Protein Encoded by the AxinFu Allele Effectively Down-regulates Wnt Signaling but Exerts a Dominant Negative Effect on c-Jun N-terminal Kinase Signaling*

Axin plays an architectural role in many important signaling pathways that control various aspects of development and tumorigenesis, including the Wnt, transforming growth factor-β, MAP kinase pathways, as well as p53 activation cascades. It is encoded by the mouse Fused (Fu) locus; the AxinFu allele is caused by insertion of an IAP transposon. AxinFu/Fu mice display varying phenotypes ranging from embryonic lethality to relatively normal adulthood with kinky tails. However, the protein product(s) has not been identified or characterized. In the present study, we conducted immunoprecipitation using brain extracts from the AxinFu mice with specific antibodies against different regions of Axin and found that a truncated Axin containing amino acids 1–596 (designated as AxinFu-NT) and the full-length complement of Axin (AxinWT) can both be generated from the AxinFu allele. When tested for functionality changes, AxinFu-NT was found to abolish Axin-mediated activation of JNK, which plays a critical role in dorsoventral patterning. Together with a proteomics approach, we found that AxinFu-NT contains a previously uncharacterized dimerization domain and can form a heterodimeric interaction with AxinWT. The AxinFu-NT/AxinWT is not conducive to JNK activation, providing a molecular explanation for the dominant negative effect of AxinFu-NT on JNK activation by wild-type Axin. Importantly, AxinFu-NT exhibits no difference in the inhibition of Wnt signaling compared with AxinWT as determined by reporter gene assays, interaction with key Wnt regulators, and expression of Wnt marker genes in zebrafish embryos, suggesting that altered JNK signaling contributes, at least in part, to the developmental defects seen in AxinFu mice.

Axin is a multifunctional protein that controls many important regulatory pathways, including ␤-catenin-mediated canonical Wnt signaling, transforming growth factor-␤, mitogen-activated protein (MAP) 3 kinase, and p53 pathways (1)(2)(3)(4)(5)(6)(7). In Wnt signaling, Axin provides a platform for the assembly of the ␤-catenin degradation complex to keep the activity of ␤-catenin in check, which otherwise would lead to developmental defects such as axis duplication and outgrowth of tissues or tumorigenesis. It also forms complexes with the upstream MAP kinases MEKK1 or MEKK4 to activate JNK and p38 MAP kinases (8,9). A remarkable aspect of Axin activation of MAPKs is that Axin has to form a homodimer or a higher order of complex formation (6,9). In addition, binding of GSK-3␤, casein kinases, or dishevelled can attenuate Axin activation of JNK by competing against MEKK binding or disrupting Axin homodimerization, suggesting that Axin activation of the MAPKs is highly conformation-dependent (10,11). We recently showed that Axin-mediated JNK activation plays a critical role in dorsoventral patterning, in that Axin exerts two opposing roles, one to down-regulate ␤-catenin and cause ventralization and the other to activate JNK leading to dorsalized embryos when overexpressed (12).
Axin is encoded by the Fused (Fu) locus (13,14). Mutant alleles of the Fused locus include Fu Fused (later termed as Axin Fu , Ref. 14), Fu Knobbly , Fu Kinky , and Fu Tg1 (13,(15)(16)(17)(18). The Fu Fused and Fu Knobbly alleles are each caused by insertion of intracisternal A-particle (IAP), one kind of transposons. While the IAP insertion occurs at intron 6 in the Fu Fused allele, IAP is inserted into exon 7 in Fu Knobbly . Fu Kinky , Fu Knobbly , and Fu Fused are dominant mutations, which are presumably caused by their truncated proteins resulting from the IAP insertion. Several studies have examined the nature of the mRNA transcripts from the Fu Fused allele. Vasicek et al. (19) showed that a 3.9-kb mRNA corresponding to wild-type Axin RNA could be detected in both wild-type and Fu Fused homozygous mice by Northern blotting, suggesting that the IAP insertion does not entirely prevent the generation and processing of the Fu Fused allele into a normal mature Axin transcript. However, transcripts that terminate immediately after exon 6 were also detected, suggesting the existence of other Axin transcripts that could potentially create a premature stop codon caused by the aberrant splicing between exons 6 and 7 (19,20). It is also believed that the inserted IAP may provide an internal promoter that may yield transcripts encoding a C-terminal portion of Axin.
Some Axin Fu/Fu mice die in utero, and others are viable with a kinked tail as well as deafness and defects in walking behavior (16,17,21,23). It is important to note that the phenotypic penetrance of the Fused allele is highly epigenetic, i.e. some Axin Fu/Fu homozygotes die prenatally, and some can develop into adulthood with kinked tails. In heterozygotes, some adult mice display kinky tails, and some do not. These variations in phenotypes among animals with the same genotype are likely a reflection of epigenetically regulated expression of the mutated Axin gene. It was reported that the presence or absence of the characteristic phenotype, a kinky tail, correlates with differential DNA methylation at the retrotransposon within Axin Fu or the overall levels of methylation of the Axin Fu gene (23). Consistently, Waterland et al. (24) reported that the supply of methyl donor to female mice before and during pregnancy resulted in increased DNA methylation at Axin Fu and thereby reduced half the incidence of tail kinking among Axin Fu/ϩ offspring. Nevertheless, the fact that some Axin Fu/Fu homozygotes die prenatally suggest that Axin Fu-NT can cause lethal defects during early development of mouse embryos.
However, the protein products from any of the Fused alleles have not been characterized. To understand the molecular nature of the naturally occurring mutant of Axin would shed new light on the multifunctionality and mechanisms thereby of Axin. In this study, we first generated a new polyclonal antibody against a C-terminal region of mouse Axin, in addition to the previously used C2b antibody raised against amino acids 348 -500 (5), to identify the protein products encoded by Fu Fused mRNA. As a result, we have identified a truncated protein of Axin that contains amino acids 1-596, which is designated as Axin Fu-NT , but were unable to detect any of the proposed truncate containing a C-terminal portion of Axin. We found that heterozygous and homozygous Axin Fu mice all express the fulllength complement of Axin (Axin WT ). When cotransfected, Axin Fu-NT abrogated JNK activation by wild-type Axin, indicating that Axin Fu-NT exerts a dominant negative effect on this aspect of Axin function. When Axin Fu-NT and wild-type Axin were co-expressed, homodimerization between wild-type Axin was disrupted by Axin Fu-NT , suggesting Axin Fu-NT possesses a domain that interacts with the wild-type Axin. GST pull-down of brain extracts from Axin Fu mice using GST-AxinNT400 that contains the N-terminal 400 amino acids, also showed that endogenous Axin can form a complex with the N-terminal Axin fragment. This provides a strong explanation as to how Axin Fu-NT abrogates Axin-mediated JNK activation. Surprisingly, Axin Fu-NT down-regulates Wnt signaling as effectively as Axin WT . When tested in zebrafish embryos, Axin Fu caused severe ventralized phenotypes that can be partially rescued by co-injection of JNK. Our current study has thus characterized the molecular nature of the Axin Fu mice, providing new insight into the structure and function of Axin.

EXPERIMENTAL PROCEDURES
Axin Fu Mice and Genotyping-Inbred 129P4/RrRk Axin Fu/ϩ mice were purchased from the Jackson Laboratory. Genomic DNA was isolated from the mouse ear. Mice were genotyped for the Axin Fu and wild-type alleles by multiplex PCR using the following primers: P23: 5Ј-cggagctattccgagaacg-3Ј, G245: 5Ј-gaccagagcccaagaaaaaccc-3Ј and IAP forward: 5Ј-gcgcatcactccctgattg-3Ј according to protocols from Dr. Frank Costantini (Columbia University).
Preparation of Mouse Brain Extract-For brain protein extract preparation, the mouse brain was homogenized in 2 ml of radioimmune precipitation assay buffer (5) containing 0.1% SDS with Polytron centrifuged at 1700 ϫ g for 30 s on ice. Samples were sonicated three times for 10 s each and centrifuged at 18,000 ϫ g rpm for 30 min at 4°C. The supernatant was collected, and the protein concentration was determined using the Bio-Rad Protein Array. Equal amounts of protein were subjected to immunoprecipitation with C2b or CT36 antibody raised against Axin.
Fish and Microinjection-Zebrafish embryos of the Tubingen strain were incubated in HoltfreterЈs solution at 28.5°C. Capped mRNAs were synthesized using T7 Cap Scribe (Roche Applied Science) according to the protocol described previously (26). Digoxigenin-UTP labeled antisense RNA probes were generated by in vitro transcription and used for whole mount in situ hybridization (27,28).
Cell Culture and Transient Transfection-HEK293T, HEK293 were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, antibiotics, and 2 mM L-glutamine, and maintained at 37°C, 5% CO 2 , in a humidified incubator. Transfection was performed using PEI. For transfection in 60-mm dishes; transfection mixture was prepared by adding DNA into 240 l of HBS (150 mM sodium chloride, 20 mM HEPES, pH 7.4), and 180 l of 10 M PEI (Cat. 23966, Polysciences, dissolved in HBS), immediately followed by mixing. The mixture was left at room temperature for about 30 min before dropping into cells. During the 30-min incubation, the cell medium was changed with 1.0 ml of DMEM (fetal bovine serum (FBS)-free). After 6 -12 h of transfection, the FBS-free DMEM medium was changed with complete DMEM medium.
Immunoprecipitation, Immunokinase Assay, and Western Blotting-Cells were lysed with lysis buffer (9) at 36-h posttransfection, and the lysates were immunoprecipitated with the indicated antibodies for about 3 h. Then beads were spun down and washed with lysis buffer three times. The proteins were eluted with 2ϫ SDS sample buffer and along with the total cell lysates were analyzed by Western blotting with different antibodies. The immunokinase assay on JNK activation was performed as previously described (9).
GST Pull-down Assay-GST-AxinNT400, which contains the Axin N-terminal-half (amino acids 1-400), was expressed in BL21 bacterial cells transformed with pGEX-AxinNT400 bacterial expression plasmid and was purified with glutathioneagarose beads (Sigma). Expression was induced with 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside for 4 h at 37°C. Approximately 0.2 mg of the GST fusion protein bound to agarose beads was added to each brain extract from adult Axin ϩ/ϩ and Axin Fu/Fu mice, and incubated for 3 h with gentle rotation, followed by washing three times with lysis buffer. The proteins were eluted with SDS sample buffer.
Transcription Reporter Assay-For the LEF-1 reporter assay, 0.5 g of LEF-1 reporter gene and 0.1 g of LEF-1 were cotransfected with HA-Axin or HA-Axin Fu-NT or both of them into HEK293T cells. Wnt1 (0.1 g) was also transfected as indicated. At 30-h post-transfection, cells were lysed, divided into two portions, and each measured for the activities of luciferase and ␤-galactosidase. The ratio of luciferase activity to ␤-galactosidase activity varied less than 10% among samples. Data are presented as means plus standard deviation from three separate experiments performed in duplicate.

Detection of a Truncated N-terminal Fragment of Axin in Axin Fu
Mice-In some of the heterozygous (Axin Fu/ϩ ) mice (36/134), the tails appeared normal, whereas all of the homozygotes showed kinked tails with some of the kinked tails also having bifurcation at the distal end. These observations are consistent with previous reports showing that, Axin Fu/ϩ and Axin Fu/Fu adult mice display variable phenotypes characterized by kinked tails (13,16,17). Based on previous work on the IAP insertion position and the detection of normal mature Axin mRNA transcript as well as aberrant transcripts, we outlined putative protein products resulting from the IAP insertion in the Axin Fu allele (Fig. 1A). Normal mature Axin mRNA must be produced by splicing out the IAP sequence between the donor sequence of exon 6 and the acceptor site in exon 7. It was also speculated that the IAP-disrupted intron 6 may not be properly spliced out, and as a consequence the translation is extended into the intron sequence and prematurely terminated immediately after exon 6 junction as diagrammed in Fig. 1A. In addition, it was hypothesized that the long terminal repeat sequence in the transposon may serve as a promoter to drive transcription of the Axin sequence downstream of intron 6 to the 3Ј terminus, resulting in a potential truncated protein containing amino acids 597-832 of Axin (19,23). However, none of the putative Axin Fu protein species has been identified or characterized.
We set out to identify the naturally occurring Axin truncates in the Axin Fu mice. For convenience of characterization of the Axin Fu truncates, we first generated cDNA fragments corresponding to the predicted Axin Fu-NT and Axin Fu-CT respectively, as diagrammed in Fig. 1A. In parallel, we generated an additional polyclonal antibody against a C-terminal region (amino acids 676 -832) of Axin, which was designated as Axin antibody CT36. The antibody against the N-terminal Axin (antibody C2b) had been previously reported (5). The C2b and CT36 antibodies specifically react with Axin Fu-NT and Axin Fu-CT that were expressed in HEK293T cells (Fig. 1B). C2b recognized HA-Axin and HA-Axin Fu-NT , but not Axin Fu-CT ; conversely, CT36 only reacted with the full-length Axin and HA-Axin Fu-CT , but not Axin Fu-NT . We carried out immunoprecipitation experiments with total brain protein extracts from adult Axin ϩ/ϩ , Axin Fu/ϩ (with normal tail appearance), Axin Fu/ϩ (kinked tail), and Axin Fu/Fu mice. After immunoprecipitation with C2b and CT36 antibodies, the precipitated protein samples were separated on PAGE gels along with total lysates of HEK293T cells expressing untagged Axin WT (full-length, the first lanes, Fig. 2A), Axin Fu-NT and Axin Fu-CT (the last two lanes, on the right of Fig.  2A), which were constructed as diagrammed in Fig. 1A and were used as molecular weight references. The separated proteins were then subjected to immunoblotting with two commercial antibodies from Santa Cruz Biotechnology, S-20 (against the N-terminal region) and R-20 (against the C-termi-nal region). The antibodies are indicated on the left of each panel ( Fig. 2A). In the wild-type mice, a protein with a molecular mass of ϳ110 kDa that corresponds exactly to the control full-length Axin protein (lane 1 of Fig. 2A) expressed in HEK293T cells was detected in immunoprecipitates of C2b. In Axin Fu mice, heterozygous or homozygous, a second specific protein species that co-migrated with the constructed Axin Fu-NT was also detected by S-20 that was raised against the N-terminal region of Axin. These results showed that the proposed Axin Fu-NT indeed exists in mice containing the Axin Fu allele.
We also carried out immunoprecipitations with the CT36 antibody, followed by Western blotting with the different antibodies indicated. In the immunoprecipitates, only the fulllength Axin protein was specifically detected by all the antibodies; no specific Axin species corresponding to the putative Axin Fu-CT was detected ( Fig. 2A). However, it remains possible that the failure to detect the Axin Fu-CT is due to the high background of the Western blot. Protein levels of tubulin in the mouse brain extracts were detected by immunoblotting and served as an internal loading control.
Axin Fu-NT Effectively Attenuates Wnt Signaling-To characterize the functional properties of the Axin Fu-NT protein, we first tested whether it affects Wnt signaling. The Axin Fu-NT and TOPFLASH reporter were co-transfected with or without Wnt1, into HEK293T cells, and the luciferase activities at 30-h post-transfection were measured. As shown in Fig. 3B, Axin Fu-NT attenuated Wnt signaling almost as effectively as the wild-type Axin. Similar results were obtained using a similar reporter, the LEF-1 luciferase plasmid (Fig. 3A). Consistently, the Axin Fu-NT protein interacted with GSK-3␤, ␤-catenin, and APC with similar affinities compared with wild-type Axin (Fig.  3, C-E). We also examined the interaction between Axin Fu-NT and CKI␣, and the results demonstrated that Axin Fu-NT also effectively interacts with CKI␣, albeit with slightly lesser affinity (Fig. 3F). These results indicate that Axin Fu-NT is fully capable of forming the degradation complex and promoting the GSK3␤-mediated, phosphorylation-dependent, ␤-catenin degradation. When injected into zebrafish embryos, Axin Fu-NT could down-regulate the expression of dharma/boz that is a specific maternal ␤-catenin target (30,31), as effectively as Axin WT (Fig. 3G, upper panel). At the shield stage, the zygotic ␤-catenin target tbx6 (32) in zebrafish was also impaired by Axin Fu-NT to the same extent as Axin WT (Fig. 3G, lower panel). The statistics of the phenotypic changes in those differently injected embryos is shown in Fig. 3H.
Identification of an N Terminus Proximal Dimerization Domain of Axin-The Axin Fu allele is a gain-of-function mutation. We therefore tested whether Axin Fu-NT can form a dimeric interaction with the wild-type Axin. We performed a co-immunoprecipitation assay in HEK293T cells transfected with HA-tagged wild-type Axin and Myc-tagged Axin Fu-NT . When co-expressed, HA-Axin was detected in the immunoprecipitate by anti-Myc (Fig. 4A), suggesting that Axin Fu-NT forms a heterodimer with wild-type Axin. To verify that, we also transfected differentially tagged Axin Fu-NT into HEK293T and carried out immunoprecipitation. HA-Axin Fu-NT was detected in the immunoprecipitate by anti-Myc antibody (Fig. 4B).

FIGURE 2. Identification of Axin mutant protein(s) in the Axin Fu mouse brain extract.
A, brain extracts from wild-type, Axin Fu/ϩ (normal tail appearance), Axin Fu/ϩ (kinked tail), and Axin Fu/Fu mice were prepared as described under "Experimental Procedures." Then equal amounts of the extracts were subjected to immunoprecipitation with C2b, CT36, or control IgG antibodies. Wild-type Axin and Axin mutant transcripts were detected by anti-Axin antibodies against the N-terminal (S-20) or the C-terminal (R-20) part of Axin as indicated. Meanwhile, lysates from HEK293T cells overexpressing untagged wild-type Axin, Axin Fu-NT , and Axin Fu-CT were analyzed as molecular weight references. Wild-type Axin could be detected in all kinds of Axin Fu mouse brain extract (marked with an asterisk). However, a second specific protein species corresponding to Axin Fu-NT was detected by S-20 (marked with an asterisk) in Axin Fu heterozygous and homozygous brain extracts, which were immunoprecipitated with C2b antibody. Tubulin of the brain extracts are shown here as loading controls. B, lowest panel shows the genotype of the mice; the 352-bp band was generated by the Fused allele, and the 248-bp band was generated by the wild-type allele.
These results indicate that Axin Fu-NT could strongly interact with wild-type Axin, unexpectedly revealing a previously unidentified dimerization domain in the N-terminal region of Axin. To further confirm that the N-terminal of Axin could form a homodimer, we carried out a GST pull-down assay with GST-AxinNT400 corresponding to the N-terminal-half (amino acids 1-400) and the brain extracts separately from adult Axin ϩ/ϩ and Axin Fu/Fu mice. Consistently, pull-down with GST-AxinNT400 and the brain extract of Axin Fu/Fu mice produced two Axin protein species corresponding to the full-length wild-type Axin and Axin Fu-NT , but only the wild-type Axin was detected from the pull-down using the extract of Axin ϩ/ϩ mice (Fig.  4C). The Axin proteins were not seen in the pull-down using GST alone.
Dominant Negative Effect of Axin Fu-NT on Axin-induced JNK Activation-Next, we tested if Axin Fu-NT has an effect on JNK activation. Whereas wild-type Axin robustly activated JNK in HEK293T cells, Axin Fu-NT failed to activate JNK, in agreement with our observations that Axin strictly requires its C-terminal regions for JNK activation (11). When Axin and Axin Fu-NT were co-expressed, JNK activation was almost completely abolished (Fig.  5A), indicating that Axin Fu-NT has a dominant negative effect on JNK activation by the wild-type Axin. One possible means for Axin Fu-NT to exert its inhibitory effect is to sequester MEKK1. Indeed, when assayed for its ability to bind MEKK1 by immunoprecipitation experiments, Axin Fu-NT was found to strongly interact with MEKK1 (Fig. 5B). We previously found that Axin can utilize MEKK4, in addition to MEKK1, to activate JNK (8,9). We therefore tested whether Axin Fu-NT interacts with MEKK4, and whether overexpressed Axin Fu-NT sequesters MEKK1 and MEKK4 against their binding to the wild-type Axin.
In the presence of increasing amounts of Axin Fu-NT , levels of Axin Fu-NT co-precipitated with MEKK1-CT or MEKK4-CT were gradually increased, whereas lesser wild-type Axin was co-immunoprecipitated with MEKK1-CT or MEKK4-CT (Fig. 5, C and E). These results indicated that Axin Fu-NT indeed prevents MEKK1 or MEKK4 from binding to the Axin WT , but that MEKK4 forms a physical interaction with Axin Fu-NT . When Axin Fu-NT was co-expressed with MEKK4-CT, we detected a strong direct interaction of MEKK4-CT with Axin Fu-NT (Fig. 5D). This new observation suggested that the previous Axin deletion mutants that failed to interact with MEKK4 (8) might not actually lose the domain for MEKK4 interaction, but rather that the domains deleted comprised conformations that disallow MEKK4 binding. The LEF-1 reporter was co-transfected with HA-Axin and HA-Axin Fu-NT into HEK293T cells in the presence or absence of Wnt1 as indicated. At 30-h post-transfection, cells were harvested, and the luciferase activity was measured as described previously. Axin Fu-NT could inhibit LEF-1 transcriptional activity as effectively as Axin. B, similar experiments were performed with the TOPFLASH reporter. C-F, Axin Fu-NT interacts with GSK3␤ (C), ␤-catenin (D), APC (E), and CKI␣ (F). HEK293T cells were transfected as indicated and incubated for 36 h. Cell lysates were subjected to immunoprecipitation with anti-Myc antibody and followed by Western blot with the antibodies indicated. Axin Fu-NT binds to GSK3␤, APC, and ␤-catenin as efficiently as wild-type Axin, whereas it shows less affinity for the interaction with CKI␣. G, Axin Fu-NT inhibits the expression of maternal and zygoticspecific ␤-catenin target genes in zebrafish. The upper panel is the animal pole views of maternal ␤-catenin target dharma/boz expression at 30% epiboly stage in GFP control mRNA, Axin WT mRNA, and Axin Fu-NT mRNA, respectively. The lower panel is the animal pole views of the expressions of zygotic ␤-catenin target tbx6 at shield stage in GFP control mRNA, Axin WT mRNA, and Axin Fu-NT mRNA, respectively. H, statistical data for Axin Fu-NT -down-regulated ␤-catenin targets dharma/boz and tbx6 in G. Axin Fu-NT could down-regulate maternal ␤-catenin target dharma/boz as effectively as wild-type Axin in zebrafish embryos at 30% epiboly stage (left panel). Axin Fu-NT could down-regulate zygotic ␤-catenin target tbx6 as effectively as wild-type Axin in zebrafish embryos at the shield stage (right panel). Each number (n) represents total embryos from two independent experiments, and 600 pg of each mRNA were injected for each embryo.

Axin Fu-NT Causes Ventralized Phenotypes in Zebrafish
Embryos-Our recent study indicated that Axin, besides ventralizing activity by facilitating ␤-catenin degradation, displays a dorsalizing activity that is mediated by Axin-induced JNK activation (12). To understand the biological function of Axin Fu-NT in vivo, we injected Axin Fu-NT mRNA into zebrafish embryos and observed the effect of its ectopic expression on embryonic development. In embryos each injected with 400 pg of Axin Fu-NT mRNA, 68.2% of all embryos (n ϭ 44) at middle gastrulation (shield stage) displayed reduced expression of the dorsal mesoderm marker goosecoid (gsc), while embryos injected with the control GFP mRNA expressed goosecoid properly on the dorsal side (Fig. 6A). When injected alone, 400 pg of ␤-catenin induced an expanded expression of the dorsal marker gsc, co-injection of Axin Fu-NT mRNA (400 pg) strongly abrogated the dorsalizing effect of ␤-catenin. Similarly, JNK mRNA injection (300 pg) induced dorsalization, which was also reversed by Axin Fu-NT mRNA. However, in the embryos coinjected with JNK and Axin Fu-NT , it is evident that the dorsal marker gsc was expressed at much higher levels than those in embryos injected with Axin Fu-NT alone. These results indicate that Axin Fu-NT could effectively attenuate ␤-catenin signaling and that JNK could on its own cause an expanded expression of the dorsal marker, which is consistent with our previous observations (12,33,34).
In contrast, in embryos injected with Axin Fu-NT mRNA, expression of the ventral margin-specific gene eve1 at the shield stage was increased, expanding to the whole blastomere margin in 88.9% (Fig. 6B). Injection of ␤-catenin or JNK mRNA reduced the expression of the ventral marker, which was reversed by co-injection of Axin Fu-NT mRNA (Fig. 6B). Of particular note, Axin Fu-NT mRNA induced expanded expression of this marker around the whole blastomere can be rescued only by JNK, conforming to the notion that JNK and ␤-catenin are two independent inducers for dorsalization. Based on the zebrafish experiments, JNK assays, as well as Wnt reporter gene assays, we summarized the functions of Axin Fu-NT in Fig. 6C, showing that Axin Fu-NT can exert a dual negative effect on dorsalization, one aspect of which is to effectively down-regulate ␤-catenin function and the other to dominant negatively attenuate JNK signaling mediated by the wild-type Axin.

DISCUSSION
In the present study, we have identified the protein product of the Axin Fu/Fu allele, Axin Fu-NT , and have characterized the Similar experiments were performed as described in A. C, GST pull-down with GST-AxinNT400. Bacterially expressed GST control or GST-AxinNT400 was immobilized on glutathione-Sepharose beads and incubated separately with the brain extract from Axin ϩ/ϩ and Axin Fu/Fu mice for 3 h. The ability of GST-AxinNT400 to pull-down Axin WT and Axin Fu-NT was analyzed by Western blot with anti-Axin (C2b) antibody after SDS-PAGE. Two protein species, corresponding to the wild-type Axin and Axin Fu-NT , were pulled down from the mutant mice by GST-AxinNT400, whereas, as expected, only the protein species corresponding to the wild-type Axin was detected in the pull-down from the wild-type mice. Input represents one-eighth of the brain extract used for GST pull-down. molecular nature of the truncated Axin. Consistent with the previous observation that the Fused allele can give rise to a normal 3.9-kb mRNA for wild-type Axin, in addition to the aberrant fused-specific mRNA species, we also detected Axin of normal size in both heterozygous and homozygous Axin Fu mice. We showed that Axin Fu-NT is virtually intact in the inhibition of Wnt signaling as determined by LEF reporter assay and based on biochemical properties of Axin Fu-NT including its ability to bind to ␤-catenin, GSK3␤, APC, and CKI␣. Moreover, zebrafish embryos injected with Axin Fu-NT mRNA displayed defects characteristic of impaired Wnt signaling, including reduced expression of the dorsal marker goosecoid, and increased expression of the ventral marker eve1. Moreover, the expressions of dharma/boz and tbx6, specific target genes for ␤-catenin, in the Axin Fu-NT mRNA-injected embryos were also found decreased to the same degree seen in embryos injected with Axin WT mRNA. These observations are consistent with previous studies using various Axin deletion mutants (35). However, it is difficult to reconcile with the report that the most notable phenotypic change in Axin Fu/Fu embryos is the formation of an ectopic axis (14). We therefore cannot rule out the possibility that Axin Fu-NT may in some way inadequately regulate Wnt signaling; in particular, the missing C-terminal region is critical for interaction with regulatory factors such as Dishevelled (36,37). On the other hand, as Axin has been shown to play pleiotropic roles in signaling pathways that control various aspects of development, we speculate that the defects seen in the Axin Fu/Fu mice may be caused by perturbation of more than one signaling pathway. Another possibility is that in mice, JNK signaling may exert a negative role in dorsal axis formation as reported in Xenopus (38), such that the deficiency of Axin-mediated JNK signaling in mutant mice may lead to axis duplication.
We found that Axin Fu-NT drastically inhibits JNK activation, which is in accordance with our previous studies showing that an intact C-terminal of Axin is required for JNK activation. Mechanistically, Axin Fu-NT was unexpectedly found to possess an additional dimerization domain, through which it forms a dimeric complex with the wild-type Axin, rendering the existing wild-type Axin unable to activate JNK. Moreover, Axin Fu-NT sequesters MEKK1 and MEKK4 to prevent their interaction with wild-type Axin, thereby exerting a dominant negative effect on Axin-mediated JNK activation. JNK signaling has been implicated in numerous biological processes ranging from stress responses to the formation of planar cell polarity during development (39 -41). Compound mutant mice lacking JNK1 and JNK2 genes die on embryonic day 10.5 (E10.5) because of defective closure of the neural tube in the hindbrain (41)(42)(43). Overactivity of JNK signaling induced by ectopic expression of Axin in zebrafish embryos can lead to enlarged head and other dorsalized phenotypes, demonstrating that the Axin-mediated JNK signaling plays an important role in dorsoventral patterning including head formation (12). These findings emphasize the importance of JNK signaling during early development. It is therefore reasonable for us to speculate that the developmental abnormalities seen in homozygotes of the Axin Fu mice, including neurological defects and embryonic lethality, may be caused by improper Axin-mediated JNK signaling, at least to a certain extent. It is important to note that the dominant negative nature of Axin Fu-NT in JNK signaling, through forming Axin Fu-NT /Axin WT , conforms to the Axin Fu allele being a gain-of-function mutation.
The ventralized phenotypes observed in zebrafish embryos overexpressing Axin Fu-NT are therefore likely caused by a dual ventralizing effect. Axin Fu-NT induces ventralization not only through effectively down-regulating ␤-catenin signaling as assayed using two different Wnt reporters, but also via abrogation of Axin-mediated JNK activation. Taken together, we have identified the protein product of the Axin Fu allele that carries the ability to interact with wild-type Axin, and the ability to interact with MEKK1 and MEKK4. These characteristics of Axin Fu-NT render it unlikely for wild-type Axin to form a homodimer with bound MEKK that is required for JNK activation. Our current study has thus emphasized the importance of Axin/JNK signaling in development and shed new light on the structure and function of Axin that is a master scaffold for many important signaling pathways.