Quantitative Phosphoproteome Profiling of Wnt3a-mediated Signaling Network

The complexity of canonical Wnt signaling comes not only from the numerous components but also from multiple post-translational modifications. Protein phosphorylation is one of the most common modifications that propagates signals from extracellular stimuli to downstream effectors. To investigate the global phosphorylation regulation and uncover novel phosphoproteins at the early stages of canonical Wnt signaling, HEK293 cells were metabolically labeled with two stable isotopic forms of lysine and were stimulated for 0, 1, or 30 min with purified Wnt3a. After phosphoprotein enrichment and LC-MS/MS analysis, 1057 proteins were identified in all three time points. In total 287 proteins showed a 1.5-fold or greater change in at least one time point. In addition to many known Wnt signaling transducers, other phosphoproteins were identified and quantitated, implicating their involvement in canonical Wnt signaling. k-Means clustering analysis showed dynamic patterns for the differential phosphoproteins. Profile pattern and interaction network analysis of the differential phosphoproteins implicated the possible roles for those unreported components in Wnt signaling. Moreover 100 unique phosphorylation sites were identified, and 54 of them were quantitated in the three time points. Site-specific phosphopeptide quantitation revealed that Ser-20 phosphorylation on RRM2 increased upon 30-min Wnt3a stimulation. Further studies with mutagenesis, the Wnt reporter gene assay, and RNA interference indicated that RRM2 functioned downstream of β-catenin as an inhibitor of Wnt signaling and that Ser-20 phosphorylation of RRM2 counteracted its inhibition effect. Our systematic profiling of dynamic phosphorylation changes responding to Wnt3a stimulation not only presented a comprehensive phosphorylation network regulated by canonical Wnt signaling but also found novel molecules and phosphorylation involved in Wnt signaling.

The Wnt family of secreted signaling molecules is highly conserved among animal species and has been implicated in three major pathways, the canonical pathway and the two noncanonical pathways including planar cell polarity and calcium pathway (1). The canonical Wnt signaling pathway plays an important role in developmental processes including cell adhesion (2), morphology (3), proliferation (4), and migration (5). Mutational deregulation of the Wnt cascade is closely associated with various tumors and other diseases (6,7). Protein phosphorylation is one of the most important mechanisms for signaling propagation, and there is no exception for canonical Wnt signal transduction (8 -13). Many regulators or downstream components in the canonical Wnt signaling pathway are known to be phosphorylated or dephosphorylated in the transduction of signals from extracellular Wnt stimulation to intracellular effectors. These include Wnt receptors low density lipoprotein receptor-related protein-6 (8) and Frizzled 3 (9) and signaling components glycogen synthase kinase-3␤ (GSK3␤) 1 (10), adenomatous polyposis coli protein (APC) (11), Dishevelled (Dvl) (12), and ␤-catenin (13). For example, as a key component in canonical Wnt signaling, ␤-catenin is phosphorylated by serine/threonine kinases in the absence of Wnt stimulation, and the hyperphosphorylation leads to its ubiquitination and degradation by the proteasome. When Wnt is present, ␤-catenin is free from phospho-rylation, which results in its stabilization and translocation into the nucleus to regulate gene transcription (13).
Previous work has demonstrated the indispensable role of reversible phosphorylation in regulation of the canonical Wnt signaling cascade by studying one or several specific proteins (10,12). The newly developed proteomics strategies make possible the global characterization of a signaling pathway. Recently phosphoproteomes of many signaling pathways, including those activated by epidermal growth factor, fibroblast growth factor, interferon-␣, insulin, and transforming growth factor ␤, have been analyzed by using immunoaffinity purification and/or IMAC enrichment or [ 32 P]orthophosphate radioactive isotope labeling followed by mass spectrometric identification (14 -18). These global studies have unraveled the relationships between many novel phosphoproteins or phosphorylation sites and specific signaling pathways, therefore enriching our view of comprehensive phosphorylation regulation in signal transduction. However, many of the earlier approaches would not allow quantitative characterization of the dynamic changes in phosphorylation events and distinguishing induced phosphorylation events from constitutive phosphorylation events (15,18). Although some of the approaches are able to quantitate phosphorylation changes, the quantitation is based on two-dimensional gel electrophoresis or a chemical derivation method (14,16,17). Recently a mass spectrometry-based quantitation approach termed stable isotope labeling by amino acids in cell culture (SILAC) has been introduced (19). Combined with a phosphoprotein or phosphopeptide enrichment method, SILAC has been widely applied to profile dynamic phosphorylation changes in signal transduction (20,21). In the present study, we investigated the differential phosphoproteome upon stimulation of a canonical Wnt protein, Wnt3a, by using SILAC in combination with phosphoprotein enrichment and LC-MS/MS analysis.

Stable Isotope Labeling with Amino Acids in Cell Culture-HEK293
cells were grown in RPMI 1640 medium containing [ 12 C 6 ]lysine ("light") or [ 13 C 6 ]lysine ("heavy") supplemented with 10% dialyzed fetal bovine serum plus antibiotics. Detailed instructions for this protocol are available upon request. Briefly the HEK293 cells were adapted to grow in isotope-containing medium supplemented with dialyzed serum and passaged for six generations to ensure a complete replacement prior to initiating these experiments. The heavy cells were stimulated with Wnt3a (100 ng/ml) for 1 or 30 min. The light, unstimulated cells were divided to serve as a common zero point.
Phosphoprotein Purification and Western Blotting-A phosphoprotein purification kit (23)(24)(25) from Qiagen (Valencia, CA) was applied to enrich phosphoproteins according to the manufacturer's instructions. Briefly light and heavy cells were lysed in the Lysis Buffer, and the cell lysate was centrifuged at 10,000 ϫ g at 4°C for 30 min to remove insoluble material. After centrifugation, the protein concentration of the cell lysate was quantitated by Bradford assay, and different cell lysates (light or heavy) were combined at an equal sample amount. The extracted proteins of combined cell lysates were diluted to a concentration of 0.1 mg/ml with Lysis Buffer, and a total of 25 ml of the extracted proteins was applied to a Lysis Buffer-equilibrated phosphoprotein purification column at room temperature. After washing the column with 6 ml of Lysis Buffer, the phosphoproteins were eluted with 2 ml of Phosphoprotein Elution Buffer. All the buffers and the phosphoprotein purification column were provided in the kit by the manufacturer (Qiagen).
The eluted phosphoproteins were dialyzed, lyophilized, and resolved by 7.5-17.5% gradient SDS-PAGE. The gel was stained using Coomassie Brilliant Blue stain to visualize the gel lanes. The two gel lanes ("0 min-1 min," the combined cell samples stimulated with Wnt3a for 0 or 1 min; "0 min-30 min," the combined cell samples stimulated with Wnt3a for 0 or 30 min) were cut into 26 slices, respectively. The excised slices were subjected to in-gel trypsin digestion as described before (26).
For verification of phosphoprotein enrichment and SILAC results, HEK293 cells were grown and treated the same way as the sample for SILAC analysis except for stable isotope [ 13 C 6 ]lysine (heavy) labeling. Briefly the cells were grown in 10-cm 2 dishes. One dish of cells was left untreated as control, and two dishes of cells were treated with Wnt3a ligand for 1 and 30 min, respectively. Cell lysis and phosphoprotein enrichment of the three samples were separately performed according to the manufacturer's instruction (Qiagen). Eluted phosphoproteins were dialyzed, lyophilized, and reconstituted by an equal volume (250 l) of 4ϫ sample buffer (40% glycerol, 8% SDS, 250 mM Tris-HCl, 4% ␤-mercaptoethanol). To validate the efficacy of phosphoprotein enrichment in the elution fraction, 10 g of proteins of the total cell lysate, elution fraction, and flow-through fraction from the phosphoprotein purification kit were separated by 12.5% SDS-PAGE followed by silver staining (26) or Western blotting analysis using mixed phosphoserine, phosphothreonine, and phosphotyrosine monoclonal antibodies. For Western blotting verification of SILAC results, the eluted phosphoprotein fractions were loaded at the same volume of 10 l. After transfer, the nitrocellulose membranes were blocked by 1ϫ Net-gelatin (150 mM NaCl, 5 mM EDTA, 50 mM Tris-HCl, pH 7.5, 0.05% Triton X-100, 0.25% gelatin) and then incubated with the corresponding primary antibodies followed by horseradish peroxidase-conjugated or IRDye 800CW-conjugated affinity-purified secondary antibodies. The membranes were subjected to fluorescent chemiluminescence detection and exposed to films or scanned by the Odyssey Infrared Imaging System 9120 (LI-COR, Lincoln, NE) according to the manufacturer's instructions.
LC-MS/MS Analysis-The peptide mixtures from each gel slice were separated by reverse phase HPLC followed by tandem mass spectrometry analysis. Reverse phase HPLC was performed using an Agilent 1100 Capillary system (Agilent Technologies) on a C 18 column (150-m inner diameter, 100-mm length; Column Technology Inc., Fremont, CA). The pump flow rate was 1.6 l/min. Mobile phase A was 0.1% formic acid in water, and mobile phase B was 0.1% formic acid in acetonitrile. The tryptic peptide mixtures were eluted using a gradient of 2-55% B over 135 min.
The mass spectral data shown in this study were acquired on a linear quadrupole ion trap (LTQ) mass spectrometer (Thermo, San Jose, CA) equipped with an electrospray interface operated in positive ion mode. The temperature of the heated capillary was set at 170°C. A voltage of 3.0 kV applied to the ESI needle resulted in a distinct signal. The normalized collision energy was 35.0. The number of ions stored in the ion trap was regulated by the automatic gain control. Voltages across the capillary and the quadrupole lenses were tuned by an automated procedure to maximize the signal for the ion of interest. The mass spectrometer was set such that one full MS scan was followed by 10 MS/MS scans on the 10 most intense ions from the MS spectrum with the following Dynamic Exclusion TM settings: repeat count, 2; repeat duration, 30 s; exclusion duration, 90 s.
Database Search, Phosphorylation Site Identification, and Quantitative Analysis by RelEx-The BioWorks TM 3.1 software suite was used to generate the peak lists of all acquired MS/MS spectra, and then they were automatically searched against the Internation Protein Index (IPI) human database (version 3.04) (27) containing 49,078 protein entries using the SEQUEST (University of Washington, licensed to Thermo Finnigan) searching program. Trypsin was designated as the protease, and up to two missed cleavages were allowed. Carbamidomethylation was searched as a fixed modification, and phosphorylation of serine/threonine/tyrosine residues (ϩ79.98 Da) and isotope-labeled lysine (ϩ6.00 Da) were allowed as variable modifications. The mass tolerance of the CID spectra was set as Ϯ1.0 Da. For non-phosphopeptide identification, a rigorously accepted SE-QUEST result must have a ⌬C n score of at least 0.1 (regardless of charge state). The cross-correlation score must be Ն1.9 for a ϩ1 tryptic peptide, Ն2.2 for a ϩ2 tryptic peptide, and Ն3.75 for a ϩ3 tryptic peptide (28).
For phosphorylation site identification, first, defined phosphopeptides should be above the increased X corr threshold setting of Ն2.0 for ϩ1 tryptic peptide, Ն2.5 for ϩ2 tryptic peptide, and Ն4.0 for ϩ3 tryptic peptide (previously reported X corr thresholds for phosphopeptide filtering were set as 1.9, 2.2, and 3.75 for singly, doubly, and triply charged ions, respectively (29), or 2.5 and 3.3 for doubly and triply charged ions, respectively (30)). Second, a phosphorylation site(s) was (were) considered to be unique only when a certain phosphopeptide had a ⌬C n score of at least 0.1, as ⌬C n Ն 0.1 is significant for discriminating the first (top) candidate peptide from the second candidate peptide (28). Third, all the spectra of the quantitated peptides and phosphopeptides were manually inspected according to the criteria (31) that an MS/MS spectrum of good quality must have its fragment ion peaks clearly above baseline noise, show sequential members of the b-or y-ion series with phosphorylation site included, and show intense proline-directed fragment ions. In addition, the phosphoric acid neutral loss peaks were checked for phosphorylation site identification (30).
To eliminate redundancy, we first used the IPI database, which offers complete nonredundant data sets built from the Swiss-Prot, TrEMBL, Ensembl, and RefSeq databases (27). Second, by using a homemade software, protein group X was removed if all identified peptides assigned to protein group X were not unique (the peptides were also assigned to another protein group). To sort out a single protein member from a protein group, we chose the protein from the Swiss-Prot database and with the highest sequence coverage. If the peptides were assigned to proteins from other databases, we chose the protein with the highest sequence coverage. In addition, to esti-mate the rate of incorrect identifications (false positives), all the filtered spectra were subjected to database searching against a composite database containing human protein sequences in both the forward (correct) and reverse (incorrect) orientation. The reverse database was positioned after the forward database. After database searching, the false-positive rate of peptide identification was calculated to be 1.17% according to the algorithm developed by Peng et al. (32). Then only proteins identified by at least two peptide hits were selected for further analysis. Protein false-positive rate (Pr_FPR) was calculated to be 1.50% according to the following equation: Pr_FPR ϭ (2 ϫ Number of proteins identified from reverse database)/ Number of proteins identified from forward and reverse database. For quantitative analysis, the peptides without lysine or lysine-containing peptides that could not be assigned to a single protein group were discarded. Only those lysine-containing peptides that could be assigned to single proteins were sent to the RelEx program (33) as candidates for determining the ratio of 13 C-peptide/ 12 C-peptide. The RelEx (33) parameters for chromatogram filters were set at: smooth point, 7; minimum signal to noise ratio, 3; minimum 0.7 correlation at 1; and minimum 0.4 correlation at 10. The mean peptide ratio for the protein was calculated following the removal of outliers using a Dixon's Q-test integrated by RelEx (33).
k-Means Clustering-The k-means clustering approach (34) was used to analyze the identified differential phosphoproteins. Before clustering analysis, protein ratios were transformed into the log scale (base 2), which would convert the distribution of relative abundance values for all quantitated proteins into a more symmetric, almost normal pattern (35). We set the number of clusters to 4 to classify the data set of differential phosphoproteins. We used the Cluster 3.0 freeware software package (M. J. L. de Hoon, Laboratory of DNA Information Analysis, Human Genome Center, Institute of Medical Science, University of Tokyo). Repeated (10 -100) k-means clustering of proteins was based on the Pearson correlation coefficient of their expression profiles (ratios in three time points). Then the prototype (mean profile) was plotted using a python 2D plotting library, Matplotlib (The MathWorks).
Network Modeling-All differentially regulated proteins were searched against the Human Protein Reference Database (HPRD) (www.hprd.org/), Kyoto Encyclopedia of Genes and Genomes (KEGG) database (36), and the literature to find protein-protein interactions (PPIs) with known canonical Wnt signaling proteins. PPI networks were visualized by using Cytoscape (37).
Plasmids-Full-length HA-tagged RRM2 was generated by RT-PCR (SuperScript III from Invitrogen). An XbaI and an XhoI site were introduced, respectively, into the sense and the antisense primer. Using XbaI/XhoI digestion, the human RRM2 cDNA was cloned into pCMV-HA-XbaI/XhoI vector. Based on pCMV-HA-RRM2WT, S20A and S20E were generated by PCR. All the plasmids were verified by multiple restrictive endonuclease digestions and DNA sequencing. The lymphoid enhancer factor (LEF) reporter genes (38) were kindly provided by Dr. Grosschedl (University of California San Francisco), and the NF-B reporter gene (39) was a generous gift from Dr. Anning Lin (Chicago University).
RT-PCR-Total RNA was isolated from HEK293 cells using a TRIzol kit (Invitrogen), and the RT reaction was conducted using the SuperScript III (Invitrogen) synthesis system following the manufac-turer's instructions. The specific primers 5Ј-CT AGT CTA GAG CTC TCC CTC CGT GTC-3Ј (sense) and 5Ј-C CAG CTC GAG GAA GTC AGC ATC CAA GGT-3Ј (antisense) were used to PCR amplify fulllength RRM2 cDNA.
Cell Culture, Transfection, and Reporter Assay-HEK293 cells were maintained in Dulbecco's modified Eagle's medium (Invitrogen) sup-plemented with 10% fetal bovine serum (Invitrogen). Transient transfection in cells was performed using Lipofectamine reagent and Plus reagent (Invitrogen) as suggested by the manufacturer. The transfection was stopped after 3 h. Generally for the LEF-reporter gene assay in HEK293 cells, the total amount of expression plasmid was 250 ng/well (24-well plate) including 5 ng of HA-LEF, 20 ng of LEF-

FIG. 1. Skematic for SILAC and verification of phosphoprotein enrichment and SILAC analysis.
A, overall experimental procedure. Briefly HEK293 cells were metabolically labeled with two stable isotopic forms of lysine, and the cells labeled with heavy isotope ([ 13 C 6 ]lysine) were stimulated with Wnt3a for 1 or 30 min, whereas the unstimulated cells labeled with the light isotopic form of lysine ([ 12 C 6 ]lysine) were used as control. Stimulated heavy cells were combined with unstimulated light cells at equal amounts of cell lysate. The combined cell lysates were subjected to phosphoprotein enrichment, SDS-PAGE fractionation, and subsequent mass spectrometry analysis. The characters of a-f in different shapes represent different peptides. B, specificity and sensitivity of phosphoprotein enrichment validated by silver staining (a) and Western blotting using mixed phosphoserine, phosphothreonine, and phosphotyrosine monoclonal antibodies (b). C, separation of enriched phosphoproteins by 7.5-17.5% gradient SDS-PAGE. The gel was stained using Coomassie Brilliant Blue stain to visualize the gel lanes. The two gel lanes (0 min-1 min, the combined cell samples stimulated with Wnt3a for 0 or 1 min; 0 min-30 min, the combined cell samples stimulated with Wnt3a for 0 or 30 min) were cut into 26 slices, respectively, which were further subjected to in-gel trypsin digestion and LC-MS/MS analysis. M, molecular mass markers in kDa. D, Western blotting verification of selected proteins. SILAC ratios of individual proteins are indicated. Five of eight proteins were differentially regulated upon Wnt3a stimulation, including four canonical Wnt signaling components, ␤-catenin, nucleophosmin, MEK1, and PP2Ac, and one candidate protein, PKC␣. The other three proteins are hnRNP C, HSP90, and RRM2, whose overall phosphorylation remained almost constant upon stimulation. luciferase reporter, 25 ng of GFP, and LacZ or RRM2 at different amounts (10 or 25 ng), and LacZ was used to make up the total. After 18 h of transfection, cells were stimulated by Wnt3a-conditioned medium for 6 h before collection. For the NF-B luciferase assay, 25 ng of GFP, 20 ng of NF-B reporter, and 25 ng of RRM2 or IB␣-SR (I kappa B alpha-super repressor) were used. After 18 h of transfection, cells were stimulated by 10 ng/ml TNF␣ for 6 h before the luciferase assay. The cell lysate was measured for fluorescence intensity emitted by GFP proteins in a FL600 fluorometer (BIO-TEK Inc., Winooski, VT), and then the luciferase substrate (Roche Applied Science Luciferase Assay kit) was added for determining the luciferase activities using a MicroLumate Plus luminometer (PerkinElmer Life Sciences). Luminescence intensity was normalized against fluorescence intensity.
Detection of Free ␤-Catenin Level by Cell Fractionation and Western Blotting-HEK293 cells in a 6-well plate were transfected with 100 ng/well LacZ, RRM2, or GSK3␤ with LacZ to make up the total amount of 1 g/well. After 18 h of transfection, Wnt3a-conditioned medium was added and incubated for 3 h. The cells were washed once with ice-cold PBS and scraped into hypotonic buffer containing 10 mM HEPES (pH 7.9), 1.5 mM MgCl 2 , 1 mM EDTA, 10 mM KCl, 0.5 mM DTT, and a mixture of protease inhibitors. After incubating on ice for 10 min, cells were homogenized in a homogenizer and centrifuged at 15,000 ϫ g for 10 min. The supernatant was collected as the cytosol fraction. The ␤-catenin level in the cytosol fraction was detected by Western blotting according to the aforementioned protocol.

Quantitative Analysis of Differentially Regulated Proteins upon Wnt3a
Stimulation-To obtain a global and dynamic view of protein phosphorylation in canonical Wnt signaling, the proteome of HEK293 cell populations was metabolically labeled with heavy ( 13 C) or light ( 12 C) stable isotopic forms of lysine. The cells labeled with heavy isotope were stimulated with Wnt3a for 1 or 30 min, whereas the untreated cells labeled with the light isotopic form of lysine were used as control. The same number of stimulated and control cells were lysed, and the lysates were then combined at an equal amount. The combined cell lysates were subjected to phosphoprotein enrichment, SDS-PAGE fractionation, and subsequent mass spectrometry analysis (Fig. 1A). To validate the sensitivity and specificity of phosphoprotein enrichment in the elution fraction from the phosphoprotein purification kit (23)(24)(25), we used silver staining and Western blotting using mixed phosphoserine, phosphothreonine, and phosphotyrosine monoclonal antibodies to compare total protein or phosphoprotein staining patterns of the total cell lysate, elution fraction, and flow-through fraction. As shown in Fig. 1B, phosphoproteins were significantly enriched in the elution fraction, which showed a phosphoprotein staining pattern similar to that of total proteins, whereas the flow-though fraction was nearly devoid of phosphoproteins.
Enriched phosphoproteins of SILAC labeled samples were separated by 7.5-17.5% gradient SDS-PAGE, and each lane was cut into 26 slices (Fig. 1C). By mass spectrometry analysis, a total of 1057 proteins were identified and quantitated from all three samples collected at the three time points 0, 1, and 30 min. To further confirm the results from the quantitative phosphoproteome analysis, we chose eight proteins for Western blotting verification. The samples used in the verification experiment were prepared in the same way as those used for SILAC analysis. As shown in Fig. 1D, SILAC results were very consistent with Western blotting analysis for all eight proteins. Upon Wnt3a stimulation, phosphorylation of PKC␣, nucleophosmin (40,41), and MEK1 (42) showed increases, whereas phosphorylation of ␤-catenin (13) and PP2Ac (43) showed decreases (Fig. 1D). Except for PKC␣, all the other four regulated proteins have been reported to be involved in Wnt signaling (13, 40 -44). However, only phosphorylation regulation of ␤-catenin has been previously reported in Wnt signaling (13,44). The other three proteins, hnRNP C, HSP90, and RRM2, remained almost unchanged in both SILAC analysis and Western blotting verification (Fig.  1D), suggesting that the overall phosphorylation levels of these proteins were not significantly changed by Wnt3a stimulation. Putting all the data together, we believe that the combination of phosphoprotein enrichment and SILAC is a reliable approach to profile phosphoproteome alteration in signal transduction.
Functional Characterization of Differentially Regulated Proteins-The distribution of -fold changes for all 1057 quantitated proteins is shown in Fig. 2, which displayed an almost symmetrical pattern for both the 0 min-1 min combined sam-  Table I. Although it has been reported that the SILAC technique can quantitate changes smaller than 10% (45,46) and a majority of housekeeping proteins, such as ribosomal proteins, proteasome proteins, and cytoskeleton proteins, were found to have a -fold change between 0.83 and 1.2 in the current study (data not shown), we chose 1.5 as the conservative threshold for a significant ratio of change. In total, 287 proteins showed a 1.5-fold or greater change in at least one time point. The detailed information on these 287 proteins is given in Supplemental Table 1. All peptides used for quantitating these 287 proteins are listed in Supplemental Table 2, and peptides from proteins assigned by a single peptide are listed in Supplemental Table 3. The MS/MS spectra of all quantitated peptides from the 287 differentially regulated proteins are shown in Supplemental Fig. 1.
Among the 58 signaling proteins, 39 proteins have been implicated in various signaling cascades (Supplemental Table  1) of which 11 proteins have been reported previously to be involved in canonical Wnt signaling. Their functional roles in Wnt signaling and previous studies on their phosphorylation regulation are detailed in Table II. In addition, 33 proteins have been implicated in noncanonical Wnt signaling or other signaling pathways, 13 of which participate in more than one signaling pathway (Fig. 3B and Supplemental Table 1). For example, MEK1 is a downstream effector or regulator of mitogen-activated protein kinase (MAPK), insulin, and canonical Wnt signaling (42,47,48).
In addition to 11 differentially regulated Wnt signaling pro-teins, 10 Wnt-related proteins remained almost constant upon Wnt3a stimulation (data not shown), indicating that phosphorylation of these proteins might not be significantly regulated by Wnt3a or that these proteins respond in the later stage of Wnt signaling. Twenty-two more Wnt signaling proteins were also identified in the present study, but their profile patterns could not be defined (data not shown). One of the reasons was that many of these proteins were only identified in one of the combined samples (either 0 min-1 min or 0 min-30 min). The other reason was the lack of quantitation information due to the absence of the lysine residue in peptides or an unat-  tainable signal to noise ratio when processed by RelEx (33).
Profile Patterns of Differentially Regulated Proteins by k-Means Clustering-Based on the time course study, we were able to obtain a temporal profile of differential phosphorylation regulation of canonical Wnt signaling at 0, 1, and 30 min. k-Means clustering (34) of these 287 proteins, which showed a 1.5-fold or greater change, revealed four different profile patterns, and the number of proteins in each profile group was 85 (29.62%), 49 (17.07%), 92 (32.06%), and 61 (21.25%), respectively (Fig. 4A). Eleven differentially regulated Wnt signal proteins were included in the different profile groups. Phosphoproteins of Cluster 1, such as a Wnt downstream effector, ␤-catenin (13,44), remained almost unchanged upon 1 min of Wnt3a stimulation, whereas they showed a significant decrease upon 30 min of stimulation (Fig. 4, A and B). Phosphoproteins of Cluster 2, such as a positive regulator of Wnt signaling, Ran-binding protein 3 (RanBP3), which is upstream of ␤-catenin (49), showed a quick increase upon 1 min of Wnt3a stimulation but decreased to almost basal level after 30 min of stimulation (Fig.  4, A and B). Phosphoproteins of Cluster 3, such as a positive regulator of Wnt signaling, MEK1 (42), showed a continuous increase upon Wnt3a stimulation (Fig. 4, A and B). In contrast to Cluster 2, Cluster 4 phosphoproteins, such as a negative regulator of Wnt signaling, occludin (50), showed a quick decrease upon 1 min of Wnt3a stimulation and increased back to almost basal level after 30 min of stimulation (Fig. 4, A and B).
Interaction Networks of Differentially Regulated Proteins-To explore possible relationships between known Wnt signaling components and proteins whose phosphorylation status may be regulated by Wnt3a as identified by the SILAC and MS approach, these 287 proteins were subjected to searches in the HPRD (51), KEGG database (36), and literature for PPI information. The search revealed that 17 candidate proteins were connected to the canonical Wnt signaling network (Fig. 5), implicating their involvement in transduction of Wnt signals. For example, it has been reported that COP9 signalosome complex subunit 7a (SGN7a) interacts with Cullin 1 (Cul1) and Ring box 1 (Rbx1) (52) and can be phosphorylated by casein kinase II and protein kinase D (53). Cul1 and Rbx1 are components of the SCF (Skp1/Cul1/F box) ubiquitin ligase complex and initiate degradation of phosphorylated ␤-catenin (54). The ubiquitin conjugation pathway can be regulated by the COP9 signalosome complex in a phosphorylationdependent manner (52,53). In the present study, SGN7a showed increased phosphorylation upon Wnt3a stimulation (Figs. 4A and 5). We hypothesized that more SGN7a molecules may become phosphorylated and regulate the activity of the ubiquitin conjugation pathway upon Wnt3a stimulation. It might be another manner of modulation, in addition to regulation of the formation of a "destruction complex" composed of GSK3␤, APC, and Axin (13), to help accumulate "free" ␤-catenin.
Site-specific Phosphorylation Regulated by Wnt Signaling-Based on SEQUEST X corr and ⌬C n filtering (see "Experimental Procedures"), 489 phosphorylation sites were identified in 326 peptides (data not shown). After further rigorous manual inspection (see "Experimental Procedures"), 100 unique phosphorylation sites were confidently identified in 90 phos- phopeptides (Table III). Among these 100 sites, 74 phosphorylation sites in 67 phosphopeptides were from the 0 min-1 min sample, whereas 43 phosphorylation sites in 38 phosphopeptides were from the 0 min-30 min sample. Seventeen phosphorylation sites in 15 phosphopeptides were identified in both samples. We were able to quantitate 54 phosphorylation sites in 47 phosphopeptides. Among these quantitated sites, 41 phosphorylation sites in 36 peptides were from the 0 min-1 min sample, and 17 phosphorylation sites in 15 pep-tides were from the 0 min-30 min sample. Three and 10 of these phosphopeptides showed at least a 1.5-fold increase and decrease upon 1 min of Wnt3a stimulation, whereas four and six phosphopeptides showed at least a 1.5-fold increase and decrease upon 30 min of stimulation (Table III) Among the 100 identified phosphorylation sites, 16 sites were found previously in other studies, and one potential phosphorylation site (Ser-570 in SLTSS#LENIFSR (where # indicates phosphorylation modification) of TBC1 domain fam-ily member 4) was confirmed in the present study (according to Swiss-Prot database or HPRD) (Supplemental Table 4). The majority (83 of 100, 83%) of the phosphorylation sites were novel. Among four phosphopeptides that were quantitated in   both combined samples, two phosphopeptides showed an increase upon Wnt3a stimulation (Table III and Supplemental  Table 5), Ser-94 on E1B-55 kDa-associated protein (E1B-AP5) and Ser-20 on RRM2. Although these proteins showed almost unchanged phosphorylation level (Fig. 1D and Supplemental  Tables 6 and 7), specific phosphorylation sites were found to be significantly regulated upon Wnt3a stimulation based on SILAC direct labeling and quantitation of the phosphopeptides (Fig. 6, Table IV, and Supplemental Table 5). For example, 30 min of Wnt3a stimulation induced a 1.56-fold increase in RRM2 phosphorylation at residue Ser-20, whereas 1 min of stimulation did not induce any significant change of Ser-20 phosphorylation (Fig. 6 and Table IV).
Mutational Analysis of Ser-20 Phosphorylation on RRM2-Because RRM2 has never been reported to be involved in Wnt signaling, we first investigated whether RRM2 has a role in regulation of canonical Wnt signaling. We cloned the fulllength RRM2 cDNA from HEK293 cells and tested its effect on Wnt activity by using a LEF-luciferase reporter gene assay (38). Interestingly overexpression of RRM2 led to a marked decrease in the LEF-luciferase reporter activity stimulated by Wnt3a-conditioned medium in a dose-dependent manner (Fig. 7A, upper panel). To examine where RRM2 may function in the Wnt signaling pathway, the effect of RRM2 on Wnt3ainduced ␤-catenin stabilization was tested. As shown in Fig.  7A (lower panel), expression of RRM2 did not inhibit Wnt3ainduced ␤-catenin stabilization, whereas a positive control, GSK3␤ (10), which functions downstream of Wnt and upstream of ␤-catenin, remarkably inhibited the accumulation of ␤-catenin. This result suggests that RRM2 may function downstream of ␤-catenin and regulate its activity without affecting its stability.
To demonstrate that RRM2 specifically inhibits Wnt signaling, we tested the effects of RRM2 on TNF␣-mediated NF-B signaling using an NF-B reporter gene assay (39) and found that RRM2 had no effect on TNF␣-induced NF-B activity (Fig. 7B). A positive control, IB␣-SR (55), which is a constitutive active form of IB␣, was able to suppress the NF-B activity.
To further confirm the inhibitory function of RRM2, two RRM2 small interfering RNAs were generated. Both of them markedly reduced the expression of endogenous RRM2 (Fig.  7C). ⌬N-␤-Catenin is constitutively active because it lacks the N-terminal 45 amino acids and cannot be phosphorylated and degraded (56). Knockdown of RRM2 in HEK293 cells by these siRNAs led to up-regulation of basal and ⌬N-␤-catenin-induced LEF-reporter activities (Fig. 7C). Thus, the results shown in Fig. 7C confirmed that RRM2 may function as an innate suppressor of canonical Wnt signaling downstream of ␤-catenin in vivo. Interestingly it was observed that siRRM2 had less effect on Wnt3a-stimulated activity than on ⌬N-␤catenin-induced activity (Fig. 7C). Considering that phosphorylation of Ser-20 on RRM2 was up-regulated upon Wnt3a stimulation ( Fig. 6 and Table IV), we hypothesized that Ser-20 phosphorylation induced by the Wnt signal may counteract the inhibition of endogenous RRM2 as siRNA does.
To prove this hypothesis, we generated two RRM2 mutants with substitution of Ala or Glu for Ser-20 by site-directed mutagenesis. They were named S20A and S20E, respectively. We then examined the effects of these mutants on the transcriptional activity of ␤-catenin/LEF. As shown in Fig. 7D, S20A showed a stronger whereas S20E showed a weaker inhibitory effect on the reporter gene activity induced by Wnt3a and ⌬N-␤-catenin than WT RRM2 did (Fig. 7D). Consistent with our hypothesis, these results suggested that RRM2 may act as an inhibitor downstream of ␤-catenin and that Wnt signaling can relieve the RRM2-induced inhibitory effect on ␤-catenin-LEF/T cell factor (TCF) transcriptional activity by stimulating its phosphorylation at Ser-20.

DISCUSSION
Phosphorylation changes in signal transduction are not easily measured because the biological and technical limitations in phosphorylation identification, such as instability, low abundance, highly heterogeneity, and ion suppression (57,58), prevent the straightforward analysis of phosphorylation sites. In the present study, we used a combined strategy that comprised phosphoprotein enrichment, SILAC, and time course analysis to profile the differential phosphoproteome in the early activation of canonical Wnt signaling.
By using such a combined strategy, we not only can measure the phosphorylation level by quantitating most non-phosphopeptides of the enriched phosphoproteins but also directly identify and quantitate site-specific phosphorylations. A total of 287 phosphoproteins showed a 1.5-fold or greater change upon 1 or 30 min of Wnt3a stimulation ( Fig. 4A and Supplemental Table 1). Moreover 100 unique phosphorylation sites were identified, and 54 of them were quantitated (Table  III). For example, although we could not identify phosphorylation sites from ␤-catenin, the phosphorylation of ␤-catenin was quantitated according to non-phosphorylation peptides. Phosphorylation of ␤-catenin was found to be almost unchanged upon 1 min of Wnt3a stimulation and showed a marked decrease upon 30 min of stimulation (Table II and Fig.  4B). This finding was not only confirmed by Western blotting analysis (Fig. 1D) but was also consistent with previous studies on phosphorylation changes of ␤-catenin in canonical Wnt signaling (13,44). On the other hand, the overall phosphorylation level of RRM2 was not significantly altered by Wnt3a treatment (Fig. 1D and Supplemental Table 6), whereas Ser-20 phosphorylation on RRM2 showed an approximately 1.56-fold increase upon 30 min of Wnt3a stimulation ( Fig. 6 and Table IV).

Implications for New Components Involved in Canonical
Wnt Signaling-A recent genome-wide RNA interference screen that led to the identification of 238 potential Wnt signaling regulators has enriched this signaling network to a great extent (59). In the present study, we found 287 differentially regulated proteins whose phosphorylation events may be involved in canonical Wnt signaling cascades. Eleven proteins previously known to be Wnt signaling components were identified (Table II), whereas only phosphorylation regulation of ␤-catenin in this signaling has been reported previously (13,44). Hence our study has suggested that phosphorylation changes of the 10 other proteins may be an additional regulatory or responsive mode upon Wnt3a stimulation.
In addition to 11 proteins involved in Wnt signaling, 276 candidate proteins can be viewed as potential targets of Wnt signaling. Recently a growing body of research has shown that canonical Wnt signaling is involved in cross-talk with several signal pathways such as Sonic hedgehog (Shh), Notch, Hox, fibroblast growth factor, and transforming growth factor ␤ (60 -63), and some of these cross-talks are dependent on phosphorylation modulation (61,63). Among 276 candidate proteins, 28 proteins are reported to be involved in a spectrum of signaling pathways (Fig. 3B). These proteins are likely interaction nodes for canonical Wnt and other signaling pathways. For example, protein phosphatase 2A (PP2A) is a multifunctional serine/threonine phosphatase that has been reported to be involved in Wnt signaling (43), and it has also been implicated in calcium signaling through interaction with a calmodulin-binding protein, striatin (64). In the present study, PP2A and striatin were both identified and, more interestingly, showed decreased phosphorylation upon Wnt3a stimulation (Figs. 4A and 5). Therefore, it is likely that PP2A and striatin constitute a signal integration platform for calcium and canonical Wnt signaling.
Profile Patterns of Differentially Regulated Proteins-Rather than a static analysis, a SILAC-based quantitative proteomics approach has been able to define phosphorylation changes in a complex signal transduction process (20,21,46). Activation profiles of individual proteins were constructed according to -fold change of stimulated state over control state. In the present study, 287 proteins were differentially regulated by Wnt3a stimulation and were characterized into four distinct activation modes (Fig. 4A). Known regulators or effectors of Wnt signaling exhibited different profiles upon Wnt3a stimulation (Fig. 4B). In contrast to other previously reported posi-

FIG. 7. RRM2 negatively regulates canonical Wnt signaling, and Ser-20 phosphorylation counteracts the inhibition effect of RRM2.
A, RRM2 suppressed Wnt activity, which was detected by luciferase reporter assay (upper panel). HEK293 cells were transfected with LEF-luciferase and RRM2 (10 or 25 ng/well), and 18 h later the cells were stimulated with Wnt3a-conditioned medium for 6 h and then assayed. RRM2 did not affect the free ␤-catenin abundance (lower panel). After 18 h of transfection of LacZ, RRM2, or GSK3␤, HEK293 cells were stimulated with Wnt3a-conditioned medium for 3 h. The cells were collected, and the abundance of endogenous free ␤-catenin, ␤-tubulin, and RRM2-HA/GSK3␤-HA was determined by Western blotting analysis. B, RRM2 had no effect on TNF␣-mediated NF-B activity. After 18 h of transfection of NF-B-luciferase and RRM2 or IB␣-SR, HEK293 cells were stimulated with TNF␣ (10 ng/ml) for 6 h and then assayed. C, the knockdown effect of siRRM2 was analyzed by Western blotting, and knockdown of RRM2 by siRNA released its inhibition effect on basal activity and ⌬N-␤-catenin activity but had little influence on Wnt activity. HEK293 cells were transfected with siRNA and the LEF-reporter gene system with and without ⌬N-␤-catenin. Wnt3a-conditioned medium was added after 48 h of transfection. The luciferase activity was measured 6 h later. D, biological activities of RRM2 WT and mutants (S20A and S20E) were monitored using LEF-luciferase reporter under the stimulation of Wnt3a-conditioned medium or expression of ⌬N-␤-catenin. The plasmids (RRM2 WT, S20A, and S20E) were expressed in HEK293 cells, respectively, and the cells were co-transfected with the LEF-reporter gene system with and without ⌬N-␤-catenin. GSK3␤ was used as a positive control; it functions downstream of Wnt and upstream of ␤-catenin. 18 h post-transfection, Wnt3a-conditioned medium was added, and luciferase activity was determined 6 h later. Luc, luciferase; L, a low dose of RRM2 (10 ng/well); H, a high dose of RRM2 (25 ng/well). tive regulators, the phosphorylation of PP2Ac was downregulated upon Wnt3a stimulation (Fig. 4B). This indicated an up-regulated activity of PP2Ac in canonical Wnt signaling because activity of PP2Ac is negatively correlated with its phosphorylation status (according to Swiss-Prot annotation). Occludin, which is reported to negatively regulate canonical Wnt signaling, showed decreased phosphorylation upon Wnt3a stimulation (Fig. 4B). Three effectors were up-regulated in canonical Wnt signaling, but ␤-catenin was not (Fig.  4B); its phosphorylation is negatively regulated by Wnt signaling (44).
Interestingly the proteins with similar profiles may represent a group of co-regulated proteins in canonical Wnt signaling. For example, ␤-catenin showed almost no change upon 1 min of Wnt3a stimulation but showed decreased phosphorylation upon 30 min of stimulation (Fig. 4B). A recent study has identified ezrin-radixin-moesin-binding phosphoprotein 50 (EBP50) as a positive regulator of canonical Wnt signaling (65). It has two PDZ domains and interacts with ␤-catenin through its carboxyl PDZ domain both in vitro and in vivo (65). Shibata et al. (65) also showed that EBP50 promotes ␤-catenin-mediated transactivation only in cells in which ␤-catenin is already stabilized and suggested that EBP50 may work with stabilized ␤-catenin for transcriptional regulation. In the present study, EBP50 showed a marked up-regulation upon 30 min of Wnt3a stimulation; this is in line with the accumulation of "stabilized" ␤-catenin at this time point (Fig. 4B). Consistent with other reports (65), the present finding suggests that EBP50 and ␤-catenin may be co-regulated in canonical Wnt signaling. Besides EBP50, five other proteins identified in the present study have also been reported as positive regulators (Table II). However, some of them display profile patterns different from that of EBP50 (Fig. 4B). For example, RanBP3 has been reported to enhance the nuclear export of active ␤-catenin (49), and in the present study RanBP3 was found to quickly increase upon Wnt3a stimulation, much earlier than the change of ␤-catenin (Fig. 4B). Unlike EBP50 as a positive regulator functioning at the level of ␤-catenin, the profile pattern indicates that RanBP3 is most likely to function upstream of ␤-catenin. Thus, different profile patterns of the quantitated phosphoproteins might suggest their distinct functional roles in canonical Wnt signaling.
Identification of Phosphorylation Sites Associated with Canonical Wnt Signaling-Through rigorous manual inspection (see "Experimental Procedures"), in total we identified 100 unique phosphorylation sites in 90 peptides and quantitated 54 phosphorylation sites/47 phosphopeptides among which 27 sites/23 phosphopeptides were differentially influenced upon Wnt3a stimulation (Table III). These phosphorylation sites/phosphopeptides are likely to be downstream effectors in canonical Wnt signaling. E1B-AP5 is a member of the hnRNP family and plays an important role in mRNA processing and transcription regulation (66,67). E1B-AP5 has been found to interact with p53 and inhibit p53 transcriptional ac-tivity (68), and in another study, Levina et al. (69) revealed that activated p53 can down-regulate ␤-catenin in a way in which phosphorylation of ␤-catenin is enhanced. It is likely that E1B-AP5 may act as a positive regulator by inhibiting the activity of p53 in Wnt signaling. In our analysis, Ser-94 phosphorylation in phosphopeptide GRS#PQPPAEEDEDDFDDT-LVAIDTYNCDLHFK (where # indicates phosphorylation modification) of E1B-AP5 showed an approximately 1.8-fold increase upon 30 min of Wnt3a stimulation (Supplemental Table 5). Therefore, it is worth studying the functional role of Ser-94 phosphorylation of E1B-AP5 in canonical Wnt signaling.
Ser-20 Phosphorylation on RRM2 Counteracts the Inhibition Effect of RRM2 on Canonical Wnt Signaling-Ribonucleoside-diphosphate reductase, consisting of M1 and M2 subunits, is known as a catalytic enzyme responsible for the de novo conversion of ribonucleoside diphosphates to deoxyribonucleoside diphosphates, which are essential for DNA synthesis and repair (70,71). A previous study showed that RRM2 is phosphorylated at serine 20 by p34cdc2 kinase. (72). However, the enzyme activity of this protein is not affected by Ser-20 phosphorylation (72). To our knowledge, the biological significance of Ser-20 phosphorylation on RRM2 has not been characterized so far. In the present study, our data suggested that RRM2 may act as an inhibitor downstream of ␤-catenin and that Wnt signaling can relieve the RRM2-induced inhibitory effect on ␤-catenin-LEF/TCF transcriptional activity by stimulating its phosphorylation at Ser-20 (Table IV and Figs. 6  and 7). Thus, besides the well known role for RRM2 in DNA synthesis (70,71), we found a new function for Ser-20 phosphorylation as well as RRM2 in regulation of canonical Wnt signaling.
It has been reported that ␤-catenin is phosphorylated by a destruction complex composed of GSK3␤, APC, and Axin in the absence of Wnt stimulation, and the hyperphosphorylation leads to its degradation by the proteasome (73,74). When Wnt activates its cell surface receptors consisting of Frizzled and low density lipoprotein receptor-related protein-5/6, the phosphorylation of ␤-catenin is inhibited (75). This causes the cytoplasmic accumulation and nuclear translocation of ␤-catenin, which activates target gene transcription by binding to LEF/TCF (76). It is a widely accepted model that the strength of the canonical Wnt signal can be manipulated by altering the protein levels of ␤-catenin (13,44,77). In addition to regulating the stability of ␤-catenin, our present study suggested another regulatory mechanism for canonical Wnt signaling, that the transcriptional activity of ␤-catenin can be modulated by RRM2 and its Ser-20 phosphorylation. When the Wnt signal is absent, RRM2 may suppress transcriptional activity of the remaining low level ␤-catenin, resulting in a relatively high threshold for target gene transcription. Upon Wnt stimulation, Ser-20 phosphorylation of RRM2 was triggered. This phosphorylation may block the inhibitory role of RRM2 and release ␤-catenin from the control of RRM2, ena-bling the smooth transduction of the Wnt signal. It has been found that Ser-20 of RRM2 can be phosphorylated by p34cdc2 kinase (72). It would be of great interest to investigate whether p34cdc2 or another kinase is involved in canonical Wnt signaling to induce Ser-20 phosphorylation of RRM2. Further study of the molecular mechanisms by which RRM2 suppresses transcriptional activity of ␤-catenin and by which Ser-20 phosphorylation of RRM2 releases such an inhibitory effect would provide significant new insight into canonical Wnt signaling.
In summary, this study extended our knowledge of canonical Wnt signaling by identifying previously uncharacterized downstream signaling proteins, establishing profile patterns for individual proteins, and demonstrating the positive role of Ser-20 phosphorylation of RRM2 in Wnt signaling. Our comprehensive study of phosphorylation regulation in canonical Wnt signaling should serve as a valuable resource for future research in the field.
As can be seen, although phosphoprotein enrichment was applied in the study, only 100 phosphorylation sites were significantly identified in 90 phosphopeptides. Following phosphoprotein enrichment, further application of phosphopeptide enrichment/fractionation techniques such as IMAC (20), strong cation exchange (78), and strong cation exchange-strong anion exchange (79) would improve identification of many more phosphorylation sites and phosphopeptides from signal proteins. Moreover combination of such quantitative phosphoproteome profiling with subcellular fractionation could be valuable for deciphering both time course and spatial gradients of signaling network in cell signaling studying. * This work was supported by National Natural Science Foundation Grants 30425021 and 30521005; Ministry of Science and Technology of China Grants 2002CB513000, 2006CB910700, and 2007CB914500; Shanghai Key Project of Basic Science Research Grant 04DZ14005; and grants from the Chinese Academy of Sciences and the Proteomage Project from FP6 of the European Union. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www. mcponline.org) contains supplemental material.
ʈ Both authors contributed equally to this work. § § To whom correspondence should be addressed. E-mail: zr@sibs.ac.cn.