A Novel Mechanism of Latency in Matrix Metalloproteinases*

Background: Animal and plant matrix metalloproteinases (MMPs) are kept zymogenic through large prodomains and a cysteine-switch mechanism. Results: Bacterial MMP karilysin has only a short N-terminal peptide upstream of the catalytic domain, which lacks cysteines. Conclusion: This peptide inhibits through an aspartate-switch mechanism and also exerts other functions of authentic prodomains. Significance: Karilysin is kept latent by a novel mechanism for MMPs. The matrix metalloproteinases (MMPs) are a family of secreted soluble or membrane-anchored multimodular peptidases regularly found in several paralogous copies in animals and plants, where they have multiple functions. The minimal consensus domain architecture comprises a signal peptide, a 60–90-residue globular prodomain with a conserved sequence motif including a cysteine engaged in “cysteine-switch” or “Velcro” mediated latency, and a catalytic domain. Karilysin, from the human periodontopathogen Tannerella forsythia, is the only bacterial MMP to have been characterized biochemically to date. It shares with eukaryotic forms the catalytic domain but none of the flanking domains. Instead of the consensus MMP prodomain, it features a 14-residue propeptide, the shortest reported for a metallopeptidase, which lacks cysteines. Here we determined the structure of a prokarilysin fragment encompassing the propeptide and the catalytic domain, and found that the former runs across the cleft in the opposite direction to a bound substrate and inhibits the latter through an “aspartate-switch” mechanism. This finding is reminiscent of latency maintenance in the otherwise unrelated astacin and fragilysin metallopeptidase families. In addition, in vivo and biochemical assays showed that the propeptide contributes to protein folding and stability. Our analysis of prokarilysin reveals a novel mechanism of latency and activation in MMPs. Finally, our findings support the view that the karilysin catalytic domain was co-opted by competent bacteria through horizontal gene transfer from a eukaryotic source, and later evolved in a specific bacterial environment.

strains (21). Structural studies supported the view that the catalytic domain of this MP is the result of horizontal gene transfer of a member of the ADAM/adamalysin family, which has 38 orthologs in humans (8, [22][23][24][25], from a mammalian host to this bacterium, which thrives in the intestinal tract (26,27).
Returning to MMPs, karilysin from the human periodontopathogen Tannerella forsythia is the only bacterial family member to have been analyzed biochemically to date (9, 28 -33). In addition to karilysin, only MmpZ from Bacillus anthracis has been functionally assessed at the genetic level through knockout studies in B. anthracis cells, but it has not been isolated or characterized (34). Similarly to vertebrate MMPs, karilysin showed preference for medium-sized to bulky hydrophobic residues (leucine, tyrosine and methionine) in the specificity pocket, S 1 Ј (Ref. 30; for active-site cleft subsite nomenclature, see Ref. 35). It inactivates antimicrobial peptide LL-37 and integrants of the complement system, including ficolin-2, ficolin-3, C4, and C5, by proteolysis and may thus contribute to evasion of the innate host immune response (29,31). Karilysin is sequentially and evolutionarily closer to MMPs from winged insects that are transmission vectors of human diseases (47% sequence identity with Dm1 from Aedes aegypti and Anopheles gambiae; (9)) and mammals (44% identity with human MMP-11, -13, and -20 (9)) than to the few other bacterial sequences found in genomic sequences. Accordingly it was likewise suggested that it may be the result of horizontal gene transfer of an MMP gene from an animal to an intimate bacterial pathogen, which inhabits a biofilm on the tooth surface in humans (9).
The metzincins are characterized by a consensus sequence responsible for binding of the catalytic zinc ion (CSBZ), H-E-X-X-H-X-X-(G/N)-X-X-(H/D) (amino acid one-letter code; X stands for any residue), and a conserved methionine-containing turn, the "Met-turn" (1)(2)(3)(4)(5)36). In MMPs, the CSBZ encompasses three histidine zinc ligands, the general base/acid glutamate for catalysis, and a structurally relevant glycine (3). In addition, the distinct MMP paralogs are multidomain proteins that display a disparate domain organization that is the result of successive polyplication, gene fusion, and exon shuffling (11). The only domains common to all animal and plant MMPs are a signal peptide, which is removed after secretion, a prodomain and a catalytic domain, as found, e.g. in human MMP-7 and MMP-26, and in plant MMPs (12,16,18).
Most peptidases are biosynthesized as zymogens containing prosegments, which are required for latency maintenance to prevent unbridled activity but also sometimes to assist in proper folding of the usually downstream catalytic moieties (37)(38)(39)(40). Metzincin exceptions lacking prosegments include the archaemetzincins, for which no hydrolytic activity has so far been reported, i.e. they might not need to be kept latent (41,42); the toxilysin EcxA from Escherichia coli, whose soluble expression requires co-expression with its cognate EcxB subunit, thus pointing to a chaperone-like function for this ancillary subunit (43)(44)(45); the cholerilysin StcE from E. coli, for which an N-terminal immunoglobulin-like domain may assist the downstream catalytic moiety in proper folding (46); and igalysins, where an all-␤-domain of similar topology to immunoglobulin-like domains is likewise found at the N terminus of the catalytic moiety (see Protein Data Bank (PDB) access codes 4DF9 and 3P1V and Ref. 5).
MMP prodomains (see Table 1 in Ref. 47) span 60 -90 residues and include a conserved sequence motif, P-R-C-G-(V/N)-P-D, engaged in a "cysteine-switch" or "Velcro" mechanism of latency (10, 16, 48 -51). It has been suggested that this mechanism may be shared by variants within other metzincin families, for which conserved cysteines were described upstream of the catalytic domain, such as the ADAMs/adamalysins (motif P-K-M-C-G-V (8, 52-54)), leishmanolysins (motif H-R-C-I-H-D (2)), and pappalysins (motif C-G (55)). In contrast, the 472 residues encoded by the karilysin gene (see UniProt sequence database access code D0EM77) only comprise a short 14-residue potential propeptide, which lacks cysteines, between the 20-residue signal peptide and the 161-residue mature catalytic moiety (Fig. 1A). A C-terminal domain of 277 residues of unknown function and sequence unrelated to any domain found in eukaryotic MMPs completes the protein. This strongly suggests a potentially different mechanism of latency maintenance, hitherto unseen not only in MMPs but also in metzincins in general, as the shortest prosegments described to date are those of members of the astacin family, which span Ͼ34 residues (7, 56 -58).
We had previously determined the structure of the catalytic domain of karilysin (termed Kly18 (9)). To shed light on the molecular determinants of the first mechanism of latency maintenance of a bacterial MMP, in this work we assayed the possible function of the propeptide in folding, stability, and activity inhibition of Kly18. We further solved the x-ray crystal structure of an active-site mutant of a construct spanning the propeptide and Kly18 affecting the catalytic glutamate, pKly18-E156A, to circumvent autolysis. The mechanism derived was supported by site-directed mutagenesis and it is discussed in the context of general MMP latency maintenance.

EXPERIMENTAL PROCEDURES
Protein Production and Purification-The gene coding for full-length wild-type T. forsythia prokarilysin without the 20-residue signal peptide (hereafter pKly; 52 kDa; residues Gln 21 -Lys 472 according to UP D0EM77, see also Fig. 1A) was cloned at BamHI and XhoI restriction sites into vector pGEX-6P-1 (GE Healthcare) as described elsewhere (30). The resulting vector, pKAR1 (see Table 1 for an overview of vectors and constructs used), confers resistance toward ampicillin and attaches an N-terminal glutathione S-transferase (GST) moiety followed by a human rhinovirus 3C proteinase (HR3CP) recognition site (L-E-V-L-F-Q-2-G-P; HR3CP cleavage leaves two extra residues, underlined, at the N terminus of the recombinant protein after digestion; three extra residues, L-G-S, are further present due to the cloning strategy). Single-residue point mutants pKly-Y35A and pKly-E156A (pKAR2 and pKAR3, respectively) were generated using the QuikChange Site-directed Mutagenesis Kit (Stratagene) according to the manufacturer's instructions as described (30). Double mutant pKly-D25A/Y35A (pKAR4) was similarly generated using pKAR2 as a template. Genes coding for the E156A-mutated catalytic domain of karilysin, with and without the propeptide (hereafter pKly18-E156A and Kly18-E156A; 20 and 18 kDa; residues Gln 21 -Ser 201 and residues Tyr 35 -Ser 201 , respectively), were also cloned into vector pGEX-6P-1 (pKAR5 and pKAR6, respectively). Genes coding for pKly18 and its mutant proteins pKly18-Y35A and pKly18-D25A/Y35A were cloned into the same vector (pKAR7, pKAR8, and pKAR9, respectively) following a strategy previously described (59). Genes coding for pKly18, pKly18-E156A, Kly18, and Kly18-E156A were, furthermore, cloned at NcoI and XhoI restriction sites into vector pCRI-7a (59), which confers resistance toward kanamycin and does not attach fusion proteins (pKAR10-pKAR13, respectively). In these cases, the cloning strategy entailed that residues M-G were attached at the N terminus. All constructs were verified by DNA sequencing.
Proteins encoded by vectors pKAR1-pKAR9 were produced by heterologous overexpression in E. coli BL21(DE3) cells, which were grown at 37°C in Luria-Bertani medium supplemented with 100 g/ml of ampicillin. Cultures were induced at an A 600 of 0.8 with 0.2 mM isopropyl ␤-D-thiogalactopyranoside and incubated overnight at 18°C. Purification of wild-type and mutant pKly, and subsequent autolysis of the former to obtain Kly18, was achieved as described elsewhere (30). In turn, pKly18-E156A, Kly18-E156A, pKly18-Y35A, and pKly18-D25A/Y35A were purified as follows. After centrifugation at 7,000 ϫ g for 30 min at 4°C, the pellet was washed twice in 1ϫ PBS, and resuspended in the same buffer supplemented with EDTA-free protease inhibitor mixture tablets and DNase I (both Roche Diagnostics). Cells were lysed using a cell disrupter (Constant Systems, Ltd.) at 1.35 Kbar, and the cell debris was removed by centrifugation at 40,000 ϫ g for 1 h at 4°C. The supernatant was filtered (0.22 m pore size; Millipore), and incubated with glutathione-Sepharose 4B resin (GE Healthcare). The sample was washed first in 1ϫ PBS and then in buffer A (50 mM Tris-HCl, 150 mM NaCl, pH 7.5), and eluted by incubation and cleavage with HR3CP at a 1:20 enzyme:substrate (w/w) ratio for 48 h at 4°C. The protein was concentrated by ultrafiltration, and finally purified by size-exclusion chromatography on 16/600 or 10/300 Superdex 75 columns (GE Healthcare) previously equilibrated with buffer B (20 mM Tris-HCl, pH 8.0) or buffer C (20 mM Tris-HCl, 150 mM NaCl, pH 7.5).
Proteins encoded by vectors pKAR10 -pKAR13 were produced in E. coli BL21(DE3) cells, which were grown at 37°C in Luria-Bertani medium supplemented with 30 g/ml of kanamycin. Cultures were induced at an A 600 of 0.8 with 0.2-1 mM isopropyl ␤-D-thiogalactopyranoside and incubated either for 5 h at 37°C or overnight at 18°C. Cells were harvested by centrifugation at 7,000 ϫ g for 30 min at 4°C, washed in buffer A, resuspended in the same buffer, and further lysed in an ice-bath using a digital sonifier (Branson). After centrifugation at 15,000 ϫ g for 30 min at 4°C, both cell debris and supernatant were analyzed by 15% Tricine-SDS-PAGE stained with Coomassie Blue.
Protein identity and purity were assessed by mass spectrometry using an Autoflex Bruker apparatus and N-terminal sequencing through Edman degradation at the Proteomics Facility of Centro de Investigaciones Biológicas (Madrid, Spain). Ultrafiltration steps were performed with Vivaspin 15 and Vivaspin 4 filter devices of 5-kDa cut-off (Sartorius Stedim Biotech). Approximate protein concentration was determined by measuring A 280 in a spectrophotometer (NanoDrop) using the calculated absorption coefficients E 0.1% ϭ 2.32 and 2.42 for pKly18-E156A and Kly18-E156A, respectively.
Autolytic Activation and Propeptide Inhibitory Activity Assays-Mutants pKly-Y35A (from pKAR2), pKly-D25A/Y35A (pKAR4), and pKly18-Y35A (pKAR8) were incubated in buffer B at 37°C and at 0.4 mg/ml final protein concentration for up to 120 h to assay autolysis. Reactions were stopped at specific time points by boiling aliquots in reducing/denaturing buffer, and samples were further analyzed by 10% or 15% Tricine-SDS-PAGE stained with Coomassie Blue. Kly18, obtained by autolysis from pKAR1-encoded protein, was incubated at 0.025 g/ml of final protein concentration for 30 min with 0.1-10 mM peptide Q-R-L-Y-D-N-G-P-L-T (purchased from GL Biochem Ltd.), which mimics the propeptide sequence. Proteolytic activity was subsequently measured at 37°C in buffer C on substrate Mca-R-P-K-P-V-E-Nva-W-R-K(dnp)-NH 2 (Bachem; at 10 M) in a microplate fluorimeter (Infinite M200, Tecan).
Thermal Shift Assays-Aliquots were prepared by mixing 7.5 l of ϫ300 Sypro Orange dye (Molecular Probes) and 42.5 l of either pKly18-E156A (from pKAR5) or Kly18-E156A (pKAR6) at 1-2 mg/ml in buffer C in the absence and presence of 1-5 mM   (60)) was determined for both proteins from the inflection point of each curve using iQ5 software.
Crystallization and Data Collection-Crystallization assays of pKAR5-encoded pKly18-E156A protein were carried out at the IBMB/IRB Crystallography Platform by the sitting-drop vapor diffusion method using 96 ϫ 2-well MRC plates (Innovadyne). A TECAN Freedom EVO robot was used to prepare reservoir solutions, and a Phoenix/RE (Art Robbins) robot or a Cartesian Microsys 4000 XL (Genomic Solutions) was used for nanodrop dispensing. Crystallization plates were stored in Bruker steady-temperature crystal farms at 4 and 20°C. Successful conditions were scaled up to the microliter range in 24-well Cryschem crystallization dishes (Hampton Research). The best crystals were obtained at 20°C from drops containing protein solution (3.75 mg/ml in buffer B) and 100 mM Bistris propane, 200 mM potassium thiocyanate, 20% (w/v) polyethylene glycol 3350 (pH 7.5) as reservoir solution from 2:1-l drops. Crystals were cryo-protected with 20% (v/v) glycerol. Diffraction datasets were collected at 100 K from liquid-N 2 flash cryocooled crystals (Oxford Cryosystems 700 series cryostream) on a Pilatus 6M pixel detector (from Dectris) at beam lines ID23-1 and ID29 of the European Synchrotron Radiation Facility (ESRF, Grenoble, France) within the Block Allocation Group "BAG Barcelona." Crystals contained two molecules per asymmetric unit. Diffraction data were integrated, scaled, merged, and reduced with programs XDS (61) and XSCALE (62) (see Table 2 for data processing statistics of the best dataset).
Structure Solution and Refinement-The structure of pKly18-E156A was solved by likelihood-scoring molecular replacement with the program PHASER (63) using the coordinates of the protein part only of mature wild-type Kly18 (PDB access code 4IN9 (64, 65)) as searching model. Two solutions were found at final Eulerian angles (␣, ␤, ␥, in°) of 285.8, 56.7, 97.2, and 76.9, 91.3, 284.2; and fractional cell coordinates (x, y, z) of 0.120, Ϫ0.017, 0.100, and 0.997, 0.332, and 0.608, respectively. These solutions gave initial Z-scores of 8.5 and 8.3 for the rotation functions, and 6.6 and 6.2 for the translation functions, respectively, as well as a final log-likelihood gain of 1,120. A subsequent density improvement step with ARP/wARP (66) rendered an electron density map that enabled straightforward chain tracing. Thereafter, manual model building with COOT (67, 68) alternated with crystallographic refinement with programs PHENIX (69) and BUSTER/TNT (70,71), which included TLS refinement and automatically determined noncrystallographic restraints, until completion of the model. Both final model chains A and B contained residues Gln 21 -Pro 199 , as well as two zincs and one calcium each. Segment Val 36 -Gly 39 of chain A was continuous in the final Fourier map but ambiguous as to the position of the side chains. In addition, segments Gln 38 -Ser 40 and Ser 54 -His 57 of chain B were traced based on weak electron density to preserve chain continuity. Pro 122 and Pro 123 were linked by a cis-peptide bond. Three glycerol mole-cules and 226 solvent molecules completed the structure (see Table 2).
Miscellaneous-Figures were prepared with CHIMERA (72). Structural superpositions were performed with SSM (73) within COOT, and with LSQKAB (74) and ROTMAT within the CCP4 suite of programs (75). Model validation was performed with MOLPROBITY (76). The interaction surface buried at the interface between the propeptide and the mature enzyme moiety was calculated with CNS version 1.3 (77). The final coordinates of pKly18-E156A are deposited with the PDB with code 4R3V.

RESULTS AND DISCUSSION
Roles of the Propeptide in Vitro and in Cellula-Wild-type karilysin is secreted as a zymogen with a 14-residue N-terminal propeptide , which is cleaved off at position Asn 34 -Tyr 35 during maturation (Fig. 1A). This is the primary activation cleavage and it releases an active 48-kDa form (Kly48 (30)). In recombinant protein production, subsequent cleavages within the C-terminal domain give rise to Kly38 and, finally, to a stable form of 18 kDa (Kly18), which corresponds to the isolated mature catalytic domain (CD) (5,28,30,33). These cleavages were shown to be autolytic as activation was repressed by general chelating MP inhibitors and in the inactive active-site variant, E156A, which ablated the catalytic glutamate of the CSBZ (1,5,30,79,80). In addition, cleavage-site mutant Y35A, which does not match the specificity of the enzyme, was activated only slowly when compared with the wild-type (30,33).
To assess whether the propeptide had a chaperone-like function on the downstream catalytic moiety, we cloned the genes encoding pKly18-E156A and Kly18-E156A in a vector that does not attach a fusion protein at the N terminus that would assist  FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8 in proper folding (pKAR11 and pKAR13, respectively; see "Experimental Procedures"). We found that the active-site mutant pKly18-E156A was successfully overexpressed in soluble form (Fig. 1B). In contrast to the zymogen, Kly18-E156A was produced only in insoluble form (Fig. 1B). Moreover, when expressed from the pKAR6 vector, which attaches an N-terminal glutathione S-transferase fusion protein (see Table 1), Kly18-E156A was obtained with ϳ10 times lower yield than the proprotein (vector pKAR5). We conclude that the propeptide plays a major role in proper folding of Kly18 as previously described for other MPs such as fragilysin (26,27), funnelin metallocarboxypeptidases (79,81,82), and ADAMs/adamalysins (54) but not for mammalian MMPs (83).

Structure of T. forsythia Prokarilysin
We further examined the effect of the propeptide in response to denaturation by a thermal shift assay following the thermofluor approach (60). Purified pKly18-E156A (pKAR5) showed two unfolding transitions compatible with unfolding of propeptide and CD, with T m values of 60 Ϯ 0.5 and 74 Ϯ 0.5°C (Fig. 1C). In contrast, the unfolding of purified Kly18-E156A This result is in agreement with the important role of calcium in Kly18 activity, as addition of 2-5 mM CaCl 2 is reported to enhance activity about three times (30). Thus, regardless of calcium, the 14-residue propeptide redounded to a dramatic increase in T m , underpinning that it plays a major role in the thermal stability of the zymogen. Finally, we assayed the effect of a decapeptide spanning propeptide sequence Gln 21 -Thr 30 on the activity of purified mature Kly18 (from pKAR1) in the presence of a fluorogenic peptide substrate (Fig. 1D). We observed a weak but consistently concentration-dependent inhibitory effect as previously shown for other MPs when their propeptides or prodomains were added in trans, among others funnelins (79,81), ADAMs/adamalysins (84), and mammalian MMPs (85)(86)(87). Summarizing, the propeptide of karilysin is the shortest currently described to date for an MP, and it exerts all roles, which collectively or selectively had been previously described for peptidase propeptides or prodomains: latency maintenance, folding assistance during biosynthesis, stability to thermal denaturation, and inhibition of peptidolytic activity (38,39,81).
Overall Structure of pKly18 -Due to rapid autolytic processing of recombinant wild-type prokarilysin (30), crystals of pKly18 could only be obtained for an inactive variant affecting the catalytic glutamate (pKly18-E156A), as already reported for other MP zymogens (88 -92). This protein crystallized as monoclinic crystals diffracting to 2-Å resolution with two molecules per asymmetric unit. These were essentially identical (C␣-atom root mean square deviation ϭ 0.53 Å) except for segments Asn 53 -His 57 and Asn 34 -Gly 39 . The latter flanks the primary activation cleavage point and is flexible. It is stabilized through an interaction with segment Asn 87 -Asn 89 of the second molecule present in the asymmetric unit of the crystal although in different conformations in molecules A and B, so the discussion hereafter is centered on molecule A if not otherwise stated. When two values are indicated, they refer to both molecules.
The protein reveals a compact, almost spherical shape of ϳ40 Å in diameter and is subdivided into three moieties ( Fig.  2A): the N-terminal propeptide (Gln 21 -Asn 34 ), and a CD split into a larger N-terminal upper subdomain moiety (Tyr 35 -Gly 162 ; NTS) and a smaller C-terminal lower subdomain moiety (Ile 163 -Pro 199 ; CTS, see also Refs. 9 and 28)) if viewed in the standard orientation for MPs (35). NTS and CTS conform to the overall fold of vertebrate MMPs (47,93) and are separated by a shallow active-site cleft. The NTS is an ␣/␤-sandwich consisting of a twisted five-stranded pleated ␤-sheet (strands ␤I-␤V; see Fig. 2A), which is parallel for its first four strands and antiparallel for its lowermost one, ␤IV. The sheet accommodates on its concave side two ␣-helices (the "backing helix" ␣A and the "active-site helix" ␣B; for numbering and extension of repeating secondary structure elements, see Fig. 2C of Ref. 9). The right-handed twist of the helices coincides with the righthanded twist of the sheet and both helices' axes intersect the strands of the sheet at an angle ⍀ Ϸ Ϫ35° (94). The two helices pack against each other interacting through Ala 66 -Ala 70 of ␣A and Leu 149 -Ala 154 of ␣B at a crossing angle ⍀ Ϸ Ϫ50°, which corresponds to a class II helix interaction (94). The loop connecting strands ␤III and ␤IV (L␤III␤IV) contains the "S-loop" (Gly 100 -Leu 115 ), which encompasses first a binding site for a structural zinc (Zn 998 ) and, downstream, a binding site for a structural calcium (Ca 997 ; see Fig. 2B Fig. 2B and its legend for details). The presence of calcium is consistent with its crucial role in catalysis (30) and in protein stability (see Fig. 1C). Such calcium is found in mammalian MMP structures (47,93), but it was not found in previous mature Kly18 structures (see below and Refs. 9 and 28). At Gly 162 of the CSBZ, the polypeptide chains take a sharp turn and enters the CTS ( Fig. 2A), which mainly contains the "C-terminal helix" ␣C and the Met-turn, centered on Met 173 , which forms a hydrophobic base for the catalytic metal-binding site and is required for its integrity in MMPs and metzincins in general (47).
The active-site cleft contains the catalytic zinc ion (Zn 999 ) at half-width coordinated by the three histidines of the CSBZ (His 155 , His 159 , and His 165 ) through their N⑀2 atoms at distances 2.00 -2.05Å (Fig. 2, A and C). The cleft is top-framed on its non-primed side (see Refs. 35 and 95)) by the "upper-rim strand" ␤IV of the NTS ␤-sheet, which in MMPs binds substrates in extended conformation from left to right through antiparallel ␤-ribbon-like interactions. On its primed side, the cleft is top-framed by the final stretch of the S-loop, termed the "bulge-edge segment" (Thr 112 -Ley 115 ), and bottom-framed by the segment bridging the Met-turn and helix ␣C. This segment includes the "S 1 Ј-wall forming segment" (Pro 175 -Tyr 177 ) at the front and the "specificity loop" (Gly 179 -Gln 183 ) at the back. Together with the first turn of the active-site helix ␣B, the latter structural elements contribute to the size and chemical nature of the S 1 Ј pocket, which confers specificity to Kly18 and also MMPs in general (47,93), here for medium-sized to bulky hydrophobic residues (30). Side chains participating in pocket shaping include Leu 115 , Ala 116 , Thr 151 , Val 152 , His 155 , Leu 172 , Tyr 177 , and Lys 181 .
Inhibition by the Propeptide-The 14-residue propeptide starts at the front right and runs in extended conformation across the active-site cleft, thus blocking access to the cleft, though in the opposite direction to a substrate, i.e. right to left (Fig. 2, A and C). This reverse orientation of the propeptide in the cleft may contribute to attenuate autolysis, as previously suggested for zymogens of cysteine peptidases and mammalian MMPs (39). The interaction with the CD buries a surface of 2,100 Ϯ 35 Å 2 , which is much larger than the average of monomeric protein-protein domain intra-chain interfaces (1,193 Å 2 (96)) but is slightly lower than the range of typical MMP-protein inhibitor interaction surfaces (2,400 -2,700 Å 2 ; see Ref. 97). The interaction includes 13 hydrogen bonds, a double salt bridge, one metallorganic bond, and hydrophobic carbon-carbon contacts between eight residues from the propeptide and 11 from the CD (see Table 3). Segments involved include almost the entire propeptide (Arg 22 -Gly 31 ) and, from the CD, mainly Asn 111 -Tyr 120 from the bulge-edge segment and the upper-rim strand, and Pro 175 -Tyr 177 from the S 1 Ј-wall forming segment. Further involved are Tyr 106 , Ala 124 , and Glu 138 and the zincliganding histidine side chains. Four inter-main chain hydrogen bonds form on the primed side of the cleft (two with the S 1 Јwall forming segment and two with the bulge-edge segment and strand ␤IV) and three more on the upstream non-primed side (with ␤IV and L␤IV␤V; Fig. 2C). In particular, Arg 22 contacts the base of the S-loop: it doubly salt bridges Glu 138 , which is also one of the calcium ligands (see above, Table 3 and Fig.  2B), and hydrogen bonds three carbonyl oxygens of the S-loop, Asn 111 , Gly 113 , and Thr 112 , which, again, is also a calcium ligand. In addition, the Arg 22 carbonyl oxygen binds the S 1 Јwall forming segment and its side chain performs a hydrophobic interaction with Leu 115 . Accordingly, this residue plays a major role in the stabilization of the Ca 997 site and, thus, the zymogen in general, which explains its enhanced stability in response to thermal denaturation (see above). In addition, superposition of pKly18-E156A onto mature Kly18 in complex with a tetrapeptidic cleavage product in the primed side (see below) and human MMP-8 with a modeled substrate traversing its cleft based on inhibitor structures (98) indicates that Arg 22 occupies the S 3 Ј position of the cleft.
However, the most important interaction of the propeptide with the CD is exerted by Asp 25 , which approaches the catalytic zinc from the top and monodentately occupies through its O␦1 atom the fourth position of the tetrahedral coordination sphere of the metal (2.00/2.04 Å apart; Fig. 2C) further to His 155 , His 159 , and His 165 N⑀2 atoms. The preceding carbonyl group of Tyr 24 binds strand ␤IV, and its aromatic side chain penetrates the deep hydrophobic S 1 Ј pocket, mainly interacting with the FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8

JOURNAL OF BIOLOGICAL CHEMISTRY 4733
His 155 ring face-to-face. The -rings are ϳ3.5 Å apart and parallel but slightly displaced along the ring planes to form a halfoverlapping sandwich, which gives rise to an optimal -stacked structure (99). Downstream in the chain, Pro 28 is in a pocket, probably S 2 , framed by His 159 , Glu 164 , and Tyr 120 , the latter two interact through a tight hydrogen bond (Tyr 120 O-Glu 164 O⑀2,

Structure of T. forsythia Prokarilysin
2.61 Å). Residue Leu 29 is surrounded by the side chains of Tyr 106 , His 117 , and Phe 119 , which may feature S 3 (Fig. 2C). After Gly 31 , the polypeptide abandons the active-site cleft moving outward to reach the primary activation cleavage point, Asn 34 -Tyr 35 (Fig. 2A), after which the chain folds back toward the molecular moiety and enters strand ␤I of the NTS ␤-sheet.
A Novel Activation Mechanism in MMPs-Previous work had yielded three structures of mature wild-type Kly18 in com-plexes with tri-and tetrapeptidic cleavage products, as well as an inhibitory tetrapeptide in the non-primed side of the cleft (PDB 2XS3, 2XS4, and 4IN9 (9, 28)). These were obtained both in the presence and absence of magnesium and showed deviating chain traces for segment Asn 53 -His 57 (L␤I␣A) in the two molecules found in the asymmetric unit of the magnesium unbound structure (PDB 2XS3 (9)) and in the single molecules found in magnesium-bound (PDB 2XS4 (9)) and inhibitor-  109)) distance values for oxygens and nitrogens. C, close-up of A in wall-eye stereo centered on the catalytic zinc after a horizontal ϳ30°r otation upwards. Selected hydrogen and ionic bonds (see also Table 3) are depicted as green lines. Residues and ions labeled in A are not labeled here for clarity. The propeptide is shown in cyan to distinguish it from the mature catalytic moiety (in tan/yellow/orange) and its chain direction is pinpointed by a cyan arrow and labels of the N-and C-terminal parts depicted. D, superposition in wall-eye stereo of pKly18-E156A (ribbon in tan for the mature enzyme moiety and in brown for the propeptide, zinc ions in magenta, and calcium ion in red; stick model for the side chains of Ser 20 -Tyr 35 with carbons in brown) and Kly18 (ribbon and zinc ions in pink, see PDB 2XS3, molecule A (9)), which was obtained in a product complex with peptide A-F-T-S bound to the primed side of the cleft (stick model with carbons in gold). Tyr 35 is shown for both structures. E, detail of D in wall-eye stereo depicting the large rearrangement of the N terminus at Tyr 35 after maturation cleavage at Asn 34 -Tyr 35 . The ␣-amino group of Tyr 35 makes a salt bridge with the side chain of Asp 187 in the mature enzyme. Aside from Tyr 120 and Glu 164 (significantly) and Pro 122 -Ala 129 (slightly; see black arrows), maturation does not entail major conformational rearrangement of the rest of the structure. FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8 bound crystals (PDB 4IN9 (28)). In addition, significant differences were also found in the second half of the S-loop including the bulge-edge segment, which was metal-free in all structures, as the aforementioned magnesium, which coincides with a potassium site in the inhibitor-bound form, was found on the opposite surface of the CD (see Fig. 1, A and C, in Ref. 9,and Fig. 1A in Ref. 28), in a place that suggests little if any functional or structural relevance. In these structures, either an outward-or an inward-folded flap was found for the S-loop (Fig. 1E in Ref. 9 and Fig. 1D in Ref. 28), which suggests intrinsic flexibility of this protein segment to adapt to different substrates. Among the distinct mature Kly18 coordinates, molecule A of the magnesium-unbound structure (PDB 2XS3) was chosen here for comparison with pKly18-E156A as it showed the lowest divergence in the overall chain trace (Fig. 2, D and E).

Structure of T. forsythia Prokarilysin
Superposition revealed that the mature CD is preformed in the zymogen and, with some notable local exceptions (see below), is simply uncovered during maturation by removal of the propeptide, as found in mammalian MMPs (47) and other MPs such as funnelins (79,82). Removal occurs through cleavage at Asn 34 -Tyr 35 , which is solvent exposed on the molecular surface and thus readily accessible for processing ( Fig. 2A). This explains why the wild-type zymogen undergoes rapid autolysis, so it cannot be isolated intact (see Ref. 30 and first section of "Results and Discussion"). This was the first cleavage observed in vitro, thus termed primary activation cleavage site, and no further cleavage was detected either within the propeptide or in the CD. The site is consistent with most vertebrate MMPs being activated at X-F/Y bonds, which are found at similar regions in all structures (10). Propeptide removal occurs under loss of a number of protein-protein interactions (see Table 3 and the preceding section), which explains why the mature enzyme is less stable to thermal denaturation (see first section of "Results and Discussion"). In particular, Arg 22 plays a key role in stabilizing the Ca 997 site (see above), and its removal may contribute to cation-site and S-loop flexibility, leading to metal loss. This site is easily created from the unbound form by two glycinemediated main chain rotations (peptide flip of bond Thr 112 -Gly 113 and ϳ70°rotation of peptide bond Gly 110 -Asn 111 ), so as to orient the carbonyl oxygens toward the interior, and cation binding should largely compensate for the energetic cost of such minor rearrangement. However, the finding that none of the mature Kly18 structures, which were partially obtained in the presence of calcium (9), contained an intact calcium site supports the requirement of Arg 22 as an additional stabilizing factor for site integrity.
Activation further entails that the position occupied by Asp 25 O␦1 in the ligand sphere of the catalytic zinc (see the preceding section) is taken over by a catalytic solvent molecule, which renders a competent active site following an "aspartate-switch" mechanism. Such a competent zinc environment has also been reported for several mature MPs (see e.g. Refs. 64, 80, and 100). To date, aspartate-switch zymogenic mechanisms have been described only for astacins (7,88) and fragilysins (26), which are only distantly related MPs grouped with MMPs within the metzincins. To verify the function of Asp 25 in latency in pKly18, we used mutant pKly18-Y35A (from pKAR8), as the wild-type form (pKAR7) was insoluble. Although this mutant was pro-duced with a yield similar to that of pKly18-E156A and was stable for several days, mutant pKly18-D25A/Y35A (pKAR9) was insoluble. We further assessed the function of Asp 25 in full-length karilysin using the slowly autolytic mutant pKly-Y35A (pKAR2), as the reaction in the wild-type is too rapid (30). While pKly-Y35A was essentially intact after 5 days at 37°C, pKly-D25A/Y35A (pKAR4) had been entirely transformed into the 38-and 18-kDa forms after this time (Fig. 1, E and F). Taken together, these results support the essential role of Asp 25 in latency maintenance.
As to further changes upon maturation, segment Pro 122 -Ala 129 from L␤IV␤V is slightly shifted downwards by ϳ2 Å and the side chains of Tyr 120 and Glu 164 rotate toward the zinc site (Fig. 2E). Activation only entails major rearrangement of the new N-terminal segment Tyr 35 -Ser 40 , on the left surface (Fig. 2,  D and E), which is rotated downward around bonds C-C␣ and C␣-N of Ser 40 . In this way, this segment nestles in a surface cavity framed by helix ␣C and the first segment of the CTS between Gly 162 , and the "family specific residue," which is a serine in MMPs (1, 101) (here Ser 166 ). This entails that the new ␣-amino group of Tyr 35 , which is translated 25 Å, establishes an intra-molecular salt bridge with Asp 187 of ␣C, which is vaguely reminiscent of the activation of trypsin-like serine peptidases (102). Asp 187 , in turn, is itself further bound to Ser 166 and is adjacent to a second aspartate, Asp 188 , which binds two main chain amides of the Met turn. This electrostatic network is characteristic of physiologically relevant mature MMPs, also referred to as "superactive forms" (47,103). With the exception of the mature N-terminal fragment, the rest of this electrostatic network is already present in the zymogen (Fig. 2E).
The prodomain globular core serves as a scaffold to place a downstream peptide, which runs in extended conformation in the opposite direction to a bound substrate and thus blocks the active-site cleft (Fig. 3, A and B). This peptide encompasses the conserved motif involved in cysteine-switch or Velcro latency characteristic of animal and plant MMPs (48 -50), 100 P-R-C-G-N-P-D 106 (MMP-2 residues in italics; see PDB 1EAK and UP P08253), which is equivalent to pKly18 segment 23 L-Y-D-N-G-P-L 29 (Fig. 3, C and D). Both the cysteine-and aspartate-switch motif show an intricate electrostatic network producing a unique scaffold to interact with the mature catalytic domain moiety. In contrast to pKly18, where the first cleavage occurs in the primary activation cleavage site, however, classical mammalian pro-MMPs are activated by conformational changes in the prodomain induced by cleavage in a so-called "bait region" by several peptidases such as trypsin, plasmin, and other MMPs. Activation follows a "stepwise activation" process to eventually yield the final cleavage site X-F/Y accessible for processing and dissociation of cysteine and zinc to generate a functional active site (48, 49, 51, 106 -108). As in Kly18, after cleavage at Asn 109 -Tyr 110 , the new N terminus is rearranged and participates in the electrostatic network centered on the conserved aspartate of helix ␣C, Asp 346 in MMP-2.
Conclusions-This examination of the structure and function of the zymogen of the first bacterial MMP to be studied biochemically has uncovered several features of the activation mechanism of pKly18, which are shared with animal and plant MMPs: (i) the relevant cleavage site is X-F/Y; (ii) the scissile bond is located in similar regions of the structure; (iii) activation entails rearrangement of the segment equivalent to Tyr 35 -Ser 40 to yield a salt bridge between the new ␣-amino group and the first of two conserved aspartates in helix ␣C; (iv) this aspartate is bound to the family-specific serine; (v) the aspartate immediately downstream binds two main chain amides of the Met turn; (vi) the inhibitory segments run across the cleft in the opposite direction to a genuine substrate and metal blockage occurs through the side chain of an intervening residue, not through a chain terminus; and (vii) the catalytic moiety is largely preformed in the zymogen. All these features are related to the highly conserved CD itself. In contrast, all features of the mechanism related to the segment preceding this conserved CD diverge: (i) in pKly the propeptide spans just 14 residues and does not contain repetitive secondary structure elements, whereas eukaryotic MMPs feature a true protein prodomain that folds into a pseudosymmetric three-helix bundle followed by a segment in extended conformation; (ii) no relevant sequence similarity is found between the proregions; (iii) in eukaryotic MMPs activation occurs through a cysteine-switch mechanism exerted by residues from a conserved sequence motif, whereas in pKly18 this motif is absent and activation follows an aspartate-switch mechanism; (iv) multiple cleavages are apparently required in eukaryotic MMPs to liberate the CD, whereas a single cleavage suffices in pKly; and (v) the prodomain is not required for (re)folding of the catalytic moieties in eukaryotic MMPs, whereas it is in karilysin. In addition, pKly shares parts of its mechanism of latency with otherwise unrelated MPs from the astacin and fragilysin families. Accordingly, this overall novel mechanism unveiled for MMPs supports previous hypotheses, according to which Kly18 originated from an animal MMP CD co-opted through horizontal gene transfer by T. forsythia. This transfer was fostered by the intimate coexistence of the latter with the human blood-irrigated gingival crevice. Subsequently, Kly18 would have evolved in a bacterial envi-  (89); MMP-2 residues in italics with superscripted numbering), shown only for its CD (Tyr 110 -Asp 52 in cyan; the fibronectin type-II domains spanning Gln 219 -Asp 92 have been omitted, the black arrows pinpoint the insertion points) and prodomain (Pro 43 -Asn 109 in pink, without the first 11 residues in extended conformation). The orientation displayed corresponds to that of Fig. 2A after applying a horizontal rotation of 15°. Residues of the conserved motif (Pro 100 -Asp 106 ) key for structural integrity of the inhibitory segment are depicted for their side chains. B, close-up of A after removal of prodomain segment Pro 43 -Asn 66 to provide insight into the interactions of the conserved motif. Key electrostatic interactions are shown as green lines. The catalytic glutamate, Glu 404 , is replaced by a glutamine, the histidines from the CSBZ are His 403 , His 407 , and His 413 . C and D, scheme depicting the interaction modi of the propeptides of pro-MMP-2 through a cysteine-switch mechanism (C) and pKly18 through an aspartate-switch mechanism (D). The catalytic zinc ions are shown as magenta spheres and relevant interactions are shown as yellow dashed lines. FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8 ronment, where it was furnished with unique flanking domains that contribute to a mechanism of zymogenicity similar to distantly related MPs only (9).