Altering the Proteoglycan State of Transforming Growth Factor β Type III Receptor (TβRIII)/Betaglycan Modulates Canonical Wnt/β-Catenin Signaling*

Hyperactive Wnt/β-catenin signaling is linked to cancer progression and developmental abnormalities, making identification of mechanisms controlling Wnt/β-catenin signaling vital. Transforming growth factor β type III receptor (TβRIII/betaglycan) is a transmembrane proteoglycan co-receptor that exists with or without heparan and/or chondroitin sulfate glycosaminoglycan (GAG) modifications in cells and has established roles in development and cancer. Our studies here demonstrate that TβRIII, independent of its TGFβ co-receptor function, regulates canonical Wnt3a signaling by controlling Wnt3a availability through its sulfated GAG chains. Our findings revealed, for the first time, opposing functions for the different GAG modifications on TβRIII suggesting that Wnt interactions with the TβRIII heparan sulfate chains result in inhibition of Wnt signaling, likely via Wnt sequestration, whereas the chondroitin sulfate GAG chains on TβRIII promote Wnt3a signaling. These studies identify a novel, dual role for TβRIII/betaglycan and define a key requirement for the balance between chondroitin sulfate and heparan sulfate chains in dictating ligand responses with implications for both development and cancer.


Hyperactive Wnt/␤-catenin signaling is linked to cancer progression and developmental abnormalities, making identification of mechanisms controlling Wnt/␤-catenin signaling vital. Transforming growth factor ␤ type III receptor (T␤RIII/betaglycan) is a transmembrane proteoglycan co-receptor that exists with or without heparan and/or chondroitin sulfate glycosaminoglycan (GAG) modifications in cells and has established roles in development and cancer.
Our studies here demonstrate that T␤RIII, independent of its TGF␤ co-receptor function, regulates canonical Wnt3a signaling by controlling Wnt3a availability through its sulfated GAG chains. Our findings revealed, for the first time, opposing functions for the different GAG modifications on T␤RIII suggesting that Wnt interactions with the T␤RIII heparan sulfate chains result in inhibition of Wnt signaling, likely via Wnt sequestration, whereas the chondroitin sulfate GAG chains on T␤RIII promote Wnt3a signaling. These studies identify a novel, dual role for T␤RIII/betaglycan and define a key requirement for the balance between chondroitin sulfate and heparan sulfate chains in dictating ligand responses with implications for both development and cancer.
Type III TGF-␤ receptor (T␤RIII)/betaglycan is a transmembrane proteoglycan with loss resulting in embryonic lethality in mice (15). Beyond its roles in regulating TGF-␤ signaling, T␤RIII also controls several other pathways to inhibit cell migration, invasion, cell growth, and angiogenesis in both in vitro and in vivo cancer models (16 -22) and regulating differentiation through FGF2 signaling (23). Mechanistically, T␤RIII regulates these pathways either by altering the actin cytoskeleton, via T␤RIII/␤-arrestin2 cytoplasmic interactions (24), or by GAG chain interactions with FGF2 (23). Overall, T␤RIII also acts as a tumor suppressor in prostate (19), lung (25), pancreatic (18), and breast cancer (16,21,26,27) but has been shown to promote metastasis in specific mesenchymal stem-like breast cancers (28), indicating its complex roles for T␤RIII in cancer.
Although the T␤RIII core can bind TGF-␤ superfamily members with high affinity (22,29,30) the extracellular domain also contains two sites of heparan and chondroitin sulfate GAG chain modifications, resulting in T␤RIII existing in multiple forms in vivo (30 -32). Given that Wnt glycoproteins have a high affinity for both heparan and chondroitin GAG chains on proteoglycans (13,33), we initiated studies to determine the possible role of T␤RIII in canonical Wnt3a signaling.
We found, using both cancer and normal epithelial cells and a combination of loss and gain of function approaches, that T␤RIII suppresses Wnt3a signaling both at the signal reception level and through inhibition of ␤-catenin transcriptional activity by binding Wnt3a via its sulfated GAG chains. In contrast, T␤RIII chondroitin sulfate chains promote Wnt3a signaling, suggesting that the composition of the GAG chains may significantly alter the cellular response to T␤RIII and thereby Wnt signaling. Consistent with a lack of a role for T␤RIII GAG chains in T␤RIII functions as a TGF-␤ co-receptor (30), T␤RIII suppression of canonical Wnt3a signaling is independent of TGF-␤ signaling and independent of the T␤RIII cytoplasmic domain interactions described previously (24,34,35). These results demonstrate an intricate mode of Wnt3a signaling regulation by T␤RIII mediated largely by its heparan and chondroitin chains, laying the foundation to advance the current understanding of the various roles that proteoglycans, with different GAG chains, have in maintaining cellular homeostasis, specifically through control of Wnt availability and signaling.

T␤RIII Suppresses Wnt/␤-Catenin Activity at the Level of Signal Reception-To investigate the role of T␤RIII in signaling by
Wnt glycoproteins, which have high affinities for both HSPG and CSPG (33,36,37), we expressed T␤RIII in the ovarian cancer cell line OVCA429 that we and others have established as expressing low levels of T␤RIII ( Fig. 1A and Refs. 17 and 24). Conversely, we reduced the expression of T␤RIII by shRNAmediated knockdown in the ovarian cancer cell line SKOV3, which expresses higher levels of T␤RIII (Fig. 1A, iii and Ref. 38). We examined whether T␤RIII can affect canonical Wnt signaling as determined by phosphorylation of co-receptor LRP6, one of the first steps initiated by the binding of Wnt to their signaling co-receptors (39). We found that although Wnt3a robustly phosphorylated LRP6 at serine 1490 (40) in OVCA429 cells (low T␤RIII levels), transiently increasing T␤RIII expression in OVCA429 cells suppressed Wnt-induced LRP6 phosphorylation in a T␤RIII dose-dependent manner (Fig. 1B). Total LRP6 levels remained stable in T␤RIII-expressing OVCA429 cells FIGURE 1. T␤RIII suppresses Wnt/␤-catenin activity at the level of signal reception. A, T␤RIII mRNA expression by qRT-PCR analysis to detect endogenous T␤RIII levels in OVCA429 and SKOV3 cells (i), overexpression of T␤RIII in OVCA429 cells as indicated (ii), and expression in SKOV3 cells of the indicated shRNA and rescue conditions generated as described under "Experimental Procedures" (iii). Ct values are normalized in graph i to endogenous T␤RIII levels in SKOV3 cells (lane 2), in graph ii to GFP (lane 1), and in graph iii to Scrambled (Scr) T␤RIII levels (lane 1). Quantitations represent the average of two independent biological trials, each conducted in triplicate. B, OVCA429 cells transiently expressing increasing doses of T␤RIII (5 and 10 multiplicities of infection (MOI) of T␤RIIIexpressing adenoviral constructs) or control (GFP) were stimulated with 50 ng ml Ϫ1 Wnt3a for 1 h followed by immunoblotting of lysates for phospho-LRP6 (Ser-1490) (pLRP6), LRP6, T␤RIII, and ␤-actin. C, SKOV3 cells transiently expressing shRNA to T␤RIII (shT␤RIII-1) or Scrambled control ("Experimental Procedures") and transiently transfected with rat T␤RIII (rT␤RIII) or control vector (pcDNA 3.1) for 24 h for rescue of T␤RIII expression (as seen in A, right panel) and then stimulated with 50 ng ml Ϫ1 Wnt3a for 1 h followed by immunoblotting of lysates for phospho-LRP6 (Ser-1490) (pLRP6), LRP6, and ␤-actin. D, SKOV3 cells transiently expressing a second independent shRNA to T␤RIII (shT␤RIII) or Scrambled control ("Experimental Procedures") were stimulated with 50 ng ml Ϫ1 Wnt3a for 1 h followed by immunoblotting of lysates for phospho-LRP6 (Ser-1490) (pLRP6), LRP6, and ␤-actin. B-D, quantitations represent pLRP6:LRP6 ratios and are normalized to the untreated sample. All panels (A-D) represent at least two independent biological trials.
when compared with OVCA429 cells with low levels of T␤RIII (Fig. 1B). In SKOV3 cells, which express high levels of T␤RIII (Fig. 1A, i), reducing T␤RIII expression using shRNA resulted in increased LRP6 phosphorylation when compared with Wnt3a-stimulated SKOV3 control cells expressing high endogenous T␤RIII (Fig. 1, C and D). To confirm that the effect of shT␤RIII was specific to T␤RIII, we utilized shRNA-resistant rat T␤RIII (21,23) to rescue T␤RIII expression and examined Wnt-induced LRP6 phosphorylation. We found that rescue of T␤RIII expression in shT␤RIII cells (Fig. 1A, iii) suppressed Wnt-induced LRP6 phosphorylation compared with cells containing endogenous T␤RIII (Fig. 1C). Total LRP6 levels were not significantly altered by shRNA to T␤RIII or transient expression of rat T␤RIII in SKOV3 cells when compared with control cells (Fig. 1C). Consistently, a second shRNA to T␤RIII (shT␤RIII-2) also resulted in increased LRP6 phosphorylation when compared with Wnt3a-stimulated control cells (Fig. 1D).
These results indicate that T␤RIII may regulate Wnt signaling at the signal reception level by suppressing canonical Wnt signaling.
Activation of the canonical Wnt pathway leads to stabilization and accumulation of cytosolic ␤-catenin, which then enters the nucleus and regulates Wnt target genes (10). Consistent with reduced LRP6 phosphorylation, Wnt-induced ␤-catenin cytosolic accumulation was significantly reduced in the presence of T␤RIII (Fig. 2, A and B).
Upon ␤-catenin accumulation and stabilization, activation of TCF/LEF-sensitive transcription by ␤-catenin provides a robust readout of the Wnt-stimulated canonical pathway (41). To test whether T␤RIII-mediated changes on LRP6 phosphorylation and ␤-catenin accumulation would translate to downstream effects on TCF/LEF activity, we analyzed the activity of a TCF/LEF-sensitive reporter, which contains multiple ␤-catenin binding sites (42). We found that Wnt3a significantly increased TCF/LEF reporter activity in OVCA429 cells (Fig.  2C). Increasing T␤RIII expression in these cell lines resulted in a significant suppression of Wnt3a-induced activation of the TCF/LEF reporter compared with control Wnt-treated cells (Fig. 2C). Similar to trends seen in OVCA429 cells, overexpressing T␤RIII in SKOV3 cells (high T␤RIII) resulted in suppression of Wnt3a-induced TCF/LEF activity compared with control Wnt-treated cells (Fig. 2C). Side-by-side analysis of Wnt3a-stimulated TCF/LEF activity in SKOV3 (high T␤RIII) and OVCA429 (low T␤RIII) cells in the same experiment revealed lower Wnt3a-induced TCF/LEF activity in SKOV3 cells when compared with Wnt3a-treated ovarian cancer OVCA429 cells (Fig. 2D), which we hypothesized was in part due to higher endogenous T␤RIII expression in SKOV3 cells (Fig. 1A, left graph). This hypothesis was confirmed in SKOV3 cells using shRNA to T␤RIII (Fig. 1A, right graph), which resulted in enhanced Wnt-induced TCF/LEF reporter activity compared with control cells (Fig. 2E). This increased Wnt signaling in shT␤RIII cells was suppressed upon restoring T␤RIII expression using shRNA-resistant rat T␤RIII (Fig. 2E), consistent with increased LRP6 activation observed in SKOV3 cells upon knockdown of T␤RIII (Fig. 1C). Regulation of TCF/LEF reporter activity by T␤RIII was not restricted to ovarian cancer cells, as T␤RIII expression also repressed Wnt-induced TCF/ LEF reporter activity in 4T1 (breast cancer) cells (Fig. 4D), indicating a broad-based impact of T␤RIII on Wnt signaling regulation.
TGF-␤ Signaling Does Not Limit T␤RIII the Ability to Suppress Wnt/␤-Catenin Signaling-To begin elucidating the mechanisms by which T␤RIII regulates Wnt signaling, we examined whether the presence of TGF-␤, a high affinity ligand for the T␤RIII core domain (43)(44)(45), impacts the ability of T␤RIII to suppress Wnt signaling. We found that both TGF-␤1 and TGF-␤2 enhanced Wnt-induced LRP6 phosphorylation and TCF/LEF activity (Fig. 3, A and B) in OVCA429 cells and, to a lesser extent, in SKOV3 cells (high T␤RIII) (Fig. 3C, lanes  1-4), indicating a cooperative role for TGF-␤ ligands in Wnt signaling that may be repressed by T␤RIII. Treating T␤RIII knockdown SKOV3 cells (shT␤RIII) with TGF-␤ resulted in an enhancement of Wnt3a-TGF-␤ cooperativity compared with control T␤RIII-expressing SKOV3 cells treated with Wnt3a and TGF-␤ (Fig. 3C, lanes [5][6][7][8]. Because TGF-␤2 binds the core domain of T␤RIII with higher affinity than TGF-␤1 (46), and it showed the most robust enhancement of Wnt3a-induced TCF/LEF activity (Fig. 3A), this ligand was chosen to determine TGF-␤ signaling-mediated changes on the suppression of Wnt3a-induced TCF/LEF activity by T␤RIII. We found that Wnt-induced TCF/LEF activity, both in the absence and presence of TGF-␤2, was dampened by T␤RIII expression in OVCA429 cells (Fig. 3D).
To confirm that T␤RIII does not require TGF-␤ signaling receptors to suppress Wnt signaling, we first utilized SB431542 (inhibitor of T␤RI kinase activity) and analyzed Wnt-induced TCF/LEF activity in OVCA429 cells. We found that inhibition of T␤RI suppressed Wnt signaling independent of T␤RIII expression in control cells (Fig. 3E). However, inhibition of T␤RI did not affect the ability of T␤RIII to suppress Wnt-induced TCF/LEF activity in OVCA429 cells when compared with control cells (Fig. 3E), indicating that repression of Wnt signaling by T␤RIII is independent of T␤RI kinase activity. Several TGF-␤-independent roles for T␤RIII have been reported through its interactions with the type II TGF-␤ receptor, T␤RII (43). However, transient expression of T␤RII lacking its cytoplasmic domain (T␤RII-⌬Cyto), and therefore unable to interact with T␤RIII (43, 47), did not affect the ability of T␤RIII to suppress Wnt-induced TCF/LEF activity when compared with control cells (Fig. 3F). Similar to what is shown in Fig. 3E, the removal of the T␤RII cytoplasmic domain (T␤RII-⌬Cyto) in GFP-expressing cells led to a suppression of Wnt-induced TCF/LEF activity when compared with control cells (Fig. 3F). These T␤RIII-independent observations of the effects of T␤RII-⌬Cyto and SB431542 on TCF/LEF activity may point to autocrine TGF-␤-Wnt signaling mechanisms unrelated to the ability of T␤RIII to suppress Wnt-dependent Wnt signaling. Collectively, these data suggest that even in the presence of the high affinity ligand TGF-␤2, the absence of TGF-␤ signaling, and T␤RIII-T␤RII interaction, T␤RIII is still able to suppress Wnt signaling.
GAG Chains of T␤RIII Regulate Wnt Signaling-Wnt glycoproteins have been shown to have a high affinity for GAG chains on transmembrane proteoglycans (33), and the extracellular T␤RIII domain contains two sites of heparan and chon-droitin sulfate GAG chains (23,48). To determine whether the chains on T␤RIII are involved in suppressive effects on Wnt signaling, we expressed full-length T␤RIII (T␤RIII), T␤RIII lacking GAG chain modifications (T␤RIII-⌬GAG) (30), or control vectors in OVCA429 cells and assessed the levels of phosphorylation of LRP6, cytosolic ␤-catenin accumulation, and TCF/LEF activity induced by exogenous Wnt3a. We found that, unlike full-length T␤RIII, T␤RIII-⌬GAG failed to suppress LRP6 phosphorylation in OVCA429 cells (Fig. 4A). Consistently, T␤RIII-⌬GAG did not suppress Wnt3a-dependent ␤-catenin cytoplasmic accumulation compared with fulllength T␤RIII; instead, ␤-catenin cytoplasmic accumulation in the presence of T␤RIII-⌬GAG resembled the cytoplasmic ␤-catenin levels observed in Wnt3a-treated control cells (Fig.  4B). T␤RIII-⌬GAG cells also failed to suppress TCF/LEF activity when compared with full-length T␤RIII (Fig. 4C). The effect of the T␤RIII GAG chains on Wnt signaling was not restricted to ovarian cells, as T␤RIII-⌬GAG also failed to suppress Wnt signaling when compared with full-length T␤RIII in the murine mammary 4T1 cells (Fig. 4D). Values are normalized to control GFP. B, cytoplasmic fractions obtained after subcellular fractionation ("Experimental Procedures") of OVCA429 cells transiently expressing T␤RIII or GFP stimulated with 50 ng ml Ϫ1 Wnt3a for 1 h followed by immunoblotting of lysates for ␤-catenin, GAPDH (positive cytoplasmic marker), and E-cadherin (negative cytoplasmic marker), representative at least two independent biological trials. C-E, the indicated cells expressing T␤RIII, shT␤RIII, or shT␤RIII with rT␤RIII, as described in Fig. 1, B and C, were transfected with a Wnt-responsive luciferase reporter and SV40 control vector and left untreated or treated with 50 ng ml Ϫ1 Wnt3a for 24 h. Luciferase activity was then measured as described under "Experimental Procedures." All values are normalized to the untreated sample and represent the average of two independent biological trials, each conducted in duplicate. Data were analyzed using two-tailed Student's t test and represent the mean Ϯ S.E. Scr, Scrambled; n.s., not significant; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001.

JOURNAL OF BIOLOGICAL CHEMISTRY 25719
To test whether the extracellular domain (ECD) of T␤RIII was sufficient to suppress Wnt-induced signaling, we used two parallel approaches. We treated OVCA429 cells, in the absence and presence of Wnt3a, with either conditioned media (CM) from cells expressing only the T␤RIII ECD (Sol-T␤RIII-1) (16,18,44) or CM from cells expressing full-length T␤RIII containing soluble T␤RIII in the media due to shedding (Sol-T␤RIII-2) (30, 44) (Fig. 4E). CM from control vector (GFP) expressing cells was used as control (GFP-CM, Fig. 4E). These conditions were compared with OVCA429 cells expressing full-length T␤RIII in the same experiment (Fig.  4E). We found that both the shed and soluble forms of T␤RIII were able to significantly suppress Wnt-induced TCF/LEF activity to the same extent as they expressed fulllength T␤RIII (Fig. 4E). To control for possible artifacts associated with infection of vectors, we also tested media from uninfected cells (Fig. 4F) and found that infection with GFP did not impact TCF/LEF activity (Fig. 4F). Taken together, these data confirm that T␤RIII ECD, with its GAG chains, is sufficient to suppress Wnt-induced signaling.
T␤RIII Interacts with Wnt, and the Balance between Sulfated Heparan and Chondroitin Chains Determines T␤RIII Ability to Regulate Wnt/␤-Catenin Signaling-To determine whether T␤RIII binds Wnt3a, we used co-immunoprecipitation of recombinant Wnt3a and T␤RIII, a methodology commonly used to study Wnt interactions with its receptors (13,49). We found a Wnt dose-dependent interaction between T␤RIII and Wnt3a in OVCA429 cells (Fig. 5A). Consistent with the extracellular domain of T␤RIII as sufficient to suppress Wnt signaling (Fig. 4E), we found that soluble T␤RIII was also able to interact with Wnt3a, as determined by using CM from COS-7 cells expressing full-length T␤RIII and HA-tagged Wnt3a (Fig.  5B). To determine whether the T␤RIII-Wnt3a interaction is mediated through the T␤RIII GAG chains as suggested by our Wnt signaling assays (Fig. 4), we incubated OVCA429 cell lysates with recombinant Wnt3a and performed co-immunoprecipitation in cells expressing T␤RIII, T␤RIII-⌬GAG, or control (see "Experimental Procedures"). We observed immunoprecipitation of Wnt3a and T␤RIII reduced to background levels in cells expressing T␤RIII-⌬GAG (Fig. 5C). These data indicate that the interaction/binding capacity of T␤RIII-⌬GAG is significantly less than full-length T␤RIII. These findings are consistent with T␤RIII-⌬GAG being unable to inhibit Wnt3a signaling (Fig. 4, A-C).
Because T␤RIII represses Wnt signaling and appears to interact with Wnt3a through its GAG chains, we aimed to test whether the regulation of Wnt signaling by T␤RIII GAG chains was dependent on the sulfation state of the T␤RIII GAG chains. We treated T␤RIII-expressing OVCA429 cells with sodium chlorate, a competitive inhibitor of ATP-sulfurylase, which resulted in proteoglycans arriving at the cell surface bearing nonsulfated heparan sulfate or chondroitin sulfate chains (50). We found that non-sulfated GAG chains on T␤RIII significantly stimulated Wnt-induced TCF/LEF activity (Fig. 6A).
Treatment with sodium sulfate, which overcomes the effects of sodium chlorate and restores sulfation of proteoglycans (50), decreased Wnt-induced TCF/LEF activity compared with T␤RIII-expressing OVCA429 cells treated only with sodium chlorate (Fig. 6A). These results demonstrate that the sulfation of the T␤RIII GAG chains is required for T␤RIII-mediated sup- Quantitations represent pLRP6:LRP6 ratios and are normalized to the untreated sample. B, the indicated OVCA429 cells were stimulated with 50 ng ml Ϫ1 Wnt3a for 1 h and immunostained for ␤-catenin (red). Scale bars: 20 m. The graph represents quantitation of ␤-catenin fluorescence at the membrane versus cytoplasm ("Experimental Procedures"). n Ն 30 cells/condition. Western analysis shows T␤RIII and ␤-actin levels. Data were analyzed using analysis of variance followed by a post hoc Shapiro-Wilk test and represent the mean Ϯ S.E. All values were normalized to the untreated sample. C-F, the indicated cells were stimulated with 50 ng ml Ϫ1 Wnt3a alone (C and D); 50 ng ml Ϫ1 Wnt3a in the presence of CM from cells expressing only T␤RIII-ECD (Sol-T␤RIII-1) (E), CM from cells expressing full-length T␤RIII (Sol-T␤RIII-2), or CM from control GFP-expressing cells for 24 h (E); or 50 ng ml Ϫ1 Wnt3a in the presence of control media from untransfected cells or from cells transiently expressing GFP (F). Cells were lysed after 24 h, and luciferase activity was measured in 1% serum as described under "Experimental Procedures." All values were normalized to the untreated sample. All luciferase data (C-F) are representative of at least two independent biological trials, each conducted in duplicate. Data were analyzed using two-tailed Student's t test and represent the mean Ϯ S.E. of a biological replicate. All values are normalized to the untreated sample, and all figures represent at least two independent biological trials. n.s., not significant; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. DECEMBER  Because the GAG chains on T␤RIII comprise both heparan sulfate (HS) and chondroitin sulfate (CS) chains (32), we aimed to isolate the individual effects of the different GAG chains of T␤RIII on Wnt signaling. To do this, we first determined whether the suppressive role of T␤RIII in Wnt signaling was conserved in parental CHO K1 cells, where T␤RIII expresses both HS and CS chains (51). Although CHO cells have a modest response to Wnt stimulation as observed previously (33,52) and by us here (Fig. 6, B and C), we observed a significant decrease in Wnt signaling upon T␤RIII-expression in CHO K1 cells compared with control cells (Fig. 6B), consistent with our observations in ovarian and breast cancer cells (Figs. 2 and 4).

T␤RIII Regulates Wnt Signaling
To determine the role of T␤RIII CS chains in Wnt signaling, we utilized the CHO cell line derivative pgsD-677; these cells lack both N-acetylglucosaminyltransferase and glucuronyltransferase activities and are unable to synthesize heparan sulfate but can produce high amounts of chondroitin sulfate (51). We increased T␤RIII expression in pgsD-677 (⌬HS) cells (as described under "Experimental Procedures") and examined Wnt-induced TCF/LEF activity. Strikingly, we observed a significant increase in Wnt signaling in T␤RIII-expressing pgsD-677 cells compared with control cells (Fig. 6C). Furthermore, the removal of the T␤RIII CS chains with chondroitinase (Fig.  6D, right panel, Ch) reduced Wnt-induced TCF/LEF activity in T␤RIII-expressing pgsD-677 cells (Fig. 6D). Because pgsD-677 cells express only CS GAG chains (51), we tested whether CS chains promote Wnt signaling in cells that make both HS and CS GAG chains. Similar to our results in pgsD-677 cells (Fig.  6D), we found that T␤RIII was able to further repress Wnt signaling in OVCA429 cells treated with chondroitinase as compared with control cells (2ϫ repressed, Fig. 6E). In contrast, heparanase treatment of T␤RIII-expressing OVCA429 cells resulted in increased TCF/LEF activity compared with heparanase (Hp)-untreated cells (5ϫ increased, Fig. 6F). These data suggest that HS and CS chains on T␤RIII contribute, in an opposing fashion, to the availability of Wnt for signaling. Therefore, we propose that the HS chains of T␤RIII are responsible for Wnt3a sequestration and subsequent T␤RIII-mediated suppression of Wnt3a signaling. In contrast, T␤RIII CS chains increase Wnt availability and signaling (Fig. 6G).

Discussion
We provide novel evidence for the T␤RIII/betaglycan-mediated regulation of canonical Wnt signaling through distinct functions of its heparan-and chondroitin-sulfated GAG chains. Our studies demonstrate that the HS chains of T␤RIII are responsible for the suppression of Wnt3a signaling, most likely via sequestering Wnt, in contrast with the CS chains of T␤RIII, which promote Wnt signaling. Based on our findings, we propose that Wnt interactions with the HS chains on T␤RIII result in the sequestration of Wnt away from LRP6 and Frizzled, which decreases the levels of signaling-productive complexes between the ligand and its receptors. This hypothesis was confirmed upon examining the inability of T␤RIII to suppress Wnt FIGURE 5. T␤RIII interacts with Wnt through its GAG chains. A, OVCA429 cells transiently expressing full-length T␤RIII were immunoprecipitated (IP) and analyzed for Wnt3a-T␤RIII interactions using the Wnt pulldown assay described under "Experimental Procedures" and previously (13). All values were normalized to the IgG sample. B, conditioned medium from COS-7 cells transiently expressing full-length T␤RIII and Wnt3a-HA was immunoprecipitated using anti-T␤RIII and immunoblotted using anti-HA and anti-T␤RIII to analyze Wnt3a-T␤RIII interactions as described under "Experimental Procedures" and previously (21,27,77). C, OVCA429 cells transiently expressing GFP (control), full-length T␤RIII, or T␤RIII-⌬GAG were immunoprecipitated using anti-T␤RIII and immunoblotted (IB) using anti-Wnt3a and anti-T␤RIII to analyze Wnt3a-T␤RIII interactions using the Wnt pulldown assay described under "Experimental Procedures" and previously (13) All values were normalized to the GFP sample, and all figures represent at least two independent biological trials.
signaling upon removal of its GAG chains (Fig. 4). Mechanistically, our pulldown assays in T␤RIII-expressing cells (Fig. 5) indicate an interaction between T␤RIII and Wnt glycoproteins, which have a high affinity for polyanionic compounds such as heparin (53), and reveal that the GAG chains significantly increase Wnt-T␤RIII interaction to suppress Wnt signaling. . The balance between sulfated heparan and chondroitin chains on T␤RIII determines the ability of T␤RIII to regulate Wnt/␤-catenin signaling. A, OVCA429 cells transiently expressing full-length T␤RIII, transfected with a Wnt-responsive luciferase reporter and a SV40 control vector were pretreated with 50 mM NaClO 3 with or without 10 mM Na 2 SO 4 as indicated for 2 h. Cells were then stimulated with 50 ng ml Ϫ1 Wnt3a, and luciferase activity was measured as described under "Experimental Procedures." All values were normalized to the Wnt-treated sample. B-F, CHO K1, pgsD-677, or OVCA429 cells expressing T␤RIII or GFP were transfected with a Wnt-responsive luciferase reporter and a SV40 control vector and pretreated with 100 mU ml Ϫ1 chondroitinase (Ch) (D and E) or 20 mU ml Ϫ1 heparanase (Hp) (F) for 2 h before overnight incubation with 50 ng ml Ϫ1 Wnt3a. Luciferase activity was measured as described under "Experimental Procedures." All values were normalized to the untreated sample. Western blotting analysis (D) shows T␤RIII expression in pgsD-677 cells after CS chain removal using 100 mU ml Ϫ1 chondroitinase. All data represent at least two independent biological trials. Data were analyzed using two-tailed Student's t test and represent the mean Ϯ S.E. G, model of canonical Wnt/␤-catenin signaling regulation by T␤RIII. n.s., not significant; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. DECEMBER 2, 2016 • VOLUME 291 • NUMBER 49

JOURNAL OF BIOLOGICAL CHEMISTRY 25723
Strikingly, the T␤RIII CS chains promote Wnt3a signaling in the absence of its HS chains (Fig. 6, B and D). To support this conclusion, chondroitinase treatment in pgsD-677 and OVCA429 cells resulted in a loss of Wnt signaling, thus indicating an exciting new role for the chondroitin chains of T␤RIII in stimulating Wnt signaling.
The role of GAG chains in Wnt signal transduction may also depend on the core protein and specific biochemical cues, as our data indicate opposing functions for T␤RIII HS and CS chains in Wnt signaling. In support of our hypothesis, it has been shown that exogenous chondroitin sulfate, heparin, and GAG are unable to stimulate Wnt3a signaling, whereas endogenous CSPG promote Wnt signaling in mouse L-cell fibroblasts, suggesting that the core proteins of CSPG may be involved in regulating Wnt3a activity (36). We speculate that the localization, sulfation, and/or chain length of GAG chains attached to core proteins could contribute to differences in ligand availability and signaling.
Studies have shown also that cell context can determine the role that proteoglycans and GAG chains play in cancer progression. The enzymatic elimination of chondroitin sulfate molecules in primary breast tumors, for example, increases lung metastases in mice (54), whereas the digestion of cell surface CS on lung cancer cells injected into tail veins leads to a reduction in the number of tumor cells able to populate and metastasize (55). These results suggest that CS molecules may have opposing roles during cancer progression: an anti-metastatic function in primary tumor tissue and a pro-metastatic role during extravasation (circulating cancer cell interaction with endothelial cells) (56). Other proteoglycans have also been shown to function as either tumor promoters or suppressors depending on the protein core, GAG chains attached, associated molecules, proteoglycan localization, and tumor type (57). Perlecan, for example, can both promote tumor invasiveness (58) and inhibit angiogenesis (59), whereas glypicans and syndecans may promote local cancer cell growth and metastatic potential in some cancer tissues (37, 60) but inhibit tissue growth, invasion, and metastasis in others (61,62). Together, these data show a requirement for the proteoglycan core domain and cellular environment in deciding GAG chain function.
In addition to the contributions made by the proteoglycan core domain and environment, the sulfation state of the proteoglycan also plays a major role in its ability to regulate signaling pathways. Upon treatment of our T␤RIII-expressing OVCA429 cells with sodium chlorate, an ATP-sulfurylasecompetitive inhibitor that causes proteoglycans to arrive at the cell surface bearing nonsulfated heparan sulfate or chondroitin sulfate chains (50), we found T␤RIII unable to repress Wnt signaling, indicating that the sulfation of T␤RIII GAG chains is required for proper Wnt signal regulation by T␤RIII (Fig. 6A), consistent with previous reports for glypican-1 (33). Studies in Drosophila have also shown that, upon treatment of Drosophila cells with sodium chlorate or in the absence of an HS N-deacetylase/N-sulfotransferase, cells are completely deficient in HS chain sulfation and Wingless (Wg) signaling is disrupted (63)(64)(65)(66). HS chain sulfation plays a vital role in regulating FGF signaling as well. Consistently, the HS chains of T␤RIII can also regulate FGF signaling and play a critical role in tumor progression (23).
Previous reports indicate that FGF signal transduction is dependent on sulfation of the 2-O and 6-O positions on HS chains, which control FGF1 binding to heparin and FGF1-dependent dimerization and activation of the FGFR1 receptor, respectively (67)(68)(69). In articular cartilage, studies reveal a Wnt signaling promoter role for CS chains that is dependent on the sulfation of the CS chain (70). Taken together, these studies, combined with our data, suggest that sulfation plays a significant role in growth factor signaling regulation by GAG chains on proteoglycans.
It is possible that different expression levels of ␤1,4-N-acetylgalactosaminyltransferase-I (␤4GalNAcT-I) and/or ␣1,4-Nacetylglucosaminyltransferase-I (␣4GlcNAcT-I), which initiate the synthesis of CS or HS chains, respectively, may also contribute to the T␤RIII proteoglycan state and subsequent effects on Wnt signaling. Moreover, within a single core protein, Ser-Gly residues in a hydrophobic pocket might signal heparan sulfate attachment, whereas Ser-Gly residues in an exposed hydrophilic environment might signal chondroitin sulfate attachment. These different local environments could achieve selectivity by modulating the activity of ␤4GalNAcT-I and ␣4GlcNAcT-I (71). Other biochemical cues may include the location of N-linked glycosylation sites (Asn-Phe-Ser) as described for syndecan-1 (72). Attachment of an N-linked sugar at a GAG chain attachment site would likely prevent subsequent recognition by the xylosyltransferase and GAG chain attachment to the T␤RIII core protein.
The precise mechanism by which CS chains of T␤RIII increase Wnt availability remains to be determined. Future studies into the biochemical cues involved in determining the proteoglycan state of HSPG such as T␤RIII, as well as the role of T␤RIII in regulating Wnt signaling, will help shed light on Wnt signaling regulation and increase our understanding of the diverse roles that proteoglycans like T␤RIII play in signaling and disease.
Immunoprecipitation and Western Blotting-Immunoprecipitation and Western blotting were performed using standard techniques as described previously (21,27,77). For co-immunoprecipitation in COS-7 cells, T␤RIII-expressing cells were transfected with the indicated Wnt3a-HA construct, and the culture medium was collected 48 h after transfection under serum-free conditions. T␤RIII was then immunoprecipitated by incubating the cell lysates overnight with anti-human T␤RIII antibody. The next day, protein G-Sepharose beads were added to the lysates for 2 h at 4°C. The beads were then washed three times with cold PBS and resuspended in sample buffer. The amount of T␤RIII or Wnt3a bound to the beads was detected by Western blotting with anti-human T␤RIII or Wnt3a antibodies.
Wnt3a-T␤RIII Pulldown Assay-This assay was performed as described previously (13,49). Briefly, OVCA429 cells were lysed in non-denaturing COIP lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM of NaCl, 1% Nonidet P-40, 10% glycerol, 1 mM DTT, 25 mM NaF, 1 mM Na 3 VO 4 and 1ϫ protease inhibitor mixture (catalog No. P8340, Sigma-Aldrich)). T␤RIII-HA was then immunoprecipitated by incubating the cell lysates overnight with an anti-human T␤RIII antibody. The next day, protein G-Sepharose beads were added to the lysates for 2 h at 4°C. Beads were then washed three times with PBS and incubated with 20 nM Wnt3a-conditioned medium for 2 h at 4°C. After two more washes with PBS, the beads were resuspended in sample buffer, and the amount of Wnt3a bound to T␤RIII was detected by Western blotting using anti-Wnt3a and anti-T␤RIII antibodies.
Luciferase Assay-The indicated cells were seeded in 24-well plates and co-transfected with a Luciferase reporter vector containing a ␤-catenin-responsive promoter (to drive luciferase expression (TOPFlash, catalog No. 12456, Addgene)) and SV40 (Renilla internal control vector). One day after transfection and infection, cells were incubated overnight with 50 ng ml Ϫ1 Wnt3a and then lysed. Luciferase activity (Luciferase assay system, Promega) was measured by calculating the ratio between luciferase and Renilla activities (to normalize for transfection efficiency) and then normalizing the values to the untreated sample.
Immunofluorescence and Intensity Analysis-The indicated cells were seeded onto coverslips in 12-well plates at a density of 5 ϫ 10 4 cells/well. After infections and treatment with 50 ng ml Ϫ1 Wnt3a, cells were washed with ice-cold PBS and fixed with 100% methanol for 10 min followed by PBS washes. Cells were permeabilized with 0.1% Triton X-100 in PBS and then blocked with 3% BSA or 0.2% gelatin in PBS for 30 min at room temperature followed by an overnight incubation at 4°C with a rabbit anti-␤-catenin antibody. After extensive washing with PBS, the cells were incubated with an Alexa-conjugated secondary antibody (Molecular Probes, Eugene, OR). Cells were mounted in mounting medium and analyzed under an Olympus IX81 motorized inverted microscope (Shinjuku, Tokyo, Japan). Fluorescence intensity for the ␤-catenin was analyzed using ImageJ 1.50d software (National Institutes of Health) by drawing a fixed line of interest over the membrane and cytoplasm followed by averaging the maximum intensities obtained from the plot profile plugin. To estimate the change in ␤-catenin localization after Wnt treatment in the presence and absence of T␤RIII, the ratio between the membrane and cytoplasmic fractions of ␤-catenin fluorescence was calculated. The statistical significance of the data was analyzed in SigmaPlot version 11 software. p values Ͻ 0.05 were considered to be statistically significant.
Subcellular Fractionation-The indicated cells were seeded in 12-well plates and infected to express T␤RIII. 48 h postinfection, the cells were treated with 50 ng ml Ϫ1 Wnt3a for 1 h and then lysed. Subcellular fractionation of ␤-catenin, the cytoplasmic marker GAPDH, and the plasma membrane marker E-cadherin was carried out using the cell fractionation kit (Cell Signaling Technology) according to the manufacturer's instructions.