Negative Regulation of LRP6 Function by Casein Kinase I ϵ Phosphorylation*

Wnt signaling acts in part through the low density lipoprotein receptor-related transmembrane proteins LRP5 and LRP6 to regulate embryonic development and stem cell proliferation. Up-regulated signaling is associated with many forms of cancer. Casein kinase I ϵ (CKIϵ) is a known component of the Wnt-β-catenin signaling pathway. We find that CKIϵ binds to LRP5 and LRP6 in vitro and in vivo and identify three CKIϵ-specific phosphorylation sites in LRP6. Two of the identified phosphorylation sites, Ser1420 and Ser1430, influence Wnt signaling in vivo, since LRP6 with mutation of these sites is a more potent activator of both β-catenin accumulation and Lef-1 reporter activity. Whereas Wnt3a regulates CKIϵ kinase activity, LRP6 does not, placing CKIϵ upstream of LRP6. Mutation of LRP6 Ser1420 and Ser1430 to alanine strengthens its interaction with axin, suggesting a mechanism by which CKIϵ may negatively regulate Wnt signaling. The role of CKIϵ is therefore more complex than was previously appreciated. Generation of active CKIϵ may induce a negative feedback loop by phosphorylation of sites on LRP5/6 that modulate axin binding and hence β-catenin degradation.

Wnts are secreted proteins that regulate morphogenesis by establishing cell polarity and stimulate cell proliferation and cell migration in response to Wnt concentration gradients in the embryo (1,2). Wnt signaling is important both in development and adulthood, particularly for its role in regulating proliferation and differentiation of stem cells. Understanding Wnt signaling is also of interest due to its up-regulation in many cancers (2)(3)(4)(5).
Wnt signal transduction is initiated when the ligand interacts with two integral membrane proteins: a member of the frizzled family of seven pass transmembrane proteins and one of two co-receptors, the low density lipoprotein receptor-related proteins LRP5 and LRP6. This interaction recruits axin to LRP5/6 and leads to inhibition of ␤-catenin phosphorylation by GSK3 (6 -8). Consequently, ␤-catenin is not targeted to the proteasome and instead accumulates and translocates to the nucleus, where it interacts with TCF/LEF proteins to stimulate transcription of Wnt-responsive genes. Depending on cell type and type of receptors present, distinct Wnts can also activate additional signaling pathways, including cell migration via RhoA, and changes in calciummediated signaling (9 -11).
The casein kinase I family plays both a positive and negative role in Wnt-␤-catenin signaling. CKI␣ 4 interacts with axin and phosphorylates serine 45 of ␤-catenin. CKI␣ is a potent negative regulator of ␤-catenin, and its inhibition (e.g. by RNA interference) leads to decreased phosphorylation and marked accumulation of ␤-catenin, leading to activation of downstream signaling (12)(13)(14). In overexpression screens, other members of the CKI family, CKI⑀ and the closely related CKI␦, were found to be positive regulators of the pathway, acting upstream of axin and GSK3 and stabilizing ␤-catenin (15)(16)(17)(18)(19). How CKI⑀ regulates ␤-catenin accumulation is less clear. CKI⑀ phosphorylates multiple proteins in the ␤-catenin degradation complex, including axin, Dvl (disheveled), and the APC (adenomatous polyposis coli) protein, and has previously been shown to destabilize the destruction complex, potentially through phosphorylation of axin or disheveled (19,20). Notably, the kinase activity of CKI⑀ is also regulated by Wnt. In resting cells, both endogenous and overexpressed CKI⑀ are inhibited by autophosphorylation of their own carboxyl-terminal autoinhibitory domain. The addition of Wnt ligand leads to rapid dephosphorylation and activation of CKI⑀. Similarly, withdrawal of Wnt ligand rapidly induces phosphorylation and inactivation of CKI⑀. The relevant substrates of activated CKI⑀ are not known. Interestingly, overexpression of a constitutively active variant of CKI⑀ (CKI⑀ MM2) only modestly increases ␤-catenin stabilization compared with wild type CKI⑀, suggesting that activated wild type CKI⑀ may function in downstream events distinct from its role in activation of canonical Wnt signaling (21,22). Such events may stem from the ability of wild type CKI⑀ to autocatalytically regulate its activity.
LRP5 and LRP6 are co-receptors that bind Wnts in conjunction with members of the frizzled receptor family and are required to induce ␤-catenin stabilization and canonical Wnt signaling (23)(24)(25). The LRPs are single-pass transmembrane proteins with an extended extracellular domain with sequence similarity to the low density lipoprotein receptor family and a unique intracellular domain. Exactly how these receptor coreceptor pairs transmit their signal across the membrane is not fully understood. Truncation mutants of LRP5/6 that lack their extracellular domain are constitutively active in Wnt signaling, suggesting that this domain plays an inhibitory role against signaling in the absence of Wnt (6 -8). One known requirement for signaling is Wnt-responsive phosphorylation of a series of five highly conserved PPP(S/T)P (PPPSP) motifs in the intracellular domain. Only a single copy of this motif is required for signaling, and mutation of the S/T residues in all five PPPSP motifs eliminates LRP6 ability to signal. Phosphorylation of the PPPSP motifs stimulates axin binding (6,8,26,27). It is postulated that this recruitment of the axin complex to the membrane via phosphorylated LRP5/6 protects ␤-catenin from phosphorylation events that target it for destruction (8,28).
Here we investigate the ability of CKI⑀ to interact with and phosphorylate LRP5 and LRP6. CKI⑀ but not CKI␣ associates with and phosphorylates the intracellular domain of LRP5/6 in vitro and in vivo. Using in vitro phosphorylation and mass spectroscopy, two juxtamembrane serines distinct from previously identified CKI sites in LRP6 were identified as CKI⑀ substrates. Mutation of these novel CKI⑀ target sites significantly stimulates both LRP6-mediated ␤-catenin accumulation and downstream signaling and is associated with increased binding of axin to LRP6. These findings are consistent with our previous finding that CKI⑀ expression destabilizes the ␤-catenin destruction complex and demonstrate that CKI⑀ can both positively and negatively regulate the canonical Wnt pathway.
Monoclonal antibodies anti-Myc (9E10) and anti-HA (12CA5) were obtained from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA). Mouse monoclonal antibodies against ␤-catenin and CKI⑀ were from Transduction Laboratories. Rabbit polyclonal anti-VSVG antibody was obtained from Covance Research Products. The anti-CKI⑀/␦ affinitypurified rabbit polyclonal antibody UT31 has been described previously and recognizes the shared extreme N terminus of CKI⑀ and CKI␦ (30). Donkey anti-rabbit IgG-horseradish peroxidase and goat anti-mouse IgG-horseradish peroxidase were from Pierce. Protein A-agarose was from Invitrogen. Thioredoxin-tagged CKI⑀⌬319 (lacking the autoinhibitory C-terminal 111 amino acids) and His 6 -CKI⑀ were expressed in Escherichia coli and purified as described (31,32). Protease inhibitor mixture tablets were obtained from Roche Applied Science. protein phosphatase was obtained from Upstate Biotechnology, Inc. (Lake Placid, NY).
Cell Culture and Transfection-Human embryonic kidney 293 (HEK 293), 293T (HEK 293T), and HEK 293 STF cells (293 cells with an integrated "Super-Top Flash" TCF-luciferase reporter (33); a gift of K. Zhang) were grown in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% heat-inactivated fetal bovine serum and 100 IU/l penicillin and 100 g/ml streptomycin (Invitrogen). Cells were maintained in a humidified incubator at 37°C and 5% CO 2 . For transient transfections, cells were seeded in individual wells of a 6-well dish treated with poly-L-lysine (Sigma). At 60 -80% confluence, cells were transfected using Lipofectamine and Plus reagent (Invitrogen) according to the manufacturer's instructions. Total transfected DNA per well was adjusted to 1 g with empty vector. For immunoprecipitation experiments, 500 ng each of LRP5/6, CKI⑀, CKI␣, or Axin expression plasmids were used. For kinase-induced mobility shift experiments, 500 ng of LRP6 expression constructs and 50 -500 ng of CKI⑀, CKI␣, and GSK3␤ were transfected. 48 -64 h after transfection, IC261 (50 g/ml in Dulbecco's modified Eagle's medium) was added where indicated for up to 5 h prior to cell lysis. After IC261 treatment, growth medium was aspirated, and the transiently transfected cells were washed twice with phosphate-buffered saline and lysed in 200 l of lysis buffer. One of two types of lysis buffer was used with equivalent results: buffer I (0.1% Nonidet P-40, 150 mM NaCl, 20 mM HEPES, pH 7.5, 1 mM EDTA, 2 mM dithiothreitol) or buffer II (1% Nonidet P-40, 0.1% SDS, 150 mM NaCl, 50 mM HEPES, pH 8.0, 1 mM EDTA, 1 mM dithiothreitol). Both were supplemented with 1ϫ Complete protease inhibitor mixture and phosphatase inhibitors (1 mM sodium fluoride, 1 mM ␤-glycerolphosphate, 1 mM sodium orthovanadate). Cells were then mechanically sheared, and the lysates were centrifuged at 18,000 ϫ g for 10 min at 4°C.
Immunoprecipitation and Western Blot-Protein concentrations in lysates were determined by Bradford assay using bovine serum albumin as a standard. Total protein (0.3-1.0 mg) and lysate volumes were then made equivalent, and VSVG-, HA-, and Myc-tagged proteins were immunoprecipitated with 1 g of anti-VSVG antibody, 2 g of 12CA5, or 2 g of 9E10 antibody. Cell lysates and antibodies were incubated in lysis buffer (400-l final volume) for 1 h on ice to allow formation of antigen-antibody complexes. Samples were then incubated with 15 l of protein A-agarose beads (equilibrated in lysis buffer) for 1 h or overnight at 4°C with gentle agitation. Finally, the beads were collected by 1 min of centrifugation at 100 ϫ g in a tabletop centrifuge (4°C) and washed four times with lysis buffer. SDS-PAGE and immunoblotting was performed as previously described (21,34). Western blotting with an anti-actin antibody (Sigma) was performed to ensure that equal amounts of protein were analyzed in each sample.
In Vitro Kinase Assay-MBP-tagged LRP6 ⌬N (wild type and mutants) were affinity-purified after expression in E. coli. 2 g of purified protein was added to kinase buffer (25 mM Tris-HCl, pH 7.5, 20 mM NaF, 1 mM dithiothreitol, and 150 M ATP) with or without recombinant CKI⑀(⌬319). The kinase reactions were incubated at 30°C for 1 h and stopped by the addition of Laemmli sample buffer. Proteins were resolved on 10% acrylamide gels. After protein transfer to nitrocellulose membranes, MBP-LRP6 was visualized by Western blotting using anti-MBP antibody. IC261 was added when indicated.
In-gel Digestion and Mass Spectrometry Analysis-SDS-polyacrylamide gel bands of phosphorylated MBP-LPR6 were sliced into 1-mm cubes and digested with 12.5 ng/l trypsin or chymotrypsin (Roche Applied Science), and peptides were extracted according to standard procedures (35). The samples were brought up to 20 l with 0.1% acetic acid for LC/MS/MS analysis using conventional C18 liquid chromatography. Each gel band solution sample was loaded onto a 360-m outer diameter ϫ 75-m inner diameter microcapillary fused silica tubing packed with C18 irregular 5-20-m-sized resin (Polymicro Technologies, Phoenix, AZ). The precolumn was then connected to a 360-m outer diameter ϫ 50-m inner diameter analytical column (Polymicro Technologies) packed with C18 regular 5-m-sized resin constructed with an integrated electrospray emitter tip (36). Additionally, immobi-lized metal affinity chromatography was utilized to enrich for phosphopeptides (37). Samples were then analyzed by gradient elution (Agilent 1100 series, Santa Clara, CA) directly into a Finnigan LTQ quadrupole ion trap mass spectrometer (Thermo Electron, San Jose, CA) at a flow rate of 60 nl/min. The nanoflow HPLC gradient used was 0 -60% acetonitrile in 0.1% acetic acid in 60 min. All MS/MS data were searched using the Sequest TM program and validated manually.
Determination of Cytoplasmic ␤-Catenin Levels-Immunoblot analysis of cytosolic ␤-catenin was carried out following standard methods (17) to evaluate the extent of Wnt pathway activation in response to LRP6 expression. HEK 293 cells seeded in individual wells of a 6-well dish were transfected with 25-250 ng of DNA expressing LRP6. DNA amounts were equalized with empty plasmid so that 1 g of total DNA was transfected. At 48 h after transfection, the cells were washed once with phosphate-buffered saline and once with hypotonic lysis buffer (40 mM HEPES, pH 7.5, 1 mM EDTA, 1 mM EGTA). 200 l of lysis buffer (supplemented with proteinase and phosphatase inhibitors) was added, and cells were incubated on ice for 10 min. Cells were then collected and mechanically sheared, and lysates were clarified by centrifugation at 15,700 ϫ g for 10 min at 4°C. 12.5 g of supernatant was analyzed for ␤-catenin levels by immunoblotting.
LEF1-luciferase Reporter Assay-pCS2-VSVG-LRP6, pCS2-VSVG-LRP6 ⌬N, and their point mutants were transfected (25-100 ng) into 293 STF cells in 6-well dishes along with a Renilla luciferase control plasmid (2 ng). pCS2 empty vector was used to equalize total DNA per well. At 46 h post-transfection, cells were harvested with passive lysis buffer (Dual-Luciferase Reporter assay system; Promega), and firefly and Renilla luciferase activity were measured using the Dual-Luciferase Reporter assay system (Promega) in a microtiter plate luminometer (Dynex Technologies) according to the manufacturer's instructions. To standardize for transfection efficiency, firefly luciferase activity was normalized to Renilla luciferase activity. The -fold increase of Lef-1-dependent activity was calculated by dividing each activity by that observed from cells transfected with empty pCS2 plasmid. Data are presented as mean Ϯ S.D. of experiments performed in triplicate.

RESULTS
CKI⑀ Interacts with LRP5 and LRP6 in Vivo-CKI⑀ and the Wnt coreceptors LRP5 and -6 are both positive regulators of Wnt signaling. To investigate whether CKI⑀ acts directly on LRP5/6 to regulate Wnt signaling, the ability of the proteins to interact in vivo was assessed. In reciprocal co-immunoprecipitation experiments, CKI⑀ interacted with both LRP5 and LRP6 (Fig. 1, A and B). The interaction with CKI⑀ was specific, since antibodies against LRP5/6 did not precipitate CKI⑀ in the absence of LRP5/6, and antibodies against tagged CKI⑀ did not precipitate LRP5/6 in the absence of kinase. Furthermore, no interaction of LRP5/6 with CKI␣ was detected (Fig. 1C). The interaction between CKI⑀ and LRP5/6 did not appear to require prior phosphorylation of the substrate by CKI⑀, since inclusion of the CKI⑀ inhibitor IC261 had no effect on the interaction (Fig. 1D), and kinase-inactive CKI⑀ K38A bound equally as well as wild type CKI⑀ (data not shown).
CKI⑀ Interacts with and Phosphorylates LRP6 in Vitro-Visual analysis of the sequence of the cytoplasmic domains of LRP5/6 revealed a potential CKI⑀ phosphorylation consensus site after each of the five PPPSP motifs. The consensus site pSXX(S/T) (where pS represents phosphoserine) (38) has been found in the human circadian clock protein PER2 and in the transcription factor NFAT 1 (39,40). To assess the ability of CKI⑀ to phosphorylate LRP6 and to further characterize the CKI⑀/ LRP6 interaction in vitro, the cytoplasmic domain of LRP6 (LRP6 ⌬N)(containing amino acids 1394 -1613 of human LRP6) was expressed and puri-fied from E. coli as an MBP fusion protein. We also expressed and purified MBP-LRP6 ⌬N(M5), a mutated version of LRP6 ⌬N in which each of the five PPPSP sites have been mutated to PPPAP. Both LRP6 ⌬N and LRP6 ⌬N M5 showed similar concentration and time-dependent decreases in electrophoretic mobility when incubated with purified CKI⑀(⌬319) (Fig. 2, A and B). (This truncated kinase is constitutively active due to deletion of the autoinhibitory carboxyl terminus). Quantitation of the stoichiometry of in vitro phosphorylation shows that CKI⑀ adds ϳ3-4 mol of phosphate/mol of MBP-LRP6 when the reaction is driven to completion (Fig. 2B), suggesting that the interaction between CKI⑀ and LRP6 is stable and several sites are phosphorylated.
In control experiments, MBP alone did not interact with CKI⑀ and was not a substrate (data not shown). Since LRP6 ⌬N M5 was phosphorylated with kinetics and stoichiometry similar to those of the wild type protein in vitro, this strongly suggested that CKI⑀ could not initiate phosphorylation of the PPPSP motifs. (Subsequent studies have indicated that LRP6 primed by GSK3 phosphorylation at the PPPS/TP site can be phosphorylated by CKI␥ (26, 27)). The in vivo co-immunoprecipitation results combined with the in vitro phosphorylation data suggest a strong and direct interaction between CKI⑀ and the cytoplasmic domain of LRP6. To directly test this, MBP-LRP6 ⌬N was incubated with CKI⑀, and the interaction was assessed after precipitation of the MBP with amylose beads. As Fig. 2C shows, bacterially expressed and purified CKI⑀ ⌬319 both interacted with and phosphorylated bacterially expressed LRP6 ⌬N, indicating that the interaction is direct and that the carboxyl terminus of CKI⑀ is not required for the interaction. The CKI⑀-dependent mobility shift was dependent on kinase activity, since it was completely blocked by inclusion of the CKI⑀ inhibitor IC261 (Fig. 2D).
Further in vitro phosphorylation assays of LRP6 ⌬N and LRP6 ⌬N M5 with GSK3␤, protein kinase A, or casein kinase II demonstrated a distinct and much less pronounced kinase-dependent electrophoretic mobility shift (Fig. 2E). Consistent with the previous suggestion that GSK3 may be the kinase that activates LRP5 (8), the M5 mutation abrogated the GSK3-induced shift. Whereas the cytoplasmic domain of LRP6 underwent a smaller mobility shift after incubation with either protein kinase A or casein kinase II, the shift was not altered by mutation of the PPPSP motifs. Furthermore, incubations of LRP6 ⌬N and the M5 derivative with CKI⑀ combined with GSK3, protein kinase A, or casein kinase II produced no obvious supershift relative to that produced by CKI⑀ alone (Fig. 2C), indicating that the gel mobility shift is due to CKI⑀ alone and is not dependent upon prior priming phosphorylation of LRP6. The lack of detectable supershifting also indicates that CKI⑀ does not visibly prime LRP6 for phosphorylation by GSK3, protein kinase A, or casein kinase II.

LRP6 Is Phosphorylated in Vivo in a CKI⑀-dependent Manner-The preceding studies established that CKI⑀ and the cytoplasmic domain of LRP6 interact in vitro and in vivo and that LRP6 is a substrate for CKI⑀ in vitro.
To test if CKI⑀ also mediated LRP6 phosphorylation in vivo, cells were transfected with plasmids expressing VSVG-tagged LRP6 ⌬N (containing amino acids 1370 -1613 of human LRP6), alone or together with plasmids expressing CKI␣, CKI⑀, or GSK3␤. Lysates were then probed with anti-VSVG antibodies. Lysates from CKI⑀-transfected cells displayed multiple mobility-shifted bands corresponding to variously phosphorylated species of LRP6 ⌬N (Fig. 3A). Co-expression of LRP6 ⌬N with CKI␣ and GSK3␤ did not induce detectable mobility shifts. The mobility shift of VSVG-LRP6⌬N was dependent on the amount of transfected CKI⑀ plasmid and was blocked by the CKI⑀ inhibitor IC261 (Fig. 3B).
Identification of Phosphorylation Sites-Taken together, the data suggest that CKI⑀ phosphorylates LRP6 both in vivo and in vitro on sites distinct from the previously identified PPPSP motif. To identify specific CKI⑀ phosphorylation sites, LRP6 ⌬N was phosphorylated in vitro with recombinant CKI⑀ ⌬319, purified, proteolytically digested, and subjected to immobilized metal affinity chromatography-LC/MS/MS (Fig.  4A). Three phosphorylation sites were identified, Ser 1420 , Ser 1430 , and either Thr 1557 or Ser 1558 . Serines 1420 and 1430 are amino-terminal to the five PPPSP sites targeted by GSK3 and near the predicted single pass transmembrane domain of LRP (amino acids 1371-1393) (Fig. 4C). Alignment of the regions in LRP5/6 demonstrates strong conservation of Ser 1430 in vertebrates from humans to Xenopus. Serine 1420 is 100% conserved in LRP6 but is absent in LRP5 (Fig. 4B). The high degree of conservation suggests functional relevance of phosphorylation at these sites. The difference between LRP5 and LRP6 is consistent with distinct functions in vivo for the two co-receptors. Mutation of Thr and Ser at amino acids 1557/1558 at the extreme carboxyl terminus had no detectable effect on LRP6 function and was not further studied.

Phosphorylation Sites Identified by LC/MS/MS Are Responsible for the Observed Mobility
Shift-To test if the putative phosphorylation sites were in fact CKI⑀ phosphorylation sites, the effect of serine 3 alanine point mutations in MBP LRP6 ⌬N was assessed (Fig. 4D). Wild type and single residue mutant LRP6 underwent a mobility shift in the presence of CKI⑀, whereas the double mutation of Ser 1420 and Ser 1430 to Ala eliminated the CKI⑀-induced mobility shift. Because these sites are close together, it is possible that they each are capable of inducing the same denaturation-resistant conformational change. Another possible explanation is that these sites both prime further phosphorylation at other unidentified sites.
Phosphorylation Site Mutants Activate Wnt Signaling-CKI⑀ has multiple substrates in the Wnt pathway, and overexpression of CKI⑀ has the net result of stabilizing ␤-catenin and up-regulating Lef-1-dependent signaling. We therefore tested the effect of mutation of these CKI⑀ phosphorylation sites on LRP6 function. Wild type and phosphorylation site mutant LRP6 were transiently expressed in HEK 293 cells containing a stably integrated ␤-catenin/Lef-1-responsive promoter driving luciferase expression (293 STF cells). Wild type LRP6 expression stimulated a 5-fold increase in luciferase expression. Unexpectedly, both the S1420A and the S1420A/S1430A mutants stimulated activity in the pathway 2.2-and 3.8-fold above wild type (Fig. 5A). To confirm this, a form of LRP6 lacking most of the extracellular domain (LRP6 ⌬N) was tested as well (Fig. 5B). As reported, the wild type LRP6 ⌬N construct was a more potent activator of the Wnt pathway, and its activity was again increased 2.3-and 4.7-fold by single and double mutation of the identified phosphorylation sites. Confirming the results of the reporter   ⌬N (lanes 1-3), LRP6 ⌬N S1420A (lanes 4 -6), LRP6 ⌬N S1430A (lanes 7-9), and LRP6 ⌬N S1420A/S1430A (lanes 10 -12) were used as substrates in an in vitro kinase assay (60 min, 30°C). Lanes 1, 4, 7, and 10 show load controls of purified MBP-LRP6, whereas lanes 2, 5, 8, and 11 show MBP-LRP6 incubated with kinase buffer in the absence of kinase. The phosphorylation-induced gel shift was detected by Western blotting with anti-MBP antibodies.
assays, mutant LRP6 stabilized free ␤-catenin more effectively than wild type (Fig. 5C). The increase in ␤-catenin stability and downstream signaling was not due simply to an effect of the mutations on LRP6 abundance, since the phosphorylation point mutants did not alter the expression levels of LRP6 (both wild type and LRP6 ⌬N) (Fig. 5C, bottom). This result suggests that phosphorylation of serines 1420 and 1430 negatively regulates Wnt signaling and supports a previously unappreciated role for CKI⑀ in down-regulation of Wnt signaling.
CKI⑀ Is Positioned Upstream of LRP6-We and others have shown biochemically and epistatically that CKI⑀ acts downstream of Wnt, and the normally low activity of cellular CKI⑀ is increased severalfold by canonical Wnt ligands (21,41). Since Wnt signals through interaction with cell surface receptors including members of the frizzled family and LRP6, we tested if LRP6 also stimulates cellular CKI⑀ activity. Wild type and phosphorylation mutants of LRP6 were co-expressed with CKI⑀, and kinase activity was then measured in an immunoprecipitation-kinase assay (21). Unexpectedly, expression of wild type and phosphorylation site mutants of LRP6 had no effect on CKI⑀ activity (Fig. 6A). LRP6 M5, which is unable to activate signaling, also failed to activate CKI⑀. In contrast to the lack of effect of LRP6 on CKI⑀ activity, co-expression of the known CKI⑀ activator Wnt3a with CKI⑀ boosts CKI⑀ activity close to 2.5-fold (Fig. 6A) (21).
In a parallel experiment, coexpression of CKI⑀ with wild type and phosphorylation site mutants of LRP6 activated signaling to the ␤-cate- . Similar results were obtained in multiple independent experiments as well as with independently created mutant constructs. B, Wnt signaling is similarly activated by CKI⑀ phosphorylation site mutants of LRP6 ⌬N. 293 STF cells were transfected with empty vector (lane 5), 25 ng of LRP6 ⌬N wild type, and phosphorylation site mutants as above. C, LRP6 S1420A/S1430A stabilizes ␤-catenin more efficiently than the wild type LRP6. Cells transfected in parallel were lysed with hypotonic buffer and analyzed for free ␤-catenin or with radioimmune precipitation buffer and analyzed for expression of LRP6. . CKI⑀ activity was assessed 48 h after transfection by a CKI⑀ immunoprecipitation assay as previously described (21). Cumulative data obtained from two experiments are presented. Data for each sample were normalized to the total amount of immunoprecipitated CKI⑀, and the kinase activity obtained from the control immunoprecipitate was subtracted. A.U., arbitrary units. B, LRP6 constructs tested in the immunoprecipitation kinase assay are biologically functional. 293 STF cells were transfected with 25 ng of CKI⑀ alone or with various LRP6 constructs as indicated. 48 h post-transfection, cells were lysed, and the -fold activation of luciferase expression was determined. Data are presented as mean Ϯ S.D. from three separate wells. Lower panel, mutation of CKI⑀ phosphorylation sites in LRP6 does not alter protein expression. Wild type and mutant VSVG-LRP6 was expressed in HEK 293 cells (250 ng plasmid/well), and protein abundance was assessed by immunoblotting lysates 48 h post-transfection.
nin/Lef-1 reporter as expected. Cotransfection of cells with CKI⑀ and wild type LRP6 stimulated luciferase activity 2.1-fold over that of cells transfected with CKI⑀ alone. Each of the phosphorylation site mutants tested ranged from 2.4-to 3.6-fold more activating than wild type LRP6. As previously reported, the control LRP6 M5 was unable to activate signaling (Fig. 6B). Differences in LRP6-dependent signaling were not due to changes in protein expression (Fig. 6B). The lack of influence of LRP6 and mutants tested on CKI⑀ activation demonstrates that LRP6 is either downstream or parallel to CKI⑀ in the Wnt signaling pathway. The ability of Wnt signaling (and other stimuli) to activate CKI⑀ and thereby decrease LRP6 function may be an additional mechanism to adjust the intensity of LRP6-dependent signaling.
Phosphorylation of LRP at Ser 1420 and Ser 1430 Inhibits Axin Binding-Previous studies have indicated that axin is recruited to the cytoplasmic domain of LRP6 by Wnt signaling. Concurrently, CKI⑀ is activated and can phosphorylate multiple substrates, including Dvl, axin, and LRP6. CKI⑀ activation can cause dissociation of axincontaining complexes (19). We therefore examined the effect of CKI⑀ expression and mutation of CKI⑀ phosphorylation sites on the interaction of both truncated (LRP6 ⌬N; Fig. 7A) and full-length LRP6 (Fig. 7B) with axin. Specific interactions between LRP6 ⌬N or full-length LRP6 and axin were identified (e.g. Fig. 7, A (lane 3) and B  (lane 3)). These interactions were enhanced by individual mutation of CKI⑀ phosphorylation site Ser 1420 or Ser 1430 (Fig. 7A, lanes 5 and  7) and further enhanced by double mutation of these sites (S1420A/ S1430A; Fig. 7, A (lane 9) and B (lane 7)). The increased recruitment of axin to LRP6 by mutation of these CKI⑀ phosphorylation sites is consistent with the stabilizing effect of the mutations on ␤-catenin and their enhancement of reporter gene activation (Fig. 5). The increased co-precipitation of axin was not due to changes in LRP6 abundance, since essentially equal amounts of LRP6 were present in each sample. Equal expression of axin was also seen under all conditions. The effect of CKI⑀ expression on the axin-LRP6 interaction was also examined. In the presence of overexpressed CKI⑀, less axin precipitated with LRP6. The effect of CKI⑀ on the LRP6-axin complex was not mitigated by mutation of the CKI sites in LRP6, suggesting, as previously reported, that CKI⑀ phosphorylates axin and additional proteins in the axin complex to further modulate binding.

DISCUSSION
The evolving model for activation of the Wnt-␤-catenin signaling pathway is that extracellular Wnt ligands interact with both a frizzled receptor, and perhaps via a separate domain, with a LRP5/6 co-receptor. When this ternary complex is formed, the cytoplasmic domain of LRP5/6 becomes phosphorylated on multiple copies of a PPPSP signaling motif that then recruits axin to the membrane and inhibits the phosphorylation-mediated degradation of ␤-catenin. Simultaneously, and perhaps independently, CKI⑀ and disheveled are activated, and distinct downstream signaling pathways regulating cellular calcium and cell motility are altered. Accumulating data suggest that a key role of LRP5/6 is to regulate the ␤-catenin branch of the Wnt response, whereas disheveled proteins play an important role in Wnt-regulated noncanonical pathways. In this study, we examined the role of CKI⑀ in regulating LRP5/6 phosphorylation and activity. We find that LRP5/6 and CKI⑀ interact both in vivo and in vitro, suggesting that the interaction is direct. Via in vitro phosphorylation of LRP6 followed by LC/MS/MS, three distinct CKI⑀ phosphorylation sites on LRP6 were identified, all of which were distinct from the previously identified PPPSP signaling motif. Whereas phosphorylation of each PPPSP motif by GSK3␤ creates a predicted CKI⑀ phosphorylation site, our LC/MS/MS analysis of in vitro phosphorylation sites did not detect their phosphorylation. Mutation of two of the identified CKI⑀ phosphorylation sites increased the activity of LRP6 severalfold as assessed by both ␤-catenin stabilization and activation of a ␤-catenin-responsive promoter. Consistent with this, the mutations also enhanced the binding of axin to LRP6. Whereas Wnt signaling activates CKI⑀, we find that LRP6 does not activate CKI⑀.
Taken together, the data suggest that activated CKI⑀ phosphorylates specific sites on LRP6 (namely, amino acids 1420 and 1430) that inhibit LRP6-axin binding. Since CKI⑀ may be regulated by diverse stimuli, one possible function of the inhibitory pathway is modulation of the strength of the Wnt 3 LRP6 3 ␤-catenin pathway. Wnt-mediated CKI⑀ activation could provide negative feedback on LRP6-axin interaction. Alternatively, other signaling that activates CKI⑀ could also lead to LRP6 1420/1430 phosphorylation and decreased Wnt sensitivity. Given that members of the CKI family each have pleiotropic physiological functions (in Wnt signaling and other pathways), the results of this FIGURE 7. LRP6 S1420A and S1430A mutants bind axin more efficiently than the wild type LRP6. Wild type (WT) and phosphorylation site mutants of VSVG LRP6 ⌬N (A) or VSVG LRP6 FL (B) were expressed in HEK 293 cells with axin-HA and CKI⑀ as indicated. In either case, LRP6 was omitted from lane 1, and axin was omitted from lane 2. 48 h post-transfection, LRP6 was immunoprecipitated (IP) with anti-VSVG antibodies. Immunoprecipitated complexes were analyzed by immunoblotting with anti-HA (first panel), anti-VSVG (second panel), and anti-CKI⑀ (third panel) antibodies. Blotting of lysates with anti-HA shows that total HA-axin abundance was not altered by LRP6 and CKI⑀ expression (fourth panel). study also emphasize that specific functions of phosphorylation are difficult to sort out without identification and functional characterization of specific phosphorylation sites on individual target proteins.
One of the phosphorylation sites, Ser 1430 , was conserved in both LRP5 and LRP6, whereas a second site, Ser 1420 , was present only in LRP6. This is consistent with the apparently distinct biological roles of LRP5 and LRP6. LRP5 and LRP6 are expressed with distinct physiological patterns, and each of these coreceptors has been associated with unique human pathologies. LRP6 is expressed in all tissues, and mice lacking the LRP6 gene have severe and early developmental abnormalities that are a composite of several Wnt gene deletions (23). Mice lacking the LRP5 gene display reduced bone density and abnormalities in cholesterol and glucose metabolism, whereas activating mutations in LRP5 are associated with osteopetrosis (42)(43)(44)(45). The differing pathologies may not be entirely due to differing expression locales. Our observation that a phosphorylation site at 1420 is highly conserved in LRP6 and that the absence of this site is equally highly conserved in LRP5 is consistent with other data indicating that these receptors are functionally nonequivalent. The apparent negative regulatory role of this site suggests that LRP5 and LRP6 may differ in their sensitivity to modulation by negative feedback via activation of CKI⑀.
CKI⑀ exhibits opportunistic pleiotropy in that it has been conscripted to play roles in multiple pathways (whatever its original function), including Wnt and NFAT signaling and regulation of circadian rhythm (46). Within the Wnt signaling pathway, CKI⑀ is known to interact with and phosphorylate a multiplicity of substrates (including ␤-catenin, axin, disheveled, APC, TCF-4, and here LRP5/6). Because initial studies have indicated that CKI⑀ is a positive regulator, this multiplicity of CKI⑀ activities has probably been presumed to contribute to activation. Our data suggest that this picture is too simple and that CKI⑀ can also function as a negative regulator of Wnt signaling. Phosphorylation of LRP5/6 on juxtamembrane sites negatively regulates LRP5/6 activity by limiting axin interaction, whereas overexpression of CKI⑀ also inhibits the interaction of axin with LRP6. Our previous studies (19) also demonstrated that axin complexes were destabilized by CKI⑀ activity. Thus, CKI⑀ activities in the Wnt pathway exemplify combinatorial pleiotropism in that a single gene product mediates distinct, and in this case opposing, effects depending upon the protein partners with which it interacts.