Feedback interactions between MKP3 and ERK MAP kinase control scleraxis expression and the specification of rib progenitors in the developing chick somite

During vertebrate embryogenesis, epithelial somites are generated sequentially from pre-segmented paraxial mesoderm (psm). In amniotes, somites are patterned by signals from the notochord, neural tube, surface ectoderm and lateral plate to give rise to the dermomyotome and the sclerotome. The definitive myotome forms from the dermomyotome(Buckingham, 2001).

Signalling between different somite compartments results in further patterning and it has been shown that growth factors from the myotome are important for the specification of sclerotome derived cell lineages, including ribs and tendons (Brent et al.,2003; Huang et al.,2003; Tallquist et al.,2000). Detailed clonal analysis of avian somites has revealed the origin of proximal and distal ribs (Evans,2003). Furthermore, ablation experiments suggested that the mesenchymal cells at the rostral and caudal edges close to the dermomyotome,which express the bHLH transcription factor scleraxis, contain distal rib progenitors (Hirao and Aoyama,2004). Scleraxis marks a subpopulation of sclerotomal cells and is involved in regulating gene expression within mesenchymal cell lineages that give rise to cartilage and connective tissues(Cserjesi et al., 1995; Schweitzer et al., 2001). In maturing somites, scleraxis marks axial tendon progenitors in the dorsal sclerotome and it has been postulated that myotome-derived fibroblast growth factor (FGF) signals are required to activate scleraxisexpression by an indirect mechanism (Brent et al., 2003). Recent work has implicated the Ets domain transcription factors Pea3 and Erm in the regulation of scleraxisexpression (Brent and Tabin,2004). This demonstrates that close interactions between the myotome and sclerotome are pivotal for the emergence of a discrete population of scleraxis positive cells within the somite.

We were interested in investigating further the role of the FGF signalling pathway in the specification of scleraxis positive somite progenitors. We have previously identified a role for the dual specificity phosphatase, MKP3, in negatively regulating the ERK MAP kinase pathway in limb and early neural development (Eblaghie et al., 2003) and this new study focuses on the function of this pathway in somite patterning and cell specification.

Mitogen-activated protein kinase (MAP kinase) cascades are effectors for many growth factor signals implicated in developmental processes, including appendage outgrowth and organogenesis. The `classical' Ras/MAP kinase cascade in which extracellular signal regulated kinases (ERK1 and ERK2) are activated by phosphorylation of the T-X-Y motif within the activation loop of the kinase is a major effector of signalling in mammalian cells(Kouhara et al., 1997). The developmental outcome of ERK signalling relies, at least in part, on the competing actions of upstream activators and inhibitory MAP kinase phosphatases (MKPs). Indeed, the level of FGF signalling has been suggested to play a deterministic role in cell fate and survival in a number of different systems (Hajihosseini et al.,2004; Partanen et al.,1998; Storm et al.,2003; Sato and Nakamura,2004; Tsang et al.,2004).

The dual-specificity MAP kinase phosphatase, MKP3 (also known as PYST1), is a specific and potent regulator of the ERK class of MAP kinases(Groom et al., 1996; Muda et al., 1996). This specificity for the ERK 1/2 MAP kinases is mediated by specific protein-protein interaction and subsequent ERK-dependent catalytic activation of MKP3 (Camps et al., 1998; Muda et al., 1998). We have previously isolated the chicken and mouse orthologues of Mkp3 and studied their embryonic expression. We found dynamic patterns of Mkp3messenger RNA expression in important signalling centres and known sites of FGF/FGF receptor signalling, which are associated with cell proliferation and patterning in developing mouse and chick embryos(Dickinson et al., 2002; Eblaghie et al., 2003).

We demonstrate that in differentiating somites Mkp3 is expressed in scleraxis-positive progenitor cells. This expression is regulated by FGF signalling via the classical ERK MAP kinase pathway. During somite patterning, MKP3 regulates ERK MAP kinase activity by dephosphorylation demonstrating that this enzyme is part of a negative feedback loop controlling the levels of phosphorylated ERK (dpERK) in this tissue. Finally, we show that decreasing or increasing the levels of dpERK in chick somites, by mis-expression of human MKP3 or a constitutively active MEK1, respectively,results in loss of Mkp3 and scleraxis expression and affects rib formation. We propose that the tight regulation of dpERK levels is crucial for the activation of scleraxis expression and we show that FGF signalling is important for the specification of distal rib precursors in somites.

Probes for in situ hybridisation were as described previously(Eblaghie et al., 2003) for Mkp3; (Schweitzer et al.,2001) for scleraxis;(Minowada et al., 1999) for sprouty2; (Niswander et al.,1994) for Fgf8 and Fgf4. In situ hybridisation was essentially as described previously(Schmidt et al., 2000);however, proteinase K treatment was omitted. After the staining reaction,embryos were de-stained in high detergent mix, 5×TBST (for 100 ml of a 10 × solution: 8 g NaCl, 25 ml 1 M Tris-HCl pH 7.5, 0.2 g KCl, 10 ml Tween-20) to reduce background.

To detect myosin heavy chain protein with MF-20 antibody (Developmental Studies Hybridoma Bank) and transcripts for Mkp3, sprouty2 or scleraxis, we performed in situ hybridisation with immunohistochemistry in combination. We followed the method described previously (Edom-Vovard et al.,2001). The proteinase K step was omitted and replaced with 0.1%Triton-X 100 15 minutes at room temperature. NBT/BCIP colour reaction was enhanced with 10% polyvinylalcohol (Sigma). Scleraxis transcripts were detected with Fast Red under a coverslip at 37°C. Sections were incubated with MF-20, 1/1000 dilution overnight at 4°C and detected with an `Alexa' secondary fluorescent antibody (488 nm, Molecular Probes).

Heparin beads (Sigma H-5263) were soaked for 1 hour at room temperature in recombinant FGF (R&D Systems) at the following concentrations: FGF4(50μg/ml), FGF8 (1mg/ml) and FGF2 (400μg/ml). AG-1 X2 beads (BioRad)were incubated in one of the following pharmacological inhibitors: SU5402,from 5 to 10 mM; SB203580 and LY294002 both 20 mM, all from Calbiochem; and PD184352, 20 mM (a gift from Philip Cohen, Dundee)(Davies et al., 2000). All compounds were dissolved in DMSO. Beads were soaked for 1 hour at room temperature in the dark then washed twice in PBS and implanted adjacent to cervical somites of HH13 embryos or adjacent to forelimb or flank level somites of HH17 embryos. In addition, PD184352 or DMSO controls were diluted 1/10 with PBS and injected directly into thoracic somites. Embryos were fixed in 4% paraformaldehyde and processed for in situ hybridisation 1, 5 or 24 hours after the operation.

Fertile chicken White Leghorn eggs were obtained from Needle farm (Sussex)and incubated at 37.5°C until the desired Hamburger-Hamilton stage was reached. Electroporation was performed in ovo using a TSS20 Ovodyne Electroporator (Intracel, UK). Expression plasmids for hMKP3-GFP,hMKP3^ΔKIM-GFP, caMEK1 (=MKK^E/E), sFREK:Fc and dnFgfR1c have been described previously(Eblaghie et al., 2003; Marics et al., 2002; Yang et al., 2002; Cowley et al., 1994). Most plasmids encoded GFP fusion proteins or produced GFP from an IRES. Alternatively, a GFP expression plasmid was co-electroporated to mark transfected somites. Positive electrode was platinum and negative electrode was sharpened tungsten wire. Eggs were windowed and black ink was injected underneath the blastoderm to visualise the embryos. DNA (3 mg/ml in water) was injected underneath the myotome of flank somites at HH18-20, electrodes were placed on either side of the embryo and 40 V were applied for 50 mseconds,with 5 pulses spaced 500 mseconds apart. For RCAS-mediated gene expression,concentrated virus was injected into presegmented mesoderm. RCAS-sFREK:Fc has been described previously (Marics et al.,2002). The spread of infection was examined using a gag antisense probe. Eggs were sealed and incubated for the times indicated and processed for in situ hybridisation, Alcian Blue staining, western analysis or RNA extraction.

hMKP3-GFP encodes a fusion protein, caMEK1 was co-electroporated with a GFP expression plasmid. GFP-labelled somites were pooled, protein was extracted using standard protocols (NP-40 lysis buffer with protease and protein phosphatase inhibitors, Roche), equal amounts were loaded on 10%polyacrylamide gels. Primary antibodies were applied overnight at 4°C,excess was washed and secondary antibodies coupled to HRP (Jackson Laboratories) were applied for 1 hour at room temperature. Primary antibodies used: dpERK (Cell Signaling); anti-GFP, (Clontech); and α-tubulin(SIGMA).

RNA was harvested from GFP-expressing somites. cDNA was prepared using random hexamer primers as described previously(Münsterberg et al.,1995). cDNA (1 μl) was used in PCR reactions, human MKP3-GFP(30 cycles), chick Mkp3 (35 cycles), scleraxis (30 cycles) and GAPDH (25 cycles) (Münsterberg et al.,1995). Primers for chick scleraxis: forward,5′-ACGTGAATTCCACACACACCGAACCACGGAC-3′; reverse,5′-ACGTGAGCTCATTATACGAACTGCTCAGGC-3′. Primers specific for human MKP3-GFP fusion: forward, 5′-ACGTCCATGGTAGATACGCTCAGACCCG-3′;reverse, 5′-ACGTAAGCTTTTACTTGTACAGCTCGTCC-3′. Specific primers for chick Mkp3: forward,5′-ACGTGCGGCCGCATGCTAGATACGTTCAGACCCGTC-3′; reverse,5′-ACGTGAATTCTCACGTGGACTGCAGGGAGTCCACC-3′. Specificity of MKP3 primers was tested on 1 ng of plasmid template. Amplification conditions for plasmid and cDNA: denaturation at 94°C for 5 minutes, followed by 30 to 35 cycles of 94°C for 30 seconds, annealing temperatures of 60°C and 55°C for both hMKP3-GFP and cMkp-3 and for scleraxis, respectively for 30 seconds and 72°C for 2 minutes. A 5 μl sample of the 30 μl reactions was analysed on a 1-2% agarose gel.

Whole-mount in situ hybridisation demonstrated that Mkp3 was expressed in a number of regions known for FGF signalling activity, including developing somites. Transcripts were first detected in paraxial mesoderm of Hamburger-Hamilton (HH) stage 15 embryos(Hamburger and Hamilton,1951). High levels of Mkp3 transcripts were expressed in the posterior half of the psm and the tail bud. Expression was diffuse in epithelial somites and then detected in a triangular shaped region abutting the neural tube as somites matured (Fig. 1A,J). As somites continued to differentiate, Mkp3expression became restricted to anterior and posterior somite edges giving rise to a `twin-stripe' pattern (Fig. 1A; Fig. 2G,H). This latter pattern persisted throughout development until at least HH28 (data not shown). A detailed comparison to other known somite differentiation markers revealed that Mkp3 transcripts correlated with transcripts of the bHLH gene scleraxis. Using our protocol, we could detect expression of scleraxis in epithelial somites at HH15(Fig. 1K; red arrow in Fig. 1B) similar to Mkp3 and earlier than reported previously(Brent et al., 2003). In more mature somites, scleraxis transcripts became restricted to anterior and posterior somite edges (Fig. 1B). Another known negative regulator of FGF signalling, sprouty2,was also expressed along the anterior and posterior somite edges giving rise to a broad stripe (Fig. 1C). In addition, sprouty2 was expressed in a thin stripe in the centre of each somite(Fig. 1L). Immunohistochemistry combined with in situ hybridisation on frontal sections confirmed that Mkp3 transcripts were localised in the dorsal sclerotome region where scleraxis transcripts were found(Fig. 1D,E). Two somites contributed to one domain of Mkp3 expression, as shown for scleraxis (Brent et al.,2003). Sprouty2 was also expressed along somite boundaries(Fig. 1F). The somitic expression of Mkp3 was associated with the appearance of Fgf4 and Fgf8 transcripts within the myotome(Fig. 1G,H). We found that the expression of Mkp3 and scleraxis in progenitor cells along the anterior and posterior somite edges was more closely associated with the first appearance of Fgf8 transcripts. Two FGF receptors, CEK1 and FREK are also expressed broadly in developing somites(Fig. 1I; data not shown). This pattern suggested that MKP3 is involved in modulating the FGF mediated regulation of scleraxis and thus could play a role for the specification of scleraxis-positive progenitor cells in the somite.

FGFs have previously been shown to induce ectopic expression of Mkp3 in the neural plate and limb bud mesenchyme(Eblaghie et al., 2003). In order to determine whether FGFs were sufficient to activate Mkp3expression in paraxial mesoderm, we grafted beads soaked in FGFs to the dorsal somite. At stage HH13, FGF4 beads induced very high levels of ectopic Mkp3 transcripts in somites (Fig. 2A). Similarly, FGF2 and FGF8 beads caused ectopic Mkp3expression (data not shown). Transverse sections showed high levels of Mkp3 transcripts throughout somite tissue(Fig. 2B). Beads soaked in FGF10, which belongs to a separate subgroup of FGF ligands based on receptor specificity, did not induce Mkp3 expression. However, FGF10 beads could initiate formation of an ectopic limb bud (data not shown). Experiments using beads soaked in the FGF receptor inhibitor, SU5402, demonstrated that ectopic upregulation of Mkp3 expression required receptor tyrosine kinase activity (Fig. 2C). Furthermore, blocking ERK MAP kinase activity or PI3 kinase activity with PD184352 or LY294002, respectively, prevented induction of ectopic Mkp3 transcripts by FGF beads(Fig. 2D,E). Placing a bead soaked in DMSO next to an FGF4 bead had no effect (not shown).

This showed that FGFs known to act on mesenchymal cells were sufficient to induce ectopic Mkp3 transcripts via either ERK or PI3-kinase. Next,we wanted to investigate whether FGF signalling was necessary to regulate the expression of Mkp3 in the dorsal sclerotome. Application of a bead soaked in SU5402 led to a dramatic loss of Mkp3 transcripts beneath the bead (Fig. 2G). A bead soaked in LY294002 did not affect endogenous Mkp3 expression in contrast to ectopic Mkp3 expression, which was sensitive to PI3-kinase inhibition (Fig. 2I,E). By contrast, a bead soaked in the ERK MAP kinase inhibitor PD184352 did result in a significant loss of endogenous Mkp3transcripts, while DMSO beads or pharmacological inhibitors that blocked other pathways had no effect (Fig. 2H; data not shown). In order to confirm in vivo the effect of PD184352, which inhibits MKK upstream of ERK, the somites beneath the bead were dissected, protein was extracted and analysed by western blot for the presence of dpERK. We observed the complete loss of dpERK in the presence of a PD184352 bead compared with readily detectable levels in untreated somites(Fig. 2F). Injection of PD184352 directly into somites resulted in complete loss of Mkp3transcripts, while injection of DMSO had no effect(Fig. 2J,K). Thus, endogenous expression of Mkp3 was dependent on the presence of active,phosphorylated ERK MAP kinase and the residual expression observed in Fig. 2H is most likely due to the limited diffusibility of PD184352.

In somites, FGFs induce scleraxis, possibly by activating the Ets domain transcription factor Pea3 (Brent and Tabin, 2004). Therefore, we asked whether scleraxisexpression was dependent on dpERK and found that injection of PD184352 resulted in loss of detectable transcripts(Fig. 2L,M). To corroborate these findings, somites of HH18 embryos were electroporated with a plasmid encoding a dominant-negative FGF receptor where the cytoplasmic tyrosine kinase domain had been replaced with EYFP(Fig. 2N). We also used a vector encoding a secreted extracellular domain of FREK (cFGFR4) fused to the immunoglobulin Fc domain, which mediates dimerisation(Fig. 2O). Both of these mutant receptors have previously been shown to inhibit FGF-mediated signalling(Marics et al., 2002; Yang et al., 2002), even though they probably do not discriminate between different FGF receptors and receptor isoforms. Expression of both these constructs by targeted electroporation into somites resulted in loss of Mkp3 expression in this tissue (Fig. 2N,O). By contrast, electroporation of a GFP vector alone had no effect(Fig. 2P). In addition, when we expressed the murine orthologue of sprouty2, which has been shown to inhibit the FGF MAPK pathway at the level of Ras and MEK(Hanafusa et al., 2002; Sasaki et al., 2003), we also observed a loss of Mkp3 expression(Fig. 2Q). Together, these experiments show that somitic expression of Mkp3 depends on FGF receptor-mediated signalling via phosphorylated ERK MAP kinase.

To demonstrate that MKP3 is part of a negative-feedback loop regulating the levels of dpERK in somites, we examined the dynamics of MKP3-dpERK interactions in embryos. We looked at both the induction of Mkp3expression and the levels of dpERK in response to FGF beads over time. Exposure of HH17 somites to FGF2, FGF4 or FGF8 resulted in high-level ectopic expression of Mkp3 within 1 hour, reaching a maximum at 5 hours(Fig. 3A,B; data not shown). Transcripts were seen in all somite regions(Fig. 3F). Interestingly, after an exposure of 24 hours, we noted a clear loss of endogenous Mkp3transcripts in the vicinity of the beads. Ectopic expression immediately around the bead was still detectable, suggesting that active FGF was released from beads at these later time points (Fig. 3C). We next examined how the initial induction and following loss of Mkp3 transcripts correlated with levels of active ERK by western analysis of dissected somites at the same time points(Fig. 3D). After exposure for 1 hour to FGF4, the levels of dual phosphorylated ERK were significantly increased compared with control somites. However, after a 5 hour exposure the levels of dpERK were reduced to the same levels as in control somites, most probably owing to the increased levels of MKP3 protein produced from the ectopic transcripts within the somite at this point. After 24 hours, dpERK was no longer detected, even though the levels in untreated somites had increased. This correlated extremely well with the apparent inhibition of Mkp3expression by FGFs after 24 hours. Interestingly, increased levels of scleraxis transcripts were only detected after 5 hours of exposure to an FGF bead and these were more restricted to the sclerotome compartment(Fig. 3E,G). This is consistent with observations by others (Brent et al.,2003; Brent and Tabin,2004) and in addition indicates differences between Mkp3and scleraxis transcriptional regulation and the competence of cells to express these genes.

In the following experiments, we investigated the function of MKP3 for transcriptional regulation and cell specification in developing somites. We transfected somites with an expression construct encoding a human MKP3 EGFP fusion protein [hMKP3-GFP; previously known as Pyst1-EGFP(Eblaghie et al., 2003)]. Somites were identified by GFP fluorescence, dissected and the levels of dpERK were analysed by western blot. This showed readily detectable levels of dpERK protein in untreated somites or in somites electroporated with empty EGFP vector (Fig. 4A, lanes 1,2). However, hMKP3-GFP expression resulted in a loss of dpERK(Fig. 4A, lane 3). Next, we used human and chick specific primers to determine the levels of Mkp3transcripts in electroporated somites by RT-PCR. Specificity of the two primer pairs was confirmed against plasmid template(Fig. 4B). PCR performed on cDNA obtained from electroporated somites showed that chick Mkp3transcripts were present in somites, which expressed the empty EGFP vector(Fig. 4C, lane 1). However,hMKP3-GFP expression resulted in a complete loss of endogenous chick Mkp3 transcripts. In addition, the levels of scleraxistranscripts were reduced significantly as a result of hMKP3-GFP overexpression(Fig. 4C, lane 2).

Immunohistochemistry on frontal sections of HH24 embryos showed that high levels of dpERK protein were readily detected in the rostral part of each somite in neural crest derived dorsal root ganglia (DRG), consistent with previous reports in mouse embryos (Corson et al., 2003). Phosphorylated ERK MAP kinase was also detected in the region where Mkp3 and scleraxis-positive progenitor cells were detected (Fig. 4D). However, in order to visualise this staining, the contrast had to be increased significantly, indicating that the levels of dpERK along the anterior and posterior edges of the dorsal sclerotome are very low, which is entirely consistent with MKP3 phosphatase activity in these cells.

We next investigated how manipulating the levels of active ERK MAP kinase influenced somite patterning and in particular the specification of scleraxis-positive progenitor cells. Electroporation of the hMKP3-GFP fusion protein into somites of HH18 embryos resulted in high levels of hMKP3-GFP expression throughout the dermomyotome, myotome and dorsal most part of the sclerotome, as visualised by detection of GFP and MyoD in sections (Fig. 4E). In situ hybridisation using the chick Mkp3 probe, which crossreacted with human transcripts, also demonstrated high-level expression(Fig. 4F). In agreement with RT-PCR results, we found that expression of hMKP3-GFP led to a loss of scleraxis expression (Fig. 4G). The same result was obtained after expression of chick Mkp3 from a different expression vector (data not shown). Expression of EGFP alone had no effect on scleraxis(Fig. 6F). To exclude the possibility that expression of the MKP3 phosphatase at high levels could affect non-specifically other phospho-proteins required for scleraxisexpression we used a mutant of MKP3 that does not contain the conserved kinase-interaction motif `KIM' (Nichols et al., 2000). This mutant protein has a normal basal activity in vitro but is unable to undergo catalytic activation in response to ERK2 or to inactivate ERK2 upon transfection into mammalian cells(Karlsson et al., 2004). Electroporation of the KIM mutant did not affect scleraxis expression in any of the embryos (Fig. 4H), even when expressed at high levels(Fig. 4I). These experiments suggest that active ERK MAP kinase, the levels of which are regulated by MKP3,is required for scleraxis expression in somitic progenitor cells.

Next, we wanted to investigate how high levels of dpERK would affect the expression of Mkp3 and scleraxis (Figs 5, 6). We have used a mutant form of MKK in which activating residues have been mutated to create a constitutively active kinase (MKK^E/E=caMEK1)(Cowley et al., 1994). Western blot analysis of somites or neural tube electroporated with a caMEK1 expression plasmid demonstrated the presence of very high levels of dpERK compared with control tissues (Fig. 5B). Ectopic Mkp3 transcripts were detected in the developing neural tube as a result of raised levels of dpERK. In electroporated somites, however, caMEK1 expression resulted in the loss of Mkp3 and scleraxis transcripts(Fig. 5A, Fig. 6D). This may indicate that a particular level of ERK signalling is required to drive transcription of these genes. Alternatively, because electroporation introduced the transgene into the dermomyotome and myotome as well as the dorsal sclerotome(Fig. 4E), it was conceivable that high dpERK activity in these compartments led to the loss of Mkp3 by an indirect mechanism. To investigate this possibility, we used markers for these compartments and found that Pax3 was unaffected, suggesting a normal dermomyotome(Fig. 5C). However, although loss of dpERK as a result of hMKP3-GFP electroporation had no apparent effect on the myotome, as shown by normal MyoD expression(Fig. 5D,E), raising the levels of active ERK MAP kinase by electroporation of caMEK1 caused a loss of MyoD transcripts (Fig. 5F). This was consistent with the idea that a MyoD-dependent signal could synergise with dpERK to regulate Mkp3 and scleraxis. Indeed, we found that Fgf4 expression, which is regulated by MyoD (Iwahori et al.,2004), was affected (Fig. 5H). However, loss of MyoD was also observed after implanting a bead soaked in FGF4 or FGF8 (Fig. 5G, data not shown). In this situation, ectopic Mkp3 and scleraxis transcripts were induced(Fig. 3A,B,E). This finding therefore suggests that the loss of MyoD does not correlate with the loss of Mkp3 and scleraxis.

To investigate the effects of aberrant FGF/ERK MAP kinase signalling on somite differentiation and to examine the long-term consequences for scleraxis-positive progenitor cells, we used RCAS mediated mis-expression of sFREK:Fc, which would antagonise FGF signalling in many somites due to widespread infection (Fig. 6I). Alcian Blue staining revealed the loss of multiple rib skeletal elements on the infected side; however, at fairly low frequency possibly owing to limited diffusion of sFREK:Fc(Fig. 6H). Next, we wanted to affect dpERK levels more directly using MKP3-GFP and caMEK1 viral misexpression. Sections showed that RCAS infection was restricted to cells in the dermomyotome and myotome (Fig. 6J) and thus, mesenchymal cells in the sclerotome were not targeted efficiently. Because hMKP3-GFP and caMEK1 act in a cell-autonomous fashion this was not a feasible approach. Instead, we used electroporation of hMKP3-GFP and caMEK1 expression plasmids(Fig. 4E) and even though effects were restricted to electroporated somites (typically two to three thoracic somites), we saw a dramatic phenotype. As shown before, decreasing dpERK levels by MKP3 resulted in the loss of scleraxis expression(Fig. 4C,G; Fig. 6A). At 10 days of development, distal ribs had failed to form from targeted somites in 60% of the embryos. We electroporated somites that gave rise to the first or the seventh rib (Fig. 6B,C). Increasing dpERK levels by caMEK1 also resulted in the loss of scleraxis expression and affected distal rib formation (60%; Fig. 6D,E). Electroporation of GFP alone did not disturb scleraxis expression or rib formation in any of the embryos analyzed (Fig. 6F,G), indicating that this phenotype is specifically due to perturbation of dpERK levels.

Here, we describe the expression and regulation of the ERK MAP kinase specific phosphatase, MKP3. In addition, we investigate its function in somite patterning and differentiation. Our results suggest that regulation of ERK MAP kinase signal strength by MKP3 establishes a crucial level of FGF signalling important for appropriate expression of the effector gene scleraxis. Scleraxis has previously been implicated in the specification of axial tendon progenitors (Brent et al.,2003), and we present evidence that it may also be required for the differentiation of distal ribs.

This study uncovered a function of the dual specificity ERK MAP kinase phosphatase, MKP3, for the specification of mesenchymal progenitors in the somite sclerotome. We showed that Mkp3 was expressed in a twin-striped pattern, which closely matched the emergence of scleraxis transcripts along the anteroposterior somite edges(Fig. 1). This pattern suggested a link between the modulation of FGF signalling by MKP3 and scleraxis expression. We demonstrated that somitic expression of Mkp3 in the dorsal sclerotome was itself dependent on active ERK MAP kinase. This implied that FGF signalling in dorsal sclerotome cells is modulated by a negative feedback loop, which involves MKP3 and ERK MAP kinase. Indeed, we found that the levels of Mkp3 transcripts detected in response to FGF beads can cycle between extensive overexpression after a short exposure to complete loss of endogenous Mkp3 message after 24 hours(Fig. 3A-C; data not shown). We showed by western blot analysis that this dynamic response correlated with an increase in active dpERK protein after 1 hour and the loss of dpERK after 24 hours (Fig. 3D). This suggests that high levels of MKP3 protein resulted in depletion of dpERK from somites,which then lead to a loss of further endogenous Mkp3 transcription. This was confirmed by RT-PCR analysis (Fig. 4C). In mouse embryos, implanting an FGF4 bead into the primitive streak had similar effects after 24 hours because of expansive Spry2expression (Davidson et al.,2000) and it is likely that in chick somites sprouty and other negative regulators impact upon this feedback loop. Interestingly, the expression of Mkp3 immediately around the bead was not affected by the feedback mechanism (Fig. 3C), as an ERK MAP kinase independent pathway, i.e. PI3-kinase(Fig. 2E) could also regulate Mkp3 expression in response to FGF beads(Kawakami et al., 2003; Echevarria et al., 2005).

MKP3 is a cell-autonomous modifier of FGF signalling and this strongly suggested that FGFs signal to the dorsal sclerotome to control scleraxis expression. We also show that dpERK is localised in the region where scleraxis positive cells are found(Fig. 4D). It is possible that other signalling pathways cooperate with ERK MAP kinase and converge to control the spatiotemporal expression of Mkp3 in developing somites(Moreno and Kintner, 2004; Rintelen et al., 2003; Tsang et al., 2004). For example, our results using caMEK1, which led to a loss of MyoD, would be consistent with the idea that a MyoD-dependent factor contributes to the expression of Mkp3 and scleraxis(Fig. 5F). However, MyoD was also lost after application of an FGF bead(Fig. 5G), which induced ectopic expression of Mkp3 and scleraxis, and this argues against a MyoD-dependent mechanism. Furthermore, scleraxis is expressed in MyoD- and Myf5-null mice(Tajbakhsh et al., 1996; Brent et al., 2005). In mouse,FGF4 is important for scleraxis expression(Brent et al., 2005). We show that Fgf4 expression is lost after caMEK1 treatment(Fig. 5H); this could provide a simple explanation for the loss of scleraxis under these conditions. However, scleraxis expression depended on ERK activity(Fig. 2L, Fig. 4C,G, Fig. 6A), which was directly stimulated in somites after caMEK1 electroporation(Fig. 4E). Therefore, loss of MyoD and Fgf4 could not explain the loss of scleraxis transcription.

We have focused here on MKP3, which is one of many negative regulators of the MAP kinase transduction pathway. Other members of this group include Sef,Spred and Sprouty. Expression of Sef and Sprouty is induced by activation of the MAP kinase cascade itself, and as MKP3 regulates MAP kinase activity directly it could be pivotal in controlling these antagonists, which act at different levels in the pathway (Ozaki et al., 2001; Furthauer et al.,2002; Tsang et al.,2002; Yusoff et al.,2002; Kovalenko et al.,2003). This suggests that MKP3 acts at a crucial level in the FGF signal transduction cascade regulating all downstream events that depend on ERK MAP kinase, including activation of transcription factors and phosphorylation of cytoplasmic targets. Furthermore, other signalling cascades, including the retinoic acid and Wnt/β-catenin pathways, are able to regulate the expression of Mkp3 in other tissues(Moreno and Kintner, 2004; Tsang et al., 2004). This provides a possible role for other pathways, yet to be described in somites,to limit Mkp3 expression to the dorsal sclerotome and would explain how FGF responsive genes can be restricted.

In this study, overexpression experiments in chick embryos demonstrated the close interdependence of dpERK and MKP3, which established a tightly controlled level of active ERK MAP kinase in cells exposed to FGFs. There seem to be discrepancies between the FGF bead experiments and caMEK1 electroporation. Mkp3 was expressed when high levels of dpERK were present in response to a bead (Fig. 3A). Equally, caMEK1 induced an increase of dpERK but led to a loss of Mkp3 transcription (Fig. 5A). The most likely explanation for these apparently conflicting results is our finding that the bead-mediated upregulation of Mkp3can go through ERK MAP kinase and PI3-kinase(Fig. 2D,E). By contrast,expression of scleraxis is probably dependent on dpERK only, and we and others have found that scleraxis is upregulated in response to an FGF bead after 5 hours (Fig. 3E) (Brent and Tabin,2004). Electroporation of hMKP3-GFP led to loss of dpERK and concomitant loss of scleraxis(Fig. 4A,C,G). We propose that scleraxis can only be induced at a certain level of dpERK, set by the MKP3-ERK feedback loop. This is consistent with the delayed induction of scleraxis by FGF beads relative to Mkp3(Fig. 3E), which correlates with a specific level of dpERK (Fig. 3D) and with the observed loss of scleraxis when dpERK levels are either too high (Fig. 6D) or too low (Fig. 4C,G; Fig. 6A). In Drosophila, the puckered gene, a member of the same family as Mkp3, functions in a negative feedback loop to modulate JNK MAP kinase activity. In both puckered overexpression and loss-of-function experiments, a similar defect in dorsal closure resulted, reminiscent of our data (Martin-Blanco et al.,1998). In addition, negative-feedback regulation has been proposed to confer multistability on ERK MAP kinase activity(Markevich et al., 2004) and the response of scleraxis to an intermediate level of dpERK would be in agreement with this model. Our finding that MKP3 is capable of inhibiting all downstream effectors involved in scleraxis induction, presumably by depleting active ERK MAP kinase, suggests that within the somite, the MKP3-dpERK negative feedback loop is crucial for establishing specific signal strength.

This work demonstrated a functional importance of MKP3 for the correct expression of scleraxis. This in turn is pivotal for the specification of cells in the dorsal sclerotome. Scleraxis marks both tendon and rib progenitors in the sclerotome of thoracic somites. Based on our data, we speculate that scleraxis functions in the specification of distal rib chondrocytes at an early stage for the following reasons. First, scleraxis has been shown to increase aggrecan expression and stimulate chondrogenesis in cell culture (Liu et al.,1997). Second, scleraxis is transiently co-expressed with sox9, a chondrogenic marker in rib primordia, in mouse at 12.5 dpc(Asou et al., 2002; Brent et al., 2005). However,after 13.5 dpc expression diverges, which we interpret as scleraxis acting at an early stage of rib specification. Similarly, in chick, scleraxisis no longer expressed in condensing ribs and the expression pattern is consistent with an early requirement for scleraxis in rib development(Brent et al., 2003). Third,when we altered the signal strength of ERK MAP kinase, distal ribs did not form (Fig. 6). Of the somitic markers analyzed, only scleraxis expression was consistently lost under these conditions. Fourth, in electroporated embryos, one rib was typically affected, consistent with the loss of scleraxis expression from the anterior border of one somite and the posterior border of the next(Fig. 6D). Thus, we speculate that one domain of scleraxis formed by two successive somites, gives rise to one rib structure. This is in agreement with experiments demonstrating that the distal rib arises through a process involving resegmentation(Aoyama and Asamoto, 2000).

The mechanism leading to rib loss remains to be investigated and could involve a failure of progenitor cells to become specified. For example, scleraxis may act together with BMP signalling and forkhead transcription factors in cell specification(Buchberger et al., 1998; Kramer et al., 2000; Sudo et al., 2001). Alternatively, progenitor cells may undergo apoptosis or they might fail to migrate. In this context, it is interesting to note that the migration of cells from the lateral somite into the somatopleure is dependent on the regulation of paxillin by ERK activity(Ishibe et al., 2004).

Interestingly, in mouse, the expression of Fgf8, scleraxis and Mkp3 is different suggesting a species-specific change in the function of MKP3. In chick, Fgf8 transcripts are found in the central myotome but in mice they colocalise with scleraxis at the rostral and caudal somite edges, indicating that FGF8 signals in an autocrine fashion(Crossley and Martin, 1995). In mouse, Mkp3 is not expressed in the dorsal sclerotome but in the dermomyotome/myotome (Dickinson et al.,2002; Klock and Herrmann,2002). Thus, it would be interesting to investigate whether Mkp3 knockout mice have normal ribs and tendons.

We thank Grant Wheeler and Mohammad Hajihosseini for stimulating discussions, and Alasdair Reid for assistance in the early stages of the project. We thank Christophe Marcelle, Philip Cohen, Delphine Duprez and Andrew Chantry for reagents. We thank the John and Pamela Salter Charitable Trust for support. T.G.S. was funded by an MRC studentship, D.S. is funded by the Wellcome Trust (ref. 070699) and M.P. is funded by the BBSRC (ref. 83/G17536). Stephen Keyse is funded by Cancer Research UK.