Vol. 17, No. 1, pp. 7-30, January 1, 2003

REVIEW
The ADAMs family of metalloproteases: multidomain proteins with multiple functions

Van Andel Research Institute, Grand Rapids, Michigan 49503, USA

	Introduction

Top Introduction ADAMs are members of... Expression, subcellular... The shedding of cytokines... Growth factor and growth... The cleavage of other... The regulation and specificity... Cell fate determination and... Fertilization Myoblast fusion Human disease Conclusions References

The ADAMs family of transmembrane proteins belongs to the zinc protease superfamily. Members of the family have a modular design, characterized by the presence of metalloprotease and integrin receptor-binding activities, and a cytoplasmic domain that in many family members specifies binding sites for various signal transducing proteins. The ADAMs family has been implicated in the control of membrane fusion, cytokine and growth factor shedding, and cell migration, as well as processes such as muscle development, fertilization, and cell fate determination. Pathologies such as inflammation and cancer also involve ADAMs family members. Excellent reviews covering various facets of the ADAMs literature-base have been published over the years and we recommend their examination (Black and White 1998; Schlondorff and Blobel 1999; Primakoff and Myles 2000; Evans 2001; Kheradmand and Werb 2002). In this review, we will first discuss the properties of each of the domains of the ADAMs. We will then go on to describe the involvement of ADAMs in selected biological processes. Then, we will highlight recent interesting findings suggesting roles for ADAMs in human disease. Finally, we look to the future and discuss some of the open issues in ADAMs function and regulation.

	ADAMs are members of the zinc protease superfamily

Top Introduction ADAMs are members of... Expression, subcellular... The shedding of cytokines... Growth factor and growth... The cleavage of other... The regulation and specificity... Cell fate determination and... Fertilization Myoblast fusion Human disease Conclusions References

Zinc proteases are subdivided according to the primary structure of their catalytic sites and include gluzincin, metzincin, inuzincin, carboxypeptidase, and DD carboxypeptidase subgroups (Hooper 1994). The metzincin subgroup (to which the ADAMs belong) is further divided into serralysins, astacins, matrixins, and adamalysins (Stocker et al. 1995). The matrixins comprise the matrix metalloproteases, or MMPs. These enzymes are the principle agents responsible for extracellular matrix degradation and remodeling, and play important roles in development, wound healing, and in the pathology of diseases such as arthritis and cancer (Chang and Werb 2001). Adamalysins are similar to the matrixins in their metalloprotease domains, but contain a unique integrin receptor-binding disintegrin domain (Fig. 1). It is the presence of these two domains that give the ADAMs their name (a disintegrin and metalloprotease). The domain structure of the ADAMs consists of a prodomain, a metalloprotease domain, a disintegrin domain, a cysteine-rich domain, an EGF-like domain, a transmembrane domain, and a cytoplasmic tail. The adamalysins subfamily also contains the class III snake venom metalloproteases and the ADAM-TS family, which although similar to the ADAMs, can be distinguished structurally (Fig. 1).

View larger version (21K): [in this window] [in a new window]   Figure 1.   The topography of the ADAMs and related metalloproteases. Generalized domain structures of the ADAMs, SVMPs, ADAM-TS, and MMP families are shown. Note that ADAM-TS family members have a variable number of thrombospondin-like (TS) motifs. The MMP shown is of the gelatinase class. Other subclasses of MMPs lack hemopexin-like sequences and/or fibronectin type II-like sequences. The subclass of MT-MMPs have transmembrane domains and cytoplasmic tails in addition to the domains shown.

	Expression, subcellular location, and domain activity of the ADAMs family

Top Introduction ADAMs are members of... Expression, subcellular... The shedding of cytokines... Growth factor and growth... The cleavage of other... The regulation and specificity... Cell fate determination and... Fertilization Myoblast fusion Human disease Conclusions References

In humans, there are 19 adam genes, as shown in Table 1. In the literature, this family is often also referred to as the MDC family, indicating the presence of metalloprotease, disintegrin, and cysteine-rich domains. Furthermore, individual family members often have two or more names. For clarity, in this review, we will use the ADAM nomenclature for each mammalian family member. However, the alternative names of the individual ADAMs are also provided in Table 1. In total, there have been at least 34 adam genes described in a variety of species. Up-to-date registries of all ADAMs family members in different species can be found at: http://www.uta.fi/~loiika/ADAMs/HADAMs.htm, http://www.uta.fi/%7Eloiika/ADAMs/MMADAMs.htm, and http://www.people.virginia.edu/%7Ejw7g/Table_of_the_ADAMs.html.

Expression

ADAMs are found in vertebrates, as well as in Caenorhabditis elegans, Drosophila, and Xenopus. They are not present in Escherichia coli, Saccharomyces cerevisiae, or plants. The fission yeast Schizosaccharomyces pombe has what may be an early progenitor of the ADAMs family, although its properties have not been studied. Expression profiles of the ADAMs can vary con siderably. In mammals, many of them (including ADAMs 2, 7, 18, 20, 21, 29, and 30) are exclusively or predominantly expressed in the testis and/or associated structures. Other members (ADAMs 8, 9, 10, 11, 12, 15, 17, 19, 22, 23, 28, and 33) show a more broad somatic distribution. Yet, there are clear differences in the pattern of expression of these enzymes (Table 1). ADAMs 9, 12, and 19 were originally cloned from myoblasts (Yagami-Hiromasa et al. 1995), but have been shown subsequently to be more broadly expressed.

Several ADAMs are expressed in multiple splice forms (Table 1). For example, ADAM22 (Sagane et al. 1998), ADAM29 and ADAM30 (Cerretti et al. 1999) have two to three forms that vary in the lengths of their cytoplasmic tails, although no functional differences in these isoforms have been reported. In other cases, alternative splicing produces proteins with markedly different activity. For example, ADAM12 has two splice forms: One, called L, produces a membrane-bound protein, the other, called S, diverges just upstream of the transmembrane domain, which results in a shorter form that is secreted from the cell (Gilpin et al. 1998). Recent studies indicate that ADAM12-S has functional IGFBP-3 and IGFBP-5 protein cleavage activity (Shi et al. 2000). Because ADAM12 is overexpressed during pregnancy, it is possible that ADAM12-S is responsible for increasing the pool of IGF in the bloodstream during pregnancy through IGFBP proteolysis. Alternative splicing of adam28 yields isoforms with different subcellular localization patterns and tissue expression. Murine ADAM28 may have three forms, two larger ones predicted to encode membrane-anchored proteins and expressed in the epididymis and lung, as well as a smaller one predicted to encode a secreted protein with testis-specific expression (Howard et al. 2000). Meanwhile, human ADAM28 has two forms. In this case, the secreted form is preferentially expressed in the spleen, whereas the membrane-bound form is lymph node specific (Roberts et al. 1999). ADAM9 and ADAM10 are also alternatively spliced and have both secreted and membrane-associated forms (Yavari et al. 1998; Hotoda et al. 2002). Finally, there is evidence that adam11 and adam33 genes produce alternative spliced products (Katagiri et al. 1995; Van Eerdewegh et al. 2002).

To date, there have been few studies reporting the phenotypes of mice lacking individual ADAMs family members. However, ADAM2, ADAM3, ADAM9, ADAM10, ADAM17, and ADAM23-deficient mice have been described. Mice lacking either ADAM2 or ADAM3 are viable and healthy with normal development, although male mice are infertile (Cho et al. 1998; Nishimura et al. 2001). The molecular basis for the infertility will be discussed later. Despite the ubiquitous, and in certain tissues high expression of ADAM9, null mice develop normally, are viable and fertile, and do not show any major pathologies (Weskamp et al. 2002). This perhaps suggests redundancy with other members of the ADAMs family. Mice harboring a germ-line mutation in the metalloprotease domain of ADAM17 exhibit perinatal lethality (Black et al. 1997; Peschon et al. 1998). Those born alive usually die within several hours. The more obvious defects in newborns include open eyelids, stunted vibrissae, and wavy hair. Histological studies of mutant fetuses reveal defects in epithelial maturation and organization that impairs the development of the digestive, respiratory, and hormonal systems (Peschon et al. 1998). A gene-trapping analysis of murine genes involved in brain-wiring patterns revealed that homozygous deletion of ADAM23 results in tremor and ataxia (Leighton et al. 2001). There is a recent report describing the targeted disruption of the adam10 gene (Hartmann et al. 2002). ADAM10--deficient mice die by day 9.5 of embryogenesis with pronounced defects in the neural and cardiovascular systems. ADAM12, ADAM15, and ADAM19 null embryos have been created, but the phenotypes, if any, have yet to be reported (Schlondorff et al. 2001).

Subcellular location

The prevailing data indicate that ADAMs are probably synthesized in the rough endoplasmic reticulum and mature in a late Golgi compartment (Lum et al. 1998; Roghani et al. 1999; Hougaard et al. 2000; Howard et al. 2000; Schlondorff et al. 2000; Kang et al. 2002). Maturation involves the removal of the prodomain from the ADAM precursor protein, which is thought to make the ADAMs metalloprotease competent. However, the intracellular localization in which the recognition and cleavage of ADAMs substrates occurs is still a matter of debate. Where studied, it appears that the bulk of the ADAM protein resides in a region near the nucleus, in some cases colocalizing with Golgi markers (Lum et al. 1998; Lammich et al. 1999; Hougaard et al. 2000; Schlondorff et al. 2000; Kang et al. 2002). However, ADAMs can also be detected on the cell surface. In many cases (e.g., ADAMs 9, 10, 15, 17, and 28), the cell surface form appears to be processed and thus catalytically active (Black et al. 1997; Lum et al. 1998; Lammich et al. 1999; Roghani et al. 1999; Howard et al. 2000; Schlondorff et al. 2000). There is one report describing regulated transition from intracellular compartments to the cell surface. The alternatively spliced short form of ADAM12 (ADAM12-S) transits through the endomembrane system and is secreted, but the large form, ADAM12-L, is retained in the trans-Golgi network (Hougaard et al. 2000; Kadota et al. 2000). Removal of the cytoplasmic domain of ADAM12-L blocks retention and leads to cell surface accumulation, suggesting the presence of a retention signal within a region composed of the transmembrane domain and cytoplasmic tail (Hougaard et al. 2000). It has been speculated that an intracellular pool of ADAM12 could be released in a regulated manner for cell surface expression, and perhaps activity, by removing these constraints. Nevertheless, several ADAMs family members may be active intracellularly. For example, most of the mature form of ADAM15 is resistant to trypsinization treatment, indicating a predominantly intracellular pool (Lum et al. 1998; Howard et al. 2000). Furthermore, the metalloprotease activity of ADAMs 10, 17, and 19 can occur within intracellular compartments (Skovronsky et al. 2000; Shirakabe et al. 2001). Such differences in subcellular localization and activity may ultimately depend on the cell type, the ADAM, and the substrate involved.

The prodomain

The N terminus of ADAMs contains a signal sequence that directs ADAMs into the secretory pathway and a prodomain that functions in ADAMs maturation (Fig. 1). Primarily, the prodomain keeps the metalloprotease site of ADAMs inactive, through a cysteine switch (Van Wart and Birkedal-Hansen 1990; Becker et al. 1995). A conserved cysteine residue within the prodomain preferentially coordinates the required active site zinc atom, and thereby sequesters the metalloprotease domain in an inactive conformation. Pharmacological inhibitors of the early secretory pathway like brefeldin A and monensin block the processing of ADAM9 and ADAM15, thus positioning the location of ADAMs processing and activation at the trans-Golgi network (Lum et al. 1998; Roghani et al. 1999; Howard et al. 2000; Kang et al. 2002). This location is consistent with the localization of furin and other proprotein convertases (PCs; Nakayama 1997). PCs cleave the prodomain from the rest of the protein at a conserved Rx(R/K)R motif, effectively releasing the prodomain and switching the zinc coordination to the metalloprotease domain, thereby making it available for catalytic activity. Support for such a mechanism comes from a number of experimental studies.

Firstly, furin cleaves ADAM15 in vitro (Lum et al. 1998). Secondly, overexpression of PCs such as PC7 and furin increase the amount of processed ADAM10 in vivo, and processing is blocked by the addition of a peptide analog of the PC cleavage site (Anders et al. 2001). Thirdly, mutation of PC cleavage sites blocks the processing of ADAM10, ADAM12, and ADAM19 to their mature, active forms (Loechel et al. 1998; Anders et al. 2001; Kang et al. 2002). Together, these data indicate that, in vivo, cleavage of the prodomain is a prerequisite for the generation of an active protease. Further support for the model comes from the observation that, in vitro, the inhibitory effect of a PC cleavage site mutation is overcome by treatment with NEM, a sulfhydryl reactive compound that alkylates the cysteine residue, thereby switching the coordination of zinc to the active site of the metalloprotease and bypassing the cleavage dependence of protease activation (Loechel et al. 1999). Mutation of the cysteine residue in the prodomain of ADAM12 to an alanine or histidine also leads to protease activation independently of prodomain cleavage (Anders et al. 2001). Finally, the application of cysteine switch peptides to cells inhibits ADAM9 and ADAM17, presumably because the peptide competes for zinc coordination in trans (Roghani et al. 1999). It is likely that this mechanism of maturation and activation applies to most ADAMs metalloproteases. However, there are cases in which ADAMs may undergo autocatalytic activation. This is most clearly shown with ADAM8 and ADAM28, in which activity-blocking mutations in the metalloprotease domains produces only the precursor form of the protein in transfected cells (Howard et al. 2000; Schlomann et al. 2002).

The secondary function of the prodomain is to chaperone the proper folding of ADAMs, particularly the metalloprotease domain. This has been suggested by studies showing that the removal of the prodomain of ADAM17 generates a protease-inactive protein (Milla et al. 1999). Similarly, an ADAM10 construct lacking its prodomain is catalytically inactive in vivo. But cotransfection of this form together with a construct expressing just the prodomain of ADAM10 generates protease activity (Anders et al. 2001). Additionally, a form of ADAM12-S lacking a prodomain, unlike wild-type protein, is not secreted from the cell, but instead remains in the early endomembrane system (Loechel et al. 1999). Deletion of both the prodomain and the metalloprotease domain allows for secretion of the protein. Taken together, these data suggest that prodomain deletion constructs are synthesized in an inactive form because they are improperly folded during synthesis. Hence, the prodomain appears to be necessary for maintaining the latency of these enzymes, and it assists in the proper folding of ADAMs, in the structuring of the catalytic active site, and in the proper transit of ADAMs throughout the secretory pathway.

The metalloprotease domain

The crystallization of the metalloprotease domain of several metzincin family members, including ADAM17, has allowed the mechanism of proteolytic activity to be more accurately defined (Maskos et al. 1998). The active site contains zinc and water atoms that are necessary for the hydrolytic processing of protein substrates, and which are coordinated by three conserved histidine residues and a downstream methionine. The methionine lies in a Met turn motif that loops around to face the consensus HExxHxxGxxH site. There is remarkable conservation within this catalytic site among the various metzincins, however, individual proteins within the family have distinguishing structural features that may impart specificity for substrates and protease inhibitors (Stocker et al. 1995). On the basis of this structural definition, 12 of the human ADAMs are predicted to be proteolytically active (Table 1), although in only half of these cases has protease activity been measured.

Is there any selectivity to the protease activities of individual ADAMs? This has not yet been examined directly in vitro, although examination of the substrate specificities of some ADAMs in cell-based assays has suggested that this may be the case. Mixture-based oriented peptide libraries have been used to examine the cleavage site motifs for six members of the related MMP family (Turk et al. 2001). The data obtained suggest that MMPs do have some innate specificity for particular cleavage sites, although it is to be expected that features such as colocalization will also play a role in substrate selection. It will be important to test ADAMs specificity using such peptide libraries.

Inhibitors of ADAMs metalloprotease activity fall into four broad classes: those that inhibit by denaturation; those that inhibit by Zn-chelation; small molecule inhibitors of catalysis; and proteinacious inhibitors called TIMPs. The first two categories represent nonselective inhibitors such as reducing agents or zinc chelating agents. The third class arose from efforts to develop inhibitors of both MMPs and ADAMs, and comprise hydroxamate-based inhibitors that bind competitively to the active site. These have proved to be useful tools for studying ADAMs and MMPs (for review, see Moss et al. 2001). The crystal structure of ADAM17 bound to a compound called IC-3 suggests that hydroxamate inhibitors replace Zn-coordinating water molecules in the active site (Maskos et al. 1998). Other chemical features, such as a peptide chain that aligns along the side of the active site, and a hydrophobic moiety within an active site pocket, define their fit and potency as metalloprotease inhibitors.

Several hydroxamate-based inhibitors have been tested in preclinical and clinical settings (Hidalgo and Eckhardt 2001). One subset, which includes batimastat and marimastat, were designed to mimic the cleavage site of collagen, an MMP substrate. Other hydroxamate inhibitors include synthetic side chains that maximize fit into the catalytic site (e.g., CGS-27023). However, despite their potency, these small molecule inhibitors are not always selective for MMPs. A direct comparison of the selectivity of some of the more commonly used metalloprotease inhibitors indicates that they are equally potent inhibitors of ADAMs (Roghani et al. 1999; for review, see Moss et al. 2001). Batimastat and Ro-31-9790 inhibit ADAM17 better than several MMPs (Barlaam et al. 1999; Amour et al. 2000). Furthermore, clinical testing of marimastat revealed side effects that may well be due to inhibition of other adamalysins (Hidalgo and Eckhardt 2001). Efforts continue in generating specific and selective metalloprotease inhibitors, and many of these new compounds exhibit better selectivity for either MMPs or ADAMs (Kottirsch et al. 2002; Sawa et al. 2002), and discriminate between various ADAMs-dependent growth factor shedding activities (Parkin et al. 2002). Nevertheless, caution must be used in interpreting studies that use inhibitors as the sole criterion for implicating either MMP or ADAMs in a given biological process.

Tissue inhibitors of metalloproteases, or TIMPs, are endogenous regulators of MMPs (Brew et al. 2000). There are four known TIMPs in vertebrates, all of which exhibit high potency for MMP inhibition. The N-terminal domain of TIMPs fits like a wedge into the catalytic site of MMPs, whereas the C-terminal domain probably imparts binding specificity. However, the TIMPs are not totally selective for MMPs. TIMP-3 also inhibits ADAM17 (Amour et al. 2000) and ADAM12 (Loechel et al. 2000), as well as ADAM-TS4 and ADAM-TS5 (Kashiwagi et al. 2001). ADAM10 is inhibited by both TIMP-1 and TIMP-3 (Amour et al. 2000). But not all ADAMs are sensitive to TIMP3; ADAM8 and ADAM9 processing of myelin basic protein is not inhibited by any TIMP (Amour et al. 2002). As with the hydroxymates, TIMPs with targeted mutations that alter their specificity have been generated. It will be interesting to test these proteins in cell-based assays, as well as their therapeutic potential in preclinical models.

There is also an activator of metalloproteases, 4-aminophenylmercuricacetate or APMA. The mechanism for its activation was originally thought to be through displacement of the prodomain from the metalloprotease domain through preferential association with the key cysteine switch residue. However, mutational studies of MMP prodomains suggest that the activation mechanism may be more complicated and related to prodomain conformational changes (Galazka et al. 1999). APMA also activates ADAMs-dependent growth-factor shedding, presumably through a similar mechanism to that of MMP activation (Milla et al. 1999; Merlos-Suarez et al. 2001).

The disintegrin domain

The disintegrin domain is named for its presence in the snake venom metalloproteases (SVMPs), in which it is involved in binding of platelet integrin receptors. This prevents the association of platelets with their natural ligands such as fibrinogen, and results in a block in platelet aggregation at the wound site. (This disintegrin-mediated interaction of SVMPs along with the breakdown of basement membrane components by their metalloprotease activity leads to the severe hemorrhaging caused by bites from snakes harboring these toxins.)

The disintegrin domain of ADAMs proteins is ~90 amino acids long. Structurally, there is little known about the disintegrin domain of ADAMs, although one can gather information from the structural studies of SVMP crystals and other integrin receptor ligands (Gomis-Ruth et al. 1994). The disintegrin domains of SVMPs mimic the ligand site of matrix proteins like fibronectin for integrin receptors. Like fibronectin, many have an RGD consensus sequence within a 13 amino acid stretch called the disintegrin loop, which projects from the surface of the protein and confers binding to IIb3 and v3 integrin receptors (Blobel et al. 1992). The disintegrin domain of human ADAM15 also has a conserved RGD sequence, and associates with v3 and 51 in an RGD-dependent manner (Nath et al. 1999; Eto et al. 2002). Most ADAMs (including murine ADAM15) do not have the RGD sequence in their disintegrin loop, even though the disintegrin domains of many ADAMs do associate with integrin receptors. How does this association take place?

There is a subfamily of integrin receptors, the 4/9 subfamily, that do not recognize RGD, but instead bind to aspartic acid-containing sequences in proteins like fibronectin, VCAM-1, MadCAM-1, and tenascin-C (Zhu and Evans 2002). The disintegrin loops of each ADAM, except for 10 and 17, also have similar aspartic acid-containing sequences. Many ADAMs share a sequence (Rx6DEVF) in the disintegrin domain that, when mutated, inhibits the association with 91integrins (Eto et al. 2002). Consistent with this, all ADAMs tested except ADAM10 and ADAM17 can bind to 91 (Eto et al. 2002). But ADAMs can also associate with other integrin receptors. For example, ADAM28 binds 41 (Bridges et al. 2002), ADAM15 associates with v3 and 51 (see previous paragraph), and several ADAMs can associate with 61 integrin receptors (Chen et al. 1999a,b; Nath et al. 2000). The structural basis for these associations has yet to be defined.

The cysteine-rich and EGF-like domains

If there were a black box in the study of ADAM domains, it would be the cysteine-rich and EGF-like domains. What is known about these domains does not appear to provide any of the unifying functional themes that characterize the metalloprotease and disintegrin domains. It was originally noted that ADAMs 1, 3, 12, and 14 have a motif in their cysteine-rich domain that is very similar to sequences found in viral fusion peptides (Blobel and White 1992). This, coupled with the observations that ADAMs 1, 3 and 12 participate in cell fusion reactions, led to the proposal that the cysteine-rich domain is involved in membrane fusion. But this hypothesis has not been borne out by experimentation. It is more likely that the cysteine-rich domain complements the binding capacity of the disintegrin domain, and perhaps imparts specificity to disintegrin domain-mediated interactions. For example, the cysteine-rich (and perhaps also the disintegrin) domain of ADAM12 promotes the adhesion of fibroblasts and myoblasts (Zolkiewska 1999). Also, the disintegrin and cysteine-rich domains of ADAM13 bind to both fibronectin and to 1-containing integrin receptors, and binding can be inhibited with antibodies to the cysteine-rich domain (Gaultier et al. 2002).

Perhaps the most compelling piece of data concerning a cysteine-rich domain-specific function is that it acts as a ligand for the cell-adhesion molecule syndecan. This was first discovered in a study showing that the cysteine-rich domain of ADAM12 supports the in vitro binding of several different tumor cell lines, as well as a variety of nontumorigenic cells of bone and muscle origin, through the interaction with cell surface heparan-sulfate proteoglycans (Iba et al. 1999, 2000). Transfection of syndecans renders cells competent for adhesion to the cysteine-rich domain of ADAM12. Furthermore, affinity chromatography experiments suggest an interaction between syndecan-4 and ADAM12 (Iba et al. 2000). Other work has suggested coordination between syndecans and integrins in the ADAMs-dependent mediation of cell attachment and cell spreading respectively.

The cytoplasmic tail

The cytoplasmic tails of the ADAMs family are highly variable both in length and in sequence (Fig. 2). This domain contains specialized motifs that have been postulated to be involved in the inside-out regulation of metalloprotease activity, the outside-in regulation of cell signaling, and/or the control of maturation and subcellular localization. The most common motifs are PxxP binding sites for SH3 domain-containing proteins. These SH3-binding sites are present in human ADAMs 7, 8, 9, 10, 12, 15, 17, 19, 22, 29, and 33 (Fig. 2; Table 2). Several ADAMs also have potential phosphorylation sites for serine-threonine and/or tyrosine kinases. Not only might this regulate ADAM function directly, but the resulting phosphotyrosine residues could also provide ligands for SH2 domain-containing proteins. Consequently, ADAMs may serve adaptor functions to assemble complexes of proteins at critical sites of functional activity. The list of proteins that have been shown to interact with ADAMs either by in vitro or in vivo assays is shown in Table 2.

View larger version (54K): [in this window] [in a new window]   Figure 2.   The cytoplasmic tails of the ADAMs. Shown is an alignment of the cytoplasmic tails of all human ADAMs, made using CLUSTALW. PxxP motifs, predicted to be able to associate with SH3 domains, are shaded in blue.

ADAM15 has an extensive array of protein-protein interaction sites, including eight possible SH3-binding domains and four potential sites for tyrosine phosphorylation/SH2-association (Fig. 2). Perhaps not surprisingly, ADAM15 associates with a number of different proteins including adaptors (endophilin I, SH3PX1, and Grb2), and three Src family tyrosine kinases (Src, Lck, and Hck; Howard et al. 1999; Poghosyan et al. 2001). Most of these associations have only been reported in vitro, although an in vivo association between ADAM15 and Lck was observed in JURKAT cells. ADAM15 is also a substrate for Lck and Hck, and its tyrosine phosphorylation greatly impacts the specificity of its associations. For example, the association of ADAM15 with Hck is stimulated by Hck-dependent ADAM15 phosphorylation. The SH2 domain of Hck can associate with ADAM15, presumably via one of the four tyrosines in the tail of ADAM15. Tyrosine phosphorylation also impacts the association of ADAM15 with Lck and Grb2, although, in this case, it may be more related to conformational changes in the presentation of nearby SH3-binding domains, and not necessarily in the formation of an SH2 domain attachment site. A physiological context for the association of ADAM15 with these proteins has not yet been reported.

ADAM12 has 10 possible SH3-binding domains and 2 potential sites for tyrosine phosphorylation (Table 2). Like ADAM15, ADAM12 has also been reported to associate with Src, the related kinase Yes, and Grb2 (Kang et al. 2000; Suzuki et al. 2000). In C2C12 myoblasts, this is mediated by the SH3 domain of Src and the most membrane proximal of the proline motifs in ADAM12. ADAM12 is also a substrate for Src at its C-terminal tyrosine residue. An association of ADAM12 with the p85 subunit of phosphatidylinositol (PI) 3-kinase has been reported to activate PI-3 kinase (Kang et al. 2001). ADAM12 also associates with -actinin-1 and 2, through interaction of either the spectrin-like repeats or the C-terminal EF hand-containing region of -actinin with the membrane proximal portion of the tail of ADAM12 (Galliano et al. 2000).

ADAM9, like ADAM15, binds to endophilin I and SH3PX1 (Howard et al. 1999). In this case, the interactions appear to favor the unprocessed, intracellular forms of these ADAMs. Given the potential function of endophilin I and SH3PX1 in vesicle sorting, it is speculated that these interactions may play a part in the regulation of ADAM maturation and/or subcellular localization. The membrane proximal region of the tail of ADAM9 associates with the catalytic domain of protein kinase C (PKC; Izumi et al. 1998). The tail of ADAM9 can be phosphorylated by PKC in vitro at one or more serine and threonine residues. It is speculated that PKC helps to recruit ADAM9 to specific sites on the plasma membrane and upon phosphorylation/activation of ADAM9, HB-EGF shedding occurs (see below).

Yeast 2-hybrid analysis was used to detect an association between MAD2 and ADAM17 and between MAD2 and ADAM9 (Nelson et al. 1999). MAD2 is a component of the spindle assembly checkpoint mechanism. MAD2 shows 25% similarity to MAD2. These associations occur in regions of ADAMs that have PXXP motifs, yet MADs lack discernible SH3 domains. Perhaps these associations are involved in coordinating ADAM function with the cell cycle, or in a regulatory process by which MADs compete for binding of SH3 domain-containing proteins (Nelson et al. 1999).

	The shedding of cytokines and cytokine receptors

Top Introduction ADAMs are members of... Expression, subcellular... The shedding of cytokines... Growth factor and growth... The cleavage of other... The regulation and specificity... Cell fate determination and... Fertilization Myoblast fusion Human disease Conclusions References

Cytokines and their receptors play important roles during development and in inflammatory processes. They are elicited both as a response to infection by microorganisms and also during disease processes such as rheumatoid arthritis and cancer. ADAMs-mediated shedding of both cytokines and cytokine receptors has been reported, and much attention has been given to the possible use of ADAMs inhibitors as therapeutics for various diseases. We will discuss selected examples of cytokine and cytokine receptor shedding below.

Cytokines

Perhaps the most well-studied member of the ADAMs family is TNF converting enzyme (TACE) or ADAM17 (Black 2002). TNF is a major immunomodulatory and proinflammatory cytokine. It exists in two forms, a 223 amino acid precursor form anchored to the membrane, and a 146 amino acid soluble form, generated by precursor cleavage and comprising the ectodomain. ADAM17 was originally identified by screening for an activity that functioned in TNF peptide cleavage assays, and could release TNF into the culture medium of live cells. Confirmation of its activity came with the observation that TNF processing was significantly inhibited by targeted disruption of the adam17 gene (Black et al. 1997; Moss et al. 1997). However, processing of TNF can still occur in cells derived from ADAM17-deficient mice. This residual sheddase activity is inhibited by the metalloprotease inhibitor IC-3, suggesting that other ADAMs family members may also process TNF (Reddy et al. 2000). Both ADAM9 and ADAM10 cleave TNF peptides in vitro (Lunn et al. 1997; Rosendahl et al. 1997; Roghani et al.1999; Amour et al. 2000), although expressing ADAM10 in ADAM17 null cells does not reconstitute full TNF shedding (Reddy et al. 2000). It is likely that the exact composition of TACE activity is dependent on the expression levels of ADAM17 and other ADAMs, in particular tissues and cell types. Nevertheless, a major role for ADAM17 in cytokine biology is supported by the observation that its expression increases during the inflammatory response (Patel et al. 1998; Colon et al. 2001). Because the production of TNF has been implicated in the development of several inflammatory diseases, including arthritis and chachexia, understanding the mechanism of shedding of cytokines by ADAMs is of great therapeutic relevance.

Fractalkine is a chemoattractant cytokine that is thought to function in both adhesion (when membrane anchored) and chemoattraction (when soluble) of inflammatory cells bearing its receptor (Rossi and Zlotnik 2000). Fractalkine shedding from cell surfaces is inhibited by GM6001, and is therefore dependent on metalloprotease activity. ADAM17 null cells fail to shed fractalkine in response to a TPA stimulus (Garton et al.2001), suggesting that ADAM17 is the responsible enzyme.

Cytokine receptors

The activity of TNF receptors is also controlled by sheddases. Proteolysis of the p75 TNF receptor is blocked by a dominant-negative ADAM17 construct lacking its metalloprotease domain. Both the p55 and p75 TNF receptors are processed in ADAM17 null cell lines reconstituted with functional ADAM17, but not in the null cells (Solomon et al. 1999; Reddy et al. 2000; Zhang et al. 2001). Thus, both the cytokine and its receptor appear to be processed predominantly by ADAM17.Because in one case the activity of ADAM17 has a positive effect (by generating active cytokine), whereas in the other case the effect would be to inhibit cytokine action (by removing active receptor from the cell surface), it will be interesting to determine how the relative amounts of these activities impact inflammatory as well as other TNF-regulated processes.

Interleukin1 (IL-1) receptor-II shedding also does not occur in ADAM17-deficient fibroblasts (Reddy et al. 2000). Reconstitution of these cells with ADAM17, but not ADAM10, restores shedding. A chimera in which the disintegrin, cysteine-rich, and EGF-like domains of ADAM17 are replaced with those of ADAM10 fails to process the IL-1 receptor-II, suggesting that one or more of these domains modulate protease activity. ADAM17 also reconstitutes TPA-inducible shedding of the IL-6 receptor in ADAM17 deficient cells (Althoff et al. 2000). But constitutive shedding of the IL-6 receptor is measurable in ADAM17 null cells, indicating an additional role for a second metalloprotease in IL-6R shedding.

	Growth factor and growth factor receptor processing

Top Introduction ADAMs are members of... Expression, subcellular... The shedding of cytokines... Growth factor and growth... The cleavage of other... The regulation and specificity... Cell fate determination and... Fertilization Myoblast fusion Human disease Conclusions References

The production of peptide growth factors in response to diverse stimuli is frequently controlled at the transcriptional level. But many of these growth factors are then produced in membrane-associated precursor forms. Another level of control is therefore the regulation of the processing of these precursor forms to allow the shedding of active growth factors. This shedding often involves ADAMs family proteins. Growth factor receptors can also be shed from the cell surface, thus terminating signaling. Here too, members of the ADAMs family play a role. In this section, we will describe some cases of both growth factor and growth factor receptor processing.

Growth factors

Many ligands for the EGF receptor family are shed from cell surfaces in response to specific signals. One such ligand is heparin-binding EGF (HB-EGF), which activates the EGF receptor (Massague and Pandiella 1993). The precursor for HB-EGF is composed of signal peptide, heparin-binding, EGF-like, transmembrane, and cytoplasmic domains. Membrane-anchored HB-EGF is biologically active, binding to 31 integrin receptors on neighboring cells and transmitting a nondiffusable signal that inhibits cell proliferation (Raab and Klagsbrun 1997). However, HB-EGF is also processed in response to activation of protein kinase C (PKC) by phorbol esters (e.g., TPA), generating a soluble, mitogenically active protein (Gechtman et al. 1999). PKC appears to be the enzyme responsible for TPA-stimulation of HB-EGF shedding (Izumi et al. 1998). HB-EGF-dependent activation of the EGF receptor also occurs when G protein-coupled receptors are activated, in a process often referred to as EGF receptor transactivation (Daub et al. 1996; Luttrell et al. 1997). The release of soluble HB-EGF during this EGF receptor transactivation is mediated by metalloproteases. (Prenzel et al. 1999)

Which metalloprotease is involved in the regulated shedding of HB-EGF? ADAM9 has been implicated in some experiments. For example, PKC directly associates with, and phosphorylates, the cytoplasmic domain of ADAM9 (Izumi et al. 1998). Overexpression of ADAM9 stimulates HB-EGF shedding in the absence of TPA stimulation, whereas expression of metalloprotease mutants of ADAM9 inhibits TPA stimulated HB-EGF shedding. Yet, it was reported recently that fibroblasts derived from ADAM9 null mice still process HB-EGF (Weskamp et al. 2002). HB-EGF shedding does not occur in cells derived from ADAM17-deficient mice, and ADAM17 can cleave an HB-EGF peptide in vitro (Merlos-Suarez et al. 2001; Sunnarborg et al. 2002). ADAM10 has also been reported to mediate the transactivation of EGF receptor in Cos7 and PC-3 cells (Yan et al. 2002). A case may also be made for ADAM12, which appears to regulate HB-EGF shedding and EGF receptor transactivation in cultured cardiomyocytes (Asakura et al. 2002). Further studies will clearly be needed to determine which ADAMs are responsible for regulating HB-EGF shedding in response to signals such as TPA and G protein-coupled receptor ligands in different cell types.

TGF is also a ligand for members of the EGF receptor family and is also produced in a membrane-associated precursor form. ADAM17 is thought to be the major sheddase for TGF. The most compelling evidence supporting this hypothesis comes from an analysis of genetically modified mice. As mentioned earlier, mice harboring a germ-line mutation in the metalloprotease domain of ADAM17 have defects in epithelial maturation and organization (Peschon et al. 1998). Similar defects are observed in mice lacking the EGF receptor and in mice null for TGF (Luetteke et al. 1993; Miettinen et al. 1995; Sibilia and Wagner 1995; Threadgill et al. 1995). Fibroblasts and keratinocytes derived from ADAM17-deficient mice shed reduced levels of TGF into the medium, and TGF is cleaved by ADAM17 in vitro (Peschon et al. 1998; Sunnarborg et al. 2002). ADAM17 can also cleave peptides of many EGF family members including amphiregulin, betacellulin, epiregulin, and HB-EGF (Sunnarborg et al. 2002). However, ADAM17 may not be the only TGF sheddase, as the metalloprotease activator APMA still stimulates TGF shedding in ADAM17-deficient cells (Merlos-Suarez et al. 2001).

Neuregulin (NRG) is another member of the family of EGF-like ligands, and acts predominantly on the receptors HER2, HER3, and HER4. It regulates acetylcholine receptor synthesis, Schwann cell growth, and the early development of the heart and central nervous system (Garratt et al. 2000). Several lines of evidence suggest that ADAM19 mediates the shedding of neuregulin (NRG; Shirakabe et al. 2001). Firstly, the two proteins are coexpressed in dorsal root ganglia during mouse embryogenesis and during the differentiation of neural precursor cells (Kurisaki et al. 1998). Secondly, ectopic expression of wild-type ADAM19 enhanced the release of NRG from transiently transfected L929 cells, whereas expression of ADAM19 metalloprotease-deficient mutants blocked NRG shedding (Shirakabe et al. 2001). It is thought that NRG processing occurs in the Golgi compartment because it is sensitive to the ER-to-Golgi transport inhibitor brefeldin A, but is insensitive to the intra-Golgi transport inhibitor monensin. ADAM9 could also process NRG when tested in a transfection system, but produced smaller and presumably inactive products (Shirakabe et al. 2001). Whether ADAM19 is the only member of the ADAMs family that is involved in the physiological processing of NRG will have to await analysis of cells deficient in ADAM19.

Growth factor receptors

HER4/erbB4 is a receptor tyrosine kinase of the EGFR family. It is predominantly activated by neuregulin, but HB-EGF, betacellulin, and epiregulin are also suitable ligands. HER4 is important for early development of the cardiovascular and central nervous systems, and its expression has also been linked to cancer development, particularly of the breast (Hackel et al. 1999). The HER4 JM-a isoform is shed in response to phorbol ester stimulation, and shedding is inhibited with the metalloprotease inhibitors BB-94, BB-3103, and TAPI-2 (IC-3), (Vecchi and Carpenter 1997; Rio et al. 2000). Shedding is essentially absent in cells lacking ADAM17. Furthermore, ADAM17 re-expression reconstitutes both constitutive and TPA-induced HER4 shedding.

Colony stimulating factor-1 (CSF-1) receptor functions in the survival, proliferation, and differentiation of mononuclear phagocytes. Shedding of the receptor by PKC activators generates a soluble form that is still competent for ligand binding, and down-modulates CSF-1 signaling and inactivates macrophages. CSF-1 receptor shedding is blocked by metalloprotease inhibitors as well as by an antibody to the catalytic site of ADAM17 (Rovida et al. 2001). The reduction of cell surface CSF-1 receptor by TPA treatment is also effectively abolished in ADAM17 deficient monocytes.

The hepatocyte growth factor receptor Met is also a sheddase target. BB-94 blocks EGF and LPA-dependent Met shedding. TIMP3 blocks EGF, LPA, and TPA-dependent Met shedding, but TIMPs 1 and 2 are ineffective. Because all known MMPs, but not ADAMs, are inhibited by TIMP1 or TIMP2, whereas TIMP3 also inhibits some ADAMs, it is likely that either an unknown MMP or an ADAM is responsible for Met shedding (Nath et al. 2001).

The nerve growth factor (NGF) receptor, also known as TrkA, is shed in response to TPA or osmotic stress (Diaz-Rodriguez et al. 2002). Both stimuli result in an activation of MAP kinase pathways, ERKs in the case of TPA, and both ERKs and p38 in the case of osmotic stress. MEK inhibitors partially reduce TPA-induced shedding, as does transfection of a dominant-negative ERK2 construct. TPA-dependent NGF receptor shedding does not occur in ADAM17 deficient fibroblasts and can be reconstituted by transfection of wild-type ADAM17. In contrast, the osmotic pathway is unaffected by ADAM17 expression, thus implicating different metalloproteases. Interestingly, ERK can phosphorylate the cytoplasmic domain of ADAM17 in vitro at threonine 735. Stimulation with TPA, EGF, or NGF also leads to an increase in ADAM17/ERK association and ERK-dependent ADAM17 phosphorylation in vivo, which is blocked by MEK inhibitors, dominant-negative ERK2, and dominant-negative ADAM17.

	The cleavage of other molecules

Top Introduction ADAMs are members of... Expression, subcellular... The shedding of cytokines... Growth factor and growth... The cleavage of other... The regulation and specificity... Cell fate determination and... Fertilization Myoblast fusion Human disease Conclusions References

Insulin-like growth factor-binding proteins

The six members of the (IGFBP) family bind and sequester mitogenic insulin-like growth factors (Ferry et al. 1999). The soluble form of ADAM12 (ADAM12-S) binds to and processes purified IGFBP 3 and 5 (but not other IGFBPs). In keeping with a role for an ADAMs family protease, this processing is inhibited by TIMP-3 (but not the other TIMPs; Loechel et al. 2000; Shi et al. 2000; ). It is likely that ADAM12-S represents the serum protease that is up-regulated during pregnancy and degrades IGFBP-3, thus freeing IGF in its fully active form (Ferry et al. 1999). Because IGFBP-3 is a major IGF-binding protein, its regulation by ADAM12-S also has broad implications in diseases related to low IGF levels such as osteoarthritis and diabetes, as well as perhaps the high levels of IGF-1 seen in some tumors.

Prions

Normal processing of cellular prion precursor yields a product called N1, which is nontoxic. In contrast, aberrant processing of the precursor leads to cerebral deposition of an insoluble, protease-resistant protein, which can lead to neurodegenerative diseases. The formation of N1 in HEK293 cells can be partially inhibited by the metalloprotease inhibitors TAPI and BB-3103 (Vincent et al. 2001). ADAM10-deficient fibroblasts show a marked reduction in both constitutive and TPA-inducible production of N1, whereas ADAM17-deficient fibroblasts show normal constitutive N1 shedding, and only a slight reduction in the inducible pathway. Thus, ADAM10 and ADAM17 may function in prion processing and could have positive effects in preventing formation of insoluble prions.

Amyloid precursor protein

The accumulation of -amyloid peptides in the cerebral cortex is a crucial step in the pathogenesis of Alzheimer's disease (Gandy and Petanceska 2000). The -amyloid peptides are formed by stepwise processing of the amyloid precursor protein (APP) by  and -secretases. An alternative pathway of APP secretion is through the action of -secretase activity. This pathway is TPA-inducible. The soluble APP (sAPP) that results from -secretase activity has positive neurotrophic effects, and opposes the harmful effects of -amyloid formation.

Overproduction of either membrane bound, or an alternatively spliced and soluble form, of ADAM9 in COS cells enhances sAPP production in a TPA-dependent and metalloprotease inhibitor-sensitive manner (Koike et al. 1999; Hotoda et al. 2002), suggesting that ADAM9 is an -secretase. But although purified ADAM9 can process APP more efficiently than other peptide substrates in vitro, the processing is not at the  site (Roghani et al. 1999). Furthermore, there are no differences in APP processing between wild-type fibroblasts and fibroblasts lacking ADAM9 (Weskamp et al. 2002), making it unlikely that ADAM9 is the sole -secretase. In keeping with this, overproduction of ADAM10 in HEK cells also enhances sAPP production, and both constitutive and TPA-inducible sAPP production is inhibited by the metalloprotease inhibitor BB3103 and by a point mutation in the catalytic site of the metalloprotease domain of ADAM10 (Lammich et al. 1999). ADAM10 also cleaves APP in vitro at the  site (Lammich et al. 1999; Amour et al. 2000). However, as with ADAM9, embryonic fibroblasts derived from ADAM10-deficient mice are still capable of sAPP production (Hartmann et al. 2002). Finally, ADAM17 also has -secretase activity in vitro (Buxbaum et al. 1998). Furthermore, metalloprotease inhibitors such as TAPI-1 block both constitutive and inducible sAPP production, overproduction of ADAM17 increases sAPP production, and ADAM17-deficient cells cannot produce sAPP (Buxbaum et al. 1998; Slack et al. 2001). It is quite possible that there is redundancy in the regulation of sAPP secretion, or that individual ADAMs regulate secretion independently. Slack and coworkers argue that ADAM17 is the principle mediator of constitutive sAPP production and that a different ADAM functions during inducible sAPP production (Slack et al. 2001). APP expression aligns better with that of ADAM10 than ADAM17 in brain tissue and neurons (Marcinkiewicz and Seidah 2000). In any case, these findings warrant the analysis of ADAMs activators for the possible treatment of Alzheimer's disease.

Extracellular matrix components

It is often thought that degradation of extracellular matrix (ECM) components is the sole purview of the matrix metalloprotease family. But there have been reports of ECM degradation by ADAMs family members. For example, ADAM10 cleaves type IV collagen in vitro (Millichip et al. 1998), ADAM15 cleaves type IV collagen and gelatin in vitro (Martin et al. 2002), and ADAM13-expressing cells degrade fibronectin (Alfandari et al. 2001). It is speculated that such activity may assist in cell migration (see below).

	The regulation and specificity of sheddase activity

Top Introduction ADAMs are members of... Expression, subcellular... The shedding of cytokines... Growth factor and growth... The cleavage of other... The regulation and specificity... Cell fate determination and... Fertilization Myoblast fusion Human disease Conclusions References

As we have seen, ADAM-mediated shedding of diverse membrane proteins occurs both constitutively and in response to a variety of stimuli, including unidentified serum factors, peptide growth factors, changes in intracellular calcium concentration, hypertonicity (osmotic stress) and PKC activation. Given this diversity of signals and substrates for sheddase action, it is possible that shedding is controlled by multiple regulatory mechanisms (Pandiella and Massague 1991). We will consider these mechanisms in this section.

TPA-induced shedding has been widely studied. In the TPA-stimulated processing of HB-EGF, PKC is involved, as described earlier (Izumi et al. 1998). Other studies have also implicated the MAP kinase pathway. For example, TPA-treatment of CHO cells leads to temporally coincident increases in ERK phosphorylation and HB-EGF shedding, which is inhibited by a MEK inhibitor (Gechtman et al. 1999). In support of multiple regulatory mechanisms, experiments have shown that calcium-modulating agents fail to impact TPA-induced HB-EGF shedding. Conversely, PKC inhibitors do not affect shedding induced by intracellular calcium-modulating agents (Dethlefsen et al. 1998). Similarly, TPA and pervanadate both induce L1 adhesion protein shedding, but the TPA pathway is insensitive to Src inhibitors, whereas the pervanadate pathway is Src dependent, but unaffected by inhibition of PKC (Gutwein et al. 2000). Further, L-selectin is shed in response to TPA and hypertonicity, but only hypertonic shedding is sensitive to p38 MAPK inhibitors (Rizoli et al. 1999).

Growth factors such as fibroblast growth factor (FGF), platelet-derived growth factor (PDGF), and EGF treatment of CHO cells stimulate TGF shedding (Fan and Derynck 1999). Inhibitors of the corresponding receptors, transient transfections of dominant-negative receptor constructs, and MEK inhibitors all block growth factor-dependent TGF shedding. In contrast, the constitutive shedding of TGF is inhibited by a p38 antagonist. The two MAPK pathway inhibitors together completely block both growth factor-dependent and independent shedding processes. Curiously, no shedding activity occurs when cells are in suspension even though TPA-stimulated MAPK activation is completely normal under these conditions (Gechtman et al. 1999; Gutwein et al. 2000). Thus, the MAPK pathway may be necessary for TGF shedding, but is clearly not sufficient. L-selectin shedding in response to F-MetLeuPhe stimulation of neutrophils is also inhibited by MEK and p38 inhibitors, as is TNF and L-selectin release from CHO cells in response to FGF, PDGF, serum, and TPA stimulation (Gutwein et al. 2000).

In an attempt to define the molecular components that regulate shedding, a mutational analysis in CHO cells was undertaken, screening for mutants that no longer shed TGF in response to TPA, calcium, and serum. Two such cell lines were obtained and further analyzed. Interestingly, constitutive shedding was unaffected in these mutants (Arribas and Massague 1995). Remarkably, it was not just the shedding of TGF that was suppressed, the shedding of APP, L-selectin, and the IL-6 receptor were also inhibited (Arribas et al. 1996). It was later shown that these two mutant cell lines belong to the same complementation group; that is, a single gene is affected. These mutant cell lines, called M2, are neither defective in PKC signaling nor can be complemented by ADAM17 transfection, thus indicating that there is some other component regulating inducible shedding processes in these cells (Merlos-Suarez et al. 1998). These data suggest that, even though different shedding responses involve different proteins and pathways, most of these pathways converge at some point on a single protein. Obviously, it will be of great interest to identify this protein.

How is the generalized proteolysis of membrane proteins avoided when ADAMs are activated? Part of the regulation could be imparted by the specificity of the ADAMs for their targets. This was addressed by replacing the extracellular juxtamembrane sequence of betaglycan (which does not undergo regulated shedding) with sequences from the equivalent region of either TGF or APP. Both of the resulting betaglycan chimeras are shed in response to TPA, increased calcium, and serum (Arribas et al. 1997). Because the juxtamembrane sequences of TGF and APP are not highly related, it is thought that the secondary structure of the juxtamembrane region is the important recognition element for the sheddase. Presumably, not all membrane proteins share this secondary structure.

There have also been suggestions that the ADAMs have intrinsic substrate specificity, based on a short peptide sequence, which would limit their proteolytic activity on most surface proteins (Hattori et al. 2000). However, as we have described earlier, it is likely that multiple ADAMs family members process a single cell surface molecule, and a single ADAM can process several growth factors, adhesion molecules, and/or cytokines. Regulation of sheddase activity likely depends on the expression profile of ADAMs family members, the stimulus being used, and some innate substrate specificity of the ADAMs.

	Cell fate determination and cell migration

Top Introduction ADAMs are members of... Expression, subcellular... The shedding of cytokines... Growth factor and growth... The cleavage of other... The regulation and specificity... Cell fate determination and... Fertilization Myoblast fusion Human disease Conclusions References

During the development of multicellular organisms, multiple cell types are generated from a population of initially equivalent cells. These cells must then migrate and establish the patterns that ultimately lead to the formation of tissues and organs during embryogenesis. This cell fate determination involves both stimulatory and inhibitory cell-cell interactions, which have been most fully studied in the case of the nervous system. In this section, we will describe the involvement of ADAMs in cell fate determination and cell migration.

Delta and Notch

The Notch/Delta signaling system is involved in the development of the nervous system (Mumm and Kopan 2000). It was first described in Drosophila, but it has since been shown that this system is conserved in worms and mammals. During the development of the nervous system, neural precursor cells inhibit adjacent cells from adopting a similar neurogenic fate through a process called lateral inhibition. Both notch and delta were genetically defined as being involved in this process.

Biochemical analyses have shown that Notch functions as a cell surface receptor for membrane-bound Delta, and other members of the DSL ligand family (Serrate, Lag2 in Drosophila) on adjacent cells. The binding of Delta to Notch signals cleavage of Notch in its ectodomain, proximal to the membrane, at a so-called site 2. Once this cleavage has occurred, a presenilin-dependent activity cleaves Notch in the transmembrane region, releasing the active cytosolic fragment of Notch to associate with and activate transcription factors of the CSL family (CBF1, Suppressor of Hairless, Lag1).

Delta can also be a substrate for sheddase activity. For example, transiently transfected Schneider-2 (S2) cells shed Delta co