PDFlib PLOP: PDF Linearization, Optimization, Protection Page inserted by evaluation version www.pdflib.com – [email protected]
The Molecular Biology of Heparan Sulfate Fibroblast Growth Factor Receptors" MICHAEL C. KIEFER, MASAYUKI ISHIHARA) STUART J. SWIEDLER? KEVIN CRAWFORD, JAMES C. STEPHANS, AND PHILIP J. BARRC Chiron Corporation 4560 Horton Street Emetyville, California 94608 hGlycomed Inc. 860 Atlantic Avenue Alameda, California 94501 The ligand-affinity or "panning" method of expression cloning has been used extensively for the isolation of cDNAs that encode mammalian cell surface proteins. Since the initial development of the technique,',' numerous cDNAs have been isolated based on the interactions of their encoded products with previously characterized monoclonal antibodies. Such cloned DNAs have included cDNAs that encode the cell surface antigens CD28,' CD2,' ELAM-1,' and C D W ~ ~ . ~ In an attempt to develop a general method for the isolation of growth factor receptor cDNAs using this technique, we targeted the molecular cloning of receptors for the fibroblast growth factors (FGFs). Members of this family of polypeptide growth factors are thought to have significant potential in several medically relevant areas, including soft tissue repair,' angiogenesis,' and neurogenesis' and for certain opthalmic indicationsnMore recently still, the FGF-FGF receptor system has been implicated in the entry of the herpes simplex virus (HSV) into susceptible cells.y,'" Currently, the FGF family includes the prototypic acidic and basic FGFs,"-" the product of the int-2 proto~ncogene,'~.~~ a factor isolated from sarcoma tissue DNA (hst or KS-FGF),'"'' FGF-5," FGF-6," and the keratinocyte growth factor.*" The acidic and basic FGFs are mitogenic for a number of cell types including those of mesenchymal, epithelial, or neural origin.".*' Consequently, studies directed toward in vivo wound healing and nerve regeneration have been The multiplicity of characterized FGFs and their diverse spectrum of activities have suggested the possibility that several receptors might exist for this growth factor family. Indeed, for the acidic and basic FGFs themselves, two classes of receptor have been described.'&*' These receptor classes are distinguished by their differing affinities for FGF. For example, binding of basic FGF to a high-affinity site on baby hamster kidney (BHK) cells occurs with a dissociation constant in the range of 20 pM, whereas basic FGF binds to the low-affinity site with a dissociation constant of around 2 nM, and is released with 2 M sodium ~hloride.'',~ Much progress in the molecular characterization of the high-affinity FGF receptors has been reported. Since the initial discovery that these membrane-bound 'This work was supported by Chiron Corporation and Glycomed Inc. 'Author to whom correspondence should be addressed. 167
ANNALS NEW YORK ACADEMY OF SCIENCES
IUEFER et al.: HSPG-FGF RECEPTORS
tyrosine kinases were encoded by the F L G gene (Fns-like gene), the molecular The complexity of this group biology of this receptor family has been well of proteins is illustrated by the multiplicity of cDNAs that have been isolated using FLG Additional FLG-related cDNAs have also been shown to encode receptors that can bind FGF.3*.36 In contrast, however, the molecular biology of the low-affinity receptors for the FGFs has not been well defined. Here, we describe the initial isolation of a cDNA encoding a cell surface heparan sulfate proteoglycan that binds basic fibroblast growth factor.” This cDNA was isolated from a hamster kidney cell line that had been shown previously to be a rich source of both low- and high-affinity F G F receptors. We also describe the characterization of a cDNA that encodes a human homologue of the hamster proteoglycan, isolated from a human liver cell line.
MOLECULAR CLONING OF A HAMSTER HEPARAN SULFATE PROTEOGLYCAN (HSPG) FGF RECEPTOR
Construction of a cDNA expression library from the BHK-21 cell line in the expression vector EBO-pCD-XN has been described previou~ly.’~ The cDNA library was stably introduced into the human lyrnphoblastoid cell line WI-L2-729 H F 2 (ATCC CRL 8062) by electroporation and selection on hygromycin. The WI-L2-729 HF2cell line does not bind significantly to basic-FGF-coated culture dishes (FIGURE I). Repeated panning (three times) of the stably transfected WI-L2-729 HF2cells on basic-FGF-coated culture dishes allowed for the enrichment of cells that contained two distinct families of recombinant plasmids with similar insert restriction patterns. WI-L2-729 HFLcells stably transfected with purified plasmids from each family were found to bind efficiently to basic-FGF-coated cultures dishes (FIGURE1) whereas no binding was observed in areas of the dish not coated with basic F G F (data not shown). The binding was markedly reduced in the presence of free basic FGF, heparin, and heparan sulfate, whereas epidermal growth factor, platelet-derived growth factor, acidic FGF, chondroitin sulfate, dermatan sulfate, and keratan sulfate had little effect. Treatment of the transfected cells with heparinase abolished 98% of the binding (FIGURElB).‘7 Inserts from each subgroup (RO-12 and RO-5) were excised and sequenced. The two families were shown to encode identical proteins and differed only in the site of attachment of poly(A) sequences within their 3’-untranslated regions. RO-12 and RO-5 cDNA sequences differed only in that RO-5 was polyadenylated 14 residues beyond the internal polyadenylation signal. The cncoded protein sequence is shown
FIGURE 1. A Binding of HSPG-FGF-transformed WI-L2-729 HF, cells to basic FGF. Binding assays were performed essentially as described for ~ a n n i n g . Cells ’ ~ were photographed ( 1 4 ) or counted after the first panningcycle. Photographs are of BHK-21 cells ( l ) , WI-L2-729 HF2cells (2). RO-5-transformed WI-L2-729 HF, cells (3), RO-12-transformed WI-L2-729 HS2 cells (4). B: Binding of control (bar C) and RO-12-transformed WI-L2-729 HF, cells with no additions (bar RO) and in the presence of keratan sulfate (bar KS), chondroitin sulfate (bar CS), dermatan sulfate (bar DS), heparan sulfate (bar HS), heparin (bar H), basic FGF (bar bFGF), or epidermal growth factor (bar EGF) at the indicated concentration (kg/ml). Also shown are transformed cells pretreated with heparinase (H’ase; bar +) or heparinase buffer (bar -), respectively, as de~cribed.’~ The values shown represent the mean of three independent experiments. (Reprinted from Reference 37 with permission.)
ANNALS NEW YORK ACADEMY OF SCIENCES
(FIGURE2A), and also a schematic representation of the structure of the hamster HSPG-FGF receptor (FIGURE 3A). MOLECULAR CLONING OF A HUMAN HOMOLOGUE OF THE HAMSTER HSPG-FGF RECEPTOR
The hamster HSPG-FGF receptor cDNA (RO-5) was used as a probe to isolate a human cDNA from a cDNA library that was constructed from RNA isolated from the human hepatoma cell line Hep G2. Multiple positive signals from the plated cDNA library and northern blot analysis indicated a moderate abundance of human HSPG-FGR receptor mRNA. Restriction enzyme analysis of three isolated cDNAs showed similar patterns for each clone. This suggested that each individual clone encoded the same human HSPG-FGF receptor. Two hybridizing mRNA species of approximately 3.2 kilobases (kb) and 2.4 kb were seen on the northern blot. The same size rnRNAs were identified previously in BHK-21 cells.” One clone (RO-4) was analyzed by DNA sequencing, and the structure of the encoded protein is shown (FIGURE2B and 3B). We have noted previously” the homology of our hamster HSPG-FGF receptor clone to a cDNA encoding murine ~ y n d e c a n ? an ~ . ~integral ~
60 60 60 59 120 120 120 119 180 178 178 179 236 237 237 238 296 297 297 298 (A) (8) (C)
FIGURE 2. Comparison of the hamster HSPG-FGF receptor amino acid sequence (A) with the homologous sequences encoded by our liver-derived human cDNA clone (B), human syndecan (C), and murine syndecan (D). Regions of amino acid sequence that are identical in all four core proteins are shaded. The predicted transmembrane domains are overlined, and a predicted signal peptidase cleavage site is arrowed. Since the amino terminus of each mature protein has not been defined experimentally, numbering is based on the sequence of each proposed precursor protein.
KIEFER et 01.: HSPG-FGF RECEPTORS
FIGURE 3. Schematic representations of the structures of the hamster HSPG-FGF receptor,” the human-liver-derived homologue (present study) (B), murine ~ y n d e c a n ~(C), ~ . ’ ~and the human lung fibroblast cell surface HSPG, 48K5, defined by a partial cDNA clone‘2 (D). Proposed signal peptide sequences are noted (Sig), as are transmembrane domains (TM). Also highlighted, are the particularly conserved cytoplasmic domains (shaded) that have been and implicated in the binding of actin.’8,w Asparagine-linked glycosylation sites are marked (e), consensus GAG attachment sites” is shown as vertical bars. The fourth serine-glycine pair of the hamster core protein (A) is not surrounded by acidic residues, and consequently may not be utilized as a GAG attachment site.
membrane proteoglycan, supporting the proposal that syndecan is the murine equivalent of the HSPG-FGF re~eptor.’~ In further support of this notion, our encoded human core protein is almost identical to the human syndecan described by Mali et aLa The only amino acid difference between these two proteins occurs in the proposed signal peptide sequence (Pro19 to Leu) (FIGURE2B and C), and may represent either a cloning artefact or a polymorphism. STRUCTURE COMPARISON OF THE HSPG-FGF RECEPTOR AND HOMOLOGUES An alignment of the amino acid sequences of the hamster HSPG-FGF receptor, the human liver-derived homologue, human syndecan and murine syndecan (FIGURE 2) reveals several highly conserved regions. The cytoplasmic domain shows 100% amino acid sequence identity in all of the molecules. This high level of conservation suggests an important role for the cytoplasmic domain and is probably related to its actin binding f~nction.~’ The transmembrane (TM) regions of the hamster and human proteins display 100% sequence identity whereas the murine TM differs by only one amino acid (Gly to Ala) resulting in a 96% sequence similarity. The signal peptides also display a high degree of homology ( 291%). As was noted above, the single amino acid difference (Pro to Leu) between the human HSPG-FGF and human syndecan was found in this
ANNALS NEW YORK ACADEMY OF SCIENCES
region. Thus, the mature core proteins are identical. The extracellular domains of the core proteins are less well conserved with a range of sequence identity between 68% and 81%. Most of the sequence divergence in this domain is located centrally with the flanking regions that contain the GAG attachment sites being more conserved (see FIGURE3). The overall sequence identity between the hamster and mouse, the hamster and human, and the mouse and human proteins is 85%, 76%, and 77%, respectively. The above-described domains of the core proteins-which include a signal peptide, an extracellular domain that contains consensus GAG attachment sites,,' a transmembrane region, and a cytoplasmic domain homologous to previously reported actin binding domains-have been r e p ~ r t e d . Schematic ~~" representations of the core proteins and these various domains are shown in FIGURE 3. Also shown (FIGURE 3D) is a diagram of the known sequence of 48K5.,*This structure, derived from the sequence of a partial cDNA clone, corresponds to the human lung fibroblast HSPG of M, 48,000.42 Mali et al. have demonstrated a 56% identity between 48K5 and human syndecan in the transmembrane and cytoplasmic domaimm Also, Kiefer et al. noted 68% identity in the cytoplasmic domains of 48K5 and hamster HSPG-FGF receptor." Particularly evident in these domains is a striking conservation of tyrosine residues:'
BIOCHEMICAL CHARACTERIZATION OF EXPRESSED HSPG-FGF RECEPTORS
In order to demonstrate that the expressed proteins serve as cell surface proteoglycan core proteins, we immunopre~ipitated~~SO,-Iabeled proteins from trypsinized and nontrypsinized transfected cells with an antibody directed against a carboxyl-terminal peptide (FIGURE 4). Using this method, WI-L2-729 HF, cells that were transfected with the hamster cDNA were shown to express the HSPG-FGF receptor as judged by the significant increase in "SO, incorporation above control levels. Furthermore, the trypsin sensitivity of the incorporated counts indicated the extracellular nature of the sulfate incorporation. To further characterize the glycosaminoglycan (GAG) side chains of the immunoprecipitated molecule, we treated samples with either nitrous acid or chondroitinase AC. The release of 77% of the incorporated counts with nitrous acid indicated the predominance of heparan sulfate incorporation into the expressed protein. Similarly, chontroitinase AC treatment showed that 22% of the incorporated 3sS0, was incorporated as chondroitin side chains. This is consistent with previous observations, where the incorporation of both heparan and chondroitin sulfates into murine syndecan was demonstrated.'8 In similar experiments with the human-liver-derived cDNA, we measured incorporation of '5S0, into immunoprecipitable expressed protein. Surprisingly, incorporation levels were low when compared with the corresponding hamster protein (FIGURE 5) and in addition, trypsin treatment of these transfected cells showed that the majority of incorporated "SO, remained intracellular (data not shown). Consistent with this finding, we also found that in contrast to the hamster HSPG-FGF receptor transfected cells, these human-cDNA-transfected cells were not capable of binding to basic-FGF-coated plates. In other studies, we showed that trypsinsensitive 3sS0,incorporation into heparan sulfate of cells that were transfected with the hamster cDNA was fivefold over control levels for cells transfected with a control plasmid.
KIEFER el af.: HSPG-FGF RECEPTORS
Trypsin FIGURE 4. Characterization of GAG incorporation into the hamster HSPG-FGF receptor expressed in WI-L2-729 HF, cells. A: Cells that were stably transfected with the expression vector EBO-pCD-XN containing hamster HSPG-FGF receptor cDNA (R012) were labeled with 50 FCi 'rSO,/ml for 18 hours, washed, incubated with (+) and without (-) trypsin, and solubilized. HSPG-FGF receptor was isolated by imrnunoprecipitation with rabbit polyclonal antiserum directed against a synthetic C-terminal peptide of hamster HSPG-FGF receptor (amino acids 287-305) and protein A Sepharose. B: Aliquots from the untreated sample in A (trypsin -)were treated with either buffer, 0.25 M nitrous acid pH 1.5 or chondroitinase AC. Heparan sulfate (HS) and chondroitin sulfate (CS) released from the HSPG-FGF receptor by nitrous acid and chondroitinase AC treatment, respectively, were separated from the proteoglycan by paper chromatography and the counts per minute determined. The percentage of the total cpm found in either HS or CS is shown.
CONCLUSIONS Our results show that the hamster cDNA that was isolated by panning on FGF-coated plates most likely encodes the low-affinity FGF receptor of BHK-21 cells. Furthermore, the homology of our hamster clone to the previously described murine syndecan indicates that, as proposed by Bernfield and S a n d e r s ~ nsyndecan ,~~ is most likely the murine low-affinity FGF receptor. Surprisingly, however, the characteristics of the expressed human syndecan, did not parallel those of the expressed hamster HSPG-FGF receptor. The inability of the human molecule to
ANNALS NEW YORK ACADEMY OF SCIENCES
induce binding of the transfected cells to basic FGF-coated plates is most likely a reflection of the low expression levels of human syndecan coupled with low levels of incorporation of GAG side chains into extracellularly located polypeptide. It remains to be established whether or not the expression differences using this cell line result from a species difference or if other, related molecules are responsible for low-affinity FGF binding in humans in viva It is also interesting to note that a three-immunoglobulin-domain form of the human high-affinity FGF receptor” was also unable to bind basic FGF-coated plastic dishes when expressed using this system. Again, this observation might be a reflec-
FIGURE 5. Comparison of GAG incorporation into human syndecan and hamster HSPG-FGF receptor expressed in WI-L2-729 HF, cells. Labeling of cells and immunoprecipitations were performed as in FIGURE 4. Cells were stably transfected with the expression vector EBOpcD-XN containing no cDNA (control), human syndecan cDNA, isolated from human liver cell line Hep G2 (human), and hamster HSPG-FGF receptor cDNA (hamster). The human syndecan cDNA (clone R04) was modified by PCR amplification to contain Xbal and Not1 restriction enzyme sites which flanked the coding region and facilitated subcloning of the PCR product into EBO-pcD-XN.
tion of low levels of expression and/or transport to the cell surface of this FLG protein. Alternatively, it is possible that multivalent interactions with bound FGF, as one might expect from the expressed hamster HSPG-FGF receptor, are required for the adherence of the appropriate transfected cells during panning. It is unlikely that this type of multivalent interaction would occur with the high-affinity FGF receptor. Either of these scenarios can explain the lack of detection of FLG or FLG-related cDNA clones during our initial planning studies,)’ and serve to demonstrate that this procedure might not represent a general cloning strategy for high-affinity growth factor receptors. In further support of this notion, Harada et al. were able to clone
KlEFER el al.: HSPG-FGF RECEPTORS
the receptor for the lymphokine interleukin 4 (IL-4) by a panning methodology.4' In this case, however, it was found to be necessary to chemically cross-link biotinylated IL-4 to the expressed receptor prior to panning on immobilized antibiotin antibodies in order to select for the appropriate transfected cells.
SUMMARY Two distinct classes of cell surface FGF-binding proteins have been identified. These receptors differ in both mode of interaction and in affinity for the FGFs. cDNAs that encode the low-affinity receptor were isolated from a hamster kidney cell line cDNA library by expression cloning. Transfected cells that contained these heparan sulfate proteoglycan F G F receptor cDNAs were enriched for by panning on basic FGF-coated plates. The analogous human cDNA was isolated from a hepatoma cell line cDNA library. The homology of our hamster cDNAs to the previously described murine integral membrane proteoglycan syndecan, together with an exact amino acid sequence match of our human-cDNA-encoded product to human syndccan, clearly indicates the identity of these independently isolated protcoglycans. Further confirmation that the expressed molecule serves as a proteoglycan core protcin was achieved by immunoprecipitation of "SO,-labeled material from solubilizcd transfected cells. Nitrous acid treatment and chondroitinase digestion revealed that 77% of the label was associated with heparan sulfate chains and 22% with chondroitin sulfate chains. These heparan sulfate chains contributed to the fivefold increasc in the total heparan sulfate found to be present on the surface of the transfected cells compared with cells transfected with a vector lacking the cDNA insert.
ACKNOWLEDGMENTS We thank A. Saphire for technical assistance and P. Anderson for preparation of the manuscript. REFERENCES I. 2. 3. 4. 5. 6.
7. 8. 9.
AKUFFO. A. & B. SEED.1987. Proc. Nat. Acad. Sci. USA 84: 8573-8577. Si:.tr), B. & A. ARUFFO.1987. Proc. Nat. Acad. Sci. USA 84: 3365-3369. B~VII.ACOUA, M. P., S. STENGEIJN, M. A. GIMBKONE, JK. & B. SEED. 1989. Science 243: 1160~1165. SIAMENKOVIC, 1.. M. AMIOT, J . M. PESANUO & B. SEED. 1Y89. Cell 5 6 1057-1062. BUNIKOCK, P.. K. D. JENTZSCH & G. HEIIEK.1982. Exp. Pathol. 21: 62-67. FOLKMAN. J. & M. KLAGSBURN. 1987. Science 235: 442-447. SEIVLKS, J., B. HAUSMANN, K. UNSICKEK & M. BERRY.1987. Neurosci. Lett 76: 157-162. FKkDJ-RtY(jK(>BtL!?r, D., J. PI-OUET,T. DELAYRE,C. BAUDOUIN, F. B ~ U K R E&T D. LAPAI.IJS.1987. Curr. Eye Res. 6 1205-1209. KANEK, R. J., A. BAIRD,A. MANSUKHANI, C. BASIl.IC0, B. D. SUMMERS, R. A. FIDKKIEWICL & D. P. HNJAK.1990. Science 2 4 8 1410-1413. BAIKI),A., R. Z. FLORKIEWICZ, P. A . MAHEK,R. J. KANEK& D. P. HAJJAK.1990. Nature 348: 344-346. JAYt, M., R. HOWK,W. BUKGESS, G. A. RICCA,I.-M. CHUI, M. W. RAVEKA. S. J. 0 ' B R l t . N . W. S. MODI,T. MACIAG & W. N. DKOHAN. 1986. Science 233: 541-545. ABKAHAM, J. A., A. MEKGIA, J. L. WHANG, A. TUMOLO,J. FREDMAN, K. A . HJERRIII, D. GOSPODAROWICZ & J. C. FIDDES.1986. Science 233: 545-548.
ANNALS NEW YORK ACADEMY OF SCIENCES
13. ABRAHAM, J. A,, J. L. WHANG, A. TUMOLO, A. MERGIA, J. FRIEDMAN, D. GOSPODAROWICZ & J. C. FIDDES.1986. EMBO J. 5 2523-2528. M. DIXON,G. PETERS& C. DICKSON. 1986. EMBO J. 14. MOORE,R., G. CASEY,S. BROOKES, 5: 919-924. & G. R. MARTIN.1986. Proc. Nat. 15. JAKOBOVITS, A,, G. M. SHACKLEFORD, H. E. VARMUS Acad. Sci. USA 8 3 7806-7810. F. KERN,A. GRECO,M. ITMA” & C. BASILICO. 1987. 16. DELLIBow, P., A. M. CURATOLA, Cell 5 0 729-737. 17. TAIRA,M., T. YOSHIDA, K. MIYAGAWA, H. SAKAMOTO, M. TERADA & T. SUGIMURA. 1987. Proc. Nat. Acad. Sci. USA 84: 2980-2984. 1988. Mol. Cell Biol. 8: 3487-3495. 18. ZHAN,X.,B. BATES,X.Hu & M. GOLDFARB. 19. MARICS,I., J. ADELAIDE, F. RAYBAUD, M.-G. MATITI, F. COULIER, J. PANCHE,0. DE LAPEYRIERE & D. BIRNBAUM. 1989. Oncogene 4 335-340. 20. FINCH,P. W., J. S. RUBIN,T. MIKI,D. RON& S. A. AARONSON. 1989. Science 2 4 5 752755. K. 1987. FASEB J. 1: 434-440. 21. THOMAS, 22. GOSPODAROWICZ, D. 1987. Methods Enzymol. 147: 106-1 19. 23. CUEVAS, P., J. BURGOS& A. BAIRD.1988. Biochem. Biophys. Res. Commun. 1 5 6 611617. 1985.J. Biol. Chem. 2 6 0 13860-13868. 24. NEUFELD, G. & D. GOSPODAROWICZ. H. Hou, H. HOSHI,P.-E. MANSSON & W. L. MCKEEHAN. 1988. J. 25. KAN,M. D., J. DISORBO, Biol. Chem. 263: 11306-11313. 26. WALICKE, P. A., J.-J. FEIGE& A. BAIRD.1989. J. Biol. Chem. 264: 412W126. 27. MOSCATELLI, D. 1987. J. Cell. Physiol. 131: 123-130. D. B. & D. MOSCATELLI. 1989. J. Cell. Biol. 109 1-6. 28. RIFKIN, 29. RUTA,M., R. HOWK,G. RICCA,W. DROHAN,M. ZABELSHANSKY, G. LAUREYS, D. E. BARTON, U. FRANCKE, J. SCHLESSINGER & D. GIVOL.1988. Oncogene 3: 9-15. L. S. COUSENS, V. A. FRIED& L. T. WILLIAMS. 1989. Science 30. LEE,P. L., D. E. JOHNSON, 245: 57-60. 1989. Proc. Nat. Acad. Sci. USA 8 6 5449-5453. 31. PASOUALE, E. & S . J. SINGER. 32. REID,H. H., A. F. WILKS& 0. BERNARD. 1990. Proc. Nat. Acad. Sci. USA 87: 1596-1600. & P. SARMIENTOS. 1990. Nucleic Acids Res. 18: 1906. 33. ISACCHI, A., L. BERGONZONI 1990. Mol. Cell. Biol. 10 4728-4736. 34. JOHNSON, D. E., P. L. LEE,J. Lu & L. T. WILLIAMS. C. GEORGE-NASCIMENTO, 0. B. MASON,L. J. 35. KIEFER, M. C., A. BAIRD,T. NGUYEN, & P. J. BARR.Growth factors. (In press.) BOLEY,P. VALENZUELA S., K. E. PAULSON & H. HANAFUSA. 1988. Mol. Cell. Biol. 8: 5541-5544. 36. KORNBLUTH, K. CRAWFORD, K. OKINO& P. J. BARR.1990. Proc. Nat. 37. KIEFER,M. C., J. C. STEPHANS, Acad. Sci. USA 87: 6985-6989. S., J. JALKANEN, S. O’FARRELL & M. BERNFIELD. 1989.J. Cell Biol. 1 0 8 154738. SAUNDERS, 1556. 1990. Phil. Trans. R. SOC.London Ser. B 327: 171M. & R. D. SANDERSON. 39. BERNFIELD. 186. & M. JALKANEN. 1990. J. Biol. Chem. 40. MALI, M., P. JAAKKOLA, A.-M. ARVILOMMI 265: 6884-6889. 41. BOURDON, M. A., T. KRUSIUS,S. CAMPBELL, N. B. SCHWARTZ & E. RUOSLAHTI. 1987. Proc. Nat. Acad. Sci. USA 84: 3194-3198. P.,J. ZHANG,J.-J. CASSIMAN, H. VANDEN BERGHE & G. DAVID.1989. J. Biol. 42. MARYNEN, Chem. 264: 7017-7024. 43. HARADA, N., B. E. CASTLE,D. M. GORMAN, N. ITOH,J. SCHREURS, R. L. BARRETT,M. & A. MIYAJIMA. 1990. Proc. Nat. Acad. Sci. USA 87: 857-861. HOWARD