The Tuberin-Hamartin Complex Negatively Regulates β-Catenin Signaling Activity

Tuberous sclerosis complex (TSC) is characterized by the formation of hamartomas in multiple organs resulting from mutations in the TSC1 or TSC2 gene. Their protein products, hamartin and tuberin, respectively, form a functional complex that affects cell growth, differentiation, and proliferation. Several lines of evidence, including renal tumors derived from TSC2+/− animals, suggest that the loss or inhibition of tuberin is associated with up-regulation of cyclin D1. As cyclin D1 can be regulated through the canonical Wnt/β-catenin signaling pathway, we hypothesize that the cell proliferative effects of hamartin and tuberin are partly mediated through β-catenin. In this study, total β-catenin protein levels were found to be elevated in the TSC2-related renal tumors. Ectopic expression of hamartin and wild-type tuberin, but not mutant tuberin, reduced β-catenin steady-state levels and its half-life. The TSC1-TSC2 complex also inhibited Wnt-1 stimulated Tcf/LEF luciferase reporter activity. This inhibition was eliminated by constitutively active β-catenin but not by Disheveled, suggesting that hamartin and tuberin function at the level of the β-catenin degradation complex. Indeed, hamartin and tuberin co-immunoprecipitated with glycogen synthase kinase 3 β and Axin, components of this complex in a Wnt-1-dependent manner. Our data suggest that hamartin and tuberin negatively regulate β-catenin stability and activity by participating in the β-catenin degradation complex.

that overexpression of TSC1 and TSC2 negatively regulates cell proliferation and induces G 1 /S arrest (2)(3)(4). In the case of tuberin, there appears to be an inverse correlation between tuberin level and p27(Kip1) expression and stability (5). Correspondingly, evidence supports a link between tuberin and cyclin D1 expression. Cortical tubers microdissected from TSC patients showed elevated cyclin D1 mRNA expression in the giant cells (6). Antisense inhibition of TSC2 in Rat1 fibroblasts resulted in up-regulation of cyclin D1 protein (3). Renal cortical tumors from the Eker rat model for TSC express elevated cyclin D1 compared with unaffected kidney tissue (7). As an in vivo target of the ␤-catenin pathway, cyclin D1 mRNA is responsive to the activity of the Tcf/LEF family of transcription factors (8,9). This raises the possibility that TSC1 and TSC2 negatively regulate ␤-catenin signaling and, thereby, modulate the expression of cyclin D1.
␤-Catenin is a highly conserved 95-kDa protein that participates in cell-cell adhesion through its association with members of the membrane-bound cadherin family, and in cell proliferation and differentiation as a key component of the Wnt/ Wingless pathway (reviewed in Ref. 10). In its signaling role, ␤-catenin shuttles between the cytoplasm and the nucleus where it binds the Tcf/LEF family of transcription factors to activate downstream target genes (reviewed in Ref. 11). In the absence of the secreted factor, Wnt, ␤-catenin is phosphorylated by GSK3␤ and is targeted for ubiquitination and degradation. Upon Wnt stimulation, Disheveled (Dsh) is activated and blocks the ability of GSK3␤ to phosphorylate ␤-catenin. Other components of this degradation complex include Axin, serving as a scaffolding protein, and APC, a tumor suppressor protein. Disruption at multiple levels of this pathway has been shown to be oncogenic in humans and rodents. In this study, ␤-catenin protein levels were found to be elevated in renal tumors from Eker rats. Overexpression of tuberin and hamartin in cells down-regulated ␤-catenin levels, its half-life and its activity. Furthermore, we showed that TSC1 and TSC2 proteins co-immunoprecipitated with GSK3␤ and Axin, supporting a role of hamartin and tuberin in modulating the ␤-catenin pathway.
␤-Catenin Steady-state Levels-HEK293T cells were transfected with increasing concentrations of TSC1 and TSC2 (wild-type or ⌬Y1571H mutant) vectors or with control vector (pcDNA3) (500, 1000, and 1500 ng) either with or without Wnt-1 vector using LipofectAMINE Plus reagents according to the manufacturer's instructions. Empty plasmids were added accordingly to normalize total DNA transfected. The cells were harvested in lysis buffer (10 mM Tris-HCl, pH 7.5, 100 mM NaCl, 0.5% Nonidet P-40, 0.5 g/ml leupeptin, 1.0 g/ml pepstatin, 0.2 mM PMSF) and lysed by freeze thawing at 48 h after transfection. Samples were analyzed by Western blotting.
Pulse-Chase Experiments-Vectors encoding TSC1, TSC2, or TSC2 (⌬Y1571H) mutant were co-transfected into HEK293T cells with or without Wnt-1. After 48 h, pulse-chase using [ 35 S]methionine incorporation was performed on transfected cells as outlined by Williams et al. (21). Protein content in cell lysates was analyzed using the BCA protein assay. Equal amounts of cell lysate were adjusted to equal volumes, pretreated with ConA-Sepharose (1:1 in immunoprecipitation buffer, see below) (Amersham Biosciences, Uppsala, Sweden) and then subjected to immunoprecipitation for ␤-catenin. Samples were resolved by SDS-PAGE and detected by autoradiography. Radiolabel band intensities were determined using a PhosphorImager Storm 840 (Amersham Biosciences).

Expression of ␤-Catenin in Eker Rat Kidney Tumors-The
Eker rat contains a germ-line mutation in the TSC2 gene (12) and spontaneously develops renal cortical epithelial tumors that have been shown to possess biallellic inactivation of TSC2 FIG. 1. Regulation of ␤-catenin levels by tuberin. A, homogenates from three different Eker rat renal tumors and corresponding unaffected kidney tissue were separated by SDS-PAGE and blotted for ␤-catenin and actin (loading control). B, steady-state levels of ␤-catenin with increasing TSC1 and TSC2 transfection. HEK293T cells were co-transfected with Wnt-1 and increasing amounts of vector(s) (500, 1000, and 1500 ng). Empty vector was added to normalize total DNA transfected. Cells were harvested at 48 h after transfection, and protein levels were analyzed by Western blotting. ␤-Catenin band intensities were measured using the NIH Image processing program and normalized against the density of corresponding ␣-tubulin bands. The accumulation of ␤-catenin relative to the non-stimulated sample point was plotted. C, determination of ␤-catenin half-life was performed as described under ''Experimental Procedures.'' Radiolabel intensities of [ 35 S]methionine-labeled ␤-catenin was measured, normalized to the zero time point, and plotted against time (0, 30, 60, and 120 min). Simultaneously exposed gels were measured and compared. The halflife of ␤-catenin was determined from the slope of each graph. Specific polyclonal antibodies were used to detect hamartin (4050) and tuberin (L3-2). Commercially available antibodies were used to detect ␤-catenin, actin, and ␣-tubulin. TSC1/TSC2 Inhibit ␤-Catenin Pathway due to loss of heterozygosity, nonsense mutation, or null mutation (22). In a previous study, cyclin D1 levels were shown to be elevated in these kidney tumors compared with unaffected kidney tissue (7). Since cyclin D1 gene is a known target of the ␤-catenin signaling pathway and the accumulation of ␤-catenin has been shown to activate the transcription of the cyclin D1 gene (8,9). ␤-Catenin levels were examined in Eker rat kidney tumors. Tumors from three separate Eker rats were dissected from unaffected tissue, homogenized, and analyzed by Western blotting for ␤-catenin expression. As shown in Fig. 1A, ␤-catenin levels are higher in tumor samples compared with corresponding unaffected kidney tissue, reflecting the trend observed for cyclin D1. Thus ␤-catenin appears to accumulate upon TSC2 inactivation implying that tuberin may affect ␤-catenin levels.
Next, we examined the effect of hamartin and tuberin expression on ␤-catenin half-life. HEK293T cells were transfected as described above and then subjected to pulse-chase following [ 35 S]methionine incorporation. Lysates collected at specific time points were treated with ConA-Sepharose to remove cad-herin-bound ␤-catenin leaving only the free pool of ␤-catenin for immunoprecipitation. Immunoprecipitates were resolved on SDS-PAGE, and the radiolabel intensity of [ 35 S]methionine incorporated in to ␤-catenin was measured from the gel to determine the half-life. Without Wnt-1 stimulation, ␤-catenin half-life was unchanged either with or without expression of hamartin and tuberin (data not shown). With Wnt-1 stimulation, the half-life of ␤-catenin in vector control samples was 1.7 h (Fig. 1C). This was the same for ␤-catenin in samples overexpressing hamartin and mutant tuberin (⌬Y1571H) (Fig.  1C). However, the half-life of ␤-catenin was reduced to about 1 h upon expression of hamartin and wild-type tuberin (Fig.  1C). This 41% decrease in ␤-catenin half-life is consistent with the steady-state data showing an effect of the TSC proteins on ␤-catenin level. In both situations, modulation of ␤-catenin by wild-type tuberin occurred under condition of Wnt stimulation. This function is disrupted in the presence of a disease-causing TSC2 mutation.
Hamartin and Tuberin Inhibit Wnt-1-stimulated ␤-Catenin Transcriptional Activity-To determine whether modulation of ␤-catenin levels by hamartin and tuberin affects its transcriptional activity, the ability of Wnt-1 to activate a Tcf/LEFluciferase reporter construct (TOPFLASH) was examined in transient transfection assays using HEK293T cells. These cells were co-transfected with hamartin and/or tuberin constructs along with a vector for ␤-galactosidase to account for transfection efficiency. Parallel assays were performed using FOP-FLASH, a mutant reporter, to monitor background activity. Upon stimulation with Wnt-1, cells with vector control revealed a 12-fold increase in luciferase activity relative to the nonstimulated cells (Fig. 2A, compare lanes 1 and 2). Co-expression of both wild-type tuberin and hamartin in Wnt-1 stimulated cells significantly reduced reporter activity (Fig. 2A, lane 3), while expression of hamartin or tuberin alone had only minor effects ( Fig. 2A, lanes 5 and 6). Importantly, co-expression of the tuberin ⌬Y1571H mutant with hamartin did not suppress Wnt-1-stimulated TOPFLASH activity (Fig. 2A, lane 4) consistent with the effects on ␤-catenin protein levels described above (Fig. 1, B and C). Under the same conditions, hamartin and tuberin had no effects on FOPFLASH activity ( Fig. 2A, lanes 7  and 8). Our data suggest that hamartin and tuberin, functioning as a complex, are capable of inhibiting Wnt-1 stimulated ␤-catenin-dependent transcriptional activity. FOPFLASH control was also assayed along with ␤-galactosidase as transfection control. Luciferase activity from each sample was normalized and expressed as a value relative to ␤-galactosidase activity. Data from each graph are averages from at least three separate assays and are expressed as a value relative to ␤-galactosidase activity in each sample. Ectopic expression of hamartin and tuberin was detected by Western blotting using polyclonal antibodies (4050 and L3-2, respectively). The relative low levels of hamartin transgene expression are due to reduced protein stability in the absence of tuberin co-expression (27) (*, unpaired t test, p Ͻ 0.05).

TSC1/TSC2 Inhibit ␤-Catenin Pathway Hamartin and Tuberin Function within the Wnt/␤-Catenin
Signaling Pathway-To determine at what level in the Wnt/␤catenin signaling pathway hamartin and tuberin act, we examined the effects of TSC1 and TSC2 on TOPFLASH activity when stimulated by different components of the Wnt pathway. The CA-␤-catenin mutant with its serine/threonine residues (Ser-33, Ser-37, Thr-41, Ser-45) replaced with alanine residues, thus preventing its phosphorylation and degradation (16), acts as a downstream stimulus and activates the TOPFLASH reporter by over 6-fold in HEK293T cells (Fig. 2B, compare lanes  1 and 2). Co-expression of hamartin and tuberin was ineffective in reducing CA-␤-catenin stimulation of the Tcf/LEF reporter (Fig. 2B, compare lanes 2 and 3). Western blot of samples confirmed equal expression of CA-␤-catenin (data not shown). As a control, a dominant negative mutant of Tcf-4 (⌬N-Tcf-4) that lacks the N-terminal ␤-catenin binding domain (18) was able to inhibit CA-␤-catenin activity completely (Fig. 2B, lane  6). These results suggest that hamartin and tuberin act upstream of ␤-catenin.
Next, TOPFLASH activity was measured in the presence of ectopically expressed Dsh, an effector that is stimulated by the Wnt-Frizzled receptor complex upstream of ␤-catenin (see Ref. 10). Transient overexpression of Dsh stimulates TOPFLASH activity by ϳ3-fold in control vector transfected cells (Fig. 2C,  compare lanes 1 and 2). This activity was inhibited to near baseline levels upon overexpression of hamartin and tuberin (Fig. 2C, lane 3). Hamartin alone did not reduce the activity, while tuberin alone slightly decreased activity (Fig. 2C, lanes 4  and 5). Again, ⌬N-Tcf-4 reduced activity toward unstimulated levels (Fig. 2C, lane 6). Together, these results are consistent with tuberin and hamartin exerting an effect on the Wnt signaling pathway at a level between Dsh and ␤-catenin (i.e. the ␤-catenin degradation complex).
Hamartin and Tuberin Interact with Components of the ␤-Catenin Degradation Complex-The ␤-catenin degradation complex is comprised of several proteins, including APC, Axin, and GSK3␤, and is responsible for the regulation of cytoplasmic ␤-catenin (see Ref. 10). To determine whether hamartin and tuberin physically interact with components of the ␤-catenin degradation complex, co-immunoprecipitation assays were performed in HEK293T cells ectopically expressing hamartin and tuberin along with GSK3␤. Anti-tuberin antibodies brought down GSK3␤ only in samples where both were overexpressed (Fig. 3A, panel i, lane 4). This band was not observed in samples where the GSK3␤ construct was co-expressed with vector control (Fig. 3A, panel i, lane 2) or in samples without the GSK3␤ construct (Fig. 3A, panel i, lanes 1 and 3). As expected, hamartin co-immunoprecipitated with tuberin in sample where both were overexpressed (Fig. 3A, panel ii, lanes 3 and 4). Conversely, immunoprecipitation of GSK3␤ brought down both tuberin and hamartin only in samples where all three were overexpressed (Fig. 3A, panels iv and v, lane 4). The expression of ectopic proteins was verified in cell lysates (Fig. 3A, panels  vii, viii, and ix). Compared with the level of overexpression, the amount of interacting protein was relatively small, suggesting that only a fraction of tuberin/hamartin and GSK3␤ can associate with one another.
If hamartin and tuberin interact with the GSK3␤ that function in the ␤-catenin degradation complex, one would predict that other components of the complex such as Axin would co-immunoprecipitate with hamartin and tuberin. To test this hypothesis, hamartin and tuberin were ectopically expressed in the presence of c-Myc-tagged Axin and then subjected to immunoprecipitation analyses. Using anti-tuberin and anti-c-Myc antibodies, tuberin and tagged Axin co-precipitated in samples where they were both overexpressed (Fig. 3B, panels i and v,   FIG. 3. Co-immunoprecipitation of tuberin and hamartin with components of the ␤-catenin degradation complex. A, HEK293T cells were transfected with vectors for pcDNA3, TSC1, TSC2, or GSK3␤. Cell lysates were immunoprecipitated (IP) either with monoclonal antibody for GSK3␤ or with tuberin polyclonal antibody (C20). Specific proteins were detected by immunoblotting (IB) using antibodies toward tuberin, hamartin, or GSK3␤. B, similarly, HEK293T cells were transfected with vectors for pcDNA3, GSK3␤, c-Myc-tagged Axin, or hamartin and tuberin. Immunoprecipitations were performed using anti-c-Myc antibody (9E10) for c-Myc-tagged Axin or polyclonal antibody for tuberin (L3-2). Co-immunoprecipitated protein was detected on Western blots using specific antibodies. C, endogenous tuberin was immunoprecipitated from HEK293T cell lysates and analyzed by immunoblotting for GSK3␤. Tuberin was immunoprecipitated using tuberin polyclonal antibody (C20), while monoclonal antibody was used to detect GSK3␤. Blots in each panel are from the same exposure. lanes 5 and 6). Hamartin also co-immunoprecipitated with Axin (Fig. 3B, panel vi, lanes 5 and 6). Tuberin, hamartin, and c-Myc-tagged Axin were not detected in vector control samples (Fig. 3B, panels i, iii, v, and vi, lanes 1-3) or in samples without the tagged construct (Fig. 3B, panels i, iii, v, and vi, lane 4). We conclude that the tuberin-hamartin complex can associate with GSK3␤ and Axin possibly as part of the ␤-catenin degradation complex.
Finally, to examine whether tuberin interacts with the endogenous GSK3␤ complex, co-immunoprecipitation assays were performed in HEK293T cells with and without Wnt stimulation. A band corresponding to GSK3␤ was found to coimmunoprecipitate with tuberin but not in preimmune serum control samples (Fig. 3C, compare lanes 1 and 3). Furthermore, the amount of co-immunoprecipitated GSK3␤ was reduced upon Wnt-1 stimulation (Fig. 3C, compare lanes 1 and 2), suggesting that this interaction can be modulated by Wnt-1.

DISCUSSION
The TSC1 and TSC2 tumor suppressor genes have recently been implicated to play a role in negatively regulating mTOR in the PI3K signaling cascade (23,24). As a result, tumors secondary to the inactivation of these genes have elevated levels of p70 S6 kinase activity that is reversible by rapamycin, a specific mTOR inhibitor (7). While this pathway may explain some of the complex phenotype exhibited by TSC pathology (i.e. cell size abnormalities), alteration in other cellular functions may involve additional mechanisms. Up-regulation of cyclin D1 mRNA and protein has been noted in CNS and renal lesions (6, 7), but a recent study using rapamycin in the Eker rat model of TSC failed to show a significant change in cyclin D1 or p27 levels despite anti-tumor response (7). Here, we describe an alternative pathway that may be relevant to the abnormalities observed in cell proliferation and differentiation. Our data show that in vivo levels of ␤-catenin are elevated upon disruption of TSC2 in renal tumors from Eker rats. Also, hamartin and tuberin reduce Wnt stimulation of ␤-catenin half-life and its downstream Tcf/LEF transcriptional activation. The results further suggest that tuberin and hamartin associate with the GSK3␤ degradation complex in a Wnt-1-dependent manner. These findings are in contrast to those of Kugoh et al. (25) who reported a lack of change in Tcf-dependent luciferase activities in tuberin-null cells compared with controls. However, these experiments were conducted in the absence of Wnt stimulation.
In our model, TSC1 and TSC2 complex with GSK3␤ to promote ␤-catenin degradation. Upon Wnt stimulation, tuberin dissociates from this complex resulting in stabilization of cytoplasmic ␤-catenin. As a known target of ␤-catenin signaling, the cyclin D1 gene contains Tcf responsive elements in its promoter and its expression is, in part, dependent on ␤-catenin activity (8,9). We postulate that in the absence of functional tuberin or hamartin, stimulation of the ␤-catenin pathway will be unopposed resulting in the sustained transcriptional activation of cyclin D1. However, at this point, we cannot exclude the influence of the PI3K/Akt/mTOR pathway on cyclin D1 regulation. There is growing evidence that there exists substantial cross-talk between the Wnt and PI3K signaling pathways. For example, Wnt stimulation increases Akt activity resulting in GSK3␤ phosphorylation and ␤-catenin stabilization (26). This function of Akt is distinct from its effects on the PI3K pathway and is dependent on the recruitment of Dsh to the GSK3␤-␤catenin-Axin complex (26). It is conceivable that upon stimulation, activated Akt phosphorylates tuberin and disables it from the complex. Accordingly, tuberin acting downstream of Akt may play a role in coordinating signals tranduced through the Wnt and PI3K pathways, thus providing a mechanism for the pleiotropic effects of the TSC1 and TSC2 genes in tuberous sclerosis. The molecular components and their regulation by which TSC1 and TSC2 interact with the GSK3␤ complex remain to be identified.