Metastasis-associated kinase modulates Wnt signaling to regulate brain patterning and morphogenesis

Vertebrate embryonic development proceeds through a cascade of inductive interactions, which specify major body axes and generate the three germ layers, ectoderm, mesoderm and endoderm, in the embryo. The Wnt pathway is among the few major signaling pathways that specify cell fates during development and is implicated in cell and tissue polarization(Logan and Nusse, 2004). Before gastrulation, the pathway is used to establish the Spemann organizer, a major inductive center in the embryo, which is responsible for dorsal development and neural induction (Harland and Gerhart, 1997; Moon and Kimelman, 1998; Sokol,1999; Tao et al.,2005). After the initial induction of the central nervous system,localized sources of Wnt ligands induce and maintain a secondary organizing center at the midbrain-hindbrain boundary, which mediates regional patterning of the brain (Chi et al.,2003; Hidalgo-Sanchez et al.,2005; Liu and Joyner,2001). Besides organizer specification and brain patterning, Wnt signaling is involved in somitogenesis, eye and neural crest development,cardiogenesis, and endoderm differentiation(Bainter et al., 2001; Cavodeassi et al., 2005; Ciani and Salinas, 2005; Gamse and Sive, 2000; Logan and Nusse, 2004; Wu et al., 2003). Thus, Wnt signaling is re-used at many developmental stages in different embryonic tissues.

The pathway leads to two major outcomes: β-catenin-dependent(canonical) activation of target genes and the regulation of actin cytoskeleton and cell polarity via a poorly understood process (noncanonical signaling). In the canonical pathway, Wnt ligands and their cognate Frizzled receptors signal through Dishevelled to stabilize β-catenin and induce its association with the TCF transcription factors to activate Wnt target genes. By contrast, Wnt signaling to the cytoskeleton occurs through the activation of small GTPases (Habas et al.,2003; Habas et al.,2001), Rho-associated kinases(Marlow et al., 2002; Winter et al., 2001),Jun-N-terminal kinases (Boutros et al.,1998; Lisovsky et al.,2002) and intracellular Ca²⁺ release(Sheldahl et al., 2003; Slusarski et al., 1997). Noncanonical Wnt signaling regulates changes in cell shape and motility during gastrulation, neurulation and organogenesis(Ciani and Salinas, 2005; Saneyoshi et al., 2002; Sokol, 1996; Wallingford et al., 2002; Winklbauer et al., 2001; Zohn et al., 2003). Both canonical and noncanonical branches of pathway involve Dishevelled, a protein phosphorylated in response to Wnt signaling(Yanagawa et al., 1995),although how Dishevelled directs signals to different molecular targets remains unclear (Wallingford and Habas,2005).

Here, we report that a SNF1 (sucrose non fermenting 1)-related protein kinase, known as metastasis-associated kinase (MAK)(Gardner et al., 2000b; Korobko et al., 1997),phosphorylates Dishevelled, inhibits the canonical Wnt pathway and upregulates non-canonical Wnt signaling during early development. SNF1-related kinases are involved in the metabolic response to nutritional and environmental stress,cell cycle, cell polarity and vertebrate development(Becker and Brendel, 1996; Hardie et al., 1998; Ruiz et al., 1994). MAK is distantly related to the yeast Kin-1 and to the C. elegans PAR-1. Kin-1 is crucial for growth polarity and cytoskeletal organization in fission yeast (Drewes and Nurse, 2003; Levin et al., 1987; Tassan and Le Goff, 2004). PAR-1 regulates cell polarity in C. elegans and Drosophilaembryos and mammalian cells (Bohm et al.,1997; Pellettieri and Seydoux,2002; Shulman et al.,2000; Tomancak et al.,2000) and has been implicated in Wnt signaling(Ossipova et al., 2005; Sun et al., 2001). Our experiments demonstrate that MAK regulates morphogenetic movements, eye development and midbrain patterning in Xenopus embryos and suggest that MAK may function as a molecular switch between the canonical and noncanonical Wnt pathways.

A plasmid containing a full length Xenopus MAK cDNA corresponds to the IMAGE clone 5156600 (Accession Number CA793236), obtained from Open Biosystems. Two different, but highly related MAK genes have been reported in Xenopus databases(Ruzov et al., 2004). The expression pattern of the second gene (Accession Number AY318878)(Ruzov et al., 2004) differs from what we report here for xMAK. To generate pXT7-GFP-xMAK, xMAK cDNA was amplified by PCR with 5′-AAGAATTCGATGCCGGCTGCCGCTG-3′ and 5′-GCCTCGAGGCAAGGACAAATCTGTT-3′, and subcloned into pXT7-GFP(Itoh et al., 2000) that has been digested with EcoRI and SalI. Flag-xMAK constructs were generated by PCR and subcloned into pCS2-Flag. Untagged mMAK cDNA in the pCTX vector (S.Y.S., unpublished) was used as a transcription template. Myc-MAK constructs were generated from the wild-type or kinase-dead (K91>R) mMAK cDNA (Korobko et al., 2004) by PCR with the primers 5′-AACTCGAGATGCCGGCAGCGG-3′ and 5′-AGGTTAACACTGGCCCTTGACACCGTC-3′ and subcloned into the XhoI and EcoRV sites of pCTX-Myc. Details of cloning are available upon request. Other expression constructs were pCS2-nβgal(Turner and Weintraub, 1994),Xfz3 (Shi et al., 1998), rFz2,membrane-tethered DsRed (Unterseher et al., 2004), Myc-Xdsh and β-catenin(Sokol, 1996), HA-Xdsh(Itoh et al., 2000), Xwnt8(Sokol et al., 1991) andΔN-Fz8 (Lisovsky et al.,2002).

In vitro fertilization, culture and microinjections of Xenopuseggs were essentially as described (Sokol,1996). Stages were determined according to Nieuwkoop and Faber(Nieuwkoop et al., 1967). Morpholino oligonucleotides were from GeneTools (Oregon) and had the following sequences: MAK MO, 5′-CGGCATCCCCAGTGGTGTAGATCTC-3′; β-Cat MO,5′-TTTCAACCGTTTCCAAAGAACCAGG-3′; control MO,5′-ATCGACTTCCTCCGAAACGGACATG-3′. RNAs for microinjection were synthesized using mMessage mMachine kit (Ambion). Axis induction assays were carried out by injecting HA-Xdsh, Xwnt8 or β-catenin RNA into a single vegetal ventral blastomere at the four- to eight-cell stage at indicated doses and assessed when the injected embryos reached stage 36-40. To monitor axis extension defects, MAK RNA, MAK-KD or GFP RNA were injected into two dorsovegetal blastomeres of four-cell embryos (2 ng each injection) and the injected embryos were allowed to develop until sibling embryos reached stage 32. In other cases, sites of injection are specified in corresponding figure legends. For lineage tracing, 20-40 pg of RNA encoding nuclearβ-galactosidase (pCS2-nβgal) was injected together with morpholinos or mRNAs. β-Galactosidase activity was visualized with the Red-Gal substrate (Research Organics).

For animal cap elongation assay, four-cell embryos were injected four times in the animal region of each blastomere with mMAK or mMAK-KD RNA (2 ng per injection), animal caps were dissected at stage 8 to 9, and cultured with or without 50 ng/ml of human recombinant activin A until the sibling embryos reached stage 15.

Embryos were injected into animal pole region of one ventral blastomere at the four-cell stage with 20 pg of the -833pSia-Luc reporter DNA(Fan et al., 1998) per embryo,alone or with the indicated amounts of mRNAs. Lysates were prepared from embryos at stage 10.5 and assayed for luciferase activity as previously described (Fan et al., 1998). For every experimental group, measurements were carried out for triplicate samples, each consisting of five embryos. Values shown are averages±s.d., which are representative of at least three different experiments.

For subcellular localization of MAK-GFP constructs, RNAs encoding MAK-GFP(500 pg) and mDsRed (2 ng) were co-injected with or without Fz3 or Fz2 RNA (1 ng) into the animal pole region of two- to four-cell embryos. Animal cap explants were dissected at stage 9-9.5, fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS) for 35-45 minutes, washed and mounted for observation in the Vectashield mounting medium with DAPI (Vector) as described(Itoh et al., 2005). Fluorescence was visualized on a Zeiss Axiophot microscope with Apotome attachment and images were acquired with Axiocam HR camera.

Digoxigenin-labeled antisense RNA probes were synthesized from linearized plasmids, encoding En2 (Brivanlou and Harland, 1989), Krox20(Bradley et al., 1993), Otx2 (Pannese et al.,1995), Gbx2 (von Bubnoff et al., 1996) and xMAK, using a digoxigenin-labeling mixture (Boehringer Mannheim). Whole-mount in situ hybridization was carried out according to Harlan(Harland, 1991) with modifications as described previously(Hikasa and Sokol, 2004). For in situ hybridization, embryos were rehydrated in 1×PBS, 0.1% Tween 20. The staining reaction was carried out for sense and anti-sense probes for the same duration. 5-Bromo-4-chloro-3-indolyl phosphate (Sigma) and Nitro blue tetrazolium (Sigma) were used for chromogenic reactions. After staining, some embryos were dehydrated, embedded into paraplast and sectioned at 10 μm for image analysis at higher resolution.

Total RNA was extracted from embryos or animal caps by proteinase Kphenol extraction as described (Itoh and Sokol,1997). cDNAs were made from DNase-treated RNA using the Superscript first strand synthesis system (Invitrogen). RT-PCR was performed on total RNA isolated from stage 10.5 embryos as previously described(Itoh and Sokol, 1997). Primers for RTPCR were: xMAK, 5′-ACCAGAAGATGGTCGA-3′,5′-TTCCAACTGATGAAACT-3′; chordin,5′-AACTGCCAGGACTGGATGGT-3′,5′-GGCAGGATTTAGAGTTGCTTC-3′; Xnr3,5′-CGAGTGCAAGAAGGTGGACA-3′,5′-ATCTTCATGGGGACACAGGA-3′; siamois,5′-CTCCAGCCACCAGTACCAGATC-3′,5′-GGGGAGAGTGGAAAGTGGTTG-3′; gsc,5′-TTCACCGATGAACAACTGGA-3′,5′-TTCCACTTTTGGGCATTTTC-3′; MyoD,5′-AGCTCCAACTGCTCCGACGGCATGAA-3′,5′-AGGAGAGAATCCAGTTGATGGAAACA-3′; Dkk1,5′-CACCAAGCACAGGAGGAA-3′, 5′-TCAGGGAAGACCAGAGCA-3′; EF-1α, 5′-CAGATTGGTGCTGGATATGC-3′,5′-ACTGCCTTGATGACTCCTAG-3′; FGFR,5′-TTGAAGTCTGATGCGAGTGA-3′,5′-GGGTTGTAGCAGTACTCCAT-3′. One quarter of each PCR reaction was electrophoresed in a 5% polyacrylamide gel, stained with ethidium bromide and photographed under ultraviolet light.

Immunoprecipitation and western analyses were carried out with embryo lysates as described (Gloy et al.,2002). To prepare embryo lysates at stage 10+, each blastomere of four-cell embryos was injected with different mRNAs. For immunoprecipitation,20 μl of 9E10 (anti-Myc) or 12CA5 (anti-HA) hybridoma supernatants (ATCC)were used per sample. Protein amount equivalent to one half embryo was loaded per lane for embryo lysates, and the equivalent of four to nine embryos for immunoprecipitated proteins. Monoclonal M2 antibodies (anti-Flag) and polyclonal anti-β-tubulin antibodies were from Sigma, secondary HRP-conjugated antibodies were from Jackson ImmunoResearch.

In vitro translation reactions were performed using rabbit reticulocyte lysates from the Retic Lysate IVT kit (Ambion) or TnT system (Promega)according to manufacturer's instructions. MAK kinase reactions were performed essentially as described (Ossipova et al.,2005). For in vitro blocking studies, 600 pM of MAK MO or COMO were incubated with 0.3 μg of xMAK RNA or mMAK RNA in 10 μl of distilled water for 30 minutes at room temperature, following by in vitro translation reaction for 90 minutes. Half of each reaction was loaded on an SDS-polyacrylamide gel. Radiolabeled lysates were electrophoresed as described(Sambrook et al., 1989), gels were fixed in 50% methanol and 10% acetic acid, dried and exposed to Kodak XAR-5 film for autoradiography.

In vitro JNK kinase assays were performed essentially as described(Lisovsky et al., 2002). Four-cell stage embryos were injected into each cell with GFP, mMAK, or mMAK-KD RNA at 1.5 ng per injection, or ΔN-Fz8 RNA (0.5 ng). Embryos were lysed at stage 14 in 100 μl of buffer containing 40 mM HEPES (pH 7.5),50 mM KCl, 5 mM EDTA, 5 mM EGTA, 50 mM β-glycerophosphate, 2 mM DTT, 1 mM sodium vanadate, 50 mM sodium fluoride, 1% Triton X-100, 10 M PMSF, 10μg/ml leupeptin, 1 μg/ml aprotinin, 1 μg/ml pepstatin and 1 μg/ml antipain. The reaction mixture (25 μl) contained 1 μl of total embryo lysate, 0.5 μg of GST-Jun(1-135) and the kinase buffer (20 mM HEPES, pH 7.5, 10 mM MgCl₂, 1 mM sodium vanadate, 2 mM DTT, 25 mMβ-glycerophosphate, 100 M cold ATP). The reactions were allowed to proceed for 30 minutes at 30°C. Proteins were separated by 12% SDS-PAGE,phosphorylated Jun-GST was detected with anti-phospho-Jun-specific antibodies(Cell Signaling Technology).

To gain insight into possible developmental functions of MAK in vertebrate embryos, we first analyzed the expression of a Xenopus homologue of MAK (xMAK) during embryonic development. Reverse transcription-based polymerase chain reaction (RT-PCR) has been carried out using total RNA isolated from different developmental stages(Fig. 1A). No maternal expression of xMAK was detected in eggs and early embryos. Zygotic xMAK transcripts appeared after the onset of zygotic transcription at the midblastula transition, and remained expressed throughout later development,with the highest levels during neurulation (stages 13-19, Fig. 1A).

We next studied the spatial distribution of xMAK transcripts by whole-mount in situ hybridization (Fig. 1B-M). Consistent with the RT-PCR data, no signal was detected in stage 7 embryos. At stage 10, xMAK expression was predominantly in prospective ectoderm and involuting mesoderm, although the signal may have been weaker in the vegetal region owing to poor probe penetration(Fig. 1C). At neurulation, xMAK transcripts were present in the deep (sensorial) layer of epidermal ectoderm and somitogenic mesoderm(Fig. 1E,N,O). xMAKtranscripts were also detected in the developing eyes and as a well-defined stripe in the brain (Fig. 1D). The comparison with two additional markers, Engrailed 2(En2), a marker for the midbrain-hindbrain boundary (MHB), and Krox20, which is expressed in rhombomeres 3 and 5(Fig. 1H), revealed that the brain-specific expression of xMAK is restricted to the MHB. At later stages, xMAK expression was observed in the eyes, branchial arches,the otic vesicle and the tailbud (Fig. 1J,K). No staining was detected with the sense probe used as a negative control (Fig. 1F,G,I,L,M). These results demonstrate that xMAKexpression is dynamically regulated during development.

To evaluate activity of MAK in a gain-of-function assay, mRNAs encoding tagged MAK or its kinase-dead mutant (MAK-KD) have been injected in the dorsal margin of four-cell embryos (Fig. 2A). We observed that MAK RNA, but not the control GFP RNA, caused strong morphogenetic abnormalities, eye and anterior brain deficiencies in injected embryos. In a typical experiment, MAK RNA produced axis extension defects in 82% of injected embryos (n=27) at stage 38. MAK-KD RNA did not have a significant effect on gastrulation or neurulation movements,indicating that MAK enzymatic activity is necessary for the morphogenetic defects observed. The majority (93%) of MAK RNA-injected embryos had incomplete or absent retinal pigmentation, and some were missing anterior head structures (Fig. 2A). MAK-KD RNA injections also produced mild eye deficiencies at lower frequency(Fig. 2A). These observations suggest that MAK may be involved in the control of morphogenetic movements,head and eye development.

To evaluate whether MAK selectively interferes with morphogenetic movements or cell fate specification, we analyzed elongation of animal cap explants treated with activin, a common model for convergent extension(Sokol, 1996; Symes and Smith, 1987). Whereas MAK-KD RNA-injected and control uninjected explants elongated in response to activin, explants expressing wildtype MAK failed to elongate under identical culture conditions(Fig.2B). Both proteins were expressed equally well in injected embryos, demonstrating that the difference in activity is not due to different expression levels(Fig. 2C).

To evaluate whether the effect of MAK on convergent extension movements in animal pole explants may be due to a block in cell responsiveness to activin,we assessed mesodermal markers in animal caps cultured until stage 14 using RT-PCR (Fig. 2D). Neither MAK nor MAK-KD inhibited the induction of MyoD and Chordin by activin. This observation suggests that MAK has a specific effect on convergent extension, rather than on the ability of animal pole cells to respond to activin. Together, these observations demonstrate that active MAK is a potent regulator of morphogenetic movements and eye development in early embryos.

The strong effect of the wild-type MAK on convergent extension movements suggested that MAK may be involved in non-canonical Wnt signaling. Several noncanonical Wnt signaling components, including Dsh, Frizzled, Prickle and Strabismus, have been reported to activate JNK in mammalian cells and Xenopus embryonic ectoderm(Boutros et al., 1998; Lisovsky et al., 2002; Park and Moon, 2002; Takeuchi et al., 2003). As JNK activation is often associated with non-canonical Wnt signaling, we measured Jun phosphorylation in lysates of embryos injected with wild type MAK or MAK-KD RNA. We observed that MAK, but not MAK-KD, activated JNK in injected embryos (Fig. 3), suggesting a role for MAK in noncanonical Wnt signaling. Both constructs were expressed at equal levels (data not shown). An activated form of Fz8, previously reported to induce JNK activity (Lisovsky et al.,2002) served as a positive control in this assay. The effects of MAK on morphogenetic movements and the ability to activate JNK are consistent with the postulated function of MAK in noncanonical signaling.

To gain further insight into the molecular function of MAK, we studied whether upstream Wnt signaling components influence the distribution of MAK in the cell. For these experiments, we analyzed the subcellular localization of GFP-tagged xMAK in animal cap cells of stage 10 embryos. xMAK-GFP exhibited a punctuated cytoplasmic distribution pattern(Fig. 4A). Co-expression of Frizzled 3 RNA dramatically altered this pattern and recruited xMAK-GFP to the cell membrane. This effect was not observed upon coexpression of Frizzled 2 or 7 RNA (Fig. 4A and data not shown), indicating that the effect of Frizzled 3 is specific. These observations suggest that the subcellular localization of MAK depends on Frizzled signaling.

Dsh is another protein reported to be recruited by Frizzled to the cell membrane, and this recruitment has been considered to be crucial for non-canonical signaling (Axelrod et al.,1998; Boutros et al.,1998; Rothbacher et al.,2000; Yang-Snyder et al.,1996). Therefore, we tested whether MAK can physically associate with Dsh. We found that MAK co-immunoprecipitated with Dsh in lysates of injected embryos (Fig. 4B). To assess whether this association is functional, we performed an immune complex kinase assay in vitro and demonstrated that MAK can phosphorylate Dsh(Fig. 4C). To exclude a possibility that another kinase is responsible for Dsh phosphorylation, we used MAK-KD as a control. As expected, the MAK-KD mutant did not have any kinase activity towards Dsh. This finding shows that Dsh may represent a molecular substrate for MAK in vivo.

Together, these experiments show that MAK is recruited to the cell membrane by Frizzled 3 and can associate with and phosphorylate Dsh. These observations provide a possible mechanism for MAK action in the Wnt pathway and during early development.

To further investigate a role for MAK in development in loss-of-function studies, an antisense morpholino oligonucleotide (MAK MO) has been designed to suppress MAK RNA translation in a sequence-specific manner. Indeed, MAK MO,but not a control MO (COMO), inhibited in vitro translation of xMAKRNA in rabbit reticulocyte lysates (Fig. 5A). MAK MO did not suppress the translation of mMAK RNA that lacks MO target sequence. More importantly, MAK MO caused a reduction of xMAK protein levels in vivo in injected embryos(Fig. 5B).

To assess a developmental role for xMAK, we examined the phenotype of embryos injected with MAK MO into both dorsal blastomeres at the four-cell stage. At stage 38, the majority of MAK-depleted embryos, but not those injected with the same dose of COMO, exhibited shortened trunks and tails and severe retinal defects (Fig. 5C, Table 1). The requirement of MAK for eye development correlates with the observed expression of MAK in the eye field (Fig. 1D). The effect of MAK MO on retinal development was cell-autonomous (Fig. 5D,E). The eye defect was dose dependent, as lower doses of MAK MO had a less pronounced effect on eye pigmentation (data not shown), and it was partially rescued by co-injection of mMAK mRNA, lacking MO target sequence(Fig. 5G,H). These observations indicate that the effect of MAK MO is specific and demonstrate a role for MAK in axis elongation and eye development. Thus, the same processes are affected in both gain-of-function and loss-of-function experiments, indicating that MAK levels are under a tight control during normal development.

Our data indicate that MAK is an activator of JNK and may be involved in noncanonical Wnt signaling, which is essential for the control of morphogenetic movements and eye development. We next wanted to assess whether MAK may regulate canonical Wnt signaling, which is crucial for early dorsal development in vertebrate embryos (Harland and Gerhart, 1997). To test this possibility, we assessed the effect of MAK on secondary axis induction by Dsh(Sokol et al., 1995). In a representative experiment, Dsh RNA induced complete body axes with eyes and head structures in 81% of injected embryos (n=27, Fig. 6A,B). By contrast, when MAK RNA was co-injected, only 16% embryos (n=32) had complete secondary axes with head structures. Coinjection of the control nβgal RNA or MAK-KD RNA did not significantly alter the axis-inducing activity of Dsh(Fig. 6B). Consistent with these results, MAK, but not MAK-KD, RNA dose-dependently suppressed Dsh-mediated induction of the pSiaLuc reporter, a direct readout of the canonical pathway (Fig. 6C) (Fan et al.,1998).

To position MAK in the Wnt/β-catenin pathway, MAK RNA was co-injected with Xwnt8 RNA and with β-catenin RNA. Whereas MAK efficiently blocked Xwnt8 signaling activity, no effect onβ-catenin-dependent reporter stimulation has been observed(Fig. 6C). These findings indicate that MAK is an efficient inhibitor of Wnt signaling, and this inhibition requires MAK kinase activity. Moreover, MAK appears to inhibit Wnt signaling upstream or parallel to β-catenin.

At neurula stages, xMAK transcripts are present in the isthmus,suggesting a role for MAK in regional brain patterning. To test this possibility, we analyzed embryos with altered levels of MAK by in situ hybridization with region-specific brain markers. MAK MO or MAK RNA was injected into a single dorsal-animal blastomere of the four to eight-cell embryo. At stage 20, the injected embryos were fixed and hybridized with probes for the anterior brain marker Otx2(Blitz and Cho, 1995; Pannese et al., 1995), the midbrain-hindbrain boundary marker En2(Brivanlou and Harland, 1989; Joyner and Martin, 1987) and the hindbrain marker Gbx2 (Millet et al., 1999; von Bubnoff et al., 1996). Embryos were scored based on the difference in marker intensity between injected and uninjected sides(Fig. 7A-O). We have found that MAK MO, but not the control MO, caused unilateral suppression or loss of Otx2 and En2 in 71% (n=58) and 28% (n=50)of injected embryos, respectively (Table 2). This effect was accompanied by the upregulation of the posterior neural marker Gbx2 in 71% (n=62) of injected embryos. These results demonstrate a posteriorizing effect of MAK MO on the developing midbrain-hindbrain boundary.

In a complementary gain-of-function experiment, embryos have been injected with wild-type MAK RNA and MAK-KD RNA. MAK RNA caused an effect opposite to that observed for MAK MO. MAK RNA inhibited Gbx2 in 53%(n=32) of injected embryos, whereas Otx2 and En2expression bands shifted to a more posterior position in 38% (n=42)and 37% (n=41) of injected embryos, respectively(Table 2), suggesting that MAK anteriorizes neural tissue. Together with the observed expression of MAK at the midbrain-hindbrain boundary (Fig. 1D), these findings are consistent with the hypothesis that MAK is involved in the establishment of the isthmic organizer(Millet et al., 1999; Wurst and Bally-Cuif,2001).

As MAK regulates the Wnt pathway, and this pathway is essential for Spemann organizer function, the observed effects of MAK on brain development may be a secondary consequence of the impaired organizer function. We evaluated this possibility by comparing expression levels of several organizer markers(Stennard et al., 1997) in injected embryos at the beginning of gastrulation. RT-PCR analysis did not reveal any significant changes for Chordin (Chd), Goosecoid (Gsc), Xenopus nodal-related 3(Xnr3), Siamois (Sia), Dickkopf1(Dkk1) and Cerberus (Cer) in embryos depleted of MAK or overexpressing MAK (Fig. 7P). By contrast, β-catenin MO significantly downregulated organizer markers, demonstrating high sensitivity of this assay. These observations suggest that MAK functions to establish the midbrain-hindbrain boundary (MHB) by influencing neural tissue directly, rather than by affecting the organizer.

The formation of the isthmic organizer at the midbrain-hindbrain boundary is known to require canonical Wnt signaling(Hidalgo-Sanchez et al., 2005; Kunz et al., 2004). Interestingly, the observed effect of MAK on the midbrain-hindbrain boundary closely resembles the effect of β-catenin MO(Wu et al., 2005). This similarity and the potential inhibitory role for MAK in Wnt signaling(Fig. 6) indicate that MAK may regulate midbrain-hindbrain boundary by inhibiting the canonical pathway.

To test this hypothesis, we attempted to rescue the effect of MAK MO by depleting β-catenin. We studied the expression of Otx2 and Gbx2 by in situ hybridization in embryos co-injected into a single animal dorsal blastomere with MAK MO and β-catenin MO(Fig. 8). Marker expression was scored by comparing injected and uninjected sides of stage 20 embryos, which have been identified by lineage tracing with β-galactosidase. MAK MO induced Gbx2 in 86% (n=14) of injected embryos and downregulated Otx2 in 70% (n=23) of injected embryos. Upon β-catenin MO co-injection, the induction of Gbx2 was suppressed, whereas Otx2 levels did not change in the majority (65%, n=40) of injected embryos. We conclude that β-catenin MO rescued normal brain patterning in MAK-depleted embryos. Thus, the effects of MAK depletion on Otx2 and Gbx2 expression are probably due to canonical pathway upregulation at the midbrain-hindbrain boundary, which is consistent with an inhibitory role for MAK in Wnt/β-catenin signaling during midbrain patterning.

This study identifies MAK, a member of the SNF1 kinase family, as a novel modulator of Wnt signaling. By the onset of gastrulation, MAK is present in mesodermal and ectodermal tissues in a broad pattern. At neurula stages, the major regions of MAK expression are in the developing eyes, the isthmus and the somites. Consistent with this pattern, our loss- and gain-of-function experiments suggest that during early development MAK is required for convergent extension movements, eye development and the midbrain-hindbrain boundary specification. We note that all these developmental processes are known to be controlled by Wnt signaling(Cavodeassi et al., 2005; Liu and Joyner, 2001; Sokol, 2000). In mouse embryos, MAK has been implicated in mammary gland development, another process regulated by Wnt signaling (Gardner et al., 2000a). In our experiments, Frizzled 3 dramatically altered MAK intracellular localization by recruiting it to the cell membrane(Fig. 4A), demonstrating that MAK can respond to Wnt signaling. The effects of MAK on convergent extension and on canonical Wnt signaling require its kinase activity, suggesting that phosphorylation of a putative MAK substrate plays a major role in signaling outcome. In a search for possible MAK substrates, we found that MAK associates with Dsh and is able to phosphorylate Dsh in vitro. As Dsh has been implicated in gastrulation movements (Sokol,1996) and brain patterning(Hamblet et al., 2002; Itoh and Sokol, 1997), it may be a relevant molecular target of MAK during these developmental processes.

The involvement of MAK in midbrain-hindbrain boundary regulation is most probably due to the inhibition of the canonical β-catenin signaling in the isthmus, as β-catenin MO suppresses the effect of MAK MO(Fig. 8). By contrast,inhibition of convergent extension movements in activin-treated explants overexpressing MAK and in embryos injected with MAK RNA or MAK MO (Figs 2, 5) may relate to a noncanonical branch of Wnt signaling. In agreement with this hypothesis, MAK stimulates JNK activity, a common downstream target of noncanonical Wnt signaling(Veeman et al., 2003). Together, these experiments indicate that MAK functions to stimulate non-canonical Wnt signaling, but negatively regulates the canonical pathway. Thus, MAK acts in a manner opposite to that of PAR-1, a related SNF1-like kinase, which activates the canonical pathway but suppresses the noncanonical pathway by downregulating JNK (Ossipova et al., 2005; Sun et al.,2001).

In the isthmic organizer, canonical Wnt signaling may be responsible for the transcriptional activation of several target genes, including those encoding Wnt pathway components, e.g. Wnt1(McMahon and Bradley, 1990),Axin (Hedgepeth et al., 1999),LEF1/TCF (Kunz et al., 2004; Molenaar et al., 1998) and Frodo (Gillhouse et al.,2004). Consistent with the hypothesis that MAK modulates Wnt signaling in the brain, the loss-of-function experiments with embryos injected with MAK MO resulted in inhibition of the anterior brain marker Otx2and the isthmus marker En2, but caused upregulation of Gbx2,the posterior brain marker (Fig. 7). MAK injections did not substantially upregulate Otx2and En2, but shifted these markers more posteriorly(Fig. 7J,K). Together with the lack of effect of MAK on organizer markers(Fig. 7P), this indicates that MAK regulates midbrain-hindbrain boundary specification, rather than causes a more general effect on anteroposterior patterning or Spemann organizer formation. In these experiments, MAK phenocopies the effect of β-catenin MO injection (Wu et al.,2005), confirming our hypothesis that it is due to MAK-mediated inhibition of canonical Wnt signaling. The restricted effect of MAK on the midbrain-hindbrain boundary suggests that the isthmic organizer differs from other brain regions in the responsiveness to Wnt/β-catenin signaling,perhaps by expressing region-specific competence factors. This hypothesis is supported by our observation that brain-specific expression of xMAK is restricted to the midbrain-hindbrain boundary.

MAK does not seem to be able to affect dorsal mesoderm markers when overexpressed in the organizer (Fig. 7P). This finding suggests that MAK is a selective inhibitor of Wnt signaling in MHB, but not at the earlier developmental stage. One reason is that MAK is not present in fertilized eggs and does not function in early axial patterning. It is also possible that the upstream components of Wnt signaling including MAK are not involved in organizer formation in early embryogenesis (Harland and Gerhart,1997; Sokol,1996), whereas they may be crucial at an earlier stage, e.g. during oocyte development (Tao et al.,2005). As secreted Wnt antagonists and Dsh interfering mutants do not influence organizer development(Glinka et al., 1998), it has been hypothesized that the pathway is activated by a ligand-independent mechanism (Sokol, 1996).

At present, how MAK redirects Wnt signaling from the canonical to the noncanonical mode is unclear. A simple hypothesis is that it does so by phosphorylation of Dsh. Thus, the identification of MAK-specific phosphorylation sites in Dsh will be important in future studies of MAK. Alternatively, MAK may have other molecular targets that are crucial for signaling, as other protein kinases, including casein kinase 1 and GSK3,function in Wnt signaling by phosphorylating multiple substrates(Lee et al., 2001; Zeng et al., 2005).

We thank Keiji Itoh and Sangeeta Dhawan for helpful discussions and critical comments on the manuscript. We are especially grateful to Igor Korobko for mMAK plasmids and for sharing ideas that allowed this project to start. This study has been supported by NIH grants to S.Y.S.