Mechanobiology June 26th - June 2nd 2016

Mechanobiology: June 26th  - June 2nd 2016

Wnt3a-stimulated LRP6 phosphorylation is dependent upon arginine methylation of G3BP2
Rama Kamesh Bikkavilli, Craig C. Malbon

Summary

Wnt signaling is initiated upon binding of Wnt proteins to Frizzled proteins and their co-receptors LRP5 and 6. The signal is then propagated to several downstream effectors, mediated by the phosphoprotein scaffold, dishevelled. We report a novel role for arginine methylation in regulating Wnt3a-stimulated LRP6 phosphorylation. G3BP2, a dishevelled-associated protein, is methylated in response to Wnt3a. The Wnt3a-induced LRP6 phosphorylation is attenuated by G3BP2 knockdown, chemical inhibition of methyl transferase activity or expression of methylation-deficient mutants of G3BP2. Arginine methylation of G3BP2 appears to be a Wnt3a-sensitive ‘switch’ regulating LRP6 phosphorylation and canonical Wnt–β-catenin signaling.

Introduction

Wnt–β-catenin signaling is essential for normal embryonic development and adult tissue homeostasis (Polakis, 2000; Moon et al., 2002; Logan and Nusse, 2004; Malbon, 2005a; Malbon, 2005b; Reya and Clevers, 2005; Polakis, 2007). Wnt signaling is initiated upon binding of Wnt to Frizzled and its co-receptor low-density lipoprotein receptor-related protein 5 or 6 (LRP5/6). The signal is then propagated downstream, mediated by a phosphoprotein scaffold, dishevelled (Dvl), leading to post-transcriptional and post-translational mechanism-mediated stabilization of β-catenin (Malbon and Wang, 2006; Angers and Moon, 2009; Bikkavilli and Malbon, 2010; Bikkavilli and Malbon, 2011). Nuclear accumulation of the stabilized β-catenin follows stimulatory activation of Lef/Tcf-sensitive transcription of developmentally regulated genes (Behrens et al., 1996; Molenaar et al., 1996).

Wnt binding to Frizzled and LRP6 produces a stable complex that aggregates into LRP6 signalosomes (He et al., 2004; MacDonald et al., 2008; Niehrs and Shen, 2010). The LRP6 signalosome formation stimulates phosphorylation of LRP6 at multiple sites through distinct kinases, including casein kinase 1γ (CK1γ) and glycogen synthase kinase-3β (GSK3β) (Tamai et al., 2004; Davidson et al., 2005; Zeng et al., 2005). Wnt-stimulated LRP6 phosphorylation provokes recruitment of Axin to LRP6. This preferential binding of Axin to phosphorylated LRP6 releases β-catenin from its regulatory degradation and promotes its stabilization (Tamai et al., 2004; Zeng et al., 2005). The Wnt-stimulated LRP6 phosphorylation is also tightly regulated. Dikkopf (DKK), for example, antagonizes Wnt signaling through LRP6 binding and internalization (Brott and Sokol, 2002; Semënov et al., 2008; Binnerts et al., 2009; Li et al., 2010). Similarly, CK1γ and GSK3β have also been shown to regulate LRP6 activity in response to Wnt (Zeng et al., 2005). A recent study also highlighted an important role for phosphatidylinositol 4,5-bisphosphate in regulation of LRP6 phosphorylation (Pan et al., 2008).

Although, Dvl proteins play an important role in regulating LRP6 phosphorylation (Bilic et al., 2007), the mechanism involved has not been examined. Dvl proteins are dynamic, scaffolding molecules (Malbon and Wang, 2006; Gao and Chen, 2010), which can oligomerize (Schwarz-Romond et al., 2007) and/or polymerize into large supermolecular structures (Yokoyama et al., 2010). However, the cellular signals that provoke their formation remain unclear. We previously reported an important role for arginine methylation as a ‘molecular switch’ in regulating the disassembly of G3BP1 from the Dvl3-based supermolecular complexes (Bikkavilli and Malbon, 2011). In the present study, we identified a role for arginine methylation and its substrate Ras GTPase-activating protein-binding protein 2 (G3BP2) in regulating Wnt3a-stimulated LRP6 phosphorylation. Depletion of G3BP2, chemical inhibition of protein methyl transferase activity, and expression of methylation-deficient G3BP2 attenuated Wnt3a-induced LRP6 phosphorylation, revealing previously unknown modes of regulation of the Wnt pathway.

Results

G3BP2 is a novel Dvl3-associated protein

We performed proteomic analysis of Dvl3-based supermolecular complexes (Yokoyama et al., 2010) following affinity pull-downs (Bikkavilli and Malbon, 2010; Bikkavilli and Malbon, 2011). Many peptides (in blue in Fig. 1A) constituting more than 50% of sequence identified Ras GTPase activating protein-binding protein 2 (G3BP2) as a Dvl3-associated protein (Fig. 1A). The primary structure of G3BP2, similar to its G3BP1 isoform identified previously (Bikkavilli and Malbon, 2011), has an N-terminal nuclear transport factor 2-like domain (NTF2), an RNA recognition motif (RRM domain) and a glycine–arginine-rich (GAR) region (Fig. 1B).

Fig. 1.

G3BP2 is a Dvl3-associated protein. (A) Protein sequence of G3BP2. The peptide sequences of G3BP2 identified in Dvl3-based complexes are in blue. (B) Schematic of the organization of G3BP2 and G3BP1 showing a conserved NTF2-like domain, an RNA recognition motif, a glycine–arginine-rich motif and SH3-binding motifs. (C) Cell lysates from F9 cells transfected with Myc–G3BP2 were immunoprecipitated with either control mouse IgG or anti-Dvl3 antibodies followed by immunoblotting with anti-Myc antibodies to detect G3BP2. (D) F9 cells were transiently transfected with Myc-G3BP2 either alone or together with HA-Dvl3-GFP2, followed by cell lysis and affinity pull-downs with anti-HA antibodies. Dvl3–G3BP2 interaction was made visible by probing the blots with anti-Myc antibodies. (E) F9 cells were transiently transfected with either HA-Dvl1-GFP2, HA-Dvl2-GFP2 or HA-Dvl3-GFP2 alone or with Myc-G3BP2. Cells were then lysed and affinity pull-downs were carried out with anti-HA antibodies. Dvl3–G3BP2 interaction was made visible by probing the blots with anti-Myc antibodies. (F) HEK 293 cells were transfected with Myc-G3BP2 for 24 hours. The cells were then treated with Wnt3a (10 ng/ml) for the indicated periods, lysed and affinity pull-downs were carried out with anti-Dvl3 antibodies. Dvl3–G3BP2 interaction was made visible by probing the blots with anti-Myc antibodies.

Dvl3 pull-downs performed with lysates of F9 teratocarcinoma cells expressing Myc–G3BP2 confirmed the Dvl3–G3BP2 association (Fig. 1C). Such interactions were not obtained in control IgG pull downs (Fig. 1C). Pull-downs performed with anti-hemagluttinin (HA) antibodies on lysates of F9 cells expressing exogenous Dvl3 (HA–Dvl3) and Myc–G3BP2 also demonstrated a positive Dvl3–G3BP2 association (Fig. 1D). Interestingly, G3BP2 showed greater binding to Dvl3 than to the other isoforms, Dvl1 and Dvl2 (Fig. 1E). Dvl3–G3BP2 association is not confined to F9 cells; HEK293 cells also displayed Dvl3–G3BP2 association (Fig. 1F). G3BP2 association with Dvl3-based supermolecular complexes increased in cells treated with Wnt3a (Fig. 1F).

G3BP2 depletion blocks Wnt–β-catenin signaling

Because the Dvl3–G3BP2 association was Wnt3a sensitive (Fig. 1E), we wondered whether G3BP2 contributed directly or indirectly to Wnt–β-catenin signaling. A small interfering RNA (siRNA) was designed against G3BP2 and tested for its effects on G3BP2 expression and on three read-outs stimulated by Wnt: β-catenin stabilization, Lef/Tcf-sensitive gene transcription and primitive endoderm (PE) formation (Fig. 2). Treatment of clones stably expressing Fz1 receptor with G3BP2-specific siRNA effectively knocked down (>90%) G3BP2 expression [Fig. 2A; the additional band (~52 KDa) corresponds to G3BP2β, a spliced isoform of G3BP2α (French et al., 2002; Irvine et al., 2004)]. Loss of G3BP2 sharply attenuated Wnt3a-stimulated β-catenin accumulation (Fig. 2B). G3BP2 knockdown also attenuated Wnt3a-stimulated Lef/Tcf-sensitive gene transcription (Fig. 2C). Wnt3a induces PE formation as shown by positive staining of cytokeratin endoA, a hallmark for PE formation (Liu et al., 2002; Bikkavilli et al., 2008a; Bikkavilli and Malbon, 2010). Loss of G3BP2 abolished Wnt3a-induced PE formation (Fig. 2D). G3BP1 isoform was shown previously to regulate β-catenin mRNA (Bikkavilli and Malbon, 2011). We next investigated whether G3BP2 could also modulate β-catenin mRNA levels. Unlike the inhibitory effects on Wnt-stimulated β-catenin, Lef/Tcf-sensitive gene transcription and PE formation, loss of G3BP2 resulted in elevated β-catenin mRNA levels (Fig. 2E). We also tested whether G3BP2, similar to its G3BP1 isoform (Bikkavilli and Malbon, 2011), could bind β-catenin mRNA in vitro, using northwestern blotting (Fig. 2F). Digioxigenin (DIG)-labeled β-catenin 3′ untranslated region (UTR) probe bound to both the G3BPs (Fig. 2F). These observations suggest an important role for G3BP2 in Wnt–β-catenin signaling.

G3BP2 is methylated in response to Wnt3a treatment

The C-terminus of G3BP2 contains GAR motifs. Such GAR motifs are prominent targets for protein arginine methyl transferases [PRMTs (Bedford and Richard, 2005; Bedford, 2007; Bedford and Clarke, 2009; Lee and Stallcup, 2009)], so we characterized the expression profile of PRMTs (PRMT1–8) in F9 cells. Reverse transcriptase-polymerase chain reaction (RT-PCR) of mouse total RNA with PRMT-specific primers revealed the expression of five PRMTs: 1, 2, 5, 7 and 8 (Fig. 3A). The expressed PRMTs are classified as type I (PRMT1 and 8), type II (PRMT5 and 7) and type IV (PRMT2), based on their abilities to catalyze either symmetrical or asymmetrical di-methyl arginine formation (Bedford, 2007). We cloned the expressed PRMTs into tagged expression vectors (pCMV-HA) to determine whether PRMTs could interact and/or methylate G3BP2. HA-tagged PRMTs were then expressed in F9 cells; whole-cell lysates were probed for PRMT–G3BP2 interaction using pull-downs (Fig. 3B). HA-affinity pull-downs revealed a specific interaction of G3BP2 with four PRMTs (PRMT1, 2, 7 and 8; Fig. 3B).

Fig. 2.

G3BP2 regulates Wnt–β-catenin signaling. (A) F9 cells were treated with either control siRNA (100 nM) or siRNA specific to mG3BP2 (100 nM) for 48 hours. (B,C) G3BP2 knockdown efficiency was visualized by probing the blots with anti-G3BP2 antibodies. F9 cells were treated with either control siRNA (100 nM) or siRNA specific for G3BP2 (100 nM) for 48 hours and the lysates were assayed either for cytosolic β-catenin levels (B) or Lef/Tcf-sensitive gene transcription (C). Values are means ± s.e.m. obtained from three independent experiments; **P<0.01; versus control (−Wnt3a). ##P<0.01; versus control (+Wnt3a). The lower panel in B shows representative blots. (D) F9 cells were treated with either control siRNA (100 nM) or siRNA specific to G3BP2 (100 nM) for 48 hours prior to stimulation with Wnt3a (20 ng/ml) for 4 days. Formation of primitive endoderm was assayed by immunocytochemistry with anti-cytokeratin A antibody. Scale bars: 50 μm. (E) F9 cells were treated with either control siRNA (100 nM) or siRNA specific to G3BP2 (100 nM) for 48 hours and the β-catenin and cyclophilin A mRNA levels from total RNA were quantified using real-time PCR. The data are presented as β-catenin mRNA levels normalized to the cyclophilin A mRNA levels (means ± s.e.m.) from two independent experiments whose results were in high agreement. **P<0.01 versus control (control siRNA). (F) Northwestern analysis of β-catenin mRNA interaction with G3BP1 or G3BP2. Myc–G3BP1 or Myc–G3BP2 immunoprecipitated from F9 cell lysates were separated on SDS-PAGE gels and transferred onto nitrocellulose membranes. Northwestern analysis was then performed using DIG-labeled β-catenin UTR. The upper panel is a northwestern blot and the lower panel is an immunoblot with anti-Myc antibodies.

We tested whether PRMT1, 2, 7 and 8 catalyze G3BP2 methylation (Fig. 3C). Using PRMTs purified from F9 cells and recombinant G3BP2 (rG3BP2) isolated from Escherichia coli, in vitro methylation was assayed in the presence of the methyl donor, [3H]S-adenosyl L-methionine (Fig. 3C). PRMT1, 7 and 8 each readily catalyzed G3BP2 methylation (Fig. 3C). Remarkably, despite an ability to bind G3BP2, PRMT2 failed to catalyze G3BP2 methylation (Fig. 3C). We extended these findings to G3BP2 methylation in vivo, using metabolic labeling of cells with [3H]L-methyl methionine (Fig. 3D). Cells transiently transfected with either pCMV-Myc or Myc-G3BP2 and metabolically labeled were subjected to pull-down. Anti-Myc antibody pull-downs, SDS-PAGE and fluorography demonstrated methylation of G3BP2 and increased methylation of G3BP2 in the presence of Wnt3a (Fig. 3D).

G3BP2 regulates LRP6 phosphorylation

We investigated a role for G3BP2 at the most proximal point of Wnt signaling, LRP6 phosphorylation. Expression of LRP6ΔN (LRP6 lacking most of its extracellular domain) has been shown to stimulate β-catenin stabilization and Lef/Tcf-sensitive gene transcription (Brennan et al., 2004; Tamai et al., 2004; Zeng et al., 2005). Expression of LRP6ΔN indeed induced robust Lef/Tcf-sensitive gene transcription, as did Wnt3a alone (Fig. 4A). Knockdown of G3BP2 was found to attenuate LRP6ΔN-induced Lef/Tcf-sensitive gene activation (Fig. 4A). Knockdown of G3BP2 strongly suppressed Wnt3a-complimented LRP6ΔN action (Fig. 4A), suggesting G3BP2 regulation of Wnt3a action through LRP6.

Fig. 3.

PRMTs catalyze G3BP2 methylation. (A) PRMT expression in F9 cells. Total RNA from F9 cells was isolated and RT-PCR was performed using mouse PRMT-specific primers. (B) Lysates (1 mg) of F9 cells transfected with either pCMV-HA or HA-PRMTs were incubated with 1 μg GST–G3BP2 and immunoprecipitated with anti-HA antibodies and protein-G resin. The immunoblots were later probed for G3BP2–PRMT interaction with anti-G3BP2 antibodies. The levels of input protein (GST–G3BP2) are also shown. (C) F9 cells were transiently transfected with either pCMV-HA or HA-PRMTs. After 24 hours, the cells were lysed and affinity pull-downs were performed with anti-HA specific antibodies. Immunoprecipitated PRMTs were used in an in vitro methylation reaction, in combination with GST-G3BP2 and [3H]S-adenosyl L-methionine. (D) F9 cells were transiently transfected with either pCMV-Myc or Myc-G3BP2. After 24 hours, the cells were treated with Wnt3a (10 ng/ml), followed by metabolic labeling with [3H]-L-methyl methionine. After 3 hours, the cells were lysed and affinity pull-downs with anti-Myc-specific antibodies were performed. The methylation status of G3BP2 was then revealed by SDS-PAGE followed by fluorography. After fluorography the blots were probed with anti-Myc antibodies (lower panel).

Wnt3a stimulation has been shown to robustly catalyze LRP6 (Ser1490) phosphorylation (Tamai et al., 2004; Zeng et al., 2005; Bilic et al., 2007; Pan et al., 2008). We examined whether G3BP2 could modulate LRP6 activation through protein phosphorylation. The phosphorylation of LRP6 was ascertained by making use of antibodies specific to phosphorylated LRP6 (Ser1490). (Fig. 4B). Knockdown of G3BP2 (Fig. 4B).

Wnt3a-stimulated LRP6 phosphorylation: the role of PRMTs

If a link between protein methylation and LRP6 phosphorylation existed, we could probe it using methyl transferase-specific inhibitors, such as 5′-methylthioadenosine [MTA (Avila et al., 2004; Côté and Richard, 2005)] (Fig. 4D) and adenosine, periodate oxidized [Adox (Chen et al., 2004; Herrmann et al., 2005)] (Fig. 4E). We pre-treated cells with either MTA or Adox and used Wnt3a-stimulated LRP6 phosphorylation as a read-out. MTA effectively blocked LRP6 phosphorylation in response to Wnt3a (Fig. 4D). Adox had a similar effect on LRP6 phosphorylation (Fig. 4E). A role for protein methylation in Wnt3a-induced LRP6 phosphorylation was clear (Fig. 4D,E).

We ascertained whether methylation of G3BP2 was essential for phosphorylation of LRP6. A panel of methylation-deficient mutants of G3BP2 was created based on similarity with the G3BP1 isoform (Bikkavilli and Malbon, 2011) and published literature (Ong et al., 2004). The mutants (Fig. 1A, K substituted for R: R418K, R432K, R438K, R452K, R457K, R468K) were later employed to probe their effects on LRP6 phosphorylation. Expression of two G3BP2 mutants, R432K and R452K, was found to attenuate Wnt3a-induced LRP6 phosphorylation (Fig. 4F). Expression of either R418K or R457K, by contrast, produced only a small decrease in Wnt3a-induced LRP6 phosphorylation (Fig. 4F). Expression of R438K or R468K had no apparent effects on LRP6 phosphorylation (Fig. 4F).

Fig. 4.

G3BP2 regulates LRP6 phosphorylation. (A) F9 cells were treated with either control siRNA or G3BP2-specific siRNA for 24 hours, followed by another round of transfection with either control vector or LRPΔN plasmid. After 24 hours, the cells were stimulated with Wnt3a (10 ng/ml) or left unstimulated for 7 hours and the lysates were assayed for Lef/Tcf-sensitive gene transcription. Values are means ± s.e.m. from three independent experiments. *P<0.05; **P<0.01; versus control (−Wnt3a), ##P<0.01; versus control (+Wnt3a). (B,C) F9 cells were treated with either G3BP2-specific siRNA (B) or G3BP1-specific siRNA (C). After 48 hours, the cells were stimulated with Wnt3a (20 ng/ml) for 30 minutes and the cells were then lysed. The lysates were later assayed for phosphorylated LRP6 (p-LRP6). Upper panel: mean values ± s.e.m. obtained from three independent experiments; lower panel: representative blots. **P<0.01; versus control (−Wnt3a). ##P<0.01; versus control (+ Wnt3a). (D,E) F9 cells were treated with the methyl transferase inhibitor MTA (250 μM; D) or Adox (15 μM; E) for 16 hours prior to stimulation with Wnt3a (20 ng/ml) for 30 minutes. Cell lysates were later assayed for phosphorylated LRP6. Upper panel: mean values ± s.e.m. obtained from three independent experiments; lower panel: representative blots. **P<0.01; versus control (−Wnt3a). ##P<0.01; versus control (+Wnt3a). (F) F9 cells were transiently transfected with either pCMV-Myc, Myc-G3BP2 (WT) or its methylation-deficient mutants. After 24 hours, the cells were stimulated with Wnt3a (20 ng/ml) for 30 minutes, lysed and assayed for phosphorylated LRP6 levels. Left panel: mean values ± s.e.m. obtained from three independent experiments; right panel: representative blots. #P<0.05, ##P<0.01 versus control (pCMV-Myc). (G) HEK 293 cells were transiently transfected with pcDNA3.1-HA-hGSK3β either alone or with Myc-G3BP2 for 24 hours. After 24 hours, the cells were starved in DMEM medium without serum for 2 hours, and then lysed. The cell lysates were later assayed for phosphorylated LRP6. Right panel: mean values ± s.e.m. obtained from three independent experiments; left panel: representative blots. **P<0.01 versus pCMV-Myc; #P<0.05 versus GSK3β alone.

To investigate how G3BP2 might modulate protein kinases acting on LRP5/6, we overexpressed GSK3β. Overexpression of GSK3β stimulates LRP6 Ser1490 phosphorylation (Fig. 4G) (Zeng et al., 2005). Expression of G3BP2 strongly suppressed GSK3β-stimulated LRP6 Ser1490 phosphorylation (Fig. 4G). Therefore, G3BP2 appears to modulate GSK3β-mediated LRP6 phosphorylation (Fig. 4G).

The G3BP2-Dvl interaction is methylation dependent

We next investigated whether G3BP2 methylation impacts Dvl3–G3BP2 association (Fig. 5A). Cells expressing wild-type (WT) or the methylation-deficient mutants of G3BP2 were employed in pull-down assays of Dvl3-based supermolecular complexes (Fig. 5A). Wnt3a stimulation was found to increase the association of both WT and the R418K mutant of G3BP2 with Dvl3-based complexes (Fig. 5A). Expression of the mutants R432K, R438K, R452K or R468K, in sharp contrast, attenuated Wnt3a-stimulated association of G3BP2 with Dvl3-based complexes (Fig. 5A). Methylation of G3BP2 (at R432, R438, R452 or R468) appears indispensable for proper assembly of G3BP2 into Dvl3-based complexes in response to Wnt3a (Fig. 5A).

Expression of methylation-deficient mutants of G3BP2 block Wnt–β-catenin signaling

Would expression of G3BP2 mutants unable to form proper Dvl3-based complexes impact Wnt3a activation of Lef/Tcf-sensitive gene expression? Expression of R432K and R452K mutants of G3BP2 suppressed the ability of Wnt3a to stimulate Lef/Tcf-sensitive gene transcription (Fig. 5B). Expression of WT or the R457K mutant of G3BP2, by contrast, showed no effect on Wnt3a action (Fig. 5B). Expression of the R418K mutant actually resulted in a slight increase in Wnt3a-stimulated Lef/Tcf-sensitive gene expression (Fig. 5B). Expression of G3BP2 mutants was probed in SW480 cells also. SW480 cells (derived from an adenocarcinoma of colon) have high levels of β-catenin (Sinner et al., 2007). Expression of four of the mutants (R432K, R438K, R452K and R468K) in SW480 cells effectively suppressed Lef/Tcf-sensitive gene transcription (Fig. 5C). Expression of R418K and R457K, like WT, stimulated Lef/Tcf-sensitive gene transcription (Fig. 5C).

Finally, we performed gene rescue experiments to determine whether G3BP2 methylation is crucial for proper activation of Wnt signaling (Fig. 5G). For this purpose we made a siRNA-resistant G3BP2 construct by engineering silent mutations (underlined) within the siRNA target site (5′-GGGAGAGAGTTTGTACGCCAATATT-3′). Depletion of G3BP2 (as shown earlier, Fig. 2C) suppressed Wnt-stimulated Lef/Tcf-sensitive gene transcription (Fig. 5D). Only G3BP2 cDNAs that were resistant to siRNA treatment, but not the methylation-deficient mutant of siRNA-resistant G3BP2 (R432K, R452K and R468K) or that of G3BP1, rescued Wnt-stimulated Lef/Tcf-sensitive gene transcription in G3BP2-depleted cells (Fig. 5D). Clearly, G3BP2 methylation mediates regulation of Wnt signaling.

Discussion

Wnt induces robust LRP6 phosphorylation. Several kinases, most importantly GSK3β, have been reported to phosphorylate LRP6 on a PPPS/TP motif (Tamai et al., 2004; Zeng et al., 2005; Bilic et al., 2007; Pan et al., 2008; Červenka et al., 2011). We reveal a novel role for protein arginine methylation in regulating LRP6 phosphorylation as well as an unanticipated role for G3BP2 in the regulation of LRP6 phosphorylation and Wnt–β-catenin signaling. The identification of G3BP2 evolved from sequence homology with G3BP1 (Irvine et al., 2004). G3BP1 and G3BP2 protein sequences are 60% identical, have an N-terminal nuclear transport factor 2-like domain, an RNA recognition motif (RRM) and RGG motifs (Irvine et al., 2004). Despite high homology at the protein level, G3BP1 and 2 are not functionally redundant, i.e. lost function of G3BP2 cannot be rescued by G3BP1 expression (Fig. 5D).

In the current work, G3BP2 is shown to be a positive regulator of Wnt signaling. G3BP2 acts at the level of LRP6 phosphorylation by modulating the activity of GSK3β (Fig. 4). Wnt3a stimulates G3BP2 methylation (Fig. 3). In addition, Wnt3a-stimulated LRP6 phosphorylation is shown to be sensitive to chemical inhibition of methyl transferase activity as well as to expression of methylation-deficient mutants of G3BP2 (Fig. 4), revealing a previously unknown and important link between arginine methylation and LRP6 phosphorylation. Furthermore, similar to the G3BP1 isoform (Bikkavilli and Malbon, 2011), G3BP2 also regulates β-catenin mRNA levels. Unlike G3BP1 depletion, which results in a concurrent increase in both the mRNA and protein levels of β-catenin, G3BP2 depletion provokes elevated β-catenin mRNA (Fig. 2). These observations suggest that G3BP2, in addition to regulating LRP6 phosphorylation, might also play an important role in β-catenin mRNA regulation, the details of which remain obscure. The role of G3BP2 methylation is not limited to these functions because G3BP2 and PRMTs are frequently overexpressed in many human tumors (Guitard et al., 2001; Liu et al., 2001; French et al., 2002; Mathioudaki et al., 2008). Dysfunctional G3BP2 methylation may play a key role in specific disease states [HeLa cells, R457 and R468 (Ong et al., 2004)].

Fig. 5.

Arginine methylation of G3BP2 stimulates dishevelled binding and regulates Wnt signaling. (A) F9 cells were transfected with either pCMV-Myc, Myc–G3BP2 (WT) or its methylation-deficient mutants for 24 hours. Then cell lysis and affinity pull-downs were performed with anti-Dvl3 antibodies. G3BP2–Dvl3 interaction was made visible by probing the blots with anti-Myc antibodies. (B) F9 cells were transiently transfected with pCMV-Myc, Myc-G3BP2 (WT) or its methylation-deficient mutants for 24 hours. After 24 hours, the cells were stimulated with Wnt3a (10 ng/ml) for 7 hours and the lysates were assayed for Lef/Tcf-sensitive gene transcription. (C) SW480 cells were transfected with either pCMV-Myc, Myc-G3BP2 (WT) or its methylation deficient mutants together with 10 ng of M50 luciferase reporter, using Fugene transfection reagent (Promega). After 24 hours, the cells were lysed and the lysates were assayed for Lef/Tcf-sensitive gene transcription. For B and C, top panels: mean values ± s.e.m. obtained from three independents; bottom panels: representative blots. **P<0.01; versus control (pCMV-Myc), ##P<0.01; versus control (wild-type). (D) Rescue experiments were performed in F9 cells by transfecting pCMV-Myc, or plasmids encoding either the siRNA-resistant wild-type or the siRNA-resistant methylation-deficient triple mutant (R432K, R452K, R468K) of G3BP2 or wild-type G3BP1 into G3BP2-depleted cells (as described in the Materials and Methods). After 24 hours, the cells were stimulated with Wnt3a (25 ng/ml) for 7 hours and the lysates were assayed for Lef/Tcf-sensitive gene transcription. Left panel: mean values ± s.e.m. obtained from three independent experiments; right panels: representative blots. ##P<0.01; versus control (control siRNA + Wnt3a).

Overexpression of wild-type G3BP2 resulted in a modest decrease in Wnt3a-induced LRP6 phosphorylation (Fig. 4F), with no significant decrease in Wnt3a-induced Lef/Tcf-sensitive gene transcription (Fig. 5B,C). We propose that overexpression of G3BP2 might not induce a corresponding increase in Wnt3a-stimulated methylation of G3BP2 by PRMTs (as PRMTs themselves were not over expressed). Wnt3a-dependent G3BP2 methylation might be affecting more than one Wnt-sensitive process. Differential sensitivities of these processes to G3BP2 methylation best explain the observed differences in effects upon Wnt-stimulated LRP6 phosphorylation versus Lef/Tcf-sensitive transcription.

Dvl proteins have conserved domains (DIX, PDZ and DEP) and a C-terminus that provides docking sites for many proteins (Malbon and Wang, 2006; Gao and Chen, 2010). In vivo, Dvl proteins appear as either large aggregates or polymers (Schwarz-Romond et al., 2007). These supermolecular size complexes (>5–10 MDa) become larger in response to Wnt stimulation (Yokoyama et al., 2010). We report herein a link between arginine methylation and assembly of Dvl3-based complexes. G3BP2 itself is shown to be a novel Dvl3-associated protein (Fig. 1). Wnt3a stimulation induces rapid recruitment of G3BP2 into Dvl3-based complexes. G3BP2 association to Dvl3-based complexes is also methylation sensitive (Fig. 5A). Methylation of G3BP2 at R432, R438, R452 and R468 is crucial, not only for G3BP2 docking onto Dvl3, but also for proper activation of Wnt–β-catenin signaling. G3BP2 methylation at R432 and R452, by contrast, was sufficient for Wnt-stimulated LRP6 phosphorylation. Methylation of G3BP2 at R438 and R468 might play a role in β-catenin mRNA regulation. The effects of methylation-deficient mutants in SW480 cells also support this hypothesis. In addition, the role of LRP6 phosphorylation in β-catenin mRNA regulation deserves in-depth study and cannot be dismissed prematurely. Furthermore, methylation of arginine residues flanking proline-rich motifs has been shown to alter binding by SH3 domains (Bedford et al., 2000; Bedford, 2007). Interestingly, R432 and R438 of G3BP2, which are crucial for Dvl3 binding, LRP6 phosphorylation and Lef/Tcf-sensitive gene transcription, have a flanking PxxP motif (RGPGGPRG).

Fig. 6.

Schematic representation of G3BP2-mediated regulation of LRP6 phosphorylation and Wnt signaling. Wnt3a stimulation provokes methylation of G3BP2 by PRMTs, resulting in increased association of methylated G3BP2 with the Dvl3-based complexes. The Dvl3-based complexes (including methylated G3BP2 and kinases CKI and GSK3β) shuttles to the plasma membrane. LRP6 phosphorylation by GSK3β follows Axin recruitment to the membrane, turning on the Wnt signal.

LRP6 Ser1490 phosphorylation by GSK3β triggers Axin recruitment to the membrane (Tamai et al., 2004; Zeng et al., 2005; Bilic et al., 2007; Pan et al., 2008). Dvl plays a crucial role in trafficking Axin to the cell membrane (Bilic et al., 2007). Earlier, cytoplasmic activation/proliferation-associated protein 2 (Caprin-2) was shown to regulate LRP5/6 action through recruitment of Axin to LRP5/6 (Ding et al., 2008). In the current study, we also identified several peptides of caprin in both Dvl3-based as well as G3BP2-based complexes (data not shown). Therefore, we suggest that the G3BP2-based complex, which includes Caprin might well regulate LRP6 phosphorylation by G3BP2-regulated GSK3β-mediated phosphorylation, and/or Caprin-mediated Axin recruitment to the membrane (Fig. 6). In any case, we reveal arginine methylation as a regulator of Wnt3a-stimulated LRP6 phosphorylation. Wnt3a activates methylation of G3BP2 and recruitment of methylated G3BP2 into Dvl-based supermolecular complexes. The Dvl3–G3BP2 complex later promotes GSK3β-catalyzed phosphorylation of LRP6 on Ser1490. Thus arginine methylation of G3BP2 is revealed as a key element in Wnt–β-catenin signaling.

Materials and Methods

Constructs

Mouse Dvl3 isoform was engineered in-house with GFP2 and HA tags. Mouse G3BP2 was amplified from cDNAs of F9 cells and was subcloned into the pCMV-Myc plasmid in frame with Myc tag sequence. Site-directed mutagenesis was performed on Myc-G3BP2 plasmid using the QuikChange Site Directed Mutagenesis Kit (Stratagene) to obtain Myc–G3BP2 mutants (R418K, R432K, R438K, R452K, R457K and R468K). To generate GST-tagged G3BP2, G3BP2 was subcloned into pGEX4T1 plasmid in frame with the GST protein. To construct siRNA-resistant G3BP2, silent mutations were engineered into the siRNA target site (5′-GGGAGAGAGTTTGTACGCCAATATT-3′; mutated bases are underlined) by using the QuikChange Site Directed Mutagenesis Kit, as described above. Myc-G3BP1 and pCDNA3.1-mβ-catenin 3′-UTRs were generated as described previously (Bikkavilli and Malbon, 2011). Mouse PRMTs were also amplified from cDNAs of F9 cells and were subcloned into pCMV-HA vector in-frame with HA tag sequence. The primers used for cloning are summarized in supplementary material Table S1.

Cell culture

Mouse F9 teratocarcinoma cell stocks were obtained from ATCC (Manassas, VA) and were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 15% heat-inactivated fetal bovine serum (Hyclone) at 37°C in a 5% CO2 incubator. The F9 cells stably expressing rat Fz1 receptor and M50 reporter were generated as described earlier (Bikkavilli et al., 2008b) and employed as a standard in all the experiments. The use of this stable cell line as the starting point for transient transfections reduced variability and offered greater consistency by reducing the number of plasmids required for simultaneous transfections. Human embryonic kidney (HEK) 293 cells were also obtained from ATCC and were maintained in DMEM supplemented with 10% FBS. SW480 cells were obtained from ATCC and were maintained in L15 medium (Mediatech) supplemented with 10% FBS, at 37°C in air (100%).

Co-immunoprecipitation and immunoblotting

For co-immunoprecipitation experiments, F9 clones stably expressing Rfz1 were transiently transfected for 24 hours with 6 μg of plasmid vectors in 100 mm culture dishes. After 24 hours, the cells were lysed in 1 ml lysis buffer (20 mM Tris pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM Na4P2O7, 1 mM β-glycerophosphate, 1 mM Na3VO4, 1 μg/ml leupeptin, 1 μg/ml aprotonin and 1 μg/ml phenylmethylsulphonyl fluoride). The lysates were cleared by centrifugation at 20,000 g for 15 minutes, twice. The supernatants were transferred into new tubes and protein concentrations were determined by the Lowry method (Lowry et al., 1951). Immunoprecipitations were performed using either rat anti-HA high affinity (Roche), mouse-monoclonal anti-Dvl3 (sc 8027; Santa Cruz), mouse-monoclonal anti-Myc antibodies (M4439; Sigma) and protein-G resin (L00209, GenScript, Piscataway, NJ). For LRP6 phosphorylation assays, the cells were lysed in a lysis buffer (137 mM NaCl, 20 mM Tris pH 8.0, 100 mM NaF, 10 mM Na2MO4, 10% glycerol, 1 mM Na2VO4, 1% NP40). For immunoblotting, total lysates (30–60 μg protein/lane) were subjected to electrophoresis in 8% SDS-PAGE gels. The resolved proteins were transferred electrophoretically to nitrocellulose membrane. The blots were incubated with primary antibodies overnight at 4°C and the immunocomplexes were made visible using a secondary antibody coupled to horseradish peroxidase and developed using the enhanced chemiluminescence method. The antibodies were purchased from the following sources: anti-HA high affinity antibody (Roche Applied Science), anti-β-catenin, anti-Myc and anti-β-actin antibodies were from Sigma-Aldrich. Phospho-LRP6 (Ser1490) antibody (#2568) was from Cell Signaling Technology. Anti-LRP6 antibody (sc15399) was from Santa Cruz.

In vitro methylation assays

In vitro methylation assay using bacterially expressed GST–G3BP2 was performed as described previously (Bikkavilli and Malbon, 2011). Briefly, F9 cells were transiently transfected with HA-PRMT1, 2, 5, 7 and 8 (6 μg) in 100 mm culture dishes. After 24 hours of transfection, the cells were lysed in a lysis buffer (20 mM Tris pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 2.5 mM Na4P2O7, 1 mM β-glycerophosphate, 1 mM Na3VO4, 1 μg/ml leupeptin, 1 μg/ml aprotonin and 1 μg/ml phenylmethylsulphonyl fluoride). PRMTs from the lysates were immunoprecipitated with anti-HA antibodies and protein-G resin (L00209, GenScript). After 16 hours the immunoprecipitates were washed three times in RIPA buffer (20 mM Tris pH 8.0, 150 mM NaCl, 5 mM EDTA and 1% Triton X-100) and once in methylation buffer (50 mM Tris pH 8.5, 20 mM KCl, 10 mM MgCl2, 1 mM β-mercaptoethanol, 100 mM sucrose). Finally, the bound PRMTs were incubated with 15 μl methylation reaction buffer containing 4 μg GST–G3BP2 and 1 μCi of S-adenosyl-L-[methyl-3H]methionine (NEN Radiochemicals; 250 μCi, 9.25 MBq), at 30°C for 1 hour. After 1 hour, the reaction was stopped by the addition of 5 μl 6×SDS sample loading buffer, and the solution was boiled and separated on a SDS-PAGE gel. Proteins on the gel were transferred onto nitrocellulose membranes using semi-dry transfer and amplified (Autofluor, National Diagnostics) 2 hours. Fluorography was then performed.

In vivo methylation assays

In vivo methylation assay for G3BP2 was performed as described previously (Bikkavilli and Malbon, 2011). Briefly, F9 cells were transiently transfected with pCMV-Myc or Myc-G3BP2 (6 μg) in 100 mm culture dishes and grown to confluency (24 hours). After 24 hours, the cells were washed once with PBS and protein translation was inhibited by treatment with 100 μg/ml cycloheximide and 40 μg/ml chloramphenicol in DMEM medium with 10% FBS for 30 minutes at 37°C. After 30 minutes, the cells were washed once with methionine-free DMEM. Cell labeling mixture consisting of methionine-free DMEM supplemented with 100 μg/ml cycloheximide, 40 μg/ml chloramphenicol and 60 μCi L-[methyl-3H]methionine was added to the cells in the absence or presence of Wnt3a (10 ng/ml) and incubated at 37°C for 3 hours. After 3 hours, the cells were lysed in a lysis buffer (1×PBS, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 1 μg/ml leupeptin, 1 μg/ml aprotonin and 1 μg/ml phenylmethylsulphonyl fluoride). Immunoprecipitations were performed on the lysates using anti-Myc antibodies and protein-A Sepharose CL-4B (17-0780-01, GE Life Sciences) for 16 hours at 4°C. After 16 hours, the immunoprecipitates were washed three times in RIPA buffer (20 mM Tris pH 8.0, 150 mM NaCl, 5 mM EDTA and 1% Triton X-100) and the beads were resuspended in 2×SDS sample loading buffer, boiled and separated on an SDS-PAGE gel. The gel was then transferred to a nitrocellulose membrane, amplified (Autofluor, National Diagnostics, 2 hours), and fluorography was performed.

Cytosolic β-catenin accumulation assay

To separate the cytosolic β-catenin from membrane-associated β-catenin, lysates were treated with concanavalin-A (Con A)–Sepharose (GE Life Sciences), as described previously (Aghib and McCrea, 1995). Briefly, confluent F9 cultures were treated with Wnt3a for 4 hours (Fig. 3B,E,F) and lysed in a lysis buffer (20 mM Tris pH 7.4, 150 mM NaCl, 5 mM EDTA, 50 mM NaF, 40 mM Na4P2O7, 50 mM K2HPO4, 10 mM Na2MoO4, 2 mM Na3VO4, 1% Triton X-100, 0.5% NP40, 1 μg/ml leupeptin, 1 μg/ml aprotonin and 1 μg/ml phenylmethylsulphonyl fluoride). The lysates were transferred into 1.5 ml Eppendorf tubes and rotated at 4°C for 15 minutes and then centrifuged at 20,000 g for 15 minutes. The supernatants were transferred into new tubes, their protein concentrations were determined and the concentration was adjusted to 2.5 mg/ml with lysis buffer. Con-A–Sepharose (60 μl) was added to each tube and rotated at 4°C for 1 hour. After a brief centrifugation, the supernatants were transferred to new tubes and 30 μl of Con-A–Sepharose was added to each tube and rotated at 4°C for another hour. Finally, after a brief centrifugation, the supernatants were transferred to new tubes and their protein concentration was determined. β-catenin accumulation was analyzed by probing the blots with anti-β-catenin antibodies and normalized by probing the same blots with anti-actin antibodies.

Knockdown protocol

Short interfering RNAs (siRNAs) targeting mouse G3BP2 (5′-GGGCGGGAGUUUGUGAGGCAAUAUU-3′) and mouse G3BP1 (5′-CCAAGAUGAGGUCUUCGGUGGCUUU-3′), and control siRNA (5′-UCUGUGAUUUGAAAGACUAGCCAAG-3′) were obtained from Invitrogen. F9 cells expressing Rfz1 were treated with 100 nM siRNA using Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's protocol. Briefly, each siRNA (100 nM) was incubated with 5 μl Lipofectamine 2000 for 20 minutes in 200 μl Optimem medium (Invitrogen), and the mixture was then added to 1 ml growth medium in a 12-well plate in which F9 cells expressing Rfz1 were cultured to 80% confluency. After siRNA treatment for 48 hours the cells were assayed for β-catenin mRNA, cytosolic β-catenin stabilization, LRP6 phosphorylation, Lef/Tcf-sensitive gene transcription or PE formation.

For gene rescue experiments, RFz1 cells were treated with 100 nM siRNAs for 4 hours. After 4 hours, the siRNA complexes were removed and fresh DMEM medium supplemented with 20% FBS were added. After an additional 18 hours, the G3BP2-depleted cells were transfected with pCMV-Myc, siRNA-resistant G3BP2, methylation-deficient mutant of siRNA-resistant G3BP2, or Myc–G3BP1 using Lipofectamine and plus reagents as suggested by the manufacturer. After 24 hours, the cells were treated with Wnt3a (25 ng/ml) for 7 hours and the lysates were assayed for Lef/Tcf-dependent luciferase activities. The luciferase activities of cells treated with control siRNA and stimulated with Wnt3a were regarded as 100% when determining the percentage reduction in luciferase activities of G3BP2-siRNA-treated cells.

Lef/Tcf transcription assays

F9 cells stably expressing Rfz1 and super 8xTOPFLASH (M50) luciferase reporter were seeded into 12-well plates. Following incubation with control or G3BP2 siRNAs for 48 hours, the cells were treated with recombinant Wnt3a or left untreated for 7 hours (R&D Systems). Cells were then directly lysed on the plates by the addition of 1× cell culture lysis reagent (Promega). Lysates were collected into chilled microfuge tubes on ice and centrifuged at 20,000 g for 5 minutes. The supernatant was transferred into a new tube and directly assayed as described below. The lysate (20 μl) was mixed with 100 μl of luciferase assay buffer (20 mM Tricine pH 7.8, 1.1 mM MgCO3, 4 mM MgSO4, 0.1 mM EDTA, 0.27 mM coenzyme A, 0.67 mM luciferin, 33 mM DTT and 0.6 mM ATP) and the luciferase activities were measured with a luminometer (Berthold Lumat LB 9507). The samples were assayed in triplicate and the luciferase activities were normalized by protein content of the samples.

Indirect immunofluorescence

F9 cells stably expressing Rfz1 were transfected with either control siRNA or G3BP2 siRNA as described above. After 24 hours, the cells were re-seeded into 24-well plates, incubated at 37°C for 4 hours, followed by stimulation with Wnt3a (day 1). On day 2, the cells were given a second siRNA treatment followed by a Wnt3a treatment. Finally, after further treatment of the cells with Wnt3a on days 3 and 4, they were fixed with 3% paraformaldehyde at room temperature for 5 minutes, followed by three washes with MSM-PIPES buffer (18 mM MgSO4, 5 mM CaCl2, 40 mM KCl, 24 mM NaCl, 5 mM PIPES, 0.5% Triton X-100, 0.5% NP40). The cells were then incubated with the TROMA-1 antibody (Developmental Studies Hybridoma Bank, University of Iowa) at 37°C for 30 minutes. After three washes with MSM-PIPES buffer, the cells were incubated with an anti-mouse antibody coupled to Alexa Fluor 488 (Invitrogen) at 37°C for 30 minutes. Finally, the cells were washed in blotting buffer (560 mM NaCl, 10 mM KH2PO4, 0.1% Triton X-100, 0.02% SDS) and images were captured using a Zeiss LSM510 inverted fluorescence microscope.

Northwestern analysis

Digioxigenin (DIG)-labeled 3′-UTR probes of β-catenin were synthesized in vitro using T7 RNA polymerase (Roche Applied Sciences), as per the manufacturer's recommendations, in the presence of rNTPs and DIG-UTP and pcDNA3.1vectors harboring β-catenin UTR as a template. For northwestern analysis, immunoprecipitated proteins were resolved on SDS-PAGE gels and electrophoretically transferred to nitrocellulose membranes (blots). The blots were blocked in Tris-buffered saline (TBST; 50 mM Tris pH 7.4, 150 mM NaCl and 0.1% Tween 20) containing 5% non-fat milk at 4°C overnight with gentle rocking. DIG-labeled probes (1 μg/ml in TBST buffer with milk) were added to the blots and incubated at room temperature with gentle rocking for 2 hours. After 2 hours, the blots were washed three times in TBST at 5 minute intervals. The binding of RNA probes to G3BP1 was then revealed by probing the blots with anti-DIG alkaline phosphatase fragments diluted (1:1000) in TBST with 5% milk (11093274910, Roche Applied Science), followed by colorimetric detection of alkaline phosphatase using Nitro Blue tetrazolium chloride (NBT) and 5-bromo-4-chloro-3′-indolyphosphate p-toluidine salt (BCIP) substrates according to the manufacturer's recommendation (Roche DIG-RNA detection kit; Roche).

RNA isolation and RT-PCR

Total RNA from F9 cells treated with either control siRNA or G3BP2-specific siRNA were isolated using RNA STAT 60 reagent (Tel-test Inc., Friendswood, TX) according to the manufacturer's instructions. After determining the RNA concentrations using a spectrophotometer, first strand cDNA synthesis was performed using 250 ng total RNA and Superscript II reverse transcriptase (Invitrogen) and random primer hexamers. Real-time quantitative PCR amplification was performed using the DNA engine Opticon continuous fluorescence detection system (MJ Research Inc., Boston, MA). For a 20 μl PCR, 8 μl cDNA template (previously diluted to 1:15 with water) was mixed with 6.25 pmol of forward and reverse primers and 2× SYBR green PCR master mix (Qiagen). The Light Cycler was programmed such that it included an initial activation step of 95°C for 15 minutes followed by 40 cycles of denaturation at 95°C for 30 seconds, annealing at 60°C for 30 seconds and extension at 72°C for 1 minute. Each cDNA sample was analyzed in triplicate and the absolute amounts of β-catenin template in the immunocomplexes were determined using an external standard. Briefly, a standard curve was generated using cycle threshold (Ct) values obtained from real-time PCR using Dvl2-specific primers (supplementary material Table S1) and pGFP2-N2-mDvl2 plasmid (1, 0.1, 0.01 and 0.001 ng/reaction). The Ct values of real-time PCR for cDNA from each RNA sample was then substituted in the equation generated from the corresponding standard curve to calculate the amount of the amplicon. The results are presented as fold increases over control values.

Data analysis

Data were compiled from at least three independent, replicate experiments, each performed on separate cultures and on separate occasions. The results are presented as fold increases over control (untreated) values. Student's t-tests were used for assessing variance between experimental groups. Statistical significance was set at P<0.05. The indirect immunofluorescence and phase-contrast images are of representative fields.

Acknowledgements

We thank Antonius Koller (Technical Director, Proteomics Center, SUNY, Stony Brook) for his help with sample preparation and generation and analysis of MS–MS spectra. We thank Xi He (Children's Hospital, Boston) for the generous gift of LRP6ΔN expression vector, Randall T. Moon (University of Washington, Seatle) for the generous gift of RFz1 and M50 expression vectors and Jim Woodget (Mount Sinai Hospital, Toronto, Canada) for the generous gift of pcDNA-HA-hGSK3β. We would also like to thank Hsien-Yu Wang and Nedialka Markova for critical reading of the manuscript. We also thank members of the Malbon and Wang laboratories for their critical comments and helpful suggestions.

Footnotes

  • * Present address: Division of Pulmonary Sciences and Critical Care, School of Medicine, Anschutz Medical Campus, University of Colorado Health Sciences Center, Denver, Colorado 80045, USA

  • Funding

    This work was supported by US Public Health Service Grant from the National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health [grant number DK30111 to C.C.M.). Deposited in PMC for immediate release.

  • Supplementary material available online at http://jcs.biologists.org/lookup/suppl/doi:10.1242/jcs.100933/-/DC1

  • Accepted January 16, 2012.

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial Share Alike License (http://creativecommons.org/licenses/by-nc-sa/3.0), which permits unrestricted non-commercial use, distribution and reproduction in any medium provided that the original work is properly cited and all further distributions of the work or adaptation are subject to the same Creative Commons License terms.

References

View Abstract