Activation Mechanism of Meprins, Members of the Astacin Metalloendopeptidase Family*

Meprins are mammalian zinc metalloendopeptidases with protease domains structurally related to astacin, the prototype of the “astacin family” of metalloproteases. Mature, active astacins are produced by proteolytic removal of an activation peptide to generate a new NH2-terminal residue. Structural studies indicate that the NH2-terminal ammonium group inserts into a water-filled cavity adjacent to the active site to form a salt bridge with a Glu residue that is conserved in all astacins. A similar interaction is known to play a crucial role in the activation of trypsin, resulting in the hypothesis that this salt bridge is required for the activation of astacin-like proteases. In this study, we have used the mouse meprin α subunit as a model to test this hypothesis of zymogen activation of the astacins. Mutants were generated to vary the NH2-terminal residue of the mature meprin α subunit (Asn78) and its putative salt bridge partner (Glu178). In addition, mutants creating NH2-terminal extensions and truncations were expressed in human embryonic kidney 293 cells. The recombinant proteins were activated by limited protease digestion and assayed for enzymatic activity and thermal stability. Point mutations of Asn78resulted in enzymes with activity comparable to the wild-type enzyme, indicating that the structure of this side chain is not essential for activity. NH2-terminal extension mutants of meprin α retained partial activity, with greater decreases against peptide relative to protein substrates. A mutant with a deletion of Asn78 to disrupt salt bridge formation with Glu178 had full activity, indicating that the putative salt bridge with Glu178 is not essential for enzyme activity. However, all changes in meprin α subunit NH2-terminal structure were found to decrease the thermal stability of the enzyme. These observations and additional data indicate that the zymogen activation mechanism of meprin and other astacins differs from that of the trypsin family of enzymes, and has some features in common with matrixins. It is proposed that prosequence removal of astacins allows the formation of hydrogen bonds involving the two NH2-terminal residues that are critical for enzyme structure.

Meprins are oligomeric zinc metalloendopeptidases composed of ␣ and/or ␤ subunits that are evolutionarily related, but differ in function (1)(2)(3). Meprin ␤ subunits are integral mem-brane proteins; ␣ subunits are secreted from cells unless associated with ␤ subunits (3,4). Meprin A (EC 3.4.24.18) can exist as secreted ␣ homooligomers (5) or as membrane-bound ␣/␤ heterooligomers (3,4). Meprin B (EC 3.4.24.63) is an ectoenzyme composed exclusively of membrane-bound ␤ subunits (1,2). Meprins are highly expressed in the microvillar membranes of the mammalian kidney and intestine (1,6,7) and are especially abundant in rodent brush border membranes (8). Meprins A and B participate in the processing and degradation of peptides and proteins at the cell surface (1). Meprins may also be involved in the remodeling of the extracellular matrix in response to renal injury (9). It has been shown that meprin B has an activity identical to that of the kinase splitting membranal protease, an enzyme that inactivates protein kinase A by a cleavage in its catalytic domain (10).
The isolation and nucleotide sequence analysis of meprin ␣ and ␤ subunit cDNAs revealed that the two subunits are 42% identical in amino acid sequence and share a similar arrangement of structural domains, including a protease domain related to astacin, a crayfish digestive protease (1,11,12). The "astacin family" of metalloendopeptidases comprises a group of proteins involved in a wide range of biological functions in organisms as diverse as the sea urchin, Drosophila, and mammals (1). Examples include Tolloid, a protein involved in dorsal/ ventral patterning in Drosophila (13), and bone morphogenetic protein-1 (procollagen C-proteinase), an enzyme required for the post-translational processing of procollagen types I-III (14). Other astacin-like enzymes ("astacins") have been identified by the cloning of their cDNAs, but have yet to be characterized at the protein level (1).
Astacins are initially synthesized as inactive precursors with an NH 2 -terminal prosequence (1,15). NH 2 -terminal sequence analyses of mature astacin, meprin A, and the procollagen C-proteinase indicate that the active forms of these enzymes have undergone a processing event that removes the prosequence (11,14,15). The NH 2 -terminal residue of mature astacins is a small, uncharged amino acid, usually Ala or Asn (1). The three-dimensional structure of crayfish astacin reveals that the NH 2 -terminal residue of the mature enzyme resides in a water-filled cavity adjacent to the zinc binding site, where its ammonium group forms a water-mediated salt bridge with a Glu side chain that is conserved in all astacins (15). A molecular modeling study predicted that this arrangement is also possible for the meprin ␣ subunit catalytic domain (16), but the charged groups are bridged by two water molecules (Fig. 1). A similar interaction has been observed to be required for the activation of trypsin (17), leading to the prediction that formation of this salt bridge is essential for the activation of astacins (15,16). To test this hypothesis, we have expressed and characterized a series of mutants of the mouse meprin ␣ subunit. Point mutations have been introduced to change the side chain of Asn 78 , the NH 2 -terminal residue of mature meprin, and Glu 178 , the two residues implicated in salt bridge formation. Mutagenesis has also been used to extend and truncate the mature ␣ subunit NH 2 terminus to disrupt the interaction of the terminal ammonium group with the conserved Glu 178 . Recombinant proteins were expressed and tested for activity and thermal stability.

EXPERIMENTAL PROCEDURES
Cell Culture and Expression of Recombinant Proteins-Human embryonic kidney 293 cells (American Type Culture Collection CRL 1573) were maintained in Dulbecco's modified Eagle's medium (DMEM) 1 supplemented with 10% fetal bovine serum (HyClone Laboratories, Inc., Logan, UT), 50 units/ml penicillin, and 50 g/ml streptomycin ("complete DMEM"). Cells were incubated at 37°C in an atmosphere containing 5% CO 2 . Unless otherwise noted, all media and cell culture reagents were purchased from Life Technologies, Inc. For the expression of meprin ␣ subunits, wild-type and mutant cDNAs were cloned into the expression plasmid pCB3 (a generous gift of Dr. Colleen Brewer, University of Texas Southwestern Medical Center, Dallas, TX). pCB3 is a derivative of pCMV-1 (18), with transcription driven by the promoter of a human cytomegalovirus immediate early gene. Plasmids used for transfection experiments were purified using reagents and columns purchased from Qiagen, Inc. (Chatsworth, CA). Expression plasmids were transfected into 293 cells by a calcium phosphate precipitation procedure (19) and were cotransfected with the plasmid pVA1 (20) to enhance mRNA translation in the adenovirus-transformed 293 cells. After transfection, cells were grown for 24 h in complete DMEM, then switched to serum-free OptiMEM I (Life Technologies, Inc.), and grown for an additional 36 -48 h before the medium was collected for analysis of secreted meprin ␣ subunits.
Analysis of Astacin and Meprin Three-dimensional Structures-The structural coordinates of astacin and a molecular model of the meprin ␣ subunit protease domain were obtained from the Brookaven Protein Data Bank. The files for astacin (1AST 5) and meprin (1IAF 8) were imaged and manipulated using InsightII™ software (Biosym/MSI, San Diego, CA) on a Silicon Graphics (Mountain View, CA) workstation.
Analysis of Recombinant Meprin ␣ Subunits-For each recombinant protein preparation, serum-free conditioned medium from four 100-mm plates was concentrated 40-fold using Centriprep 30 cartridges (Ami-con, Inc., Beverly, MA); diluted 10-fold in 10 mM Tris-HCl, 0.15 M NaCl, pH 8.0; and reconcentrated. Protein concentrations were measured using a bicinchoninic acid assay (Pierce) and found to be 0.8 -1.0 mg/ml in the 40-fold concentrated media samples. Meprin ␣ subunit zymogens were activated by a limited digestion with either L-1-tosylamido-2phenylethyl chloromethyl ketone-treated trypsin (Sigma) or endoproteinase Arg-C (Boehringer Mannheim). For trypsin activation, proteins were diluted to a concentration of 0.5 mg/ml in 10 mM Tris-HCl, 0.15 M NaCl, pH 8.0; trypsin was added to 10 g/ml, and the mixture was incubated at 24°C for 30 min. Trypsin was then inhibited by the addition of soybean trypsin inhibitor (Sigma) to 40 g/ml. Conditions for activation by endoproteinase Arg-C were as described previously (21), except that 0.1 mg/ml endoproteinase Arg-C was included in the reaction, and incubation was at 37°C for 30 min. After the incubation, endoproteinase Arg-C was inhibited by the addition of 3,4-dichloroisocoumarin (Sigma) to 200 M.
Meprin protease activity was measured using the substrate azocasein (Sigma) as described previously (22). Peptidase activity was measured using the fluorogenic bradykinin analog BKϩ (21). Assays were performed in the buffer 10 mM Tris-HCl, 0.15 M NaCl, pH 8.0, at 24°C in a total volume of 300 l, using 10 M BKϩ. Product fluorescence was measured with a Hitachi F2000 fluorescence spectrophotometer using an excitation wavelength of 320 nm and an emission wavelength of 417 nm. Arylamidase activity of meprin was determined using the substrate glutaryl-Ala-Ala-Phe-4-methoxy-2-naphthylamide (glut-AAF-MNA, purchased from Sigma). Assays were as for BKϩ, except that reactions contained 50 M glut-AAF-MNA, and the respective excitation and emission wavelengths were 340 nm and 425 nm. Estimates of k cat /K m for BKϩ and glut-AAF-MNA were calculated using the equation: where V o is the initial rate observed under the assay conditions described above, in which [S]Ͻ ϽK m (23).
[E] was calculated using a meprin ␣ subunit molecular mass of 8.8 ϫ 10 4 g/mol. The specific activities reported are averages of triplicate assays of at least two independent protein preparations per recombinant protein.
Apparent K i values for the inhibition of meprin hydrolysis of BKϩ by actinonin were determined as follows. The concentration of actinonin was varied from 0.05 M to 0.5 M in the presence of constant BKϩ (10 M) and constant amounts of meprin. Data were plotted as the inverse of initial rate versus actinonin concentration and fit to straight lines using a computer program. The x axis intercepts of the resulting lines were taken to be ϪK i .
Immunoblot Analysis of Recombinant Proteins-Samples of conditioned media from 293 cells were subjected to SDS-PAGE (24) using 7.5% polyacrylamide gels after reduction of samples with 2-mercaptoethanol. The proteins were then electrophoretically transferred to polyvinylidene difluoride (PVDF) membranes (Fisher). PVDF membranes were successively incubated with an antibody raised against rat meprin A, horseradish peroxidase-conjugated donkey anti-rabbit IgG, and reagents for enhanced chemiluminescence (Pierce). The membranes were then exposed to x-ray film (NEN Life Science Products). Quantitation of recombinant meprin ␣ subunits was aided by densitometry of films from immunoblots with known amounts of meprin A (purified from mouse kidney) electrophoresed alongside the recombinant samples. Gel scanning and densitometry software was from PDI, Inc. (Huntington Station, NY).
Amino-terminal Sequence Analysis-Concentrated conditioned media samples containing the proteins N78del, N78,79del, N78NN, N78NAN, or N78NAAN were activated by trypsin as described above. Each protein was then purified by chromatography on a MonoQ column (Pharmacia, Uppsala, Sweden) followed by preparative SDS-PAGE and transfer to PVDF membranes. The membranes were stained with Amido Black, and the meprin bands were excised and delivered to the Macromolecular Core Facility of this institution for six cycles of automated Edman degradation.
Mutagenesis-All mutagenesis of the mouse meprin ␣ subunit cDNA (11) was done using polymerase chain reaction (PCR) with mutagenic oligonucleotide primers and Taq DNA polymerase (Fisher). Every PCR reaction made use of the antisense primer GGAGACTCCTGGCCTTG-CAGAAGG (complementary to residues 1905-1928 of the mouse meprin ␣ cDNA) and the appropriate mutagenic primer. Before PCR products were subcloned into the meprin ␣ subunit cDNA, they were screened for unintended errors and verified as correct by DNA sequence analysis using dideoxy sequencing reagents purchased from Amersham Life Science. DNA restriction and modification enzymes were purchased from Fisher. Prior to construction of mutations at Asn 78 , a mutation generating an AatII restriction site (GACGTC) upstream of this region was made to facilitate the cloning of other DNA fragments. This mutation converted Ile 71 of the prosequence to Val (ATC to GTC),  (16)). Structures were imaged using the In-sightII program on a Silicon Graphics workstation. Hydrogen bonds (dashed lines) were calculated and the distances (in Å) measured using the "H-bond" and distance measuring functions of InsightII. For astacin (left), the charged groups of Ala 1 , Glu 103 , and Arg 106 are able to participate in salt bridges mediated by a single structural water molecule (Wat 501 ). For meprin ␣ (right), Glu 178 and Arg 181 potentially form a salt bridge mediated by Wat 505 . Meprin's Asn 78 ammonium group can be linked to Glu 178 indirectly through two water molecules (Wat 501 and Wat 505 ).
and was found to have no effect on the expression level or specific activities of the meprin ␣ subunit. The I71V mutant (having no changes in the mature form of the enzyme) is therefore considered to be equivalent to "wild-type" and was utilized for the cloning and expression of all Asn 78 mutants.
The mutagenic primer "N78Z": (CGCGTCGACGTCCTCCTACCACG-GACACGG(G/A)NTGCCATGCGAGATCCCTCAAGC) was used in a PCR reaction to generate point mutations at Asn 78 . The primer is 8-fold degenerate at codon 78 ((G/A)NT) to produce a number of mutants; Ala, Ser, Thr, and Ile mutations were subcloned and used for the expression studies.
Extensions of the NH 2 terminus of the mature meprin ␣ subunit were generated by primers of the design: CGCGTCGACGTCCTCCTAC-CACGGACACGG(X)AATGCCATGCGAGATCCCTCAAGC, where (X) represents codons inserted between Arg 77 , the trypsin cleavage point, and Asn 78 , the NH 2 terminus of wild-type meprin ␣ subunits. For the primer "N78NN," (X) is AAT; for primer "N78NAN," (X) is AACGCC; for "N78NAAN," (X) is AACGCAGCC. These primers generate NH 2 -terminal extensions of Asn, Asn-Ala, and Asn-Ala-Ala, respectively. All DNA fragments with mutations at codon Asn 78 were subcloned into the meprin ␣ cDNA (containing the I71V mutation) using AatII and SphI restriction sites.
Mutations at Glu 178 of the meprin ␣ subunit were generated in a PCR reaction using the mutagenic primer "E178X": GCCATGCATGCTCT-GGGATTCTTCCAT(G/C)(A/C)(G/C)CAGTCAAGGACTGACCGGGAT, which is 8-fold degenerate at codon 178. Products of the resulting PCR reaction contained the mutations E178A, E178Q, and E178D. The products were subcloned into the meprin ␣ cDNA using SphI and NdeI restriction sites.

RESULTS
Asn 78 Point Mutations Decrease the Stability but Not the Activity of the Meprin ␣ Subunit-The wild-type meprin ␣ subunit and the mutant N78S, N78T, and N78I cDNAs were transfected into 293 cells, and the conditioned culture medium was harvested for analysis of secreted proteins. As shown by the immunoblot analysis in Fig. 2, the wild-type and mutant proteins were secreted as 94-kDa zymogens. Medium harvested from 293 control cells (transfected with the pCB3 vector containing no cDNA insert) contained no immunoreactive proteins. Zymogen activation of meprin subunits was achieved by limited trypsin digestion to remove the prosequences by cleavage at Arg 77 (3,4), generating the mature 88-kDa enzymes. The mutants appear to be as stable as the wild-type enzyme to the trypsin treatment, as no degradation other than prosequence removal was detected. The meprin ␣ subunit and mutants have similar activities against protein, peptide, and arylamide substrates (Table I). These activities were assayed using the substrates azocasein, BKϩ, and glut-AAF-MNA, respectively, as described under "Experimental Procedures." It is evident that the Asn 78 point mutants do not differ significantly from the wild-type enzyme in their ability to degrade azocasein. The mutants also have comparable activities against the peptide substrate BKϩ and the arylamide substrate glut-AAF-MNA, although it was observed that the specific activity of the N78T mutant was only 50 -65% that of the wild-type ␣ subunit for these substrates. This difference was observed consistently in three separate preparations of the N78T mutant. However, substitution of the small, polar Asn side chain with the bulky, hydrophobic side chain of Ile had no deleterious effect on enzymatic activity. Therefore, the structure of the side chain of the meprin ␣ subunit NH 2 -terminal residue has relatively little influence on the mature enzyme as judged by measurements of enzymatic activity.
The effects of the Asn 78 point mutations were further examined by investigating the thermal stability of the mutants (Fig.  3). The activated wild-type and mutant proteins were incubated for varying periods of time at 53°C, then assayed for activity using the BKϩ substrate. It was observed that the wild-type enzyme is the most thermostable, retaining 84% of its initial activity after a 30-min incubation at 53°C. The Asn 78 mutants are not as stable at this temperature, and stability appears to decrease with increasing bulk of the mutant side chain. The t 1/2 values for inactivation calculated from the data are: 116 min for wild-type ␣, 54 min for the N78S mutant, 36 min for the N78T mutant, and 16 min for the N78I mutant. It is apparent that mutations at Asn 78 having little effect on enzyme activity decrease the stability of the mature meprin ␣ subunit.
Meprin ␣ Glu 178 Mutants Have Little or No Protease Activity-Mouse meprin ␣ residue Glu 178 is analogous to Glu 103 of astacin, residing in an internal water-filled cavity (15,16). This  Prior to assays, meprin zymogens were activated by limited trypsin digestion as described under "Experimental Procedures." One unit of azocasein activity is defined as the degradation of 1.1 g of azocasein/min. BKϩ and glut-AAF-MNA activities are expressed as both specific activity (mol/min ⅐ mg) and as the ratio k cat /K M (M Ϫ1 s Ϫ1 ). BKϩ assays contained 10 M substrate; glut-AAF-MNA assays contained 50 M substrate. The values are averages of triplicate assays obtained from at least two independent protein preparations. The amount of meprin included in assays was determined by densitometry as described under "Experimental Procedures." WT, wild-type. residue is conserved in all members of the astacin family, and has been proposed to participate in the water-mediated salt bridge with the NH 2 -terminal ammonium group predicted to be required for zymogen activation. We therefore generated mutants of Glu 178 in the meprin ␣ subunit and examined the properties of the recombinant enzymes to test the effects on activity and stability. Fig. 4 shows an immunoblot analysis of proteins secreted into the conditioned medium of 293 cells transfected with the mutants E178D and E178Q. Unlike the wild-type ␣ subunit, the Glu 178 mutants were degraded by the limited trypsin digestion generally used for zymogen activation (10 g of trypsin/ml), indicating structural changes in these mutant proteins relative to the wild-type ␣ subunit. Virtually no mature (88 kDa) enzyme was seen for the E178D mutant; degradation products are visible in the 50 -70 kDa range. Similar degradation products were observed for the E178Q mutant, although some of the 88-kDa form of the enzyme is evident. When meprin ␣ subunits were incubated with endoproteinase Arg-C, mature enzyme forms of 88 kDa for the Glu 178 mutants were observed with no evidence of extensive degradation (Fig.  4). The E178A mutant exhibited sensitivity to protease treatment similar to the E178D mutant (data not shown). After limited digestion of the wild-type and E178 mutant zymogens by endoproteinase Arg-C, two of the three Glu 178 mutants (E178A and E178D) were found to have no detectable protease or peptidase activity (Table II). E178Q was only partially active, with about 10% of wild-type activity against both azocasein and BKϩ. The inactivity of the E178D mutant is noteworthy, because the negative charge of the side chain is conserved. However, the position of this charge is predicted to be moved about 1 Å relative to that of wild-type Glu due to the shorter side chain, indicating that the precise positioning of this negative charge is critical for enzyme structure. The partially active mutant E178Q maintains the general structure of the side chain with a polar amide group, but its low activity confirms the importance of the Glu charge. NH 2 -terminal Truncations and Extensions Decrease the Thermal Stability, but Do Not Abolish the Activity of the Meprin ␣ Subunit-The Asn 78 point mutations would not necessarily disrupt the formation of a salt bridge between the terminal ammonium group and Glu 178 , because it is likely that the relatively large water cavity is able to accommodate the mutant side chains. Therefore, a series of mutants were prepared to create extensions and truncations of the mature meprin ␣ NH 2 terminus after activation by limited trypsin digestion. It is expected that these mutants will change the position of the NH 2 -terminal ammonium group relative to Glu 178 in the mature enzyme, thereby preventing formation of the salt bridge proposed to be required for zymogen activation. The NH 2 termini expected for the mature enzymes after trypsin activation are diagrammed in Fig. 5 (top). NH 2 -terminal extensions of 1, 2, and 3 residues were generated by N78NN, N78NAN, and N78NAAN, respectively. All three mutations retain Asn as the NH 2 -terminal residue, with Ala spacers making up the remainder of the extensions for N78NAN and N78NAAN. The mutants N78del and N78,79del delete residue Asn 78 and the dipeptide Asn 78 -Ala 79 , respectively, generating NH 2 termini of Ala 79 and Met 80 in the mature enzymes after limited trypsin digestion. The NH 2 -terminal structures of all five activated truncation and extension mutants were determined by NH 2terminal sequence analysis of the purified recombinant enzymes (data not shown). The expected NH 2 termini were observed for all mutants, but species resulting from trypsin cleavage at Arg 81 were detected for the three extension mutants, accounting for 10% (N78NN) to 25% (N78NAN and (N78NAAN) of the activated proteins. The immunoblot of Fig.  5 (bottom) shows the truncation and extension mutants after trypsin activation. All mutants were expressed at levels comparable to the wild-type enzyme, were converted to the mature enzyme mass of 88 kDa, and exhibited no evidence of extensive degradation by the trypsin treatment.
The activated NH 2 -terminal extension and truncation mutants were assayed for activity with the substrates azocasein, BKϩ, and glut-AAF-MNA (Table III). Of this group of mutants, none were found to be completely inactive, but only N78del had

TABLE II Enzymatic activities of meprin ␣ subunits mutated at Glu 178
Proteins secreted into the conditioned medium of 293 cells transfected with wild-type and mutant cDNAs were activated by endoproteinase Arg-C digestion and assayed as described under "Experimental Procedures." One unit of azocasein activity is defined as the degradation of 1.1 g of azocasein/min. BKϩ activity is expressed as both specific activity (mol/min ⅐ mg) and the ratio k cat /K M (M Ϫ1 s Ϫ1 ). ND, no activity was detected. WT, wild-type. activity comparable to wild-type for all three substrates tested. N78,79del and all three extension mutants retained a substantial amount of azocasein degrading activity, with specific activities ranging from 36% to 46% of the wild-type value. This group of four mutants displayed an even greater decrease in activity against the peptide substrates, with specific activity values only 10 -17% of wild-type, depending on the substrate and the mutant considered. The greater loss of activity against peptide compared with the protein substrate is interesting, indicating that the two activities are affected to different degrees by changes in enzyme structure. The NH 2 -terminal truncation and extension mutants were also tested for their thermal stability. Initial experiments showed that these mutants are highly unstable at 53°C, so the experiments were conducted at 49°C to obtain inactivation rates. After incubation at 49°C for various times, the samples were assayed using BKϩ as the substrate. The wild-type enzyme is extremely stable at 49°C, having a calculated t 1/2 for inactivation of 1030 min. As shown in Fig. 6, the N78del mutant is considerably less stable, with a t 1/2 of 15 min. The other mutants exhibit even lower stability at 49°C, with t 1/2 values of approximately 3 min. These values are: 3.5 min for N78NN, 2.6 min for N78NAN, 4.3 min for N78NAAN, and 3.0 min for N78,79del. Therefore, this group of four mutants was found to have nearly identical properties when both catalytic activity and thermal stability are considered. Comparison of the mutant proteins before and after heat treatment by immunoblot analysis confirmed that the observed losses of activity were due to denaturation, not to autolysis or degradation by other proteases present in the samples.
The above deletion and truncation mutants were also compared for their sensitivity to inhibition of BKϩ hydrolysis by actinonin, a peptide hydroxamate. Apparent K i values for actinonin of 0.10 M and 0.13 M were obtained for wild-type ␣ and the N78del mutant, respectively; higher K i values of 0.29, 0.33, 0.25, and 0.27 M were obtained for N78,79del, N78NN, N78NAN, and N78NAAN. The similarities in enzymatic activity, thermal inactivation rates, and inhibition constants for the latter group of four mutants is an indication that they have all undergone a similar change in structure, even though they possess different mutations.

DISCUSSION
The results herein for meprin are inconsistent with a trypsin-like mechanism of zymogen activation, because neither the structure of the NH 2 -terminal side chain nor conservation of a salt bridge involving this residue were found to be essential for enzyme activity. The activation of trypsinogen is well characterized by both structural and mutagenesis studies. A region termed the "activation domain," including the S1 binding site and the oxyanion hole, is disordered in trypsinogen, but completely formed in mature, active trypsin (17,25). The crucial conformational change is triggered by cleavage of the zymogen activation peptide, creating a new NH 2 -terminal residue (Ile16, chymotrypsin numbering system). The new terminal ammonium group forms a salt bridge with Asp 94 , stabilizing the protein by 3 kcal/mol (17,26). A recent study has shown that the contribution of the Ile side chain to stability is critical, accounting for 5 kcal/mol of stabilization energy to the activation domain; trypsin mutants I16A and I16G have greatly decreased rates of acylation in amide bond hydrolysis (26). This contrasts with the full activity seen for the meprin ␣ Asn 78 point mutants. Thus, although there are several structural features of mature astacins that resemble the NH 2 -terminal structure of trypsin, the crucial features for activation differ.
Point mutations of meprin Glu 178 were found to have marked effects on enzyme structure. The mutants E178A, E178D, and E178Q were all substantially degraded by the limited trypsin digestion generally used to activate meprin zymogens. Of the mutants activated by digestion with endoproteinase Arg-C, only E178Q had detectable activity, 10% that of the wild-type enzyme. Although these mutants would be expected to disrupt a salt bridge between residue 178 and the NH 2 terminus, interpretation of these results is difficult due to the fact that Glu 178 (Glu 103 in astacin) is also positioned to form a salt bridge with the conserved residue Arg 181 (Arg 106 in astacin). Glu 178 of the meprin ␣ subunit and its counterpart in other astacins are  directly adjacent to a His residue involved in the coordination of the catalytic zinc ion. Minor disruptions in structure at this point would be expected to have effects on active site structure and catalysis. This and the potential involvement of Glu 178 in several ionic interactions may explain its absolute conservation in all members of the astacin family. It is not possible to determine from these experiments exactly which of these interactions are disrupted by the Glu 178 mutations. The NH 2terminal extension and truncation mutants were designed to more directly assess the importance of the NH 2 -terminal salt bridge for zymogen activation. The astacin NH 2 terminus is inserted into an internal cavity containing eight ordered water molecules. The charged groups of astacin residues Glu 103 and Arg 106 participate in salt bridges mediated by some of the ordered water molecules (15,16). The residues Ala 1 and Ala 2 are also completely inserted into this cavity. In the meprin ␣ model described by Stöcker et al. (16), this cavity is of comparable size, with the ammonium group of Asn 78 located 4.88 Å and 5.32 Å from the oxygen atoms of the Glu 178 side chain. The mutant N78NN would extend the meprin ␣ NH 2 terminus by about 3.5 Å. It is possible that the water cavity can accommodate the extra residue and allow a conformation that conserves the salt bridge, but this seems unlikely for the extensions N78NAN and N78NAAN, which would extend the NH 2 terminus by some 7-10 Å. Similar mutations of the plasmin (trypsin-like) protease domain apparently disrupted the crucial salt bridge, decreasing activity about 10 3 -fold (27). The relatively minor decreases in activity observed for these meprin mutants indicate that the proposed salt bridge is not critical for active site structure and activity. The results also demonstrate that processing of the enzyme's terminus to a precise residue is not absolutely required for catalytic activity, because the mutant proteins retain significant activity, particularly against the substrate azocasein.
The NH 2 -terminal truncation mutant N78del was found to have catalytic activity virtually identical to the wild-type enzyme. This mutation is expected to position the NH 2 terminus an additional 3.6 Å away from Glu 178 , making the formation of a salt bridge virtually impossible without gross rearrangements in enzyme structure. The analogous mutation of trypsin (deletion of Ile 16 ) decreased activity against amide substrates 10 7 -fold, 2 and similar truncations of the plasmin protease domain lowered activity to undetectable levels (27). Therefore, we conclude that the formation of a salt bridge between the NH 2 terminus and Glu 178 is not essential for the activation of meprin, and such an interaction is probably not required for the activation of other astacins. Despite activity comparable to wild-type meprin ␣, evidence for a structural change in the N78del mutant lies in its decreased thermal stability. The decreased stability of this protein could be explained by the loss of one or more of several interactions that could be contributed by the wild-type NH 2 -terminal residue. Besides the potential salt bridge with Glu 178 , the molecular model of meprin ␣ (16) shows that a hydrogen bond forms between the backbone carbonyl oxygen of the terminal residue (Asn 78 ) and the backbone amide group of residue Phe 176 (Fig. 7). This hydrogen bond is also present in the astacin structure (between Ala 1 and Tyr 101 ), possibly stabilizing the mature forms of both enzymes.
Astacins are part of a superfamily of structurally-related metalloproteinases known as the "metzincins" (28). Matrixins (matrix metalloproteinases or MMPs), reprolysins (e.g. snake venom metalloproteinases and tumor necrosis factor-␣-converting enzyme), and serralysins (bacterial enzymes) are also metzincins. The latency of proforms of matrixins and reprolysins is maintained by a "cysteine switch" mechanism (29 -31), in which a cysteine residue of the prosequence coordinates the active site zinc ion to prevent catalysis. Astacins do not possess a recognizable cysteine switch peptide; the meprin ␣ subunit prosequence contains no cysteine residues at all. Therefore, the 2 L. Hedstrom, personal communication.
FIG. 6. Thermal inactivation of meprin ␣ subunit NH 2 -terminal truncation and extension mutants. Trypsin-activated samples were incubated at 49°C for the indicated time periods before assay with the peptide substrate BKϩ (at 24°C) as described under "Experimental Procedures." Initial activity is defined as the activity measured in the absence of heat treatment. Each data point is the average of measurements obtained from two separate thermal inactivation experiments performed in parallel.

FIG. 7.
Hydrogen bonds involving the main chain of NH 2 -terminal residues of mature astacin and the meprin ␣ subunit. Structural coordinates are from the Brookaven Protein Data Bank files 1AST 5 (astacin) and 1IAF 8 (meprin ␣ subunit protease domain molecular model (16)). Structures were imaged using the InsightII program on a Silicon Graphics workstation. Hydrogen bonds (dashed lines) were calculated and the distances (in Å) measured using the "H-bond" and distance measuring functions of InsightII. The hydrogen bonds evident in the astacin structure between the backbone atoms of Ala 1 -Tyr 101 and Ala 1 -Arg 135 are also permitted between the analogous atoms of the meprin ␣ model. latency of astacins must be maintained by a different mechanism. Several activated forms of both MMP-1 (interstitial collagenase) and MMP-8 (neutrophil collagenase) have been observed, corresponding to NH 2 termini beginning with residues 79, 80, and 81 (32)(33)(34)(35). All of these enzyme forms are active, but enzymes beginning with residue 79 have specific activities 3-12-fold higher than the proteins having shortened NH 2 termini (34,35). Partial activity (15-20%) of collagenases and stromelysin retaining up to 13 residues of the prosequence has been observed (34,36). It is evident that both astacins and matrixins can exist as partially active NH 2 -terminally extended and shortened forms, and are fully active only when the NH 2 terminus is precisely processed. The structural basis of the above observations has been addressed by determination of the three-dimensional structures of neutrophil collagenase protease domains having NH 2 termini beginning with Phe 79 and Met 80 (37,38). No obvious differences were noted in the active sites of the two proteins, however it was noted that the NH 2terminal seven residues of the less active Met 80 form were disordered, whereas they were ordered in the Phe 79 protein (37). A similar loss of NH 2 -terminal structure could explain the decreased activity of the extended and truncated meprin mutants observed in this study.
It was observed that the mutant N78,79del, which lacks the first two residues of the mature ␣ subunit, displays phenotypes indistinguishable from the three NH 2 -terminal extension mutants. Nearly identical values were obtained for activity against protein, peptide and arylamide substrates, thermal stability, and sensitivity to inhibition by actinonin, raising the possibility that all four mutants have undergone the same structural change. Examination of the astacin three-dimensional structure and the meprin ␣ catalytic domain structural model reveals that for the second residue of the mature enzymes (Ala 2 of astacin, Ala 79 of meprin ␣) the backbone carbonyl oxygen and amide groups make two hydrogen bonds with the backbone atoms of Arg 135 (astacin) and Thr 210 (meprin ␣, Fig. 7). Thus, the two terminal residues of both enzymes contribute three hydrogen bonds linking the NH 2 terminus to distant regions of the polypeptide chain (residues 101 and 135 in the astacin numbering system). Because both terminal residues are deleted in the N78,79del mutant, these three hydrogen bonds cannot form. It is possible that localized structural perturbations in the NH 2 -terminal extension mutants eliminate these hydrogen bonds in the proteins, resulting in structures and phenotypes for all three extension mutants that are similar to the N78,79del mutant. If so, this observation could explain why astacins are processed to a precise residue in their mature forms, a processing event producing an enzyme with maximal activity and stability.