The Dkk3 gene encodes a vital intracellular regulator of cell proliferation

Members of the Dickkopf (Dkk) family of Wnt antagonists interrupt Wnt-induced receptor assembly and participate in axial patterning and cell fate determination. One family member, DKK3, does not block Wnt receptor activation. Loss of Dkk3 expression in cancer is associated with hyperproliferation and dysregulated ß-catenin signaling, and ectopic expression of Dkk3 halts cancer growth. The molecular events mediating the DKK3-dependent arrest of ß-catenin-driven cell proliferation in cancer cells are unknown. Here we report the identification of a new intracellular gene product originating from the Dkk3 locus. This Dkk3b transcript originates from a second transcriptional start site located in intron 2 of the Dkk3 gene. It is essential for early mouse development and is a newly recognized regulator of ß-catenin signaling and cell proliferation. Dkk3b interrupts nuclear translocation ß-catenin by capturing cytoplasmic, unphosphorylated ß-catenin in an extra-nuclear complex with ß-TrCP. These data reveal a new regulator of one of the most studied signal transduction pathways in metazoans and provides a novel, completely untapped therapeutic target for silencing the aberrant ß-catenin signaling that drives hyperproliferation in many cancers.


Introduction
The Dickkopf family of secreted glycoproteins is composed of four members that first appeared in early metazoans as key regulators of the Wnt/ß-catenin signaling pathway [1][2][3][4]. Three family members DKK1, DKK2 and DKK4 bind to the LRP5/6 and Kremen subunits of the receptor [5] and prevent assembly of a functional Wnt receptor complex [6][7][8]. The a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 EYIELIEIARNSTQDRILEMKVMEFFMKVYGYRGKHLGGSRKPDGAIYTVGSPIDYGVIVDT KAYSGGYNLPIGQAREMQRYVEENQTRNKHINPNEWWKVYPSSVTEFKFLFVSGHFKGNYKA QLTRLNHITNCNGAVLSVEELLIGGEMIKAGTLTLEEVRRKFNNGEINF-COOH. The homologous recombination donor DNA for the mouse target Dkk3 locus was assembled by PCR amplification. PCR primers are listed in (Table 1).
The 824 bp 5' homology arm (nt7415-nt8239) and 722 bp 3'homology arm (nt8342-nt9064) were appended to a loxP-TagCFP-pA-loxP cassette to create a linear 2.7 kb Dkk3b HR Donor DNA ZFN and HR Donor DNA. Validation was performed in the C8D1A cells (ATCC) derived from the C57Bl/6J mouse cerebellum. C8D1A cells were transfected with Dkk3 targeted ZFN plasmids and genomic DNA was isolated 48 h later. The 681 bp target locus was PCR amplified, heat denatured, re-annealed and indel formation evaluated by Cel-1 assay [36] (Transgenomic, Inc.). HR repair of the ZFN generated DSB was validated using ssDNA oligonucleotides to insert a unique restriction site at the DSB [37]. C8D1A cells were transfected with a 96-mer ssDNA oligo with a unique EcoRI site bracketed by 45 nt long homology arms (5'CAGTCTTGGCACCTATAGAAGAGGGGAAGAGAAGGCAGCCCCTTTTGAATTCTTGTAACT GAAAGAAACCGTTTGTGTCTTTGATTTGATAGATG) along with the Dkk3 targeted ZFN plasmids using Fugene6. The Dkk3 target sequence was PCR amplified from gDNA isolated 48 h after transfection. HR mediated repair of the ZFN generated DSB was confirmed by the EcoRI restriction of the PCR amplified target sequence.
Predicted off-target sites were identified by PROGNOS [38]. Potential off-target sites were PCR amplified from liver gDNA (founder #19) and cloned into the pCR4 vector (Life Technologies). Eight to twelve independent clones of each potential off-target site were DNA sequenced in both directions.
The transcription initiation site in exon 2 of the mouse Dkk3 gene was captured by 5'RACE of mRNA purified from the mutant mouse Dkk3 CFP cerebral cortex using the SMARTer 1 RACE 5'/3' kit from Clontech laboratories following manufacturer's instructions. Briefly, mRNA was purified using the Dynabeads 1 mRNA Direct™Kit from ThermoFisher according to manufacturer's instructions. Mouse brain mRNA (500 ng) was used for 1 st cDNA synthesis and the 5'UTR of the Cfp mRNA was amplified from the 1 st strand cDNA by 5'RACE PCR using the kit upstream UPM primer (5'-GTAATACGACTCACTATAGGGCAAGCAGTGGTAT CAACGCAGAGT) and two, independent, Cfp-specific downstream primers each with an appended infusion cloning sequence (bolded nucleotides): one located 384 nucleotides downstream from the Cfp ATG start site CFP384R (5'-gattacgccaagcttCTCGCCCTTCAGC TCGACGCGGTTCACC) and a second located 234 nucleotides downstream of the Cfp ATG start site CFP234R (5'-gattacgccaagcttCATGTGCTCGGGGTAGCGGGCGAAGC). Specific PCR products of~500 bp and~300 bp, respectively, were gel isolated, cloned into the pRACE vector (TaKaRa) and 15 clones from each Cfp primer were DNA sequenced.
Transient transfections were done with Fugene6 according to manufacturer's instructions. Cell lysates were prepared 48 h after transfection. Each transfection experiment repeated at least 3 times and each assay point was determined in triplicate.
Tunable ectopic protein(s) expression was done using the Tet-ON expression system (Life Technologies). The Tet Repressor Protein (TR) was integrated in the genome of PC3 cells by lentivirus infection (pLenti3.3/TR, Life Technologies) and cell selected with G418. TR expressing PC3 cells were transfected with pTREX-DKK3-HA or pTREX-Flag-IBS and ectopic protein expression induced by addition of increasing concentrations of doxycycline to the growth medium.

Luciferase reporter assays
Promoter analysis was done with PCR isolated segments of intron 2 of the rat Dkk3 gene. The Tcf promoter [41] (M50 Super 8x TOPflash) was a gift from Randall Moon (Addgene plasmid # 12456)); the E-Cad promoter [42] was a gift from Eric Fearon (Addgene plasmid # 19290)); and the E2F promoter [43] was the gift of Jason Chen. All promoter cDNAs were cloned into the pGL4 firefly luciferase vector (Promega). Cells were transfected with the individual promoter-reporter plasmids using Fugene6. After 48 h, cell lysates were prepared and luciferase assays were done according to manufacturer's instructions (Promega). ß-catenin signaling was evaluated in cells co-transfected with the Wnt expression plasmid, pcDNA-Wnt1 [44]; pcDNA-Wnt1 was a gift from Marian Waterman (Addgene plasmid # 35905). Each promoterreporter experiment was performed in triplicate, and each experiment was repeated at least three times.

Total RNA isolation, and real time PCR
Total cellular RNA was isolated using Trizol Reagent (Life Technologies) or RNeasy (Qiagen) according to the manufacturer instructions. RT-PCR was performed with SuperScript First-Strand Synthesis System (Life Technologies) and primed with either Oligo(dT) or random hexamers. First strand cDNAs primed with either Oligo(dT) or random hexamers yielded identical results. Initially, real time PCR (MJ Research Thermal Cycler) was done with a SYBR Green PCR kit (Qiagen). All subsequent experiments were done using multiplexed TaqMan

Co-immunoprecipitation binding studies
Cells expressing the epitope-tagged target proteins were lysed in 1X IP buffer (ThermoFisher Sci) with 150 mM NaCl, protease and phosphatase inhibitors. Lysates were incubated with anti-epitope antibody for 3 hrs at 4˚C and immune complexes collected on ProteinA/G-agarose beads (Sigma). Immune complexes were washed five times with lysis buffer and eluted in Laemmli buffer containing 1% SDS. Cell extracts and immune complexes were separated by SDS-PAGE and immunoblot analysis done with anti-Flag, anti-Myc and anti-HA antibodies.
Native ß-catenin and ß-TrCP in HeLa cells were co-immunoprecipitated from cells treated for increasing times with TAT-Dkk3b. Anti-Dkk3b affinity DynaBeads 1 (7.2 μg/mg resin) and control rabbit IgG affinity DynaBeads 1 (8.4 μg/mg resin) were prepared by coupling affinity purified anti-Dkk3b IgG [46] and rabbit IgG, respectively, to Dynabeads M-270 (Ther-moFisher) according to manufacturer's instructions. HeLa cell lysates were prepared in 1X IP buffer (ThermoFisher) supplemented with 150 mM NaCl and 1X protease inhibitors. Cell lysates (200 μg protein) were incubated with 1 mg of anti-Dkk3b or Control affinity beads, in duplicate. Immune complexes were collected using a DynaMag™magnet. Acid eluates were then separated on 10% SDS-PAGE gels and target proteins identified by immunoblot. All native co-immunoprecipitation experiments were repeated three times.

TAT-fusion protein synthesis
Dkk3b cDNA was cloned into the pTAT-HA plasmid [47], a gift from Steven Dowdy (Addgene plasmid # 35612). TAT-Dkk3b expression was induced with isopropyl ß-D-1-thiogalactopyranoside in BL21(DE3) cells. TAT-Dkk3b present in inclusion bodies was solubilized by denaturation in PBS containing 6 M urea, purified by affinity chromatography on Ni-NTA resin, and eluted with 150 mM imidazole in 6M Urea. Urea was rapidly removed by G10 spin column gel filtration and TAT-Dkk3b was stored at -20˚C at 1 mg/ml in PBS at pH 6.0.

The Dkk3 gene encodes multiple transcripts
The unaltered ß-catenin signaling in the Dkk3 knockout mouse (Dkk3 tm1Cni ) [20] led us to reexamine the biological relevance of an intracellular~30 kDa DKK3 isoform (D2p29) that showed dynamic, microfilament dependent intracellular trafficking in rat astrocytes [46,48]. Amino acid sequence alignment showed that the secreted DKK3 and D2p29 (designated hereafter Dkk3b) shared all but the N-terminal 71 residues that comprise the signal peptide sequence and N-glycosylation sites.
The Dkk3 tm1Cni mutant mouse was generated by replacement of the majority of exon 2 of the Dkk3 gene with an in-frame LacZ-stop cassette [20]. Exon 2 encodes the N-terminal 71 amino acids responsible for directing DKK3 to the secretory vesicle, and also harbors a biologically important CpG island ( Fig 1A).
The first codon in exon 3 of the Dkk3 gene encodes the initiator methionine of Dkk3b from frogs to man, suggesting that Dkk3b is generated from a second Dkk3b transcript, possibly initiating within the 6 kb intron 2 ( Fig 1A). Using exon specific qPCR to quantify Dkk3 locus transcripts containing exon 2 or exon 3 (Fig 1B, S1 Fig), we found that all Dkk3 transcripts present in mouse astrocyte mRNA had Dkk3 exon 3 codons, while only~60% generated a PCR product with Dkk3 exon 2 specific PCR primers, suggesting that~40% of the total Dkk3 mRNA lacked the exon 2 codon and could presumably encode Dkk3b (S1 Fig). The specificity of the exon-specific qPCR of these Dkk3 exons was validated using total RNA from Dkk3 tm1Cni mutant mouse brain; no Dkk3 transcripts with exon 2 codons were detected and only Dkk3 transcripts with exon 3 codons were found (Fig 1B).
Promoter: luciferase reporter assays were used to search for transcriptional regulatory elements located in intron 2 of the rat Dkk3 gene. Robust promoter activity is found in the 750 nucleotides of intron 2 immediately upstream of exon 3 and progressive deletion from the 5' end of this intron 2 fragment localized a functional promoter (TSS2) to the 250 nucleotides just upstream of the 5' end of exon 3 (Fig 1C). The essential TATA box was located at -35 nucleotides upstream from the 5' end of exon 3 of the Dkk3 gene (Fig 1C, see construct  P259D); in mouse and human Dkk3 genes a TATA box is located~90 nucleotides upstream of exon 3 and the transcriptional start site was determined to be at -92 nucleotides upstream of exon 3 by 5'RACE analysis of the Cfp mRNA surrogate of DKK3b (see S2 Fig). Chromatin immunoprecipitation (ChIP) of rat astrocyte gDNA revealed that the native TSS2 in the Dkk3 gene bound RNA pol II (Fig 1D) and the TATA Box Binding Protein (TBP) (Fig 1D) indicating that the TATA box at -35nt in intron 2 (see Fig 1C) is functional.

Dkk3b is a biologically active product of the Dkk3 gene in vivo
To explore the biological significance of Dkk3b, we used targeted homologous recombination (HR) driven by artificial nucleases to selectively disrupt this transcript within the mouse genome. A promoter trap knock-in strategy was devised using zinc finger nucleases [33] (ZFNs) to create a double stranded break (DSB) in intron 2 of the Dkk3 gene between TSS2 and exon 3 that would facilitate the HR mediated insertion of a floxed CFP-stop cassette. Splice junctions at exon 3 were preserved so that TSS1-driven transcripts encoding the secreted DKK3 could produce full-length, spliced mRNA (Fig 2A-2C). The CFP insert diverts TSS2-driven transcription to the CFP surrogate resulting in the selective, functional loss of Dkk3b transcripts in homozygous animals. Immortal C8D1A cells, isogenic to the C57Bl/6j mouse, Cel-I assays, and single stranded oligonucleotide directed HR repair were used to validate the targeting strategy and ZFN-facilitated donor DNA HR. ZFN-generated DSBs and HR repair in the presence of the donor DNA resulted in weak expression of CFP in C8D1A cells ( Fig 3A). These modified cells retain an intact exon 2 CpG island(s) upstream of the edited Dkk3 locus [49][50][51]. Since CpG island methylation is common in immortalized cells [52] and presumably depresses expression of REIC (Reduced Expression in Immortal Cells)-a synonym for DKK3 [11]-we examined whether the methyltransferase inhibitor, azacytidine, would enhance expression of the TSS2-driven CFP decoy. Inhibition of DNA methyltransferase activity in the gene-edited C8D1A cfp/wt cells resulted in a >5-fold increase in CFP expression ( Fig 3A).
C57Bl6 mouse zygotes were then injected with ZFN Dkk3b mRNAs and a linear HR donor DNA to create the Dkk3b CFP knock-in mouse. Thirty-five of 65 (54%) injected one cell embryos produced viable pups, and DNA sequencing of the target locus confirmed that 1 male (#17) and 2 females (#18 & #19) (8.6%) had HR mediated insertion of a floxed CFP decoy inserted 35 nucleotides upstream from exon 3 (see Fig 2C) with intact splice junctions. No offtarget mutations were found for the 10 highest predicted candidate target sites in founder #19 ( Table 2). F1 progeny from crosses of the Dkkb3 CFP/+ male and a wild type female showed Mendelian inheritance patterns characteristic of a single segregating allele ( Table 3). The tissue distribution of Dkk3b expression was evaluated using the~26 kDa CFP surrogate in the Dkk3b CFP/+ mutant mouse. Immunoreactive CFP was found throughout the Dkk3b CFP/+ mouse (Fig 3D) illustrating the ubiquitous nature of TSS2 activity of the Dkk3 gene.
The mRNA for the CFP surrogate of Dkk3b was then used to establish the TSS2-driven transcription initiation site by 5'RACE of polyadenylated mRNA isolated from the cerebral cortex of the mutant Dkk3 CFP mouse. The Cfp mRNA began at nt8284 of the Dkk3 gene,~90 nt upstream of the ATG start site of exon 3 of the Dkk3 gene (S2 Fig). The 5'UTR of the Cfp transcript also harbored the imbedded LoxP site used to excise the promoter trap reporter illustrating that the TSS2 is functional at this mouse gene locus.
No viable homozygous mutant offspring were found after mating of heterozygous Dkk3b CFP/wt mice (Table 3), and no homozygous mutant blastocyst implants were found as early as embryonic day 5.5 (n = 34 embryos) suggesting that DKK3b is essential for development at or near the time of embryo implantation. This outcome differs markedly from that of the Dkk3 tm1Cni mouse and suggests that at least one wild type Dkk3b allele is required for survival. The penetrance of the lethal phenotype for the single segregating Dkk3b CFP allele was confirmed in out-crosses on the CD1 background (Table 3). Target locus modification confirmed using PCR primers anchored outside of the HR region (LF and RR) and overlapping in the CFP cds (LR and RF). PCR products (LF:LR and RF:RR) were sequenced in both directions. Genotyping PCR primers indicated by arrows (see Table 1 for sequences). Excision of the floxed CFP cassette (ΔCFP) in the unfertilized mutant oocyte using a Sox2 promoter-driven Cre recombinase [53,54] rescued the lethal phenotype of the Dkk3b CFP mutation and left behind a diagnostic single 34 bp loxP remnant at the target locus. Bi-allelic, gene-edited Dkk3b ΔCFP/CFP offspring were recovered from crosses of Dkk3b ΔCFP/+ to a Dkk3b CFP/+ mice (Table 3) and showed Mendelian inheritance confirming that embryonic  lethality resulted from the interruption of Dkk3b transcription rather than any tightly linked cis gene defect(s). Since embryonic lethality prevented the generation of homozygous mutants, ex vivo gene editing of heterozygous mutant MEFs was used to confirm the promoter trap knock-in strategy. Bi-allelic disruption of the Dkk3b locus was done in heterozygous mutant MEFs (Dkk3b CFP/+ ) using a second round of ZFN-initiated, HR repair to insert an mCherry reporter into the remaining wild type allele. Forty-eight hours after ex vivo gene editing, viable MEFs with bi-allelic mutations at the target locus expressing both fluorescent proteins were found attached to the dish, as well as free floating in the growth medium. Mutant Dkk3b CFP/mCherry MEFs, both attached and free floating, were pooled and sorted by FACS. Immunoblots of sorted cell lysates showed that the Dkk3b CFP/mCherry MEFs continued to express the~65 kDa glycosylated DKK3 protein, while the 30 kDa DKK3b was absent (Fig 3B). Exon-specific qPCR confirmed preservation of the secreted Dkk3 transcript and the selective loss of the Dkk3b transcript in Dkk3b CFP/mCherry MEFs (Fig 3C). Attempts to propagate these mutant MEFs was prevented by the severe attachment defect associated with the DKK3b-deficient cells. These data suggest that DKK3b is necessary for cell-cell interactions, required for embryogenesis, and that TSS2-driven CFP is a surrogate for Dkk3b expression.
RNAi knock down was used to explore the role of DKK3b on MEF attachment, cell proliferation and its relationship to ß-catenin signaling. RNAi knockdown of the Dkk3/Dkk3b and/or ß-catenin transcripts in MEFs was confirmed by qPCR ( Fig 4A). The effects of Dkk3b knockdown on ß-catenin dependent gene expression was examined using the ß-catenin responsive TOPflash luciferase reporter. Dkk3b KD MEFs showed a 10-fold increase in basal TOPflash activity compared that of the unaltered control and non-silencing KD cells (Fig 4B). Simultaneous knockdown of Dkk3/Dkk3b and ß-catenin expression completely blocked this increased signaling. Similar to the Dkk3b-null MEFs (see above), >70% of the viable Dkk3/Dkk3b KD cells were found free-floating after 3 days in culture (Fig 4C). Cell attachment and proliferation during this 3-day growth period was unaffected in cells expressing a non-silencing shRNA control or in ß-catenin KD cells (Fig 4D). The attachment defect in Dkk3/Dkk3b KD cells was completely reversed by simultaneous knockdown of ß-catenin mRNA (Fig 4D and 4E). Replacement of cellular DKK3b by addition of a cell penetrating TAT-DKK3b protein to the growth medium [55] completely restored cell attachment of the Dkk3/Dkk3b knockdown MEFs to that of the controls, but provided no additional benefit to either the ß-catenin KD cells or in ß-catenin KD:Dkk3/Dkk3b KD MEFs (Fig 4D and 4E).

DKK3b modulates cell proliferation and ß-catenin signaling
The relationship between DKK3b and the ß-catenin signaling pathway was further defined by cell proliferation, promoter-driven reporter assays, and cell migration analysis in prostate and breast cancer cells. Tet-inducible constructs of the intracellular DKK3b and the secreted DKK3 were used to provide fine control over exogenous expression levels and to avoid the untoward effects of over-expression. In the DKK3/DKK3b-deficient, PC3 prostate cancer line, expression of DKK3b arrested cell proliferation of the (Fig 5A) at the G0/G1 phase of the cell cycle ( Fig 5B) and led to the loss of DKK3b expressing cells by 24-36 h of induction (Fig 5A). Unlike prior over-expression studies [11,19,56,57], induction of equivalent levels of secreted DKK3 did not alter PC3 cell proliferation (Fig 5A and 5B).
These data show that DKK3b has the anti-proliferative activity in cancer cells that were previously associated with gene product(s) from the Dkk3 locus [11,56].
The relationship between DKK3b and ß-catenin signaling was defined by cell proliferation and promoter-driven reporter assays in HEK293 cells with reduced endogenous DKK3b. Basal cell proliferation was unaffected by ectopic DKK3b produced either by transient transfection and by addition of the TAT-DKK3b (Fig 6A and 6B). On the other hand, Wnt-stimulated cell proliferation was progressively slowed, but not arrested in cells treated with DKK3 by either transient transfection or by TAT-Dkk3b (Fig 6B). The more robust decrease in proliferation observed with TAT-DKK3b treatment is likely due to the more uniform delivery of the tumor suppressor to the cell monolayer. Concentrations of TAT-DKK3b !2.5 μg/ml completely silenced Wnt-stimulated cell proliferation without altering basal cell proliferation (Fig 6B, compare basal and Wnt stimulated proliferation at 2.5 μg/ml TAT-DKK3b).
Primary and downstream promoter-luciferase reporter assays determined the effects of DKK3b on ß-catenin-driven gene expression. HEK293 cells were co-transfected with Wnt1 and promoter-driven firefly luciferase constructs paired with a control CMV-driven renilla luciferase cDNA, and then treated with TAT-DKK3b for 24 h. Wnt1 stimulated cells showed a 65-fold increase in TOPflash activity and TAT-DKK3b completely arrested expression of this canonical ß-catenin reporter (Fig 6C). DKK3b also modulated downstream ß-catenin regulated pathways that reduce cell adhesion (ECad promoter [58]) and promote cell cycle progression (E2F promoter [59]). Wnt1 silenced E-Cad promoter activity by 90%, but the presence of TAT-DKK3b restored promoter activity to basal levels (Fig 6C). Similarly, Wnt1 increased E2F-promoter activity 6-fold, but the presence of TAT-DKK3B maintained E2F-promoter activity at baseline levels ( Fig 6C). Taken together, these data show that DKK3b modulates multiple aspects of ß-catenin signaling.

DKK3b blocks nuclear translocation of ß-catenin
Yeast two hybrid screens showed that DKK3 interacted with the E3 ubiquitin protein ligase, ß-TrCP, and that this complex captured cytoplasmic dephosphorylated ß-catenin [17], although DKK3b regulation of cell proliferation the source of the essential intracellular effector produced by the Dkk3 gene was not identified [17,18]. DKK3b provides just such an intracellular effector.
Co-IP studies were done with lysates of HEK293 cells constitutively expressing epitopetagged DKK3b, ß-TrCP and the constitutively active S33Y mutant of ß-catenin that evades the destruction complex. When all three proteins were present, immune precipitates of Flag-S33Y ß-catenin also contained both ß-TrCP and DKK3b, while control IgG precipitates failed to capture any epitope tagged targets (Fig 7A). Similarly, immune precipitates of myc-ß-TrCP contained both S33Y ß-catenin and DKK3b; immune precipitates of HA-DKK3b contained both S33Y ß-catenin, and ß-TrCP (Fig 7A). Co-IP studies of cell lysates expressing only two of the three binding partners (Myc-ß-TrCP and HA-DKK3b) or (FLAG-S33Y ß-catenin and HA-DKK3b) failed to show any interaction between these epitope-tagged binding partners ( Fig 7B). These data suggest that DKK3b captures unphosphorylated ß-catenin in a complex with ß-TrCP and that all three partners are required to assemble this complex.
The role of ß-TrCP in the DKK3b:ß-TrCP:ß-catenin complexes on nuclear trafficking and gene expression was examined in ß-TrCP KD HeLa cells. Immunoblot analysis confirmed shRNA dependent loss of the ß-TrCP transcript in KD cells (Fig 7C). To maximize nuclear localization of ß-catenin, control and ß-TrCP KD cells were treated with LiCl, a chemical mimic of Wnt that stabilizes cytosolic ß-catenin by inhibiting its phosphorylation by GSK-3 in the destruction complex [60]. The effects of the DKK3b:ß-TrCP:ß-catenin complex on ß-catenin-driven gene expression in ß-TrCP KD HeLa cells was then evaluated using the TOPflash assay. In both resting and LiCl-stimulated control and non-silencing KD cells, addition of TAT-DKK3b silenced TOPflash activity (Fig 7D). In ß-TrCP KD cells, TAT-DKK3b had no effect on TOPflash activity, whereas rescue of the ß-TrCP KD by transfection with mouse ß-TrCP cDNA restored the ability of TAT-DKK3b to arrest TOPflash activity in both resting and LiCl-stimulated cells. These data show that the DKK3b:ß-TrCP:ß-catenin complex alters the ability of the ß-catenin to enter the cell nucleus and blocks ß-catenin-driven gene expression.
The effects of DKK3b on the dynamics ß-catenin nuclear trafficking was determined in LiCl stimulated HeLa cells. HeLa cells lack native Dkk3/Dkk3b expression and show accelerated ß-catenin-driven cell proliferation [61]. As expected, HeLa cell ß-catenin was distributed throughout the cell interior with marginal localization at the cell periphery (Fig 8A). In the absence of DKK3b, LiCl led to the rapid accumulation of ß-catenin in the nucleus reaching maximal levels after 60 min in >70% of the treated cells, and this remained constant for 18 h when LiCl was present (Fig 8A and 8B). Addition of TAT-DKK3b along with LiCl completely blocked the accumulation of nuclear ß-catenin for up to 18 h.
The time dependent efflux of nuclear ß-catenin due to TAT-DKK3b was determined in HeLa cells stimulated with LiCl for 18 hrs. Addition of TAT-DKK3b to these maximally stimulated cells resulted in the rapid loss of nuclear ß-catenin, beginning within 30 min and reaching maximal suppression by 60 min (Fig 8C). TAT-DKK3b-suppressed nuclear ß-catenin below levels in untreated HeLa cells for at least 3 h. Total HeLa cell ß-catenin levels were unaffected by TAT-DKK3b (Fig 8D) indicating that the loss of ß-catenin from the nucleus in TAT-DKK3b treated cells was due to cellular redistribution rather than enhanced ß-catenin degradation. Anti-DKK3b-IgG-affinity beads were used to capture DKK3b associated native proteins in the LiCl-stimulated HeLa cell lysates (Fig 8E). The loss of nuclear ß-catenin in LiCl-stimulated cells was accompanied by the time-dependent accumulation of both ß-catenin and ß-TrCP in a complex with the TAT-DKK3b (Fig 8E). Total cellular levels of both ß-catenin and ß-TrCP were unaltered during the 3 h TAT-DKK3b treatment period, while intracellular DKK3b showed a time-dependent accumulation in the complex that paralleled the loss of ß-catenin from the cell nucleus ( Fig 8E). Thus, the DKK3b:ß-TrCP:ß-catenin inhibitory complex is formed rapidly, interrupts the dynamics of nuclear import/export, and defines the molecular basis for the silencing of ß-catenin signaling by DKK3b.
Since DKK3b silenced both TOPflash activity and captured the native transfactor in an inhibitory complex with DKK3b:ß-TrCP, we determined the effects of TAT-DKK3b on native gene expression in resting and LiCl-stimulated HeLa cells. Like the TOPflash assay, addition of TAT-DKK3b to unstimulated HeLa cells led to a 40-60% decrease in both native   to control ß-catenin translocation to the cell nucleus, and that this action depends on the the ability of DKK3b to capture the native transactivator in an extranuclear complex with ß-TrCP.

Discussion
DKK3 is the enigmatic member of an ancient family of secreted glycoproteins that regulate the Wnt/ß-catenin pathway by interrupting the assembly of a functional Wnt liganded receptor [4,11]. It is the only unambiguous tumor suppressor in the family, and a diverse literature links DKK3 and tumor suppression to the ß-catenin pathway [11]. Unfortunately, the inability of DKK3 to block Wnt receptor assembly due to steric hindrance [4,12] poses a significant challenge to our understanding of its tumor suppressor function. Yeast two-hybrid screens [17] showed that DKK3 formed a complex with ß-TrCP, and dephosphorylated ß-catenin that prevented transfactor nuclear translocation [17], but Lee and co-workers did not identify the authentic intracellular DKK3 isoform responsible for this biology.
The recognition that the Dkk3 gene encodes a second gene product, DKK3b-a vital intracellular protein that regulates ß-catenin trafficking-provides the missing component that connects the Dkk3 gene to its regulation of the ß-catenin signaling pathway. DKK3b acts on the ß-catenin signaling pathway independent of the Wnt modulated destruction complex and modulates a diverse array of cellular functions, such as cell proliferation, cell attachment, embryogenesis and gene expression. Positioned downstream of the Wnt regulated degradation complex, DKK3b modulates ß-catenin trafficking to the nucleus. In astrocytes, DKK3b rapidly shuttles between the perinuclear space and the plasma membrane using myosin motors and actin fibers [48,62], suggesting that it may carry ß-catenin back the cell periphery for reuse in the adherens complex remodeling. The ability of DKK3b to repair the attachment defect observed in DKK3b-null MEFs suggests that DKK3b plays a key role in the delivery of ß-catenin to its plasma membrane reservoir.
The Dkk3 gene joins a growing list of mammalian genes that use multiple alternate promoters to achieve functional diversity [63,64]. Our identification of a second TSS element in the Dkk3 gene and the demonstration that it generates a second transcript encoding a vital intracellular isoform was unexpected. Using targeted gene editing to abolish Dkk3b expression, while preserving expression of its sister (Dkk3), we found that DKK3b was essential for early embryonic development at or near implantation, dramatically different from the Dkk3 tm1Cni mutant mouse where secreted Dkk3 expression was inactivated [20]. Despite the presence of the entire Wnt/ß-catenin pathway from oocyte to the late blastocyst stage [65,66], active canonical Wnt signaling is not observed in pre-implantation blastocysts [67], and is first observed post-implantation at E6.5 [68]. Using a Wnt reporter mouse expressing a functional mutant of ß-catenin that evades the destruction complex, Kemler et al [69] found the stabilized ß-catenin in the cytoplasm of pre-implantation embryos, but that it did not traffic to the nucleus and no Wnt reporter expression was observed. They suggested that cellular mechanism(s) other than the canonical destruction complex [69] were responsible for keeping the stabilized ß-catenin out of the cell nucleus before implantation. Since DKK3b is an important gatekeeper of ß-catenin nuclear translocation, its loss in the zygote is likely to unleash ß-catenin signaling that has detrimental consequences on the developing embryo both before implantation and at this critical developmental event.
were scored as in B, and the data reported as means ± SD of 3 independent experiments. (D) Immunoblot of the effects of TAT-DKK3b on ß-catenin levels in HeLa cell lysates. Data reported as means ± se, n = 3. (E) Time dependent assembly of the TAT-DKK3b:ß-TrCP:ß-catenin complex in LiCl-stimulated HeLa cells. Cell lysates were incubated with anti-DKK3b-IgG or NRS-conjugated DynaBeads, and co-precipitating proteins determined by immunoblot with anti-target antibodies. https://doi.org/10.1371/journal.pone.0181724.g008 DKK3b-deficient MEFs showed both elevated ß-catenin signaling and cell attachment defect(s); two events likely to interrupt orderly development of the embryo. These disruptions in the signaling pathway were rescued by replacement with ectopic DKK3b or by simultaneous knockdown of ß-catenin. These data illustrate that the loss of DKK3b directly leads to aberrant ß-catenin signaling and disrupts cell-cell and cell-substrate interactions.
Unlike its ubiquitous expression in somatic cells, silencing of the Dkk3 gene is common in cancer and ectopic over-expression of DKK3 arrests tumor cell growth [16][17][18][19]. Analysis of intracellular product of the Dkk3 gene, revealed that DKK3b arrests Wnt stimulated cell proliferation and selectively silences ß-catenin dependent gene expression in both immortal and cancer cells. The ability a purified cell-penetrating DKK3b protein to fully restore control of ßcatenin signaling by partnering with ß-TrCP to capture the transfactor in an extra nuclear complex offers a promising new therapeutic target for ß-catenin driven hyper-proliferative disease(s) like cancer.
An essential partner in the molecular mechanism of DKK3b action is ß-TrCP, an Fbox protein with WD40 repeats that captures a broad range of protein targets through a consensus dephosphorylated 6 amino acid long degron domain in the target protein(s) that is located 10 to 20 residues downstream of the lysine used for ubiquitin conjugation [70]. DKK3b captured dephosphorylated ß-catenin in a complex with ß-TrCP, prolonged the biological half-life of the transfactor, and prevented its translocation to the nucleus (Fig 9). Direct analysis of the dynamics of ß-catenin nuclear entry and exit showed that the tripartite complex trapped the transfactor outside the nucleus providing a clear molecular mechanism for the DKK3b-dependent regulation of this pathway. Since the N-terminal localized degron of ßcatenin docks with ß-TrCP, the ability of DKK3b stabilize transient interaction(s) between the dephos-degron of ß-catenin and ß-TrCP during "kiss-and-run" interaction(s) [71], suggests that other ß-TrCP ligands that similarly engage this E3 ligase are potential regulatory targets. These include members of most of the kinase signaling cascades that impact cell growth, motility, and apoptosis [71][72][73][74]. Further studies are required to determine if DKK3b captures other ß-TrCP interacting proteins and impacts their signaling pathways.
DKK3b is a novel and essential component of the Wnt/ß-catenin signaling pathway that plays a key role in both development and cancer. It serves as a gatekeeper for ß-catenin nuclear entry, directly modulates this pro-proliferative signaling molecule, and provides an important new point of control that impacts the regulatory pathways responsible for differentiation, lineage specification, pluripotency and oncogenesis.