Ptk7 promotes non-canonical Wnt/PCP-mediated morphogenesis and inhibits Wnt/β-catenin-dependent cell fate decisions during vertebrate development

Wnt signalling controls a diverse range of developmental processes from tissue specification to axial morphogenesis (Moon et al., 1997; Wodarz and Nusse, 1998). Pathway activation is carefully controlled, and misregulated Wnt activity has been implicated in a variety of developmental abnormalities (Ikeya et al., 1997; Yamaguchi et al., 1999a; Hamblet et al., 2002; Wallingford and Harland, 2002). Wnt signalling is also required for stem cell maintenance and adult tissue homeostasis, with perturbations frequently implicated in cancer formation (Logan and Nusse, 2004; Polakis, 2007).

Wnt signal transduction pathways are highly conserved among metazoans. In the canonical Wnt/β-catenin pathway, Wnt ligands bind Frizzled (Fz) receptor/low density lipoprotein receptor-related protein (LRP) complexes at the cell surface. Wnt signals, transduced through Dishevelled (Dvl), inhibit GSK-3β-APC-Axin complex-directed proteosomal degradation of β-catenin. Nuclear translocation of stable β-catenin leads to the transcriptional activation of genes involved in cell replication, growth, apoptosis and self-renewal. During embryogenesis, canonical Wnt/β-catenin signalling also controls diverse cell fate determination events, including dorsal-ventral patterning (De Robertis et al., 2000; De Robertis and Kuroda, 2004; Schier and Talbot, 2005), cardiac and central nervous system differentiation (Yamaguchi et al., 1999b; Ryu et al., 2001), and the specification of paraxial mesoderm cell fate from a bi-potential neural/mesodermal progenitor cell pool within the vertebrate tailbud (Martin and Kimelman, 2012; Nowotschin et al., 2012).

Multiple β-catenin-independent pathways can also be activated upon Wnt binding to Fz receptors (Kohn and Moon, 2005; Jenny and Mlodzik, 2006), the best characterised of which is the planar cell polarity (PCP) pathway. PCP coordinates the uniform orientation of cell structure and cell movement within the plane of a tissue (Zallen, 2007). In vertebrates, this is required to drive convergence and extension (C&E) - polarised cell movements that narrow and extend the body axis (Keller, 2002; Wallingford and Harland, 2002; Yin et al., 2009). Mutation of vertebrate PCP genes block these cell movements, resulting in broader and shorter tissues, and neural tube closure defects (Simons and Mlodzik, 2008; Yin et al., 2009).

A core ‘cassette’ of proteins establish PCP across animal species. As in Wnt/β-catenin signalling, PCP requires Fz activation and subsequent membrane localisation of Dvl. In vertebrates, Fz activation is thought to occur through binding of Wnt5 and Wnt11 ligands (Yin et al., 2009). Other proteins, specific to PCP, regulate Fz-Dvl activity. They include the transmembrane protein Van Gogh (Vang) and the cytosolic effector protein Prickle (Pk) (Jenny et al., 2003). Asymmetric subcellular localisation of PCP molecules is key to cell polarity, creating regionalised activation of Rho kinase, and local reorganisation of the actin cytoskeleton (Strutt and Strutt, 2009).

Differential activation of alternative Wnt pathways is not entirely ligand dependent; in some contexts, Wnt5a can activate Wnt/β-catenin signalling (Clevers, 2006). Rather, the cellular complement of Fz co-receptors dictates Wnt pathway choice (van Amerongen et al., 2008). Lrp6 activates canonical Wnt/β-catenin responses (Tamai et al., 2000), whereas the ROR transmembrane kinases can specifically bind Wnt5a and inhibit Wnt/β-catenin signalling (Mikels and Nusse, 2006). Ror2 can also form a Wnt-induced complex with the PCP regulator Van Gogh-like 2 (Vangl2) to affect tissue polarity (Gao et al., 2011). Furthermore, regulatory proteins can shift the balance between alternative Wnt pathway activation. The metastasis-associated transmembrane protein Wnt-activated inhibitory factor 1 (Waif1), for example, inhibits Wnt/β-catenin signalling and activates non-canonical Wnt pathways by modifying Lrp6 subcellular localisation (Kagermeier-Schenk et al., 2011).

Protein tyrosine kinase 7 (Ptk7; also known as colon carcinoma kinase-4, CCK4) is an atypical receptor tyrosine kinase that has also been implicated in vertebrate Wnt signal transduction (Lu et al., 2004; Shnitsar and Borchers, 2008; Paudyal et al., 2010; Peradziryi et al., 2011; Puppo et al., 2011). Initially identified in metastatic colon cancer, PTK7 expression is also upregulated in acute myeloid leukaemia (AML) and multiple other cancer types (Mossie et al., 1995; Easty et al., 1997; Endoh et al., 2004; Müller-Tidow et al., 2004; Gorringe et al., 2005; Gobble et al., 2011). Full-length Ptk7 contains seven extracellular immunoglobulin (Ig)-like domains, a transmembrane domain and a catalytically inactive kinase domain; the Ptk7 ATP-binding motif is mutated (Mossie et al., 1995) and tyrosine phosphorylation (activation) is not observed following overexpression in COS-1 or HEK293 cells, or in in vitro kinase assays (Mossie et al., 1995; Jung et al., 2004). Orthologues have been identified in mouse, Xenopus, chick (KLG), Drosophila (Off-track, OTK) and Hydra (Lemon), and have been implicated in various morphogenetic processes in each system (Chou and Hayman, 1991; Lee et al., 1993; Mossie et al., 1995; Miller and Steele, 2000; Winberg et al., 2001; Jung et al., 2004). Mouse Ptk7 mutant embryos die perinatally and display shorter and wider body axes, loss of mesodermal cell polarity, and neural tube closure defects, abnormalities attributed to defective PCP (Lu et al., 2004; Yen et al., 2009; Paudyal et al., 2010). Similar phenotypes are observed upon morpholino (MO) knockdown of PTK7 in Xenopus (Lu et al., 2004).

There is also evidence that Ptk7 might regulate canonical Wnt/β-catenin signalling. Ptk7 co-immunoprecipitates (co-IPs) with Wnt3a and Wnt8 (Peradziryi et al., 2011), requiring a Ptk7 and Fz7 receptor complex that has been demonstrated to be both dependent (Shnitsar and Borchers, 2008) and independent (Peradziryi et al., 2011) of interactions with Dvl. The Ptk7 pseudokinase domain can also interact with β-catenin in yeast-two-hybrid assays, and mammalian cell co-IPs. Moreover, overexpression of PTK7ΔICD constructs, which lack the intracellular pseudokinase domain, suppresses Wnt/β-catenin dorsal organiser gene expression in Xenopus (Puppo et al., 2011).

Functional investigations into the role for Ptk7 in vertebrate Wnt/β-catenin signalling remain inconclusive. Wnt/β-catenin patterning defects have not been reported in three independent analyses of Ptk7 mutant mice (Lu et al., 2004; Yen et al., 2009; Paudyal et al., 2010). In Xenopus, MO knockdown of PTK7 suppresses Wnt/β-catenin Spemann organiser formation and siamois promoter activation (Puppo et al., 2011), suggesting that Ptk7 potentiates canonical Wnt signalling. However, MO knockdown of PTK7 in Xenopus animal cap explants enhances the ability of exogenous Wnt8 to activate β-catenin signalling (Peradziryi et al., 2011), suggesting that Ptk7 inhibits the Wnt/β-catenin pathway.

In order to clarify the function of Ptk7 in vertebrate embryonic patterning and morphogenesis, we cloned the single zebrafish ptk7 orthologue and generated a loss-of-function allele. Through analysis of maternal-zygotic (MZ) ptk7 mutant embryos, we demonstrate a requirement for Ptk7 in potentiating PCP-dependent morphogenesis, and establish a clear role for Ptk7 in attenuating canonical Wnt/β-catenin signalling in vivo. Finally, we demonstrate that the highly conserved pseudokinase domain is not required for Ptk7 activity in axial morphogenesis and mesodermal cell fate determination. Rather, a plasma membrane-tethered Ptk7 extracellular domain fragment is sufficient to rescue these PCP and canonical Wnt-dependent activities in MZptk7 mutant embryos.

Wild-type and mutant zebrafish strains were raised under standard conditions (Westerfield, 2007). The vangl2 mutant allele tri(tk50f) contains a deletion in the vangl2 coding sequence (Jessen et al., 2002).

ZiFiT software (http://zifit.partners.org/ZiFiT/) was used to identify potential target sites in ptk7. Using the Oligomerized Pool Engineering (OPEN) system (Maeder et al., 2008; Foley et al., 2009) method, two selections (one for each half-site) were made. A bacterial one-hybrid assay was used to test specificity of the engineered arrays against a sequence-specific pH3U3 reporter vector expressed in USOΔhisΔpyrΔrpoZ cells (Meng et al., 2008). The zinc-finger cassette was inserted into pCS2+ expression vectors. The sequences for left and right zinc-fingers are listed in supplementary material Table S1.

mRNA (80 pg) for each ZFN was injected into one-cell-stage embryos (F0). F1 genomic DNA was screened for mutations in ptk7 using a PCR-based assay: forward, 5′-CGAAGGCCGCTGAGGATGA-3′; reverse, 5′-CAGAAAACGCATGAAGTGACCAGC-3′. SfaN1 endonuclease (NEB) was used to identify potential mutations.

Ptk7 was amplified from 24 hours post-fertilisation (hpf) wild-type cDNA made using oligo(dT)_12-18 primer (Invitrogen) and SuperscriptII (Invitrogen) as per manufacturer’s instructions. Ptk7 was cloned into pCS2+ using Gateway technology (Invitrogen). Ptk7ΔICD, ptk7 egfrTM, ptk7ΔECD and ptk7 ECD were amplified from full-length ptk7 (supplementary material Table S2). The mMESSAGE mMACHINE System (Ambion) was used to make mRNA. All injections were performed at the one-cell stage.

Antisense RNA probes were prepared by in vitro transcription (DIG RNA Labeling Kit, Roche) in the presence of digoxigenin-11-UTP from linearized DNA templates. Embryos were cleared in 100% methanol, mounted in benzylbenzoate:benzylalcohol (2:1), and imaged on an Axio Imager.M1 (Zeiss) compound microscope.

Total RNA was extracted using TRIzol reagent (Invitrogen) according to manufacturer’s recommendations. First-strand cDNA was made using SuperScript II reverse transcriptase (Invitrogen) and oligo(dT)_12-18 primer (Invitrogen). For primer sequences, see supplementary material Table S2. SYBR green (Applied Biosystems) was used according to manufacturer’s recommendations. All analyses were carried out in triplicate using a Light Cycler 480 (Roche) platform. Fifty amplification cycles were performed, with each cycle consisting of 95°C for 10 seconds, 60°C for 20 seconds and 72°C for 30 seconds. Amplification and dissociation curves generated by the Light Cycler 480 Software release 1.5.0 SP4 were used for gene expression analysis. Ct values were obtained for each gene and normalised to Gapdh. Fold change was calculated relative to wild-type expression according to the equation: 2^-ΔΔCt. Standard error was calculated as standard deviation of the fold change according to the equation: stdev_foldchange=(ln2)(stdev_ΔΔCt)(2^-ΔΔCt), where stdev_ΔΔCt=√(stdev of reference)² + (stdev of gene of interest)². All graphs are representative of two independent experiments with three technical replicates each. Statistical significance was calculated using Student’s t-test.

Live embryos were immobilised on a coverslip in 0.8% agarose, and imaged using a Zeiss LSM 710 microscope. z-stacks were collected and processed using ImageJ software.

TopFlash was conducted in human HEK293T cells as previously described (Angers et al., 2006; Lui et al., 2011). Lentiviruses containing the superTopFlash β-catenin-dependent luciferase reporter (Firefly luciferase) and pRL-TK (Renilla luciferase) were produced and used to establish stable HEK293T Wnt-reporter lines. Cells were seeded on 24-well plates, followed by cDNA transfection with polyethylenimine (PEI). Media was replaced with a 1:1 mix of fresh DMEM:Wnt3A or DMEM:Control conditioned media. Cells were assayed 16 hours after stimulation, performed in accordance with the Dual Luciferase assay specifications (Promega) using an Envision 2103 Multilabel Plate Reader (PerkinElmer).

Full-length zebrafish ptk7 coding sequence was cloned from a 24 hpf cDNA library using primers designed against predicted open reading frame sequence (supplementary material Fig. S1). Zebrafish Ptk7 protein contains 1061 amino acids and consists of an extracellular domain with seven immunoglobulin-like loops, a single transmembrane domain and an intracellular domain with structural features typical of a tyrosine kinase (van der Geer et al., 1994) (Fig. 1A). However, as with all vertebrate Ptk7 orthologues, the DFG triplet believed to be essential for catalytic activity is modified in zebrafish Ptk7 (supplementary material Fig. S1), classifying it as a pseudokinase. Overall, zebrafish Ptk7 is highly conserved with chicken (Chou and Hayman, 1991), Xenopus (Lu et al., 2004) human and mouse (Mossie et al., 1995; Jung et al., 2004) proteins, with greatest conservation observed within transmembrane and intracellular pseudokinase domains (Fig. 1A). We were unable to find evidence of additional ptk7 homologues in the zebrafish genome through nucleotide BLAST searches against the Zv9 assembly.

Using whole-mount RNA in situ hybridisation (WISH), we observed ptk7 expression at the two-cell stage, prior to mid-blastula transition (MBT) and initiation of zygotic transcription, indicating strong maternal contribution (Fig. 1B). ptk7 expression is ubiquitous at the onset of gastrulation (shield stage; Fig. 1C), and is pronounced in axial, paraxial and tailbud lineages during late-gastrula stages (Fig. 1D). Post-gastrulation, ptk7 is expressed in the head, neural tube and somites, with the highest level of expression observed in the tailbud (Fig. 1E). ptk7 expression within the tailbud persists until 24 hpf (Fig. 1F). The expression of zebrafish ptk7 is similar to that observed in Xenopus and mouse (Lu et al., 2004), encouraging us to examine Ptk7 function further using zebrafish as our model system.

As a first investigation into Ptk7 function, we injected high levels of ptk7 mRNA into single-cell-stage embryos (within 10 minutes of fertilisation) and analysed its effect on embryonic patterning and morphogenesis. Consistent with published results, overexpression of ptk7 caused phenotypes associated with abnormal PCP (Fig. 2B), including defects in axial extension as well as dorsal curvatures of the posterior tail (observed with vangl2 overexpression) (Borovina et al., 2010). However, overexpression of ptk7 also yielded dorsalised embryos (32%, n=100; Fig. 2C), a phenotype not associated with PCP signalling mutants.

Canonical Wnt/β-catenin signals strongly influence dorsal-ventral pattern, first promoting dorsal organiser formation prior to MBT and then opposing its maintenance during late-blastula stages (De Robertis et al., 2000; De Robertis and Kuroda, 2004; Schier and Talbot, 2005). At 4 hpf, we could not detect changes in expression of bozozok (boz; dharma - Zebrafish Information Network) and chordin (chd), direct targets of β-catenin at the dorsal organiser (Wylie et al., 1996; Fekany et al., 1999; Solnica-Krezel and Driever, 2001), in ptk7-injected embryos (supplementary material Fig. S2A). However, later shield-staged embryos displayed lateral expansion of chd (47%, n=30; Fig. 2D,E) as well as reduced expression of the ventral Wnt-target gene vox (Ramel and Lekven, 2004) (77%, n=30; Fig. 2F,G). Moreover, ptk7 overexpression resulted in reduced expression of the β-catenin target axin2, as assayed by WISH (Fig. 2H,I) and qRT-PCR (supplementary material Fig. S2B). Our results suggest that ptk7 overexpression may attenuate the ventralising activity of canonical Wnt signals during early embryogenesis.

To investigate further the role for zebrafish Ptk7 in canonical Wnt signalling, we examined the effects of ptk7 overexpression on exogenous Wnt/β-catenin activity. In zebrafish embryos, ectopic expression of wnt8 produces phenotypes characteristic of Wnt/β-catenin activation; namely, defects in dorsoventral patterning and posteriorisation of the neuroectoderm (Lekven et al., 2001) (Fig. 2J). Strikingly, co-injection of ptk7 mRNA largely rescued the loss of forebrain and eye-field associated with wnt8 overexpression (94%, 52/55; Fig. 2J), suggesting that Ptk7 can inhibit exogenous Wnt/β-catenin activity.

To determine which domains of Ptk7 protein are required for this function, we tested the ability of deletion and substitution mutants (Fig. 2K) to rescue wnt8 overexpression defects. An intracellular domain deletion construct (ptk7ΔICD) rescued wnt8-induced eye and forebrain defects to the same extent as full-length ptk7 (96%, n=50; Fig. 2L), as did substitution of the Ptk7 transmembrane domain with that of Egfr (ptk7EgfrTM; 94%, n=51; Fig. 2L). Expression of an extracellular domain deletion mutant (ptk7ΔECD) failed to rescue wnt8 overexpression (0/26; Fig. 2L), as did expression of a secreted extracellular domain (ptk7ECD) fragment (0/28; Fig. 2L). Our results indicate that the highly conserved Ptk7 pseudokinase and transmembrane domains are not required to inhibit Wnt/β-catenin activity. Rather, a membrane-tethered Ptk7 extracellular domain is crucial for this function.

Ptk7 expression similarly abrogated Wnt3a-stimulated TOP-Flash luciferase activity in HEK293T cells. Deletion of the intracellular domain (ptk7ΔICD) and substitution of the transmembrane domain (ptk7EgfrTM) did not impair protein function (supplementary material Fig. S3). However, ptk7ΔECD expression failed to inhibit Wnt3a-stimulated luciferase activity (supplementary material Fig. S3), confirming that the extracellular domain is necessary to inhibit exogenous Wnt/β-catenin signalling.

In contrast to our zebrafish studies, Ptk7ECD inhibited luciferase activity to the same extent as full-length protein in HEK293T cultures (supplementary material Fig. S3). This suggests functional differences between in vivo and in vitro Wnt overexpression systems. Of note, overexpression of Ptk7 extracellular fragments in zebrafish embryos had no effect on patterning and morphogenesis (supplementary material Fig. S4B), whereas Ptk7ΔICD, in which the extracellular domain remains tethered to the plasma membrane, produced both PCP and dorsoventral patterning phenotypes at frequencies similar to full-length Ptk7 (supplementary material Fig. S4B).

To determine the requirement for Ptk7 in zebrafish embryonic patterning and morphogenesis, we targeted mutations into the genomic ptk7 locus using specifically engineered ZFNs. Web-based ZiFiT software (http://zifit.partners.org/ZiFiT/) was used to identify an optimal target site within the N-terminus of the pseudokinase domain, upstream from the predicted ATP-binding pocket. OPEN (Oligomerized Pool ENgineering) combinatorial-based selection methods were employed to generate multiple ZFNs (Maeder et al., 2008; Foley et al., 2009), and efficient/specific binding proteins were selected using a bacterial one-hybrid screen (Meng et al., 2008). mRNA (80 pg) coding for each of two ZFN pairs was injected into one-cell stage embryos, F0 fish were raised to adulthood, and F1 genomic DNA was screened for mutations in ptk7 using PCR-based assays (Fig. 3B). Using this strategy, we isolated a mutant allele (ptk7^hsc9) harbouring a 10-bp deletion that results in a frame shift and the incorporation of multiple premature termination codons.

In contrast to mouse and Xenopus studies, Ptk7^hsc9 mutant zebrafish develop normally until 3-4 weeks post-fertilisation, at which time mutant larvae acquire axial curvatures (100%, n=34; Fig. 3C,D). The presence of significant maternal ptk7 transcript in homozygous mutant embryos, however, led us to consider the possibility that maternal contribution might rescue early developmental defects.

Maternal zygotic ptk7^hsc9 (MZptk7^hsc9) mutant embryos were generated through routine crosses of Ptk7^hsc9 mutant adults. Because the ptk7^hsc9 mutation introduces multiple premature termination codons into exon 16 of 20, we investigated whether mutant transcript was targeted by non-sense-mediated decay (NMD). Using WISH and qRT-PCR, we did not detect ptk7 transcripts in one-cell-stage MZptk7^hsc9 embryos (100%, n=15; Fig. 3E,F), suggesting strong loss of maternal contribution. At gastrula to early segmentation stages, we observed significant downregulation of ptk7^hsc9 transcript levels (Fig. 3J-L). We did detect small amounts of ptk7 in the tailbud of 10- to 12-somite-stage MZptk7^hsc9 mutants (Fig. 3G,H); however, at 24 hpf, no transcript was observed (Fig. 3I). Our data suggest that ptk7^hsc9 transcripts are subject to NMD, indicating a strong loss-of-function allele.

MZptk7^hsc9 embryos appear shorter and wider than wild-type embryos (100%, n=100; Fig. 4A-D), indicating possible defects in PCP-mediated C&E movements. Morphometric analyses of myoD (myod1; somite) and krox20 (egr2; hindbrain) marker expression confirmed a significant compression of the MZptk7^hsc9 rostral-caudal axis (Fig. 4E), with concomitant expansion of the MZptk7^hsc9 mediolateral axis (Fig. 4F). At tailbud stages, the neural plate and notochord of MZptk7^hsc9 embryos (marked by dlx3 and ntl expression, respectively) are expanded mediolaterally, and the prechordal plate (hgg1; ctsl1b) does not extend to its normal anterior position (100%, n=16; Fig. 4G-J). Furthermore, MZptk7^hsc9 embryos demonstrate neural tube morphogenesis defects similar to those observed in MZvangl2 embryos (a core PCP signalling mutant) (Ciruna et al., 2006), with cells accumulating ectopically at the neural midline (Fig. 4K,L). Although MZptk7^hsc9 embryos are viable through pharyngula and hatching stages, the majority die at 1-2 weeks post-fertilisation. Surviving MZptk7^hsc9 adult zebrafish (9%, n=90) appear similar to zygotic ptk7^hsc9 mutant parents.

Cells engaged in C&E movements become elongated and oriented mediolaterally (Keller et al., 2000; Topczewski et al., 2001). Using confocal microscopy, we analysed the shape and orientation of MZptk7^hsc9 and wild-type gastrulae cells (Fig. 4M-O). In the dorsal ectoderm of wild-type gastrulae, cells are elongated [length-to-width ratio (LWR) of 2.0±0.46, n=74; Fig. 4M] with strong mediolateral bias in their orientation with respect to the midline (72% of cells within ±15° with respect to the mediolateral axis; Fig. 4N,O). MZptk7^hsc9 mutant cells were significantly less elongated (LWR of 1.2±0.24, n=94, P<0.001; Fig. 4M) and exhibited less mediolateral bias in their orientation (22% of cells, P<0.001; Fig. 4N,O), consistent with a loss of PCP.

To investigate further how loss of Ptk7 affects PCP, we visualised the subcellular localisation of GFP-tagged Prickle (GFP-Pk), a cytosolic effector protein that localises asymmetrically along the anterior membrane of zebrafish neuroepithelial and mesodermal cells in response to PCP signals (Carreira-Barbosa et al., 2003; Veeman et al., 2003; Ciruna et al., 2006; Yin et al., 2008). Although GFP-Pk localised to discrete puncta at the anterior plasma membrane of wild-type neuroepithelial cells (47/61 cells, n=4 embryos; Fig. 4P-R), GFP-Pk puncta were largely lost from the anterior membrane in MZptk7^hsc9 mutants (puncta present in 2/136 cells, n=6 embryos; Fig. 4P-R). Our results suggest a conserved requirement for Ptk7 in regulating PCP.

Both hyperactivation and loss of PCP signalling often yield similar embryonic phenotypes (Moon et al., 1993; Park and Moon, 2002; Carreira-Barbosa et al., 2003). To determine whether Ptk7 plays a positive or negative role in PCP, we examined the effect of ptk7 overexpression on exogenous non-canonical Wnt activity. Although injection of high levels of wnt5 (Kilian et al., 2003) or ptk7 (Fig. 2A-C) mRNA interfere with zebrafish C&E movements, injection of low wnt5b or ptk7 mRNA levels do not disrupt embryonic morphogenesis (Fig. 5A,B). Strikingly, we observed that co-expression of low levels of both wnt5b and ptk7 mRNA yielded strong defects in C&E, as quantified through morphometric analyses of myoD expression (Fig. 5A,B). This observation suggests that Ptk7 potentiates Wnt5 signalling activity in vivo.

Similarly, embryos injected with high levels of wnt11 mRNA displayed cyclopia (a phenotype associated with activated non-canonical Wnt signalling) (Ungar et al., 1995; Lu et al., 2011), whereas injection of lower wnt11 mRNA levels did not adversely affect embryogenesis (Fig. 5C,D). However, we observed that co-injection of ptk7 mRNA with low levels of wnt11 mRNA caused a striking increase in both the incidence and severity of cyclopia, as scored in live embryos at 48 hpf (Fig. 5C,D). The synergistic effect of exogenous ptk7 and wnt11 expression on PCP-mediated morphogenetic processes suggests that Ptk7 regulates PCP by potentiating non-canonical Wnt signal activity.

We have also shown that overexpression of ptk7 attenuates Wnt/β-catenin signalling, suggesting an inhibitory role for Ptk7 in canonical Wnt signal transduction (Fig. 2). However, MZptk7^hsc9 mutant embryos do not display the gross embryological defects typically associated with mis-regulated Wnt/β-catenin activity, such as dorsalisation, ventralisation, or mispatterning of the CNS (supplementary material Fig. S5).

To investigate further the role for Ptk7 in dorsoventral patterning, we used qRT-PCR to quantify β-catenin-dependent dorsal organiser gene expression in wild-type versus MZptk7^hsc9 mutant embryos. At 4 hpf, we observed a small but significant upregulation of bozozok (boz) and chordin (chd) expression in MZptk7^hsc9 mutants (Fig. 6A), suggesting an inhibitory role for ptk7 in dorsal organiser formation. MZptk7^hsc9 mutants also demonstrated a slight expansion in boz expression at 50% epiboly as detected by WISH (100%, n=20) and qRT-PCR (Fig. 6B). However, by gastrulation, we could no longer detect significant differences in the dorsal expression of chd and goosecoid (gsc) (Fig. 6C,D), nor did we observe changes in ventral axin2 or vox expression (Fig. 6C,D). Therefore, although ptk7 might play a role in restricting dorsal organiser formation, genetic loss of ptk7 does not have a significant effect on subsequent establishment and/or maintenance of dorsoventral domains in the zebrafish gastrula.

Post-gastrulation, Wnt/β-catenin signalling is required for posterior vertebrate axis development (Galceran et al., 1999; Agathon et al., 2003; Martin and Kimelman, 2008), and drives an ongoing process of mesodermal fate specification within stem cell pools of the tailbud (Martin and Kimelman, 2012). As ptk7 expression is highly enriched in the tailbud, we investigated the post-gastrula role for Ptk7 in canonical Wnt/β-catenin activity. Using WISH, we observed upregulated levels of axin2 expression within the presomitic mesoderm of 10- to 12-somite stage MZptk7^hsc9 embryos (100%, n=17; Fig. 7A,B). Furthermore, axin2, lef1 and cyclin D1 expression levels were all significantly upregulated in MZptk7^hsc9 embryos at both bud and 10- to 12-somite stages, as assayed by qRT-PCR (Fig. 7C,D). Our data indicate that Ptk7 is required to attenuate canonical Wnt/β-catenin signal transduction in the developing tailbud.

To determine whether Ptk7 regulates canonical Wnt-dependent cell fate decisions in the tailbud, we used WISH to examine tbx6 expression (a direct target of Wnt/β-catenin signals, and a marker of presomitic mesoderm) in control versus MZptk7^hsc9 mutant embryos. Strikingly, the tbx6 expression domain was expanded in the tailbud of 10- to 12-somite stage MZptk7^hsc9 embryos (Fig. 7E-I). Similarly, WISH analysis of mespb expression (which defines the anterior limit of presomitic mesoderm) also revealed a rostral extension in the MZptk7^hsc9 presomitic mesoderm domain, and a significant expansion of the tailbud (supplementary material Fig. S6).

To determine whether observed patterning defects are an artefact of abnormal PCP-mediated tailbud morphogenesis, we examined Wnt/β-catenin target gene expression and mesodermal fate specification in vangl2 mutant embryos. No significant differences in axin2, lef1 or cyclin D1 expression were observed (supplementary material Fig. S7A), indicating that loss of PCP is not sufficient to promote canonical Wnt/β-catenin activity. Furthermore, no significant expansion of tbx6 expression was observed within the tailbud of vangl2 mutants (supplementary material Fig. S7B,C). Our data suggest that the expansion of mesodermal fate observed in MZptk7^hsc9 embryos is not a consequence of abnormal PCP. Rather, our data further supports the role for Ptk7 as an inhibitor of the Wnt/β-catenin signalling pathway, with essential functions in regulating canonical Wnt-dependent cell fate decisions in posterior stem cell pools of the vertebrate tailbud.

In contrast to MZptk7^hsc9 mutants, zygotic ptk7 mutant embryos appear normal. This indicates that maternal ptk7, which is present at the one-cell stage, is sufficient to rescue embryonic patterning and morphogenesis through to juvenile stages. We therefore reasoned that careful titration of ptk7 mRNA into one-cell-stage embryos might rescue MZptk7^hsc9 embryogenesis. Indeed, injection of full-length ptk7 mRNA constructs could largely rescue both C&E (Fig. 8A) and tailbud patterning defects (Fig. 8B) associated with MZptk7^hsc9 mutant embryos. Therefore, we decided to perform structure-function analysis of Ptk7 by assaying the differential ability of deletion and substitution constructs to rescue MZptk7^hsc9 mutant phenotypes.

The kinase domain of Ptk7, which interacts with Dvl, RACK1 and β-catenin, is thought to be essential for Ptk7 activity despite an apparent lack of catalytic activity (Lu et al., 2004; Shnitsar and Borchers, 2008; Puppo et al., 2011). Moreover, Xenopus overexpression studies suggest that Ptk7 kinase deletion mutants possess dominant-negative activity, disrupting both PCP C&E movements (Lu et al., 2004; Shnitsar and Borchers, 2008) and Wnt/β-catenin dorsal organiser gene expression (Puppo et al., 2011). However, we found that zebrafish ptk7ΔICD expression constructs (lacking the entire intracellular domain) rescued both MZptk7^hsc9 axial extension and tbx6 tailbud patterning defects to the same extent as did full-length ptk7 expression (Fig. 8A,B). Similarly, we found that ptk7EgfrTM expression (substituting the transmembrane domain for that of Egfr) also rescued MZptk7^hsc9 patterning and morphogenesis (Fig. 8A,B). Our results indicate that the pseudokinase domain of Ptk7 is not required for non-canonical/PCP or canonical Wnt/β-catenin signalling activity.

However, deletion of the Ptk7 extracellular domain (ptk7ΔECD) abrogated its ability to rescue MZptk7^hsc9 axial extension and mesodermal fate specification defects (Fig. 8A), indicating a clear requirement for the extracellular domain in regulating PCP and Wnt/β-catenin signal transduction. Of interest, expression of a secreted Ptk7 extracellular fragment (ptk7ECD) failed to rescue MZptk7^hsc9 mutant phenotypes (Fig. 8A,B). Therefore, our results clearly indicate that the plasma membrane-tethered extracellular domain of Ptk7 is both required and sufficient to regulate PCP-mediated morphogenesis and Wnt/β-catenin cell fate decisions during zebrafish embryonic development.

The Wnt pathway comprises a complex network of signalling molecules, which integrate extracellular information using multiple cell surface proteins. The presence or absence of a single component can dictate the context of cellular response, favouring activation of canonical Wnt/β-catenin, non-canonical Wnt/PCP, or other alternative Wnt signalling pathways. Recent studies implicate Ptk7 as a novel Wnt co-receptor that can influence signal specificity during vertebrate embryogenesis. However, conflicting reports have indicated both positive and negative roles for Ptk7 in regulating Wnt/β-catenin activity.

In this study, we cloned and characterised the zebrafish ptk7 orthologue. Zebrafish ptk7 is strongly expressed in one-cell-stage embryos, it is expressed ubiquitously throughout gastrulation, and strong ptk7 expression is maintained in the tailbud throughout segmentation stages. Using ZFNs assembled through OPEN combinatorial-based selection methods (Maeder et al., 2008; Foley et al., 2009), we generated a ptk7 mutant allele (hsc9) containing a 10-bp deletion within the pseudokinase domain. ptk7^hsc9 transcript is subject to non-sense-mediated decay, indicating a strong loss-of-function allele.

Zygotic ptk7^hsc9 mutant embryos develop normally, with no appreciable defects in embryonic patterning or morphogenesis. However, maternal zygotic ptk7^hsc9 mutant embryos display axial C&E defects, abnormal neural tube morphogenesis, and loss of PCP. Our results provide further evidence that Ptk7 is an essential regulator of vertebrate PCP. More importantly, we demonstrate that Ptk7 enhances the activity of exogenous Wnt5 and Wnt11 ligands, indicating that Ptk7 regulates PCP by potentiating non-canonical Wnt signal transduction.

Although the molecular mechanisms by which Ptk7 promotes PCP remain ill defined, overexpression studies in Xenopus have implicated a key role for the highly conserved Ptk7 pseudokinase domain: it interacts with receptor of activated protein kinase C 1 [RACK1, a known effector of Dvl required for Xenopus neural tube closure (Wehner et al., 2011; Kinoshita et al., 2003)], and it is required for Ptk7-mediated translocation (and subsequent hyper-phosphorylation) of Dvl to the plasma membrane (Shnitsar and Borchers, 2008). In this study, we used an in vivo rescue-based strategy to perform structure-function analyses of Ptk7. Strikingly, we found that full-length Ptk7 and Ptk7ΔICD constructs have equivalent abilities to rescue MZptk7^hsc9 C&E defects, whereas deletion of the Ptk7 extracellular domain eliminated protein function. Therefore, the kinase domain is dispensable for Ptk7 activity in PCP-directed C&E. Rather, our data indicates that a plasma membrane-tethered Ptk7 extracellular domain (Ptk7ΔICD) is sufficient to promote normal PCP, and highlights the importance of extracellular protein-protein interactions for Ptk7 function.

Our findings complement recent studies of mammalian Ptk7, which demonstrate that Ptk7 is not required for Dvl2 membrane localisation in planar polarised cells of the mouse mesoderm or auditory epithelium (Yen et al., 2009; Lee et al., 2012). Therefore, functions for the Ptk7 kinase domain in Dvl localisation and PCP might have evolved differently in Xenopus. Alternatively, Ptk7 might have context-dependent PCP functions. The specific role for Ptk7 intracellular domains in oriented cell division (Gong et al., 2004), radial intercalation (Roszko et al., 2009), cilia orientation (Borovina et al., 2010; Happé et al., 2011), and facial branchiomotor neuron migration (Jessen et al., 2002) remain to be determined.

We have demonstrated that overexpression of zebrafish Ptk7 disrupts dorsoventral patterning, inhibits Wnt/β-catenin-dependent gene expression in early gastrula-stage embryos and inhibits the effects of wnt8 overexpression. Furthermore, loss of Ptk7 activity in MZptk7^hsc9 mutant embryos results in increased Wnt/β-catenin target gene transcription at early segmentation stages. Ptk7 is strongly expressed in the tailbud, where Wnt/β-catenin signals specify paraxial mesoderm cell fate from bi-potential neural/mesodermal progenitor cell pools (Martin and Kimelman, 2012). In MZptk7^hsc9 embryos, we observe an expanded domain of pre-somitic mesoderm differentiation that is independent of PCP-mediated morphogenesis defects. Our data therefore indicate that Ptk7 attenuates Wnt/β-catenin signalling during vertebrate embryogenesis, with essential roles in regulating Wnt/β-catenin-dependent stem cell fate decisions in the developing tailbud.

Wnt/β-catenin patterning defects have not been reported in three independent analyses of Ptk7 mutant mice (Lu et al., 2004; Yen et al., 2009; Paudyal et al., 2010). However, these studies focused predominantly on PCP-related phenotypes. Of note, Paudyal et al. (Paudyal et al., 2010) did report a shortened spinal cord and body axis in chuzoi Ptk7 mutant mouse embryos, which would be consistent with misregulated Wnt/β-catenin signalling during posterior paraxial mesoderm differentiation and tailbud outgrowth.

In Xenopus, studies of Ptk7 function using antisense MO methodologies have yielded conflicting results. Peradziryi et al. (Peradziryi et al., 2011) demonstrated that knockdown of Ptk7 in animal cap explants stimulated β-catenin target gene transcription in response to exogenous Wnt signals. This study supports our findings that Ptk7 inhibits Wnt/β-catenin signalling in vivo; however, endogenous Wnt/β-catenin patterning defects were not reported for Xenopus PTK7 MO-injected embryos. In a second study, MO knockdown of PTK7 activity caused defects in Spemann organiser formation, suggesting that Ptk7 potentiates canonical Wnt signalling (Puppo et al., 2011). Although this finding is difficult to reconcile with our analysis of MZptk7^hsc9 mutant embryos, PCP factors such as Vangl2 have been implicated in the asymmetric distribution of maternal transcripts within Xenopus oocytes, which is required for dorsoventral pattern formation (Cha et al., 2011). As Ptk7 is a well-established regulator of PCP, defects observed in the latter study may reflect a similar requirement for maternal Ptk7 in Xenopus oocyte polarity.

Of interest, despite an early expansion of Wnt/β-catenin-dependent dorsal organiser gene expression in MZptk7^hsc9 mutant embryos, we did not observe later dorsoventral or anterior CNS patterning defects. As Wnt/β-catenin signals also restrict dorsal organiser activity in later-staged blastula, it is possible that potentiation of ventralising Wnt/β-catenin signals in MZptk7^hsc9 mutants might compensate for early expansion of the dorsal organiser. Alternatively, there may be functional redundancy between Ptk7 and other Wnt/β-catenin inhibitors (van de Water et al., 2001; Kagermeier-Schenk et al., 2011; Xie et al., 2011). For example, the transmembrane protein Waif1 has recently been shown to attenuate Wnt/β-catenin signalling while promoting PCP (Kagermeier-Schenk et al., 2011). Zebrafish waif1 (also known as tpbg) is broadly expressed during gastrulation but becomes restricted to the neuroectoderm during segmentation (Kagermeier-Schenk et al., 2011) and these domains overlap with ptk7 expression.

We have also demonstrated that the Ptk7 pseudokinase domain is dispensable for Wnt/β-catenin signal inhibition, as assayed by rescue of MZptk7^hsc9 tailbud patterning defects. Rather, a plasma membrane-tethered Ptk7 extracellular domain is sufficient for this activity. Our results contrast recent work by Puppo et al. (Puppo et al., 2011), which suggests that physical interactions between the pseudokinase domain and intracellular β-catenin might provide a mechanistic link between Ptk7 and canonical Wnt signal transduction pathways. However, Ptk7 can form a complex with Fz (Shnitsar and Borchers, 2008; Peradziryi et al., 2011), and the extracellular domain of Ptk7 can selectively bind canonical ligands Wnt3a and Wnt8 (Peradziryi et al., 2011). This suggests that Ptk7 may function as a canonical Wnt co-receptor to affect pathway specificity in vivo. Our work highlights the importance of the Ptk7 extracellular domain in both Wnt/β-catenin and non-canonical Wnt/PCP signal transduction pathways during vertebrate embryonic development.

Wnt/β-catenin signalling plays an oncogenic role in the development of colon cancer, hepatocellular carcinoma and multiple other tumours (Clevers, 2006). By contrast, non-canonical Wnt pathways antagonise Wnt/β-catenin signalling, and have been speculated to drive morphogenetic cell behaviours associated with tumour invasion and metastasis - pathogenic events that account for ∼90% of cancer deaths (Wang, 2009). Our data demonstrate that Ptk7 not only promotes PCP, but also inhibits Wnt/β-catenin signalling during vertebrate development. Therefore, our results could have significant implications for Ptk7 misregulation observed in colon carcinoma, AML and multiple other cancer types (Mossie et al., 1995; Easty et al., 1997; Endoh et al., 2004; Müller-Tidow et al., 2004; Gorringe et al., 2005; Golubkov et al., 2010; Prebet et al., 2010; Gobble et al., 2011). For example, during cancer progression Ptk7 could initiate a switch from β-catenin-mediated tumour formation and cell proliferation towards PCP-directed changes in cell adhesion, polarity and motility that are associated with metastasis.

Zygotic ptk7^hsc9 mutant embryos display no obvious defects in embryonic patterning or morphogenesis. Rather, maternal Ptk7 is sufficient to rescue early development. This maternal rescue will now permit us to investigate the role for Ptk7 (and consequently, misregulated Wnt/β-catenin and PCP signalling) in juvenile growth, patterning and adult physiology. For example, the late onset of axial deformity observed in zygotic ptk7^hsc9 mutants suggests that they occur independently of early tailbud patterning defects. Indeed, aberrant activation of Wnt/β-catenin signalling has also been implicated in a wide variety of human pathologies (Clevers and Nusse, 2012). Zygotic ptk7^hsc9 mutant zebrafish may therefore prove to be a useful model for determining the consequences of abnormal Wnt signal transduction in adult/late-onset disease.

We thank Drs Scot A. Wolfe, Nathan D. Lawson and Keith Joung, the Howard Hughes Medical Institute and Massachusetts Institute of Technology for reagents utilized in the generation of zebrafish ptk7 ZFNs.