The Lymphocyte Metalloprotease MDC-L (ADAM 28) Is a Ligand for the Integrin α4β1

The interaction of lymphocytes with other cells is critical for normal immune surveillance and response. MDC-L (ADAM 28), a member of the ADAM (a disintegrinand metalloprotease) protein family, is expressed on the surface of human lymphocytes. ADAMs possess a disintegrin-like domain similar in sequence to small non-enzymatic snake venom peptides that act as integrin antagonists. We report here that the disintegrin domain of MDC-L is recognized by the leukocyte integrin α4β1. Recombinant Fc fusion proteins possessing the disintegrin domain of MDC-L supported adhesion of the T-lymphoma cell line, Jurkat, in a concentration- and divalent cation-dependent manner. Adhesion of Jurkat cells to the disintegrin domain of MDC-L was inhibited by an anti-MDC-L monoclonal antibody (mAb), Dis1-1. The epitope for mAb Dis1-1 was localized within 59 residues of the disintegrin domain. Recombinant expression of this 59-residue fragment of the disintegrin domain also supported cell adhesion. Adhesion of Jurkat cells to the MDC-L disintegrin domain was specifically inhibited by anti-α4 and anti-β1 function-blocking mAbs. Furthermore, adhesion of various cell lines to MDC-L correlated with expression of the integrin α4-subunit. Transfected K562 cells expressing α4β1 adhered to the disintegrin domain in contrast to non-transfected K562 cells. We further investigated the binding of recombinant MDC-L disintegrin domain (rDis-Fc) in solution. The rDis-Fc was found to bind to Jurkat cells in solution in a concentration-dependent and saturable manner. Both adhesion and solution binding of rDis-Fc was inhibited by the α4β1 ligand mimetic CS-1 peptide. Additionally, recognition of the MDC-L disintegrin domain required “activation” of lymphocyte β1 integrins. The interaction of MDC-L with α4β1 may potentially regulate metalloprotease function by targeting or sequestering the active protease on the cell surface. These results suggest a potential role for the lymphocyte ADAM, MDC-L, in the interaction of lymphocytes with α4β1-expressing leukocytes.

A disintegrin and metalloprotease (ADAM) 1 is a family of recently identified cell surface and secreted glycoproteins that possess both proteolytic and adhesive properties (1,2). Prototypical members of this family are composed of a prodomain, metalloprotease, disintegrin-like, cysteine-rich, EGF-like, transmembrane, and cytoplasmic domains. The metalloprotease domain is homologous to the reprolysins, members of the metazincin superfamily that include the matrix metalloproteases, the astacins, and serralysins (3). Although the ADAMs may serve other proteolytic functions, their role in the ectodomain shedding of specific cell surface proteins has been well documented (4,5). ADAMs 9 (meltrin-␥), 10 (kuzbian), and 17 (tissue necrosis factor-␣-converting enzyme) have been shown to function in ectodomain shedding of membrane-anchored heparin-binding EGF-like growth factor, amyloid precursor protein, and tissue necrosis factor-␣, respectively (6 -10). These ADAMs and other members of this protein family have also been implicated in the shedding of other proteins such as L-selectin, FasL, and VCAM-1 (11)(12)(13).
In addition to the proteolytic functions mentioned, several ADAMs have been shown to interact with the integrin family of cell surface receptors. Integrins are a widely distributed superfamily of noncovalent heterodimeric glycoproteins that play a vital role in cellular adhesion, migration, and signal transduction (14,15). The ADAM disintegrin-like domains are homologous to small non-enzymatic peptides isolated from the venom of snakes that function as antagonists of integrins (16,17). The direct interaction of the snake venom disintegrin peptides with integrins led to the hypothesis that the disintegrin-like domains of ADAMs may function as integrin ligands. The disintegrin-like domains of ADAMs 1-3 expressed on the surface of sperm interact with the integrin ␣ 6 ␤ 1 in association with CD9 and CD98 on the egg surface (18 -23). Recognition of ADAMs 2 and 3 by ␣ 6 ␤ 1 requires the residues DECD located within a region of the disintegrin domain, designated the disintegrin loop, that corresponds to an extended loop in the snake venom peptides that typically contains an RGD sequence required for integrin binding (21)(22)(23). Human ADAM 15 contains an RGD sequence within the disintegrin loop region and is recognized by the integrins ␣ v ␤ 3 and ␣ 5 ␤ 1 (24 -26). However, mouse ADAM 15 lacks the RGD sequence found in the human homologue. Both human and mouse ADAM 15 as well as ADAM 12 were shown to be recognized by the integrin ␣ 9 ␤ 1 via residues outside the disintegrin loop (27). Other ADAM disintegrin-like domains reported to bind integrins are ADAM 9, which is recognized by ␣ 6 ␤ 1 and ␣ v ␤ 5 (28,29), and ADAM 23, which is recognized by ␣ v ␤ 3 (30).
At present 21 human ADAMs have been identified; however, the function of most remains unknown. MDC-L (ADAM 28) is a member of the ADAM protein family possessing a prototypical domain structure (31,32). MDC-L is predominantly expressed on the surface of human lymphocytes (31); however, others have also reported expression within mouse epididymis (32). The metalloprotease domain of MDC-L possesses catalytic activity, as demonstrated by the ability of MDC-L to cleave the substrate myelin basic protein (33). The MDC-L metalloprotease also appears to be involved in the proteolytic processing of its prodomain (32). However, the physiological substrate for the MDC-L metalloprotease has not been identified. Furthermore, the adhesive properties of MDC-L have not been investigated. To shed light on the function of MDC-L expressed on the surface of lymphocytes, we sought to determine whether the disintegrin domain of MDC-L supported cell adhesion.
Leukocytes express a variety of integrins, including the highly expressed members of the ␤ 2 integrin family and ␣ 4 ␤ 1 (14). The integrin ␣ 4 ␤ 1 is predominantly expressed by leukocytes and is involved in embryogenesis, hematopoiesis, and the immune and inflammatory responses. Ligands for ␣ 4 ␤ 1 include the alternatively spliced connecting segment of fibronectin (CS-1) and vascular cell adhesion molecule 1 (VCAM-1) (34 -37). The binding sequences in CS-1 and VCAM-1 are IELD-VPST and QIDSP, respectively (38 -41). Similar to several other members of the integrin family, ␣ 4 ␤ 1 is able to adapt multiple conformations resulting in affinity modulation, "activation," of the integrin (42)(43)(44). We report here that the disintegrin-like domain of MDC-L is recognized by ␣ 4 ␤ 1 in an activation-dependent manner. The interaction of MDC-L with ␣ 4 ␤ 1 suggests a role in lymphocyte-leukocyte adhesion and localization of the active metalloprotease.
Production of Anti-MDC-L mAbs-Hybridoma cell lines were generated at the Hybridoma Center for Agriculture and Biological Sciences at Oklahoma State University, Stillwater, OK. MDC-L disintegrin-like domain and EGF-like domain maltose-binding protein (MBP) fusion proteins were produced as described (31). The purified MDC-L fusion proteins were mixed and used as immunogens. Supernatants from the hybridomas grown in selective media were first assayed in an enzymelinked immunosorbent assay (ELISA) for immunoglobulins that reacted with either disintegrin-like domain MBP fusion protein or EGF-like domain MBP fusion protein (48). Supernatants positive for both proteins were considered to be against MBP and not pursued. Positive supernatants were then subjected to a second ELISA screen using rDis/EGF-Fc-coated microtiter well plates. Two hybridomas, designated Dis1-1 and EGF 1-2 based on their initial reactivity, were subcloned until clonal. These antibodies were produced in serum-free culture and purified on protein A-Sepharose (Amersham Biosciences). Both mAbs were of the IgG 1 isotype (Mouse Ig Screening Kit, Roche Molecular Biochemicals).
Recombinant MDC-L Fusion Proteins-Manipulation of recombinant DNA was by standard techniques (49). Restriction enzymes, T4 DNA ligase, and Taq polymerase were purchased from Roche Molecular Biochemicals. Oligonucleotides were synthesized by the Molecular Biology Resource Facility at the University of Oklahoma Health Sciences Center. The rDis-Fc and rDis/EGF-Fc constructs were expressed using the Insect Select System (Invitrogen, Carlsbad, CA). First, cDNA en- Jurkat cell (2 ϫ 10 5 cells/well) adhesion to wells coated with 10 g/ml rDis-Fc was examined in the presence of mAb Dis1-1 (E) and mAb EGF1-2 (q) at various concentrations. Results shown are the average Ϯ S.D. of triplicate determinations. b, the epitope for mAb Dis1-1 is located within 59 residues (Gly 418 -Glu 476 ) of the MDC-L disintegrin domain. Microtiter wells coated with 5 g/ml rDis/EGF-Fc, rDis-Fc, rDis, and Mini-Dis (residues Gly 418 -Glu 476 ) were incubated with 1 g/ml mAb Dis1-1 (white bars) and mAb EGF1-2 (black bars). After washing, bound mAb was disclosed with peroxidase-conjugated goat anti-mouse IgG-specific antibody and developed with o-phenylenediamine. Results shown are the average absorbance (490 nm) Ϯ S.D. of triplicate determinations. c, SDS-PAGE (12%) of nickel-nitrilotriacetic acid-agarose-purified (5 g/lane) rDis and Mini-Dis performed under reduced conditions and stained for total protein with Coomassie Blue. Molecular weight standards are shown on the left of the gel. d, Jurkat cells (2 ϫ 10 5 cells/well) were incubated for 1 h at 37°C with rDis-Fc (Ⅺ), rDis (q), and Mini-Dis (E) at various concentrations. After washing away non-bound cells, the number of adherent cells was determined using endogenous acid phosphatase as described under "Experimental Procedures." Adherent cells/well ϭ adherent cells (recombinant protein) Ϫ adherent cells (BSA) . Results shown are the average Ϯ S.D. of triplicate determinations.
coding MDC-L residues Thr 407 -His 664 with unique XbaI and BglII sites on the ends was generated by PCR, and the product was subcloned into XbaI/BglII-digested baculovirus cloning vector pAc-GP67 (BD Phar-Mingen) to fuse the GP67 signal sequence to the amino terminus of the MDC-L protein sequence. The cDNA encoding the heavy chain Fc region of human IgG 3 was subcloned from a lymph node cDNA library (CLONTECH, Palo Alto, CA) by PCR using gene-specific primers. The isolated IgG 3 heavy chain Fc (starting at residue Ala 130 ) lacking the hinge region was then fused in frame to the carboxyl-terminal end of the GP67-MDC-L cDNA at residue His 664 (rDis/EGF-Fc) or residue Gly 500 (rDis-Fc) by PCR overlap extension (50,51). The PCR products encoding the GP67-MDC-L-Fc constructs were then subcloned into the TOPO pIB/V5-His vector (Invitrogen). The sequence of all constructs was verified by complete nucleotide sequencing of both strands. The pIB-rDis-Fc and pIB-rDis/EGF-Fc vectors were then used to transfect High Five cells using Insect Plus TM Liposome (Invitrogen). Stable transfectants were selected in media containing blasticidin. Recombinant protein was purified from the conditioned media by polyethylene glycol concentration and purification on protein G-Sepharose. The fusion proteins were eluted from the protein G-Sepharose in 100 mM citric acid, pH 3.0, and immediately neutralized by addition of 1 M Tris-Cl, pH 9.0. Purified protein was analyzed by SDS-PAGE. Mass spectrometry analysis indicated that the amino acid composition of secreted rDis-Fc was consistent with the amino terminus having the sequence NЈ-ADP-GYLLD from the GP67 secretion signal fused to the amino terminus of the MDC-L disintegrin-like domain starting at residue Thr 407 .
The hexahistidine MDC-L disintegrin domain rDis (residues Ser 406 -Gly 500 ) and MiniDis (residues Gly 418 -Glu 476 ) fusion proteins were produced in Escherichia coli using the pQE30 expression vector (Qiagen Inc., Valencia, CA). DNA encoding the human MDC-L DNA was PCR-amplified using primers that created novel BamHI and HindIII sites and then subcloning the BamHI/HindIII-digested PCR product into BamHI/HindIII digest pQE30. The ligated DNA was used to transform E. coli M15[pREP4]. The correct coding sequence was verified by nucleotide sequencing of the insert in the plasmid. Cells harboring the pQE30 disintegrin-like domain plasmids were grown, and recombinant protein expression was induced as described (46). The cell pellet was disrupted in lysis buffer (8 M urea, 0.1 M NaH 2 PO 4 , 0.01 M Tris-Cl, pH 8.0), centrifuged at 11,000 ϫ g for 20 min at RT, and the supernatant subjected to nickel-nitrilotriacetic acid-agarose affinity chromatography. The column was washed in lysis buffer at pH 6.3, and bound protein was eluted in lysis buffer, pH 6.3, containing 100 mM EDTA. The recombinant protein was then subjected to rapid dialysis in PBS. The purified fusion proteins were examined by SDS-PAGE, and the correct amino acid composition was verified by electrospray mass spectrometry.
For production of alanine substitution mutants at positions Glu 471 and Asp 473 , the Quick Change Site-directed Mutagenesis Kit (Stratagene, La Jolla, CA) was employed. After verification of the expected mutation by complete nucleotide sequencing of the rDis-Fc coding region, High Five cells were transfected, and recombinant protein was purified and analyzed as described above.
Cell Adhesion Assay-The assays for cell adhesion to immobilized substrates were based on that described for adhesion to fibronectin (52). Immulon-2 96-well plates (Dynatech Laboratories, Inc. Chantilly, VA) were coated with the designated protein at various concentrations in 0.1 M NaHCO 3 , pH 8.4, overnight at 4°C. The wells were blocked with 2% bovine serum albumin (BSA) in 0.1 M NaHCO 3 , pH 8.4, at RT for 2 h. Cells were prepared by washing three times in HEPES Tyrode's buffer (5 mM HEPES, 150 mM NaCl, 12 mM NaHCO 3 , 2.6 mM KCl, 5 mM D-glucose, 0.2 mg/ml BSA, 0.5 mM MgCl 2 , 1 mM CaCl 2 , 1 mM MnCl 2 ). The washed cells were resuspended in the HEPES Tyrode's buffer, and 100-l aliquots were added to the wells. Where indicated, the inhibitors or stimuli were diluted in HEPES Tyrode's buffer and mixed with the cells just prior to addition to the wells. Cells were allowed to adhere for 1 h at 37°C in an atmosphere of 5% CO 2 . The wells were gently washed three times with HEPES-Tyrode's buffer to remove non-adherent cells. The remaining adherent cells were visually examined and quantitated by colorimetric analysis using endogenous cellular acid phosphatase by incubating 1 h at 37°C in 100 l of 1% Triton X-100, 50 mM sodium acetate, pH 5.0, 6 mg/ml p-nitrophenyl phosphate. The wells were developed by addition of 50 l/well of 1 N NaOH and then read in a plate reader at 405 nm. The number of adherent cells was determined from a standard curve generated using known numbers of the same cells.
Soluble rDis-Fc Binding-Cells were washed twice in modified Tyrode's (5 mM HEPES, pH 7.4, 150 mM NaCl, 12 mM NaHCO 3 , 2.6 mM KCl, 1 mg/ml BSA, 1 mM MgCl 2 , and 1 mM CaCl 2 ) with or without 1 mM  General Procedures-Protein concentrations were determined by the BCA assay (Pierce). ELISAs were based on standard methods (48). Recombinant proteins were diluted in 0.1 M NaHCO 3 , pH 8.4, and Immulon-2 96-well plates coated at 5 g/ml. Nonspecific binding was blocked with 2% blotto in PBS. The wells were washed 3 times in PBS containing 0.05% Tween 20, and 50 l of mAbs Dis1-1 or EGF 1-2 diluted in PBS to 1 g/ml were added. After washing the plates in PBS/Tween, goat anti-mouse biotin-conjugated antibody (Pierce) was then added 1:5000 dilution in PBS. The plates were again washed, and bound biotinylated immunoglobulin was detected by incubation with avidin-biotin-horseradish peroxidase (Vector Laboratories, Inc., Burlingame, CA) and developed using o-phenylenediamine. Analysis of ␣ 4 ␤ 1 expression on transfected K562 cells was performed with FITCconjugated anti-human CD49d mAb 1382f (Chemicon International) in PBS. Sequence alignments were performed using Vector NTI Align X software (InforMax, Bethesda, MD).

The Disintegrin-like Domain of MDC-L Supports Adhesion of
Lymphocytes-Lymphocyte-cell interactions, including those with other lymphocytes, are critical for normal immune surveillance and regulation of the immune response. The lymphocyte-expressed protein MDC-L (ADAM 28) is a member of the ADAM family of metalloproteases that possess disintegrin-like domains postulated to act as integrin ligands. To investigate whether membrane-associated MDC-L is capable of interacting with leukocyte adhesion receptors, a recombinant Fc-fusion protein (rDis/EGF-Fc) lacking the metalloprotease and cytoplasmic regions (Fig. 1a) was expressed and examined for the ability to support adhesion of a T-lymphoma cell line. Due to the extensive number of disulfide bonds predicted to exist within the disintegrin-like and other domains of MDC-L, the recombinant proteins were produced using transfected insect cell lines so that proper folding and appropriate disulfide bonds would be expected to occur during secretion. The purified rDis/ EGF-Fc protein, which migrated as a single monomeric band on SDS-PAGE under non-reduced conditions (Fig. 1b), supported adhesion of Jurkat cells in a concentration-dependent and saturable manner (Fig. 1c). The adhesive component of MDC-L was further refined to the disintegrin-like domain. Jurkat cell adhesion to microtiter wells coated with a disintegrin domain-Fc fusion protein (rDis-Fc) exhibited a concentration dependence similar to that observed for rDis/EGF-Fc (Fig.  1). A recombinant prodomain/metalloprotease Fc-fusion protein (rPro/Met-Fc) did not support adhesion (Fig. 1c), indicating that the Fc component was not responsible for the observed result.
A mAb, designated Dis1-1, raised against a mixture of recombinant MDC-L domains was found to inhibit adhesion of Jurkat cells to rDis-Fc in a concentration-dependent manner (Fig. 2a). A mAb, EGF1-2, from the same fusion and of the same isotype as Dis1-1 did not inhibit cell adhesion (Fig. 2a). The epitope for mAb Dis1-1 was localized between MDC-L residues Gly 418 and Glu 476 based on positive reactivity in an ELISA with various MDC-L recombinant proteins, including the MDC-L disintegrin domain and a region loosely based on the smaller snake venom disintegrin peptides (Mini-Dis) produced in E. coli (Fig. 2, b and c). Because mAb Dis1-1 inhibited adhesion of Jurkat cells to rDis-Fc, the recombinant disintegrin domain (rDis) and Mini-Dis were tested for the ability to support Jurkat cell adhesion. Unlike the recombinant protein rDis-Fc, the E. coli produced Mini-Dis and rDis may not possess the correct disulfide bonding associated with the disintegrin-like domain, because these were expressed within the reduced environment of the bacterial cytosol. However, both the rDis and the 59-residue Mini-Dis supported Jurkat cell adhesion in a concentration-dependent manner comparable with rDis-Fc suggesting sufficient refolding and disulfide bond formation in the E. coli produced proteins for receptor recognition (Fig. 2d).
These results indicate that an adhesive component in MDC-L is located in the disintegrin domain between residues Gly 418 and Glu 476 . ␣ 4 ␤ 1 Recognition of MDC-L-Several studies (18 -30) have demonstrated the potential for integrin-ADAM interactions. The adhesion of Jurkats to rDis-Fc and the Mini-Dis was dependent on the presence of divalent cations as attachment was inhibited by 5 mM EDTA. Taken together, this suggested that a member of the integrin family was involved in adhesion of Jurkats to the disintegrin domain of MDC-L. To elucidate which integrin(s) may be required for attachment to the disintegrin domain of MDC-L, function-blocking anti-integrin ␤-subunit mAbs were tested for the ability to inhibit Jurkat cell adhesion. Two distinct function-blocking anti-␤ 1 mAbs were found to inhibit Jurkat cell adhesion to rDis-Fc, whereas a function-blocking anti-␤ 2 mAb had no effect (Fig. 3a). Functionblocking anti-␣-subunit mAbs known to form a heterodimer with ␤ 1 and found to be expressed on the surface of Jurkat cells by flow cytometry (data not shown) were tested for the ability to inhibit adhesion to rDis-Fc. Only the anti-␣ 4 mAb P1H4 significantly inhibited Jurkat cell adhesion to rDis-Fc (Fig. 3a). This was not due to the quantity of ␣ 4 ␤ 1 receptors expressed on the surface of Jurkats because a non-inhibitory anti-␣ 4 mAb had no effect on adhesion to the Mini-Dis (Fig. 3b). These results indicate that ␣ 4 ␤ 1 participates in the adhesion of Jurkats to the disintegrin domain of MDC-L.
Cell adhesion to the disintegrin domain was found to be dependent on the presence of ␣ 4 ␤ 1 and not unique to the Jurkat cell line. The hepatocellular carcinoma cell line HepG2 and two cell lines of hematopoietic origin were examined with respect to adhesion on rDis-Fc, Engelbreth-Holm-Swarm sarcoma laminin, type I collagen, and the RGD containing 9th and 10th type III repeats (3fn9 -10) of fibronectin (Fig. 4). All three cell lines adhered to recombinant 3fn9 -10; however, only the ␣ 4 ␤ 1 -expressing Burkett's lymphoma cell line, Daudi, demonstrated substantial adhesion to rDis-Fc. Transfected CHO cells expressing human ␣ IIb ␤ 3 , ␣ v ␤ 3 , ␣ 2 , ␣ 3 , ␣ 5 , and ␣ 6 did not significantly attach to the Mini-Dis above background (data not shown). However, a transfected K562 cell line expressing the human ␣ 4 -subunit was able to adhere to the Mini-Dis in a concentration-dependent manner, whereas non-transfected K562 cells did not recognize the Mini-Dis (Fig. 5).
Cell Surface ␣ 4 ␤ 1 Binds the MDC-L Disintegrin Domain in Solution-Ligands recognized by ␣ 4 ␤ 1 include the alternatively splice CS-1 region of fibronectin and VCAM-1. A synthetic CS-1 peptide (EILDVPST) inhibited ␣ 4 ␤ 1 -dependent adhesion of Jurkats to the Mini-Dis in a concentration-dependent manner suggesting that MDC-L binds to the same ligand-binding site as the fibronectin CS-1 region (Fig. 6). Also consistent with attachment to the rDis-Fc being ␣ 4 ␤ 1 -dependent, an RGD-containing peptide required a 10-fold higher concentration (500 g/ml) than the CS-1 peptide to obtain 50% inhibition of Jurkat cell adhesion. Soluble VCAM-1 inhibited adhesion to the Mini-Dis by only 25% at the maximum concentration tested (20 g/ml).
Because the interaction of ␣ 4 ␤ 1 with MDC-L would be expected to occur at the cell surface, we examined whether soluble rDis-Fc would bind ␣ 4 ␤ 1 on the Jurkat cell surface by flow cytometry. Increasing concentrations of rDis-Fc were combined with cells in solution, and the bound rDis-Fc was quantitated using FITC-labeled goat anti-human Fc-specific antibody. The rDis-Fc bound in a concentration-dependent and saturable manner (Fig. 7a). Approximately 350 nM rDis-Fc was required to achieve 50% maximal binding. As seen in the adhesion experiments, the binding of rDis-Fc in solution was inhibited by the CS-1 peptide (Fig. 7b). Thus, the disintegrin domain of MDC-L was recognized by ␣ 4 ␤ 1 when presented in solution.
The Putative Disintegrin Loop Is Not Required for ␣ 4 ␤ 1 Binding-Previous studies investigating the interaction of snake venom disintegrin peptides and disintegrin-like domains of ADAMs have used peptides to localize the integrin recognition site to a region designated as the "disintegrin loop" (21)(22)(23). A synthetic peptide (KGGVCRPAKDECDL) based on the putative disintegrin loop of MDC-L did not inhibit adhesion of Jurkats to rDis-Fc at 1 mg/ml (data not shown). To rule out potential conformational requirements that the peptide may not mimic, two residues, Glu 471 and Asp 473 , within the DECD sequence of rDis-Fc were individually substituted with alanine and compared with the native sequence for the ability to support cell adhesion (Fig. 8). Neither alanine substitution had any significant effect on Jurkat cell adhesion, suggesting that the residues required for ␣ 4 ␤ 1 recognition of MDC-L are located outside the putative disintegrin loop.
Binding of MDC-L Requires Activation of ␣ 4 ␤ 1 -Ligand binding to ␣ 4 ␤ 1 has been shown to be regulated by affinity modulation (42)(43)(44)53). Because our experiments had been carried out in the presence of 1 mM MnCl 2 , ␣ 4 ␤ 1 would be expected to be in the high affinity state (42,53). We investigated whether ␣ 4 ␤ 1 recognition of the MDC-L disintegrin domain was dependent on activation of the integrin. As shown in Fig. 9, both cell adhesion and soluble binding of rDis-Fc required the presence of MnCl 2 . However, an anti-␤ 1 -activating mAb, QE2E5 (45), significantly enhanced binding of Jurkats to rDis-Fc in the absence of MnCl 2 , indicating that adhesion required affinity modulation of a ␤ 1 integrin and was not an effect of MnCl 2 alone (Fig. 9a). Furthermore, binding of rDis-Fc to Jurkat cells in the absence of MnCl 2 could be induced by fucoidin crosslinking of L-selectin (data not shown) (43). These experiments demonstrated that ␣ 4 ␤ 1 binding of the MDC-L disintegrin domain is regulated by affinity modulation of the integrin. DISCUSSION In this study we have demonstrated that the disintegrin-like domain of MDC-L (ADAM 28), a member of the ADAM protein family expressed on the surface of lymphocytes, is recognized by the integrin ␣ 4 ␤ 1 . The disintegrin-like domain of MDC-L alone or in association with the cysteine-rich and EGF-like domains supported attachment of ␣ 4 ␤ 1 -expressing cell lines. Cell adhesion and receptor binding of recombinantly expressed MDC-L disintegrin-like domain was inhibited by functionblocking anti-␣ 4 and anti-␤ 1 mAbs, as well as the ␣ 4 ␤ 1 ligand mimetic peptide, CS-1. Similar to Jurkat binding of fibronectin and VCAM-1, recognition of the MDC-L disintegrin domain required exogenous activation of ␣ 4 ␤ 1 by addition of Mn 2ϩ or an anti-␤ 1 "activating" mAb (44). This is the first description of ␣ 4 ␤ 1 recognition of an ADAM protein and suggests that ␣ 4 ␤ 1 -MDC-L binding may be involved in lymphocyte-leukocyte interactions and/or regulating proteolytic activity by sequestering or targeting the active protease.
Small non-enzymatic disintegrin peptides from snake venoms act as antagonists of integrin function, and the disintegrin-like domains of several mammalian ADAMs have been shown to interact with ␣ 6 ␤ 1 , ␣ v ␤ 3 , ␣ 5 ␤ 1 , ␣ 9 ␤ 1 , and ␣ v ␤ 5 (18 -30). We hypothesized that the disintegrin-like domain of MDC-L (ADAM 28) may also interact with this receptor family. Because human MDC-L was predominantly expressed on the surface of lymphocytes and the interaction of lymphocytes with other leukocytes is critical for normal immune response, attachment of the Jurkat T-lymphoma cell line to a recombinant MDC-L protein lacking the metalloprotease domain and cytoplasmic region was initially examined. By using various anti-integrin function-blocking mAbs, ligand mimetic peptides, and transfected cell lines, we determined that ␣ 4 ␤ 1 was specifically responsible for the attachment of Jurkats to MDC-L disintegrin-like domain coated onto a plastic surface. However, ␣ 4 ␤ 1 -MDC-L interactions would likely occur with MDC-L in the context of the cell membrane or possibly with a secreted form of this ADAM. Analysis of soluble rDis-Fc binding to Jurkats by flow cytometry demonstrated that this disintegrin-like domain bound ␣ 4 ␤ 1 in a concentration-dependent and saturable manner with an apparent K d of ϳ350 nM. Although this is only a preliminary estimate and requires a more rigorous examination, binding of rDis-Fc by ␣ 4 ␤ 1 appears less robust than recombinant soluble VCAM-1 D1-D2 2 but comparable with fibronectin binding analyzed in a similar manner (44). Our Fc fusion proteins lacked the hinge region and therefore did not form covalently linked dimers that might lead to enhanced avidity of the recombinant protein. It appears from the results presented here that MDC-L is capable of binding ␣ 4 ␤ 1 with reasonable avidity suggesting that this interaction may have important physiological ramifications.
We identified an anti-MDC-L mAb, Dis1-1, that inhibited adhesion of Jurkats to the disintegrin-like domain of MDC-L. The epitope for this mAb was localized between MDC-L residues (Gly 418 -Glu 476 ), and this 59-residue region was able to support adhesion of cells expressing ␣ 4 ␤ 1 . There were no sequences homologous to the known ␣ 4 ␤ 1 recognition motifs QIDSP (VCAM-1), EILDVPST (fibronectin), or MLDG (Echis carinatus EC3 disintegrin peptide) (58) within the disintegrinlike domain of MDC-L. This suggests that a unique ␣ 4 ␤ 1 recognition site exists in MDC-L. Interestingly, another recently identified ligand for ␣ 4 ␤ 1 , ICAM-4, was not recognized via an LDV motif; however, the recognition site in ICAM-4 was not identified (59). The integrin recognition site in several of the snake venom disintegrin peptides has been localized to sequences within an extended loop of these peptides. Similarly, ␣ 4 ␤ 1 recognition site is located between residues Gly 418 and Glu 476 and outside the disintegrin loop.
It is becoming apparent that the affinity states of many integrins are physiologically regulated. The affinity of ␣ 4 ␤ 1 can be increased by the exogenous addition of Mn 2ϩ or activating antibodies. Similar to results using soluble VCAM-1 and fi-bronectin, MDC-L disintegrin-like domain binding to the surface of Jurkats also requires exogenous ␣ 4 ␤ 1 activation. This is similar to results found for ␣ 9 ␤ 1 -dependent adhesion to recombinant ADAMs 12 and 15 but in contrast to ␣ 6 ␤ 1 recognition of ADAM 2 (19,27). ADAM 2 on the surface of sperm has been reported to have a divalent cation requirement distinct from  9. Recognition of the disintegrin domain of MDC-L requires activation of the integrin ␣ 4 ␤ 1 . a, adhesion of Jurkat cells (2 ϫ 10 5 cells/well) to 10 g/ml rDis-Fc was examined in HEPES/Tyrode's buffer containing 1 mM MnCl 2 (white bars) or without MnCl 2 (black bars) in the absence (Alone) and presence of 5 g/ml of the function-blocking mAb P1H4 or 5 g/ml activating mAb QE2E5. Adherent cells/well ϭ adherent cells (rDis-Fc) Ϫ adherent cells (BSA) . Results shown are the average Ϯ S.D. of triplicate determinations. b, binding of rDis-Fc (20 g/ml) to Jurkat cells in solution was examined in the presence (solid line curve) and absence (dashed curve) of 1 mM MnCl 2 by flow cytometry. Bound rDis-Fc was disclosed by addition of FITC-labeled goat anti-human IgG-specific antibody. Jurkat cells with no added rDis-Fc are shown as the shaded curve. This experiment was repeated three times with similar results. ␣ 6 ␤ 1 recognition of laminin (19). Thus, MDC-L binding to ␣ 4 ␤ 1 may be regulated in a manner similar to soluble VCAM-1 and fibronectin.
Because MDC-L and ␣ 4 ␤ 1 both exhibit a predominantly leukocyte expression pattern, interaction between these two cell surface molecules may occur on the same cell surface or at a site of cell-cell contact. The metalloprotease domain of mouse (33) and human 3 MDC-L are active and autocatalytic with regard to cleavage of the regulatory prodomain. The interaction of MDC-L with ␣ 4 ␤ 1 on the same cell surface may act to target this potential sheddase to proteins such as VCAM-1. Interestingly, as shown for fibronectin, MDC-L may interact with a site on ␣ 4 ␤ 1 distinct from VCAM-1 allowing simultaneous binding (36). Conversely, binding of ␣ 4 ␤ 1 on the same cell surface may function to sequester the active protease away from potential target proteins such as L-selectin or FasL. If MDC-L is capable of interacting with ␣ 4 ␤ 1 on a neighboring cell, it may function as a counter-receptor and target the active protease to substrates at the site of cell-cell contact. Determining which of these roles is possible will depend on further analysis of the MDC-L-␣ 4 ␤ 1 interaction. Determining whether ␣ 4 ␤ 1 binding of MDC-L occurs on the same cell surface and/or between cells may shed light on the physiologically relevant substrate for this metalloprotease.
The results presented here demonstrate that the disintegrinlike domain of MDC-L is recognized by the integrin ␣ 4 ␤ 1 and that binding requires affinity modulation of this receptor. This interaction suggests a potential adhesive and proteolytic role for this ADAM in inflammatory and immune processes. Insight into the function of this ADAM should come from further characterization of the ␣ 4 ␤ 1 -MDC-L interaction on the cell surface and identifying the physiological relevant substrate for the metalloprotease.