Cellular Heparan Sulfate Negatively Modulates Transforming Growth Factor-β1 (TGF-β1) Responsiveness in Epithelial Cells*

Cell-surface proteoglycans have been shown to modulate transforming growth factor (TGF)-β responsiveness in epithelial cells and other cell types. However, the proteoglycan (heparan sulfate or chondroitin sulfate) involved in modulation of TGF-β responsiveness and the mechanism by which it modulates TGF-β responsiveness remain unknown. Here we demonstrate that TGF-β1 induces transcriptional activation of plasminogen activator inhibitor-1 (PAI-1) and growth inhibition more potently in CHO cell mutants deficient in heparan sulfate (CHO-677 cells) than in wild-type CHO-K1 cells. 125I-TGF-β1 affinity labeling analysis of cell-surface TGF-β receptors reveals that CHO-K1 and CHO-677 cells exhibit low (<1) and high (>1) ratios of 125I-TGF-β1 binding to TβR-II and TβR-I, respectively. Receptor-bound 125I-TGF-β1 undergoes nystatin-inhibitable rapid degradation in CHO-K1 cells but not in CHO-677 cells. In Mv1Lu cells (which, like CHO-K1 cells, exhibit nystatin-inhibitable rapid degradation of receptor-bound 125I-TGF-β1), treatment with heparitinase or a heparan sulfate biosynthesis inhibitor results in a change from a low (<1) to a high (>1) ratio of 125I-TGF-β1 binding to TβR-II and TβR-I and enhanced TGF-β1-induced transcriptional activation of PAI-1. Sucrose density gradient analysis indicates that a significant fraction of TβR-I and TβR-II is localized in caveolae/lipid-raft fractions in CHO-K1 and Mv1Lu cells whereas the majority of the TGF-β receptors are localized in non-lipid-raft fractions in CHO-677 cells. These results suggest that heparan sulfate negatively modulates TGF-β1 responsiveness by decreasing the ratio of TGF-β1 binding to TβR-II and TβR-I, facilitating caveolae/lipid-raft-mediated endocytosis and rapid degradation of TGF-β1, thus diminishing non-lipid-raft-mediated endocytosis and signaling of TGF-β1 in these epithelial cells.

TGF-␤ 2 is a family of 25-kDa disulfide-linked dimeric proteins. It has three members in mammals (TGF-␤ 1 , TGF-␤ 2 , and TGF-␤ 3 ), which share ϳ70% sequence homology (1)(2)(3). TGF-␤ exhibits bifunctional growth regulation; it inhibits growth of most cell types, including epithelial cells, endothelial cells, and lymphocytes, and stimulates proliferation of mesenchymal cells such as fibroblasts. The growth regulatory activity of TGF-␤ has been implicated in many physiological and pathological processes (e.g. embryonic development, morphogenesis, carcinogenesis, autoimmune diseases, and Alzheimer disease) (1)(2)(3). One other prominent activity of TGF-␤ is transcriptional activation of extra-cellular matrix synthesis-related genes, which has been implicated in tissue fibrosis. TGF-␤ also exhibits chemotactic activity toward monocytes and neutrophils and is involved in the process of inflammation (1)(2)(3).
Among the TGF-␤ receptor types, T␤R-III (betaglycan) is the most abundant in many cell types. Due to its high density, T␤R-III is capable of modulating TGF-␤ binding to other TGF-␤ receptor types and of regulating TGF-␤-induced cellular responses (12,13). Ectopic expression of wild-type T␤R-III augments TGF-␤-induced cellular responses in myoblasts (12,13). On the other hand, in epithelial cells, ectopic expression of wild-type T␤R-III attenuates cellular responses induced by TGF-␤, whereas expression of T␤R-III mutants lacking proteoglycans promotes TGF-␤-induced cellular responses (13). The distinct effects of wild-type and mutant T␤R-III on TGF-␤ 1 -induced cellular responses in these two cell types appear to be due to the different sizes of proteoglycans in the T␤R-III molecules in these cells (13). The proteoglycan sizes of T␤R-III in myoblasts are smaller than those in epithelial cells. These observations suggest that T␤R-III with larger proteoglycans negatively modulates TGF-␤ 1 -induced cellular responses, whereas T␤R-III lacking proteoglycans or containing smaller proteoglycans enhances TGF-␤ 1 -induced cellular responses. However, which proteoglycan (heparan sulfate or chondroitin sulfate) in T␤R-III is involved in modulating TGF-␤-induced cellular responses remains unknown.
Proteoglycans are large unbranched carbohydrates comprised of repeating disaccharide units. They are components of the extracellular matrix and the external cell-surface membrane and are also present in secretory granules. They have six forms: heparan sulfate, chondroitin sulfate, heparin, dermatan sulfate, keratan sulfate, and hyaluronic acid. Proteoglycans have been shown to regulate cell growth, differentiation, and migration (18,19). Since TGF-␤ is also known to regulate these processes, it would be important to define the molecular mechanisms by which heparan sulfate or chondroitin sulfate (or both) modulate TGF-␤ 1 -induced cellular responses. In this communication, we demonstrate that deficiency or enzymic removal of heparan sulfate in epithelial cells increases the ratio of TGF-␤ 1 binding to T␤R-II and T␤R-I, attenuates degradation of TGF-␤ 1 , and augments TGF-␤ 1 -induced cellular responses.

I-TGF-␤ 1 Affinity Labeling of Cell-surface TGF-␤ Receptors in CHO and Mv1Lu Cells-CHO cells (CHO-K1 and CHO-677 cells) and
Mv1Lu cells were grown on 6-well cluster dishes in DMEM-F-12 medium and DMEM containing 10% fetal calf serum, respectively, as described (16,17,20). 125 I-TGF-␤ 1 affinity labeling of cell-surface TGF-␤ receptors was performed using the bifunctional cross-linking agent DSS as described previously (11,14,16,17). The affinity-labeled TGF-␤ receptors were then analyzed by 5 and 7.5% SDS-PAGE under reducing conditions and autoradiography. In some experiments, the affinity-labeled TGF-␤ receptors were immunoprecipitated by specific antibodies to T␤R-I, T␤R-II, and T␤R-III as described previously (17). The immunoprecipitates were then analyzed by 7.5% SDS-PAGE under reducing conditions and autoradiography.
[methyl-3 H]Thymidine Incorporation and Northern Blot Analyses-Cells were grown to near confluence on 24-well cluster dishes and then treated with several concentrations of TGF-␤ 1 at 37°C for 18 or 2 h for [methyl-3 H]thymidine incorporation into cellular DNA or for plasminogen activator inhibitor-1 (PAI-1) expression analysis, respectively. [methyl-3 H]Thymidine incorporation into cellular DNA and Northern blot analysis of PAI-1 and glyceraldehyde-3-phosphate dehydrogenase (G3PDH) in CHO-K1, CHO-677, and Mv1Lu cells were carried out as described previously (16,17). The relative level of PAI-1 mRNA was estimated based on the ratio of PAI-1 mRNA and G3PDH mRNA intensities on the autoradiogram. The relative intensities of both mRNAs on the autoradiogram were quantitated by a PhosphoImager.
Treatment of Mv1Lu Cells with Heparitinase or a Heparan Sulfate Biosynthesis Inhibitor-Mv1Lu cells were grown near confluence on 6-well cluster dishes, washed twice with serum-free DMEM, and treated with vehicle only, heparitinase (100 milliunits/ml) (11), or a heparan sulfate biosynthesis inhibitor p-nitrophenyl-␤-D-xylopyranoside (3 mM) in DMEM containing 0.1% fetal calf serum (21). After 3 h at 37°C (for heparitinase) and 72 h at 37°C (for the inhibitor) in serum-free DMEM containing 1% bovine serum albumin, 125 I-TGF-␤ 1 affinity labeling of cell-surface TGF-␤ receptors and Northern blot analysis of PAI-1 or G3PDH mRNA expression was performed as described above.
Western Blot Analysis of T␤R-I and T␤R-II in CHO Cells-Cell lysates of CHO-K1 and CHO-677 cells (ϳ50 g of protein) were sub-jected to 7.5% SDS-PAGE under reducing conditions and then electrotransferred to nitrocellulose membranes. After being incubated with 5% nonfat milk in Tris-buffered saline plus Tween (TBST) (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.05% Tween 20) for 1 h at room temperature, the membranes were further incubated with specific polyclonal antibodies to T␤R-I and T␤R-II in TBST/nonfat milk at room temperature for 1 h and washed three times with TBST for 10 min each. Bound antibodies were detected using peroxidase-conjugated anti-rabbit IgG and visualized using the ECL system (Santa Cruz Biotechnology).
Cellular Degradation of Receptor-bound 125 I-TGF-␤ 1 in CHO-K1, CHO-677, and Mv1Lu Cells-Cells grown on 24-well cluster dishes were pretreated with 25 g/ml nystatin at 37°C for 1 h in serum-free DMEM-F-12 medium and then incubated with 100 pM 125 I-TGF-␤ 1 in the presence and absence of 100-fold excess unlabeled TGF-␤ 1 . The presence of 100-fold-excess unlabeled TGF-␤ 1 was used to estimate nonspecific binding. After 2.5 h at 0°C, cells were washed and warmed to 37°C in DMEM-F-12 medium (for CHO cells) or DMEM (for Mv1Lu cells) containing 0.2% bovine serum albumin (22). After several time periods in the presence and absence of nystatin (25 g/ml), the conditioned medium was precipitated with 10% trichloroacetic acid at 4°C for 0.5 h. The trichloroacetic acid-soluble radioactive material, which contained the degradation products (e.g. amino acids or peptides) of cell-surface bound 125 I-TGF-␤ 1 after internalization and degradation and were released from cells, was counted. The trichloroacetic acidsoluble radioactive material derived from the specific binding of 125 I-TGF-␤ 1 was estimated as the percentage of the total specific binding of 125 I-TGF-␤ 1 determined before incubation at 37°C.
Sucrose Density Gradient Analysis-Sucrose density gradient analysis was performed at 4°C as described previously (23). Briefly, cells were grown to near confluence in 100-mm Petri dishes. After washing with ice-cold HEPES buffer (16,17), cells were scraped into 0.85 ml of 500 mM sodium carbonate, pH 11.0. The cell pellets were homogenized with a tight-fitting Dounce homogenizer followed by three 20-s bursts of ultrasonic disintegrator on ice. The homogenates were adjusted to 45% sucrose using 90% sucrose in MES-buffered saline, pH 6.5 (25 mM MES and 0.15 M NaCl) and placed at the bottom of an ultracentrifuge tube. Two solutions (1.7 ml each) of 35 and 5% sucrose were laid sequentially on the top of the 45% sucrose solution. After ultracentrifugation at 35,000 rpm with Beckman SW Ti55 rotor for 16 -20 h, 10 0.5-ml fractions were collected from the top of the tubes, and a portion of each fraction was analyzed by SDS-PAGE followed by Western blot analysis using specific antibodies to T␤R-I, T␤R-II, or caveolin-1.

Deficiency in Heparan Sulfate Augments TGF-␤ 1 Responsiveness in
CHO Cells-CHO wild-type and mutant cells, which are epithelial cells defective in heparan sulfate and chondroitin sulfate synthesis, have provided an excellent system for defining the roles of proteoglycans in cellular responses to stimuli in CHO cells (20,24). To define the roles of heparan sulfate and chondroitin sulfate in TGF-␤-induced cellular responses (TGF-␤ responsiveness) in CHO cells, we determined the effects of increasing concentrations of TGF-␤ 1 on cell growth (as determined by measurement of DNA synthesis and cell number) and PAI-1 expression in CHO-K1 and CHO-677 cells. CHO-K1 cells are wild-type CHO cells. CHO-677 cells are CHO mutant cells, which are defective in heparan sulfate synthesis but not chondroitin sulfate synthesis (20). As shown in Fig. 1A, TGF-␤ 1 inhibited DNA synthesis in CHO-677 cells more potently than in CHO-K1 cells. At 25 pM, TGF-␤ 1 inhibited DNA synthesis by ϳ60% in CHO-677 cells and by ϳ40% in CHO-K1 cells (Fig. 1A, panel a). TGF-␤ 1 also inhibited cell growth more potently in CHO-677 cells than in CHO-K1 cells. TGF-␤ 1 (100 pM) inhibited cell growth by ϳ75 and ϳ25% in CHO-677 and CHO-K1 cells, respectively ( Fig. 1A, panel b). CHO-677 cells also responded more strongly to TGF-␤ 1 -induced expression of PAI-1 when compared with wild-type CHO-K1 cells (Fig. 1B, panel a). At 20 pM, TGF-␤ 1 stimulated PAI-1 expression by ϳ2.2-fold in CHO-677 cells (Fig. 1B, panel b). In contrast, TGF-␤ 1 at the same concentration only slightly stimulated PAI-1 expression by 1.2-fold in CHO-K1 cells (Fig. 1B, panel b). These results suggest that the defect in heparan sulfate biosynthesis augments TGF-␤ 1 responsiveness in CHO cells.
Deficiency in Heparan Sulfate Increases the Ratio of TGF-␤ 1 Binding to T␤R-II and T␤R-I in CHO Cells-125 I-TGF-␤ 1 affinity labeling of cells ( 125 I-TGF-␤ 1 binding and cross-linking with the bifunctional reagent DSS) followed by SDS-PAGE has been commonly used for determining 125 I-TGF-␤ 1 binding to individual cell-surface TGF-␤ receptor types (11,14,17). To determine the effect of the deficiency in heparan sulfate on 125 I-TGF-␤ 1 binding to TGF-␤ receptors in CHO cells, CHO-K1 and CHO-677 cells were affinity-labeled with 125 I-TGF-␤ 1 (100 pM) and analyzed by 7.5% SDS-PAGE under reducing conditions and autoradiography. As shown in Fig. 2A, in CHO-677 cells, 125 I-TGF-␤ 1 affinity-labeled T␤R-III exhibited a distinct band with molecular mass of ϳ120 kDa (T␤R-III*), representing the 125 I-TGF-␤ 1 affinity-labeled core protein of T␤R-III ( Fig. 2A, panel a, lane 3). In CHO-K1 cells, 125 I-TGF-␤ 1 affinity-labeled T␤R-III (ϳ270 -300 kDa) migrated at the top of the separating gel on 7.5% SDS-PAGE ( Fig. 2A, panel a, lane  2). 125 I-TGF-␤ 1 affinity-labeled T␤R-I and T␤R-II migrated as distinct bands with molecular masses of 68 and 88 kDa, respectively, on SDS-PAGE in CHO cell lines including CHO-LRP-1 Ϫ cells are CHO mutant cells lacking T␤R-V (22) (Fig. 2A, lane 1). The lack of T␤R-V appeared to affect 125 I-TGF-␤ 1 binding to other TGF-␤ receptor types in CHO cells. Quantative analysis of 125 I-TGF-␤ 1 affinity labeling of TGF-␤ receptors in these CHO cells revealed that CHO-677 cells exhibited an increase of TGF-␤ 1 binding to T␤R-II and a decrease of 125 I-TGF-␤ 1 binding to T␤R-I when compared with CHO-K1 cells ( Fig. 2A, panel b). The ratio (ϳ1.4) of 125 I-TGF-␤ 1 affinity-labeled T␤R-II and T␤R-I appeared to be higher in CHO-677 cells than that (0.6) in wild-type cells (CHO-K1 cells) ( Fig. 2A, panel c). This result suggests that the deficiency in heparan sulfate alters binding of TGF-␤ 1 to T␤R-I and T␤R-II. To further define this, CHO-677 and CHO-K1 cells were incubated with increasing concentrations of 125 I-TGF-␤ 1 ; 125 I-TGF-␤ 1 affinity labeling of TGF-␤ receptors in these cells was then performed and analyzed by 5% (Fig. 2B) and 7.5% (Fig. 2C). SDS-PAGE under reducing conditions and autoradiography. As shown in Fig. 2, B and C, 125 I-TGF-␤ 1 bound to T␤R-I, T␤R-II, and T␤R-III or T␤R-III* in a concentration-dependent manner in CHO-K1 cells (lanes 1-5) and CHO-677 cells (lanes 6 -10). The half-maximum concentration (ϳ50 pM) of 125 I- TGF-␤ 1 for binding to the core protein (ϳ120 kDa) of T␤R-III (T␤R-III*) in CHO-677 cells was similar to that of 125 I-TGF-␤ 1 for binding to T␤R-III (ϳ270 -300 kDa) in CHO-K1 cells (Fig. 2B, panels a and b). CHO-K1 and CHO-677 cells also did not exhibit significant differences in the half-maximum concentrations (ϳ50 pM) of 125 I-TGF-␤ 1 for binding to T␤R-I and T␤R-II (Fig. 2C, panels b and c). This suggests that the affinities of TGF-␤ 1 for binding to T␤R-I and T␤R-II are similar in CHO-K1 and CHO-677 cells. However, the total binding of 125 I-  (Fig. 2B, panel a, and Fig. 2C, panel a, lanes  6 -10 versus lanes 1-5 and Fig. 2C, panel c versus panel b). At 100 pM 125 I-TGF-␤ 1 , the ratio of 125 I-TGF-␤ 1 affinity-labeled T␤R-II and T␤R-I in CHO-677 cells was estimated to be ϳ2 (Fig. 2C, panel d). In contrast, at the same concentration of 125 I-TGF-␤ 1 , the ratio of 125 I-TGF-␤ 1 affinity-labeled T␤R-II and T␤R-I was estimated to be 0.7 in CHO-K1 cells (Fig. 2C, panel d). CHO-677 and CHO-K1 cells exhibited similar ratios of T␤R-II and T␤R-I protein levels as determined by Western blot analysis (Fig. 3, lane 2 versus lane 1). The multiple bands of T␤R-II shown in Fig. 3 were also demonstrated previously (25). These results suggest that the deficiency in heparan sulfate alters the total binding of TGF-␤ 1 to T␤R-I and T␤R-II without significantly affecting the affinities of TGF-␤ 1 binding to T␤R-I and T␤R-II and their protein expression in these CHO cells. The similar half-maximum concentrations of 125 I-TGF-␤ 1 for binding to T␤R-III and T␤R-III* (Fig. 2B, panel b) in CHO-K1 and CHO-677 cells suggest that the binding affinities of 125 I-TGF-␤ 1 for T␤R-III and T␤R-III* are similar in these cells. Since T␤R-III is known to present the ligand TGF-␤ to T␤R-II and then T␤R-I, and since T␤R-III is also known to form hetero-oligomeric complexes with T␤R-I and T␤R-II in the presence of TGF-␤ (11,26,27), the lack of heparan sulfate in T␤R-III* (in CHO-677 cells) may affect the formation of the hetero-oligomeric complexes of T␤R-III/T␤R-II/T␤R-I, which contain different percentages of T␤R-I and T␤R-II (28). To test this possibility, we performed immunoprecipitation using specific antibodies to T␤R-I, T␤R-II, and T␤R-III following 125 I-TGF-␤ 1 affinity labeling of CHO-677 and CHO-K1 cells. As shown in Fig. 4, CHO-677 cells exhibited more 125 I-TGF-␤ 1 affinity-labeled T␤R-II in the TGF-␤ receptor complexes when compared with CHO-K1 cells (lanes 2, 4, and  6 versus lanes 1, 3, and 5).

Enzymic Removal of Heparan Sulfate Increases the Ratio of TGF-␤ 1 Binding to T␤R-II and T␤R-I and Augments TGF-␤-induced PAI-1 Expression in Mv1Lu
Cells-To see if absent or decreased heparan sulfate alters TGF-␤ 1 binding to T␤R-I and T␤R-II and resultant TGF-␤ responsiveness in other epithelial cell systems, mink lung epithelial cells (Mv1Lu cells) were treated with heparitinase for 3 h at 37°C or with a heparan sulfate biosynthesis inhibitor p-nitrophenyl-␤-D-xylopyrano-side for 72 h at 37°C and then affinity-labeled with 125 I-TGF-␤ 1 at 0°C. The 125 I-TGF-␤ 1 affinity-labeled TGF-␤ receptors were analyzed by 7.5% (Fig. 5A) and 5% (Fig. 5B) SDS-PAGE under reducing conditions and autoradiography. Heparitinase or p-nitrophenyl-␤-D-xylopyranoside treatment of the cells appeared to convert ϳ270 -350-kDa T␤R-III to fast-migrating forms (Fig. 5B, lanes 2 and 3 versus lane 1). Accordingly, heparitinase or p-nitrophenyl-␤-D-xylopyranoside treatment resulted in the increase of 125 I-TGF-␤ 1 binding to T␤R-II (Fig. 5, A-C) and T␤R-V (Fig. 5B, lanes 2 and 3 versus lane 1). The ratios of 125 I-TGF-␤ 1 binding to T␤R-II and T␤R-I in cells treated with vehicle only, heparitinase and p-nitrophenyl-␤-D-xylopyranoside were estimated to be ϳ0.6, ϳ1.6, and ϳ1.1, respectively (Fig. 5D). Heparitinase or p-nitrophenyl-␤-D-xylopyranoside treatment did not affect the protein levels of T␤R-I and T␤R-II in Mv1Lu cells as determined by Western blot analysis (data not shown). These results suggest that enzymic removal or synthesis inhibition of heparan sulfate increases the ratio of 125 I-TGF-␤ 1 affinity-labeled T␤R-II and T␤R-I in Mv1Lu cells as observed in CHO-677 cells (versus CHO-K1 cells). To see whether heparitinase treatment affects TGF-␤ 1 -induced cellular responses in Mv1Lu cells, the cells were treated with increasing concentrations of TGF-␤ 1 following treatment with heparitinase. The PAI-1 expression was then determined by Northern blot analysis. As shown in Fig. 6, A and B, heparitinase treatment augmented TGF-␤ 1 -induced expression of PAI-1 in Mv1Lu cells. The half-maximum concentration (ED 50 ) of TGF-␤ 1 to induce PAI-1 expression are similar (ϳ4 -5 pM) in cells treated with and without heparitinase. This is consistent with the observation that the deficiency of heparan sulfate causes increase and decrease of TGF-␤ 1 binding to T␤R-II and T␤R-I, respectively, without altering the affinities of TGF-␤ 1 binding to T␤R-II and T␤R-I in CHO-677 cells. These results support the hypothesis that the ratio of TGF-␤ 1 binding to T␤R-II and T␤R-I positively correlates with the magnitude of TGF-␤ 1induced cellular responses (28).

Cells Lacking Heparan Sulfate or Treated with Heparitinase Exhibit
Decreased Cellular Degradation of TGF-␤ 1 -TGF-␤ receptor-mediated signaling is known to occur in endosomes (29 -31). The TGF-␤ receptor complex with a higher ratio of 125 I-TGF-␤ 1 binding to T␤R-II and T␤R-I (e.g. in CHO-677 cells and Mv1Lu cells treated with heparitinase) is hypothesized to undergo clathrin-mediated endocytosis and transduces signaling in endosomes (28 -31). On the other hand, the TGF-␤ receptor complex with a low ratio (Ͻ1) of 125 I-TGF-␤ 1 binding to T␤R-II and T␤R-I (e.g. in CHO-K1 and Mv1Lu cells) is hypothesized to undergo rapid degradation via caveolae/lipid-raft-mediated endocytosis (28 -31). To test this hypothesis, we examined the cellular degradation of 125 I-TGF-␤ 1 bound to cell-surface receptors in CHO-677 cells, CHO-K1 and Mv1Lu cells. As shown in Fig. 7, 125 I-TGF-␤ 1 bound to cell-surface TGF-␤ receptors underwent rapid degradation, as determined by measurement of specific trichloroacetic acid-soluble radioactive material in the conditioned medium, in wild-type CHO cells (Fig.  7A) as compared with CHO-677 cells (Fig. 7B). The cellular degradation of 125 I-TGF-␤ 1 bound to cell-surface TGF-␤ receptors appeared to be inhibited by nystatin in CHO-K1 and Mv1Lu cells but not in CHO-677 cells (Fig. 7, A and C versus B). Nystatin is a cholesterol-binding compound and has been used to inhibit caveolae/lipidraft-mediated endocytosis/degradation (29 -32). This result suggests that caveolae/lipid-raft-mediated endocytosis is involved in rapid degradation of TGF-␤ receptor complexes in CHO-K1 and Mv1Lu cells, whereas clathrin-mediated endocytosis/signaling/degradation, which is not inhibited by nystatin, mainly occurs in CHO-677 cells. To test this, we performed sucrose density gradient analysis of T␤R-I and T␤R-II in CHO-K1, CHO-677, and Mv1Lu cells. As shown in Fig. 8, a significant fraction of T␤R-I and T␤R-II was localized in caveolae/lipid-raft fractions (fractions 4 and 5) in CHO-K1 and Mv1Lu cells. The T␤R-I localization in Mv1Lu cells was not  shown because the antibody to T␤R-I used in the experiment did not react well with mink T␤R-I antigen. By contrast, the majority of T␤R-I and T␤R-II were localized in non-lipid-raft fractions (fractions 7 and 8) in CHO-677 cells. These results are consistent with the notion that caveolae/lipid-raft-mediated endocytosis facilitates rapid degradation of TGF-␤ and attenuates TGF-␤ responsiveness (28 -31).

DISCUSSION
Members of the TGF-␤ family are among the most potent growth factors or cytokines known. They are active at subpicomolar concentrations. TGF-␤ is regulated at several levels including: 1) transcription (1)(2)(3)10), 2) activation of the latent form of TGF-␤ (33,34), 3) endocytosis of TGF-␤ receptor complexes (28 -31), and 4) post-TGF-␤ receptor signaling (35)(36)(37)(38). Among these, the molecular mechanisms underlying the regulation of endocytosis and signaling of TGF-␤ receptor complexes have been less studied until recently. TGF-␤-induced signaling has recently been shown to occur in endosomes (29 -31). Potassium chloride depletion, which inhibits clathrin-mediated endocytosis, has been shown to attenuate TGF-␤ receptor internalization and TGF-␤induced cellular responses in Mv1Lu cells (29 -31). TGF-␤ receptors at the cell surface undergo ligand-independent internalization and recycling via clathrin-coated and caveolin-positive vesicles (28 -31). In the absence of ligand, TGF-␤ receptors undergo constitutive internalization and recycling (30). Following ligand binding, ligand-activated TGF-␤ receptor complexes are internalized into endosomes (where they mediate signaling) and caveolin-positive vesicles (where they are subjected to rapid degradation). The biochemical properties of TGF-␤ receptor complexes designated for endosomes and caveosomes (caveolin-1-positive vesicles) are unknown (29 -31). Here we demonstrate that cells defective in biosynthesis of heparan sulfate or treated with a heparan sulfate biosynthesis inhibitor respond more strongly to TGF-␤ 1 -induced cellular responses than wild-type cells or untreated cells. These cells exhibit a higher ratio (Ͼ1) of TGF-␤ 1 binding to T␤R-II and T␤R-I and a decrease of degradation of TGF-␤ 1 bound to TGF-␤ receptors as compared with those observed in wild-type or untreated cells. These results support the hypothesis that the ratio of T␤R-II and T␤R-I in the TGF-␤ receptor complex provides a signal for determining TGF-␤ responsiveness (28).
A model is proposed to demonstrate how the formation of distinct TGF-␤ receptor complexes at the cell surface determine the cellular response to TGF-␤ 1 . In this model ( Fig. 9) modified from the models reported previously (28,29), T␤R-III presents TGF-␤ 1 to T␤R-II at the cell surface. T␤R-I is recruited to form T␤R-III⅐T␤R-II⅐T␤R-I ternary complexes that have two forms: I and II. Complex I, which contains more T␤R-II than T␤R-I (as determined by 125 I-TGF-␤ 1 affinity labeling), undergoes clathrin-mediated endocytosis and transduces signaling in endosomes. Complex II, which contains more T␤R-I than T␤R-II, undergoes caveolae/lipid-raft-mediated endocytosis and rapid degradation. The formation of these complexes can be regulated by the following: 1) the proteoglycan moiety of T␤R-III: larger proteoglycan moieties in T␤R-III facilitate the formation of Complex II (13), and no proteoglycan or a small proteoglycan moiety in T␤R-III facilitates the formation of Complex I (13); 2) altered expression of endoglin: endoglin is a TGF-␤-binding, proteoglycan-containing membrane protein and shares with T␤R-III a limited amino acid sequence homology (39), and the increased expression of endoglin facilitates the formation of Complex II, leading to the attenuation of TGF-␤-induced cellular responses (40); and 3) altered expression of T␤R-I or T␤R-II. Increased expression of T␤R-II or T␤R-I facilitates the formation of Complex I or II and enhances or attenuates TGF-␤ 1 -induced cellular responses, respectively (41)(42)(43). Blobe et al. (12) reported that stable transfection of L6 myoblasts with T␤R-III cDNA yields a higher ratio (Ͼ1) of TGF-␤ binding to T␤R-II and T␤R-I and enhances TGF-␤ responsiveness.  Cells were subjected to sucrose density gradient ultracentrifugation as described (24). The fractions were analyzed by Western blot analysis using antibodies to T␤R-I, T␤R-II, and caveolin-1. The arrows indicate the locations of T␤R-I, T␤R-II, and caveolin-1. Fractions 4 and 5 were caveolar/lipid-raft fractions whereas fractions 7 and 8 were non-lipid-raft (or clathrin) fractions.
Eickelberg et al. (13) demonstrated that, in LLC-PK (which are epithelial cells, unlike L6 myoblasts), overexpression of T␤R-III produces a lower ratio (Ͻ1) of TGF-␤ binding to T␤R-II and T␤R-I and attenuates TGF-␤ responsiveness. Since T␤R-III in LLC-PK1 cells exhibits higher molecular weight glycosaminoglycan chains than that in L6 cells, they suggested that the sizes of glycosaminoglycan chains in the T␤R-III molecule affect the ratio of TGF-␤ binding to T␤R-II and T␤R-I and thus affect TGF-␤ responsiveness. However, which proteoglycan (heparan sulfate or chondroitin sulfate) is involved in modulating TGF-␤ responsiveness has not previously been determined. Here we demonstrate that epithelial cells deficient in heparan sulfate or treated with heparitinase or a heparan sulfate biosynthesis inhibitor exhibit a higher ratio of TGF-␤ 1 binding to T␤R-II and T␤R-I, decreased degradation of TGF-␤ 1 , and enhanced TGF-␤ 1 responsiveness. This suggests that heparan sulfate negatively modulates TGF-␤ 1 responsiveness by facilitating formation of Complex II in epithelial cells.
In murine embryonic fibroblasts, the T␤R-III gene ablation does not significantly affect the ratio of TGF-␤ 1 binding to T␤R-II and T␤R-I in T␤R-III-null murine embryonic fibroblasts (44), which is Ͼ1.0. 4 This suggests that the formation of Complex I (with a high ratio of TGF-␤ 1 binding to T␤R-II and T␤R-I) is a default process and that the formation of Complex II (e.g. induced by heparan sulfate) is a regulatory event. TGF-␤ is known to stimulate the expression of proteoglycans, which may play a feedback regulatory role in TGF-␤ actions (45,46). Certain carcinoma cells may acquire resistance to TGF-␤ growth suppression by up-regulation of heparan sulfate expression (47)(48)(49). FIGURE 9. Heparan sulfate modulation of T␤R-I⅐T␤R-II complex (Complex I and Complex II) formation in epithelial cells. According to the dominance hypothesis (28), there are two major T␤R-I⅐T␤R-II complexes (Complex I and Complex II) present on the cell surface. Complex I contains more T␤R-II than T␤R-I, whereas T␤R-II is less than T␤R-I in Complex II. The ratio of T␤R-I and T␤R-II in the complexes can be determined by 125 I-TGF-␤ affinity labeling (as shown on the top of the figure). Complex I preferentially undergoes clathrin-mediated endocytosis of TGF-␤, resulting in promoting signaling (e.g. Smad protein phosphorylation) and cellular responsiveness. Complex II mainly undergoes caveolae/lipid-raft-mediated endocytosis, resulting in rapid degradation of TGF-␤ and less cellular responsiveness. In epithelial cells expressing heparan sulfate (HS) (e.g. CHO-K1 and Mv1Lu cells), the majority of the T␤R-I/T␤R-II complexes are present as Complex II, which exhibits nystatininhibitable endocytosis and degradation of TGF-␤. In epithelial cells lacking heparan sulfate (e.g. CHO-677 cells and Mv1Lu cells treated with heparitinase), the T␤R-I⅐T␤R-II complexes mostly exist as Complex I, which exhibits nystatin-uninhibitable endocytosis, signaling, and degradation of TGF-␤.