receptor tyrosine phosphatase ψ is required for Delta/Notch signalling and cyclic gene expression in the presomitic mesoderm

The body plan of most higher organisms is made up of serially repeated elements, or segments. In vertebrate embryos, the most obvious metameric structures are the somites. They constitute the basis of the segmental pattern of the body, give rise to the axial skeleton and the muscles and dermis of the trunk, and impose segmentation on the vascular and peripheral nervous system. Somites are formed sequentially from the presomitic mesoderm at a rate that is species-specific (e.g. every 30 minutes in zebrafish and every 90-120 minutes in chick and mouse).

The periodic production of somites along the anteroposterior axis of the vertebrate body involves a molecular oscillator, the `segmentation clock',which can be visualised through the cyclic activation of a small set of regulatory genes (for a review, see Maroto and Pourquié, 2001). These oscillations result in dynamic wave-like domains that sweep across the presomitic mesoderm (PSM) in a posterior-to-anterior direction, narrowing as they approach its anterior end. The oscillation becomes arrested in each cell as it passes from the presomitic to the somitic region of the mesoderm. One temporal oscillation occurs in the PSM for each somite that is formed, and mutations or treatments that perturb oscillatory gene expression also disrupt segmentation(Evrard et al., 1998; Henry et al., 2002; Holley et al., 2000; Hrabe Angelis et al., 1997; Jiang et al., 2000; Kusumi et al., 1998; Oates and Ho, 2002; Zhang and Gridley, 1998).

In each cycle, these cycling genes are first expressed in the tailbud, and expression is subsequently propagated through the posterior PSM. When it reaches the anterior PSM, it becomes stabilized, and is localized to either the rostral or caudal part of the future somite. Based on these observations and others, the PSM has been subdivided into three different regions in which the oscillator responds to different regulatory cues: the posterior undetermined zone, the anterior committed zone (within which cycling is still seen) and a differentiating anterior most zone, within which somite boundaries and compartments are established (Gajewski et al., 2003; Morales et al.,2002; Saga and Takeda,2001).

Most oscillatory genes are related to Notch signalling and dependent on Notch signalling for their cyclic expression. In the mouse and chick, these include lunatic fringe (lfng), which modulates the efficiency of Notch signalling (Aulehla and Johnson, 1999; Forsberg et al., 1998; McGrew et al.,1998), and various hairy-related genes [hairy1,hairy2 and Hey/Hesr/HRT2 in chick(Jouve et al., 2000; Leimeister et al., 2000; Palmeirim et al., 1997); Hes1, Hes7 and Hey1 in mouse (Bessho et al., 2001a; Jouve et al., 2000; Nakagawa et al., 1999)] that are transcriptional targets of Notch signalling and encode basic helix-loop-helix (bHLH) repressor proteins. In the zebrafish PSM, three genes have so far been shown to have cyclic expression: the Notch ligand deltaC (Jiang et al.,2000) and the hairy-related genes, her1 and her7 (Henry et al.,2002; Holley et al.,2000; Oates and Ho,2002). Mutations in these cycling genes and other Delta/Notch components result in defective somite segmentation: intersomitic clefts fail to form or are late and irregular. In zebrafish, her1 and her7 appear to cross regulate each other, and it has been proposed that a negative feedback loop involving these genes constitutes the oscillator(Henry et al., 2002; Holley et al., 2002; Lewis, 2003; Oates and Ho, 2002).

In this study we present a novel regulator in the control of the somitogenesis clock, RPTPψ, a member of the type IIB family of receptor tyrosine phosphatase. We describe the cloning of zebrafish RPTPψ and its expression pattern during early zebrafish development, and provide evidence that RPTPψ is required for normal oscillatory gene expression in the PSM. We show that RPTPψbehaves as a positive regulator of her1 and her7 expression,acting either upstream of or in parallel with Delta/Notch signalling. We also find that RPTPψ is required for convergent extension, a process of cell-rearrangement during gastrulation, raising the possibility that RPTPψ functions in Notch and Wnt signalling.

Zebrafish embryos were obtained by natural spawnings and maintained at 28.5°C in system water. Embryos were fixed in 4% paraformaldehyde using an automated device. aei^tr233 was used to study embryos mutant in Notch signalling (van Eeden et al., 1996; Jiang et al.,1996). Embryos were staged according to Kimmel et al.(Kimmel et al., 1995).

A chick RPTPψ cDNA was used to screen a zebrafishλZapII cDNA library (Haddon et al.,1998) from which several positive cDNA clones were isolated, of which the longest clone (clone 21) spanned sequence nucleotides 1621-4565.

The missing 5′ sequence was obtained by reverse transcription-PCR from 24 hpf embryo total RNA by using the 5′/3′ RACE kit(Boehringer) according to the manufacturer's protocol. Specific primers used for 5′-RACE were: antisense 5′-RACE-A1,5′-CCTTCTTGCCCTCGGTGTTGGCGAG-3′ and antisense nested 5′-RACE-A2, 5′-CTCCTCAGTCTGAAACATGACCTCC-3′. The full-length sequence of RPTPψ was deposited in the GenBank database under the Accession Number AY555586.

Whole-mount in situ hybridisation was performed essentially as previously described (Haddon et al.,1998). For all experiments using multiple genotypes, hybridisation was carried out in parallel and colour development allowed to run for the same amount of time. The embryos were photographed using a Leica DC500 camera. Digoxigenin-labelled RNA antisense probes were generated with a Stratagene RNA transcription kit. Enzymes for linearization and transcription for probe synthesis were as follows: RPTPψ, EcoRI/T7; deltaC,XbaI/T7; her1, XhoI/T3; her7, SpeI/T7;mespa, EcoRI/T7; mespb, HindIII/T3; papC, ApaI/T3;fgf8, EcoRV/SP6; ntl, HindIIII/T7; spt, EcoRI/T7;dlx3, EcoRI/T7; hgg1, XhoI/T3.

Morpholinos (Genetools) were designed with sequences complementary to RPTPψ cDNA in a location just upstream or covering the initiating start codon based on the company's recommendations. The morpholino sequences were: RPTPmo1, 5′-CGCAGGTATTCATTTTCCGTTGTTA-3′; RPTPmo2, 5′-GTTGGGAAAACAAGTCGAAATCATT-3′; 5-m (5-mispair control oligonucleotide to RPTPmo1),5′-CGgAGcTATTgATTTTCCcTTcTTA-3′; her1mo,5′-CGACTTGCCATTTTTGGAGTAACCA-3′. Morpholinos were solubilised and diluted as described by Nasevicius and Ekker(Nasevicius and Ekker, 2000)and injected into one- or two-cell stage embryos at a total amount of 1-8 ng/embryo.

To test the specificity and efficiency of the RPTPψmorpholinos in knocking down the respective protein, we used in vitro transcription and translation of RPTPψ (TNT Coupled Reticulocyte Lysate System, Promega) performed according to the manufacturer's protocol with the following modifications: in a 25 μl reaction, 0.5 μg of RPTPψ cDNA and various amounts of morpholino antisense oligos(25-250 nM) were added to the TNT mix, containing all of the required components for in vitro transcription and translation, and incubated at 30°C for 90 minutes. Five microlitres from the reaction mix were resolved by SDS/PAGE (NuPAGETM, 4-12% Bis-Tris Gel; Invitrogen), and ³⁵S-labeled proteins were visualised by autoradiography.

We have previously described the cloning and expression of a chick gene encoding receptor protein tyrosine phosphatase ψ (RPTPψ)(Aerne et al., 2003). We showed that chick RPTPψ is expressed uniformly throughout the PSM and in a dynamic fashion in nascent somites, consistent with a potential role in somitogenesis (Aerne et al.,2003).

To analyse the molecular function of RPTPψ during somitogenesis, we used zebrafish, owing to the accessibility of its embryos and the ease of its genetic manipulations. A partial zebrafish cDNA clone was obtained by screening a zebrafish cDNA library with a chick RPTPψprobe under low stringency. The missing 5′ end was obtained by 5′RACE (see Materials and methods).

The predicted RPTP protein consists of a 740 amino acid extracellular region, a single transmembrane domain and a 666 amino acid intracellular region. The extracellular sequence contains a MAM (meprin/A5/PTPμ) domain,an immunoglobulin-like domain and four fibronectin type III-like repeats,characteristics of members of the RPTP type IIB family (or MAM domain subfamily) of receptor tyrosine phosphatases (for a review, see Stoker and Dutta, 1998). Comparison of the derived amino acid sequence with other vertebrate receptor tyrosine phosphatases clearly identifies the full-length clone as zebrafish RPTPψ, showing 73-78% homology to human, mouse and chick RPTPψ(Aerne et al., 2003; Wang et al., 1996; Yoneya et al., 1997). Fig. 1A shows a schematic representation of the zebrafish RPTPψ protein domains.

We determined the sites of RPTPψ expression during early zebrafish development by in situ hybridisation(Fig. 1B). Low level RPTPψ expression is seen throughout the embryo during the first day of development. At later stages (from 10-24 hours), RPTPψ is transcribed at slightly increased levels in the somites, and in the pronephric duct, the midbrain hindbrain boundary, the otic vesicle and the retina. Beyond 26 hours post-fertilisation, when somite formation is complete, RPTPψ is no longer expressed throughout the embryo but, instead,becomes restricted to the retina, the forebrain-midbrain, the midbrain hindbrain boundary, the otic vesicle and the branchial arches.

To examine the effects of reduced RPTPψ activity on segmentation, we used two anti-RPTPψ morpholinos (RPTPmo1 or RPTPmo2) targeted to independent regions of the 5′ end of the RPTPψ mRNA (Fig. 2A). Antisense morpholino oligos are specific inhibitors of translation that act by binding to complementary sequences on mRNA and inhibiting ribosome access (Nasevicius and Ekker, 2000; Summerton and Weller, 1997). In the absence of a specific antibody that recognises the RPTPψ protein, we tested the potency and specificity of RPTPψ morpholinos in an in vitro transcription and translation system. Each RPTPψ morpholino oligo inhibits protein translation in a dose-dependent manner (Fig. 2B, lanes 5-8). Inhibition by unrelated or mispaired control morpholinos is negligible, even at 250 nM(Fig. 2B, lanes 3,4). These data suggest that morpholino treatment significantly and specifically reduces RPTPψ protein levels. In the experiments described below, the phenotypic effects of RPTPmo2 were indistinguishable from those of RPTPmo1, whereas the mispaired control oligonucleotide did not produce any phenotype.

Injection of RPTPψ morpholinos into the one- or two-cell zebrafish embryo results in severe disruption of the segmental pattern of the embryo. The first few somites are relatively normal, but subsequent somite boundaries are indistinct and irregular, like those in embryos mutant for Delta/Notch signalling (Fig. 3;data not shown). The expression pattern of the somite mesodermal marker myod in RPTPmo embryos reveals a highly penetrant loss of boundary integrity in the disrupted region and a variation in apparent segment size throughout the paraxial mesoderm (Fig. 3B). The number of segments affected and the frequency and severity of boundary defects is dependent on the concentration of injected oligonucleotide. In the extreme, somites are completely lost(Fig. 3B).

We also see a slight shortening of the body axis and broadening of the notochord, suggestive of a disruption of convergent extension movements during gastrulation (Fig. 3A,B, and see later). RPTPmo embryos show neuronal degeneration from the first day of development, with cell death occurring mainly in the brain area (data not shown). RPTPmo embryos die 2-3 days post-fertilisation. Presumably, lethality results from requirements in the later expression domains.

The disruption of somitogenesis observed in the RPTPmo-injected embryos could be due to interference with specification and maturation of the PSM. Alternatively, processes during somite patterning, such as the establishment of segment polarity or the timing and maintenance of the somite oscillator, could be defective. To exclude some of these possibilities, we analysed the integrity of the presomitic mesoderm by examining markers for paraxial mesoderm formation (spadetail; spt) and maturation(fgf8).

spt is required for the convergence of mesodermal cells towards the dorsal side during gastrulation, and in the specification of cardiac and presomitic mesoderm (Amacher et al.,2002; Griffin and Kimelman,2002). Once cells of the paraxial mesoderm are formed, they undergo a maturation process, which is determined by a gradient of fgf8, with high levels in the posterior and low levels in the anterior presomitic mesoderm. When fgf8 levels drop below a threshold level, the segmentation clock slows down and somitogenesis is initiated(Dubrulle et al., 2001; Dubrulle and Pourquié,2004; Sawada et al.,2001). In wild-type embryos, spt is expressed strongly in adaxial and tailbud cells, and more weakly in presomitic and lateral mesoderm cells. In RPTPmo embryos, the levels and pattern of sptexpression appear normal (Fig. 4), indicating that RPTPψ is not required for specification of presomitic mesoderm tissue. Similarly, the gradient and level of fgf8 expression is not affected by morpholino treatment(Fig. 4), arguing that the disrupted segmentation seen in RPTPmo embryos is not due to impaired mesoderm maturation.

Somite boundary formation depends on polarisation of presomites into anterior and posterior compartments. To test if this regionalisation is affected by reduction in RPTPψ function, we assayed the expression patterns of markers of rostral and caudal half-segment identity.

Zebrafish papC (pcdh8 – Zebrafish Information Network), a rostral segment polarity marker, is expressed during segmentation in four bilateral pairs of bands in the anterior paraxial mesoderm, and more weakly and uniformly in the rest of the PSM(Yamamoto et al., 1998). The anteriormost bands are located at the anterior borders of the newest somite formed (SI) and the forming somite (S0). Stronger, posterior bands are located in successively less mature somite primordia (S-1, S-2)(Fig. 5A). papCexpression in RPTPmo-injected embryos is very similar to that in somite mutants of the Delta/Notch signalling pathway [e.g. after eight (aei), a mutation in deltaD(Jiang et al., 2000); Fig. 5D,G]. Expression is strong but non-metameric in the region corresponding to newly formed and nascent somites, with marked random variability of intensity from cell to cell. Expression in the rest of the PSM is normal and diffuse.

papC transcription is dependent on Mesp genes, mespa and mespb, which code for bHLH transcription factors involved in anteroposterior specification within the presumptive somites(Sawada et al., 2000). In wild-type embryos, mespa and mespb are segmentally expressed in one to three stripes at the anterior of the PSM, each corresponding to one band of papC expression (Fig. 5B,C). mespb expression in treated embryos is reduced to a single, broad domain resembling the expression of mespb in aei mutants (Fig. 5C,F,I). mespa expression is completely lost in the morphant embryos, an extreme version of the aei phenotype in which expression is very weak but still detectable(Fig. 5B,E,H).

Reduced RPTPψ activity also disrupts expression of markers of caudal half-segments, such as myod and deltaC(Fig. 3B, Fig. 8F). Together, these results show that RPTPψ is required for the specification of anteroposterior polarity within somites.

Somite compartmentalisation and boundary formation depend on the segmentation oscillator. To analyse whether the somitic defects observed in RPTPψ mutants derive from a defective segmentation clock, we analysed oscillator behaviour in RPTPmo-injected embryos.

In wild-type embryos, cycling genes show dynamic patterns of expression in the PSM, except at its anterior where somites are formed and expression becomes stable and compartment specific. At high doses of injected morpholino oligonucleotide, cyclic expression in the PSM is lost: expression of deltaC, her1 and her7 is no longer dynamic, and only one static pattern is observed (Fig. 6). For deltaC, expression in these embryos is moderate in the posterior part of the PSM, relatively low in the middle part and high in the anterior part of the PSM. her1 and her7, however,show uniform expression throughout the PSM(Fig. 6). Thus, dynamic expression of all known cyclic zebrafish genes is disrupted in the treated embryos, indicating that RPTPψ is directly involved in the operation of the somitogenesis clock.

The anterior and posterior PSM differ in their threshold requirements for RPTPψ. At lower morpholino doses, cycling continues in the posterior PSM but the normally sharp anterior boundaries of the deltaC domains become diffuse. The most anterior stripe becomes weaker and less distinct as interstripe cells begin to express deltaC(Fig. 6). A similar dose effect can be seen for her1 and her7 expression(Fig. 6). Thus, dynamic expression of cycling genes in the anterior PSM is more sensitive to changes in RPTPψ levels, consistent with domain-specific regulation of cycling genes (Gajewski et al., 2003; Morales et al., 2002; Saga and Takeda, 2001).

To consider how RPTPψ affects the segmentation clock, we analysed cyclic gene expression in RPTPmo embryos in more detail and compared it with that in other known `clock arrested' embryos, e.g. aei mutant and her1 morphant embryos(Fig. 7A).

We considered, in particular, the posterior PSM, where clock circuitry is not yet affected by differentiation. her1 and her7 are downregulated and non-dynamic in RPTPmo embryos, as is deltaC expression. As Her proteins are repressors, it seems unlikely that their lowered levels are directly responsible for reduced deltaCexpression. Similar effects are seen in aei embryos, suggesting that RPTPψ is required to promote Notch signalling(Fig. 7A). Indeed, injecting RPTPmo oligonucleotides does not exacerbate the aeisegmentation phenotype (Fig. 7B), arguing that their major effect is via Notch signalling.

In her1mo embryos, levels of her7 expression are reduced in the cycling PSM, although not as drastically as in RPTPmo embryos(Fig. 7A). her1levels, on the other hand, are greatly increased(Fig. 7A), but this is probably due to transcript stabilisation by the antisense oligonucleotide(Gajewski et al., 2003; Oates and Ho, 2002). Indeed,this increase is abolished by co-injection of RPTPmo, such that the embryos resemble those injected with RPTPmo alone(Fig. 7A). These results indicate that RPTPψ activity is needed for efficient transcription of both her1 and her7, and suggest that RPTPψ acts upstream or in parallel to Delta/Notch signalling.

The situation in the more anterior PSM, where somite differentiation and boundary formation would normally occur, is rather more complex. There, RPTPψ activity is again needed for expression of her1and her7, but deltaC expression is broadened into a single,anterior stripe (Fig. 6, Fig. 7A). Thus, RPTPψ is required for final, transient repression of deltaC prior to its compartment-specific expression and formation of the somite boundary.

Reduction of RPTPψ activity also results in shortened anteroposterior and broadened mediolateral axes (Fig. 3A,B, Fig. 8). This phenotype is characteristic of a failure of convergent extension, a process of cell polarisation and intercalation that leads to lengthening and narrowing of the embryonic body during gastrulation. An alteration in axial proportions is confirmed by staining for distal-less3 (dlx3) and no tail (ntl), which mark the boundaries of the neuroectoderm and nascent notochord, respectively (Akimenko et al., 1994; Schulte-Merker et al., 1992). These markers reveal that the neural plate in RPTPmo embryos is broader and shorter, and that the notochord is wider and slightly undulated (Fig. 8A,B,D,E).

In addition, staining for a marker for the anteriormost prechordal plate, hgg1 (hatching gland gene 1; ctlsb –Zebrafish Information Network) (Vogel and Gerster, 1997) reveals that anterior migration of the prechordal mesoderm is impaired in the treated embryos. hgg1 expression normally lies rostral to dlx3, in the periphery of the neural plate; in RPTPψ morphants, hgg1 is located more caudally,overlapping the edge of the broader neural plate(Fig. 8B,E). Impairment of convergence movements is further indicated by the presence of laterally widened somites as shown by deltaC expression in RPTPmo-injected embryos compared to the wild-type control(Fig. 8C,F).

Defects in convergent extension might directly explain the lack of dynamic expression in RPTPmo-treated embryos, e.g. if the segmentation clock depends on novel neighbourhood relationships arising during cell intercalation. This explanation seems unlikely because segmentation seems to be more sensitive than convergent extension to reduced RPTPψ activity,judged by the different levels of RPTPmo required to observe the described phenotypes (Fig. 6). Nevertheless, we tested this idea by examining cyclic gene expression in embryos mutant for knypek (kny), which encodes a glypican that promotes non-canonical Wnt signalling during convergent extension(Topczewski et al., 2001). The body axis is shortened in kny mutant embryos, but segmentation appears normal, and expression of cyclic genes (e.g. deltaC) is still dynamic (Fig. 8G), indicating that non-canonical Wnt signalling is not required for periodic gene expression in the PSM.

RPTPs play a significant role in antagonising the activities of protein tyrosine kinases, thereby limiting the amplitude and duration of signalling. Several such phosphatases have been shown to play a role in embryogenesis, in particular in the developing nervous system, but also during gastrulation and hematopoiesis (den Hertog,1999; Stoker and Dutta,1998; Van Vactor,1998). However, so far, little is known about their biochemical function, ligands or downstream signalling pathways. We have described the zebrafish RPTPψ gene, and analysed its expression and activity in early embryos. We show that embryos with reduced RPTPψ activity lack oscillatory gene expression in the PSM, and that their neural plates and PSMs are shortened and widened. We discuss these results in terms of requirements for RPTP activity in segmentation and convergent extension.

Our experiments show that antisense-mediated reduction of RPTPψactivity leads to the loss of oscillatory behaviour of cycling genes and to severe reduction of both her1 and her7 transcription. deltaC is also downregulated in the posterior PSM of RPTPmoembryos. Thus, RPTPψ appears to be required for effective Notch signalling in the PSM. All these expression patterns resemble those in zebrafish embryos defective for Delta-Notch signalling, consistent with RPTPψ acting upstream of, or in parallel with, this pathway.

her1 and her7 code for bHLH transcriptional repressors of the Hairy/E(spl) family, genes encoding which are directly activated by Notch signalling in a variety of developmental contexts, including segmentation(Oates and Ho, 2002; Takke et al., 1999). Hes1 and Hes7, her1 and her7counterparts in mouse, negatively regulate their own expression both in cultured cells and in vivo (Bessho et al.,2003; Hirata et al.,2002). Based on this observation, Lewis showed, using mathematical modelling, that an auto-regulatory feedback loop involving her1 and her7 provide a possible molecular basis for an intracellular oscillator (Lewis, 2003).

Reducing RPTPψ levels decreases her1/7 expression in the cycling PSM, and this reduction is independent of her1/7 activity(Fig. 6, Fig. 7A). Therefore, RPTPψ appears to be required to activate Notch target gene transcription during cycling. This effect appears to be independent of effects on Notch ligand expression because RPTPmo also reduces her1expression in aei embryos (Fig. 7B). Overall levels of deltaC, her1 and her7,and also mesp gene expression are much reduced in both the cycling and anterior PSM (Fig. 5, Fig. 6, Fig. 7A,B). However, RPTPψis also needed for repression of deltaC during somite boundary formation – two anterior stripes in wild-type embryos become a single,broad stripe – perhaps because of the failure of anterior herexpression. This latter, indirect requirement for RPTPψ may reflect differing regulatory circuits operating in different regions of the posterior PSM (Gajewski et al., 2003; Morales et al., 2002; Saga and Takeda, 2001).

Nevertheless, it is still not clear to what extent Delta-Notch signalling is required for the oscillation itself. For example, the first few somites are still formed in zebrafish and mouse embryos defective in Notch signalling. One possibility is that Notch signalling is required only to synchronise neighbouring PSM cells (Jiang et al.,2000), and that an upstream segmentation clock drives cyclic Notch signalling.

One pathway that could account for the latter possibility is that of Wnt signalling. Aulehla et al. (Aulehla et al.,2003) showed recently that axin2, which encodes a negative regulator of Wnt signalling, displays oscillating expression in the mouse PSM, alternating with that of lfng and hes7. They argue that wnt3a is necessary for cyclic expression of both axin2 and the oscillating Notch signalling activity, but that Notch signalling is not required for axin2 oscillation. This implies that axin2 oscillations reflect cyclic Wnt signalling that is distinct from, and possibly upstream of, cyclic Notch signalling in the mouse PSM.

It is not yet clear if cyclic Notch signalling in zebrafish embryos is driven by an upstream Wnt clock. axin2 expression appears not to cycle in the zebrafish PSM (B.A., unpublished), although Wnt signalling components other than axin2 could be cycling in zebrafish and thereby generate cyclic Wnt activity. In any case, RPTPψ, like Wnt3a in the mouse,is required for Delta/Notch signalling.

In addition to its role in segmentation, RPTPψ seems to be required for convergent extension during gastrulation. RPTPmo embryos have a shorter and broader body axis, a phenotype characteristic of convergent extension mutants.

Convergent extension has been shown to depend on the so-called,non-canonical Wnt signalling pathway (for a review, see Tada et al., 2002). Unlike canonical Wnt signalling, which targets the nucleus and directs changes in gene transcription, non-canonical Wnt signalling is independent ofβ-catenin-mediated transcriptional activity, and directs morphogenetic processes such as changes in cell shape and cell migration. How Wnt signalling is translated into convergent extension movements during gastrulation is poorly understood, but it clearly involves changes in the adhesive properties of cells, e.g. via regulated decreases in the activity of cell adhesion molecules such as cadherins (Kuhl et al.,1996; Marsden and deSimone,2003). Non-canonical Wnt signalling seems not to be required for the segmentation clock, as we have shown that oscillator behaviour is normal in kny mutants (Fig. 8G). Similarly, no role for Notch signalling in convergent extension is known.

How might the dual effect of RPTPψ on the somite oscillator and convergent extension be explained? The multiplicity of kinases in the vertebrate genome implies that PTPs have a relatively broad range of substrate specificities. One possibility, therefore, is that RPTPψ affects factors from independent pathways (e.g. Wnt and Notch) that regulate convergent extension and somitogenesis. Alternatively, RPTPψ might affect a single pathway/component that impinges on both convergent extension and somitogenesis. Human and mouse RPTPψ have been shown to associate withβ-catenin and to dephosphorylate β-catenin both in vivo and in vitro(Cheng et al., 1997; Wang et al., 1996; Yan et al., 2002). Both these processes could be modulated by RPTPψ, e.g. by acting on tyrosine phosphorylation levels of β-catenin, which is crucial for both instability of the β-catenin/cadherin bond and for enhanced binding to TBP and the Tcf complex (Piedra et al.,2001; Roura et al.,1999). Thus, RPTPψ has the potential to promote adhesion and negatively regulate β-catenin-dependent transcriptional activity(Balsamo et al., 1996; Balsamo et al., 1998). It is therefore possible that changes in RPTPΨ activity impinges both on adhesion and migration processes during convergent extension movements, and on Wnt-directed transcriptional regulation of the somite oscillator.

Clearly, further experiments are needed to pinpoint the targets of RPTPψ activity in both processes. In any case, our study adds an unexpected and novel component to the somitogenesis clock, which, until recently, exclusively implicated members of the Delta/Notch signalling pathway.

We acknowledge members of the Ish-Horowicz and Lewis laboratories for discussion, in particular Francois Giudicelli and Julian Lewis for their comments on the manuscript. We are especially grateful to Phil Taylor for excellent fish-keeping. B.A. was supported by a fellowship from the Swiss National Science Foundation.