Marked differences between metalloproteases meprin A and B in substrate and peptide bond specificity.

Meprin A and B are highly regulated, secreted, and cell-surface metalloendopeptidases that are abundantly expressed in the kidney and intestine. Meprin oligomers consist of evolutionarily related alpha and/or beta subunits. The work herein was carried out to identify bioactive peptides and proteins that are susceptible to hydrolysis by mouse meprins and kinetically characterize the hydrolysis. Gastrin-releasing peptide fragment 14-27 and gastrin 17, regulatory molecules of the gastrointestinal tract, were found to be the best peptide substrates for meprin A and B, respectively. Peptide libraries and a variety of naturally occurring peptides revealed that the meprin beta subunit has a clear preference for acidic amino acids in the P1 and P1' sites of substrates. The meprin alpha subunit selected for small (e.g. serine, alanine) or hydrophobic (e.g. phenylalanine) residues in the P1 and P1' sites, and proline was the most preferred amino acid at the P2' position. Thus, although the meprin alpha and beta subunits share 55% amino acid identity within the protease domain and are normally localized at the same tissue cell surfaces, they have very different substrate and peptide bond specificities indicating different functions. Homology models of the mouse meprin alpha and beta protease domains, based on the astacin crystal structure, revealed active site differences that can account for the marked differences in substrate specificity of the two subunits.

Meprin A and B are zinc metalloendopeptidases composed of evolutionarily related ␣ and/or ␤ subunits. They are members of the astacin family and are highly expressed in brush border membranes of the intestine and renal proximal tubules (1,2). Meprins are particularly abundant in mouse juxtamedullary nephrons and constitute ϳ5% of total protein in renal brush border membranes (3). The meprin ␣ and ␤ subunits are expressed early in embryonic development of mouse kidney and intestine, by day 11, and have different patterns of expression in the suckling phase and after weaning (4). Homologous en-zymes are found in rat and human kidney and intestine (1,5,6). Meprins are also expressed in leukocytes of intestinal lamina propria and in cancer cells and are consequently implicated in inflammation and cancer growth and metastasis (7,8).
Mouse kidney meprin A (EC 3.4.24.18) is a homooligomer of ␣ subunits, or a heterooligomer of ␣ and ␤ subunits (2,9). Mouse kidney meprin B (EC 3.4.24.63) is a homooligomer of ␤ subunits (10). The multidomain ␣ and ␤ meprin subunits are highly glycosylated and form disulfide-linked dimers and higher order oligomers by noncovalent interactions (11,12). Meprins containing at least one ␤ subunit remain membranebound by virtue of a transmembrane domain located near the carboxyl terminus of ␤ subunits (1). Mature meprin ␣ homooligomers contain no transmembrane domain and are found in mouse urine (13). The expression of the meprin ␣ subunits in mice is strain-dependent (1). Random-bred mice (such as ICR) and many inbred strains of mice (e.g. C57BL/6) express both meprin ␣ and ␤ in the adult kidney, and these strains possess heterooligomeric forms of meprin A. Some inbred mouse strains (such as C3H/He) only express meprin ␤ in the adult kidney and therefore have only meprin B. Furthermore, meprin ␤ subunits in mouse kidney exist primarily in the proenzyme form, thus these subunits are latent (14). This is in contrast to the rat kidney enzyme, as well as meprins from mouse, rat, and human intestine where the propeptide is removed and the enzymes are fully active (1,15). Therefore, meprin A and B isolated from adult mouse kidney are novel in that they can be used to determine the activities of the ␣ and ␤ subunits, respectively.
Identification of substrates for proteases is a valuable step toward elucidation of physiological function and provides a knowledge-based approach to inhibitor design. Previous studies have shown that a variety of biologically active peptides and proteins are hydrolyzed by meprins in vitro. For example, meprins cleave blood pressure regulators such as bradykinin, metabolism mediators such as parathyroid hormone, signaling molecules such as protein kinase A, and basement membrane proteins such as entactin (1,2,16,17). There has been no systematic study, however, of the enzymatic differences between meprin A and B from any species or of the contributions of the individual subunits to activity. Enzymological data for meprin B are particularly lacking even though meprin B appears to be a more essential protease than meprin A as indicated by expression patterns in mammalian tissues. The work herein describes previously unidentified meprin substrates and provides the first detailed comparison of meprin ␣ and ␤ substrate and peptide bond specificity.

EXPERIMENTAL PROCEDURES
Isolation of Meprin-Mouse meprin A and B were isolated from kidney brush border membranes of ICR and C3H/He male mice, respectively. For kinetic analyses, meprin B was activated using trypsin as previously described (18). Concentrations of meprins are based on a subunit molecular mass of 90 kDa. Meprin A contains active ␣ subunits and inactive ␤ subunits. Trypsin-treated meprin B has only ␤ subunit activity.
Peptide Library Screen-An amino-terminally acetylated dodecamer peptide mixture (1 mM) consisting of a roughly equimolar mixture of the 19 naturally occurring L-amino acids, excluding cysteine, was incubated with either meprin A (33 nM) or B (10 nM) in 25 mM HEPES, pH 7.4, 100 mM NaCl, 5 mM CaCl 2 at 37°C until 5-10% of the peptides were digested. The mixture contained the inhibitors pepstatin, leupeptin, L-trans-epoxysuccinyl-leucylamide-(4-guanidino)-butane, 3,4-dichloroisocoumarin, and bestatin to prevent hydrolysis by trace amounts of protease impurities. Hydrolysis was terminated by heating to 100°C for 2 min, and 10 l of the mixture was subjected to Edman degradation-based aminoterminal peptide sequencing. The data in each sequencing cycle was normalized to the total molar amount of amino acids in that cycle so that a value of 1 indicates the average value. Undigested peptides and the amino-terminal fragments of digested peptides are amino-terminally blocked and therefore do not contribute to the sequenced pool. The relative amount of a given amino acid residue present in a particular sequencing cycle indicates the preference for that residue at a particular position relative to the cleavage site. Values for a given amino acid were corrected by renormalizing to the average relative amount of that residue present across the first nine sequencing cycles to correct for the distribution of amino acids in the starting mixture.
Identification of Novel Meprin Substrates-Naturally occurring peptides (100 M) were incubated with meprin A or B (2 nM) for 4 h. The incubation was carried out at 37°C in 20 mM Tris-HCl, 150 mM NaCl, pH 7.5, containing the same inhibitors as above. The reaction was terminated by the addition of EDTA to 10 mM and trifluoroacetic acid (TFA) to 0.05% v/v. Samples were then subjected to high performance liquid chromatography (HPLC) analysis using a POROS R2H (4.6 ϫ 100 mm) column and a Brownlee Newguard 7-m RP-18 (3.2 ϫ 15 mm) guard column. Peptides were eluted with a linear gradient of acetonitrile in 0.1% TFA at a constant flow rate of 3.2 ml min Ϫ1 . Peptides were monitored at A 220 . Percent hydrolysis was then calculated by monitoring the decrease in the substrate peak area, compared with time zero.
Kinetic Constant Determination-Kinetics for peptide hydrolysis by meprins were determined by quantitative HPLC analysis of reaction mixtures using the same buffer conditions as the peptide screen. Meprin concentrations in reactions were between 1 and 4 nM depending on efficiency of proteolysis. Hydrolysis was limited to 20% and was calculated by monitoring loss of substrate peak area. Velocity (M min Ϫ1 ) was plotted against the average substrate concentration for the reaction (  (19). The K m values of meprin A hydrolysis of neuropeptide Y and meprin B hydrolysis of neuropeptide Y, secretin, peptide YY, and kinetensin were too high for accurate measurement. Therefore, the k cat and K m constants were not individually determined. Instead the specificity constant k cat /K m was determined using the method of Fersht (20) where

a simplification of the Michaelis-Menten equation that applies when K m Ͼ Ͼ [S].
Identification of Cleavage Sites-Peptides (50 M) were incubated with meprin A or B (2 nM) at 37°C in 20 mM Tris-HCl, 150 mM NaCl, pH 7.5, in a final volume of 100 l for 5 min to 6 h depending on the efficiency of hydrolysis. Total peptide was subjected to HPLC using a Spheri-5 ODS 5 micron (4.6 ϫ 250 mm) column and a Brownlee Newguard 7-m RP-18 (3.2 ϫ 15 mm) guard column. Peptides were eluted with a linear gradient of acetonitrile in 0.1% TFA at a constant flow rate of 1.2 ml min Ϫ1 . Appropriate peptide peaks were extensively dried and then dissolved in a matrix solution of ␣-cyano-4-hydroxycinnamic acid in 50% acetonitrile and 0.3% TFA. This was then spotted on a PerSeptive Biosystems stainless steel plate. Products and substrates were identified using a Perceptive Biosystems Linear Voyager matrix-assisted laser desorption ionization time of flight (MALDI-TOF) machine with continuous extraction. The accelerating voltage was set at 18,750, and the laser intensity was between 175 and 300 in positive mode except for sCCK8 NH2 detection, which was in the negative mode. Between 60 and 240 scans were averaged. The machine was calibrated using the PerSeptive Biosystems Sequazyme Peptide Mass Standards kit. Data were evaluated using Grams/386 software version 3.04 (Galactic Industrial Corp.). Where required, amino-terminal sequence analysis was performed to identify peptides. Neuropeptide Y and peptide YY and their products were not separated by HPLC due to poor resolution. Instead, total peptide was bound to a ZipTip (Millipore) in 1% TFA, washed extensively to remove nonpeptide impurities, and eluted in 70% acetonitrile. Total peptide was then dissolved in ␣-cyano-4-hydroxycinnamic acid solution and treated as above.
Meprin Activity against Extracellular Matrix Proteins-The extracellular matrix proteins gelatin, collagen I and IV, laminin, and fibronectin (20 g) were incubated with meprin A or B (20 nM) for 18 h at 25°C in 20 mM Tris-HCl, 150 mM NaCl, pH 7.5, with a final volume of 40 l. Proteolysis was terminated by the addition of EDTA to 10 mM. The sample was boiled in SDS-PAGE sample buffer containing ␤-mercaptoethanol, and proteins were resolved on a 4 -15% PAGE gradient gel. Protein was visualized with Coomassie Brilliant Blue.
Degradation of Osteopontin by Meprin-Recombinant mouse osteopontin (50 g ml Ϫ1 ) was incubated with meprin A or B (10 nM) at 37°C in 20 mM Tris-HCl, 150 mM NaCl, pH 7.5, in a final volume of 40 l. A 6-l sample was removed at various times and immediately mixed with an equal volume of 20 mM EDTA to terminate hydrolysis. The sample was boiled in SDS-PAGE sample buffer containing ␤-mercaptoethanol and subjected to electrophoresis on a 4 -15% gradient gel. Osteopontin was detected by Western blot analysis using goat antimouse osteopontin as a primary antibody (1:1000), rabbit anti-goat IgG peroxidase conjugate as a secondary antibody (1:5000), with SuperSignal West Dura Extended Duration Substrate used as a horseradish peroxidase substrate for detection purposes.
Inhibition Profiles of Meprins-The inhibition profiles of meprin A and B were compared using actinonin and Pro-Leu-Gly-hydroxamate (PLG-(NHOH)). Meprin A or B (2 nM) was preincubated with various concentrations of inhibitor for 20 min before addition of substrate. Reactions were performed in 20 mM Tris-HCl, 150 mM NaCl, pH 7.5, and sCCK8 NH2 (50 M) was used as substrate. Substrate hydrolysis was limited to 25% and was determined by quantitative HPLC analyses. The concentrations of inhibitors ranged from 12.5 nM to 50 M for actinonin and from 250 nM to 500 M for PLG-(NHOH). The level of inhibition was determined by comparing the decrease in substrate concentration in the presence and absence of inhibitor.
Homology Models of Meprin ␣ and ␤ Protease Domains-The protease domain structures of the mouse ␣ and ␤ subunits were determined by knowledge-based homology modeling using the program Modeller (21). The input consisted of a sequence alignment of the catalytic domains of meprin and astacin, and the coordinates obtained from the Rutgers Protein Data Bank for the crystal structure of astacin (accession number 1AST (22)). The catalytic zinc was modeled by coordinating it to the equivalent residues that were determined for astacin. The model quality was assessed using the program PROCHECK (23) and was found to be comparable in stereochemical quality to a low resolution crystal, typical of homology models. Surface representations were prepared using WebLab Viewer 3.7 (Molecular Simulations Inc., San Diego, CA).

Meprin A and B Have Distinct Specificities for Preferred Amino Acids at or Near the Cleavage Site of Substrates as
Determined by Peptide Libraries-Peptide libraries were used to identify preferred amino acids of substrates at or near the cleavage site and to map the primed subsite 2 binding region of meprins (25). A completely random dodecapeptide mixture acetylated at the amino terminus was partially digested with meprin A or B, and the digested peptides were subjected to amino-terminal peptide sequencing. For meprin B the strongest specificity was seen at the P1Ј site of the substrate where acidic residues were preferred ( Fig. 1). Aspartic acid at the amino terminus of peptide products was particularly prevalent with a signal 2.5-to 41-fold greater than other residues. Glutamic acid was also selected but to a lesser extent. The selectivity decreased further away from the scissile bond; with 41-, 3.4-, 2.1-, and 1.5-fold differences between the highest and lowest relative molar amounts for the P1Ј, P2Ј, P3Ј, and P4Ј subsites, respectively. There was no clear preference for a single amino acid at the P2Ј through P4Ј sites, aside from a slight preference for proline in P3Ј. However, there was a selection against basic residues at P2Ј and P3Ј as well as a slight selection against aspartic acid and arginine side chains in P4Ј.
Meprin A had a completely different profile to meprin B for amino acid preference at the P1Ј subsite (Fig. 2). There was a selection for small and aromatic residues ( The only similarity between the enzymes at the P1Ј site was a strong selection against proline. The profiles at the P2Ј subsite were also dissimilar for the two enzymes. Proline was the most preferred amino acid at the P2Ј position for meprin A, with at least a 2-fold larger signal than other residues. There was also a slight selection against leucine, asparagine, and basic amino acids in the P2Ј subsite. Proline was also the preferred residue at the P3Ј site, although to a lesser extent than the P2Ј subsite. Charged residues and tyrosine were slightly disfavored. The specificity signature at P3Ј was strikingly similar to that of meprin B, and there was little or no specificity at P4Ј. The data indicated that, for both enzymes, at least three primed subsites contribute to substrate binding specificity (Figs. 1 and 2). In addition, meprin A and B have very different specificities in the P1Ј and P2Ј subsites.
Comparison of Naturally Occurring Peptides as Substrates of Meprins-Specific peptides were tested as meprin substrates to further define substrate preferences (Table I). The peptides were chosen based on the knowledge of meprin peptide bond specificity and the peptide localization in vivo. In an initial screen of potential substrates it was noted that peptides of the gastrointestinal tract (e.g. GRP-(14 -27), sCCK8 NH2 , and gastrin 17) were relatively good substrates of meprins. The screen employed 25 bioactive peptides that fell into four groups. The first group, GRP-(14 -27), sCCK8 NH2 , secretin, glucagon, neuropeptide Y, and cerulein were susceptible to hydrolysis by both meprin A and B. GRP-(14 -27) was efficiently hydrolyzed by both meprins under the conditions used; the intact peptide was almost completely hydrolyzed after 4 h. The second group, bombesin, neurotensin, luteinizing hormone releasing hormone (LHRH), bradykinin, ␣-melanocyte stimulating hormone (␣-MSH), substance P, valosin, parathyroid hormone fragment 13-34, vasoactive intestinal peptide, and angiotensin I were only susceptible to meprin A. The third group of peptides was susceptible to meprin B only; orcokinin and gastrin 17 were 2 The nomenclature for the interaction of proteases with their substrates is from Schecter and Berger (24). The substrate amino acid residues are called P (for peptide), the subsites on the protease that interact with the substrate are called S (for subsite). The residues on the amino-terminal side (also known as the unprimed residues) of the scissile bond (bond that is cleaved during hydrolysis) are numbered P1 through P6 counting outward. The residues on the carboxyl-terminal side (also known as primed residues) of the scissile bond are numbered P1Ј through P6Ј. Thus, hydrolysis occurs between the P1 and P1Ј residues. The subsites on the protease are termed S1 through S6 and S1Ј through S6Ј to complement the substrate residues that interact with the enzyme. particularly well hydrolyzed together with peptide YY and kinetensin. The final group of peptides was resistant to both enzymes; these peptides were [Lys 8 ]-vasopressin, somatostatin, kassinin, oxytocin, and ␣-neurokinin. The screen identified 11 peptides not previously known to be cleaved by meprins. Novel meprin A activity was seen against GRP-(14 -27), sCCK8 NH2 , secretin, glucagon, cerulein, bombesin, and vasoactive intestinal peptide. In addition, novel substrates of meprin B identified were GRP-(14 -27), sCCK8 NH2 , secretin, glucagon, neuropeptide Y, cerulein, orcokinin, peptide YY, and kinetensin.
Kinetic Parameters for Meprin Cleavage of Bioactive Peptides-A kinetic study was conducted to better characterize the substrates of meprins identified in the initial screen. The kinetic constants, k cat and K m as well as the specificity constant k cat /K m (Table II) were determined by directly fitting data to the Michaelis-Menten equation by nonlinear regression analysis as described under "Experimental Procedures." All peptides tested exhibited typical Michaelis-Menten kinetics (data not shown). Velocity was determined by monitoring the loss of substrate peak area rather than the appearance of substrate. This approach was taken due to the presence of multiple cleavage sites as evident by more than two product peaks for most substrates (data not shown). The K m values for meprin A ranged from 116 for GRP- (14 -27) to 425 M for bradykinin; for meprin B the range was from 7.1 for gastrin and 211 M for sCCK8 NH2 . Meprin A had k cat values between 11.8 for secretin and 88.3 s Ϫ1 for GRP- (14 -27). The k cat values of meprin B lay between 12.4 for gastrin and 26.8 s Ϫ1 for glucagon. Of all the peptides tested in this study, meprin B hydrolysis of gastrin 17 was found to give the highest specificity constant (17.5 ϫ 10 5 M Ϫ1 s Ϫ1 ). This was predominantly due to the low K m value of 7.1 M. This value is over 6-fold lower than any other K m determined for either enzyme. The fluorogenic bradykinin analog substrate, 2-aminobenzoyl-RPPGFSPFRK-(dinitrophenyl)-G is commonly used to study meprin A activity (27). However, bradykinin is a relatively poor substrate due to a high K m (425 M (26)). This high K m leads to the low specificity constant of 5.1 ϫ 10 4 M Ϫ1 s Ϫ1 . Rates of hydrolysis for neuropeptide Y by meprin A and neuropeptide Y, secretin, peptide YY, and kinetensin by meprin B were linear with respect to substrate concentration at high peptide concentrations, therefore, the individual k cat and K m values were not directly determined.
FIG. 2. Preferred amino acids in substrate subsites P1 to P4 for meprin A. The same acetylated dodecamer peptide library mixture (1 mM) described in Fig. 1 was incubated with meprin A (33 nM) until the library was between 5 and 10% digested. Amino-terminal sequencing of the resulting products allowed for the characterization of amino acid preference for meprin A at each primed subsite. Instead the specificity constant k cat /K m was determined directly. As expected, these values were the lowest seen for all substrates, the lowest value being 0.133 ϫ 10 5 M Ϫ1 s Ϫ1 for meprin B hydrolysis of secretin.

Meprins Have Distinct Differences in Peptide Bond
Specificity-Peptides that were relatively good substrates of meprins were further characterized to determine peptide bonds cleaved. The peptides were digested with either meprin B (Table III) or meprin A (Table IV), and products were identified by MALDI-TOF. Cleavage was evident at more than one site in all instances except meprin B hydrolysis of orcokinin, sCCK8 NH2 , peptide YY, and kinetensin and meprin A hydrolysis of bradykinin (Tables III and IV). In general meprin B appeared to be a more specific protease than meprin A. Bonds with an acidic residue in the P1Ј position were cleaved by meprin B in most instances (Table III). Out of 20 scissile bonds determined, 13 sites were amino-terminal to an acidic residue. Residues seen multiple times at P1Ј were aspartic acid (9 times), glutamic acid (4 times), and glycine (twice), consistent with the peptide library data (Fig. 1). Meprin B cleaved gastrin 17, orcokinin, glucagon, sCCK8 NH2 , secretin, peptide YY, and neuropeptide Y with an acidic residue at P1Ј (Table III). However, GRP-(14 -27) and kinetensin, which lack acidic residues, were also hydrolyzed. This demonstrated a distinct preference, but not an absolute requirement, for acidic residues in the P1Ј site of substrates by meprin B. The data also indicated a lack of substrates with a basic residue in the P2Ј position, which is consistent with peptide library data in which both lysine and arginine are strongly selected against (Fig. 1). Meprin B is able to accommodate all types of amino acids in P1 showing a lower stringency for specific residues in this subsite. Glutamic acid is seen the most often (6 out of 20 times). Proline was frequently seen in the unprimed region (8 times). Interestingly, meprin B was able to act as both an amino-and carboxypeptidase (e.g. orcokinin and sCCK8 NH2 , respectively). In general, meprin A tended to cleave bonds that have small uncharged or hydrophobic residues in their P1 and P1Ј positions (Table IV). The presence of a proline residue in meprin A substrates was frequent. Proline was seen at P4 (4 times), P3 (twice), P2Ј (5 times), and P3Ј (twice). Proline was never seen at P1 or P1Ј.
The Hydrolytic Efficiency of Meprin B Is Dependent on Peptide Length-Previous work indicated that meprin A has a preference for substrates with a minimum of eight amino acids (28). To test the peptide length requirement for efficient hydrolysis by meprin B, derivatives of sCCK8 NH2 were used as substrates (Table V). Meprin B cleaved sCCK8 NH2 at a single site toward the carboxyl terminus (Table III). The sCCK8 NH2 derivatives allowed the determination of the unprimed subsites contribution, peripheral to the scissile bond as well as the contribution of the carboxyl terminus toward substrate recognition. The data indicated that meprin B had a preference for a minimum of six amino acids. CCK4 NH2 and CCK5 NH2 were relatively poor substrates (17 and 15% hydrolyzed, respectively, compared with 56% for sCCK8 NH2 ), whereas CCK6 NH2 and larger peptides were relatively good substrates. The effects of modifications within sCCK8 NH2 were also examined. The absence of sulfation at the tyrosine residue resulted in a poorer substrate compared with the parent peptide (35 and 56% hydrolysis, respectively). Blockage at the amino termini (Boc-CCK8 NH2 ) increased the susceptibility to hydrolysis (59% hydrolyzed compared with 35% for CCK8 NH2 ). The presence or absence of amidation at the carboxyl terminus resulted in larger differences toward hydrolysis. The free acid (CCK8) was 76% hydrolyzed compared with 35% hydrolysis for CCK8 NH2 . Cerulein, a CCK analog, was hydrolyzed 60% in 4 h. This peptide is identical to sCCK8 NH2 except for a methionine in place of a threonine residue, carboxyl-terminal to the sulfated tyrosine, and an amino-terminal extension of pyroglutamic acid and glutamine. The free acid was hydrolyzed to the greatest extent of all CCK derivatives tested.
Meprin Degradation of Extracellular Matrix Proteins-To determine the ability of meprins to degrade extracellular matrix components, gelatin, fibronectin, collagens I and IV, and laminin were incubated with meprins, and the products were subjected to SDS-PAGE (Fig. 3, upper panel). Gelatin proved to be the best substrate under the conditions used. After an 18-h incubation with meprin A or B, intact gelatin was extensively hydrolyzed by both enzymes. Fibronectin was sensitive to both enzymes, yielding similar patterns of hydrolysis. The major protein bands above 200 kDa seen in the control were hydrolyzed by both meprin A and B, yielding bands that migrated slightly faster and reproducibly in both cases. Collagens I and IV and laminin were resistant to hydrolysis at 25°C (data not shown). Collagen I was sensitive to hydrolysis by meprin B at 37°C, possibly due to some local unfolding of the triple helix (data not shown).
Meprin B Hydrolysis of Osteopontin-To test the ability of meprin B to hydrolyze proximal to acidic residues in the context of a protein rather than a short peptide, osteopontin was used as a substrate (Fig. 3, lower panel). This 294-amino acid protein has a high percentage of acidic residues (23%). Meprin B effectively degraded the protein in a time-dependent fashion. Over a 30-min incubation the intact protein decreased markedly as detected by Western blot analysis. Meprin A showed little if any ability to degrade osteopontin under the conditions used; the 60-min time point shows slightly less protein than the control.
Inhibitor Profiles of Meprins-The active sites of meprins were further mapped using the inhibitors PLG-(NHOH) and actinonin. The crystal structure of astacin complexed with PLG-(NHOH) has shown that the inhibitor binds the unprimed subsite region (accession number 1QJJ (29)). The hydroxamate moiety ligates the zinc, whereas the peptide moiety binds the unprimed subsites and proline binds P3, leucine P2, and the glycine P1. The IC 50 values toward meprin A and B were similar to one another with values of 30 and 50 M, respectively (Fig. 4). This was in the same range as the reported K i value toward astacin, which has a K i of 16 M (29). Unexpectedly, actinonin was a much more potent inhibitor of both meprin A and B compared with astacin, even though it has a very similar structure to PLG-(NHOH). The IC 50 values were over 300-and 125-fold higher against meprin A and B, respectively, at 100 and 400 nM, respectively. In contrast, the K i toward astacin is 8-fold lower at 130 M (30).

DISCUSSION
The work herein clearly demonstrates marked differences in the preferences of meprin A and B for the P1Ј and P2Ј subsites of substrates. Meprin B is predominantly an Asp/Glu-N peptidase as shown by the peptide library studies and individual substrate data, selecting acidic residues in the P1Ј site ( Fig. 1 and Table III). By contrast, meprin A selects a variety of small and hydrophobic amino acids in the P1Ј site and clearly prefers proline residues in the P2Ј sites ( Fig. 2 and Table IV). Both proteases have extended binding sites and prefer substrates of at least 6 amino acids (Ref. 28 and Table V). The different specificities of the two subunits implicate diverse functions.
Amino acid sequence analyses and homology models of the ␣ and ␤ protease domains yield insights into the substrate specificity differences and activities of meprins compared with other astacin family members, such as crayfish astacin and mouse bone morphogenic protein-1 (BMP-1; Fig. 5 and Ref. 1). There are nine basic residues in the meprin ␤ sequence that are not basic in the equivalent positions in the meprin ␣ protease domain. Three of the basic residues are within the active site cleft of mouse meprin ␤ (Arg-147, Lys-185, and Lys-214; Fig. 5) and have the potential to form a salt bridge with acidic residues of substrates. For example, Phe-161 in meprin ␣, six amino acids carboxyl-terminal to the zinc HEXXH binding site, is equivalent to the Arg-147 position in meprin ␤. Astacin and meprin ␣ have aromatic residues at this position, whereas meprin ␤ and BMP-1 have basic residues. Meprin ␤ and BMP-1 have a preference for acidic residues in P1Ј of substrates (Fig.  1, Table III  Peptides (50 M) were incubated with meprin B (2 nM) in 20 mM Tris-HCl, 150 mM NaCl, pH 7.5, for between 5 and 300 min. Peptide products were separated by HPLC, collected, and identified by MALDI-TOF and amino-terminal sequencing. Cleavage sites are indicated by an arrow (2). In instances of multiple cleavage sites double arrows (s) represent the major site(s) of cleavage. The subscript "NH2" represents amidation at the carboxy terminus; the asterisk represents sulfation at the tyrosine; p represents pyro. TABLE IV Cleavage sites in peptides by meprin A Peptides (50 M) were incubated with meprin A (2 nM) in 20 mM Tris-HCl, 150 mM NaCl, pH 7.5, for between 5 and 300 min. Peptide products were separated by HPLC, collected, and identified by MALDI-TOF and amino-terminal sequencing. Cleavage sites are indicated by an arrow (2). In instances of multiple cleavage sites double arrows (s) represent the major site(s) of cleavage. The subscript "NH2" represents amidation at the carboxy terminus; the asterisk represents sulfation at the tyrosine; p represents pyro; Ac-represents N-acetylated at the amino terminus.  with a P1Ј aspartic acid. Thus, it appears that meprin ␤ and BMP-1 are similar proteins with respect to activity. Residues Tyr-199 and Lys-185 of mouse meprin ␣ and ␤, respectively, form the floor of the proposed S1Ј subsites (Fig. 5). Thus, this is a potential site of interaction between meprin ␤ and acidic side chains of substrates such as gastrin. The CCK free acid, CCK8, is a better substrate for meprin B than CCK8 NH2 , which is consistent with the preference of meprin ␤ for an acidic moiety in or around the P2Ј position of substrates.
The Tyr-199 residue of mouse meprin ␣ protrudes into the active site cleft. This may explain why prolines are preferred at proximal sites to the scissile bonds in substrates. Prolines impart rigidity and stability to peptide backbones and have the ability to form cis peptide bonds within the backbone of substrates. This may allow the peptide to kink around the protruding tyrosine, thus allowing the peptide to fit in the cleft. Consequently, the cis conformation of a peptide bond may be preferred by meprins.
The inhibition of meprins by PLG-(NHOH) was expected, because both enzymes select for prolines in the unprimed substrate sites (e.g. bradykinin and gastrin 17 for meprin A and B, respectively). The crystal structure of astacin complexed with PLG-(NHOH) shows that the inhibitor binds along a ␤-edge strand, which consists of Trp-114 through Tyr-116 (accession number 1QJJ (29)). The tyrosine is not conserved in the meprins; Met-143 and Ser-128 are in the equivalent positions in meprin ␣ and ␤, respectively. Therefore the pocket in astacin is likely smaller than in both meprins. Actinonin is bulkier than PLG-(NHOH) in the region that is predicted to bind along the ␤-edge strand. Actinonin is a much better inhibitor of meprins than it is of astacin, perhaps due to steric hindrance with astacin that would not occur with the meprins. The improved inhibition of meprins by actinonin over PLG-(NHOH) could reflect the presence of an additional methylene group within the carboxyl-terminal residue of the peptide backbone in acti- nonin. This may provide more optimal spacing between the metal chelating hydroxamate moiety and the peptide backbone and side chains of the inhibitor.
Meprins are the only known endopeptidases in brush border membranes that degrade proteins; other proteases in these specialized cell surface membranes either degrade only small polypeptides (e.g. neprilysin) or are exopeptidases such as angiotensin converting enzyme or leucine aminopeptidase (32). Thus, meprins may initiate degradation of proteins in the lumen of the kidney proximal tubule, act as sheddases in this location, or activate/inactivate polypeptide hormones in the glomerular filtrate. In the intestine, meprins are most active in the ileum where there is an active immune system (33). The location at this site and proteolytic capacity of meprins implicates them in the hydrolysis of proteins and formation of peptides that are presented to antigen-producing cells. The presence of meprins in leukocytes and cancer cells also implies functions for meprins in cytokine activation or degradation and in hydrolysis of basement membrane components.
Peptides of the gastrointestinal tract appear to be among the best substrates identified for meprins and are of particular interest because of the expression of meprins in the intestine. Gastrin 17, cerulein, and sCCK8 NH2 have an identical stretch of five amino acids at their carboxyl termini, the four terminal amino acids confer the total biological activity of these peptides (34). Therefore meprins have the ability to inactivate these important regulatory molecules. The gastrointestinal peptides regulate the movement, secretory activity, and growth of the intestinal tract and pancreas, and thus the concentration of these peptides must be highly regulated.
The kidney is known to play a major role in the clearance of many plasma polypeptides such as glucagon and sCCK8 NH2 (e.g. Ref. 35). Patients with chronic renal failure also have elevated levels of peptides in the blood that are involved in gut motility, hunger, and satiety. Neurotensin, peptide YY, substance P, vasoactive intestinal peptide, GRP- (14 -27), and gastrin are all elevated during chronic renal failure probably due to a decrease in metabolism of circulating peptide (36,37). This points toward a role of renal brush border proteases, including meprins, in the catabolism of circulating peptides and thus the recapture of amino acids and/or the alteration of urinary peptides. The finding that osteopontin is a substrate for meprin B is intriguing. Experimental hydronephrosis results in an accumulation of osteopontin protein within the lumen of the proximal tubule at a time that meprin protein is markedly decreased (38,39). The absence of meprins may account for some of the accumulation of peptides and proteins that contribute to a cascade of events that lead to fibrosis and end-stage renal disease.