Differential Expression of Cell Surface Heparan Sulfate Proteoglycans in Human Mammary Epithelial Cells and Lung Fibroblasts *

Treating the liposome-intercalatable heparan sulfate proteoglycans from human lung fibroblasts and mammary epithelial cells with heparitinase and chondroitinase ABC revealed different core protein patterns in the two cell types. Lung fibroblasts expressed heparan sulfate proteoglycans with core proteins of -35,48190 (fibroglycan), 64 (glypican), and 125 kDa and traces of a hybrid proteoglycan which carried both heparan sulfate and chondroitin sulfate chains. The mammary epithelial cells, in contrast, expressed large amounts of a hybrid proteoglycan and heparan sulfate proteoglycans with core proteins of -35 and 64 kDa, but the fibroglycan and 125-kDa cores were not detectable in these cells. Phosphatidylinositol-specific phospholipase C and monoclonal antibody (mAb) S1 identified the 64-kDa core proteins as glypican, whereas mAb 2E9, which also reacted with proteoglycan from mouse mammary epithelial cells, tentatively identified the hybrid proteoglycans as syndecan. The expression of syndecan in lung fibroblasts was confirmed by amplifying syndecan cDNA sequences from fibroblastic mRNA extracts and demonstrating the cross-reactivity of the encoded recombinant core protein with mAb 2E9. Northern blots failed to detect a message for fibroglycan in the mammary epithelial cells and in several other epithelial cell lines tested, while confirming the expression of both glypican and syndecan in these cells. Confluent fibroblasts expressed higher levels of syndecan mRNA than exponentially growing fibroblasts, but these levels remained lower than observed in epithelial cells. These data formally identify one of the cell surface proteoglycans of human lung fibroblasts as syndecan and indicate that the expression of the cell surface proteoglycans varies in different cell types and under different culture conditions.

Differential Expression of Cell Surface Heparan Sulfate Proteoglycans in Human Mammary Epithelial Cells and Lung Fibroblasts* (Received for publication, June 17, 1991) Veerle Lories, Jean-Jacques Cassiman, Herman Van den Berghe, and Guido David$ From the Center for Human Genetics, University of Leuuen, B-3000 Leuuen, Belgium Treating the liposome-intercalatable heparan sulfate proteoglycans from human lung fibroblasts and mammary epithelial cells with heparitinase and chondroitinase ABC revealed different core protein patterns in the two cell types. Lung fibroblasts expressed heparan sulfate proteoglycans with core proteins of -35,48190 (fibroglycan), 64 (glypican), and 125 kDa and traces of a hybrid proteoglycan which carried both heparan sulfate and chondroitin sulfate chains. The mammary epithelial cells, in contrast, expressed large amounts of a hybrid proteoglycan and heparan sulfate proteoglycans with core proteins of -35 and 64 kDa, but the fibroglycan and 125-kDa cores were not detectable in these cells. Phosphatidylinositol-specific phospholipase C and monoclonal antibody (mAb) S 1 identified the 64-kDa core proteins as glypican, whereas mAb 2E9, which also reacted with proteoglycan from mouse mammary epithelial cells, tentatively identified the hybrid proteoglycans as syndecan. The expression of syndecan in lung fibroblasts was confirmed by amplifying syndecan cDNA sequences from fibroblastic mRNA extracts and demonstrating the cross-reactivity of the encoded recombinant core protein with mAb 2E9. Northern blots failed to detect a message for fibroglycan in the mammary epithelial cells and in several other epithelial cell lines tested, while confirming the expression of both glypican and syndecan in these cells. Confluent fibroblasts expressed higher levels of syndecan mRNA than exponentially growing fibroblasts, but these levels remained lower than observed in epithelial cells. These data formally identify one of the cell surface proteoglycans of human lung fibroblasts as syndecan and indicate that the expression of the cell surface proteoglycans varies in different cell types and under different culture conditions. The proteoglycans (PGs)' from the cell surface bind a *These investigations have been supported by Grant 3.0066.87 from the National Science Foundation of Belgium, by a grant "Geconcerteerde Acties" from the Belgian Government, and by the "Interuniversity Network" for Fundamental Research sponsored by the Belgian Government (1987)(1988)(1989)(1990)(1991). 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.
$ Research Director of the National Science Foundation of Belgium. To whom correspondence should be addressed Center for Human Genetics, Campus Gasthuisberg O&N, Herestraat 49, B-3000 Leuven, Belgium.
Epithelial and mesenchymal cell types, however, produce distinct extracellular matrices, assume different relationships to these matrices, and often respond to different growth factors . Syndecan is expressed by several simple and stratified epithelia (Hayashi et al., 1987), but it is hardly detectable in mesenchymal cell types, except perhaps during development, where it is transiently expressed in condensing mesenchymes Vaino et al., 1989;Solursh et al., 1990). Moreover, previous investigations on mouse mammary epithelial cells and human fetal lung fibroblasts have yielded different grades of complexity for the cell surface PGs in these two cell types. In the mouse mammary epithelial cells two major core proteins were identified, i.e. that of syndecan and the 38-kDa core protein of a distinct HSPG . Human fibroblasts, in contrast, expressed at least four major structurally distinct HSPGs  and two CSPGs . Two of these fibroblast HSPGs, fibroglycan (Marynen et al., 1989) and glypican  have been cloned and shown to be distinct from syndecan. These data imply that different cell types or different situations may engage different proteoglycans.
To further investigate this possibility we have characterized the hydrophobic cell surface PGs from human mammary epithelial cells and attempted to relate these to the cell-surface PGs from human fibroblasts and mouse mammary epithelial cells. It appeared that the human mammary epithelial cells synthesized large amounts of a hybrid HS/CSPG which is related to the mouse epithelial hybrid PG syndecan. This hybrid proteoglycan was also produced by confluent human fetal lung fibroblasts. The glycosyl phophatidylinositol-linked HSPG, glypican, was also expressed by both cell types, but fibroglycan and one other membrane HSPG from fibroblasts were not detectable in the mammary epithelial cells, indicating that distinct human cell types display distinct patterns of cell surface PGs.

MATERIALS AND METHODS
Cell Culture-The human fetal lung fibroblasts, and the human (HBL100) and mouse (NMuMG) mammary gland epithelial cell lines were cultured on plastic substrata in a (1:l) mixture of Ham's F-12 and Dulbecco's modified Eagle's medium containing 10% (v/v) fetal calf serum. For metabolic labeling of the proteoglycans confluent monolayers were incubated with 5 pCi/ml carrier-free HZ3'S04 during 24-48 h (Lories et al., 1986).
Isolation and Radiolabeling of the Hydrophobic Cell Surface Proteoglycans-Only confluent cultures were used to avoid variations in proteoglycan expression due to different growth states of the cells. The cultures were rinsed with phosphate-buffered saline and extracted with ice-cold Triton X-100 buffer. Triton X-100 buffer contained 0.5% (v/v) Triton X-100, 20 mM Tris-CI, pH 8.0, 150 mM NaCI, and several protease inhibitors as described previously (Lories e t al., 1986). The extracts were cleared by centrifugation (10.000 X g; 60 min) and concentrated by adsorption on DEAE-Sepharose Fast Flow. The PG fractions were isolated from the DEAE eluates by ionexchange chromatography on Mono Q. For experiments which included autoradiography the isolated PGs were adsorbed on DEAE beads and '"I-iodinated using the chloramine-T procedure (Lories et al., 1988). Hydrophobic PGs were isolated by incorporation into lipid vesicles, followed by gel filtration on Sepharose CL4B columns in the absence of detergent (Lories et al., 1987). Further fractionations by immunoaffinity chromatography using mAbs 2E9, S1, and 1C7 were as described before .
Gel Electrophoresis and Western Blotting-Electrophoresis was performed in SDS-polyacrylamide gradient (4-16% T , 2.5% or 10% C) gels. Running conditions, sample preparation, and autoradiography were as described previously (Lories et al., 1988). For immunostaining, the electrophoresed samples were immunoblotted (0.5 A, 3-4 h) to Zeta-probe or to Hybond-N membranes. Immunostaining, using the appropriate antibodies was as described before . Apparent molecular weights were determined by comparison to prestained molecular weight markers (GIBCO BRL, Ghent, Belgium).
Isolation and Sequencing of Human Syndecan cDNA-Sequences coding for human syndecan were amplified from cDNA obtained in the first strand synthesis reaction during the construction of a human lung fibroblast cDNA library (David et al., 1990). The reaction mixture contained 2 units of Taq polymerase, 1 ng of cDNA, and 140 pmol of the sense and antisense primers. These were 42-mers identical, respectively, to the murine syndecan cDNA sequence (from residue 240 to 281) and to the complement of this sequence (from residue 1131 to 1172), as reported by Saunders et al. (1989). After 40 thermal cycles (1 min of denaturation a t 94 "C, 1 min of annealing a t 55 "C, 1 min of extension at 68 "C), the amplification products were analyzed in 1.1% agarose gels and detected by ethidium bromide staining. The -900-base pair amplification produce was electroeluted, ethanol precipitated, phosphorylated, and ligated in the SmaI restriction site of pGEM-3Z (Promega Corporation, Madison, WI). The termini of five independent clones and, after subcloning, both strands of the PstI fragments of three of these clones were sequenced by the dideoxy chain termination method (Sanger et al., 1977), using supercoiled plasmid, a modified T7 DNA polymerase (Pharmacia LKB Biotechnology, Uppsala, Sweden), T7 and SP6 primers, and both dGTP and C7-deaza dGTP.
Construction of the Expression Plasmids and Identification of the Recombinant Protein-For expression as a recombinant P-galactosidase-syndecan fusion protein, one of the fully sequenced syndecan cDNA clones obtained by polymerase chain reaction (HUSYN-2) was linearized with EcoRI and blunted with Klenow enzyme. The insert and short flanking sequences were liberated from the vector by cleavage with Sal1 and ligated to a SmaI and SalI-restricted pEX 3 expression vector (Genofit, Geneva, Switzerland). In this construct (HUSYN-pEX3) the human syndecan cDNA sequence is inserted, in frame, into P-galactosidase-coding sequences through the intermediate of a short open sequence derived from the multiple cloning site of the pGEM-3Z vector. Transformed POP 2136 cells were selected a t 28 "C on ampicillin plates. Exponentially growing cultures were initiated from single colonies and induced by shifting the incubation temperature from 28 to 42 "C. After treatment with lysozyme, the cells were solubilized in hot SDS and the extracts analyzed I)? polyacrylamide gel electrophoresis. Western blotting and immunostaining with the anti-HSPG monoclonal antibodies 281-2, 2E9, a n ( 1 1C7, and with a monoclonal anti-P-galactosidase antibody (Promew Corporation) were as reported before .
Isolation and Analysis of RNA-The isolation of poly(A)' RNA from confluent human lung fibroblasts and from confluent human mammary epithelial cells was as described before (Marynen et nl., 1989). Total RNA samples from cells a t different stages of confluence were prepared by the acid guanidinium thiocyanate phenol-chloraform extraction protocol (Chomczynski and Sacchi, 1987).
For Northern analysis, aliquots of 3 pg of poly(A)+ RNA or 10 pg of total RNA/lane were separated in formaldehyde-containing 1.2% agarose gels. The RNA was transferred to Nytran membranes and cross-linked to the membrane by exposure to UV light. The blots were prehybridized (4 h a t 42 "C) in 50% formamide, 0.1% SDS, 5 X SSPE, 5 X Denhardt's, 100 pg/ml denatured salmon sperm DNA, and 100 pg/ml heparin. Hybridization (16 h a t 42 "C) was in the same buffer containing 4.106 cpm '"P-oligolabeled cDNA/ml. The filters were washed in 2 X SSPE, 0.1% SDS at room temperature (2 X 15 min) and at 42 "C (1 X 30 rnin), and in 0.1 X SSPE, 0.1 SDS at 55 "C (1 X 20') and 65 "C (1 X 25'). Methylene blue-stained RNA markers (GIBCO BRL) and 18 S and 28 S ribosomal RNA were used to determine the molecular size of the hybridizing RNA species.

RESULTS
Distinctive Cell Surface Proteoglycan Patterns-The hydrophobic cell surface PGs synthesized by the human mammary epithelial cells were purified from detergent extracts of the cells by ion-exchange chromatography and liposome incorporation and were radioiodinated to analyze their protein cores. During gel electrophoresis these "'I-labeled hydrophobic proteoglycans ran as a high molecular mass (>200 kDa) smear (Fig. 1, lane 1 ) . After treatment with heparitinase, two major core proteins of -35 and 64 kDa and a residual smear (200-90 kDa) were obtained (Fig. 1, lane 2). This smear was converted into an 88-kDa protein band when the PG fraction was digested with both condroitinase ABC and heparitinase ( Fig. 1, lune 3 ) . Besides this 88-kDa core the combined digestion yielded two additional proteins of 115 and 130 kDa. These were derived from chondroitin sulfate proteoglycan as they were also observed after chondroitinase digestion only (Fig. 1, lane 4 ) . Reduction of the disulfide bonds did not change the electrophoretic behavior of the ""I-bands (results not shown). Thus, cultured human mammary epithelial cells synthesized multiple types of cell surface PG, expressing core proteins of -35 and 64 kDa which carried only HS chains, core proteins of 115 and 130 kDa which carried only CS
chains, and an 88-kDa core protein which carried both HS and CS glycosaminoglycan chains. Similar analyses on radioiodinated cell surface heparan sulfate proteoglycans from human lung fibroblasts (Fig. 1, lanes 5-8) yielded several core protein bands as reported before (Lories et al., 1987. These included prominent bands of -125 and 48 kDa, not seen in the detergent extracts of the epithelial cells. These results were suggestive for differential expression of proteoglycan in epithelial and fibroblastic cells. Immunological Relationships between Human Epithelial and Fibroblast Proteoglycans-A panel of mAbs raised against the cell surface HSPGs of human lung fibroblasts  was used to investigate possible relationships between the PGs from fibroblastic and epithelial cells. From initial immunodot blot assays it appeared that the core protein epitopes recognized by mAb 2E9 and by mAb S1 (anti-glypican) were detectable in the detergent extracts of the human epithelial cells, but that the epitopes of mAb 6G12 (antifibroglycan) and mAb 1C7 were absent (not shown).
The antibodies were then used to further immunopurify the radioiodinated epithelial and fibroblastic PGs prior to electrophoresis and autoradiography. The epithelial PGs recognized by mAb 2E9 migrated as a smear >88 kDa after heparitinase digestion (Fig. 2, lane 5) and as an -88-kDa band after combined heparitinase and chondroitinase ABC digestion (Fig. 2, lane 6). In contrast, the '251-PGs of fibroblasts which bound to mAb 2E9 yielded both a 125-kDa core protein and a smear >82 kDa after heparitinase digestion (Fig. 2, lane 3 ) . This smear was converted into a sharp band after combined heparitinase and chondroitinase digestion (Fig. 2, lane 4 ) , indicating (more clearly than in experiments with whole proteoglycan extracts as shown in Fig. 1) that fibroblasts also synthesized a HS/CS hybrid PG. Although both hybrids shared the mAb 2E9 epitope, the core protein of the fibroblast hybrid PG appeared to be somewhat smaller in size than the core protein of the epithelial hybrid PG (82 versus 88 kDa). Interestingly, mAb 1C7, which did not react with extracts from epithelial cells, bound only the "125 K"-HSPG from fibroblast PG extracts (Fig. 2, lanes 1 and 2) suggesting that the 125 kDa and the -82-kDa cores contained structurally distinct peptides.
The mAb S1-reactive proteoglycans from epithelial cells and fibroblasts, finally, resembled each other very closely. They migrated as 64-kDa core proteins after heparitinase digestion, and supplementary chondroitinase treatment did not change their migration (Fig. 2, lanes 7-20).
in immunostainings of Western blots (not shown). Thus, human mammary epithelial cells and fibroblasts seemed to share a hybrid-PG and glypican (the HSPG with a 64-kDa core), but fibroglycan which yields the 48-kDa core (Marynen et al., 1989), and the HSPG with the 125-kDa core protein appeared unique to fibroblasts (see Table I).
Glypiated Epithelial Proteoglycans-To test whether, like glypican in fibroblasts , some epithelial proteoglycans might be membrane-anchored through a glycosyl phosphatidylinositol moiety, the epithelial proteoglycans were treated with PI-specific phospholipase C and tested for their residual hydrophobic properties (Fig. 3). PI-PLC nearly completely abolished the ability of the mAb S1-reactive epithelial proteoglycans to intercalate into liposomes, causing them to elute in the included volume of a Sepharose CL4B column (Fig. 3A). Control preparations of S1-reactive proteoglycans which had not been treated with PI-PLC, coeluted with the liposomes, mostly in the excluded volume of the column (Fig. 3A). The epithelial proteoglycans which had not been retained on the S1 column, in contrast, coeluted with the liposomes, whether treated with PI-PLC or not (Fig. 3B). Heparitinase and chondroitinase ABC digestion followed by SDS-polyacrylamide gel electrophoresis and autoradiography confirmed that the 64-kDa core protein had been enriched in the S1-bound and depleted from the non-bound proteoglycan fractions, and indicated that the PI-PLC treatment had no detectable effect on the apparent size of the core protein (not shown). Thus, as with fibroblasts, the PI-PLC-sensitive proteoglycans were primarily represented by those that harbored the S1 epitope and the 64-kDa core.
Relationships between the Human Hybrid Proteoglycans and Syndecan-To test whether the human hybrid PGs might be related to syndecan (Saunders et al., 1989), Western blots of partially purified cell surface PGs from mouse mammary epithelial cells were assayed with mAb 3G10, which reacts with A4-uronate generated by heparitinase and therefore stains the heparitinase-generated core of any HSPG ; with mAb 2E9, the anti-human HSPG mAb, and with mAb 281-2, an anti-mouse syndecan mAb (Jalkanen et al., 1985). mAb 3G10 revealed multiple "HS stub"-carrying core proteins, among which an -84-kDa core protein which ran as a smear after heparitinase digestion (Fig. 4, lane 2) but as a sharp band after combined heparitinase and chondroitinase digestion (Fig. 4, lane 3). The two anti-core antibodies mAb 281-2 and mAb 2E9 reacted only with the -84-kDa core protein of the hybrid proteoglycan (Fig. 4, lanes 7 and 9), indicating that syndecan and the human hybrid proteoglycans shared the mAb 2E9 epitope.
The presence of a syndecan homologue in human fibroblasts was confirmed by amplifying the corresponding cDNA from human lung fibroblasts using the polymerase chain
The hydrophobic proteoglycans were isolated from the S1 eluate ( A ) and fall through fractions ( B ) by liposome incorporation, and retested for their ability to intercalate into phospholipid vesicles either after treatment with phosphatidylinositol-specific phospholipase C (O), or after treatment with enzyme buffer only (0). Association with the liposomes was tested by fractionating the proteoglycan-liposome mixture over Sepharose-CL4B in 4 M guanidinium chloride buffer. All proteoglycan fractions eluted in the included volume of the column if detergent was added to the samples and elution buffer (not shown).  (lanes 1 and 5 ) or were digested with heparitinase only (lanes 2 and 6), with both heparitinase and chondroitinase ABC (lanes 3, 7, and  9), or with chondroitinase ABC only (lanes 4 and 8) before electrophoresis and transfer to nylon membranes. The Western blots were stained with mAb 3G10 (lanes 1-4), mAb 2E9 (lanes 5-8), and mAb 281-2 (lane 9). reaction technique and primers derived from the murine syndecan cDNA sequence (Fig. 5). The amplified human fibroblast cDNA sequences coded for a protein which was 75% homologous to the mouse syndecan core protein (Saunders et  al., 1989) and identical, except for one amino acid, to the R u m -2 ~t g a g~~e g c g g c g c t c t~c t c t g g c t c t 9~~t~c~~~ V 0

A T P R P R K T T Q L P T T H Q A S T T
ACAGt XUXACCGCCCAKA"""-  (lanes 1 and 2 ) , with mAb 2E9 (lanes 3 and 4 ) , and with mAb 1C7 (lanes 5 and 6 ) . protein sequence predicted by human syndecan cDNA clones isolated from a HBL-100 library with the aid of the murine syndecan probe .

K D K G S Y S L K K P K Q A H G G A Y Q ~a~c c c a c c~9 c a g g a~9 a g t t c t a~~c K P T K Q E E P Y A
Consistently, supernatants of lysed POP2136 cells which had been transfected with pEX plasmids containing a human fibroblast syndecan cDNA insert and induced by heat shock contained the mAb 2E9 epitope. Western blotting indicated that the 2E9 epitope occurred in an -180-kDa protein (Fig.  6, lane 3) which was also traced by an anti-P-galactosidase mAb (Fig. 6, lane I ). The P-galactosidase-syndecan fusion protein did not react with mAb 1C7 (Fig. 6, lane 5 ) nor did the lysate of control cells infected with pEX vector only react with mAb 2E9 (Fig. 6, lane 4 ) .
Northern Blot Amlysis-To study the expression of the different cell surface proteoglycans at the transcriptional level, Northern blots of poly(A)+ RNA prepared from human lung fibroblasts and from human mammary epithelial cells were hybridized with 32P-oligolabeled 48K3, a fibroglycanspecific cDNA (Marynen et al., 1989), with 64K3, a glypicanspecific cDNA , and with HUSYN-2, one of the human syndecan-specific cDNAs, and submitted to autoradiography (Fig. 7). Mammary epithelial cells did not contain a mRNA which hybridized with the 48K3 probe, whereas in human lung fibroblasts a major -2.3 kb and a minor -3.5-kb mRNA were recognized (Fig. 7, lanes 1 and 2). The 64K3 probe detected a major 3.7-kb mRNA in both human epithelial cells and fibroblasts (Fig. 7, lanes 3 and 4 ) . Epithelial poly(A)+ RNA also contained a second, less abundant 64K3-labeled RNA of 1.9 kb. This band was also observed in poly(A)+ RNA from fibroblasts after prolonged exposure (not shown). The human syndecan probe revealed two mRNA bands of 2.5 and 3.3 kb in both cell types (Fig. 7,  lanes 5 and 6 ) , but the syndecan message appeared to be relatively less abundant in lung fibroblasts than in the epithelial cells. In fibroblasts there was a clear difference in the relative abundance of the syndecan message depending on the growth state of the cells. Fibroblasts from confluent and postconfluent cultures expressed higher levels of syndecan message than cells from exponentially growing cultures (Fig. 8).
Although not necessarily reflecting syndecan, immunodot blot assays suggested that the increase in syndecan message was paralleled by enhanced expression of the 2E9 epitope (not shown). Under similar circumstances no such differences were seen with the "48K" (Fig. 8) and the "64K probes (not shown). Growing HBL-100 cells, in contrast, yielded strong signals for both syndecan and glypican, but the 48K message remained undetectable (not shown). Finally, Northern analysis of total RNA samples from a panel of cultured human cells revealed large individual differ- Human lung fibroblasts were plated at one-tenth of their confluent density. The RNA was extracted after 1, 2, 3, 4, 7, and 9 days of culture. Ten-pg aliquots of total RNA from each of the extracts were fractionated in agarose gels and hybridized to the EcoRI-PstI (bases 1-1389) generated fragment of the 48K3 probe (FIB) and to the HUSYN-2 probe ( S Y N ) . Only the major (2.3 kb) fibroglycan message is shown. ences in the message levels for these three proteoglycans, extending the findings of proteoglycan heterogeneity and variability, as observed in human fibroblasts and mammary epithelial cells, to other cell lines (Fig. 9).

DISCUSSION
In the present study we have isolated and characterized the cell surface HSPGs of human mammary epithelial cells and related the human epithelial PGs to previously described human fibroblast and mouse epithelial PGs by using a panel of anti-fibroblast HSPG mAbs and PG-specific cDNA probes. The results are indicative of differential expression of the cell surface proteoglycans (Table I).
Cell Surface Proteoglycans with Restricted Distributions-The major cell surface HSPG of human lung fibroblasts has a core protein with apparent M , of 48,000 (see Fig. 1). From cDNA sequencing it appears to have the characteristics of an integral membrane protein featuring a membrane spanning hydrophobic domain, a cytoplasmic domain at the carboxyl terminus, and an NH2-terminal extracellular domain with the attachment sites for the heparan sulfate side chains (Marynen et al., 1989). This proteoglycan reacts with mAbs 6G12 and 10H4 , and has been named "fibroglycan" (David, 1990).
Detergent extracts of human mammary epithelial cells showed no reactivity with the 48K-specific mAbs 6G12 or mAb 10H4 and contained no core proteins of 48 kDa, suggesting that these cells did not synthesize a HSPG related to fibroglycan. This was confirmed by Northern blot analysis of both proliferating and stationary human mammary epithelial cells, which also ruled out that in these epithelial cells the core protein might be expressed in an alternative glycoprotein form that would behave differently from the proteoglycans during the purification procedure. Of all human cells analyzed in the panel, skin and lung fibroblasts, neuroblastoma, and retinoblastoma cells had the highest levels of fibroglycan mRNA (Fig. 9). The fibroglycan message was also detectable in glioma, sarcoma, teratoma, vascular endothelial, and HEP 3B cells, but it was virtually absent from the other cell lines, suggesting that high levels of this proteoglycan may be characteristic for cells of mesenchymal and neuroectodermal origin. Syndecan and glypican also showed variable levels of expression, but their messages were detectable in most of the cell lines tested. With the exception of HEP 3B cells, all cell lines with relatively high levels of fibroglycan mRNA contained relatively low levels of syndecan mRNA, suggesting that these two proteoglycans may have opposite patterns of expression. Further experiments are needed to confirm this and to assess whether additional proteoglycan forms with a restricted expression might exist, since a 125-kDa core and the 1C7 epitope were also not detected in the mammary epithelial cells (Fig. 1).
Glypican Expression in Epithelial Cells-The cell surface HSPG from human fibroblasts which is characterized by the 64-kDa core protein and the S1 epitope is associated with the cell surface through a glycosyl phosphatidylinositol anchor and has therefore been named glypican . Glypican cDNA clones isolated from fibroblasts predict a protein with a short stretch of hydrophobic amino acids at its carboxyl terminus and apparently no cytoplasmic domain (David, 1990). They also indicate that glypican is molecularly distinct from fibroglycan (Marynen et al., 1989) and from syndecan (Saunders et al., 1989;Mali et al., 1990). Apparently, human mammary epithelial cells also express glypican. They synthesize a 64-kDa protein which carries HS chains (Fig. 1, lane 2) and the S1 epitope (Fig. 2, lane 7). Moreover, enzyme susceptibility tests confirmed the presence of the phospholipid anchor in the epithelial forms (Fig. 3). The expression of glypican in both cell types was confirmed by Northern blot analysis, revealing a major -3.7-kb mRNA in both epithelial and fibroblast poly(A)+ RNA preparations, and perhaps a minor 1.9-kb mRNA in the epithelial preparations (Fig. 7). This smaller mRNA could also be detected in the fibroblast preparation after prolonged exposure of the filter but was not detected with a 64K3 probe from which the 5"untranslated sequences had been removed. Although probably nonspecific, the origin of this minor signal remains to be further unraveled and its occurrence implies that the possibility of variant forms of glypican should be considered, especially in epithelial cells. Partial or even complete resistance to phospholipase C is not uncommon among proteins with known glycosyl phosphatidylinositol anchors, probably due to additional modification of the lipid moiety, but the 5-15% of residual hydrophobic glypican following the enzyme treatment could be representative of such variant forms. A cell surface HSPG with a core protein of about 64 kDa and with a phospholipid membrane anchor has also been isolated from Schwann cells (Carey and Stahl, 1990), and may represent the equivalent of glypican. The screening of the cell panel indeed indicates that glypican is widely distributed.
Syndecan Expression in Fibroblasts-Both human mammary epithelial cells and fibroblasts produce hybrid HS/CS cell surface proteoglycans with 82-88-kDa core proteins that migrate only as sharp protein bands after both heparitinase and chondroitinase ABC digestion. In fibroblasts, the hybrid PG is a relatively minor component of the total labeled PG fraction, and when heparitinase-digested total PG samples are additionally digested with chondroitinase ABC the change in the migration pattern of the core protein is not easily detected (Fig. 1, compare lane 6 to lane 7). It is only when mAb 2E9-immunopurified samples are doubly digested that the hybrid PG core protein is clearly visualized (Fig. 2, lanes  3 and 4 ) .
These human hybrid proteoglycans share the 2E9 epitope with syndecan (Fig. 4), a hybrid cell surface PG first identified in mouse mammary epithelial cells (Rapraeger et al., 1985;David and Ven den Berghe, 1985), and seem to represent the human homologue of this molecule. Direct evidence for the presence of syndecan in human lung fibroblasts was obtained by amplifying a cDNA that is homologous to syndecan from lung fibroblast mRNA extracts, and demonstrating the reaction of mAb 2E9 with the encoded recombinant protein (Fig.  6). These data complement the results of Mali et al. (1990) who recently isolated a syndecan-homologous sequence from a HBL-100 cDNA library. Their reported sequence codes for a protein which is identical, except for one amino acid at position 19, to the sequence predicted by the lung fibroblast cDNAs (Fig. 5). The single difference may represent a structural polymorphism, rather than a Taq polymerase error, since it was identified in five independent isolates. The variant codon substitutes a proline for a leucine in the human fibroblast sequence and occurs in a region which has been highly conserved in mouse and man. Incidentally or not, the murine protein also features a proline at this position (Saunders et al., 1989). Then, comparison of the tyrosine contents of the predicted sequences for human fibroglycan (Marynen et al., 1989), syndecan , and glypican  indicates that, at similar specific activities of labeling, the amount of '251/mole of core protein for fibroglycan will be twice, and for glypican nearly three times as high as for syndecan. Interpretation of the relative intensities of the bands on the autoradiograms in this light underscores that syndecan represents a significant fraction of the cell surface proteoglycans from fibroblasts and the major form in mammary epithelial cells.
Synthesis of syndecan, or at least of a proteoglycan which cross-reacts with mAb 281-2 in mesenchymal cells has been described before. Immunocytochemistry indeed suggests that during tooth (Vaino et al., 1989) and limb (Solursh et al., 1990) morphogenesis, syndecan is transiently detectable in the condensing mesenchyme. In these instances, synthesis appears to be induced by epithelial-mesenchymal tissue interactions. Apparently, confluent in vitro cultured human fetal lung fibroblasts are able to synthesize syndecan in the absence of such tissue interactions. The enhancement of the expression of syndecan in dense fibroblast cultures and in condensing mesenchymal tissues suggests, however, that the expression of this proteoglycan may be modulated, and related to or regulated by the formation of cell-cell contacts. These findings are reminiscent of the changes in syndecan expression during the ontogeny of B lymphocytes (Saunders et al., 1989) and of the important spatial and quantitative fluctuations in syndecan expression during the steroid-dependent cyclic changes of vaginal epithelial differentiation (Hayashi et al., 1988). The fluctuations in syndecan expression during culture are also consistent with prior findings which had noticed differences in the composition and physical properties of the cell sulfate proteoglycans in confluent and proliferating fibroblasts (Coster et al., 1986).
Northern blot analysis using the polyermase chain reaction product as a probe revealed the presence of two syndecan mRNAs of 2.5 and 3.3 kb in both human lung fibroblasts and human mammary epithelial cells. The syndecan message was clearly less abundant in fibroblasts than in epithelial cells, but in confluent fibroblasts the signal was quite distinct, indicating that the polymerase chain reaction product did not result from the amplification of an extremely rare fibroblast transcript. The sizes of the two syndecan mRNAs are reminiscent of the 2.4-and 3.5-kb syndecan mRNAs detected in murine cells and tissues with the original murine probe (Saunders et al., 1989) and of the mRNA bands detected in HBL-100 cells by others using c-DNA probes isolated from these cells . The origin of these double mRNA bands is unknown, but they are thought to represent variations in noncoding parts of the messages (Saunders et al., 1989). In this context it is interesting that a cell surface heparan sulfate proteoglycan was found, in fibroblasts only, which shows partial immunological cross-reaction with syndecan, as it shares the 2E9 epitope, but which is distinguishable from syndecan by the absence of CS chains, by the size of its core protein and by the presence of the 1C7 epitope in its peptide moiety . Whereas these data imply that the 125-kDa core protein could be a structural variant of the syndecan core protein, it seems unlikely that this variant is encoded by one or the other of these two syndecan messages, as both messages occur in similar relative abundances in the epithelial cells where the lC7-reactive proteoglycan is not expressed. Otherwise one would have to postulate post-transcriptional or post-translational controls on the expression of this proteoglycan. More knowledge on the primary structure of the 125-kDa core protein should resolve this question. Finally, whereas in human cells the 3.3kb band appeared to be slightly more intense than the 2.4-kb band (Fig. 7), the mRNA bands in murine cells and tissues created the opposite pattern with the 2.4-and 3.5-kb band occurring in 3:l proportions (Saunders et al., 1989). Again, the significance of this finding is unknown.
In conclusion, the present data suggest that different cells and a given cell under different circumstances may express different proteoglycans or combinations of proteoglycans at their cell surfaces. It is tempting to speculate that these differences might allow the establishment of different cellmatrix contacts or support responses to different growth factors, or that the different core protein genes and their products might be differentially regulated providing the cells with means to respond differentially to factors that can modulate the abundance of heparan sulfate at the cell surface and thereby affect cell contacts and cell growth.