Profein Science (1995), 4:1247-1261. Cambridge University Press. Printed in the USA Copyright 0 1995 The Protein Society

The astacin family of metalloendopeptidases

JUDITH S. BOND’ AND ROBERT J. BEYNON2 ’ Department of Biochemistry and Molecular Biology, The Pennsylvania State University College of Medicine, Hershey, Pennsylvania 17033 Department of Biochemistry and Applied Molecular Biology, University of Manchester Institute of ‘Science and Technology, Manchester M60 1QD. United Kingdom (RECEIVEDMarch 23, 1995; ACCEPTED April19, 1995)

Abstract The astacin family of metalloendopeptidases was recognized as a novel family of proteases in the 1990s. The cray- fish enzyme astacin was the first characterized and is one of the smallest members of the family. More than 20 members of the family have nowbeen identified. They have been detected in species ranging from hydra to hu- mans, in mature andin developmental systems. Proposed functions of these proteases include activation of growth factors, degradation of polypeptides, and processing of extracellular proteins. Astacin family proteases aresyn- thesized with NH,-terminal signal and proenzyme sequences, and many (such as meprins, BMP-1, folloid)con- tain multiple domains COOH-terminal to the protease domain. They are eithersecreted from cells or are plasma membrane-associated enzymes. They have some distinguishing features in addition to the signature sequencein the protease domain: HEXXHXXGFXHEXXRXDR. They have a unique typeof zinc binding, with pentacoor- dination, and a protease domain tertiary structure that contains common attributeswith serralysins, matrix me- talloendopeptidases, and snake venom proteases; they cleave peptide bonds in polypeptides such as insulin B chain and bradykinin andin proteins such as casein and gelatin; and theyhave arylamidase activity. Meprins are unique proteases in the astacin family, and indeed in the animal kingdom,in their oligomeric structure; they are dimers of disulfide-linked dimers and arehighly glycosylated, type I integral membrane proteins thathave many attributes of receptors or integrins with adhesion, epidermal growth factor-like, and transmembrane domains. The 01 and /3 subunits are differentially expressed and processed to yield latent and active proteases as well as membrane- associated and secreted forms. Meprins represent excellent models of hetero- and homo-oligomeric enzymes that are regulated at the transcriptional and posttranslational levels. Keywords: astacin; bone morphogenetic protein-I ; brush border membranes; development; meprin; metallo- endopeptidases; secreted and plasma membrane proteases; tolloid

The great abundance and variety of proteases in nature have One of thefamilies that emerged in the 1990s is the “astacin stimulated scientists to look forcommonalities in theseen- family” (Dumermuth et al., 1991). Recognition of this family zymes and to classify them. By the end of the 1980s there was was a consequence of cloning and sequencing of mammalian an accepted system to classify endopeptidases on the basis of meprins (metalloproteases from brush border membranes of mechanism of action andthree-dimensional structure (Neurath, rodents and humans). Theoriginal members of the family iden- 1989). Four mechanistic classes of proteases (serine, cysteine, tified were: the crayfish digestive enzyme astacin, bone morpho- aspartic, and metallo)were identified, andsix families (two ser- genetic protein-I (BMP-1) from human bone, meprins from ine families: chymotrypsin and subtilisin; one cysteine: repre- mouse kidney and humanintestine (the latter was originally des- sented by papain; one aspartic: penicillopepsin; two metallo: ignated asN-benzoyl-L-tyrosyl-p-aminobenzoic acid hydrolase carboxypeptidase and thermolysin families). Each family had or PPH), and UVS.2, a partial sequence fromXenopus laevis characteristic active site residues and was thought to have de- embryos. The family was named the “astacin family” because scended from a common ancestorby divergent evolution. How- a protease from thecrayfish Astacus astacus L., the enzyme now ever, as a result of the informationexplosion that occurred along called astacin (EC 3.4.24.21), was the first to be sequenced and with the advances in molecular and cell biology, as many as60 biochemically characterized (Titani et al., 1987). It has sub- families of peptidases (endo- andexopeptidases) have now been sequently been crystallized and the three-dimensional structure proposed on thebasis of similarities in primary structure (Rawl- solved, adding a new dimension to our understanding and ex- ings & Barrett, 1993). ploration of these metalloendopeptidases (Bode et al., 1992). Since the discovery of this family, 17 additional members have ~~~ ~ ~~~ ~~ ~ ~~ Reprint requests to: Judith S. Bond, Department of Biochemistry and been cloned and sequenced, and thereis evidence for others. It Molecular Biology, Pennsylvania State University Collegeof Medicine, is the aim of this review to summarize information about this Hershey, Pennsylvania 17033; e-mail: [email protected]. new family, point out the unique features of family the and in- 1247 1248 J.S. Bond and R.J, Beynon dividual members, and suggest how future investigations can all of the proteins of the family studied thus far are secreted lead to answers to fundamental questions. or plasma membrane bound. The prosequences varygreatly in size, containing up to 5 19 amino acids for Drosophila lolloid- relaled-/ (DrTlr-/),and are likely to be important for regulat- Members of the astacin family ing activity and perhaps expression of the proteases. Regarding the latter point, for example, thelarge prosequence of DrTlr-/ Domain slrudures has been suggested to prevent expression of this gene product Astacin family members are characterized by a unique 18-amino in early stages of embryogenesiswhen cell cycles are very short acid signature sequence, HEXXHXXGFXHEXXRXDR, which (Nguyen et al., 1994). begins with an HEXXH zinc binding motif found in most me- The smallest members of the family have no domains COOH- talloendopeptidases. The signature sequence is part of an ap- terminal to the protease domain. They are: crayfish astacin (EC proximately 200-amino acid sequence, which is the entire mature 3.4.24.21)and theteleostean choriolysin L (LCE; EC3.4.24.66) crayfish astacin and the catalyticor protease domain of all the and choriolysin H (HCE 1 and 2; EC 3.4.24.67). They are se- members of the family (Fig. 1). Signal and prosequences are also creted proteaseswith molecular masses of approximately 25 kDa. common features of family members, with the possible excep- Most of the known astacin family members contain one or tion of QuCAM-I; these NH2-terminal domains have yet to be more copies ofan EGF(_epidermal growth factor)-like(E) and/or found for this latter protein. Astacin,which until recently was a CUB (complement subcompon&s Clr/Cls, embryonic sea only studied as a mature protein that begins with the protease urchin protein Uegf, BMP-I) domain. These are non-catalytic, domain, is now known to containa prepro segment of 49 resi- “interaction” domains, which may promote protein-protein and dues (W. Stocker & R. Zwilling, pers. comm.). The transient substrate interactions. EGF-like domains consist of approxi- signal peptides direct the proteins into the endoplasmicreticu- mately 40 amino acids and have 6 highly conserved cysteine res- lum during biosynthesis, which is consistent with the finding that idues, which most probably are linked by intradomain disulfide

QuCAM-1 . -. .

HMP1 I Pro El

DrTld, HuTld. MuBMP-1 D:w E DrTlr-1 0 I1 Pro Mu,Rt.Hu Meprin-a c mflc

Mu.Rt.Hu Meprin-p lpro ~ - I ., - E blcl

0 200 400 600 800 1000 Amino acids

Fig. 1. Domain structure of astacin family members. Domains are aligned relative to the astacin protease domain.All sequences were deduced from cDNA sequences except for astacin in which the protease domain sequence was determined from the ma- ture protein and the prepro sequence deduced from the cDNA.References: astacin, Titani et al. (1987); HCEI, HCE2, and LCE, Yasumasu et al. (1994); QuCAM-I, Elaroussi and DeLuca (1994); HMPI, Sarras (pers. comm.); BPI0 (blastula protease-10). Lepageet al. (1992a);SpAN, Reynoldset al. (1992); SuBMP, Hwangetal. (1994); HuBMP-I, Wozneyet al. (1988); XeBMP-I, Maeno et al. (1993); DrTld (Drosophila tolloid), Shimell et al. (1991); HuTld, Takahara et al. (1994); MuBMP-I, Fukagawa et al. (1994); DrTlr-l (Drosophila rolloid-relared-l), Nguyen et al. (1994); Mu meprin-a, Jiang et al. (1992); Rt meprin-a, Cor- beil et al. (1992); Hu meprin-a, Dumermuthet al. (1993); Mu meprin-0, Gorbeaet al. (1993);Rt meprin-0, Johnson and Hersh (1992); Hu meprin-0, Sterchi (pers. comm). Signal peptides are represented by the NH2-terminal open boxes; Pro, prosequences; Protease, catalytic domains; E, EGF-like domains; CUB, Clr/s complement-like domains; MAM, adhesion domains; X, un- known; I, inserted domain; TM, putative transmembrane domains;C, cytoplasmic domain (seetext for further explanations). Species are indicated as: Qu, quail; Su, sea urchin; Hu, human; Xe, frog; Dr, Drosophila; Mu, mouse; Rt, rat. The astacin family 1249 bonds. EGF-like domains exist in one or more copies on many 34% identical. Thea subunit containsa 56-amino acidinserted secreted proteases, such as the coagulation proteases FactorsX domain (I) that has no counterpartin the P subunit, and there and IX, protein C, andtissue plasminogen activator, and many is no homology between the putative transmembrane or cyto- plasma membrane proteinssuch as thrombomodulin,L-selectin, plasmic domains of a and (Marchand et al., 1995). The cyto- neurexin la, and Neu differentiation factor (Campbell& Bork, plasmic domains are deduced to be 6 residues for a and 26 1993). CUB domains(sometimes referred to as Clr/Clsrepeats) residues for 0. The cytoplasmic domain is highlypositively consist of approximately 110 amino acids,4 conserved cysteine charged (seven of a sequence of nine amino acids are Arg/Lys residues, and conserved hydrophobicity patterns typical of an residues). This sequence may help tostabilize the protein in the antiparallel P-sheet. It has been suggested that they form anti- membrane and mayinteract with cytoskeletal or other cytoplas- parallel P-barrels similarto those in immunoglobulins (Bork& mic proteins. This domain contains two potential phosphory- Beckmann, 1993). They have been foundin complement serine lation sites: a protein kinase C phosphorylation site (TRR) and proteinases, spermadhesins, neuronal recognition molecule A5, a calmodulin kinase I1 phosphorylation site (RANT) (Kennelly and the tumor necrosis factor-induced protein Tsg6. & Krebs, 1991). The domain structuresof the human (Hu) andXenopus (Xe) Several other astacin family members have been identified BMP-1 and human and Drosophila (Dr) tolloid (Tld) and from DNA sequences that are not shown in Figure 1. For ex- tolloid-related-l (Tlr-I)are very similar except that Tld and Tlr-1 ample, partial Cuenorhabditis elegans sequences (CeR15 1.5, generally have three extra COOH-terminal domains (2 CUBS CeF42A10.8, CeCOSDl1.6) have been identified from genomic and 1 EGF-like domain). However, mouse (Mu) BMP-1 has the DNA sequencing of cosmid clones; the primary sequences of the same domain structure as Tld (Fukagawa et al., 1994). There protease domain of two of these are included in Figure 2. UVS.2, is now evidence that the shorterBMP-1 variants and thetolioids a partial cDNA sequence from X. laevis embryos, also identi- are alternately spliced forms of a single gene in mice and humans fied as an astacin family member,is not included in the figures (Takahara et al., 1994). It is likely that splice variants of these (Sato & Sargent, 1990; Dumermuth et al., 1991). enzymes exist in most species. In addition,isoforms of thesepro- The domain structureof the astacin family representsa pat- teases exist as a result of multiple genes. For example, DrTld and tern seen in several protease families, such as the coagulation DrTlr-1 are the productsof two differentgenes (Nguyen et al., proteases, complement proteases, proprotein convertases, and 1994), and differentgenes have been identified for BMP-1 and matrixins (Furie & Furie, 1988; Matrisian, 1992; Steiner et al., tolloid (or xolloid) in Xenopus (R. Albano & L. Dale, pers. 1992; Sim et al., 1993). The various membersof these families comm.). result from gene duplication, evolution, gene fusion, and exon Sea urchin (Su) SPAN andBPlO have a Thr-rich (Tr) domain shuffling. For all of these families, a conserved proteasedomain and SuBMP has a Ser/Thr-rich (S/Tr) domain that have not is associated with a variety of noncatalytic domains that yield been identified in other members of the family. In SpAN, the different proteins, with somewhat different enzymatic activities, 40-amino acid domain contains four repeats of the sequence regulation, expression, and interactionswith small ligands and STTTLQTT (Reynoldset al., 1992); in BPlO, the 25-amino acid other proteins. The structures of the precursor processing en- domain contains 14 threonine residues (Lepage et al., 1992a); dopeptidases of the subtilisin family have some striking similar- in SuBMP, the 66-aminoacid domain contains 22 serine or thre- ities to those of the astacin family(Steiner et al., 1992). These onine residues (Hwang et al., 1994). These regions may contain proteases (e.g., yeast kexin, mammalian furin, PC2, PC3/PCI, sites for 0-glycosylation. PACE) are synthesized with preprosequences followed by highly The full-length cDNA sequences of the a and @ subunits of conserved subtilisin-like catalytic domains(47-7270 identical). meprins are about45% identical and predict domains COOH- Only furin and kexin contain TM domains; the others arelikely terminal to the protease domain, designated as MAM (meprin to be soluble enzymes, though they interact with membranes and subunits, 4-5 protein, and receptor protein tyrosine phospha- may have signals that sort them to secretory granules. There are tase p), X, I, E (EGF-like), TM (putative transmembrane), and several other COOH-terminal domains,including Ser/Thr-rich, C (cytoplasmic). The MAM domain is found in several mem- P, amphipathic helix, Cys-rich, and cytoplasmic domains that brane proteins and hasbeen proposed to act as an adhesion do- are present in some of the members. Specific expression of var- main (Beckmann & Bork, 1993). Enteropeptidase (also termed ious members can actto generatea diversity of products from enterokinase; EC3.4.21.9), the intestinal enzyme that converts identical precursors and are thought toprovide finely tuned reg- trypsinogen to trypsin, also containsa MAM domain (Kitamoto ulation of physiological and developmental processes. et al., 1994). MAM domains are120-180 amino acidsin length, and there are14 amino acids,including 4 cysteine residues,that Proposed functions and tissue distribution are conserved. The X domains of meprin subunits contain approximately Astacin family enzymes havebeen proposed to have roles in ma- 180 amino acids. No counterparts (matching sequences) have ture and developmentalsystems. They areexpressed in a tissue- been found in the sequence databases. This domain may be specific manner in mature organisms, and are temporally and an “extender” region to position the activesite of meprin sub- spatially expressed in developmental systems (Table 1). units away from the cell membrane. In addition, at least two Astacin is synthesized in the crayfish hepatopancreas, anor- sites of N-glycosylation of mouse meprin a have been localized gan that has intestinal, hepatic, and pancreatic functions(Vogt to this domain (Jiang et al., 1992). Mouse meprin CY and @ et al., 1989). It is stored extracellularly as an active proteinase contain approximately 30% carbohydrate that is attributed to in the stomach-like cardia, andis thought to be a digestive en- N-glycosylation.The cDNA-deduced amino acid sequences zyme in a variety of crustaceans. COOH-terminal to theX domain differ in thea and P subunit Several of the astacin family members are implicatedin the of meprin, except that both contain EGF-like domains that are hatching process of embryos andin skeleton formation. There 1250 J.S. Bond and R.J. Beynon

-~ I I I I I , I I I I I 10 20 30 40 50 60 70 80 90 100 110 Ce5D11.6 Ce7D10.4 HCEl HCE2 LCE QU” 1 HMPl Astacin MUBMP-1 Dfllr-1 DrTld SuBMP XeBMP-1 HuBMP-1 SPAN BPlO MuMep-a RtMep-a HuMep-a MuMep-b RtMep-b HuMep-b

130 140 150 12 0 140 130 160 180 170 190 210 200 CeSD11.6 Ce7D10.4 HCEl HCEZ WE QUCA” 1 HMpl Astacin MuBMP- 1 DrTlr-1 Dflld SuBMP XeBMP-1 HUBMP- 1 SPAN BP10 MuMep-a RtMep-a HuMep-a MuMep-b RtMep-b HuMep-b

Fig. 2. Sequence comparisonof the protease domainsof 22 members of the astacin family.Sequences were aligned by the pro- gram Megalign from DNASTAR, using the CLUSTAL algorithm and a PAM250 residue weight table. Alignments highlight residues that are in the majority for all sequences. Abbreviations and referencesas for Figure 1. Included in ;his alignment (first two sequences) are two partial sequences from nematodes, C. elegans, predicted from genomic DNA sequencing (GenBank).

is good evidence to indicate that thefish enzymes HCEl and 2 from human bone (Wozney et al., 1988). The complex was ca- and LCE work in concert to degrade theegg shell/chorion of pable of inducing cartilage and bone formation. BMP-2 and 3 embryos. In Oryzias latipes, these enzymes are storedin zymo- are members of the TGF-0 superfamily; these are growth factors gen granules and aresecreted/activated at the appropriatetimes active in skeletodbone formation. The associationof BMP-1, to serve this function (Yasumasu et al., 1992). HCEs actdirectly a putative metalloendopeptidase, with the complex led to the on the “hard” chorion to digest protein into Pro-rich pieces; suggestion that it is required to activate latent TGF-0. The coun- LCE further degrades the remaining polypeptides (Lee et al., terpart of mammalian BMP-I and theTGF-0s in Drosophila are 1994). CAM-1 of quail embryos is also implicated in degrada- tolloid and decapentaplegic (dpp),respectively (Shimell et al., tion of eggshell matrix proteins (Elaroussi& DeLuca, 1994). It 1991). Tolloid and dpp are intimately involved in dorsal-ventral is induced by 1,25-dihydroxyvitamin D-3 and is proposed to pattern formation in early stages of Drosophila development: break down extraembryonic chorioallantoic membrane, mobi- dpp null embryos are ventralized and lack all dorsal epidermal lizing Ca+’ for bone formation and weakening the shell in structures (Ferguson & Anderson, 1992), and dpp is also criti- preparation for hatching. cal for developmentof the midgut. Because tolloid and TGF-(3 The hydra HMPl enzyme is thought tobe involved in pattern molecules are highly conserved in other phyla, it has been pro- formation and morphogenesis (M.P. Sarras, Jr.,pers. comm.). posed the interactionbetween these molecules is a general phe- Immunohistochemical studies indicateit has a role in head re- nomenon. The discovery that mouse BMP-4 can substitute for generation and formation of tentacle batterycells, and that it dpp in Drosophila supports this contention (Padgett et al., degrades extracellular matrix proteinsof the developing tentacle. 1993). There is evidence that the BMP-1 variants andtolloids inter- Tolloid-related-I (DrTlr-1; also called tolkin) is also impli- act with and additionally activateTGF-(3 (transforming growth cated in activating growth factors but not dpp, because Tlr-I factor-P)-like molecules. Human BMP-I was originally discov- cannot substitute fortolloid. Tlr-1 is an essential gene; null al- ered in a complex as oneof three proteins (BMP-1-3) isolated leles of Tlr-l result in death in the larval stage (Nguyen et al., The astacin family 1251

Table 1. Tissue distribution and proposed functions of astacin family members " - ~~~~ ~ ~- ~ ~~ ~~ ~~ ~~~

Member Species/tissue Natural substrates (suggested) Functions (suggested)

~ ~______~~ ~ Astacin Crayfish stomach Ingested proteins Digestion of food Choriolysin H (HCEl and 2) Fish hatching gland cells Hard chorion, egg envelope Embryo hatching Choriolysin L (LCE) Fish hatching gland cells HCE-digested chorion peptides Embryo hatching CAM-1 Quail embryos Extraembryonic chorioallantoic Releases Ca2+ from eggshell protein; membrane hatching HMPl Hydra tentacles Extracellular matrix proteins; Pattern formation; formation of fibronectin, type 1 gelatin tentacle cells SPAN and BPlO Sea urchin; early stages of Latent univin Early developmental decisions embryogenesis BMP- 1 Animal embryos; Latent TGF-/3 growth factors Pattern formation; biomineralization; mouse and human tissues bonekartilage formation Tolloid Drosophila embryo; mouse Latent decapentaplegic (DPP) Dorsal/ventral patterning; early embryo and human embryo development Toolloid-related-I Drosophila embryo Latent growth factors Larval and pupal development Meprin A (hetero-oligomers Mouse (strain-specific) Urinary peptides; PTH; MSH; Urinary peptide processing; extracellular of 01 and p; homo- kidney; rat and human bradykinin; extracellular protein processing and digestion oligomers of CY subunits) kidney and intestine; matrix proteins embryos Meprin B (homo-oligomers Mouse (strain-specific) Extracellular proteins Extracellular protein interactions; of p subunits) kidney and intestine; remodeling embryonic

1994). The protease,first two CUB, and firstEGF-like domains the International Union of Biochemistry and Molecular Biology of Tld and Tlr-I are 62% identical, but they have some differ- (Webb, 1992), has also been referred to as endopeptidase-2 and ences, for example, in the prosequence and in inserted sequences endopeptidase 24.18 in the rat, and PPH in humans. Meprins in the first CUB domain (7 amino acids for Tlr-I, 26 amino acids are present at high concentrations in brush border membranes, for Tld) (see Fig. 1). It has been suggested that the different specialized epithelial cell plasma membranes, of mammalian kid- CUB and EGF-like domains inTld and Tlr-I result in different ney and intestine, and have also been detected in salivary glands interactions with proteins and make them functionally differ- and thyroid by immunohistochemical techniques (Craig et al., ent (R.W. Padgett, pers. comm.). 1987, 1991; Barnes et al., 1989). The mRNAs for the meprina Mammalian tolloids and BMP-I areexpressed at high levels and @ subunits are abundantin kidney and intestinaltissue, but in developing bones of mouse embryos andin placenta (Taka- have not been detected in many other mature mammalian tis- hara et al., 1994). In situ hybridization demonstrated high lev- sues, such as liver,muscle, brain, spleen, and lung (Gorbea els of tolloid transcripts, but not BMP-I, in the floor plate of et al., 1993; Jiang et al., 1993). Meprins have also been detected the neural tube of thedeveloping nervous system. Low levels of on the apical surface of neuroepithelial cells in the inner ear, both tolloid and BMP-I were found in adult human tissues such nasal conchae, ependymal layer of the brain ventricles, and as heart, liver, placenta, lung, skeletal muscle, kidney, and pan- choroid plexus in rat embryos (Spencer-Dene et al., 1994). Sug- creas; mammalian tolloid, but not BMP-I, was detected in adult gested functions for meprins in mature systems include degra- brain. A third tolloid isoform was not detectedin adult human dation of polypeptides such as parathyroid hormone (PTH), tissues but was found in trophoblast giant cells in mouse pla- a-melanocyte-stimulating hormone (MSH), bradykinin,lulibe- centa with the other isoforms. rin (LHRH), substance P, and TGF-a (Stephenson & Kenny, Sea urchin (Su) BPlO and SPAN are expressed in very early 1988; Choudry & Kenny, 1991; Wolz et al., 1991; Yamaguchi blastula stages, andit has been suggested they play a role indif- et al., 1991). In addition, meprins are proposed to degradeex- ferentiation of ectodermal lineages and subsequent patterning tracellular matrix proteinssuch as type IV collagen, laminin, fi- (Lepage et al.,1992b; Reynolds et al., 1992). SPAN expression bronectin, and gelatin (Kaushal et al., 1994). Meprins have also parallels univin, a member of the TGF-0 superfamily; this pairbeen suggested to process urinary peptides and proteinsin ro- may be the equivalent oftolloid and dpp in Drosophila (Sten- dents and thereby affect release of pheromones and behavior zel et al., 1994). SuBMP is expressed maximally at the hatched (Bond & Beynon, 1986; Beynon et al., 1995). blastula stage, with modest decreases at later stages of develop- ment (Hwang et al., 1994). It is detected in ectodermal and primary Primary structure of the catalytic domain mesenchyme cells just before theonset of primitive skeleton for- mation and is suggested to function similarly to HuBMP-I in The primary structureof astacin was determined by amino acid bone formation. sequencing; all the other amino acid sequences were deduced Meprin subunits form oligomers thatresult in meprin A (EC from the cDNA sequenceswith some confirmatory peptidese- 3.4.24.18; hetero-oligomers of a and 0 subunits, or homo- quencing (Fig. 2). The signature sequence for the family spans oligomers of a subunits) or meprin B (EC 3.4.24.63; a homo- from astacin His92 to Asp 109 (Fig. 2, positions 101-1 18); this oligomer of 6 subunits). Meprin A, the name recommendedby sequence contains the three imidazole-zinc ligands. There are 1252 J.S. Bond and R.J. Beynon four conserved cysteine residues that are known to form intra-(1993) to be involved in internal bonds maybe particularly im- domainsulfhydryl bridges in astacin: Cys 42/Cys 198 and portant to thefolding and stability of the bacterial enzyme. The Cys 64/Cys 84 (in Fig. 2 these are positions 43/209or 21 1 and enzyme contains an unusual 0-linked oligosaccharide, andini- 72/93). Because of the conservation of cysteine residues in all tial experiments indicate a preference for cleavage of peptides members of the family,it is likely that equivalent disulfide link- with aspartic acid in the P1’ position. Most significantly, the ages exist in the protease domains of the otherfamily members. identification of flavastacin demonstrates thatthis family was Another conserved region, SXMHY(astacin Ser 145 to Tyr149; present before the divergence of prokaryotes and eukaryotes. Fig. 2, positions 156-160), contains the tyrosine residue involved in zinc binding and the methionineinvolved in a “Met-turn” of Tertiary structure of the protease domain thepeptide chain (Jiang & Bond, 1992; Bode et al., 1993; Stocker et al., 1995). Stocker and colleagues (1993) identified The X-ray crystal structureof astacin has been solved to 1.8 A resolution (Bode et al., 1992). Astacin has a compact bilobal other specific residues that are conserved,in comparisons of 10 structure with long, deep active-site cleft that divides it into members of the astacin family, and are probably crucial to the a overall structure of the protease domain. Comparisonof the 22 two parts (Fig. 4). The NH,-terminal 100 amino acids are organized five-strandedpleated @sheets and two long family members herein indicates that there are28 residues that into are totally conserved. Seventeen of those are accounted for a-helices. One of the helices forms the top partof the active site within the ]&residue family signature sequence, the sequence of astacin and includes the zinc ligands His 92, His 96, and His 102 (positions 101, 105, and 11 1 in Fig. 2). The catalytic zinc containing thezinc-binding Tyr, andby the conserved cysteines. ion is at the bottom of the cleft, and Tyr149 (position 160) and The other 1 I conserved residues are primarily involved in inter- a water molecule (SOL 300) that also interactswith Glu 93 (po- nal bonds in astacin. These include (astacin residue numbers sition 102) are the fourth and fifth ligands for zinc. The trigo- with Fig.2 positionnumbers in parentheses): Pro 16 (16), nal bipyramid or pentacoordination of zinc (Fig. 5) is also found Phe 45 (46), Gly 63 (71), Ser 66 (74), Gly 69 (77), Gln 75 (84), in the serralysins and is quite different from thatof the thermo- Phe 128 (137), Tyr 141 (152), Phe 154(165), Asp 186(199), and lysin family, in which two histidines and a glutamic acid resi- Tyr 194 (207). due areligands to theactive-site zinc ion. The zinc coordination The amino acid sequences of the family member protease of astacin is discussed in detail by Gomis-Ruth et al. (1994) and domainsare 29-99% identical(Fig. 3A). The fish enzymes Stocker et al. (1995). HCEl and HCE2 are products of two differentgenes that are The COOH-terminal half of astacin has little defined second- highly conserved, i.e., 95% identical (Yasumasu et al., 1994). ary structure except for a three-turn a-helix before a disulfide The domains that are over 80% identical often represent spe- bond, which ends the domain and connects it to the NH2- cies differences for the samegene products (e.g., human, mouse, terminal half. The “Met-turn” at astacinMet 147 (Fig. 2, posi- and Xenopus BMP- I protease domains are 93-99% identical; tion 158) has been identified in crystal structures of adamalysins, mouse,rat, human meprin-as are 84-90‘70 identical,and serralysins, and matrixins, and these have been classified col- meprin-ps are 89% identical). lectively as “metzincins” (Stocker et al., 1995). The phylogenetic tree based on theprotease domains (Fig. 3B) indicates several clusters or branches in the family. The meprins form one branch, SPAN and BPI0 from sea urchins form an- Oligomeric structure and membrane association other, BMP-I and tolloids form a third branch, and the fish Although meprin subunits are the only astacinfamily members enzymes (HCEI, HCE2, LCE) and QuCAM-I form a fourth that are known to form homo- andhetero-oligomers and to con- branch. Phylogenetic trees comparing the EGF-like domainsof tain putative COOH-terminal transmembrane domains, there these enzymes yield the same branching picture for the appro- are others thatassociate with membranes and macromolecules. priate members, indicating that the protease and EGF-like do- For example, HCE is tightlyassociated with the chorion mains evolved together (Bond & Jiang, 1995). The finding that (Yasumasu et al., 1989~). Sea urchin BMPis tightly associated meprin a and subunits are approximately 50% identical in the with embryo plasma membranes; however, the interaction could same species, but each subunit in the mouse, rat, and humanis be disrupted by carbonate salts (Hwanget al., 1994). The orig- approximately 85% identical, implies that the gene duplication inal finding that human BMP-Iis present in a “complex” with that resulted in a and p preceded speciation. growth factor BMPs alsoimplies somewhat stable interactions Until recently, astacin family members have only been de- of the protease with the growth factors. Thenoncovalent inter- tected in the animal kingdom. However,a member of the fam- actions of astacin family members with membranes and com- ily (flavastacin)has now been identified in Flavobacterium plexes may serve to restrict movement of the active proteases and meningosepticurn (Tarentino et al., 1995). The gene for flav- concentrate activity at specific sites. Investigations of the fac- astacin hasbeen cloned and sequenced and theenzyme isolated. tors that determine these interactions may have relevance to The amino acid sequence of the protease domainis 18% iden- other proteolytic systems that function at thecell surface, such tical with the astacin sequence and24-29% identical with me- as complement proteases, plasminogen activators, and matrix prin subunit sequences. The signature sequence is conserved metalloproteases (Vassalli et al., 1991; Sim et al., 1993; Strongin (except for a substitution of an Ile fora Phe at position109 of et al., 1995). Fig. 2), the Met of the “Met-turn” and the zinc-binding Tyr are The quaternary structureof meprin A and its membrane as- conserved, and 9 of the 11 other conserved residues thought sociation are unique for known proteases and for brush border to be involved in internal bonds in the family are present in enzymes (Fig. 6). Investigations of meprinsA and B have been flavastacin. Interestingly, flavastacin lacks the four conserved aided by the fact that these enzymes arehighly expressed in ro- cysteine residues of the other astacinfamily members. The con- dent kidney; thea subunit, for example, hasbeen estimated to servation of most of the residues identified by Stocker et al. comprise 5% of the mouse brush border membrane proteinin 1253 The astacin family

Ce5D11.6 Ce7D10.4 HCEl HCE2 LCE QUCAM-1 HMPl Astacin MUBMP-1 DrTlr- 1 DrTld SuBMP XeBMP-1 HUBMP-1 SPAN BP10 MuMep-a RtMep-a HuMep-a MuMep-P RtMep-P

B MuMep-a HuMeD-a

SPAN I' BPI 0 MUBMP-1 Fig. 3. Comparisons of the sequences d HuBMP-I in the protease domains. Analyses were XeBMP-1 produced by Megalign(DNASTAR) SuBMP based on the alignment in Figure 2. A: DrTlr-1 Percent amino acid sequence identities. DrTld B: Phylogenetic tree based on sequence Astacin alignments.Abbreviations as in Fig- HCEl ure 2. HCE2 LCE QUCAM-1 Ce5D11.6 Ge7D10.4

50 40 30 20 10 0 Distance (substitution events)

strains that express it (Bond & Beynon, 1986). In addition, some Purified preparations of meprins inbred strainsof mice express both cy and 0subunits in the kid- ney, whereas others express only 0,providing good sources of Papain digestion of brush border membranes does not affect the rneprin A and B, respectively. Kidney brush border membranes size of the mouse or rat meprincy subunit, but decreases the size and papain-solubilized forms of meprins have been studied using of the (I subunit from 110 to 90 kDa (mouse) or from 80 to ICR and BALB/c mice and Sprague-Dawley rats for meprin A 74 kDa (rat) by cleaving the subunit near the end of the X and C3H/Heand micemeprin for B. domainthusand releasing enzymethe membranefromthe 1254 J.S. Bond and R.J. Beynon

Fig. 4. Ribbon structure of astacin. Images were pro- duced from the PDB file lAST (Bode et al., 1992) using the program SETOR (Evans, 1993). This view looks into theactive site cleft, which runs from left to right in the image. Catalytic zinc atom is represented by a green sphere.

(Marchand et al., 1994). Gel filtration and electrophoresis ex- ments with papain-purified preparations of mouse meprin A periments of papain-solubilized meprin A and B in the presence have now established that this enzyme consists of tetramers, di- and absence of reducing agents indicated that these enzymes are mers of disulfide-linked dimers. Sedimentation equilibrium anal- disulfide-linked oligomers. Analyticalcentrifugation experi- yses, for example, indicated the following molecular weights for

Fig. 5. Zinc coordination in the astacins. Zinc atom (magenta) is complexed by three histidine residues (His 92,96, and 102 of astacin; positions 101, 105,and 11 1 of Fig. 2), a glutamate residue (Glu 93; position 102) through a water molecule (SOL 300), and a tyro- sine residue (Tyr 149; position 160) from the “Met- turn.” Glu 103 (position 112)is oriented by a salt bridge via SOL 501 to the NH2-terminal residue; in the proenzyme forms, this orientation is not possible and the enzymes are inactive. I255

membrane-bound meprin A Secreted meprin A membrane-bound meprin B I

” W W ww ww

Fig. 6. Slcmbranc-associatcd and sccrcted forms of Incprin. Mcprin subunits are shown schematically: CY subunits with open domain \tructurc\ and 13 subunit\ \vith shadctl structures. Intersubunit disulfide bonds are indicated by horizontal lines between S1A51 tlomain\.

mouse meprin A: in phosphate buffer. 360,000 +. 4,000; dena- bridges either directly bridge the subunit to membrane-bound tured in 4 h4 guanidine-HCI, 167,000 800; performic acid os- (3 or form intrasubunit conformational units(e.g.. in the MAM idized and denatured, 86,OOO * 1,100. Thus, the denatured single domain) that adhere to membrane-bound proteins. Meprin B (S-S bridges disrupted) subunits are86 kDa; if disulfide bridges is also tetrameric, and the data indicate that thisis a dimer of are not disrupted, the subunits are covalentlylinked as dimers; S-S-bridged 0 homodimers. and the disulfide-linked dimers associate noncovalently to form Studies with transfected meprin cDNAshave led to a similar tetramers under nondenaturing conditions. picture of meprin organization (Milhiet et al., 1994). When full- length meprin CY transcripts were transfected into COS-I, 293, or MDCK (Madin Darby canine kidney) cells, meprin CY sub- Menlhrrrne a~s(.soc.iatiotl/sec.retionqf rveprins units were secreted, and very little remained associated with cells Studies with brush border membranes have also indicated an (Corbeil et al., 1993; Dumermuth et al., 1993). In contrast, oligomeric structure for meprins in vivo. SDS-gel electropho- transfection of full-length (3 subunit transcripts led to plasma resis and native gel analyses indicate that disulfide-linked CYCY membrane-bound (3 subunits (Johnson & Hersh, 1994). When homodimers and cui3 heterodimers are the predominant cova- meprin CY and /3 transcripts were cotransfected, both were asso- lently linked dimers in ICR kidney membranes in vivo (they ciated with cells a! the cell membrane. This typeof data is con- then associate to tetrameric forms of n2P2and ~,/3~),and that sistent with the model for meprins (Fig. 6). In contrast to the C3H/He membranes contain only /3[3 dimers and tetramers situation in vivo, however, homodimers of CY subunits were not (Gorbea et al., 1991; hlarchand et al., 1994). The meprin CYCY detected in the media or membrane of recombinant systems homodimers can be released from membranes by 7 M urea, but when meprin CY and 0transcripts were cotransfected. In mouse ab heterodimers and /3/3 homodimers remain membrane asso- kidney, the CY subunit is synthesized at twice the rate of the b’ sub- ciated, indicating that the latter are intesral membrane proteins, unit, which may well result in an excess of (Y subunits (Hall whereas (YCY homo-oligomers arenoncovalently associated with et al., 1993). The differential rate of synthesis of the two sub- the membrane (Fig. 6). Specific antibodies to COOH-teminal units explains the observation of a3P,forms of meprin in the domains of the subunits indicated that brush border forms of kidney membranes and a4 forms in the urine of mice. (x, but not 13, are truncated before the EGF-like domain. More Thus, although the cDNA sequence of the meprin CY sub- than 9OWo of the meprin CY subunits can be released from the unit contains a COOH-terminal domain that would predict a membrane by reducing agents such as 0-mercaptoethanol. In membrane-spanning anchor, the mature a subunit is truncated contrast, brush border membrane meprin /3 cannot be removed at the end of the X domain in kidney and in mammalian recom- from membranes by reducing agents. This indicates that CY as- binant systems. Transfection and mutagenesis studies of the sociation with membranes is dependent on disulfide bridges; S-S CY subunit have shown that COOH-terminal proteolytic process- 1256 J.S. Bond and R.J. Beynon ing is determined by the I domain (Figs. 1,7). Removal of this ER retention information (P. Marchand & J.S. Bond, unpubl. domain by genetic engineering results in no COOH-terminal work). The meprin a subunit TM domain, but not the p TM do- processing of meprin a,and notransport of the subunit to the main, contains repeating glycine residues, like the TM domain cell surface; the subunit remains inthe endoplasmic reticulum of class I1 molecules of the major histocompatibility complex (Marchand et al., 1995). This is in contrast to the meprin p sub- (MHC) (Cosson & Bonifacino, 1992). The glycines could form unit, which is not COOH-terminally processed in vivo, and is a surface in the transmembrane helix capable of interacting with found asa type I integral membrane protein localized to the cell other proteins. For meprin a,the interaction of the TM with ER surface. However, if the I domain is introduced into thep sub- membrane proteins may be responsible for retention in the ER. unit in chimeric cDNA, then @ is proteolytically processed and It is not known whether the oligomeric nature of meprins in- secreted into the medium. fluences proteolytic activity, but no allosteric effects have been The fact that removal of the I domain from meprin a results observed. Milhiet etal. (1994) were able to show that both a and in retention of the subunit in the ER, indicates that theCOOH- p subunits contribute to activity in hetero-oligomeric meprins. terminal domains of a contain a retention signal. Recent exper- Replacing rat meprin-a Glu 157 (position 102 of Fig. 2; the cat- iments indicate that it is the a TM domain that contains the alytic residueof the HEXXHmotif) by a Val, using site-directed

A. a subunit (wild-type)

ER Lumen Golgi Cell Surface

B. fl subunit (wild-type)

c. a subunit with theI domain deleted

D. fl subunit containing theI domain of a

Fig. 7. Diagrammatic representationof the biosynthesis of meprin subunits. Meprin-a subunits (red) transfected into COS-1 cells were secreted (A), whereas meprin-P subunits were transportedto the cell surface but remained membrane bound(B). The I domain of meprin-a is indicated by the egg-shaped structure between the bulkof the protein (rectangle containingthe pro- protease-MAM-X domains) and the EGF-like-TM-cytoplasmicdomains. If the I domain is deleted from meprin-a, the protein does not leave the ER (C). If the I domain is inserted into the 0 transcript, then the proteolytically truncated 0 subunit is secreted into the medium (D). ER, endoplasmic reticulum. The astacin family 1257 mutagenesis, abolished a enzymatic activity but only decreased ducing agents probably inhibitby chelating metals or reducing activity of (YO hetero-oligomers by 40%. disulfide bonds in the enzymes. No inhibition is observed by It is clear that meprins, like many other membrane pro- tissue inhibitors of metalloendopeptidases (TIMPs) norby phos- teins, have soluble counterparts (Beynon et al., 1995). Other phoramidon (rhamnosephosphate-Leu-Trp; a substituted dipep- examples include tumor necrosis factor receptors, TGF-a, tide isolated from Streptomyces that inhibits thermolysin-like ELAM-Iadhesion molecules, immunoglobulins, and angio- metalloendopeptidases). Astacin family enzymes are not inhib- tensin I-converting enzyme (peptidyl-dipeptidase A) (Ehlers& ited by inhibitors of serine, cysteine, or aspartic proteinases (e.g., Riordan, 1991). The subcellular site of the COOH-terminal PMSF, 3,4-dichloroisocoumarin, iodoacetate, E-64, leupeptin, processing for meprins is the endoplasmic reticulum, but the pepstatin), or by angiotensin I-converting enzyme inhibitors such proteases responsible for processing are unknown. Proteolytic as captoprilor thiorphan. Thebest inhibitor found for meprin processing of the meprin 01 subunit is not autolytic and is not A thus far is actinonin, a peptide hydroxamate with an inhibi- due to furin-typeenzymes (Marchand et al., 1995). There is ev- tion binding constant of about1 pM (Wolz& Bond, 1995). Ac- idence for an active proteolyticsystem in the ER becausemis- tinonin also inhibits crayfish astacin, but it is 1,000-fold less folded or incompletely assembled proteins arerapidly degraded, effective against this enzyme compared to mouse meprin A. and antigen processingis proposed to occurin this subcellular compartment. Investigations of meprin processing may be use- Substrates ful for identifying ER proteasesinvolved in limited proteolysis. A variety of peptide and protein substrates have been used to assay astacin family members (Wolz & Bond, 1995). Some of Enzymatic activities the more commonlyused protein substrates include: azocasein/ Our knowledge of astacin family members as enzymesderives casein, gelatin, and iodinated oxidized insulin B chain. The from the investigations of those that havebeen purified (asta- types of synthetic peptides that have been used as substrates cin, HCEl and2, LCE, HMPI,meprins). It should be clear that are: dansyl-Pro-Lys-Arg-Ala-Pro-Trp-Val(astacin cleaves at the CUB-containing proteins have not yet been isolated but the Arg-Ala bond) and nitrobradykinin (Arg-Pro-Pro-Gly- rather have been studied at the mRNA and DNA level. Some Phe(N0,)-Ser-Pro-Phe-Arg) (astacin and meprin A cleave at generalities about enzymatic propertiesof the family are sum- the Phe(N02)-Ser bond). Peptide amideshave also been used marized in the next several paragraphs. assubstrates: succinyl-Leu-Leu-Val-Tyr-4-methylcoumaryl- 7-amide(MCA) for rat meprin (Yamaguchi et al.,1991), for human meprin (Sterchi pH optima N-benzoyl-L-Tyr-p-aminobenzoate et al., 1988), Glu-Ala-Ala-Phe-MCA,succinyl-Ala-Ala-Ala- Astacin family enzymes are generally maximally active in the MCA, and succinyl-Ala-Ala-Val-Ala-p-nitroanilide(Johnson & neutral to alkaline pHrange. Optimal pH values range from 7.5 Hersh, 1994). Cleavage of the latter substrates is at thearylamide to 9.5 and may be partially determined by the substrate. bond, an amide bond but not a peptide bond. This is notable because, although arylamidase activity is common for serine proteases, it is not usually exhibited by metalloendopeptidases. Metal requirements In general, a wide variety of peptide bonds canbe cleaved in The enzymes can be inhibitedby metal chelatorssuch as EDTA peptides and proteinsby astacin family members; these enzymes and 1,lO-phenanthroline. Zinc has been established as the ac- are neither sequence specific nor highly specific for residues tive site metal in meprin A (1.1 mol zinchol subunit), astacin flanking the scissile bond. For example, the oxidized insulin B (0.97 k 0.03 mol zinc/mol protein), and HCEILCE (approxi- chain has been used to determine the peptide bond specificity mately 1.3 pg/mg protein) (Butler et al., 1987; Stocker et al., of mouse meprin A and B. Meprin A cleaved 10 bonds; 3 ma- 1988; Yasumasu et al., 1989a, 1989b). Reactivation studies have jor cleavages were Gly 20-Glu 21, Phe 24-Phe 25, Phe 25- indicated that Zn2+, Cu2+, and Co2+ reactivatecan astacin apo- yr 26. Meprin B cleaved 4 bonds; one, Cys(S0;) 19-Gly 20, enzyme; Zn2+, and Cu2+ can reactivate the fish enzymes and was distinct, and the others were common for both enzymes meprin A apoenzyme (Wolz & Bond, 1995). Calcium ions have (His 5-Leu 6, Leu 6-Cys(SO,) 7, Ala 14-Leu 15). This pattern also been found tobe associated with meprinA (2.75 mol/mol of cleavage indicated a preference for bonds flankedby neutral subunit) and the fish enzymes(3-5 pg/mg protein). The func- or hydrophobic amino acid residues, but negative charges in the tion of calciumis unknown, but it has been suggested to stabi- flanking residues can also be accommodated. Another substrate lize the enzymes. The ability of chelators to inhibit activity may is parathyroid hormone (PTH),which has been suggested as a vary for the enzymes. For example, meprins can be totally in- natural substrate forkidney meprin A in vivo (Yamaguchi et al., hibited by incubation with 1-10 mM EDTA for15 min, whereas 1993). Many bonds arecleaved by the rat enzyme in hPTH(38- astacin must be incubated for days with this chelator to ob- 84), including: Gly 68-Glu 69, Arg44-Asp 45, Asp 56-Asn 57, serve inhibition; inactivation of meprins and astacinby 2 mM Glu 55-Asp 56, and Leu 67-Gly 68. These studies indicate a I, 10-phenanthroline, however, can occur afterminutes (Stocker preference for cleavage at bonds containing polaror small neu- et al., 1988). tral aminoacids, not at bonds flankedby hydrophobic residues. Taken together, these data indicate thathydrolysis can occur at Inhibitors a variety of flanking amino acids. Mapping the active sites of astacin and meprinA revealed an There areno known specific natural inhibitors of astacin fam- extended substrate binding region accommodating seven or more ily members. Reducing agents such ascysteine and glutathione amino acid residues (Stocker et al., 1990; Wolz et al., 1991). As- can inhibit in vitro at concentrations 1 of mM and greater.Re- tacin has a distinct preference for alanine in the PI’ position, 1258 J.S. Bond and R.J. Beynon although other residues with short aliphaticside chains such as present at thecell surface predominantly as proforms, and the serine, threonine, andglycine are also accommodated in this po- enzyme is inactive against proteins in this form (Gorbea et al., sition. The crystal structure of astacinrevealed that this is due 1993). This allows for a high concentrationof a proteolytic en- to thepresence of Pro 176 (position 188 on Fig. 2) at the SI’ site, zyme that can beactivated at the membrane. The maturea sub- which the P1’ residue must align with (Stocker etal., 1993). Me- unit, but not the mature 6 subunit, of meprin A from ICR or prin A can accommodate many otherresidues in PI’, including C57BL/6 mousekidney is largely activated (i.e., the prosequence bulky hydrophobic residues, due to the substitutionof Gly 176 is removed). Thus, there is differential proteolytic processing of (position 190) and deletion of Tyr 177 (position 189). These the NH2-terminus, aswell as the COOH-terminus,of the a and changes result in a more open configurationof meprin compared p subunits in mouse kidney. The enzymes responsible for NH2- to astacin. The activities of both enzymes are greatly affected terminal processing, and therefore activation, and the subcel- by residues three and four amino acids away from cleavage the lular compartmentin which activation occurshave not yet been site, indicating that the extended amino acid chainof the sub- identified. Interestingly, secreted (urinary) forms of meprin A, strate interacts with the enzyme and influences binding and/or and transfected forms of mouse, rat,or human a or fi subunits turnover of the enzymehubstrate complex. Prolineresidues in expressed in COS-1, 293, or MDCK cells in culture, are all ex- positions two and three aminoacids away from thecleavage site pressed predominantly as inactive proenzymes, indicating they greatly affect activity, again demonstrating multiple interactions are notexposed to or are not vulnerable to the activation pro- of the enzyme subsites with substrate residues. teases. All the prosequences contain basic amino acids at the One area for future investigation that should be fruitful is COOH-terminus, although different motifs(such as single Lys molecular modeling of the interactions of substrates and inhib- or Arg residues, R/KXR, RRXR) are foundin different mem- itors with the active sites of astacin family members based on bers of the family. This implies that trypsin-like, furin-like,or the astacin structure. One question of interest to us, for exam- other processing enzymes that cleave at basic residues are in- ple, is why meprin A and B are so different in substrate speci- volved in activation. Removal of the prosequence is likely a ma- ficity. Meprin A hydrolyzes a-melanocyte-stimulating hormone jor formof posttranslational regulation of astacin family (kca,/K,r,= 1 X lo6 M” SKI) and many peptides and aryl- proteinase activity. amides, whereas meprin B is a very poor peptidase toward sub- strates of fewer than IO amino acids (Wolz & Bond, 1995). Interaction domains/CUB domains Examination of the amino acid sequences of the a and sub- units of meprins in the light of the three-dimensional structure It has been proposed that CUB domains promote multimeric of astacin (Stocker et al., 1993) indicates that there aresignifi- complex formation and aidin substrate recognition. For com- cant differences inthe sequences of the twomeprin subunits that plement proteases, there is evidence that Clr/s mediates calcium- could account for substrate specificity differences. For exam- dependent tetrameric complex formation between two Clr-Cls ple, the Phe of position 165 of Figure 2 has been identified as dimers and theassociation of this complexwith Clq to form the the putative S2’ binding site of astacin (residue 154), and theres- mature C1 molecule (e.g., Sim et al.,1993). Childs and O’Con- idues flanking this site are quite different in the a and 0 sub- nor (1994), studying Drosophila tolloid, found that loss-of- units; positions 162-167 are PFSFNK for a, and KTAFQ- for function mutations map to CUB repeats.By analogy with com- p. The S1’ binding site for astacin (residue175) corresponds to plement proteins, the CUB domainsin astacin family enzymes the area containing positions 187-189 on Figure 2; positions 186- could be interaction domains that promote multimeric com- 187 are N(T/A/S) fora and ED for/3 subunits. There aresev- plexes. The differences in functions of tolloid and tolloid- eral otherconsistent differences between the subunits thatcould related-I in Drosophila have been attributed to the differences explain activity differences. We anticipate that modeling predic- in amino acid sequences ofthe CUB andEGF-like domains that tions will eventually be tested experimentally by the production could result in specific interactions with different latent growth of correctly folded recombinant proteases alteredby site-directed factors (R. Padgett, pers. comm.). mutagenesis, and that this typeof study will aid in identifying physiological substrates and inhibitors for the enzymes. Genes- Localization, duplication, structure The gene structuresof the fish enzymes have been analyzed in Effects of noncatalytic domains on enzymatic activity 0. latipes (Yasumasu et al., 1994). There areeight copies of the HCE genes: five for HCEl and three for HCE2.Six (three of Acfivafion/prosequences each) of the genes form a cluster within approximately 25 kb of Astacin family enzymes are synthesized as inactive proenzymes, the genomic DNA; the two others are located separately. There and removal of the prosequences constitutes amajor mechanism are no intronsin the HCEgenes. In contrast, thereis only one for activation. The crystal structure of astacin indicates that copy of the LCE gene. It is within 3.6 kb and containseight ex- there is a salt bridge between the NH2-terminal amino acidof ons and seven introns. For LCE, the TATA box consensus se- the mature protease domain (Ala1) and Glu 103 (position 112 quence is 28 bp upstream from the transcription startsite. The of Fig. 2), which is next to the thirdzinc ligand (astacin residue 5’-flanking regions of all the HCE genes are 80-95% similar for His 102, position 11 1) near the active site. This bridge maybe 200-400 bp. There is no similarity between the HCE and LCE critical for enzymatic activity and, because the prosequence 5’ regions within 1.5 kb. These enzymes are both found in se- would prevent the formation of the salt bridge, it would inhibit cretory vesicles in fish and are bothnecessary to digest the egg activity. Indeed, there is substantial experimental evidence from envelope. Larger amountsof HCE relative to LCE protein are the meprins that supports this proposal. The p subunits of found in zymogen granules of the hatching gland cells, and it the mature form of meprinB from C3H/He mouse kidney are has been suggested that this reflects the larger gene copy num- The astacin family 1259 ber for HCE. The strong similarity ofHCE and LCE proteins important roles in growth, development, and remodeling, and indicates a common ancestral gene; if so, it is interesting that their relationship to healthy and diseased states is yet to be their gene structures have evolved to differ so markedly. investigated. The BMP-1 gene has been localized to the proximal half of mouse chromosome14 and to human chromosome 8p21(Ceci Acknowledgments et al., 1990; Yoshiura et al., 1993). There is keen interest in determining whether this is a candidate gene for diseases- This work was supported byNIH grant DK 19691 (J.S.B.) and by grants especially those that are caused by dysfunction of bone and from the BBSRC and theWellcome Trust (R.J.B.). We thank Dr. Petra Marchand for providing Figure 7 and for commentson the manuscript. cartilage. The structuralgenes for meprin subunits have been localized in the mouse and human genomes, and althoughit is likely the References a and p genes derived from a common ancestral gene, they have evolved to reside on different chromosomes (Gorbea etal., 1993; Barnes K, lngram J, Kenny AJ. 1989. Proteins of the kidney microvillar mem- Jiang et al., 1993; Bond et al., 1995). The structural gene for the brane. Structural andimmunochemical properties of rat endopeptidase-2 and its immunohistochemical localization in tissues of rat and mouse. meprin (Y subunit is on mouse chromosome17 and human chro- Biochem J 264:335-346. mosome 6p near the centromere; this geneis linked to the ma- Beckmann G, Bork P. 1993. An adhesive domain detected in functionally jor histocompatibility complex (MHC) in both genomes. The diverse receptors. Trends Biochem Sci 18:40-41. Beynon RJ, Oliver S, Stateva L, Evershed RP, Robertson D. 1995. Soluble structural gene for meprin @ is on chromosome 18 of both the forms of kidney meprin-a. In: Stocker W, Zwilling R, eds. The astacins- mouse and human (18q12.2-q12.3) genomes. The structural Structure and function of a new protein family. Forthcoming. genes and linkages to neighboring genes are conserved for mouse Bode W, Gomis-Ruth RF, Huber R, Zwilling R, Stocker W. 1992. Structure and human, and the localization of genes the in the twospecies of astacin and implications for activation of astacins and zinc-ligation of collagenases. Nature 358:164-167. sets the stage for the identification of the elements that deter- Bode W, Gomis-Ruth FX, Stocker W. 1993. Astacins, serralysins, snake mine the differences intissue-specific expression, as well as venom and matrix metalloproteinases exhibit identical zinc-binding en- association with disease genes. Regulation of expression is mark- vironments (HEXXHXXGXXH and Met-turn) and topologies and should be grouped into a common family, the “metzincins.” FEBS Lett edly different in tissues, species, and strains. For example, al- 33/:134-140. though both (Y and p are expressed in human intestine, only p Bond JS, Beynon RJ. 1986. Meprin:A membrane-bound metallo-endo- is expressed in mouse intestine; both are expressed at high lev- peptidase. Curr Topics Cell Regu128:263-290. Bond JS, Jiang W. 1995. Astacin family of metalloendopeptidases. In: NM els in kidney of many random-bred and inbred strainsof mice, Hooper, ed. Zinc metalloproteases in health and disease. London: Taylor but only p is expressed in kidney of some inbred mouse strains & Francis. (such as C3H/He and CBAmice); and both are expressed at low Bond JS, Rojas K, Overhauser J, Zoghbi HY, Jiang W. 1995. The structural levels in human kidney (Gorbea et al., 1993; Yamaguchi et al., genes MEPlA and MEPIB,for the a and 0 subunits of themetalloendo- peptidase meprin map to human chromosome 6p and18q, respectively. 1994). There must be some coordinateregulation of the two sub- Genomics 25:300-303. units because they are found ascomplexes, but at least @ can be Bork P, Beckmann G. 1993. The CUB domain: Awidespread module in de- expressed independently. velopmentally regulated proteins. J Mol Biol231539-545. Butler PE, McKay MJ, Bond JS. 1987. Characterizationof meprin, a membrane-bound metalloendopeptidase from mouse kidney. Biochem Prospects for the future J 241:229-235. Campbell LD, Bork P. 1993. Epidermal growth factor-like modules. Curr Astacin family members represent proteases that are widely dis- Opin Struct Biol3:385-392. Ceci JD, Kingsley DM, Silan CM, Copeland NG, JenkinsNA. 1990. An in- tributed in nature, highly regulated at the transcriptional and terspecific backcross linkage map of the proximal half of mouse chro- posttranslational levels, and active in specialized functions in de- mosome 14. Genomics 6:673-678. veloping and maturesystems. Exploration of the structure, func- Childs SR, O’Connor MB. 1994. Two domains of the tolloid protein con- tribute to its unusual genetic interaction with decapentaplegic. Dev Eiol tion, and regulation of these enzymesis at an early stage, and I62:209-220. we may anticipate futurerevelations as we have onlyuncovered Choudry Y, Kenny AJ. 1991. Hydrolysis of transforming growth factor-n the “tip of the iceberg” for this interesting family. by cell-surface peptidases in vitro. Biochem J 280:57-60. The three-dimensional structure of only one member of the Corbeil D, Milhiet PE, Simon V, Ingram J, Kenny AJ, Boileau G, Crine P. 1993. Rat endopeptidase-24.18 a subunit is secreted into the cultureme- family is known, and we expect that the structures of other dium as a zymogen when expressed by COS-I cells. FEBS Lett 335: members, or at least the protease domainsof members, will be 361-366. solved in the near future. It will be particularly interesting to Cosson P, Bonifacino JS. 1992. Role of transmembrane domain interactions in the assembly of class I1 MHC molecules. Science 258:659-661. compare the structure of flavastacin to astacin, tosee how the Craig SS, Mader C, Bond JS. 1991. Immunohistochemical localization of absence of sulfhydryl bridges affects the tertiary structure. It will the metalloproteinase meprin in salivary glands of male and female mice. be possible to produce substantial amountsof the developmen- J Histochem Cytochem 39:123-129. tal proteins, aswell as separate domains of the proteins,by re- Craig SS, Reckelhoff JF, Bond JS. 1987. Distribution of meprin in kidneys from mice with high- and low-meprin activity. Am JPhysi0/253:535-540. combinant techniques to study structure-functionrelationships. Dumermuth E, Eldering JA, Grunberg J, Jiang W, Sterchi EE. 1993. Clon- Investigations of these enzymes will no doubt contribute to ing of the PABA peptide hydrolase a subunit (PPHa) from human small our fundamental knowledge of the biosynthesis, trafficking, intestine and its expression in COS-I cells. FEES Lett 335:367-375. Dumermuth E, Sterchi EE, Jiang WP, Wok RL, Bond JS, Flannery AV, Bey- proteolytic processing, and regulation of cell surface and se- non RJ. 1991. The astacin family of metalloendopeptidases. JBiol Chem creted proteins and of proteolytic enzymes. 266:21381-21385, Finally, there is a great deal to learn about the natural sub- Ehlers MRW, Riordan JF. 1991. Membrane proteins with soluble counter- strates of astacin family members and about the interactions parts: Role of proteolysis in the release of transmembrane proteins,Bio- chemistry 30:10065-10074. that the proteins are involved in at the cell surface and in the Elaroussi MA, DeLuca HF. 1994. A new member to the astacin family of extracellular milieu. It is likely that these proteases will have metalloendopeptidases: A novel 1,25-dihydroxyvitamin D-3-stimulated 1260 J.S. Bond and R.J. Beynon

mRNA from chorioallantoic membraneof quail. Biochim Biophys Aria Milhiet PE, CorbeilD, Simon V, Kenny AJ, Crine P, Boileau G. 1994. Ex- 1217:l-8. pression of rat endopeptidase-24.18 in COS-Icells: Membrane topology Evans SV. 1993. SETOR: Hardware lighted three-dimensional solid model and activity. Biochem J 300:37-43. representations of macromolecules. J Mol Graphics 11:134-1 38. Neurath H. 1989. The diversity of proteolytic enzymes.In: Beynon RJ, Bond Ferguson EL, Anderson KV. 1992. Decapeniaplegic acts as a morphogen to JS, eds. Proleolyiic enzymes: A practical approach. Oxford, UK: IRL organizedorsal-ventral pattern in the Drosophila embryo. Cell 71: Press. 451-461. Nguyen T, Jamal J,Shimell MJ, Arora K, O’Connor MB.1994. Character- Fukagawa M, Suzuki N, HoganB, Jones CM. 1994. Embryonic expression ization of tolloid-related-I:A BMP-1-like product thatis required during of mouse bone morphogenetic protein-I (BMP-I), whichis related to the larval and pupal stagesof Drosophila development. Dev Biol166:569- Drosophila dorsoventral gene tolloid and encodes a putative astacin me- 586. talloendopeptidase. Dev Biol 163:175-183. Padgett RW, Wozney JM, Gelbart WM.1993. Human BMP sequences can Furie B, Furie BC. 1988. The molecular basis of blood coagulation. Cell confer normal dorsal-ventral patterning in the Drosophila embryo. Proc 53505-518. Nail Acad Sci USA 90:2905-2909. Gomis-Ruth FX, Grams F, Yiallouros I, Nar H, Kusthardt U, Zwilling R, Rawlings ND, Barrett AJ. 1993. Evolutionary families of peptidases. Bio- Bode W, Stocker W. 1994. Crystal structures, spectroscopic features, and chem J290:205-218. catalytic properties of cobalt (11), copper (11). nickel (I]), and mercury ReynoldsSD, Angerer LM, Palis J, Nasir A, Angerer RC. 1992. Early (11) derivatives of the zinc endopeptidase astacin. J Biol Chem 269: mRNAs, spatially restricted along the animal-vegetal axis of sea urchin 17111-17117. embryos, include one encoding a protein related tolloid to and BMP-I . Gorbea CM, Flannery AV, Bond JS. 1991. Homo- and heterotetrameric Development 114:769-786. forms of the membrane-bound metalloendopeptidases meprin A andB. Sato SM, Sargent TD. 1990. Molecular approach to dorsoanterior develop- Arch Biochem Biophys 290:549-553. ment in Xenopus laevis. Dev Biol 137:135-141. Gorbea CM, Marchand P, Jiang W, Copeland NG, Gilbert DJ, Jenkins NA, Shimell MJ, Ferguson EL, Childs ST, O’Connor MB. 1991. The Drosoph- Bond JS. 1993. Cloning, expression, and chromosomal localization of ila dorsal-ventral patterning gene iolloid is related to human bone mor- the mouse meprin 0 subunit. J Biol Chem 268:21035-21043. phogenetic protein-I. Cell 67:469-481. Hall JL, Sterchi EE, Bond JS. 1993. Biosynthesis and degradation of me- Sim RB, Day AJ, Moffatt BE, Fontaine M. 1993. Complement factor I and prins, kidney brush border proteinases. Arch Biochem Biophys 307: cofactors in control of complement system convertase enzymes.Meih- 73-77. ods Enzymol223: 13-35. Hwang S, Partin JS, Lennarz WJ.1994. Characterization of a homolog of Spencer-Dene B, Thorogood P, Nair S, Kenny AJ, Harris M, Henderson B. human bone morphogenetic protein I in the embryo of the sea urchin, 1994. Distribution of, and a putative role for, the cell-surface neutral Sirongylocenlroius purpuraius. Developmeni 120:559-568. metalloendopeptidases during mammalian craniofacial development. De- Jiang W, Bond JS. 1992. Families of metalloendopeptidases and their rela- velopmeni 120:3213-3226. tionships. FEES Lel~312:110-114. Steiner DF, Smeekens SP, OhagiS, Chan SJ. 1992. The new enzymology of Jiang W, Gorbea CM, FlanneryAV, Beynon RJ, Grant GA, Bond JS.1992. precursor processing endoproteases. J Biol Chem 267:23435-23438. The a subunit of meprin A. Molecular cloning and sequencing, differ- Stenzel P, Angerer LM, Smith BJ, Angerer RC, Vale WW. 1994. The univin ential expression in inbred mouse strains, and evidence for divergent evo- gene encodes a member of the transforming growth factor-beta super- lution of the a and 0 subunits. J Biol Chem 267:9185-9193. family with restricted expression in the sea urchin embryo. Dev Biol Jiang W, Sadler PM, Jenkins NA, Gilbert DJ, Copeland NG, Bond JS.1993. 166:149-158. Tissue-specific expression and chromosomal localization of the a sub- Stephenson SL, Kenny AJ. 1988. The metabolism of neuropeptides. Hydro- unit of mouse meprin A. JBiol Chem 268:10380-10385. lysis of peptides by the phosphoramidon-insensitive rat kidney enzyme Johnson GD, Hersh LB. 1992. Cloning a rat meprin cDNA reveals the en- “endopeptidase-2”and by ratmicrovillar membranes. Biochem J zyme is a heterodimer. JBiolChem 267:13505-13512. [Erratum in JBiol 255:45-5 I. Chem 268:17647.] Sterchi EE, Naim HY, Lentze MJ, Hauri HP, Fransen JA. 1988. N-benzoyl- Johnson GD, Hersh LB. 1994. Expression of meprin subunit precursors. L-tyrosyl-p-aminobenzoic acid hydrolase: A metalloendopeptidase of the Membrane anchoring through the0 subunit and mechanism of zymo- human intestinal microvillus membrane which degrades biologically ac- gen activation. J Biol Chem 269:7682-7688. tive peptides. Arch Biochem Biophys 265:105-118. Kaushal GP, Walker PD, Shah SV. 1994. An old enzyme with a new func- Stocker W, Gomis-Ruth FX, Bode W, Zwilling R. 1993. Implications of the tion: Purification and characterization of a distinct matrix-degrading me- three-dimensional structure of astacin for the structure and function of talloproteinase in rat kidney cortex and its identification as meprin. J the astacin family of zinc-endopeptidases.Eur J Biochem 214:215-231. Cell Biol 126:1319-1327. Stocker W. Grams F, Baumann U, Reinemer P, Gomis-Ruth FX,McKay DB, Kennelly PJ, Krebs EG. 1991, Consensus sequences of substrate specificity Bode W. 1995. The metzincins-Topological and sequential relations be- determinants for protein kinases and protein phosphatases.J Biol Chem tween the astacins, adamalysins, serralysins, and matrixins (collagenases) 266:15555-15558. define a superfamily of zinc-peptidases. Proiein Sci 4:823-840. Kitamoto Y, Yuan X, Wu Q, McCourt DW, Sadler JE. 1994. Enterokinase, Stocker W, Ng M, Auld DS. 1990. Fluorescent oligopeptide substrates for the initiatorof intestinal digestion, is a mosaic protease composed of a kinetic characterization of the specificityof Asiacus protease. Biochem- distinctive assortment of domains. Proc Nail Acad Sei USA 91:7588- isiry 29:10418-10425. 7592. Stocker W, Wolz RL, Zwilling R. 1988. Asiarus protease, a zinc metallo- Lee KS, Yasumasu S, Nomura K, luchi I. 1994. HCE, a constituent of the enzyme. Biochemistry 27:5026-5032. hatching enzymes of Oryzias latipes embryos, releases unique proline- Strongin AY, Collier I, Bannikov G, Marmer BL, Grant GA, Goldberg GI. rich polypeptides fromits natural substrate, the hardened chorion.FEES 1994. Mechanism of cell surface activationof 72-kDa type IV collagen- Leti 339:281-284. ase. J Biol Chem 270:5331-5338. Lepage T, Ghiglione C, Gasche C. 1992a. Spatial and temporal expression Takahara K, Lyons GE, Greenspan DS. 1994. Bone morphogenetic protein-I pattern during sea urchin embryogenesis of a gene coding for a prote- and a mammalian tolloid homologue (mTld) are encodedby alternatively ase homologous to the human protein BMP-1 to and the productof the spliced transcripts which are differentially expressed in some tissues.J Drosophila dorsal-ventral patterning gene tolloid. Development 114: Biol Chem 269:32572-32578. 147-164. Tarentino AL, Quinones G, Grimwood BG, Hauer CR, Plummer TH Jr. Lepage T, Sardet C, Cache C.1992b. Spatial expression of the hatching en- 1995. Molecular cloning and sequence analysis of flavastacin: An 0- zyme gene in the sea urchin embryo. Dev Biol 150:23-32. glycosylated prokaryotic zinc metalloendopeptidase. Arch Bochem Bio- Maeno M, XueY, Wood TI, Ong RC, Kung H.1993. Cloning and expres- phys 319:281-285. sion of cDNA encodingXenopus laevis bone morphogenetic protein-I Titani K, Torff HJ, Hormel S, Kumar S, Walsh KA, Rod1 J, Neurath H, during early embryonic development. Gene 134:257-261. Zwilling R. 1987. Amino acid sequence of a unique protease from the Marchand P, Tang J, Bond JS. 1994. Membrane association and oligomeric crayfish Asiacus fluviaiilis.Biochemistry 26:222-226. organization of thea and 0 subunits of mouse meprin A.J Biol Chem Vassalli JD, Sappino AP, Belin D. 1991. The plasminogen activator/plas- 269:15388-15393. min system. J CIin Invesi 88:1067-1072. Marchand P, Tang J, Johnson GD, Bond JS.1995. COOH-terminal proteo- Vogt G, Stocker W, Storch V, Zwilling R. 1989. Biosynthesis of Asiacus pro- lytic processing of secretedand membrane forms of theCY subunit of the tease a digestive enzyme from crayfish. Histochemistry 91:373-381. metalloprotease meprin A: Requirement of the I domain for processing Webb EC, ed. 1992. Enzymenomenclature. San Diego, California: Academic the endoplasmic reticulum. J Biol Chem 270:5449-5456. Press, Inc. p 41 1. Matrisian LM. 1992. The matrix-degrading metalloproteinases. BioEssays Wolz RL, Bond JS. 1995. Meprins A and B. MethodsEnzymol 248: 14:455-463. 325-345. The astacin family 1261

Wolz RL, Harris RB, Bond JS. 1991. Mapping the active site of meprin-A of low choriolytic enzyme (LCE), a component of the hatching enzyme with peptide substrates and inhibitors. Biochemistry 309488-8493. of the teleost, Oryzias latipes. J Biochem I05:212-218. Wozney JM, Rosen V, Celeste AJ, Mitsock LM, Whitters MJ, KrizRW, He- Yasumasu S, luchi I, Yamagami K. 1994. cDNAs and thegenes of HCE and wick RM, Wang EA. 1988. Novel regulators of bone formation: Molec- LCE, two constituents of the medaka hatching enzyme.Dev Growrh & ular clones and activities. Science242:1528-1531. Differ 36:241-250. Yamaguchi T, Fukase M, SugimotoT, Kid0 H, ChiharaK. 1994. Purifica- Yasumasu S, Katow S, Hamazaki TS, luchi 1, Yamagami K. 1992. Two con- tion of meprin from human kidney and its role in parathyroid hormone stituent proteases of a teleostean hatching enzyme: Concurrent synthe- degradation. Biol Chem Hoppe-Seyler 375:821-824. ses and packaging in the same secretory granulesin discrete arrangement. Yamaguchi T, Kid0 H, Fukase M, FujitaT, Katunuma N. 1991. A membrane- Dev Biol149:349-356. bound metalloendopeptidase from rat kidney hydrolyzing parathyroid Yasumasu S, Katow S, Umino Y, Iuchi I, Yamagami K. 1989c. A unique hormone. Purification and characterization.Eur J Biochem 200563-571. proteolytic action of HCE, a constituent proteaseof a fish hatching en- Yamaguchi T, Kid0 H, Kitazawa R, Kitazawa S, Fukase M, Katunuma N. zyme: Tight binding to its natural substrate, egg envelope. Biochem Bio- 1993. A membrane-bound metallo-endopeptidase from rat kidney: Its phys Res Commun 162:58-63. immunological characterization. Biochemisfry 113:299-303. Yoshiura K, Tamura T, Hong HS, OhtaT, Soejima H, KishinoT, Jinno Y, Yasumasu S, luchi I, Yamagami K. 1989a. Purification and partial charac- Niikawa N. 1993. Mapping of the bone morphogenetic protein I gene terization of high choriolytic enzyme(HCE), a component of the hatch- (BMPI) to 8~21:Removal of BMPI from candidacy for the bone dis- ing enzyme of the teleost, Oryzias latipes. J Biochem 105:204-211. order in Langer-Giedion syndrome. Cytogenet Cell Genet 64:208-209. Yasumasu S, luchi I, Yamagami K. 1989b. Isolation and some properties