We hypothesized that the intracellular LRP6 region acts as a direct GSK3β inhibitor in Wnt/β-catenin signal propagation. Within the intracellular region of LRP6, the miniC region of LRP6, especially the PPPSPxS motif, might be the GSK3β inhibitory sequence; replacement of the cytoplasmic domain of LRP6 with this region can induce Wnt/β-catenin signaling even in the absence of the LRP6 extracellular region [12], [13].

To test our hypothesis, we constructed a GFP-fusion protein based on LRP6-miniCL (Fig. 1A), which is similar to the miniC construct (this protein is called GFP-miniCL; Fig. 1). We also constructed a mutant of LRP6-miniCL containing an Ala substitution at Ser1490 as a GFP-fusion construct (GFP-miniCLMT; Fig. 1). We transfected the constructs into A549 (Figs. 1B and C) and HepG2 cells (Figs. S1 and S2) and analyzed the distribution, expression, and phosphorylation state of each construct, and their effects on β-catenin. Both wild type and mutant miniCL were evenly distributed throughout the cytoplasm (GF; Fig. 1B and Fig. S1). However, only wild-type miniCL dramatically increased the total amount of cellular β-catenin and promoted the translocation of β-catenin into the nucleus (β-cat; Fig. 1B and Fig. S1). We consistently observed an increased level of β-catenin with the miniCL transfection in western blot analyses (β-catenin; Fig. 1C and Fig. S2). These observations are consistent with the results of Cselenyi et al., who used the LRP6 cytoplasmic region containing the Ser/Thr rich cluster and five PPPSPxS motifs, and variant in which the first Ser residue of each PPPSPxS motif is replaced by Ala [17]. To confirm the activation of the Wnt/β-catenin signaling pathway, we measured the transcriptional activity of β-catenin using luciferase in 293 cells. Forced cytoplasmic expression of wild type miniCL, but not the mutant miniCLMT, significantly induced TOP-flash activity (Fig. 1D). These results suggest that the LRP6 miniCL works as an inhibitor of GSK3β regardless of its cellular localization, and that the PPPSPxS motif plays a crucial role in this effect.

The wild-type miniCL, but not the mutant miniCLMT, also suppressed phosphorylation of glycogen synthase (GS; p-GS in Fig. 1C) without the induction of phosphorylation of Ser9 in GSK3β (p-S9-GSK3β; Fig. 1C). Suppression of GS phosphorylation normally occurs when GSK3β is phosphorylated at Ser9 by AKT as a part of the insulin/IGF-1 signaling pathway [18]. Our data indicate that the miniCL protein can inhibit GSK3β independently of AKT activity when this LRP6 fragment is overexpressed in the cytoplasm, implying that it can act as a general inhibitor of GSK3β. Note that under physiological conditions, such inhibition would not occur due to the localization of the LRP6/GSK3β/Axin complex at the plasma membrane (see below).

Since phosphorylation of the PPPSPxS motif in the miniCL region of LRP6/5 is correlated with the inhibition of GSK3β, we asked whether the PPPSPxS motif can directly inhibit GSK3β in vitro. We synthesized one dually-phosphorylated (NPPPpSPApTERSH, where pS or pT designates a phosphorylated Ser or Thr residue), two singly-phosphorylated (NPPPSPApTERSH and NPPPpSPATERSH) and one non-phosphorylated (NPPPSPATERSH) peptides derived from the first PPPSPxS motif of LRP6 (Fig. 1A). To measure the GSK3β activity toward β-catenin in vitro, we utilized a β-catenin N-terminal fragment (residues 1–133), whose Ser45 was prephosphorylated by CK1, as a substrate (Ha et al., 2004). Each synthetic peptide was added to a reaction mixture containing the purified GSK3β enzyme and the substrate β-catenin N-terminal fragment. The GSK3β activity was determined through measurement of the intensity of the supershifted (retarded) β-catenin bands (Fig. S3), which represent the phosphorylated product species (Fig. S4), on an SDS-polyacrylamide gel (Figs. S3, S4, S5, S6, S7). The dually-phosphorylated peptide showed a strong inhibitory effect on GSK3β in a concentration-dependent manner (1–100 µM), while the non-phosphorylated peptide did not show any inhibitory effect on the GSK3β activity at concentrations up to 100 µM (Fig. 2A). The two singly-phosphorylated peptides showed distinct inhibitory effects. Whereas the peptide phosphorylated at the second Ser/Thr residue [PPPSPx(p)S; Fig. 2A] exhibited only a marginal inhibitory effect at 100 µM, the peptide phosphorylated at the first Ser/Thr [PPP(p)SPxS; Fig. 2A] exhibited a strong inhibitory effect close to that of the dually-phosphorylated peptide (Fig. 2A). Our results demonstrate that the dually-phosphorylated PPPSPxS motif directly inhibits GSK3β, and that phosphorylation at the first Ser residue plays a more important role. This corresponds well with the results of Zeng et al., who reported that the first Ser residue in the PPPSPxS motif has a more important role in Wnt/β-catenin signaling than the second Ser residue [13]. Since the non-phosphorylated peptide does not inhibit, the inhibition is not due to competition between this potential alternative substrate and the β-catenin substrate.

Based on steady-state kinetic analyses [19], the dually phosphorylated LRP6 peptide competitively inhibits GSK3β with an apparent inhibition constant (K_i) of 13 µM. Thus, the phosphorylated LRP6 motif is a much stronger inhibitor of GSK3β than the phosphorylated N-terminal fragment of GSK3β (K_i ∼700 µM), which is a known pseudosubstrate inhibitor of GSK3β [20] (Fig. 2B).

It was noted above that cytoplasmically expressed miniCL inhibits phosphorylation of GS. To further test whether the dually-phosphorylated PPPSPxS peptide can act as a general inhibitor of GSK3β, we tested its effect on GSK3β-mediated phosphorylation of an Axin fragment (mouse Axin amino acids 512–650), in which Ser614 is the only site phosphorylated by GSK3β in vitro and in vivo [21]. As shown in Fig. 2C, the phosphorylated PPPSPxS peptides inhibited GSK3β-mediated phosphorylation of the Axin fragment as efficiently as on the primed β-catenin N-terminal fragment substrate. This result is consistent with the observation that the intracellular region of LRP6 decreased GSK3β phosphorylation of an unprimed substrate Tau protein and the primed substrate full-length β-catenin in vitro [16]. In addition, the phosphorylated PPPSPxS peptides did not show any inhibitory effect on CK1 when we use the unphosphorylated β-catenin fragment as a substrate (data not shown), confirming that the phosphorylated PPPSPxS peptide inhibits only GSK3β. Collectively, these data suggest that the phosphorylated PPPSPxS motif of LRP6/5 can act as a general inhibitor of GSK3β in vitro.

We next sought to determine if the phosphorylated LRP6 motifs are capable of inhibiting GSK3β in the Axin complex. An in vitro kinase assay with Axin-bound GSK3β was performed by adding the purified GSK3β binding domain (GBD) of Axin to the reaction mixture. Although full-length Axin increases GSK3β activity toward full-length β-catenin through its scaffolding of GSK3β and β-catenin [22], the Axin GBD domain itself slightly inhibited GSK3β (Fig. 2D), which is consistent with the reduced activity of Fratide-bound GSK3β [23]. As shown in Fig. 2D, the dually-phosphorylated PPPSPxS peptide inhibits Axin-bound GSK3β similarly to free GSK3β. This result is consistent with the crystal structure of GSK3β complexed with Axin, which shows that Axin does not occupy the active site of GSK3β [22]. The in vitro inhibition of Axin-bound GSK-3β indicates that the phosphorylated PPPSPxS motifs of LRP6/5 can be an in vivo inhibitor of GSK3β in Wnt/β-catenin signaling.

It is well established that Wnt signaling reduces GSK3β activity during development of the Xenopus embryo [24], and that the suppressed activity of GSK3β induces a second body axis in the Xenopus embryo in the response to Wnt [25]. To examine the in vivo effect of the PPPSPxS motif on Wnt/β-catenin signaling, we injected the synthetic peptides into Xenopus embryos, along with a small amount of Xenopus Wnt8 that is insufficient to induce second body axis formation on its own. The dually-phosphorylated peptide [PPP(p)SPx(p)S; Fig. 3] strongly induced the second body axis, whereas the non-phosphorylated peptide [PPPSPxS; Fig. 3] did not. The peptide phosphorylated at only the first Ser residue [PPP(p)SPxS; Fig. 3] induced the second body axis with an efficiency comparable to that of the dually-phosphorylated peptide. This result indicates that the phosphorylated PPPSPxS motif can induce Wnt signaling in vivo. Although the amount injected is likely to be higher than physiological concentrations, it is interesting that microinjection of each peptide in the absence of Xenopus Wnt8 did not induce a second body axis (data not shown). It is likely that a slight reduction in the GSK3β activity due to the activity of the exogenous Wnt on the endogenous LRP6/5 is required for the peptide to cause a sufficient inhibition of GSK3β to trigger Wnt/β-catenin signaling. Moreover, even if the amount of peptide injected is well above physiological concentrations, the effect is specific to the phosphorylated peptide. Thus, the peptide injection experiments, when considered with the experiments using purified proteins, indicate that the phosphorylated PPPSPxS motif directly inhibits GSK3β and leads to the activation of β-catenin.

It was recently reported that injection of a recombinant portion of the LRP6 cytoplasmic region containing the Ser/Thr rich cluster and the five PPPSPxS motifs induced a second body axis as effectively as the phosphorylated PPPSPxS peptide in the absence of the exogenously-added Wnt ligand, even when the LRP6 protein was not prephosphorylated by GSK3β and CK1 [17]. It seems likely that the longer fragment of the LRP6 cytoplasmic region is phosphorylated by endogenous kinases in the embryo, and has a stronger inhibitory effect on GSK3β due to the multiple PPPSPxS motifs. Also, as we show below, the Ser/Thr-rich region upstream of the PPPSPxS motifs, may have an important role in this process.

The evidence above suggests that the phosphorylated PPPSPxS motifs of the LRP6/5 intracellular region competitively inhibit GSK3β. The intracellular region of LRP6/5 has been implicated in binding both GSK3β and Axin [6], [12], [13], [16]. Whereas GSK3β bound to LRP6 or 5 in the absence of Axin in yeast two-hybrid experiments [13], [16], GSK3β dramatically increased the interaction of Axin with LRP6 or 5 in cells and in vitro [6]. Moreover, both the GBD and DIX domains of Axin were present in constructs that interacted with LRP5 in yeast two-hybrid experiments [6]. The Axin-binding site found in vertebrate GSK3 [22] is conserved in the yeast homolog. These observations suggest that GSK3 can associate directly with LRP6 and bridge LRP6 to Axin.

We first investigated whether the intracellular region of LRP6 binds to GSK3β in a phosphorylation-dependent manner. To detect binding, we used recombinant GST-fused LRP6 fragments (miniCL, miniCLMT, PPPSPxS, PPPAPxS; Fig. 1A) and GSK3β proteins in a far western binding assay. This assay is appropriate as the LRP6 cytoplasmic region is expected to be natively unstructured and therefore should not suffer from denaturation artifacts. Prior to the binding reaction, the GST-fused LRP6 fragments were phosphorylated by GSK3β and CK1 and transferred to a PVDF membrane. The membrane was incubated with GSK3β, and bound GSK3β was detected with an anti-GSK3β antibody. GSK3β bound to GST-miniCL and GST-miniCLMT, but not GST-PPPSPxS or GST-PPPAPxS (Fig. 4). Comparison with LRP6 that had not been phosphorylated revealed that phosphorylation enhanced the binding by four to five fold (Fig. 4A Right). Since the S1490A mutation does not affect binding, these results suggest that the Ser/Thr rich cluster present in miniCL plays more important role in mediating the phosphorylation-dependent interaction of LRP6 with GSK3β. Also, given the measured K_i value of the single PPPSPxS motif, it was expected that the single PPPSPxS motif might exert a limited role in recruiting the GSK3β.

To investigate whether the Axin-bound GSK3β is able to bind to the phosphorylated LRP6 fragments (miniCL), we incubated GSK3β with excess Axin GBD protein to saturate its binding to GSK3β and then tested the binding of GSK3β and LRP6 using the same far western blot analysis (Fig. 4B). Binding of GSK3β to the LRP6 fragment was not affected by the presence of Axin, indicating that the Axin GBD, GSK3β, and LRP6 miniCL proteins form a ternary complex in vitro. Furthermore, we observed that an Axin fragment including the GBD was bound to the LRP6 miniCL fragment only when GSK3β was co-incubated (Fig. 4C), demonstrating that GSK3β mediates the association of the Axin fragment with the LRP6 intracellular region. Taken together, we propose that the GSK3β plays an important role in mediating Axin binding to LRP6/5 in vivo mainly through the Ser/Thr rich cluster of LRP6/5.

DNA fragments encoding LRP6 miniCL (residues 1470–1510 of human LRP6) were obtained from a human cDNA library by PCR, and inserted into the pEGFP-C1 vector (Clontech) using HindIII and BamHI sites of to generate GFP-tagged protein, resulting in pEGFP-miniCL. The LRP6 mutant (S1490A) was generated by PCR-mediated mutagenesis, resulting in pEGFP-miniCLMT.

Antibodies were used according to the manufacturer's instructions. Antibodies used include mouse monoclonal anti-β-catenin (Santa Cruz), anti-GFP (Santa Crutz), and rabbit polyclonal anti-GSK3β (Cell signaling), anti-phospho GSK3β (Ser9) (Cell signaling), anti-phospho LRP6 (Ser1490) (Cell signaling), anti-phospho-glycogen synthase (Ser641) (Cell signaling), anti-GS (Cell signaling), anti-phospho-β-catenin (Ser45) (Cell signaling), and anti-phospho-β-catenin (Ser33/37/Thr41) (Cell signaling).

pEGFP-miniCL and pEGFP-miniCLMT vectors were transfected into A549 cells and HepG2 cells using jetpei according to the manufacturer's protocol (Polyplus transfection). In brief, 2 µg of DNA and jetpei mixtures were added to A549 cells and HepG2 cell and incubated for 3 h under serum-free conditions, and then equal volumes of DMEM containing 20% serum were added. After 24 h incubation, cells were harvested using RIPA (containing protease cocktail). Subsequently, 20 µg of protein lysates were applied to SDS-PAGE. After transferring the proteins to a PVDF membrane, samples were incubated with the appropriate antibodies, according to standard western blot protocol.

To address the transcriptional activity of β-catenin, TOP-flash (responsive luciferase vector) and FOP-flash (unresponsive vector), which were kindly provided by Drs. B. Vogelstein and K. Kinzler (Johns Hopkins Univ.), were transfected into 293 cells. Using 10 µl of cell lysate, we performed the luciferase assay using the luciferase assay kit (Promega).

Transfected cells were fixed with 100% methanol for 20 min at −20°C. After washing with PBS and subsequent blocking with blocking buffer (PBS+0.05% BSA), fixed cells were incubated with anti-β-catenin antibody for 2 h at room temperature. After washing with PBS, cells were incubated with a rhodamine-conjugated anti-mouse antibody for 2 h. To visualize nuclei, cells were stained with DAPI for 5 min. After washing with PBS, expression of β-catenin, GFP, and DAPI were analyzed using fluorescence microscopy.

The catalytic domains of mouse GSK3β (residues 27–393) and CK1ε (residues 1–319) were expressed and purified as previously described [32]. The N-terminal region (residues 1–133) of β-catenin was expressed in Escherichia coli as a GST-fusion protein with a TEV protease cleavage site and then purified using Glutathione-agarose. Subsequently, the GST tag was cleaved using recombinant TEV protease and was removed by repeated HiTrapQ anion-exchange chromatography. The N-terminal region of β-catenin was then phosphorylated by incubation with the catalytic domain of CK1ε (0.28 mg/ml) at 37°C for 20 min in 50 mM Tris buffer (pH 8.0) containing 10 mM MgCl₂, 10 mM 2-mercaptoethanol, and 2 mM phenylmethylsulfonylfluoride. The phosphorylated β-catenin fragment was then further purified using HiTrapQ anion-exchange chromatography to remove the CK1ε enzyme. The mouse Axin GBD (residues 512–530), and variants of human LRP6 (miniCL, residues 1470–1510; miniCLMT, residues 1470–1510 with substitution at Ser1490 with Ala; LRP6 PPPSPxS, residues 1485–1497; and LRP6 PPPAPxS, residues 1485–1497 with the substitution at Ser1490) were expressed in E.coli as GST-fusion proteins and purified using Glutathione-agarose and a HiTrapQ anion-exchange chromatographic column. The final buffer for the GST-fusion proteins was changed to 20 mM Tris buffer (pH 8.0) using Centriprep (10 kDa cutoff; Millipore). His-tagged Axin fragment (512–650 in the mouse Axin numbering) was expressed in E.coli, and purified using Ni-NTA affinity and HitrapQ anion exchange chromatographic columns.

Peptides for inhibition assays were prepared by solid-phase synthesis. The amino acid sequence of each peptide is shown in Fig. 1A. The N-terminus and C-terminus of all peptides were modified by biotinylation and amidation, respectively. Biotinylation was accomplished using sulfo-NHS-SS-biotin (Pierce), which allowed reductive cleavage of the 15-mer from the biotin group. The identity and purity of the synthesized peptides were confirmed by mass spectrometry.

To measure the GSK3β activity in the presence of each synthetic peptide dissolved in 10% dimethylsulfoxide, 7.5 µl of the reaction mixture was incubated for 30 min at 37°C. This reaction mixture contained 50 mM Tris (pH 8.0), 10 mM MgCl₂, 10 mM 2-mercaptoethanol, 1 mM ATP, 2 ng of recombinant GSK3β catalytic domain, 7 µM of the primed β-catenin N-terminal fragment, and the same volume (0.3 µl) of varying concentrations of each synthetic peptide. After the reaction was stopped by the addition of SDS-loading buffer and immediate boiling, the resultant mixture was applied to a 15% SDS-polyacrylamide gel for analysis. As shown in Fig. S3, S4, S6, and S7, the intensities of the supershifted bands (substrate, product a, product b) on the Coomassie-stained gel were integrated using the program IMAGEQUANT (Molecular Dynamics, USA). To calculate the GSK3β activity, the intensities of the two product bands were combined according to the number of phosphate groups incorporated by GSK3β. The lower band (product a) contains one phosphate group (×1), while the upper (product b) band contains two or three phosphate groups (×2.5). To examine the effect of Axin GBD on the inhibitory role of the dually-phosphorylated peptide toward GSK3β activity, each concentration of GST-fused Axin GBD protein was added in 7.5 µl of the reaction mixture with GSK3β.

Thirty µl of the reaction mixture containing 50 mM Tris (pH 8.0), 10 mM MgCl₂, 10 mM 2-mercaptoethanol, 1 mM ATP, 8 ng of recombinant GSK3β catalytic domain, 6 µM of the primed β-catenin N-terminal fragment, and varying concentrations of dually-phosphorylated peptide were incubated at 37°C. At each time point, 7.5 µl was taken from the reaction mixture for SDS-PAGE. The GSK3β activity was measured using a Coomassie-stained SDS-polyacrylamide gel, as described above. The initial parts of the activity curves, which show no curvature, were obtained and their slopes were calculated as initial velocities. The apparent K_i value was determined according to the equation: (V₀/V_i)−1 = [peptide]/K_i, where V₀ and V_i are the initial velocities of the reaction in the absence and in the presence of the dually-phosphorylated peptide [19].

Xenopus laevis was purchased from Xenopus I and Nasco. Eggs were obtained from Xenopus laevis primed with 800 U of human chorionic gonadotropin (Sigma). In vitro fertilization was performed as described previously [33], and the developmental stages of the embryos were determined according to Nieuwkoop and Faber [34]. Microinjection was carried out in 0.33×Modified Ringer's (MR) containing 4% Ficoll-400 (GE healthcare) using a Nanoliter Injector (WPI). Injected embryos were cultured in 0.33×MR until stage 8 and then transferred to 0.1×MR until they had reached the appropriate stage.

To phosphorylate GST-fusion proteins, 15 µg of each GST-fusion protein was incubated for 4 h at 37°C with 0.2 µg of recombinant GSK3β and 0.4 µg of recombinant CK1ε in 10 µl of 50 mM Tris (pH 8.0) buffer containing 5 mM ATP, 10 mM MgCl₂ and 10 mM 2-mercaptoethanol.

GST-fusion proteins (GST-miniCL, GST-miniCLMT, GST-PPPSPxS, GST-PPPAPxS) were subjected to SDS-PAGE and transferred to a PVDF membrane. The membrane was then blocked for 1 h with 5% skim milk in TBST (20 mM Tris (pH 7.6), 137 mM NaCl, 0.05% Tween 20), and rinsed with TBST. Then the membrane was incubated for 3 h with the 1 µg of the recombinant GSK3β in 1 ml of 5% skim milk in TBST, and was washed 3 times with TBST. The binding of GSK3β was detected with rabbit anti-GSK3β antibody and secondary HRP-conjugated goat anti-rabbit IgG (Pierce).