Distinct requirements for Gab1 in Met and EGF receptor signaling in vivo

Abstract

Gab1 is a multiadaptor protein that has been shown to be required for multiple processes in embryonic development and oncogenic transformation. Gab1 functions by amplifying signal transduction downstream of various receptor tyrosine kinases through recruitment of multiple signaling effectors, including phosphatidylinositol 3-kinase and Shp2. Until now, the functional significance of individual interactions in vivo was not known. Here we have generated knockin mice that carry point mutations in either the P13K or Shp2 binding sites of Gab1. We show that different effector interactions with Gab1 play distinct biological roles downstream of Gab1 during the development of different organs. Recruitment of phosphatidylinositol 3-kinase by Gab1 is essential for EGF receptor-mediated embryonic eyelid closure and keratinocyte migration, and the Gab1–Shp2 interaction is crucial for Met receptor-directed placental development and muscle progenitor cell migration to the limbs. Furthermore, we investigate the dual association of Gab1 with the Met receptor. By analyzing knockin mice with mutations in the Grb2 or Met binding site of Gab1, we show that the requirements for Gab1 recruitment to Met varies in different biological contexts. Either the direct or the indirect interaction of Gab1 with Met is sufficient for Met-dependent muscle precursor cell migration, whereas both modes of interaction are required and neither is sufficient for placenta development, liver growth, and palatal shelf closure. These data demonstrate that Gab1 induces different biological responses through the recruitment of distinct effectors and that different modes of recruitment for Gab1 are required in different organs.

Gab1 belongs to a class of signaling proteins that function as docking/scaffolding proteins downstream of tyrosine kinase and cytokine receptors (1, 2). Other members of this group include the insulin receptor substrates, p62dok family members, and FGF receptor substrates (3–5). These docking proteins are substrates of tyrosine kinases and interact with further adaptors and enzymes to assemble multiple signaling complexes. Gene ablation studies in the mouse have shown that Gab1 is essential for embryonic development: Gab1^−/− embryos die between embryonic day 14 (E14) and E18 and display defects in placental development, reduced size liver, malformation of limb and diaphragm musculature, open eyelids, and skin and heart defects (6, 7). Conditional ablation of Gab1 in the liver further showed that, in this organ, Gab1 participates in liver regeneration and that it modulates insulin–receptor signaling (8, 9).

Gab1 has an N-terminal pleckstrin homology domain, a Met receptor tyrosine kinase binding site (MBS), two Grb2 binding sites, and multiple tyrosine residues that when phosphorylated enable the recruitment of signaling effectors such as phosphatidylinositol 3-kinase (PI3K), the tyrosine phosphatase Shp2, the Crk/CrkL adaptor proteins, and RasGAP (10–15). The Grb2 adaptor is implicated in coupling Gab1 to Met as well as to other upstream tyrosine kinases and scaffolding proteins, like the EGF receptor, PDGF receptor, and FGF receptor substrate 2 (1, 2, 16–18). In addition, Gab1 can directly associate with Met through a central 13-aa MBS of Gab1. The MBS is unique for Gab1 and is not present in the other mammalian Gab family members, Gab2 and Gab3 (2).

The significance of Gab1 protein interactions have been primarily examined in vitro by using cell culture experiments and overexpression systems. For instance, the interaction of Gab1 and PI3K appears to be required for cell migration and survival of PC12 neuronal precursor cells downstream of the NGF receptor TrkA (19). The Gab1–Shp2 interaction has been shown to be required for Met-induced branching morphogenesis of epithelial MDCK cells in three-dimensional cultures and for ErbB2- and EGF receptor-dependent transformation of fibroblasts (12, 14, 20, 21). The in vivo function of the various Gab1 effector interactions remains unknown, and it is unclear whether these interactions are differentially required in the various Gab1-regulated biological events.

To investigate the physiological significance of Gab1 protein interactions in vivo, we generated a series of knockin mice with point mutations in the downstream effector binding sites PI3K and Shp2 and in the binding sites that connect Gab1 to upstream receptors, the Grb2 binding site and the MBS. We show here that the different Gab1 mutations adversely affect distinct receptor tyrosine kinase signaling pathways and result in distinct phenotypes. In particular, we show that the interaction of PI3K with Gab1 is required for Tgf-α/EGF receptor-induced keratinocyte migration and for eyelid closure. Recruitment of Shp2 by Gab1 is central to Met signaling, because mice display phenotypes characteristic of impaired Met signaling, namely embryonic lethality, defects in placenta development, and reduced muscle progenitor cell migration to the limbs. The interactions required for coupling Gab1 to the Met receptor also varied depending on biological context. Thus, we could show that, although either the direct or the indirect interaction of Gab1 with Met is sufficient for Met-dependent muscle precursor cell migration, both modes of interaction are required for placenta development, liver growth, and palatal shelf closure. We have therefore shown that, in vivo, both upstream and downstream interactions with the Gab1 adaptor protein are flexible and are differentially required in different biological contexts.

Results

Generation of Gab1 Docking Site Mouse Mutants.

To generate mice that express Gab1 docking site mutants instead of wild-type Gab1, we inserted cDNAs encoding amino acids 198–695 of either wild-type Gab1 (WT cDNA) or Gab1 mutants by homologous recombination in ES cells into exon d of the Gab1 locus (Fig. 1a). The cDNAs encode well characterized Gab1 mutants, Gab1ΔMet (V490A), Gab1ΔGrb2 (Δ341–348/L524P), Gab1ΔPI3K (Y448F/Y473F/Y590F), or Gab1ΔShp2 (Y628F) with a C-terminal myc epitope tag (10, 12–14, 17, 22). The floxed neomycin resistance cassette was excised in vivo by crossing with Cre deleter strain [supporting information (SI) Fig. 5a]. Heterozygous Gab1 knockin mice were intercrossed to obtain homozygous mutant knockin mice. Mice expressing wild-type Gab1 cDNA (Gab1^wt/wt) or the Gab1ΔPI3K mutation (Gab1^{ΔPI3K/ΔPI3K}) were viable and fertile and born close to the expected Mendelian ratio of 25% (SI Table 1). Gab1^{ΔShp2/ΔShp2} mice were found only up to E12.5 and at a reduced ratio. Gab1^ΔGrb2/Grb2 and Gab1^ΔMet/ΔMet embryos lived consistently up to E16.5 but were never born.

Fig. 1.

Generation of Gab1 knockin mutants. (a) The schematic representation of the Gab1 protein (lane 1) shows pleckstrin homology domain (PH), Grb2 binding sites (Grb2), MBS, and phosphotyrosine residues, which bind PI3K (orange) or Shp2 (purple). In the genomic locus of Gab1 (lane 2), exon d encodes amino acids 198–400 of Gab1. In the targeting vector (lane 3), amino acids 198–695 of wild-type Gab1 or Gab1 mutants are inserted. Gray boxes mark the position of external probes used for Southern blot analyses. Xb, XbaI; Nd, NdeI; H, HindIII. The structure of the knockin allele is shown in lane 4. (b) Western blot analysis from E12 embryos. Protein extracts were blotted with anti-Gab1 and α-tubulin antibodies (control). Gab1 expression is displayed as percent of Gab1+/+ mice. (c) Gab1ΔPI3K and Gab1ΔShp2 proteins are deficient in binding p85 PI3K or Shp2, respectively. Mouse embryonic fibroblasts (MEFs) were prepared from Gab1 knockin mice and stimulated with Tgf-α as indicated. Protein complexes were precipitated by using antibodies against the C-terminal Myc tag of the Gab1 knockin proteins and detected by Western blotting. An aliquot of the lysates was subjected to Western blotting for input control.

Northern blot analysis showed Gab1 mRNA expression in multiple tissues in Gab1^wt/wt and control (+/+) mice (SI Fig. 5b). Gab1 protein levels were examined by Western blotting of lysates prepared from E12 embryos. In all knockin animals, including those that express wild-type Gab1 cDNA, Gab1 protein was reduced to ≈40–50% of wild-type animals (Fig. 1b). We prepared fibroblasts from E12 knockin mice and immunoprecipitated Gab1 proteins using the myc epitope tag. Upon stimulation with Tgf-α, recombinant wild-type Gab1 precipitated both p85 PI3K and Shp2 (Fig. 1c, lane 3), whereas Gab1ΔPI3K and Gab1ΔShp2 failed to bind PI3K or Shp2, respectively (Fig. 1c, lanes 4 and 5). The embryonic fibroblasts were also examined for defects in the activation of downstream signaling pathways, like the Erk1/2 or PI3K/Akt pathways (SI Fig. 5c). Upon stimulation with TGF-α, Erk1/2 rapidly became activated in all embryonic fibroblasts analyzed. Erk1/2 activity remained elevated in Gab1 control- or Gab1ΔPI3K-expressing cells for >2 h. However, in Gab1ΔShp2-expressing cells, Erk1/2 phosphorylation returned faster to basal levels, demonstrating a deficit in sustained activation of these kinases. pAkt activation was similar in control Gab1- and Gab1ΔShp2-expressing cells but was reduced by ≈30% in Gab1ΔPI3K-expressing cells. This shows that the interaction of Shp2 and PI3K with Gab1 is required for full activation of Erk/MAPK and PI3K/Akt pathways, respectively.

Gab1–PI3K Interaction Is Required for Embryonic Eyelid Closure.

Gab1^{ΔPI3K/ΔPI3K} mice were viable, but frequently (14 of 22) they were born with open eyelids (Fig. 2b). In wild-type mice, eyelids are usually closed by E16.5 (Fig. 2a) and do not open until ≈2 weeks after birth. Gab1 is expressed in the eyelid epithelium (SI Fig. 6a). To test whether the PI3K signaling pathway is activated in normal eyelid development, we carried out antibody staining for phosphorylated Akt. In control animals at E16.5, strong phospho-Akt staining marks the periderm and the epithelium in the area of eyelid fusion (SI Fig. 6c). At E16.0, just before the eyelids are closed in control mice, there is strong phospho-Akt staining in the epithelium at the very tip of the upper eyelids (Fig. 2e) and lower eyelids (data not shown) of control mice. However, phospho-Akt staining in the eyelid epithelium of Gab1^{ΔPI3K/ΔPI3K} mice was reduced by 60% (Fig. 2f; quantification in SI Fig. 6b). These data suggest that, in the eyelid, Gab1 is required for full activation of the PI3K–Akt pathway. Apoptosis and cell proliferation were examined by cleaved caspase 3 and phospho-Histone 3 staining, respectively (SI Fig. 6 d–h). Apoptosis was not detectable in the eyelid epithelium of Gab1^{ΔPI3K/ΔPI3K} mice at that stage, and cell proliferation was comparable to controls.

Fig. 2.

PI3K signaling by Gab1 is important for eyelid closure and keratinocyte migration. Gab1^{ΔPI3K/ΔPI3K} mice are born with open eyelids (b; compare with heterozygous control in a). (c–f) Coronal sections of the eyelid region of E16.5 (c and d) and E16.0 (e and f) mice. Phospho-Akt staining is stronger in the eyelid epithelium of control than in Gab1^{ΔPI3K/ΔPI3K} mice (e and f). Phospho-Akt-staining is red. [Scale bars: 0.25 mm in c (for c and d) and 0.1 mm in e (for e and f).] el, eyelid; le, lens; co, cornea. Impaired in vitro wound closure of primary Gab1^{ΔPI3K/ΔPI3K} keratinocytes in response to Tgf-α is shown. After wounding, Gab1^{ΔPI3K/ΔPI3K} (g, i, and k) or Gab1^ΔPI3K/+ (h, j, and l) keratinocytes were cultured in serum-free medium (a and b) and treated with 20 ng/ml Tgf-α (i and j) or with 20 units/ml HGF/SF (k and l). Pictures were taken 2 days after wounding. The dotted line indicates the position of the initial wound.

We prepared keratinocytes from the skin of newborn Gab1^{ΔPI3K/ΔPI3K} mice and controls and examined them in a scratch-wound closure assay in cell culture (23, 24). Scratch wounds of control cells were virtually closed 2 days after treatment with Tgf-α (Fig. 2 g and i). In contrast, Gab1ΔPI3K-expressing cells showed impaired wound closure in response to Tgf-α (Fig. 2 h and j; quantification in SI Fig. 6h). Wound closure in the presence of HGF/SF was normal in Gab1^{ΔPI3K/ΔPI3K} keratinocytes (Fig. 2 k and l). These data demonstrate that binding of PI3K by Gab1 is essential for closure of the eyelid epithelia and for migration of keratinocytes induced by Tgf-α, but not HGF/SF.

Gab1–Shp2 Association Is Required for Placenta Development and Muscle Progenitor Cell Migration.

We investigated placenta development in the Gab1^{ΔShp2/ΔShp2} embryos. At E12.5, the labyrinth layer of Gab1^{ΔShp2/ΔShp2} placenta is 40% thinner than that of Gab1^wt/wt embryos and appears less organized (Fig. 3 a and b; quantification in c) as revealed by in situ hybridization staining using dlx3 as a marker for labyrinth trophoblast cells (6). In contrast, the spongiotrophoblast layer of Gab1^{ΔShp2/ΔShp2} mice is well developed (Fig. 2 d and e).

Fig. 3.

Placenta development and muscle progenitor cell migration are disturbed in Gab1^{ΔShp2/ΔShp2} mice. (a–e) Analysis of placenta stage E12.5. Paraffin sections were prepared from placentas of Gab1^wt/wt (a and d) and Gab1^{ΔShp2/ΔShp2} (b and e) mice. In situ hybridizations were carried out with the dlx3 probe (a and b), a marker for trophoblast cells of the labyrinth layer (Lb), and with the flt1 probe (d and e), a marker for the spongiotrophoblast layer (Sp). For quantification, the area fractions of labyrinth layer or spongiotrophoblast layer were compared with the corresponding areas of Gab1^wt/wt placenta. There was a significant difference in the size of the labyrinth layers (*, P < 0.05) (c). Whole-mount in situ hybridization of stage E10.5 embryos with an lbx1-specific probe of control Gab1^wt/wt (f and i) and Gab1^{ΔShp2/ΔShp2} (g and j) mice. Arrows mark myogenic progenitor cells in the limb buds. The arrowhead marks the hypoglossal stream of cells that migrates into the branchial arches (f and g). Vibratome transverse sections at forelimb level reveal reduced lbx1 staining in the limb buds of Gab1^{ΔShp2/ΔShp2} mice (h–j). For quantification, lbx1-positive areas were measured in corresponding limb sections. **, P < 0.001. (Scale bars: 0.5 mm.)

We also investigated whether limb muscles that are derived from migratory precursors (25) are affected by the homozygous Gab1ΔShp2 mutation. At E10.5, migratory cells in the limb buds or the hypoglossal stream of Gab1^{ΔShp2/ΔShp2} embryos were significantly reduced compared with Gab1^wt/wt embryos, as revealed by whole-mount in situ hybridization with lbx1 (Fig. 3 f and g, marked by arrows and arrowheads, respectively). Sections of the forelimb bud showed that, in Gab1^{ΔShp2/ΔShp2} mice, muscle progenitor cells are approximately three times less abundant and have not moved as far distally as in Gab1^wt/wt mice (Fig. 3 h–j and SI Fig. 7 a–c). Double immunostaining of limb bud sections with Lbx1 and cleaved caspase 3 or phospho-Histone 3 (26) did not reveal significant differences in apoptosis or cell proliferation of Gab1^{ΔShp2/ΔShp2} or Gab1^wt/wt muscle progenitor cells (SI Fig. 7 d and e). Staining for MyoD, a myogenic differentiation factor (27), revealed differentiating muscle precursor cells in limbs of E12.5 Gab1^{ΔShp2/ΔShp2} mice, but at reduced numbers (data not shown). These results demonstrate that binding of Gab1 to Shp2 is required for placenta development and the correct migration of muscle precursors into the limbs and other organs.

Intact MBS and Grb2 Binding Site of Gab1 Are Required for Placenta Development and Liver Growth.

Gab1^{ΔGrb2/ΔGrb2} and Gab1^ΔMet/ΔMet mice are nonviable and develop a placenta with a thin, disorganized labyrinth layer: At E14.5, the thickness of the labyrinth layers amounts to approximately half the thickness of controls (Fig. 4 a–c; quantification in SI Fig. 8a). We crossed Gab1^ΔGrb2/+ and Gab1^ΔMet/+ mice to obtain heterozygous compound mutants. The labyrinth layer phenotype was not rescued in Gab1^ΔGrb2/ΔMet mice (Fig. 4d; quantification in SI Fig. 8a). The livers of Gab1^{ΔGrb2/ΔGrb2} and Gab1^ΔMet/ΔMet mice were reduced in size: At E14.5 the liver-to-body-weight ratio was 8.0% in Gab1^wt/wt control mice but only ≈5.5% in Gab1^ΔGrb2/Grb2 mice, Gab1^ΔMet/ΔMet mice, and Gab1^ΔGrb2/ΔMet compound mutants (Fig. 4i). In contrast, emigration of Gab1^{ΔGrb2/ΔGrb2} and Gab1^ΔMet/ΔMet muscle progenitor cells to the limbs was comparable to Gab1^wt/wt mice (data not shown). At E14.5, limb muscles of Gab1^{ΔGrb2/ΔGrb2} and Gab1^ΔMet/ΔMet mice were formed properly, in contrast to Gab1^−/− mice (SI Fig. 8 b–i; quantification in j and k). Together, these data indicate that coupling of Gab1 to Met, via the Grb2 binding site and the MBS, is required for placenta formation and liver growth, but for limb muscle formation one interaction site alone is sufficient.

Fig. 4.

Placenta defect, cleft palate, and small size liver in Gab1^{ΔGrb2/ΔGrb2} and Gab1^ΔMet/ΔMet mice. Placenta stage E14.5 was analyzed as in Fig. 3 a–e. There is no rescue of the thin labyrinth layer phenotype in Gab1^ΔGrb2/ΔMet mice (d). [Scale bar: 0.5 mm in e (for a–h).] Liver-to-body-weight ratios of different Gab1 knockin embryos were determined at E14.5 (i). Cleft palate incidence was determined at E16 (l) and E15 (m). There was a significant difference in cleft palate incidence between Gab1^{ΔGrb2/ΔGrb2} and Gab1^{ΔGrb2/ΔGrb2}/Met^+/− mice and controls (Gab1^ΔGrb2/+, Met^+/−, or Gab1^ΔGrb2/+/Met^+/−). *, P < 0.05. Coronal sections of stage 16 embryos reveal a closed palate in Gab1^{ΔPI3K/ΔPI3K} mice (l), and open palates were found in Gab1^{ΔGrb2/ΔGrb2} mice (m) and Gab1^ΔMet/ΔMet mice (n). [Scale bar: 0.5 mm in l (for l–n).]

Cleft Palate in Gab1^{ΔGrb2/ΔGrb2} and Gab1^ΔMet/ΔMet Mice.

Coronal sections of Gab1 knockin embryos at E16.0 revealed a new phenotype that had previously not been associated with Met signaling. This was the presence of open secondary palates in approximately one-third of Gab1^{ΔGrb2/ΔGrb2} and Gab1^ΔMet/ΔMet embryos (Fig. 4 l–n; quantification in j). Analysis of the conventional Gab1^−/− mice showed also a significant incidence of open secondary palates (Fig. 4j). Gab1 expression was prominent in the epithelium of the palatal shelves, similar to Met (SI Fig. 9 b and c). To test for genetic interaction, we crossed a Met null allele (28) into a Gab1^{ΔGrb2/ΔGrb2} background. The frequency of cleft palate was significantly enhanced in the compound Gab1^{ΔGrb2/ΔGrb2}/Met^+/− mutants (Fig. 4k). Our data indicate that indeed the Met and Gab1 pathways have an important function in palatal shelf closure.

Discussion

Here we have performed a comprehensive in vivo analysis of Gab1 docking site functions through a gene knockin approach in the mouse. Mouse mutants containing specific mutations in Gab1 were screened for phenotypes in different tissues and at different stages of development. We have demonstrated differential requirements of Gab1 effectors and upstream receptor interactions in different biological contexts.

Gab1–PI3K Signaling Is Required for EGF Receptor Signaling and Eyelid Closure.

Previous experiments using MDCK epithelial cells have suggested that PI3K signaling is essential for HGF/SF- and Met-induced cell migration (29). Surprisingly, unlike Met or Gab1 knockout mice (6, 7, 28), Gab1^{ΔPI3K/ΔPI3K} mice are viable, limb muscle formation is normal, and liver growth is unaffected (SI Fig. 10). Instead, Gab1^{ΔPI3K/ΔPI3K} mice are born with open eyes at birth, a phenotype that is not associated with Met signaling but has previously been reported in mice with perturbed EGF receptor signaling (30–32). We showed that a downstream component of PI3K signaling, Akt (33), is strongly activated in the area of eyelid fusion in wild-type embryos but greatly reduced in Gab1^{ΔPI3K/ΔPI3K} mutant embryos. The recruitment of PI3K by Gab1 may therefore be an important signaling event by the EGF receptor during eyelid fusion. We did not find any indication of enhanced apoptosis or decreased proliferation in the eyelid epithelium of Gab1^{ΔPI3K/ΔPI3K}. Instead, we found that keratinocytes derived from Gab1ΔPI3K homozygous mice have an impaired response to Tgf-α in scratch-wound cell migration assays, suggesting that recruitment of PI3K by Gab1 is involved in EGF receptor-induced cell migration.

Gab1–Shp2 Interaction Is Essential for Met Signaling in Embryonic Development.

Gab1^{ΔShp2/ΔShp2} mice showed the most severe phenotypes: embryonic lethality, a thin labyrinth layer, and impaired migration of muscle progenitors. These phenotypes are also seen in Gab1 and Met knockout mice (6, 7, 28). Thus, a major function of Gab1 in embryonic development is the recruitment and activation of Shp2 downstream of Met (SI Fig. 10). A similar importance of Shp2 recruitment by multiadaptor proteins has been shown for FGF receptor substrate 2α, which is structurally related to Gab1 (3). The Shp2 binding sites of FGF receptor substrates 2 are critical for eye development and FGF receptor signaling (4, 34, 35). The functional importance of the Gab1–Shp2 association is conserved in lower vertebrates. In Drosophila, the primary function of the Gab1 family member Dos is to bind the Shp2 ortholog Corkscrew (36, 37), whereas the Gab1-like protein Soc1 in Caenorhabditis elegans is required for the recruitment of PTP-2 (Shp2) (38).

Embryonic lethality of Gab1^{ΔShp2/ΔShp2} mice, however, occurs earlier than that of Gab1^−/− mice, between E12.5 and E14.5 rather than between E14.5 and E18.5 (6). Lower viability could result from the thinner placental labyrinth layer present in Gab1^{ΔShp2/ΔShp2} mice as compared with Gab1^−/− embryos. Recent in vitro experiments have shown that Gab1 is also a substrate of the tyrosine phosphatase Shp2. In particular, Shp2 dephosphorylates the binding sites on Gab1 for RasGAP, preventing the recruitment of this negative regulator of Ras and enabling sustained activation of the Ras/MAP kinase pathway (13). Mutation of the Shp2 binding site of Gab1 can enhance recruitment and activation of RasGAP, thereby inhibiting Ras signaling even further (13).

Grb2 Binding Site and MBS on Gab1 Are Differentially Required for Coupling to the Met Pathway.

Among the family of docking proteins, Gab1 is unique in that it possesses a direct binding site for Met (1, 2, 39). In addition, as occurs downstream of other receptors, Gab1 can bind Met indirectly via the Grb2 adapter (SI Fig. 10). Remarkably, we find that mice expressing the Gab1 mutants lacking the MBS or the Grb2 binding sites show identical phenotypes, namely placenta defects and small livers. Furthermore, compound mutants, which express one Gab1ΔGrb2 and one Gab1ΔMet allele, do not rescue the Gab1-regulated placenta or liver phenotypes. Given that the MBS mediates solely Met signaling, this observation underscores the importance of Gab1 downstream of Met, but not other tyrosine kinases, during embryogenesis and demonstrates that direct and indirect recruitment of Gab1 is required in placenta and liver development. Because Gab1 expression in the Gab1 knockin mouse mutants is ≈50% lower than in wild-type mice, we cannot exclude that the mutant phenotypes are exaggerated because of the lower expression level of Gab1 knockin mutants, in particular in liver and placenta. In that respect, it is surprising that Gab1^{ΔGrb2/ΔGrb2}, Gab1^ΔMet/ΔMet, and compound mutants Gab1^{ΔGrb2/ΔGrb2}/Met^+/− and Gab1^ΔMet/ΔMet/Met^+/− show normal limb muscle formation, an event that is also controlled by Met and Gab1 (6, 28). Thus, even at reduced levels of Met and Gab1, recruitment of Gab1 either through direct interaction with Met or indirectly through Grb2, is sufficient for normal signaling in limb muscle cells. This shows that the requirements for the recruitment of Gab1 downstream of Met are variable and depend on the biological context. The in vivo function of Grb2 in Met signaling was controversial. A hypomorphic Met knockin mouse, which carries a point mutation in the Grb2 binding site of Met, showed a reduction in muscles that are derived from migratory progenitor cells, suggesting that Met–Grb2 is essential for limb muscle formation (40, 41). However, this phenotype disappeared when the neomycin was deleted (42), thus questioning the original conclusion. In addition, replacing the multiple docking site of Met with two Grb2 binding sites caused no phenotype in Met knockin mice, indicating that Grb2 is sufficient to recruit all of the necessary downstream effectors of Met (41). Our Gab1 mutant analysis now reveals that the mechanism of recruitment of Gab1 to Met depends on the tissue. For functional Met signaling in placenta and liver formation, efficient coupling of Gab1 to Met by both the Grb2 binding site and the MBS is required.

A New Function of Gab1: Palatal Shelf Closure.

One-third of Gab1^{ΔGrb2/ΔGrb2} and Gab1^ΔMet/ΔMet embryos showed incomplete closure of secondary palates, a phenotype that had not been previously associated with Met or Gab1 signaling. This phenotype is specific, because it did not manifest in Gab1^{ΔPI3K/ΔPI3K} mice. Furthermore, we detected open secondary palates in the classical Gab1 knockout mice. Again, Gab1^ΔMet/ΔMet and Gab1^{ΔGrb2/ΔGrb2} mice display identical phenotypes, indicating that the Met pathway is involved and that recruitment of Gab1 through both sites is required. Crossing-in one Met null allele into the Gab1^{ΔGrb2/ΔGrb2} background significantly enhanced the incidence of open secondary palates, confirming the genetic interaction of Gab1 and Met in palatal shelf closure. The analysis of Met^−/− mice in respect to palatal shelf closure was inconclusive, because Met^−/− animals were grossly growth-retarded at E15.5. Despite this retardation, we found Met^−/− animals with closed secondary palates. The fact that incomplete penetrance of the cleft palate phenotype was found in Egf receptor^−/− mice (43) and Gab1^−/− and Met mutants suggests that these pathways are redundant. FGF receptors 1, 2, and 3, PDGF receptor α, and Grb2 mutant mice display more severe craniofacial defects, including clefting of the secondary palate (44–46). In humans, the etiology of cleft palate is complex and involves major and minor genetic and environmental factors (47). Our data suggest that Gab1 and Met are additional candidate genes in the etiology of human cleft palates.

Our functional analysis of Gab1 mutants in mouse development has shown that downstream signaling and recruitment of this multiadaptor to various receptors is a complex process and that distinct interactions are required depending on the biological context. The Met kinase pathway is a validated therapeutic target in cancer therapy and a prognostic marker for metastasis (39, 48). Gab1 has also been shown to be required for cell transformation by oncogenic Met (49). We show here that the principal function of Gab1 in vivo is the recruitment of Shp2 downstream of the Met receptor. Thus, the Shp2, MBS, or Grb2 docking sites of Gab1 may be good targets for pharmacological intervention to treat human malignancies.

Materials and Methods

Generation of Gab1 Knockin Mice.

For the construction of the targeting vector, genomic fragments isolated from a λ FIXII 129/Ola library were introduced into pTVFlox-0 vector (6, 50). The knockin targeting vector contains as 5′ flanking sequence and NdeI and EcoRV genomic fragment, which includes a part of exon d (codons 123–339). The cDNAs (codons 340–695) of Gab1 mutants were inserted into the EcoRV site in exon d. The cDNAs contain a C-terminal myc epitope tag plus poly(A) tail of bovine growth hormone (12). The targeting vector was introduced into E14–1 ES cells as described previously (50). Positive cell clones were identified by Southern hybridization using the 5′ and 3′ external probes (Fig. 1a) and the neomycin resistance gene (neo) as internal probe. Two independent ES cell clones of each mutation were microinjected into C57BL/6 blastocysts to generate chimeras and backcrossed with C57BL/6 animals as described previously (50). Germ-line transmission was verified by Southern blotting. Animals were genotyped by PCR using the primers neos (GCCTTCTATCGCCTTCTTGAC), ID3 (TGGGTTTACTTGCTTGTGTGCC), and wts (GAACAAGTTGCGGAAAGGTAAAGCC). The neo cassette was excised by crossing with Cre deleter strain (51). Knockin mice without the neomycin resistance gene cassette were genotyped with wts, BGH5 (AAGGGGGAGGATTGGGAAGA), and ID5 (CCCCTGTCTCCGTCCCATAC) primers. Mice were analyzed in a mixed 129/C57BL/6/CD1 background.

Antibodies, Western Blot Analyses, Histology, and in Situ Hybridization.

The following antibodies were purchased: myc (9E10), α-tubulin (T-9026), and anti-skeletal fast myosin antibody (M4276) from Sigma–Aldrich (St. Louis, MO); Met (sc-162) and Shp2 (sc-280) from Santa Cruz Biotechnology (Santa Cruz, CA); p85 PI3K (06-195), Gab1 (06-579), and phospho-Histone 3 (06-570) from Millipore (Billerica, MA); pAkt from Cell Signaling Technology (Danvers, MA); and alkaline phosphatase-conjugated anti-mouse secondary antibody and Cy3-labeled goat anti-rabbit secondary antibodies from Jackson Immuno Research (West Grove, PA).

Mouse fibroblasts were prepared from E12 embryos (52). Cells were cultured in DMEM plus 10% serum and plated onto tissue culture dishes coated with 0.1% gelatin. Immunoprecipitations and Western blots were carried out as described (12). Cells were starved in 0.5% serum overnight. After stimulation with 20 ng/ml Tgf-α for 2 min, cells were lysed in lysis buffer (50 mM Tris, pH 7.5/150 mM NaCl/1% Triton X-100/1 mM EDTA) supplemented with a proteinase inhibitor mixture (1836170; Roche; Penzberg, Germany), 1 mM PMSF, and 1 mM vanadate. Proteins were immunoprecipitated and separated by 8% SDS/PAGE. Protein extracts from E12.5 embryos were lysed by sonication. For histology, embryos were fixed in 4% formaldehyde/PBS and embedded in paraplast as described (6). Sections (7 μm) were prepared, deparaffinized with xylene, and rehydrated with graded ethanol. For immunostainings, sections were unmasked with citrate buffer as described by Cell Signaling Technology. Yo-Pro-1-iodide (Y3603; Invitrogen, Carlsbad, CA) was used to stain nuclei. Sections were embedded in Immumounting medium (Thermo Electron). In situ hybridizations were carried out as described previously (6, 41). Antisense probes were labeled by in vitro translation and incorporation of digoxigenin-11-UTP (Roche Molecular Biochemicals). Bound probes were detected with alkaline phosphatase-labeled anti-digoxigenin (Fab fragment; 11093274910; Roche) and 4-nitroblue tetrazolium chloride/5-bromo-4 chloro-3 inodyl phosphate (11383213001; Roche) or BM purple AP substrate (11442074001; Roche) and counterstained with Pyronin G (Sigma–Aldrich). For vibratome sections, stained embryos were fixed in 4% formaldehyde/PBS overnight, embedded in 20% gelatin, refixed, and sectioned at 30 μm. For lacZ stainings, embryos were frozen in Tissue Tek (4583; Sakura, Torrance, CA), and 12-μm sections were fixed in 0.2% glutaraldehyde and stained with X-Gal solution as described (7). Pyronin G was used as counterstain. Sections were examined by light microscopy by using a Zeiss Axioskop, and embryos were examined with the Zeiss StemiSV11 microscope and scanned with the video camera Axiocam HRc (Zeiss, Jena, Germany) using Axiovision 3.1 software. Cells were examined with the Axiovert 135 microscope (Zeiss) and photographed by using RT Monochrome Spot camera (Diagnostic Instruments) and spot.basic 4.6. software program. Immunofluorescent stainings were examined with the confocal microscope LSM 510 META (Zeiss) and LSM 5 Pascal Version 3.2. Images were processed by using Photoshop 7.0 and Illustrator 10 software (Adobe Systems, San Jose, CA).

Isolation and Culture of Primary Keratinocytes.

Primary keratinocytes were prepared as described previously (23, 24). After dispase treatment, the epidermis was separated from the dermis and digested in 0.25% trypsin/PBS. Cells were cultured in defined keratinocyte serum-free medium with defined growth supplements (Gibco BRL), 10 μg/ml EGF (Sigma–Aldrich), and 10⁻¹⁰ M cholera toxin (Sigma–Aldrich). For in vitro scratch assays, cells were grown to confluency and kept in growth factor-free medium for 2 days. The monolayer was wounded with a disposable pipette tip, and growth factors were added.

Statistical Analysis.

Data were analyzed with the SPSS statistical program package. In experiments with only two conditions, Student's t test was applied. In experiments of multiple conditions, the multiple test of ANOVA was used. In the following pairwise test, the error probability was corrected by Bonferroni. The cross-tabulations were computed, and a χ² test of Person and the exact test by Fisher were applied vice versa.

Abbreviations

PI3K

phosphatidylinositol 3-kinase

MBS

Met receptor tyrosine kinase binding site

En

embryonic day n.