Shp‐1, Shp‐2 and corkscrew comprise a small family of cytoplasmic tyrosine phosphatases that possess two tandem SH2 domains. To investigate the biological functions of Shp‐2, a targeted mutation has been introduced into the murine Shp‐2 gene, which results in an internal deletion of residues 46–110 in the N‐terminal SH2 domain. Shp‐2 is required for embryonic development, as mice homozygous for the mutant allele die in utero at mid‐gestation. The Shp‐2 mutant embryos fail to gastrulate properly as evidenced by defects in the node, notochord and posterior elongation. Biochemical analysis of mutant cells indicates that Shp‐2 can function as either a positive or negative regulator of MAP kinase activation, depending on the specific receptor pathway stimulated. In particular, Shp‐2 is required for full and sustained activation of the MAP kinase pathway following stimulation with fibroblast growth factor (FGF), raising the possibility that the phenotype of Shp‐2 mutant embryos results from a defect in FGF‐receptor signalling. Thus, Shp‐2 modulates tyrosine kinase signalling in vivo and is crucial for gastrulation during mammalian development.
Protein tyrosine kinases mediate the biological effects of a wide range of polypeptide growth factors. Tyrosine phosphorylation is reversed by tyrosine phosphatases, which comprise a large family of transmembrane and cytoplasmic enzymes (Sun and Tonks, 1994; Hunter, 1995). Such phosphatases can attenuate positive signals emanating from tyrosine kinases. Alternately, since tyrosine phosphorylation can be a negative regulator of signalling, as in the C‐terminus of Src family kinases (Cooper and Howell, 1993), tyrosine phosphatases may also directly stimulate intracellular signalling by removing such inhibitory phosphorylation events.
A principal mechanism by which receptor tyrosine kinases (RTKs) stimulate downstream signalling pathways involves the recognition of phosphotyrosine sites on either the activated receptors or their substrates by the SH2 or PTB domains of cytoplasmic effector proteins (Cohen et al., 1995; Pawson, 1995). A small family of SH2‐containing tyrosine phosphatases may regulate signalling events immediately proximal to activated receptors for growth factors, cytokines and antigens. This group of phosphatases contains two vertebrate members, Shp‐1 (previously termed Hcp, PTP1C, SH‐PTP1, SHP; Adachi et al., 1996) and Shp‐2 (previously termed Syp, PTP1D, SH‐PTP2, PTP2C; Freeman et al., 1992; Ahmad et al., 1993; Feng et al., 1993; Vogel et al., 1993), and Drosophila csw, a likely homolog of Shp‐2 (Perkins et al., 1992).
Genetic investigation of csw indicates that this phosphatase is a positive signal transducer downstream of the Torso RTK and is required for normal development of terminal structures in the Drosophila embryo (Perkins et al., 1992). Through its SH2 domains, csw binds to activated Torso receptors and becomes tyrosine phosphorylated at its C‐terminus (Cleghon et al., 1996). csw has also been identified in a genetic screen for genes involved in photoreceptor cell development mediated by the sev RTK (Simon et al., 1991). Wild‐type (WT) csw is essential for R7 photoreceptor cell development, and a membrane targeted form of this phosphatase can drive R7 cell differentiation in the absence of sev (Allard et al., 1996). csw does not bind directly to sev, but rather associates with an intermediate docking protein, dos (Herbst et al., 1996; Raabe et al., 1996).
The Shp‐1 gene is disrupted by the motheaten murine mutations (Shultz et al., 1993; Tsui et al., 1993), which cause excessive erythropoiesis, deregulated mast cell function and leukocyte hyperactivity. This genetic evidence for a negative regulatory role is complemented by several studies indicating that Shp‐1 inhibits cell signalling mediated by interleukin‐3, erythropoietin, steel factor, Fcγ and antigen receptors (Yi et al., 1993; Cyster and Goodnow, 1995; D'Ambrosio et al., 1995; Klingmuller et al., 1995; Paulson et al., 1996).
Shp‐2 is widely expressed and becomes tyrosine phosphorylated after cell stimulation with a myriad of growth factors and cytokines (Feng et al., 1993; Vogel et al., 1993; Welham et al., 1994). There is some contention concerning the role of Shp‐2 in tyrosine kinase signalling and the relative contributions of its SH2 domains, catalytic domain and C‐terminal tail to its downstream effects (Milarski and Saltiel, 1994; Noguchi et al., 1994; Rivard et al., 1995; Tang et al., 1995; Bennett et al., 1996; Marengere et al., 1996). Two tyrosine phosphorylation sites located at the C‐terminus of Shp‐2 are consensus binding sites for the SH2 domain of the Grb2 adaptor protein (YXNX; Bennett et al., 1994; Songyang et al., 1994). Thus, Shp‐2 may be a positive effector of signal transduction by itself acting as an adaptor protein to link Grb2 and mSos1 to Ras, thereby leading to activation of the Raf/MAP kinase cascade (Bennett et al., 1994; Li et al., 1994; Noguchi et al., 1994; Welham et al., 1994). However, overexpression of Shp‐2 protein in which the C‐terminal tyrosines have been mutated to phenylalanine was shown to have no effect on MAP kinase activation or induction of DNA synthesis in response to growth factor stimulation (Bennett et al., 1996). In contrast, overexpression of phosphatase‐inactive mutants has suggested that Shp‐2, but not Shp‐1, plays a positive role in Ras/MAP kinase activation downstream of the receptors for epidermal growth factor, platelet derived growth factor (PDGF) and insulin. These experiments indicate that substrates specific for the Shp‐2 catalytic domain may be essential for its positive role in signal transduction.
The N‐terminal region of Shp‐2 exerts an inhibitory effect on the catalytic activity of its phosphatase domain, which can be relieved by the binding of specific phosphopeptides to the SH2 domains (Lechleider et al., 1993a; Pluskey et al., 1995; Eck et al., 1996; Pei et al., 1996). Thus, the SH2 domains of Shp‐2 may play a dual role in regulating its enzymatic activity and in directing the phosphatase to the proximity of its substrates. We have introduced a targeted mutation into the murine Shp‐2 gene which results in the deletion of 65 amino acids within the N‐terminal SH2 domain. We find that WT Shp‐2 is essential for embryonic survival and appears to play a key role during gastrulation in the organization of axial mesodermal structures and posterior development. Furthermore, we have used primary fibroblast‐like cells isolated from mutant embryos to explore the role of Shp‐2 in the regulation of MAP kinase activation elicited by fibroblast growth factors (FGFs) and to investigate the function of its SH2 domains in controlling signal transduction in vivo.
Clones of the Shp‐2 locus were obtained by screening an inbred 129/Sv strain mouse genomic library with a probe to the 5′ coding region of the Shp‐2 cDNA (Feng et al., 1993). To generate a mutation in the Shp‐2 locus, exon 3, encoding amino acids 46–110, was deleted by homologous recombination in embryonic stem (ES) cells and replaced by a neomycin resistance (neor) cassette (Figure 1A). Aggregation chimeras were generated and germline transmission of the mutant allele was obtained. Animals were routinely genotyped by Southern blot analysis (Figure 1B) or by Shp‐2‐specific PCR (Figure 1C). Shp‐2+/− heterozygotes were long lived (>18 months) and fertile in 129 inbred, 129×CD1 hybrid and CD1 outbred backgrounds. However, no Shp‐2−/− homozygous offspring were born following intercrosses between heterozygous males and females, indicating a recessive lethal phenotype (see below).
Since exons 2 and 4 are in‐frame, it was possible that aberrant splicing around the inserted neor cassette could yield a mutant Shp‐2 protein containing an internal deletion of residues 46–110, which comprise the majority of the N‐terminal SH2 domain (SH2‐N) and the short inter‐SH2 region. To investigate the coding potential of the mutant allele, protein lysates of embryos isolated at 8.0 days of gestation (E8.0) were analyzed by Western blotting using either antibodies raised to the tandem SH2 domains (residues 2–216), or to the C‐terminus (residues 576–593). Although the N‐terminal antibodies did not recognize a polypeptide in the Shp‐2−/− embryo lysates (Figure 1D), an abnormally sized gene product was observed when the C‐terminal antibody was used (Figure 1E). In heterozygous specimens, the C‐terminal antibody recognized both the WT 64 kDa and the Shp‐2Δ46–110 57 kDa polypeptides. Densitometry indicated that the mutant protein is expressed at ∼25% of WT levels, possibly due to inefficient splicing around the neor cassette.
To determine whether the transcript from the mutant Shp‐2Δ46–110 allele was spliced to join exons 2 and 4, RT–PCR analysis of total RNA from heterozygous specimens was performed (Figure 1F and data not shown). PCR oligonucleotide primers flanking exon 3 amplified an internally deleted cDNA of the size expected. Sequencing confirmed the in‐frame loss of codons for amino acids 46–110. Therefore, the mutant allele encodes a protein which retains the catalytic phosphatase domain and the C‐terminal tail, but would be expected to be deficient in the binding of phosphotyrosine motifs as it lacks a functional SH2‐N domain.
The Shp‐2Δ46–110 mutation is a recessive embryonic lethal
To characterize the lethality of homozygous mutant embryos, litters were dissected between E8.5 and E16.5 of development (Table I). No homozygous Shp‐2Δ46–110 embryos were observed between E11.5 and E16.5, though visible resorbing dead embryos that could correspond to the mutant progeny class were detected. Homozygous Shp‐2Δ46–110 mutant embryos could be detected at E10.5 of gestation, however, these embryos were necrotic (data not shown). The expected ratio (∼25%) of Shp‐2Δ46–110 homozygotes was observed at both E9.5 (27%) and E8.5 (20%) of development. These mutant embryos appeared viable, but their growth was retarded when compared with normal littermates and they exhibited severe developmental abnormalities (Figure 2). Examination of >100 different Shp‐2Δ46–110 homozygotes at these stages revealed a variable phenotype. Approximately 70% of the mutants formed an identifiable anterior–posterior (A–P) axis, while the remaining 30% appeared more severely affected and had arrested development prior to the formation of recognizable embryonic structures (see below). Those forming a visible A–P axis appeared extremely abnormal. All had severe posterior truncations and had not initiated the process of turning. They also had few if any somites, disorganized neuroectoderm, abnormal formation of the midline structures and perturbed development of the vasculature, including the heart and yolk sac (see Figure 2 and below).
Defective gastrulation in Shp‐2 mutant embryos
Initial characterization of the Shp‐2Δ46–110 mutant phenotype indicated that the overall organization of the body plan was affected, consistent with a defect in gastrulation and embryonic patterning. During gastrulation, nascent mesodermal cells migrating out of the primitive streak differentiate into distinct cell types including those which form the axial, paraxial and lateral mesoderm. The overall organization of the embryonic body plan depends to a considerable extent on the normal development of a specialized embryonic structure called the node, which in turn specifies cells forming the notochord, an axial mesoderm midline structure crucial for axis formation. To investigate whether the node or notochord were affected in the Shp‐2Δ46–110 homozygous embryos, the cordon‐bleu (cobl)–LacZ transgene (Gasca et al., 1995) was employed to label axial mesoderm structures (Figure 3). In WT E8.5 specimens, cobl is expressed in the notochord, which runs the entire length of the embryo, from the midbrain to the tail bud (Figure 3A). In sharp contrast to WT embryos, Shp‐2Δ46–110 mutants isolated at E8.5 (Figure 3B) were abnormally small and had failed to elongate posteriorly and develop a notochord, appearing more similar in size to WT specimens isolated at E7.5 (Figure 3C). In WT E7.5 specimens, cobl–LacZ was solely expressed in the node and in a few mid‐line cells anterior to the node. The node‐like structure in the E8.5 Shp‐2Δ46–110 mutants appeared larger than in WT E7.5 specimens, indicating that cells may have accumulated at this structure, perhaps from a failure to differentiate and/or migrate properly. The intensity of cobl–LacZ staining at E8.5 was variable between different Shp‐2Δ46–110 mutants. In the extreme situation, some embryos did not express any cobl–LacZ, nor did they form a distinguishable node (data not shown).
E9.5 Shp‐2Δ46–110 mutant embryos were identified and stained for cobl–LacZ (Figure 3D). Although these mutant embryos had developed a defined A–P axis, all exhibited varying degrees of posterior truncations that were associated with abnormalities in the development of the node/tail bud and the notochord. By E9.5, WT and heterozygous embryos had formed 20–24 pairs of somites and had a smoothly curving notochord which expressed cobl–LacZ from the midbrain to the chordoneural hinge at the posterior tip of the tail (Figure 3E). Some Shp‐2Δ46–110 mutants were able to form somites (Figure 3F and G), whereas others failed to make any recognizable somites (Figure 3H). All three of the mutant embryos exhibited an abnormal notochord, as highlighted by the pattern of cobl–LacZ expression. The well developed Shp‐2Δ46–110 mutant embryo shown in Figure 3F had a relatively normal notochord that almost reached the posterior tip of the tail. However, at the posterior tip of this embryo, there was a buildup of cobl–LacZ expressing cells, from which there were several abnormal projections of notochord‐like material (* in Figure 3F–H). In a more severely affected Shp‐2Δ46–110 mutant embryo, the neural tube and the notochord were kinked and cobl–LacZ expression did not extend to the posterior tip of the embryo (Figure 3G). Instead, a large mass of cobl‐positive cells extending anteriorly from the most posterior extent of the notochord was present. An even more defective mutant (Figure 3H) exhibited a dramatic posterior truncation of the notochord. This embryo also had an expansion of cobl–LacZ expressing cells and displayed an abnormal structure of labelled cells that projected outside of the embryo on its dorsal surface.
Transverse sections highlight the defects observed in the Shp‐2Δ46–110 mutant embryos. In WT E9.5 sections, cobl–LacZ was expressed in cells corresponding to the notochord, the floorplate of the neural tube and the roof of the gut (Figure 3I). In the moderately well developed mutants, cobl–LacZ expression labelled the notochord; however, additional labelled cells were also identified (Figure 3J–L). Instead of labelling a small group of cells at the midline, cobl–LacZ expression in mutant embryos was frequently expanded and observed to accumulate and protrude within the gut of the embryos.
To confirm that these abnormal tissues corresponded to axial mesoderm structures, an independent marker for the node and the notochord was analyzed. The Brachyury (T) gene is solely expressed within axial mesoderm (Wilkinson et al., 1990) and therefore should directly highlight defects in the development of this tissue. Whole‐mount RNA in situ hybridization studies were performed on Shp‐2Δ46–110 mutants using a T probe. This analysis corroborated the results obtained with cobl (Figure 3M–P). Of particular significance was the identification in E9.5 mutants of accumulations of T‐positive cells protruding into the gut (Figure 3N) and of multiple or split notochords (Figure 3O and P). Furthermore, at E8.5 some mutant embryos did not stain for T at all and others stained an abnormal or delayed primitive streak, which is reminiscent of expression observed for WT E7.0 embryos (data not shown).
The combined results of cobl and T expression indicate that the development of axial mesodermal structures are severely affected in Shp‐2 mutant embryos. One possible outcome of an inability to properly gastrulate would be a phenotype in which the embryos are so severely affected that they fail to form an A–P axis. Indeed, we found that ∼30% of the Shp‐2 homozygotes failed to develop an A–P axis, or any recognizable embryonic structures at all (Figure 4). These embryos were observed between E7.5 and 10.5 of development and expressed a wide array of markers for different cell types, including cobl–LacZ (Figure 4A), an endothelial marker Flk1–LacZ (Figure 4B and see below; Shalaby et al., 1995) and a marker mainly restricted to the developing nervous system, Nuk–LacZ (Figure 4C and D; Henkemeyer et al., 1996). These severely affected embryos may represent the class of Shp‐2Δ46–110 mutants which failed to develop a node, as observed in several E8.5 embryos which did not stain for cobl–LacZ or T expression. The failure of some Shp‐2Δ46–110 embryos to develop any axis indicates a crucial role in early gastrulation for Shp‐2, which can be overcome to some degree by many of the Shp‐2Δ46–110 mutant embryos, potentially due to the retention of partial function of the Shp‐2Δ46–110 protein.
Vascular defects in Shp‐2 mutants
Upon dissection of mutant embryos, we noted that the yolk sac appeared abnormally thin and wrinkled. We used Flk1–LacZ as a marker to label endothelial cells in Shp‐2Δ46–110 mutants (Figure 5). By E9.5, embryos stained for Flk1–LacZ expression showed a complex network of blood vessels throughout the embryo, which was also present in mutant embryos (Figure 5A and B), demonstrating their ability to form a network of endothelial cells. Flk1‐LacZ expression in E9.5 yolk sacs of mutant embryos indicated that in situ vasculogenesis and the production of the primary capillary plexus occurred normally. However, the yolk sac blood vessels failed to re‐organize into a highly vascularized network in the mutants (Figure 5C and D) and remained in the honeycombed pattern reminiscent of E8.5 WT embryos.
Transverse sections of these embryos confirmed the presence of a vascular network within the mutants (Figure 5E and F). Even though the heart formed somewhat abnormally, it was often beating at the time of dissection. Sections through the yolk sac (Figure 5G and H) showed that blood vessels did form within the mutant yolk sac and that primitive blood cells were present. However, thick‐walled blood vessels, which are the main arteries and veins delivering nutrients to the embryo, did not form in the mutant yolk sacs. The allantois (Figures 5B, 3F and H) did not develop properly nor connect to the placenta, leading to death and resorption of these embryos by E10.0–11.0 of gestation. These studies indicate that WT Shp‐2 protein is not required for endothelial differentiation, but rather is needed for certain aspects of vascular endothelial cell organization.
Shp‐2 is required for RTK signalling
The phenotypes observed in the Shp‐2Δ46–110 mutant embryos are likely due to a deregulation of tyrosine kinase signalling during development. To address whether the Shp‐2Δ46–110 protein could function downstream of activated growth factor receptors, primary cell cultures from homozygous mutant, heterozygous and WT embryos were established. In order to test biochemically the binding properties of the Shp‐2Δ46–110 protein, we analyzed its interactions with the β receptor for PDGF (PDGFR), as both the SH2‐N and the SH2‐C domains of Shp‐2 bind to the activated form of this growth factor receptor (Lechleider et al., 1993b; Case et al., 1994; Feng et al., 1994). Cultures were stimulated with PDGF, Shp‐2 proteins were immunoprecipitated with the C‐terminal antibody and precipitates were probed with either anti‐phosphotyrosine or anti‐Shp‐2 antibodies (Figure 6A). Upon addition of PDGF, both the WT and mutant forms of Shp‐2 became tyrosine phosphorylated. This result, however, does not indicate that Shp‐2Δ46–110 can still bind to the PDGFR, as a mutant form of the receptor (Y1009F) that can no longer associate with Shp‐2 still induces tyrosine phosphorylation of this protein (Kazlauskas et al., 1993). Indeed, we found that although autophosphorylated PDGFR was present in anti‐Shp‐2 precipitates from WT or heterozygous cells, the activated receptor could not be detected in association with the Shp‐2Δ46–110 protein (Figure 6A). Consistent with this observation, a GST fusion protein containing both Shp‐2 SH2 domains bound in vitro to the activated PDGFR, while the corresponding mutant protein lacking residues 46–110 did not (Figure 6B). These results suggest that the Shp‐2Δ46–110 protein is impaired in its ability to bind the autophosphorylated PDGFR.
Shp‐2 has been implicated in the activation of MAP kinase, which serves as a downstream effector of Ras proteins. To assess whether MAP kinase becomes activated by the PDGFR in the absence of WT Shp‐2 protein, gel shift assays, which correlate directly with the activation of this enzyme, were performed. An enhanced and potentiated MAP kinase activation in response to PDGF stimulation was observed in the Shp‐2Δ46–110 cells as compared with WT cells (Figure 7A). At 5 min post‐stimulation, 98% of MAP kinase was in the phosphorylated active state in the mutant cells, whereas only 67% of MAP kinase was activated in WT cells (Figure 7C, left graph). Similarly, at 80 min after addition of PDGF, 32% of MAP kinase remained activated in mutant cells, as compared with 20% in WT cells. These results show that Shp‐2 protein is involved in regulating the intensity of MAP kinase activation during PDGF signalling.
Mutation within the murine PDGFRs induce embryonic phenotypes which effect later stages of murine development than that exhibited by Shp‐2 mutant embryos. This suggests that the Shp‐2 mutant phenotype results from dysregulation of growth factor receptor pathways important earlier in development than the PDGFR pathways. FGFs have been implicated in controlling numerous developmental decisions, including those made during gastrulation. Furthermore, disruption of the Fgf‐R1 gene by homologous recombination (Yamaguchi et al., 1994; Deng et al., 1994) leads to a recessive lethal phenotype that is similar to, though distinct from, the Shp‐2Δ46–110 mutant phenotype described here (see Discussion). In addition, activation of MAP kinase is known to be crucial for the induction of mesoderm by FGF (Gotoh et al., 1995; LaBonne et al., 1995; Umbhauer et al., 1995). RT–PCR analysis from total RNA isolated from primary cells indicated that the Fgf‐R1 gene was expressed in the primary cell cultures from both WT and mutant embryos (data not shown). To address whether the Shp‐2 protein is involved in signalling downstream of FGF receptors, WT and mutant primary cultures were subjected to a time course of FGF stimulation followed by analysis of MAP kinase mobility. The MAP kinase gel shift of the Shp‐2Δ46–110 mutant cells in response to FGF was greatly reduced and transient in comparison with WT cells (Figure 7B). Densitometry analysis indicated that at 5 min after addition of FGF to the WT cells, ∼70% of MAP kinase was shifted, compared with only 17% in the mutant cells (Figure 7C, right graph). Maximal MAP kinase activation of both WT (80% shifted) and mutant (57% shifted) cells was observed within 10 min of FGF stimulation. WT and mutant cells also showed a sharp distinction in their ability to achieve a sustained level of MAP kinase activation following FGF stimulation. In WT cells, 70% of the MAP kinase remained in the activated state 80 min after FGF stimulation. In contrast, only 5% of MAP kinase was phosphorylated in the mutant cells at this time. These results indicate that Shp‐2 is required in a positive fashion for robust and sustained activation of MAP kinase by FGF, consistent with the supposition that Shp‐2 and FGF receptors act in the same signalling pathway during gastrulation.
We describe effects on murine embryonic development and MAP kinase activation resulting from a mutation in the gene encoding the Shp‐2 SH2 domain‐containing tyrosine phosphatase. In the absence of WT Shp‐2 protein, the production of axial mesoderm is severely perturbed as indicated by abnormalities in the node, notochord and the A–P axis. These embryos also exhibit poorly developed somites and kinked or unclosed neural tubes. Extra‐embryonic mesoderm also requires functional Shp‐2 protein, as endothelial cells within the yolk sac of Shp‐2 mutants remain in the primitive honeycombed pattern rather than re‐organizing into a highly vascularized network. Finally, the allantoic mesoderm is underdeveloped and fails to fuse with the maternal circulation, leading to the death and resorption of these mutant embryos between E10 and 11.0 of gestation. Tyrosine kinase receptors, such as those for the FGF family, are known to function in the process of mesoderm induction. We show that the Shp‐2Δ46–110 mutation leads to a failure of cells to fully activate MAP kinase in response to FGF. Moreover, in WT cells, FGF led to an extended duration of MAP kinase activation which lasted >80 min, whereas in Shp‐2 mutant cells only a transient peak of MAP kinase activation was observed which was rapidly down‐regulated to near basal levels. Interestingly, not all tyrosine kinase receptors require Shp‐2 to fully activate MAP kinase, as stimulation of the same Shp‐2Δ46–110 mutant cell cultures with PDGF hyper‐activated signalling through this receptor leading to heightened MAP kinase activation. Our results suggest that Shp‐2 plays a critical role in controlling tyrosine kinase signalling in the mouse embryo.
The Shp‐2Δ46–110 allele
The Shp‐2Δ46–110 mutant allele encodes a 57 kDa protein, which is expressed at ∼25% of wild type Shp‐2 protein levels and results from in‐frame splicing around the inserted neor cassette. The Shp‐2Δ46–110 protein lacks residues responsible for phosphopeptide recognition by the SH2‐N domain (Lee et al., 1994), as well as the inter‐SH2 region, which forms part of the hydrophobic core that orients the tandem SH2 domains (Eck et al., 1996). Deletion of residues 46–110 in the N‐terminal SH2 domain would be expected to entirely destroy the phosphotyrosine binding properties of the SH2‐N domain, and might also influence the activity of the C‐terminal SH2 domain. Indeed, we show that disruption of this region severely compromises the ability of Shp‐2 to bind the PDGFR.
The Shp‐2Δ46–110 mutation is not a dominant inhibitory allele of Shp‐2, as heterozygous animals have no obvious phenotype, while homozygotes have a striking embryonic defect. While we cannot rule out the possibility that Shp‐2Δ46–110 is a hypomorphic allele, it is clear that the mutation induces a severe phenotype, and may be tantamount to a genetic null. Indeed, it is interesting to note that a protein null mutation within the Shp‐2 gene leads to resorption of the homozygous mutant embryos at the same time of gestation as the Shp‐2Δ46–110 allele (Arrandale et al., 1996). Although the embryonic phenotype of the protein null mutation has not been analyzed, these results suggest that the Shp‐2Δ46–110 mutation creates a severe loss of function allele. The failure of the Shp‐2Δ46–110 protein to support gastrulation indicates that the mutant Shp‐2 protein lacking an intact SH2‐N is defective in signalling. Moreover, the remaining SH2‐C sequence is not sufficient to direct a stable association of the phosphatase with the PDGFR. The ability to bind upstream phosphotyrosine‐containing proteins, such as activated receptors, appears to be crucial for signal transduction by SH2‐containing proteins. For example, recent genetic evidence indicates that a point mutation which disrupts the csw binding site in the Drosophila Torso RTK phenocopies loss of function Torso or csw alleles, indicating this receptor–phosphatase interaction is essential for receptor signalling (Cleghon et al., 1996). Here, we find that a mutant Shp‐2 protein that lacks a functional SH2‐N domain and does not bind to upstream targets is severely compromised in its biological function.
Shp‐2 is required for gastrulation and mesoderm formation
Gastrulation is a complex process that involves co‐ordination of cell division, differentiation and migration and results in the formation of the third major cell layer, the mesoderm (reviewed in Beddington and Smith, 1993). During gastrulation, several populations of mesoderm are induced, including axial mesoderm (node and notochord), paraxial mesoderm (somites) and extra‐embryonic mesoderm (yolk sac and allantoic mesoderm). In homozygous Shp‐2Δ46–110 mutant embryos, formation of all these sub‐populations was affected to some degree. Most importantly, staining of embryos for cobl or T expression showed that axial mesoderm was severely perturbed in Shp‐2Δ46–110 mutant embryos. Abnormal projections of cobl‐ or T‐positive cells could be observed, suggesting that Shp‐2 is required for proper organization or migration of mesodermal cells during gastrulation.
Some Shp‐2Δ46–110 embryos failed to express cobl or T, and may represent a population of mutants that later failed to develop an A–P axis. These results suggest that Shp‐2 may also act at an earlier stage during formation of the node. The defects observed in mutant embryos suggests that the Shp‐2 phosphatase acts downstream of one or more receptor (or receptor‐linked) tyrosine kinases to control the formation, movement or organization of mesodermal cells. Tyrosine kinase and Ras signal transduction cascades appear to be required for gastrulation to proceed normally; in addition to this study on Shp‐2, mutations within the Fgf‐R1, csk, Fak and the Gap/Nf1 (double mutant) loci have all been shown to interfere with mesoderm formation (Immamoto and Soriano, 1993; Nada et al., 1993; Deng et al., 1994; Yamaguchi et al., 1994; Furuta et al., 1995; Henkemeyer et al., 1995). For all of these mutations, mutant embryos implant and initiate gastrulation, but subsequently fail to complete this process and die prior to E10.5. Our results indicate that Shp‐2 is a critical target of tyrosine kinases during gastrulation, and may act by regulating phosphorylation events that in turn control the Ras/MAP kinase pathway.
Dominant‐negative studies of Fgf‐R1 and Shp‐2 proteins have been performed by micro‐injection of mutant RNAs into Xenopus embryos (Amaya et al., 1991; Tang et al., 1995). There are similarities and differences between the outcome of these studies compared with results obtained from genetic analysis of mouse mutants. In both species, embryos initiated gastrulation, but did not complete this process normally and exhibited severe posterior truncations, abnormal notochord development and disorganized neural tubes. Although not apparent in the Xenopus studies, mutation of either Shp‐2 or Fgf‐R1 in mouse embryos frequently affected the development of anterior structures, including the formation of branchial arches, head mesenchyme and closure of the neuroectoderm. Furthermore, expression of dominant‐negative Fgf‐R1 or Shp‐2 in Xenopus led to an apparent reduction in axial mesoderm, whereas the Fgf‐R1 and Shp‐2 mutations in mice often lead to an expansion of axial mesoderm. The molecular reasons for these differences are unclear, however, they may reflect differences between the experimental design of injection of dominant‐negative RNAs as compared with targeted gene mutations, or differences in the inherent developmental patterning between Xenopus and mice.
Shp‐2 modulates RTK signalling
We have used cells derived from homozygous mutant embryos to investigate the role of Shp‐2 in the MAP kinase pathway. It has been reported that Shp‐2 is required for the induction of DNA synthesis and c‐fos transactivation downstream of the PDGFR (Rivard et al., 1995). Conversely, it has also been reported that Shp‐2 is not involved in Elk‐1 transactivation or S phase entry in response to PDGF (Bennett et al., 1996). The experiments reported here show that activation of MAP kinase in response to PDGF was enhanced in Shp‐2 mutant embryonic cells as compared with their WT counterparts. This is consistent with a model in which the WT Shp‐2 protein, via its SH2 domains, is recruited to and subsequently dephosphorylates the activated PDGFR, leading to receptor inactivation and down‐regulation of the Ras/MAP kinase cascade. Indeed, phosphotyrosine blots of lysates from PDGF stimulated WT or mutant cells indicated that Shp–2 is involved in down‐regulating the phosphotyrosine content of the PDGFR (data not shown). In contrast to its apparent inhibitory role in PDGFR signalling to MAP kinase, functional Shp‐2 protein is required for both the initial and sustained activation of MAP kinase in response to FGF. Although the mechanism through which Shp‐2 might enhance signalling through the MAP kinase pathway is not known, it is possible that dephosphorylation of tyrosine sites that engage the SH2 domains of proteins that antagonize the Ras pathway, such as GAP, would contribute in a positive manner to Ras/MAP kinase activation.
A linear signalling pathway has been described (Pawson, 1993), through which a variety of distinct growth factor receptors activate the Ras/MAP kinase cascade. However, these observations do not explain how different receptors utilize this same pathway to induce distinct cellular responses such as proliferation, differentiation or migration. In PC12 cells, the decision to proliferate or differentiate in response to signalling by RTKs is controlled, at least in part, by the potency and duration of MAP kinase activation (Marshall, 1995). Our results suggest that Shp‐2 may represent a widely utilized target which modulates the intensity and duration of MAP kinase activation, depending on the nature of the specific receptor that has been stimulated. Therefore, Shp‐2 plays a key role downstream of RTKs in defining the specificity of cellular responses to external cues. The profound phenotype of Shp‐2 mutant embryos suggests that the interactions of Shp‐2 with receptors, and its role in controlling intracellular signalling, are of biological significance.
Materials and methods
Shp‐2 gene targeting
A library of 129/Sv strain mouse genomic DNA was screened with the Shp‐2 cDNA. Positive clones were characterized by restriction mapping and sequence analysis to identify intron–exon structure. A targeting vector was constructed by inserting 1.7 and 5.5 kb fragments from the 5′ end of the Shp‐2 gene into pPNT (Tybulewicz et al., 1991). Genomic sequences were cloned into the vector in the opposite transcriptional orientation to the neor cassette. The linearized targeting vector was electroporated into the ES cell line R1 (Nagy et al., 1993) and colonies were isolated following selection in G418 and gancyclovir (Wurst and Joyner, 1993). Double resistant cell clones were expanded and purified genomic DNA was screened for homologous recombination by Southern blotting with the 3′ external probe. Eight out of 200 cell lines screened exhibited a 4.3 kb mutant‐specific NcoI fragment. Similar blots were also screened with a neor probe to check for multiple insertion events (data not shown). Germline transmitting chimeras of three independent Shp‐2+/− cell lines (8.14, 5.20 and 6.23) were obtained by aggregating ES cells with morula‐staged CD1 embryos (Nagy et al., 1993). Subsequent germline mice were analyzed in an inbred (129Sv), mixed (129Sv×CD1) and CD1 (>95%) outbred backgrounds with no observable difference in phenotype.
The Shp‐2 mutation was routinely genotyped using the following oligonucleotides:
Shp‐2 E1AS: GTA GGA GCC CTA TAG AAT CTG
PCRneoβ2: TAC CCG GTA GAA TTG ACC TGC AG
Shp‐2 10: GAG TCA CAC AGA TCG TAT GCA TCC CA
Shp‐2 11: GAT ACG CCT TCT CTC AAT GGA C
Shp‐2 E1AS and PCRneoβ2 detect the mutant allele of 250 bp, and Shp‐2 10 and Shp‐2 11 oligonucleotides amplify the WT allele of 311 bp.
Western blot analysis
E8.0 embryos were lysed in 50 μl (WT) or 25 μl (mutant) TxLB (1% Triton X‐100, 138 mM NaCl, 20 mM Tris pH 8.0, 2 mM EDTA, 10% glycerol, plus protease inhibitors). Denatured lysates were resolved on 10% polyacrylamide gels and transferred to Immobilon‐P membrane (Millipore). Membranes were blocked in 5% milk powder in TBST (0.05% Tween‐20) and Western blots were performed with either anti‐Shp‐2 SH2 (Feng et al., 1993), anti‐GAP (UBI) or anti‐Shp‐2 C‐terminus (Santa Cruz) polyclonal antibodies, followed by protein A–HRP (Bio‐Rad) and chemiluminesence. Immunoblotting was analyzed with a Molecular Dynamics computing densitometer (model 300A).
For β‐galactosidase staining, male mice compound heterozygous for the Shp‐2 mutation and a LacZ (cobl‐, Flk1‐ or Nuk‐) marker were mated to Shp‐2 heterozygous females. In the course of crossing the Shp‐2 and Flk1 mutations, a genetic linkage between these two loci was identified with a recombination frequency of 0.032. Collection, X‐gal staining, sectioning and photography was performed as described (Henkemeyer et al., 1996). Whole‐mount in situ hybridization was performed as described (Conlon and Rossant, 1992).
Establishing cell lines
E9.5 embryos were dissected and, due to the decreased cell number in Shp‐2Δ46–110 mutants as compared with WT embryos, mutant embryos from each litter were pooled in order to plate approximately the same cell numbers as for the WT cultures. Single cell suspensions were prepared by treatment with 0.1% trypsin for 2–5 min. Cells were plated in DMEM, 15% FCS, 5% embryo fibroblast conditioned medium. Genotyping of the cell lines was performed by PCR and Western blotting analysis. All experiments were performed on cells between passage 5 and 9, prior to crisis, by comparing cell cultures from either littermates or equivalent passage number. PDGF and FGF time courses were performed a minimum of three times on two independent cell lines of each genotype, and representative results are shown.
Growth factor stimulations of cell cultures, analysis of phosphotyrosine induction and MAP kinase gel mobility
The GST fusion of the tandem SH2 domains with the internal deletion of amino acids 46–110 was cloned by amplification of the mutant mRNA from heterozygous cells. The PCR product was cloned into pGEX4T‐2 and sequenced on both strands with the ALF automated DNA sequencer (Pharmacia). Sequence analysis revealed that amino acid 46 of the Shp‐2Δ protein is encoded by a chimeric codon generated by in frame splicing between codons 46 and 111. The WT GST‐SH2 N+C, expression and purification has been previously described (Feng et al., 1993). Embryo fibroblasts or NIH 3T3 cells were grown to sub‐confluency, serum starved for 48 h and stimulated for the times indicated with 100 ng/ml PDGF‐BB (UBI). For immunoprecipitations (IPs) or GST mixing experiments, cells were lysed in TxLB, protein concentrations determined by bicinchoninic acid assay (Pierce) and equal amounts of lysate were precipitated with anti‐Shp‐2 antibody or 2 μg of GST fusion. IPs/mixes were washed three times in TxLB and were eluted from the beads by boiling for 5 min in 1× sample buffer and proteins were resolved on 10% acrylamide gels. Membranes were blocked in 1% gelatin and probed with anti‐phosphotyrosine 4G10 (UBI). Goat anti‐mouse–HRP (Bio‐Rad) was utilized as the secondary reagent. Embryo fibroblast cells were grown to sub‐confluency and serum starved for 70 h. For FGF stimulations, a media change into DMEM + 1 μg/ml heparin was performed prior to cell stimulations. Cells were then stimulated with a mixture of both acidic and basic FGF (Promega and UBI; final concentration 100 ng/ml). Cells were lysed in 1× sample buffer, boiled, DNA was sheared and proteins were resolved on 12% acrylamide: 0.12% bisacrylamide gels and Western blotted with anti‐MAP kinase monoclonal antibody (UBI).
We thank K.Harpal for help with histology, V.K.Lai for help with establishing cell lines, A.Cheng for help with gene targeting, D.Hu for help with sequencing, S.Holland for critical review of the manuscript and J.Rossant for many invaluable discussions. Predoctoral support for T.M.S. and postdoctoral support for M.H. was from the Medical Research Council of Canada. G.S.F. has a career development award from the American Diabetes Association. This work was supported by grants from Bristol Myers Squibb and the MRC, a Terry Fox programme grant from the National Cancer Institute of Canada (NCIC) and a Howard Hughes International Research Scholar Award to T.P. T.P. is a Terry Fox Cancer Research Scientist of the NCIC.
- Copyright © 1997 European Molecular Biology Organization