sticks. The diverting loop (green), the dichloromethane (Biosolve) under vigorous shaking for 2 hr. Peptides were S-loop (slate blue), and the zinc ion are indicated. cleaved from the resin with triﬂuoroacetic acid (TFA), triisopropylsilane (TIS; both from Fluka), and water (95:2.5:2.5, v/v/v) at room temperature. Products For the product-complex of Zmp1(E143A) with the substrate peptide Ac- were precipitated with ice-cold n-hexane (Merck)/diethyl ether (Biosolve)

EVNPPVPD-NH2, the protein (11 mg/ml) was mixed with a 7-fold molar excess (1:3, v/v), collected by centrifugation, washed four times, and lyophilized of the peptide in TBS and incubated at 20C for 30 min before plate set-up. from water with tert-butyl alcohol (Fluka) (3:1, v/v). Purification was carried Crystals appeared in various conditions after a few days. The best diffracting out by preparative reverse-phase high-performance liquid chromatography crystals were obtained after seeding from a condition containing 1.8 M ammo- (HPLC) on a 90-A˚ C18 Jupiter column (Phenomex), and analyzed using a nium phosphate dibasic, 0.1 M Tris (pH 8.5), with a protein to precipitant ratio liquid chromatography-electrospray ionization-mass spectrometry instrument of 2:1 in 3-ml drops. Data were collected from crystals that were cryoprotected (LCQ, Finnigan MAT, Thermo Scientific). The final purity of the HPLC-purified with 20% glycerol. For the substrate peptide complex of Zmp1(E143AY178F), and lyophilized peptide was >95%. protein (12 mg/ml) was mixed with a 10-fold molar excess of the peptide Ac- For the synthesis of the internally quenched fluorogenic peptide Ac- EVNPPVP-NH2 in TBS and incubated at 20 C for 30 min. Conditions with clear K(Dabcyl)EVNPPVPDE(EDANS)-CONH2, Fmoc-L-glutamic acid-g-[2-(1-sulfo- drops were seeded with seeds diluted 1:250 in the original reservoir solution nyl-5-naphthyl)-aminoethylamide (Fmoc-Glu(EDANS)-OH) (3 eq.) was coupled 2.1 M ammonium phosphate dibasic, 0.1 M Tris (pH 8.0) (prepared as three times with 3 eq. of O-(7-azabenzotriazol-1-yl)-1,1,3,3-tetramethyluro- described above), and best diffracting crystals were obtained from 2.25 nium hexafluorophosphate-DIPEA (HATU-DIPEA; Fluka) on Rink amide resin ammonium phosphate dibasic, 0.1 M Tris (pH 9.0) as precipitant with a protein (30 mg, resin loading 0.5 mmol/g) for 3 hr. Afterward the remaining amine to precipitant ratio of 2:1 in 3-ml drops. Data were collected from crystals cry- groups on the resin were blocked by acetylation using 10 eq. of Ac2O and oprotected with 30% sucrose-supplemented precipitant solution. 10 eq. of DIPEA for 15 min. Subsequently the peptide was N-terminally extended by automated multiple solid-phase peptide synthesis as described Data Collection and Structure Determination for the substrate peptide. Fmoc-ε-(4,4-dimethylazobenzene-40-carbonyl)-L- Crystals were screened in-house using a rotating anode MicroMax-007 HF lysine (Fmoc-L-Lys(Dabcyl)-OH) (3 eq.) was manually coupled using either (Rigaku, Japan) equipped with a Mar345 image plate (MAR Research, 3 eq. of HATU-DIPEA or 3 eq. of Oxyma-DIC until Kaiser tests were negative Germany). High-resolution data were collected at the Swiss Light Source, (Kaiser et al., 1970). The FRET peptide was further treated as described above. Paul-Scherrer-Institute, Villigen, Switzerland, on beamlines X06DA or X06SA using Pilatus 2M and Pilatus 6M detectors (Dectris). All diffraction data were ACCESSION NUMBERS processed using XDS (Kabsch, 2010). For wild-type Zmp1unbound (crystal ˚ form I), a peak dataset at a wavelength of 1.283 A was collected in the in- The accession numbers for the refined structures reported in this paper verse-beam mode. The structure was solved by single-wavelength anomalous are PDB: 5A0P (Zmp1unbound), 5A0S (Zmp1 E143Aunbound), 5A0R (Zmp1 dispersion using phases computed from the two zinc sites by the E143Aproduct), and 5A0X (Zmp1 E143A/Y178Fsubstrate). phenix.autosol routine of the PHENIX package (Adams et al., 2010) and an almost complete model was automatically built. The model was refined using SUPPLEMENTAL INFORMATION iterative cycles of phenix.refine (Afonine et al., 2012) and manual model build- ing in Coot (Emsley et al., 2010). For the orthorhombic crystal form Zmp1 E143Aunbound and the monoclinic crystal forms Zmp1 E143Aproduct and Supplemental Information includes two figures and can be found with this Zmp1 E143A/Y178Fsubstrate, native datasets were collected at a wavelength article online at http://dx.doi.org/10.1016/j.str.2015.06.018. of 1.0 A˚ . The structure was solved by molecular replacement with one mono- mer from crystal form I as search model by the program Phaser (McCoy et al., AUTHOR CONTRIBUTIONS 2007), and refined as described above. All data collection and refinement statistics are given in Table 1. Conceptualization, M.S., U.B.; methodology, M.S., U.B., I.N.; investigation, M.S., C.P.; writing, M.S., U.B.,; funding acquisition, U.B., I.N.; supervision, Activity Assay Using an Internally Quenched Fluorogenic Peptide M.S., I.N., U.B. The reaction of the fluorogenic substrate peptide containing an N-terminal Lys-Dabcyl group and a C-terminal Glu-EDANS group was measured by the ACKNOWLEDGMENTS increase in fluorescence at 485 nm that occurs upon cleavage of any peptide bond in the peptide sequence (Dabcyl)KEVNPPVPDE(EDANS). The substrate We thank the staff at the beamlines X06SA and X06DA at the Swiss Light (50 mM) was incubated for up to 60 min at 37C with 1 mM Zmp1 wild-type or its Source, Paul-Scherrer-Institute, Villigen, Switzerland for support during syn- variants, respectively, in a 96-well plate (black pureGrade S, Brand) using the chrotron data collection. We are grateful to Monika Gompert for excellent

1640 Structure 23, 1632–1642, September 1, 2015 ª2015 Elsevier Ltd All rights reserved technical support. The project was supported by the University of Cologne and (1998). Proteolytic inactivation of MAP-kinase-kinase by anthrax lethal factor. grant INST 216/682-1 FUGG from the German Research Council. A PhD Science 280, 734–737. fellowship from the International Graduate School in Development Health Emsley, P., Lohkamp, B., Scott, W.G., and Cowtan, K. (2010). Features and and Disease to C.P. is acknowledged. The research leading to these results development of Coot. Acta Crystallogr. D Biol. Crystallogr. 66, 486–501. has received funding from the European Community’s Seventh Framework George, W.L., Sutter, V.L., and Finegold, S.M. (1977). Antimicrobial agent- Program (FP7/2007-2013) under grant agreement No. 283570 (BioStruct-X). induced diarrhea–a bacterial disease. J. Infect. Dis. 136, 822–828.

Received: May 11, 2015 George, R.H., Symonds, J.M., Dimock, F., Brown, J.D., Arabi, Y., Shinagawa, Revised: June 21, 2015 N., Keighley, M.R., Alexander-Williams, J., and Burdon, D.W. (1978). Accepted: June 22, 2015 Identification of Clostridium difficile as a cause of pseudomembranous colitis. Published: July 23, 2015 Br. Med. J. 1, 695. Gomis-Ruth, F.X., Botelho, T.O., and Bode, W. (2012). A standard orientation REFERENCES for metallopeptidases. Biochim. Biophys. Acta 1824, 157–163. Goulas, T., Arolas, J.L., and Gomis-Ruth, F.X. (2011). Structure, function and Adams, P.D., Afonine, P.V., Bunkoczi, G., Chen, V.B., Davis, I.W., Echols, N., latency regulation of a bacterial enterotoxin potentially derived from a Headd, J.J., Hung, L.W., Kapral, G.J., Grosse-Kunstleve, R.W., et al. (2010). mammalian adamalysin/ADAM xenolog. Proc. Natl. Acad. Sci. USA 108, PHENIX: a comprehensive Python-based system for macromolecular struc- 1856–1861. ture solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213–221. Grams, F., Dive, V., Yiotakis, A., Yiallouros, I., Vassiliou, S., Zwilling, R., Bode, Afonine, P.V., Grosse-Kunstleve, R.W., Echols, N., Headd, J.J., Moriarty, W., and Stocker, W. (1996). Structure of astacin with a transition-state N.W., Mustyakimov, M., Terwilliger, T.C., Urzhumtsev, A., Zwart, P.H., and analogue inhibitor. Nat. Struct. Biol. 3, 671–675. Adams, P.D. (2012). Towards automated crystallographic structure refinement Hege, T., and Baumann, U. (2001). Protease C of Erwinia chrysanthemi: the with phenix.refine. Acta Crystallogr. D Biol. Crystallogr. 68, 352–367. crystal structure and role of amino acids Y228 and E189. J. Mol. Biol. 314, Bartlett, J.G. (2006). Narrative review: the new epidemic of Clostridium diffi- 187–193. cile-associated enteric disease. Ann. Intern. Med. 145, 758–764. Hensbergen, P.J., Klychnikov, O.I., Bakker, D., van Winden, V.J., Ras, N., Bartlett, J.G., Chang, T.W., Gurwith, M., Gorbach, S.L., and Onderdonk, A.B. Kemp, A.C., Cordfunke, R.A., Dragan, I., Deelder, A.M., Kuijper, E.J., et al. (1978). Antibiotic-associated pseudomembranous colitis due to toxin-produc- (2014). A novel secreted metalloprotease (CD2830) from Clostridium difficile ing clostridia. N. Engl. J. Med. 298, 531–534. cleaves specific proline sequences in LPXTG cell surface proteins. Mol. Cell. Proteomics 13, 1231–1244. Beall, F.A., Taylor, M.J., and Thorne, C.B. (1962). Rapid lethal effect in rats of a Hooper, N.M. (1994). Families of zinc metalloproteases. FEBS Lett. 354, 1–6. third component found upon fractionating the toxin of Bacillus anthracis. J. Bacteriol. 83, 1274–1280. Kabsch, W. (2010). XDS. Acta Crystallogr. D Biol. Crystallogr. 66, 125–132. Bouza, E. (2012). Consequences of Clostridium difficile infection: understand- Kaiser, E., Colescott, R.L., Bossinger, C.D., and Cook, P.I. (1970). Color test for ing the healthcare burden. Clin. Microbiol. Infect. 18 (Suppl 6 ), 5–12. detection of free terminal amino groups in the solid-phase synthesis of peptides. Anal. Biochem. 34, 595–598. Brazier, J.S. (1998). The epidemiology and typing of Clostridium difficile. J. Antimicrob. Chemother. 41 (Suppl C ), 47–57. Lee, R.A., Razaz, M., and Hayward, S. (2003). The DynDom database of protein domain motions. Bioinformatics 19, 1290–1291. Brunger, A.T. (1997). Free R value: cross-validation in crystallography. Methods Enzymol. 277, 366–396. Madala, P.K., Tyndall, J.D., Nall, T., and Fairlie, D.P. (2010). Update 1 of: proteases universally recognize beta strands in their active sites. Chem. Rev. 110, Cafardi, V., Biagini, M., Martinelli, M., Leuzzi, R., Rubino, J.T., Cantini, F., PR1–PR31. Norais, N., Scarselli, M., Serruto, D., and Unnikrishnan, M. (2013). Identification of a novel zinc metalloprotease through a global analysis of Maize, K.M., Kurbanov, E.K., De La Mora-Rey, T., Geders, T.W., Hwang, D.J., Clostridium difficile extracellular proteins. PLoS One 8, e81306. Walters, M.A., Johnson, R.L., Amin, E.A., and Finzel, B.C. (2014). Anthrax toxin lethal factor domain 3 is highly mobile and responsive to ligand binding. Acta Chen, V.B., Arendall, W.B., Headd, J.J., Keedy, D.A., Immormino, R.M., Crystallogr. D Biol. Crystallogr. 70, 2813–2822. Kapral, G.J., Murray, L.W., Richardson, J.S., and Richardson, D.C. (2010). Makris, A.T., and Gelone, S. (2007). Clostridium difficile in the long-term care MolProbity: all-atom structure validation for macromolecular crystallography. setting. J. Am. Med. Dir. Assoc. 8, 290–299. Acta Crystallogr. D Biol. Crystallogr. 66, 12–21. Matthews, B.W. (1988). Structural basis of the action of thermolysin and Coutte, L., Alonso, S., Reveneau, N., Willery, E., Quatannens, B., Locht, C., related zinc peptidases. Acc. Chem. Res. 21, 333–340. and Jacob-Dubuisson, F. (2003a). Role of adhesin release for mucosal colonization by a bacterial pathogen. J. Exp. Med. 197, 735–742. McCoy, A.J., Grosse-Kunstleve, R.W., Adams, P.D., Winn, M.D., Storoni, L.C., and Read, R.J. (2007). Phaser crystallographic software. J. Appl. Crystallogr. Coutte, L., Willery, E., Antoine, R., Drobecq, H., Locht, C., and Jacob- 40, 658–674. Dubuisson, F. (2003b). Surface anchoring of bacterial subtilisin important for maturation function. Mol. Microbiol. 49, 529–539. Neurath, H., and Walsh, K.A. (1976). Role of proteolytic enzymes in biological regulation (a review). Proc. Natl. Acad. Sci. USA 73, 3825–3832. Crooks, G.E., Hon, G., Chandonia, J.M., and Brenner, S.E. (2004). WebLogo: a sequence logo generator. Genome Res. 14, 1188–1190. O’Connor, J.R., Johnson, S., and Gerding, D.N. (2009). Clostridium difficile infection caused by the epidemic BI/NAP1/027 strain. Gastroenterology 136, Cunningham, D.F., and O’Connor, B. (1997). Proline specific peptidases. 1913–1924. Biochim. Biophys. Acta 1343, 160–186. Okimoto, N., Tsukui, T., Hata, M., Hoshino, T., and Tsuda, M. (1999). Diederichs, K., and Karplus, P.A. (1997). Improved R-factors for diffraction Hydrolysis mechanism of the phenylalanine-proline peptide bond specific to data analysis in macromolecular crystallography. Nat. Struct. Biol. 4, 269–275. HIV-1 protease: investigation by the ab initio molecular orbital method. Diederichs, K., and Karplus, P.A. (2013). Better models by discarding data? J. Am. Chem. Soc. 121, 7349–7354. Acta Crystallogr. D Biol. Crystallogr. 69, 1215–1222. Pannifer, A.D., Wong, T.Y., Schwarzenbacher, R., Renatus, M., Petosa, C., Doerfel, L.K., Wohlgemuth, I., Kothe, C., Peske, F., Urlaub, H., and Rodnina, Bienkowska, J., Lacy, D.B., Collier, R.J., Park, S., Leppla, S.H., et al. (2001). M.V. (2013). EF-P is essential for rapid synthesis of proteins containing Crystal structure of the anthrax lethal factor. Nature 414, 229–233. consecutive proline residues. Science 339, 85–88. Pavlov, M.Y., Watts, R.E., Tan, Z., Cornish, V.W., Ehrenberg, M., and Forster, Duesbery, N.S., Webb, C.P., Leppla, S.H., Gordon, V.M., Klimpel, K.R., A.C. (2009). Slow peptide bond formation by proline and other N-alkylamino Copeland, T.D., Ahn, N.G., Oskarsson, M.K., Fukasawa, K., Paull, K.D., et al. acids in translation. Proc. Natl. Acad. Sci. USA 106, 50–54.

Structure 23, 1632–1642, September 1, 2015 ª2015 Elsevier Ltd All rights reserved 1641 Pellizzari, R., Guidi-Rontani, C., Vitale, G., Mock, M., and Montecucco, C. (2012). Bacillus anthracis protease InhA regulates BslA-mediated adhesion (1999). Anthrax lethal factor cleaves MKK3 in macrophages and inhibits the in human endothelial cells. Cell. Microbiol. 14, 1219–1230. LPS/IFNgamma-induced release of NO and TNFalpha. FEBS Lett. 462, Turk, B.E., Wong, T.Y., Schwarzenbacher, R., Jarrell, E.T., Leppla, S.H., 199–204. Collier, R.J., Liddington, R.C., and Cantley, L.C. (2004). The structural basis Rawlings, N.D., Waller, M., Barrett, A.J., and Bateman, A. (2014). MEROPS: for substrate and inhibitor selectivity of the anthrax lethal factor. Nat. Struct. the database of proteolytic enzymes, their substrates and inhibitors. Nucleic Mol. Biol. 11, 60–66. Acids Res. 42, D503–D509. Ude, S., Lassak, J., Starosta, A.L., Kraxenberger, T., Wilson, D.N., and Jung, Schechter, I., and Berger, A. (1967). On the size of the active site in proteases. K. (2013). Translation elongation factor EF-P alleviates ribosome stalling at pol- I. Papain. Biochem. Biophys. Res. Commun. 27, 157–162. yproline stretches. Science 339, 82–85. Smith, H., Tempest, D.W., Stanley, J.L., Harris-Smith, P.W., and Gallop, R.C. (1956). The chemical basis of the virulence of Bacillus anthracis. VII. Two com- Vanhoof, G., Goossens, F., De Meester, I., Hendriks, D., and Scharpe, S. ponents of the anthrax toxin: their relationship to known immunising aggres- (1995). Proline motifs in peptides and their biological processing. FASEB J. sins. Br. J. Exp. Pathol. 37, 263–271. 9, 736–744. Tallant, C., Marrero, A., and Gomis-Ruth, F.X. (2010). Matrix metalloprotei- Vazeux, G., Wang, J., Corvol, P., and Llorens-Cortes, C. (1996). Identification nases: fold and function of their catalytic domains. Biochim. Biophys. Acta of glutamate residues essential for catalytic activity and zinc coordination in 1803, 20–28. aminopeptidase A. J. Biol. Chem. 271, 9069–9074. Thomas, C., Stevenson, M., and Riley, T.V. (2003). Antibiotics and hospital- Vitale, G., Bernardi, L., Napolitani, G., Mock, M., and Montecucco, C. (2000). acquired Clostridium difficile-associated diarrhoea: a systematic review. Susceptibility of mitogen-activated protein kinase kinase family members to J. Antimicrob. Chemother. 51, 1339–1350. proteolysis by anthrax lethal factor. Biochem. J. 352, 739–745. Tonello, F., Naletto, L., Romanello, V., Dal Molin, F., and Montecucco, C. (2004). Tyrosine-728 and glutamic acid-735 are essential for the metallopro- Yiallouros, I., Grosse Berkhoff, E., and Stocker, W. (2000). The roles of Glu93 teolytic activity of the lethal factor of Bacillus anthracis. Biochem. Biophys. and Tyr149 in astacin-like zinc peptidases. FEBS Lett. 484, 224–228. Res. Commun. 313, 496–502. Zheng, L., Baumann, U., and Reymond, J.L. (2004). An efficient one-step site- Tonry, J.H., McNichol, B.A., Ramarao, N., Chertow, D.S., Kim, K.S., Stibitz, S., directed and site-saturation mutagenesis protocol. Nucleic Acids Res. 32, Schneewind, O., Kashanchi, F., Bailey, C.L., Popov, S., and Chung, M.C. e115.