We describe cDNA clones for a cell surface proteoglycan that bears both heparan sulfate and chondroitin sulfate and that links the cytoskeleton to the interstitial matrix. The cDNA encodes a unique core protein of 32,868 D that contains several structural features consistent with its role as a glycosamino-glycan-containing matrix anchor. The sequence shows discrete cytoplasmic, transmembrane, and NH2-terminal extracellular domains, indicating that the molecule is a type I integral membrane protein. The cytoplasmic domain is small and similar in size but not in sequence to that of the beta-chain of various integrins. The extracellular domain contains a single dibasic sequence adjacent to the extracellular face of the transmembrane domain, potentially serving as the protease-susceptible site involved in release of this domain from the cell surface. The extracellular domain contains two distinct types of putative glycosaminoglycan attachment sites; one type shows sequence characteristics of the sites previously described for chondroitin sulfate attachment (Bourdon, M. A., T. Krusius, S. Campbell, N. B. Schwartz, and E. Ruoslahti. 1987. Proc. Natl. Acad. Sci. USA. 84:3194-3198), but the other type has newly identified sequence characteristics that potentially correspond to heparan sulfate attachment sites. The single N-linked sugar recognition sequence is within the putative chondroitin sulfate attachment sequence, suggesting asparagine glycosylation as a mechanism for regulating chondroitin sulfate chain addition. Both 5' and 3' regions of this cDNA have sequences substantially identical to analogous regions of the human insulin receptor cDNA: a 99-bp region spanning the 5' untranslated and initial coding sequences is 67% identical and a 35-bp region in the 3' untranslated region is 81% identical in sequence. mRNA expression is tissue specific; various epithelial tissues show the same two sizes of mRNA (2.6 and 3.4 kb); in the same relative abundance (3:1), the cerebrum shows a single 4.5-kb mRNA. This core protein cDNA describes a new class of molecule, an integral membrane proteoglycan, that we propose to name syndecan (from the Greek syndein, to bind together).

Cultured mouse mammary (NMuMG) cells produce heparan sulfate-rich proteoglycans that are found at the cell surface, in the culture medium, and beneath the monolayer. The cell surface proteoglycan consists of a lipophilic membrane-associated domain and an extracellular domain, or ectodomain, that contains both heparan and chondroitin sulfate chains. During culture, the cells release into the medium a soluble proteoglycan that is indistinguishable from the ectodomain released from the cells by trypsin treatment. This medium ectodomain was isolated, purified, and used as an antigen to prepare an affinity-purified serum antibody from rabbits. The antibody recognizes polypeptide determinants on the core protein of the ectodomain of the cell surface proteoglycan. The reactivity of this antibody was compared with that of a serum antibody (BM-1) directed against the low density basement membrane proteoglycan of the Englebarth-Holm-Swarm tumor (Hassell, J. R., W. C. Leyshon, S. R. Ledbetter, B. Tyree, S. Suzuki, M. Kato, K. Kimata, and H. Kleinman. 1985. J. Biol. Chem. 250:8098-8105). The BM-1 antibody recognized a large, low density heparan sulfate-rich proteoglycan in the cells and in the basal extracellular materials beneath the monolayer where it accumulated in patchy deposits. The affinity-purified anti-ectodomain antibody recognized the cell surface proteoglycan on the cells, where it is seen on apical cell surfaces in subconfluent cultures and in fine filamentous arrays at the basal cell surface in confluent cultures, but detected no proteoglycan in the basal extracellular materials beneath the monolayer. The amino acid composition of the purified medium ectodomain was substantially different from that reported for the basement membrane proteoglycan. Thus, NMuMG cells produce at least two heparan sulfate-rich proteoglycans that contain distinct core proteins, a cell surface proteoglycan, and a basement membrane proteoglycan. In newborn mouse skin, these proteoglycans localize to distinct sites; the basement membrane proteoglycan is seen solely at the dermal-epidermal boundary and the cell surface proteoglycan is seen solely at the surfaces of keratinocytes in the basal, spinous, and granular cell layers. These results suggest that although heparan sulfate-rich proteoglycans may have similar glycosaminoglycan chains, they are sorted by the epithelial cells to different sites on the basis of differences in their core proteins.

The proteoglycan (PG) on the surface of NMuMG mouse mammary epithelial cells consists of at least two functional domains, a membrane-intercalated domain which anchors the PG to the plasma membrane, and a trypsin-releasable ectodomain which bears both heparan and chondroitin sulfate chains. The ectodomain binds cells to collagen types I, III, and V, but not IV, and has been proposed to be a matrix receptor. Because heparin binds to the adhesive glycoproteins fibronectin, an interstitial matrix component, and laminin, a basal lamina component, we asked whether the cell surface PG also binds these molecules. Cells harvested with either trypsin or EDTA bound to fibronectin; binding of trypsin-released cells was inhibited by the peptide GRGDS but not by heparin, whereas binding of EDTA-released cells was inhibited only by a combination of GRDS and heparin, suggesting two distinct cell binding mechanisms. In the presence of GRGDS, the EDTA-released cells bound to fibronectin via the cell surface PG. Binding via the cell surface PG was to the COOH-terminal heparin binding domain of fibronectin. In contrast with the binding to fibronectin, EDTA-released cells did not bind to laminin under identical assay conditions. Liposomes containing the isolated intact cell surface PG mimic the binding of whole cells. These results indicate that the mammary epithelial cells have at least two distinct cell surface receptors for fibronectin: a trypsin-resistant molecule that binds cells to the sequence RGD and a trypsin-labile, heparan sulfate-rich PG that binds cells to the COOH-terminal heparin binding domain. Because the cell surface PG binds cells to the interstitial collagens (types I, III, and V) and to fibronectin, but not to basal lamina collagen (type IV) or laminin, we conclude that the cell surface PG is a receptor on epithelial cells specific for interstitial matrix components.

The cell surface proteoglycan on normal murine mammary gland (NMuMG) epithelial cells consists of a lipophilic domain, presumably intercalated into the plasma membrane, and an ectodomain that binds via its glycosaminoglycan chains to matrix components, is released intact by proteases and is detected by monoclonal antibody 281-2. The antibody 281-2 also detects a proteoglycan in the culture medium conditioned by NMuMG cells. This immunoactive proteoglycan was purified to homogeneity using DEAE-cellulose chromatography, isopycnic centrifugation, and 281-2 affinity chromatography. Comparison of the immunoreactive medium proteoglycan with the trypsin-released ectodomain revealed that these proteoglycans are indistinguishable by several criteria as both: (a) contain heparan sulfate and chondroitin sulfate chains; and (b) are similar in hydrodynamic size and buoyant density; (c) have the same size core protein (Mr approximately 53 kD); (d) are nonlipophilic as studied by liposomal intercalation and transfer to silicone-treated paper. Kinetic studies of the release of proteoglycan from the surface of suspended NMuMG cells are interpreted to indicate that the immunoreactive medium proteoglycan is derived directly from the cell surface proteoglycan. Suspension of the cells both augments the release and inhibits the replacement of cell surface proteoglycan. These results indicate that the cell surface proteoglycan of NMuMG cells can be shed by cleavage of its matrix-binding ectodomain from its membrane-associated domain, providing a mechanism by which the epithelial cells can loosen their proteoglycan-mediated attachment to the matrix.

The cell surface proteoglycan on normal murine mammary gland mouse mammary epithelial cells consists of an ectodomain bearing heparan and chondroitin sulfate chains and a lipophilic domain that is presumed to be intercalated into the plasma membrane. Because the ectodomain binds to matrix components produced by stromal cells with specificity and high affinity, we have proposed that the cell surface proteoglycan is a matrix receptor that binds epithelial cells to their underlying basement membrane. We now show that the proteoglycan surrounds cells grown in subconfluent or newly confluent monolayers, but becomes restricted to the basolateral surface of cells that have been confluent for a week or more; Triton X-100 extraction distinguishes three fractions of cell surface proteoglycan: a fraction released by detergent and presumed to be free in the membrane, a fraction bound via a salt-labile linkage, and a nonextractable fraction; the latter two fractions co-localize with actin filament bundles at the basal cell surface; and when proteoglycans at the apical cell surface are cross-linked by antibodies, they initially assimilate into detergent-resistant, immobile clusters that are subsequently aggregated by the cytoskeleton. These findings suggest that the proteoglycan, initially present on the entire surface and free in the plane of the membrane, becomes sequestered at the basolateral cell surface and bound to the actin-rich cytoskeleton as the cells become polarized in vitro. Binding of matrix components may cross-link proteoglycans at the basal cell surface and cause them to associate with the actin cytoskeleton, providing a mechanism by which the cell surface proteoglycan acts as a matrix receptor to stabilize the morphology of epithelial sheets.

Mouse mammary epithelial cells, of the normal murine mammary gland (NMuMG) cell line, bear a heparan sulfate-rich proteoglycan (HSPG) on their surfaces. A hybridoma (281-2) secreting a monoclonal antibody that recognizes this HSPG was produced by fusion of SP-2/0 myeloma cells with spleen cells from rats immunized with NMuMG cells. The 281-2 monoclonal antibody is directed against the core protein of the cell surface HSPG, as demonstrated by (a) recognition of the isolated proteoglycan but not its glycosaminoglycan chains, (b) co-localization of 281-2-specific antigen and radioactive cell surface HSPG on gradient polyacrylamide gel electrophoresis and on isopycnic centrifugation, and (c) abolition of immunofluorescent staining of the NMuMG cell surface by the intact, but not the protease-digested ectodomain of the cell surface HSPG. The antibody is specific for cell surface HSPG and does not recognize the HSPG that accumulates extracellularly beneath the basal cell surface. Therefore, the 281-2 antibody may be used to isolate the cell surface HSPG and to explore its distribution in tissues.

When mouse mammary epithelial cells are cultured on a plastic substratum, no basal lamina forms. When cultured on a type I collagen gel, the rate of glycosaminoglycan (GAG) synthesis is unchanged, but the rate of GAG degradation is markedly reduced and a GAG-rich, basal lamina-like structure accumulates. This effect of collagen was investigated by comparing the culture distribution, nature, and metabolic stability of the 35S-GAG-containing molecules produced by cells on plastic and collagen. During 48 h of labeling with 35SO4, cultures on collagen accumulate 1.4-fold more 35S-GAG per microgram of DNA. In these cultures, most of the extracellular 35S-GAG is immobilized with the lamina and collagen gel, whereas in cultures on plastic all extracellular 35S-GAG is soluble. On both substrata, the cells produce several heparan sulfate-rich 35S-proteoglycan fractions that are distinct by Sepharose CL-4B chromatography. The culture types contain similar amounts of each fraction, except that collagen cultures contain nearly four times more of a fraction that is found largely bound to the lamina and collagen gel. During a chase this proteoglycan fraction is stable in cultures on collagen, but is extensively degraded in cultures on plastic. Thus, collagen-induced formation of a basal lamina correlates with reduced degradation and enhanced accumulation of a specific heparan sulfate-rich proteoglycan fraction. Immobilization and stabilization of basal laminar proteoglycan(s) by interstitial collagen may be a physiological mechanism of basal lamina maintenance and assembly.