Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Published December 1, 1982 Binding of Soluble Type I Collagen to Fibroblasts: Specificities for Native Collagen Types, Triple Helical Structure, Telopeptides, Propeptides, and Cyanogen Bromide-derived Peptides BURTON D. GOLDBERG and ROBERT E. BURGESON Department of Pathology, New York University Medical Center, New York, New York 1001@and Department of Pediatrics, Los Angeles County Harbor-- University of California Los Angeles Medical Center, Torrance, California 90509 The binding of type I collagen to fibroblasts is considered to be specific because the radiolabeled ligand is readily displaced by unlabeled type I collagen but not by noncollagenous proteins, nor by every molecule containing some collagen structure (1). In the present study, binding inhibition measurements have further defined ligand specificities with respect to collagen types, native triple helical assembly, and regions of procollagen and collagen chains that contain binding determinants. MATERIALS AND METHODS Binding I n h i b i t i o n Assays Swiss mouse 3T3 fibroblast cultures were used in standard binding assays as described (1, 2). Unlabeled lathyritic rat skin (LRSC) and other unlabeled collagenous proteins were dissolved in 0.05 M Tris HCI-0.15 M NaCI pH 7.5 buffer at 1-2 mg/ml. Protein concentrations of the stock solutions were computed from the difference in absorbances at 215 and 225 nm of appropriate dilutions of the stock (3). For each experiment, increasing amounts (0-100 #g) of unlabeled LRSC and of putative binding proteins were each premixed with a fixed amount of ~zr'I-LRSC (~ 130,000 cpm) in 2 ml of binding buffer (Dulbecco's modification of Eagle's medium with half bicarbonate concentration, 0.05 M HEPES, pH 7.2 and 2.5 mg/ml bovine serum albumin). The solutions were warmed at 37°C for 30 min, cooled to room temperature (26°C) and added to washed, replicate confluent cultures of 3T3 fibroblasts. After a 2-h incubation at 26°C, the cell layers were rapidly washed and lysed, and the radioactivity specifically bound to the ceils was measured. Percentage inhibition of binding at each protein concentration was computed by: 1 c cpm bound with unlabeled protein ~ × ~ ~ protein,/ 102 Limited Digestion o f LRSC with Pepsin and a-Chymotrypsin Pepsin was purchased from Sigma Chemical Co., St. Louis, MO and achymotrypsin from Worthington Biochemical Corp., Freehold, NJ. Pepsin digestion was done according to the procedures of Leibovich and Weiss (4) and Helseth and Veis (5). Digestion with a-chymotrypsin was done by the procedure of Bornstein et al. (6). Control samples without enzyme additions were carried through all the procedures. Samples were finally dissolved in Tris buffer and tested as inhibitors of binding as described above. SDS PAGE Samples were dissolved in sample buffer containing 2% SDS, 10% glycerol. 0.001% bromphenol blue in 0.0625 M Tris HCI, pH 6.8. After heating at 100°C TttE JOURNAL OI C E i l BrOLOGY • VOLUME 95 DECEMBER 1982 7 5 2 - 7 5 6 752 © The Rockefeller University Press • (X~21-9525/82/12/0752/05 $100 Downloaded from on June 18, 2017 ABSTRACT Unlabeled collagenous proteins were quantified as inhibitors of binding of native, soluble, radioiodinated type I collagen to the fibroblast surface. Collagen types IV, V, a minor cartilage isotype (1~2c~3c~), and the collagenlike tail of acetylcholinesterase did not inhibit binding. Collagen types II and III behaved as competitive inhibitors of type I binding. Denaturation of native collagenous molecules exposed cryptic inhibitory determinants in the separated constituent ~ chains. Inhibition of binding by unlabeled type I collagen was not changed by enzymatic removal of the telopeptides. Inhibitory determinants were detected in cyanogen bromide-derived peptides from various regions of helical ~1(I) and al(lll) chains. The aminoterminal propeptide of chick procd(I) was inhibitory for binding, whereas the carboxyterminal three-chain propeptide fragment of human type I procollagen was not. The data are discussed in terms of the proposal that binding to surface receptors initiates the assembly of periodic collagen fibrils in vivo. Published December 1, 1982 for 2 min, the samples were applied to a 3% stacking/6%separating slab gel (7) and, after development,the gels were stained with Coomassie Blue. Materials Collagenousproteinswere generouslydonated by the followinginvestigators: Dr. Terrone L. Rosenberry,Case Western Reserve University,Cleveland,OH, the collagenliketail of acetylcholinesterase;Dr. Jerome Seyer, Veterans Administration MedicalCenter, Memphis,TN, cyanogenbromidepeptidesof collagen; Dr. George Martin,National Instituteof Dental Research. Bethesda, MD, LRSC and rodent collagen types 11 and IV; Dr. Bjrrn R. Olsen, Department of Biochemistry,Collegeof Medicineand Dentistryof New Jersey-RutgersMedical School, Piscataway,NJ, the aminoterminalpropeptide of chick pro a l(I). The syntheticpeptidesDNP-Pro-GIn-GIy-Ile-AIa-GIy-GIn-D-Arg and DNPPro-Leu-Gly-lle-Ala-Gly-Arg-NH2were purchased from the Peptide Institute, Protein Research Foundation, Osaka, Japan. RESULTS Specificities: Collagen Structure, Collagen Types, and Native Helicity 100 ~ g6o if_ 4o z ~ 2O =.c_o.~-----,~ , 5 10 15 /ig PROTEIN q 20 1(30 80 g 6c m_ 4( g o 2C 0--0 5 10 15 ~g PROTEIN 20 F~GURE 1 Acetylcholinesterase as a competitor of 12SI-LRSC binding. Increasing amounts of unlabeled LRSC and the collagenlike tail of acetylcholinesterase were each mixed with 1.36 × 105 cpm of ~251-LRSC in binding buffer. Binding studies were performed in replicate cultures and the percentage inhibitions produced by the unlabeled proteins were computed. @, LRSC. O, acetylcholinesterase. FIGURE 2 Native collagen types as inhibitors of '251-LRSC binding. Composite of binding inhibition experiments performed with the following native collagen types: type I, LRSC; type II from a rat chondrosarcorna; type III, from human placenta; type IV, from the mouse EHS sarcoma; la2a3a cartilage isotype, from human cartilage, and type V, from human placenta. 2 0 ~ t , Q ~ FIGURE 3 Type II inhibition: scatchard plot. In15 creasing amounts (0-100 o " /~g) of unlabeled LRSC were premixed with a fixed amount of '2%/RSC (0.16 5[ ~ pmol, 118,000 cpm) in the L J , presence (O) and absence 0 0.5 to 1.5 2.0 2.5 3.0 (O) of 5 #g (17.5 pmol) of B (PICOMOLES) unlabeled type II collagen. The respective mixtures were presented to replicate plates of 3T3 fibroblasts and specific binding measured after 2.5 h at 26°C. Picomoles of type I collagen (radiolabeled and unlabeled) bound (B) and free (F) were calculated. 0 3O "& 2O I 0 I 040 2 I 60 L (PtCOMOLES) I 80 FIGURE 4 Type ll i n h i b i t i o n : Hanes-Woolf plot. Data from experiment of Fig. 3. (L) Picomoles of type I collagen ligand (labeled and unlabeled) added. (B) Picomoles of type I collagen ligand (labeled and unlabeled) bound. (O), 5/xg (17.5 pmol) of unlabeled type II collagen added as inhibitor. (z~), in absence of type II inhibitor. GOLDBERG AND 6URGESON Specificities for Collagen Binding 753 Downloaded from on June 18, 2017 Potential competitors for 12~I-LRSC binding sites were compared to the unlabeled LRSC in inhibition assays as described in Materials and Methods. That general collagen structure does not ensure binding to or blocking of type I sites is shown in Fig. 1. In this instance the putative competitor was the collagenlike tail of acetylcholinesterase (8, 9). Additions of up to 20 t~g of this unlabeled protein did not significantly inhibit binding of the radioactive ligand, whereas 20/xg of unlabeled LRSC produced 78% inhibition of binding. C l q , the complement component containing some collagen structure, also fails to inhibit type I binding (1). Native Collagen types II, III, IV, V, and a minor isotype present in cartilage but distinct from type II collagen (10) were tested in binding inhibition assays (Fig. 2). These data show that unlabeled type II and type I collagens cause equivalent inhibition whereas unlabeled type III collagen is a somewhat less efficient inhibitor of binding. The much lower levels of inhibition produced by the other collagen types are considered to be of doubtful significance. Graphical analyses of binding data were performed to determine whether collagen types II and III acted as competitive or noncompetitive inhibitors of type I binding. Fig. 3 presents Scatchard plots of type I binding with and without the addition of 5 #g (17.5 pmol) of type II collagen. Calculation of the slopes showed that inhibition was accompanied by an increase in the binding dissociation constant (KD) from 7.5 X 10-9 M to 1.7 x 10-8 M. The common abscissal intercept for the two lines indicates that the n u m b e r of cell binding sites was not changed by the inhibition. Fig. 4 is a Hanes-Woolf plot (11, 12) of the same data and it, like the Scatchard plot, is typical for a form of competitive inhibition. Qualitatively similar plots were obtained for inhibition by type III collagen. The collagen types tested in the experiments of Fig. 2 were native triple helical molecules. It is of interest to know whether native helicity critically determines binding specificity. If triple helicity is critical for binding, the separated (denatured) a chains should yield different binding inhibition data than the assembled parent molecules. In the experiments presented in Fig. 5, the three denatured a chains from the minor cartilage isotype and the denatured a 1 and a2 chains of type V collagen were individually tested in b i n d i n g inhibition assays. Whereas the parent triple helical molecules did not inhibit binding (see Fig. 2), each chain class of the cartilage isotype and the a2(V) chain gave equivalent or significant inhibition relative to the LRSC standard. Denatured reduced and alkylated type III chains were better inhibitors of type I binding than native III, and nondenatured reduced and alkylated type IV collagen did not inhibit binding (data not shown). We conclude that triple helicity can confer binding specificity and that denaturation of a chains can expose additional binding determinants. Published December 1, 1982 Are Telopeptides Required for Binding Potential Binding Determinants in Cyanogen Bromide-derived Helical Peptides Peptides generated by cyanogen bromide cleavage of helical I0(] I g~ oc 2a g6c aIci . . . . • a2(~) 4c 2C / oi- _~o.-.o"" I 5 I I [ i0 15 20 #g PROTEIN FIGURE 5 Denatured (x chains as competitors of '2SI-LRSC binding. Cartilage isotype chains l(x, 2a, and 3~x and a l and c~2 chains of type V collagen were resolved I~y carboxymethyl cellulose chromatography under denaturing conditions (13). The individual chain classes and native LRSC (I) were tested as inhibitors of '251-LRSC binding. ~ Bc 5 10 15 20 25 #g PROTEIN al(1) RAT NI I I I B 0.J6 754 THE JOURNAL OF CELL BIOLOGY,VOLUME 95, 1982 I 3 0.56 at(I} HUMAN NlZl i I FIGURE 6 Gel electrophoresis of LRSC and chymotrypsin-modified LRSC. Lane (A): LRSC digested with chymotrypsin at pH 7.5, 15°C, 24 h, at enzyme/substrate ratio of 1/ 10. Reaction stopped by cooling to 4°C, addition of protease inhibitors and acidification. Collagen separated by three cycles of precipitation with NaCI to 7.5%. Lane (B): Control LRSC. Same protocol, enzyme omitted. 5 p,g of collagen applied to both lanes. FIGURE 7 Binding inhibition by LRSC and chymotrypsin-modified LRSC. Samples of Fig. 6 used as inhibitors of 12%LRSC binding. Chymotrypsin-modified LRSC (O) and control LRSC (O), activated at 37°C, 30 min. Chymotrypsin-modified LRSC ([7) and control LRSC (IlL w i t h o u t thermal activation. 100 _g 6c ~ 4c ~_ 20 0° (]1 (I) CHICK a al(~) HUMAN N I 7 I c t.O 3 l i I 6 I I 6A I 6B c 0.50 0.05 3 I I 6 I I 8 0.039 0.27 0 I I 4 0 c 0.4@ O. f8 S O. f t t 0 O.OG? N I I4 I I Downloaded from on June 18, 2017 The telopeptides are short (10-25 residues) nonhelical extensions at the ends of the collagen a chains. Collagen molecules in solution can self-assemble to form native fibrils (14) and the initiation ("nucleation") of such self-assembly seems to depend upon a thermally induced conformational change in the amino terminal telopeptides (5). Collagen binding to the fibroblast also requires thermal activation of the ligand, and such binding has been proposed as an initiating event for fibrillogenesis (1, 2). Accordingly, experiments were performed to determine whether the telopeptides were necessary for binding of collagen to the fibroblast. The tetopeptides were first removed from unlabeled LRSC by limited digestion with a-chymotrypsin or pepsin. Fig. 6 shows the SDS-polyacrylamide gel patterns given by a-chymotrypsin-digested LRSC and the control preparation. The enzyme-digested collagen contains somewhat fewer covalently cross-linked molecules (y and fl forms) and has a major population of al and a2 chains with slightly greater mobilities than the control preparation. This result shows that the enzyme digested the telopeptides and did not alter the triple helix in most of the molecules. Binding inhibition assays were performed with enzyme-digested and control LRSC, with and without prior thermal activation (37°C, 30 min) of the samples. Fig. 7 shows that, with heat activation, molecules lacking telopeptides bound as well as intact collagen. In the absence of heat activation, neither preparation displayed significant binding. The latter result shows that the requirement for heat activation is independent of radioiodination of the ligand. Similar data were obtained when the telopeptides were removed by limited digestion with pepsin. We conclude that the telopeptides are not necessary for thermal activation and binding of type I collagen to fibroblasts. regions o f a l type I and III chains were evaluated as inhibitors of type I binding. Fig. 8 depicts the positions in the a chains of the peptides tested and quantifies inhibition by each peptide relative to the type I LRSC standard. The values given below the lines are the ratio of LRSC (#g)/peptide ~g) which produced 30% inhibition of t25I-LRSC binding in the same experiment. For all of the experiments, 1-2/~g of LRSC regularly produced 30% inhibition of binding. Inhibition ratios <0.1 are not considered as significant, and a ratio value of zero means that addition of 100 ~g of the peptide produced <30% inhibition of binding. By the test of inhibition, binding determinants were identified in more than one peptide in each species and type of a l chain, and these peptides were variably distributed along the chains from amino to carboxy termini. Peptides of similar size from the same regions of type I a chains of different species varied in binding capacities, for example, human a l(I)-CB8 as compared to the homologous peptides of rat and chick. The rodent a I(I)-CB7 peptide which includes residues 552822 of the al chain is notable for being equivalent on a concentration basis to LRSC as an inhibitor of binding. The same peptide has been reported to contain the determinants for attachment of fibronectin to ctl (I) but the following observations indicate that ~zSI-LRSC is not binding to fibronectin on the fibroblast surface (a) aI(I)-CB7 has been claimed to contain the sole binding site for fibronectin (15), but the data of Fig. 8 show that peptides positioned elsewhere in al(I) chains can significantly inhibit binding of ~25I-LRSC. (b) Critical determinants for binding of a I(I)-CB7 to fibronectin have been localized to the cleavage site for animal collagenase (residues 775-776) and a chymotrypsin-sensitive site (residues 779-780) (15). However, two synthetic peptides which have identical sequences to residues 773-780 and 775-778 of al(I), respectively, and which serve as substrates for animal collagenase (16), did not inhibit x2~I-LRSC binding (binding inhibition I ~ O. 10 I 0.67 9 C 0.54 FIGURE 8 Cyanogen bromide-derived helical peptides as inhibitors of binding. Linear schematic of helical a chains of coltagen types I and III indicating positions, standard numbering (above lines), and relative sizes of peptides produced by cyanogen bromide cleavages. Only the peptides tested in binding assaysare numbered. Numbers below lines are the ratio of p.g LRSC//Lg peptide that produced 30% inhibition of 1251-LRSC binding. N and C, amino and carboxytermini of the chains, respectively. Published December 1, 1982 ratios of zero, calculated as for experiments of Fig. 8). Moreover, human cd(III)-CB5, which also contains the coUagenase cleavage site, produced minimal inhibition of binding (see Fig. 8). Potential Binding Determinants in Propeptides Collagen is secreted as a precursor molecule (procollagen) that has large polypeptides (propeptides) covalently linked to the terminal telopeptides of the three helical ot chains of collagen. Type I procollagen could bind to the cell surface through the demonstrated determinants in its central triple helical region but it is possible that propeptide residues could also contribute to procollagen binding. A reasonable candidate for a propeptide binding region is the short, collagenlike segment ("Col 3") in the aminoterminal propeptides of type I procollagen (see reference 17 for review). When the aminopropeptide isolated from the chick pro al(I) chain (18) was tested as an inhibitor of 125I-LRSC binding, it gave a binding inhibition ratio of 0.19, indicating some affmity for the cell binding site. The carboxyterminal propeptides of type I procoUagen do not contain any collagenlike structure and they failed to inhibit 125I-LRSC binding in the standard assay system. The protein tested was a three-chain, disulfide-linked fragment of 100,000 daltons isolated from human type I procollagen (19). Quantitative inhibition assays have further defined the specificity of binding of native, soluble radioiodinated type I collagen to the fibroblast surface. The binding is not inhibited by every collagenous molecule; thus, the collagenlike tail of acetylcholinesterase and native collagen types IV, V and cartilage lct2~t3a do not inhibit binding. Only native collagen types II and III inhibit type I binding, and graphical analyses indicate that they act as competitive inhibitors. These inhibitions can be ascribed to interactions with type I cell binding sites or to complexes formed directly with type I collagen. Radioiodinated type II collagen binds to the fibroblast (unpublished data), so there is independent evidence for a cell binding site for this collagen type. In terms of receptor models, competitive inhibition can mean that collagens I, II, and III bind to the same cell receptor site, to different sites that interact allosterically, or to both shared and independent sites. Interactions of collagens I, II, and III with a common receptor on the fibroblast can be interpreted in the following context. Collagen types I and III are synthesized by the fibroblast and the mesenchymal precursor of the differentiated chondrocyte. Although the latter cell type synthesizes only type II collagen, it is prone to dedifferentiate and to switch to the synthesis of collagen types I and III (20). The data suggest, therefore, that the same cell binding complex is used in the course of these phenotype conversions. Previously published and less complete binding inhibition data supported the view that the primary structure of collagen a chains determined the binding specificity of the assembled molecule (1). We currently report that c~chains separated from collagen types V and cartilage la2a3a inhibit type I binding but that the triple helical molecules formed from such chains are not inhibitory in the assays. We now conclude, therefore, that tertiary or quartenary molecular structure can critically modify the binding specificities of the collagens. Enzymatic removal of telopeptides from native unlabeled type I collagen did not alter its capacity for inhibiting 125ILRSC binding nor obviate the requirement for thermal acti- We thank Sheila Heitner and Josa Williamson for their expert assistance. This study was supported by the National Institutes of Health grant HL 17551 and the Bear Foundation. Received for publication 26 April 1982, and in revised form 10 August 198Z REFERENCES I. Goldberg, B. 1979. Binding of soluble type I collagen molecules to the fibroblast plasma membrane. Cell. 16:265-275. 2. Goldberg, B. D. 1982. Binding of soluble type I collagen to fibroblasts: effects of thermal activation of ligand, ligand concentration, pinocytosis, and cytoskeletal modifiers. J. Cell BioL95:747-751. GOLDBERG AND BURGESON Speoficiti~s for Collagen Binding 755 Downloaded from on June 18, 2017 DISCUSSION vation of the ligand. Potential binding sites in helical al(I) and al(III) chains from different species were partially mapped by testing a limited number of cyanogen bromide-derived peptides as inhibitors. Inhibitory and noninhibitory peptides were thus identified, and inhibitory activity was shown to be subject to species variation. As all of the peptides shared the repeating collagen triplet Gly-X-Y and were not much different in their content of proline and hydroxyproline, these elements of collagen primary structure must be insufficient as determinants of binding specificity. Every al(I) and txl(III) chain was shown to contain two or more inhibitory peptides and, considered as a group, such peptides were variously positioned in the chains. It seems reasonable to suppose that some or all of these regions contribute to binding determinants expressed by the assembled triple helical molecules. Fibronectin is a reasonable candidate as the molecule responsible for the binding of collagens to the fibroblast surface. However, the ligand-binding specificities described above do not correspond to those reported for collagen-fibronectin interactions. In contrast to our data, fibronectin binds to the collagenlike tail of acetylcholinesterase (21), has different binding affinities for the various collagen types (22-24), and has been reported to bind exclusively to a single region of the td(I) chain (15). Because of these discordancies and other data (1), we do not favor the view that fibronectin is responsible for the binding of native, soluble x2SI-LRSC to the fibroblast surface. The reported data are interpreted to mean that ligand determinants distributed over the central collagen triple helix are sufficient to bind the interstitial procollagens and collagens (types I, II and III) to the fibroblast plasma membrane. Such binding is proposed as an in vivo mechanism for initiating fibril assembly and for specifying the spatial order of that assembly (1, 2). The essential element of the hypothesis is that binding to a membrane component is determinate for fibrillogenesis, and so the model can include those cases where procollagen fibrils apparently assemble within intraceUular vacuoles (25, 26). The binding model allows for modulations of fibril assembly by cell receptor metabolism as well as by ligand-ligand interactions and, if there are independent binding sites for each collagen type, the model provides a mechanism for segregation of type-specific fibrils. The model assumes that, upon achieving a limiting size, the soluble surface-bound aggregates detach from the cell and that they then combine to form insoluble fibrils. Perhaps it is at this later stage of fibrillogenesis that collagen-fibronectin and proteoglycan interactions become important. The binding model as formulated is quite compatible with proposals by others that type I collagen fibrils form by stepwise additions of molecular aggregates (see reference 27, for review). Published December 1, 1982 3. Segel+ I+ H. 1976. Biochemical calculations. J. Wiley and Sons, lnc++ NY+ 334. 4 Leibovich, S. J., and J. B+ Weiss. 1970, Electron microscope studies of the effects ofendoand exopeptidase digestion on tropocollagen. Biochim. Biophys. Aeta, 214:445~154. 5 Helseth, D. L.+ Jr., and A. Veis. 1981. Collagen self-assembly in vitro..L BioL Chem. 256:7118-7128. 6. Bornstein, P., A. H. Kang. and K. A. Piez. 1966. The limited cleavage of native collagen with chymotrypsin, trypsin, and cyanogen bromide. Biochemistry. 5:3803-3812. 7. Laemmli, U. K. 1970. Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature (Lond). 227:68~685. 8. Rosenberry, T. L., and J. M, Richardson. 1977. Structure of 18S and 14S acetylcholinesterase. Identification of coUagen-like subunits that are linked by disulfide bonds to catalytic subunits. Biochemistry. 16:3550-3558. 9. Mays, C.. and T, L, Rosenberry. 1981. Characterization of pepsin-resistant collagen-like tail subunit fragments of 18S and 14S acetylcholine-esterase from Electrophorus electricus. Biochemistry. 20:2810-2817. 10. Burgeson, R. E., and D. W. Hollister. 1979. Collagen heterogeneity in human cartilage: identification of several new collagen chains. Biochem. Biophys. Res. Commun. 87:11241131. I I. Hanes, C. S. 1932. Studies on Plant Amylases. Biochem..L 26:1406-1421. 12. Scgel, 1. H. 1975. Enzyme Kinetics John Wiley and Sons, Inc. NY. 210. 13. Burgeson, R. E., F. A. E1 Adli, I. I. Kaitila, and D. W. Hollister. 1976. Fetal membrane collagens: identification of two new collagen alpha chains. Proc. NatL Acad. Sci. U. S. A 73:2579 2583. t4. Gross, L+ and D. Kirk. 1958. The heat precipitation of collagen from neutral salt solutions: some rate-regulating factors..L BioL Chem. 233:355 360+ 15. Kleinman, H. K., E. B. MeGoodwin, G. R. Martin, R. J. Klebe, P P. Fietzek, and D. E. WoolIey. 1978. Localization of the binding site for cell attachment in the ~ I(I) chain of collagen. £ BioL Chem. 253:5642-5646. 16. Masui+ Y., T. Takemoto, S. Sakakibara, H. Hori. and Y. NagaL I977. Synthetic substrates for vertebrate collagenase. Biochem Med. 17:215-221. 17. Fessler, J. H., and L. I. Fessler. 1978. Biosynthesis of procollagen. Annu. Rev. Biochem. 47:129-162. 18. Pesciotta, D. M., M. H. Silkowitz, P. P. Fietzek, P. N. Graves, R. A. Berg, and B. R. Olsen. 1980. Purification and characterization of the amino-terminal propeptide of Pro c,l(1) chains from embryonic chick tendon procollagen. Biochemisto,. 19:2447-2454. 19. Sherr, C. J.. M. B. Taubman, and B. Goldberg. 1973. Isolation of a disulfide-stabilized, three-chain polypeptide fragment unique to the precursor of human collagen..L BioL Chem. 248:7033 7038. 20. yon der Mark, K. 1980. Immunological studies on collagen type transition in chondrogenesis. Cur. Top. Dey. BioL 14:t99-225. 21+ Emmerling, M. R., C. D. Johnson, D. F. Mosher, B. H Lipton. and J. E. Lilien 1981 Cross-linking and binding of fibronectin with asymmetric acetylcholinesterase. Biochemistry 20:3242-3247. 22. Dessau, W., B. C. Adelmann. R. Timpl, and G. R. Marlin. 1978. Identification of the sites in collagen a-chains that bind serum anti-gelatin factor (cold insoluble globulin) Biochem. J. 169:55 59. 23. Jilek, F., and H. Hormann. 1978. Cold-insoluble globulin (fibronectin), affinity to soluble collagen of various types. Hoppe-Seyler's Z. Physiol. Chem. 359:247-250. 24. Engvall, E., E. Ruoslahti, and E. J. Miller. 1978. Affinity of fibronectln to collagens of different genetic types and to fibrinogen. J. Exp. Meal 147:1584-1595. 25+ Trelstad, R. L. 1971. Vacuoles in the embryonic chick corneal epithelium, an epithelium which produces collagen. J. Cell BioL 48:689~694. 26. Weinstock, M.. and C. P. LeBlond 1974. Synthesis, migration, and release of precursor collagen by odontoblasts as visualized by radioautography after [:'H]pruline administration..L Cell BioL 60:92-127 27. Trelstad+ R. L. 1982. Multistep assembly of type 1 collagen fibrils. Cell 28:197 198. Downloaded from on June 18, 2017 756 THE JOURNAL OF CELL BIOLOGY • VOLUME 95, 1982