Betaglycan Inhibits TGF-β Signaling by Preventing Type I-Type II Receptor Complex Formation

Transforming growth factor (TGF)-β is a multifunctional growth factor with important roles in development, cell proliferation, and matrix deposition. It signals through the sequential activation of two serine/threonine kinase receptors, the type I and type II receptors. A third cell surface receptor, betaglycan, serves as a co-receptor for TGF-β in some cell types, enhancing TGF-β-mediated signaling. We have examined the function of betaglycan in renal epithelial LLC-PK1 cells that lack endogenous betaglycan. We demonstrate that the expression of betaglycan in LLC-PK1 cells results in inhibition of TGF-β signaling as measured by reporter gene expression, thymidine incorporation, collagen production, and phosphorylation of the downstream signaling effectors Smad2 and Smad3. In comparison, the expression of betaglycan in L6 myoblasts enhances TGF-β signaling, which is consistent with the published literature. The effects of betaglycan in LLC-PK1 cells are not mediated by ligand sequestration or increased production of a soluble form of the receptor, which has been reported to serve as a ligand antagonist. We demonstrate instead that in LLC-PK1 cells, unlike L6 cells, expression of betaglycan prevents association between the type I and type II TGF-β receptors, which is required for signaling. This is a function of the glycosaminoglycan modifications of betaglycan. Betaglycan in LLC-PK1 cells exhibits higher molecular weight glycosaminoglycan (GAG) chains than in L6 cells, and a GAG− betaglycan mutant does not inhibit TGF-β signaling or type I/type II receptor association in LLC-PK1cells. Our data indicate that betaglycan can function as a potent inhibitor of TGF-β signaling by a novel mechanism and provide support for an essential but complex role for proteoglycan co-receptors in growth factor signaling.

TGF-␤ 1 is an essential regulator of development, cell proliferation, and matrix deposition (1)(2)(3). Biological effects of TGF-␤ are achieved through the sequential activation of two high affinity serine/threonine kinase receptors, the type II (T␤RII) and the type I (T␤RI) TGF-␤ receptors. TGF-␤ signaling is initiated by the binding of TGF-␤ to T␤RII followed by the formation or stabilization of T␤RI-T␤RII complexes, transphosphorylation of T␤RI by T␤RII, and phosphorylation of the receptor-associated cytoplasmic effector molecules Smad2 and Smad3 by T␤RI (1,3). A third TGF-␤ transmembrane receptor, the type III receptor or betaglycan, is a widely expressed heparan and chondroitin sulfate proteoglycan that binds TGF-␤ with high affinity through its protein core but has no apparent intrinsic signaling activity (4 -7).
Betaglycan is believed to be a co-receptor for TGF-␤. In several cell lines, including rat L6 myoblasts lacking endogenous betaglycan, it significantly increases the affinity of T␤RI and T␤RII for TGF-␤, thereby enhancing the response to TGF-␤ (4,5). In addition, betaglycan eliminates differences in efficacy among the three TGF-␤ isoforms. TGF-␤2, which binds to T␤RI and T␤RII with low affinity and has minimal growth inhibitory effects in the absence of betaglycan, behaves similarly to TGF-␤1 and TGF-␤3 in the presence of betaglycan (8 -10). In normal development, betaglycan is required for the mesenchymal transformation of AV cushion myocytes, probably because it enhances signaling by local TGF-␤2 (11). A recent report suggests that the mechanism of this enhancement of TGF-␤2 signaling is an interaction between the cytoplasmic domain of betaglycan and the autophosphorylated T␤RII kinase domain (12).
There is increasing evidence, however, that the function of betaglycan is more complex. Although normally membraneanchored, betaglycan can undergo proteolytic processing in vivo, resulting in the secretion of a soluble ectodomain. This ectodomain can both enhance and inhibit signaling, depending on the concentration of TGF-␤ present (13), although in most cases it is believed to bind and sequester TGF-␤, functioning in effect as a receptor antagonist (5, 8, 14 -16). Full-length, membrane-anchored betaglycan has only been reported to enhance signaling by TGF-␤. Recently, however, betaglycan has been shown to play an inhibitory role in the function of the TGF-␤ superfamily member activin by binding to the activin antagonist inhibin, enabling it to recruit the activin type II receptor and prevent activin binding (17).
Interestingly, many cell types and tissues express high levels of betaglycan, yet are poorly responsive to TGF-␤ (4,18,19). 2 To better understand the function of betaglycan, we studied its biological effects in renal epithelial LLC-PK 1 cells. We demonstrate that these cells have very low levels of endogenous betaglycan and that expression of betaglycan in these cells inhibits TGF-␤ signaling. This inhibition is not caused by enhanced secretion of soluble betaglycan and sequestration of ligand. Rather, betaglycan, as a function of its glycosaminoglycan modifications, prevents the physical association of T␤RI and T␤RII and thereby prevents the initiation of downstream signaling.

EXPERIMENTAL PROCEDURES
Constructs-An N-terminally HA-tagged rat betaglycan receptor (20) was inserted into vector pcDNA3 (Invitrogen) for subsequent use and is designated T␤RIIIwt. A cytoplasmic deletion mutant lacking residues 812-853 (the entire cytoplasmic tail; T␤RIII⌬cyto) was constructed by PCR mutagenesis, as was a mutant with serine to alanine mutations at positions 535 and 546 (T␤RIII⌬gag), eliminating the two glycosaminoglycan attachment sites (16). Sequences of all mutants were confirmed by dideoxy sequencing (Yale Keck Center). Primer sequences for PCR reactions are available upon request.
Western Blot Analysis-Cells were seeded onto 60-mm cell culture dishes and allowed to reach 80% confluence. Cells were then stimulated with TGF-␤1 or -␤2 at the indicated concentrations, and nuclear and cytosolic extracts were prepared. In brief, cell pellets were resuspended in 200 l of low salt buffer (20 mM Hepes, pH 7.9, 10 mM KCl, 0.1 mM NaVO 4 , 1 mM EDTA, 1 mM EGTA, 0.2% Nonidet P-40, 10% glycerol, and a set of proteinase inhibitors, Complete TM ). After 10 min of incubation on ice, the samples were centrifuged at 8,000 ϫ g for 1 min (4°C), and the supernatants (cytosolic extracts) were immediately frozen in a dry ice/ethanol bath. Pelleted nuclei were resuspended in 120 l of high salt 2 O. Eickelberg and M. Kashgarian, unpublished observations. buffer (20 mM Hepes, pH 7.9, 420 mM NaCl, 10 mM KCl, 0.1 mM NaVO 4 , 1 mM EDTA, 1 mM EGTA, 20% glycerol, and supplemented with Complete TM ), and nuclear proteins were extracted by shaking on ice for 30 min. Samples were centrifuged at 13,000 ϫ g for 10 min (4°C), and the supernatants were taken as nuclear extracts. Protein concentrations of all samples were determined by standard Bradford assay. Aliquots of cell extracts were applied to 7.5% SDS-PAGE gels and run at 25 mA of constant current for 3 h at room tempterature. After electrophoresis, proteins were blotted on nitrocellulose transfer membranes overnight with 1 mA/cm 2 at 4°C. Membranes were then probed with antibodies against the indicated Smads at room temperature for 1 h. Specific bands were visualized using the ECL system (Pierce).
Thymidine Incorporation-LLC-PK 1 cells were seeded into 24-well plates in the presence or absence of TGF-␤1 (40 pM) for 16 h, as indicated. Cells were pulsed with 0.5 Ci/ml [ 3 H]thymidine for 4 h, and incorporation of radioactive label into newly synthesized DNA was analyzed by trichloracetic acid precipitation and liquid scintillation counting.
Collagen Deposition-LLC-PK 1 or L6 cells were cultured to subconfluence in 12-well plates, stimulated with TGF-␤1 (40 pM), and pulsed with [ 3 H]proline (0.5 Ci/ml) for 24 h. Supernatants were then removed and cells lysed and extracellular matrix proteins were fixed in 70% ethanol. Collagen deposition was determined by differential collagenase digestion and liquid scintillation counting, as described (25).
Affinity Labeling and Immunoprecipitation-Iodination of TGF-␤1 (R & D Systems) and affinity labeling of L6, Mv1Lu, and LLC-PK 1 cell lines were performed as described (21). For labeling of soluble betaglycan, cells were incubated in serum-free media for 24 h. Media were collected and centrifuged to remove detached cells and were concentrated using a Centricon 10 microconcentrator (Amicon). After equalization of protein concentration, detergent lysates of labeled cells or culture supernatants were immunoprecipitated with antibodies recognizing the HA tag of betaglycan, the type II TGF-␤ receptor (␣-IIC), or the type I receptor (21). For double immunoprecipitation experiments, affinity-labeled LLC-PK 1 expressing betaglycan were immunoprecipitated with antibodies against the HA epitope and the ectodomain of betaglycan (R & D Systems) and were eluted by boiling in 0.5% SDS. A portion of the eluant was then diluted in lysis buffer and reprecipitated with the same antibodies. Immunoprecipitates were separated by SDS-PAGE, and specific bands were visualized by autoradiography.

RESULTS
Betaglycan Inhibits TGF-␤ Signaling-We studied TGF-␤ signaling in LLC-PK 1 cells, which are highly responsive to TGF-␤ but have very low expression of betaglycan as determined by TGF-␤ affinity labeling (Fig. 1a), Western blotting, immunohistochemistry, and Northern analysis (Ref. 4, and data not shown). Cells were stably transfected with full-length wild-type rat betaglycan (T␤RIIIwt), with a mutant lacking the cytoplasmic domain (T␤RIII⌬cyto), or with expression vector alone (pcDNA). TGF-␤ responsiveness in these cell lines was then analyzed using the three luciferase reporter constructs p3TP-lux, pAP-1, and pSBE4. As demonstrated in Fig. 1 significantly increased reporter gene expression (14.4-fold, 5.5fold, and 10.4-fold for p3TP-lux, pAP-1, and pSBE4, respectively, p Ͼ 0.001 for all) (Fig. 1, b-d). TGF-␤1-mediated reporter gene expression was significantly reduced, however, in cells transfected with either T␤RIIIwt or T␤RIII⌬cyto (Fig. 1, b-d, p Ͼ 0.001). Similar inhibitions for all three reporter gene constructs were observed in LLC-PK 1 cells transiently transfected with betaglycan expression constructs and in cells expressing non-epitope-tagged betaglycan constructs (data not shown).
Betaglycan expression also inhibited TGF-␤-mediated biological effects. TGF-␤-mediated inhibition of DNA synthesis was reduced in LLC-PK 1 cells stably expressing either fulllength or truncated betaglycan compared with those expressing vector alone ( Fig. 2A). Similarly, TGF-␤1-mediated collagen deposition was suppressed when cells expressed either T␤RIIIwt or T␤RIII⌬cyto (Fig. 2B). In contrast, L6 myoblasts, which also lack endogenous betaglycan (4, 26), demonstrate enhanced TGF-␤1-induced collagen deposition in the presence of either betaglycan construct (Fig. 2C). This is consistent with published data using L6 and other mesenchymal cell lines that demonstrate increased TGF-␤ signaling associated with betaglycan expression (4,8). Taken together, our results thus far demonstrate that betaglycan is a potent inhibitor of TGF-␤ signaling in LLC-PK 1 cells and that this response is independent of the cytoplasmic domain.
Dose-response relationships demonstrate that inhibition of TGF-␤ activity results from an overall loss of TGF-␤ responsiveness in cells expressing betaglycan and not from decreased ligand affinity (Fig. 3). As assayed by p3TP-lux reporter gene activity, the EC 50 for TGF-␤1 was ϳ4 pM in LLC-PK 1 cells expressing vector or either betaglycan construct. If betaglycan reduced affinity or sequestered ligand, it would be expected to increase the EC 50 rather than reduce efficacy. Similar EC 50 and dose-response relationships were observed with TGF-␤2 treatment (data not shown).
Betaglycan-mediated inhibition of TGF-␤ signaling appears to be a direct effect. In the TGF-␤ signaling pathway, phosphorylation of the cytoplasmic signaling molecules Smad2 and Smad3 by T␤RI is the first step in signal transduction. LLC-PK 1 cells expressing betaglycan demonstrate a clear reduction in Smad2/3 phosphorylation compared with cells expressing vector alone (Fig. 4). Of note, total levels of Smad2 and Smad3 are not altered by the expression of betaglycan.
Betaglycan Inhibition of TGF-␤ Signaling Is Not Caused by Increased Generation of Soluble Betaglycan-The full-length membrane-anchored betaglycan can undergo proteolysis with secretion of a soluble ectodomain. In some systems, soluble betaglycan can bind and sequester active TGF-␤, thereby reducing the cellular effect. Several lines of evidence demonstrate that this is not the mechanism for betaglycan inhibition of  1, 4, 11, and 14), T␤RIII⌬cyto (lanes 2 and 12), or T␤RIIIwt (lanes 3, 5, 13, and 15) were cultured to subconfluence. Cells were then placed in fresh media and incubated for another 24 h. Conditioned culture supernatants were then collected and concentrated. The concentrated medium and the remaining adherent cells were affinity-labeled with 125 I-TGF-␤1 (200 pM). Labeled medium and total cell lysates were immunoprecipitated with an antibody against the HA epitope. Immunoprecipitants from cell lysates (left panel) and culture supernatants (right panel) were then subjected to SDS-PAGE and autoradiography. Faster migrating bands seen in LLC-PK 1 cell lysate and supernatant lanes represent betaglycan species (see Fig. 7b).
TGF-␤ signaling in LLC-PK 1 cells. In the dose-response curve shown in Fig. 3, a change in EC 50 rather than a change in efficacy would be expected if soluble rather than membraneanchored betaglycan mediated inhibition of TGF-␤ signaling. Additionally, we analyzed cell lysates and culture supernatants from both L6 and LLC-PK 1 cells stably expressing betaglycan. As shown in Fig. 5, L6 and LLC-PK 1 cells expressing full-length betaglycan produce comparable amounts of the soluble ectodomain, suggesting that the production of the soluble receptor is unlikely to account for the inhibition of TGF-␤ signaling associated with betaglycan expression in LLC-PK 1 cells. Consistent with this observation, neither supernatant collected from betaglycan-expressing LLC-PK 1 cells (data not shown) nor recombinant-soluble betaglycan ectodomain inhibited TGF-␤ signaling in untransfected LLC-PK 1 cells that do not express betaglycan (Fig. 6A). At TGF-␤1 concentrations ranging from 4 to 400 pM, no inhibition of TGF-␤1 signaling was observed by the addition of recombinant soluble betaglycan. In a control experiment to demonstrate the effects of ligand sequestration, addition of neutralizing anti-TGF-␤ antibody led to a significant and dose-dependent inhibition of p3TP-Lux activity in LLC-PK 1 cells (Fig. 6B).
Betaglycan Inhibits Association of T␤RI-T␤RII-It has recently been shown that betaglycan can mediate inhibition of the superfamily member activin (17). This occurs through the formation of a complex between betaglycan, the activin type II receptor, and the activin antagonist inhibin. This complex prevents both binding of activin and recruitment of the activin type I receptor into an active signaling complex (17). To determine whether betaglycan inhibits TGF-␤ signaling in LLC-PK 1 cells by a similar mechanism, L6 and LLC-PK 1 cells were affinity-labeled with 125 I-TGF-␤1, and cell lysates were immunoprecipitated with antibodies against the three TGF-␤ receptors (Fig. 7A). In L6 cells, immunoprecipitation with antibodies against T␤RI or T␤RII demonstrates co-immunoprecipitation of the two receptors, independent of betaglycan expression (compare lanes 1 and 2 with 4, 5, 7, and 8). In LLC-PK 1 cells expressing vector alone, immunoprecipitations demonstrate the existence of similar T␤RI-T␤RII complexes (lanes 10 and 11). In LLC-PK 1 cells expressing full-length or truncated betaglycan, however, T␤RI and T␤RII do not co-immunoprecipitate, regardless of whether anti-type I or anti-type II receptor antibodies are used. No labeled T␤RI could be detected by such co-immunoprecipitation experiments (lanes 13, 14, 16, and 17), although Western blot analysis revealed no change in the total levels of T␤RI and T␤RII with betaglycan expression (data not shown). Double immunoprecipitations with antibody against betaglycan indicate that the faster migrating species seen coimmunoprecipitating with betaglycan are primarily betaglycan species, not T␤RI or T␤RII (Fig. 7B). Additionally, these species have slightly different molecular weights than T␤RI and T␤RII. Immunoprecipitations with antibody against T␤RII demonstrate co-immunoprecipitation of T␤RII and betaglycan (Fig. 7A, lanes 14 and 17, and evident on shorter exposures), suggesting the preferential formation of T␤RII-betaglycan complexes compared with T␤RII-T␤RI complexes. This lack of T␤RI-T␤RII complexes in betaglycan-expressing LLC-PK 1 cells provides a mechanism to explain the observed decreases in phosphorylation of Smad2 and Smad3 (Fig. 4) and inhibition of TGF-␤ signaling (Figs. 1-3).
Glycosaminoglycan Chains Regulate the Function of Betaglycan in LLC-PK 1 Cells-We observed significant differences in the molecular weights of betaglycan expressed by LLC-PK 1 and L6 cells (Figs. 5 and 7). Both cell lines were transfected with the same cDNA, suggesting that differences in post-translational modification, specifically glycosaminoglycan chain ad-dition, might account for variations in molecular weight. To test whether these differences might explain the different functions of betaglycan in L6 and LLC-PK 1 , we constructed a mutant betaglycan (T␤RIII⌬gag) lacking both GAG attachment sites (although N-linked glycosylation sites remained). This mutant is expressed on the cell surface and binds TGF-␤, as has been reported for a similar GAG mutant of betaglycan ( Fig.  8A and Ref. 16). In rat L6 cells, T␤RIII⌬gag functions similarly to T␤RIIIwt, as assayed by luciferase reporter gene activity (data not shown). In LLC-PK 1 , however, T␤RIII⌬gag lacks the inhibitory effect of T␤RIIIwt on TGF-␤ signaling (Fig. 8B). These data demonstrate that the high molecular weight GAG modifications of betaglycan observed in LLC-PK 1 but not L6 cells render it an inhibitory co-receptor in TGF-␤ signaling. DISCUSSION Our results show that the expression of betaglycan inhibits TGF-␤ signaling in LLC-PK 1 cells although it enhances TGF-␤ signaling in L6 cells. This is the first demonstration that fulllength betaglycan can inhibit TGF-␤ signaling depending on the cell type, and it indicates that the function of this receptor is more complicated than previously assumed. The cleaved soluble ectodomain of betaglycan has been reported to inhibit TGF-␤ signaling by sequestering ligand, but this is clearly not the underlying mechanism for signaling inhibition by the fulllength receptor because of the following: (i) the levels of soluble receptor are similar in both L6 and LLC-PK 1 cells; (ii) the direct addition of soluble receptor to wild-type LLC-PK 1 cells does not inhibit TGF-␤ signaling; and (iii) the kinetics of inhibition are not consistent with ligand sequestration by the soluble receptor. Furthermore, we could not detect direct interactions between T␤RII and either a recombinant soluble betaglycan or the soluble receptor produced by LLC-PK 1 (data not shown). Rather, our data demonstrate that membraneanchored betaglycan prevents the association between T␤RI and T␤RII. Interestingly, this mechanism is similar to the one reported for inhibin antagonism of activin signaling. In that case, inhibin binds to betaglycan and the activin type II receptor, thereby preventing recruitment of the activin type I receptor and formation of a functional signaling complex (17).
A recent report suggests that betaglycan enhancement of TGF-␤2 signaling in L6 cells is a function of interactions between the cytoplasmic domain of betaglycan and the autophosphorylated T␤RII kinase domain (12). We have consistently observed the inhibition of TGF-␤ signaling in LLC-PK 1 cells with the expression of both the full-length betaglycan and a   Fig. 7, demonstrating cell surface expression and ligand binding ability of the T␤RIII⌬gag construct. B, LLC-PK 1 cells were transiently transfected with vector control (pcDNA), T␤RIIIwt, T␤RIII⌬cyto, or T␤RIII⌬gag. Expression of T␤RIIIwt or T␤RIII⌬cyto significantly reduced reporter gene expression directed by p3TP-lux in response to TGF-␤1 (100 pM), whereas T␤RIII⌬gag had no effect on baseline or TGF-␤1-induced reporter gene activity (*, p Ͻ 0.001 compared with pcDNA; **, p Ͻ 0.001 compared with T␤RIIIwt and T␤RIII⌬cyto, respectively). Results presented were obtained in quadruplicate and are representative of five independent experiments. Similar results were observed in stably transfected LLC-PK 1 cells and using the pSBE4 reporter construct. truncated receptor lacking its cytoplasmic domain. Betaglycanmediated inhibition of TGF-␤ signaling was also observed with both TGF-␤1 and TGF-␤2, regardless of the presence of the cytoplasmic domain, suggesting that the inhibition described here is a function of the ecto-and transmembrane domains of betaglycan.
Endoglin, a TGF-␤-binding protein with 63% sequence similarity to betaglycan in its transmembrane and cytoplasmic domains (27), inhibits the response to TGF-␤ in many cell lines, including L6E9 rat myocytes, a cell line similar to the L6 used here (28). Domain-swapping studies in these cells localized the inhibitory effects of endoglin and the stimulatory effects of betaglycan to either the transmembrane or extracellular domain (29), consistent with our results. Endoglin typically coimmunoprecipitates with both T␤RI and T␤RII, and binds TGF-␤ only in the presence of T␤RII (28,29), but the effect of endoglin on T␤RI-T␤RII complex formation is unclear. Endogenous endoglin and betaglycan co-immunoprecipitate in human microvascular endothelial cells (30), raising the possibility of additional mechanisms of regulating TGF-␤ activity (11).
We have clearly demonstrated that betaglycan GAG chains account for its inhibitory activity in LLC-PK 1 cells because mutations to remove the GAG attachment sites changed the effect of betaglycan on TGF-␤ signaling. Correspondingly, T␤RIII⌬gag, in contrast to the wild-type receptor, had no effect on T␤RI/T␤RII association in co-immunoprecipitation assays in LLC-PK 1 cells. Variable modification of heparan sulfate proteoglycans has been shown to influence the biological activity of FGF-2 (31), although betaglycan is to our knowledge the first receptor shown to have opposing actions depending on its GAG chains. Betaglycan has two GAG sites that have been shown to be specific for either heparan sulfate or chondroitin sulfate GAG chains (16). Although we have not determined whether either one or both sites are responsible for the effects we see, it may well be the heparan sulfate site (Ser-535) because heparan sulfate chains on betaglycan have been shown in osteoblasts to be variable (6). We are currently investigating the relative contributions of heparan and chondroitin sulfate chains to betaglycan function. We are also investigating the structural basis for the GAG effect and the possibility that GAG chain modification represents a mechanism whereby other growth factors influence TGF-␤ signaling.
The actual nature of the interactions between betaglycan and T␤RI and T␤RII is not well understood. It has been clearly established that betaglycan and the type II receptor co-immunoprecipitate (9, 32), although immunofluorescence co-patching experiments with live cells suggest that the percentage of each receptor in a joint complex is very small. It should be noted, however, that the betaglycan-associated GAG chains in the COS7 cells used in the studies of Henis et al. (20) are of low molecular weight, and it is possible that co-patching studies performed with a cell line demonstrating higher molecular weight betaglycan GAG chains would produce different results. The degree of betaglycan/T␤RII coimmunoprecipitation is low even in LLC-PK 1 cells, where T␤RI/T␤RII interactions are disrupted. The details of the interaction between betaglycan and the other TGF-␤ receptors are poorly understood in any cell line, regardless of whether betaglycan enhances or inhibits TGF-␤ signaling.
Our data have significant implications for the understanding of betaglycan function and the regulation of TGF-␤-induced responses in different cell types. In particular, our results suggest an explanation for observed differences in TGF-␤ responsiveness between epithelial and mesenchymal cells. Mesenchyme-derived cells including mesangial cells, which are generally poorly responsive to TGF-␤, express high levels of betaglycan. 3 There are examples in the literature that correlate increased TGF-␤ responsiveness with decreased betaglycan expression, or vice versa (33). These are physiologically relevant models for further studies investigating the function of betaglycan, as are cell types in which betaglycan expression changes dramatically during differentiation (34). Cell lines and tissue types in which betaglycan has an inhibitory role need to be identified, as the loss of betaglycan in these cells predicts a significant enhancement of the cellular response to TGF-␤.