Dacts are phosphoproteins that migrate at higher than expected molecular weight on SDS-PAGE

Some previous studies and commercial antibody sources have reported apparent molecular weights for full-length Dact1 proteins as less than 100 kD [2, 9] consistent with bioinformatic predictions based on primary sequence information but inconsistent with our previously published biochemical data [1, 8]. Using SDS-PAGE, recombinantly expressed full-length Dact1 and Dact2 consistently migrate between 100-120 kD [1, 8] and Dact3 migrates between 75-100 kD [4, 23]. Part of the apparent discrepancy between bioinformatic prediction and experimental observation is due to phosphorylation in vivo [1, 11], as demonstrated by a downward mobility shift when cell lysates containing Dact proteins are pan-dephosphorylated (Figure 1A-C, left panels). Since even pan-dephosphorylated Dact proteins migrate at a larger than expected size, we checked for evidence of other post-translational modifications that can variably affect apparent molecular weight by SDS-PAGE, such as glycosylation. However, treatment of Dact paralogs with an enzymatic deglycosylation cocktail caused no shift in their apparent molecular weight (Figure 1A-C, right panels), nor could we detect any evidence of glycosylation using dye-based methods such as periodic acid-Schiff staining (data not shown).

All murine Dact paralogs form complexes with CK1δ/ε homologs

One of the initial reports identifying Dact1 in Xenopus laevis documented complex formation with CK1ε when the protein was expressed in mammalian cell lines [1]; a later study showed that CK1δ-mediated phosphorylation of the X. laevis Dact1 protein alters its Wnt/β-catenin signaling activity in a cell-free system [17]. We tested whether interaction with CK1δ/ε was specific to Dact1 or a general feature of all Dact family members. When recombinantly expressed in HEK293 cells, all three murine Dact paralogs formed complexes with murine CK1δ (Figure 2A). We reasoned that if this interaction were functionally important it might occur with more divergent members of the CK1δ/ε family, such as the single CK1δ/ε homolog doubletime/discs overgrown (dbt/dco) from Drosophila melanogaster, in which no Dact homolog has yet been identified. Indeed, all three murine Dact paralogs formed robust complexes with Drosophila dbt/dco (Figure 2B).

Similarly, Protein Kinase A (PKA) has recently been reported to associate with human DACT1 in HEK293T cells, regulating its activity in Wnt/β-catenin signaling [11]. Concordantly, we found that the catalytic subunit of Protein Kinase A formed complexes with all three murine Dact family members when co-expressed in HEK293T cells (Figure 2C). Protein Kinase C (PKC) has not previously been tested for interactions with Dact proteins, but has been implicated repeatedly in different types of Wnt signaling [24–29]. We found that it formed complexes with all three Dact paralogs when expressed in HEK293T cells-most robustly with Dact2, followed by Dact1 (Figure 2D).

Of the serine/threonine kinases tested, the most robust and conserved interactions were with CK1δ/ε, PKA, and PKC. In contrast, Casein Kinase 2a1 (CK2a1) formed a weak complex only with Dact1 (Figure 2E). Casein Kinase 2a2 (CK2a2) showed no appreciable complex formation with any murine Dact family member (Figure 2E, F). Casein Kinase 2b (CK2b) formed complexes only with Dact1 and Dact2 (Figure 2G). GSK3β, which is central to Wnt/β-catenin signaling and has been reported to interact with Dact1 [1, 2], in our assays formed complexes only weakly with Dact1 and not appreciably with either Dact2 or Dact3 (Figure 2H). GSKα behaved indistinguishably from GSKβ in this respect (Figure 2I).

All murine Dact paralogs form complexes with all Dvl homologs

Though homologous in the sequences and positions of a few well-conserved domains, the three mammalian Dact paralogs are nevertheless only modestly conserved across their overall primary sequence (~20% identity), and have distinct though overlapping domains of tissue expression during development and in the adult [4]. In contrast, the three mammalian Dvl paralogs are more conserved at the primary sequence level (> 60% identity) and are ubiquitously or near-ubiquitously expressed during development and in adult tissues [30–32]. This, combined with evidence that different Dact paralogs have distinct signaling functions in vivo [16, 22], raises the question of whether some Dact paralogs might preferentially associate with only a subset of co-expressed Dvl proteins, or perhaps not associate with Dvl proteins at all. We tested this hypothesis and found that all three murine Dact paralogs formed complexes with all three murine Dvl paralogs (Figure 3A-C). Furthermore each Dact paralog formed complexes with each Dvl paralog indiscriminately, with the sole exception that Dact2 reproducibly showed a particularly strong interaction with Dvl3 (Figure 3C, lane 2 vs. lanes 1 & 3). As with CK1δ/ε, all three Dact paralogs also formed complexes with the D. melanogaster Dvl homolog, dsh (Figure 3D).

All Dact paralogs form complexes with Vangl proteins; TGFβ receptor interaction is relatively weaker

In the mouse embryo, constitutive loss of Dact1 leads to post-translational upregulation of the Vangl2 transmembrane protein in cells undergoing epithelial-to-mesenchymal transition at the primitive streak with consequences on gastrulation and subsequent morphogenic events in the posterior mesoderm and endoderm [8]. This finding in genetically-engineered mice led to our discovery that in addition to the Dvl proteins that bind to Vangl2 [33, 34], Dact1 binds to Vangl2 via independent domain interactions [8]. There are two paralogous Vangl proteins in mammals (Vangl1 and Vangl2) that at least partially overlap in function [35]. We accordingly tested the hypothesis that all Dact paralogs can form complexes with Vangl paralogs. We found that all three Dact proteins formed robust complexes with Vangl1 (Figure 4A). However, to our surprise there were some differences in the affinity of each murine Dact protein for Vangl2. Specifically, by coIP assay Dact1 formed the most robust complexes with Vangl2 (both in HEK cells, not shown; and in HEK293T cells, shown), followed by Dact3, and then by Dact2 which formed complexes with Vangl2 at levels just detectable above background (Figure 4B). As with the CK1δ/ε and Dvl protein families, all three murine Dact paralogs readily formed complexes with the sole D. melanogaster Vangl family member, Vang/Stbm (Figure 4C).

Dact2 has been implicated in TGFβ signaling via binding, endocytosis, and lysosomal degradation of the Alk4/5 subtype of TGFβ receptor proteins [22]. Combined with the observations above regarding Dact protein binding to the Vangl transmembrane protein family, this raises the possibility that Dact proteins might be involved in endocytic turnover and degradation of multiple classes of transmembrane protein. We therefore sought to replicate complex formation between Dact2 and Alk5, and also asked whether all Dact proteins interact similarly with TGFβ receptors. Relative to the Vangl proteins, we observed weaker complex formation between murine Dact proteins and Alk5. In HEK293 cells we were unable to detect complex formation between Alk4 or Alk5 and any Dact protein (Figure 4D and data not shown). In HEK293T cells we could replicate weak complex formation between both the wild type and a constitutively active point-mutated form of Alk5 [36] (Figure 4E, F); the coIP of Alk5 was weakly positive with Dact1, and negative with Dact3 (Figure 4D-F).

Complex formation with catenin proteins is relatively weak and most conserved for p120ctn

When co-expressed in tissue culture cells Dact1 can form complexes with β-catenin [1, 2] and this interaction has been mapped to the β-catenin armadillo repeat region [2], a structurally-conserved protein-interaction domain shared with other members of the catenin superfamily as well as with other proteins [37]. Dact1 has also been shown to bind and regulate the catenin p120ctn [10]. We therefore tested interactions between the three murine Dact paralogs and representatives from each major class of the catenin superfamily. No Dact paralogs formed complexes with α-catenin (Figure 5A), which lacks armadillo repeats. In contrast, Dact2 and Dact3 formed complexes, albeit weakly, with β-catenin in HEK293T cells; Dact2 exhibited the stronger β-catenin coIP (Figure 5B). Dact2 also showed the strongest coIP with δ-catenin; Dact1 interacted weakly whereas complex formation between δ-catenin and Dact3 was not detectable above background (Figure 5C). Among members of the catenin superfamily, the Dact interaction that was most conserved was with p120ctn (Figure 5D). Notably, even positive coIPs with catenin superfamily members were less robust than those with CK1δ/ε, Dvl, or Vangl family members.

A subset of Dact proteins weakly complexes with LEF/TCF proteins and with HDAC1

The Dact1 homologs from X. laevis and H. sapiens have been reported to form complexes with a subset of the LEF/TCF transcription factors that act as transcriptional regulators downstream of Wnt/β-catenin signaling and some other pathways [2, 7, 10]. We sought to replicate this finding and to test its specificity for Dact1 versus the other two Dact paralogs. Using the 293T cell line, we detected a positive coIP only for murine Dact2; this interaction was positive across all members of the LEF/TCF family examined (Figure 6A-D).

Another nuclear protein that has been reported to interact with DACT1 from H. sapiens is HDAC1 [2]. Using the HEK293T cell line and the murine Dact paralogs, we could replicate this finding for Dact1, but found that the coIP was stronger between Dact2 and HDAC1, whereas with Dact3 it was not detectable above background (Figure 6E, left). Because the previously published experiment was performed with human homologs in HEK293T cells, we replicated this for both the short and long isoforms of human DACT1 (Figure 6E, right).

All Dact proteins homo- and hetero-dimerize

Given numerous efforts by several independent groups to experimentally identify novel Dact interacting proteins, it is curious that no binding partner for one of the principal conserved Dact domains has been identified, specifically the leucine zipper region near the N-terminus. The leucine zipper is a well-defined structural motif that forms an amphipathic alpha helix or coiled-coil with a hydrophobic stripe along one side; this acts as a protein interaction or dimerization domain [38, 39]. Given the established ability of leucine zippers to mediate dimerization and the lack of a putative partner for this domain in Dact family members, we hypothesized that this conserved domain might mediate Dact homo- and/or hetero-dimer formation.

We tested this hypothesis using the same experimental strategy used above to assess other potential interactions: we co-expressed alternately tagged murine Dact paralogs in HEK293 or 293T cells and performed coIPs, pulling down complexes with one epitope tag and probing gel-separated precipitated protein complexes with the other. We found that all Dact paralogs form complexes with themselves and with other Dact paralogs (Figure 7A-C). In general coIPs involving Dact homo-interactions were moderately more strongly positive than hetero-interactions (Figure 7A-C, summarized in Figure 7D). Using two panels of Dact1 deletion constructs, one incorporating successive deletions at the N-terminus (Figure 7E) and the other incorporating successive deletions at the C-terminus (Figure 7F) we confirmed that the leucine zipper region of Dact1 is both necessary and sufficient for this association, consistent with leucine-zipper mediated dimerization (summarized in Figure 7G).

Overview

Our data indicate that the most robust interactions for all mouse Dact paralogs are with members of the Dvl and Vangl protein families; these interactions, along with interactions with several kinases, are conserved across all members of the Dact gene family. Somewhat surprisingly, the Dvl, Vangl, and Casein Kinase 1δ/ε (CK1δ/ε) proteins derived from the fruit fly Drosophila melanogaster, in which a Dact paralog has yet to be identified, also readily formed complexes with mammalian Dact paralogs. We also discovered that all Dact proteins can form complexes with themselves and with each other, and their conserved leucine zipper domains are necessary and sufficient for this interaction, suggesting dimerization. This has implications for functional cooperation between Dact family members, particularly in those tissues where the paralogs are co-expressed. It also raises the possibility that mutant or overexpressed Dact proteins could cause dominant effects by association and interference with wild type Dact proteins and their partners. Taken together, our biochemical findings suggest that all Dact family members participate in conserved kinase-regulated biochemistry involving Vangl and Dvl. This suggests a role within, or upstream of, PCP or a molecularly related pathway. It further suggests that some mutations in the human DACT loci could contribute to pathogenesis by disrupting this conserved pathway in a dominant or semi-dominant manner.

Functional Implications of Dact Phosphorylation

We suspect that the smaller sizes reported for Dact1 homologs in some studies and commercial antibody literature may variously represent poorly resolved size-markers, partial proteolysis products, and/or non-specific antibody cross-reactivity to more abundant cellular proteins. Dact proteins all clearly interact with several kinases, including not only CK1δ/ε and PKA, but also PKC and possibly other kinases as well. Phosphorylation and other post-translational modifications of Dact proteins may regulate function [11, 17]; this idea is certainly worthy of further empirical exploration not restricted to Wnt/β-catenin signaling, as that may not be the sole or even the primary physiological role for this protein family. For example, we and others have not yet tested whether Dact proteins can interact with or are modified by tyrosine kinases, some of which have recently been shown to play important roles in PCP signaling [40]. We note that at least one highly conserved peptide motif in the Dact proteins, located just C-terminal to the leucine zipper domain, contains not only two serines but also an invariant tyrosine in all family members [4].

Conserved binding partners suggest conserved function in a conserved pathway

It is also possible that the pathway involving Dact proteins in vertebrates is not synonymous with the PCP pathway in Drosophila. A divergent signaling pathway might regulate a catenin protein, such as p120ctn which was positive in our coIP assay with every Dact paralog (Table 1). The p120ctn protein plays a role at the plasma membrane in cytoskeletal and adhesive events [43], at the nucleus in gene transcription [10], and has recently been shown to interact with CK1ε and the Wnt receptor complex in Wnt/β-catenin signaling [44]. Given all this, a transient interaction with Dact proteins reflected by a comparatively weak coIP, but that regulates p120ctn localization or stability, could account for at least some conflicting observations of Dact function derived from different model systems. Alternatively, a more robust and specific functional interaction might exist between Dact proteins and an unidentified armadillo repeat containing protein, of which there are an abundance of candidates both within and without the catenin superfamily [37].

Based on the robustness of interactions between Dact, Dvl, and Vangl proteins, in those cells where these proteins are coexpressed they might be expected to form a stable or semi-stable complex. A logical future direction is to determine the subcellular localization of this putative complex and to identify other colocalized proteins. This will provide clues about whether Dact family members play a primary role in intercellular signaling, extracellular adhesion, cytoskeletal polarity, or perhaps in the protein trafficking that underlies one or more of these cell biological processes [8, 9, 12, 18, 45–47]. Indeed, given interactions documented here and elsewhere between Dact proteins and two widely divergent types of transmembrane protein [8, 22], as well as evidence that Dvl proteins play a role in endocytic regulation of transmembrane receptors [46, 48], a role for Dact proteins in transmembrane protein trafficking merits further investigation. The relatively stronger coIPs of Dact2 with Dvl3 and Alk5, and of Dact1 with Vangl2, support prior suggestions that there is some functional divergence between Dact paralogs [16, 22], but this should also be reconsidered in light of the new biochemical evidence presented here that Dact paralogs can physically interact. This suggests that Dact paralogs may functionally cooperate or compete in those cells where they are coexpressed.

Implications of Dact Dimer Formation

The discovery reported here that Dact paralogs can heterodimerize has implications for their physiological function. Although the mammalian Dact proteins do display distinct patterns of expression, there are many developing and mature tissues in which two or all three paralogs are co-expressed [4]. To the extent that coexpressed Dact proteins form active heterodimers they must functionally cooperate in these tissues. Despite some limited differences, our side-by-side comparison found conserved coIP interactions between every Dact paralog tested and the most robust partnering proteins. Taken together, the conserved coIP profiles and dimerization data suggest that Dact paralogs are likely to participate in shared biochemistry and have convergent physiological functions. If Dact paralogs do differ in endogenous activity, then in those cells where they are coexpressed they might mediate signaling pathway cross-talk and/or antagonism-either through non-productive heterodimer formation or through competition for common binding partners.

The discovery that Dact proteins dimerize also raises important issues for biochemical data interpretation. Immunoblotting and immunohistological data suggest that levels of endogenous Dact proteins are low even in those tissues where the mRNA is present and where knock-down or knock-out causes phenotypes [8, 18–21, 23]. In healthy tissues the levels of Dact proteins may be tightly regulated because, as self-associating scaffold proteins, if their levels are elevated they can aggregate with themselves, their partners, and with other more loosely associated proteins. In that case, non-physiological effects on biochemical pathways, including Wnt/β-catenin signaling, may occur in heterologous and in vitro assays in which these proteins are not maintained in their native cell biological context or concentrations. Indeed, functional studies in genetically engineered mice [8, 19, 20] so far do not support previous findings that Dact proteins play roles in Wnt/β-catenin signaling. Nevertheless, it remains possible that the lack of observed impacts on Wnt/β-catenin signaling in single-hit Dact mutant mice is due to redundancy between paralogs with respect to this pathway. This will be resolved once phenotypic and signal pathway consequences can be assessed in a mouse line in which all three Dact genes have been simultaneously eliminated. All that said, if Dact proteins are ultimately determined not to physiologically modulate Wnt/β-catenin signaling in healthy tissues, it will remain possible that they do influence this signaling pathway in cancerous and other diseased tissues where their levels or subcellular localization are dysregulated via mutation or epigenetic mechanisms [2, 15, 23, 49, 50].

Clinically, the discovery that their translation products homo- and hetero-dimerize raises the possibility that missense mutations in any of the three human DACT genes could cause genetically dominant or semi-dominant effects by interfering with functions of wild type homologs produced from unaffected alleles in the same individual. Given evidence that these proteins participate in a conserved biochemical pathway with demonstrated critical roles in urinary and lower gastrointestinal system development [8, 21], in neural differentiation and synaptogenesis [19], and in oncogenesis and metastasis [15, 23, 49], human genetic variants at these loci may have important clinical ramifications.

Cell culture, Transfections, and CoIPs

Performed as described [1, 51] with the following modifications. Two different protocols were employed depending on desired stringency. In cases where candidate interactors were not found to detectably coIP with Dact proteins in HEK293 cells (ATCC product number CRL-1573), the experiment was repeated in HEK293T/17 cells (ATCC product number CRL-11268); in some cases only the HEK293T/17 cell line and associated protocol was attempted. Where employed, the HEK293T/17 cell line and coIP protocol is specified in the text and figures as "HEK293T". In brief:

Phosphatase Treatment

Whole cell extracts from transfected cells in lysis buffer without phosphatase inhibitors were treated with lambda protein phosphatase (New England Biolabs, catalog # P0753) for 30 minutes at 30°C. Reactions were blocked by boiling in Laemmli protein gel loading buffer and resolved by SDS-PAGE.

Deglycosylation

Whole cell extracts from transfected cells in lysis buffer were treated with a protein deglycosylation mix (New England Biolabs, catalog # P6039) according to manufacturer's instructions. Reactions were blocked by boiling in Laemmli protein gel loading buffer and resolved by SDS-PAGE.

cDNAs and Expression Plasmids

The three murine Dact cDNAs employed in this study have been described previously [4]. The human short DACT1 isoform (GenBank NM_001079520.1) was obtained by RT-PCR from HEK293T cells, and the long DACT1 isoform (GenBank NM_016651.5) was synthesized from the shorter clone using overlapping PCR. The human GSK3α cDNA was obtained from Dr. Junichi Sadoshima (UMDNJ). All other cDNAs were obtained commercially from Open Biosystems (M. musculus clones), from the Bloomington Stock Center (D. melanogaster clones), or were generated in the Cheyette laboratory by RT-PCR from total mouse embryonic mRNA. For transfection and expression in cells, all Dact cDNAs were subcloned into vector p3XFLAG-CMV-10 (Sigma-Aldrich, catalog # E7658) whereas all putative interactor cDNAs were subcloned into vector pcDNA3.1(-) (Invitrogen, catalog # V795-20). The sequence of each cDNA employed was confirmed by Sanger sequencing.

Antibodies