THE J O U R N A L OF BIOLOGICAL CHEMISTRY Val. 266. No. 6, Issue of February 25, pp. 3579-3585,1991 Printed in U.S.A.

The CS5 Peptide Is a Second Site in the IIICS Region of Fibronectin Recognized by the Integrin CY&

INHIBITION OF CY&^ FUNCTION BY RGD PEPTIDE HOMOLOGUES’

(Received for publication, July 25, 1990)

A. Paul Mould$, Akira Komoriyae, Kenneth M. Yamadall, and Martin J. HumphriesS From the $Department of Biochemistry and Molecular Biology, University of Manchester Medical School, Manchester M13 9PT, United Kingdom, the §Laboratory of Cell Biology, Division of Cytokine Biology, Center for Biologics Evaluation and Research, Food and Drug Administration, and the llMembrane Biochemistry Section, Laboratory of Molecular Biology, National Cancer Institute, National Institutes of Health, Bethesda, Maryland 20892

The alternatively spliced type I11 connecting segment (IIICS) region of fibronectin contains two distinct sites that support the adhesion of melanoma cells. These sites are contained within the synthetic peptides CS1 and CS5 (residues 1-25 and 90-109 of the IIICS, respectively). Recently, the cellular receptor for the CS1 site has been identified as the integrin heterodimer a4D1. In this report, we have investigated the role of the CS5 sequence in melanoma cell adhesion and the identity of its receptor. Adhesion to CS5, when pre- sented to cells as an immobilized IgG conjugate, was blocked by antifunctional monoclonal antibodies di- rected against either the a4 or Dl integrin subunits, but not by antibodies against other subunits, implying that a481 is also the receptor for CS5. In peptide inhibition experiments, CS5 was inhibitory for melanoma cell spreading on both CS5-IgG and CS1-IgG conjugates; conversely, CS1 inhibited spreading on both CS1-IgG and CS5-IgG. In both cases, peptide inhibition could be outcompeted by increasing the concentration of sub- strate-bound conjugate. These results suggest that CS1 and CS5 are recognized by the same or overlapping sites on a4/31. The minimal active sequence within CS5, the tetrapeptide Arg-Glu-Asp-Val (REDV), is some- what related to the Arg-Gly-Asp-Ser (RGDS) sequence that represents a major active site in the central cell- binding domain (CCBD) of fibronectin. When RGDS peptide homologues were tested for their ability to inhibit spreading of melanoma cells on CS1- and CS5- IgG conjugates, GRGES, GRGES, and REDV were found to be inhibitory, while G R E S had no effect. In contrast, spreading on a fibronectin fragment contain- ing the CCBD was inhibited by GRGES only. GRGDS was also able to elute a4D1 specifically from a CS1 affinity column, confirming directly that a4f11-IIICS interactions are sensitive to peptides containing this recognition motif. Because the minimal active se- quence within CS1 is the tripeptide Leu-Asp-Val (LDV; Komoriya et aZ., manuscript submitted for publication), these findings together define a new adhesive recog- nition sequence, X-Asp-Y, used by a4B1 for binding to fibronectin. The central aspartate residue in this tri- peptide is almost always essential, but some flexibility in the amino acid residues at X (glycine, leucine, or glutamic acid) and Y (serine or valine) is tolerated.

* These studies were supported by grants from the Wellcome Trust and the Wigan and District Cancer Research Committee (to M. J. H.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Potential models for the interaction of the IIICS region with are discussed.

The interactions of cells with extracellular matrix compo- nents are important in determining behavior during cell growth, differentiation, and migration. Many of these inter- actions are mediated by integrin receptors, which are a family of structurally related, cell surface heterodimers (Hynes, 1987; Ruoslahti and Pierschbacher, 1987; Ginsberg et al., 1988; Akiyama et al., 1990; Hemler, 1990). Based on the nature of their /3 subunit, integrins have been classified into at least five groups. For each class, the 0 subunit can complex with a number of different (Y subunits, and the resulting dimers exhibit different ligand-binding specificities.

The adhesion of cells to fibronectin has been studied exten- sively, and several regions of the molecule have been shown to support cell attachment (for reviews, see Mosher, 1988; Humphries et al., 1990; Hynes, 1990). One domain that is recognized by a wide variety of cell types is located close to the center of the fibronectin subunit. Within this central cell- binding domain (CCBD),’ two sequences appear to act syn- ergistically to generate full cell adhesion activity. One is the tetrapeptide RGDS (Pierschbacher and Ruoslahti, 1984a; Ya- mada and Kennedy, 1984) and the second, located by deletion mutagenesis, lies approximately 20 kDa NH2-terminal to RGDS (Obara et al., 1988). The integrin a& has been iden- tified as a major receptor for the CCBD, although at least five other integrin receptors may also recognize this region (for a review, see Humphries et at., 1990).

A second cell adhesive domain lies in the type I11 connecting segment (IIICS, also known as V), a region which is subject to complex alternative splicing of pre-mRNA (Kornblihtt et al., 1985). This domain was identified originally as an attach- ment site for murine melanoma cells (Humphries et aL, 1986), but subsequently it has been shown to support the adhesion of neural crest cells and their derivatives and lymphocytes (Dufour et al., 1988; Humphries et al., 1988; Wayner et at., 1989; Ferreira et al., 1990; Garcia-Pardo et al., 1990). A series of six overlapping synthetic peptides (designated CS1 to CS6) spanning the entire IIICS region have been examined for their ability to promote the spreading of melanoma cells directly and to inhibit spreading on the parent fibronectin molecule. Two peptides, CS1 and CS5, were found to possess significant activity, although CS1 was about 2 orders of magnitude more

I The abbreviations used are: CCBD, central cell-binding domain of fibronectin; IIICS, type I11 connecting segment; mAb, monoclonal antibody.

3579

3580 Integrin cu&-Fibronectin IIICS Interactions

active than CS5 (Humphries et ai., 1986, 1987). CS1 and CS5 are located in separate, independently spliced segments of the IIICS that are expressed in a tissue-specific manner (Schwarz- bauer et aL, 1983, 1985; Kornblihtt et al., 1985; Paul et al., 1986; Sekiguchi et al., 1986; Hershberger and Culp, 1990). This suggests that post-transcriptional modification of fibro- nectin pre-mRNA may regulate the adhesive activity of the mature protein. The minimal active sites within CS1 and CS5 have been narrowed down to the peptides LDV and REDV, respectively (Humphries et al., 1986).' The REDV sequence is present as RGDV in both rat and bovine fibronectins (Schwarzbauer et al., 1983; Skorstengaard et al., 1986), and hence it is notable that both cell-binding domains in fibro- nectin contain at least one active site based on an RGD-type motif and a second site that apparently is unrelated.

Recent studies have identified the adhesion receptor for the CS1 sequence (Wayner et al., 1989; Guan and Hynes, 1990; Mould et al., 1990). Using affinity chromatography on a matrix containing the CSl peptide, both the human mela- noma cell (Mould et al., 1990) and murine lymphocyte (Guan and Hynes, 1990) receptors recognizing this sequence have been isolated and identified as the integrin a4&. In a comple- mentary approach, monoclonal antibodies (mAb) directed against either a4 or PI blocked melanoma cell (Mould et al., 1990) or lymphocyte (Wayner et al., 1989; Guan and Hynes, 1990) adhesion to peptide or protein ligands containing the CS1 sequence. In this report, we now demonstrate that a& is also the adhesion receptor for the CS5 sequence in the IIICS, and we provide evidence that CS1 and CS5 may be recognized by the same or mutually exclusive sites on this integrin.

EXPERIMENTAL PROCEDURES

Materials-Human plasma fibronectin was purchased from the Blood Products Laboratory, Elstree, Hertfordshire, United Kingdom. 110-kDa cell-binding and 38-kDa heparin-binding fragments of fibro- nectin were prepared by thermolysin and trypsin digestion, respec- tively, using established procedures (Zardi et al., 1985; Garcia-Pardo et ab, 1987). The synthetic peptides CS1 (Asp-Glu-Leu-Pro-Gln-Leu- Val-Thr-Leu-Pro-His-Pro-Asn-Leu-His-Gly-Pro-Glu-Ile-Leu-Asp- Val-Pro-Ser-Thr), KKT-CS1-VQK (Lys-Lys-Thr-Asp-Glu-Leu-Pro- Gln-Leu-Val-Thr-Leu-Pro-His-Pro-Asn-Leu-His-Gly-Pro-Glu-Ile- Leu-Asp-Val-Pro-Ser-Thr-Val-Gln-Lys), CS2 (Pro-Ser-Thr-Val- Gln-Lys-Thr-Pro-Phe-Val-Thr-His-Pro-Gly-Tyr-Asp-Thr-Gly-Asn- Gly-Ile-Gln), CS5 (Gly-Glu-Glu-Ile-Gln-Ile-Gly-His-Ile-Pro-Arg- Glu-Asp-Val-Asp-Tyr-His-Leu-Tyr-Pro), and CS6 (Leu-Tyr-Pro- His-Gly-Pro-Gly-Leu-Asn-Pro-Asn-Ala-Ser-Thr) were synthesized using an Applied Biosystems 430A peptide synthesizer and purified as described previously (Humphries et a/., 1986,1987). GRGDS (Gly- Arg-Gly-Asp-Ser), GRGES (Gly-Arg-Gly-Glu-Ser), and REDV (Arg- Glu-Asp-Val) were purchased from Peninsula Labs. (Belmont, CA). GRDGS was purchased from Bachem Inc. (Saffron Walden, Essex, United Kingdom). CS1-IgG, CS2-IgG, and CS5-IgG peptide-protein conjugates were prepared as described (Humphries et al., 1987). The following monoclonal antibodies were obtained as gifts: (a) 12F1 IgG (recognizing a2; from V.L. Woods, University of California San Diego, CA), (6) P3E3 ascites (recognizing a*; from E.A. Wayner, Oncogen, Seattle, WA), (c) mAbl6 and mAbl3 IgG (recognizing a5 and 01, respectively; from S. K. Akiyama and S. S. Yamada, Howard Univer- sity Cancer Center, Washington, D.C., and National Cancer Institute, Bethesda, MD), ( d ) 439-9B ascites (recognizing P4; from S. J. Kennel, Oak Ridge National Laboratory, Oak Ridge, TN), and ( e ) OX8 ascites (recognizing rat CD8; from A. Ager, University of Manchester, United Kingdom).

Cell Spreading"A375-SM cells, a human metastatic melanoma cell line (provided by I. J. Fidler, M.D. Anderson Hospital and University of Texas, Houston, TX) were cultured in Eagle's minimal essential medium containing 10% fetal calf serum, minimal essential medium vitamins, nonessential amino acids, sodium pyruvate, and

* A . Komoriya, L. J. Green, M. Mervic, S. S. Yamada, K. M. Yamada, and M. J. Humphries, manuscript submitted for publication.

glutamine (all from GIBCO) as described (Kozlowski et a/., 1984). Cell-spreading assays were performed in 96-well microtiter plates (Costar) essentially as described previously (Humphries et al., 1986). Briefly, wells were coated for 60 min at room temperature with 100- pl aliquots of adhesion factors diluted with Dulbecco's phosphate- buffered saline, and then sites on the plastic for nonspecific cell adhesion were blocked for 30 min a t room temperature with 100 rl of 10 mg/ml heat-denatured bovine serum albumin (Humphries et al., 1986). A375-SM cells were detached with 0.05% trypsin, 0.02% EDTA, resuspended to 2 X 105/ml in serum-free Dulbecco's minimal essential medium, and allowed to recover for 10 min at 37 "C. For experiments examining the effects of antibodies or peptides on spreading, 50-pl aliquots of the cell suspension were added to wells together with 50 pI of antibodies or peptides diluted with Dulbecco's phosphate-buffered saline and then were incubated in a humidified atmosphere of 5% CO, for 90 min at 37 "C. The cells were then fixed with 3% glutaraldehyde and counted using phase-contrast microscopy for the degree of spreading as described (Humphries et a/., 1986). Each point was obtained by counting 300 cells/well from a number of randomly selected fields. No cell spreading was observed on wells coated only with heat-denatured bovine serum albumin.

Elution of a 4 0 1 from KKT-CS1-VQK-Sepharose-KKT-CS1-VQK peptide was coupled covalently to CNBr-activated Sepharose 4B (Pharmacia-LKB Biotechnology Inc.) according to the manufactur- er's instructions with the exception that 4 mg of peptide was coupled to 3 ml of packed beads. Three confluent 225-cmZ flasks of A375-SM cells were detached with 3 mM EDTA in Hanks' balanced salt solution and iodinated in suspension by the lactoperoxidase method of Lebien et al. (1982). The cells were then extracted with 2 ml of 2% Triton X-100, 25 mM Tris-HCI, pH 7.4, 150 mM NaC1, 1 mM MnC12, 1 mM CaC12, 2 mM phenylmethylsulfonyl fluoride, 10 pg/ml leupeptin, and 5 mg/ml bovine serum albumin. MnC12 was included in all extraction and chromatography buffers because of its reported ability to enhance the affinity of interaction between integrin receptors and their ligands (Gailit and Ruoslahti, 1988). Bovine serum albumin was included as a carrier. The A375-SM extract was centrifuged at 40,000 X g for 30 min; and the supernatant solution was precleared by mixing with 2 ml of packed, underivatized Sepharose 4B for 30 min and then rotated end over end overnight at 4 "C with 3 ml of KKT-CS1-VQK-Sepha- rose beads. The suspension was packed into a 1.6-cm diameter chro- matography column and washed at 4 "C with 60 ml of 1% octylf3-D- glucopyranoside, 25 mM Tris-HCI, pH 7.4, 150 mM NaC1, 1 mM MnC12, 1 mM CaCIz, 2 mM phenylmethylsulfonyl fluoride, and 10 pg/ ml leupeptin (column buffer) at 3 ml/h. The column was eluted sequentially with 6 ml of 0.5 mg/ml G R E S and 6 ml of 0.5 mg/ml G R B S dissolved in column buffer. The column was also washed with 4 ml of column buffer after each peptide elution. 2-ml fractions were collected and 1-ml aliquots precipitated with 10% trichloroacetic acid, redissolved in Laemmli sample buffer (Laemmli, 1970), and subjected to electrophoresis in the absence of reducing agent on a 7.5% polyacrylamide gel.

RESULTS AND DISCUSSION

The Receptor for CS5 Is Integrin a4fi1-0f the two sites in the IIICS that support cell attachment and spreading, CS1 was found to be the stronger by at least 2 orders of magnitude in previous studies with B16-Fl0 melanoma cells (Humphries et al., 1986, 1987). Similarly, CS1 was also much more active than CS5 in promoting the spreading of the human melanoma cell line, A375-SM. The maximal levels of spreading were approximately 80% on CS1-IgG and 30% on CS5-IgG (data not shown). A control conjugate, CS2-IgG, did not support spreading. The concentrations of conjugates required for half- maximal spreading were 5 and 30 pg/ml for CS1-IgG and CS5-IgG, respectively. Thus, although CS5 is a weaker adhe- sive site for human melanoma cell adhesion compared to CSI, it does support significant spreading, and the difference in activity between the two peptides does not appear to be as great as that reported previously for mouse melanoma cell adhesion.

The identification of the adhesion receptor for the CS1 sequence has been achieved recently by affinity chromatog- raphy on the immobilized peptide (Guan and Hynes, 1990; Mould et al., 1990). However, when these experiments were

Integrin a&-Fibronectin IIICS Interactions 3581

loo I 80

40

A

Y

a ANTI-a4 (DILUTION) v1

w 0 l o 0 1

0 0.1 1 1 o/c ANTI-81 @ d m 0

FIG. 1. Effects of anti-al (A) and anti+, ( B ) antibodies on A375-SM cell spreading on CS1-IgG (open circles) and CS5- IgG (closed circles). The antibodies used were P3E3 (ascites rec- ognizing a,) and mAbl3 (IgG recognizing PI). Controls were 12F1 (open triangle, IgG recognizing ap used a t 100 pg/ml), 439-9B (closed triangle, ascites recognizing p4 used a t 1:20 dilution), OX8 (open square, control ascites recognizing rat CD8 used at 1:20 dilution), and normal rat IgG (closed square, used a t 100 pglml). For clarity, the effects of control antibodies on CSI-IgG- and CS5-IgG-mediated spreading are shown in either A or 3, respectively, but not in both panels. Error bars = S.D. C = controls. The coating concentrations of conjugates were 125 pg/ml for CS1-IgG and 150 pg/ml for CS5- IgG. In a separate experiment, mAbl6 (anti-a6), when tested at concentrations up to 100 pg/ml, was found to have no significant inhibitory effect on CS1-IgG- or CS5-IgG-mediated spreading (for CS1-IgG, control spreading was 81 * 6%, and spreading in the presence of mAbl6 was 83 ? 2%; for CS5-IgG, control spreading was 27 * 1%, and spreading in the presence of mAbl6 was 32 * 5%).

repeated with immobilized CS5, no radiolabeled receptor ma- terial was retained by the column. This was true even when the chromatography was performed in the presence of Mn2+ t o enhance the affinity of interaction of integrin receptors with their ligands (Gailit and Ruoslahti, 1988). The inability of CS5 to retain its receptor even though the peptide is active in adhesion assays indicates that the affinity of interaction of CS5 with the cell surface is low, and below the threshold needed for affinity isolation of a receptor. The low maximal level of spreading supported by CS5 (plateau at 30%; see above) is also consistent with this interpretation.

As an alternative approach to identifying the CS5 receptor, however, we investigated the ability of antifunctional mono- clonal antibodies directed against integrin subunits to inhibit A375-SM spreading on CS5-IgG comparedwith CS1-IgG (Fig. 1). Spreading on both conjugates was sensitive to monoclonal antibodies P3E3 and mAbl3 (directed against a., and Dl, re~pectively).~ The other anti-integrin antibodies tested were without effect (anti-a2, anti-a,, and anti-&; Fig. 1). These results strongly suggest that CS5, like CS1, is recognized by the integrin a&.

CSl and CS5 Share Common or Mutually Exclusive Binding

'' In further experiments, two other monoclonal antibodies directed against ea, B5G10 (from M.E. Hemler) and 8F2 (from C. Morimoto), also inhibited spreading on both conjugates (data not shown).

Sites on a4P1-To investigate further the relationship between CS1 and CS5 in binding to (Y&, we next examined the ability of each peptide to inhibit spreading on CS1-IgG and CS5-IgG (Fig. 2). As expected, both peptides were able to autoinhibit spreading on their respective IgG conjugate. However, both peptides were also cross-inhibitory, i.e. CS1 peptide inhibited spreading on CS5-IgG and CS5 peptide inhibited spreading on CS1-IgG. Control peptide CS6 had no effect on spreading on either conjugate (Fig. 2). The concentrations of peptides required for half-maximal inhibition of spreading were 10 gg/ ml CS1 on CS5-IgG, 100 gg/ml CS1 on CS1-IgG, and 1.5 mg/ ml CS5 on either conjugate.

The cross-inhibition by CS1 and CS5 suggests that they may be recognized ( i ) by the same site on a&, ( i i ) by mutually exclusive, overlapping sites, or ( i i i ) by distant, but allosteri- cally connected sites. To test these possibilities further, the ability of substrate-bound CS1 or CS5 (as IgG conjugates) to outcompete the inhibition of spreading by the opposing CS peptide was examined (Table I). This approach has been used previously to analyze the noncompetitive inhibition of fibro- nectin-mediated cell adhesion caused by soluble collagen (Na- gata et al., 1985). At high concentrations of substrate-bound CS1 or CS5, the levels of cell spreading approached those obtained in the absence of CS5 or CS1 inhibitor. The inhibi- tion of spreading by CS1 on CS5-IgG was never fully com- peted, however, perhaps due to the extreme sensitivity of CS5-IgG-mediated spreading to CS1 (see Fig. 2). If CS1 and CS5 did bind to spatially separate sites on a&, it would not be expected that the inhibition of adhesion induced by one peptide would be overcome by an excess of the other. The competitive nature of the inhibition by CS1 and CS5, there- fore, suggests instead that the two peptides either share a common binding site or bind a t two overlapping sites on a&.

a& Interactions with CS1 and CS5 Are Inhibited by RGD 100 r

I

0 10 100 1000 PEPTIDE (pg/ml)

FIG. 2. Inhibition of A375-SM cell spreading on CS1-IgG (solid lines) and CS5-IgG (dashedlines) by CS1 (open circles), CS5 (closed circles), and CS6 (closed triangle on CS1-IgG; open triangk on CS5-IgG). The coating concentrations of conju- gates were the same as in Fig. 1.

TABLE I Effects of the concentration of CSl-ZgG and CS5-IgG conjugates

coated onto the substrate on CS5- and CS1-mediated inhibition of A375-SM cell spreading

Conjugate concentration

Inhibition by Inhibition by 0.5 mg/ml 5 r d m l

CS5 on CS1-I& CS1 on CS5-lgG

d m l 0.2 0.9 2.7 9.0

27.0 90.0

270.0

% 67 57 15 9 0 4 1

% N.D." N.D. N.D.

62 40 25 18

a N.D., not determined.

3582 Integrin a&-Fibronectin IIICS Interactions

I A

0.4 1 c \ \

0 0.2 0.4 0.6 0.8 1.0

PEPTIDE (rng/ml)

FIG. 3. Inhibition of A375-SM cell spreading on CSl-IgG (A) and CS5-IgG ( B ) by RGD peptide homologues. The abscissa shows concentration of GRGDS (open circles), GRGES (closed cir- cles), G R E S (open triangles), and REDV (closed triangles). The coating concentrations of the conjugates were the same as in Fig. 1. Results are presented relative to the degree of spreading achieved in the absence of peptide inhibitors; this was 76% for CS1-IgG and 32% for CS5-IgG.

Peptide Homologues-RGD peptides have been reported pre- viously to inhibit spreading of B16-FlO melanoma cells on fibronectin completely (Humphries et al., 1986). Because these cells recognize the IIICS region almost exclusively, this sug- gests that IIICS-IIICS receptor interactions are sensitive to RGD peptides. To investigate further the nature of the CS1 and CS5 binding sites on arb1, the ability of a series of RGD peptide homologues to inhibit A375-SM cell spreading on CS1-IgG and CS5-IgG was tested (Fig. 3). The pattern of inhibition on both conjugates was essentially identical; GRGDS and CRGES were both inhibitory (half-maximal inhibition at approximately 200 pg/ml), while G R E S was without effect. The peptide REDV, which represents the minimal active site in CS5, was moderately inhibitory for spreading on both conjugates. This pattern of inhibition is the same as that previously reported for B16-Fl0 cell spread- ing on either fibronectin or unconjugated CS5 (Humphries et al., 1986). To ensure that the peptides were not causing nonspecific inhibition of cell spreading, we also compared the pattern of inhibition of spreading by RGD-containing pep- tides on two protease fragments of fibronectin, a thermolysin fragment of 110 kDa containing the CCBD (Zardi et al., 1985) and a 38-kDa tryptic fragment containing the COOH-termi- nal heparin-binding domain and the first 67 residues of the IIICS ( i e . CS1-CS3; Garcia-Pardo et al., 1987). This experi- ment allows direct comparison of the peptide sensitivity of CCBD receptors (principally a&) with the receptor for the

As shown in Fig. 4, the pattern of inhibition of spreading on the 38-kDa fragment and CS1-IgG was identical and the

' In antibody inhibition experiments, A375-M spreading on the 38- kDa fragment was inhibited completely by antibodies P3E3 (recog- nizing CY') and mAbl3 (recognizing PI). In contrast, spreading on the 110-kDa fragment was not affected by P3E3, but was blocked by mAbl3 and mAbl6 (the latter directed against as; Mould and Hum- phries, unpublished work).

IIICS (a4P1).4

A

a

0.4 1 \

1 .o

0.8

0.6

0.4

0.2

0 O 0.2 0.4 0.6 0.8 1.0

PEPTIDE (mg/ml)

FIG. 4. Inhibition of A375-SM cell spreading on 110-kDa ( A ) and 38-kDa ( B ) fragments of fibronectin and CSl-IgG (C) by RGD peptide homologues. The key to the peptides is the

kDa), 25 pg/ml (38 kDa), and 125 pg/ml (CS1-IgG). Results are same as in Fig. 3. The coating concentrations were 20 pg/ml (110

presented relative to the degree of spreading achieved in the absence of peptide inhibitors; this was 89% for 110 kDa, 42% for 38 kDa, and 43% for CS1-IgG. In a separate experiment, CS1 was found to have no effect on the adhesion of A375-SM cells to the 110-kDa fragment (control spreading was 84 -t 4%, spreading in the presence of 1 mg/ ml CS1 was 81 k 2%).

same as in Fig. 3, but distinctly different from that on the 110-kDa fragment. GRGPS was a potent inhibitor of spread- ing on the 110-kDa fragment with half-maximal inhibition at approximately 30 pg/ml (60 p ~ ) ; the other peptides tested were either noninhibitory ( G R E S ) or only slightly inhibi- tory (GRGES and REDV). The specific requirement for the (G)RGDS sequence to inhibit CCBD receptors has been shown previously (Pierschbacher and Ruoslahti, 1984, a and b, 1987; Yamada and Kennedy, 1985). The results in Fig. 4 demonstrate that the effects of GRGDS, GRGES, and REDV on A375-SM cell adhesion are specific and suggest that the ability of these peptides to inhibit spreading on IIICS ligands is due to inhibition of the a,@, complex. Although RGD- containing peptides disrupt adhesion to the IIICS, a similar cross-inhibition is not seen with larger, more active polypep- tide reagents from the CCBD (Humphries et al., 1986). Simi- larly, the high activity CS1 peptide has no effect on CCBD- dependent adhesion (Fig. 4, legend). Taken together, these findings are consistent with the conclusion that both cell- binding domains employ similar mechanisms of interaction with their respective receptors (since generic RGD-containing peptides block the function of both regions), but that in their native state in the parent molecule the active sites in each domain do not compete with each other for receptor recogni- tion.

As a final test of the involvement of an RGD-type motif in

Integrin cY&-Fibronectin IIICS Interactions 3583

G R E S G R B S 14 15 16 17 18 19 20 21 22 23

220-

100-

68-

43-

FIG. 5. Affinity chromatography of detergent-extracted, surface-iodinated A375-SM cells on KKT-CSl-VQK-Sepha- rose. Fractions 15-17 represent material eluted with control G R G S peptide, and fractions 20-22 are material eluted with active G R e S peptide. The two bands observed correspond to the subunits of‘ the tu4/$ integrin dimer.

binding to the IIICS, we investigated whether the pro- totype RGD-containing peptide, GRGDS, could elute a4PI from a CS1 affinity column. Although A375-SM cells contain a number of integrins, only is detected after elution of this affinity matrix (a) with the specific competitor CS1 peptide or ( b ) with EDTA which disrupts the interaction of all known integrins with their ligands (Mould et al., 1990).5 Triton X-100 extracts of surface-iodinated A375-SM cells were applied to a KKT-CS1-VQK-Sepharose column, un- bound material was removed by washing, and the column eluted sequentially with 0.5 mg/ml G R E S and 0.5 mg/ml GRG&S. As shown in Fig. 5, when the control peptide G R E S was added to the elution buffer, no receptor was eluted above background. In contrast, elution with G R g S caused significant elution of the receptor, thereby confirming that the interaction of a4pI with CS1 is sensitive to RGD- containing peptides.

IIICS ligand interactions explains the previous observation that these peptides inhibit B16-Fl0 cell spreading on fibro- nectin (Humphries et al., 1986). It originally was suggested that RGD peptides were able to inhibit spreading because of sequence similarity with the REDV active site in CS5. How- ever, it is now known that CS5 is only present a t very low levels in the plasma fibronectin used for these spreading assays (Hershberger and Culp, 1990). Furthermore, since CS1 is the dominant adhesive site in the IIICS, it now appears that these previous results were due to inhibition of CS1-a4pI interactions by RGD peptides. Our results also confirm other reports (Dufour et al., 1988; Ager and Humphries, 1990) in which GRGES was used as a diagnostic inhibitor of IIICS receptor-ligand interactions. However, our results conflict with a number of studies suggesting that the IIICS receptor is RGD-independent (Liao et al., 1989; Wayner et al., 1989; Garcia-Pardo et al., 1990; McCarthy et al., 1990) or only slightly sensitive to RGD peptides (Guan and Hynes, 1990). While some of these discrepancies may be explained by dif- ferences in the cell type or assay system used, we favor the explanation that, since IIICS receptor-ligand interactions are relatively resistant to GRGDS compared to CCBD receptors (half-maximal inhibition at 350 and 30 pg/ml, respectively; Fig. 4), the concentrations of RGD peptides used in these reports were too low. Indeed, many of these investigations

The ability of RGD peptide homologues to inhibit

‘A. P. Mould and M. J. Humphries, unpublished work.

used cell attachment assays which are known to be less sensitive to the effects of RGD peptides than spreading assays (compare Pierschbacher and Ruoslahti, 1984b with Piersch- bacher and Ruoslahti, 1984a; Yamada and Kennedy, 1984). Superficially, our results also appear to conflict with those of Guan and Hynes (1990) who showed that GRGDSP did not affect binding of a4P1 to CS1-Sepharose at a concentration of 0.3 mg/ml. However, in our system, although neither G R E S nor GRGJS eluted any receptor at a molar concentration equivalent to 0.3 mg/ml GRGDSP (0.25 mg/ml; data not shown), a4p1 was specificially eluted by GRGDS at higher concentrations (0.5 mg/ml) (Fig. 5).

Models for cu4pl-IIICS Interactions-As discussed above, there are several possible explanations for the ability of CS1 and CS5 to cross-inhibit each other’s biological activity. Pos- sible models include: model I, in which CS1 and CS5 interact with exactly the same site on ~y4pI; model 11, CS1 and CS5 recognize different but overlapping (hence mutually exclusive) sites on &@I; and model 111, CS1 and CS5 recognize different sites on (Y4pI but each site can be inhibited in a cis fashion by the opposing peptide. Further experiments will be necessary to distinguish unequivocally between these possibilities. How- ever, model I11 can probably be discounted because CS1 and CS5 appear to be competitive inhibitors of each other’s activ- ity (Table I). At the present time, we favor model I because (a) three different monoclonal antibodies directed against a4 block recognition of both CS1 and CS5 by ( b ) a4pl interactions with CS1 and CS5 have the same pattern of inhibition with RGD peptide homologues (Fig. 3); if different sites were involved in binding CS1 and CS5, the pattern of inhibition would be expected to differ; and (c) the minimal active site in CS1, LDV, is related in sequence to REDV (the active site in CS5); the similarity between these two active peptides suggests that they could be recognized by the same site on a&. The ability of RGD peptide homologues to inhibit binding of CS1 and CS5 to this site would then be explained by their ability to mimic the active LDV/REDV sequences. In summary, although our data favor a model in which CS1 and CS5 are recognized by the same site on (~4pI, we cannot currently exclude the possibility that CS1 and CS5 bind to overlapping sites (model 11).

The essential structural feature common to both ligands and inhibitors of a& is the presence of a central aspartate residue (with the sole exception of GRGES peptide) within a triplet of amino acids (i.e. X-Asp-Y). The flanking X and Y residues are less well conserved, and some flexibility in these amino acids is tolerated (e.g. X can be leucine, glycine, or glutamic acid, and Y can be valine or serine). This situation contrasts markedly with the motif recognized by aspl and a number of other integrins (including the vitronectin receptor and platelet glycoprotein IIb/IIIa) which have a strict require- ment for the RGD tripeptide (Ruoslahti and Pierschbacher, 1987; Ginsberg et al., 1988).

Studies employing a chemical cross-linking approach have recently localized binding sites for RGD peptides on both the a and p subunits of the vitronectin receptor and platelet glycoprotein IIb/IIIa (D’Souza et al., 1988, 1990; Smith and Cheresh, 1988, 1990). In the case of the a subunits (av and a l l b ) , the cross-linked sites were close to putative divalent cation-binding sites which bear strong homology to an EF- hand-type repeat (D’Souza et al., 1990; Smith and Cheresh, 1990). Interestingly, the binding sites on the p subunit of these two receptors (p3) also bear homology to an EF-hand structure, although these sequences do not fit the consensus as well as the sites in the a subunits (D’Souza et al., 1988; Smith and Cheresh, 1988). Recent analyses of a mutant ~ I I I , @ R

3584 Integrin a&-Fibronectin IIICS Interactions

integrin derived from patients with the Cam variant of Glanz- mann’s thrombasthenia have confirmed the importance of these putative cation-binding sites (Loftus et al., 1990). The mutant receptor lacks both ligand recognition and interaction with divalent cations, a phenotype caused by a point mutation in the p3 subunit that converts one of the aspartate residues in the EF-hand consensus to tyrosine. Taken together, these results have now led to the hypothesis that interaction with receptor-bound cations may be a common mechanism for ligand binding to integrins (Edwards et al., 1988; Loftus et al., 1990).

Comparison of the sequence of the a subunit divalent cation-binding sites with the EF-hand consensus reveals an interesting difference. The amino acid occupying the impor- tant -2 coordination position is always a small hydrophobic residue in integrin a subunits, whereas the consensus for a functional EF-hand is either aspartate or glutamate (Kretsin- ger, 1980). This suggests that the aspartate residue in adhesive recognition signals such as RGD, REDV, LDV, and LGGAKQAGDV (an antiadhesive peptide found in fibrino- gen, the aspartate of which is reported to be critical for function; Hawiger et al., 1982; Ruggeri et al., 1986) may play a functional role in cell adhesion by providing the final coordination group for integrins to chelate divalent cation. Presumably the conformation of the EF-hand repeats in different integrin subunits combined with the conformation of the different aspartate-containing active sites in adhesion proteins will be important in determining the adhesive spec- ificity of various receptors. The LGGAKQAGDV and RGDS peptides have already set a precedent for two aspartate- containing peptides binding to the same or mutually exclusive binding sites on an integrin since each competes for the binding of the other to platelet glycoprotein IIb/IIIa (aII&; Lam et al., 1987; Santoro and Lawing, 1987; D’Souza et al., 1990). The same now appears to be true for the interaction of LDV (in CS1) and REDV (in CS5) with (~4p1.

The dual interaction of CS1 and CS5 with a& is of partic- ular interest because both sequences are located in independ- ently spliced segments of the IIICS, and fibronectin variants containing either site alone or both sites have been identified (Kornblihtt et a!., 1984; Bernard et al., 1985; Sekiguchi et al., 1986; Oyama et al., 1989; Hershberger and Culp, 1990). If both sequences do interact with divalent cation-binding sites in

this raises the interesting question of whether both sequences bind to exactly the same EF-hand repeat or whether they interact with neighboring repeats. A fuller understanding of the molecular basis of the interaction of the IIICS with a& may also provide insight into the functional relevance of the CS5 sequence. Since the adhesive activity of CS5 is so much lower than that of CS1, it is currently unclear what its contribution is to the activity of the whole IIICS region. Intriguing possibilities are that binding of a4pI to CS1 or CS5 might transduce different signals to the cell interior or alter- natively that interaction with the low affinity CS5 site might allow rapid migration while interaction with the stronger CS1 site may facilitate immobilization.

In addition to studying IIICS-a4pI interactions, it will also be important to identify the binding sites for other sequences known to interact with a&. These include sites in the COOH- terminal heparin-binding domain of fibronectin (Wayner et al., 1989) and the endothelial cell surface molecule VCAM-1 (Elices et al., 1990). In both cases, the peptide sequences responsible for a& binding activity have yet to be deter- mined, although VCAM-1 does contain the LDV tripeptide in its extracellular domain (Osborn et al., 1989).

Acknowledgments-We would like to express our thanks to I. J.

Fidler for A375-SM cells and to V. L. Woods, E. A. Wayner, S. K. Akiyama, S. S. Yamada, S. J. Kennel, and A. Ager for providing antibodies.

REFERENCES Ager, A., and Humphries, M. J. (1990) Znt. Zmmunol. 2, 921-928 Akiyama, S. K., Nagata, K., and Yamada, K. M. (1990) Biochirn.

Bernard, M. P., Kolbe, M., Weil, D., and Chu, M. L. (1985) Biochem-

D’Souza, S. E., Ginsberg, M. H., Burke, T. A., Lam, S. C.-T., and

D’Souza, S. E., Ginsberg, M. H., Burke, T. A., and Plow, E. F. (1990)

Dufour, S., Duband, J.-L., Humphries, M. J., Obara, M., Yamada, K.

Edwards, J. G., Hameed, H., and Campbell, G . (1988) J. Cell Sci. 89,

Elices, M. J., Osborn, L., Takada, Y., Crouse, C., Luhowsky, S.,

Ferreira, 0. C., Garcia-Pardo, A., and Bianco, C. (1990) J. Exp. Med.

Gailit, J., and Ruoslahti, E. (1988) J. Biol. Chem. 263 , 12927-12932 Garcia-Pardo, A., Rostagno, A., and Frangione, B. (1987) Biochem.

Garcia-Pardo, A., Wayner, E. A., Carter, W. G., and Ferreira, 0. C.

Ginsberg, M. H., Loftus, J. C., and Plow, E. F. (1988) Thromb.

Guan, J.-L., and Hynes, R. 0. (1990) Cell 6 0 , 53-61 Hawiger, J., Timmons, S., Kloczewiak, M., Strong, D. D., and Doolit-

Hemler, M. E. (1990) Anrzu. Reu. Zmmunol. 8, 365-400 Hershberger, R. P., and Culp, L. A. (1990) Mol. Cell. Biol. 10, 662-

671 Humphries, M. J., Akiyama, S. K., Komoriya, A., Olden, K., and

Yamada, K. M. (1986) J. Cell Biol. 103,2637-2647 Humphries, M. J., Komoriya, A., Akiyama, S. K., Olden, K., and

Yamada, K. M. (1987) J. Biol. Chem. 262,6886-6892 Humphries, M. J., Akiyama, S. K., Komoriya, A., Olden, K., and

Yamada, K. M. (1988) J. Cell Biol. 106, 1289-1297 Humphries, M. J., Mould, A. P., and Yamada, K. M. (1990) in

Receptors for Extracellular Matrix Proteins (Mecham, R. P., and McDonald, J . A., eds) Academic Press, New York, in press

Hynes, R. 0. (1987) Cell 48,549-554 Hynes, R. 0. (1990) Fibronectins, Springer-Verlag, New York Kornblihtt, A. R., Vibe-Pedersen, K., and Baralle, F. E. (1984) Nucleic

Kornblihtt, A. R., Umezawa, K., Vibe-Pedersen, K., and Baralle, F.

Kozlowski, J . M., Hart, I. R., Fidler, I. J., and Hanna, N. (1984) J.

Kretsinger, R. H. (1980) CRC Crit. Rev. Biochem. 8, 119-174 Laemmli, U. K. (1970) Nature 227,680-685 Lam, S. C.-T., Plow, E. F., Smith, M. A., Andrieux, A., Ryckwaert,

J.-J., Marguerie, G., and Ginsberg, M. H. (1987) J. Biol. Chem.

Lebien, T. W., BouC, D. R., Bradley, J. G., and Kersey, J . H. (1982) J. Zmmunol. 129,2287-2292

Liao, N.-S., St. John, J., McCarthy, J . B., Furcht, L. T., and Cheung, H. T. (1989) Exp. Cell Res. 181 , 348-361

Loftus, J. C., O’Toole, T. E., Plow, E. F., Glass, A., Frelinger, A. L., and Ginsberg, M. H. (1990) Science 249,915-918

McCarthy, J. B., Skubitz, A. P. N., Zhao, Q,, Yi, X.-Y., Mickelson, D. J., Klein, D. J., and Furcht, L. T. (1990) J. Cell Biol. 110 , 777- 787

Biophys. Acta 1031,91-110

istry 24,2698-2704

Plow, E. F. (1988) Science 242, 91-93

J. Biol. Chem. 265,3440-3446

M., and Thiery, J. P. (1988) EMBO J. 7, 2661-2671

507-513

Hemler, M. E., and Lobb, R. R. (1990) Cell 6 0 , 577-584

171,351-356

J. 241,923-928

(1990) J. Zmmunol. 144, 3361-3366

Haemostasis 59 , l -12

tle, R. F. (1982) Proc. Natl. Acad. Sci. U. S. A. 79 , 2068-2071

Acids Res. 12,5853-5868

E. (1985) EMBO J. 4, 1755-1759

Natl. Cancer Znst. 72,913-917

262,947-950

Mosher, D. F. (ed) (1988) Fibronectin, Academic Press, New York Mould, A. P., Wheldon, L. A., Komoriya, A., Wayner, E. A., Yamada,

K. M., and Humphries, M. J. (1990) J. Biol. Chem. 265 , 4020- 4024

Nagata, K., Humnhries, M. J., Olden, K., and Yamada, K. M. (1985) Cell Biol. 101,386-394

Obara. M.. Kane. M. S.. and Yamada. K. M. (1988) Cell 53,649-657 Osborn, L:, Hesiion, C.; Tizard, R., Vassallo,’C., Luhowsky, S., Chi-

Oyama, F., Murata, Y., Suganuma, N., Kimura, T., Titani, K., and Rosso, G., and Lobb, R. (1989) Cell 5 9 , 1203-1211

Sekiguchi, K. (1989) Biochemistry 2 8 , 1428-1434

Integrin cY&-Fibronectin IIICS Interactions k. 3585

Paul, J. I., Schwarzbauer, J. E., Tamkun, J. W., and Hynes, R. 0.

Pierschbacher, M. D., and Ruoslahti, E. (1984a) Nature 309,30-33 Pierschbacher, M. D., and Ruoslahti, E. (198413) Proc. Nutl. Acad.

Pierschbacher, M. D., and Ruoslahti, E. (1987) J. Biol. Chem. 262,

Ruggeri, Z. M., Houghton, R. A., Russell, S. R., and Zimmerman, T. S. (1986) Proc. Nutl. Acud. Sci. U. S. A. 83, 5708-5712

Ruoslahti, E., and Pierschbacher, M. D. (1987) Science 238,491-497 Santoro, S. A., and Lawing, W. J. (1987) Cell 48,867-873 Schwarzbauer, J. E., Tamkun, J. W., Lemischka, I. R., and Hynes,

Schwarzbauer, J. E., Paul, J. I., and Hynes, R. 0. (1985) Proc. Natl.

(1986) J. Biol. Chem. 261, 12258-12265

Sci. CJ. S. A. 8 1 , 5985-5988

17294-17298

R. 0. (1983) Cell 35,421-431

Acad. Sci. U. S. A. 82, 1424-1428

Sekiguchi, K., Klos, A. M., Kuraohi, K., Yoshitake, S., and Hakomori,

Skorstengaard, K., Jensen, M. S., Sahl, P., Petersen, T. E., and

Smith, J. W., and Cheresh, D. A..-(1988) J. Biol. Chem. 263, 18726-

Smith, J. W., and Cheresh, D. A. (1990) J. Biol. Chem. 265, 2168-

Wayner, E. A., Garcia-Pardo, A,, Humphries, M. J., McDonald, J . A.,

Yamada, K. M., and Kennedy, D. W. (1984) J. Cell Biol. 9 9 , 29-36 Yamada, K. M., and Kennedy, D. W. (1985) J. Cell. Biochem. 2 8 ,

Zardi, L., Carnemolla, B., Balza, E., Borsi, L., Castellani, P., Rocco,

S. (1986) Biochemistry 25,4936-4941

Magnusson, S. (1986) Eur. J . Biochem. 1 6 1 , 441-453

18731

2172

and Carter, W. G. (1989) J. Cell Biol. 109, 1321-1330

99-104

M., and Siri, A. (1985) Eur. J. Biochem. 1 4 6 , 571-579