Proenzyme Structure and Activation of Astacin Metallopeptidase*

Proteolysis is regulated by inactive (latent) zymogens, with a prosegment preventing access of substrates to the active-site cleft of the enzyme. How latency is maintained often depends on the catalytic mechanism of the protease. For example, in several families of the metzincin metallopeptidases, a “cysteine switch” mechanism involves a conserved prosegment motif with a cysteine residue that coordinates the catalytic zinc ion. Another family of metzincins, the astacins, do not possess a cysteine switch, so latency is maintained by other means. We have solved the high resolution crystal structure of proastacin from the European crayfish, Astacus astacus. Its prosegment is the shortest structurally reported for a metallopeptidase, and it has a unique structure. It runs through the active-site cleft in reverse orientation to a genuine substrate. Moreover, a conserved aspartate, projected by a wide loop of the prosegment, coordinates the zinc ion instead of the catalytic solvent molecule found in the mature enzyme. Activation occurs through two-step limited proteolysis and entails major rearrangement of a flexible activation domain, which becomes rigid and creates the base of the substrate-binding cleft. Maturation also requires the newly formed N terminus to be precisely trimmed so that it can participate in a buried solvent-mediated hydrogen-bonding network, which includes an invariant active-site residue. We describe a novel mechanism for latency and activation, which shares some common features both with other metallopeptidases and with serine peptidases.

The proteolytic activity of most metallopeptidases (MPs) 3 is regulated, and it is only exerted where and when required (1). Such control may occur through modulation of gene expression, compartmentalization, allostery, or inhibition by protein inhibitors. Another regulatory mechanism is zymogenic latency, which is provided by mostly N-terminal prosegments. These block access of substrates to the active-site cleft, and they are removed by limited proteolysis during maturation (2,3). Such prosegments often fold independently and guide on their part the folding process of the cognate protease domains. They may also act as intramolecular chaperones or inhibitors of the mature enzymes in trans and in intracellular sorting of the zymogen (2). Therefore, the study of the molecular mechanisms by which MPs maintain latency is indispensable to an understanding of their basic mode of action. It also paves the way for the design of inhibitors that mimic the latent state so as to modulate MP activity as part of therapeutic approaches. Detailed three-dimensional structural information can contribute much to this understanding (4). However, among the MPs, only funnelins, lysostaphins, thermolysins, and matrix metalloproteinases (MMPs) have been structurally analyzed for their zymogens. Results reveal that the mature enzyme moieties are already in a competent conformation. Notwithstanding, each group displays a distinct mechanism of latency maintenance (5)(6)(7)(8)(9)(10)(11).
The metzincins are a clan of MPs characterized by a consensus sequence responsible for binding of the catalytic zinc ion (CSBZ), HEXXHXX(G/N)XX(H/D), and a conserved methionine-containing turn (Met-turn) (12)(13)(14)(15)(16). This clan can be subdivided into several families, including the MMPs. In the latter, the CSBZ encompasses three histidine zinc ligands, the general base/acid glutamate for catalysis, and an essential glycine (15). A total of 23 MMP paralogs are present in humans (15). Most of them comprise a prosegment upstream of the catalytic domain, and, to date, structures of pro-MMP-1, -2, -3, and -9 have been reported (17)(18)(19)(20)(21). They show 60 -90-residue globular prosegments, which comprise a three-helix bundle that creates a scaffold to place a segment in extended conformation to block the active-site cleft. This segment includes a conserved "cysteine switch" or "Velcro" sequence motif (22)(23)(24), PRCGXPD, which runs through the active-site cleft in the opposite direction to a substrate bound along the cleft. The cysteine S␥ atom binds the catalytic zinc ion, thus replacing the catalytic solvent molecule present in mature MMPs. The competent conformation of the cysteine switch peptide is maintained by a double salt bridge between the arginine and the aspartate of the motif. It was hypothesized that this mechanism is shared, with some variation, by other metzincin families for which conserved cysteine switch motifs were found or suggested, such as the ADAMs/ adamalysins (conserved motif PKMCGV (25,26)), leishmanolysins (motif HRCIHD (14)), and pappalysins (motif CG (27)).
In addition, the metzincins also include the astacins, which owe their name to the prototypical digestive enzyme, astacin, from the European freshwater crayfish, Astacus astacus L. (13, 28 -30). Astacins are found throughout the animal kingdom and in bacteria. Several of these enzymes are crucial for embryonic development, tissue differentiation, and extracellular matrix assembly, and they are thus therapeutic targets. They consist of ϳ200-residue catalytic moieties preceded by prosegments of varying length (Fig. 1). Several family members have additional downstream domains engaged in substrate recognition, membrane anchors, or signaling. The prototypical crayfish astacin is synthesized as a preproenzyme in the midgut gland (31). The 15-residue signal peptide is removed during secretion, and the proenzyme, which comprises a 34-residue prosegment and a 202-residue catalytic domain, is only transiently found within the ducts between the hepatopancreas and the stomach. Once in the stomach, the zymogen is activated, and the mature enzyme participates in collagenolysis and gelatinolysis during digestion (32). Optimal astacin substrates comprise at least five amino acids with small aliphatic residues in PЈ 1 , proline in PЈ 2 , bulky hydrophobic residues in PЈ 3 , and basic residues in P 1 and P 2 (peptide substrate and enzyme subsite nomenclature according to Refs. 33 and 34). Cleavage is best performed at pH 6 -8. A comprehensive review on the enzyme was written by Stöcker and Yiallouros (31).
The prosegment of most astacins contains a short sequence with a conserved aspartic acid residue (Fig. 1). However, there is no chemical evidence for a function of this aspartate in latency maintenance; site-directed mutations of this residue to alanine or asparagine and subsequent expression of the mutant proteins in bacteria resulted in non-foldable, unstable proastacin variants (35). Therefore, we solved the high resolution structure of proastacin in an attempt to shed light on this potentially novel mechanism of latency among MPs.

EXPERIMENTAL PROCEDURES
Protein Preparation and Crystallization-Recombinant proastacin E93MA/I91ML (UniProt Q9U918; numbering is based on the mature enzyme (M); see below) was expressed in Escherichia coli BL21(DE3) cells as inclusion bodies, purified by Ni 2ϩ -nitrilotriacetic acid affinity chromatography, and folded by dilution and removal of reducing agents and guanidinium chloride as described (32). The resulting protein in 50 mM AMPSO buffer, pH 9.0, was concentrated in Amicon Ultra-15 centrifugal filter units of 10 kDa cut-off (Millipore, Bad Schwalbach, Germany) to a volume of about 1 ml. The protein was applied to a HEMA-Bio gel filtration column (MZ Analysentechnik, Mainz, Germany) attached to a Kontron high pressure liquid chromatography unit to separate monomeric proastacin from an oligomeric fraction with a flow rate of 0.5 ml/min. The composition of the fractions was analyzed by native PAGE and SDS-PAGE. Crystallization assays were performed with the monomeric protein fraction by the sitting drop vapor diffusion method. Reservoir solutions were prepared by a Tecan robot, and 200-nl crystallization drops were dispensed on 96 2-well MRC plates (Innovadyne) by a Cartesian (Genomic Solutions) nanodrop robot at the High-Throughput Crystallography Platform of the Barcelona Science Park. Best crystals appeared in a Bruker steady temperature crystal farm at 4°C with protein solution (10 mg/ml in 50 mM AMPSO, pH 9.0) and 20% polyethylene glycol 8000, 0.1 M (NH 4 ) 2 SO 4 , 0.01 M MgCl 2 , 0.05 M MES, pH 5.6, as reservoir solution. These conditions were efficiently scaled up to the microliter range with 24-well Cryschem crystallization dishes (Hampton Research). Crystals were cryoprotected with 16% polyethylene glycol 8000, 20% glycerol, 0.1 M (NH 4 ) 2 SO 4 , 0.01 M MgCl 2 , 0.05 M MES, pH 5.6. A complete diffraction data set to 1.45 Å resolution was collected at 100 K (provided by an Oxford Cryosystems 700 series cryostream) from a single liquid N 2 flash-cryocooled crystal on a MarCCD detector at beam line ID23-2 of the European Synchrotron Radiation Facility (Grenoble, France) within the Block Allocation Group "BAG Barcelona." The crystal was primitive orthorhombic, with one molecule per asymmetric unit. Diffraction data were integrated, scaled, merged, and reduced with programs XDS (36) and SCALA (37) within the CCP4 suite of programs (see Table 1).
Structure Solution and Refinement-The structure of proastacin was solved by Patterson search methods with the program PHASER (38) by using the coordinates of mature astacin (Protein Data Bank code 1AST) (39,40) as searching model. A single solution was found at 358.0, 23.2, and 149.4 (␣, ␤, and ␥ in Eulerian angles) and 0.353, 0.094, and Ϫ0.624 (x, y, and z as fractional unit cell coordinates) after rigid body refinement. This solution gave a Z score of 17.8 for the rotation function and 25.5 for the translation function, as well as a final log likelihood gain of 1133. A subsequent density improvement step with ARP/wARP (41) with all data rendered an electron density map that enabled straightforward chain tracing. Subsequently, manual model building on a Silicon Graphics work station with TURBO-Frodo (42) alternated with crystallographic refinement with program REFMAC5 until completion of the model (see Table 1). This model contained all of the protein residues of the prosegment (Ser 1P to Gly 34P ; prosegment residues bear the suffix "P") and the mature protease moiety (Ala 1M -Arg 201M ; suffix "M") (see Fig. 1 for the numbering convention) and the catalytic zinc ion. Two segments, Ala 26P -Gln 29P and Asp 129M -Gly 138M , were flexible and traced on the basis of weak electron density, the latter with two alternative conformations for Asp 129M -Pro 135M .
Sequence Alignment-Amino acid sequences were aligned by using ClustalX (43) with default parameters and then manually readjusted with GENEDOC (44) in accordance with structural constraints emanating from the present data and from observations by Stöcker et al. (28). Before alignment, each sequence was processed in silico with SIGNAL P (available on the World Wide Web) to remove the signal peptides (45). (46) and TURBO-Frodo. The interaction surface between the prosegment and the mature enzyme moiety was calculated with CNS version 1.2 (47) as the half of the surface area buried at the complex interface. Close contacts were ascertained with the latter program and the PISA server.

RESULTS AND DISCUSSION
Overall Structure of Proastacin-In order to prevent autolysis, proastacin was recombinantly overexpressed as a mutant, in which the general base/acid glutamate, Glu 93M , had been replaced by alanine to create a catalytically compromised variant (see Ref. 32). The properly folded monomeric fraction of this protein crystallized and rendered diffraction data to 1.45 Å resolution, which yielded a detailed and accurate picture of the zymogenic structure of proastacin ( Fig. 2A). It is an ellipsoid of overall dimensions 55 ϫ 45 ϫ 28 Å divided into three major regions: an N-terminal prosegment (residues Ser 1P -Gly 34P ), an upper subdomain (Ala 1M -Gly 99M ), and a lower subdomain (Phe 100M -Arg 201M ) according to the standard orientation for MPs (Figs. 1 and 2B). The two subdomains, of similar size, constitute the mature enzyme moiety. They resemble the two valves of a clamshell that delimitate a deep narrow horizontal active-site cleft, which explains the preference of astacin for elongated substrates.
The Mature Enzyme Moiety-The upper subdomain contains a twisted five-stranded ␤-sheet (strands ␤3-␤7; Fig. 2B), with all strands except the fourth (␤6) parallel to each other and to any substrate that is bound in the cleft. As determined for the mature enzyme structure, protein and peptide substrates are bound left-to-right, mainly on the non-primed half of the cleft (i.e. N-terminally of the scissile bond in peptide substrates).
Binding involves antiparallel inter-main-chain interactions with ␤6, which constitutes the upper rim strand of the cleft and forms the lower edge of the upper subdomain (39,40). The concave face of the ␤-sheet accommodates two helices, the backing helix (␣2) and the active-site helix (␣3) (Fig. 2B). The latter comprises the first two zinc-binding histidine residues (His 92M and His 96M ; Fig. 2C) and the general base/acid glutamate (here alanine) of the CSBZ. The end of the helix at Gly 99M is also the boundary with the lower subdomain, which contains the third zinc-binding histidine residue, His 102M . This subdomain comprises a flexible segment, Asp 129M -Gly 138M ( Fig. 2D; see below) and a few repetitive secondary structure elements, mainly a short two-stranded ␤-sheet (strands ␤8 and ␤9; see Fig. 2B) and a C-terminal helix (␣4) at the end of the polypeptide chain. The latter helix has a kink at Tyr 194M , which bends it about 50°. The C terminus of the protein resides at the molecular surface, and two disulfide bonds (Cys 42M -Cys 198M and Cys 64M -Cys 84M ) contribute to the overall stability and rigidity of the molecule (Figs. 1 and 2). A structural hallmark of metzincins is the Met-turn (12)(13)(14)(15), which spans Ser 145M -His 148M in astacin and contains a methionine at position 3 that is fully conserved (even for its side-chain conformation) among all structurally characterized metzincin prototypes (12, 14 -16). This conservation may be attributable to the maintenance of the central active-site core structure of these enzymes.
A Novel Prosegment Structure-With 34 residues, proastacin has the shortest metallopeptidase prosegment structurally analyzed to date, and it has a unique fold (5)(6)(7)(8)(9)(10)(11). In contrast to other (metallo)peptidases, it does not have an intramolecular chaperone function in astacin, as revealed by refolding experiments with the heterologously expressed recombinant mature protein (32,48,49). The prosegment is elongated and stabilized through several intramolecular interactions. It runs across the front surface of the mature enzyme moiety blocking the cleft, which is consistent with previous functional studies (32) (Fig. 2,  B and E). The interaction with the catalytic domain buries an interface of 1,580 Å 2 and is based on 75 close contacts (Ͻ4 Å), among them 20 hydrogen bonds, two organometallic interactions, and hydrophobic contacts between 27 pairs of residues (see Table 2). The N terminus of the prosegment at Ser 1P is anchored to the lower left of the lower subdomain through a solvent-mediated hydrogen bond with Ile 156M O, within the segment connecting the Met-turn and the C-terminal helix. The polypeptide extends up along the molecular surface to the active-site cleft. From Glu 6P on, the chain runs through the active-site cleft in reverse orientation to a substrate approximately to Arg 25P and adopts a helical conformation (helix ␣1) until Tyr 12P (Figs. 1 and 2B). This segment nestles in the primed side of the cleft (i.e. C-terminally of the scissile bond in peptide substrates), approximately until subsite SЈ 1 , which is partially occupied by the side chain of Leu 11P . This helical structure may contribute to the prevention of autolysis because substrates usually bind in extended conformation to active-site clefts (2). Possibly as a result of the aforementioned helix and the flexible segment, Asp 129M -Gly 138M , within the lower subdomain, the two subdomains are more separated in proastacin than in the unbound mature enzyme (Protein Data Bank code 1AST) (40). In contrast, in the complex of mature astacin with a reaction  (50). Among the residues of helix ␣1, three alanines (Ala 7P , Ala 8P , and Ala 10P ) and Tyr 12P interact with protein residues of the mature enzyme moiety ( Fig. 2E and Table 2). At Asn 14P , which contacts preceding and downstream carbonyl oxygens of the prosegment through its N␦2 atom, the chain protrudes outward from the protein moiety and enters a wide loop that ends at Asp 21P . This loop comprises two subsequent 1,4-turns (Asn 14P -Met 17P and Met 17P -Gly 20P ), which together with hydrophobic interactions of Met 17P with both Lys 23P and Trp 65M give rise to a scaffold. This is optimized for positioning Asp 21P to approach the catalytic zinc ion (Fig. 2E). The loop structure is further stabilized by a hydrophobic interaction between the side chains of Phe 18P and Leu 24P and by three key electrostatic interactions of the side chain of Glu 19P (see Table 2) with the third zinc-binding histidine, His 102M , with Thr 105M , and with the penultimate residue of the prosegment, Arg 32P (Glu 19P O⑀2-Arg 32P N1, 3.01 Å; see Fig. 2E). This last salt bridge is vaguely reminiscent of a similar arrangement found in MMP prosegments (see above). Asp 21P coordinates the catalytic zinc ion from the top in a bidentate manner. Together with the mature enzyme ligands, this results in a distorted octahedral coordination sphere, which is unusual for zinc (51) (Fig. 2C). The mature protein ligands are the N⑀2 atoms of His 92M , His 96M , and His 102M as well as Tyr 149M O. The O␦2 atom of Asp 21P replaces the zincbound solvent molecule in the mature enzyme (ϳ0.5 Å away). The side chain of Tyr 149M , which undergoes a "tyrosine switch" motion upon substrate binding (31), is closer to the conformation in the unbound mature enzyme than in the bound enzyme, where it stabilizes one of the two gem-diolate oxygens of the FIGURE 1. Sequence alignment of the astacin family. Aligned are the prosegments and catalytic domains of representative astacin family members: AAS_ASTA, astacin from the crayfish A. astacus (UniProt P07584); HSA_MEP␣, human meprin ␣ (Q16819); HSA_MEP␤, human meprin ␤ (Q16820); HVU_HMP2 from the freshwater hydroid Hydra vulgaris (Q9XZG0); LPO_ASTM from the horseshoe crab Limulus polyphemus (B4F320); CEL_NAS13 from the nematode Caenorhabditis elegans (Q20191); TPA_MYO1 myosinase from the giant squid Todarodes pacificus (Q8IU46); ATE_TLL, tolloid-like proteinase from the spider Achaearanea tepidariorum (Q75UQ6), HSA_BMP1, human bone morphogenetic protein 1 (P13497); SPU_SPAN, proteinase from the sea urchin Strongylocentrotus purpuratus (P98068); OLA_HCE1, high choriolytic enzyme from the medaka fish Oryzias latipes (MEROPS data base access code MER001105); HSA_OVAST, human embryonic ovastacin (Q6HA08). Highlighted in black are the regions containing the conserved aspartic residue (in yellow) within the prosegment, the activation site (the mature N terminus after cleavage is shown in yellow) at the interface between the prosegment and the catalytic domain, the CSZB (the astacin hallmark glutamate is highlighted in yellow), and the Met-turn region. Also highlighted are the segment mainly engaged in shaping of subsite SЈ 1 and the four conserved cysteine residues. The final activation cleavage site (Gly 34P -Ala 1M in astacin) is pinpointed by scissors. Regular secondary structure elements of proastacin are indicated above the alignment in orange (prosegment) and blue (mature protease domain) as cylinders (helices) and arrows (strands). Selected residues of the pro-domain and the catalytic domain of astacin are numbered and carry the suffixes P and M, respectively (1P-34Pϩ1M-201M). tetrahedral reaction intermediate (50). However, the distance of the phenolic oxygen to the cation in proastacin is slightly greater than in the unbound mature structure, 3.01 versus 2.53 Å, possibly due to the nearby side chain of Asp 21P (Fig. 2C).
The four residues after Asp21P (Ile 22P -Arg 25P ) run in extended conformation and in opposite orientation to a substrate along the cleft between subsites S 2 and S 5 (Fig.  2). They bind through two intermain-chain bonds to strand ␤6 above the cleft (Asp 21P O-Ser 66M N and Ser 66M O-Lys 23P N; see Table  2). This reverse orientation of the prosegment in the cleft may contribute to the prevention of selfcleavage and was previously described in cysteine protease and MMP zymogens (2). The polypeptide chain reaches the molecular surface after Arg 25P , whose side chain is anchored via three interactions with the mature enzyme backbone (Table 2). Thereafter, a flexible stretch from Ala 26P to Gln 29P leads to a short ␤-hairpin structure made up by strands ␤1 (Ala 31P -Val 33P ) and ␤2 (Ala 2M -Leu 4M ; see Fig. 2B). The tip of the hairpin contains the main maturation cleavage point (Gly 34P -Ala 1M ), which is completely buried within the molecular structure.
A Novel Two-step Mechanism of Activation-Based on biochemical studies, proastacin activation in the crayfish is understood to be a twostep process entailing removal and degradation of the prosegment (32). In addition, maturation involves conformational changes in the first six residues of mature astacin, Ala 1M -Asp 6M , and major rearrangement of segment Asp 129M -Gly 138M (Fig. 3, A and B), which is flexible in the zymogen and is hereafter referred to as the "activation domain." This is analogous to serine proteinases, for which chain segments grouped under this term show statistical or thermal disorder in the zymogens, partially obstruct the substrate-binding cleft, and lead to incompetent oxyanion holes. In these enzymes, the activation domain becomes rigid and functional upon activation (2,52).
In the first activation step, crayfish trypsin would perform initial cleavages and render a series of premature astacin species (32). This step would be facilitated by the flexibility of Ala 26P -Gln 29P within the final stretch of the prosegment. The last tryptic cut was shown to occur at Arg 32P -Val 33P , and this would probably disrupt a salt bridge, Arg 32P -Glu 19P , which is essential for structural integrity of the prosegment (see above). The arginine residue is not accessible on the molecular surface of the zymogen, but either the previous tryptic cleavages or the flexible activation domain beneath could provide the space and motion for the minor conformational changes required to enable access by trypsin (Fig. 2E). Subsequent removal of most of the prosegment would offer sufficient space for the activation domain to adopt its competent conformation by undergoing a hitherto unseen flap motion of up to 10 Å (measured at Tyr 133M C␣) (Fig. 3B). In addition, the activation domain would become rigid and well defined, as denoted by the electron density and average B-factors in the mature astacin structure (data not shown; see Refs. 39 and 40), in this way contributing to a functional substrate-binding cleft and active site (Fig. 3C). The initial multiple tryptic cleavages are reminiscent of MMP activation, during which trimming clips by other proteinases in a so-called "bait region" are observed (53). This contrasts with funnelins and trypsin-like serine proteinases, in which the first cut during activation generates the mature N terminus (2,54).
In the second activation step, the premature astacin variants, which are catalytically active (32), would produce subsequent cleavages, eventually giving rise to the competent N terminus at Ala 1M . The final autolytic cleavage at Gly 34P -Ala 1M , which matches the substrate specificity of astacin (see above), was unambiguously shown to occur in cis (32). Although Ala 1M N is about 12 Å away from the catalytic zinc in the proenzyme structure, the structural rearrangement of the activation domain triggered by the previous tryptic cleavages could easily lead the new provisional N-terminal segment, Arg 32P -Asp 6M , which is too long to adopt the definitive structure (see below), to protrude from the molecular surface and become accessible for intramolecular cleavage. After the second set of cleavages, segment Ala 1M -Asp 6M would again insert into the molecular moiety. In MMPs, similar trimming yields the competent N terminus, needed to form a salt bridge with a conserved aspartate of the final C-terminal helix (53,55).
The second cleavage step is indispensable because the new N terminus is buried within the mature enzyme body until Ile 3M in a cavity framed by segments Asp 131M -Gln 142M , Phe 100M -Met 107M , and Thr 185M -Gln 189M (Protein Data Bank code 1AST) (39,40). Residues from these segments are involved, together with seven solvent molecules and the first N-terminal residues, in an intricate, completely buried hydrogen-bonding network that is key for structural integrity of the enzyme (see Fig. 3 of Ref. 39) and that is incompatible with N-terminally elongated polypeptide chains. In particular, the ␣-amino group of Ala 1M is coordinated by three solvent molecules, which are further bound to Asp 131M O␦2 and His 102M O; to Tyr 101M O; and to Gln 189M O⑀1, Arg 106M N1, and Glu 103M O⑀1, respectively. The last interaction is of particular importance because Glu 103M is the immediately downstream neighbor of the third zinc-binding histidine. It is unique for and invariant within the astacin family (28). Comparison of zymogen and mature enzyme further reveals that upon cleavage at Gly 34P -Ala 1M , the main chain must undergo a 180°rotation around the ⌿ mainchain angle of the new N-terminal residue to establish the aforementioned solvent-mediated interactions (Fig. 3B). This is again reminiscent of serine proteinases, in which a salt bridge FIGURE 2. Structure of proastacin. A, representative detail in stereo view of the initial omit F obs -electron density map (contoured at 1) obtained after density improvement with ARP/wARP superposed with the final refined model from Ala 7P to Asn 14P of the prosegment. B, Richardson-type plot of proastacin depicting the prosegment in orange and the mature enzyme moiety in cyan. The orientation corresponds to the standard orientation characteristic for MPs (i.e. with the view into the active-site cleft, which runs from the left (non-primed side) to the right (primed side) according to Abramowitz et al. (33)). Repetitive secondary structure elements of the structure (see also Fig. 1) are shown as ribbons (helices ␣1-␣4) and arrows (strands ␤1-␤9), the catalytic zinc ion is shown as a magenta sphere, and the two disulfide bonds are shown as yellow sticks. The latter are labeled, as are the termini of the molecule. The final activation cleavage site (Gly 34P -Ala 1M ) is pinpointed by scissors. For clarity, only one conformation has been displayed for segment Asp 129M -Pro 135M . C, close-up view of B depicting the catalytic zinc ion with its six ligands, which are labeled. The respective bonding distances (in Å) are shown below each residue label. The Met-turn methionine is also shown and labeled. D, detail in stereo view of segment Phe 128M -Tyr 136M , which was modeled in two alternative occupancies (magenta and green), superimposed with the final (2mF obs Ϫ DF calc )-type electron density map contoured at 0.75 . E, close-up view of B in stereo view to illustrate the major interactions between the prosegment and the mature enzyme moiety. Participating residues are labeled, except those already labeled in B and C. Relevant solvent molecules are displayed as green spheres.

TABLE 2
Direct interactions between the pro-segment and the catalytic moiety between the newly formed N terminus and the aspartate residue upstream of the catalytic serine renders a functional enzyme (52). In contrast to the latter, however, maintenance of the hydrogen-bonding network of the new N terminus seems to be mainly required for structure and stability in astacins. Mutants of proastacin, in which Glu 103M had been replaced with glutamine and alanine, evinced equivalent catalytic efficiency but much lower thermal stability (32). This structural rather than functional importance is supported by the position and conformation of Glu 103M being actually superimposable in the zymogen and the mature structures. A similar scenario is conceivable for N-terminally elongated pre forms of astacin, which would bear intrinsic activity to permit the last steps in maturation but would be much less stable than the mature end product. This is corroborated by the finding that N-terminally extended variants of the astacin family member meprin were enzymatically active but thermally labile (32,56).

CONCLUSIONS
Proastacin maturation entails stepwise degradation of the prosegment to ensure irreversibility and to prevent competitive inhibition of the mature enzyme. Once the enzyme is activated, it is kept in check in vivo by protein inhibitors of the requisite specificity, such as ␣ 2 -macroglobulin (57). Previously structurally documented MP activation processes, as seen in MMPs, revealed preformed protease moieties, which merely had to be uncovered by removal of the prosegment. By contrast, activation is accompanied by major rearrangement in astacin and probably in the other members of the astacin family. Such rearrangement affects the activation domain and the new N-terminal segment, which becomes completely buried and salt-bridged to active-site residues. These interactions and a precisely trimmed N terminus are pivotal for the structural integrity and stability of astacins (31,32).
Despite having distinct chain traces and sequences, proastacin activation shares the following properties with that of MMP zymogens: (i) the prosegments run in opposite direction to a substrate across the cleft; (ii) the catalytic zinc-bound solvent molecule is replaced by a sidechain atom of a conserved residue from the prosegment; (iii) FIGURE 3. Proposed mechanism of activation. A, superimposition in stereo view of the C␣-traces of proastacin (prosegment in orange, catalytic moiety in cyan) and mature astacin (purple) in standard orientation. The catalytic zinc ion of proastacin is shown as a magenta sphere for reference. Segments involved in activation (magenta arrow) are indicated. Only one conformation has been displayed for segment Asp 129M -Pro 135M . B, close-up view of A in stereo view showing only the first residues of the mature moiety and the activation domain of both the proenzyme (cyan sticks, blue labels) and the mature enzyme (purple sticks and labels) as well as the four last proenzyme residues (orange sticks and labels). C, schematic diagram illustrating the transition between the zymogen, with a flexible activation domain in the lower subdomain of the molecule, and the rigid mature enzyme with the buried mature N terminus. the new N terminus establishes a salt bridge with a residue from the mature enzyme; and (iv) activation entails initial cleavage by other proteinases, although the last step is autolytic. As for other MPs, no comparable accumulating similarities are encountered with the zymogens and activation processes of funnelins, lysostaphins, and thermolysins. With trypsin-like serine proteinases, proastacin activation shares (i) an internal salt bridge of the new N terminus with an acidic residue directly adjacent to a main catalytic residue and (ii) an activation domain that undergoes a disorder-order transition to render a functional substrate-binding cleft. Accordingly, the present novel activation mechanism contains elements from two distinct classes of peptidases and contributes to a better understanding of enzyme maturation processes.