Dpr Acts as a Molecular Switch, Inhibiting Wnt Signaling when Unphosphorylated, but Promoting Wnt Signaling when Phosphorylated by Casein Kinase Iδ/ε

The Wnt pathway is a key regulator of development and tumorigenesis. Dpr (Dact/Frodo) influences Wnt signaling in part through the interaction of its PDZ-B domain with Dsh's PDZ domain. Studies have shown that XDpr1a and its close relative, Frodo, are involved in multiple steps of the Wnt pathway in either inhibitory or activating roles. We found that XDpr1a is phosphorylated by casein kinase Iδ/ε (CKIδ/ε), an activator of Wnt signaling, in the presence of XDsh. Abrogating XDpr1a's ability to bind XDsh through mutation of XDpr1a's PDZ-B domain blocks CK1δ/ε's phosphorylation of XDpr1a. Conversely, XDsh possessing a mutation in its PDZ domain that is unable to bind XDpr1a does not promote XDpr1a phosphorylation. Phosphorylation of XDpr1a and XDsh by CKIδ/ε decreases their interaction. Moreover, the phosphorylation of XDpr1a by CKIδ/ε not only abrogates XDpr1a's promotion of β-catenin degradation but blocks β-catenin degradation. Our data suggest that XDpr1a phosphorylation by CKIδ/ε is dependent on the interaction of XDpr1a's PDZ-B domain with XDsh's PDZ domain, and that the phosphorylation state of XDpr1a determines whether it inhibits or activates Wnt signaling.

Dsh has three highly conserved domains: an N-terminal DIX domain (which is also present in axin), a central PDZ domain (first identified in PSD, Discs large, and ZO1), and a C-terminal DEP domain (first identified in Dsh, Egl-10, and Pleckstrin) [13,14].
Dsh's DIX and PDZ domains are likely to function in canonical Wnt signaling, while its PDZ and DEP domains are likely to function in noncanonical Wnt signaling [15][16][17]. Dsh is thought to act as a molecular scaffold in signal transduction processes, since at least eighteen binding partners of Dsh have been identified [14]. PDZ domains are protein-protein interaction domains and therefore are important for the scaffolding function of Dsh. Xenopus Dapper1a (XDpr1a) and the highly related protein Functional regulator of disheveled in ontogenesis (Frodo), are novel mediators of Wnt signaling which were isolated in independent yeast two-hybrid screens for proteins that interact with Xenopus Dsh (XDsh) [18,19]. XDpr1a and Frodo have two defined motifs: an N-terminal leucine zipper (LZ) domain, and a C-terminal PDZ-binding (PDZ-B) domain [18,19]. The LZ domain is unnecessary for XDpr1a's interaction with XDsh, since deletion of the LZ domain does not influence XDpr1a's ability to associate with XDsh, although it does affect the ability of XDpr1a to inhibit Wnt signaling [18]. However, deletion or mutation of the PDZ-B domain impedes XDpr1a or Frodo from associating with XDsh [18,19]. Recently, a domain was identified in the central region of human Dpr1 that also mediates Dvl binding, but the role of this region in XDpr1a is not known [20].
XDpr1a and Frodo are the result of a recent gene duplication in Xenopus and intriguingly, studies have shown that XDpr1a and Frodo are involved in multiple steps of the Wnt pathway in either inhibitory or activating roles [21,22]. In many instances, XDpr1a and/or Frodo negatively regulate Wnt signaling. XDpr1a forms a complex with XDsh, axin, GSK3b, CKIe, and b-catenin; exogenous XDpr1a increases axin and GSK3b, and reduces CKIe, in this complex, resulting in reduced b-catenin abundance and reduced activation of Wnt responsive genes [18]. Furthermore, Dpr inhibits Wnt signaling both from the cytoplasm and the nucleus. Mammalian Dpr1/Dact inhibits expression of Wntresponsive reporters through its promotion of Dvl degradation in a lysosome-dependent pathway, and by inhibiting the binding of LEF1 with b-catenin, but promoting the binding of LEF1 with a corepressor, histone deacetylase 1 (HDAC1) [20,23].
In contrast, XDpr1a and/or Frodo also activate Wnt signaling. Reducing Frodo abundance with antisense morpholino oligonucleotides (MOs) inhibits Xwnt8-and XDsh-induced body axes and reduces Wnt-dependent reporter activity in Xenopus [18,19]. In addition, Dpr1 enhances Wnt8's ventralization and posteriorization activities in zebrafish [21,22]. Using a similar experimental system, however, Frodo and XDpr1a were found to either inhibit or activate Wnt signaling, dependent on the point at which the pathway was activated [22]. Frodo and XDpr1a MOs reduce b-cateninindependent transcription induced by TCF-VP16, suggesting that Frodo and XDpr1a are required for Tcf-mediated transcription, and Frodo and XDpr1a MOs increase Xwnt8-induced siamois reporter activity, suggesting that Frodo and XDpr1a inhibit Wnt signaling. The molecular mechanisms underlying these apparent differences in function have not yet been explored.
Phosphorylation plays a key role in the regulation of Wnt signaling, and CKI family members appear to have several targets in the Wnt pathway. CKIa phosphorylates b-catenin to negatively regulate the pathway [24]. CKId and CKIe are two highly related CKI isoforms likely to have similar functions, since their kinase domains are 98% identical and they have 53% identical C-terminal tails, unique to CKId and CKIe, that inhibit their function when autophosphorylated. CKId and CKIe activate Wnt signaling through their putative targets of Dsh and/or Lef/Tcf, while CKIc activates Wnt signaling through LRP5/6 phosphorylation [24]. We have shown that CKId and CKIe both interact directly with Dvl-1, suggesting that they influence Wnt signaling using a common mechanism [8]. CKId directly phosphorylates numerous components of the b-catenin degradation complex in vitro, i.e., Dsh, APC, axin, and b-catenin, and CKIe phosphorylates axin and b-catenin in vivo [8]. In addition, we found that CKIe dissociates protein phosphatase 2A (PP2A) A and C subunits from the b-catenin degradation complex both in vitro and in vivo, and that CKId and CKIe both act upstream of the B56a regulatory subunit of PP2A in Xenopus body axis formation, which is a Wnt-dependent process [8].
The circumstances that determine whether Dpr inhibits or activates Wnt signaling are not known. An appealing hypothesis is that posttranslational modifications of Dpr may regulate its role in Wnt signaling. Since phosphorylation is a common mechanism to alter a protein's activity, we analyzed the phosphorylation of Dpr and found that XDpr1a is phosphorylated by CKId/e in the presence of XDsh. CKId's phosphorylation of XDpr1a reduces its interaction with XDsh and transforms XDpr1a from a promoter to an inhibitor of b-catenin degradation. Our data suggest that XDpr1a inhibits Wnt signaling when unphosphorylated and bound to XDsh, but activates Wnt signaling when phosphorylated by CKId/e and associated with other Wnt players.

XDsh promotes a CKId-induced phosphorylation of XDpr1a
To study the phosphorylation of XDpr1a, we first examined the ability of two Wnt pathway kinases to phosphorylate XDpr1a in vitro, GSK3b and CKId/e. We found that CKId, but not GSK3b, was able to phosphorylate XDpr1a, as exemplified by a gel shift (Fig. 1A, compare lane 2 with lane 1, and data not shown). XDpr1a was initially isolated because of its interaction with XDsh, and XDsh is phosphorylated by CKId/e in vitro [8,18,19], so we determined if XDsh influences the phosphorylation state of XDpr1a. We found that XDsh induced an upward mobility shift of XDpr1a similar to the shift seen in the presence of CKId (Fig. 1A, compare lane 3 with lane 1). Note that XDpr1a migrates slower than its calculated molecular weight of 91 kD, likely due to as yet uncharacterized posttranslational modifications. Intriguingly, the presence of both CKId and XDsh resulted in a hypershift of XDpr1a (Fig. 1A, compare lane 4 with lanes 1-3), suggesting that XDsh promotes the phosphorylation of XDpr1a by CKId. This also suggests that the modest gel shift seen in the presence of CKId alone (Fig. 1A, lane 2) was due to limiting amounts of endogenous Dsh in reticulocyte lysates, whereas the modest gel shift in the presence of XDsh alone (Fig. 1A, lane 3) was due to limiting amounts of endogenous CKId in reticulocyte lysates, and that the supershift in the presence of XDsh and CKId (Fig. 1A, lane 4) occurs when neither XDsh nor CKId are limiting. As reported previously [8,18,19], we found that CKId also phosphorylates XDsh (  is likely due to limiting amounts of endogenous XDsh and CKId in the reticulocyte lysates used in the in vitro transcription and translation, respectively. B. XDsh-mediated CKId phosphorylation of XDpr1a has little effect on XDpr1a abundance. Phosphorylation reactions were carried out as in A., but with the inclusion of luciferase as a loading control. XDpr1a and luciferase bands were quantitated, and the XDpr1a signal was normalized to that of luciferase. The luciferase-normalized signals were then normalized to that of XDpr1a alone. Error bars signify standard deviation (n = 3 trials). doi:10.1371/journal.pone.0005522.g001 Because mobility shifts of phosphorylated proteins can cause the broadening of SDS-PAGE bands and make it difficult to estimate protein abundance determinations by eye, we quantitated the XDpr1a signal from CKId phosphorylation reactions to determine if CKId-mediated phosphorylation of XDpr1a affects XDpr1a abundance. We found that the abundance of XDpr1a is relatively constant in the presence of XDsh and/or CKId (Fig. 1B). This suggests that while CKId phosphorylates XDpr1a, it has little effect on XDpr1a abundance. To verify that the XDsh/CKIdmediated XDpr1a gel shift was due to phosphorylation, we carried out CKId phosphorylation reactions in the presence of [c-33 P]ATP. The presence of XDsh and CKId induced incorporation of [c-33 P]ATP into XDpr1a (Fig. 2A, compare lane 4 to 3), which indicates XDpr1a is phosphorylated under these conditions, and induced a gel-shift that comigrates with the [ 35 S]methioninelabeled XDpr1a gel-shift ( Fig. 2A, compare lane 4 to 3 and lane 2 to 1), showing that the gel-shifted XDpr1a is phosphorylated. The gel-shift of XDpr1a represents a 4.060.7% increase in molecular weight, or approximately 6 kD. The data confirm that the XDsh/ CKId-mediated gel-shift of XDpr1a is due to XDpr1a phosphorylation. In summary, XDsh promotes the phosphorylation of XDpr1a by CKId, perhaps by bridging XDpr1a to CKId.

CKIe phosphorylates XDpr1a in vivo
To examine the phosphorylation of XDpr1a by CKId/e under more physiological conditions, we monitored the effects of modulating CKId/e activity on XDpr1a phosphorylation in vivo. HEK293 cells were transfected with Myc:XDpr1a alone or with HA:XDsh and CKIe, and metabolically labeled with [ 32 P]orthophosphoric acid. CKIe induced an XDpr1a gel-shift and promoted increased incorporation of [ 32 P]orthophosphoric acid into XDpr1a (Fig. 2B). The gel-shift of XDpr1a represents a 9.160.8% increase in molecular weight, or approximately 13 kD, somewhat higher than the in vitro shift of 6 kD, suggesting that XDpr1a phosphorylation is more robust in vivo. This result extends our in vitro data and shows that XDpr1a is phosphorylated by CKIe in vivo.

PDZ-B mutants of XDpr1a are not phosphorylated by CKId
Mutational analyses have shown that the PDZ-B domain of XDpr1a/Frodo interacts with the PDZ domain of XDsh [18,19]. If XDsh is required to promote the phosphorylation of XDpr1a by CKId, then mutants of XDpr1a with reduced XDsh binding may not be phosphorylated by CKId. We tested this hypothesis by determining if mutation or deletion of the XDpr1a PDZ-B domain abrogated the ability of XDsh to promote CKId-mediated XDpr1a phosphorylation. We used in vitro transcription/translation to synthesize wild-type and mutant XDpr1a proteins, as well as XDsh, followed by a phosphorylation reaction in the presence of purified CKId. We tested three XDpr1a mutants, one that binds XDsh's PDZ domain (XDpr1aDLZ), and two that do not (XDpr1aDMTTV and XDpr1aMNTV) [18]. The XDpr1aDLZ protein, lacking the N-terminal 129 amino acids including the leucine zipper motif, served as a control and behaved similarly to wild-type XDpr1a, undergoing a gel shift indicative of hyperphosphorylation in the presence of CKId (Fig. 3A, compare lane 2 with lane 1, and lane 4 with lane 3). XDpr1aDMTTV lacks the PDZ-B domain and did not undergo a mobility shift in the presence of CKId, suggesting that it is not phosphorylated by CKId (Fig. 3A, compare lane 6 with lane 5). Since deletion of the XDpr1a PDZ-B domain inhibited XDsh's promotion of XDpr1a phosphorylation by CKId, we examined whether a point mutation within the PDZ-B motif would affect XDsh-dependent CKId phosphorylation of XDpr1a. We found that XDpr1aMNTV, containing a T822N point mutation in its PDZ-B domain, behaved similarly to XDpr1aDMTTV, and did not exhibit a mobility shift (Fig. 3A, compare lane 8 with lane 7). These data suggest that an intact PDZ-B domain in XDpr1a is required for XDsh-dependent phosphorylation of XDpr1a by CKId.

Mutation of the XDsh PDZ-B binding domain blocks XDpr1a phosphorylation
To further show the importance of XDsh in the phosphorylation of XDprIa by CKId, we examined whether a mutation in XDsh that reduces its interaction with XDpr1a's PDZ-B domain still promotes the phosphorylation of XDpr1a by CKId. The region of XDsh's PDZ domain that binds to XDpr1a's PDZ-B was identified by X-ray crystallography [18]. An N317T point mutation in XDsh's PDZ-B binding domain (*aXDsh) diminishes its interaction with XDpr1a, whereas a triple mutation in a PDZ domain loop upstream of XDsh's PDZ-B binding domain (272QSNE275 to 272AANA275, ***b-bXDsh) does not [18]. We examined the ability of *aXDsh and ***b-bXDsh to promote the phosphorylation of XDpr1a by CKId. XDpr1a exhibited a mobility shift indicative of hyperphosphorylation in the presence of XDsh and ***b-bXDsh (Fig. 3B, compare lanes 2 and 3 to lane 1), whereas XDpr1a did not exhibit a mobility shift in the presence of *aXDsh (Fig. 3B, lane 4 compared to lane 1). The inability of an XDsh protein containing a point mutation in its PDZ-B binding domain to promote the phosphorylation of XDpr1a by CKId suggests that XDsh must retain its ability to bind XDpr1a in order to promote XDpr1a phosphorylation.

CKId reduces the interaction between XDpr1a and XDsh
CKId/e destabilizes the b-catenin degradation complex [8], and Dpr and Dsh are both components of this complex, so CKId may disrupt the interaction between Dpr and Dsh as well. We tested this hypothesis using an in vitro coimmunoprecipitation assay. We immunoprecipitated Myc-tagged XDpr1a from a reaction containing HA-tagged XDsh in the absence or presence of CKId. The presence of CKId dramatically decreased the immunoprecipitation of XDsh with XDpr1a (Fig. 4A, compare  lanes 2 and 1), resulting in a concomitant increase of XDsh in the immunosupernatant (Fig. 4A, compare lanes 4 and 3). Because of the CKId-induced mobility shift of XDsh, determination of the extent of the reduction of XDsh coimmunoprecipitation in the presence of CKId is difficult to make by eye, therefore we quantitated the coimmunoprecipitatipon of XDsh. We found that CKId reduced the coimmunoprecipitation of XDsh by XDpr1a by approximately one-half (Fig. 4B). This suggests that XDpr1a and XDsh interact transiently, dissociating soon after XDpr1a and/or XDsh is phosphorylated by CKId. Therefore, CKId reduces the interaction between XDsh and XDpr1a and is likely to play a significant role in Wnt signaling activation.  XDpr1a promotes b-catenin degradation in the absence of CKId, but inhibits b-catenin degradation in its presence XDpr1a has been shown to reduce b-catenin abundance both in Xenopus embryos and in mammalian tissue culture [18]. To investigate the functional consequences of XDsh-mediated CKId phosphorylation of XDpr1a, we utilized an in vitro b-catenin degradation assay using Xenopus egg extracts [8,25,26]. When unmodified XDpr1a was added to Xenopus egg extracts, the rate of b-catenin degradation increased, reducing the half-life of b-catenin approximately two-fold, from 1.8 to 0.9 hours, indicating an inhibition of canonical Wnt signaling (Fig. 5). However, when XDpr1a was preincubated with CKId, b-catenin degradation was blocked, indicative of Wnt pathway activation (Fig. 5). These data suggest that XDpr1a acts as a molecular switch, inhibiting Wnt signaling when unphosphorylated, but promoting Wnt signaling when phosphorylated by CKId.

Discussion
XDpr1a is a member of a conserved family of novel Dsh binding proteins. XDpr1a's PDZ-B domain interacts with XDsh's PDZ domain [18]. Here, we show that XDsh mediates the phosphorylation of XDpr1a by CKId. In addition, we found that an intact PDZ-B domain in XDpr1a, as well as an intact PDZ-B binding domain in XDsh, is required for XDsh-dependent phosphorylation of XDpr1a by CKId. This suggests that XDpr1a and CKId/e do not interact directly and/or robustly with one another, and that XDsh is required to link XDpr1a and CKId/e. Epigenetic regulation of Dpr expression is associated with tumorigenesis. Human Dpr1/DACT1 is often downregulated by allelic loss or promoter methylation in hepatocellular carcinomas [27,28]. There is reduced DACT3 expression in human colon tumors due to histone modifications, resulting in increased Wnt signaling activity; this suppression of DACT3 expression is relieved in colon cancer cell lines by treatment with histone methylation and deacetylase inhibitors [28,29]. In addition, Dpr is upregulated by the treatment of breast cancer patients with DNA methylation and histone deacetylase inhibitors [28,29]. Each of these results suggests that the dominant function of Dpr with regard to tumor formation is inhibition of Wnt signaling, and that the loss of this function is associated with tumorigenesis. Our data show that XDpr1a is phosphorylated by CKId/e, and that this phosphorylation reduces XDpr1a's interaction with XDsh. Frizzled-1 overexpression, which activates Wnt signaling, results in disparate localizations of XDsh and XDpr1a, causing the membrane localization of XDsh but not XDpr1a [18]. CKId's ability to reduce XDsh/XDpr1a binding may play a part in Wnt's localization of XDsh and XDpr1a to disparate locations. Intriguingly, we found that XDpr1a promotes b-catenin degradation when unphosphorylated but blocks b-catenin degradation when phosphorylated by CKId. Overall, our data suggest that XDpr1as acts as a molecular switch in Wnt signaling. In the absence of CKId/e activity, XDpr1a is bound to XDsh and inhibits Wnt signaling, whereas XDpr1a promotes Wnt signaling when phosphorylated by CKId/e. Cong et al proposed that the N-terminal region of Dsh, containing the DIX domain, is required for its canonical Wnt signaling activity, and that Dsh's C-terminal region, containing the PDZ and DEP domains, structurally blocks this function [30]. In addition, they suggest that this inhibition is relieved by the phosphorylation of Dsh by CKIe. Our data suggest that Dpr is a missing link in the inhibition of Dsh function. We propose that Dpr inhibits Dsh's function in canonical Wnt signaling by binding to Dsh's PDZ domain, and preventing Dsh's DIX domain from activating Wnt signaling. Further, the phosphorylation of XDpr1a and XDsh by CKId/e relieves this suppression by phosphorylating both XDpr1a and XDsh and reducing their interaction.
We propose a model for the regulation of Wnt signaling based on our data, as well as that from several other labs, in which the phosphorylation state of Dpr acts as a molecular switch to determine whether Dpr inhibits or activates Wnt signaling. In the absence of Wnt, unphosphorylated XDpr1a binds to XDsh in the b-catenin degradation complex and inhibits Wnt signaling. When Wnt binds to LRP5/6 and fz coreceptors, the b-catenin degradation complex moves to the membrane through the interactions of axin with LRP5/6 and Dsh with fz, [2][3][4][5][6]. This results in CKId/e-mediated phosphorylation of XDsh, XDpr1a, and other substrates [24]. The b-catenin degradation complex then partially dissociates, with b-catenin, APC, and XDsh remaining at the membrane, while PP2A A and C subunits [8] and XDpr1a are released from the complex. Phosphorylated XDpr1a then associates with distinct Wnt pathway binding partners and activates Wnt signaling. Future experimentation will determine if the phosphorylation state of XDpr1a affects its association with other known Wnt pathway protein partners. Our data suggest that LEF1, HDAC, and b-catenin may differentially associate with unphosphorylated XDpr1a, while Tcf3 may differentially associate with CKId/e-phosphorylated XDpr1a.

In vitro phosphorylation assay
In vitro transcription and translation was performed with [ 35 S]methionine (Amersham Biosciences Corp., Piscataway, NJ) using TNT SP6 Quick Coupled Transcription/Translation System (Promega, Madison, WI) according to the manufacturer's instructions. Unlabeled proteins were prepared with TNT SP6 Quick Coupled Transcription/Translation System using cold methionine. XDpr1a and XDsh TNT reactions were mixed at a ratio of 1:1 in a reaction also containing 5 mM MgCl 2 and 0.5 mM ATP. Rat CKId lacking its autoinhibitory C-terminal domain (New England Biolabs, Ipswich, MA) or GSK3b (New England Biolabs, Ipswich, MA), was added to a final concentration of 1.1 mM, and the phosphorylation reactions were carried out for 1.5 hours at 30uC. Where specified, [ 35 S]methionine-labeled luciferase was added to the phosphorylation reactions, and bands were quantitated after denaturing SDS-PAGE using a Molecular Dynamics PhosphorImager and ImageQuant software. Where specified, [c-33 P]ATP was added to the CKId reactions followed by a 1.5 hour incubation at 30uC.

In vitro coimmunoprecipitations
[c-33 P]ATP-labeled Myc:XDpr1a was immunoprecipitated in the presence of anti-Myc beads for 2 hours at room temperature. Immunoprecipitates were washed three times with 50 mM Tris-HCl pH 7.4, 137 mM NaCl, followed by SDS-PAGE and visualization by autoradiography. XDpr1a's molecular weight was determined in the absence or presence of CKId and XDsh from three experimental trials using GelScape (www.gelscape. ualberta.ca:8080/htm/index.html).
[ 35 S]methionine-labeled Myc:XDpr1a was immunoprecipitated in the presence of anti-Myc antibodies and protein A agarose at room temperature for 2 hours. Immunoprecipitates were washed three times with 50 mM Tris-HCl pH 7.4, 137 mM NaCl, followed by denaturing SDS-PAGE, visualization by autoradiography, and quantitation using a Molecular Dynamics PhosphorImager and ImageQuant software. The charted coimmunoprecipitation data is the percentage of XDsh in the immunoprecipitate (versus total input), divided by the percentage of XDpr1a in the immunoprecipitate (versus total input), normalized to 1.0 in the absence of CKId (n = 4 trials).

b-catenin degradation assays
Xenopus egg extracts were prepared, RNA was synthesized and translated, and degradation assays were carried out as described previously with minor modifications [8,25,26]. Myc:XDpr1a or bgalactosidase was preincubated with or without CKId following its translation in egg extracts. Anti-Myc immunoprecipitates were washed prior to being added to fresh egg extract for the degradation assay, which contained 40 mM IC261 to inhibit any potential carryover CKId activity. [ 35 S]b-catenin was synthesized using TNT T7 coupled wheat germ extract system (Promega, Madison, WI). Degradation assays were performed six times, with aliquots removed at 0, 0.5, 1.0, and 2.0 hours. Aliquots were resolved using SDS-PAGE, imaged using a Molecular Dynamics PhosphorImager, and quantitated using ImageQuant software.