The interacting binding domains of the beta(4) integrin and calcium-activated chloride channels (CLCAs) in metastasis.

CLCA (chloride channel, calcium-activated) proteins are novel pulmonary vascular addresses for blood-borne, lung-metastatic cancer cells. They facilitate vascular arrest of cancer cells via adhesion to beta4 integrin and promote early, intravascular, metastatic growth. Here we identify the interacting binding domains of endothelial CLCA proteins (e.g. hCLCA2, mCLCA5, mCLCA1, and bCLCA2) and beta4 integrin. Endothelial CLCAs share a common beta4-binding motif (beta4BM) in their 90- and 35-kDa subunits of the sequence F(S/N)R(I/L/V)(S/T)S, which is located in the second extracellular domain of the 90-kDa CLCA and near the N terminus of the 35-kDa CLCA, respectively. Using enzyme-linked immunosorbent, pull-down, and adhesion assays, we showed that glutathione S-transferase fusion proteins of beta4BMs from the 90- and 35-kDa CLCA subunits bind to the beta4 integrin in a metal ion-dependent manner. Fusion proteins from fibronectin and the integrins beta1 and beta3 served as negative controls. beta4BM fusion proteins competitively blocked the beta4/CLCA adhesion and prevented lung colonization of MDA-MB-231 breast cancer cells. A disrupted beta4BM in hCLCA1, which is not expressed in endothelia, failed to interact with beta4 integrin. The corresponding CLCA-binding domain of the beta4 integrin is localized to the specific determining loop (SDL). Again enzyme-linked immunosorbent, pull-down, and adhesion assays were used to confirm the interaction with CLCA proteins using a glutathione S-transferase fusion protein representing the C-terminal two-thirds of beta4 SDL (amino acids 184-203). A chimeric beta4 integrin in which the indicated SDL sequence had been replaced with the corresponding sequence from the beta1 integrin failed to bind hCLCA2. The dominance of the CLCA ligand in beta4 activation and outside-in signaling is discussed in reference to our previous report that beta4/CLCA ligation elicits selective signaling via focal adhesion kinase to promote metastatic growth.

Distinct vascular addresses, so-called addressins, have been implicated in playing a major role in the colonization of select organs by blood-borne cancer cells (for reviews, see Refs. [1][2][3]. A recently discovered pulmonary vascular addressin is the multifunctional hCLCA2 1 molecule (4), a member of a putative Ca 2ϩ -activated chloride channel family (for reviews, see Refs. 5 and 6). This molecule has been identified by a monoclonal antibody selected for its ability to block the adhesion between lung-metastatic cancer cells and lung matrix-modulated bovine aortic endothelial cells expressing the functional counterpart of hCLCA2, i.e. bCLCA2 (previously named Lu-ECAM-1) (7)(8)(9). Human CLCA2 is expressed on the luminal surface of endothelial cells lining distinct pulmonary vascular branches (e.g. arterioles and postcapillary venules) (4,7,9) where it mediates vascular arrest of lung metastatically competent cancer cells and promotes early intravascular tumor colony growth (4,10). The surface molecule by which lung-metastatic cancer cells recognize hCLCA2 (and its murine equivalents mCLCA1 and mCLCA5) is the ␤ 4 integrin, establishing for the first time a cell-cell adhesion function for this integrin that involves an entirely new adhesion partner (4). Accordingly, all lung-metastatic human and mouse cancer cell lines available to us (e.g. MDA-MB-231, 4T1, CSML-100, B16-F10, and LLC) prominently expressed this integrin on their surfaces, while tumor cells that were unable to colonize the lungs upon tail vein injection of syngeneic or xenogeneic immunocompromised mice (e.g. MCF7, T47D, 67NR, and CSML-0) did not express ␤ 4 integrin and were unable to bind to endothelial CLCAs. 2 Adhesion to endothelial CLCAs is augmented by an increased surface expression of the ␣ 6 ␤ 4 integrin in cancer cells selected in vivo for enhanced lung colonization (4) but abolished by the specific cleavage of the ␤ 4 integrin with matrilysin (4). ␤ 4 / hCLCA2 adhesion-blocking antibodies directed against either of the two interacting adhesion molecules inhibit lung colonization (4,7,8), while overexpression of the ␤ 4 integrin in Kirsten murine sarcoma virus-transformed Balb/c/3T3 tumor cells significantly increases the lung metastatic performance (4). Association of the ␤ 4 integrin with a metastatic cancer phenotype was further underscored by cDNA microarray analyses in a murine pulmonary metastasis model (11). The ␤ 4 integrin showed a 5.6-fold overexpression in a tumor cell line that was selected for increased lung metastatic performance and that had a spontaneous lung metastatic rate of 93.3% versus the parental cell line that exhibited only a 33.3% spontaneous lung metastatic rate (11).
Despite intense scrutiny (4,10), the role of ␤ 4 /CLCA in metastasis of human cancers has been met with skepticism. This skepticism is fueled by the "unusual" binding interaction of the ␤ 4 integrin involving a putative chloride channel protein to achieve pulmonary vascular arrest of blood-borne cancer cells and to promote early metastatic growth and by the notion that ␤ 4 is perceived as a "non-obligatory" molecule in metastasis. However, integrins are rapidly emerging as important regulators of ion channels (for a review, see Ref. 12). In some cases, such regulation involves the direct interaction between integrin and channel protein as exemplified by the binding of the ␤ 1 integrin to the RGD domain of inwardly rectifying K ϩ channels (GIRK1 and GIRK2) (13). Moreover increasing numbers of reports identify plural roles for ion channels that reach beyond their channel function (for reviews, see Refs. 5, 6, and 12). A classical example of this functional diversity is the Na ϩ /H ϩexchanger. In addition to regulating intracellular pH homeostasis and cell volume, the Na ϩ /H ϩ -exchanger NHE1 acting independently of its transporter function is critical for the dynamic reorganization of the cortical cytoskeleton in response to extracellular signals, regulating the assembly of focal adhesions, the formation of actin stress fibers, and the cell shape and serving as an anchor of the actin-based cytoskeleton to the plasma membrane (for a review, see Ref. 14). Such a multifunctional role has also been reported for the cystic fibrosis transmembrane conductance regulator (15,16) and is rapidly emerging for CLCA family members (17,18). In addition to their putative Ca 2ϩ -activated chloride secretion (19), CLCA molecules also serve as tumor suppressor genes (20 -22) and as potent adhesion and signaling molecules in metastasis (3)(4)(5)(6)(7)(8)(9)(10).
A similar controversy surrounds the status of the ␤ 4 integrin as an obligatory molecule in cancer metastasis, albeit the ␤ 4 integrin has been associated with malignant progression and metastasis in several human cancers (for a review, see Ref. 23). In breast cancer, the association between ␤ 4 integrin and metastasis has been studied most thoroughly, and the reported data are consistent with a mechanistic role of the ␤ 4 integrin in the complex process of metastasis. These data include the following. (i) The ␤ 4 integrin is associated consistently with breast cancers originating from basal cells, which are well known for their aggressive behavior including metastasis, but rarely with the more "benign" tumors of luminal origin (24). (ii) Individuals that express both ␤ 4 and laminin-5 in their primary tumors have the poorest prognosis among breast cancer patients (25). (iii) More than 50% of dormant cancer cells isolated from bone marrow express the ␣ 6 and/or ␤ 4 integrin subunits (26). (iv) Lymph node metastases originating from ␤ 4 -negative primary breast cancers often stain positive for the ␤ 4 integrin (27). (v) The MDA-MB-231 human breast cancer cell line known for its invasive and metastatic behavior prominently expresses ␤ 4 integrin (4). Other cancers also provide support for a link between ␤ 4 and metastasis including the expression of ␤ 4 at the invasive front of gastric cancers (28), the de novo expression and association with lymph node metastasis of ␤ 4 integrin in papillary thyroid cancers (29), and the coexpression of ␤ 4 and its ligand laminin-5 in colon cancers (30).
Here we identify the interacting binding domains of the ␤ 4 integrin and CLCA-type vascular addressins. We show that ␤ 4 recognizes a common binding motif that is present in both the 90-and 35-kDa subunits of CLCA addressins including hCLCA2, mCLCA1, mCLCA5, and bCLCA2 (Lu-ECAM-1) (for cloning and partial structural and functional characterization of these molecules, see Refs. 5, 6, and [17][18][19]. The ␤ 4 recognition site is located within the specific determining loop (SDL) of the I-domain of the ␤ 4 integrin subunit (31)(32)(33). A dramatic inhibition of metastasis of MDA-MB-231 breast cancer cells by a fusion protein containing the ␤ 4 -binding motif of hCLCA2 further advances the role of the ␤ 4 /CLCA adhesion in pulmonary metastasis.
␤ (4-1-4) Chimeric Integrin-Amino acids 184 -203 of the SDL of the ␤ 4 I-domain were substituted for the corresponding sequence of the ␤ 1 integrin (amino acids 197-219) (31) by PCR using the unique restriction sites NdeI in the RcCMV vector backbone and BspMI in the ␤ 4 cDNA with the high fidelity DNA polymerase Herculase (Stratagene, La Jolla, CA). The sequence of the chimeric integrin was verified.

Cell Lines and Transfections
The MDA-MB-231L breast cancer cell line was from Dr. J. A. Price (The University of Texas M. D. Anderson Cancer Center, Houston, TX), 4T1 was from Dr. F. R. Miller (Karmanlos Cancer Institute, Detroit, MI), and human embryo kidney (HEK) 293 cells were from ATCC (Manassas, VA). All cell lines were grown in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal bovine serum. HEK293 cells were transiently transfected with Myc-tagged hCLCA2, ␣ 6 , ␣ 6 ϩ ␤ 4 , and ␣ 6 ϩ ␤ 4-1-4 , or vector alone using LipofectAMINE TM Plus as described by the manufacturer (Invitrogen). Transfection rates assessed by green fluorescent protein co-transfection were 40 -50%. Cells were used in the various assays 48 h after transfection unless otherwise stated.

Purification of hCLCA2 and ␤ 4 Integrin
Myc-tagged hCLCA2 was immunopurified from transfected HEK393 cells 48 h after transfection, and the ␤ 4 integrin was immunopurified from MDA-MB-231L cells as described previously (4,10). Cells were lysed in Tris-buffered saline (TBS) lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM phenylmethylsulfonyl fluoride, 0.01% aprotinin, 1 mM benzamidine, and 1% octyl-␤-glucoside (OG)) (for 1 h at 4°C), and lysates were centrifuged at 15,000 rpm (for 20 min at 4°C) to remove insoluble materials. Precleared supernatants were mixed with ␣-Myc mAb9E10 (hCLCA2) or ␣-␤ 4 mAb3E1, respectively, and incubated for 4 h at 4°C. Protein G-Sepharose beads were then added to the reaction mixtures and incubated overnight at 4°C. Immune complexes were washed extensively with cold TBS lysis buffer (0.5% OG), and bound protein (hCLCA2 and ␤ 4 integrin) was collected in 100 mM Tris-HCl, pH 8.0, containing 150 mM NaCl, 100 mM glycine, and 0.5% OG. The purity was determined by SDS-PAGE followed by silver staining and/or Western blotting with ␣-Myc mAB9E10 or rabbit ␣-␤ 4 pAbH101, respectively (4, 10). The 35-kDa subunit of hCLCA2 was purified from extracts of HEK293 cells that had been transfected with a double tagged hCLCA2 cDNA construct containing a Myc tag at its N terminus and an HA tag at its C terminus. The 35-kDa protein was purified by ␣-HA immunoaffinity chromatography, while the 90-kDa hCLCA2 protein was purified by ␣-Myc immunoaffinity chromatography from the flow-through of the ␣-HA column.

CLCA Proteins Contain Binding Domains for ␤ 4 -expressing
Tumor Cells-To identify the CLCA sequence that is responsible for the ␤ 4 /CLCA-mediated adhesion of lung-metastatic human (MDA-MB-231) and mouse (4T1) breast cancer cells (4), we first examined the binding behavior of the ␤ 4 /CLCA adhesion-blocking mAb6D3 (7-9) using a series of polypeptides encompassing the length of the 90-kDa subunit of the CLCA prototype bCLCA2 (Lu-ECAM-1) (36) (Fig. 1A). Polypeptides were prepared as GST fusion proteins in E. coli, and the fusion proteins were purified on a glutathione column (Fig. 1B). Antibody 6D3 was able to bind and immunoprecipitate the fusion proteins GST-HX, GST-VX, GST-PX, and GST-BX but not GST-HV, GST-HP, and GST-NE (Fig. 1C). The shortest bCLCA2 fragment recognized by mAb6D3 was GST-BX, localizing the antibody-binding domain to the second extracellular domain of bCLCA2 (36). Next we examined whether the same bCLCA2 fragments that supported binding of the ␤ 4 /CLCA adhesion-blocking antibody also harbor the binding domain for ␤ 4 -expressing MDA-MB-231 cells. To do so, we coated wells of microtitration plates with GST-BX, GST-PX, GST-NE, and GST-HP and seeded coated wells with MDA-MB-231 cells. Analogous to the binding characteristics of the adhesion-blocking antibody, MDA-MB-231 bound to GST-PX and GST-BX but not to GST-HP and GST-NE (Fig. 1D). Binding of MDA-MB-231 to GST-PX and GST-BX was equally as strong as the adhesion to the 90-kDa natural processing product of bCLCA2. Unexpectedly, adhesion of MDA-MB-231 as well as mAb6D3 (data not shown) was not restricted to the 90-kDa protein but was also mediated by the 35-kDa subunit of bCLCA2 (Fig. 1D). Curious whether this adhesion behavior was specific for bCLCA2 or extended to other CLCA proteins, we tested the adhesion of the 90-and 35-kDa subunits of hCLCA2 for adhesion to MDA-MB-231 cells. Both products bound the cancer cells (Fig. 1D) but not mAb6D3, which is consistent with its specificity for bCLCA2 and mCLCA1 (7)(8)(9). Identical binding data were also obtained for other lung-metastatic cancer cell lines including 4T1, B16-F10, and CSML-100 using the 90-and 35-kDa subunits of either bCLCA2 (Lu-ECAM-1) or hCLCA2 in static adhesion assays (data not shown).
The ␤ 4 Integrin-binding Motif of CLCAs-The above adhesion data suggest that the 90-and 35-kDa subunits of hCLCA2 (and bCLCA2) harbor a common binding motif for the ␤ 4 integrin. To test this hypothesis, we used the PROTOMAT motif search program (37). The sequences AFSRISSGTG in the 90-kDa and GFSRVSSGGS in the 35-kDa subunits of hCLCA2 were identified as the single, common motif ( Fig. 2A). The first sequence is located at amino acid residues 479 -488 of hCLCA2 (39), placing it within the GST-BX fragment of bCLCA2 (AFS-RISSRSG) recognized above as the shortest bCLCA2 fragment to mediate binding of lung-metastatic cancer cells (Fig. 1D). The second sequence is located at amino acids 740 -749 of hCLCA2 located near the N terminus of the 35-kDa hCLCA2 (39). To prove that this motif is binding MDA-MB-231 cells via the ␤ 4 integrin, we generated a HA-tagged GST fusion protein of the 90-kDa ␤4-binding motif of hCLCA2 (termed ␤ 4 BM hCLCA2(90) ) and tested its binding ability for the ␤ 4 integrin by ELISA. ␤ 4 BM hCLCA2(90) bound to immobilized ␤ 4 integrin but not to ␤ 1 integrin, ␤ 3 integrin, fibronectin, or BSA. The same result was achieved in pull-down assays (Fig.  2B). Immobilized on glutathione-conjugated agarose beads, ␤ 4 BM hCLCA2(90) pulled down ␤ 4 but not ␤ 1 and ␤ 3 from solutes (Fig. 2C), and, in reverse, ␤ 4 integrin but not ␤ 1 integrin, immobilized by anti-integrin antibodies on protein G-conjugated agarose beads, pulled down soluble ␤ 4 BM hCLCA2(90) (Fig. 2D). Identical results were obtained with the 35-kDa ␤ 4 -binding motif of hCLCA2 (termed ␤ 4 BM hCLCA2 (35) ) (data not shown). To test whether the adhesion of ␤ 4 BM hCLCA2(90) to ␤ 4 integrin was dose-dependent, we coated wells of microtitration plates with a standard concentration of immunopurified ␤ 4 integrin (10 g/ml) and determined the adhesion of increasing concentrations of ␤ 4 BM hCLCA2(90) by ELISA. Our data showed a linear increase in adhesion of ␤ 4 BM hCLCA2(90) from 1 to 1,000 ng (Fig. 3A). This adhesion was dependent upon the presence of Mn 2ϩ , but not Mg 2ϩ or Ca 2ϩ , in the assay medium (Fig. 3B).
␤ 4 BM hCLCA2(90) Binds to Lung-metastatic Cancer Cells and Inhibits Adhesion to hCLCA2-To establish hCLCA290-␤ 4 BM as a ␤ 4 /hCLCA2 adhesion-blocking polypeptide, we first examined the ability of the polypeptide to bind to the surface of lung-metastatic MDA-MB-231 cancer cells. ␤ 4 BM hCLCA2(90) was incubated with tumor cells for 20 min at room temperature, and bound polypeptide was detected by ␣-GST antibody and quantified by FACS analysis. Data showed strong binding of ␤ 4 BM hCLCA2(90) to tumor cell surfaces, while the control polypeptide P14 (37) did not adhere (Fig. 4A, a). The FACS histogram generated by bound ␤ 4 BM hCLCA2(90) was similar to that generated by ␣-␤ 4 antibody staining of MDA-MB-231 cells (Fig. 4A, b), concurring with the interaction between ␤ 4 BM hCLCA2(90) and the ␤ 4 integrin. In accordance, ␤ 4 BM hCLCA2(90) as well as ␤ 4 BM hCLCA2 (35) immobilized on the well bottom of microtitration plates supported adhesion of MDA-MB-231 cancer cells to the same extent as full-length, immunopurified hCLCA2, while BSA and GST did not support tumor cell adhesion (Fig. 4B). Finally ␤ 4 BM hCLCA2(90) and ␤ 4 BM hCLCA2 (35) were tested for their abilities to block the adhesion of lung-metastatic MDA-MB-231 cells to hCLCA2 in vitro. Both ␤ 4 BM hCLCA2(90) and ␤ 4 BM hCLCA2 (35) , preincubated with hCLCA2-coated wells for 20 min at room temperature, completely blocked the adhesion of MDA-MB-231 cells to hCLCA2 (Fig. 4C). The control polypeptides P14 and PEDA were unable to block tumor cell adhesion to hCLCA2. Identical results were obtained for lung-metastatic 4T1 murine breast cancer cells (data not shown). Eight weeks later, animals in the control group exhibited signs of respiratory distress, and the experiment was terminated. Autopsy and lung colony counting revealed a median number of Ͼ100 (from 77 to Ͼ100) tumor colonies in the control group and zero (from 0 to 2) colonies in the ␤ 4 BM hCLCA2(90) -treated animal group (Fig. 5A). Gross examination of the lungs showed numerous tumor nodules throughout the lungs as well as in mediastinal and bronchial lymph nodes in GST-treated mice and normal, tumor-free lung in seven of eight ␤ 4 BM hCLCA2(90)treated mice. Histological examination of the lungs confirmed the gross findings. There was massive tumor involvement in the lungs of GST-treated mice but no evidence of metastatic disease in seven of eight ␤ 4 BM hCLCA2(90) -treated mice. The remaining ␤ 4 BM hCLCA2(90) -treated mouse had two small lung colonies. This outcome was not the result of diminished growth and survival rates of tumor cells exposed to ␤ 4 BM hCLCA2(90) polypeptide (data not shown).
The Specific Determining Loop of the ␤ 4 Integrin Harbors the CLCA-binding Domain-To identify the ␤ 4 sequence that interacted with hCLCA2, we generated a GST-␤ 4 (184 -203)-HA fusion protein (GST-␤ 4 for short) that corresponded to a pre-dicted loop of the ␤ 1 and ␤ 3 integrins shown to be involved in ligand binding (31) (Fig. 6A). This sequence comprises the N-terminal two-thirds of the SDL region of the ␤ 4 integrin subunit (32,33). The corresponding sequence of the ␤ 1 integrin was used to prepare a control GST fusion protein (GST-␤ 1 (197-219)-HA (GST-␤ 1 for short)). These fusion proteins were tested first for their ability to bind hCLCA2 in a modified ELISA. Wells of microtitration plates were coated with GST-␤ 4 , GST-␤ 1 , or GST (all at 10 g/ml), and coated wells were probed for hCLCA2 adhesion. Human CLCA2 adhesion to GST-␤ 4 -coated wells was more pronounced than the binding of hCLCA2 to high protein-binding plastic, while GST-␤ 1 -and GST-coated wells did not support hCLCA2 binding (Fig. 6B). These binding data were confirmed in pull-down assays in which GST-␤ 4 and GST-␤ 1 bound to glutathione-agarose beads were tested for their abilities to pull down hCLCA2 from lysates of hCLCA2-Myc-transfected HEK293 cells. As expected, only GST-␤ 4 but

FIG. 3. Adhesion of ␤ 4 BM hCLCA2(90) to ␤ 4 integrin.
A, ␤ 4 BM hCLCA2(90) , but not GST, adheres to ␤ 4 integrin-coated dishes (15 g/ml) in a dose-dependent manner (assay medium: phosphate-buffered saline ϩ 1 mM MnCl 2 ). B, Mn 2ϩ , but not Mg 2ϩ and Ca 2ϩ , promotes adhesion of ␤ 4 BM hCLCA2(90) (50 ng/ml) to ␤ 4 integrin (15 g/ml). *, p Ͻ 0.01 relative to GST control. not GST-␤ 1 was able to pull down hCLCA2 (Fig. 6C). To examine whether GST-␤ 4 and ␤ 4 BM hCLCA2(90) were the interacting binding domains of the ␤ 4 integrin subunit and hCLCA2, microtitration plates were coated with skim milk (blocking agent), ␤ 4 BM hCLCA2(90) , or the control polypeptide PEDA. Biotinylated GST-␤ 4 selectively bound to ␤ 4 BM hCLCA2(90) but not to PEDA (Fig. 6D). GST-␤ 1 did not bind to any of the three substrates (data not shown). Similarly the chimeric ␤ 4 protein ␤ 4-1-4 in which the C-terminal two-thirds of the ␤ 4 SDL domain were replaced with the corresponding region of the ␤ 1 integrin subunit failed to bind to hCLCA2 (Fig. 6E). Finally synthetic peptides of ␤ 4 (184 -203) and ␤ 1 (207-213) were evaluated for their ability to block the adhesion of MDA-MB-231 and 4T1 breast cancer cells to hCLCA2 and mCLCA1, respectively. The ␤ 4 polypeptide, but not the ␤ 1 polypeptide, blocked adhesion of both MDA-MB-231 and 4T1 cells to the respective human and mouse CLCA proteins (Fig. 7, A and D). Polypeptides were equally efficient in their inhibitory activities when they were preincubated with CLCA-coated wells prior to seeding of tumor cells or when they were present throughout the adhesion assay. Polypeptides had no effect on the binding of tumor cells to placental (Fig. 7B) and EHS (Fig. 7, C and E) laminins.
The 90-kDa Protein of hCLCA1 Harbors a Disrupted ␤ 4 BM and Fails to Bind ␤ 4 Integrin-In contrast to the highly conserved ␤ 4 BMs of the 90-kDa subunits of hCLCA2, mCLCA5, mCLCA1, and bCLCA2, the hCLCA1 90-kDa protein exhibited a disrupted ␤ 4 BM of the sequence AFGALSSGNG in which the amino acids RS are substituted by GA (Fig. 8A). However, hCLCA1 contains a relatively well conserved ␤ 4 BM motif in its 35-kDa processing product (CFSRTSSGGS) (Fig. 8A). Thus, ␤ 4 integrin should not be able to bind to the 90-kDa hCLCA1 protein but might bind to the unprocessed 125-kDa and the processed 35-kDa proteins of hCLCA1. To examine this premise, we transfected HEK293 cells with Myc-tagged hCLCA1 and purified the protein by ␣-Myc immunoaffinity chromatography. Four fractions were collected from the affinity column. Fractions 1 and 4 contained only the 90-kDa processing product, while fractions 2 and 3 contained the 90-kDa processing product as well as the 125-kDa full-length, unprocessed hCLCA1 (the untagged 35-kDa hCLCA1 was lost in the column flow-through) (Fig. 8B). To test these fractions for adhesion of MDA-MB-231 cells, wells of microtitration plates were coated with the four fractions, then seeded, and incubated for 20 min with MDA-MB-231 cells. Tumor cells strongly bound to fractions 2 and 3 but failed to bind to fractions 1 and 4, indicating that they did not recognize the disrupted "␤ 4 BM" sequence but recognized the sequence of the conserved ␤ 4 BM in the 35-kDa fragment of full-length hCLCA1 protein (Fig. 8C). A pull-down assay using GST-␤ 4 immobilized on glutathione-agarose beads confirmed these data showing the inability of ␤ 4 to pull down the 90-kDa hCLCA1 but an excellent pull-down of the 90-kDa hCLCA2 (Fig. 8D). DISCUSSION The CLCA family comprises a multifunctional group of proteins (for reviews, see Refs. [17][18][19]. One of their functions is the cell-cell adhesion of select CLCA molecules (for reviews, see Refs. 5 and 6). Expressed on the surface of pulmonary endo- thelial cells, CLCA molecules (e.g. hCLCA2, mCLCA1, mCLCA5, and bCLCA2) have been shown to serve as vascular addresses for hematogenously disseminating, ␤ 4 -expressing cancer cells (4). By engaging in high affinity bonds with tumor cell ␤ 4 integrin, CLCA molecules mediate vascular arrest and promote early, intravascular, metastatic growth (4, 10). Here we have disclosed the interacting binding domains of the two molecules, using two independent, yet complementary approaches. In the first approach, we took advantage of the availability of an anti-bCLCA2 monoclonal antibody that blocked the adhesion of lung-metastatic, ␤ 4 -expressing cancer cell lines (e.g. B16-F10, 4T1, R3230AC-MET, and MDA-MB-231) to bCLCA2-expressing lung matrix-modulated bovine aortic endothelial cells (7,8). This antibody identified the second extracellular domain of bCLCA2 to harbor its binding site (amino acids 448 -511). Not surprisingly, this region also supported binding of lung-metastatic MDA-MB-231 cancer cells. In the second approach, we examined the 90-and 35-kDa subunits of hCLCA2, which both supported adhesion of MDA-MB-231 cancer cells, for a common binding motif using the PROTOMAT algorithm (38). A single common motif was identified in the two proteins. Complementing the findings of our scanning of bCLCA2 fragments for MDA-MB-231 adhesion, the motif of the 90-kDa protein was located in the second extracellular domain of hCLCA2 (amino acids 479 -488). Detailed biochemical and functional analyses conduced with GST fusion proteins that harbored the motif of the 90-(AFSRISSGTG) or 35-kDa (GFS-RVSSGGS) subunits of hCLCA2 showed that these sequences indeed serve as binding domains for the ␤ 4 integrin. Both motif polypeptides of the hCLCA2 subunits blocked adhesion between tumor cells and hCLCA2 and prevented lung and lymph node colonization of MDA-MB-231 cells using a standard lung colony assay. The importance of the identified motif in ␤ 4 adhesion is underscored by the binding behavior of hCLCA1. This molecule has a disrupted ␤ 4 -binding motif of the sequence AFGALSSGNG. Substitution of two successive amino acids with uncharged polar (Ser) and basic (Arg) side chains with amino acids with nonpolar side chains results in failure to bind ␤ 4 integrin. Thus, the ␤ 4 -binding motif has the following predicted sequence: F(S/N)R(I/L/V)(S/T)S.
Interestingly the ␤ 4 -binding domain of CLCA proteins (hCLCA2, mCLCA5, mCLCA1, and bCLCA1) is located within a von Willebrand's factor A (vWFA)-like domain (41)(42)(43)(44). This domain was discovered by searching the Conserved Domain Database with the structure-sensitive CD Search BLAST tool of the National Center for Biotechnology Information (NCBI) Web site (www.ncbi.nlm.nih.gov:80/Structure/cdd/wrpsb.cgi) using the complete amino acid sequence of hCLCA2 as the query. The remaining CLCA family members produced the same domain match despite considerable overall sequence di-  ). B, hCLCA2 binding assay. Myc-tagged hCLCA2 (3 g/ml) is bound to uncoated wells (black column), GST-(gray column), GST-␤ 1 (197-219)-(dashed column), or GST-␤ 4 (184 -203)-coated wells. Bound hCLCA2 is detected by anti-Myc antibody. hCLCA2 binds to GST-␤ 4 (184 -203)-coated wells as well as uncoated wells (positive control) but not to wells coated with GST-␤ 1 (197-219) or GST. C, pull-down assay. GST-␤ 4 and GST-␤ 1 fusion polypeptides are immobilized on glutathione beads and then tested for pull-down of hCLCA2 from lysates of transfected HEK-293 cells. Notice that only GST-␤ 4 is able to pull down hCLCA2. D, GST-␤ 4 binds to GST-␤ 4 BM hCLCA2(90) . The wells of microtitration plates were coated with ␤ 4 BM hCLCA2(90) or control polypeptide PEDA (10 g/ml each) and tested for binding of biotinylated GST-␤ 4 (30 g/ml) by ELISA. E, chimeric ␤ 4-1-4 fails to bind hCLCA2. ␤ 4-1-4 was generated as described under "Experimental Procedures" and transfected together with ␣ 6 into HEK293 cells (positive control: ␤ 4 ϩ ␣ 6 ; negative control: ␣ 6 ). Lysates from transfected HEK293 cells were incubated with anti-␤ 4 pAbH101-conjugated protein G beads (overnight at 4°C). Beads were then washed and incubated with immunopurified Myc-tagged hCLCA2. Bound material was detected by Western blotting using anti-Myc mAb 9E10. WB, Western blot. tive N-terminal, conserved metal ion-dependent adhesion site (MIDAS) of the sequence DXSXS, which is present in all CLCA members (34,36,45) except hCLCA2 where the domain sequence is DVSSK (39). The identified binding site for the ␤ 4 integrin is located near the C-terminal end of the vWFA domain. Whether the MIDAS domain of CLCAs is of importance in the ␤ 4 /CLCA adhesion is questionable since MDA-MB-231 cells adhere equally well to a GST-bCLCA2 fusion protein, which contains both the MIDAS domain and ␤ 4 BM (e.g. GST-PX), as they do to the GST-BX fusion protein of bCLCA2 or the 35-kDa subunit of bCLCA2 and hCLCA2, which all contain only the ␤ 4 -binding motif but not the MIDAS domain.
The result that the ␤ 4 domain that interacts with the identified ␤ 4 -binding motif of CLCA molecules is located within the SDL domain is not surprising as this non-conserved loop sequence of ␤ integrins has been associated with ligand binding for ␤ 1 and ␤ 3 integrin subunits (31,32,40) and more recently with the binding interaction of the ␤ 4 integrin with laminin-5 (33). Corresponding with the swapping of ␣ V ␤ 1-3-1 , in which the sequence CTSEQNC of the SDL region of the ␤ 1 subunit was replaced with the corresponding CYDMKTTC sequence of the ␤ 3 integrin subunit (31), replacement of SDL amino acids 184 -203 in the ␤ 4 integrin subunit with the corresponding region of the ␤ 1 integrin subunit (residues 197-219) resulted in loss of adhesion of the chimeric ␤ 4 integrin to hCLCA2. The binding interaction between ␤ 4 and CLCA molecules or the ␤ 4 BMs thereof did not require partnership of the ␤ 4 with ␣ 6 (4), which is consistent with previous studies that SDL was not required for ␣ 6 ␤ 4 heterodimer formation (32), and thus appears to be "immune" to ␣-␤ interface-mediated SDL conformational changes required for ligand binding of other integrins (32). Moreover the binding interaction between CLCAs and fulllength ␤ 4 integrin required the presence of Mn 2ϩ , confirming the importance of the MIDAS domain within the I-domain-like structure of ␤ 4 for proper ligand binding (47).
Previous work in our laboratory has associated ␤ 4 ligation to CLCA molecules (e.g. hCLCA2 and mCLCA1) with distinct outside-in signaling via focal adhesion kinase (10). Focal adhesion kinase activation and autophosphorylation occurred in a manner independent of the cooperation of receptor tyrosine kinases (e.g. c-Met and ErbB2 (48,49)) and did not involve phosphorylation of the ␤ 4 integrin (10). A similar mode of focal adhesion kinase activation has recently been reported for ␤ 4 ligation to the transforming growth factor-␤-inducible gene-h3 (␤ig-h3) in astrocytoma cells (46). Identical to signaling elicited by ␤ 4 ligation to hCLCA2, activation of focal adhesion kinase by ␤ 4 ligation to ␤ig-h3 was abolished by an anti-␤ 4 antibody. Most intriguingly, however, was the observation that the amino acid sequence of ␤ig-h3 contained a sequence motif (AFSRASQ) similar to that identified as the ␤ 4 -binding motif of CLCA FIG. 7. Inhibition of the ␤ 4 /CLCA adhesion with a ␤ 4 SDL polypeptide. Adhesion assays were performed as described in detail previously (4,33). In brief, wells of microtitration plates were coated with substrate (hCLCA2 (A), mCLCA1 (D), EHS laminin (C and E), or placental laminin (B)) overnight at 4°C at the indicated concentration, then seeded with MDA-MB-231 (A, B, and C) or 4T1 (D and E) breast cancer cells, and incubated for 20 min at 37°C. The number of bound tumor cells was determined by a colorimetric method (4,33). Polypeptide ␤ 4 (184 -203) and ␤ 1 (197-219) were added to substrate-coated wells and incubated for 30 min at room temperature. Polypeptides were either removed by washing prior to the addition of tumor cells (adhesion blocking) or were present throughout the tumor cell adhesion assay (adhesion competition). Notice a complete inhibition of adhesion (by blocking or competition) was observed for both MDA-MB-231 and 4T1 cells with the ␤ 4 polypeptide but not the ␤ 1 polypeptide. No effect was recorded for the binding to placental and EHS laminins. *, p Ͻ 0.01 relative to adhesion to substrate alone. molecules (AFSRISS). In contrast, laminin-5 ligation to ␤ 4 SDL induces association of the ␤ 4 cytoplasmic tail with the adaptor protein Shc and an as yet unidentified 105-kDa phosphoprotein in outside-in signaling (33). Since laminin-5 binding to ␤ 4 was sensitive to point mutations K177A and Q182L (33), located at the N terminus of the SDL domain and proximal to the loop sequence analyzed in this study (amino acids 184 -203), and since neither the ␣3, ␤3, nor ␥2 chains of laminin-5 contained a ␤ 4 BM similar to that in CLCAs, it is likely that CLCAs and laminin-5 interact with different SDL motifs. However, this needs to be confirmed since changes in loop configuration resulting from the introduced point mutations might have negatively affected the "true" laminin-5 binding site. Nonetheless these data suggest that the ligand type, with which ␤ 4 interacts at the cell surface, is not only responsible for the activation of the cytoplasmic domain of the ␤ 4 integrin subunit but is the determining factor in the character of the elicited outside-in signaling (32,33). This notion is supported by the observation that ␤ 4 incapacitated for ligand binding was only able to signal when clustered artificially by anti-␤ 4 antibodies (33). The discovery of new ␤ 4 ligands (e.g. CLCAs (for reviews, see Refs. 5 and 6) and ␤ig-h3 (46)) has unmasked a new, as yet unrecognized role of the ␤ 4 integrin in cell signaling that appears to be dependent upon a distinct ligand interaction. Thus, future studies will have to address the modality by which different, yet natural ligands (e.g. laminin-5, CLCAs, ␤ig-h3, and possibly others) activate the long cytoplasmic tail of ␤ 4 and elicit distinct signaling responses in health and disease.