Mutations in the Human naked cuticle Homolog NKD1 Found in Colorectal Cancer Alter Wnt/Dvl/β-Catenin Signaling

Background Mutation of Wnt signal antagonists Apc or Axin activates β-catenin signaling in many cancers including the majority of human colorectal adenocarcinomas. The phenotype of apc or axin mutation in the fruit fly Drosophila melanogaster is strikingly similar to that caused by mutation in the segment-polarity gene, naked cuticle (nkd). Nkd inhibits Wnt signaling by binding to the Dishevelled (Dsh/Dvl) family of scaffold proteins that link Wnt receptor activation to β-catenin accumulation and TCF-dependent transcription, but human NKD genes have yet to be directly implicated in cancer. Methodology/Principal Findings We identify for the first time mutations in NKD1 - one of two human nkd homologs - in a subset of DNA mismatch repair-deficient colorectal tumors that are not known to harbor mutations in other Wnt-pathway genes. The mutant Nkd1 proteins are defective at inhibiting Wnt signaling; in addition, the mutant Nkd1 proteins stabilize β-catenin and promote cell proliferation, in part due to a reduced ability of each mutant Nkd1 protein to bind and destabilize Dvl proteins. Conclusions/Significance Our data raise the hypothesis that specific NKD1 mutations promote Wnt-dependent tumorigenesis in a subset of DNA mismatch-repair-deficient colorectal adenocarcinomas and possibly other Wnt-signal driven human cancers.


Introduction
Activation of ''canonical'' Wnt/b-catenin signaling in nearly all human colorectal adenocarcinomas (CRC) makes the Wnt pathway a promising yet untapped therapeutic target [1]. The prevailing paradigm for canonical Wnt signaling was deduced in part through elegant developmental studies of the fruit fly Drosophila melanogaster and the amphibian Xenopus laevis: Absent the Wnt signal, a ''destruction complex'' composed of the proteins Apc, Axin, GSK3b, and CK1 phosphorylates b-catenin, leading to b-catenin ubiquitination and proteasomal degradation [2]. Binding of Wnt ligands to Frizzled/Lrp coreceptors activates the scaffold protein Dishevelled (Dsh; Dvl1, Dvl2, Dvl3 in mammals), leading to sequestration and degradation of Axin, which allows bcatenin to accumulate, enter the nucleus, and bind TCF transcription factors to regulate target genes [2].
A majority (60-85%) of human CRC exhibit activated canonical Wnt signaling due to truncating mutations in APC that stabilize b-catenin [3]. Alternatively, mutations in b-catenin (CTNNB1) that block phosphorylation and degradation are found in some CRC that lack APC mutation [4]. CRCs display at least two types of genomic instability: chromosomal instability (CIN) associated with mutant Apc and p53 and giving rise to aneuploidy, and microsatellite-instability (MSI) caused by defective DNA mismatch repair (MMR) and resulting in mutations in simple sequence repeats (SSR) throughout the genome [5,6]. Mutation of SSRs in the coding or splice junction regions of key regulatory genes can create point mutant or truncated proteins that promote cancer progression; indeed, MMR deficiency and MSI are characteristic of tumors in patients with hereditary nonpolyposis colorectal cancer syndrome {HNPCC; a.k.a. Lynch Syndrome (OMIM 120435)} and of 13-17% of sporadic CRC [7]. APC mutations are prevalent in CIN-CRC [3], but the Wnt pathway gene mutation spectrum in MSI-CRC is less well characterized, with mutations in the Axin homolog AXIN2 and the TCF-family transcription factor TCF7L2 identified in ,25% and ,35% of MSI-CRC, respectively [8,9]. APC mutation is less frequent in MSI-CRC than in CIN-CRC [10,11], while activating mutations in CTNNB1, though widespread throughout the spectrum of human cancer, are rare in MSI-CRC [11]. These data suggest that additional mechanisms activate Wnt/b-catenin signaling in MSI-CRC.
The Naked cuticle (Nkd) protein family attenuates canonical Wnt signaling by binding and possibly destabilizing Dsh/Dvl proteins [12][13][14][15][16][17]. Drosophila nkd mutants develop lethal segmentation defects very similar to those seen in apc or axin mutants [12,18,19] (Fig. 1A). We therefore hypothesized that alteration of nkd gene activity in mammals might activate Wnt signaling and cause cancer. Here we identify novel mutations in the human NKD1 gene in MSI-CRC that alter Wnt signaling and reduce Nkd/Dsh interactions. Our data suggest that specific NKD1 mutations alter Wnt/b-catenin signaling in a minority of MSI-CRC as well as possibly in other bcatenin signal-dependent tumors in which mutations in the known Wnt regulators are infrequent. Wild type has alternating denticle bands (arrow) and naked cuticle (arrowhead), with each mutant lacking denticle bands. (B) NKD1 locus has 10 exons. Nkd1 schematic (orange) includes N-terminal myristoylation, EFX, 30aa (blue), and carboxy-terminal His-rich motifs. Exon 10 sequences around poly-(C) tracts (red) above native (black) and mutant (blue) residues are shown. (C) NKD1 electropherograms showing wild-type (WT) poly-(C) 7 , cell line TC7 with C-deletion (C6), cell line RKO with Cinsertion (C8), and cell line HCT15 with G.A mutation (arrow) 39 of poly-(C) 7 . (D, E) a-Nkd1 blots of whole cell extracts (D) and Triton X-100 soluble and insoluble fractions (E) from cell lines with indicated NKD1 mutation. Arrows designate Nkd1 proteins. b-actin is loading control in D. CCD841 has full length Nkd1, with a minor degradation product at ,35 kDa also seen with transfected NKD1 (e.g. Fig. 1E, 5C), whereas SW480 with more abundant but wild-type Nkd1 has several degradation products. doi:10.1371/journal.pone.0007982.g001

NKD1 mutations in colorectal adenocarcinoma
We identified three different NKD1 exon 10 coding region mutations in 5/11 CRC cell lines and 2/40 sporadic CRC tumors with MSI (Table 1), but no NKD1 coding region or splice junction mutations in 5/5 CRC cell lines and 50/50 tumors without MSI. Two mutations, either a deoxycytidine (C) deletion or insertion due to polymerase slippage within an exon-10 poly-(C) 7 tract, result in the synthesis of truncated proteins of 345 or 298 amino acids (aa) (Fig. 1B,C). A (C) 7 -adjacent missense mutation (G.A) converts Arg-288, conserved in Nkd2, to His (Fig. 1B,C). NKD1 mutations were not detected in 32/40 MSI-CRC tumors that harbor mutations in other Wnt-pathway genes including APC, CTNNB1, AXIN2, and TCF7L2, but were found in 2 of the remaining 8/40 tumors without lesions in these known Wntpathway genes (p = 0.036, one-tailed Fisher's exact test) (Table 1; see Materials and Methods). These data indicate a mutual exclusivity among mutations in NKD1 and other Wnt-pathway genes in our cohort of MSI-CRC tumor samples, suggesting that the NKD1 mutations are of pathological significance.
Both the colonic epithelial cell line CCD841 and the CRC cell line SW480, the latter with a C.T mutation at APC codon #1338 [20], encode a wild-type Nkd1 (470 aa) of 53.2 kDa that migrates at ,50 kDa on western blot (Fig. 1D, E). Cell lines Co115 and TC7, each with the (C) 6 mutation, have a ,39 kDa band that corresponds to the 38.1 kDa Nkd1 C6 , while cell line RKO, with the (C) 8 mutation, has a ,33 kDa band that corresponds to the 33.4 kDa Nkd1 C8 ; all three cell lines with truncated Nkd1 also have less abundant full-length Nkd1, suggesting that the wild-type NKD1 locus does not undergo loss-of-heterozygosity (Fig. 1D). By western blot with Nkd1 antibodies we are unable to distinguish wild-type from mutant Nkd1 in cell line HCT15, which harbors NKD1 R288H (Fig. 1D).
Mutant Nkd1 proteins are defective at inhibiting Wnt/Dvl signaling Mouse Nkd1 can inhibit axis duplication induced by ectopic Wnt signaling in Xenopus embryos [16]. As shown in Fig. 2A,  [45]. Presence (+) or absence (2) of APC and CTNNB1 lesions in each cell line is as described [48]. doi:10.1371/journal.pone.0007982.t001 .90% of Xenopus embryos injected with XWnt8 mRNA developed partial-to-complete axis duplication; consistent with Nkd1's activity as a Wnt antagonist, wild-type human NKD1 mRNA coinjection reduced axis duplication frequency to 42%, with only 3% complete duplications. In contrast, co-injection of each mutant human NKD1 resulted in axis duplication frequencies similar to that observed with XWnt8 injection alone ( Fig. 2A). As in cell lines, misexpressed Nkd1 C6 and Nkd1 C8 were more abundant than wildtype Nkd1 or Nkd1 R288H by western blot (not shown). In agreement with the Xenopus results, wild-type Nkd1, but none of the three mutants, suppressed Dvl-induced activation of the TCFreporter TOPflash in HEK-293 cells (Fig. 2B). Expression of mutant Nkd1 alone slightly increased basal TOPflash activity and had no effect on endogenous Xenopus axis formation (not shown), indicating that the effect of Nkd1, like that of nkd in the fly, depends on Wnt signaling [23].

NKD1 mutations stabilize b-catenin and promote cell proliferation
Consistent with the NKD1 mutations activating Wnt signaling in CRC, cytoplasmic and nuclear levels of b-catenin are higher in Co115 cells than in CCD841 cells (Fig. 3A). Accordingly, nuclear b-catenin was prominent in Co115 cells but not in CCD841 cells (Fig. 3B, C). HEK-293T cells transfected with Nkd1 C6 , Nkd1 C8 , or Nkd1 R288H had higher levels of cytosolic b-catenin than cells transfected with wild-type Nkd1 or a control (Fig. 3D), indicating that each mutant Nkd1 protein can stabilize b-catenin.
Since canonical Wnt signaling promotes cell proliferation [2], we assayed the accumulation of cells uniformly expressing comparable levels of either wild-type or mutant Nkd1. Retroviral expression of each mutant Nkd1 protein in CCD841 cells increased cell numbers compared to wild-type Nkd1 or empty vector control (Fig. 3E). Conversely, expression of wild-type Nkd1 in Co115 cells reduced cell numbers (Fig. 3F). These data indicate that the NKD1 mutations can promote b-catenin stabilization and colonic cell proliferation.

Altered subcellular localization and Dvl colocalization of truncated Nkd1
We were unable to detect endogenous Nkd1 in cell lines by immunocytochemistry, so to further investigate the relationship between Nkd1 localization and activity we examined the localization of tagged proteins in HEK-293 cells. Nkd1 localizes in a punctate, predominantly cytoplasmic distribution similar to Drosophila Nkd GFP when expressed in fly salivary gland (Fig. 4A, F). In contrast, Nkd1 C6 and Nkd1 C8 distribute diffusely in cytoplasm ( Fig. 4B and not shown), consistent with their enhanced detergent solubility relative to Nkd1 (Fig. 1E), while Nkd1 R288H aggregates like wild-type Nkd1 (not shown).
Dvl proteins localize to dynamic, cytoplasmic and plasma membrane-associated aggregates that have been proposed to amplify Wnt/b-catenin signaling [24]. Consistent with in vitro Nkd/Dsh association, expression of Nkd1 GFP in HEK-293 cells or fly Nkd GFP in salivary gland gave rise to intracellular aggregates with colocalized Nkd and Dsh/Dvl ( Fig. 4C-C0, F-F0, and not shown). In contrast, each Dvl co-synthesized with each truncated Nkd1 GFP (C6 or C8) formed aggregates, but colocalization was instead observed at aggregate interfaces, with Dvl aggregates typically surrounding truncated Nkd1 aggregates ( Fig. 4D-D0 and not shown). Although the significance of these localizations vis-àvis Wnt signaling is unclear, the truncated Nkd1 proteins identified in CRC exhibited a reduced ability to colocalize with Dvl proteins as compared to full-length Nkd1.

Mutant Nkd1 proteins are defective at binding Dvl proteins
Next we investigated the biochemical mechanism of defective Wnt signal inhibition by mutant Nkd1 proteins. The Nkd EFX motif binds the basic/PDZ region of Dsh/Dvl proteins [14]. Surprisingly, each Nkd1 mutant, despite having an intact EFX motif, bound each Dvl protein less than wild-type Nkd1 by yeasttwo-hybrid (Y2H) assay (Fig. 5A). Nkd1 truncation at the (C) 7 -tract (Nkd1 1-286 ) also reduced Dvl binding (Fig. 5A), indicating that reduced Dvl binding was not due to frameshift-induced unique Ctermini in the two truncated mutants. Each truncated Nkd1 protein retains near its C-terminus a 30aa amphipathic a-helical motif that is highly conserved (28/30 aa) in Nkd2 [14]. In fly Nkd, a similarly positioned 30aa motif is critical for function and nuclear localization [25], but the role of the 30aa motif in vertebrate Nkd proteins remains unknown. Deletion of the 30aa motif in Nkd1 1-286 restored Dvl binding, whereas further deletion of the EFX motif eliminated Dvl binding (Fig. 5A). These data suggest that one function of the vertebrate Nkd 30aa motif is to oppose Nkd1-EFX/Dvl interactions, which is itself apparently opposed by further C-terminal sequence that is deleted in our MSI-CRC tumors.
We confirmed the Y2H results by GST-pulldown and coimmunoprecipitation experiments. As shown in Fig. 5B, each Dvl protein exhibited reduced binding to GST-Nkd1 C6 and GST-Nkd1 C8 as compared to wild-type GST-Nkd1, while GST-Nkd1 R288H showed reduced associations with Dvl1 and Dvl2, but less so with Dvl3, similar to that seen by Y2H. When expressed in HEK-293 cells, Nkd1 C6 and Nkd1 C8 accumulated to higher levels than Nkd1 and Nkd1 R288H (not shown); by normalizing input lysates so that equal amounts of wild-type Nkd1 and each mutant Nkd1 protein were immunoprecipitated, we observed a two-fold reduction in the amount of Dvl3 co-immunoprecipitated by Nkd1 C6 or Nkd1 C8 as compared to Nkd1 or Nkd1 R288H (Fig. 5C). Thus, the NKD1 mutations reduce Nkd1/Dvl associations in vitro and in vivo.

Mutant Nkd1s are defective at altering Dvl levels
Dvls can be ubiquitinated and degraded by the proteasome, and the binding of Nkd or other Dvl-binding proteins leads to Dvl turnover [17,[26][27][28]. The NKD1 mutations might therefore compromise the ability of Nkd1 to destabilize Dvls. By expressing Drosophila Nkd GFP at different levels in adjacent cells of the thirdinstar Drosophila salivary gland, we consistently observed an inverse relationship between levels of Nkd GFP and Dsh ( Fig. 6A-A0). Since dsh transcription is not known to be regulated in Drosophila, Nkd GFP is likely destabilizing Dsh, as observed when Nkd1 was overexpressed in cultured mammalian cells [17]. Next, we cotransfected wild-type or each mutant Nkd1 with Dvl1, Dvl2, or Dvl3 into HEK-293 cells and examined the levels of each Dvl protein by western blot. As shown in Fig. 6B, Dvl1 and Dvl2, and to a lesser extent Dvl3, were less abundant when co-expressed with wild-type Nkd1 than when expressed alone. Neither empty-vector nor co-expressed GFP affected the levels of each Dvl (not shown). In contrast, the amount of Dvl1 detectable with co-expression of each mutant Nkd1 was similar to the Dvl1-only transfected control (Fig. 6B). Dvl2 levels were partially reduced by each mutant Nkd1, while Dvl3 levels were reduced less by coexpression of either truncated Nkd1 than by wild-type Nkd1. Similar to the inverse relationship between Nkd GFP and Dsh levels in Drosophila salivary gland ( Fig. 6A-A0), fine punctate Dvl1 immunoreactivity in cells with high levels of Nkd1 GFP appeared reduced compared to adjacent cells with lower levels of Nkd1 GFP (Fig. 6C,C9). Taken together, our data support the hypothesis that the specific alteration of Nkd1's ability to promote Dvl turnover might activate Wnt/b-catenin signaling during CRC tumor progression.

Discussion
Mutation of the tumor suppressor APC elevates Wnt/b-catenin signaling in the majority of the .1 million new cases of CRC diagnosed annually world-wide [3,29]. We hypothesized that mutations in other Wnt antagonists might elevate signaling in the subset of CRC, particularly MSI-CRC, without mutations in known Wnt regulators. We report three cancer-associated human NKD1 mutations that alter Wnt/b-catenin signaling and disrupt Nkd1/Dvl binding. Based on the frequency of NKD1 mutation (5%) identified in our cohort of MSI-CRCs, we estimate that NKD1 mutations occur in up to ,1% of newly diagnosed CRC, or ,10,000 cases per year.
Since MSI tumors are prone to mutation throughout the genome, the question arises of whether the NKD1 mutations drive tumor progression or are merely ''bystander'' mutations. A National Cancer Institute workshop [30] proposed five criteria to distinguish bona-fide target genes from bystander mutations, including a) high mutation frequency, b) biallelic inactivation, c) a role in a growth suppressor pathway, d) inactivation of the same growth suppression pathway in tumors without MSI through mutation in the same gene or in another gene within the same pathway, and e) functional suppressor studies, although the validity of these criteria in evaluating rare or novel driver mutations has been questioned {e.g. [6]}. Our work suggests that the NKD1 mutations fulfill four of the five criteria -a, c, d, and e. While the frequency of NKD1 mutation was relatively low compared to that of known Wnt pathway genes, a mutual exclusivity among mutations in NKD1 and other Wnt pathway genes was statistically significant in tumor samples. The absence of NKD1 mutations in our sample of tumors without MSI could be due to the rare nature of specific Nkd1 truncation in tumors with intact MMR, or due to our small sample size. However, other genes in the Wnt signaling pathway such as APC are frequently mutated, deleted, or methylated in tumors with intact MMR. Biallelic inactivation of NKD1 was not observed, but the presence of wild type Nkd1 in tumor cell lines, as well as the ability of mutant Nkd1 to stabilize b-catenin, suggests that the NKD1 mutations might act dominantly (see below). Finally, the Nkd family of proteins inhibits canonical Wnt signaling, and this activity is defective in all three mutant Nkd1 proteins, suggesting that the Nkd1 mutations alter Wnt/b-catenin signaling in vivo.
We further demonstrate that Nkd can limit Dsh/Dvl abundance in both mammalian cell culture and fly systems, with fly Nkd additionally having unknown but essential nuclear functions [25,31] (Fig. 6D). We propose that the mutant human Nkd1 proteins, each with a reduced ability to bind and limit the abundance of Dvls, increase Wnt-dependent Dvl activity, thereby attenuating b-catenin degradation and increasing TCF-dependent transcription of target genes that promote proliferation (Fig. 6E). However, preventing direct Nkd1/Dvl association is apparently insufficient to promote neoplasia, as deletion of the Dvl-binding Nkd1 EFX motif neither rendered mutant mice susceptible to cancer nor potentiated the frequency of Apc mutation-driven intestinal adenomas [32]. Similarly, mice carrying N-terminal truncating Nkd1 and/or Nkd2 mutations did not develop spontaneous tumors [33]. Since Nkd is integral to feedback loops in flies and vertebrates [12,34], we previously hypothesized that in mammals a lack of Nkd activity can be compensated by redundant feedback mechanisms, whereas in flies no such compensation is possible given the absence of genes encoding extracellular Wnt signaling antagonists [33].
The non-random pairing of dissimilar APC mutations in tumors {e.g., protein truncation near the mutation cluster region (MCR) with allelic deletion or methylation}, coupled with functional studies of mutant Apc proteins [35,36], has suggested that the cell must retain some ability to regulate Wnt/b-catenin signaling during tumor progression -the so-called ''just right'' hypothesis [37]. A consideration of the known roles for Wnt/b-catenin signaling during colorectal carcinoma progression provides some rationale for this hypothesis: in early stages, increased signaling promotes stem cell renewal and alters the migration of crypt epithelial cells [38], whereas later it acts as a switch to regulate epithelial to mesenchymal transitions during invasion and metastasis [39]. Thus, tumor progression might require transient up-or down-regulation of target gene expression depending on mutational load and local environmental conditions. Given that wild-type Nkd1 protein persists in the NKD1-mutant cell lines tested, we hypothesize that the mutant Nkd1 proteins -each of which retains multiple functional motifs (N-terminal myristoylation, EFX, and 30 aa motifs) -activate Wnt signaling in vivo, perhaps analogous to the manner in which Apc proteins truncated near the MCR activate Wnt signaling [36].
Despite the overwhelming evidence that abnormal Wnt/bcatenin signaling causes cancer, the role of Dsh/Dvl proteins in neoplasia remains obscure. Wnt signaling can promote Dsh/Dvl accumulation [40], and Dvl overexpression can mimic activation of the Wnt/b-catenin signaling axis [41], suggesting that Dvl hyperactivity, like b-catenin stabilization due to mutation, could be a primary cause of elevated Wnt signaling in cancer. Indeed, Dvl amplification and overexpression has been identified in neoplasia {e.g. lung cancer [42]}, but Dvl accumulation in cancer could also be a secondary consequence of unopposed Wnt ligand-driven autoactivation of signaling [43]. Given the crucial roles for Dsh/ Dvls in ''non-canonical'' Wnt pathways that govern planar-cellpolarity and cell migration in vertebrates [44], the NKD1 mutations might also alter Dvl activity in non-canonical Wnt pathways that control cell polarity or migration during cancer progression. Future experiments will focus on understanding how the mutant Nkd1 proteins alter Wnt signaling during cancer progression in vivo.

Ethics statement
This study was conducted according to the principles expressed in the Declaration of Helsinki. The study was approved by the Institutional Review Board of Mayo Clinic hospitals. All patients provided written informed consent for the collection of samples and subsequent analysis. Frog husbandry, in vitro fertilization, and embryo culture and staging were performed according to standard protocols and all animals were handled in strict accordance with good animal practice as defined by the American Association for Laboratory Animal Science, and the Xenopus studies were approved by the Animal Care and Use committee at the University of Pennsylvania.   a-Dvl3 (middle). b-actin is loading control (lower). Note that less Dvl3 is IP'd by Nkd1 C6 or Nkd1 C8 as compared to full-length Nkd1. No co-IP was observed between wild-type or mutant Nkd1s and Dvl1 or Dvl2, reminiscent of the lack of stable co-IP between Drosophila Nkd and Dsh [13]. doi:10.1371/journal.pone.0007982.g005 LS411) was provided by R.H.. DNA from 5 CRC cell lines without MSI (SW480 from U. Verma, UTSW; SW837, SW620, HT29, and Caco-2, from ATCC) was isolated as described [8]. Forty primary MSI-CRC and 50 CRC without MSI were collected at Mayo Clinic [8]. Following microdissection of tumor cells from sections of tumor specimens, DNA was extracted using the Easy-DNA TM kit (Invitrogen).
Xenopus injections pCS2+ EGFP, XWnt8, and Nkd1 (wild type/mutant) plasmids were linearized and mRNAs transcribed with SP6 mMessage mMachine (Ambion). Axis duplication assays were performed by injecting 1 pg XWnt8 +/2250 pg NKD1 (wild-type or mutant) with 200 pg EGFP (lineage tracer) mRNAs into one ventral blastomere at the 4-cell stage. Embryos were collected at stage 10, and lysates were analyzed by western blot with a-HA (Sigma) at 1:1,000. The remaining embryos were cultured until stage 35, fixed, and scored for axis phenotype: Single axis: wild-type; Single axis (short A/P): embryos without secondary axes, but with marked reduction in the A/P axis; Partial Axis Duplication: embryos with ectopic trunk structures, but no ectopic head tissues; Full Axis Duplication: embryos bearing ectopic trunk and anterior structures including cement gland and at least one eye or forebrain vesicle.

TOPflash assay
HEK-293 cells were cultured in DMEM+10% horse serum, 100 u/ml pen-strep, 1 mM sodium pyruvate, 2 mM L-glutamine, and 0.1 mM nonessential amino acids. 5610 4 cells were seeded into each well of a 24-well plate and co-transfected with 0.3 mg plasmid including 0.1 mg TOPflash (Upstate Biotechnology) +/20.1 mg . Arm complexes with Pan(TCF) to activate target genes including nkd. Nkd promotes Dsh turnover to partially inhibit signaling, and employs the nuclear import factor Imp-a3 to enter the nucleus and further inhibit signaling through unknown mechanisms [31]. (E) In NKD1mutant CRC, the mutant Nkd1 protein no longer binds and promotes Dvl turnover, stabilizing b-catenin and activating TCF-dependent transcription of target genes. doi:10.1371/journal.pone.0007982.g006 pCMV-Dvl2 GFP +/20.1 mg each pCMV-Nkd1 expression construct or empty vector +1 ng pRL-TK as a control for transfection efficiency. TCF reporter activity was measured 36 hr posttransfection with the Dual-Luciferase Reporter Assay System (Promega), and luciferase activities were measured using a Lumat LB-9507 luminometer (Berthold). Assays were performed in triplicate.

Detection of endogenous Nkd1 and b-catenin
Whole cell lysates: 2610 6 cells were harvested with 5 mM EDTA in PBS, pelleted at 300 g for 5 min at 4uC, lysed in NETN-100 lysis buffer (100 mM NaCl, 1 mM EDTA, 20 mM Tris pH 8.0, 0.5% NP-40 containing complete protease inhibitor cocktail (Roche), 10 mM NaF, and 10 mM b-glycerolphosphate), and lysates were centrifuged at 10,000 g for 10 min at 4uC. Protein concentrations were determined by Bradford assay (Bio-Rad). Supernatants were resolved by 10% SDS-PAGE. Triton X-100 lysates: 3610 5 cells were rinsed in PBS, and incubated with 0.5 ml 50 mM MES buffer, pH 6.8 (2.5 mM EGTA, 5 mM MgCl 2 and 0.5%Triton X-100) for 3 min at room temperature. Triton X-100soluble lysates were then aspirated into a fresh tube. The remaining material, scraped into 0.5 ml MES buffer and vortexed for 1 min, constituted the insoluble fraction. 50 mg each extract was resolved by 10% SDS-PAGE. After Hybond membrane (GE Healthcare) transfer, western blot was performed at 4uC overnight with rabbit polyclonal a-Nkd1 (Ab1, Cell Signaling Technology) at 1:1,000 with 0.1% Triton X-100 and 5% nonfat powdered milk (as blocking agent) in TBS, followed by incubation at room temperature for 1 hr with HRP-conjugated secondary antibodies (Pierce) at 1:1,000. Signals were visualized by the SuperSignal West Chemiluminescent Substrate kit (Pierce). The specificity of a-Nkd1 antisera was confirmed by western blot of nkd1-mutant tissues [33]. For Fig. 3D, HEK-293T cells were transfected with 4 mg each plasmid (wild type or each mutant Nkd1), and 48 hrs later subcellular fractionation was performed as described [46]. Concentrations of extract fractions were determined by BCA Protein Assay Kit (Pierce) and confirmed by immunoblots.
Cell proliferation assay 2610 3 cells/well were plated in triplicate into 96-well plates and infected 2 hrs later with retrovirus expressing wild-type or each mutant Nkd1, or retrovirus control (titers$10 7 /ml). Medium was replaced with 100 ml fresh medium 24 hrs post infection. Cell numbers were counted using the MTS assay (Promega).
Lysates containing each GST-Nkd fusion protein were prepared from BL21pLys E. coli (Amersham Biosciences). Each lysate was incubated with glutathione-Sepharose-4B beads for 1 hr at 4uC, and then washed 3x with DT80 buffer [47]. 35 [S]-Methioninelabeled Dvl1, Dvl2, and Dvl3 proteins were synthesized using TNT T7 coupled reticulocyte lysate system (Promega) and incubated with beads for 2 hr at 4uC in DT80 buffer. The beads were washed 4x with DT300 buffer, and labeled proteins were eluted in SDS-PAGE sample buffer, boiled 5 min, and then separated by SDS-PAGE on 8% gels. Gels were vacuum dried, and Dvl1-3 protein bands were quantitated using a Typhoon phosphorimager with ImageQuant software (Molecular Dynamics).

Immunofluorescence staining and microscopy
For b-catenin detection, CRC cell lines cultured on 12-well slides (Erie Scientific Co.) were fixed in 4% formaldehyde in PBS for 15 min, permeabilized with 0.2% Triton X-100 (1xPBS) for 5 min, and blocked in 3% BSA (1xPBS) for 30 min. The cells were incubated with mouse monoclonal a-b-catenin antibody (BD Transduction Laboratories) followed by Alexa 594-conjugated donkey anti-mouse IgG (Molecular Probes). Images were acquired with an Olympus BX41.